Professional Documents
Culture Documents
Quantum Dots
SEMICONDUCTORS
AND SEMIMETALS
Volume 60
Semiconductors and Semimetals
A Treatise
SEMICONDUCTORS
AND SEMIMETALS
Volume 60
Volume Editor
MITSURU SUGAWARA
OPTICAL SEMICONDUCTOR DEVICE LABORATORY
FUJITSU LABORATORIES LTD.
ATSUGI, JAPAN
ACADEMIC PRESS
San Diego London Boston
New York Sydney Tokyo Toronto
This book is printed on acid-free paper. @
COPYRIGHT 0 1999 BY ACADEMICPRESS
ALL RIGHTS RESERVED.
NO PART OF THIS PUBLICATION MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM OR BY ANY
MEANS, ELECTRONIC OR MECHANICAL, INCLUDING PHOTOCOPY, RECORDING, OR ANY INFORMATION
STORAGE AND RETRIEVAL SYSTEM, WITHOUT PERMISSION IN WRITING FROM THE PUBLISHER.
The appearance of the code at the bottom of the first page of a chapter in this book indicates
the Publisher’s consent that copies of the chapter may be made for personal or internal use of
specific clients. This consent is given on the condition, however, that the copier pay the stated
per-copy fee through the Copyright Clearance Center, Inc. (222 Rosewood Drive, Danvers,
Massachusetts 01923), for copying beyond that permitted by Sections 107 or 108 of the US.
Copyright Law. This consent does not extend to other kinds of copying, such as copying for
general distribution, for advertising or promotional purposes, for creating new collective works,
or for resale. Copy fees for pre-1999 chapters are as shown on the title pages; if no fee code
appears on the title page, the copy fee is the same as for current chapters. 0080-8784/99 $30.00
ACADEMIC PRESS
525 B Street, Suite 1900. San Diego, C A 92101-4495. USA
http://www.apnet.corn
ACADEMIC PRESS
24-28 Oval Road, London NWl 7DX, UK
http://www.hbuk.co.uk/ap/
PREFACE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
LISTOF CONTRIBUTORS . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
V
vi CONTENTS
MITSURU SUGAWARA
This Page Intentionally Left Blank
List of Contributors
...
Xlll
This Page Intentionally Left Blank
SEMICONDUCTORS AND SEMIMETALS. VOL. 60
CHAPTER 1
. . . . . . . . . . .
I. INTRODUCTION . . . . . . . . . . . . . . . . . 1
11. ELECTRONIC . . . .
NANO-STRUCTURES
STATES OF SEMICONDUCTOR QUANTUM 3
111. INTERBAND OPTICAL TRANSITION . . . . . . . , . . . , . , . . , , . . 11
1. Linear and Nonlinear Opticul Susceptibiliiy . . . . . . . . . . . . . . . 12
2. Spontaneous Emission of Photons . . . , . . . , , . , . . , . , . , . 21
3. Rate Equations for Laser Opcrutions . , . . . . , , . . . , . . . , . 21
IV. EXCITON OPTICAL PROPERTIES . . . . . . , . . . , , . . . . . . . . . 30
1. State Vectors . . . . . . . . . . . . , . . . . . . . . . . . . . . 32
2. Effective-Mass Equations . . . . . . . . , , . . . . . . . . . . . . 36
3. Exciton-Photon Interactions . . . . . . . , , . . . . . . . . . . . . 43
4. Opticul Absorption Spectra . . . . . , . , . . . . . . . . . . . . . 47
5. Spontaneous Emission of Photons in Quantum Wells and Mesoscopic
Quantum Disks . . . . . . . . . . . . . . . . . , . . , , . . , . 55
6. Spontaneous Emission of Photons in Quantum Disks Placed in a Planar
Microcavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
I. The Coulomb Effect on Opticd Guin Specisu . . . . . . , , . . , . . . I7
V. QUANTUM-DOT LASERS . . . . . . . . . . . . . . . . . , . , , , . . 83
1. The Effect of Carrier RdcJXUliOn Dynunnc.s on Luser Per:/Om~nnce . . . . . 86
2. Efect of Homogeneous Brmlrning of Opticul Gain on Lasing Emission
Spcctsu.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3. Bi-Exciton Spontaneous Emission unrl Lasing . . . . . . . . . . . . . . 97
VI. SUMMARY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Appendix.. . . . . . . . . . . . . . . . . . . . . . . . . . , , . . . . . 107
References . . . . . . . . . . . . . . . . . . . . . . . . , . . . . . , . . 112
I. Introduction
InP
InGaAs
20 nm
H
Bulk
r3L V=L3
Quantum wire
Energy
=t Energy
X
Quantum well Quantum dot
OW
Well Barrier
D = L2
Energy Energy
FIG. 1.2. Schematic view and graph of quantum nano-structures and their density of states.
1
V2 + V(r) Fk(r) = EFk(r)
Bulk Materials
h2k2
E = E(k) = ~
2m*
1
Fdr) = -exp( - i k r) - (1.4)
Jv
where V = L3 is the crystal volume. The envelope wave function is normal-
ized in the crystal. The wave vector of k = ( k x ,k,, k,) satisfies the periodic
boundary conditions as k , = (2nn,)/L, k, = (2nn,)/L, and k, = (2nn,)/L;
n x , n y , and n, are integers. The density of states per unit volume, which is
the number of states between the energy of E and E + dE, is given as
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR NANO-STRUCTURES7
QUANTUM
D(E) = hxG[E(k) - El
Vk
-
-- x-
v '
(243
x R, [: dkk2G[E(k) - El
Quantum Wells
The confinement potential for the square quantum well is given as V(r) =
V(z), where V ( z ) = co when IzI 3 L,/2; V(z) = 0 when IzI < L,/2; and L, is
the well width (Fig. 1.3(a)). Under the infinite potential height, the wave
function vanishes at the boundaries between the well and barrier layers. The
eigenvalue is
1
Fk(r)= -cp,,(z) exp(--ikll*rll)
@
and
Here, D = L2 is the area of the quantum well; k = (k", n,n/L,); the in-plane
wave vector is kll = ( k x ,k y ) ; r = (rll, z); rll = (x,y); and n, = 1,2,3,. . . . The
wave function in the z direction becomes stationary. The minimum energy
and the energy separation between each quantized state increase as the well
width decreases. Figure 1.3(b) illustrates the parabolic dispersion curve in
the conduction and valence bands. The conduction-band electron effective
mass is written as mQ. The superscript 11, indicates that the mass is for the
in-plane carrier movement. Since the valence-band effective mass is negative
in most cases, it is more convenient to introduce a positive particle of a hole
8 M~TSURU
SUGAWARA
'T Conduction
Et
transition
________----____---
(4
I
I
band
(b)
I
FIG. 1.3. (a) Confinement potential for the square quantum well, V ( z ) , and the quantized
energies for an electron in the conduction band. En*-,and for a hole in the valence band, En&=.
E, is the band gap of the well material. (b) The band dispersion of the conduction band and
valence band in the quantum-well plane. mr" is the electron effective mass, mrll is the hole
effective mass, and m:I1 is the reduced mass. Interband optical transition occurs almost
perpendicularly in the dispersion curve.
with the opposite effective mass of m;. Then, Eq. (1.6) holds also for holes
with the energy origin at the valence-band maximum and with the energy
positive in the downward direction.
EneZ(or E,J and En,= (or En,J in Fig. 1.3(a) are quantized energies for
electrons and holes, respectively. The quantized wave functions are de-
scribed as cpne,(z) (or cp,,Jz)) for electrons, and cp,,,(z) (or cp.,,(z)) for holes.
Letting LQW be the quantum-well thickness, that is, the sum of the well and
barrier regime thickness- the density of states per unit volume is
The confinement potential for the square quantum wire is given as V(r) =
+
V(y) V(z), where V(y) is the additional confinement potential, with
V ( y ) = GO under lyl 3 L,/2 and V ( y ) = 0 under lyl < LJ2; and L, is the
y-direction length of the wire cross-section. The eigenvalue to the infinite
barrier model is
c = E ( k ) = ,h2k2
,.En,,+En==~[k:+(~)2+~~)2] (1.10)
2m*
where EnYhas the same form as that of En=of Eq. (1.6) and ny,II,= 1,2,3.. . .
The electron energy further increases due to the additional confinement. The
corresponding normalized wave function is
2N .
D(E) G[E(k) - E l
ny,n,.k,
2Nwi
=--X-XQ1
L 2n
L
1
ny,nz
1; dkG[E(k) - El
-
Nwi
~~
J2m* c 1
[eV- K~] (1.12)
n h ny.nz J E- E . -
~ E,,=
where Q, = 2.
Quantum Dots
The interband optical transition between the conduction band and the
valence band reflects the density of states derived above. Figure 1.4(a) shows
the optical absorption spectra of the quantum well of Fig. l.l(a) (Sugawara,
1993a). The absorbance is measured by
(1.16)
where li, is the intensity of the incident light beam; I,, is that of the
transmitted light beam; and N , is the number of well layers-10 in this
case. Clearly observable is the steplike optical absorption spectrum due to
the two-dimensionality of quantum wells. (Strictly speaking, this clear
steplike absorption continuum arises from the parabolic nature of the
ground-state heavy-hole valence band, which is caused by enhanced split-
ting against the light-hole valence subband under the in-plane compressive
strain.) In addition, we see peaky spectra due to the electron-heavy-hole
exciton resonance (le-hh) at the absorption edge and the electron-light-
hole resonance (le-lh) on the absorption continuum at a shorter
wavelength. We also observe higher-order electron-heavy-hole exciton
resonance, as (2e-hh). The numbers of 1 and 2 represent the z-direction
quantum number, nz, which should be equal between the conduction and
valence bands, for the allowed optical transition (see the appendix).
Figure 1.4(b) shows the photoluminescence spectra of the quantum dots
of Fig. l.l(b) (Mukai and Sugawara, 1999). As the excitation laser power
increases, peaky emissions appear one by one, caused by the interband
transition between the discrete energy levels of the conduction and valence
bands. This is experimental proof of the zero-dimensionality of electrons and
holes in quantum dots.
In the following sections, formulae to describe the optical response of
quantum nano-structures are derived.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 11
NANO-STRUCTURES
Pe-hh 1e-hh
0
t le-lh
- 0
1.2 1.4 1.6
Wavelength (pm)
(4
FIG. 1.4. (a) Optical absorption spectrum at 77K of the In,,,Ga,,,,As/InP strained
quantum wells of Fig. l,l(a). The steplike optical absorption spectrum due to the two-
dimensional character of the density of states is clearly observed. Resonant spectra are due to
the ground-state electron-heavy-hole exciton (le-hh) at the absorption edge, the electron-
light-hole exciton (le-lh) on the absorption continuum at the shorter wavelength, and the
higher-order electron-heavy-hole exciton (2e-hh) (from Sugawara et al., 1993a. Copyright
1993 by The American Physical Society). (b) Photoluminescence spectra at 5 K of the
In,,,Ga,,,As/GaAs quantum dots of Fig. l.l(b) for various excitation level. Emissions from the
discrete energy levels due to three-dimensional quantum confinement are clearly observed
(from Chapter 4).
making the materials quite useful for optical devices. “Direct” means that
both the conduction-band minimum and the valence-band maximum are at
the same electron wave vector position in the band dispersion diagram,
making possible the electron-hole recombination by the emission of
photons, with a wave vector selection rule satisfied. This is also the case in
quantum nano-structures consisting of direct band-gap materials and hav-
ing type-I band lineup, where both electrons and holes are confined in the
materials. A variety of material systems fall into this category, such as
GaAs/AlGaAs, InGaAsP/InP, InGaP/AlGaInP, and InGaN/GaN.
The optical responses of semiconductors are categorized as optical
absorption, stimulated light emission, spontaneous light emission, and
various nonlinear optical processes. All are applied to various optical
devices-for example, optical absorption to photodetectors, modulators,
and solar batteries; stimulated emission to lasers and optical amplifiers;
spontaneous emission to light-emitting diodes; nonlinear processes to
switches and wavelength conversion devices. Depending on the required
photon energy in practical application fields such as optical transmission
systems and optical storage, we choose appropriate semiconductor ma-
terials, from infrared to red to blue to violet. As seen in Fig. 1.4, the optical
response varies, depending on the confinement dimension.
This section presents the theoretical bases of the interband optical
transitions in quantum nano-structures. In addition, a prescription to
simulate lasing operation using rate equations is briefly summarized. The
excition effect due to the electron-hole Coulomb interaction is not taken
into account here, but is treated in Section IV. By comparing the results in
this section with those in Section IV, readers will see how the exciton effect
influences semiconductor optical response.
1. LINEARAND NONLINEAR
OPTICALSUSCEPTIBILITY
Here, I represents the band index, k represents the electron wave vector, and
12, k) represents the Dirac state vectors to describe the electron states in the
conduction and valence bands. The dimension of materials need not be of
concern for the moment. The state vector satisfies the orthogonality relation
Of (A‘, k’II, k) = 6;.;,6kk. and the complete relation of Cj.,k 12, k)(A, kl = 1.
By taking the time differential of Eq. (1.17) and using the time-dependent
Schrodinger equation, the density operator satisfies the equation of motion
as
i
aP(t)
-=
at
--[H,+H,,p(t)]
h
+( - (1.18)
(1.19)
A&, + C.C.]
t ) = fe,[Aq,u(w)e''~"-"') (1.21)
k, z k, = k (1.24)
Linear Susceptibility
The linear susceptibility and the linear optical absorption and gain
spectra are derived here. Let us use a two-band model, that is, we consider
the interband transition between one conduction band, denoted by c, and
one valence band, denoted by v. The contributions from other bands can be
added up later. One optical mode with a fixed wave vector, polarization, and
frequency is taken into account because this is a situation in which we
observe the linear optical absorption and gain of semiconductors. Thus, Eq.
(1.19) becomes
where u&k is the frequency of the conduction-band electron with k, and wsk
is the frequency of the valence-band electron with k. Equation (1.20)
becomes
H I = I*cr,k(t)lcrk)(v, kl + &,k(L)lv? k)<c, kl (1.26)
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 15
NANO-STRUCTURES
with
where
(1.28)
(1.29)
and
See the appendix for the interband transition matrix element of P:L,.k.The
density matrix of Eq. (1.17) becomes
The nondiagonal terms, pCDand pVc,are related to the polarization, and the
diagonal terms, p,, and p,,, represent the population of the conduction and
valence bands, respectively.
From Eqs. (1.18), (1.25), (1.26), and (1.32), we get the motion of the
nondiagonal and the diagonal terms as
and
16 MITSURUSUGAWARA
Here, o,,,k = w,,k - uL..k; r,, is the transverse relaxation rate due to scat-
tering; rTIis the longitudinal relaxation rate of the conduction-band elec-
tron; and r[ is the longitudinal relaxation rate of the valence-band electron.
Due to intraband scattering. the relationship of jc,(k, t ) = -jJk, t ) does
not always hold. These equations of motion can be solved by the conven-
+ + +
tional perturbation expansion of p = p(O) p ( l ) p(') . . . , where the
number in parentheses represents how many times the electron has interac-
ted with the electromagnetic field. The zero-th order term, p(O), represents
the initial state of materials. Equation (1.18) can be separated according to
the perturbation order. By expanding the density operator into the Fourier
series of p(t) = X u ,p(o')e-iu'*,we obtain the first-order perturbation term as
(1.36)
(P:f:(o)/Z) = = -
Tr[p(’)(o)P] p$)(o)( jlee;rli)Qd
i.j=c.r
where P is the polarization per unit volume in the e, direction, and the
relationship of
(1.38)
1
( i = c,v) (1.39)
= exp[(hw,., - Pi)/kBT] +1
1 OPTICAL
PROPERTIES QUANTUM
OF SEMICONDUCTOR 17
NANO-STRUCTURES
(1.41)
where
( 1.43)
(1.44)
Let us perform the summation on the wave vector in Eq. (1.42) for the
parabolic band with the dispersion of
(1.45)
is the reduced effective masses; ELd) is the quantized energy in the conduction
band; ELd’ is that of the valence band (see Fig. 1.3, depicting for quantum
wells); and k(d) is the electron wave number in the free direction. By
substituting the delta function for the broadening function of Eq. (1.43), we
get an analytical expression for bulk (d = 3), quantum wells (d = 2), and
quantum wires (d = 1) as (Haug and Koch, 1994)
(i = c,v) (1.48)
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 19
NANO-STRUCTURES
For quantum dots, by omitting the wave vector from Eq. (1.42), we get
where indexes c and v represent discrete energy states in the conduction and
valence bands. By comparing the results in Section I1 with Eqs. (1.47) and
(1.49), we see that the interband optical absorption spectrum directly reflects
the density of states or, more correctly, the joint density of states between
the conduction band and the valence band. The steplike absorption continu-
um in Fig. 1.4(a) in quantum wells fits well with Eq. (1.47) for d = 2, .f, = 0,
and S, = 1. The sign of the absorption coefficient turns from positive to
negative when f, - f, < 0; in other words, the optical gain appears.
Nonlinear Susceptibility
and
p1.t;'(w) = 0 (1.52)
( P g ( 0 = 2w, - cu1)/2)
(1.55)
In the numerical coefficient in Eq. (1.55), 3! represents the order of the
interaction and 2! indicates that two of the three electromagnetic fields
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 21
NANO-STRUCTURES
1
w2- w,, + ir,,
-
1
]
- TCt,
(ol - o,,~
[m2/V2] (1.56)
Taking the light “2” with the frequency of w 2 as the pump beam, and the
light “1” with w 1 as a probe beam, Eq. (1.55) describes the polarization to
generate a signal beam with a frequency of w, = 2w2 - wl. If the two pump
beams proceed in the opposite directions, the signal is the phase conjugate
wave of the probe beam. See Chapter 7 for experiments on nondegenerate
four-wave mixing for wavelength conversion and conjugate wave generation
in semiconductor lasers. The bandwidth of is dominated by the dephas-
ing rate of the polarization To,, which determines the response speed of the
nonlinear process (see Chapter 7 ) .
Using f 3 ’ of Eq. (1.56) under the degenerate case of o2= ol, the optical
gain, including the nonlinear effect, is given as
(1.58)
2. SPONTANEOUS EMISSION
OF PHOTONS
+
AqJr, t ) = A4,~(o,)e,[efq"h,(t)e-iq%,'(t)] (1.59)
where
(1.60)
(1.61)
Let us describe the temporal variation between the excited state with the
electron in the conduction band and no photon Ic,k;O) and the ground
state with the electron in the valence band and one emitted photon
Iu, k; lq,u)using a time-dependent Schrodinger equation. The wave function
for the system is written under the rotating wave approximation with only
the resonant terms considered as
Let us assume the initial condition where one electron is in the conduction
PROPERTIES
1 OPTICAL OF SEMICONDUCTOR
QUANTUM 23
NANO-STRUCTURES
band and the field is in the vacuum state with a(0) = 1 and b,,(O) = 0. The
electron drops into the valence band by spontaneously emitting a photon
with an energy of oqand a momentum of q. By substituting Eq. (1.62) into
the time-dependent Schrodinger equation of
and
( 1.66)
The decay rate, y d , includes all the dissipation processes besides spontaneous
emission, such as nonradiative recombination at crystal defects or impurities
and scattering into other electronic states. Integrating Eq. (1.65) and then
substituting it into Eq. (1.64), according to Weisskopf and Wigner (1930),
we obtain
(1.68)
24 MITSURUSUGAWARA
where gw,ois the coupling constant at the frequency of o.This result shows
that the coupling to the continuous optical mode always causes an exponen-
tial irreversible spontaneous emission. It can be visualized as an electron
coupling to various optical modes with different frequencies and losing its
memory as soon as it emits a photon due to the interference of the different
optical modes. From Eq. (1.68), we get the spontaneous emission rate as
Z
4
FIG. 1.5. (a) S-polarized mode with its electric vector normal to the plane defined by q. and
the z-axis and P-polarized mode with its electric vector in the plane defined by q and the z-axis.
The UJ plane corresponds to the quantum-well plane. (b) Wave vector of the electromagnetic
field, q = (q,,.4,). The xy plane corresponds to the quantum-well plane. The wave number of
the electromagnetic field emitted by the interband transition is n,q,,k/c, where w , , , ~is the
frequency dillerence between the k states of the conduction band and the valence band, In
exciton emission, q = n,tu,/c, where w.,' IS the exciton energy.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 25
NANO-STRUCTURES
The next step is the summation on the possible optical modes that satisfy
the energy conservation rule. Let us transfer the summation only on 9, into
the integration, using the mode number of (LC/27c)dq,between q, + dqz and
dq,. This is done to clarify the difference between the electron-hole
transition and the exciton spontaneous emission treated in Section IV.5.
Thus, we obtain
where
( I .74)
vector is taken within qll < qc to satisfy the energy conservation rule (see
Fig. lS(b)). Then, by integrating Eq. (1.72) into the quantum-well plane, we
get
(1.75)
( 1.77)
3. RATE EQUATIONS
FOR LASEROPERATIONS
(1.78)
28 MITSURUSUGAWARA
FIG. 1.6. (a) Cavity laser. (b) Schematic plot output power and carrier number versus
injected current in a cavity laser. Above the threshold of If,,.the injected current is transformed
into photons, keeping the carrier number clumped. The nonlinear gain causes gain saturation
at high output power, leading to a decrease in the quantum efficiency and an increase in the
carrier number.
and
(1.79)
where ai is the internal loss. Carriers injected into the active region
recombine radiatively with the lifetime of rsp or nonradiatively with the
lifetime of z,, giving the total lifetime as
Fabry-Perot lasers, since the separation of the cavity mode is usually less
than 1 nm.
The output power is proportional to the cavity-photon number and is
given as
where ga)' is the differential optical gain and N , , is the carrier number for
the transparency where the gain is zero. This formula is often used to express
the optical gain in bulk semiconductor lasers. For quantum-well lasers, an
empirical formula using a logarithmic dependence of carrier is usually
employed (Chuang, 1993). The solutions are
N = z,I/e (1.85)
and
s=o (1.86)
(1.87)
30 MITSURUSUGAWARA
and
above the threshold. The threshold current is given from Eqs. (1.87) and
(1.88) with S = 0 as
(1.89)
Figure 1.6(b) is a schematic graph of the output power and carrier numbers
as functions of the injected current. The carrier number is clumped above
the threshold, since the injected electrons are transformed to photons
through stimulated emission. An increase in the injection current above the
threshold of Eq. (1.89) leads to an almost linear increase in output laser
power until its saturation occurs. The saturation of power and the increase
in carrier number are due to the nonlinear gain saturation effect through the
third-order nonlinear gain coefficient.
A procedure to design strained quantum-well lasers with low-threshold
currents is as follows. First, the band dispersion curves for the conduction
and valence bands are calculated by means of the k * p perturbation method,
the curves are quite sensitive to the amount of strain and the well width.
Then the spontaneous emission lifetime (z), and the optical gain are
calculated as a function of carrier density through Eqs. (1.42) and (1.76).
Next, the strain and well-width dependence of the threshold currents are
obtained. In strained InGaAs/InP quantum-well lasers, it has been shown
that either direction of strain -that is, bi-axial compressive or tensile -
lowers the threshold currents (Sugawara and Yamazaki, 1994).
In this simple description of laser operation, it is assumed that all of the
carriers are injected into the active region without a time delay. This is not
a good assumption, especially in quantum-dot lasers, since the carrier
relaxation is retarded due to the complete discrete level formed in quantum
dots. The laser simulation including retardation of carrier relaxation is
presented in Section V.l.
1. STATEVECTORS
(1.90)
where ak,, is the electron annihilation operator in the valence band, and
k h --ku
- (1.91)
The conduction-band effective mass and the wave vector are renamed for
the electrons as m,*and k,. The position vectors of an electron and a hole
are re and r,, respectively.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 33
NANO-STRUCTURES
(1.92)
To determine the expansion coefficient, Ck,kh, in Eq. (1.92), let us replace the
creation operators of an electron and a hole with the field operator, using
the relationship below (Haken, 1973, for example):
(1.93)
and
r r 1
(1.95)
(1.96)
34 MITSURUSUGAWARA
and
Then, by substituting Eqs. (1.96) and (1.97) into Eq. (1.95), we get the
exciton state vector as
(1.98)
Comparing Eqs. (1.98) and (1.99), it is clear that the Coulomb interaction
makes up the exciton state by summing the electron and hole pair state with
the amplitude of A(k,, k,) around the band edge.
The Coulomb potential between an electron and a hole is given by
(1.100)
where E, is the static dielectric constant. The potential depends only on the
difference between the electron and hole position vectors. Therefore, let us
change the coordinate vectors into the relative motion vector as
r = re - rh (1.101)
R=
+
m$re mtrh
(1.102)
M*
where
M* = m: + mt (1.103)
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 35
NANO-STRUCTURES
K = k, + k,, (1.104)
and
(1.105)
we get
Quantum Wells
(1.107)
and
where ri = (r\I,zi);ki = (k\I, niJ; and i = e, h. Then, from Eqs. (1.95), (1.107),
and (1.108), we get
(1.109)
36 MITSURUSUCAWARA
where $ex(rll, R") is the in-plane envelope wave function separated from the
z-direction component. By transforming the variables, the Fourier compo-
nent is written as
The vectors related to the relative motion and center of motion are all
two-dimensional.
Quantum Dots
(1.111)
and
The integration in Eq. (1.1 13) represents the mixing of different quantized
states via the Coulomb interaction. Without the Coulomb interaction, Eq.
(1.1 13) is reduced to Eq. (1.99).
2. EFFECTIVE-MASS
EQUATIONS
where H ; is the electron kinetic energy; H i is the hole kinetic energy; V(r, r’)
is the Coulomb potential energy of Eq. (1,100) between two particles; and
the field operator working on both the conduction and valence bands is
By substituting Eq. (1.116) into Eq. (1.1 19, the Hamiltonian becomes
where
for the electron, with the energy origin taken at the valence-band top,
(1.119)
( 1.120)
38 MITSURUSUGAWARA
The commutation relation of the field operators are used (Haken, 1973, for
example). V,, is the electron-electron interactions; V,, is the hole-hole
interaction; V,, is the electron-hole interaction; and V,, is the exchange
interaction.
By substituting the exciton state vector of Eq. (1.95) into the effective-
mass equation of Eq. (1.1 14), we get the exciton effective-mass equation as
m:e4
E,, = E, -
32rc2h2&;n2
+-k2M*
2K2
(1.122)
(1.123)
(1.125)
Here, several other cases are briefly described: quantum wells, quantum
wells with in-plane confinement potential; and quantum wells under a
magnetic field.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 39
NANO-STRUCTURES
Quantum Wells
It is assumed that quantum wells are type I; that is, both an electron and
a hole are confined in the same well layer. Though the asymmetry of the
effective mass is taken into account for the direction perpendicular to and
parallel to the well layer, band nonparabolicities are neglected. The super-
scripts of 11 and I are attached to the effective masses and position vectors
to describe the parallel and perpendicular directions, respectively. By adding
the quantum-well confinement potential to Eq. (1.121), an effective-mass
equation results:
where peh = [r"' + (z, - z , ) ~ ] " ~is the distance between an electron and a
hole, and ((ze) and V,(z,) are the z-direction quantum-well confinement
potentials for an electron and a hole, respectively. When the confinement
potential in the z direction is much greater than Coulomb potential energy,
the wave function can be separated as
This is a good approximation for actual quantum wells with several tens to
hundreds meV potential barriers. The in-plane relative motion and the
center-of-mass motion can be separated. The exciton resonance energy is
given as
(1.128)
where Enpl and En,,=are the electron and hole quantized energies in the
quantum well. The energy of ELm for the relative motion is determined by
( 1.129)
and
40 MlTSURU SUGAWARA
+
where n = 1,2,3,. . . and rn = 0, 1,. . . , +(n - 1). The optically active
exitons are s excitons with rn = 0. For the ground-state optically active 1s
excitons (n = 1 and rn = 0), a variational wave function of
(1.131)
( 1.132)
(1.133)
O(x) is the Heaviside step function (O(x) = 1 for x 3 0 and O(x) = 0 for
x < 0). If the z-direction confinement potential is high enough to confine
excitons in the in-plane two-dimensional regions, Eq. (1.134) is reduced to
approximately
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 41
NANO-STRUCTURES
These are exact when the z-direction potential height is infinite. Thus, the
wave function can be separated as
using the z-direction quantized electron and hole wave functions. The
+ +
exciton resonance energy is given as E,, = Eg Enez En,,= E,. +
The binding energy E , is determined by
Then,
and
(1.140)
where $, is given by Eq. (1.130). For the ground-state s-state optically active
excitons (n = 1 and rn = 01, we solve Eq. (1.141) by a variational method
using the trial wave function of
where a', b', and c' are variational parameters, two of which are independent
because of the normalizing condition. For center-of-mass motion, the
42 MITSURUSUCAWARA
effective-mass equation is
(1.144)
where
(1.150)
(1.151)
The expectation value of the third term (the Zeeman term) and the fifth term
of Eq. (1.150) is zero for optically active S-state excitons. The fourth term,
the diamagnetic energy term, dominates the magnetic-field-induced shifts of
exciton resonance in quantum wells.
INTERACTIONS
3. EXCITON-PHOTON
(1.152)
(1.154)
and
(1.155)
+
The wave vector selection rule is q = k, ke and is already applied for the
interband matrix element of PgSk. Note that the band-edge Bloch functions
of u,(r) for the conduction band and ul,(r) for the valence band are added
in Eqs. (1.154) and (1.155), but were dropped in the effective-mass equations.
Let us derive the exciton-photon interaction Hamiltonian under the
Dirac representation. The exciton state with n photons is
Coupled modes between excitons and photons are not taken into account
here. The Dirac representation for the optical transition between the exciton
state and the ground state is given as
(1.160)
1 OPTICAL
PROPERTIES QUANTUM
OF SEMICONDUCTOR 45
NANO-STRUCTURES
and
Quantum W e h
(1.162)
and
(1.163)
(1.165)
46 MITSURUSUGAWARA
and
Quuntum Dots
(1.167)
and
(1.168)
(1.170)
and
P:; =
kex
P&A(q,kex) % P&
s d3R$,,(0, R)e-'q'R2 P;"
s d3R$,,(0, R)
(1.172)
The last approximation holds when the dot volume is smaller than the
exciton resonant wavelength and the electric dipole approximation holds -
that is, eciq'Rz 1. Under this approximation, the square of the matrix
1 OPTICAL PROPERTIES OF SEMICONDUCTOR QUANTUM 47
NANO-STRUCTURES
ABSORPTION
4. OPTICAL SPECTRA
Let us use a density-matrix equation to derive the formula for the optical
absorption caused by excitons. The procedure is almost same as that used
in deriving linear susceptibility without the Coulomb interaction in Section
111.1. Slight differences are the base functions and the initial condition. The
base functions are taken as the exciton state, le), and the ground state, Is),
where
The system is initially in the ground state. Note that, as the number of
excitons increases, the system is transferred to the electron-hole plasma
state, making it impossible to use the exciton base functions. The classical
expression for the electromagnetic field of Eq. (1.21) is used. The kinetic
energy term is
H, +
= ~cJ~ele)(el ho,Ig)(gl (1.174)
with
+
,ueg(t) = +[peg(o)e-iw' peg(-o)e'"'] (1.176)
where
(1.177)
and
(1.178)
48 MITSURUSUGAWARA
(1.180)
where we, = oj, - 01,. The polarization of the system is given by the
diagonal trace as
(1.182)
is used. The summation over the exciton states is also taken into account.
From its imaginary part, the optical absorption coefficient is given as
(1.18 3)
Comparing Eq. (1.142) and Eq. (1.183), the exciton effect on the optical
absorption spectra appears in the transition matrix of P:; and the transition
energy of a,,.These are determined by the exciton effective-mass equations,
The optical absorption spectrum consists of a series of resonance, the
continuum spectrum whose intensity is affected by the Coulomb interaction,
and the smoothly connecting part in between. The enhancement ratio of the
continuum spectrum strength and the strength of the noninteracting elec-
tron-hole pair is known as the Somrnerfeld factor. The Coulomb effect on
the optical absorption spectra is schematically summarized for one- two,
and three-dimensional cases by Ogawa (1995).
For quantum wells, using Qd = (DLQw)-', Eq. (1.1 85) becomes
(1.185)
( 1.186)
B,(hw - hw,,) =
hLg/n (1.1 87)
( h -~ +
When probing light is applied perpendicularly to the quantum-well layer,
ql1= 0, and thus KII = 0 from the wave vector selection rule (see Eq. (1.166).
The exciton optical absorption strength is proportional to Id(rl1= 0)l2,
which represents the probability of finding the electron and the hole at the
same position. For the ground-state excitons in quantum wells, Ig5(rl1= 0)l2
is inversely proportional to the square of the two-dimensional exciton
radius-that is, to as seen in Eq. (1.131). Equation (1.187) shows that
the exciton optical absorption spectrum has a Lorentzian profile, with an
FWHM of 2hTeg, which increases with an increase in temperature due to
scattering by phonons.
Figure 1.7 shows the calculation of the ground-state exciton resonance
characteristics for various 111-V and 11-VI semiconductor quantum wells
(Sugawara, 1992): the well-width dependence of (a) radius, ,Iex; (b) binding
energy, E,; calculated by Eq. (1.126); and (c) integrated intensity, Sex,
calculated by Eq. (1.185). The exciton radius decreases as the well width
decreases due to the quantum-confinement effect, and increases in wells
narrower than 2 nm due to the breakdown of the confinement effect -that
is, the spread of the z-direction quantized-state wave function to the barriers.
In conjunction with this, the binding energy increases as the well width
decreases to about 2nm, and then goes down. The variation in the
integrated intensity shows almost the same tendency as the variation in the
binding energy. The decrease in narrow quantum wells is due to both the
increase in the radius and the decrease in the electron-hole overlapping
integral. Figure 1.7(d) shows the band-gap dependence of the integrated
intensity in 5-nm quantum wells. The dashed line serves as a visual guide.
Wider-gap semiconductors have a smaller radius, a greater binding energy,
50 M ITSURu SUGAWARA
'ZnMnSsRnSe
-
210 , , , , s , , , . r . 2
I
8
s ! . s! ZnSelZnMn%sQ
X
Y ZnSaRnMnSe
z 6 -
.-
1:-
tn
5 4- CdZnTmTm /
.-
CI
C
9'
C'
J&aPIIIGaInP-
cn
C InGaAsilnP
c
'0
C
-
Y
0
C
I
2.
.
d""
*.g
GaAa(AIGaAs .
d%GaAallnP
FIG. 1.7. Characteristics of 1s exciton resonance calculated for various 111-V and 11-VI
semiconductor quantum wells (from Sugawara. 1992). (a) Well-width dependence of the radius,
&. (b) Well-width dependence of the binding energy, E,. (c) Well-width dependence of the
integrated intensity of the optical absorption spectra. (d) Band-gap dependence of the
integrated intensity in 5-nm quantum wells. The dashed line serves as a visual guide.
and stronger resonance. This is due to their larger effective masses and
smaller dielectric constants.
Though the quantum wells used in the calculation in Fig. 1.7 have
actually been grown, so far there are no systematic experimental analyses of
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 51
NANO-STRUCTURES
which is proportional to the square of the magnetic field (dashed line). The
approximation of the low-field limit holds at most up to 1 or 2 Tesla. Under
higher magnetic fields, diamagnetic shifts are no longer proportional to the
square of the magnetic field and deviate from the straight line, indicating
that the in-plane exciton wave function shrinks (Aex decreases) under the
magnetic field.
The integrated intensity of the le-hh exciton resonance is plotted as a
function of the magnetic field in Fig. 1.8(c). Here, the curves are the
calculation of Eq. (1.185). Actually, the calculation was done in wave vector
space and band nonparabolicities were taken into account. The second-
order k * p term in Eq. (A. 21) is set at D' = - 6 as a fitting parameter, giving
52 MITSURU SUGAWARA
3- ----4.2Tesla .\
4l '
8 .- -7.OTesla ' \
--.x=0.35, Lz=9.7nm
"
1.40 1.& 1.!ill
(a) Wavelength (pn) (b) Wavelength (pm)
2 0 h . I - , . , . A
E
2
2.
i
u)
u
OO 100 200 300 400
(c) Temperature (K)
FIG. 1.9. Experiments on exciton optical absorption spectra (from Sugawara et al., 1990.
Copyright 1990 by The American Physical Society). (a) Measured (solid line) optical absorp-
tion spectrum of In,,,,Ga,~,,As/InP quantum wells at 4.2 K. The dashed line is Eq. (1.190)
with an FWHM of 4.7meV. (b) Measured (solid line) optical absorption spectrum of
In,,,,Ga,,,,As/lnP quantum wells at 295 K. The dashed line is calculated by Eq. (1.189) with
r,' = 150 fs. (c) FWHM of the ground-state electron-heavy-hole exciton resonance spectrum
as a function of temperature for three samples with almost the same composition and well
width (10-nm In,~,,Gao~,,As/lnP quantum wells). The solid curves are calculated by Eq.
(1.189),assuming the temperature-dependent phonon scattering rate of Eq. (1.191). The dashed
line is the thermal broadening due to Eq. (1.187).
with the LO phonon energy of hw,, = 30 meV and r;,,' = 65 fs. The dashed
line is the FWHM of Eq. (1.187). Calculated values for the three samples
are in good agreement with measurements, supporting the assumption
1 OPTICAL PROPERTIES OF SEMICONDUCTOR NANO-STRUCTURES55
QUANTUM
5. SPONTANEOUS EMISSION
OF PHOTONS IN QUANTUM
WELLS
AND MESOSCOPIC
QUANTUM DISKS
Let us consider a quantum a well with well width, L,, and an exciton
resonance wavelength, Leg, that satisfies the following relation:
Here, uB is the exciton Bohr radius in bulk materials and L2 = D is the area
of quantum-well crystals in the x-y plane. Under this condition, electron and
hole energies are quantized in the z direction, and excitons move freely in
the x-y plane. We also assume that the length of the crystal in the z
direction, L, is much larger than A,,. The electromagnetic field is quantized
in the crystal with the volume of V, = L3 and with the refractive index of n,.
The basic rules to derive a formula of exciton spontaneous emission
lifetime in quantum wells are as follows: (1) The selection rule of the in-plane
wave vector is taken into account between the electromagnetic field, 411, and
the exciton center-of-mass motion, KII, that is, KII = qll (Eq. (1.166)); (2) only
excitons with the wave vector within a critical value, qc, can radiate
spontaneously because of the energy conservation rule. The emission is in
the direction of (q”, q,), with q, = ,,/(n,o,g/c)2 - q1Iz.This is illustrated in
Fig. 1.5(b).
The procedure to derive the formula is almost identical to the one used
for the electron-hole pair in Section 111.2. The wave function for the
exciton-photon system is written as
1 OPTICAL
PROPERTIES NANO-STRUCTURES57
QUANTUM
OF SEMICONDUCTOR
where I@,,,O) is the exciton state with no photon and lag,lq,b)is the
ground steate with one photon. We assume the initial condition where one
exciton with the resonant frequency of we, is in a quantum well and the field
is in the vacuum state with a(0) = 1 and bq.b(0)= 0. The exciton drops into
the ground state by spontaneously emitting a photon with an energy of wq
and a wave vector of q. For the exciton-photon interaction Hamiltonian,
Eqs. (1.159) and (1.165) with n = 0 are used. By substituting Eq. (1.193) into
the Schrodinger equation of Eq. (1.63) and combining the equations of
motion for a(t) and b,,(t), we get
(1.195)
The rate, yd, includes all the dissipation processes besides the spontaneous
emission, such as acoustic phonon scattering into nonradiative states in K
space, ionization due to longitudinal optical phonons, and nonradiative
recombination processes at crystal defects or impurities.
As long as the exciton interacts with the continuous optical mode, Eq.
(1.194) is reduced to Fermi's golden rule, as seen in Eq. (1.68). This is also
the case for excitons in quantum wells, which interact with a definite optical
mode in the in-plane direction and with continuous optical modes in the z
direction. So, we again get Fermi's rule as
4t)
b(t) = - 'z,[ + yd] 2 (1.196)
with
Since exciton states are formed mainly from the states around the inverse of
the relative motion radius, kll is replaced by’,1 in the polarization factor.
Considering the optical-mode density in the z direction, as in Section 111.2,
we obtain the spontaneous emission rate as
where ylI and q 2 are given in Eqs. (1.73) and (1.74). Comparing Eq. (1.72)
for the electron-hole pair and Eq. (1.199) for the exciton, the most
remarkable difference is the wave-vector selection rule.
The spontaneous emission spectrum is given as
( 1.200)
FIG. 1.10. Calculation of the exciton spontaneous emission lifetime (from Sugawara, 1995.
Copyright 1995 by The American Physical Society). (a) Quantum disk with a width L,, which
is less than 2u,, and a radius R,. The origin of the coordinate axis is the center of the disk,
and the z-axis is perpendicular to the layer. (b) Calculated spontaneous emission lifetime (solid
line) of the optically active ground-state excitons as a function of w, and R, by Eq. (1.208). The
disk is In,~,,Ga, 4 7 A with
~ a width of Lz = IOnm, and the surrounding barrier is InP. As R,
exceeds L,,/2n,, the lifetime gradually increases from a constant value. Under a strong
confinement of R , < A,,, the lifetime increases by two orders of magnitude. The horizontal
dashed line is the lifetime of free excitons with K" = 0 in quantum wells (23 ps), calculated from
Eq. (1.199), and the dotted line is that calculated from the electric dipole approximation of Eq.
(1.209).The lifetime calculated from Eq. (1.208) smoothly connects the two limiting cases.
60 MITSURUSUGAWARA
(The subscripts for the quantum numbers are added for clarity.) The
expansion is done in the whole crystal. From the periodic boundary
condition, KII = 27c(n,/L,n,/L), where ni is an integer (i = x and y). Note
that Eq. (1.201) is the superimposition of free-exciton wave functions with
various wave vectors. For the optically active excitons with 1 = 0 (see Eq.
(1.204) below), the Fourier coefficient in Eq. (1.201) is given as
p exp( - fl2K1I2/4)Lk(p2;Il2)
~
(1.202)
k!
(1.204)
Note that in & I , ~ , I Eq. (1.198), for free excitons is replaced by lCk,(qll)lz.
Using the same procedure as before, the spontaneous emission rate on the
basis of Fermi's golden rule is given as
Using Eq. (1.202) for optically active excitons with 1 = 0, Eq. (1.207)
becomes
(1.210)
where AEf, = E i - Et0 and AEL", = E*,, - E\o. The denominator repre-
sents all the carrier distribution, and the numerator represents the carrier
distribution only to the optically active states. The lifetime increases as the
temperature increases due to the distribution of excitons to the inactive
states. As the quantum-disk radius decreases, the lifetime becomes insensi-
tive t o temperature due to the increase in the interlevel separation. In
quantum dots with an energy separation much larger than k,T = 25 meV,
the lifetime is kept almost constant up to room temperature.
10
2.
.-
c
u)
2
c
.-
8
5 05
0c
.g-
0
I
0
E o
1 42 146 150
(4 Wavelength (pn)
Free excitm
(c)
0; A 40 M,
Temperature (K)
' eb ’A
(b) Magnetic field (Ta)
FIG. 1.11. Experiments on exciton spontaneous emission (from Sugawara, 1995. Copy-
right 1995 by The American Physical Society). (a) Photoluminescence spectra of 10-nm
In,,,,Gao,,,As/InP quantum wells resulting from the spontaneous emission of the ground-state
electron-heavy-hole excitons. (b) Diamagnetic shifts of exciton spectra up to 12 Tesla for
photoluminescence at 2,4.2,40, and 100 K and optical absorption at 4.2 K. The magnetic field
is applied perpendicular to the quantum-well layers. The error bars represent the shifts of
self-assembled In,~,,Ga,,,As quantum dots in Fig. l.l(b). Values of w, are indicated in Fig.
l.lO(b). (c) Peak intensity of photoluminescence as a function of temperature. The intensity of
localized excitons (below 20 K) is almost constant; the intensity of free excitons decreases as
temperature increases. The exciton localization significantly enhances the spontaneous emission
lifetime.
V,II -
e2 ( e 2 B 2 I rn:lk.o~)
+-
4 7 - c ~ 8m,*11
~ ~ ~ ~2 ~ ~
]
rli2 Yr = E,Y, (1.214)
(1.215)
radius increases and the exciton's in-plane coupling to the photon mode
becomes stronger. When the disk radius is much larger than the exciton
resonant wavelength and the cavity photon decay rate is smaller than the
exciton-photon coupling constant, the exciton-photon system transits from
the weak coupling to the strong-coupling regime, leading to oscillating
spontaneous emission decay. The temporal evolution of the emission spec-
trum is evaluated to show coupled-mode spectrum splitting.
Figure 1.12(a) shows the coordinate system, where the x- and y-axes lie
in the mirror plane and the z-axis is normal to the mirrors. The mirrors are
assumed to be spaced a distance, L , apart, and have an area, D = L2, where
the length of mirrors in the x and y directions is infinite compared to the
exciton resonant wavelength, ;leg. The planar cavity is assumed to be ideal
in the sense that it has a reflectivity, R,, that is constant not only as a
function of wavelength but also as a function of the incident angle. The
radiation field is quantized in the cavity with a volume of V , = L L 2 and a
refractive index of n,. The quantum disk is placed in the middle of the cavity
with its center at z = 0.
Figure 12(b) represents the wave number space of the cavity-disk system.
Due to the resonant condition, only excitons with an in-plane wave number
less than q, = n,o,,/c can radiate spontaneously. When the wave number of
the cavity mode in the z direction is fixed at q, and the resonant condition
of weg= c(ql12 + qZ)'i2/n, is satisfied, the excitons emit photons in the
direction of (K1l,qz). For real Bragg mirrors with alternate high and low
refractive indices, the photon emitted in the direction outside a critical angle
connects with the electromagnetic-field background in free space. This
means that excitons with the in-plane wave vector below a critical value, qs,
are connected into the cavity mode and excitons with qs < K d y, are
connected into the open-side mode.
It is assumed that the photons repeatedly reflected between the mirrors
interact with the disk every time they pass the middle of the cavity. This
assumption is particularly important when the cavity photon proceeds in
the declined direction and the disk radius is smaller than the resonant
wavelength. Here, to satisfy this condition, it is assumed that disks are
packed in the z = 0 plane to the extent that the distance between neighbor-
ing disks is significantly smaller than the exciton resonant wavelength.
However, the overlap of exciton wave functions between neighboring wells
is still assumed to be negligible. The photons emitted in the declined
direction due to the nonzero in-plane wave vector component depart from
the initial disk and interact with another one. This is the exciton transfer
between the disks in the in-plane direction.
The inclusion of the cavity effect in the exciton decay problem is not
straightforward, although the basic strategy to expand the exciton wave
68 MITSURUSUGAWARA
function in K" space need not be changed. First, the z-direction optical-
mode density should be changed from the continuous one in a free space to
a definite one in a cavity. Second, the formula for the coupling constant
should be modified to include the standing-wave electric field in the cavity.
Finally, instead of the simple Fermi's rule, a general expression that is valid
regardless of coupling strength, should be used, since the coupling strength
varies continuously from the strong-coupling regime to the weak-coupling
regime depending on the disk radius. All of these processes bring consider-
able changes in the resultant formulae, as shown below.
The exciton decay in a planar microcavity is typically characterized by
the following coupling constants: the dipole coupling constant between the
exciton and the cavity-mode photon, y; the decay rate of the cavity-mode
photon via mirror losses, K ; the spontaneous emission rate into the open-
side modes due to the electromagnetic field background, yo; and the exciton
dissipation rate besides spontaneous emission, yd. The photon decay rate is
given as
C
K = ---lnR, (1.216)
n, L’
As shown in Fig. I.l2(b), yo depends on the critical in-plane wave vector, qs,
above which excitons are connected to the open-side mode.
The vector potential in a planar cavity is modified from Eq. (1.59) by
internal reflection and interference. For the S-polarized mode (its electric
vector is normal to the plane defined by q and the z-axis), we obtain
For the P-polarized mode (its electric vector is in the plane defined by q and
the z-axis), we obtain
(1.220)
70 MITSURUSUGAWARA
The wave function for the exciton-photon system is the same as in Eq.
(1.193). Again, the initial condition is that one exciton is in a quantum disk
and the field is in the vacuum state with a(0) = 1 and b,,(O) = 0. The
exciton drops into the ground state by spontaneously emitting a photon
with an energy of ho, and a momentum of hq. Using the rotating wave
approximation, the equations of motion for the probability amplitude of a(t)
and b,,(t) are found as
and
(1.222)
where
Depending on the way the exciton interacts with the electromagnetic field,
Eq. (1.225) gives the following types of solutions:
Decay into the open-side mode in the cavity: The components with the
wave vector in the range of q, < q d q, contribute to this decay with a
rate of yo.
1 OPTICAL OF SEMICONDUCTOR
PROPERTIES QUANTUM 71
NANO-STRUCTURES
Weak coupling with the cavity mode: For K >> g, the emitted photon
escapes from the cavity and is barely reabsorbed by the quantum disks.
In this weak coupling regime, a ( t ) can be removed from the integral.
Thus, the remaining integral gives, for t >> C1,
(1.226)
where
and
( 1.228)
a(t - s)f(s)eCKSI2
ds - (t+ $) a(t) (1.229)
72 MITSURUSUGAWARA
where
where
c(1,2= (y,
- - -+-++
2 2
IC( j ,
2
) +' [(h---
-2 2
ti
2
-
i(jqy ~cP]'" (1.232)
Q2 = 2 1 lgq,aI2 (1.233)
o=P.S
6, = oq- ueg, q = (Kll, qz), and cl.z are numerical coefficients determined
by the initial conditions of a(0) = 1 and h(0) = -yd/2. The factor of two in
Eq. (1.233) represents the two z directions, that is, the front and back sides
of the cavity. For resonant excitons with 6, = 0, Eq. (1.231) becomes
cos(@t/2)+ -sin(Qt/2)]
20 (1.234)
with
(1.235)
From Eqs. (1.221) and (1.222), the coefficient for each optical mode is
given as
_- 1
(1.238)
Numerical Calculations
Let us use the half-wavelength cavity with the cavity space as L' =
Aeg</2nr.By changing ( around unity, the slight detuning from the exact
<
half-wavelength cavity of = 1 is expressed. From Eqs. (1.219) and (1.220),
the cavity mode matching the exciton resonance energy is
2nmn,
q; = (rn = 0,l) (1.239)
i
~
&g
Only the m = 1 mode is taken into account since the rn = 0 mode is coupled
to the exciton with q = q, (>q,) and should be treated as the open-side
decay in a real planar microcavity. For simplicity, the dissipation rate into
the open-side mode is not taken into account here. The spontaneous
emission of the optically active ground-state excitons is calculated for the
In,,,,Ga,,,,As/InP disk with a thickness of L, = 10nm. Free excitons in
quantum wells have a resonant wavelength of Leg = 1.45 pm, a resonant
frequency of weg= 1.30 x l 0 l 5 s-' at 4.2K, and an in-plane radius of
16 nm. The exciton resonant wavelength in the cavity is approximately given
as Aeg/nr2 400 nm. The dissipation rate of yd is neglected so that the cavity
effect can be clearly seen.
14 MITSURUSUGAWARA
( 1.240)
Time (ps)
(a)
FIG. 1.13. Calculation of exciton spontaneous emission in the microcavity (from Sugawara,
1997). ( a ) Probability of ln(t)12 for K = 0 excitons in the 10-nm In, ,,Ga, ,,As/InP quantum
wells as a function of r. The photon lifetimes in the half-wavelength cavity (i = I ) are
ti-’ = 240fs, 8Ofs, 51 fs, 23fs, and 11 fs. (b) Probability of la(f)I2 for K = 0 excitons in the
10-nm In,,,,Ga,,,,As/InP quantum wells when I < - ' = 240fs at a variety of detunings i. (c)
Probability of la(t)12 for the ground-state excitons in the In, ,,Ga,,,,As/InP quantum disks in
'
the half-wavelength cavity (( = 1) when ti- = 2.4 p.
(1.241)
101 ' . ' ' ! , ' ' ' I , ' ' . I ' ' ' ' 1
7. THECOULOMB
EFFECTON OPTICALGAINSPECTRA
That excitons take part in lasing has long been discussed in wide band
gap materials like 11-VI semiconductors with large binding energies (Koch
et al., 1978, Haug and Koch, 1977, Klingshirn and Haug, 1981). This topic
is once again of broad interest, given that green to blue semiconductor
quantum-well lasers have been produced primarily owing to the success of
high-density p-type doping. Room-temperature operation of ZnSe-based
(Haase, 1991) and GaN-based (Nakamura et al., 1996), lasers has had a
great impact on semiconductor optical technology, since short-wavelength
lasers are the key to next-generation high-density optical data storage
systems. Studies of the lasing mechanism in wide-gap quantum-well lasers
have become critical in the creation of practical lasers and in their improved
performance.
Exciton-related lasing in 11-VI wide-gap semiconductor quantum wells
was first suggested by experiments showing that lasing in ZnCdSe/ZnSe
multiple quantum wells occurred around the tail of the free-exciton optical
78 MITSURUSUGAWARA
absorption resonance (Ding et al., 1992, 1993). Ding and his colleagues
proposed a mechanism by which excitons created by photo excitation diffuse
into the low-energy tail of inhomogeneous broadening and cause population
inversion, leading to lasing. Experimental data have since suggested various
kinds of exciton-related lasing mechanisms in quantum wells (for example,
Kuroda et al., 1992; Jen et al., 1993; Kawakami et al., 1994; Masumoto et
al., 1993; Kreller et al., 1995; Kozlov et al., 1996).
We should discount the simple explanation that the large exciton-binding
energy of wide-gap materials stabilizes exciton states, resulting in exciton
lasing. This is because excitons as bound states between an electron and a
hole cannot exist under the population inversion, as the semiconductor
Bloch equation predicts (Haug and Koch, 1994; Uenoyama, 1995). Thus, in
order for bound-state excitons t o generate a gain, the exciton emission
energy should be shifted from the resonant exciton absorption energy
through some mechanism. Possible emission processes are: (1) radiative
recombination assisted by exciton-exciton, exciton-optical phonon, and
exciton-electron (hole) scattering, (2) exciton localization; and (3) bi-ex-
citon-exciton transition. All of these processes need the participation of a
third partner besides the photon and excitons: scattering particles; local
potential minima; bi-excitons. In this sense, exciton lasing occurs in very
special circumstances. See Galbraith, 1995 on the calculation of the scatter-
ing-related mechanism. He showed that, in the ZnSe bulk active layer, the
exciton-electron scattering process can cause a room-temper-
ature optical gain of up to 100cm-’ at a lower carrier density than the
Mott density of about 5 x 1 0 ” ~ m - ~This . calculation suggests that
exciton-related lasing is possible if the cavity loss is well restricted with a
high-quality long cavity and high-reflectivity mirrors. Lasing due to bi-
exciton-exciton transition is discussed in Section V.2.
Though exciton lasing will occur only in special cases, the Coulomb
interaction itself considerably affects optical gain. Here, the semiconductor
Bloch equation is summarized to see how the optical gain spectrum in
Section 111.1 is changed by the Coulomb interaction. The advantage of
Coulomb enhancement of gain for wide-gap laser performance is discussed.
(1.242)
where He, is the unperturbed electron and hole Hamiltonian of Eq. (1.115)
and H I is the Hamiltonian resulting from the interaction with the photons
of Eq. (1.152). The polarization of the system is given as
4
P(t) = - Qd d3r($+(r, t)ee; r$(r, t))
(1.243)
where
( 1.244)
is called a pair function and Qd is the inverse of the crystal volume as defined
before. Here, the two-band model and the k-selection rule are adopted, as
in Section 111.1. The problem is to determine the pair function of Eq. (1.244)
by solving the Heisenberg equation of Eq. (1.242). The four-operator terms
coming from the Coulomb interaction term in Eq. (1.242) are approximately
factorized into the products of the two operators: one two-operator term
describing the carrier density and the other describing the interband pair
function. The product of two operators with different wave vectors vanishes
by taking the time average since they have different rotation frequencies.
The semiconductor Bloch equation thus derived represents the kinetic
motion of the pair function under the quasi-equilibrium condition. Its
80 MITSURUSUGAWARA
Fourier-transformed equation is
(1.245)
where
(1.248)
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 81
NANO-STRUCTURES
I I I I
'E -1 E
0 0
U
0
0 -12
m
r F
v w
.-c -3
d
-5 -28
-0.1 0.0 0.1 0.2 -0.1 0.1 0.3
(a) fi%-Ego (ev) (b) h0-Ea (ev)
FIG. 1.15. (a) Computed TE gain spectra at T = 300 K for 6-nm GaN-AI, ,,Ga,,s6N at
various carrier densities. (Reprinted with permission from Chow et al., 1996. Copyright 1996,
American Institute of Physics.) (b) Gain spectra at the carrier density of 7 x l O " c r ~ - ~
computed with (solid curve) and without (dashed curve) Coulomb enhancement. E,, is the
unexcited bulk material band-gap energy. (Reprinted with permission from Chow et al., 1996.
Copyright 1996, American Institute of Physics.)
The reason for Eq. (1.248) is that, the mixing of different wave-vector
states under the Coulomb interaction, causes QL.,ck to spread in k space
and there is a finite range around k, where Qu,ck can be regarded as almost
constant. The term V, represents the summation of t $ - k / in the area
and has a positive real value. The term V, is the rest of the summa-
tion of Yk-k,Qc,ck(u)on k'. When the Coulomb interaction is zero, that
is, Yk-k'l = 0, only a resonant term of Qc.,ck responds to the electro-
magnetic field. Due to the mixing of different k states through the
Coulomb interaction, the nonresonant k state takes part in the polar-
ization. The k-state range that gives V, depends on the occupation
factor of fr,k - fl,k. In the case of j C . k- h , k = 0, the range is approxi-
mately Ik - k'l < l/Aex, where Aex is the exciton radius. As the occupa-
tion factor increases, the range decreases. When the population inversion
is achieved with L..k - fL.k > 0, the polarization concentrates on the res-
onant k state and V, approaches zero, since the Coulomb interaction
changes the sign and becomes repulsive, as seen in Eq. (1.245). However, V'
still survive, under this condition due to the states with A , k - f;.,k < 0 in the
higher energy region. This is the origin of the Coulomb, or excitonic,
enhancement of gain. Then, from Eqs. (1.245) and (1.248), the pair function
82 MITSURUSUGAWARA
is written as
Using the imaginary part of the polarization derived from Eqs. (1.247) and
(1.249), the optical gain becomes
E
- " Boundkate"
.-0 0 exciton
-
CI
Q
=I
Renormalization
r...\i
a
0
n .E
d Unbound stale
-’4 Energy
FIG. 1.16. Schematic of optical absorption and gain spectra under the Coulomb interaction.
V. Quantum-Dot Lasers
A study of single dot optical properties has been one of the most exciting
subjects in the field of semiconductor quantum-dot research since the advent
of nanocrystals embedded in glass dielectric matrices (Ekimov et al., 1985)
and self-assembled quantum dots on semiconductor substrates. A key to
observing single dot light emission is to reduce the spatial resolution to
about 1 pm or less using, for example, a microscope objective and also to
reduce the number of dots in the observation area to less than, say, a
hundred. Various unique nature of semiconductor quantum dots has been
revealed; a very narrow natural emission linewidth less than 0.1 meV at low
temperatures (Zrenner et al., 1994; Brunner et al., 1994; Grundmann et al.,
1995; Gammon et al., 1996; Empedocles et al., 1996; Marzin et al., 1997), the
broadening and subsequent extinction of the emission line as temperature
increases due to homogeneous broadening (Ohta et al., 1997), an intermit-
tent photoluminescence like random telegram signal explained as a result of
carrier transfer between a single dot and a deep trap state (Nirmal et al.,
1996; Efros et al., 1997). Do these single-dot properties have any correlation
with the performance of quantum-dot lasers? If the area density of self-
assembled dots is 5 x 10" cm in the active layer of a cavity laser with a
~
cavity length of 300pm and a stripe width of 3pm, the number of dots
amounts to 450,000. This dot ensemble have the fluctuation of size and/or
semiconductor compositions, and thus, the fluctuation in the resonance
energy, which is called inhomogeneous broadening. Because of this large
number of hundreds of thousands in the cavity and the energy fluctuation,
people would think that single dot optical properties are masked in the
operation of conventional cavity lasers. It is pointed out here and in the
latest news chapter of this volume that lasing emission spectra are greatly
influenced by the magnitude of homogeneous broadening of optical gain of
single dots, and that the laser performance is never a mere ensemble of
optical properties of single dots.
Also unique in quantum dots is the Coulomb interaction among a limited
number of carriers confined in quantum dots. The quantum-dot discrete
state contains up to two electrons and two holes, and each quantum dot in
the ensemble has a different photo-emission energy depending on the
number of carriers it contains. The ground state of the filled two-electron
and two-hole state with opposite spin directions has lower energy than the
one-electron and one-hole states have, as shown both theoretically (Hu,
1990, 1996) and experimentally (Kamada et al., 1997), and can be regarded
as a bi-exciton state. Bi-exciton lasing was observed in CuCl quantum dots
(Masumoto et al., 1993).
This section describes how the carrier relaxation lifetime, homogeneous
broadening of the optical gain of single dots, and the bi-exciton effect
manifest themselves in quantum-dot laser operation. First, carrier-photon
86 MITSURUSUGAWARA
rate equations for quantum-dot lasers are derived to simulate the relation-
ship of output power versus injected current and the small-signal modula-
tion response. Criteria for the carrier relaxation lifetime, as well as
inhomogeneous broadening linewidth, dot density and crystal quality, are
clarified in how they achieve low-t hreshold, high-efficiency, high-power, and
high-speed operation. It is shown that the carrier relaxation lifetime forms
a hierarchy in quantum-dot laser performance. Then, the optical gain
formula taking into account both inhomogeneous broadening and homo-
geneous broadening is derived. How lasing emission spectrum varies de-
pending on the magnitude of homogeneous broadening is discussed. After
discussing a framework of bi-exciton lasing, we develop an optical gain
formula for the bi-exciton-exciton transition and modify the rate equations
to describe bi-exciton lasing in quantum dots. Also, we simulate the
spontaneous emission and lasing properties of bi-excitons and we discuss
the effect of carrier relaxation rate and oscillator strength enhancement on
laser operations.
Figure 1.17 is an energy diagram of the laser active region, including the
self-assembled quantum dots of Fig. l.l(b), and the relaxation process of
carriers into the quantum-dot ground state. It is assumed that only a single,
discrete electron and hole ground state is formed inside a quantum dot, and
that a charge neutrality always holds in each dot. The energy axis represents
the energy difference between the conduction-band and valence-band edges.
Holes and electrons are treated together, which is enough to check how the
retardation of carrier relaxation affects laser performance. The injected
carriers diffuse through the separate confinement heterostructure (SCH)
layer, relax into the quantum well, and then relax into the dots. Some
carriers recombine radiatively or nonradiatively both outside (in the quan-
tum-well region) and inside the dots. Above the lasing threshold, carriers in
the ground state emit photons into the lasing mode primarily due to the
stimulated emission process. The associated time constants are diffusion in
the SCH region (TJ, carrier recombination in the SCH region (TJ, carrier
emission from the quantum well to the SCH region (T+J, carrier emission
from the quantum dot to the quantum well (re),carrier recombination in the
quantum well (T& carrier relaxation into the quantum dot (T~), and
recombination in the quantum dot (TJ Needless to say, the discussion
below also holds for SK dots accompanied by much thinner wetting layers.
1 OPTICAL
PROPERTIES OF SEMICONDUCTOR QUANTUM
NANO-STRUCTURES 87
FIG. 1.17. Energy diagram of the laser active region, including the self-assembled quantum
dots of Fig. l.l(b), and the relaxation process of carriers into the quantum-dot ground state
(from Sugawara et al., 1997~).
and
- SIT, (1.255)
(1.256)
Under the charge neutrality in each single quantum dot, we set f, = P and
f, = 1 - P. The carrier occupation probability, P , is determined by the
balance among the rates of carrier relaxation, carrier emission, and photon
emission. According to Pauli's exclusion principle, P is related to N as
P = N/(2NDI/,) (1.258)
where r o 1 is the relaxation rate when the ground state is unoccupied, that
is, P = 0. From Eqs. (1.56) through (1.58), the third-order optical gain in
quantum dots is written as
where
(1.261)
(1.262)
1 OPTICAL OF SEMICONDUCTOR
PROPERTIES QUANTUM 89
NANO-STRUCTURES
This form is often used in laser simulation to avoid negative gain. The
third-order nonlinear coefficient is
(1.263)
( 1.264)
is used.
Using a set of the above formulae and setting the time-dependent terms
of Eqs. (1.252) through (1.255) to be zero, we can calculate steady-state laser
operation as a function of the injected current. For simplicity, carrier
emission from the quantum dots is ignored; that is, Z/r >> N/T,. This
assumption will hold when the quantum-dot energy state is deep enough
compared to the thermal energy state. The parameters used are as follows.
The dots have a cylindrical shape with a radius of 7 nm and an equal height,
ho,,. = 1 eV, considering InGaAs quantum dots; three dot layers are grown;
the total optical confinement factor is r = 4%, R , = 30%, R , = 90%0,
M = 5 cm-', and L,, = 300ym; and the stripe width is 2pm. The photon
'
lifetime of the cavity is z P = 4.4 ps. The threshold gain is 670 cm from Eq.
(1.83). The spontaneous emission lifetime in quantum dots is calculated to
be z S p= 2.8 ns using Eq. (1.210) with the exciton effect neglected. The time
constants for nonlinear gain are re;' = 100 fs and rli'= z P .
Figure 1.18(a-c) shows the output power as a function of the injected
current using T~ as a parameter. The spectrum broadening is To = 20 meV,
and the coverage is 4 = 10%. which gives P = 0.84 at the lasing threshold
and T~ = 6.37, from Eq. (1.259). The recombination lifetimes are (a) T, =
T~~ = 2.8 ns and T,, = 3 ns; (b) T , = 2.8 ns and zqr = 0.5 ns; and (c) T , = 0.5 ns
and T ~ = , 0.5 ns. Case (a) corresponds to high quality both in the quantum
dots and the surrounding quantum well; case (b) corresponds to high-
quality quantum dots and a low-quality quantum well; and case (c) cor-
responds to low quality both in the quantum well and in the quantum dots.
The lowest threshold current is 110 pA in case (a) with t o = 1 ps. An
increase in injected current increases the carrier number in the quantum
well, N , , and carriers that relax into the quantum-dot ground state at a rate
of N,/rd are transformed into lasing-mode photons through the stimulated
emission process. The calculation is stopped when the quantum-well carrier
density exceeds 1 x 1012cm-2, because lasing from the quantum well
90 MITSURUSUGAWARA
F
.53
z8 2
CI
P,
0
0
(a) Injected current (mA) (b) Injected current (mA)
2 4 6
Injected current (mA)
occurs around this point and this clumps the carrier number in the quantum
well, N , , leading to saturation of the output power from the quantum-dot
ground state. (We have found experimentally that lasing from the ground
state saturates at high output power and then gradually disappears as the
lasing from the upper levels starts and its output power increases. The
quenching of lasing might be due to the elevated temperature of quantum
dots under high current injection and the resultant increase in the emission
rate, T ~ . )What is remarkable is as follows:
potential. Also, these are enough for our purpose which is to clarify how the
modulation bandwidth is limited by the relaxation lifetime. The resultant
response function is given as
1 Q (1.265)
M(wm)= (1 + o;tj)l’2(1 + c1&rf)”2 - of)’
[(oz, + y om]l / 2
2 2
(1.266)
(1.267)
(1.268)
and
(1.269)
where 0,is the relaxation oscillation frequency, y is the damping factor, and
g‘ is the differential gain. The major factors that limit modulation bandwidth
are relaxation oscillation frequency and rolloff due to retarded carrier
relaxation. The parameter that dominates relaxation oscillation frequency is
the differential gain and thus the spectrum width. The 3-dB bandwidth
limited by the relaxation lifetime of sd is given as
(1.270)
2, m
1 5 = 10%
1.4 mw\ ( P = 0.84)
h
v3 1.5-. ro=20meV -
13.8 mW
20
15
P
9
10
c
OD
'0 10 20 30 40 50
(b) Relaxation lifetime, ro (ps) (c) Output power (mw)
FIG. 1.l9. ( a ) Frequency dependency of the modulation response function with parameters
of T" = lops, To = 20meV, and 5 = 10%. (b) 3dB bandwidth as a function of T" for various
coverages. (c) Relaxation oscillation frequency ( 1;.= <ov/277)as a function of output power using
To as a parameter (from Sugawara et al., 1998).
Q
u Direct modulation lasers
2z
i! - High quality I
V
QD lasers
well lasers emitting in the infrared (1.3 and 1.55 pm grown on InP substrates
have suffered from the low temperature stability of threshold currents,
primarily due to carrier overflow from the active region to the SCH region
with relatively low potential height (Ishikawa, 1993; Ishikawa and Suemune,
1994). Carrier overflow into SCH regions with high density of states
increases the carrier number at lasing and thus accelerates the Auger
recombination process, the rate of which is proportional to the cube of the
carrier density, resulting in the consumption of carriers. At the same time,
large carrier density at lasing greatly increases the internal cavity loss,
resulting in low external quantum efficiency. In self-assembled InGaAs
quantum dots grown on GaAs substrates, carrier overflow is greatly
reduced, since high-potential barrier (SCH) and clad materials like Al-
GaInAs and AlGaInP can be grown on GaAs substrates to prevent carrier
overflow. As a result, quantum-dot lasers on GaAs substrates, especially
infrared lasers for optical transmission systems, will, in many respects, be
superior to strained quantum-well lasers on InP substrates.
If the carrier relaxation lifetime is reduced to 1 ps and, at the same time,
the spectrum linewidth is narrowed to less than 10 meV, quantum-dot lasers
can work as high-speed direct-modulation lasers with almost zero chirp.
Note that the GHz modulation of strained quantum-well lasers is done by
external quantum-well modulators using the quantum-confined Stark effect.
This is because, in strained quantum-well lasers. wavelength chirp is an
inevitable problem for high-speed modulation. A high-speed direct-modula-
tion laser, which was predicted by Arakawa and Sakaki in 1982, might be
one of the goals in quantum-dot laser research.
How long is the relaxation lifetime? To what extent is the relaxation
accelerated? As stated in the introduction to this section, much more work
should be done to find solutions to this problem. In addition to improved
theoretical and experimental studies on carrier relaxation itself, detailed
experimental analyses of quantum-dot laser performance will give a clear
picture of carrier relaxation dynamics. The thinking presented in this section
will help analyze laser operations. Finding a way to accelerate relaxation
through the design of a quantum-dot active layer might be required.
Experimental evaluation of the phonon bottleneck problem is presented in
Chapter 5.
2. EFFECTOF HOMOGENEOUSBROADENING
OF OPTICAL GAIN
SPECTRA
ON LASINGEMISSION
The formula for the optical gain spectrum is presented here, taking into
account both inhomogeneous and homogeneous broadening. Using the
96 MITSURUSUGAWARA
where Ec,, is the average energy, and the FWHM is given as To = 2.35&.
The optical gain in the laser cavity is then given by the convolution integral
of the two broadening factors as
(1.272)
In this case, dots with different energies have no correlation to each other
and gives optical gain independently since they are spatially isolated from
each other. As a result, all dots with an optical gain above the lasing
threshold start lasing emission independently, which leads to broad band
lasing emission as depicted in Fig. 1.21(a). The quantum-dot laser at this
situation behaves in the same way as if it included independent lasing media
in the same cavity.
When homogeneous broadening of the optical gain is comparable to
inhomogeneous broadening, lasing-mode photons receive gain not only
from energetically resonant dots but also from other nonresonant dots that
lie within the amount of homogeneous broadening. Since carriers of non-
resonant dots are extracted and put into the central lasing modes by the
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 97
NANO-STRUCTURES
FIG. 1.21. Relationship between lasing spectra and homogeneous broadening of single-dot
optical gain. (a) If the homogeneous broadening is a delta function, all dots with an optical
gain above the lasing threshold start lasing emission independently, which leads to broad-band
lasing spectrum. (b) When the homogeneous broadening of the optical gain is comparable to
the inhomogeneous broadening, carriers of non-resonant dots are extracted and put into the
central lasing modes by the stimulated emission. leading to lasing spectrum with one central
line.
stimulated emission, lasing emission with one line occurs as depicted in Fig.
1.21(b). Note that the calculation in Sec. V.1 implicitly assumed this
collective lasing.
See our recent experiments on the lasing emission spectra of our orig-
inally-designed columnar shaped quantum dots in the latest news chapter in
this volume, where experimental evidence for the above discussion is
presented and the origin of homogeneous broadening of optical gain is
discussed.
Excitons, in the sense of electron-hole bound states, are not formed under
the population inversion bet ween the conduction and valence bands, and
cannot generate optical gain, as shown in Section IV.7. The Coulomb
interaction affects the optical gain such that the gain’s absolute value is
enhanced, a phenomenon introduced earlier Coulomb, or excitonic, enhan-
cement, and such that the gain spectrum is red-shifted due to screened
exchange energies and Coulomb-hole self-energies. For bound-state excitons
to generate gain, the exciton emission should be shifted toward lower
energies from the exciton absorption resonance through some mechanism.
This becomes possible by the assistance of third partners besides photons
and excitons, such as scattering particles, bi-excitons, or localized states.
The bi-exciton-exciton transition process is one of the possible candi-
dates for exciton-related lasing (Shaklee et al., 1971; Leheny and Nahory,
98 MITSURU
SUGAWARA
1971; Masumoto et al., 1993; Shionoya, 1995; Kreller et al., 1995; Sugawara,
1996, 1997b, 1998; Kozlov et al., 1996). A bi-exciton, analogous to a
hydrogen molecule, is formed through the coupling of a pair of excitons. The
ground-state bi-exciton is composed of two excitons with opposite spin
directions and has a binding energy of B,,, as seen in Figure 1.22(a). The
bi-exciton binding energies increase as the ratio of effective mass of m,*/mt
increases. For example, the binding energy is calculated to be about 0.07EB
at m,*/m: = 0.2 (Akimoto and Hanamura, 1972). The oscillator strength of
Bi-exicton dispersion
4M
2M
Wave number, K
(b)
Exciton
-
Ground state
(c)
FIG.1.22. (a) Three-level model of a bi-exciton. an exciton and a ground state. B,, is the
bi-exciton binding energy. (b) Energy-wave vector dispersion relationships of excitons with a
wave vector of K, and bi-excitons with wave vector of K,. (c) Bi-exciton formation process.
Excitons excited by pumping, whether electronically or optically, collide with each other and
form bi-excitons. A bi-exciton annihilates through spontaneous emission and the stimulated
emission process, leaving one exciton behind.
1 OPTICAL PROPERTIES OF SEMICONDUCTOR QUANTUM 99
NANO-STRUCTURES
is similar to Eq. (1.255), with the nonlinear gain effect omitted for simplicity.
The bi-exciton gain is given as
where gb is the differential gain. The bi-exciton number under the thermal
equilibrium, N:!, is given by the Saha equation (Kim et al., 1994) as
with
(1.278)
(1.280)
(1.281)
c
.-00
CI
lO O
X
-Q 12
c1
lo8
10’
0 50 100 150 200 250 300
Temperature (K)
FIG. 1.23. Total exciton density. ( N , + 2N,,)/D, needed to reach transparency as a
function of temperature, using B,, and T~ as parameters. The horizontal dashed line represents
the exciton saturation density due to the phase-space-filling efiect. (Reprinted from Sugawara,
<Q 1998, with permission from Elsevier Science)
are stable and population inversion is realized at quite low densities. The
temperature range where the electron-hole system stabilizes as bi-excitons
is 10-20 K for B,, = 10 meV and below 30 to 50 K for 30 meV.
The above scenario of bi-exciton lasing seems unlikely to occur in
wide-gap semiconductors. One reason is that the bi-exciton binding energy
is around lOmeV, or 30meV at most, in 11-VI and InGaN materials,
requiring low temperatures for bi-exciton stability. The other is that spatial
energy fluctuations due to composition fluctuations lower the optical gain
at a low injection density, since only a small fraction of the active region
volume can contribute to the gain. This leads to electron-hole plasma lasing
with the local potential minima fully filled with a large number of carriers.
Similarly, quantum dots with large-size inhomogeneity cannot generate
enough optical gain for lasing, so lasing caused by electron-hole plasma
occurs from the quantum well or wetting layers. As long as spatial energy
fluctuations are not controlled to give homogeneous samples, bi-exciton
lasing is not observed.
102 MlTSURU SUGAWARA
1. In quantum wells, the bi-exciton and exciton excited states and many
exciton states appear along with the ground-state excitons. By control-
ling the size and strains of quantum dots and the band-gap discontinu-
ity against surrounding materials, we can design the number of discrete
states in the dots. If there is only one in both the conduction and
valence bands, a pure bi-exciton state is formed when the state is fully
filled with two electrons and two holes. If quantum dots contain more
than two levels, the bi-exciton and exciton-excited states and many
exciton states appear, depending on the number of carriers in the dot.
2. For bi-exciton lasing, bi-excitons should be more stable than excitons,
and the bi-exciton formation should occur faster than the bi-exciton
and exciton annihilation. The bi-exciton stability in quantum dots can
be improved by increasing the quantum-dot potential barrier height.
The counterpart of the bi-exciton formation rate in quantum dots is
the relaxation rate discussed in Section V.1. The rate is controlled to
some extent by designing the quantum-dot energy levels and by
devising injection methods.
3. The Mott density of bi-excitons is a matter to be investigated. We can
experimentally study how the carrier number around quantum dots (in
the quantum well in the model of Section V.1) affects the dots’ binding
energy and oscillator strength, by changing the dot density, the laser
output power, etc.
To take into account the bi-exciton effect in quantum dots, let us use the
1 OPTICAL
PROPERTIES QUANTUM
OF SEMICONDUCTOR 103
NANO-STRUCTURES
three-level model, where the bi-exciton and exciton states are taken as base
functions. The bi-exciton and exciton system is described by the Hamil-
tonian of (Hu, 1990; Hu et al., 1996)
where hwy is the ground-state energy, h o e is the exciton energy, hw, is the
bi-exciton energy, lg) is the ground state, le) is the exciton state, and Ib) is
the bi-exciton state. The transition between the eigenstates of the system
occurs as a result of the interaction of the electromagnetic field with the
wave vector, q, and the polarization, 0. The interaction with a field is
described as
(1.285)
Here, $c(r, R) is the exciton envelope wave function. The second term of Eq.
(1.284) represents the bi-exciton-exciton transition, where the oscillator
104 MITSURUSUGAWARA
strength is given as
where $be(rl, R,, rz, R,) is the bi-exciton envelope wave function and the
subscripts of 1 and 2 represent the first and second excitons. Qualitatively,
Eq. (1.286) shows that the oscillator strength of the bi-exciton-exciton
transition is proportional to the product of the absolute square of the
first-exciton overlap integral, the probability of finding an electron and a
hole at the same position in the second exciton, and the volume covered by
the second exciton. The oscillator strengths of Eqs. (1.285)and (1.286)can
be measured as a ratio against the oscillator strength for the electron-hole
transition of
(1.287)
where a,, is the interband optical transition frequency at the band edge. The
oscillator strength, f: and ,j":, was calculated by Takagahara (1989) as a
function of the radius of the spherical quantum dot. It increases as the dot
radius increases, a phenomenon known as mesoscopic enhancement. Accord-
ing to Takagahara's calculation, j;,/fo2 4, when the radius is equal to the
bulk exciton Bohr radius.
To calculate the population balance between ground, exciton, and bi-
exciton states, let us use the occupation probability, P, defined by Eq.
(1.258). The probability of finding a bi-exciton in quantum dots is P 2 , the
probability for a spin-up exciton is P(l - P), the probability for a spin-
down exciton is P(l - P), and the probability of not finding any excitons is
(1 - P ) 2 . Thus, we obtain
and
pL0,' - p g ) = P(l - P ) - (1 - P), = -1 + 3P - 2Pz (1.289)
The population inversion appears when P > lj2 for both transitions.
Bi-exciton lasing from quantum dots can be described using rate equa-
tions only slightly different from those in Section V.l in the following points.
Considering quantum dots that include only the ground-state exciton and
PROPERTIES
1 OPTICAL OF SEMICONDUCTOR
QUANTUM 105
NANO-STRUCTURES
bi-exciton levels, Eq. (1.284) gives the maximum optical gain at the bi-
exciton-exciton transition as
(1.290)
Again, the Gaussian broadening function arising from the size inhomogene-
ity of quantum dots is assumed to be dominant. The spontaneous emission
rate is given as
where
N
-
rr
=2N,Vu
LXX
-P+2
z, 1
2P(1 - P)
(1.291)
( 1.292)
for bi-excitons ( i = xx) and excitons ( i = x), and the emitted photon fre-
quency is (0. The relaxation rate of Eq. (1.259) is unchanged, since
~d-' =(1 - +
P ) 2 ~ i 12P(1 - P)(tG1/2) = (1 - P ) T ~ ' (1.293)
2NDP(1 - P )
I, = (1.294)
5,
(1.295)
for bi-excitons.
The spontaneous emission and steady-state laser output power as a
function of current can be calculated by the same procedure as in Section
V.l. The parameters considering wide-gap InGaN quantum dots are as
follows. The dots have a cylindrical shape with a radius of 3.6 nm (equal to
the exciton Bohr radius in bulk materials) and an equal height; 5 = 5%,
hw,,. = 3.0 eV, To = 20 meV, and L,, = 300 pm; three dot layers are grown;
the total optical confinement factor is = 4%; R , = 30%, R , = 90%, and
a = 5 cm-'; and the stripe width is 2pm. Since the bi-exciton binding
106 MITSURUSUGAWARA
B
P
I
2 3 4
(a) Current (mA) Oscillator strength, fxx/fo
Fiti. 1.24. (a) Spontaneous emission intensity due to excitons, I , . and bi-excitons, I,,, in
the quantum dot laser as a function of injected currents. The bi-exciton emission intensity
exceeds the exciton emission at a current of I.lmA. The lasing threshold is 1.4mA (from
Sugawara et al., 1998). (b) Maximum output power and the occupation, P , at the lasing
threshold as a function of oscillator strength (from Sugawara et al., 1998).
1 OPTICAL
PROPERTIES
OF SEM~CONDUCTOR
QUANTUM 107
NANO-STRUCTURES
VI. Summary
Appendix
in Eq. (1.20) should be calculated. For the interband transition between the
conduction and valence bands, using Eq. (1.21), we get
From Eqs. (1.2) and (1.4), the electron wave function in the conduction
band is
-
ik, r)u,(r)
where u,(r) is for the conduction band and u,(r) is for valence bands. Thus,
the matrix element is given as
+ q)
=
J
exp[i(k, - k,, * r]d3r
Jn u,*(r)e, pu,(r)d3r
The integration is separated into the product of the integration over the unit
cell and over the entire crystal because the envelope function varies slowly
over the unit cell. Since the Bloch states in the conduction band and the
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 109
NANO-STRUCTURES
Quantum Wells
From Eqs. (1.2) and (1.7), the electron wave function in the conduction
band is
where
I
(c, k,le'q''e,- plu, k,) = d k ~ l , ~ ~ l ~ q l I ~ n , , . f l ,u:(r)e;pu,,(r)d3r
, (A.lO)
(A.ll)
is the overlap integral between the quantized envelope wave function of the
conduction and valence bands. Here, since the well-layer thickness is much
smaller than the wavelength of the electromagnetic field, the electric dipole
approximation of e'Yz"= 1 is used. For the infinite potential well, the
110 Mr~suuuSUCAWARA
In actual quantum wells with a finite potential height, Z,lcz,na,2 is less than 1
even when ncz = nrz. The wave vector selection rule is
kII = kII - qII - kII
3
(A.13)
Quantum Dots
Since quantum dots are much smaller than the photon wavelength, the
electric dipole approximation of e'q" = 1 holds. Thus,
(A.15)
(A.18)
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 111
NANO-STRUCTUKES
(A.19)
for the split-off band resulting from the spin-orbit interaction, where X , X
and Z have p-orbital symmetry. The quantization axis of angular momen-
tum is taken in the z direction. The P-state bases of Eqs. (A.17) through
(A.19), written in ( J , M , ) representation, are taken to diagonalize the
spin-orbit interaction. The electron wave vector is assumed in the z
direction, and the base functions rotate with k vectors. We should first
calculate the matrix element with the z-axis at the k direction and then
transform the element into the xyz coordinate system. Finally, we average
the matrix elements over degenerate k vectors depending on the symmetry
of the crystal. If the system has structural asymmetry, as do quantum wells
and quantum wires, the matrix element has polarization and k-vector
dependence, according to the valence-band type related to the optical
transition.
Bulk Materials
(A.21)
where mo is the free electron mass, 15, is the band gap, A is the spin-orbit
splitting energy, rn: is the conduction-band-edge effective mass, and D'
is the second-order perturbation term (Sugawara, 1993b).
Quantum Wells
The wave vector is (k",n,n/L,). Let 6' be the angle between the z axis and
the wave vector. The character gives the polarization dependence to the
112 MITSURUSUGAWARA
(A.22)
where
and
and
(A.27)
Quuntum Dots
IP&12 = M2 (A.28)
REFERENCES
Akimoto, 0.- and Hanamura, E. (1972). J. Phys. Sot.. Jpr7. 33, 1537.
Andreani, L. C., Tassone, F., and Bassani, F. (1991). Solid Sfcite Commun. 77, 641.
Arakawa, Y., and Sakaki, H. (1982). Appl. Phys. Left. 40, 939.
Arita, M., Avramescu, A,, Uesugi, K., Suemune, I.. Numai, T.. Machida, H., and Shimoyama,
N. (1997). J p n . J . Appl. P h ~ x36. 4097.
Asada, M. (1989). IEEE J . Quantum Electron. 25, 2019.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 113
NANO-STRUCTURES
Bastard, G., Brum, J. A,, and Ferreira. R. (1991). “Electronic States in Semiconductor
Heterostructures,” in Solid State Physics 44, Academic Press, New York.
Benisty, H., Sotomayor-Torres, C. M.. and Weisbuch, C. (1991). Phvs. Reu. B. 44, 10945.
Bockelmann, U., and Bastard, G. (1990). Phys. Rev. B. 42, 8947.
Bockelmann, U., and Egeler, T. (1992). Phys. Reu. B. 46, 15574.
Bockelmann, U., Heller, W., Filoramo, A., and Roussignol, P. (1997). Phys. Rev. B. 55, 4456.
Brunner, K., Abstreiter, G., Bohm, G . , Trinkle, G., and Weimann, G. (1994). Phys. Rev. Lett.
73, 1138.
Cao, H., Jacobson, J., Bjork, G., Pau, S., and Yamamoto, Y. (1995). Appl. Phys. Lett. 66. 1107.
Carlsson, N., Georgsson, K., Montelius. L.. Samuelson, L., Seifert, W., and Wallenberg, R.
(1995). J . Cryst. Growth 156, 23.
Chang, L. L., Esaki, L., and Tsu, R. (1974). A p p l . Phys. Lett. 24, 593.
Chemla, D. S., Damen, T. C., Miller, D. A. B., Gossard, A. C., and Wiegmann, W. (1983). Appl.
Phys. Lett. 42, 864.
Chemla, D. S., Miller, D. A. B., Smith. P. W., Gossard, A. C., and Wiegmann, W. (1984). I E E E
J . Quantum Electron. QE-20, 265.
Chow, W. W., Wright, A. F., and Nelson, J. S. (1996). Appl. Pliys. Lett. 68, 296.
Chuang, S. L. (1991). P ~ J J Reo.
S . B. 43, 9649.
Chuang, S. L. (1995). Physics ufOptoelec,fronirsDeoices. John Wiley & Sons, New York.
Citrin, D . S. (1993). Phys. R ~ I JB.. 47, 3832.
Ding, J., Jeon, H., Ishihara, T., Hagerott, M., Nurmiko, A. V., Luo, H., Samarth, N., and
Furdyna, J. (1992). Phj’s. Reu. Left. 1707.
Ding, J., Hagerott, M.. Ishihara, T., Jeon, H., and Nurmiko, A. V. (1993). Phjx Rev. B. 47.
10528.
Dingle, R. (1975). Festkorperprobleme 15, 21.
Efros, A. L., and Rosen, M. (1997). Phys. Rer. Lett. 78, 1110.
Efros. A. L., Kharchenko, V. A,, and Rosen. M. (1995). Solid State Cornmun. 93, 281.
Ekimov, A. L., Efros, A. L., and Onushchenko, A. A. (1985). Solid State Commun. 56. 921.
Elliot, R. J. (1957). Phys. Rev. B. 108, 1384.
Empedocles, S. A., Norris, D. J., and Bawendi. M. G. (1996). Phys. Rec. Lett. 77. 3873.
Esaki, L., and Tsu, R. (1969). IBM Resetrrch Note. RC-2418.
Esaki, L., and TSU,R. (1970). I B M J . Res. Deiwlop. 14, 61.
Feldmann, J.. Peter, C.. Gobel, E. 0..Dawson. P., Moore, K., Foxon, C., and Elliott, R. (1987).
J . P h w . Reo. Lett. 20, 2337.
Galbraith, I. ( 1995). “Excitonic gain processes in ZnSe,” in Microscopic Theorj of Semiconduc-
tors: Qumtum Kinetics. Confinemen/ rind Ltrsers, World Scientific Publishing Co. Pte. Ltd.,
Singapore.
Gammon, D., Snow, E. S., Shanabrook, B. V., Katzer, D. S.. and Park, D. (1996). Phys. Rev.
Lett. 76, 3005.
Gerard, J.-M. (1995). “Prospect of high efficiency quantum boxes obtained by direct epitaxial
growth,” in Confined Electrons mid Photons: New Physics and Applicutions. Plenum. New
York.
Goldstein, E. V.. and Meystre, P. (1995). ”Spontaneous Emission and Laser Oscillation in
Microcavities.” CRC Press, Boca Raton. FL.
Grosse, S.. Sandmann. J. H. H. (1997). Phys. Ret:. B. 55, 4473.
Grundmann, M., Christen, J., Ledentsov. N. N., Bohrer, J.. Bimberg. D.. Ruvimov. S. S.,
Werner, P.. Richter, U.. Gosele. U., Heydenreich, J., Ustinov, V. M.. Egorov, A. Yu,
Zhukov, A. E., Kop’ev, P. S.. and Alferov, Zh. I. (1995). PIiys. Reu. Lett. 74, 4043.
Haase, M. A. (1991). Appl. Phys. Lett. 59. 1272.
Haken, H. (1973). Quuntenfeldtheorie drs Fesfkurpers. B.G. Teubner, Stuttgart.
114 MITSURUSUCAWARA
Wood, T. H., Burrus, C. A,, Miller, D. A. B.. Chemla, D. S., Damen, T. C., Gossard, A. C., and
Wiegmann, W. (1984). Appl. Phys. Lett. 44, 16.
Yamanishi, M., and Suemune, I. (1984). J p n . J . Appl. Phys. 23, L35.
Yokoyama, H., and Ujihara, K. (1995). "Spontaneous Emission and Laser Oscillation in
Microcavities." CRC Press, Boca Raton, FL.
Zhang, 2. L., Nishioka, M., Weisbuch. C., and Arakawa, Y . (1994). Appl. Phys. Lett. 64, 1068.
Zrenner, A,, Butov, L. V., Hagn, M. Abstreiter, G.. Bohm, G., and Weimann, G . (1994). Phys.
Rev. Lett. 12, 3382.
SEMICONDUCTORS A N D SEMIMETALS. VOL. 60
CHAPTER 2
Mitsuvu Sugawara
OPTIC'AL SEMICONOUCTOR DEVICES LABOKATORY
FUIITSULABORATORIES
LID
JAPAN
ATSUGI,KANAGAWA,
I. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . 117
11. THESTRANSKI-KRASTANOW GROWTH MODE. . . . . . . . . . . . . . . . 119
1. Energy- Balance Model,for Islcind Formation . . . . . . . . . . . . . . 119
2. InAs Island Growfh . . . . . . . . . . . . . . . . . . . . . . . . . 121
3. Multiple-Layer Growth anti Perpendicular Alignment of Islands . . . . . . 125
4. In-Plane Alignment of Ishncis . . . . . . . . . . . . . . . . . . . . 130
111. CLOSELY STACKED InAsJGaAs QUANTUM DOTS. . . . . . . . . . _ . _ _ 132
1. Close Stacking of InAs Islunt6 . . . . . . . . . . . . . . . . . . . . 133
2. Photoluminescence Properties . . . . . . . . . . . . . . . . . . . . 137
3. Zero-Dimensional Exciton Corlfinenwnt Evaluated by Diamagnetic Shifts. 1 140
I V . COLUMNAR InAsJGaAs QUANTUM DOTS. . . . . . . . . . . . . . . . . 143
v. S U M M A R Y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I52
I. Introduction
1. ENERGY-BALANCE
MODELFOR ISLAND FORMATION
FIG. 2.1. Three types of growth mode on a substrate; Frank-van der Merwe (FM);
Volmer-Weber (VW); and Stranski-Krastanow (SK). In the F M mode growth proceeds layer
by layer. In the VW mode growth occurs as three-dimensional islands on the substrate. The
SK mode is a combination of the FM and VW modes, where the growth of a several-
monolayer-thin film, called a wetting layer. is followed by cluster nucleation, leading to island
formation.
islands on the substrate, if the film has a higher surface energy than that of
the substrate. SK growth is a combination of the FM and VW modes, where
the growth of a several-monolayer thin film, called a wetting layer, is
followed by cluster nucleation and then to island formation. Which growth
mode occurs depends primarily on the difference in the surface energy
between the substrate and the grown material and on the strain energy
accumulated in the grown materials as a result of lattice mismatch. In the
strained-layer growth of semiconductors discussed here, the SK growth
mode has been found to dominate.
Why islands form in the lattice-mismatched epitaxy can be understood by
a simple model of energy balance (Tu et al., 1992). Let us compare the total
energy between a coherently strained film on a substrate and strain-free
islands made by the same number of atoms as in the film. The total energy
is supposed to consist of the strain energy due to lattice mismatch and the
surface energy. The strain-free islands are, for simplicity, assumed to be
cubic shaped with a side length of X . Instead of releasing the strain energy,
the cubic islands have greater surface energy due to the increased surface
area. The island energy is lower than the film energy, then, when the island
is larger than a critical value determined by the surface energy, y , and the
in-plane strain, E, as
varied between 470 and 560°C, and was 650°C for the growth of the buffer
layer. The substrate temperature was monitored by a pyrometer calibrated
by the melting temperature of In-A1 alloys.
The growth rates of InAs and GaAs were approximately 0.1 and 0.75
pm/h. In other words, it takes about 10 seconds to grow one monolayer
(ML) InAs and about 1.3 seconds to grow one ML GaAs. The arsenic
pressure used for InAs island growth was 6 x Torr and was 1.2 x
Torr for buffer-layer growth. These values were measured by an ion gauge
at the substrate position. Two arsenic cells were used, both set at 6 x lop6
Torr. The arsenic pressure was changed abruptly at the interface by
switching one of these cells on and off. After 60-second annealing of the InAs
islands, GaAs layers were overgrown.
The variation of the RHEED pattern and the diffraction intensity are
shown in Fig. 2.2. During the growth of the buffer layer, the surface was
FIG. 2.3. AFM images of the islanding surfaces when the InAs supply was 1.3, 1.6, 2.1, and
2.6 MLs. The islanding growth starts at the 1.6-ML supply, as seen in the change in RHEED
pattern. As the supply amount increases, the dot density rapidly increases and the dots coalesce.
124 YOSHIAKI
NAKATA
AND YOSHIHIRO
SUGIYAMA
increased, the island density rapidly increased, and the islands began to
coalesce when the amount of InAs exceeded approximately 2.3 ML.
Figure 2.4 shows the in-plane base diameter and the density of the islands
as a function of the amount of InAs supply (InAs nominal thickness (ML)).
As depicted in the figure, growth proceeds from the two-dimensional growth
of a wetting layer to the nucleation of InAs islands at about a 1.6-ML
supply, to the increase in the island density, and finally to the coalescence
of neighboring islands at about a 2.3-ML supply.
Figure 2.5 shows the base diameter and density of islands grown with the
2.1-ML InAs supply as a function of the growth temperature. The AFM
images of islands grown at each temperature are also shown. As temperature
increases from 470 to 56OoC,the island diameter increases from 15 to 45 nm
and the density decreases correspondingly. This temperature dependence is
due to the surface migration lengh of TnAs, which increases as the tempera-
ture increases.
Figure 2.6 shows the plan view (a) and the (1 10) cross section (b) of TEM
images for InAs islands grown with a 2.5-ML InAs supply. Observable were
lo'*
Ts =
510°C
540°C
125
0
450 500 550 600
Growth temperature ( "C) 560°C
two types of islands-small, with a diameter of 15-20 nm, and large, with
about twice the diameter -which were formed probably by coalescence.
While the height of most islands is 3 to 5 nm, that of the large-size islands
exceeds 10nm. Contrasts due to strain were observed both in the GaAs
buffer layer and in the GaAs overlayer, and they extended as the dot size
increased. There are two types of defects: V-shaped dislocations and multiple
stacking faults. In the upper middele part of Fig. 2.6(a), a large island
accompanied by line-shaped defects on both sides could be seen. These
defects are a V-shaped dislocation pair lying on two equivalent { 11l } planes.
Note that both types of defect are generated from large islands and almost
no defects are observed around the small islands (Ueda et al., 1998).
3. MULTIPLE-LAYER
GROWTH
AND PERPENDICULAR
ALIGNMENT
OF ISLANDS
The InAs islands buried in GaAs work as quantum dots because they are
smaller than the exciton wave function both in the in-plane and perpendicu-
lar directions. When we apply the TnAs island layers to the active region of
126 YOSHIAKI AND YOSHIHIROSUGIYAMA
NAKATA
FIG. 2.6. (a) the (001) plan-view and (b) the (110) cross-sectional TEM images of InAs
islands (2.5-ML InAs coverage). Observed are small islands with a diameter of about 20nm
and large islands with about twice that diameter formed by the coalescence of two islands.
While the height of most islands is 3 to 5 nm, that of the large islands exceeds 10 nm. Contrasts
due to strain are observed in both the GaAs buffer layer and the GaAs overlayer, and they
extend as the dot size increases. There are two types of defect to be observed: V-shaped
dislocations and multiple-stacking faults.
2 MOLECULAR
BEAMEPITAXIAL
GROWTH 127
FIG. 2.8. In-plane diameter and density normalized against the first-island-layer density as
a function of the intermediate-layer thickness. When 1.8-ML InAs islands were grown using
LO-nm thick intermediate GaAs layers, the average diameter of the 10th-layer islands was about
45m, 90% larger than the first-layer islands, with a 24-nm diameter. The island density
decreased from 1 x 10" em-' for the first-layer islands to 3 x 10" cm-' for the 10th-layer
islands. As the thickness of the intermediate layer was further reduced below 10 nm, the increase
in size and the decrease in density became less remarkable.
PL
77 K
Stacked
island layers
FWHM:
90 meV
ALIGNMENTOF ISLANDS
4. IN-PLANE
1. CLOSESTACKING
OF InAs ISLANDS
Figure 2.10 shows the growth sequence for the close stacking of InAs
islands. The amount of InAs supply for island formation was fixed at about
1.8 ML, and the nominal thickness of the GaAs intermediate layers was set
at 2 and 3 nm. Prior to and following InAs island growth, the sample was
annealed for 2 minutes and 1 minute, respectively. The growth rates, arsenic
pressure, and growth temperature were the same as in Section 11.1.
Figure 2.1 1 shows the RHEED-pattern intensity transition, observed at
the area indicated by an arrow in the inset patterns, during the growth of
InAs islands. The RHEED shows the streak pattern for the two-dimensional
growth in the early stage, and the change to the spot pattern for the
three-dimensional growth at above a critical amount of InAs supply. The
SK-mode islands also grow for the 3rd and 5th layer. Note that the 3rd- and
5th-layer islanding started when the growth of InAs reached about 1 ML-
about 63% of the 1st-layer islanding (approximated 1.6 ML). The reason for
the smaller critical amount for the islanding is thought to be that the strain
induced by the lower-layer islands accumulates InAs preferentially or that
segregation of InAs from lower islands. The transition of the growth mode
from two- to three-dimensional and the existence of wetting layers, as will
be shown in the TEM image (Fig. 2.14(a)), both indicate that SK-growth
islands were formed even on such thin GaAs intermediate layers.
FIG. 2.10. Growth sequence for the close stacking of InAs islands. The amount of InAs
supply for island formation was fixed at about 1.8 ML, and the nominal thickness of the GaAs
intermediate layers was set at 2 and 3nm. Prior to and following InAs island growth, the
sample was annealed for 2 minutes and 1 minute, respectively.
134 YOSHIAKINAKATA
AND YOSHIHIROSUGIYAMA
0 10 20 30 40
Time (s)
FIG. 2.1 1. RHEED pattern intensity transition observed at the area indicated by an arrow
in the inset patterns during the growth of InAs islands. The RHEED pattern shows that the
SK island growth occurred also for the 3rd and 5th layer. Note that the 3rd- and 5th-layer
island started when the growth of InAs reached about IOML, which was about 63% of the
1st-layer islanding.
The island size and the density at each layer were evaluated by ex-situ
AFM. Figure 2.12 shows AFM images of the islanding surfaces at the (a)
lst, (b) 3rd, (c) 5th, and (d) 10th layers stacked with 3-nm thick intermediate
layers. These images refer to different epilayers grown under the same
conditions. The scanned area is 250 x 250 nm'. The upper layer islands
expanded slightly as the number of stacked layers increased. Figure 2.13
shows the dependence of island size on the number (diameter) and density
of stacked layers. The average diameter of the 10th-layer islands was about
33 nm-40% larger than the 1st-layer islands, which were about 24 nm in
diameter-while the island density decreased from 1 x loi1 cm-' for the
1st-layer islands to 8 x 10'ocm-2 for the 10th-layer islands. The increase in
the diameter and decrease in the density were caused by the strain field
formed by the lower islands as in Section 1.3.
TEM photography shows the overall structural features. Figure 2.14(a) is
a (110) cross-sectional TEM image of a 5-island stacked structure grown
with 2-nm thick intermediate layers. Each island layer was accompanied by
2 MOLECULAR
BEAMEPITAXIAL
GROWTH 135
FIG. 2.12. AFM images of the islanding surfaces at the (a) Ist, (b) 3rd, (c) Sth, and (d) 10th
layers stacked with 3-nm thick intermediate layers. The upper-layer islands expanded slightly
as the number of stacked layers increased.
40 1.5
n
30
EC
U
10
0 0
0 2 4 6 8 10 12
Layer number
FIG. 2.13. The dependence of the island size (diameter) and density on the number of
stacked layers. The average diameter of the 10th-layer islands was about 33 nm-40"/0 larger
than the diameter of the 1st-layer islands (about 24 nm), while the island density decreased from
1 x 10" cm-' for the 1st-layer islands to 8 x IOL0cm-*for the 10th-layer islands.
136 YOSHIAKI NAKATA
AND YOSHIHIRO SUGIYAMA
(a)
FIG.2.14. (a) ( I 10) cross-sectional TEM image of a 5-layer stacked island structure grown
with 2-nm thick intermediate layers. Each island layer was accompanied by a wetting layer.
The islands as a whole were about 22 nm in diameter and 13 nm in height, shown in the image
as dark megaphone-like strained regions. (b) Plan view of a TEM image obtained from a
5-layer stacked island structure. Island density was 8 x 10'" cm-'. Size uniformity and
lateral ordering were improved compared to a single-island layer.
FIG. 2.15. Growth process of stacked islands. The height of the InAs island after GaAs
overgrowth decreased to the thickness of the intermediate layer. Then SK growth occurred on
top of the lower island. This phenomenon enables us to precisely control the thickness of
islands in the growth direction through the intermediate-layer thickness.
2. PHOTOLUMINESCENCE
PROPERTIES
I I I I I
EXCITONCONFINEMENT EVALUATED
3. ZERO-DIMENSIONAL BY
DIAMAGNETIC
SHIFTS
Section 111.2 indicated that the exciton states of the stacked dot form
through the wave function overlap of each SK dot component, and that the
perpendicularly stacked structure should be considered as one quantum dot.
Magnetophotoluminescence, treated here, will give us not only the evidence
of the three-dimensional confinement of excitons but also the exciton wave
function extent in quantum dots. The principle can be summarized this way.
The magnetic field confines excitons in the plane perpendicular to the field
and increases their energy. We can determine the magnitude of other
competing confinement potentials and thus the extent of the wave function
by evaluating the number of energy shifts.
The samples and the magnetic field direction are schematically shown
in Fig. 2.18(a). The samples were five-layer-stacked InAs SK islands with
3-nm GaAs intermediate layers and, for comparison, an In,,,,Ga,~,,As/
GaAs quantum-well sample grown on a GaAs substrate by MOCVD.
The In,,,,Ga,,,,As well-layer thickness was 8 nm and subjected to - 1.3%
biaxial compressive strain. Figure 2.18(b) shows the photoluminescence
spectra of the two samples with no magnetic field applied.
Figure 2.18(c) shows the diamagnetic shifts for both samples under
magnetic fields perpendicular and parallel to the sample plane at 4.2 K. In
the quantum well, while the shift under a perpendicular field reached up to
7.3 meV at a maximum field of 11.8 T, the shift under the parallel field was
much smaller (only 1.6meV at 11.8 T). The shift of the dots was almost
independent of the field direction, and the maximum shift was 2.4meV in
either direction. These results immediately give us the following insights into
the difference in the quantum-confinement characteristics between the two
samples. The asymmetrical shift in the quantum well is due to its one-
direction quantum confinement. When the field is perpendicular to the
plane, it works as a two-dimensional confinement potential for excitons in
the well layer, causing large diamagnetic shifts. When the field is parallel to
the plane, it affects only one direction in the quantum-well plane, since the
other direction is already confined by strong potential barriers. The sym-
metrical, small diamagnetic shifts in the closely stacked dots show that
excitons are three-dimensionally confined by an almost symmetrical confine-
ment potential. The extent of the exciton wave function is estimated to be
almost equal to the quantum-well thickness of 8 nm, since the shifts for the
closely stacked dots and the quantum well under the parallel configuration
are almost the same.
Detailed quantitative analyses support this simple qualitative estimation
(Sugawara et al., 1997). The extent of the wave function, approximately
8 nm, is smaller than the dot size observed by plan-view TEM (Fig. 2.14(a))
and definitely larger than the height of the single SK dot of about 3 nm. This
indicates that the wave function of the ground-state exciton extends over
142 YOSHIAKINAKATA
AND YOSHIHIRO
SUGIYAMA
Stacked islands
(A = 1.088 (pm))
B
QW
(A = 0.936 (pm))
i;-
I I I I I
InGaAdGaAs QW 4.2 K
lnAslGaAs dot
*5.*meA
W
W
0 20 40 60
A80 100 120 140
Square of magnetic field ( T 2 )
FIG. 2.18. (c)
two or three SK dots out of five. Such exciton localization in the whole
stacked structure can be attributed to the composition and/or size in-
homogeneity of each SK dot components; that is, the potential minimum is
formed by the coupling of the adjacent two or three SK dots with low
band-edge energies.
various growth conditions, annealing after the growth of the GaAs inter-
mediate layers (see the growth conditions in Fig. 2.10) was found to cause
the most serious degradation. A large number of defects and/or impurities
should have been introduced during annealing, working as nonradiative
recombination centers. On the contrary, though the emission efficiency was
fairly good even at room temperature without annealing, the spectrum
width was reduced to 50-60 meV at most, presumably due to fluctuations
in the island height and in the thickness of the spacer between the bottom
of the upper-layer islands and the top of the lower-layer islands (Endoh et
al., 1998).
To overcome this problem, we stacked SK mode islands using even
thinner intermediate layers -3 ML thick (less than 1 nm) -so that islands
would physically contact each other in the perpendicular direction. This
thickness is much less than the island height. As seen in Fig. 2.2, about 6 ML
of GaAs supply was needed to fully cover the islands.
The SK-mode islands could be grown even on such thin GaAs intermedi-
ate layers. Figure 2.19 shows the RHEED-pattern intensity transition during
the stacked island growth (the GaAs intermediate layer on the first InAs
island layers and the second InAs island layers). In this experiment, the first
island layer was grown with the InAs supply of 1.8 ML. The intermediate-
layer thickness was reduced to 3 ML from 10 ML, which is the same as that
used in the closely stacked structures in Section 111. Immediately after the
GaAs supply, the spot intensity decreased, showing that the growth surface
was becoming flat by the growth of the intermediate layers. The spot
intensity recovered to that of the first islands when the InAs supply started,
showing that the second island layers grew. It should be noted that even on
3- and 5-ML thick intermediate layers, which were much thinner than the
island height, second island layers grew as well as the first island layer.
However, the critical amount of lnAs supply for the transition from
two-dimensional layer growth to three-dimensional island growth decreased
as the intermediate layer thickness has reduced: 0.84ML for the 10-ML
intermediate layers, 0.76 ML for the 7-ML intermediate layers, 0.60 ML for
the 5-ML intermediate layers, and 0.26ML for the 3-ML intermediate
layers. These were much smaller than that for the first island layer, shown
at the bottom of Fig. 2.19.
Figure 2.20 shows the RHEED-pattern intensity transition when the InAs
islands were repeatedly stacked using the 3-ML thick intermediate layers.
The inset photographs are the AFM images of the topmost layer after 8 to
9 repetitions. The amount of InAs supply for the stacked-island formation
was changed from about 0.5 ML to 0.8 ML. When the stacked islands were
grown with an InAs supply of about 0.7ML, the RHEED intensity
oscillated with the InAs island growth (RHEED intensity increased) and
BEAMEPITAXIAL
2 MOLECULAR GROWTH 145
a
YI
U
FIG.2.21. (a) Cross-sectional TEM image of columnar-shaped quantum dots. (b) Sche-
matic structure. (c) Low magnification image.
148 YOSHIAKI NAKATA
AND YOSHIHIRO SUGIYAMA
0.9’
t I ’ I I ”
i
0 10 20 30 40
Stacked-layer number
FIG.2.23. Photoluminescence wavelength at room temperature and 77 K as a function of
the stacked layer number. As the layer number increases, the emission wavelength shifts to
longer wavelength due to the size increase.
increased, the emission wavelength becomes longer due to the size increase.
It is surprising that the longest wavelength at room temperature was
1.24 ym at the 23rd stacking- very close to the practically applicable
1.3 pm.
Structural features of the stacked islands are summarized in Fig. 2.24.
Stacked-island structures strongly depend on the intermediate layer thick-
ness, t (nm). When t > 20 nm, stacked islands form without any correlation
to the lower-layer islands. The mean island size and density are constant in
all island layers, which is useful for increasing island density per unit
volume. When t < 20 nm, stacked upper islands form in correlation to the
lower-layer islands due to strain fields induced by the lower-layer islands.
The upper-layer islands form in alignment with the first-layer islands in the
perpendicular direction, and expand with an increase of the stacked-layer
number. As the intermediate-layer thickness decreases almost to the island
height, the size expansion of the stacked islands is suppressed, and almost
equal size islands are stacked closely. In this case, the electron wave
functions between neighboring islands in the perpendicular direction over-
lap each other, and the perpendicularIy stacked structure works effectively
as a single large quantum dot. Drastic improvement of the spectrum
linewidth is observed. When the intermediate-layer thickness further de-
150 NAKATAAND YOSHIHIROSUGIYAMA
YOSHIAKI
0 10 20 30
Intermediatelayer thickness, t (nm)
FIG. 2.24. Structural features of stacked islands. Stacked island structures strongly depend
on the intermediate-layer thickness, t (nm). When f z 20nm, stacked islands were formed
without any correlation to the lower-layer islands. When t < 20 nm, stacked upper islands were
formed in correlation to the lower-layer islands due to strain field induced by the lower-layer
islands.
creases to a few monolayers of less than 1 nm, which is much smaller than
the island height, stacked islands are formed. This success is achieved by
optimizing the amount of InAs supply for the stacked-island formation. If
we use 3-ML-thick intermediate layers, the amount of InAs supply has to
be reduced to about 0.7ML. The islands in the stacked structure are in
contact with each other physically in the perpendicular direction and the
stacked structure has a columnar shape as a whole. Even when we use these
thin intermediate layers, all stacked islands are accompanied by wetting
layers, indicating that the islands grow in the SK mode.
V. Summary
Schematic
structure
are very useful in suppressing island height fluctuation because the effective
height can be controlled artificially by the number of stacked island layers.
The luminescence spectrum was remarkably improved from 80 meV to
25 meV by the close stacking method. In columnar-shaped quantum dots,
where both narrow spectrum width and high emission efficiency are ob-
tained, we have successfully achieved low-threshold and highly efficient
operation of quantum-dot lasers (Chapter 6). According to the simulation
of quantum-dot laser performance in Chapter 1, the narrow spectrum width
we have achieved is very promising for high-performance quantum-dot
lasers that are superior to strained quantum-well lasers. If we aim at
high-speed direct-modulation quantum-dot lasers, as predicted by Arakawa
and Sakaki (1982), we need to further reduce the spectrum width to 10 meV
(Chapter 1).
The in-plane size fluctuation of our columnar-shaped dots was evaluated
by AFM and found to give a broadening of about 20meV (Endoh et al.,
1998). For that reason, we are now concentrating on improving in-plane size
homogeneity.
Acknowledgments
The authors would like to thank Dr. Hiroshi Ishikawa, Dr. Hajime Shoji,
Dr. Kohki Mukai, Dr. Osamu Ueda, Dr. Akira Endoh, and Dr. Toshiro
Futatsugi, of Fujitsu Laboratories Ltd., and Professor Shunichi Muto of
152 YOSHIAKINAKATA
AND YOSHIHIRO
SUGIYAMA
Hokkaido University for their fruitful input and support. We also are
grateful to Dr. Hajime Ishikawa and Dr. Naoki Yokoyama of Fujitsu
Laboratories Ltd., for their encouragement and strong support throughout
this work.
REFERENCES
Arakawa, Y., and Sakaki, H. (1982). Appl. Phys. Lert. 40, 939.
Arita, M., Avramescu, A., Uesugi, K., Suemune, I., Numai, T., Machida, H., and Shimoyama,
N. (1997). Jpn. J . Appl. Phys. 36. 4097.
Bimberg, D., Ledentsov, N. N., Grundmann, M., Kirstaedter, N., Schmidt, 0.G., Mao, M. H.,
Ustinov, V. M., Egorov, A. Y., Zhukov, A. E., Kop’ev, P. S., Alferov, Z . I., Ruvimov, S. S.,
Gosele, U., and Heydenreich, J. (1996). J p n . J. Appl. Phys. 35, 1311.
Carlsson, N., Georgsson, K., Montelius. L.. Sarnuelson, L., Seifert, W., and Wallenberg, R.
(1995). J . Cryst. Growth 156, 23.
Endoh, A., Nakata, Y., Suguyama, Y., Takatsu, M., and Yokoyama, N. (1998). Pmc. 10th Inr.
Conf: InP and Related Materials.
Fafard, S., Leon, R., Leonard, D., Martz, J. L., and Petroff, P. M. (1994). Phys. Reu. B. 50,8086.
Fukui, T., and Saito, H. (1987). Appl. Phys. L m . 50, 824.
Gershoni, D., Ternkin, H., Dolan, G. J., Dunsmuir, J., Chu, S. N. G., and Panish, M. B. (1998).
Appl. Phys. Lett. 53, 995.
Goni, A. R., Pfeiffer, L. N., West, K. W.. Pinzuk, A., Baranger. H. U., and Storrner, H. L. (1992).
Appl. Phys. Lett. 61, 1956.
Grundrnann, M., Christen, J., Ledentsov. N. N., Bohrer, J., Bimberg, D., Ruvirnov, S. S.,
Werner, P., Richter, V., Gosele, V., Heydenreich, J., Ustinov, V. M., Egorov, A. Y., Zhukov,
A. E., Kopev, P. S., and Alferov, 2 . 1. (1995a). Phys. Rev. Lett. 20, 464.
Grundmann, N.. Ledentsov, N., Heitz, R.. Eckey, L., Christen, J., Bohrer, J., Bimberg, D.,
Ruvimov, S. S., Werner, P., Richter, V., Heydenreich, J., Ustinov, V. M., Egorov, A. Y . ,
Zhukov, A. E., Kopev, P. S., and Alferov, Z. 1. (1995b). Phys. Stat. Sol. 188, 249.
Glas, F., Guill, C., Henoc, P., and Haussy, F. (1987). Inst. Phys. Can$ Ser., No. 87; Section 2,
p. 87.
Grundmann, M., Ledentsov. N. N., Steir, 0..Bimberg, D., Ustinov, V. M. Kop’ev, P. S., and
Alferov, Z. I. (1996). Appl. Phys. Lert. 68. 979.
Hessman, D.. Castrillo, P., Pistol, M.-E.. Pryor. C., and Samuelsson, L. (1996). Appl. Phys. Lett.
69, 749.
Houzay, F., Guille, C., Moisson, J.-M., Henoc, P., and Barth, F. (1987). J . Cryst. Growth 81,67.
Ikoma, N., and Ohkouchi, S. (1995). J p n . J . Appl. Phys. 34, L724.
Jeppensen, S.. Miller, M. S., Hessrnan. D., Kowalski, B., Maximov, I., and Sarnuelson, L. (1996).
Appl. Phys. Lett. 68, 2228.
Kapon, E.. Kash, K., Olausen, E. M.. Hwang, D. M., and Colas, E. (1992). Appl. Phys. Lett.
60, 477.
Kapon, E., Simhony, S., Bhat, R., and Hwang. D. M. (1989). Appl. Phys. Lett. 55, 2715.
Kitamura, M., Nishioka, M., Oshinowo, J., and Arakawa, Y. (1995). A p p l . Phys. Lett. 66, 3663.
Konkar, A. Madhukar, A,, and Chen, P. (1998). Appl. Phys. Lett. 72, 220.
Ledentsov, N. N., Bohrer, J., Bimberg. D., Kochnev, 1. V., Maximov, M. V., Kop’ev, P. S.,
Alferov, 2 . I., Kosogov, A. O., Ruvimov. S. S., Werner, P., and Gosele, U. (1996). Appl.
Phys. Lett. 69, 1095.
2 MOLECULAR
BEAMEPITAXIAL
GROWTH 153
Leon, R., Petroff, P. M., Leonard, D., and Fafard, S. (1995). Science 267, 1966.
Leonard, D., Krishnamurthy, M., Reaves, C. M., Denbaars, S. P., and Petroff, P. M. (1993).
Appl. Phys. Lett. 63, 3203.
Leonard, D., Pong, K., and Petroff, P. M. (1994). Phys. Rev. B. 50, 11687.
Lewis, B. F., Lee, T. C., Grunthaner. F. J.. Madhukar, A,, Fernandez, R., and Masanian, F. F.
(1984). J . Vac. Sci. Techno/. B2, 419.
Marzin, J.-Y., Gerard, J.-M., Izrael, A,, and Barrier, D. (1994). Phys. Rec. Lett. 73, 716.
Marzin, J.-Y., Gerard. J.-M., Israel, A., Barrier, D., and Bastard, G. (1994). Phys. Rer. Lett. 73,
716.
Miyamoto, Y., Cao, M., Shingai, Y., Furuya, K., Suematsu, Y., Ravikumar, K. G., and Arai. S.
(1987). Jpn. J . Appl. Phys. 26, L225.
Moison, J. M., Houzay, F., Barthe, F., Leprince, L.. Andre, E., and Vatel, 0. (1994). A p p l . Phys.
Lett. 64, 196.
Mui, D. S., Leonard, D., Coldren. L. A,. and Petroff, P. M. (1995). Appl. Phys. Lett. 66. 1620.
Mukai, K., Ohtsuka, N., Sugawara, M., and Yamazaki, S. (1994). J p n . J . Appl. Phys. 33. L1710.
Muto, S. (1995). Jpn. J . Appl. Ptiyr. 34. L210.
Nakata, Y.. Sugiyama, Y., Futatsugi. T,. and Yokoyama, N. (1997). J . Cryst. Growth 175/176,
173.
Nishi, K., Anan, T., Gomyo, A,. Kohmoto, S.,and Sugou, S. (1997). Appl. Phys. Lett. 70. 3579.
Oshinowo, J., Nishioka, M., Ishida, S.. and Arakawa, Y. (1994). Appl. Phys. Lett. 65, 1421.
Petroff, P. M., Gossard, A. C., Logan, R. A., and Wiedmann, W. (1982). Appl. Phys. Lett. 41,
635.
Petroff, P. M., Gossard, A. C., and Wiegmann, W. (1984). Appl. Phys. Lett. 45. 620.
Pfeiffer, L. N., Stormer, H. L., Baldwin. K. W., West, K. W., Goni, A. R., Pinczuk, A. Ashoori,
R. C., Dignarn, M. M., and Wegscheider, W. (1993). J . Cryst. Growrh 127. 849.
Reaves, C. M., Bressler-Hill, V., Varma, S.. Weinberg, W. H., and DenBaars, S. P. (1995). Surf
Sci. 326. 209.
Reed, M. A., Bate, R. T., Bradshow. K.. Duncan, W. M., Frenseley, W. R., Lee, J. W., and Shih,
H. D. (1986). J . Vuc. Sci. Echnol. B4. 358.
Saitoh, T., Taniniura, A., and Yoh. K. (1996). J p n . J . Appl. Phys. 35, 1370.
Schaffer. W. J., Lind, M. D., Kowalczyk, S. P., and Grant, R. W. (1983). 1. Vac. Sci. Trchnot.
B1, 688.
Seifert, W., Carlsson, N., Peterson, A.. Wernersson, L.-E., and Samuelson, L. (1996). Appl. Phjs.
Lett. 68, 1684.
Shchukin, V. A,, Ledentsov, N. N., Kop’ev, P. S., and Bimberg, D. (1995). Appl. Phys. Lett. 75,
2968.
Solomon, G. S., Komarov, S. K.. Harris. J. S., Yamamoto, Y. (1997). J . Cryst. G r o w h 175/176.
707.
Solomon, G. S., Trezza, J. A., Marshall, A. F., and Harris, J. S. (1996). Pkjs. Reu. Lett. 76. 952.
Stranski, 1. N., and Krastanow, L. (1937). Aktid. Wiss. Wien Math.-Natur. ilh 146. 797.
Tackeuchi, A., Nakata, Y.. Muto, S., Sugiyama, Y., Inata, T., and Yokoyama. N. (1995). J p n .
J . Appl. Phys. 34, L405.
Tabuchi, M., Noda, S., and Sasaki, A. (1992). Sci. & Tech. Mesoscopic Structures 379.
Tanaka, S.. Hirayama, H., Ramvall, P.. and Aoyagi, Y. (1997). Piocec~cli,zgs of the 2nd
Inrernationul Corzference on Nitride Seniiconductors, 48.
Tanaka, S., Iwai, S., and Aoyagi, Y . (1996). Appl. Plzys. Lett. 69, 4096.
Tersoff, J., Tieichert, C., and Lagally, M. G. (1996). Phys. Rev. Lett. 76, 1675.
Tsui, R., Zhang, R., Shiralagi, K., and Golonkin. H. (1997). Appl. Phys. Lett. 71, 3254.
Tu,, K.-N., Mayer, 3. W., and Feldnian. C. (1992). “Electronic Thin Film Science,” Chapter 7.
Macmillan Publishing Company.
154 YOSHIAKINAKATA
AND YOSHIHIROSUGIYAMA
Ueda, O., Nakata, Y., and Muto, S. (1998). Prcir. 10th Int. ConJ InP and Related Muterial.?, p.
113.
Wegscheider, W., Pfeiffer, L. N., Pinczuk. A., West, K. W., Dignam, M. M., Hull, R., and
Leibenguth, R. E. (1995). f.Cryst. Growth 150, 285.
Xie, Q., Madhukar, A,, Chen, P., and Kobayashi, N. P. (1995). Phys. Rev. Lett. 75, 2542.
SEMICONDUCTORS AND StMIMETALS. VOL 60
CHAPTER 3
FUJITSU
LABORATORIFS
LTD.
ATSUGI, JAPAN
KANAGAWA.
Nobuyuki Ohtsuku
[NTGRATtU MATERIALS
LABVRATORY
LTU.
FUJITSULABORATORIES
A T ~ ~ J KANAGAWA.
G~. JAPAN
1. VTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . I
I. Introduction
was that short-period superlattices can break through the limit set by the
critical thickness to reach a highly efficient 1.3-pm emission. Actually, Roan
and Cheng (1991) had already succeeded in growing (GaAs),/(InAs),
short-period superlattices on GaAs substrates by MBE that showed 1.3-pm
emission at room temperature. Using the ALE approach, we also succeeded
in realizing 1.3-pm emission and in making the emission spectrum linewidth
much narrower (Ohtsuka and Ueda, 1994). However, we found that the
grown materials were not short-period superlattices but had bizarre struc-
tures, as seen in the transmission electron microscope (TEM) photograph of
Fig. l.l(b) in Chapter 1, where spherical dark regions are buried in the
quantum wells. Having done various diagnostics of the materials (see
Chapter 4), we finally concluded that the structures are quantum dots.
This chapter reviews ALS quantum-dot growth. We start by describing
the self-limiting one-monolayer growth of InAs and GaAs and the growth
of (InAs),/(GaAs), superlattices. Then we present two kinds of growth
sequences for the quantum dots: one, an In-As-Ga-As sequence based on the
concept of making monolayer superlattices by ALE; the other, a newly
developed In-Ga-As sequence. Quantum dots are characterized by their
shape, size, composition, and emission wavelength. In particular, we illus-
trate how the dots vary with alternate supply cycle, growth temperature, and
composition of the buffer layers on which the dots are grown. Finally, we
compare the growth process of ALS quantum dots with that of SK dots.
al., 1988, 1989, 1991; Sakuma et al., 1988). The structure of the growth
system is shown in Fig. 3.1. The reactor was designed through a simulation
to realize smooth gas flow streamlines without vortices. We adopted a
low-pressure atmosphere for precursors to decompose only on the heated
substrate surface. The substrate was positioned in the fast gas stream
emitted from a jet nozzle to prevent thermally decomposed source molecules
from building up in stagnant layers. Source materials were trimethylindium
(TMIn), TMIDMEA, trimethylgallium (TMGa), triethylgallium (TEGa),
and arsine (ASH,) diluted with hydrogen (HJ. Experiments on self-limiting
ALE growth were done on GaAs substrates for GaAs growth and on InAs
substrates for InAs growth.
The most severe problem we encountered in fabricating InAs/GaAs
superlattices using ALE was that the temperature range for self-limiting
growth was different for the two binary compounds-450 to 550°C for
GaAs using TMGa and 300 to 400’C for InAs using TMIn. Since there is
no overlap in ALE growth temperatures for the two materials, it appeared
difficult to fabricate the superlattices under precise self-limiting growth
(Mori and Usui, 1992). A key factor in determining the temperature range
for self-limiting growth is the decomposition temperature of the group-I11
precursors. We expected that the thermal stability of TMIn, which is a
FIG. 3.1. Growth system for pulse-jet epitaxy with a chimney reactor, in which precursors
are supplied in a fast, pulsed stream and self-limited 111-V compounds can be grown under a
wide range of conditions.
3 METALORGANIC
VAPOR PHASE GROWTH
EPITAXIAL 159
ALE-In As I
I
Tg = 460;C
0
2 4 6 8
I-
I
v1
c,
.3 TMIDMEA
c
s = I TMIn
h
c,
.M
- FWHM
8c
w
= 28 meV
.&
used in Section 11. Each supply is separated by H, purging pulses. The dots
were grown on an In,Ga, -,As buffer layer on a (001)-GaAs substrate and
were covered by an In,Ga, -,As layer having the same composition as the
buffer layer. The supply amount of TMIDMEA and TMGa was set at less
than one monolayer to effect higher emission efficiency.
Figure 3 4 a ) and (b) show typical plan-view and cross-sectional TEM
images of ALS dots, grown by 12 cycles of (InAs),/(GaAs), periodic supply
on a 300-nm GaAs buffer layer on a (001)-GaAs substrate, covered by a
30-nm-thick GaAs cap layer. The growth temperature was 460°C which
provides distinct self-limiting growth for both InAs and GaAs (see Fig. 3.3).
The growth pressure was 15 Torr. A plan-view image shows uniform
dot-like microstructures about 20 nm in diameter and an area coverage of
5-10%. The dots diameters follow a Gaussian distribution with a standard
deviation of 2.9 nrn. A cross-sectional dark-field image indicates that sphere-
shaped dots were self-assembled within the InAs/GaAs short-period layer
(dot layer) sandwiched by GaAs. They were about 10nm in height and
surrounded laterally by a quantum-well layer having almost the same
thickness as that of the dots. The thickness of the quantum-well layer was
definitely greater than that of the so-called wetting layer of the SK islands
(Krost et al., 1996; Ramachandran et al., 1997a, and b. Kitamura et al.,
162 KOHKIMUKAI.MITSURUSUGAWARA, EGAWA
AND MITSURU
FIG.3.5. (a) Plan-view and (b) cross-sectional TEM images of ALS dots, grown by 12
cycles of (InAs),/(GaAs), periodic supply on a 250-nm GaAs buffer layer on a (001) GaAs
substrate and covered by a 30-nm-thick GaAs cap layer.
1997). We observed two clear upper and lower borders between the GaAs
sandwiching layers and the dot layer. We also observed a high-resolution
TEM image around the dot in a cross-sectional sample and did not find any
dislocations or defects in the lattice image (Fig. 3.6). A weak contrast at the
border of the dots suggests that the composition and lattice distortion do
not abruptly change around the dot.
Figure 3.7 shows the spatial composition distribution of a sample cross
section evaluated by energy-dispersive X-ray microanalysis (EDX) with an
excitation electron beam focused to a diameter of about 1 nm. The sample
was thinned down to lOnm, which is less than the in-plane diameter of the
dots, to obtain a signal only from the dots. Signal intensity was calibrated
to estimate a precise indium composition using the signal intensity of a
lattice-matched In,,,,Ga,~,,As layer grown on an InP substrate as a
reference. The indium composition was 0.5 at the center of the dots (i.e.,
GROWTH
VAPORPHASEEPITAXIAL
3 METALORGANIC 163
0.27
0.53(dot)
0 0.5 1.o
Indium composition
FIG.3.7. Spatial indium composition distribution of quantum-dot cross sections evaluated
by energy dispersive X-ray microanalysis (EDX) with an excitation electron beam focused to
a diameter of about 1 nm (from Mukai et al., 1994).
164 KOHKIMUKAI,
MITSURU
SUGAWARA,
AND MITSURU
EGAWA
diameter, while the height of ALS dots is about half of the in-plane diameter.
These characteristics suggest that our dots were not self-assembled by the
conventional SK mode.
The outstanding characteristic of our quantum dots is a wide-range
wavelength tunability of between 1.2 and 1.5 pm via 1.3 pm, which is
achieved by changing the cycle of short-period growth and the indium
composition of the In,Ga,-xAs buffer layer. As shown in Fig. 3.4, emission
from the ground state of the ALS dots was observed at 1.34pm, and that
from the excited state at 1.24 pm. The FWHM of 28 meV shows a high
degree of uniformity in terms of dot size, strain, and composition.
Figure 3.8 shows (a) the photoluminescence wavelength and (b) the
in-plane diameter of quantum dots evaluated by TEM photography as a
function of the number of cycles. The buffer layer was In,,,,Ga,~,,As. By
increasing the alternate growth cycle number from 7 to 24, the emission
peak wavelength was increased from 1.29 to 1.41 pm. Figure 3.9 shows the
photoluminescence wavelength as a function of the buffer-layer indium
composition. By increasing the In composition of the buffer layer from 0
(GaAs) to 0.09 (Ino,09Gao,91As), the emission peak wavelength was in-
creased from 1.3 to 1.46 pm at the 12 cycles of InAs/GaAs. Since there was
no distinct differences among the In composition in dots measured by EDX,
we know that the wavelength variation is due to the change in dot size, as
seen in Fig. 3.8(b). Combining the above results concerning the number of
supply cycles and the indium composition of the buffer layer, we can control
the emission wavelength continuously from 1.2 to 1.5 pm.
In order to improve the quantum-dot crystal quality, we optimized
various growth conditions such as temperature, sequence, and source supply
amount, by checking the photoluminescence emission efficiency. Figure 3.10
shows the photoluminescence intensity versus growth temperature for dots
grown at 12 cycles of (InAs),/(GaAs),. The optimized growth temperature
was between 440 and 460"C, above and below which the emission intensity
rapidly decreased. We also found that the source supply amount should be
less than one monolayer. In addition, we adopted a new sequence of
In-Ga-As, which provided an emission efficiency more than two times higher
than that of the In-As-Ga-As sequence.
Under these optimized growth conditions, we made cavity lasers and
succeeded in lasing at 80 K, though the emission was from the higher-order
excited states (see Chapter 6). We could not achieve lasing from the ground
state, because of insufficient optical gain for lasing due to the low area
density of the ALS dots. Since the PJE system was not equipped with a gas
supply sufficient to fabricate laser devices, we were obliged to grow parts of
the laser structure other than the quantum-dot active region using other
growth systems. We believe that this process severely damaged the crystal,
3 METALORGANIC
VAPORPHASEEPITAXIAL
GROWTH 165
1.25
5 10 15 20 25
0 0.05 0.1
manifold, then divided into subflows and introduced to the injectors. For
the gas-handling system, we used a fast switching manifold with a pressure-
balanced vent-and-run configuration.
Self-limiting growth was observed for InAs and GaAs at common
temperatures using TMIDMEA and TMGa. Note that the growth tempera-
ture indicated in this section does not agree with the temperature of the PJE
reactor in Sections I1 and 111. Comparing temperatures for self-limiting
growth in each reactor, a growth temperature of 460°C for the PJE reactor
in Sections I1 and 111 corresponds to 520°C for the MOVPE reactor in this
section.
The most striking change from Sections I1 and 111 was that we adopted
the growth sequence In-Ga-As from the beginning. Optimized source supply
amounts were 2.0 seconds for TMIDMEA and 0.3 seconds for TMGa,
which correspond to 0.5-monolayer In and 0.1-monolayer Ga. The duration
of the ASH, supply was fixed at 7 seconds. The H, purge was inserted after
the TMGa and ASH, supply for 3 seconds and 0.5 seconds, respectively. The
supply rates of TMIDMEA, TMGa, and ASH, sources were 0.04, 4.6, and
40 cc per minute, respectively. A single cycle time of 12.8 seconds provided
a very low growth rate of about 0.04 ML per second.
168 KOHKIMUKAI, SUGAWARA,
MITSURU AND MITSURUEGAWA
1. Two TYPES
OF ALS DOT
The growth procedure was as follows. A 0.25-pm GaAs buffer layer was
grown on a (100)-GaAs substrate at 630°C. The growth was then interrup-
ted for 10 minutes and the substrate temperature was lowered under an
ASH, atmosphere. Then, after an additional 0.05-pm GaAs buffer layer
growth, the In-Ga-As layer was grown by the alternate source supply,
followed by the 30-nm GaAs cap layer. The GaAs buffer and cap layers were
grown by conventional MOVPE using TEGa and ASH,. The growth
pressure was 15 Torr throughout the structure.
Figure 3.1 l(a) shows the wavelength of a photoluminescence spectrum
peak as a function of the growth temperature for the 18-cycle In-Ga-As
supply. The emission wavelength varied considerably between 1.2 and
1.3pm depending on the growth temperature, with a 1.3-pm emission at
around 500°C. Figure 3.11(b) shows the peak wavelength of the photo-
luminescence spectra as a function of the cycle number at various growth
temperatures. The relationship between the cycle number and the emission
wavelength also varied considerably between the two temperature ranges,
that is, the dots grown at 435 and 460°C showed an almost constant
emission wavelength, while the emission wavelength of dots grown at
490-510°C increased as the cycle number increased. This suggests that there
are two types of ALS dots grown according to different growth processes.
The dots with an emission wavelength of 1.2pm are similar to SK dots in
that the emission wavelength is independent of the source supply amount.
Chapter 2 states that the diameter of SK dots saturates above a critical InAs
layer thickness and that the dots are controlled by growth temperature only.
Figure 3.12 shows how the photoluminescence spectra vary with cycle
number and growth temperature. They were measured at 300 K using a Krf
laser with an excitation power density of 60 W/cm2. Using both the peak
wavelength and intensity as standards, we classified these photolumines-
cence spectra into five types. With a small cycle number, a very weak and
broad emission was observed, which can be attributed to an emission from
the two-dimensional layer prior to three-dimensional nucleation. No dots
were observed in the TEM photograph. As the cycle number increased,
single peaked and intense emissions appeared at around 1.2 pm (we call this
emission type A), followed by double-peaked, intense emissions at around
VAPORPHASE
3 METALORGANIC EPITAXIAL
GROWTH 169
E
1
W
I-!
a
Cycle number
FIG. 3.11. (a) Wavelength of a photoluminescence spectrum peak as a function of growth
temperature for the 18 cycles of In-Ga-As sequence. The emission wavelength varied consider-
ably between 1.2 and 1.3pm, depending on the growth temperature, with a 1.3-pm emission at
around 500°C. (b) Peak wavelength of photoluminescence spectra as a function of the In-Ga-As
cycle number grown at various temperatures. The relationship between the cycle number and
the emission wavelength varies considerably between the two temperature ranges; that is, the
dots grown at 435°C and 460°C showed an almost constant emission wavelength, while the
emission wavelength of dots grown at 490-510°C increased as the cycle number increased.
170 SUGAWARA,
KOHKIMUKAI,MITSKJRU AND MITSURU
EGAWA
rn
0.9 12 I5 0.9 11 15 0.9 I2 15 0.9 12 15 1.2 12 15
Mo*cfIl -
LJ LJLid
...f$$jq 0.9
No*ck-lt@ Lid u
-
0.9 12 15
K/K/~1 a~1
0.9 1.2 1.5 0.9 1.2 15 0.9 12 15 0.9 1.2 1 5 0.9 1.2 15
4 9 -e0 . c ~ ~ /
0.9 1.2 1.5 0.9 12 15 0.0 12 15 0.9 12 1.5 0.9 1.2 1 5 0.9 12 15
N = 12 14 18 18 20 22
type A and type B photoluminescence spectra are called type A and type B
quantum dots respectively.
The photoluminescence matrix of Fig. 3.12 tells us that source supply
amount and growth temperature govern the formation of ALS quantum
dots. For a growth temperature of 5O0-55O0C, type-A quantum dots with a
density of about 6 x 10'' c m p 2 (determined from plan-view TEM) h i -
tially form after two-dimensional growth. The subsequent source supply
transforms type-A quantum dots to type-B quantum dots with an identical
density. The shorter wavelength peak in the type-B photoluminescence
spectra is the second sublevel emission of type-B quantum dots because, as
the excitation laser power increases, several peaks appear in the higher
energy regime of the second peak (see Fig. 4.11(a) in Chapter 4). The
photoluminescence linewidth for type-B quantum dots is narrow, with an
FWHM of 30-50meV, while the linewidth for type-A quantum dots is
broad, with an FWHM of 80-120 meV. At higher growth temperatures,
type-A' quantum dots are transformed into type-B' quantum dots with an
increase in source supply. The weak photoluminescence in type-A' and
type-B' quantum dots above 560°C can be attributed to their low density,
as confirmed by the TEM photograph; the density of type-A' quantum dots
at 560°C and 18 cycles was about 2 x 10" cm-2-one-third that of type-A
and type-B quantum dots. For growth temperatures below 490"C, type-A
quantum dots transform into type-B' quantum dots with weak photo-
luminescence intensity. An introduction of dislocations reduced the photo-
luminescence intensity of the type-B' dots below 490°C, as detected by the
TEM photograph, although their density was high ( - 1 x 10l1cm-2).
Figure 3.13 shows high-resolution cross-sectional TEM images of typical
type-A (14-cycle) and type-B ( 1 8-cycle) quantum dots grown at 520°C.
Type-A dots have a diameter of 15nm and a height of 5nm. The In
composition detected by EDX is 0.30 (i.e., Ino.30Gao,70As). For type B the
diameter, height, and In composition were 25 nm, 8 nm, and 0.23, respect-
ively (i.e., Ino,23Gao,77As).Judging from the smaller In composition for
type-B dots, the longer photoluminescence wavelength is due to larger dot
sizes. We notice that the quantum-well layer surrounding quantum dots in
Fig. 3 3 b ) has become thinner like the wetting layer of an SK-mode dot as
seen in Fig. 3.13(b). This is because the supply of gallium is reduced from
an early amount of almost 1 monolayer to about 0.1 monolayer.
The structure of Fig. 3.5(b) can be reproduced in this MOVPE chamber
by an In-As-Ga-As sequence (Fig. 3.14(a)) or with an InGaAs layer to cover
the dots grown by an In-Ga-As sequence (Fig. 3.14(b)). In the TEM image
of Fig. 3.14(a)), the upper and lower borders of the quantum well are
not flat, since growth conditions were not yet optimized. In Fig. 3.14(b),
type-B quantum dots grown by an In-Ga-As sequence are covered by
172 KOHKIMUKAI,
MITSURUSUGAWARA,
AND MITSURU
EGAWA
FIG.3.13. Cross-sectlonal TEM images of quantum dots: (a) Type-A quantum dots grown
at 520°C with 14 cycles, and (b) Type-B quantum dots grown at 520°C with 18 cycles. The
diameter and height of type-A quantum dots are 15 nm and 5 nm, respectively, and those of
type-B quantum dots are 25 nm and 8 nm, respectively.
2. MULTIPLE-LAYER
GROWTH
FIG. 3.14. TEM image of (a) ALS dots grown by an In-As-Ga-As sequence, and (b) ALS
dots grown by an In-Ga-As sequence embedded in an In,,,Ga,,,As layer.
FIG. 3.16. Cross-sectional TEM images of triple-layer structures of type-B quantum dots
grown at 520'C with 18 cycles. The GaAs spacers are (a) 20nm. (b) 15 nm, and (c) 10nm. The
single-layer structure of type-B quantum dots is shown in (d).
Vertically aligned quantum dots became larger toward the upper layers in
the triple-layered structure, a phenomenon also observed in SK islands in
Chapter 2. The diameter of the dots at the top layer was about 35 nm, while
the diameter of those at the bottom layer was about 25 nm.
We discovered a unique phenomenon in the multiple-layer growth of
type-B quantum dots separated with a thick GaAs spacer. As shown in Fig.
176 SUGAWARA,
KOHKIMUKAI,MITSURU EGAWA
AND MITSURU
3.16, in the case of a 20-nm GaAs spacer, the quantum-dot density in the
bottom layer increased to about twice that of the single-layer structure, and
few dots were formed on the upper layers. For type-A quantum dots, a
drastic increase in density did not occur and similar structures were simply
repeated. Since the photoluminescence spectrum of the type-B quantum-dot
triple-layer structure with a 20-nm spacer was similar to that of the
single-layer structure, the newly formed quantum dots in the bottom layer
were of type B. At the present stage, we have no idea why the two types of
dot display different multiple stacking behavior.
We have seen in this chapter that ALS dots are very different from SK
islands. Let us assign type-A quantum dots to SK islands because, with what
we saw in Chapter 2, they have many common characteristics. The type B
ALS dots emit photons with a wavelength of 1.3pm or longer, because of
thicker structure than that of the SK islands. The FWHM of the emission
spectra is 30-50 meV, which is much smaller than the 80- 120 meV of the
SK islands. The ALS dots are buried in a quantum well with clear upper
and lower borders against the GaAs. The thickness of the quantum well
decreases and becomes like that of a wetting layer in SK-mode growth as
the supply amount of group-111 sources is reduced. The area coverage of the
dots is 5-lo%, which is definitely smaller than the 20-40% of the SK
islands. As the alternate supply cycle increases, dot size and emission
wavelength increase, whereas SK islands are almost insensitive to the supply
amount of InAs. As seen in Chapter 2, further indium supply after the
formation of SK islands increases the islands density and then causes them
to coalesce.
The photoluminescence matrix of Fig. 3.12 gives us a hint as to how the
alternate supply works. We have seen that, as the supply cycle increases the
wetting layer and then the SK islands (type A) form, leading to the
appearance of ALS dots (type B). The SK islands are self-assembled
primarily by the thermodynamical balance between strain energy and
surface energy. If we supply only indium after the SK islands (type A) form,
dot size will hardly change, as in the case of usual SK growth, since the
islands have already reached an equilibrium point. A supply of gallium to
the islands decreases the total strain energy, making it possible to add
further indium to the islands and thus increase dot size in a perpendicular
direction. This process repeats itself until the dots reach another equilibrium
point.
3 VAPORPHASEEPITAXIAL
METALORGANIC GROWTH 177
We believe that the growth process is not exactly the same for the early
ALS dots grown by supply sequence of In-As-Ga-As and for the recently
adopted ALS dots grown by In-Ga-As sequence, with both sequences
providing a unique 1.3-pm emission. The plan-view TEM images in Figs.
3.17 and 3.18 indicate how the In source gas contents in a supply cycle
affects the structure of the two types of ALS dots. By the increase of source
(b)
FIG. 3.17. Plan-view TEM images of the ALS dots grown by 18 cycle of the In-Ga-As
supply sequence: (a) the standard case and (b) the case when In source gas contents in a supply
cycle was raised by 20%.
178 KOHKIMUKAI,MITSURUSUGAWARA,
AND MITSURUEGAWA
(bi
FIG. 3.18. Plan-view TEM images of the ALS dots grown by 12 cycle of the In-As-Ga-As
supply sequence: (a) the standard case and (b) the case when In source gas contents in a supply
cycle was raised by 100%.
gas contents, the numerical density increases but the diameter does not
change for the ALS dots grown by In-Ga-As sequence (Fig. 3.18), while the
numerical density does not change but the diameter increases for the ALS
dots grown by In-As-GaAs sequence (Fig. 3.17). We don’t know the details
in the growth processes, but we suppose that the supply of ASH, after the
supply of TMIDMEA makes metal indium on growth surface stable by
VAPOR PHASE
3 METALORGANIC EPITAXIAL
GROWTH 179
VI. Summary
and a narrow spectrum linewidth of 28 meV. The quantum dots also acquire
great controllability in size and emission wavelength by an adjustment in
the cycle number and the composition of buffer layers on which the dots are
grown. By comparing the ALS quantum dots with SK islands, especially
their spectrum variation with the supply cycle number, we have achieved a
picture of the growth process of ALS dots. Optical characterization of ALS
dots is presented in Chapter 4.
REFERENCES
Kondo, M., Kuramata, A., Fuji, T., Anayama, C., Okazaki, J., Sekiguchi, H.. Tanahashi. T.,
Yamazaki, S., and Nakajima, K. (1992). J . Cryst. Growth 124, 265.
Krost, A., Heinrichsdorff, F., Bimberg, D., Darhuber, A,, and Bauer, G . (1996). Appl. Phys. Lett.
68, 785.
Kuan, S. T., and Iyer, S. S. (1991). Appl. Phys. Lett. 59, 2242.
Kitamura. M., Nishioka, M., Schur. R.. and Arakawa, Y. (1997). J . Cryst. Growth 170. 563.
Ledentsov, N. N., Grundmann, M., Kirstaedter. N., Schmidt, O., Heitz, R., Bohrer, J.. Bimberg,
D., Ustinov, M. V., Shchukin, A. V.. Kop'ev, S. P., Alferov, I. Z., Ruvimov, S. S., Kosogov,
0. A., Werner, P., Richter, U., Gosele. U., and Heydenreich, J. (1996a). Solid-State
Electron. 40, 785.
Ledentsov, N. N., Bohrer, J., Bimberg, D., Kochnev, V. I., Maximov, V. M., Kop'rv, S. P.,
Alferov, I. Z., Kosogov, 0. A., Ruvimov, S. S., Werner, P., and Gosele, U. (1996b). A p p l .
Phys. Lett. 69, 1095.
Mukai, K., Ohtsuka, N., Sugawara, M., and Yamazaki, S. (1994). Jpn. J . Appl. Phps. 33. L1710.
Mukai, K., Ohtsuka, N., and Sugawara, M. (1996). Jpn. J . Appl. Phys. 35, L262.
Mnkai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1997). Appl. Surf: Sci. 112, 102.
Mon, K., and Usui, S. (1992). Appl. Phys. Lett. 60, 1717.
Ohtsuka, N., and Ueda, 0. (1994). In Gus-Phase und Surface Clieinistry in Electronic Materials
Processing Vol. 334, 225-22.
Ohtsuka, N., and Mukai, K. (1995). Proc,. I n t . Conf 1nP and Related Materials, Hokkaido, 303.
Ozeki, M., Mochizuki, K., Ohtsuka, N., and Kodama, K. (1988). Appl. Phys. Lett. 53. 1509.
Ozeki, M., Mochizuki, K., Ohtsuka, N..and Kodama, K. (1989). Thin Solid Films 174. 63.
Ozeki, M., Ohtsuka, N., Sakuma, Y.. and Kodama, K. (1991). J . Cryst. Growth 107. 102.
Ramachandran, R. T., Heitz, R., Chen, P.. and Madhukar, A. (1997a). Appl. Phys. Lett. 70, 640.
Ramachandran, R. T., Heitz, R., Kobayashi, P. N., Kalburge, A,, Yu, W., Chen, P., and
Madhukar, A. (1997b). J . Cryst. Growth 175/176, 216.
Roan, E. J., and Cheng, K. Y. (1991). Appl. Phys. Lett. 59, 2688.
Sakuma, Y., Kodama, K., and Ozeki, M. (1988). Jpn. J . Appl. Phys. Lett. 27, L2189.
Solomon, S. G., Trezza. A. J., Marshall. F. A.. and Harris. S. J. (1996). Pkys. Rev. Lett. 76, 952.
TersoK, J., Teichert. C., and Lagally, G. M. (1996). Phys. Reo. Lett. 76, 1675.
Xie, Q., Madhukar, A., Chen, P., and Kobayashi, P. N. (1995). Phys. Reii. Lett. 75, 2542.
Yao, Y. J., Andersson, G. T., and Dunlop, L. G. (1991). J . Appl. Phys. 69, 2224.
This Page Intentionally Left Blank
SEMICONDUCTORS AND SEMIMETALS, VOL. 60
CHAPTER4
I. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
11. LIGHTEMISSION FROM DISCRETE ENERGYSTATES . . . . . . . . . . . . . . 185
I, Photoluminescence, Photoluminescence Excitation, and Electroluininescence
Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
2. Wafer Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . 190
3. Microprobe Photoluminescence . . . . . . . . . . . . . . . . . . . . 192
111. CONTROLLABILITY OF QUANTUM CONFINEMENT . . . . . . . . . . . . . . 196
1. Two Methods of Controlling Quantized Energies . . . . . . . . . . . . 196
2. Magneto-Optical Spectroscopy . . . . . . . . . . . . . . . . . . . . 200
IV. RADIATIVE EMISSION EFFICIENCY. . . . . . . . . . . . . . . . . . . . 20 1
V. S U M M A R Y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
1. Introduction
Cross-sectional TEM
SK dot
ALS dot
FIG.4.1. Two types of self-assembled dots. The InAs SK dots are grown by MBE or
MOVPE via the Stranski-Krastanov (SK) mode under highly strained epitaxy. The InGaAs
ALS dots are grown by alternately supplying InAs and GaAs precursors with one or fewer than
one monolayer. While SK dots are dome- or pyramid-shaped islands on a growth surface, ALS
dots are spherical and surrounded laterally by a quantum-well layer having almost the same
thickness as that of the dots.
4 OPTICAL CHARACTERIZATION OF QUANTUM
DOTS 185
1. PHOTOLUMINESCENCE,
PHOTOLUMINESCENCE
EXCITATION
AND
ELECTROLUMINESCENCE
SPECTRA
The inset of Fig. 4.4 shows additional shoulders at the shorter wavelength
regime of the spectra. These were the first sign of excited-level photo-
emission from self-assembled dots (Mukai et al., 1994). Multiple peak
emissions from discrete states can be detected owing to a narrower spectrum
linewidth than that of the interlevel separation. Figure 4.5 shows the
excitation power dependence of the photoluminescence spectra of ALS dots
at 77 K. As the excitation power increased, the second peak appeared, and
its intensity increased to exceed the first peak. At the highest power of
306 W/cm2, a shoulder assigned to the third peak appeared. Figure 4.6
compares the photoluminescence spectrum at a rather high excitation
intensity of 100 W/cm2, and the photoluminescence excitation spectrum,
monitored at the energy of 1.0 eV (the lowest-energy first emission). It was
possible to detect two clear resonances exactly at the same energy as those
in the photoluminescence spectra. This represents direct proof that the
multiple peaks in the photoluminescence spectra are due to the discrete
states under three-dimensional quantum confinement. Let us call the first
peak in the photoluminescence spectra the ground-state emission and the
second peak the first excited-state emission.
Figure 4.7 shows the electroluminescence spectra of the ALS dots at 77 K
from the p-n junction diode with the dot active region. The area of the
188 KOHKIMUKAIAND MITSURUSUGAWARA
k 0 1. . , * I , . .
* I ' ' * ' I a .' ' I ' ' ' ' ". a '
electrode was 20 x 900 pm'. Since the electric carrier injection can provide
many more carriers into dots than the optic excitation can, peaks up to the
5th appeared at the current injection of 400mA. As the injected current
increased, the emission intensity of the first and second peaks were almost
saturated. Since the emission rate is given by NITsp, where N is the carrier
number occupying the state and t S p is the spontaneous emission lifetime. the
saturation of the emission intensity is due to the saturation of the carrier
number in the corresponding energy state. As low energy states come to be
almost occupied by carriers, the relaxation rate into the states decreases due
to Pauli blocking, and the emission starts to rise from higher energy states.
However, Pauli blocking cannot completely explain the excitation inten-
sity dependence of emission spectra in Fig. 4.7. We should note that, at
lOmA, the second and third emissions had appeared though the first
emission intensity was less than its maximum. The carrier number, N , of
each state, governing the emission intensity, is determined by the balance
among the rates of various processes, such as spontaneous emission at the
state, relaxation into the state, and thermal emission from the state. In the
ALS dots with an energy separation of 50-80 meV-much higher than the
thermal energy at 77 K-the thermal emission rate is negligible and the
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 189
I I I I I I I I
I
I I I I I I I
Wavelength (pm)
FIG.4.5. Excitation power dependence of photoluminescence spectra of ALS dots grown
using 9 cycles of alternate supply. The sample was excited by a 647.1-nm Kr’-ion laser. As the
excitation power increased, the second peak appeared and its intensity increased to exceed the
first peak. At the highest power of 306 W/cm2, a shoulder assigned to the third peak appeared.
(Reprinted from Appl. Surface Science, vol. 112, Mukai et al., “Growth and optical evaluation
of InGaAs/GaAs quantum dots self-formed during alternate supply of precursors,” pp.
102- 109, 1997, with kind permission of Elsevier Science-NL, Sara Burgerhartstraat 25, 1055
KV Amsterdam, The Netherlands.)
I I I I 1
n
3
-a
3
3 -
.3
LQ
G -
I
0.9
I
1.o
Y I
1.1
I
1.2
1I
1.3
Energy (eV)
FIG.4.6. Photoluminescence excitation spatrum at 4.2 K of the sample grown by 18 cycles
of In-Ga-As supply. The sample was excited by the monochromated light of a Halogen lamp,
and a Ge detector was used to detect luminescence from the sample surface at the energy of
1.0 eV, that is, the lowest-energy first emission. Two clear resonances could be detected exactly
at the same energy as that in the photoluminescence spectra (from Mukai et al., 1996c.
Copyright 1996 by the American Physical Society).
2. WAFERMAPPING
Wavelength (pm)
FIG. 4.7. Electroluminescence spectra of ALS dots grown by 18 cycles of In-Ga-As
alternate supply. The luminescence parallel to the surface was dispersed and detected by an
InGaAs photo-multi detector quenched to - 70°C using a conventional lock-in technique.
Electrode size was 20 x 900prn'. Up to the fifth emission peaks appeared as the injected current
increased (from Mukai et al., 1996b).
3. MICROPROBE
PHOTOLUMINESCENCE
FIG. 4.9. Near-field optical microscope. Samples were set in the cryostat and cooled to 5 K.
With a wavelength of 633nm, the excitation beam went from the laser diode through a
sharpened optical fiber to the sample surface. Luminescence from the surface was collected by
the fiber probe. The tip was controlled in close proximity to the sample by the shear-force
feedback technique. The luminescence was dispersed by a monochrometer, and the photon
signal was counted with a n InGaAs photo-multi detector kept at -8O'C.
c
0
Y
0
d
a
. 90 W/cm2
NSOM :
r . . . . l . . . . l . . . . l . . . . i
0.95 1.oo I .05 1.10 1.15
Energy (eV)
FIG.4.10. Photoluminescence spectra of the ALS dots by the near-field optical microscope.
The excitation power was estimated to be 90 W/cm2. The spectra of macro-photoluminescence
is shown by a dashed line. The two sets of multiple emission peaks in the NSOM spectra
correspond to the ground-state and the first excited-state emission.
posed by a dashed line. The two sets of emission spectra with fine structures
denoted as S, and S,, correspond to the quantum dots' ground state and
first excited state, respectively. Note that S, and S, are well separated and
that almost no emission was observed in between. This is because the
smaller probing area reduced the inhomogeneous spectrum broadening.
However, the emission spectrum of S, and S, did not become as narrow as
we expected, indicating that inhomogeneity of dots had already occurred in
the area smaller than 1 pm. Inhomogeneity in such a small area cannot be
caused by nonuniformity in growth conditions, such as wafer temperature
distribution and source gas streamline. The fine structures on S, and S, are
reproduced by repeated measurements and can be attributed to emission
from one or a few dots. Sharp peaks, denoted by the common small letter
with the subscripts 1 and 2, might be the ground-state and the excited-state
emission of the same dots.
Figure 4.1 l(a) shows the excitation-power dependence of the NSOM
spectra. Comparing that with the macroscopic photoluminescence spectra
of Fig. 4.5, we see that the emission from each state is sharper and up to the
seventh emission is resolved, due to the higher spatial resolution. We found
that the higher the sublevel order, the wider the peak width. Thus, the
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 195
higher-ordered peaks were not well isolated, perhaps because the energy
interval was not uniform in dot ensembles in the area; the discrepancy in
the interval enhances the peak broadening at the excited sublevels. Another
reason might be that an additional emission appeared at the lower-energy
side in each peak. The spectrum around the ground-state emission is
magnified in Fig. 4.11(b). As the excitation power was increased from
90W/cm2 to 9kW/cmz, the peak emission intensity of the ground state
decreased and the lower-energy emission rose. This red shift of emission
might be due to the many-body effect in quantum dots (Sugawara, 1996).
The energy states (e.g., quantized energies and density of states) and
optical characteristics of quantum nano-structures can be artificially con-
trolled by changing the size of the structures. In quantum wells, well width
and strain are designed to match target devices using well-established k . p
band calculation methods (for example, see Chuang, 1995) and the designed
structures are precisely fabricated in the atomic scale by state-of-the-art
epitaxial growth techniques like MOVPE and MBE. The ability to control
quantum-dot structures is not yet as strong as it could be in spite of its
significance for practical device applications. In self-assembled InAs SK dots
on GaAs substrates, the size and emission wavelength can be controlled only
within a narrow range by changing the growth temperature (Chapter 2).
Changing the supply amount of sources has a negligible effect because the
assembling process is driven toward a thermodynamical equilibrium point.
In this section, we will introduce two methods of controlling the size and
the emission wavelength of InGaAs ALS quantum dots over a wide range
of 1.2 to 1.5pm. Then we will evaluate the quantum-dot confinement
potential by magneto-optical measurements.
1. TWO METHODS
OF CONTROLLING QUANTIZED
ENERGIES
4.2 K
Wavelength (pm)
FIG. 4.12. Photoluminescence spectra of the ALS dots grown by various cycles of In-As-
Ga-As alternate supply. As the number of cycles increased, the peak wavelength shifted from
1.17 to 1.3 pm. The second peaks were observed at the left side of the first peak. The maximum
emission intensity was obtained by the sample grown by 9 cycles (from Mukai et al., 1996a).
(C)
-
40nm
FIG.4.13. Plan-view TEM images of ALS dots grown with (a) 9 cycles, (b) 12 cycles, and
(c) 24 cycles. The dot diameter increased as the number of cycles increased.
- 35
n 1.05 - x
- 30
%
W
1st
0
1.00 ~
- 25
-
4.2 K
- 20
0.95l' '
lI
"
' I ' ' I " " I ' " * I " " ' I " 'I
5 10 15 20 25 30 35
E
1.45 F----A IntAsfGdAs x 12 cycles
3.
W
l1!
1.35
1.3 i/
c
0 0.05
--I
0.1
Indium composition of buffer layer
FIG. 4.15. Emission wavelength as a function of the indium composition of an InGaAs
buffer layer. With 12 cycles of In-As-Cia-As alternate supply, the emission wavelength increased
from 1.3 to 1.46pm as the indium composition increased from 0 (i.e., GaAs) to 0.09 (from
Ohtsuka and Mukai, 1995).
200 KOHKIMUKAIAND MITSURUSUGAWARA
2. MAGNETO-OPTICAL SPECTROSCOPY
Exciton diamagnetic shifts are powerful tools for evaluating not only the
reduced effectivemass but also the magnitude of potential affecting excitons.
The magnetic field confines excitons in the plane perpendicular to the field
and increases their energy. We can determine the magnitude of other
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 201
0 50 100 150
Square of magnetic field (T2)
injected into the laser active region. We must pay attention to the quality
of both inside dots and outside dots, since the retarded carrier relaxation
peculiar to quantum dots enhances the opportunity for carriers to be killed
before they relax. Here, we evaluate the nonradiative channel in quantum-
dot crystals by measuring time-resolved photoluminescence decay and
photoluminescence intensity as a function of temperature.
Recombination lifetimes of the discrete emission spectra from ALS dots
are shown in Fig. 4.18. (See Chapter 5 for details on time-resolved photo-
luminescence measurements.) The measured lifetimes, T " , are almost inde-
pendent of the emission level and are about 0.8 to 1.5 nanoseconds, which
we judge to be long enough for laser application. The lifetime is slightly
shorter than the calculated spontaneous emission lifetime of z S p =2.8 ns
using Eq. (1.210) of Chapter 1, with the exciton effect (the Coulomb
enhancement effect) neglected. We are not sure whether the slightly shorter
experimental lifetime is due to a simple error caused by missing factors in
the theory or to uncertain parameters in the calculation or to nonradiative
process in the measurements. If we estimate the nonradiative lifetime, z,,,
using the relationship of T,- = t,;’ + T,;' and z, = 0.8 - 1.5 ns, we get
T,, = 0.6 - 1.0ns. The measured lifetimes are almost independent of tem-
perature up to 300 K because of the discrete quantized levels that prevent
0
+0 0
4 54 'f .$
10-'O r 0 2nd +5th
A 3rd
- Calculation 4
suggesting that there are at least two sources for the reduction in photo-
luminescence intensity.
The temperature independence of the recombination lifetimes of Fig. 4.19
tells us where the nonradiative centers are located: The centers outside the
dots degrade emission efficiency as temperature increases. Since the excita-
tion wavelength was 647.1 nm, carriers were initially excited outside the dots
and diffused into the dots to emit inside them. Assuming nonradiative
centers outside the dots, we calculated the temperature dependence of
photoluminescence intensity. We neglected the radiative recombination
outside the dots, since no emission was observed from the GaAs substrate
and the In,,, ,Ga,,.,As layer surrounding the dots. The photoluminescence
intensity can thus be described using
where G is the carrier generation rate and zd is the carrier relaxation time
from outside to the ground state of the dots. Since only the emission from
the ground state was observed, we included the intersubband carrier
relaxation time in z d . The lifetime of zz:'(T) is the nonradiative recombina-
tion lifetime outside the dots and is written as
TABLE
4.1
REPORTEDPARAMETERSOF NONRADIATIVECENTERS IN InGaAs MATERIALS. E IS THE ACTIVATION
ENERGYFOR ne THERMAL RATE,B IS THE CAPTURE
EMISSION CROSS-SECTION, Q IS THE
ACTIVATION
ENERGY FOR THE CAPTURECROSS-SECTION, AND N IS THE TRAP DENSITY. “ V P E IS
AN ABBREVIATIONFOR VAPORPHASEEPITAXY. (From Mukai et al., 1997a.)
Growth
E (W B (cm2) Q (meW N(~m-~) Method
the sets in Table 4.1. Remarkably the parameter set of the MBE-grown dots
is much different from the sets in Table 4.1, which suggests that the point
defects outside the dots cannot explain the MBE-grown dots’ near-three-
order decrease. Other origins, such as the network of dislocation lines or the
nonradiative center inside the dots should be assumed to explain the large
parameters.
The high emission efficiency of ALS dots can be attributed to their unique
growth process. One possible reason for their better optical quality is that
the quality of their interface is higher. The ALS dots are surrounded
laterally by a quantum-well layer, and the interface around them is ambigu-
ous (see Fig. 4.1). This may be because the atoms around the dots
reconstruct the interface during the growth, given that ALS dots are
supposedly formed through the repetition of an equilibrium point and a
non-equilibrium point of thermodynamical balance between strain energy
and surface energy (see Chapter 3). On the other hand, the interface of the
SK dots is abrupt, as it is exactly the border between the growth islands and
the overgrown layer. Another possible reason for the better optical quality
of the ALS dots is that they are in-situ annealed due to the low growth rate
of the alternate supply of precursors. The growth rate was about 1
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 207
V. Summary
REFERENCES
Ambrose, P. W.. Goodwin, M. P., Martin, C . J., and Keller, A. R. (1994). Phys. Reu. Lett. 72,
160.
Bimberg, D., Ledentsov, N. N., Kirstaedter, N., Schmidt, O., Grundmann, M., Ustinov, M., V.,
Egorov, Y. A., Zhukov, E. A., Maximov, V. M., Kop’ev, S. P., and Alferov, I., 2 . (1995).
Ext. Abs. 1995 Int. Con$ Solid Slate Devices and Materials, Osaka.
Brunner, K., Bockelmann, U., Abstreiter, G., Walther, M., Bohm, G., Trankle, G., and
Weimann, G. (1992). Pliys. Rev. Leu. 69, 3218.
Betzig, E., Trautman, K. J., Harris, D. T., Weiner, S. J., and Kostelak, L. R. (1991). Science 251,
1468.
Betzig, E., Finn, L. P., and Weiner, S. J. (1992). Appl. Phys. Lett. 60, 2484.
Betzig, E., and Chicheser, J. R. (1993). Science 262, 1422.
Chuang, S. L. (1995. Physics ofOptoelectronics Devices, John Wiley & Sons, New York.
Daruka, I., and Barabasi, A.-L. (1997). Phys. Rev. Lett. 79, 3708.
Diirig, U., Pohl, W. D., and Rohner, F. (1986) J . Appl. Phys. 59, 3318.
Grober, D. R., Harris, D. T., Trautrnann. K. J., Betzig, E., Wegscheider, W., Pfeiffer, L., and
West, K. (1994). Appl. Phys. Lett. 64. 1421.
Ghaemi, F. H., Goldberg, B. B., Cates, C.. Wang, D. P., Sotomayor-Torres, M. C., Fritze, M.,
and Nurmikko, A. (1995). Superlattices and Microstructures 17, 15.
Hessman, D., Castrillo, P., Pistol, E. M., Pryor, C., and Samuelson, L. (1996). Appl. Phys. Lett.
69, 749.
Leonard, D., Kishnamurthy, M., Reaves, M. C., Denbaars, P. S., and Petroff, M. P. (1993). Appl.
Phys. Lett. 63, 3203.
Marzin, Y. J., Gerard, M. J., Izrael, A,. Barrier, D., Bastard, G. (1994). Phys. Rev. Lett. 73, 716.
Moison, M. J., Houzay, F., Barthe, F., and Leprince, L. (1993). Appl. Phys. Lett. 64, 196.
Mukai, K., Ohtsuka, N., Sugawara, M., and Yarnazaki, S. (1994). Jpn. J . Appl. Phys. 33, L1710.
Mukai, K., Ohtsuka, N., and Sugawara, M. (1996a). Jpn. J . Appl. Phys. 35, L262.
Mukai, K., Shoji, H., Ohtsuka, N., and Sugawara, M. (1996b). Appl. Phys. Lett. 68, 3013.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1996~).Phys. Rev. B. 54, R5243.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara. M. (1997a). Appl. Surf Sci. 112, 102.
Mukai, K., Ohtsuka, N., and Sugawara. M. (1997b). Ext. Abs. 58th Autumn Meet. Japan
Society of Applied Physics, 4pS-8 (in Japanese).
Mirin, R. P., Ibbetson, P. J., Nishi, K., Gossard, C. A., and Bowers, E. J. (1995). Appl. Phys.
Lett. 67, 3795.
Nakata, Y., Sugiyama, Y., Futatsugi, T.. and Yokoyama, N. (1997a). J . Crystal Growth 175/176,
713.
Nakata, Y., Sugiyama, Y., Futatsugi, T., Mukai, K., Shoji, H., Sugawara, M., Ishikawa, H., and
Yokoyama, N. (1997b). Ext. Abs. 58th Autumn Meet Japan Society of Applied Physics,
2pM-4 (in Japanese).
Ohtsuka, N., and Mukai, K. (1995). Proc. Int. Con$ InP and Related Materials, Hokkaido.
Sakuma, Y., Ozeki, M., and Nakajima, K. (1993). J . Cryst. Growth 130, 147.
Saiki, T., Mononobe, S., Ohtsu, M., Saito. N., and Kusano, J. (1996). Appl. Phys. Lett. 68,2612.
Sugawara, M. (1996). Jpn. J . Appl. Phys. 35, 124.
Sugawara, M., Okazaki, N., Fuji, T., and Yamazaki, S. (1993). Phys. Rev. 8. 48, 8102.
Tabuchi, M., Noda, S., and Sasaki, A. (1992). Science and Technology of Mesoscopic Structure,
Springer-Verlag, Tokyo, p. 379.
SEMICONDUCTORS AND SEMIMETALS, VOL 60
CHAPTER5
I. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
11. A MODELOF THE CARRIER RELAXATION PROCESS IN QUANTUM DOTS . . . . . 21 1
111. EXPERIMENTS ON LIGHTEMISSION AND CARRIER RELAXATION I N QUANTUM-DOT
DISCRETE ENERGY LEVELS . . . . . . . . . . . . . . . . . . . . . . . 2 14
1. Elertroluminescence Spectra . . . . . . . . . . . . . . . . . . . . . 2 15
2. Time-Resolved Photoluminescence . . . . . . . . . . . . . . . . . . 2I I
3. Simulation of Electroluminescence Spectra . . . . . . . . . . . . . . . 225
IV. INFLUENCEOF THERMAL TREATMENT . . . . . . . . . . . . . . . . . . . 229
1. Change in Emission Spectra after Annealing . . . . . . . . . . . . . . 229
2. Competition between Carrier Relr.rution and Recombination . . . . . . . 231
V. SIMULATION OF LASERPERFORMANCE. INCLUDING THE AUGERRELAXATION
PROCESS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
VI. SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
I. Introduction
Energy
.........................
I
I Nonradiative
I channel
I
I
I
...
(Defect level)
I
-Y.--
Radiative
channel
FIG.5.1. Schematic of the phonon bottleneck. Discrete sublevels in quantum dots may
hinder carrier relaxation toward the ground level. In this case. carriers will recombine to emit
at higher levels or go to the nonradiative channel during a carrier cascade toward the ground
level.
5 PHOTON EFFECTIN QUANTUM
BOTTLENECK DOTS 211
The carrier relaxation process into quantum dots is actually two pro-
cesses, as shown in Fig. 5.2. One is the carrier relaxation from continuous
energy levels into quantum-dot discrete levels (A). The other is the relax-
ation between the discrete levels inside dots (B). In many optical experi-
ments as well as in quantum-dot lasers, carriers undergo these two processes
212 KOHKIMUKAI
AND MITSURUSUGAWARA
FIG. 5.2. Two cases of carrier relaxation in a quantum dot: (A) the relaxation from
continuous levels and (B) the relaxation between discrete levels. It is expected that case (A) will
occur faster than case (B) since the energy conservation rule is more severe for the latter case.
except when they are brought directly into the discrete levels by resonant
excitation or tunneling (Kamath et al., 1997a). Since the energy conservation
rule must be satisfied for carrier relaxation, relaxing carriers transfer a
corresponding energy to other particles such as phonons and to other
carriers. Thus, the relaxation rate strongly depends on the density of final
levels and on the number of particles other than the transition matrix
elements.
Figure 5.3 illustrates the two representative processes: (a) a single- or
multiphonon process, and (b) an Auger process. Inoshita and Sakaki (1992)
showed that two-phonon processes, including LO and LA phonons, reduce
the requirements of energy conservation, thus drastically decreasing the
lifetime to subpicosecond order. Even so, this relaxation process cannot send
carriers deep inside the dots, since the transition matrix for the relaxation
processes that include more than three phonons will be greatly reduced.
Therefore, carriers once trapped in the upper levels of the dots relax into
lower energy levels by emitting phonons step by step. The Auger process of
Fig. 5.3(b) works effectively when there are many carriers, since an electron
can find another electron or a hole into which to transfer its energy and thus
fall into the dot energy levels. This process works to further relax carriers
trapped in the discrete levels into lower energy regions, as long as there is
still a larger number of carriers outside the dots. For that reason the Auger
process is more effective for carrier relaxation into deep sublevels.
Figure 5.4 shows an experimental study of the rising characteristics of
photoluminescence for SK dots as a function of excitation power (Ohnes-
orge et al., 1996). The rising time depends on the optical excitation density.
For an excitation density between lo-' and 4, rising time stays almost
5 PHOTON BOTTLENECKEFFECT
IN QUANTUM
DOTS 213
FIG. 5.3. Schematic of two processes of carrier relaxation from continuous levels: (a) the
multi-phonon process and (b) the Auger process. Usually, the energy separation between the
continuous level and the sublevel in a dot does not exactly match the energy of a single phonon.
There can be a combination of LO and LA phonons during relaxation from the continuous
energy level, but this is unlikely for deep sublevels. The Auger process is effective when carrier
density is high.
120 4
I
T=5K
100 I
e 80
.-
a 40
0
.. 0
2oL I ' ’’mns-s’ ' ' ' 1 4 1 1 6 1 ' ' ' s m m a m ' 8 1
1. ELECTROLUMINESCENCE
SPECTRA
Figure 5.5 shows the electroluminescence spectra for (a) 300 K and (b) 77 K.
To fabricate diode structure, a 0.5-pm n-GaAs layer, a 1-pm n-InGaP layer,
an ALS-dot layer sandwiched by 100-nm GaAs layers, a l-pm p-AlGaAs
layer, and a 0.5-pm p-GaAs layer were grown on a (001) GaAs substrate.
The luminescence parallel to the sample surface was dispersed and detected
with a lock-in technique using an InGaAs photo-multi detector that was
kept at -70°C. The electrode size was 20 x 900pm’. The current-
injected area was estimated by the near field pattern, and according to this
estimation, it was spread out over 150 to 200 x 900pm2 in the quantum-dot
layer. Three emission peaks corresponding to the interband transition
between the discrete levels of the conduction and valence bands appeared at
300 K when the injected current was increased, five peaks appeared at 77 K.
In Chapter 4, we pointed out the phonon bottleneck from the emission
spectra at 77 K. We should note that, at lOmA, the 2nd and 3rd peak
emissions exceeded the first peak emission much earlier than the first
emission intensity saturated, indicating that the relaxation rate into the
ground level is comparable to the spontaneous emission rate of the excited
levels.
The electroluminescence spectra of Fig. 5.5 varied between the two
temperatures, which indicates that some features of carrier dynamics depend
on temperature. For example, when the emission intensity of the ground
level reached half maximum, the intensity of the second level was stronger
than that of the ground level at 77 K (10 mA), but the two were almost equal
at 300 K (50 m A).
Random carrier capturing into the dots was proposed by Grundmann
and Bimberg (1997) as an alternative interpretation for the multiple
emission peaks (Mukai et al., 1996b; Lipsanen et al., 1995). They calculated
the emission spectra using a Poisson distribution to count the number of
216 KOHKIMUKAIAND MITSURU
SUGAWARA
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 * 1 1 1
carriers that were trapped in quantum dots. As a result, they claimed that,
even if the retarded carrier relaxation is not taken into account, the emission
from high energy levels appears even before the lower energy level emission
is saturated. However, their calculation of the emission spectra cannot
explain the emission spectra we measured, where the third-level emission
exceeded the ground-level emission before the ground-level intensity
saturated. Above all, the random capture model offers no explanations for
the difference in the emission spectra between 77 K and 300 K.
PHOTOLUMINESCENCE
2. TIME-RESOLVED
0 2 4 6 8
Time (ns)
FIG. 5.6. Example of luminescence decay of the five levels at 250 and 20K. The higher the
temperature, the faster the decay; the lower a level, the slower the decay (from Mukai et al.,
1996b. Copyright 1996 by the American Physical Society).
A B C D
[E]I?[I I
a I
DI
I I I I
t l I I I I j
3
0
> NBO
3
0
NAO
c
.3
D
k
E
1
c
I I I
reaching the final destination of D. Moreover, two paths, B, and B,, are
possible for case B, and a single path, C,, is possible for case C. Radiative
decay is allowed in both levels on the assumption that hole quantization
energies are less than k,T due to the comparative heavy mass of holes
(Benisty et al., 1991).
The carrier number in the excited level is counted by evaluating the paths
beginning from case A and the paths beginning from case B. Let NAA(t)be
the number of A at time t; and NAB@),the number of B formed from A as
a result of the recombination of the ground-level carrier. The carrier number
in the excited level in dots when the initial situation is A is described by
where zreLis the carrier relaxation lifetime from the excited to the ground
level. Solving Eqs. (5.2) and (5.3), we obtain
(5.4)
where NAo is the initial number of case-A dots. When zrel<< z,, NAB(t)
becomes zero and the decay lifetime of N A ( t ) reduces to 2 , / 2 . When zrel is
not negligible compared with z, the decay of NAB(t)is not negligible and
the decay lifetime of NA(t) is larger than r r / 2 .
For the carrier number in the excited level when the initial situation is B,
5 PHOTON EFFECTIN QUANTUM
BOTTLENECK DOTS 221
and we obtain
where N,, is the initial number of case B. The decay lifetime of NB(t) is
(T~;: +
~ ~ - ‘ ) -and
l gets closer to sret when z~, << T ~ .As a result, the total
number of carriers in the excited level is given by
(5.10)
(5.11)
and
(5.12)
222 KOHKIMUKAI
AND MITSURUSUGAWARA
(5.13)
(5.14)
(5.15)
f
where N,, is the initial number of carriers in case C. Finally, the total
number of carriers in the ground level is given by
When treI << t,, Eqs. (5.13) through (5.15) become zero and the decay time
of Ng,(t) reduces to t,. In the same case, we saw that the decay lifetime of
NA(t) reduces to 2,/2. However, in Fig. 5.6, the later moderate decay in the
second level is not half of the decay in the ground level. This indicates that
the relaxation lifetime is not negligible compared with the recombination
lifetime and/or that the recombination lifetime may not be strictly uniform
among the sublevels.
We analyzed the measured decay curves of Fig. 5.6 by the double-
exponential formulation of
(5.18)
after deconvoluting the decay curves from the excitation pulse. The second
term of Eq. (5.18) is of course neglected in considering the decay of the
ground level. Here, from the estimation by the RIO model we assumed the
initial steep decay component to be dominated by the carrier relaxation and
the subsequent slow decay component to be by the recombination. In strict
treatments, the degeneracy and the selection rule in relaxation for the
multiple sublevels must be taken into account in solving the rate equations.
However, Eq. (5.18) is a good approximation because the initial situation of
the dots is roughly categorized into two cases: when the lower level is empty
and when the lower level is filled.
5 PHOTONBOTTLENECK
EFFECTIN QUANTUM
DOTS 223
Y
.3
4
;10
0
Y
.C
m
c
._
10p-
0 SO 100 150 200 250 300
Temperature (K)
(b)
Re 8 ,
FIG. 5.8. Carrier lifetimes of sublevels as a function of temperature: (a) recombination
lifetime and (b) relaxation lifetime. The recombination lifetime is independent of temperature
and about 1 ns in all sublevels. The higher the temperature and the higher the order of a level,
the faster the carrier relaxation (from Mukai et al., 1996b. Copyright 1996 by the American
Physical Society).
224 KOHKIMUKAIAND MITSURUSUGAWARA
(5.19)
(5.20)
and
(5.22)
where zi0 is the intrinsic carrier relaxation lifetime at the ith level. Using the
lifetimes in Fig. 5.8, we calculated the temporal evolution of the carrier
number in the sublevels (Fig. 5.9). To estimate the fastest rise under
step-by-step relaxation, we assumed ,h = 0 in the calculation and found that
the delay of rising at the lower-ordered level cannot be neglected. The delay
is about 2 ns between the fifth and the ground level, and is approximately
equal to the simple sum of the relaxation lifetimes of the sublevels.
5 PHOTON
BOTTLENECK
EFFECT
IN QUANTUM
DOTS 225
I"
0 0.5 1 1.5 2
Time (ns)
FIG. 5.9. Calculated rise of the carrier number when the direct carrier relaxation from the
continuous level is negligible. The rise slowed as the level order decreased. The delay is
approximately equal to the simple sum of the relaxation lifetimes of the sublevels.
Comparing the experimental results of Fig. 5.6 and the calculated results of
Fig. 5.9, it is clear that the assumption of the step-by-step relaxation process
cannot explain the fast rising of photoluminescence in the time-resolved
measurements. The RIO model, which assumes fast carrier setup in quan-
tum dots resulting from carrier-carrier scattering and the subsequent slow
decay inside the dots, successfully explains the decay curves of time-resolved
photoluminescence.
3. SIMULATION OF ELECTROLUMINESCENCE
SPECTRA
N5
- +-
N5 - G = O , (5.24)
55 ‘r
Ni N . N , , ,
-+>--- - 0 (i = 2, 3,4), (5.25)
Ti zr Ti+ 1
and
(5.26)
where G is the carrier injection rate from the continuous level into dots.
Here, we consider that the recombination lifetime, zr, is common to all
levels, as seen in Fig. 5.8(a), and disregard nonradiative recombination in
our high-quality samples (Chapter 4).Assuming a Gaussian distribution for
the broadening of the emission spectrum, we calculated the luminescence
spectra by
(5.27)
where E is the energy of the photon, E , is the energy at the ith level, and r
is an inhomogeneous broadening factor. Figures 5.10(a) and 5.10(b) show
the results of the calculation for 300K and 77K, respectively. During
calculation, we used an in-plane dot density of 5%, a broadening factor of
50meV, T~ = 1 ns, and D i= 2 . i ( i = 1 to 5), taking the spin into consider-
ation. The intrinsic relaxation lifetimes used were zo2 = 400, zo3 = 200,
T~~ = 50, and z05 = 20ps for 300K; and T~~ = 1000, T~~ = 500, T~~ = 200,
and z 0 5 = 90 ps for 77 K.
The calculations of Fig. 5.10 explain the measured electroluminescence
spectra of Fig. 5.5. The relaxation lifetime is short enough that the ground-
level emission is dominant for small currents. As the number of injected
carriers increases, the carrier relaxation rate decreases and emission from
higher levels occurs. Since the relaxation lifetime is closer to the radiative
recombination lifetime at 77K than at 300K, the emission from higher
levels at 77 K occurs for a smaller current. At 77 K, the calculated current
values match the measured values. At 300K, the trend of relative peak
intensities is well explained, but the values for the current by the calculation
are much smaller than the measured values. This discrepancy grows larger
as the current increases- a tendency that can be attributed to the consump-
5 PHOTON BOTTLENECK DOTS
EFFECTIN QUANTUM 221
Energy (eV)
(b)
FIG. 5.10. Calculated electroluminescence spectra at (a) 300 and (b) 77 K using the rate
equations of the five levels, The results of the calculation correlate with the measured spectra
in 5(a) and (b) (from Mukai et al., 1996b. Copyright 1996 by the American Physical Society).
Current (mA)
(b)
FIG. 5.1 1. Value of f; as a function of injected current for (a) 300 K and (b) 77 K. The
ground, second, and third levels were not completely filled, and the maximum occupation factor
of these levels depended on the temperature.
levels are not completely filled and that the characteristic features depend
on temperature. For example, the maximums of f i for the ground and the
second levels are about 80% at 300 K and about 60% at 77 K. For the third
level, these maximum are 90% at 300 K and 80% at 77 K. While the
emission from the ground level is dominant, the maximum occupation factor
at the ground level is 60% for 300 K. In quantum-dot lasers, the suppression
of carrier filling may limit gains needed for lasing at the ground level (Shoji
et al., 1995).
5 PHOTON EFFECTIN QUANTUM
BOTTLENECK DOTS 229
1. CHANGE
IN EMISSION
SPECTRA
AFTER ANNEALING
""'""'"""""""'1
h
Y
.r(
2
Y
E
H
minutes. In the as-grown sample, three peaks can be observed that are due
to the interband transition between the discrete levels of the conduction and
valence bands. The excitation power is very low, as the emission intensity of
the ground level in Fig. 5.12 is less than one-fifth of the maximum value
reached at a higher excitation. Annealing greatly reduces the integrated
photoluminescence intensity, broadens the spectra, and slightly moves the
emission peak toward higher energies. What is most remarkable is that the
relative peak intensity between the ground- and second-level emission is
greatly changed after annealing: While the ground-level emission is about
three times stronger than the second-level emission in the as-grown sample,
the two peaks are almost comparable in their intensity after 30-minute
annealing. This is not due to drastic structural changes and the resultant
reduction of the transition oscillator strength, since the emission energy shift
is only about 5 meV and the TEM image of the 30-minute-annealed samples
shows no significant changes in dot structure, size, and numerical density.
One possible explanation for this phenomenon is that the carrier dynamic
process in quantum dots is influenced by the thermal treatment, which leads
to the change in the relative intensity.
Figure 5.13 shows the energy of the ground- and second-level emission
peaks, as well as the relative emission intensity of the ground level against
the second level, as a function of the annealing temperature. The energy shift
occurs at over 600°C because of interdiffusion, and simultaneously the
relative intensity of the ground-level emission is greatly reduced. The figure
also shows that the energy shift is larger in the second peak and that the
level separation is increased by annealing. This contradicts our experiences
in the study of interdiffusion in quantum wells, where the level separation
always decreases after interdiffusion, which spreads the confinement poten-
tial in the growth direction. This unique feature can be attributed to
structural asymmetry between the in-plane and perpendicular directions in
ALS quantum dots with a disk-like shape because of the asymmetry, the
interdiffusion in the early stage occurs faster at the edge of the dot than in
the central region (Sallese et al., 1995). Since this corresponds virtually to
the decrease in the in-plane diameter of the dots, the energy separation
increases.
6 ' I I " I " ' ' 1 ' I ' ' I " ' " " I ' I ' ' ' 1 ' ' I " 'g
h
3
'a1
d
W
2
5
.H
Ei
B
E
Y
0 2 4 6 8 10 12 14 16 18 20
Time (ns)
FIG.5.14. Examples of decay curves of the ground and second levels after annealing at
615°C for 60 minutes. The decay time of the ground level decreased as the temperature rose,
but was insensitive to temperature in as-grown samples. The decay curves of the second level
were double-exponential(from Mukai el al.. 1998).
(5.28)
where znlis the nonradiative lifetime at the ground level. Figure 5.16 shows
the photoluminescence spectra as calculated using the nonradiative lifetime
as a parameter. We used a carrier injection rate of G = s - l per indi-
vidual dot, a broadening factor with a Gaussian distribution of 50 meV,
z, = 1 ns, Di = 2 . i (i = 1,2,. . . , 5 ) , T~~ = 200, zo3 = 100, T~~ = 50, and zo5 =
BOTTLENECK
5 PHOTON EFFECTIN QUANTUM
DOTS 233
j
G-
v
10-8 . . , . .. . , . . . . , . , . . , . I I . , .., . i
't
0
300
Temperature (K)
o so 100
Temperature (K)
is0
FIG. 5.15. Lifetimes up to the third level as a function of temperature: (a) recombination
200
A
-
250
:t300
lifetimes and (b) relaxation lifetimes. The solid and open symbols indicate lifetimes of as-grown
and annealed samples, respectively. The recombination lifetimes in the ground level became
temperature-dependent after annealing and decreased to a few hundred picoseconds near room
temperature. The relaxation lifetimes decreased slightly after annealing (from Mukai et al.,
1998).
occupation at the ground level, fi, is about 8%. As the nonradiative lifetime
decreases from infinity to a few hundred picoseconds, the emission intensity
of the ground level decreases and becomes comparable to that of the second
level. The value of the nonradiative lifetime in the simulation matches the
value estimated in Fig. 5.15, using the recombination lifetime of 200 ps and
the radiative lifetime of 1 ns.
The carrier dynamics in the ALS dots are summarized in Fig. 5.17. The
relative emission intensity of the ground and second levels is determined by
the equilibrium between the radiative recombination process (rl, r2), the
relaxation process (rel), and the nonradiative recombination process
(nl, n2). In an as-grown sample, the nonradiative lifetimes are negligible,
which can be expected from the temperature-independent radiative lifetimes
(see Section 111). After annealing, the recombination lifetime at the ground
level decreases due to the nonradiative channel, while the nonradiative
process of the second level is still negligible. The drastic change in relative
emission intensity demonstrates that the balance of radiative and non-
radiative recombination and relaxation rates determines the shape of the
emission spectrum. This contradicts our experience with quantum wells,
whose emission spectra generally become narrow toward a lower energy
5 PHOTON BOTTLENECK
EFFECT DOTS
IN QUANTUM 235
Energy
t
I
I
I
I
r2
channel
channel
>Radiative
channel
FIG. 5.17. Schematic of carrier dynamics in a quantum dot before and after annealing. The
relative emission intensity of the ground and second levels was determined by the equilibrium
between the radiative recombination process ( r l . r2), the relaxation process ( r e / ) , and the
nonrddiative recombination process (rrl, n2). The nonradiative lifetimes, T,, and rn2, were
negligible for the as-grown sample. After annealing, the nonradiative lifetime T , , ~ .became
effective. When carriers were not sufficiently supplied to the ground level, the emission intensity
of the ground level decreased compared to that of the second level.
From what we have seen in the experiments and their respective analyses,
we can construct a framework of the carrier relaxation problem as follows.
First, when there are a large number of carriers of 10'2-10'3 cm-2 around
quantum dots, as in the initial stage of time-resolved photoluminescence, the
carrier capture into dots occurs with a considerably short lifetime. Also,
carriers are relaxed into deep low-order levels through the Auger processes.
When there are only a few carriers of less than 10" cmP2, they are first
trapped at the high energy levels and relax toward the lower energy levels
later. During quantum-dot laser operation, the actual relaxation process
236 KOHKIMUKAIAND MITSURUSUGAWARA
varies between the above mentioned two cases as the current injection
increases.
Let us here introduce the Auger effect into the simulation presented in
Section V.l of Chapter 1. The same rate equations are used, and we take
into account the dependence of the carrier injection time into dots, zo, on
the carrier density in the quantum well Nq as
where C is the Auger coefficient. Figure 5.18(a) shows the output power as
a function of the injected current for C = 1.0 x cm4/s and C = 0
with zC0= 100 ps. The inhomogeneous broadening is To = 20 meV, and the
coverage is = 10%. The recombination lifetimes we used are zr = zSpon=
2.7 ns and z~~= 3 ns (solid lines), and 7 r = 0.5 ns and z~~ = 0.5 ns (dashed
lines). It is clear that the Auger process increases efficiency, lowers threshold
current, and enables higher output power (again, calculation stops when the
quantum-well carrier density reaches 1.0 x 10l2cm2). Figure 5.18(b) shows
the quantum-well carrier density as a function of injected current. For C =
cm4/s, zo is superimposed. We see that z0 decreases close to 10ps as
the injected current increases and that the increase in carrier density is
-- l r =0.5 ns, T ~ ~ ns
0.5 =
... . .me
witnout . rr .
Auger enecr.-. . . . -- .-
FIG.5.18. (a) Output power as a function of the injected current calculated with and
. * ..
I ne spectrum oroaaening was ro= LU mev, ana me coverage was
I .
5 = lo%, which gave P = 0.84 at the lasing threshold (from Sugawara, 1998). (b) Carrier
density in a quantum well located around the quantum dots as a function of injected current.
The carrier relaxation lifetime into dots decreased close to l o p s as the injected current
increased when C = 1 x lo-'' cm"/s (from Sugawara, 1998).
5 PHOTON BOTTLENECK
EFFECT DOTS
IN QUANTUM 231
, , 1 1 1 1 1
10-
Auger coefficient (cm4/s)
FIG. 5.19. Maximum output power as a function of the Auger coefficient for various
quantum-dot coverages at Po = 20 meV (from Sugawara, 1998).
VI. Summary
REFERENCES
Adler, F., Geiger, M., Bauknecht, A., Scholz, F., Schweizer, H., Pilkuhn, H. M., Ohnesorge, B.,
and Forchel, A. (1996). J. A p p l . Phys. 80, 4019.
Anopelto, J., Yamaguchi, A. A,, Nishi, K., Usui, A., and Sakaki, H. (1993). Jpn. J . Appl. Phys.
32, L32.
Apetz, R.,Vescan, L., Hartmann, A,, Dieker, C., and Luth, H. (1995).A p p l . Phys. Lett. 66,445.
Arakawa, Y., and Sakaki, H. (1982). Appl. Phys. Lett. 40, 939.
Arakawa, Y., Vahala, K., and Yariv, A. (1994). Appl. Phys. Lett. 45, 950.
Asada, M., Miyamoto, Y.,and Suematsu, ?. (1986). IEEE J . Quantum Electron. QE-22, 1915.
Asada, M. (1989). IEEE J . Quantum Electron. 25, 2019.
Benisty, H., Sotomayor-Torres, M. C., and Weisbuch, C. (1991). Phys. Rev. B. 44, 10945.
Benisty, H. (1995). Phys. Rev. B. 51, 13281.
Bockelmann, U., and Bastard, G. (1990). Phys. Rev. B. 42, 8947.
Bockelmann, U., and Egeler, T. (1992). Phys. Rev. B. 46, 15574.
Bockelmann, U. (1993). Phys. Rev. B. 48, 17637.
Bockelmann, U., Heller, W., Filoramo, A,, and Roussignol, P. (1997). Phys. Rev. B. 55, 4456.
Brus, L. (1986). IEEE J. Quantum Electron. 22, 1909.
Chemla, D. S., and Miller, B. A. D. (1986). O p t . Lett. 11, 522.
Eaglemann, D. J., and Cerullo, M. (1990). Phys. Reti. Lett. 64, 1943.
Efros, A. L., Khachenko, A. V., and Rosen, M. (1995). Solid Stute Cornmun. 93, 281.
Gerard, J.-M. (1 995). Con$ned Electrons and Photons: New Physics und Applicutions. Plenum,
New York, p. 357.
Grundmann, M., and Bimberg, D. (1997). Phys. Rer. B. 55,9740.
Grosse, S., Sandmann, H. H. J., Plessen, V. G., Feldmann, J., Lipsanen, H., Sopanen, M.,
Tulkki, J., and Ahopelto, J. (1997). Phys. Rev. B. 55, 4473.
Inoshita, T., and Sakaki, H. (1992). Phys. Rev. B. 46, 7260.
Kamath, K., Klotzkin, D., and Bhattacharya. P. (1997a). Proc. Confi IEEE Lasers and
Electro-Optics Society, p. 498.
Kamath, K., Chervela, N., Linder, K. K., Sosnowski, T., Jiang, H.-T., Norris, T., Singh, J., and
Battacharya, P. (1997b). Appl. Phys. Len. 71, 927.
Kurtenbach, A., Ruhle, W. W., and Eberl. K. (1995). Solid State Corn. 96, 265.
5 PHOTON BOTTLENECK DOTS
EFFECTIN QUANTUM 239
Leonard, D., Krishnamurthy, M., Reaves, M. C., DenBaars, P. S., and Petroff. M. P. (1993).
Appl. Phys. Lett. 63, 3203.
Lipsanen, H., and Sopanen, M. (1995). Phys. Rev. B. 51, 13868.
Moison, J. M., Houzay, F., Barthe, F., Leprince, L., Andre, E., and Vatel, 0. (1994). Appl. Phys.
Lett. 64, 196.
Mukai, K., Ohtsuka, N., Sugawara, M., and Yamazaki, S. (1994). Jpn. J . Appl. Phys. 33, L1710.
Mukai, K., Shoji, H., Ohtsuka, N., and Sugawara, M. (1996a). Appl. Phys. Lett. 68, 3013.
Mukai, K., Ohtsuka, N., Hajirne, S., and Sugawara, M. (1996b). Phys. Rev. B. 54, R5243.
Mukai, K., Sugawara, M. (1998). Jpn. 1. Appl. Phys. 37, 5451.
Nakayama, H., and Arakawa, Y. (1994). 7th Int. Con$ Superlattices, Microstructures, and
Microdevices, Banff, Canada.
Ohnesorge, B., Albrecht, M., Oshinowo, J., Forchel, A., and Arakawa, Y . (1996). Phys. Rev. B.
54, 11532.
Oshinowo, J., Nishioka, M., Ishida, S., and Arakawa, Y. (1994). Appl. Phys. Lett. 65, 1421.
Raymond, S., Fafard, S., Charbonneau, S., Leon, R., Leonard, D., Petroff, M. P., and Merz, L. J.
(1995). Phys. Rev. B. 52, 17238.
Raymond, S., Fafard, S., Poole, J. P., Wojs, A., Hawrylak, P., Charbonneau, S., Leonard, D.,
Leon, R., Petroff, M. P., and Merz, L. J. (1996). Phys. Rev. B. 54, 12548.
Sallese, J. M., Carlin, F. J., Gaihanou, M., and Grunberg, P. (1995). Appl. Phys. Lett. 67, 2633.
Sercel, P. C. (1995). Phys. Rev. B. 51, 14532.
Shoji, H., Mukai, K., Ohtsuka, N., Sugawara, M., Uchida, T., and Ishikawa, H. (1995). I E E E
Photo. Tech. Lett. 7, 1385.
Sugawara, M. (1998). Prnc. of Optoelectronics '98, Photonic West. San Jose.
Tabuchi, M., Noda, S., and Sasaki, A. (1992). Science and Technology of Mesoscopic Siructures
(Springer-Verlag. Tokyo, p. 379.
Tsang, W. T. (1978). Appl. Phys. Lett. 33, 245.
Uskov, A. V., Mcinerney, J., Adler, F., Schweizer, H., and Pilkuhn, M. H. (1998). Appl. Phys.
Lett. 72, 58.
Utzmeier, T., Postigo, A. P., Tamayo, J., Garcia, R., and Briones, F. (1996). Appl. Phys. Lett.
69, 2674.
Vollrner, M., Mayer, J. E., Riinle, W. W., Kurtenbach, A,, and Eberl, K. (1996). Phys. Reu. B.
54, R 17292.
Wang, G., Fafard, S., Leonard, D., Bowers, E. J., Merz, L. J., and Petroff, M. P. (1994). Appl.
Phys. Lett. 64. 2815.
Yu, H., Lycett, S., Roberts, C., and Murray, R . (1996). Appl. Phys. Lett. 69, 4087.
This Page Intentionally Left Blank
SEMICONDUCTORS AND SEMIMETALS, VOL. 60
CHAPTER6
I. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . 24 1
11. FUNDAMENTAL PROPERTIES OF Q U A N T U M - D O T LASERS . . . . . . . . . . . 243
1. Gain Characteristics . . . . . . . . . . . . . . . . . . . . . . . . 244
2. Threshold Current . . . . . . . . . . . . . . . . . . . . . . . . . 246
3. Dynamic Characteristics . . . . . . . . . . . . . . . . . . . . . . 248
111. FABRICATION OF SELF-ASSEMBLED QUANTUM-DOT LASERS. . . . . . . . . . 249
1. Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
2. Device Characteristics . . . . . . . . . . . . . . . . . . . . . . . 255
3. Limiting Factors of’Laser Performance . . . . . . . . . . . . . . . . 261
IV. KEYTECHNOLOGIES FOR THE NEXT ERA . . . . . . . . . . . . . . . . . 269
I . Close1.v Slacked Quantum-Dot Lusers . . . . . . . . . . . . . . . . . 270
2. Columnur Quantum-Dot La~ers . . . . . . . . . . . . . . . . . . . 213
3. Long- Wavelength Quantum-Doi Lasers . . . . . . . . . . . . . . . . 216
4. Quantum-Dot Vertical-Cavity Surface-Emitting Lasers . . . . , . . , , . 219
V. CONCLUSION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Acknowledgments . . . . . . . . . . . . . . . . . . , , . . , . . , . 283
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
I. Introduction
tor lasers equipped with quantum dots as the active region are expected to
exhibit ultralow threshold current, temperature-insensitive operation, nar-
row spectral linewidth, and large modulation bandwidth (Arakawa and
Sakaki, 1982; Asada et al., 1986). All of these features are attributed to the
quantum dots’ enhanced gain properties, especially the extremely high
differential gain, induced by their three-dimensionally discrete density of
states.
Despite its theoretically predicted superiority, no quantum dot laser has
been actually realized. This is because quantum structures fabricated so far
have not produced a strong carrier confinement, especially in a lateral
direction. In the last decade, two approaches have been tried for realizing
quantum dot structures. The first one is a precise patterning technique
(Hirayama et al., 1994), originally developed for quantum-wire structures. It
involves patterning of quantum wells by wet or dry etching and overgrowth
on the patterned structure, or selective growth on a patterned substrate
(Kash et al., 1986; Fukui et al., 1991). Because of recent progress in electron
beam lithography and etching techniques, InGaAs-InGaAsP quantum-dot
lasers became a reality in 1994 (Hirayama et al., 1994). However, the
threshold current density was as high as 7.6 kA/cmZ even at 77 K. In this
approach, it is generally very difficult to obain a high-quality structure
without damage, since nonradiative recombination, especially at the surface
of etched sidewalls where the active region is exposed to air, severely limits
laser performance.
The second approach is self-organization of quantum dots during highly
strained heteroepitaxy, and it has been very successful. Uniform dots with
small size, high density, and high emission efficiency, unachievable with
conventional techniques, have been easily realized in conventional molecu-
lar beam epitaxy (MBE) and metalorganic vapor phase epitaxy (MOVPE).
The self-organization technique has lead to three methods for fabricating
quantum-dot structures: (1) spontanous formation of In(Ga)As islands on
GaAs in the Stranski-Krastanow (SK) mode (Leonard et al., 1993; Kirstaed-
ter et al., 1994; Oshinowo et al.. 1994; Tackeuchi et al., 1995; Mirin et al.,
1994); (2) self-organization of dot-like disk structures on (311) B GaAs
substrates (Notzel et al., 1994); and (3) self-formation of InGaAs dots on
GaAs by alternative supply of InAs and GaAs in atomic layer epitaxy (ALE)
(Mukai et al., 1994 and 1996a). Furthermore, the self-organization of
quantum dots has been successful in many material systems, such as InAs
on InP (Fafard et al., 1996; Nishi et al., 1998); InP on InGaP (Moritz et al.,
1996); and GaSb (Hatami et al., 1995) or InSb, GaSb, and AlSb on GaAs
(Glaser et al., 1996). Attempts to create a new quantum-dot structure,
including development of new material systems and new growth mechan-
isms, are ongoing, bringing with them a deeper understanding of the growth
6 SELF-ASSEMBLED
QUANTUM
DOTLASERS 243
1. GAINCHARACTERISTICS
The basic theory of the gain characteristics of quantum dots was derived
by Asada et al. (1986). This density-matrix theory gives the linear gain of
quantum dots without inhomogeneous broadening as follows:
1o4
ox looxlooxloo A3
-
6
v
.-
C
103
2x102
1.2 1.3 1.4 1.5 1.6 1.7
Wavelength, h (pm)
FIG. 6.2. Calculated gain spectra at 300K for a quantum dot (box), quantum wire,
quantum well (film), and bulk. (Reprinted with permission from Asada et al., 1986. $'I 1986
IEEE.)
2. THRESHOLD
CURRENT
The large optical gain and differential gain result in a low threshold
current operation. Assuming that the current is injected only through the
top surfaces of cubic quantum dots, as described by a schematic image
shown in Fig. 6.3, the relation between the injection current and the carrier
density is expressed by a simple formula as follows:
J = td,eN/z, (6.4)
where J is the averaged current density defined by dividing the total current
by the whole area of the active region ( J = I/WL in Fig. 6.3); e is the
electron charge; N is the carrier density; z, is the carrier lifetime; and 5 is
the filling factor of quantum dots in the active region. Here, the active region
is assumed to consist of a single dot layer. Referring to Fig. 6.3, 5 is defined
as 5 = d,d,/WL, where W is width of active region and L is the cavity
length. The effect of interaction among dots and finite potential depth is
ignored. Generally, the carrier lifetime z, consists of radiative and non-
radiative components of z, and znr, respectively, as
(6.5)
Current, I
zr =N /[
c [: I M I ~ E ~ Lmw)
3/2Eolj2n , ~ n h - 4
imn
-U 4 (6.6)
Using Eqs. (6.1) through (6.6), the optical gain can be related to the
current density. Figure 6.4 shows calculated peak gains as a function of
current density for four types of lasers with active regions of quantum dots,
quantum wires, quantum wells and bulk (Asada et al., 1986). The structures
assumed here are the same as in Fig. 6.2, which shows the increase of peak
gain with the increased dimension of quantum confinement. The threshold
gain, g t h , is denoted as follows by considering the optical confinement
factors in the given structures:
FIG. 6.4. Calculated peak gain at 300 K for a quantum dot (box), quantum wire, quantum
well (film), and bulk. The dashed line on each curve indicates the threshold gain. The
corresponding threshold current density is also shown. (Reprinted with permission from Asada
et al., 1986. 1Ci 1986 IEEE.)
248 HAJIME
SHOJI
where u,, and a,, are the absorption losses in the active region and other
regions, respectively; tx,F,, and 5, are the optical confinement factors in the
x,y , and z directions, respectively; and R is the facet reflectivity. 5, and 5,
can be approximately defined as the filling factors in the x and z directions,
respectively, because the optical field in the cavity is much larger than that
in the quantum dot. t, is determined by the waveguide structure. For
example, assuming that tx= 5, = 0.5 for a quantum-dot structure, 4, = 1
and t, = 0.5 for a quantum-wire structure, 5, = tz= 1 for quantum-well
and bulk structures, and assuming reasonable values for other parameters
(Asada et al., 1986), the threshold gains can be calculated by Eq. (6.7). Then,
if the results are put into Fig. 6.4, the threshold current densities for the
given structures are derived. As noted in Fig. 6.4, lower threshold current
density should be achieved in quantum-dot lasers-about one tenth that of
quantum well lasers. Of course. this value strongly depends on the par-
ameters assumed, and they should be much improved by optimizing the
structure. The low threshold current in the quantum-dot laser also brings
about a lesser temperature dependence of the threshold current due to
strong confinement of the electrons and holes.
Regarding threshold gain, there is another attractive feature of the
quantum-dot laser. As described above, the internal loss expressed by the
first and second terms of Eq. (6.7) becomes smaller in the quantum-dot
laser, therefore, larger differential quantum efficiency, given by
r
In-
1
Rf
1
is expected, where qd is the differential quantum efficiency, qi is the internal
quantum efficiency, and R, and R , are the reflectivities of the front and rear
facets, respectively.
3. DYNAMIC
CHARACTERISTICS
f ' =L/ii$g
2a (6.9)
6 SELF-ASSEMBLED
QUANTUM
DOTLASERS 249
where neff is the effective refractive index of the waveguide, c is the speed of
light in the vacuum, Po is the photon density in the cavity, and zp is the
photon lifetime. It is quite clear that greater differential gain results in a
larger relaxation oscillation frequency, which potentially enhances the
modulation bandwidth of the quantum-dot laser.
Narrower spectral linewidth can be also expected in the quantum-dot
laser. The spectral linewidth, Av, is expressed by
(6.10)
(6.11)
where M , R,, vg, hw, r, nsp, n, and Po are, respectively, the linewidth
enhancement factor, the mirror loss, the group velocity of light, the photon
energy, the optical confinement factor, the spontaneous emission factor, the
effective refractive index, and the output power. A large differential gain
results in a linewidth enhancement factor of nearly zero, leading to smaller
spectral linewidth. These parameters are strongly related to the direct
modulation characteristics, and chirping under high-speed modulation
should be suppressed because of the small a, leading to long transmission
distance in fiber communication applications.
- -
application, given their small diameter of 20 nm, high density of > 1 x
10'' cm ',typical size distribution of lo%, high emission efficiency, and
~
1. FABRICATION
FIG.6.5. (a) Plan-view and (b) cross-sectional TEM images of a triple-stacked quantum-
dot structure. The grown quantum dots were 20 nm in diameter and 5 nm in height. The areal
coverage was about 10%. Each dot layer was separated by 20-nm-thick GaAs barriers. (From
c,
Shoji et al., 1997a. 1997 IEEE.)
Energy (eV)
lo-'
ground
Wetting 2nd
1 o'6
800 900 1000 1100 1200 1300
Wavelength (nm)
FIG. 6.6. Photoluminescence spectrum at 300 K for a triple-stacked quantum-dot structure.
The emission peak at around 1.13pm is from the ground state, and the full width at half
maximum of the peak is about 80meV. (From Shoji et al., 1997a. 0 1997 IEEE.)
Energy El
El E2
E3
Electro-
luminescence
kE Ei
Ei Ez E3
E
FIG. 6.7. Effect of dot size on the excited states of electrons and holes.
states are also generally observed together with those from the ground state
because of nonorthogonality due to anisotropy of dot shape, the emission
from the ground state always becomes dominant even with increasing
excitation. On the other hand, in larger dots, there are several excited states
of electrons in the conduction band, which leads to the transitions from the
excited states of electrons and holes. The result in Fig. 6.6 corresponds to
the latter case.
Device Fabrication
Refractive
t index
FIG. 6.8. Schematic of a typical In(Ga)As-GaAs quantum-dot laser, showing profiles of the
conduction-band energy and the corresponding refractive index. The dashed curve shows the
optical field in the lateral direction.
FIG. 6.9. Schematic of a fabricated ridge-wave guide quantum-dot laser. The laser contains
a triple-stacked quantum-dot structure. The width of the ridge-waveguide is about 2.5 pm.
6 SELF-ASSEMBLED
QUANTUMDOTLASERS 255
(Shoji et al., 1997a). The laser structure was grown on a (001) n-GaAs
substrate followed by a 0.9-pm-thick n-Al,,,Ga,.,As cladding layer, a
quantum-dot active layer, a 1.2-pm-thick p-Al,,,Ga,,,As classing layer, and
a 0.3-pm-thick p-GaAs contact layer. As the active layer, a triple-stacked
quantum-dot structure was employed to obtain larger optical confinement
to the active layer. In this case, the optical confinement factor to the
active-layer plane was about 6% (actual optical confinement to quantum
dots is much less than this value because of low areal coverage of quantum
dots in the active layer). A narrow ridge-waveguide structure is usually
formed by chemical or dry etching of the p-GaAs contact layer and a part
of the p-AlGaAs cladding layer. In Fig. 6.9, the width of the ridge-
waveguide was about 2.5 pm, which enabled stable control of the lateral
mode. After growing the entire structure, several laser chips with different
device parameters were prepared for measurement. Typically, cavity length
and facet reflectivity varied case by case. In addition, the number of dot
layers often varied wafer to wafer.
2. DEVICE
CHARACTERISTICS
Electroluminescence
L = 300 ;m
as-cleaved
3rd 2nd
50?* I 1
Lasing Churucteristics
300 K, pulsed
I
I I
300 600 900 1200
Cavity length (pm)
FIG. 6.11. Cavity-length dependence of threshold current density for quantum-dot lasers
with different numbers of dot layers and different facet reflectivities, Niayeris the number of
quantum-dot layers in the active layer. The threshold current densities are normalized for one
dot layer for comparison. (From Shoji et al., 1997a.)
the threshold current densities for a dot layer were smaller than 1 kA/cm2.
In these lasers, room-temperature CW operation was achieved (Shoji et al.,
1996b).
Figure 6.12 shows light output versus current characteristics for a
CW-lasing quantum-dot laser. In this case, the threshold current was about
140 mA at 20"C, which corresponded to the threshold current density of
3.1 kA/cm2. Much lower threshold current density, below 400 A/cm2 at
room temperature, has also been reported (Kamath et al., 1997b; Hein-
richsdorff et al., 1997). The difference in the threshold current density is
mainly due to the difference in the emission efficiency of quantum dots.
Reduction of nonradiative recombination as a result of improved crystal
quality leads to the improved emission efficiency, especially at room tem-
perature, which results in lower threshold current density.
Various parameters are derived from the mirror loss dependence shown
in Fig. 6.11. The internal loss, m i , the differential quantum efficiency, yld, and
the internal quantum efficiency, q i are important factors in laser quality.
Although these parameters strongly depend on the waveguide structure and
the quality of the active region, those values can be derived from a
relationship expressed by Eq. (6.8). Experimentally, the internal loss, ( x i , was
estimated to be 3-5 cm-' (Shoji et al., 1997a). Of course, this value might
depend on the density of quantum dots in the active layer. The internal
258 HAJIME
SHOJI
1.o
N,aye, = 3,L = 900 pm
HR-coated
0.8 20°C
z
v
cw
c
0.6
3
Q
c
3
0.4
E0)
7
0.2
a
100 200 300 400
Current (mA)
FIG. 6.12. Light output versus current characteristics for a quantum-dot laser under the
CW condition. Triple-stacked quantum dots are in the active layer (PIloyer = 3). and the
cavity length is 900pm ( L = 900pm). The threshold current is about 140mA at 20’C, which
corresponds to the threshold current density of 3.1 kA/cm2.
WHR) 1
50
20
10
5
100
20
/aver = 1
-300 vm
state. A key point here is that the lasing wavelength did not change
continuously from the higher-order subbands to the lower-order subbands,
but changed discontinuously, reflecting the existence of discrete quantum-
confined subbands. A relationship between the lasing wavelength and the
threshold gain for one dot layer, as shown in Fig. 6.14, is evidence that the
lasing wavelength jumps from the second state to the ground state with a
lowering of the threshold gain for one dot layer. The critical threshold gain
260 HAJIME
SHOJI
. ' - I 1 .
0 2 4 6 8 1 0
Threshold gain for one dot layer (cm-I)
FIG 6 14. Relationship between the lasing wavelength and the threshold gain for one dot
layer. The lasing wavelength jumps from the second state to the ground state when the
threshold gain for one dot layer is lowered by the use of a long cavity structure and a multiply
stacked dot layer (From Shoji et al. 1997a ( 1997 IEEE)
for one dot layer was found to be 3.5 cm - in this case, as marked by a
dashed line (Shoji et al., 1997a).
A relationship between modal gain and current density can be also
derived from the cavity-loss dependence of threshold current density, as
shown in Fig. 6.15 (Shoji et al., 1997a), where both values were normalized
for one dot layer. It was found that the differential gain for the current
density depended on the subband at which lasing occurred. The different
differential gains can be attributed to the variation in degeneracy between
the ground state and the second state-the larger degeneracy at the higher
state leads to the larger differential gain. This means that lasing at the
ground state reduces the threshold gain for one dot layer by multiply
stacking the dot layer, increases the optical gain for one dot layer by
increasing the in-plane dot density, or reduces the cavity loss by the use of
a long cavity structure and HR coating. At around the cross-point of the
two gain curves, which coincides with the critical threshold gain of 3.5 cm-
in Fig. 6.14, lasing occurred at the two subbands at the same time because
of comparable modal gains at the two subbands at a given current density.
-
Measurement of material gain was also carried out (Kirstaedter et al., 1996).
A large gain of 7 x lo4cm- was reported at a current injection level
of only 80 A/cm2.
Another important feature observed in actual quantum-dot lasers is gain
saturation (Schmidt et al., 1996). For example, as shown in Fig. 6.13,
6 SELF-ASSEMBLED DOTLASERS
QUANTUM 261
0.1 1 10
Current density for one dot layer (kA/crn2)
FIG. 6.15. Relationship between the modal gain and the current density derived from the
cavity-loss dependence of the threshold current density, where both values were normalized for
one dot layer. It was found that the differential gain for the current density depended on the
subband at which lasing occurred. (From Shoji et al., 1997a. c? 1997 IEEE.)
2nd j Ground
state \ state
20 C
Pulsed
1 A,:
' A
ncn
1000 1050
..-.?length (nm)
Current (A)
FIG. 6.16. Large kink observed in their light-current characteristics of a quantum-dot laser.
Because of gain saturation at the ground state, the lasing spectrum broadened with an increase
in the injection current, and lasing at the second state began. At that point a large kink in the
light-current curve was observed.
began to lase with increased current injection; lasing finally began at the
second state due to the gain saturation at the ground state. For that reason,
a large kink was observed in the light-current curve at which lasing at the
second state started.
Temperature Depetzdence
Type B, as-cleaved
1000 . ' ' .' .. .'. .. '. . ..' ...
' ' '
Figure 6.18 shows the EL and lasing spectra of the type-A laser in Fig.
6.17 for three temperatures: 60, 200, and 300 K (Shoji et a]., 1997a and b).
Distinct differences were observed in the spectra. At 300 K, the threshold
current was large and the EL spectra below the threshold were asymmetri-
cally shaped. This indicates that the injected carriers distribute to the second
state because of high threshold carrier density, which is caused by poor
emission efficiency of quantum dots and large thermal excitation of carriers
from the ground to the second state at room temperature. Consequently, the
EL peak, which coincided with the PL peak at a low current-injection level,
exhibited a large shift to the shorter wavelength side with an increase in
current. On the other hand, with a lowering of temperature, the FWHMs of
300 K
150.
100.
.- 7'
t
h
.mm rn a - PL .-cn
z-
E
-I
C
3
-
v R
\
0
41K
Type-8 (as-cleaved) A
x
.-c
cn
C
a,
c
currents for type-A and type-B lasers, as already shown in Fig. 6.17 (Shoji
et al., 1997a and b). Temperature dependence of P L intensity is also shown
for comparison. The type-A laser exhibited a high characteristic tempera-
ture, To, defined by
);(
I , , x exp (6.12)
of over 300K in the temperature range below 200K, and the threshold
current density was less than 600 A/cm2. In particular, the characteristic
temperature was as high as 477 K in the temperature range of 60- 110 K,
and the threshold current density was less than 450 A/cm2. However, in the
temperature region over 200 K, the characteristic temperature rapidly de-
creased to a value below 100 K . It was as small as 48 K at room tempera-
ture.
As for the type-B laser, although it also exhibited a high characteristic
temperature of over 400K in a temperature region below 100K, the
characteristic temperature started to decrease at 170 K, which was a lower
temperature compared with that in the type-A laser. The reason for this
unusual behavior of the threshold current at temperatures of 120-130 K ha5
not been determined. Another unusual behavior in the temperature range of
220-260 K -the wavelength shift between the ground state and the second
state observed at around 260K, shown in Fig. 6.17-might be attributed
to this phenomenon. As for the value of the characteristic temperature, a
much higher value of 530 K over a temperature range of 80-220 K has been
reported by Bimberg et al. (1997).
It can be assumed that the temperature dependence of the threshold
current is strongly related to that of the P L intensity; that is, the decrease
in the PL intensity with increasing temperature is due to the increased
nonradiative recombination in the quantum dots, which results in the
proportional increase in the threshold current. However, the results shown
in Fig. 6.19 do not necessarily support this assumption. The PL intensity
begins to decrease at 150 K, whie the rapid increase in the threshold current
starts at 170 K in the as-cleaved laser and at 200 K in the HR-coated laser.
Although it is qualitatively correct that a decrease in the P L intensity leads
to an increase in the threshold current, the relationship between the two
values is not so simple. Further investigation should be done on temperature
characteristics, because the energy separation between the subbands is also
related to them; nevertheless, the energy separation between the ground
state and the second state is much larger than kT of 26 meV, even at room
temperature. On the other hand, the comparison between the HR-coated
laser and the as-cleaved laser shows that a lower threshold current leads to
better temperature characteristics in a higher temperature region. Reduction
6 DOTLASERS
QUANTUM
SELF-ASSEMBLED 267
Modulation Characteristics
might limit the bandwidth. Although both factors should contribute to the
bandwidth, it is not clear at present which is the dominant factor. Further
investigation of carrier dynamics in quantum dots is needed.
3. LIMITING
FACTORS
OF LASERPERFORMANCE
1. CLOSELY QUANTIJM-DOT
STACKED LASERS
1
.e I- ' I I I t
8
8
'I
' I
in size averaging in the lateral direction. The other reason is the larger
volume of quantum dots. Because the effective vertical height becomes
greater than that of single-layer dots, the emission wavelength becomes less
sensitive to fluctuations in dot size. The red shift of the PL peak in Fig. 6.21
is further evidence of the larger volume of quantum dots due to the
electronic coupling of stacked dot layers.
The second attractive feature is enhancement of modal gain from a single
dot layer. In the closely stacked structure, optical confinement to the active
layer is enhanced by the number of stacked layers. Because the closely
stacked dots act as a single dots, the modal gain obtained from a single dot
layer becomes larger. Larger optical gain is expected in the closely stacked
structure at an injection level that compares with that of the structure
without vertical coupling. Effective dot height can be precisely controlled by
changing the number of stacking cycles.
Laser oscillation has been achieved by use of the closely stacked quan-
tum-dot structure. Shoji et al. (1997~)reported pulsed lasing at room
temperature and CW lasing at 50 K. The threshold current density at 50 K
was 520 A/cm2. As shown in Fig. 6.22, lasing occurred at the second
subband at 50 K. Bimberg et al. (1996) demonstrated a laser oscillation
using a different closely stacked structure, where the thickness of GaAs
barriers separating InAs dot layers was 6nm. Even for this rather thick
barrier, a decrease in the FWHM of the quantum-dot emission down to
L = 900 pm
as-cleaved
>
.-
c
u)
c
0)
-
c
C
2. COLUMNAR
QUANTUM-DOT
LASERS
I I
10 nm
FIG. 6.23. Cross-sectional T E M image of a columnar quantum dot.
F
E
v
0 20 40 60 80 100 120
Current (mA)
FIG. 6.24. Light output versus current characteristics of a columnar quantum-dot laser.
showing the result of a 900-pm-long device with as-cleaved facets for various temperatures up
to 70°C. The threshold current is 31 mA at 25"C, which corresponds to the threshold current
density of 500 A/cm2. (Reprinted with permission from Mukai et al., 1998. :p 1998 IEEE.)
- 20 meV
I
-
300K
Columnar dots
H I
ui
.-c
c
3
$ -
v
Wetting layer
a Wetting layer
1997a). The emission from the wetting layer at a longer wavelength was due
to electronic coupling of thin wetting layers associated with extremely thin
barrier layers separating the dot layers, which resulted in the formation of
an effectively thick wetting layer, as in a quantum well. As discussed in
Chapter 5, the lasing characteristics of quantum-dot lasers are subject to the
carrier relaxation rate from the continuous level to the quantum confined
states (Sugawara et al., 1997b). If the energy difference between the continu-
ous state in the wetting layer and the discrete levels of quantum dots is
smaller, the carrier relaxation rate should be enhanced by the assistance of
the multiphonon effect (Inoshita and Sakaki, 1992) because the number of
required multi-phonons is smaller for the shallow potential.
3. LONG-WAVELENGTH
QUANTUM-DOT LASERS
Quantum dots that emit in the long wavelength region such as 1.3 pm and
1.55pm are of great practical interest. They can be used in a variety of
practical devices, such as optical communication and optical interconnects.
Emission in the 1.3-pm region has been reported by several groups in
InAs/GaAs or in InGaAs/GaAs quantum-dot structures on GaAs substrates
(Mukai et al., 1994; Tackeuchi et al., 1995; Mirin et al., 1995). On InP
substrates, emission in the 1.4- 1.7-pm spectral region has been demon-
strated in InAs/InAlAs or InAsPnGaAsP quantum-dot structures (Fafard et
al., 1996; Nishi et al., 1998).
As described in Chapter 3, the ALS quantum-dot structure, which is
self-organized by alternately supplying monolayers of InAs and GaAs in the
atomic layer epitaxy (ALE) mode, holds promise for realizing long
wavelength emission in quantum dots. ALS dots are In,,,Ga,,,As quantum
dots typically with a diameter of 20 nm and a height of 10 nm, and they are
surrounded by a 10-nm-thick Ino,,Ga,,,As barrier in the lateral direction
and by a GaAs barrier in the growth direction (Mukai et al., 1994). The
emission wavelength of the ALS dots was about 1.3 pm at room tempera-
ture, as shown in Fig. 6.26, and the wavelength can be controlled in a range
of 1.1-1.5pm by changing the growth conditions, such as the number of
cycles for alternate supply of InAs and GaAs monolayers and the composi-
tion of the InGaAs buffer layer on which quantum dots are grown (Ohtsuka
et al., 1995). The FWHM of the ground-state PL peak was as narrow as
35 meV. These features are quite attractive for laser application.
Using the ALS dots, quantum-dot lasers were fabricated by Shoji et al.
(1995). One is schematically shown in Fig. 6.27. In the active layer, InGaAs
quantum dots were self-organized by performing 12 cycles of (lnAs),/
(GaAs), short-period growth. Figure 6.28 shows the emission spectra for
6 SELF-ASSEMBLED DOTLASERS
QUANTUM 271
Wavelength (vm)
FIG. 6.26. PL spectrum of ALS quantum dots at 300K. The emission peak is at 1.3 jtm,
and the FWHM is as small as 35meV. (From Shoji et al., 1995. Ci 1995 JEEE.)
Ino,Gao ,As
p+-GaAs contact (0.5 pm) Quantum dot
- - - - - - - - -'\
*n-In, 47Ga053Pclad (1 pm)
n-GaAs buffer (0.5 pm) \
*I1: 1
In, ,Ga, ,As barrier
n-GaAs sub.
FIG. 6.27. Schematic of a n ALS quantum-dot laser. The active region consists of single-
layer ALS dots formed by twelve cycles of (InAs),/(GaAs), short-period growth. (From Shoji
et a]., 1995. (3 199.5 IEEE).
278 SHOJI
HAJIME
1.1 A (lasing) 80 K
-
.-v)
4-
40
C
3
.@ 30
Y
m
.220
v)
5
W
4-
C
- 10
0
800 1000 1200 1400
Wavelength (nm)
FIG. 6.28. Emission spectra for various injection currents at 80 K. Although the peaks are
not as clear as in the PL spectrum shown in Fig. 6.26, discrete quantum levels in the EL spectra
are found under current injection. Due to band filling with current injection, several high-order
levels appear in the shorter wavelength region because of a higher excitation level compared
with the PL measurement. Laser oscillation is achieved at the threshold current of 1.1 A under
the pulsed condition. (From Shoji et al., 1995. 8 1995 IEEE.)
pulsed condition, and the corresponding current density was 815 A/cm2. The
lasing wavelength was 91 1 nm.
Considering the fact that the compositional wavelength of the 10-nm-
thick In,,,Ga,,,As barrier layer at 80 K was shorter than 850 nm, the laser
oscillation was from a high-order sublevel of the quantum dots. Note that
the wetting layer, which is formed in the growth of quantum dots in the SK
mode is not observed in the ALS quantum-dot structure. To confirm that
the obtained laser oscillation was really from a high-order sublevel of
quantum dots, the diamagnetic shifts of the lasing peak were also carried
out. A strong magnetic field up to 13T was applied perpendicular to the
lasers, and the experimental results showed that the energy shift of the lasing
peak was much smaller than that of the quantum well, whose compositional
wavelength was close to that of the InGaAs barrier layer surrounding the
quantum dots. This indicates that the obtained laser oscillation was from
the quantum-confined level of the quantum dots. In this experiment,
although lasing was observed up to 140 K, room-temperature operation was
not achieved because the dot density was found to be quite low (a few
percent areal coverage) and only a single-sheet dot layer was grown in the
active layer. In addition, because of the limitation in growth equipment, the
6 SELF-ASSEMBLED DOTLASERS
QUANTUM 279
active layer was exposed to the air once before growing the p-type cladding
layer, so that nonradiative centers might have been introduced. Considering
the emission intensity at room temperature, as increase in the dot density
and improved growth would lead to 1.3-pm operation and a lower threshold
current at room temperature.
Mirin et al. (1995) reported a sharp emission at 1.3 pm from In,.,Ga,,,As
quantum dots grown in the SK mode. Although these dots were larger than
the ALS quantum dots, the FWHM of the emission was as narrow as
28meV. These researchers also succeeded in laser oscillation at room
temperature (Mirin et al., 1996). Because the lasing occurred at an excited
state, the lasing wavelength was 1.2 pm. However, this is so far the longest
lasing wavelength ever reported in self-assembled quantum-dot lasers grown
on GaAs substrates.
Nishi et al. (1998) reported a possibility of long wavelength quantum-dot
lasers on InP substrates. InAs quantum dots were grown on (311)B InP
substrates in the SK mode. Although lasing was still at low temperatures,
the lasing wavelength of 1.2- 1.4 pm and the threshold current of 540 A/cmZ
were demonstrated at 77 K. If room temperature lasing can be achieved, a
much longer lasing wavelength is expected in this structure. Note that in the
case of InAs dots on InP substrates the dot diameter is typically 30 nm or
more (Fafard et al., 1996; Nishi et al., 1998), which is larger than that of
InAs dots on GaAs substrates at present. This may originate from the small
lattice mismatch between the lnAs and the substrate. Extracting the atom-
like nature from quantum dots requires a smaller size.
4. QUANTUM-DOT
VERTICAL-CAVITY
SURFACE-EMITTING
LASERS
E
Mirror
Quantum dots
Mirror
Optical field
Photon mode
t -c
(n-1)th
nth
FIG. 6.29. Photon modes and electronic states in a quantum-dot microcavity, where both
the photon mode and the electronic state become discrete. Although lasing occurs only when
the photon mode and the electronic state match, an extremely low threshold current is
expected.
FIG. 6.30. Quantum-dot VCSEL with 10-period InGaAs dots in the active region. (Re-
printed with permission from Saito et al., 1996. 6 1996, American Institute of Physics.)
1.6
Pulse
1.4
1.2
2
E 1.0
v
4-.
I 2 0.8
2
0
E 0.6
rn
._
1
0.4
0.2
0
0 20 40 60 80 100 120
Current (mA)
FIG. 6.31. Light output versus current for two polarization states of a quantum-dot
VCSEL at room temperature. The orthogonal polarization suppression ratio is 18 dB at 1-mW
output power of a [OTl] emission. (Reprinted with permission from Saito et al., 1997. 0 1997
American Institute of Physics.)
V. Conclusion
Acknowledgments
REFERENCES
Ahopelto, J., Lipsanen, H., and Sopangen. M. (1995). Proc. 7th Con/: Indium Phosphide rind
Relutrd Materiuls ( I P R M '9s) FB2-4, 19-20.
Arakawa, Y., and Sakaki, H. (1982). Appl. Phys. Lett. 40, 939-941.
Arakawa, Y., Sakaki, H., Nishioka, M., Okamoto, H., and Miura, N. (1983). Jpn. J . A p p l . Phy.~.
22, L804-L806.
Arakawa, Y., and Yariv, A. (1986). IEEE J. Qimritton Electron. QE-22, 1887-1899.
Asada, M., Miyamoto, Y., and Suematsu, Y. (1986). ZEEE J. Quantuni Electron. QE-22,
1915-1921.
Baba, T., Hamano, T., Koyama, F., and Iga. K. (1991). IEEE J. Qumturn Electrori. QE-27.
1347 1 354.
-
Benisty, H.. Sotomayor-Torres, C. M., and Weisbuch, C. (1991). Phys. ReG. B.44, 10945- 10948.
Bimberg, D., Ledentsov, N. N., Grundniann, M., Kirstaedter, N., Schmidt, 0. G., Mao. M. H.,
Ustinov, V. M., Egorov, A. Y., Zhukov, A. E., Kop'ev, P. S., Alferov, 2. I., Ruvimov, S. S.,
Gosele, U., and Heydenreich, J. (1996). J p n . J . Appl. Phys. 35, 1311-1319.
Bimberg, D., Kirstaedter, N., Ledentsov. N. N., Alferov, Z. I., Kop'ev, P. S., and Ustinov, V. M.
( I 997). IEEE J. Selected Topics in Qurintuni Electron. 3, 196-205.
Bockelmann, U., and Bastard, G. (1990). Phys. Reu. B. 42, 8947-8953.
Dallesasse, J. M., Holonyak, N.. Sugg. A. R., Richard, T. A., and El-Zein, N. (1990). Appl. Phys.
Lett. 57. 2844-2846.
284 HAJIMESHOJI
Fafard, S., Wasilewski, Z., McCaffrey, J., Raymond, S., and Charbonneau, S. (1996). Appl. Phys.
Lett. 68, 991-993.
Fukui, T., Ando, S., Tokura, Y., and Toriyama, T. (1991). Appl. Phys. Lett. 58, 2018-2020.
Glaser, E. R., Benett, B. R., Shanabrook, B. V., and Magno, R. (1996). Appl. Phys. Lett. 68,
36 14-3616.
Grundmann, M., Christen, J., Ledentsov, N. N., Bohrer, J., Bimberg, D., Ruvimov, S. S.,
Werner, P., Richter, U., Gosele, U., Heydenreich, J., Ustinov, V. M., Egorov, A. Y.,
Zhukov, A. E., Ustinov, V. M., and Alferov, Z. I. (1995). Phys. Reu. Lett. 74, 4043-4046.
Grundmann, M., Ledentsov, N. N., Stier, 0..Bimberg, D., Ustinov, V. M., and Alferov, Z. I.
(1996). Appl. Phys. Lett. 68, 979-981.
Hatami, F., Ledentsov, N. N. Grundmann, M., Bohrer, J., Heinrichsdorff, F., Beer, M., Bimberg,
D., Ruvimov, S. S., Werner, P., Gosele. U., Heydenreich, J., Ivanov, S. V., Meltser, B. Y.,
Kop’ev, P. S., and Alferov, Z. I. (1995). Appl. Phys. Lett. 67, 656-658.
Hayashi, Y., Mukaihara, T., Hattori, N., Ohnoki, N., Matsutani, A,, Koyama, F., and Iga, K.
(1995). Electron. Lett. 31, 560-561.
Heinrichsdorf, F., Krost, A., Grundmann, M., Bimberg, D., Kosogov, A. O., and Werner, P.
(1996). Appl. Phys. Lett. 68, 3284-3286.
Heinrichsdorff,F., Mao, M. H., Kirstaedter, N., Krost, A,, Bimberg, D., Kosogov, A. O., and
Werner, P. (1997). Appl. Phys. Letr. 71, 22-24.
Heitz, R., Grundmann, M., Ledentsov, N. N.. Eckey, L., Veit, M., Bimberg, D., Ustinov, V. M.,
Egorov, A. Y., Zhukov, A. E.. Kop’ev, P. S., and Alferov, Z. I. (1996). Appl. Phys. Lett. 68,
361-363.
Hiraydma, H., Matsunaga, K., Asada, M., and Suematsu, Y. (1994). Electron. Lett. 30, 142-143.
Huffaker, D. L., Shin, J., and Depper, D. G. (1994). Electron. Lett. 30, 1946-1947.
Huffaker, D. L., Baklenov, O., Graham, L. A,, Streetman, B. G., and Depper. D. G. (1997a).
Appl. Phys. Lett. 70, 2356-2358.
Huffaker, D. L., and Deppe, D. G. (1997b). .4ppl. Phys. Lett. 71, 1449-1451.
Inoshita, T., and Sakaki, H. (1992). Phq’s. Ret:. B. 46, 7260-7263.
Kamath, K., Battacharya, P., Sosnowski, T., and Phillips, J. (1996). Electron. Lett. 30,
1374-1375.
Kamath, K., Phillips, J., Jiang, H., Singh, J.. and Bhattacharya, P. (1997a). Appl. Phys. Lett. 70,
2952-2953.
Kamath, K., Klotzkin, D., and Bhattacharya, P. (1997b). Proc. 10th Ann. Meer. LEOS (LEOS
’97), ThY4.
Kash, K., Scherer, A,, Worlock, J. M., Craighead, H. G., and Tarnargo, M. C. (1986). Appl.
Phys. Lett. 49, 1043-1045.
Kirstaedter, N., Ledentsov, N. N., Grundmann, M., Bimberg. D., Ustinov, V. M., Ruvimov,
S. S., Maximov, M. V., Kop’ev, P. S., Alferov, Z. I., Richter, U., Werner, P., Gosele, U., and
Heydenreich, J. (1994). Electron. Lett. 30, 1416-1417.
Kirstaedter, N., Schmid, 0. G., Ledentsov, N. N., Bimberg, D., Ustinov, V. M., Egorov, A. Y.,
Zhukov, A. E. Maximov, M. V., Kop’ev. P. S., Alferov, Z. I. (1996). Appl. Phys. Lett. 69,
1226-1228.
Ledentsov, N. N., Bohrer, H., Bimberg, D., Kochnev, I. V., Maximov, M. V., Kop’ev, P. S.,
Alferov, Z. I., Kosogov. A. O., Ruvimov. S. S., Werner, P., and Gosele, U. (1996). Appl.
Phys. Lett. 69, 1095-1097.
Leonard, D., Kishnamurthy, M., Reaves. C. M.. Denbars, S. P., and Petroff, P. M. (1993). Appl.
Phys. Lett. 63, 3203-3205.
Mao, M. H., Heinrichsdorff, F., Kirstaedter, N., Krost, A., Ledentsov, N. N., Bimberg, D.,
Ustinov, V. M., Egorov, A. Y., Zhukov, A. E., Kop’ev, P. S., and Alfreov, Z. I. (1997). Proc.
10th Ann. Meet. LEOS (LEOS ’97).ThY5.
6 SELF-ASSEMBLED
QUANTUMDOTLASERS 285
Mirin, R. P., Ibbetson, J. P., Nishi, K., Gossard, A. C., and Bowers, J. E. (1995). Appl. Phps.
Lett. 67, 3795-3191.
Mirin, R. P., Gossard, A,, and Bowers, J. E. (1996). Electron. Lett. 32, 1732-1733.
Moritz, A., Wirth, R., Hangleiter, A., Kurtenbach, A,, and Ebert, K. (1996). Appl. Phys. L e t f .
69, 212-214.
Mukai, K., Ohtsuka, N., Sugawara. M., and Yamazaki, S. (1994). Jpn. J . Appl. Phys. 33,
L1710-Ll712.
Mukai, K., Ohtsuka, N., and Sugawara, M. (1996a). Jpn. J . Appl. Phys. 35, L262-L265.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1996b). Appl. Phys. Lett. 68, 3013-3015.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1996~).Phys. Rev. B. 54, R5243-R5246.
Mukai, K.. Shoji, H., and Sugawara. M. (1998a). Proc. lO!h In!. Conj: Indium Phosphide and
Related Muteriuls ( I P R M '98), pap. WBI-4.
Mukai, K. (1988b) (unpublished).
Nakata, Y., Sugiyama, Y., Futatsugi, T., Yokoyama, N. (1997). J . Cryst. Growth 175/176,
71 3-719.
Nishi, K., Yamada, M., Anan, T., Gomyo. A.. and Sugou, S. (1998). Ext. Ahs. Jupunese Sociery
of Applied Physics, 3 (in Japanese).
Notzel, R., Temmyo, J., Kamada, H., Furuta, T., and Tamamura, T. (1994). Appl. Phys. Lett.
65, 457-459.
Ohtsuka, N., and Mukai, K. (1995). Proc. 7th Int. Conj: Indium Phosphide und Related
Muterials ( I P R M '95), WP59.
Oshinowo, J., Nishioka, M., Ishida, S., and Arakawa, Y. (1994). Appl. Phys. Lett. 65, 1421-1423.
Pan, D., Zeng, Y. P., Wu, J., Wang, H. M., Chang, C. H., Li, J. M., and Kong, M. Y. (1997).
Appl. Phys. Lett. 70, 2440-2442.
Saito, H., Nishi, K., Ogura, I., Sugou, S.. and Sugimoto, Y. (1996). Appl. Phys. Lett. 69,
3 140-3 142.
Saito, H., Nishi, K., Sugou, S., and Sugimoto, Y. (1997). Appi. Phjs. Lett. 71, 590-592.
Schmidt, 0. G., Kirstaedter, N., Ledentsov, N. N., Mao, M. H., Bimberg, D., Ustinov. V. M..
Egorov, A. Y., Zhukov, A. E., Maximov, M. V., Kop'ev, P. S., and Alferov, Z. I. (1996).
Electron. Lett. 32, 1302-1 304.
Shoji, H., Mukai, K., Ohtsuka, N.. Sugawara, M., Uchida, T., and Ishikawa, H. (1995). IEEE
Photon. Technol. Lett. 7, 1385- 1387.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara. M., Yokoyama, N., and Ishikawa.
H . (1996a). Jpn. J. Appl. Phys. 35, L903-L905.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara, M., Yokoyama. N.. and Ishikawa,
H. (1996b). Electron. Lett. 32, 2023-2024.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara, M., Yokoyama, N., and Ishikawa,
H. (1997a). IEEE J. Selected Topics in Quantuni Electron. 3, 188-195.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama. Y., Sugawara, M., Yokoyama. N., and Ishikawa,
H. (1997b). Appl. Phys. Lett. 71, 193-195.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara, M., Yokoyama, N., and Ishikawa,
H. (1997~).Proc. 2nd ConJ Optoelutronics rind Communicution (OECC' '97). 9C2-5.
Shoji, H.. Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara, M., Yokoyama. N., and Ishikawa,
H. (1997d). Proc. 10th Ann. MCCI.L E O S ( L E O S '97), pap. TbY2.
Soda, H., Iga, K., Kitamura, C., and Suematsu, Y. (1979). Jprz. J . Appl. Phys. 18, 2329-2330.
Solomon, G. S., Trezza, J. A., Marshall, A. F., and Harris, J. S. (1996). Phys. Rer. Lett. 76,
952-955.
Solomon, G. S., Larson, M. C., and Harris, J. S. (1996). Appl. Phys. Lett. 69, 1897-1899.
Sugawara, M., Nakata, Y., Mukai, K., and Shoji, H. (1997a). Phys. Rev. B. 55, 13155-13160.
Sugawara, M., Mukai, K., and Shoji, H. (1997b). Appl. Phys. Lett. 71, 2791-2793.
286 HAJIMESHOJI
Sugiyama, Y., Nakata, Y., Imamura, K.. Muto, S., and Yokoyama, N. (1996). Jpn. J . Appl. Phys.
35, 1320-1324.
Sugiyama, Y., Nakata, Y., Futatsugi, T., Sugawara, M.. Awano, Y., and Yokoyama, N. (1997).
Jpn. J . A p p l . Phys. 36, L158-LI61.
Tackeuchi, A., Nakata, Y., Muto, S.. Sugiyama, Y., Inata, T., and Yokoyama, N. (1995). Jpn.
J . Appl. Phys. 34, L405-L407.
Temmyo, J. Kuramochi, E., Sugo, M., Nishiya, T., Kamada, H., Notzel, R., and Tamamura, T.
(1994). Proc. 14th Int. Semiconductor Luser Cot$ (ISLC '94). PD3.
Uskov, A. Y., McInerney, J., Adler, F., Schweizer. H., and Pilkuhn, M. H. (1998). Appl. Phys.
Leu. 72, 58-60.
Ustinov, V. M., Egorov, A. Y., Kovsh. A. R.. Zhukov, A. E., Maximov, M. V., Tsatsulnikov,
A. F., Gordeev, N. Y., Zaitsev, S. V.,Shernyakov, Y. M., Bert, N. A,, Kop'ev, P. S., Alferov,
Z. I., Ledentsov, N. N., Bohrer, J., Bimberg, D., Kosogov, A. O., Werner, P., and Gosele,
U. (1996). Proc. 9th Int. Cortf: MBE.
Xie, Q., Kalburger, A., Chen, A., and Madhukar, A. (1997). J . Cryst. Growth, 175/176,689-695.
SEMICONDUCTORS AND SEMIMETALS . VOL. 60
CHAPTER 7
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
11. PROPERTIES OF QUANTUM DOTS. . . . . . . . . . . . . . . . . . . . 288
1 . The Quantum Dot us a T M V - L ~ WS.ystern
I . . . . . . . . . . . . . . . 288
2 . Attractive Features of Quantum Dots.for Device Application . . . . . . . 294
111. QUANTUM DOTSFOR VERYHIGHSPEEDLIGHT MODULATION . . . . . . . . 295
1. The Need.for High-speed. Low- Wavelength-Chirp Light Sources . . . . . 295
2 . Direct Modulation of Quantuni-Dot Lasers . . . . . . . . . . . . . . 298
3 . The Quuntuin-Dot Intensity Modulator . . . . . . . . . . . . . . . . 302
IV . QUANTUM DOTSAS A NONLINEAR MEDIUM. . . . . . . . . . . . . . . . 303
1. The Needfor Large Nonlinearity with u Large Bandwidth . . . . . . . . 303
2. Ana!ysis of zl3’ . . . . . . . . . . . . . . . . . . . . . . . . . . 306
3 . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
V. PERSISTENT HOLE BURNINGMEMORY. . . . . . . . . . . . . . . . . . 314
1 . Persistent Spectrul Hole Burning Memory Using Quantum Dots . . . . . 314
2 . E.xperiinentu1 Results . . . . . . . . . . . . . . . . . . . . . . . . 316
3. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
VI . SUMMARY A N D PERSPECTIVES ON QUANTUM-DOT OPTICAL DEVICES . . . . . 319
1. Trends in Optoelectronics . . . . . . . . . . . . . . . . . . . . . . 320
2 . Usesfor Quantum Dot Opticul Devices . . . . . . . . . . . . . . . . 321
Acknowledyment . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
I . Introduction
The past several years have seen remarkable progress in the fabrication of
quantum dots and quantum-dot lasers. Thanks to new self-organized
growth technologies for example. as described in Chapter 6. ground-state
CW lasing can now be achieved at room temperature (Kirstaedter et al.,
1994; Shoji et al., 1996). Although the performance of quantum-dot lasers is
as yet far from ideal. recent gains have enlarged our expectation of applying
287
Copyright ( 1999 by Academic Press
All rights of reproduction in any form recerved
ISBN 0-1?-752169-0
ISSN 0080-8784 99 $3000
288 HIROSHI
ISHIKAWA
quantum dots to various optical devices. One of the reasons that we are so
interested in quantum dots is that the performance and functions of optical
devices based on current technologies, such as the quantum well or the
strained quantum well, have reached their limit. In this respect, quantum
dots are highly attractive because of their unique three-dimensional carrier
confinement which may provide us new optical properties. At present,
application research is concentrating mainly on quantum-dot laser, and
even here there is much to be done. We do not know what is ultimately
achievable with quantum-dot lasers because our understanding of the
carrier dynamics in quantum dots is incomplete. Nevertheless, we can start
experimental and theoretical studies to explore the advantages of quantum
dots for device applications.
In this chapter, I first review some of the attractive properties of quantum
dots and then discuss several feasible applications. In Section 11, I look at
the quantum dot as a two-level system, and summarize some of its
advantages. In Section 111, I discuss the use of quantum dots for high-speed
modulation of light for long-distance high-bit-rate communication systems.
I review the problems with conventional technologies and then discuss the
direct modulation of quantum-dot lasers and the feasibility of an external
modulator using quantum dots. In Section IV, I discuss the use of quantum
dots as a nonlinear medium. Realization of large nonlinear susceptibility
using quantum structures has long been a dream. To see if such a dream is
feasible, I perform a trial analysis of the third-order nonlinear susceptibility
of quantum dots. In Section V, I discuss the use of quantum dots for
high-density optical memories using persistent spectral hole burning. In
Section VI, I summarize and discuss the roles of quantum-dot-based devices
in the future of optoelectronics.
Because of our so-far limited knowledge of the quantum dots, my
discussions in this chapter are not concrete. I simplify my analysis to show
the feasibility of quantum dots. From an engineering point of view, such an
approach is sometimes more efficient that waiting for the establishment of a
full-blown and rigid theory and complete experimental evidence.
1. THE QUANTUM
DOTAS A TWO-LEVEL
SYSTEM
What makes quantum dots attractive is that their carriers are strongly
confined three-dimensionally in a very narrow region; this gives discrete
energy levels. There are also many unresolved problems with quantum dots:
7 APPLICATIONS
OF QUANTUM DEVICES
DOTTO OPTICAL 289
how the carriers relax to the ground state (Bockelmann and Bastard, 1990;
Bockelmann, 1993; Efros et al., 1995), the dephasing time of the ground-state
wave function, the effect of the Coulomb interaction on the emission spectra
(Hu et al., 1990), and so on. However, here I put aside such problems and
focus on some of the fundamental properties of quantum dot, assuming it
to be a simple discrete two-level system. Most of these properties can be
illustrated by this assumption.
Linear Gain
where c is the light velocity; y~is the refractive index; R is the crystal volume;
n is the dot number; e is the electron charge; E, is the vacuum dielectric
constant; Y : and 4':' are the wave function (including the Bloch part) of the
ground-state electron and hole, respectively; f;. and f , are the Fermi-Dirac
distribution function; EL and E:, are the ground-state energy of the electron
and hole, respectively; and y,, is the inverse of the dephasing time. The suffix
i denotes the ith dot, and the summation is for all dots in a volume, R. The
square of the matrix element in Eq. (7.1) can be expressed by a matrix
element for bulk (Kane, 1957) as
where, $: and Ic/i are the envelope wave functions of the ground state; m, is
the electron mass, mr6 is the electron effective mass at r6;E , is the band
gap of the quantum-dot material; and A is the spin-orbit splitting energy.
From Eq. (7.1) we see that, at resonance, where Ek - El = ho,the linear
gain is inversely proportional to hy,,, which is a broadening determined by
the dephasing time. We obtain a large gain for a small broadening of Wy,,,,
which is highly attractive for semiconductor lasers because a high gain
results in a very low threshold current.
What is interesting about the linear gain is the refractive index dispersion.
Figure 7.l(a) shows schematically the refractive index dispersion of a
quantum well. Refractive index, 17, is governed by dispersions due to various
oscillators (Adachi, 1982). Among them, the carrier-density-dependent re-
290 HIROSHIISHIKAWA
High carrier
density
Gain,
change
.. ..
,* * .
fractive index change results from the change in the optical gain spectrum
caused by band filling, band-gap renormalization, and the effect of plasma
resonance (see, for example, Bennett et al., 1990). Any change in the gain
spectrum changes the refractive index through the Kramers-Kronig relation.
In a typical quantum well with a high-density carrier injection, an increase
in carrier density causes a reduction in the refractive index a t the wavelength
of gain region. This reduction, Aq, amounts to -4 x 10-20n, where n is
the carrier density per cubic centimeter (Lee et al., 1986). Thus, a change of
5 x 10'7cm-3 in carrier density gives a refractive index change of about
0.02.
There is no plasma dispersion associated with carriers in the quantum
dots. Moreover, the gain is symmetric around the resonant wavelength.
Figure 4.l(b) shows schematically the gain and refractive index dispersion
of a quantum dot. If the shift of the ground-state transition wavelength for
a change in carrier number is very small, we obtain an almost zero index
change for a change in carrier number at the resonant wavelength. In the
actual case, a red shift of the resonant wavelength may take place when a
carrier is added to the ground state to form a bi-exciton (Hu et al., 1990),
and the addition of carriers to upper sublevels may also cause some change
in the resonant wavelength through the Coulomb interaction. These shifts
are yet to be analyzed quantitatively but they are presumed to be within a
few meV. In actual quantum dots there is an inhomogeneous broadening,
to be described later, which cancels out the change in refractive index near
the gain peak. If we can reduce the plasma dispersion effect of carriers at the
7 APPLICATIONS DOTTO OPTICAL
OF QUANTUM DEVICES 291
barrier layer and the wetting layer, we might obtain an almost zero
refractive index change at the gain peak wavelength for the change in carrier
number in the quantum dot. As will be discussed in Section 111, there is a
possibility of obtaining no wavelength chirp under high-speed modulation
in quantum-dot lasers.
4e4
--z
E& =
<
I y:l Iy I)I4( L - .t.)
(hw,- EL + EL $- ihycC)(hwp- ho,+ ihy,,)(hw - Ei
J"
+ E:;- ihy,,)
where we assume a four-wave mixing process and only the resonant terms
are shown. The x’~’is doubled taking account of the spin. In Eq. (7.2), w,,
ws, and w, are angular frequencies of the pump wave, signal wave, and
conjugate wave, respectively. ycc and y,,,.are the dephasing rate of each level.
In the degenerate case, that is, Q,, = oiS,this reduces to
f3'(wp= w s )=
We can expect a large f3) for the smaller dephasing rate y,,,, as in the case
292 HIROSHIISHIKAWA
of linear gain. Of course, this also holds for the nondegenerate case. The
large x'j) encourages us to use quantum dots as a nonlinear medium for
various device applications. The response speed, or a bandwidth of ~ ( j ) ,
depends on the dephasing rate yes. The bandwidth reduces for small y,,, as
can be read from Eq. (7.4).
We can derive the rate equation for photons using x"’ and f 3 ' under the
degenerate on-resonant case, that is, w = cop = o,= (Ef - EI)/A. We assume
a single polarization optical field with an angular frequency of w. We start
from the equation for the optical electric field E (Kuwatsuka et al., 1997) as
dS S
-= t.,T(G - ES)NS - -
dt 'ph
where, the term of photon lifetime z,, is added assuming a laser cavity and
i n
(7.7)
(7.9)
(7.10)
(7.11)
2. ATTRACTIVEFEATURES
OF QUANTUM
DOTSFOR DEVICE
APPLICATION
On the basis of their basic properties, and from the two-level model just
described, the attractive features of quantum dots for device application can
be enumerated as follows:
1. Their discrete energy levels may open up new applications for quantum
dots. In the quantum well, light with an energy greater than the band-
edge energy is absorbed. In quantum dots, there are transparent
regions between the discrete levels when the inhomogeneous broaden-
ing is not large. In Section 111, we discuss an optical intensity
modulator that makes use of this feature.
2. There is a possibility of obtaining a very small refractive index change
at the resonant wavelength for the change in carrier number in the
quantum dot. This will give a very small wavelength chirp under
high-speed direct modulation in quantum-dot lasers, as will be dis-
cussed in Section 111.
3. Discrete energy levels also affect carrier dynamics. When the LO-
phonon energy does not coincide with the level separation, carrier
energy relaxation is slow, causing the so-called phonon bottleneck
(Bockelmann and Bastard, 1990). However, the slow rate of carrier
energy relaxation means the slow excitation rate of carriers from the
dot. This will be advantageous in an infrared photodetector using the
subband transition because it gives a small dark current. Readers
interested in the quantum-dot infrared photodetector should refer to
Ryzhii (1996).
4. Large x'l) and f 3 ) will be obtained in quantum dots if we can reduce
the inhomogeneous broadening. A large x(l) gives a large linear gain,
which means that we can expect very low threshold current lasers. A
large x(3) will enable us to realize optical nonlinear devices. The
response speed of fl) and x’~’depends largely on the dephasing rate,
y,,, which is determined by the carrier dynamics. The quantum dots'
unique carrier dynamics may give a different dephasing rate from that
of conventional quantum wells. Section IV discusses x(3’ together with
carrier dynamics in optical nonlinear devices.
5. The surface density of a quantum dot grown in the typical SK mode
is about 10"/cm2. If we have both large dot density and large non-
uniformity, leading to many different transition energy levels (equival-
ent to large inhomogeneous broadening), we may be able to use this
as a huge-capacity optical memory. The application of quantum dots
in Tb/cm2 memory is discussed in Section VI.
7 APPLICATIONS DOTTO OPTICAL
OF QUANTUM DEVICES 295
1. THENEEDFOR HIGH-SPEED,
LOW-WAVELENGTH-CHIRP
LIGHTSOURCES
(7.12)
where P is the light power, r is the optical confinement factor, 0 is the gain
compression factor (as discussed in Eq. (7.6)), V,,, is the active layer volume,
yld is the differential quantum efficiency, and hv is the photon energy. The
parameter u is (Henry, 1982)
(7.13)
296 HIROSHIISHIKAWA
which is proportional to the ratio of the refractive index change to the gain
change for the change in carrier concentration N . The first term in Eq. (7.12)
gives the change related to the time change of the optical power originating
from the phase modulation of light. The second term gives the offset
wavelength shift, which depends on the optical power. The value of tl in
multiple quantum well (MQW) lasers is around 3 to 4 (Green et al., 1987).
In semiconductor lasers under very fast modulation, the second term is
mostly responsible for the chirp (Wiesenfeld et al., 1987). Its coefficient
depends largely on the device structure. Wavelength chirping is often
evaluated by the time-averaged spectral width. In quantum-well DFB lasers,
a broadened spectral width due to chirping of 0.28 nm (-20 dB full
width) was reported under 2 Gb/s NRZ (non-return zero) modulation
(Kakimoto et al., 1990), and one of 0.34nm was reported under 10Gb/s
NRZ modulation (Uomi et al., 1991). The chirp at 10Gb/s modulation is
large enough to limit the transmission distance in 10Gb/s systems. For a
fiber dispersion of - 17 ps/km/nm, 0.34-nm chirp causes a pulse broadening
of 578ps for a 100-km transmission. This value exceeds the time slot of
100ps for 10 Gb/s NRZ modulation.
(7.14)
where the derivatives are by the applied voltage, I/: The wavelength chirp
for the external modulation is given by substituting this a value in Eq. (7.12)
and eliminating the second term, which is specific to semiconductor lasers.
Figure 7.2(a) illustrates how the tl parameter is determined in the MQW
modulator. As shown in the figure, application of the electric field causes a
7 APPLICATIONS DOTTO OPTICAL
OF QUANTUM DEVICES 297
A
Exciton peak A
Signal
a. +a.
lv&o>a.]+o v = v, v=o v = v,
a.
v=o
shift of the exciton absorption peak. In association with this peak shift, the
refractive index dispersion also shifts, as shown. The refractive index change
is positive, and the gain change is negative; thus, the CI parameter is positive.
This refractive index change gives the blue chirp at the turn-on and the red
chirp at the turn-off, as can be read from Eq. (7.12). The blue-chirped
leading edge of the light pulse is faster than the red-chirped pulse end
because of a fiber dispersion. Therefore, this chirping widens the pulse width
when transmitted through an optical fiber.
In the MQW modulator, however, the CI value is typically 0.2- 1.4
(Kataoka et al., 1994), which is smaller than in the MQW laser. In addition,
there is no second term of Eq. (7.12). Therefore, the wavelength chirp in
external modulators is smaller than that of the direct modulation of
semiconductor lasers. At lO-Gb/s NRZ modulation, a wavelength chirp of
0.15 nm (-20 dB full width) was reported, but this limits the transmission
distnce to about 25 km (Haisch et al., 1994). If we can get a negative CI
parameter, we can get pulse compression in the transmission system. When
the chirp is controlled to give a narrow pulse width for a given transmission
298 HIROSHIISHIKAWA
2. DIRECT
MODULATION
OF QUANTUM-DOT
LASERS
1
(7.15)
group velocity of light, and x represents the ratio of carriers in the upper
subband and the barrier to those in the ground state (Nagarajian et al.,
1992; Ishikawa and Suemune, 1994). At first sight, one benefit of the
quantum-dot laser is that the differential gain, G’ can be made very large if
we can reduce the inhomogeneous broadening, because G’ is inversely
proportional to the inhomogeneous broadening width. This gives a high
f,,,. However, factors specific to quantum-dot lasers prevent a large fmax:
factor and the gain compression factor E. Recent experimental and
theoretical studies suggest a large carrier relaxation time to the ground state
in quantum dots of 10-1OOps (Mukai et al., 1995; Mukai et al., 1996b;
Uskov et al., 1997; Uskov et al., 1998). This is much slower than that in the
MQW laser, which is around 10- 100 fs (Seki and Yokoyama, 1994). As the
factor x is the ratio of total carriers to those in the ground state, it becomes
large because of the slow energy relaxation. This slow energy relaxation also
causes a large value of E resulting in gain saturation at high power levels.
There is a possibility that the product of Zx will be a large value which would
cancel out the large differential gain.
There also is a frequency rolloff associated with carrier transport
(Nagarajian, 1992). In SK-mode quantum dots, carriers are injected into the
barrier layer and then diffuse to a quantum dot and relax to the ground
state. These carrier transport time constants give the rolloff in modulation
response, S(o,), given by
(7.16)
where omis the modulation angular frequency, zd is the diffusion time, and
T ~ is the
, carrier
~ ~ energy
~ relaxation time. If we assume a diffusion distance
of 50 nm and that a diffusion constant of 7 cm2/s, zd is 1.8 ps, and if we
assume a Zrefax of 20 ps, Eq. (7.16) gives a rolloff of -2.1 dB at a modulation
frequency of 10GHz. This roleoff is mainly due to the energy relaxation
time.
Experimental Reports
the small signal modulation response in both types. In the SCH type, the
optical gain measured by the Hakki-Paoli method in the 4-layer stacked
quantum-dot laser was 2.5 x 10-'4cmz, which was larger than in the
single-layer laser. Nevertheless, there was no significant increase in the
modulation bandwidth, which was limited to 6.2 GHz. The tunnel injection
laser showed a larger differential gain of 6 x 10-'4cm2, however, the
bandwidth was 8.5 GHz. Also, the increase in bandwidth did not correspond
to the increase in differential gain. The researchers concluded that the
bandwidth was limited by the carrier relaxation, or the phonon bottleneck.
This result fits with our discussion that slow carrier relaxation gives a large
value of E x to limit the modulation bandwidth.
A similar result was reported by Mao et al. (1997). They measured the
relaxation oscillation frequency of InGaAs/GaAs self-organized quantum-
dot lasers, observing a saturation of the relaxation oscillation frequency
despite an increase in optical power. The relaxation oscillation frequency
saturated at 5.3 GHz at the optical power of 9.1 mW, which also suggests a
large value for the gain compression factor, 2, and the existence of the
phonon bottleneck.
Wang et al. (1997) reported a comparison of wavelength chirping in
quantum-well, quantum-wire, and quantum-dot lasers under an ultra-short
optical pulse excitation. They fabricated InGaAs/InGaAsP DFB-type lasers
emitting at 1.4pm using the etching and regrowth technology. Their
quantum-dot diameter was 60 nm, which is a little too-large for sufficient
three-dimensional carrier confinement. The researchers reported that the
wavelength chirp was smaller in the higher-dimensional carrier confinement,
as shown in Fig. 7.3. In quantum-well and quantum-wire lasers, lasing under
pulsed excitation starts at a carrier density higher than the threshold level.
Then the lasing starts at a shorter wavelength. As a depletion of carriers
refractive index increases to give longer wavelength lasing, that is, the red
chirp takes place. What is interesting is that Wang et al. reported almost
zero-frequency chirp in the quantum-dot laser except for a slight blue chirp
at the onset. Their explanation was that the chirping behavior in the
quantum-dot laser is determined predominantly by the carriers in the barrier
layer because of a very small optical confinement to the dot layer of 0.5%.
The carrier relaxation process, which is due mostly to the Auger scattering of
carriers, generates high-energy carriers in the barrier layer. This causes
carrier heating, and which might cause a chirping different from that in the
quantum-well or the quantum-wire laser. Wang et al.'s interpretation did not
refer to the c( parameter of the quantum dot, and it considered the effect of
carriers in the barrier layer to be the cause of anomalous chirp behavior. This
finding is somewhat different from what I have discussed in this section.
However, their result is encouraging in that almost zero chirp was obtained.
7 APPLICATIONS DOTTO OPTICAL
OF QUANTUM DEVICES 301
FIG. 7.3. Wavelength chirp of quantum-well, quantum-wire, and quantum-dot lasers. Three
solid lines represent lasing optical output under optical pulse excitation. The dotted lines
represent the change in the lasing wavelength (Wang et al., 1997, 0 1997, Optical Society of
America).
Summary
3. THEQUANTUM-DOT
INTENSITY
MODULATOR
c
A
1280-
s
c
0
960-
O IlO 1w t30 th 4b io 0
Fleld (kVlcm)
FIG. 7.5. Calculated absorption coefficient and the a parameter, after Sahara et al.. 1996, 8
1996 I E E E .
chirping to compress the light pulse. The positive a parameter at low field
region is not at all a problem, because the light is almost absorbed by
quantum dots under the low field. It was deduced that for a 100-pm-long
device with an optical confinement factor of 0.25, we could obtain trans-
mission changes from 0.57 to 0.017, corresponding to a 15-dB extinction
ratio.
The analysis by Sahara et al. is a preliminary one. The researchers
assumed a Stark shift the same as that of the quantum well, but in actual
quantum dots a Coulomb interaction forms the exciton or bi-exciton, which
might have some effect on the Stark shift. The assumed inhomogeneous
broadening of 34meV is based on the SK-growth mode. The progress of
growth technology will largely reduce it, and a narrower broadening will
largely reduce the absorption loss between discrete levels to give a much
larger modulation extinction. At present, practical modulators that make
use of the quantum dot are not a realized application, given the problems
to be solved. However, they are worth more detailed investigation because
of the need for controlled-chirp high-speed modulation for longer-span
transmission systems.
1. THENEEDFOR LARGENONLINEARITY
WITH A LARGE
BANDWIDTH
It has long been a dream to achieve large optical nonlinearity with large
bandwidth using semiconductor structures for optical nonlinear devices.
There are many applications waiting for this dream to become a reality:
304 HIROSHIISHIKAWA
%
0,
-
-
Semiconductor optical amplifier
(b)
FIG.7.6. Four-wave mixing in a semiconductor gain medium: (a) an optical amplifier to
which both the pump wave and the signal wave are coupled: (b) a single-mode semiconductor
laser, in which the lasing beam itself is used as a pump wave.
7 APPLICATIONSOF QUANTUM DEVICES
DOTTO OPTICAL 305
s
= 0 ' " " " " ; ' " " " " ~
>r
g -10
.-
aa
.-0
=aa -20
E
.-02 -30
aa
>
5 -40
1500 1520 1540 1560 1580 1600
Signal wavelength ( nm )
FIG. 7.7. Example of a conversion efficiency from a signal to the conjugate wave.
306 HIROSHIISHIKAWA
10-14, r
2. ANALYSISOF x ( ~ )
i-fr
l---i-r
Thermal excitation
---*--
Auger-like excitation
--om-
the carrier exchange process provided when the carrier concentration in the
barrier layer is high. The e-e or h-h scattering through the Coulomb
interaction causes the exchange of a bound electron/hole and a free
electon/hole.
To explore the feasibility of third-order nonlinear susceptibility of quan-
tum dots, I perform a trial calculation, taking the dephasing rate due to the
carrier exchange as a major cause of dephasing. I introduce some rough
approximations and estimate the third-order nonlinear susceptibility using
Eq. (7.2) of Section 11.
1
Yi(rl,r2) = exp(ik*r,)Y’,(r ,) (7.19)
Jn
~
carrier and the ground-state carrier. The final state wave function is
1
YJ(rl, rz) = exp(ik'-rl)Yd(rz) (7.20)
Jn
~
This means that the incident electron is captured at the ground state of the
quantum dot and the bound electron in the dot is excited, becoming a free
electron with a wave number of k'. The interaction Hamiltonian for the
process is a screened Coulomb potential, as
9= (qz (7.22)
This matrix element can be evaluated by the Fourier transform, Fd(k), of the
ground-state wave function as
(7.24)
Then
- d3k" (7.25)
e-~k'.r,~;(kt!)e-Ik".r,Herk.r, Fd(k"')e'k"'~'~d3r
310 HIROSHIISHIKAWA
Assuming that the band is isotropic in K space, the scattering rate is given
by Fermi’s golden rule as
w =2 . 2
h
jk
d3k I, ($r
\MI2 f(k)(l - f(k))G(E - Ef)d3k’ (7.26)
where the initial-state energy, E i , and the final-state energy, E,, are given by
(7.27)
1oi4 1 I I I
300 K
109' I I I I I
lo** 1015 1016 10'7 10'8
Carrier density in barrier ( ~ r n - ~ )
FIG. 7.10. Calculated exchange scattering rate as a function of carrier density in the barrier
layer.
quantum dots, and by summing for all dots. The dot density was at
2 x 1016/cm3,which is equivalent to the volume of 8%, The result is shown
in Fig. 7.11 as a function of the angular detuning frequency between the
pump and the signal wave. Results are shown for carrier densities in the
barrier layer of 1.85 x 10"/cm3 and 1.94 x 10"/cm3. The ycc, yo,, and y,,
for the carrier concentration of 1.85 x 1017/cm3 are 1.1 x lo", 4.08 x
lo", and 5.18 x 101's-', respectively, and, for 1.94 x 101'/cm3, 5.03 x
lo", 1.107 x and 1.159 x 10'3sp1,respectively.
3. DISCLJSS~ON
Dephasing Rate
10-13 I
- 1 . k x 1017 crn-a
FWHM=6meV --.1.94 x 1018m-a
-N
>
N
-
Y
E
I
L
X
Detuning, Aw (mas)
FIG.7.11. Calculated x ‘ ~ as
’ a function of detuning between the pump wave and the signal
wave.
shorten dephasing time, however, from the measurement and analysis of the
real carrier energy relaxation, tlelaris estimated to be 10-loops (Mukai,
1995; Mukai et al., 1996b; Uskov et al., 1997; Uskov et al., 1998). The
modulation bandwidth is limited to 5-8 GHz (Kamath et al., 1997; Mao et
al., 1997; Wang et al., 1997). These estimates indicate that dephasings
associated with the energy relaxation is not so fast. Exchange scattering
could be an important way to quicken the response of quantum dots when
the carrier density in the barrier layer is high. Also interesting is that our
dephasing rate depends on the quantum dot’s geometry. This can be seen
from the functional form of the matrix element, which is
where k”’ is the wave number of the ground-state carrier, and k is the wave
number of an incident carrier. This equation means that the maximum is
obtained when the K-space distribution of the ground-state electron wave
function reaches the wave number of the incident electron. Because the
K-space distribution of IFd(k”’)ldepends on the geometry of the quantum
DOTTO OPTICAL
7 APPLICATIONSOF QUANTUM DEVICES 313
dot, the dephasing rate may depend on the geometry. Our calculation is for
a spherical dot, so it will be interesting to extend the calculation to SK dots
whose geometry is not spherical.
Asymmetry in f3)
Thus, with a simplified model and analysis, I have calculated the x(3) of
quantum dots, and I have demonstrated the possibility of obtaining fast
responses of around 1 THz, and also a much larger value than that of
quantum wells. In addition, I have shown the possibility of obtaining a
symmetric f 3 ) with respect to pump wavelength. Of course, development of
more precise and rigid theory is required to draw decisive conclusions,
especially for the dephasing process.
Our discussion is for a four-wave mixing process, but prospective appli-
cations are not limited to that. I emphasized the small refractive index
change close to the center of the inhomogeneously broadened spectral line,
which should eliminate asymmetry in f 3 ’ . However, at the shoulder regions
of the inhomogeneously broadened line, we can expect a large change in the
refractive index due to the large x(~).Possibly we can control the refractive
index at the shoulder region by the pump power. There will be applications
for all optical switches.
So far, no concrete measurement of third-order susceptibility has been
reported. Experimental verification of the usefulness of the quantum-dot’s
nonlinearity is a very important task, together with development of a more
rigid and precise theory of nonlinearity.
1. PERSISTENT
SPECTRAL
HOLEBURNING
MEMORY
USING
QUANTUMDOTS
(7.29)
RESULTS
2. EXPERIMENTAL
FIG. 7.12. Band lineup of a sample used in the persistent hole-burning experiment by
Imamura et al. (1995, 0 1995 J p n . J . Appl. Phys.).
7 APPLICATIONS
OF QUANTUM
DOTTO OPTICAL
DEVICES 317
lo-’ I I I I I I l i
lnAs QDs, 300 K
-
Q)
c
L
.
- I I I t I I I
10-0
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
At (ms)
FIG 7 13 Difference in the photo-induced current between the write pulse and the read
pulse as a function of a time interval between the two pulses (Imamura et al , 1995, P 1995
J p n J Appl P h y s )
n+-GaAs,300 nm 1
I (001) n+-GaAssubstrate I
1 x lo-”
9
U
e 0
2
-1 x lo-”
1063 1084 106S 1066
Wavelength (nm)
FIG. 7.15. Comparison of experimental and calculated results of spectral hole burning at
5 K: (a) the experimental result, (b) a calculated fit. The writing wavelength of YAG consists
of three wavelengths, as indicated by the arrows (Sugiyama et al., 1997, 0 1997 IEEE).
with five-layer stacked quantum dots at the i region. Each layer contained
dots with a density of 5 x 10’o/cm2. The sample was irradiated by a CW
YAG laser at 1064 nm, and the reading was taken using a tunable CW TiS
laser. The YAG laser consists of several emission lines 0.1 or 0.2 nm apart
with a linewidth of 27 peV. The linewidth of the TiS laser is below 165 peV.
Measurements were performed at 5 K. After irradiation by the YAG laser
at 8-mW output, the photoresponse current was measured irradiating by the
TiS laser at 30pW. When the reverse-bias voltage to the sample was below
4 V, the researchers observed no spectral hole in the response current.
However, when they increased the inverse bias up to 5 V , an apparent
spectral hole became visible. Trace (a) in Fig. 7.15 is the measured spectral
hole under 6-V reverse bias. At a low bias voltage, the generated electron
and hole remained in the dots and recombined, producing no spectral hole.
When the bias voltage was increased, the electron tunneled out from the dot,
initiating persistent hole burning. This took place for a bias voltages larger
than 4 V.
Sugiyama and his colleagues simulated the shapes of the spectral hole.
Assuming a Gaussian-shaped TiS read light with an FWHM of 165peV,
and assuming a Lorentzian-shaped spectrum for both the YAG read light
and the quantum dot, they performed the fit with a measured spectral hole,
taking convolutions. Trace (b) in Fig. 7.15 shows the calculated result. Note
7 APPLICATIONS DOTTO OPTICAL
OF QUANTUM DEVICES 319
that the YAG emission line consists of three lines, as indicated by arrows in
the figure. The best fit was obtained for a homogeneous broadening smaller
than 80 peV.
3. DISCUSSION
Finally, I sketch the trends of optoelectronics and illustrate some uses for
quantum-dot optical devices.
1. TRENDSIN OPTOELECTRONICS
DOTOPTICAL
2. USESFOR QUANTUM DEVICES
Acknowledgment
The author would like to give hearty thanks to his collaborator, Dr. H.
Kuwatsuka, for his comments and advice. The study of Section IV quantum
dots as a nonlinear medium has been done under the management of
Femtosecond Technology Research Association (FESTA), supported by
the New Energy and Industrial Technology Development Organization
(NEDO) Japan.
REFERENCES
Bowers, J. E., Hemenway, B. R., Gunack. A. H., and Wilt, D. P. (1986). IEEE J . Quantum
Electron. QE-22, 833.
Devaux, F., Bigan, E., Ougazzaden, A., Huet, F., Carre, M., and Carenco, A. (1992). Electron
Lett. 28, 2157.
Efros, A. L. Kharchenko, V. A., and Rosen, M. (1995). Solid State Comm. 93, 281.
Green, C . A., Dutta, N. K., and Watson, W. (1987). Appl. Phys. Lett. 50, 1409.
Haisch, H., Baumert, W., Hache, C., Kiihn, E., Klenk, M., Satzke, K., Schilling, M., Weber, J.,
and Zielinski, E. (1994). Proc. 201h Eur. Con$ Optical Communication 2.
Henry, C. H. (1982). IEEE J . Quantum Electron. QE-18, 259.
Hu, Y. Z., Lindberg, M., and Koch, S. W. (1990. Phys. Rev. B. 42 1713.
Imamura, K., Sugiyama, Y., Nakata, Y., Muto. S., and Yokoyama, N. (1995). Jpn. J . Appl. Phys.
34, L1445.
Ishikawa, H., Soda, H., Wakao, K., Kihara, K., Kamite, K., Kotaki, Y., Matsuda, M., Sudo,
H., Yamakoshi, S., Isozumi, S., and Imai, H. (1987). IEEE J. Lightwave Technol. LT-5,848.
Ishikawa, H., and Suemune, I. (1994). IEEE Photonic Technol. Lett. 6, 1315.
Ishikawa, H., Shoji, H., Nakata, Y., Mukai, K., Sugawara, M., Egawa, M., Otsuka, N.,
Sugiyama, Y., Futatsugi, T., and Yokoyama, N. (1998). J . Vacuum Sci. & Technol. A16,794.
Kakimoto, S., Nakajima, Y., Sakakibara, Y., Watanabe, H., Takemoto, A,, and Yoshida, N.
(1990). IEEE J . Quantum Electron. 26, 1460.
Kamath, K., Klotzkin, D., and Bhattacharya. P. (1997). IEEE Lasers and Electo-Optics SOC.
Ann. Meet. ThY4, San Francisco.
Kane, E. 0. (1957). J . Phys. Chem. Solids 1. 249.
Kataoka, T., Miyamoto, Y., Hagimoto, K.. Sato, K., Kotaka, I., and Wakita, K. (1994). Electron
Lett. 30, 872.
Kirstaedter, N., Ledentsov, N. N., Grundmann, M., Bimberg, D., Ustinov, V. M., Ruvimov,
S. S., Maximov, M. V., Kop’ev, P. S., Alferov, Z. I., Richter, U., Werner, P., Gosele, U., and
Heydenreich, J. (1994). Electron. Lerr. 30. 1416.
Koch, T. L., and Linke, R. A. (1986). A p p l . Pkys. Lett. 48. 613.
Koyama, F., and Iga, K. (1985). Elerrron. Lerr. 21, 1065.
Kuwatsuka, H., Shoji, H., Matsuda, M.. and Ishikawa, H. (1995). Electron Lerr. 31, 2108.
Kuwatsuka, H., Shoji, H., Matsuda, M., and Ishikawa, H. (1997). I E E E J . Quantum E/ectron.
33, 2002.
Lee, Y. H., Chavez-Pirson, A,, Koch, S. W.. Gibbs, H. M., Park, S. H., Morhange, J., Jeffery,
A,, and Peyghambaridn, N. (1986). Phys. Rev. Lett. 57,2446.
Mao, M.-H., Heinrichsdorff, F., Kitstaedter, N., Krost, A,, Ledentsov, N. N., and Bimberg, D.
(1997). IEEE Lasers and Electro-Optics Sor. Ann. Meet. ThY5, San Francisco.
Morito, K., Sahara, R., Sato, K., Kotaki, Y . (1996). IEEE Phofonic Technol Lett. 8. 431.
Mukai, K., Otsuka, N., Shoji, H., and Sugawara, M. (1995). Phys. Rev. B. 54, 5243.
Mukai, K., Ohtsuka, N., and Sugawara, M. (1996a). J p n . J . Appl. Phys. 35, L262.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1996b). Appl. Phys. Lett. 68, 3013.
Muto, S . (1995). Jpn. J . Appl. Phys. 34, L210.
Nagarajian, R., Ishikawa, M., Fukushima. T.. Geel, R. S., and Bowers, J. E. (1992). I E E E J.
Quantum Electron. 28, 1990.
Nakata, Y., Sugiyama, Y., Futatsugi. T.. and Yokoyama, N. (1997). J . Cryst. Growth 175/176,
713.
Newkirk, M. A,, and Miller, B. I. (1993). Appl. Phys. Lett. 63, 1179.
Olshansky, R., Hill, P., Lanzisera, V., and Powazinik, W. (1987). I E E E J . Quantum Electron.
23, 1410.
Ogasawara, N., and Ito, R. (1988). Jpn. J . Appl. Phys. Lett. 27, 615.
Ryzhii, V. (1996). Semirond. Sci. Technol. 11. 759.
7 APPLICATIONS DOTTO OPTICAL
OF QUANTUM DEVICES 323
Sahara, R., Matsuda, M., Shoji, H., Morito, K., and Soda, H. (1996). IEEE Photonic Tecliriol.
Lett. 8, 1477.
Seki, S., and Yokoyama, K. (1994). Phys. Ren B. 50, 1663.
Seino, M. (1996). 1st Optoelectronics and Cornmunicutions Conj.' (Technical Digest) 17D-2, 182.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara, M., Yokoyama. N.. and Ishikawa,
H. (1996). Electron. Lett. 32, 2023.
Sinha, K. P., DiDomenico Jr., M. (1970). Phys. Rev. B. 1, 2623.
Soda, H., Sato, K., Nakai, K., Ishikawa, H., Imai, H. (1988). Electron Lett. 24, 1194.
Sugiyama, Y., Nakata, Y., Muto, S., Horiguchi, N., Futatsugi, T., Awano, Y., Yokoyama, N.
(1997). Electron Lett. 33, 1655.
Tucker, R. S. (1985). IEEE J. Lightn~nvr~ Technol. LT-3, 1180.
Uomi, K., Tsuchiya, T., Nakano, H., Aoki, M., Suzuki, M., Chinone, N. (1991). I E E E J .
Quantum Electron. 27, 1705.
Uskov, A. V., Adler, F., Schweizer, H., and Pilkuhn, M. H. (1997). J . Appl. Phys. 81. 7895.
Uskov, A. V., Mcinerney, J., Adler, F., Schweizer, H., and Pilkuhn, M. H. (1998). Appl. Phys.
Lett. 72, 58.
Wang, J., Griesingter, U. A., and Schweizer, H. (1997). Con$ Lusers nd Electro-Optics CWF19,
Baltimore.
Watanabe, S., Naito, T., Chikama, T. (1993). IEEE Pliotonic Technol. Lett. 5. 92.
Watanabe, S., Kuwatsuka, H., Takeda. S., and Ishikawa, H. (1997). Electron. Leti. 33. 316.
Wiesenfeld, J. M., Tucker, R. S., and Downey, P. M. (1987). Appl. Phys. Lett. 51, 1307.
This Page Intentionally Left Blank
SEMICONDUCTORS AND SEMIMETALS, VOL. 60
CHAPTER 8
Yoshiaki Nakata
QUANTUM
ELOCTRON
DEVICES LABORATORY
F U J I T SLABORATORIES
~I LTD
ATSUUI,KANAGAWA, JAPAN
10
FE
W
6
E
0
% 4
1
,a
8 2
n = 2, HFUHR, CW
n
"
I J/, I , , , I , , , I , , , , , , , I , , ,I , , ,
per single sheet of dot layer). In both lasers, lasing occurred from the top of
spontaneous emission of the ground state at 1.15pm. For a laser with
L = 900pm and an HR coating on one side of the facets, an output power
of 110mW was achieved (Fig. 8.2). Even at 70°C, an output power of 75 mW
was obtained. To our knowledge, such low threshold current and high
output power have never before been reported for edge-emitting quantum-
dot lasers.
Figure 8.3 shows the lasing emission spectra of the laser with n = 1,
L = 900pm, and an HR coating on both facets up to 300mA. The lasing
started at the ground state with a threshold current of 11.5mA. As the
current increased to 60mA, the lasing spectra became broad with spikes;
then, at 80mA, lasing from the excited state appeared at around 1.1 pm.
What is surprising is that the lasing at the ground state weakened at 130mA
and completely disappeared at 200mA. The spectrum at 300mA shows
broad-band lasing only from the excited state. This jump of lasing
wavelength from the ground to the excited state can be explained by gain
saturation at the ground state, which is due to the phonon bottleneck
and/or third-order nonlinearity, and a subsequent increase in the carrier
number in the excited state (see Chapter I). The quenching of lasing at the
80
E
W
60
Y
3 40
B
0' 20
0
0 50 100 150 200 250 300 350
Current fmA)
FIG. 8.2. Output power versus current characteristics of a columnar-dot laser with ti = 2,
L = 900pn, and a high-reflective (HR) coating on one side of the facets. (Reprinted from
Mukai et at., 1998 IEEE).
328 MITSURUSUGAWARA
ET AL.
FIG.8.3. Lasing emission spectra of the laser with n = 1, t = 900prn, and an HR coating
on both facets up to 300mA.
.=
v)
C
10
Y
$ 1
a
W
.-E
v)
10.'
C
Q)
E
- 10"
10"
1 .o 1.1
(a) Wavelength (pm)
Since carriers of nonresonant dots are extracted and put into the central
lasing modes by the stimulated emission, lasing emission with one line
occurs. As the injection current increases, the gain saturation occurs due to
the phonon bottleneck and/or third-order optical nonlinearity, which results
in new lasing lines in addition to the central line. The energy separation
between the lasing lines in Fig. 8.4b- 19 and 16meV-corresponds to the
homogeneous broadening of this laser chip.
The origin of the homogeneous broadening can be attributed to the
8 THELATESTNEWS 331
In,,,,Ga,,,,As substrates (Shoji et al., 1994; Shoji et al., 1996; Otsubo et al.,
1997). Our aim is to fabricate 1.3-pm lasers on the ternary substrates with
low lasing threshold and excellent temperature characteristics for future
optical interconnection and optical subscriber systems.
The merits of ternary substrates are twofold. The first is that their
quantum wells have deeper potential height than the quantum wells on InP,
which prevents carrier overflow into the SCH layers and results in larger
optical gain and improvement of temperature characteristics (Ishikawa,
1993; Ishikawa and Suemune, 1994). Conventional strained quantum-well
lasers emitting in the infrared-that is, 1.3 and 1.55pm-grown on InP
substrates suffer from low temperature stability of threshold currents,
primarily due to the carrier overflow from the active region to the SCH
region with relatively low potential height. The carrier overflow into SCH
regions with a high density of states increases the carrier number at lasing
and thus accelerates the Auger recombination process, the rate of which is
proportional to the cube of the carrier density, resulting in the consumption
of carriers. At the same time, large carrier density at lasing greatly increases
internal cavity loss, resulting in low external quantum efficiency.
The second merit of ternary substrates is that high-reflectivity distributed
Bragg reflector mirrors can be grown for vertical cavity surface emitting
lasers. The mirrors consisting of In,,,,Ga,~,,As and Ino,25Alo~7,As on
In,,,,Ga,,,,As substrates give a refractive index difference of 0.4-about
twice that of the combination of InP and InGaAsP (A = 1.2pm) on InP
substrates (Shoji et al., 1997). This opens the way to vertical cavity surface
emitting lasers with a lasing wavelength at 1.3pm, which are the key devices
in future optical interconnections.
The combination of ternary substrate and self-assembled dots holds
promise for 1.3-pm lasers with high-temperature stability, along with alter-
nate-supply growth in metal organic vapor phase epitaxy (Chapter 3)
and the low-growth rate method plus InGaAs covering in molecular
beam epitaxy (MBE). The reason for this is that the self-assembling of
quantum dots is a strain relaxation growth process, and the larger the In
content in substrates, the larger the amount of In introduced in self-
assembled dots.
Figure 8.5(a) is an atomic field microscope image of InAs quantum dots
on the (001) ,Ga,,,,As substrate by MBE. The 450-nm In,,,,Ga,,,,As
buffer layer and then the dots were grown. The substrate temperature was
51OoC, the growth rate was 0.1 monolayer (ML)/sec, and 1.8-ML InAs was
supplied. The dots are slightly smaller and have a larger density than dots
grown on GaAs substrates. Figure 8.5(b) shows the photoluminescence
spectra at 4.2K from dots grown on the (001) In,~,,Ga,,,,As substrate
covered by lOOnm In,,,,Ga,~,,As, and dots grown on the (001) GaAs
8 THELATESTNEWS 333
enhances the migration of growth atoms, resulting in larger size and longer
wavelength emission. Also, by applying the close-stacking technique de-
scribed in Chapter 2, room-temperature 1.3-pm emission from quantum
dots can be achieved. We are now working on it.
FIG.8.7. (a) Plan-view and (b) cross-sectional view transmission electron microscope image
of lnAs quantum dots grown at an extremely slow rate of 0.007 ML/sec and covered with 8-nm
1% ,,GaLl 8 3 A S .
Figure 8.9 shows the photoluminescence emission energy from the quan-
tum-dot ground state as a function of temperature. The samples are InGaAs
quantum dots grown by ALS on GaAs substrates (Chapter 3). The dots are
covered by GaAs or In,,,Ga,,,As. The inset is the emission spectrum of the
dots covered by In,,,Ga,~,As at 20 K, which shows clear emission lines from
the ground and second states. Unfortunately, we could not detect the
emission from the Ino,,Gao,,As-covered dots above 200 K due to their low
crystal quality. The emission energy decreased as temperature increased in
both samples, as always observed in semiconductors. However, the amount
of energy shift in the In,,,Ga,,,As-covered dots was almost half that of the
GaAs-covered dots. We posit the reason for this unique phenomenon as
follows. First, since the thermal expansion coefficient of In,,,Ga,.,As is
about 94% that of GaAs, the temperature-induced release of compressive
strain, which decreases the interband transition energy, is restricted in dots
covered by In,.,Ga,.,As. Second, the asymmetric structure surrounding the
dots when covered by In,.,Ga,.,As gives rise to a remarkable change in the
8 THELATESTNEWS 337
’-0 *I
*--a- 'I
dot shape as temperature increases. These two factors work to cancel the
temperature-induced reduction in the bandgap of quantum-dot materials.
The results tell us that the temperature-induced shift of emission energy can
be artificially reduced in quantum dots, which is quite attractive for lasers
with their lasing wavelength insensitive to temperature.
References
Mukai, K., Nakata, Y.,Shoji, H., Sugawara, M., Otsubo, K., Yokoyama, N., and Ishlkawa, H.
(1998). Electron. Lett. 34, 1588.
Murray, R.,Childs, D., Malik, S., Siverns, P., Roberts, C., Hartmann, J-M., Stavrinou, P. (1998).
Nakajima, K., and Kusunoki, T. (1996). J . Crysr. Growth 169, 217.
Nakajima, K., Kusunoki, T., and Otsubo, K. (1997). J . Cryst. Growth 173, 42.
Otsubo, K., Shoji, H., Kusunoki, T., Suzuki, T., Uchida, T., Nishijima, Y., Nakajima, K., and
Ishikawa, H. (1997). Electron. Lett. 33. 1795.
Shoji, H., Otsubo, K., Fujii, T., and Ishikawa, H. (1997). IEEE J . Quanfum Electron. 33, 238.
Shoji, H., Otsubo, K., Kusunoki, T.. Suzuki. T., Uchida, T., and Ishikawa, H. (1996). Jpn. J .
Appl. Phys. 35, L778.
Shoji, H., Uchida, T., Klusunoki, T., Matsuda, M., Kurakake, H., Yamazaki, S., Nakajima, K..
and Ishikawa. H. (1994). IEEE Photon. Technol. Letr. 6, 1170.
SEMICONDUCTORS AND SEMIMETALS, VOL. 60
Index
339
INDEX
351
352 CONTENTS IN THISSERIES
OF VOLUMES
Part B
T. Misawu,lMPATT Diodes
H. C. Okean, Tunnel Diodes
R. B. Campbell and Hung-Chi Chang, Silicon Junction Carbide Devices
R. E. Enstrom, H. Kressel, and L. Krassnur, High-Temperature Power Rectifiers of GaAs, -.P,
Part B
J. I. Pankove, Introduction
G. D. Cody, The Optical Absorption Edge of a-Si: H
N. M. Amer and W. 5.Jackson, Optical Properties of Defect States in a-Sk H
P. J. Zanzucchi, The Vibrational Spectra of a-Si: H
Y. Hamakawa, Electroreflectance and Electroabsorption
J. S. Lannin, Raman Scattering of Amorphous Si, Ge, and Their Alloys
R. A. Street, Luminescence in a-Si: H
R. S. Crandall, Photoconductivity
J. Taur, Time-Resolved Spectroscopy of Electronic Relaxation Processes
P. E. Vanier, IR-Induced Quenching and Enhancement of Photoconductivity and Photo
luminescence
H. Schade, Irradiation-Induced Metastable Effects
L. Ley, Photoelectron Emission Studies
Part C
J. I. Pankove, Introduction
J. D. Cohen, Density o f States from Junction Measurements in Hydrogenated Amorphous
Silicon
P. C. Taylor, Magnetic Resonance Measurements in a-Si: H
K. Morigaki, Optically Detected Magnetic Resonance
J. Dresner, Carrier Mobility in a-Si: H
T. Tiedje, Information about band-Tail States from Time-of-Flight Experiments
A. R. Moore, Diffusion Length in Undoped a-Si: H
W. Beyer a i d J . Uverhqf, Doping Effects in a-Si: H
H. Fritzche, Electronic Properties of Surfaces in a-Si: H
C. R. Wronski, The Staebler-Wronski Effect
R. J. Nernanich, Schottky Barriers on a-Si: H
B. Abeles and T. Tiedj,, Amorphous Semiconductor Superlattices
Part D
J. I. Pankove, Introduction
D. E. Carlson, Solar Cells
G. A. Swartz, Closed-Form Solution of ILV Characteristic for a a-Si: H Solar Cells
I. Shimizu, Electrophotography
S. Ishioka, Image Pickup Tubes
CONTENTS IN THISSERIES
OF VOLUMES 351
P. G. LeComher and W. E. Speur, The Development of the a-Si: H Field-Effect Transistor and
Its Possible Applications
D. G. Ast, a-Si: H FET-Addressed LCD Panel
S. Kaneko, Solid-state Image Sensor
M . Mutsuinuru, Charge-Coupled Devices
M. A . Bosch, Optical Recording
A . D'Amico and G. Fortunato, Ambient Sensors
H. Kukimoto, Amorphous Light-Emitting Devices
R. J. Phelun, Jr., Fast Detectors and Modulators
J. I. Punkove, Hybrid Structures
P. G. LeComher, A . E. Owen, W. E. Speor, J. Hajro, and W. K. Choi, Electronic Switching in
Amorphous Silicon Junction Devices
Part B
J. P. van der Z i d , Mode Locking of Semiconductor Lasers
K. Y. Luu und A . Yariv, High-Frequency Current Modulation of Semiconductor Injection
Lasers
C. H. Henry, Special Properties of Semiconductor Lasers
Y. Suemutsu, K. Kishino, S. Arai, and F. Koyctrm, Dynamic Single-Mode
Semiconductor Lasers with a Distributed Reflector
W. T. Tsnng, The Cleaved-Coupled-Cavity (C') Laser
Part C
R. J. Nelson und M. K. Duttu, Review of InGaAsP InP Laser Structures and Comparison of
Their Performance
N . Chinone and M. Nukumuru, Mode-Stabilized Semiconductor Lasers for 0.7-0.8- and
1.1-1.6-pm Regions
Y. Horikoshi, Semiconductor Lasers with Wavelengths Exceeding 2 pm
B. A . Dean and M. Dixon, The Functional Reliability of Semiconductor Lasers as Optical
Transmitters
R. H. Suul, T. P . Lee, and C. A. Burus, Light-Emitting Device Design
C. L. Zipfel, Light-Emitting Diode-Reliability
T. P. Lee and T. Li, LED-Based Multimode Lightwave Systems
K Ogaivu, Semiconductor Noise-Mode Partition Noise
358 CONTENTS
OF VOLUMESIN THISSERIES
Part D
F. Cupusso, The Physics of Avalanche Photodiodes
T. P. Peursull and M. A. Polluck, Compound Semiconductor Photodiodes
T. Kunedu, Silicon and Germanium Avalanche Photodiodes
S. R. Forresl, Sensitivity of Avalanche Photodetector Receivers for High-Bit-Rate Long-
Wavelength Optical Communication Systems
J. C. Cumpbell, Phototransistors for Lightwave Communications
Part E
S. Wung, Principles and Characteristics of Integrable Active and Passive Optical Devices
S. Murgufit und A. Yuriv, Integrated Electronic and Photonic Devices
T. Mukui, Y. Yumamoto, and T. Kimuru, Optical Amplification by Semiconductor
Lasers