Chapter 6

OPTICAL PROPERTIES AND IMPURITY STATES IN

NANOSTRUCTURED MATERIALS

Raúl Riera, José L. Marín

Departamento de Investigaciön en Física, Universidad de Sonora, Hermosillo,
Sonora, México
Rodrigo A. Rosas
Departamento de Física, Universidad de Sonora, Hermosillo, Sonora, México

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Optical Properties of Semiconductor Nanostructures . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1. Absorption (One-electron Approximation) in Nanostructures . . . . . . . . . . . . . . . . . 3
2.2. Absorption: A Simplified Description of Excitonic Effects . . . . . . . . . . . . . . . . . . . . 10
2.3. Photoluminescence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4. Optical Properties in Semiconductor Quantum-well Wire . . . . . . . . . . . . . . . . . . . . 22
2.5. Optical Properties in Semiconductor Spherical Quantum Dots . . . . . . . . . . . . . . . . . 27
3. Optical Properties in Nanostructures II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.1. Model and Applied Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2. Electron Raman Scattering in Quantum Well . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3. Electron Raman Scattering in Quantum Dots . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.4. Electron Raman Scattering in Cylindrical Quantum-well Wires . . . . . . . . . . . . . . . . . 47
3.5. One Phonon-assisted Electron Raman Scattering in Quantum-well Wire, Free
Standing Wire, and Quantum Dot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4. Physical Effects of Impurity States and Atomic Systems Confined in
Semiconductor Nanostructures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1. Hydrogenic Impurity in Quasi-two-dimensional Systems . . . . . . . . . . . . . . . . . . . . . 63
4.2. Hydrogenic Impurities within Asymmetric and Symmetric Quantum-well Wires . . . . . . . 67
4.3. Asymmetric Confinement of Hydrogen by Hard Spherical and Cylindrical Surfaces . . . . . 70
4.4. Confined Electron and Hydrogenic Donor States in a Spherical Quantum Dot . . . . . . . . 72
4.5. Shallow Donors in a Quantum-well Wire: Electric Field and Geometrical Effects . . . . . . 77
4.6. Hydrogenic Impurities in a Spherical Quantum Dot in the Presence of a Magnetic Field . . 79
4.7. Two-electron Atomic Systems Confined within Spheroidal Boxes . . . . . . . . . . . . . . . . 82
4.8. Quantum Systems within Penetrable Spheroidal Boxes . . . . . . . . . . . . . . . . . . . . . . 87
4.9. Confinement of One- and Two-electron Atomic Systems by Spherical Penetrable Boxes . . 88
4.10. Optical Properties of Impurity States in Semiconductor Quantum Dots . . . . . . . . . . . . 92
5. Nonconventional and Idealized Confined Systems and Geometrical Effects in
Nanostructured Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.1. Ground State Energy of the Two-dimensional Hydrogen Atom Confined within
Conical Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2. Geometrical Effects on the Ground State Energy of Hydrogenic Impurities in
Quantum Dots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.3. A Harmonic Oscillator Confined within an Infinite Potential Well . . . . . . . . . . . . . . . 103
5.4. Hydrogen Atom and Harmonic Oscillator Confined by Impenetrable Spherical Boxes . . . 105
5.5. Sublevels and Excitons in GaAs–Ga1−x Alx As Parabolic-Quantum-well Structures . . . . . 107

Handbook of Advanced Electronic and Photonic Materials and Devices, edited by H.S. Nalwa
Volume 6: Nanostructured Materials
Copyright © 2001 by Academic Press
ISBN 0-12-513756-7/$35.00 All rights of reproduction in any form reserved.

1
2 RIERA ET AL.

5.6. Energy States of Two Electrons in a Parabolic Quantum Dot in Magnetic Field . . . . . . . 110
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

1. INTRODUCTION the dot to the volume of the exciton. However, Weisbuch and Vin-
ter caution us to consider the available data on weakly emitting
In this chapter, we consider the optical processes in nanocrys- dots and wires. This suggests that emission becomes weaker as the
tals that can be interpreted in terms of creation and annihilation dimensions are reduced, contrary to expectations based on the en-
of a single electron-hole pair or exciton within a nanostructure. hancement of the oscillator strength predicted from a treatment
Size-dependent absorption and emission spectra and their fine of “bottleneck” effects in energy and momentum relaxation. This
structures as well as size-dependent radiative lifetime will be dis- chapter concentrates on the spectroscopy of dry etched nanostruc-
cussed for the different nanostructured systems. Nontrivial aspects tures, with special emphasis on the intensity of luminescence from
of exciton-phonon interactions that manifest themselves in homo- quantum wells, quantum-well wires, and quantum dots.
geneous linewidths and/or intraband relaxation processes will be The optical characterization of semiconductors has been re-
outlined. viewed several times. A publication edited by Stradling and Klip-
Optical spectroscopy of semiconductors has been a lively sub- stein [2] covers most techniques in a clear introductory manner.
ject for many decades. It has helped to shape the concepts of en- It includes luminescence, scanning electron microscopy, and lo-
ergy gaps, impurity states, and resonant states among others. Of calized vibrational mode spectroscopy. A more advanced survey
course spectroscopic techniques were in use for some consider- of techniques suitable for semiconductor quantum wells and su-
able time before semiconductors were developed, in the study of perlattices is given by Fasol et al. [3] which includes time-resolved
atomic spectra, for example. spectroscopy, Raman scattering, and the effects of a magnetic or a
In spectroscopic terms, the size of the energy gap is the key dif- electric field.
ference between metals, insulators, and semiconductors. In both The application of external perturbations such as electric fields
intrinsic luminescence and near-gap absorption, the main features and their impact on the absorption and emission spectra was dis-
of the spectrum can be associated with the magnitude of the en- cussed in Bastard’s book [1]. In general, a parameter which can
ergy gap, since the photons reflect the energy spacing between the be changed in a controllable manner and which affects at least
states taking part in the optical process. This is discussed in more one of the optical properties can be used to study the optical be-
detail in the following. havior; with such a technique the background signal is subtracted
The states involved in optical transitions must be allowed states, from the processes under study. Pollak and Glembocki [4] gave an
and hence an understanding of the density of allowed states and its excellent review of photoreflectance, electroreflectance, piezore-
correlation with optical transitions is helpful. In the case of three- flectance, analysis of line-shape, and interference effects in semi-
dimensional or bulk materials, the density of allowed states fol- conductors and quantum wells. The information which can be ob-
lows a ε1/2 -dependence on the energy of the carrier and the occu- tained is usually directly related to the energy of the optically
pation of the energy levels is governed by Fermi–Dirac statistics. active electronic levels and the scattering mechanism involved.
Lowering the dimensionality to two-dimensional leads to a step- Time-resolved spectroscopy is used among other applications
like density of allowed states. In one-dimensional, the edges of the to measure the energy relaxation of quasi-particles such as excitons
two-dimensional density of allowed states develop a spike with a in quantum wells, to probe hot phonon distributions and the life-
tail decaying as ε−1/2 to high energies, and finally in quasi-zero- time of electron-hole liquids and plasmas, and to measure trapping
dimensional the density of allowed states is reduced to a set of times of carriers by impurities or other crystalline defects. There
δ-functions. are several techniques in use such as pump-and-probe, streak cam-
Not only must the states be allowed for a given optical process eras, etc. A useful series of publications covering the applications
to occur, but they must also be connected by selection rules. For of time-resolved spectroscopy to the study of three-dimensional
example, the quantum numbers of the wave function of the elec- and two-dimensional semiconductors was edited by Alfano [5].
tron and the hole may have to be same, and for luminescence in Techniques, experimental arrangements, tool for data analysis, and
two-dimensional this would translate into the requirement 1l = 0 results are covered.
where l is the orbital angular momentum quantum number of the In general, the application of large magnetic fields results in
states. confinement of the electron and the hole wave functions. Magnetic
The strength of optical transitions is usually described by the fields were used to obtain fingerprints of the two-dimensional elec-
oscillator strength, which in three-dimensional is dimensionless tron gas in the early days of low-dimensional semiconductor struc-
and of order unity. The concept of oscillator strength conveys the tures, confirming its existence by monitoring the positions of Lan-
probability that the transition can occur, and is proportional to the dau levels as a function of the tilt angle between the magnetic field
number of k-states coupled to a given energy range. As discussed and the normal to the two-dimensional layer. Similar oscillatory
by Weisbuch and Vinter [1], the oscillator strength per atom in behavior was seen in magneto-photoluminescence; in this case a
two-dimensional does not increase but the new form of the density Landau ladder appears associated with each occupied electronic
of allowed states arising from the two-dimensional confinement subband. In undoped structures, magneto-photoluminescence ex-
results in a shared kz for electrons and holes in the same subband. citation allowed the measurement of the binding energies of ex-
Having the same wave-vector kz helps to concentrate the oscillator citons in two-dimensional and one-dimensional semiconductors,
strength, when compared to three-dimensional. Going to quasi- as discussed by Heitmann [6]. Another parameter that can be ob-
zero-dimensional, for the case where the dimensions of the dot are tained is the effective mass of the electrons, from the separation
somewhat larger than the Bohr radius of the exciton, a giant oscil- between Landau levels at a given field. A comprehensive coverage
lator strength develops, proportional to the ratio of the volume of of the use of high magnetic fields can be found in the series on the
 OPTICAL PROPERTIES IN NANOSTRUCTURES 3

proceedings of the conferences on applications of high magnetic in the crystal in a electromagnetic field,
fields in semiconductor physics. Magneto-Raman spectroscopy is !2
also a very powerful technique; in doped structures it reveals the b= 1 |e|
H p̂ − A + eϕ + V(r) (3)
spacing of energy levels, effective masses, and allows probing of 2m0 c
more exotic quasi-particles, such as the rotons demonstrated by
Pinczuk et al. [7]. where A and ϕ are the vector and scalar potential of the electro-
One of the crucial problems in semiconductor physics and par- magnetic field, respectively. Considering the Landau gauge, the
ticularly in these new low-dimensional semiconductor structures is Hamiltonian can be divided into two parts: the first one corre-
the presence of ionized impurities, which play a fundamental role sponds to the electrons motion in the crystal (the first two terms)
in transport mechanisms at low temperature and optical applica- and the second one (the last two terms) represents the electron-
tions. This represents the basis for technological fabrications of radiation interaction,
the new electronic and optical devices of nanometric dimensions. 1 2 |e| e2
The investigation of the hydrogen atom and hydrogenic impurities b=
H p̂ + V(r) − p̂ · A + A2 (4)
2m0 m0 c 2m0 c
confined in different nanostructures attracted the attention of sev-
eral research groups in the last few years, mainly because of the Taking to the first order of perturbation theory we can neglect the
diverse technological applications possessed by devices based on term that contains A2 , therefore,

ber = − |e| p̂ · A
such structures.
From the physical point of view, the energy levels of hydro- H (5)
m0 c
genic impurities and excitonic states possess certain similarities,
however, from the mathematical point of view the excitonic states The vector potential of the radiation has the form of a plane wave
are more complicated. In the hydrogenic impurity states, the ion with wave vector k and frequency ω,
h i
or core is bound to the electron through the Coulombic potential A = A0 er cos (k · r) − ωt
but it remains in a fixed site, while in the formation of the exci-
1 n h i
ton the electron-hole pair is correlated through Coulombic inter-
= A0 er exp −i(k · r) exp(iωt)
action, but it moves through the whole nanostructure. 2 h i o
+ exp i(k · r) exp(−iωt) (6)
2. OPTICAL PROPERTIES OF SEMICONDUCTOR 1 ∂A ω h i
NANOSTRUCTURES E = − = A0 er sin (k · r) − ωt (7)
c ∂t c
where er is the polarization vector of the radiation. Considering N
2.1. Absorption (One-electron Approximation) photons and volume V ,
in Nanostructures
2
An optical absorption experiment consists of measuring the atten- 1 E2 A2 ω
N h̄ω = = 0 (8)
uation of a light beam, which passes through a sample of thick- V 4π 8c 2 π
ness d. If the incident electromagnetic wave is monochromatic and
propagates perpendicularly to the sample surface (area S), which
then r
2N h̄π
is supposed to be flat, it can be shown [8, 9], that the transmitted A0 = 2c (9)
ωV
intensity is given by
substituting Eqs. (6) and (9) in Eq. (5) we obtain
I (1 − R)2 exp(−αd) r
Id = 0 (1) e 2N h̄π h i
1 − R2 exp(−2αd) b
Her = − exp −i(k · r) (er · p̂) (10)
m0 ωV
where I0 is the intensity of the incident beam, R is the sample re-
flectance, and α is the absorption coefficient at the angular fre- Under typical experimental conditions the photon wave vector k
quency ω of the electromagnetic wave. For weakly absorbing me- can safely be neglected: the wave amplitude varies over distances
dia, R is given by (the wavelength 2π/k ≈ 1 µm) which are considerably larger than
!2 any characteristic dimension of electronic origin. Thus, under this
n−1 ber , considering N = 1, reduces to
R = R0 = (2) electric dipole approximation H
n+1 r
b e 2h̄π
where n is the refractive index of the sample. For absorption mea- Her = − (er · p̂) (11)
m0 ωV
surements, which probe the vicinity of the bandgap of the het-
erostructures, the ω-dependence of and R0 in Eqs. (1) and (2) can The mass of the free electron that appears in Eq. (11) is substi-
often be neglected. tuted by the effective mass in the case of the interaction with the
To calculate the absorption coefficient α(ω), the transition secondary radiation field.
probability per unit time W for a photon to disappear is first of Thus, to the first order in A (linear absorption) and neglect-
all evaluated and, as a result, α(ω) can then be obtained from the ing any contributions arising from the spin-orbit terms, the one-
knowledge of W . electron Hamiltonian of a nanostructure can be written as
b=H
H ber
b0 + H (12)
2.1.1. Electron-photon Interaction where Hb0 is the nanostructure Hamiltonian in the absence of elec-
To determine the electron-photon interaction Hamiltonian, we tromagnetic wave. In Eq. (12), the effects associated with the time-
start from the Hamiltonian that represents the electrons motion varying magnetic field of the wave were neglected.
4 RIERA ET AL.

The electromagnetic perturbation is time-dependent. It thus in- where c is the vacuum speed of light and n is the index of refraction
duces transitions between the initial states |ii and the final states of the quantum nanostructure. In Eq. (20), we have neglected the
|f i, where |νi denotes an eigenstate of H b0 with energy εν . In the spontaneous emission due to our classical treatment of the elec-
following, we take εf > εi . tromagnetic field.
The Fermi golden rule tells us how to calculate the transition γ has dimension of a [length]−1 and pif , which is proportional
probability per unit time W eif that an electron, under the action to the dipolar matrix element erif , governs the selection rules of
b
Her , makes a transition from the state |ii to the state |f i. Since the allowed optical transitions. To exploit the general expression
εf > εi , only the +ω component of A induces transitions which of α(ω), we need to have a model for the energy levels. We choose
yields: the simplest one and we write the wave functions of the initial state
D E2 |ii in the form,
eif = 2π  f |H
W ber |i  δ(εf − εi − h̄ω) (13) ψi (r) = uνi (r)fi (r) (22)
h̄
In the case where the levels |ii and |f i are partially or com- where uνi (r) is the periodic part of the Bloch functions at the zone
pletely occupied, one has to account for the impossibility of al- center (assumed to be the same in both type of layers) for the
lowing transitions either from an empty level or toward a filled band νi . fi (r) implicitly contains the effective envelope function
one. The perturbation by the electromagnetic wave of the statisti- which describes the confined motion of the electron in the sub-
cal distributions of the levels |νi of the nanostructure is assumed to band i corresponding to the extreme νi .
be negligible. The probability per unit time that an electron makes An expression similar to Eq. (22) holds for the final state. This
a |ii → |f i transition or, equivalently, that a photon disappears is expression of the wave function is given for each kind of nanos-
thus equal to tructure.
j k As long as we are interested in optical transitions from one sub-
Wif = Weif f(εi ) 1 − f(εf ) (14) band derived from a given host band (say a valence subband) to a
subband derived from another host band (say a conduction sub-
In Eq. (14), f(εν ) is the mean occupancy of the level |νi,
band), Eq. (22) is sufficient. On the other hand, if we are dealing
n h io−1 with transitions between the subbands derived from the same host
f(εν ) = 1 + exp β(εν − µ) (15)
band (say both initial and final states are conduction subbands)
Eq. (22) becomes insufficient even at the lowest order. To be con-
where β = (kB T )−1 , T is the temperature, and µ is the chemical
sistent, either we have to use the eigenfunctions which include the
potential of the electrons.
k · p̂νν0 /(εν − εν0 ) corrections of the next order and we have to
As a result of the electronic transitions |ii → |f i, the electro-
eif dt during the time interval keep a coupling Hamiltonian between carriers and light which is
magnetic wave loses an energy h̄ωW
given by Eq. (11), or we have to keep Eq. (22) for the wave func-
dt. The transitions |f i → |ii can also be induced by the wave (stim-
tion but use the effective coupling Hamiltonian between the car-
ulated emission). They occur at the rate of
riers and the electromagnetic wave. Here, to evaluate α(ω) in the
D E2
efi = 2π  i|Hber |f  δ(εi − εf − h̄ω)
case of intraband transitions, we restrict ourselves to the parabolic
W (16) dispersion relations of the host bands. The effective coupling pro-
h̄
cedure therefore amounts to replacing 1/m0 in Eq. (11) by 1/m∗e
transitions per unit time and they contribute to the creation of
taking into account the considered confinement directions. This
photons by a quantity,
means that, under most circumstances, the selection rules for the
h i
Wfi = Wefi f(εf ) 1 − f(εi ) (17) intraband transitions will be the same, irrespective of the use of a
bare or an effective coupling between carrier and light. Thus, in
per unit time. The net energy loss of the electromagnetic wave per the derivation of the intraband transition selection rules, we keep
unit time associated with |ii ↔ |f i transitions is thus, Eq. (22) for the wave function, but we restore m∗e instead of m0
where appropriate. This is relevant for the evaluation of the mag-
h̄ω[Wif − Wfi ] (18) nitude of the absorption coefficient.
Let us denote by V0 the volume of the elementary cell of the
Finally, by summing over all states |ii, |f i, we obtain the decrease
materials involved in the considered structure. We have
in the wave energy per unit time, Z
2π 2h̄πe2 X D E2 er · pif = er · ψ∗i (r)p̂ψf (r)d 3 r (23)
 
W(ω) = δ[εf − εi − h̄ω] f |er · p̂|i  D
V
E Z
h̄ m2 ωV
0 i;f er · pif = er · uνi |p̂|uνf fi∗ (r)ff (r)d 3 r
h i V
× f(εi ) − f(εf ) (19) Z
+ δif er · fi∗ (r)p̂ff (r)d 3 r (24)
The absorption coefficient α(ω) is given by V

W(ω) X 1 where D E Z
α(ω) = =γ |er · pif |2 δ(εf − εi − h̄ω) uν1 |p̂|uνf = u∗νi (r)p̂uνf (r)d 3 r (25)
nc m0 V0
i;f
h i
In Eqs. (23)–(25), we took advantage of the rapid variations of
× f(εi ) − f(εf ) (20)
uνi , uνf over k−1 or over the characteristic lengths of variation of
with fi , ff .
4π 2 e2 D E The allowed optical transitions split into two categories, on the
γ= and pif = i|p̂|f (21) one hand, the intraband transitions (νi = νf ) which involve the
ncm0 ωV
 OPTICAL PROPERTIES IN NANOSTRUCTURES 5

dipole matrix elements between envelope functions and on the quantum-well lasers. For these highly excited continuum states,
other hand the interband transitions which occur between sub- we may neglect the reflection-transmission phenomena which take
bands originating from different extrema. The selection rules for place at z = ±L/2, where L is the quantum-well thickness. Thus:
the latter have two origins:
1
(i) the overlap integral between envelope functions selects χf (z) = √ exp(ikz) (28)
`
the quantum numbers of the initial and final subbands
(ii) the atomic-like dipole matrix element huνi |p̂|uνf i gives One should define ` in an empirical way: for a perfect heterostruc-
rise to the selection rules on the polarization of the light ture ` is identified as the length over which the heterostructure en-
wave. ergy levels are well defined, i.e., a length which, if changed slightly,
does not appreciably alter the energy levels and their associated
Let us examine the intraband and interband transitions in wave functions. For a single quantum-well structure with thick-
closer detail for different nanostructures. ness L, one should take `  L. For a superlattice comprising N
periods (thickness h), one may take ` ≥ Nh if N  1. On the
2.1.2. Intraband Transitions in Quasi-two-dimensional other hand, ` should not be too large to ensure the assumption of
Systems a plane electromagnetic wave.
Besides,
The dipole matrix element between envelope functions is r !
D E Z 2 πz L
1 χi (z) = cos |z| ≤ (29)
fi |er · p̂|ff = d 3 rχ∗ni (z) exp(−ik⊥ · r⊥ ) L L 2
S
h i
× ex p̂x + ey p̂y + ez p̂z where, for simplicity, we neglected the exponential tails of the
bound state wave function in the barrier. We evaluate hχi |p̂z |χf i
×χnf (z) exp(ik0⊥ · r⊥ ) (26) and we obtain
r !" #
or D E 2 kL 1 1
D E χi |p̂z |χf = cos − h̄k (30)
fi |er · p̂|ff = (ex h̄kx + ey h̄ky )δni ;nf δk0 ;k⊥ L` 2 k + π/L k − π/L
⊥
Z
∗ The quantity,
+ ez δk0 ;k⊥ dzχni (z)p̂z χnf (z) (27)
⊥
1 X h i
The polarizations ex , ey give rise to allowed transitions only if B= ∗ |er · pif |2 δ(εf − εi − h̄ω) f(εi ) − f(εf ) (31)
me
both the initial and the final states coincide (i.e., if ω = 0). The in- i;f
trasubband absorption (ni = nf ), which in the static limit cannot is also easily evaluated if the continuum levels have negligible pop-
be reasonably treated without including scattering mechanisms, is ulation. We find
the two-dimensional analogue of the free-carrier absorption. For !" #2
perfect heterostructures, the free-carrier absorption is forbidden 2k0 LNe 2 k0 L 1 1
B= cos − (32)
in quasi-two-dimensional electron gases for the same reason as in π 2 π + k0 L k0 L − π
bulk materials, i.e., the impossibility of conserving the energy and
momentum simultaneously during the photon absorption by an where
2m∗e
electron. Free-carrier absorption may be induced by defects (im- k20 = (h̄ω − V0 + ε1 ) (33)
purities, phonons) capable of providing the momentum necessary h̄2
for the electron transition. The effect of reduced dimensionality and V0 is the barrier height. The absorption coefficient is there-
on the free-carrier absorption was calculated [10]. fore,
The polarization ez , which corresponds to an electromagnetic !" #2
wave propagating in the layer plane with an electric field vector 8πe2 ηe 2 x 1 1
α(ω) = x cos − (34)
parallel to the growth axis of the structure, leads to optical tran- ncm∗e `ω 2 x+π x−π
sitions which are allowed between subbands provided that the in-
plane wave vector of the carrier is conserved (vertical transitions in with x = k0 L, ηe = Ne /S. At large ω, α(ω) decreases like ω−5/2 .
the k⊥ space). In addition, if the heterostructure Hamiltonian has We notice that the absorption coefficient is inversely proportional
a definite parity the initial and final subbands should be of opposite to `, but does not explicitly depend on S. These results are not sur-
parities. The occurrence of intraband transition without the partic- prising. Both the initial and the final states are delocalized in the
ipation of defects is specific to the quasi-two-dimensional materi- layer plane. The carrier interacts with the electromagnetic wave
als and originates from the existence of a z-dependent potential all over the area S. On the other hand, since the initial state is
in Hb0 . This potential makes up for the momentum necessary for confined within the quantum well, only a fraction L/` of the elec-
the intraband absorption when the electric field vector of the light tromagnetic energy is available for absorption.
is parallel to z. As a second example, let us now consider the optical absorp-
In Table I, we show the selection rules for intraband (intersub- tion which promotes the carrier from the ground (ε1 ) to the first
band) transitions. excited (ε2 ) subbands of a quantum well.
Let us now evaluate the order of magnitude of such allowed To lighten the algebra, we again approximate the wave func-
intraband transitions in a rectangular quantum well whose lowest tions by those of an infinitely deep well, i.e.,
subband (confinement energy ε1 ) is occupied by Ne electrons. r !
We first consider transitions which send the ε1 electrons deep 2 πz
χi (z) = χ1 (z) = cos (35)
into the quantum-well continuum, a possible loss mechanism in L L
6 RIERA ET AL.

Table I. Selection Rules for Intraband (Intrasubband) Transition.

Polarization ex ey ez

Propagation parallel to ez Forbidden if ω 6= 0 Forbidden if ω 6= 0 Impossible

Propagation parallel to ex Impossible Forbidden if ω 6= 0 Allowed
Propagation parallel to ey Forbidden if ω 6= 0 Impossible Allowed

r !
2 2πz A major complication arises due to the intricate valence sub-
χf (z) = χ2 (z) = sin (36)
L L band dispersions in the layer plane. The decoupling of the m∗j =
±3/2 and the m∗j = ±1/2 states at k⊥ = 0 is well known. Strictly
We obtain D E 8ih̄ speaking, it is necessary to treat the problem of the in-plane dis-
χi |p̂z |χf = − (37) persion relations at k⊥ 6= 0 rigorously to be able to evaluate α(ω).
3L
In fact, publications [11, 12] dealt with this very complicated prob-
and
64h̄2 lem. Here, we bypass this difficulty by assigning an arbitrary in-
B= (N1 − N2 )δ(ε2 − ε1 − h̄ω) (38) plane effective mass to the valence subbands while retaining the
9m∗e L2
m∗j = ±3/2, m∗j = ±1/2 decoupling. This will give us a reasonable
where N1 and N2 are the number of electrons in the initial and description of the onsets of the absorptions. We notice parenthet-
final subbands whose confinement energies are ε1 and ε2 , respec- ically that if our model predicts that a given interband transition is
tively. The absorption coefficient associated with |1i → |2i transi- allowed in the vicinity of k⊥ = 0, it cannot in practice become for-
tions is then equal to bidden at k⊥ 6= 0. Symmetrically, any forbidden transition found
256π 2 e2 h̄2 in our simplified approach might become allowed due to the va-
α(ω) = (η1 − η2 )δ(ε2 − ε1 − h̄ω) (39) lence subbands mixing between the m∗j = ±3/2, m∗j = ±1/2 states
9nc(m∗e )2 L2 ω`
at k⊥ 6= 0. However, it will present a smoother absorption edge
where η1 and η2 are the areal density of carriers found in the ε1 and it will be weaker than an allowed transition. The major weak-
and ε2 subbands, respectively. ness of our simplified model lies in its inability to account for peaks
The delta singularity originates from the exact parallelism be- in the joint density of states, which arise away from the zone center
tween the in-plane dispersion relations of the initial and final sub- when the initial and final subbands are nearly parallel. We believe
bands. Introducing nonparabolicity or any other effect which alters this situation however to be rare.
this parallelism would change the delta peak into a line of finite en- In the following, we thus adopt the following subband disper-
ergy extension. It may also be noticed that we have obtained the sion relations:
same `−1 dependence for α as in Eq. (34). In both examples, the
physical absorption mechanism takes place within the quantum- h̄2 k2⊥
εhhn (k⊥ ) = −εg − hhn − (40)
well thickness, whereas the electromagnetic energy is delocalized 2Mn∗
over the whole heterostructure.
To get an order of magnitude of the absorption coefficient asso- h̄2 k2⊥
ciated with |1i → |2i transitions, let us take m∗e = 0:07m0 , n = 3:6, εlhn (k⊥ ) = −εg − lhn − (41)
2m∗n
η1 − η2 = 1011 cm−2 , ` = 1 µm, h̄ω = ε2 − ε1 = 3h̄2 π 2 /2m∗e L2
and replace the delta function by a Lorentzian function with a full h̄2 k2⊥
width at half maximum 20 = 4 meV. We obtain α ≈ 67 cm−1 . εe;n (k⊥ ) = εn + (42)
2m∗en
The absorption coefficient associated with |1i → continuum tran-
sitions is very small (α ≈ 10−3 cm−1 ) since x in Eq. (34) is usually where Mn∗ and m∗n can eventually become negative if the hhn (or
≈10 for material parameters adapted to GaAs wells. Note finally lhn ) subband has a positive curvature in the vicinity of k⊥ = 0. We
that collective excitations were not taken into account. introduce the matrix element,
We summarize the selection rules for intraband transitions in Y −i D E −i D E −i D E
perfect heterostructures in the previous table. = S|p̂x |X = S|p̂y |Y = S|p̂z |Z (43)
m0 m0 m0

2.1.3. Interband Transitions which is related to the Kane matrix element εP (≈ 23 eV) by

We have to discuss here, on the one hand, the selection rules as- Y2
εP = 2m0 (44)
sociated with the wave polarizations and, on the other hand, the
selection rules on the subband index which arise from the enve-
lope functions overlap. In addition, interband transitions raise the 2.1.3.1. Polarization Selection Rules
question of the respective location of the initial and final states. It
is clear that the optical absorption in a type-II system (e.g., InAs– Examination of the periodic parts of the Bloch functions of the
GaSb), where electrons and holes are spatially separated, should 06 , 07 , 08 bands leads to the selection rules for the electromag-
be weaker than in type-I structures (e.g., GaAs–GaAlAs) where netic wave polarization. These are summarized in Table II for the
electrons and holes are essentially localized in the same layer. hhn → εm , lhn → εm , and (07 )n → εm transitions, respectively.
 OPTICAL PROPERTIES IN NANOSTRUCTURES 7

Table II. Selection Rules for Interband (Intersubband) Transitions Obtained from the Absolute Value of the Matrix Elements hS ↑ |er · p̂|uα i,
α = 08 , 07 ; k

Polarization ex ey ez Type of transitions

Propagation parallel to ez 5
√ 5
√ Impossible hhn → εm
2 2
Propagation parallel to ex Impossible 5
√ Forbidden hhn → εm
2
Propagation parallel to ey 5
√ Impossible Forbidden hhn → εm
2
Propagation parallel to ez 5
√ 5
√ Impossible lhn → εm
6 6
Propagation parallel to ex Impossible 5
√ 25
√ lhn → εm
6 6
Propagation parallel to ey 5
√ Impossible 25
√ lhn → εm
6 6
Propagation parallel to ez 5
√ 5
√ Impossible (07 )n → εm
3 3
Propagation parallel to ex Impossible 5
√ 5
√ (07 )n → εm
3 3
Propagation parallel to ey 5
√ Impossible 5
√ (07 )n → εm
3 3

Table II shows that for interband transitions:

(a) The heavy hole → electron transitions are three times
more intense than the light hole → electron transitions.
The 07 → electron transitions have an intermediate
strength (2/3 of those involving heavy holes).
(b) For the propagation parallel to z, corresponding to the
electric field of the wave in the layer planes, the three
possible types of transitions are allowed. On the other
hand, when the propagation occurs in the layer planes
the polarization ez is forbidden for the hhn → εm
transitions. This may have some influence on the lasing
action in quantum-well systems since, in this case, the
light propagates perpendicularly to the growth axis.
However, absorption (or spontaneous emission) studies
in this geometry should show a strong dependence of the
spectra on the exciting light polarization for the
hhn → εm transitions. An example of such a
dependence is presented in Figure 1.

2.1.3.2. Selection Rules on the Envelope Function Quantum

Numbers: Evaluation of hfi |ff i
We need to evaluate hfi |ff i. This is equal to Fig. 1. Polarization dependence of the photoluminescence excitation
D E D ED E spectrum in a In1−x Gax As–GaAs superlattice at T = 77 K. In this ex-
fi |ff = χn |χm k⊥ |k0⊥
(h) (e) periment, the excitation spectrum mimics the absorption coefficient. For
Z an electromagnetic wave propagating in the layer plane (q⊥ez ) two ab-
∗(h) (e) 1 sorption peaks are seen in the er ⊥ez polarization while only one shows up
= dz χn (z)χm (z)
S in the er k ez polarization. This is in agreement with the predictions of the
Z h i previous table. Moreover, when q k ez and thus er ⊥ ez two peaks are
× d 2 r⊥ exp i(k0⊥ − k⊥ ) · r⊥ (45) observed which coincide in energy with those seen in the er ⊥ ez , q ⊥ ez
polarization.
Thus, the first selection rule is the conservation of the in-plane
wave vector in the optical transitions:
k0⊥ − k⊥ = 0 (46) Type-I systems are (GaAs–GaAlAs; GaInAs–InP; GaInAs–
It arises from the translational invariance of the heterostructure AlInAs · · ·. These heterostructures are such that conduction and
Hamiltonian in the layer plane. valence electrons are essentially confined in the same layers. For
The evaluation of the selection rules on the subband indexes is a type-I symmetrical quantum well, the envelope functions have a
(h) (e)
obtained by calculating hχn |χm i. At this point, one should make definite parity with respect to the center of the well. The overlap
a distinction between the type-I and type-II systems. integral hχn |χm i is thus nonzero only if n + m is even.
8 RIERA ET AL.

In addition, if the valence and conduction quantum wells are

rectangular and infinitely deep, only the transitions which fulfill
n = m are allowed. In the GaAs–GaAlAs rectangular wells, the
transitions 1n = n − m = 0 are much stronger than those which
correspond to 1n even (6= 0). The transition hh3 → ε1 was iden-
tified in absorption [13] and photoluminescence excitation spec-
troscopy [14]. Other GaAs–GaAlAs wells which are not rectangu-
lar, e.g., the pseudo-parabolic wells [15] or separate confinement
heterostructures [16], were optically studied. The former struc-
tures are obtained by imposing a quadratic variation of the con-
duction and valence edges by beam chopping during the epitaxial
growth, whereas the latter structures consist of embedding a nar-
row GaAs–Ga1−x1 Al1 As quantum well into Ga1−x2 Alx2 As barri-
ers with x2 > x1 . Both kinds of heterostructures permitted the ob-
servation of more optical transitions than those found in the plain
GaAs–GaAlAs rectangular quantum wells. This led to a significant
reappraisal of the band discontinuities in the GaAs–GaAlAs sys-
tem (for reviews see [17, 18]). Fig. 2. Conduction and valence-band edge profiles in a type-II quantum
The selection rule n + m even is in fact very strong and only well.
very asymmetrical wells can lead to transitions with n + m odd.
Investigations of intentionally designed asymmetrical wells might 
yield a better determination of the band offsets, since more optical X h̄2 k2⊥
×2 δεA + ε1 + hh1 +
transitions are allowed. 2
k⊥
Type-II systems are (InAs–GaSb; InP–Al 0:48 In0:52 As). In these ! 
structures, the electrons and the holes are spatially separated. Let 1 1
us consider the case of a type-II quantum well, whose band pro- × ∗ + ∗ − h̄ω (48)
me M1
files are shown in Figure 2. We denote the well-acting (barrier-
acting) materials for electrons by A(B) and we denote their re- By performing the summation upon k⊥ , we finally obtain
spective bandgaps by εA , εB . The A material is a barrier for the
holes and since we are considering a quantum-well structure, the πe2 εP m∗e M1∗ D (h)  (e) E2
αhh1 →ε1 (ω) = ∗ ∗  χ1 χ1 
valence spectrum is continuous and twice degenerate. To simplify ncm0 ω`h̄2 me + M1
the discussion, we later consider only photon energies such that ×Y(h̄ω − εA − ε1 − hh1 ) (49)
h̄ω < εA and we assume that ε1 < 1ν . For a given bound state εn
of the conduction band, a transition should involve only one of the In Eqs. (48) and (49), population effects for both initial and
(e)
two degenerate valence states: if χm is even in z, the initial state final states are neglected (undoped quantum wells), εA is the
should also be even in z to participate to the optical transitions. It bandgap of the bulk well-acting material and Y(x) is the step func-
tion. For a lh1 → ε1 transition, a similar expression is obtained
is also clear that the overlap of the conduction and valence enve-
with M1∗ , hh1 replaced by m∗1 , lh1 , and εP replaced by εP /3 (see
lope functions is only due to the exponential tails of the conduction
Table II).
and valence states in the layers B and A, respectively: in the limit
The absorption coefficient has a staircase-like shape at the on-
of large valence and conduction band discontinuities, the overlap
(h) (e) (h) (e) set of the absorption (see Fig. 3), as expected from the joint den-
integrals hχn |χm i tend to zero. In addition, hχn |χm i should sity of states of the two two-dimensional subbands. For hhn → εm
increase with increasing n and m: the tails of the conduction and transitions, n, m 6= 1, the magnitude of the step involves the square
valence envelopes in their respective barriers become more and (h) (e)
of hχn |χm i which is smaller than one and also the in-plane ef-
more important. Notice that the opposite is true in type-I systems.
fective mass of the nth heavy-hole subband in the vicinity of the
zone center, i.e., Mn∗ replaces M1∗ in Eq. (49). We saw previously
2.1.3.3. Order of Magnitude of the Absorption Coefficient that some of the Mn∗ s may be negative (i.e., may have electron-
Comparison between Type-I and Type-II Systems like behavior near k⊥ = 0). This, in turn, affects the magnitude
of αhhn →εm . Experimentally, it is often found that the edges of
As an example, let us consider the optical transitions between higher lying heavy-hole-electron transitions are less marked than
the ground heavy-hole and electron subbands of a rectangular the hh1 → ε1 one, which is consistent with an increased broaden-
type-I quantum well (ex polarization). In Eq. (20), the quantity ing of the excited subband states.
|er · pif |2 /m0 is equal to However, the most significant effect in Eq. (49) is the blueshift
of the quantum-well fundamental absorption edge with respect to
1 ε D  E
 (e) 2 that of the bulk A material. Photons start to be absorbed at an en-
|er · pif |2 = P  χ(h) χ1  (47)
m0 4 1
ergy ε1 +hh1 , above εA . This blueshift can be tuned by varying the
quantum-well thickness. For instance, bulk GaAs starts absorbing
and the absorption coefficient αhh1 → ε1 (ω) is equal to
in the infrared part of the spectrum (εA = 1:5192 eV at low tem-
ε D (h)  (e) E2 perature). Narrow GaAs–Ga1−x Alx As quantum wells (L ≈ 30 Å)
αhh1 →ε1 (ω) = γ P  χ1 χ1  can be designed to start absorbing light only in the red part of the
4
 OPTICAL PROPERTIES IN NANOSTRUCTURES 9

where m∗ν is the heavy-hole mass along the z-direction. Note that
in principle m∗ν should be set equal the M1∗ in Eq. (52) since each of
the B layers is a bulk material. We keep the m∗ν and M1∗ notations
to be able to make a comparison with the absorption coefficient in
type-I structures later on.
We can now introduce functions χe (z) and χ0 (z) which are
even and odd with respect to z,
(h) 1 h (h) (h)
i
χe (z) = √ χr (z) + χl (z) (53)
2
(h) 1 h (h) (h)
i
χ0 (z) = √ χr (z) − χl (z) (54)
2
Fig. 3. Staircase lineshape of the absorption coefficient of a type-I quan- We are interested in calculating the absorption coefficient at the
tum well.
onset of the absorption, i.e., that which is associated with valence
(e)
→ ε1 transitions. The envelope function χ1 (z) of the electrons
in the ε1 subband can be written thus,
spectrum (εA + ε1 + hh1 ≈ 1:7275 eV if x = 0:5). The blueshift of
(e) L
the fundamental absorption edge was observed in a number of het- χ1 (z) = Ac cos(kc z) |z| ≤ (55)
 2
erostructures; for example, in GaAs–GaAlAs [13]; GaInAs–InP !
[19]; GaInAs–AlInAs [20]; GaSb–AlSb [21]; GaInA–GaAs [22]. L  L
χ1 (z) = Bc exp−κc z −
(e)
z> (56)
Finally, let us calculate the magnitude of αhh1 →ε1 in the case 2 2
of a single quantum well embedded in a structure of total thick-
ness ` = 1 µm. For h̄ω = 1:6 eV, m∗e = 0:067m0 , M1∗  m∗e , (e)
χ1 (−z) = χ1 (z)
(e)
(57)
(h) (e)
n = 3:6, εP = 23 eV, and taking hχ1 |χ1 i ≈ 1, we obtain (h) (e) (h) (e)
The overlap integral hχ0 |χ1 i vanishes while hχe |χ1 i is
αhh1 →ε1 ∼ −1
= 60 cm . Such a faint absorption can hardly be mea- equal to
sured in practice, for it leads to an attenuation of the light beam D  E r
(h)  (e) 2 kν
intensity of only 1 − exp(−αhh1 →ε1 `) ≈ αhh1 →ε1 ` = 0:6%. χe χ1 = 2 Bc (58)
Therefore, one has to use multiple quantum wells to enhance the
` kν + κ2c
2

absorption. These structures are obtained by separating N iden- The coefficient Bc can easily be related to the probability Pb (ε1 )
tical wells with N − 1 thick barriers, resulting in an absorption of finding the electron in the B layers, while in the ε1 states,
which is enhanced by a factor of N over that of a single well. The
Bc2
barriers have to be thick enough to prevent any tunnel coupling Pb (ε1 ) = (59)
between the wells, a condition which is more difficult to achieve κc
when higher lying transitions (n; m > 1) are involved. In practice The quantity B defined in Eq. (31), with m∗e replaced by m0 (since
N should be ≥10 to obtain a sizeable absorption. we are dealing with interband transitions), can now be calculated.
We stress again that the low absorption coefficient is due to the It is equal to
different spatial locations of the electromagnetic energy (delocal- Z kmax
P (ε )κc εP m∗e M1∗ S dkν k2ν
ized over `) and the electrons which actually absorb light (essen- B= b 1 (60)
tially localized within the quantum well). π 2 h̄2 m∗e + M1∗ 0 (κ2c + k2ν )2
For a type-II quantum well, the only difficulty consists in eval- where
uating the overlap integral between valence states (which are ex- 2m∗ν
k2max = (h̄ω − εA + 1ν − ε1 ) (61)
tended in z and are labeled by a wave vector kν ) and conduction h̄2
states (which are localized in the well). For that purpose, we use
and finally the absorption coefficient is found to be equal to
a rather crude approach: the valence barrier 1ν is assumed high " #
enough to be considered as impenetrable by the valence electrons. 2e2 εP Pb (ε1 ) m∗e M1∗ −x
In this case, two linearly independent valence envelope functions α(ω) = ∗ ∗ + arctan x
hh→ε1 ncm0 ωh̄2 ` me + M1 1 + x2
correspond to each kν . They are
 ! " # kmax
x= (62)
(h) 2  L  L κc
χr (z) = √ sin kν z − Y z− (50)
` 2 2 The onset of the absorption is located at
 ! " # h̄ω0 = εA − 1ν + ε1 (63)
−2 L L
(h) 
χl (z) = √ sin kν z +  Y −z − (51)
` 2 2 In the vicinity of this onset, αhh→ε1 behaves like (ω − ω0 )3/2
(Fig. 4) instead of displaying the staircase-like behavior character-
These functions are normalized to 1 over `/2(L  `) and for istic of type-I quantum wells. Thus, the indirect optical transitions
heavy holes we have in real space have the same smoothing effect on the absorption
edge as the indirect transitions in reciprocal space in the case of
h̄2 k2ν h̄2 k2⊥ bulk materials. The physical origin of the absorption (exponential
εν (kν ; k⊥ ) = −εA + 1ν − − (52)
2m∗ν 2M1∗ tail of the ε1 wave function outside its confining layer) is confirmed
10 RIERA ET AL.

The exciton picture emerges after lengthy many-body calcula-

tions [24]. All of these start by attempting to calculate an energy
difference between the ground state of a crystal, which consists of
N electrons occupying a filled valence band, and an excited state
of the crystal obtained by optically promoting an electron from the
aforesaid filled band to an empty conduction band. If there were
no electron–electron interactions, this energy difference 1ε would
simply be the sum of the bandgap εg (eventually including valence
and conduction confinement energies in the case of heterostruc-
tures) and the difference between the kinetic energies in the final
and excited states, i.e., for parabolic bands,
Fig. 4. Absorption lineshape of a type-II quantum well. !
h̄2 k2e −h̄2 k2ν
1ε = εg + − (65)
2m∗e 2m∗ν

by the presence of the proportionality factor Pb in Eq. (62). At The onset of the interband absorption would occur at h̄ωonset =
large ω, i.e., kmax  κc , α(ω)hh→ε1 levels off at the value, εg . The existence of lower excited states of the crystal are the ori-
gins of peaked structures below εg . They are obtained by rear-
πe2 εP m∗e M1∗ ranging the conduction and valence electronic densities (in a cou-
α∞
hh→ε = Pb (ε1 ) ∗ (64)
1
ncm0 ω`h̄2 me + M1∗ pled way). To the excited state whose energy with respect to the
ground state is 1ε (Eq. (65)), one associates valence and conduc-
which is just Pb (ε1 ) times the magnitude of the plateau value for tion charge distributions which are uniform in space (at the scale
absorption in type-I quantum wells (see Eq. (49)). This means that of the envelope functions) and which are spatially uncorrelated.
the use of multiple wells to enhance the single well absorption is By making wave packets of conduction states with various ke and
even more imperative in type-II than in type-I structures. Since wave packets of valence states with various kν , where ke and kν
Pb (ε1 ) is usually only a few percent, it appears that hundreds of are univocally correlated (kν = ke − K, K fixed), the potential
wells may be necessary to measure the absorption in type-II quan- energy which is gained is greater (by a decrease of the repulsive
tum wells. electron–electron interaction) than the kinetic energy which is lost
Finally, it may be added that light-hole → ε1 (lh → ε1 ) tran- (by spatially correlating the electrons).
sitions have the same threshold as heavy-hole → ε1 (hh → ε1 ) The following discussion is a survey of excitonic absorptions in
transitions (Eq. (63)). Thus, only the sum of the two contributions semiconductor heterostructures and of their associated selection
will actually be measured. In addition, transitions from the valence rules. This discussion does not pretend to be complete: although
levels to the excited conduction subbands εi , i > 1 will be more in- relatively well documented on the experimental side, the optical
tense than the hh → ε1 or lh → ε1 transitions. This arises from properties of excitons in semiconductor heterostructures have not
the smaller κc and the larger Pb (εi ) for the excited subbands. received enough attention on the theoretical side.
We shall adopt the same decoupling procedure for the exci-
tons as for the free particle states (Eqs. (22) and (40)–(42)). Thus,
2.2. Absorption: A Simplified Description of in this crude model, there exist two kinds of decoupled excitons:
Excitonic Effects those formed between heavy holes and electrons (heavy-hole ex-
A convenient picture of an exciton is an electron and a hole orbit- citons) and those formed between light holes and electrons (light-
ing around each other. This picture is quite difficult to derive from hole excitons). This decoupled exciton theory, first used by Miller
the first principles because an exciton is in fact a many-body effect. et al. [25] and Greene and Bajaj [26] in the calculation of exci-
The excitonic effects are seldom negligible in practice. They ton binding energies, neglects off-diagonal coupling terms in the
most clearly manifest themselves with the appearance of sharp valence band Hamiltonian. For our purpose, it proves to be very
peaks in the interband absorption spectra whereas, as discussed convenient, for it lightens the already heavy algebra involved in
in the previous paragraph, noninteracting electron models at most the derivation of the excitonic absorption.
only predict staircase-like structures starting at h̄ωonset . In the ex- Let us consider hh1 → ε1 excitonic transitions in a type-I single
citon (or electron-hole pairs) picture, the pair energies larger than quantum well. Other situations may be treated with minor modi-
h̄ωonset correspond to dissociated pairs. It is often forgotten that fications. The ground state of our quantum well consists of N va-
the electron-hole attraction also alters the optical response of the lence electrons occupying the hh1 subband and an empty ε1 con-
pair continuum [23]. This alteration is so deep that, in bulk mate- duction subband. The dispersion relations of hh1 and ε1 are taken
rials for instance, the calculated one-electron absorption is never as parabolic upon the in-plane wave vector k⊥ (see Eqs. (40) and
correct. Even deep in the continuum, where the large electron- (42)), and the one-electron wave functions are, respectively,
hole relative velocity makes the noninteracting particle model at (h) (h) 1
its best, it falls short of accounting for the correct absorption co- ψk (r) = uhh (r)χ1 (z) √ exp(ik⊥ν · r⊥ ) (66)
⊥ν
S
efficient, i.e., the coefficient calculated by including the electron-
(e) (e) 1
hole interaction (although approaching it asymptotically from be- ψk (r) = ue (r)χ1 (z) √ exp(ik⊥e · r⊥ ) (67)
⊥e
low [23]). We see later how this enhancement can be traced back S
algebraically. Physically, it arises from the long range nature of where uhh and ue are the heavy-hole (i.e., uhh ≡ u3/2;±3/2 ) and
the Coulombic attraction between the electron and the hole, even electron (i.e., ue ≡ iS ↑ or iS ↓) periodic parts of the Bloch
when they do not form a bound state. function at the zone center of the host layers. It is clear that the
 OPTICAL PROPERTIES IN NANOSTRUCTURES 11
r !
quadratic dispersion law for hh1 is unable to give rise to a filled e 2h̄π
× − er · p̂n
band. It is understood that at large enough k⊥ν this dispersion flat- m0 ωV
tens so that ultimately the hh1 energy will be a periodic function h i Y
(e) (h)
upon k⊥ν with a period given by the in-plane basis vectors of a suit- × Pψk (re ) ψk (rn ) (73)
⊥e ⊥n
able Brillouin zone. Actually, the knowledge of the whole disper- k⊥n 6=k⊥ν
sion of the hh1 subband is not needed, for we are only interested Fortunately, this complicated expression simplifies into [23, 24,
in Wannier excitons which are weakly bound (with respect to the 27],
bandgaps). Their wave functions are built from hh1 and ε1 states r
 X 
which involve k⊥e , k⊥ν values much smaller than the in-plane size b n e 2h̄π
ψi | Her |ψf = − er
of the Brillouin zone. We assume that for these states the quadratic m0 ωV
n
laws are valid. Z
(h)∗ (e)
× d 3 rψk (r)p̂ψk (r) (74)
⊥ν ⊥e
2.2.1. Absorption in the Absence of Coulombic Interactions, due to D  E D   E
Equivalence with the One-electron Model (h)  (e) (h)   (h)
ψk
⊥n
ψk⊥e = ψk⊥n p̂ψk⊥n = 0 (75)
b0 (see Eq. (4)), the exact initial
In the absence of e2 /rij terms in H
b
and final states of H0 are Slater determinants [23, 24, 27]. The In the case N = 2, Eq. (73) can be checked by direct computa-
initial state is unique and corresponds to the N electrons which tions. Using Eqs. (71) and (72), one obtains eight terms; of these,
occupy the hh1 subband: six vanish owing to Eq. (75) and the last two, which are identical,
cancel the 1/2!. Finally, Wif is simply
ψi (r1 ; r2 ; : : : ; rN ) !2
 (h)  r D  
 ψk (r1 ) (h) (h) E
 ⊥1 ψk (r2 ) ψk (rN ) 
::: 2π e 2h̄π  (h)   (e) 2
⊥1 ⊥1 Wif = −  ψk⊥ν er · p̂ψk⊥e 
 (h) (h) (h)  h̄ m0 ωV
1  ψk⊥ν (r1 ) ψk (r2 ) : : : ψk (rN ) 
 (68)
= √  ⊥ν ⊥ν " #
N!  :: :: ::  h̄2 k2⊥e h̄2 k2⊥ν

 : : :  ×δ εA + ε1 + hh1 + + − h̄ω (76)
 (h) (h) (h)  2m∗e 2M1∗
ψk (r1 ) ψk (r2 ) : : : ψk (rN )
⊥N ⊥N ⊥N
A final state ψf is obtained by removing a hh1 electron with This expression is exactly the same as the one obtained in the one-
wave vector k⊥ν and sending it into the ε1 subband where it electron picture, which after all is not very surprising. If the dipolar
acquires an in-plane wave vector k⊥e . Its wave function is ob- matrix element in Eq. (76) were written in terms of both the dipole
(h) (e) matrix element between uhh and ue and the overlap integral of
tained from Eq. (68) by changing the line ψk⊥ν (ri ) into ψk⊥e (ri ),
the envelope functions, we could rederive the polarization selec-
i = 1; 2; : : : ; N.
tion rules on the one hand and the selection rules on the envelope
The Slater determinants can be rewritten in the more compact
functions quantum numbers on the other hand.
form,
In particular, we would find that only “vertical” transitions are
1 X Y (h)
optically allowed, i.e., k⊥ν = k⊥e .
hr1 ; r2 ; :::; rN |ψi i = √ (−1)P P ψk (rn ) (69)
⊥n
N! P k
There is however another way to derive this selection rule. Let
⊥n
us consider the translation operator τ̂⊥ which is such that
1 X (e)
hr1 ; r2 ; : : : ; rN |ψf i = √ (−1)P ψk (re )
N! P ⊥e τ̂⊥ f(r1 ; r2 ; : : : ; rN ) = f(r1 + a⊥ ; r2 + a⊥ ; : : : ; rN + a⊥ ) (77)
Y where a⊥ is an in-plane lattice b0
×
(h)
ψk (rn ) (70) P vector. The operators τ̂⊥ and H
⊥n commute as well as τ̂⊥ and n H ber
n . The initial state |ψ i is an
i
k⊥n 6=k⊥ν
eigenstate of τ̂⊥ with an eigenvalue which can always be written as
where P is the permutation operator which exchanges the particles exp(iKin ⊥ · a⊥ ).
between themselves. The case N = 2 leads to On the other hand, from Eq. (68) we can directly obtain
1 h (h) (h)
"
X
! #
hr1 ; r2 |ψi i = √ ψk (r1 )ψk (r2 )
2 ⊥1 ⊥2 τ̂⊥ ψi (r1 ; r2 ; : : : ; rN ) = exp i k⊥n · a⊥ ψi (r1 ; r2 ; : : : ; rN )
i n
(h) (h)
− ψk (r1 )ψk (r2 ) (71) (78)
⊥2 ⊥1
1 h Thus, X
(h) (e)
hr1 ; r2 |ψf i = √ ψk (r1 )ψk (r2 ) Kin
2 ⊥1 ⊥e ⊥ = k⊥n (79)
i n
(h) (e)
−ψk (r2 )ψk⊥e (r1 ) (72) However, for a filled band,
⊥1
X
The dipolar matrix element of the ψi → ψf transition can there- k⊥n = 0 i.e.; Kin
⊥ =0 (80)
fore be written as n
 X  Z
b n 1 XX P+P 0 The final state is also an eigenstate of τ̂⊥ with eigenvalue
ψi | Her |ψf = (−1) d 3 r1 · · · d 3 rN
N! n exp(iK⊥ · a⊥ ). By directly applying τ̂⊥ on |ψf i, we obtain
n PP 0
" #
Y (h)∗ τ̂⊥ ψi (r1 ; r2 ; : : : ; rN )
× P ψk⊥n (rn ) h i
k⊥n = exp i(k⊥e − k⊥ν ) · a⊥ ψi (r1 ; r2 ; : : : ; rN ) (81)
12 RIERA ET AL.

Thus, (N − 1) + 1 electrons or as a particular combination of electron-

K⊥ = k⊥e − k⊥ν (82) hole pairs. The exciton is a truly delocalized excitation. This can
be checked
P by calculating the average in-plane electronic density
From now on we denote the final state by ψk⊥e ;k⊥e −K⊥ . The se-
(1/N) m δ(r⊥ − r⊥m ) in the excited state Eq. (84) and by com-
lection rule on the envelope function quantum number means
paring it with that found in the ground state, Eq. (68). Both are
K⊥ = P 0. This is not surprising in view of the commutativity be-
ber
n and τ̂ . As the initial state was characterized by found equal to 1/S (if only the slow spatial variations, which oc-
tween n H ⊥
P cur at the scale of the envelope function, are retained). What may
in
K⊥ = 0 it could only be coupled by n H b n to an excited state
er give rise to localized states in the exciton are the internal degrees
with the same K⊥ values, resulting in the selection rule δK⊥ ;0 . In of freedom of the pairs. This is due to the attraction between the
short, we obtained electron and the hole (or equivalently to the diminished electron–
r !2 electron repulsion) in the excited state Eq. (84). The electron-hole
2π e 2h̄π D 

 E2 D  E
   (h)  (e) 2
Wif = −  uhh er · p̂ue   χ1 χ1  δK⊥ ;0 attraction, together with the restricted class of electron-hole pairs
h̄ m0 ωV that we retain in Eq. (84) (i.e., which all have the same total wave
 ! 
vector K⊥ ), can be described in a better way if we use the exciton
h̄2 k2⊥e 1 1
×δεA + ε1 + hh1 + + ∗ − h̄ω (83) amplitude,
2 m∗e M1
1 X
β(r⊥ ) = √ A(k⊥e ) exp(ik⊥e · r⊥ ) (85)
From such an expression, one could easily derive the staircase-like Sk
⊥e
absorption spectrum, Eq. (49)), which is characteristic of subband
→ subband interband transitions in type-I quantum wells. The physical interpretation of β(r⊥ ) is that |β(r⊥ )|2 is the areal
probability density of finding the electron and the hole separated
from each other by r⊥ , while having a total wave vector K⊥ . Equa-
2.2.2. Absorption in the Presence of tions (84) and (85) are the starting points of the exciton formal-
Electron–Electron Interactions ism, fully developed in, for instance, Knox’s text book [24]. For
In the presence of Coulombic terms, single Slater determinants are our purposes, it is sufficient to know that β(r⊥ ) (after consider-
no longer exact eigenstates of H b0 . However, they remain the best able manipulations and numerous assumptions) is the solution of
single determinant approximation of the ground state if the one- 
p2 2 Z Z dzr dz |χ(r) (zr )|2 |χ(h) (z )|2
 ⊥ −e
(h) h h
electron functions ψk (r) are solutions of the coupled integrod- q l l
⊥ν
ifferential Hartree–Fock. We keep this approximation for |ψi i. 2µ∗ κ 2
r⊥ + (zr − zh ) 2
For the final state, a single determinant is also not an exact solu- 
tion. If we keep a single determinant of the form given by Eq. (70),
+ εA + ε1 + hh1 β(r⊥ ) = εβ(r⊥ ) (86)
we obtain a Hartree–Fock solution which describes a situation
where a valence electron, uniformly spread over the area S, is re-
placed by a conduction electron whose charge density is also uni- where µ∗ is the exciton reduced effective mass and κ is the relative
formly spread over S. The removal of a valence electron hardly dielectric constant of the heterostructure. The energy difference
affects the self-consistent Hartree–Fock potential experienced by between the excited state Eq. (84) and the ground state is equal to
the remaining N − 1 valence electrons (the charge density is of
the order of (1/S)). Similarly the single conduction electron only h̄2 K⊥
2
1ε = εν + (87)
slightly modifies the self-consistent potential. This results in an ex- 2(me + M1∗ )
∗

cited energy which differs from the ground state energy by almost
where εν is one of the eigenvalues of Eq. (86). The ground state
(i.e., to the order of (1/S)) the same energy 1ε as calculated in
of Eq. (86) was extensively discussed previously. We know that a
Eq. (65). A lower lying excited state can be constructed by expand-
simple trial wave function of this ground state is
ing |ψf i on the set of the single determinants |ψk⊥e ;k⊥e −K⊥ i. Be- s !
cause H b0 commutes with τ̂⊥ we know that such an expansion can 2 r⊥
β1s (r⊥ ) = exp − ∗ (88)
be restricted to a summation over k⊥e , K⊥ being a good quantum πa∗2B2D
aB2D
number. Thus, even with the inclusion of electron–electron inter-
actions (within the Hartree–Fock approximation) we can write where a∗B2D is the effective Bohr radius of the quasi-two-di-
mensional exciton. The 1s exciton state lies at an energy lower than
hr1 ; r2 ; : : : ; rN |ψf i εg + ε1 + hh1 by a quantity εb , which is the exciton binding en-
X
= A(k⊥e )ψk⊥e ;k⊥e −K⊥ (r1 ; r2 ; : : : ; rN ) (84) ergy. Thus, the minimum energy separation between the ground
k⊥e state and the excited state (characterized by the wave vector K⊥ )
is
If we now reason in terms of electron-hole states instead of h̄2 K2⊥
(N − 1) + 1 electrons states, the wave function ψk⊥e ;k⊥e −K⊥ cor- 1εmin = εA + ε1 + hh1 − εb + (89)
2(m∗e + M1∗ )
responds to an electron-hole pair with a total wave vector k⊥e −
(k⊥e − K⊥ ) = K⊥ (since the hole wave vector is minus that of which can be lower than εA + ε1 + hh1 if K⊥ is small enough.
the missing electron). Thus, the physical meaning of Eq. (84) is There exists an infinite number of bound states to Eq. (86).
that we are looking for an excited state of the heterostructure, They correspond to βν (r⊥ ) functions, which decay to zero at
which consists of a wave packet of electron-hole pairs, which have large r⊥ . Unbound solutions also exist. They form a continuum
the same total wave vector K⊥ . This wave packet is the exci- and are characterized by βν (r⊥ ) exciton amplitudes, which extend
ton. The exciton can be seen either as a particular excitation of over the whole area S, and by energies εν > εA + ε1 + hh1 .
 OPTICAL PROPERTIES IN NANOSTRUCTURES 13

In the interacting electron-hole problem, these unbound solutions

are the counterparts of the plane wave solutions in the noninter-
acting electron-hole problem.
We now evaluate Wif and we obtain the selection rules for the
excitonic absorption from the ground state to the final state which
is characterized on the one hand by K⊥ and on the other hand by
a quantum number ν which labels the solutions of Eq. (86). This is
indicated by affixing a subscript ν to the A(k⊥e ) of Eq. (84):
*   + *   +
X  X X 
 
b  ψf =
n  
b ψk −K
n
ψi  H er  A(k⊥e ) ψi  H er  ⊥e ⊥ (90)
 
n k⊥e n

The matrix element appearing on the right-hand side of Eq. (90) Fig. 5. Theoretical (unbroadened) absorption spectrum of a purely two-
was already evaluated in the previous subsection. This leaves us dimensional semiconductor (solid line). R∗y;exc is the three-dimensional ex-
with citon binding energy. Only the 1s and 2s exciton bound states contributing
*   + r to the absorption were represented. The dashed line plateau corresponds
X  2h̄π D   ED  E
 b n  e   (h)  (e) to the absorption coefficient of dissociated and uncorrelated electron-hole
ψi  Her ψf = − uhh er · p̂ue χ1 χ1
  m0 ωV pairs.
n
X
×δK⊥ ;0 Aν (k⊥e ) (91)
k⊥e An important new fact brought by the electron-hole interac-
which, together with Eq. (85), leads to tion is the appearance of the term |βν (0)|2 in Eq. (92). It means
!2 that only excitons with a nonzero amplitude at r⊥ = 0 can absorb
r
2π e 2h̄π D E2 D  E
  (h)  (e) 2 light. Since the effective two-dimensional potential which appears
Wif = −  uhh |er · p̂|ue   χ1 χ1  in Eq. (86) is radial, the βν (r) functions can be classified according
h̄ m0 ωV
 2 to the eigenvalues of L̂ and L̂z , where L̂is the angular momentum.
  In addition, a radial quantum number n (discrete for the bound
×δK⊥ ;0 S βν (0) δ(εν − h̄ω) (92)
states, continuous for the extended states) labels the eigenstates.
where εν is the energy of the eigenstates |νi of Eq. (86). Again, if Thus, only the ns excitons (corresponding to L̂ = 0) can be opti-
we neglect the electron-hole interaction, the result of the nonin- cally created.
teracting electron model is recovered: in this case βν (r⊥ ) and εν The only situation where Eq. (86) admits exact solutions cor-
are given by (e)
responds to a purely two-dimensional case (i.e., |χ1 (z)|2 =
|χ1 (z)|2 = δ(z)). In this case, εb is equal to 4R∗y;exc , where
1 (h)
βν (r⊥ ) = √ exp(ik⊥ · r⊥ )
S R∗y;exc is the three-dimensional exciton Rydberg. The magnitude
h̄2 k2⊥ of |βns (0)|2 , and thus of the ns excitonic absorption, is propor-
εν =
2µ∗
+ ε1 + hh1 + εA (93) tional to (8/π)a∗2 3 ∗
B (2n − 1) where n = 1; 2; : : : and aB is the bulk
exciton Bohr radius [28]. This shows that the magnitude of the ns
and the staircase-like shape of the absorption is readily derived. exciton absorption decreases very rapidly with n. However, the lev-
In the case of excitonic absorption, the selection rule δK⊥ ;0 els are more dense. Both effects combine to lead to a finite absorp-
is very important since it considerably restricts the possibilities of tion when h̄ω approaches εA + ε1 + hh1 from below. The absorp-
creating (or annihilating) excitons. It expresses the delocalized na- tion associated with the dissociated electron-hole hole pairs is also
ture of the exciton, an entity which corresponds to a delocalized enhanced with respect to the noninteracting electron model (see
Bloch state of the excited crystal, despite the fact that the exci- Fig. 5). It approaches the one-electron result when h̄ω exceeds
ton amplitude βν (r⊥ ) can be a spatially localized function of r⊥ . εA + ε1 + hh1 by many εb .
The δK⊥ ;0 selection rule can also be associated with the conser- Let us rewrite the hh1 → ε1 excitonic absorption coefficient in
vation of the in-plane wave vector K⊥ which, being zero in the the quasi-two-dimensional case,
ground state, must also be zero in the excited state if the dipo-
lar approximation is used to describe the interaction between the 2π 2 e2 εP D (h)  (e) E2 X  2

α(ω) =  χ1 χ1  βν (0) δ(h̄ω − εν ) (94)
electromagnetic wave and the carriers. Going beyond the dipolar hh1 →ε1 ncm0 ω` ν
approximation would give rise to the selection rule δK⊥ ;κ⊥ , where
where εP ; `; n · · · have the same meaning as in Subsection 2.1.
κ⊥ is the in-plane projection of the photon wave vector.
First of all, we focus our attention on the quantum-well thickness
As seen from Eq. (92), the excitonic absorption fulfills the same
(L) dependence of the magnitude of the 1s exciton peak. To ac-
polarization selection rules as the band-to-band absorption model
count for the broadening mechanisms, we replace the delta func-
which neglects the electron–electron interaction. This is because
tions in Eq. (94) by a Gaussian function (which sometimes fits the
the electron–electron interaction has more symmetry (spherical
observed line-shapes better than a Lorentzian one) and we use
symmetry) than the heterostructure potential. Looking at the pre-
Eq. (88) to evaluate β1s (0). For a single quantum well, we obtain
vious table we notice that the heavy-hole excitonic transitions
the peak value,
are forbidden (within our model) when the electromagnetic wave
propagates in the layer plane and when its electric field vector is peak 4π 2 e2 εP D (h)  (e) E2 1
polarized along the growth z-axis. `αhh →ε (ω) =  χ χ1  √ (95)
1 1 ncm0 ωλ2 1 0 2π
14 RIERA ET AL.

2.3. Photoluminescence
2.3.1. Introduction
The optical spontaneous emission, for example, photolumines-
cence, cannot be predicted when the electromagnetic field is de-
scribed classically. However, it is possible to relate its intensity ex-
actly to that of absorption by means of the Einstein coefficients [8,
9, 31]. Here, we are interested in presenting the theoretical con-
siderations on the line-shape of some recombination processes in
semiconductor heterostructures rather than calculating the photo-
luminescence intensities. When possible, these considerations will
be illustrated by examples. In contrast with a widespread belief,
the information one can extract from the photoluminescence con-
cerning the energy levels of a given heterostructure is often very
scarce and depends to a large extent on an a priori knowledge
Fig. 6. Measured absorption coefficient of a 76–33 Å GaAs multiple of that heterostructure. It is true that photoluminescence exper-
quantum-well sample at T = 2 K. Between the two main absorption peaks iments are much easier to perform than absorption experiments.
corresponding to excitonic ε1 − hh1 , ε1 − lh1 transitions, respectively, one It is also true that they are much more complicated to interpret.
sees a weaker absorption due to either the ε1 − hh1 2s excitons or to the Absorption and photoluminescence seem to be, at first sight,
ε1 − hh1 exciton continuum. symmetrical processes. On the one hand, a photon is absorbed pro-
moting an electron from the level |ii to the level |f i and photons
√ which are not absorbed are detected. On the other hand, the emis-
where 20 2 is the full width at e−1 amplitude of the Gaussian sion involves the transition |ii → |f i with εi > εf and the pho-
peak. For a GaA–GaAlAs quantum well, a representative figure tons, which are emitted at the energy εi −εf , can be observed. Such
peak
for `αhh →ε is obtained with n = 3:6, h̄ω = 1:6 eV, εP = 23 eV, a process is possible only if the system was initially excited in the
1 1
√ state |ii. Thus, it is not at equilibrium and one of the ways in which
a∗B2D = 70 Å, 0 2π = 3 meV, and hχ1 |χ1 i ≈ 1. We obtain
(h) (e)
this can be reached is by optical emission. The radiative channel is
peak
`αhh →ε ≈ 0:19. This is ≈32 times larger than the band-to-band in competition with the nonradiative relaxation processes (phonon
1 1
plateau value obtained in the noninteracting electron model of the emission, capture by deep centers, Auger effect etc. ...) which send
hh1 → ε1 absorption. Precise absorption measurements of the the excited carriers to lower states from which they can emit pho-
continuum edge are however difficult since it is often (energy) lo- tons or relax nonradiatively, etc. ... . It appears therefore that lumi-
cated in the low energy tail of the 1s ε1 − lh1 exciton. nescence is different from absorption because, instead of having a
When the quantum-well thickness varies, two factors affect the 100% efficiency (as in the case of absorption where one absorbed
peak value of the 1s excitonic absorption. photon creates one electron-hole pair), spontaneous emission is
(i) a∗B2D decreases with decreasing L. This strengthens the only one of the mechanisms which might occur. This is a question
excitonic absorption and witnesses an increase in the of the lifetimes of the level |ii with respect to the radiative and
two-dimensionality of the exciton. Correlatively, the nonradiative relaxations. To illustrate these notes, let us consider
energy distance between the 1s exciton peak and the the four-level system shown in Figure 7. The system is pumped
continuum increases. by |1i → |4i transitions and we only consider the luminescence
(ii) The broadening parameter 0 also increases. As shown by signal of levels |3i and |2i at energies ε3 − ε1 and ε2 − ε1 , re-
Weisbuch et al. [29], the layer thickness fluctuations spectively. The parameters τij and Ti1 are time constants which
make 0 to vary like L−3 . characterize interexcited levels nonradiative channels and radia-
tive channels from level |ii to level |1i, respectively. The effect of
Thus, depending on which mechanism prevails, the peak value
the final state population will be neglected. As level |3i is assumed
of the excitonic absorption can increase or decrease. Masumoto
to be populated nonradiatively from |4i, it suffices, since levels |1i
et al. [30] reported a complete study of the thickness dependence
and |4i are of no interest, to write the rate equations,
of the excitonic absorption in GaAs–GaAlAs multiple quantum
wells. dn3 n n n
The exciton continuum absorption (h̄ω > εA + ε1 + hh1 ) is = 0 − 3 − 3 (96)
dt τp τ32 T31
also enhanced by the electron-hole attraction compared with that
dn2 n n n
calculated with a noninteracting electron model. Note however = + 3 − 2 − 2 (97)
that this enhancement cannot be explicitly calculated dt τ32 τ21 T21
because the solutions of the Schrödinger equation for the quasi- In the steady state, dn/dt = 0 and therefore,
two-dimensional exciton Eq. (86) are not known analytically. Ex-
perimentally, a small hump is often visible in high quality GaAs n0 T31
n3 = (98)
quantum wells. It is located in energy between the 1s hh1 − ε1 and τp 1 + T31 τ32
1s lh1 − ε1 exciton peaks (Fig. 6) and can be attributed either to n T31 T21 1
the 2s hh1 − ε1 broadened exciton absorption [25] or to the onset n2 = 0 (99)
τp 1 + T31 /τ32 τ32 1 + T21 /τ21
of the hh1 − ε1 exciton continuum [26]. These two energies are so
close (1 − 2 meV) that the small hump probably arises from both The luminescence signals at energies h̄ω31 = ε3 − ε1 and h̄ω21 =
of them. ε2 −ε1 are, respectively, proportional to n3 and n2 . It is easy to see
 OPTICAL PROPERTIES IN NANOSTRUCTURES 15

Fig. 7.

from Eqs. (98) and (99) that, if τ32  T31 , almost all the popula-
tion of level |3i is used to populate |2i and does not participate in
the emission which occurs at h̄ω31 . Even if the density of states of
level |3i is more important (or, in our “atomic” scheme, if its de- Fig. 8. Calculated time evolution of the n2 and n3 populations in a three
generacy is larger) than that of |2i, the spontaneous emission from level system. The level |3i was assumed to be populated by a δ(t) pulse.
|3i can be considerably smaller than from |2i. This is to be com-
pared with an absorption experiment, which would give a larger
absorption at energy h̄ω31 than at energy h̄ω21 if the oscillator
The population n2 (t) first increases with time, (as long as the filling
strengths of the two transitions are the same and the degeneracy
of |2i from state |3i is larger than the radiative and nonradiative
of |3i is larger than that of |2i. In fact, in bulk semiconductors at
depopulations), and then decreases (see Fig. 8). It is very impor-
low temperature, the luminescence very often involves impurity
tant to notice that the luminescence decay time at energy ε2 − ε1
levels (impurity-to-band, donor–acceptor transitions ...) while the
density of states of the impurity levels is much smaller (by several only yields the total lifetime T2 of the excited level. This involves
orders of magnitude) than the density of states of the free states in both the radiative and nonradiative lifetimes. The time-resolved
the bands. These impurity levels, which are so prevalent in the pho- luminescence experiments thus need to be cautiously interpreted
toluminescence, are almost invisible in absorption spectra. While if conclusions are to be drawn on the radiative lifetimes of the ex-
the absorption is sensitive to the density of states, the photolumi- cited levels. In addition, level |3i is never pumped at a rate n0 δ(t).
nescence gives information on this quantity, which is completely The actual function g(t) displays an apparatus-limited time width,
distorted by the relaxation effects. This occurs to such an extent as does the detection system which counts the photons emitted at
that it favors the lowest lying excited states of the materials, even the energy ε2 − ε1 or ε3 − ε1 . This implies that deconvolutions of
if their density of states is very small. the experimental signal n2 (t) are required to distinguish between
Let us now consider the transient phenomena (i.e., time- the “true” n2 (t) and the time dependences caused by instrument
resolved photoluminescence) and let us assume that the pump limitations.
rate of level |3i is described by a function g(t) which is arbi-
trary, except that g(t) vanishes if t < 0. If we denote the quantity
−1 −1 −1 −1 2.3.1.1. Excitation Spectroscopy
T21 + τ21 − T31 − τ32 by 23 , we obtain the time dependences,
 !Z ! A technique closely associated with photoluminescence is the pho-
1  t t
0 t0 toluminescence excitation spectroscopy. The detection spectrom-
n2 (t) = exp − g(t ) exp dt 0
τ32 23 T3 −∞ T3 eter is set at some energy inside the emitted photoluminescence
band (ε2 − ε1 in the atomic model of Fig. 9) and the energy of
!Z ! 
t t
0 t0 the exciting light is scanned. As a result, the various excited levels
− exp − g(t ) exp dt 0  (100) of the solids |2i; |3i : : : are populated at rates g2 ; g3 ; : : :, respec-
T2 −∞ T2
tively. These rates are proportional to the absorption coefficients
!Z !
t t
0 t0 α(ε2 ); α(ε3 ) : : : . Once populated, the excited levels relax either
n3 (t) = exp − g(t ) exp dt 0 (101) radiatively or nonradiatively towards lower energies. In particular,
T3 −∞ T3
a fraction of their populations ends up in the lowest lying excited
where state |2i. In the steady state and for the situation depicted in Fig-
T3−1 = T31
−1 −1
+ τ32 T2−1 = T21
−1 −1
+ τ21 (102) ure 9, one obtains
g2
In the case where the pump state is the delta spike n0 δ(t), we ob- n2 = (105)
−1 −1
tain T21 + τ21
 ! !
n0 Y(t)  t t  for an excitation energy equal to ε2 − ε1 ,
n2 (t) = exp − − exp − (103)
τ32 23 T3 T2 g3
n2 = (106)
! (T21 + τ21 )(1 + τ32 T3−1 )
−1 −1
t
n3 (t) = n0 Y(t) exp − (104)
T3 for an excitation energy equal to ε3 − ε1 , and
16 RIERA ET AL.

Fig. 9.

g4
n2 = (107)
(T −1 +τ−1 )(1+τ32 T −1 )(1+τ43 T −1 )
21 21 3 4
g5
n2 = −1 −1 (108)
(T +τ )(1+τ32 T −1 )(1+τ43 T −1 )(1+τ54 T −1 )
21 21 3 4 5
for excitation energies equal to ε4 − ε1 and ε5 − ε1 , respectively.
In Eqs. (105)–(108), τij is now the relaxation time associated with
the nonradiative path i → j and Ti is the decay time due to all the
mechanisms which empty |ii, apart from those which contribute to
the population of |i − 1i.
The previous model is too crude to be applied as such to semi-
conductor heterostructures but it helps us to understand the dif-
ferent kinds of information given by the photoluminescence ex-
citation spectra. It should be noticed that the photoluminescence
signal at energy ε2 − ε1 , which is proportional to n2 has an ampli-
tude which is governed by two competing factors:
(i) the absorption coefficients of the exciting light of energy
ε2 − ε1 via the rates,
(ii) the relative orders of magnitude of the τij s and the Ti s.
Fig. 10. Comparison between absorption and luminescence excitation
In the limiting case where all the τij s are much shorter than spectra in a five level system. Upper figure: absorption coefficient; mid-
the Ti s, the excitation spectrum gives information equivalent to dle and lower figures: luminescence excitation spectra at ε2 − ε1 . Case (a)
that provided by absorption (see Fig. 10). The advantage of the τij  Ti : the excitation spectrum mimics the absorption spectrum; case
former technique over the latter is its sensitivity. Like photolumi- (b) resonant relaxation τ32  T3 and τij = Ti , i 6= 3.
nescence, excitation spectroscopy does not require thick samples,
e.g., it can easily be performed on a single GaAs quantum well,
whereas the absorption experiments, as shown previously, require trum shows a peak at ε3 − ε1 , which is not found in the absorp-
multiple quantum wells. tion spectrum (see Fig. 10). This peak arises because the levels |4i
The excitation spectroscopy is moreover selective, allowing the and |5i, once populated, contribute little to the population of |3i
physical origin of the photoluminescence signal in the heterostruc- and |2i, leading to a reduced photoluminescence signal (at energy
ture to be traced back. Let us suppose that the photoluminescence
ε2 − ε1 ) over what could have been inferred uniquely from the
band contains several peaks. By setting the detection wavelength at
rates g4 ; g5 .
each of these peaks, one obtains excitation spectra whose shape is
Thus, excitation spectra may or may not be equivalent to ab-
characteristic of the various excited levels which, after relaxation,
sorption spectra. There is no significant luminescence band in-
ultimately give rise to these different photoluminescence peaks.
volving the two-dimensional electron gas confined near the GaAs–
If the excitation spectra are independent of the detection wave-
GaAlAs heterointerface. This is due to the very quick relaxation
length, one may safely assume that the whole luminescence band
of holes far away in the acceptor depletion length. Strong lumi-
has a single physical origin. If, on the other hand, the excitation
nescence peaks are associated with the radiative recombination of
spectra are markedly different, one can conclude that the various
the bulk GaAs buffer layer. The resonant photon energies, which
photoluminescence lines have different physical origins. As is of-
produce structures in the excitation spectrum, are
ten the case in heterostructures, these differences arise from the
different locations of the recombining levels. h̄ω = εg + nh̄ωLO (1 + m∗e /m∗lh ) (109)
In another limiting case, the excitation spectrum is very dif- h̄ω = εg + nh̄ωLO (1 + m∗e /m∗hh ) (110)
ferent from the absorption spectrum. This happens when, for in-
stance, the τij s are all of the same order of magnitude or larger The other limiting regime, where the excitation spectrum roughly
than the Ti ’s but one, say τ32 , which is much shorter than T3 (res- mimics the absorption spectrum, is exemplified in the GaAs–
onant relaxation from |3i to |2i). In this case, the excitation spec- GaAlAs quantum wells and superlattices (see Ref. [29]). In the
 OPTICAL PROPERTIES IN NANOSTRUCTURES 17
n h io−1
vast majority of these heterostructures, any trace of peaks or struc- 1 − f(εf ) = 1 + exp −βh (εf − µh )
tures in the excitation spectra can be fairly associated with an ab-
sorption process. The lack of structures associated with optical βh = (kB Th )−1 (115)
phonons, which are so prevalent in bulk GaAs (see earlier), has
not been clearly explained to our knowledge. It calls either for a This formulation assumes that the photocreated conduction (va-
strongly diminished electron-phonon coupling in quantum wells or lence) electrons thermalize among themselves much quicker than
for a very fast carrier–carrier interaction, which would overshadow they recombine and that they acquire a temperature Te (Th ). This
the resonant emission. temperature is often different from the lattice temperature T and
depends on the density of injected carriers, external perturbations
(e.g., heating by an in-plane electric field [32]). The reader should
2.3.2. Quantum-well Luminescence (Steady State) be aware that a great deal of the photoluminescence spectra in-
Through the Einstein relationships, the spontaneous emission can terpretations are based on the assumptions of thermalized carri-
be related to the absorption. However, since we are only inter- ers. This assumption is very convenient but is seldom justified by
ested in the line-shape of the luminescence spectrum and we are detailed calculations. With this reservation in mind, let us discuss
not interested in its absolute magnitude, we discard all the pro- some emission mechanisms in more detail.
portionality constants and we treat the spontaneous emission and
absorption on the same footing. We shall denote by f(εi ) and f(εf )
the stationary distribution functions of the initial (|ii) and the fi- 2.3.2.1. Band-to-band Emission
nal (|f i) states. Thus, the luminescence signal `(ω) associated with
For an emitted light which propagates along the z-axis, we know
that particular |ii → |f i transition will be proportional to
(see Table II) that the hhn ↔ εm or lhn ↔ εm transitions are al-
h i
`i→f (ω) ∝ Wif f(εi ) 1 − f(εf ) (111) lowed provided that n + m is even and that the heterostructure po-
tential is even in z. For light emitted in the layer plane, the ez po-
where Wif is the transition probability per unit time that the carrier larization is forbidden for the εm ↔ hhn transitions and is allowed
undergoes a transition |ii → |f i due to the effect of the (dipolar) for the lhn ↔ εm transitions. We retain the quadratic dispersions,
coupling Hamiltonian between the light and the carrier: (Eqs. (40)–(42)) for the various subbands and we compute the lu-
r !2 minescence line-shape of the subband-to-subband transitions. For
2π e 2h̄π D E2
 the εm ↔ hhn recombination, we obtain
Wif = −  i|er · p̂|f  δ(εi − εf − h̄ω) (112)
h̄ m0 ωV
`m→n (ω)
The optical transition |ii → |f i is symmetrical to the absorption D E2 D E
 (h) (e) 2
one (|ii → |f i) and therefore follows the same selection rule. As ∝ uhh |er · p̂|ue  χn |χm 
for the absorption, the dipole matrix element can be split into two Z !
h̄2 k2⊥
parts (see Eqs. (23)–(25)), × d 2 k⊥ δ εg + εm + hhn + − h̄ω
2µ∗nm
D E D EZ
i|er · p̂|f = ui |er · p̂|uf d 3 rfi∗ (r)ff (r) (113) 1
× " !#
h̄2 k2⊥
The first matrix element on the right-hand side of Eq. (113) gives 1 + exp βe εm + − µe
2m∗e
the selection rule on the polarization of the emitted light (in the
layer plane of the layers or along the growth axis, see Table II) 1
× " !# (116)
and the overlap integral selects the subband indexes which govern h̄2 k2⊥
interband recombination. 1 + exp −βh −εg − hhn − − µh
2Mn∗
In a quantum well with band edge profiles which band edges
are symmetric in z, we find that, as with absorption, the only op-
tical allowed transitions between subbands preserve the parity of where µ∗nm is the reduced electron-hole mass (µ∗nm −1
the z-dependent envelope functions. Thus, using the same label- = m∗e −1 + Mn−1 ).
ing as in Subsections 2.1 and 2.2, n + m should be even. In addi- Let us introduce the chemical potentials ηe , ηh of the elec-
tion, if the broadening is neglected, the radiative recombination trons and the holes measured from εm and hhn , respectively, (see
occurs with conservation of the in-plane wave vector of the carrier Fig. 11):
(i) (f)
(k⊥ = k⊥ ). Similarly, in superlattices the optical transitions take (e) (h)
(i) (f) ηm = µe − εm ηn = −µh − hhn − εg (117)
place vertically in the superlattice Brillouin zone, (k⊥ = k⊥ ),
(i) (f)
qz = qz . They are related to the steady-state areal concentrations of elec-
The explicit form of the stationary distribution functions f(εi ) trons (nm ) and holes (pn ) in the εm and the hhn subbands by
and f(εf ) is often not known. Thus, they are (usually) taken as  !
being Fermi–Dirac distribution functions characterized by Te , µe π h̄ 2n
m
ηm = kB Te ln−1 + exp 
(e)
(118)
for conduction electrons and Th , µh for the valence electrons, kB Te m∗e
n h io−1  !
f(εi ) = 1 + exp βe (εi − µe ) π h̄2 pn
ηn = kB Th ln−1 + exp 
(h)
(119)
βe = (kB Te )−1 (114) kB Th Mn∗
18 RIERA ET AL.

The steplike onset reflects the quasi-two-dimensional

nature of the carrier motions. In practice, it is rounded
off by band-tailing, damping, etc. ... . However, the
characteristic feature is the exponential decay at large y
(i.e., ω), which may allow T ∗ to be deduced from the
luminescence spectrum. In addition, electrons and holes
are often assumed to be in thermal equilibrium with each
other (Te = Th = T ∗ ), which is likely for delocalized
carriers. When kB T ∗ becomes comparable to the energy
separation between two consecutive subbands of a given
band, the photoluminescence spectrum displays several
lines. For instance, in GaAs–GaAlAs quantum wells with
a GaAs layer thickness ≥100 Å, one observes at low
carrier injection and room temperature two
luminescence lines associated with the ε1 → hh1 and
ε1 → lh1 recombinations, respectively. By decreasing
T ∗ , the ε1 → lh1 feature disappears.
(ii) Degenerate electrons and holes
(e) (h)
(βe ηm > 3; (βh ηn ) > 3).
This situation corresponds to high-injected carrier
concentrations and/or low temperatures. The line-shape
(e) (h)
Fig. 11. Definition of the reduced electron (ηm ) or the hole (ηn ) is again very simple:
Fermi levels in terms of the conduction (µe ) and the valence (µh ) Fermi !
levels. (e) µnm
`(y) ∝ Y(y)Y ηm − ∗ y
me
!
The total steady-state concentrations n, p of the electrons and the (h) µnm
× Y ηn − y (125)
holes are, respectively: Mn∗
X X
n= nm p= pn (120) The photoluminescence signal versus ω is rectangular but the
m n edges are actually smoothed by the damping (low-energy side) and
Let us denote by y the excess emitted photon energy over εg + finite temperature effects (high-energy side). The high-energy cut-
εm + hhn : off corresponds to ymax where
" #
h̄ω = εg + εm + hhn + y (121) m∗e (e) Mn∗ (h)
ymax = Inf η η (126)
We then obtain µnm m µnm n
1
`m→n (y) ∝ " !# The previous limiting cases are illustrated in Figures 12 and 13
µnm (e) where plots of the photoluminescence line-shape versus energy are
1 + exp βe y ∗ − ηn
me calculated for band-to-band emission in a GaAs–Ga0:67 Al0:33 As
quantum well (thickness 55 Å) at different temperatures and for
1
× " !# Y(y) (122) different concentrations of injected carriers. Only the ε1 → hh1
µnm (h) and ε1 → lh1 recombinations were considered and equal car-
1 + exp βh y ∗ − ηn
Mn rier concentrations (n = p) and temperatures (Te = Th ) were as-
sumed in the calculations.
It is interesting to notice that this line-shape is obtained analyt- Band-to-band recombination takes places when excitonic ef-
ically whatever the degeneracies of the electron and holes gases fects can be discarded. This may be the case in undoped GaAs
(namely, whatever βe , βh , µe , µh ). This is a particular feature quantum wells at room temperature (see the discussion in Subsec-
of two-dimensional systems, which display a constant density of tion 2.3.2.2) but the situation is controversial [33, 34]. We show in
states if the dispersion relations are quadratic in k⊥ . Several limit- Figure 14 an example where the radiative recombination does not
ing cases can be considered and two are discussed in the following. involve excitons because their binding energies are insignificant.
(i) Nondegenerate electrons and holes In InAs–GaSb type-II superlattices at room temperature [35], the
(e) (h)
(ηn < 0; ηn < 0). spatial separation between the electrons and the holes weakens
Such distributions are obtained at low concentrations of the excitonic binding so much that the excitons, if any, are ionized
injected carriers and/or high temperatures, into free electron-hole pairs. The theoretical line-shape in these
! true superlattices is modified with respect to the formula given in
y Eq. (122) due to the subband dispersions along the growth axis
`(y) ∝ Y(y) exp − (123)
kB T ∗ and the peculiar selection rules prevailing in type-II superlattices
(see [36]): one of the two van Hove singularities of the joint den-
with sity of states is suppressed. From Figure 14, it can be seen that the
1 1 Mn∗ 1 m∗e high-energy part of the spectrum is well reproduced by the cal-
= ∗ ∗ + (124)
kB T ∗ kB Te Mn + me kB Th Mn + m∗e
∗ culations whereas the low-energy side cannot be interpreted by
 OPTICAL PROPERTIES IN NANOSTRUCTURES 19

Fig. 14. Calculated and measured band-to-band recombination in an

InAs–GaSb superlattice. Solid line: theory. Open circles: data in Ref. [35].
Fig. 12. Calculated band-to-band recombination lineshape in a GaAs– (30–50 Å), Teff = 370 K and εg = 230 meV. (Reprinted from P. Voisin,
Ga0:67 Al0:33 As quantum well for nondegenerate populations of electrons Thése de Doctorat d’Etat, Paris, 1983, unpublished.)
and holes. At T = 77 K only the ε1 − hh1 recombination is significant
while at T = 300 K the thermal population of the hh1 subband is sufficient
enough to allow the existence of a sizeable ε1 − lh1 recombination. εeffg A second example about the radiative recombination of n-type
denotes the effective bandgap of the quantum well viz. εg (GaAs) + ε1 +
modulation-doped GaAs quantum wells is presented in Ref. [37].
hh1 . For h̄ω < εeff
g , there is no band-to-band emission. lh1 − hh1 = 25 The large electron concentration effectively decreases the exci-
meV, n = p = 2 × 1010 cm−2 . tonic binding in these structures and consequently band-to-band
emission predominates. The luminescence band contains three
lines. The two high-energy lines were interpreted in terms of
ε1 → h1 and ε2 → h1 recombinations. To account for the spectral
positions of the lines, a “renormalized” GaAs bandgap had to be
invoked, the shrinkage of this gap being equal to 20 meV.
It is worth noticing that it is possible to observe the lumines-
cence spectra calculated in this subsection (and later on), only if
the absorption by the sample of the emitted photons can be ne-
glected. If `0 (ω) is the luminescence spectrum originating from a
point M at a distance z from the sample surface, the spectrum ef-
fectively observed corresponds to photons, which traveled through
the sample to reach the surface. It is given by
h i
`(ω) = `0 (ω)[1 − R] exp −zα(ω) (127)

where α(ω) and R are the absorption coefficient and the reflectiv-
ity of the considered sample, respectively. It is sensible to assume
that the spontaneous emission is homogeneous, which means that
`0 (ω) is independent of z, so that the apparent luminescence spec-
trum is given by

Fig. 13. Same as in Figure 12 except that the recombining populations [1 − exp(−hα(ω))]
`(ω) = (1 − R)`0 (ω) (128)
have much larger densities. At high temperature the electrons and the hα(ω)
holes approximately follow a Boltzmann distribution while at low temper-
where h is the distance that a photon should travel in the sample.
ature the electron and hole gases are degenerate, which affects the lumi-
The luminescence spectrum is thus modified by photon reabsorp-
nescence lineshape. For h̄ω < εeffg , there is no band-to-band emission.
tion, the low-energy side being favored with respect to the high-
n = p = 5 × 1011 cm−2 .
energy one (see Fig. 15). For photons which propagate along the
growth axis of the heterostructure this effect is weak and for single
quantum wells negligible since h ≤ 1 µm and α(ω) ≈ 60 cm−1 .
a model which assumes an ideal superlattice. Recombination in- However, the situation may be different for photons which prop-
volving impurities or defects have to be invoked to account for the agate in the layer planes because, in this case, h is of the order of
low-energy side of the recombination spectrum. several millimeters.
Notice that the effective carrier temperature is significantly
higher than the lattice temperature. This hot carrier luminescence
may have arisen due to the large kinetic energies supplied to the
2.3.2.2. Excitonic Recombination
injected carriers by the exciting laser light which is well in excess The excitonic recombination fulfills the same selection rules as the
of the effective bandgap of the superlattice. excitonic absorption, i.e., k⊥ = 0 and n+m even if the heterostruc-
20 RIERA ET AL.

concept of bound electron-hole pairs fades away to be replaced by

that of an electron-hole fluid, i.e., interacting fermions systems.
These two difficulties (relaxation of the k⊥ selection rule and
the nature of the exciton statistics) were the subjects of a consid-
erable body of literature in bulk materials (see, e.g., [31, 38]. They
were not studied much in heterostructures, both from the experi-
mental and the theoretical point of view. Therefore, we shall limit
our considerations to a presentation of some experimental results.
One of the dominant features of the optical properties of
GaAs–GaAlAs quantum wells is the strong intensity of the “free”
exciton luminescence at low temperatures with respect to the in-
tensity of the lines involving impurities. This is the opposite of what
is usually observed in bulk GaAs. This is remarkable but has not
been satisfactorily explained, at least to our knowledge. Possibly,
this may be related to the very small effective volume in which
the carriers and the electromagnetic wave interact in the case of
quantum wells. Once created the photon has fewer chances of be-
ing reabsorbed to create another exciton than in bulk materials,
where the effective volume of interaction is the whole crystal.
The main arguments used to assign the photoluminescence line
to excitonic recombination in GaAs quantum wells is its energy
position and the nature of the recombining species, as supported
by the spin orientation measurements [39]. In “good” samples,
the maximum of the photoluminescence line is very close to (and
sometimes coincides with) the maximum of the hh1 → ε1 absorp-
tion line (or photoluminescence excitation line). In such a case,
the Stokes shift s, i.e., the energy separation between these two
Fig. 15. Calculated reabsorption effects on a photoluminescence line. lines (luminescence and absorption) is equal to zero but in other
Upper figure: absorption coefficient α times sample thickness versus en- samples s can be as large as several million electronvolts. The exis-
ergy (measured from the effective bandgap). Lower figure band-to-band tence of such a shift seems to be independent of the residual dop-
luminescence intensities without the reabsorption effect (`0 ) and including ing level of the GaAs (usually p- and n-type in molecular-beam
the reabsorption effects (`). In both figures, a phenomenological damping epitaxy (MBE) and metal-organic chemical-vapor decomposition
coefficient of 0 = 5:1 and kB T = 10 meV were assumed. (MOCVD) grown materials, respectively). Thus, it seems reason-
able to conclude that the observed excitons are not bound to ex-
trinsic defects (acceptors or donors). A model of exciton trapping
ture potential is symmetric in z (see Subsection 2.2.2). In addition, on intrinsic interface defects [40] is proposed, which correlates s
only the ns excitons can radiatively recombine. This means that the to the sizes of the defects (extension along the z-axis and in the
excitonic luminescence emitted below the εm − hhn band edge layer plane). This model predicts that beyond L ≈ 100 Å in GaAs–
should consist of monochromatic lines. Most likely a single line Ga0:47 Al0:53 As quantum wells the trapped exciton binding energy
caused by the 1s exciton attached to the ε1 − hh1 bandgap will becomes smaller than 1 meV. Thus, in this model, the Stokes shift
be observed, owing to the fast relaxation of exciton states toward could only be observed in thin quantum wells. The luminescence
the ground one. We have already had difficulties to interpret the line would be associated with trapped excitons with a low density
low-energy side of the band-to-band recombination line of nonin- of states (≤ 1010 cm−2 ), with respect to that of the delocalized
teracting electrons, all the luminescence signal below the ε1 − hh1 excitons (≈ a few times 1011 cm−2 ), while the absorption, which
edge arising from defects. is essentially sensitive to the large density of states regions of the
However, the high-energy side at least could be successfully energy spectrum, would exhibit features due to the delocated ex-
described by models which neglect defects. In the case of exci- citons. The thermally activated detrapping of excitons has in fact
tonic recombination, a proper theory should take into account the been observed [41]. In should be noted that the previous model
broadening effects from the very beginning. The broadening ef- was designed for defects whose in-plane dimensions are not too
fects are twofold. First, they provide bound exciton states below large (≤ 500 Å in a GaAs–GaAlAs quantum well). It is clear that
εg + ε1 + hh1 − R∗ . Second, they relax the k⊥ = 0 selection the excitonic photoluminescence in quantum wells with defects,
rules, allowing a finite luminescence above εg + ε1 + hh1 − R∗ which extend over 1000 Å or more in the layer plane, is better
to be predicted. At low temperatures, the broadening mechanisms described by models which consider free excitons moving in “mi-
are either the exciton-defect interactions or the exciton-acoustical croquantum wells” whose areas are equal to the defect areas and
phonon interactions. Careful studies should be able to discrimi- whose thickness is equal to the local quantum-well thickness in the
nate between both kinds of effects. defect. In other words, the criterion for deciding which model is
The second difficulty with excitons is the hybrid statistics which more appropriate is to count the number ℵ of bound states (for
they obey. Diluted excitons can be treated as bosons to a first ap- the exciton center of mass) that the defect supports. For ℵ ≤ 5,
proximation, essentially because they arise from the pairing of two the motion of the exciton center of mass is size-quantized and one
fermions, the electron and the hole. Excitons retain an increasing recovers a photoluminescence due to excitons bound to these de-
fermion nature when their density increases to the point where the fects. For ℵ ≥ 5, microquantum-well models are preferable and
 OPTICAL PROPERTIES IN NANOSTRUCTURES 21

if the diffusion of excitons from one large defect to another one

is difficult, structures in the excitonic photoluminescence should
be observed. These structures appear at the local exciton energies
in these various islands. Typically, with one monolayer fluctuations
of the quantum thickness L, the photoluminescence shows three
peaks due to the excitons which move in micro-quantum wells of
thickness L, L ± a, where a = 2:83 Å is the thickness of a GaAs
monolayer.
Such structures were evidenced [42] in high quality MBE grown
GaAs–GaAlAs multiple quantum wells and superlattices. It is now
admitted that the depth of interface fluctuations in GaAs–GaAlAs
quantum wells grown by MBE can be reduced to one monolayer.
On the other hand, the in-plane extensions of the interface defects
depend significantly on the growth conditions.
Thus, a variety of experimental results concerning the Stokes
shift and the shape of the excitonic photoluminescence line may
Fig. 16. Calculated and measured excitation spectrum linewidth as a
and do occur. A series of sophisticated optical measurements
function of the confinement energy (viz. ε1 + hh1 ) in GaAs–Ga1−x Alx As
(transient gratings, Rayleigh scattering, hole burning, etc. ...) were multiple quantum wells.
undertaken by Sturge et al. [43, 44] to ascertain the nature of exci-
ton states in quantum wells. The aim is to discover whether quasi-
two-dimensional excitons can become localized by a weak disor-
der, the localization being the result of constructive interferences
of the exciton wave functions by randomly located defects. These
experiments tend to show that, at low temperature, excitons are
localized rather high in energy and that even a large part of the
absorption spectrum arises from these localized excitons.
To our knowledge, the free exciton luminescence line-shape in a
quantum well has never been calculated. One currently character-
izes the quality of quantum-well structures by the width δ at half-
maximum of this line. When δ is smaller than some million elec-
tronvolts, the sample is claimed to be good. However, it is worth
noticing that quantum wells, which exhibit broad lines, display a
very intense luminescence, which shows the arbitrary nature of the Fig. 17. The photoluminescence decay times τ (at 1/e of the maximum)
criterion. Empirically the width δ and the Stokes shift are corre- of several GaAs–Ga1−x Alx As heterostructures are plotted versus the
lated: the wider the luminescence, the larger the shift. GaAs layer thickness Lz at T = 10 K. SQW, DQW, and MQW stand for
A model of the width γ of the luminescence excitation spectrum single, double, and multiple quantum wells, respectively.
was proposed by Weisbuch et al. [29]. If the local quantum-well
thickness is L + δL inside an interface defect, the exciton absorp-
tion occurs at an energy given by by the GaAs well and the GaAs substrate in terms of excitonic re-
combination.
h̄ω(L + δL) = ε1 (L + δL) + hh1 (L + δL) Investigations of the time dependence of the photolumines-
− εb (L + δL) + εg (129) cence of GaAs–GaAlAs quantum wells give results which dif-
fer significantly from one group to the other. The simple model
thus the line-width γ is given by described in Eqs. (100)–(104) would lead us to interpret these
" # differences in terms of sample-dependent nonradiative lifetimes.
dE1 dhh1
γ ≈ δL + (130) Nonetheless, it seems agreed that the characteristic time of the ex-
dL dL citon luminescence decay at low temperature shortens when the
since the variation of the exciton binding is negligible if δL  L. quantum-well thickness L is decreased (see, e.g. [46, 47]). If one
identifies the decay time with the radiative lifetime τr , and if one
As the confinement energies vary like L−2 , γ should vary like L−3 .
assumes the recombination to be excitonic, the lifetime reduction
This behavior was observed in GaAs–GaAlAs multiple quantum
can be understood from the squeezing of the in-plane exciton Bohr
wells (see Fig. 16). Singh et al. [45] proposed a model of the pho-
radius a∗B2D with decreasing L. Indeed, the inverse of the radiative
toluminescence line-width δ. These authors correlated δ to the dis-
tribution of fluctuations in the well thickness, neglecting however lifetime τr−1 , which is proportional to the transition probability per
the effect of carrier relaxation toward lower energy states. unit time that a photon is emitted, contains (as in the case of ex-
The nature of the radiative recombination at room tempera- citonic absorption) an enhancement factor |β1s (0)|2 which varies
ture in GaAs quantum wells is rather controversial. Taking into (see Eq. (88), like a∗−2
B2D (L)). In the ideal situation where valence
account the dissociation of the excitons, it seems unlikely that this and conduction barriers are infinitely high and where all the va-
recombination is entirely due to excitons. However, Dawson et al. lence band mixings are neglected, the in-plane extension a∗B2D de-
[34] and Bimberg et al. [33] interpreted their steady-state photolu- creases from a∗B to a∗B /2 when L decreases from infinity to zero.
minescence and time-resolved cathodoluminescence lines emitted Thus, the excitonic luminescence lifetime should decrease by a fac-
22 RIERA ET AL.

tor of 4 between these two limits. It so happens that the photo- free standing etched structures, the spatial variation of n may be
luminescence decay time decreases approximately by a factor of of importance. Variations of n, caused by the absorption to be cal-
four from bulk GaAs to GaAs quantum wells with L = 40 Å (see culated, are negligible in real systems when any singularity of the
Fig. 17). This excellent agreement between theory and experiment joint density of states is suppressed by level broadening.
is probably fortuitous. In the dipole approximation, the absorption coefficient α for a
plane electromagnetic wave in a medium of refractive index n is
given [1, 68] by
2.4. Optical Properties in Semiconductor
Quantum-well Wire πe2 X D E2

α=  f |er · p̂|i  δ(εf − εi − h̄ω) (131)
2.4.1. Interband Absorption in Quantum-well Wires ncκ0 m20 ωV i;f

In this subsection the theory of interband transitions in quasi- κ0 , m0 , V , and h̄ω represent the permitivity of free space, the mass
one-dimensional semiconductor structures is described following of the free electron, the sample volume, and the photon energy, re-
Bockelman and Bastard [48]. It is based on the effective mass ap- spectively. The polarization vector er defines the orientation of the
proximation and takes into account the mixing of the heavy-hole electric field of the linearly polarized wave. It is assumed that any
and the light-hole states in the valence band. The dipole matrix initial electronic state i (final state f ) is occupied (unoccupied).
element for optical transitions between the valence band and the In quasi-zero-dimensional, quasi-one-dimensional, and quasi-
conduction band states is derived for different linear photon polar- two-dimensional systems the sample volume V has to be replaced
izations with respect to the wire orientation. The theory is applied by 1, `y (length of the wire), and `x , `z (area of the quasi-two-
to single wires and periodic arrays of coupled wires exhibiting spa- dimensional layer), respectively, to obtain an unambiguous defini-
tially direct or indirect optical transitions. The influence of sam- tion of an absorption quantity via Eq. (131). R This can be seen by
ple imperfections by calculating the optical properties of a statisti- converting the sum into integrals (`v /2π dkv ) for the directions
cal ensemble of confinement potentials exhibiting random fluctu- v that are translationally invariant. In the other spatial directions,
ations was also studied. ill-defined lengths remain in the expression of α. The proposed
The spectroscopy of optical transitions across the bandgap replacements of V in Eq. (131) avoid this problem by absorbing
(interband transitions) is a powerful and versatile method to these lengths in the definitions of the absorption quantities: three-
study the electronic structure of semiconductors. Such experimen- dimensional-α, in units of m−1 ; two-dimensional-α`z , in units of 1;
tal activities are increasing in quasi-one-dimensional structures. one-dimensional-α`x `z , in units of m; zero-dimensional-α`x`y `z ,
Some effects, which were related to a quasi-one-dimensional car- in units of m2 . In a way, the definitions of an “effective wire width”
rier motion, were already observed in photoluminescence and α`x cz for one-dimensional systems and an “absorption probabil-
photoluminescence-excitation spectra [49–55], and their depen- ity” α`z for two-dimensional systems close the gap between the ab-
dence on the light polarization [53–55] and on time [56] was sorption cross-section σabs = α`x `y `z and the absorption coeffi-
studied. Single-particle and collective excitations of the quasi- cient α, which are well known from atomic and solid-state physics.
one-dimensional electron gas were investigated by Raman spec- In Subsection 2.4.1.6, these definitions are explicitly related to the
troscopy [57, 58]. attenuation of a light wave that propagates parallel or perpendic-
On the theoretical side, there have already been some calcula- ular to one-dimensional or two-dimensional structures.
tions of the energy levels in the conduction and valence bands of
quasi-one-dimensional structures [59–63]. Quasi-one-dimensional
interband transitions were studied only in the parabolic approxi- 2.4.1.2. One-dimensional Conduction and
mation for the valence band [64–66]. A proper treatment of the Valence-band States
valence band not only leads to modified transition energies and To calculate interband transitions, the energy dispersions and wave
oscillator strengths, but is necessary to account for the different functions of the initial and the final electron states have to be
possible polarizations of the light relative to the wire structure. known. This subsection deals with the optical transitions that are
The absorption of a light beam traveling parallel to the epitaxial close in energy to the fundamental absorption across the bulk
growth axis is more easily measurable than that of a beam travel- semiconductor bandgap. We suppose that the initial and final elec-
ing perpendicular to it. It is shown [67] that the polarization depen- tron states are eigenstates separately described by the valence
dence of the absorption in the former configuration is determined band 08 and the conduction band 06 effective mass Hamiltonian,
by the heavy-hole (hh) or light-hole (lh) character and the amount respectively. In that way, we neglect the presence of the electron-
of hh − lh mixing of the quasi-one-dimensional valence subbands. hole Coulombic interaction (excitonic effects) and we suppose that
the bandgap energy as well as the splitoff energy are sufficiently
large to prevent a sizeable coupling between the 06 , 07 , and 08
2.4.1.1. Theoretical Basis
bands. As usual, the lower dimensional electronic states are built
In general, the electronic structure and the dielectric properties on these bulk bands in the envelope function approximation [1].
of lower dimensional systems can be modified in comparison with The conduction band (06 ) wave functions are written as
those of the three-dimensional crystal. We are interested in how
ψcms (r) = f c (r)ucms (r) (132)
the electronic structure finds expression in the optical properties
of quasi-one-dimensional systems. It is assumed that the refrac- where the ucms (r) are the two spin-degenerate (s = 1/2, ms =
tive index n is spatially constant. This should work fine for a light ±1/2) Bloch functions at the bottom of the 06 bulk band (k = 0).
wave propagating in as-grown or overgrown etched quantum-well In the valence band (08 ), we have to account for two different
wires that are based on semiconductor heterostructures with sim- carrier types, which originate from the k = 0 degenerate heavy-
ilar refractive indices of the host materials (e.g., GaAs–AlAs). In hole and light-hole bands of the bulk semiconductor. There, the
 OPTICAL PROPERTIES IN NANOSTRUCTURES 23

equivalent of Eq. (132) reads We use

 E
X 
ψv (r) = v (r)uv (r)
fm (133) uv3/2 = 2−1/2 (x + iy) ↑
j mj
mj  E  1/2

uv−1/2 = −6−1/2 (x − iy) ↑ − 23 |z ↓i
The uvmj are the degenerate Bloch functions at the top of the  E  1/2

08 bulk bands. The sum extends over the four expectation values uv1/2 = 6−1/2 (x + iy) ↓ − 23 |z ↑i (142)
(mj = ±3/2 for the hh and 1/2 for the lh) of the j = 3/2 multiplet.  E

For the sake of simplicity, we assume that the potential W(x; z) uv−3/2 = −2−1/2 (x − iy) ↓
that confines the carriers into the one-dimensional structure can uc1/2 = i|s ↑i u−1/2 = i|s ↓i
be written as W(x; z) = Vx (x) + Vz (z). In addition, any depen-
dence of the effective mass in a given direction on the motion per- The Luttinger parameters γ1 − γ3 describe the coupling be-
pendicular to it is neglected. According to that, the envelope wave tween the 08 and the edges of all the other host bands [70] and
functions in Eqs. (132) and (133) separate in the three spatial co- are in principle z-dependent. In Eq. (141), the matrix element c is
ordinates x, y, and z. This “decoupling approximation” describes given in the axial approximation [71].
well the common situation, where the confinement in the growth If Vz (z) and the γs are even functions of z, the envelope func-
direction (z) is stronger than that in the lateral direction (x) [59]. tions can be chosen to have a definite parity in the z-direction. The
In the conduction band, the envelope wave function, off-diagonal elements of H b0 couple functions of either opposite
8
+ +
(b; b ) or equal (c; c ) parity. This implies that there are two de-
−1/2
f c (r) = ϕe (x)`y exp(iky y)χe (z) (134) coupled types of 08 wave functions: ψ ↑ and ψ ↓. The first (sec-
ond) exhibit even, even, odd, odd (odd, odd, even, even) z-parity
and the corresponding energy ε = εxy +εz , are determined by the v ;fv v v
of their envelope functions f3/2 −1/2 ; f1/2 ; f−3/2 , respectively.
following two differential equations,
The notations ψ ↑ and ψ ↓ point out that the corresponding en-
 ! 
 h̄2 ∂2  ergy levels are degenerate at zero magnetic field B, while they split
− + k 2 + V (x) − ε ϕ (x) = 0 (135) for nonzero B.
 2m∗e (0) ∂x2
y x xy
 e The calculations presented in the following subsection suppose
 !  a symmetric GaAs–Ga0:7 Al0:3 As quantum well in the z-direction
 h̄2 ∂ 1 ∂ 
− + V (z) − ε χ (z) = 0 (136) that gives rise to a relatively large subband separation with respect
 2 ∂z m∗e (z) ∂z z z
 e to the energy range of interest (stronger confinement in z than
in x). Then we can restrict the expansion to the three lower band
m∗e (z) is the (z-dependent) effective mass at the bottom of the 06 states of that quantum well: the first hh, the first lh, and the second
band and ky is the carrier wave vector of the in-wire motion. hh states (χhh1 ; χlh1 , and χhh2 ), respectively, [72]. This approxi-
The calculation of the 08 valence band starts from the Hamil- mation leads to the following 08 wave functions,
tonian, n
↑ ↑
b   ψv↑ = ϕhh1 (x)χhh1 (z)uv3/2 + ϕlh1 (x)χlh1 (z)uv−1/2
 Hhh c b 0   3/2  o
 + blh    ↑ −1/2
b0 =  c + H 0 −b   −1/2  + ϕhh2 (x)χhh2 (z)uv−3/2 `y exp(iky y) (143)
H blh    (137)
8  b 0 H c   1/2  n
 + + b   ↓ ↓
ψv↓ = ϕhh1 (x)χhh1 (z)u−3/2 + ϕlh1 (x)χlh1 (z)uv1/2
0 −b c Hhh −3/2
o
↓ −1/2
where + ϕhh2 (x)χhh2 (z)uv3/2 `y exp(iky y) (144)
 ! 
2  2 
bhh = h̄ ∂ ∂ ∂ b0 on Eq. (143), we obtain a system of three coupled
By applying H
H (γ1 + γ2 ) − k2y + (γ1 − 2γ2 ) 8
2m0  ∂x2 ∂z ∂z  ↑
eigenvalue equations for the envelope functions ϕhh1 , ϕlh1 , and
↑
↑
+Vx (x) + Vz (z) (138) ϕhh2 and the eigenenergy ε,
 ! 
2  2  n o
blh = h̄ ∂ ∂ ∂ b − ε1 ) ϕ↑ (x); ϕ↑ (x); ϕ↑ (x) = 0
H (γ1 − γ2 ) − k2y + (γ1 + 2γ2 ) (H hh1 lh1 hh2
(145)
2m0  ∂x2 ∂z ∂z 
where
+Vx (x) + Vz (z) (139)  2 
√ ! !  εhh1 − 2mh̄∗ 
h̄2 3 ∂  
∂ ∂  hh

b = − + ky γ3 + γ (140)  ×(− ∂22 +k2y ) 2
− h̄2µ ( ∂x
∂
+ky )2 0 
2m0 ∂x ∂z ∂z 3  ∂x 
 +Vx (x) 
√ !2  
 2 
h̄2 3 γ2 + γ3 ∂  − h̄ (− ∂ +k )2 h̄ 2 
c = − + ky (141)  2µ ∂x y εlhh1 − 2m ∗ 
2m0 2 ∂x  lh 
b=
H ∂2 +k2 ) h̄2 ( ∂ +k ) 
 ×(− y 
 ∂x2 y m0 l0 ∂x 
Equation (137) was derived from the original work of Luttinger  +Vx (x) 
 
and Kohn (Eq. (V.13) of Ref. [69]) by replacing kx by −i∂/∂x,  
 h̄2 ∂ +k ) εhh2 − 2m∗ h̄ 2 
by replacing kx by −i∂/∂z, and by symmetrizing any product of  0 m0 l0 (− ∂x y 
 hh 
noncommuting factors. The phases of the periodic wave functions  ∂ 2
×(− 2 +ky )  2 

b0 .
uvmj (r) are changed with respect to Ref. [69] to obtain a real H  ∂x 
8 +Vx (x)
24 RIERA ET AL.
√
3 γ2 + γ3 or the resulting anisotropy of the transition probability in the xy-
µ−1 = hχlh1 |χhh1 i
m0 2 plane increases in proportion to (J3/2 J−1/2 + J1/2 J−3/2 ) with in-
√ * + creasing hh − lh mixing in the valence band. Let us stress that the
−1 3 ∂ ∂
l0 = χlh1 |γ3 + γ |χ (146) use of decoupled hh and lh valence-band states would result in a
2 ∂z ∂z 3 hh1
zero xy-anisotropy.
m0 m0
m∗hh = m∗lh = For the systems described by Eqs. (143)–(145), the polarization-
γ1 + γ2 γ1 − γ2 dependent interband matrix element (Eq. (149)) is simply given by
εhh1 , εlh1 , and εhh2 are the edge energies of the first hh, the " #2
XD E2
 m0 P
first lh, and the second hh subbands of the quantum well in the c v
 ψms |er · p̂|ψ  =
z-direction (solutions of the hh and lh equivalent of Eq. (136)). ms
h̄
The equation that determines ψ ↓ rather than ψ ↑ is obtained by " #
 1 2 2
reversing the signs of all terms proportional to ∂/∂x, and by re- 2
× Jhh1 + Jlh1 − √ Jhh1 Jlh1 cos 2ϕ
placing ↑ by ↓ in Eq. (145). Finally, we expand Eq. (145) in an  3 3
appropriate set of basis functions (the explicit choice depends on 
the form of Vx ) and we diagonalize the resulting eigenvalue matrix 4 
numerically. × sin2 ϑ + Jlh1 2 cos2 ϑ (150)
3 

2.4.1.3. Polarization Dependence of the Interband where

Z Z
Matrix Element
Jhh1 = dxϕhh1 (x)ϕe (x) dzχhh1 (z)χe (z)
Since the periodic part uc , uv varies rapidly over the characteris- Z Z (151)
tic length of variation of the envelope function f c , f v , the dipole Jlh1 = dxϕlh1 (x)ϕe (x) dzχlh1 (z)χe (z)
matrix element of a transition between a 06 (Eq. (132)) and a 08
(Eq. (133)) state can be written as In Eq. (150), we added the equal contributions of the two parity
D E X D E
degenerate solutions ψ ↑ and ψ ↓. By χe , we mean the ground
ψcms |er · p̂|ψv = Jmj ucms |er · p̂|uvmj
mj
state wave function of Eq. (136). χe and χhh2 have opposite par-
D E ity, hence Jhh2 equals zero and does not appear in Eq. (150). The
Jmj = f c (r)|fm
v (r)
j
(147) overlap integrals J in Eqs. (147)–(150) are supposed to be real,
which is no restriction, since it is always possible to choose real
The atomic-like dipole matrix elements give rise to the depen- solutions of the Schrödinger equation.
dence on the polarization vector er of the light wave. They are
weighed by the quantum numbers of the initial and the final states
via the overlap integrals Jmj . 2.4.1.4. Results for Some One-dimensional Structures
The symmetry of the host functions that define the uc , uv by In all the following one-dimensional model systems, the electrons
means of Eq. (142) gives rise to the selection rule,
and holes are confined in the z-direction by a symmetric GaAs
D E
s|p̂v |v0 = δvv0 im0 p/h̄ v; v0 ∈ {x; y; z} (148) quantum well embedded in Ga0:7 Al0:3 As. We suppose that the
offset of the conduction (hh and lh valence) band equals 2/3
For the Kane matrix element p [68], the GaAs value 2m0 p2 = (1/3) of the bandgap difference between Ga0:7 Al0:3 As and GaAs
22:71 eV is used. We express the polarization vector in spherical of 0.354 eV. The effective mass m∗e = 0:067m0 (0:083m0 ) and
coordinates er = (cos ϕ sin ϑ; sin ϕ sin ϑ; cos ϑ) and we evaluate the Luttinger parameters γ1 = 6:85(5:83), γ2 = 2:10(1:67),
Eq. (147) using Eqs. (142) and (148). We obtain γ3 = 2:90(2:42) are used for the well (barriers) throughout the
" #2 calculations.
XD E2
 m0 P
c v
 ψms |er · p̂|ψ  =
ms
h̄ 2.4.1.5. Infinitely Deep Square Well Model
2 2 2

 3 (J1/2 + J−1/2 ) for ϑ = 0 In a first approach, we model the lateral confinement by a rectan-



 2 2 2 2 gular well of width Lx with infinite barriers for both electrons and
 2 (J3/2 + J−3/2 ) + 6 (J1/2 + J−1/2 )
1 1
holes. The lateral wave functions are written as
× 1 (149) !

 − √ (J3/2 J−1/2 + J1/2 J−3/2 ) ∞

 3 X nx π

 π ϕ(x) = cnx sin x (152)
 × cos 2ϕ ϑ = Lx
2 nx

Equation (149) is very general. In particular, we made no assump- The basis functions sin(nx πx/Lx ) of this expansion are the
tions concerning the confining potential. Neglecting the internal eigenfunctions in the conduction band (Eq. (135)). Figure 18
anisotropy of the lattice unit cell, the absorption is independent of demonstrates how the lateral confinement transforms the two-
er in a cubic three-dimensional lattice. In two-dimensional systems dimensional valence-band dispersions of a 5-nm (a) and a 10-nm
(Vz 6= 0) the absorption depends on the angle ϑ. The summation (b) wide quantum well into one-dimensional subbands. The two-
over the in-plane wave vector eliminates the term proportional to dimensional branches (dashed lines) correspond at k = 0 to the
cos 2ϕ of Eq. (149). This term survives the integration over the first hh, the first lh, and the second hh eigenstates (at −7, −20:5,
in-wire momentum in one-dimensional systems. The magnitude and −27:7 meV in Figure 18(b)). A finite in-plane momentum k
 OPTICAL PROPERTIES IN NANOSTRUCTURES 25

Fig. 18. Solid lines: valence subbands of a quantum-well wire as a func-

Fig. 19. Edge energies of the one-dimensional valence subbands as a
tion of the in-wire wave vector, Lx = 50 nm. Dashed lines: valence sub-
bands of a quantum well as a function of the in-plane wave vector. The top function of the lateral wire width Lx . Lz is (a) 5 nm and (b) 10 nm.
of the bulk valence band corresponds to zero energy.

couples these states. In one-dimensional systems, the kinetic en-

ergy that is introduced by the lateral confinement (∂/∂x terms
in Eqs. (140) and (141)) mixes the two-dimensional band edge
eigenstates even for the zero in-wire wave vector ky . Thus, any
one-dimensional valence band state has both hh and lh contri-
butions. Nevertheless, a state with an energy in the vicinity of a
two-dimensional subband edge can display a dominant heavy- or
light-hole character. Therefore, the one-dimensional and the two-
dimensional subband dispersions exhibit a similar shape in the
vicinity of the two-dimensional subband edges.
In Figure 19, the energies at the edge of the one-dimensional
valence subbands (ky = 0) are plotted against the lateral width Lx
of the wire. Near the two-dimensional lh edge εlh1 (at −47:5 in (a)
and −20:5 in (b)) the curves show a weak dependence on Lx . This Fig. 20. Absorption of the quantum-well wire plotted versus photon en-
is due to the flatness of the two-dimensional lh dispersion (Fig. 18) ergy h̄ω for the three orthogonal light polarizations. Lx = 50 and Lz =
and indicates the existence of one-dimensional states with a strong 10 nm.
lh contribution over the whole range of Lx . The curves corre-
sponding to states of opposite parity on x cross, while the anti-
crossing occurring between the subbands of equal parity increases For lower photon energies h̄ω, the lh contribution to the corre-
with decreasing Lx . For nonzero ky , all branches anticross. sponding 08 state is much smaller, and therefore the structures
Figure 20 shows the absorption spectra of the quantum wire of the lower one-dimensional absorption edges are masked by the
of Figure 18(b) for polarization vectors parallel to the x-, y-, broadening of the h6e1 peak.
and z-axes (labeled x, y, and z, respectively). Broadening effects The dependence of the absorption on the polarization in the
caused by sample imperfections are modeled by replacing the δ xy-plane is described by the product Jhh1 Jlh1 cos 2ϕ of Eq. (150).
function of Eq. (131) by a Lorentzian with a full width of 2 meV The xy-anisotropy is expressed by (αx − αy )/(αx + αy ), where αx
at half-maximum. For a polarization in the z-direction, only the and αy are the absorption coefficients for a polarization parallel
lh parts of the 08 state contribute to the absorption (Eqs. (149) to the x- and y-axes, respectively. If the overlap integrals Jhh1 and
or (150)). The peak of the spectrum labeled z in Figure 20 is mainly Jlh1 are equal (opposite) sign, the absorption is weaker for x(y)-
due to the transition h6e1 (from the sixth valence to the first con- polarization and the xy-anisotropy is negative (positive). The mag-
duction one-dimensional subband) and partly due to the nonzero nitude of the anisotropy increases proportionally to Jhh1 Jlh1 with
ky contribution of the transitions exhibiting lower edge energies. increasing hh-lh mixing.
26 RIERA ET AL.

Fig. 21. xy-anisotropy of the dominant one-dimensional absorption

edges as a function of the lateral wire width Lx . αx and αy are the ab- Fig. 22. Edge energies of the conduction (three topmost branches) and
sorption coefficients of the indicated transitions for a polarization parallel valence subbands of a (50 × 10 nm2 cross-section) quantum-well wire as a
to the x- and y-axes, respectively. Lz = 10 nm. function of the electric field in the x-direction E. Lz = 10 nm and ky = 0.

In Figure 20, the h1e1 and h2e2 peaks are stronger for y- than
for x-polarization. The larger xy-anisotropy at the h2e2 edge re-
flects the stronger hh − lh mixing in the second valence subband.
The edge of the h6e1 transition located at about 53 meV exhibits
the opposite polarization dependence.
In Figure 21, the effect of a polarization in the xy-plane on the
peaks of the absorption spectrum (h̄ω equals the transition ener-
gies at ky = 0) is presented as a function of the wire width Lx .
We plotted the xy-anisotropy for any transition that contributes
substantially (α`x `z > 0:03 nm) to the absorption in the en-
ergy range of the lowest one-dimensional subband edges. Over
the whole range of Lx , only the h1e1, the h2e2, and the transi-
tions involving strongly lh-like valence subbands (flat curves near
the two-dimensional lh energy ε1h1 , in Fig. 19(b)) are impor- Fig. 23. Normalized xy-anisotropy of the dominant (α`x `z > 0:03 nm)
tant. The anisotropy increases with decreasing Lx due to the in- absorption peaks as a function of E. The diameters of the circles indicate
creasing hh − lh mixing. At a fixed Lx , the transitions that origi- the strength of the peaks of the different transitions averaged over the in-
nate from different valence subbands exhibit different magnitudes plane polarization [(αx + αz )`x `z /2].
of the anisotropy. A positive (negative) xy-anisotropy identifies
a transition that involves a valence subband of dominant lh(hh)
character. The field E influences the polarization-dependent absorption
The qualitative properties of the polarization-dependent ab- in two ways. First, it shifts the edge energies of the conduction and
sorption are independent of the model used for the lateral con- valence subbands (Fig. 22), and with it the spectral position of the
finement. In Subsection 2.4.1.7, we show by an analytical perturba- absorption peaks (Stark shift). Second, the lateral wave functions
tive treatment of the hh − lh coupling that the lateral anisotropy deform and separate increasingly with E. The latter effects give
characterizes the hh- or lh-like character of the one-dimensional rise to the electric field dependence of the xy-anisotropy and to the
valence subbands in the same way for any symmetric confinement variations of the strength of the one-dimensional transitions plot-
potential that gives rise to spatially direct transitions. ted in Figure 23. In the electric field range where a given transition
An electric field breaks down the inversion symmetry in the di- exhibits a sizeable strength, its transition energy and anisotropy
rection of its application. Pointing in the z-direction, it lifts the do not change significantly. The influence of spatial variations of
parity degeneracy in the valence band, and Eqs. (143)–(145) are E caused by defects in real structures can be estimated by tak-
no longer sufficient. ing the average over different spectra calculated for different E.
On the other hand, an electric field E in the x-direction can be However, for a variation of E smaller than about 2 kV/cm, the
described by simply adding eEx to the potential Vx , in Eqs. (135) dominant spectra in this average exhibit similar polarization de-
and (145). Its influence is of particular interest for real sys- pendence. This indicates that the xy-anisotropy is relatively stable
tems where charged impurities at the lateral boundaries of the with respect to this kind of sample imperfections.
quantum-well wire produce an electric field E along the x-axis,
whose magnitude may vary along the wire axis y. A simple estimate
for the order of magnitude of E is the electric field in a capacitor
2.4.1.6. Attenuation of a Light Beam in One-dimensional and
consisting of two oppositely charged two-dimensional sheets sepa-
Two-dimensional
rated by an undoped GaAs layer of width Lx . There, a density of In this subsection, the attenuation of a light beam in one-
1010 cm−2 of singly charged impurities results in E ∼ = 1:5 kV/cm. dimensional and two-dimensional structures is discussed in terms
 OPTICAL PROPERTIES IN NANOSTRUCTURES 27

of the absorption quantities defined in Subsection 2.4.1.1. The sim-

plest configuration is the perpendicular penetration through a two-
dimensional (x − y) layer. The ratio of the transmitted to the inci-
dent intensities is given by
It /(1 − α`z ) (153)
For a propagation parallel to the two-dimensional plane (for in-
stance, in the y-direction), the intensity profile of the light beam
I(z) is of importance. We define an effective beam width in the
z-direction by Z
e
Lz = dz I(z)/I0 (154)

where I0 stands for the intensity in the interior of the active struc-
Fig. 24. The two different types of lateral confinement.
ture. For the attenuation of the beam, we obtain
I(y) = I(0) exp(−α`z y/e
Lz ) (155)
In the following, we treat the one-dimensional structure (wire axis The two-component vectors in Eqs. (159) and (160) operate
in the y-direction) in a very similar way. For a beam that penetrates on the basis [χhh1 (z)uv3/2 ; χlh1 (z)uv−1/2 ] and [χhh1 (z)uv−3/2 ,
perpendicular through the wire (propagation parallel to x), we get χlh1 (z)uv1/2 ] for the solutions ψ ↑ and ψ ↓ (Eqs. (143) and (144)),
It /Ii = (1 − α`x `z /e
Lz (156) respectively.
To first order in the off-diagonal terms of H b of Eq. (145), the
For a propagation parallel to z, the parameter e Lz , has to be re- pure hh state (Eq. (159)) gets an lh contribution (from Eq. (160))
placed by eLx (defined by the equivalent of Eq. (154) for the and vice versa,
x-direction).
A light beam that propagates parallel to the wire axis is de- |ψhh i1 = |ψhh i0 + c|ψlh i0 (161)
scribed by |ψlh i1 = |ψlh i0 − c|ψhh i0 (162)
Z
I(y) = I(0) exp(−α`x `z y/e
Sx;z ) (157)
2 dxϕlh (x)(−∂2 /∂x2 )ϕhh (x)
where e
Sx;z is related to the intensity profile I(x; z) of the beam by c =
h̄
Z Z 2µ ε0hh − ε0lh
e
Sx;z = dx dz I(x; z)/I0 (158) Z
dxϕ0lh (x)ϕ0hh (x)
h̄2
In all these configurations, the attenuation is described by two = >0 (163)
well-defined, independent quantities that represent the absorp- 2µ ε0hh − ε0lh
tion properties (α`z for two-dimensional and α`x `z for one- For a stronger confinement in the z- than in the x-direction, the
dimensional structures) on the one hand, and the respective in- difference between the zeroth-order hh and lh energies (ε0hh −
tensity profiles of the light wave (e Lz ; e
Lx ; e Sx;z ) on the other hand.
ε0lh ) is positive for the uppermost valence subbands. The integral
in the numerator of Eq. (163) is positive as well, since ϕhh (x) and
2.4.1.7. Heavy-hole–Light-hole Coupling by ϕlh (x) have similar shape. The resulting positive sign of c implies
Perturbation Theory a positive (negative) factor Jhh1 Jlh1 for an interband transition in-
In this subsection, we treat the hh − lh-coupling by perturbation volving the state |ϕhh i1 (|ϕlh i1 ). Together with Eq. (150) it follows
theory and we derive analytically the sign of the factor Jhh1 Jlh1 that
that determines the effect of a lateral polarization via Eq. (150). αx − αy
> 0 for |ψlh i1
Since we are interested in the peaks of the one-dimensional joint αx + αy
density of states, we set ky to zero. (164)
αx − αy
Let us now look at interband transitions to a given one- < 0 for |ψhh i1
αx + αy
dimensional conduction-band state. When the confinement po-
tential in the x-direction has a similar shape for the holes as for
the electrons (type-I systems, see Fig. 24), only the lateral valence 2.5. Optical Properties in Semiconductor Spherical
band wave functions with the same subband index as the final Quantum Dots
conduction band state in question contribute significantly to the
electron-hole overlap integrals. These lateral functions are labeled To interpret the phonon assisted optical transitions in semicon-
ϕhh (x) for the heavy hole and ϕlh (x) for the light hole. ductor quantum dots, following Fomin et al. [73] a theory is de-
To zeroth order of perturbation (in the absence of hh − lh- veloped comprising the exciton interaction with both adiabatic
coupling), the valence band eigenfunctions are either hh states, and Jahn–Teller phonons and also the external nonadiabaticity
  (pseudo-Jahn–Teller effect). The effects of nonadiabaticity of the
ϕhh (x) exciton-phonon system are shown to lead to a significant enhance-
|ψhh i0 = (159)
0 ment of phonon-assisted transition probabilities and to multi-
or lh states, phonon optical spectra that are considerably different from the
 
0 Franck–Condon progression. The calculated relative intensity of
|ψlh i0 = (160) the phonon satellites and its temperature dependence compare
ϕlh (x)
28 RIERA ET AL.

well with the experimental data on the photoluminescence of CdSe scenario of the proper Jahn–Teller effect for the
quantum dots, both colloidal and embedded in glass. impurity centers, transitions do occur between different
There was an increasing experimental interest in multiphonon states of a degenerate exciton level in a quantum dot,
photoluminescence [74–78] and Raman scattering [79–84] of provided the exciton-phonon interaction in the basis of
nanosize quantum dots. Photoluminescence peaks due to phonon- this degenerate level is characterized by
assisted processes were observed in colloidal quantum dots (nano- noncommutative matrices.
crystals) under size selective excitation [74, 75]. Observation of (ii) Differences between energy levels of an exciton in a
multiphonon photoluminescence of quantum dots embedded in quantum dot can become comparable to the optical
glass is complicated because of a quick trapping of holes onto phonon energy in the experiments [79, 74–83].
the local surface levels [85]. Overlapping of photoluminescence Therefore, the resulting effects of external
bands related to recombination of the electron-hole pairs, which nonadiabaticity, or the so-called pseudo-Jahn–Teller
are present in different states localized at the surface, smears the effect [99], in the phonon-assisted optical transitions are
features of the spectrum due to the phonon-assisted processes of crucial importance. The term “external
[77, 86]. nonadiabaticity” is used to make a distinction between
Distinct phonon line progressions, which are caused by recom- this class of nonadiabatic phenomena and the previously
bination of the electron-hole pairs in “interior” states of a quan- described proper Jahn–Teller effect relevant to a
tum dot, (i.e., the states spatially quantized due to the confinement degenerate exciton level.
potential), were observed in the fast components of photolumi- To the best of the knowledge, in the subsection on optical spec-
nescence of CdSe quantum dots embedded in glass [76, 77] using tra of quantum dots, the effects of nonadiabaticity were ignored
time-resolved spectroscopy. when drawing an analogy between impurity centers, on the one
Existing attempts to interpret [77–79, 87, 88] the aforemen- hand, and quantum dots, on the other hand. To adequately de-
tioned experiments on the basis of the adiabatic theory of mul- scribe the multiphonon optical spectra of quantum dots, it is neces-
tiphonon transitions in deep centers by Pekar [89] and Huang and sary [101] to develop the nonadiabatic approach. This is the main
Rhys [90] meet considerable difficulties. In the framework of the goal of the present subsection.
adiabatic theory, the values of the Huang–Rhys parameter were
obtained in Refs. [87] and [88] using a spherical model Hamilto-
nian for the electron-hole pair in a quantum dot [91–93] and taking 2.5.1. Light Absorption by Quantum Dots
into account the valence-band mixing. The values of the Huang– Quantum dots are considered to be embedded in a medium with
Rhys parameter obtained in this way appear to be significantly (by a weak dispersion of refractive index. In this subsection, the quan-
1 to 2 orders of magnitude) smaller than those derived from exper- tum dots are supposed to be identical. On the basis of the Kubo
iment. This discrepancy is due to the fact that under a strong con- formula [102], the linear coefficient of absorption by an ensemble
finement the charge density of the electron-hole pair in the ground of quantum dots in the electric dipole approximation [103] can be
state is small everywhere in the quantum dot. To achieve agree- represented as
ment with the experimental data, additional mechanisms are in-  !
tuitively introduced, which ensure separation of the electron and 8πωl N  h̄ωl 
hole charges in space: (i) additional built-in charge [79, 87, 94], α(ωl ) = 1 − exp −
3h̄cn(ωl ) kB T
(ii) traps that would localize holes [75], and (iii) different bound-
Z ∞ h i
ary conditions for electrons and holes [80].
The main idea of the adiabatic approach consists in the as- × Re dt exp i(ωl + iδ)
0
sumption that a stationary state of charge carriers is formed for * ! ! +
i b ˆ+ i b ˆ
each instantaneous position of ions (i.e., charge carriers follow the × exp Ht d exp − Ht d ; δ → 0+ (165)
ion motion adiabatically). In the framework of this approach, the h̄ h̄
exciton-phonon interaction leads only to a modification of the exci-
where ωl is the light frequency, N is the concentration of quantum
ton wave function, but does not give rise to transitions between dif-
ferent exciton states. However, two circumstances are to be men- dots, n(ωl ) is the refractive index of the medium, dˆ is the dipole
tioned, which imply that the exciton-phonon systems in quantum moment operator, and h i denotes the averaging over the statistical
dots are essentially nonadiabatic. ensemble of the exciton-phonon systems. The total Hamiltonian of
the exciton-phonon system,
(i) The states of an exciton in a quantum dot including the
b=H
H bph + H
bex + H bint (166)
ground state are, generally speaking, degenerate. In this
connection, it is worthwhile to recall, [95–99] that the contains the exciton Hamiltonian H bex , contains the phonon
electron-vibrational interaction in the impurity centers Hamiltonian Hbph , and contains the exciton-phonon interaction
with a degenerate electron level may cause a dynamic Hamiltonian, X
(γ̂λ b̂λ + γ̂λ+ b̂+
Jahn–Teller effect [100], or, equivalently, so-called bint =
H λ) (167)
internal nonadiabaticity. Namely, if the electron
λ
interaction with some vibrational modes is described by
noncommutative matrices calculated in the basis of a where b̂+
λ and b̂λ are the creation and annihilation operators for
degenerate electron level (those vibrational modes are the phonons of the λth vibrational mode, the interaction operators
usually called the nonadiabatic or Jahn–Teller modes), γ̂λ , are specified later.
there is a nonadiabatic mixing of electron states Note that in the experiments [74–78] the exciton energy is
belonging to this level. In a direct analogy to this much larger than both the phonon energy and the value kB T .
 OPTICAL PROPERTIES IN NANOSTRUCTURES 29

This means, in particular, that the probability of thermal gener- where ωβ is the Franck–Condon frequency of generation of an
ation of an exciton is vanishing. For optical transitions leading ˆ is the dipole matrix element
exciton in the state |βi; dβ ≡ hβ|d|0i
to generation of an exciton and starting from the exciton vac- of a transition between the exciton vacuum state and the state |βi,
uum |0i whose energy is chosen as zero, the initial states of the and
exciton-phonon system are described by the wave functions |0i|ni X D E
h iph ≡ ρn n||n (174)
where |ni are eigenstates of the phonon Hamiltonian H bph . Fur-
n
ther, the final states of the exciton-phonon system are not occu-
pied, so that in Eq. (165), one can omit the term that is propor- denotes the averaging over the phonon ensemble.
tional to exp(−h̄ω/kB T ) and describes a correction due to stimu- Using Eq. (167) and neglecting the dispersion of the frequency
lated emission of light with frequency ω. Finally, the frequency of of the optical phonons, one obtains the following result of the av-
the exciting radiation is much larger than the phonon frequency; eraging over the equilibrium phonon ensemble in Eq. (173),
hence the interaction of this radiation with free phonons can be ( Z Z t1
D E 1 X t h
ˆ
neglected, i.e., d|ni ˆ Hence, the transformation of the av-
= |nid. Û(t) = T̂ exp − dt1 dt2 (n̄ + 1)
ph 2
h̄ λ 0 0
erage,
h i
* ! ! + × exp −iω0 (t1 − t2 ) γ̂λ (t1 )γ̂λ+ (t2 ) + n̄
i b ˆ+ i b ˆ
exp Ht d exp − Ht d )
h̄ h̄ h i i
+
*  !* ! + + × exp iω0 (t1 − t2 ) γ̂λ (t1 )γ̂λ (t2 ) (175)
X  i b ˆ 
 i b ˆ+
= ρn nexp H t 0|d exp − Ht d|0 n
 h̄ ph h̄  where
n
*  *  ! h i−1
D E  X  n̄ = exp(h̄ω0 /kB T) − 1 (176)
  i b
= 0|dˆ+ |β β ρn nexp Hph t
  h̄
n The absorption coefficient α(ωl ) of Eq. (173) with Eq. (175) can
! +! +
i b  

D
ˆ
E be analytically calculated at arbitrary exciton-phonon coupling
× exp − Ht n β0 β0 |d|0 (168) when (i) the exciton levels are nondegenerate (this case was an-
h̄  
alyzed by Pekar [89] and Huang and Rhys [90]) or (ii) the vibra-
where |βi labels eigenstates of the exciton Hamiltonian, ρn is the tional modes interacting with an exciton are adiabatic, i.e., the ma-
probability of the phonon state |ni in the equilibrium statistical trices |hβ|γ̂λ |β0 i| in the basis of the degenerate level can be simul-
ensemble of the phonon systems, taneously diagonalized (this case is referred to as a static Jahn–
! Teller effect [95–98].
1 εn However, these analytical approaches cannot be applied for the
ρn = exp − (169)
Zph kB T exciton-phonon systems in quantum dots because these systems
are essentially nonadiabatic, as already stated before. The fact that
Zph is the partition function of the free phonon subsystem, and the exciton-phonon interaction in quantum dots is weak enables
εn is the energy of the phonon state |ni. Using the well-known one to adequately describe the effect of the nonadiabaticity on
Feynman relations [104] for operator exponentials, one obtains phonon-assisted optical transitions. Here, we will concentrate on
! ! ! the intensity of phonon satellites in the optical spectra.
i b i b i b
exp H t exp − Ht = exp − Hex t Û(t) (170) Under the condition of weak exciton-phonon interaction,
h̄ ph h̄ h̄
max |hβ|γ̂λ |β0 i|
where the evolution operator, η≡ 1 (177)
h̄ω0
Z t !
b =T
U(t) b exp − i b (t )
dt H (171)
contributions to the absorption coefficient from transitions, ac-
h̄ 0 1 int 1 companied by a change of the number of phonons by K, can be
calculated to leading (Kth) order in the small parameter η2 . To
b,
is represented in terms of the chronological ordering operator T realize this approximation, it is enough to keep only those terms
while in the expansion of the evolution operator (175) where all the in-
" # teraction operators γ̂λ (t) are placed either before or after all the
i b 
b
A(t) = exp b
Hex + Hph t operators γ̂λ+ (t). Hence, the absorption coefficient (173) takes the
h̄ form,
" #
 
×A bexp − i H bph t
bex + H (172) α(ωl )
h̄ 8πωl N
=
b in the interaction representation. Hence, 3h̄cn(ωl )
denotes the operator A 
Eq. (165) takes the form, X ∞ X (+)
Z ∞ ×Im (n̄ + 1)K ϕK (β−K ; : : : ; βK ; ωl )
8πωl N X h i 
α(ωl ) = Re ∗d 0
dβ dt exp i(ωl − ωβ + iδ)t K=0 β−K ;:::;βK
3h̄cn(ωl ) β
0

β;β0 ∞
X X 
D D E  E + n̄K
(−)
ϕK (β−K ; : : : ; βK ; ωl ) (178)
 b  0 
× β U(t) β (173)
ph K=1 β−K ;:::;βK
30 RIERA ET AL.

where βi with i = −K; : : : ; K label various intermediate and fi- coefficient on the frequency interval ($ν − 1ω/2; $ν + 1ω/2,
nal exciton states relevant to the K-phonon optical transition; the Z $ν +1ω/2
1
other denotations are αν = α(ω0 )dω0 (185)
1ω $ν −1ω/2
(±)
ϕK (β−K ; : : : ; βK ; ωl )
= ξK (β−K ; : : : ; βK ) can be expanded in powers of η2 because the integration in
Eq. (185) eliminates the problem of the previously mentioned
Y
K
1 divergences of the phonon-line intensities. The leading terms of
× (179)
ωβk ± (K − |k|)ω0 − ωl − iδ such an expansion are determined by the residua of the func-
k=−K (±)
tions ϕK in the poles ωl = ωβ0 ± Kν ω0 + iδ, the frequen-
ν
ξ0 (β0 ) = |dβ0 |2 (180) cies ωβ0 satisfying the inequality (183) at K = Kν . As results
from the definition of the parameter Kν , all the other 2K, poles
(±)
ωl = ωβk ±(Kν − |k|)ω0 +iδ (with |k| ≥ 1) of the function (ϕK )
ξK (β−K ; : : : ; βK ) ν
∗ lie far away from the interval ($ν − 1ω/2; $ν + 1ω/2) and cor-
dβ d
−K βK respond already to other groups of spectral lines. In combination
=
h̄2K with inequalities (183) and (184), this leads to the following prop-
(±)
X
K K h
Y i erty of residua of the functions ϕK (179): The residua determin-
× 1 − δni ni0 (1 − δii0 ) ing the leading terms in the average absorption coefficient αν , con-
n1 ;:::;nK =1 i;i0 =1 tain in their denominators no small factors that would obey the in-
K XD ED E equalities |ωβ − ωβ0 − kω0 | ≤ 1ω (where k = 1; 2; 3; : : :). This
Y
× β−k |γ̂λ |β−k+1 βnk −1 |γ̂λ+ |βnk K ≥ 1 (181) property implies, consequently, that the contributions to the aver-
k=1 λ age absorption coefficient due to the residua that contain small de-
nominators of the aforementioned type can be neglected. There-
When the inequalities, fore, to correctly (within a relative accuracy ≈ η2 ) describe the val-
ues of the average absorption coefficient on the narrow frequency
|ωβ − ωβ0 − kω0 |  ηω0 (182)
intervals ($ν −1ω/2; $ν +1ω/2), the function α(ωl ) (178) should
are satisfied for any |βi; |β0 i, and k = 1; 2; 3; : : :, the intensities of be represented as
absorption lines can be expanded in powers of parameter η2 . To ( ∞
8π 2 ωl N X X (+)
the leading (Kth) order in this parameter, the intensity of a line at α(ωl ) = (n̄ + 1)K fβK δ(ωβ + Kω0 − ωl )
the frequency ωl = ωβ0 ± Kω0 is determined by the residuum of 3h̄cn(ωl )
β K=0
(±) )
the function ϕK in the pole ωl = ωβ0 ± Kω0 + iδ. The residuum X∞
−
of the same function [see Eq. (179)] in each of the other 2K poles + n̄K fβK δ(ωβ − Kω0 − ωl ) (186)
ωl = ωβk ± (K − |k|)ω0 + iδ with |k| ≥ 1 describes a relatively K=1
small correction ≈ η2K to the (K − |k|)-phonon-line intensity de- (+) (−)
Here the coefficients fβK (fβK ) defined by
termined by its leading term, which is of the order of η2(K−|k|) .
Under the resonance conditions |ωβ − ωβ0 − kω0 | → 0 (±)
fβ0 = |dβ |2 (187)
(where k = 1; 2; 3; : : :), the phonon-line intensities calculated
(±)
X
within the framework of perturbation theory may diverge and fβK = δβ0 β ξK (β−K ; : : : ; βK )
hence a straightforward expansion of the absorption coefficient β−K ;:::;βK
α(ωl ) in powers of η2 is not adequate. The following procedure
also allows one to study the resonance situation. The resonance Y
K Y(|ω
ββk ± kω0 | − 1ω)Y(|ωββ−k ± kω0 | − 1ω)
×
situation is noticeable for the fact that the optical spectra contain (ωββk ± kω0 )(ωββ−k ± kω0 )
k=1
groups of close spectral lines. Distinct phonon-line progressions
can then be observed in the optical spectra of quantum dots pro- K≥1 (188)
vided that (i) for the frequencies of optical transitions correspond- are proportional to the probabilities for the generation of an exci-
ing to one and the same νth group, with $ν , denoting the central ton in the state |βi with simultaneous emission (absorption) of K
position of spectral lines of this group, the inequalities, phonons,
n
1 x≥0
|ωβ ± Kω0 − $ν | < 1ω/2 Y(x) = (189)
(183) 0 x<0
1ω  ω0 ; K = 0; 1; 2; : : :
ωββ0 = ωβ − ωβ0 (190)
are fulfilled and (ii) the groups are separated from each other by In Eq. (188), there are three kinds of terms which contain nondi-
relatively large frequency intervals, agonal matrix elements hβi |γ̂λ |βk i in the coefficients ξK : (i) terms
|$ν0 − $ν |  1ω ν0 6= ν (184) with all the states |β−K i; : : : ; |βK i belonging to the same energy
level, which describe the influence of the internal nonadiabaticity
For the νth group, the integer parameter K obeying the inequal- (or the proper Jahn–Teller effect) on the optical transition prob-
ity (183) possesses a minimum value that is denoted as Kν . If the abilities; (ii) terms with |β−K i; : : : ; |βK i pertaining to different
interval 1ω is chosen to be much larger than ηω0 (such a choice energy levels, which take into account the effects of the external
is always possible under condition (177)), the average absorption nonadiabaticity (or the so-called pseudo-Jahn–Teller effect); and
 OPTICAL PROPERTIES IN NANOSTRUCTURES 31

∗
(iii) terms proportional to dβ d with β−K 6= βK which de- where the partition function Z relates to the states of the system
−K βK
scribe the intermultiplet mixing between the exciton states with that includes one exciton and phonons. The concentration of ex-
the same symmetry. This mixing occurs due to the exciton-phonon cited quantum dots Nex is described by
interaction. sα(ωexc )`exc
When an exciton energy level is degenerate, Eq. (188) differs Nex = (194)
h̄ωexc
from the result of the adiabatic theory, [89, 90] which gives for a
weak exciton-phonon interaction, where ωexc and `exc are the frequency and the intensity of the
   2 K monochromatic exciting radiation; the factor s = τ for the con-
|dβ |2 X  hβ|γ̂λ |βi  tinuous excitation, while s = τexc exp(−t/τ) for the excitation by
f˜βK =     (191) short light pulses of duration τexc  τ. Substituting the equation
K!  h̄ω0 
λ for α (186) in Eqs. (193) and (194), one obtains at low tempera-
the difference taking place even if one neglects the effects men- tures typical for the experiments [74–78] (kB T  h̄ω0 ),
tioned in points (ii) and (iii) earlier. For example, the shape of the ∞ X
X
absorption band edge of a spherical quantum dot can be obtained I(ωl ) = N δ(ωl − ωexc + Kω0 + ωβ0 β )
using the spherical model [91–93] for the exciton Hamiltonian (see K=0 β;β0
(±)
Subsection 2.5.5). The coefficients fβK that are proportional to X
K
the probabilities of generation of an exciton in the ground state × Aββ0 KK 0 δ(ωexc − ωβ0 − K 0 ω0 ) (195)
with emission or absorption of K-phonons of nonadiabatic (or, K 0 =0
equivalently, Jahn–Teller) modes are then given by
where
(±)
" # " #  !
fβK 4 K + 1 1  K + 1 3 32π 2 ω4l s`exc h̄ωβ (−)
= + + (+)
(192) Aββ0 KK 0 = exp − f 0f 0 0 (196)
f˜βK 3 2 2 2 2 9h̄2 c 4 Zex kB T β;K−K β K

where [x] denotes the integer part of x. The adequate account with Zex the partition function of the one exciton subsystem.
of the exciton interaction with the Jahn–Teller vibrations leads When deriving Eq. (196) it was taken into account that within the
to a significant increase of the multiphonon transition probabil- framework of the weak coupling approach developed in the pre-
ities and to a more complicated dependence of these probabili- vious subsection, Z ≈ Zex Zph . In the limit of slow relaxation
ties on the number of emitted or absorbed phonons in compari- τ0  τ, the recombination occurs from a state in which an ex-
son with the results of the theory by Pekar [89] and Huang and citon is generated, hence the photoluminescence intensity follows
Rhys [90]. Therefore, the multiphonon optical spectrum deter- from Eq. (186) to be
mined by Eq. (186) with Eq. (192) is considerably different from ∞
X
the Franck–Condon progression described by Eq. (186), where the I(ωl ) = N δ(ωl − ωexc + Kω0 )
fβK are replaced by the f˜βK given through Eq. (191).
(±)
K=0
X X
K
× BβKK 0 δ(ωexc − ωβ − K 0 ω0 ) (197)
2.5.2. Photoluminescence Spectrum
β K 0 =0
The relaxation processes during the time interval t between the
where
generation and the recombination of an exciton substantially in-
fluence the photoluminescence spectrum. Two limiting cases are 32π 2 ω4l sβ `exc (−) (+)
BβKK 0 = fβ;K−K 0 fβK 0 (198)
examined here: (i) the thermodynamic equilibrium photolumines- 2
9h̄ c 4
cence that takes place when τ0 the time of the relaxation between
with the factors sβ described at τexc  τ by
the exciton energy levels is much smaller than the radiative life-
time of an exciton τ and (ii) the opposite case relevant to slow sβ = τexc
relaxation τ0  τ. The energy level broadening due to the fi- ( ∞
)
nite values of τ0 and τ is disregarded here, i.e., the inequalities 4t X (−)
× exp − (ωβ − Kω0 )3 n(ωβ − Kω0 )fβK (199)
τ0−1 , τ−1  ω0 are supposed to be satisfied. This supposition is h̄c 3
K=0
in agreement with the theoretical estimations of the exciton life-
time (for example, the value τ ≈ 1 ns was obtained in Ref. [77] for For a specimen that contains quantum dots of different types, the
CdSe quantum dots) and with the experimental observation of ul- photoluminescence intensities (195) and (197) must be summed
tranarrow (< 0:15 meV) luminescence lines from single quantum up over those different types. When one considers an ensemble
dots [105]. The spectral broadening of the multiphonon photolu- of spherical quantum dots with a continuous distribution function
minescence lines is considered in Subsection 2.5.4. N(R) over the radii R, the aforementioned summation is realized
According to Refs. [97] and [106], the spectrum of equilibrium by an integration over radii. For the equilibrium photolumines-
luminescence can be obtained directly from the absorption spec- cence intensity, one then obtains from Eq. (195),
trum. For an ensemble of identical quantum dots, the equilibrium ∞ X X
X K
luminescence intensity can be expressed as I(ωl ) = N(Rβ0 K 0 )Aββ0 KK 0 (Rβ0 K 0 )
! K=0 β;β0 K 0 =0
h̄ω3l n2 (ωl ) Nex Zph h̄ωl  
I(ωl ) = exp − α(ωl ) (193)
2π 2 c 2 N Z kB T × δ ωl − ωexc + Kω0 + ωβ0 β (Rβ0 K 0 ) (200)
32 RIERA ET AL.

shape close to spherical. For the samples investigated in Refs. [76]

and [77], small angle X-ray measurements indicate that the func-
tion N(R) that describes the distribution of quantum dots over
radii is close to the logarithmic standard distribution,
 !2 
Ntot  ln(R/R0 ) 
N(R) = √ exp − √ (203)
2πυR  2υ 

the average radius of quantum dots being determined as

!
υ2
hRi = R0 exp (204)
2

Here R0 and υ are distribution parameters and Ntot is the total

concentration of quantum dots,
Z ∞
Ntot = N(R)dR (205)
0

Fig. 25. Scheme of the transitions for the light absorption under the ex-
In the next subsection, the function N(R) (203) is used to model
citation by the radiation of frequency ωexc (arrows up) and for the sub- the distribution of quantum dots over radii.
sequent luminescence (arrows down) in quantum dots of three different In the exciton-phonon interaction Hamiltonian Hint (167), the
radii. Thin dashed lines indicate a phonon-line progression occurring in operators γ̂λ are
the photoluminescence spectrum after the exciton relaxation from the state γ̂λ = γ̂λ (re ) − γ̂λ (rh ) (206)
into the state |β0 i. The values ω0 and ωβ0 − ωβ are shown magnified as
compared to the frequency ωexc . where re and rh are the coordinates of an electron and a hole,
respectively. In a spherical quantum dot, the interaction operators
γ̂λB (r) and γ̂λl (r) describing the electron interaction, respectively,
while in the opposite limit τ0  τ the photoluminescence intensity with the bulk and interface phonons are expressed according to
of Eq. (197) takes the form, Ref. [107] by
∞
X γ̂λ (r) ≡ γ̂nlm (r) = vnl (r)i−l−m−|m| Ylm (θ; ϕ)
(B) (B) (B)
(207)
I(ωl ) = δ(ωl − ωexc + Kω0 )
γ̂λ (r) ≡ γ̂lm (r) = vl (r)i−l−m−|m| Ylm (θ; ϕ)
K=0 (l) (l) (l)
(208)
XK X
× N(RβK 0 )BβKK 0 (RβK 0 ) (201) where Ylm (θ; ϕ) are the spherical harmonics. The radial parts of
the interaction operators are described by
K 0 =0 β
v !
u
The monochromatic excitation light with frequency ωexc is ab- u h̄ω0 1 1 jl (µnl r/R)
sorbed only by quantum dots of radii Rβ0 K 0 satisfying the equa-
(B)
vnl (r) = e t − (209)
κ0 R κ(∞) κ(0) µnl jl0 (µnl )
tion,
v
ωβ0 (R) + K 0 ω0 − ωexc = 0 (202) u !
u h̄ω0 1 1
(l)
vl (r) = e t −
The photoluminescence spectrum described by Eq. (201) contains κ0 R κ(∞) κ(0)
only one phonon-line progression with a zero phonon line at the
√ !l
frequency ωl = ωexc , while the equilibrium photoluminescence lκ(∞) r
spectrum consists of a set of phonon-line progressions shifted from × (210)
lκ(∞) + (l + 1)κ̃(∞) R
ωexc by the values ωββ0 (190), which are determined by exciton
energy losses due to the radiationless relaxation. The schemes of where jl (z) are the spherical Bessel functions, µnl is the nth zero
the relevant optical transitions in quantum dots are shown in Fig- of the function jl (z), κ0 is the permittivity of vacuum, κ(∞) and
ure 25. In particular, the equilibrium photoluminescence line at κ̃(∞) are the optical dielectric constants, respectively, of a quan-
the frequency ωl = ωexc − Kω0 − ωβ0 β (Rβ0 K 0 ) corresponds to
tum dot and of its surrounding medium, and κ(0) is the static di-
the following sequence of transitions in a quantum dot of radius
electric constant of a quantum dot.
Rβ0 K 0 (e.g., of radius Rβ0 2 in Fig. 25): (i) generation of an exciton
The exciton Hamiltonian for the spherical model [91–93] sup-
in the state |β0 i accompanied by emission of K 0 (K 0 ≤ K) phonons plemented to account for the electron-hole exchange interaction
(shown by arrow up), (ii) radiationless transition of the exciton to [108, 109] is
the state |βi (heavy dashed arrow), and (iii) radiative recombina-
tion of the exciton with emission of K − K 0 phonons (thin dashed bex = 1 2 γ γ
H p̂ + 1 p̂2 − 2 (p̂(2) · Ĵ (2) )
arrows for transitions with emission of zero to three phonons). 2m∗ e 2m0 h 9m0
2
+ Vc (re ; rh ) − εexch a30 δ(re − rh )(σ̂ · Ĵ) (211)
2.5.3. Models for Quantum Dots 3
Both the quantum dots in glass and the colloidal quantum dots Here, p̂e and p̂h are the momenta of an electron and a hole, re-
of small size (1–2 nm) are known [75–78] to have a geometrical spectively, γ1 and γ2 are the Luttinger parameters, p̂(2) and Ĵ (2)
 OPTICAL PROPERTIES IN NANOSTRUCTURES 33

are the irreducible second-rank tensor operators of the momen-

tum and the spin-3/2 angular momentum of a hole, σ̂ and Ĵ are the
spin operators of an electron and a hole, a0 is the lattice constant,
and εexch is the exchange strength constant, [110], which is equal to
320 meV in CdSe [109]. In a spherical quantum dot, the Coulomb
electron-hole interaction may be approximately described by the
Hamiltonian [107],
Vc (re ; rh )
e2
=−
4πκ0 κ(∞)
∞
(
X rel rhl
× Pl (cos θeh ) Y(rh − re ) + Y(re − rh )
l=0 rhl+1 rel+1 Fig. 26. Lowest energy levels of an exciton in a spherical CdSe quantum
) dot as a function of the radius. (a) Splitting of the energy levels due to
(re rh )l [κ(∞) − κ̃(∞)](l + 1) the electron-hole exchange interaction in quantum dots with zinc-blende
+ (212)
R2l+1 κ(∞) + κ̃(∞)(l + 1) (cubic) structure. (b) Splitting of the energy levels due to both the exchange
interaction and the anisotropic crystal field in quantum dots with wurtzite
where Pl (x)is a Legendre polynomial of degree l and θeh is the (hexagonal) structure. The degenerate level (1S; 1S3/2 ) is chosen as the
angle between re and rh . The Franck–Condon frequency for gen- origin of the energy scale.
eration of an exciton in the eigenstate |βi is determined by the
corresponding eigenvalue εβ of the Hamiltonian (211) as
εg + εβ into five sublevels with definite values of the exciton angular mo-
ωβ = (213) mentum components Nz along the crystal axis, as indicated in Fig-
h̄
ure 26(b): a twofold degenerate level with |Nz | = 2 (the curve
where εg is the energy bandgap for the substance of the quantum with the label 2), two twofold degenerate levels with |Nz | = 1
dot. (two curves with the labels 1L and 1U for the lower and up-
For typical [75–77] quantum dot radii, the exciton states are per sublevels, respectively), and two nondegenerate levels with
determined mainly by confinement, while the electron-hole inter- |Nz | = 0 (two curves with the labels 0L and 0U ). The splitting
action can be treated as a perturbation. To zeroth order in this per- of the level (1S; 1P3/2 ) characterized by relatively smaller energy
turbation, the hole states [93], designated as nS3/2 , nP1/2 , nP3/2 , intervals between sublevels [see Fig. 26(b)], while the mutual or-
nP5/2 , nD1/2 , nD5/2 , : : : are characterized by definite values of der of sublevels with different |Nz | is the same as that for the level
the hole angular momentum F = L + J (where L is the orbital (1S; 1S3/2 ).
angular momentum of a hole) and the parity; n labels the solu- In quantum dots with both the zinc-blende structure and the
tions of the equations for the radial components of the hole wave wurtzite structure, the zero-phonon generation of an exciton in
function [93]. the ground state and the radiative recombination from this state
Under size selective excitation, the frequency ωexc is chosen to are forbidden in the dipole approximation. Therefore, the inten-
lie in the absorption band tail. Thus, only the largest quantum dots sity of zero-phonon line is due to transitions from high energy lev-
are involved in the absorption processes and excitons are gener- els. As a consequence, on the basis of Eq. (200) combined with
ated in the lowest states. The exciton ground state (1S; 1S3/2 ) and Eq. (196), a decrease of temperature must result in a decrease
the state (1S; 1P3/2 ) separated from the ground state by an energy of the zero-phonon line intensity in the equilibrium luminescence
comparable to the phonon energy h̄ω0 play the main role in the spectra. Such a behavior is in agreement with the experimental
absorption and the photoluminescence. (As for an electron, the data [75].
energy of its size quantization is large and only the state 1S must
be taken into account.)
In a spherical quantum dot, both the energy level (1S; 1S3/2 ) 2.5.4. Numerical Results and Comparison with the
and the level (1S; 1P3/2 ) are eightfold degenerate to the zeroth or- Experiment
der in electron-hole interaction. In a quantum dot with zinc-blende In Figures 27 and 28, the fluorescence spectra I(ωl ) calculated by
structure, the electron-hole exchange interaction splits [108, 109] using Eq. (200) are presented together with the experimental data
each of these levels into a fivefold degenerate level with eigenvalue of Ref. [75] on colloidal CdSe quantum dots of wurtzite structure,
N = 2 of the exciton angular momentum N = σ + F and a three- respectively, at different temperatures T and average radii hRi.
fold degenerate level with N = 1 [see Fig. 26(a)]. In a quantum The following values of the parameters were used: κ(∞) = 6:23,
dot with wurtzite structure, the crystal field, described by the po- κ(0) = 9:56, m∗e = 0:13m0 , and εg = 1:75 eV from Ref. [87] and
tential [110], ! γ1 = 2:04 and γ2 = 0:58 from Ref. [112]. From a comparison
1 2 1 of the calculated absorption spectrum (186) with the experimental
V =− J − (214)
2 z 4 absorption spectrum of Ref. [75], the value 0.06 was obtained for
the parameter υ of the distribution function N(R) (203). Because
(where 1 = 25 meV for CdSe (Ref. [111]) and Jz is the spin an- of the chaotic nature of the energy spectrum of excitons in real
gular momentum component of a hole along the unique axis of quantum dots (e.g., owing to the strain or nonperfect geometrical
hexagonal lattice), together with the electron-hole exchange in- form) and also due to a limited spectral resolution of experimental
teraction, leads [78, 109] to the splitting of the level (1S; 1S3/2 ) equipment, photoluminescence lines describes by δ-functions turn
34 RIERA ET AL.

Fig. 29. Fluorescence spectra of CdSe quantum dots with wurtzite struc-
ture at the average radius hRi = 1:25 nm. Dot-dashed line displays a
Franck–Condon progression with the Huang–Rhys parameter S = 0:06
calculated in Ref. [75], the dotted line shows another Franck–Condon pro-
gression with the Huang–Rhys parameter S = 1:7, which is obtained by
fitting the ratio of one-phonon and zero-phonon peak heights to the exper-
imental value, the solid line results from the calculation by using Eq. (200)
and circles are experimental values taken from Ref. [87]. T = 1:75 K.
(Reprinted from M. Nirmal et al., Z. Phys. D, 26, 361, (1993); N. Nomura
Fig. 27. Fluorescence spectra of CdSe quantum dots with wurtzite struc- and I. Kobayashi, Phys. Rev. B, 45, 1305, (1992).)
ture at various temperatures. Solid lines were calculated by using Eq. (200)
and dashed lines represent the experimental data of Ref. [2]. Different
lines are shifted along the vertical axis for clarity. hRi = 1:2 nm. (Reprinted
from R. A. Stradling and P. C. Klipstein, “Growth and Characterisation of crystal field for a wide range of radii R is close to the optical
Semiconductors,” Hilger, Bristol, U.K., (1990).) phonon energy h̄ω0 = 25 meV as seen from Figure 26(b). There-
fore, the observed peak of the equilibrium photoluminescence
in the spectral region ω ≈ ωexc − Kω0 is caused not only by
K-phonon processes (with probability ≈η2K ) but also by (K − 1)-
phonon processes (with probability ≈ η2(K−1) ). The latter pro-
cesses are related to the generation of an exciton in the upper
group of states (resulting from the splitting of the (1S; 1S3/2 )
level) and the subsequent radiationless relaxation to the lower
group of states. This feature of the energy spectrum of the exciton-
phonon system in quantum dots, in combination with the pseudo-
Jahn–Teller effect, leads to a substantial different of the observed
spectrum of multiphonon photoluminescence from the Franck–
Condon progression, as shown in Figure 29. It is important to note
that a straightforward calculation in the framework of the adia-
batic theory leads to intensities of the phonon satellites that do
not fit the observed spectrum well for any value of the Huang–
Rhys parameter.
In Figure 30(a)–(d), the photoluminescence spectra I(ωl ) cal-
culated for a zinc-blende-type crystal lattice are given and are com-
pared with the experimental data of Ref. [77] on photolumines-
cence of CdSe quantum dots in borosilicate glass. The parameter
values m∗e = 0:11m0 and εg = 1:9 eV (Ref. [113]) were used for
Fig. 28. Fluorescence spectra of CdSe quantum dots with wurtzite struc- the zinc-blende lattice. The theoretical curves are calculated for
ture at various average radii hRi. Solid lines were calculated by using the time interval between generation and recombination of an ex-
Eq. (200) and dashed lines represent the experimental data of Ref. [2]. citon t  τ. The shape of photoluminescence peaks was modeled
Different lines are shifted along the vertical axis for clarity. T = 1:75 K. by Lorentzians of width 0 = 2 meV relevant to the limited spec-
(Reprinted from R. A. Stradling and P. C. Klipstein, “Growth and Charac- tral resolution of measurements (better than 5 meV in Ref. [77]).
terisation of Semiconductors,” Hilger, Bristol, U.K., (1990).)
From a comparison of the calculated absorption spectra (186) with
the experimental absorption spectra of Ref. [77], the values 0.15,
0.18, and 0.20 of the parameter υ were obtained for average radii
into broadened peaks. The shape of these peaks was modeled by hRi equal to 1:4, 1.8, and 2.7 nm, respectively. The experimental
Lorentzians of finite width 0 = 15 meV. photoluminescence spectra at each value of the average radius hRi
It is worthwhile to note that in CdSe quantum dots of wurtzite refer to different time intervals between the pumping pulse (of du-
structure, the magnitude of the (1S; 1S3/2 )-level splitting by the ration 2.5 ps) and the measurement, the upper curve correspond-
 OPTICAL PROPERTIES IN NANOSTRUCTURES 35

tion by light whose frequency lies near the absorption band max-
imum (no size selective excitation). In this case, a significant shift
of the photoluminescence intensity maximum from the excitation
frequency appears 10 ps after the excitation pulse [77]. Therefore,
it is the limiting case of the equilibrium photoluminescence rather
than the opposite case of τ0  τ that seems to be relevant to the
experimental data of Ref. [77].
The quantum dots in glass are characterized by broad distribu-
tion functions over the radii R [85, 86]. Therefore, even under size
selective excitation some excitons are generated in states with rela-
tively high energy (first of all, in the states (1S; 2S3/2 ), (1S; 1P5/2 ),
(1S; 1D5/2 ) whose energies differ slightly from each other). In
the case of equilibrium photoluminescence, the radiationless re-
laxation of such excitons into the lowest states (1S; 1S3/2 ) leads to
the appearance of zero phonon luminescence peaks shifted from
ωexc to lower frequencies. These peaks hide the phonon satellites
(see Fig. 30(c)). This effect becomes less pronounced when the
excitation is deeper in the absorption band tail (see Fig. 30(d)).
We show along of Subsection 2.5 that, due to the nonadiabatic-
ity of the exciton-phonon system in spherical quantum dots, differ-
ent channels of the phonon-assisted optical transitions open. First,
distinct from the adiabatic theory, which describes only phonon
transitions through intermediate exciton states coinciding with ei-
ther its initial or its final state, the present approach takes into con-
sideration also additional phonon transitions: (i) between differ-
ent exciton states belonging to the same degenerate level (proper
Jahn–Teller effect) and (ii) between exciton states of different en-
ergy (pseudo-Jahn–Teller effect). Second, in contrast to the works
based on the theory of Pekar and of Huang and Rhys, which ad-
equately account for the adiabatic phonons only, the present ap-
proach enables one to correctly describe the transitions involving
all types of phonons, including Jahn–Teller phonons.
The effect of the channels studied here leads to a consid-
erable enhancement of the phonon-assisted transition probabil-
ities in the photoluminescence of quantum dots even with rel-
atively weak electron-phonon coupling strength. The resulting
Fig. 30. Photoluminescence spectra of CdSe quantum dots embedded multiphonon optical spectra are considerable different from the
in borosilicate glass at different average radii and excitation energies: Franck–Condon progression.
(a) hRi = 1:4 nm, h̄ωexc = 2:5 eV; (b) hRi = 1:8 nm, h̄ωexc = 2:28 eV;
(c) hRi = 2:7 nm, h̄ωexc = 2:0 eV; and (d) hRi = 2:7 nm, h̄ωexc = 1:95 eV.
Thin solid lines represent the families of experimental time-resolved pho- 2.5.5. Exciton-phonon System
toluminescence spectra of Ref. [77] measured at different time intervals
The shape of the absorption band edge of a spherical quantum
between the pumping pulse and the measurement, the upper curve cor-
responding to the time interval equal to 0. Theoretical results are dis-
dot is considered using the spherical model [91–93] of the exciton
played for the equilibrium photoluminescence spectra of Eq. (200) (heavy Hamiltonian. To the zeroth order in the electron-hole interaction
solid lines) and for the photoluminescence spectra of Eq. (201) in the case the exciton ground state is fourfold degenerate with respect to the
of slow relaxation of the exciton energy (dashed lines). (Reprinted from z-component of the hole angular momentum Fz , (see Ref. [93]).
V. Jongnickel and F. Henneberger, J. Lumin., 70, 238, (1996).) The twofold spin degeneracy of the electron ground state is not
important for the analysis of the Jahn–Teller effect [99]. The wave
functions of the exciton ground state can be written as [93],
 E  E
ing to the time interval equal to 0. According to Ref. [77], the de-  
|Fz i = ψ0 (re ) 12 ; σz f0 (rh )0; 32 ; 32 ; Fz
cay of fast photoluminescence components is caused by trapping  E
of holes onto deep (binding energy ≈ 200 meV) local surface lev- 
+ g0 (rh )2; 32 ; 32 ; Fz (215)
els. Depending on the average radius of the quantum dots, the
decay time varies from some tens to some hundreds of picosec- where ψ0 (re ) is the envelope wave function of an electron in the
onds. The separation between neighboring energy levels of a non- ground state, |1/2; σz i is the Bloch wave function at the bottom
trapped exciton is smaller than the depth of local levels. There- of the conduction hand, |L; J; F; Fz i are the eigenfunctions of the
fore, the relaxation in the system of interior exciton states, which total angular momentum of a hole, and L and J = 3/2 are the
were discussed before, can be expected to proceed more quickly eigenvalues of the orbital angular momentum of a hole and its
than the hole trapping. This supposition is in agreement with the spin angular momentum, respectively. An exciton in the ground
behavior of fast photoluminescence components under the excita- state interacts only with s phonons (l = 0) and d phonons (l = 2)
36 RIERA ET AL.

(see Eqs. (207)–(210)). Using Eq. (167) with Eqs. (206)–(210), one representation as
obtains bs (t)iph hU
b ph = hU bd (t)iph
hU(t)i (225)
∞
X
bint = Î
H
(B) (B)
cn0 ân00 where
Z t !
n=0 h i
! b exp −
bs (t) = Î T iC s +
X
2 ∞
X U dt b̂s (t1 ) + b̂s (t1 ) (226)
+
(I) (I)
V̂m c2 â2m +
(B) (B)
cn2 ân2m + H.c. (216) h̄ 0 1
Z !
m=−2 n=0 iCd t h i
b b
Ud (t) = T exp − b b +
dt B(t1 ) + B (t1 ) (227)
where h̄ 0 1
Z ∞ h i  b̂ √ √ 
(B) 1 (B) 2b̂d1 2b̂d2 0
cn0 = √ drvn0 (r) ψ20 (r) − f02 (r) − g02 (r) (217) √ d0 √
4π 0  2b̂ −b̂d0 0 2b̂ 
Z ∞ b =  √ d;−1
B √ d2  (228)
1  2b̂ −b̂ − 2b̂d1 
(B)
cn2 = − √
(B)
drvn2 (r)f02 (r)g02 (r) (218) d;−2 √ 0 √ d0
5π 0 0 2b̂d;−2 − 2b̂d;−1 b̂d0
Z ∞
(I) 1 (I)
c2 = − √ drv2 (r)f02 (r)g02 (r) (219) The interaction of an exciton with totally symmetrical s phonons
5π 0 is adiabatic and the corresponding interaction matrix is diagonal.
In Eq. (216) and later, bold letters denote operator matrices in Having averaged U bs (t) (226) over the phonon ensemble, one ob-
the basis of the functions |Fz i(Fz = 3/2; 1/2; −1/2; −3/2). The tains the Fourier transform of the Franck–Condon progression.
matrices The probability of a K-phonon optical transition is then propor-
1 0 0 0 tional to f˜βK (191) with |βi = |Fz i.
 0 −1 0 0  The average hU bd (t)iph is calculated by using the approach de-
V0 =  
0 0 −1 0 veloped in Subsection 2.5.1. Namely, contributions into the ab-
0 0 0 1 sorption coefficient due to transitions accompanied by a change
 0 √2 0 0 
of the number of d phonons by K are taken in account to the
leading (Kth) order in the small parameter η2 (177). At the tem-
+ 0 0 0 0 
√ peratures corresponding to the experimental conditions [74–78]
V1 = V−1 =  (220)
0 0 0 − 2 (kB T  h̄ω0 ), one obtains
0 0 0 0
!2K
0 0 √ D E X∞ D E
2 √0  b
Ud (t) =
1 Cd b+ )K
bK (B
B
+ 0 0 0 2 ph (K!)2 h̄ω0 ph
V2 = V−2 =  K=0
0 0 0 0 × exp(−iKω0 t) (229)
0 0 0 0
are relevant to the interaction of an exciton with d phonons. The Using the definition (228), the identity
matrices Vbm with different m do not commute. Therefore, the in- b2 = Î(b̂2 + 2b̂d1 b̂d;−1 + 2b̂d2 b̂d;−2 )
B (230)
teraction Hamiltonian H bint cannot be diagonalized in the basis of d0

the functions |Fz i (215) and, in contrast to the case of the adia- bd (t)iph
can be verified. Taking in account Eq. (230), the average hU
batic approach, the wave function of the exciton-phonon system (229) is easily calculated,
cannot be represented as a product of an exciton wave function
∞
!2K " # 
with a phonon wave function. This fact expresses mathematically D E X 1 C 4 K + 1 1
bd (t)
U = Î d  + 
the internal nonadiabaticity of the exciton-phonon system. ph K! h̄ω0 3 2 2
It is convenient to introduce the effective phonon modes with K=0
" # 
the second quantization operators, K+1 3
∞
× +  exp(−iKω0 t) (231)
1 X (B) (B) 2 2
b̂s = cn0 ân00 (221)
Cs
n=0 where [x] denotes the integer part of x. Equation (192) follows
∞
!
1 (I) (I)
X (B) (B)
immediately from Eq. (231).
b̂dm = c2 â2m + cn2 ân2m (222)
Cd
n=0
where 3. OPTICAL PROPERTIES IN NANOSTRUCTURES II
∞ 
!1/2
X 
(B) 2 The Raman scattering provides a powerful tool that permits the
Cs = cn0 (223)
investigation of several physical properties of the nanostructures.
n=0
∞ 
! In particular, the electronic structure of the semiconductors and
  X  1/2
(I) 2 (B) 2 nanostructures can be studied considering several polarizations for
Cd = c2 + cn2 (224)
the incident and emitted radiation. In connection with this type of
n=0
experiment, the calculations of the differential cross-section, for
Using Eqs. (216) and (220)–(222), the evolution operator (171) Raman scattering, remain rather like a fundamental and interest-
averaged over the phonon ensemble can be written in the matrix ing source to obtain a better understanding of the nanostructures
 OPTICAL PROPERTIES IN NANOSTRUCTURES 37

characterized by their mesoscopic dimensions. Between the differ-

ent processes of Raman scattering involved in this type of inves-
tigation, the electron Raman scattering is a useful technique that
provides direct information on the structure of energy bands and
optical properties of the investigated systems.
Interband electron Raman scattering processes can be quali-
tatively described in the following way [114]: an external photon
is absorbed from the incident radiation field and after that a vir-
tual electron-hole pair is created in an intermediate crystal state by
means of an electron interband transition involving the crystal va-
lence and conduction bands. The electron (hole) in the conduction
(valence) band is subject to a second intraband transition with the
emission of a secondary radiation photon. Therefore, in the final
state we have a real electron-hole pair in the crystal and a long- Fig. 31. Diagrams contributing to electron Raman scattering for inter-
wavelength photon of the secondary radiation field. The effects of band intersubband transitions. The final state is characterized by an elec-
external applied fields on these kinds of processes were investi- tron in the subband N1 with momentum h̄ke , a hole in the subband N2
gated in [115, 116]. Particularly, the case of an external magnetic with momentum h̄kh , and a secondary radiation photon of frequencyωs .
field with the electron (or hole) transitions between Landau sub- Diagrams (a) and (b) show contributions from the electron or the hole,
band was considered in [117]. respectively. Diagrams (c) and (d) show interference processes.
The fundamental aim of the present section is to analyze the
main features of electron Raman scattering emission and exci-
tation spectra in semiconductor nanostructures, where the elec-
tron (hole) system essentially bears a low-dimensional character
of the nanostructure. Due to electron confinement, the conduc-
tion (valence) band is split in a subband system and transitions be-
tween them determine the electron Raman scattering processes.
The general relations needed for calculating of the electron Ra-
man scattering differential cross-section in the “two-step” model
for the light scattering processes is presented. Only two bands are
considered (a valence and a conduction band), which are split in an
infinite subband system under the assumption of a nanostructure
with barriers of infinite height. The effective mass and envelope Fig. 32. Band diagrams contribute to electron Raman scattering. They
function approximations are also assumed and the emission and are equivalent to the diagrams of Figure 114(a) and (b), respectively.
excitation spectra are reported.

3.1. Model and Applied Theory the system, while εa and εb correspond to the system intermedi-
ate state energies. While 0a and 0b are the corresponding lifetime
The electron Raman scattering cross-section can be written in the broadenings, Hbl(s) is the interaction of the incident (emitted) ra-
form [118], diation field with the crystal which was chosen previously in the
d2σ V 2 ω2s η(ωs ) form,
= W(ωs ; es ) (232) r
dωs d 8π 3 c 4 η(ωl ) |e| 2π h̄
b
Hjl(s) = e · p̂ (235)
where c is the light velocity in vacuum, η(ω) is the material refrac- m Vω l(s)
tion index, ωs is the secondary radiation frequency, es is the (unit) where e and m = m0 (m∗ ) are the charge and free (effective) mass
polarization vector for the secondary radiation field, W(ωs ; es ) is for the incident (emitted) radiation.
the transition rate for the emission of secondary radiation (with In Figure 31, Feynmann diagrams describing the considered
frequency ωs and polarization es ) in the solid angle d when there processes are shown. Diagrams (a) and (c) describe the emission
is in the volume V a photon (with frequency ωl ) from the incident of photon (with frequency ωs ) by a conduction electron and can
radiation field. W(ωs ; es ) is calculated assuming a refraction index be directly related to the two terms on the right-hand side of
equal to unity (because the material refraction index is explicitly Eq. (234). Diagrams (b) and (d) can be similarly interpreted, but
included in the prefactor (in Eq. (232)) by means of Fermi’s golden correspond to a hole in the valence band. In Figure 32, band dia-
rule, grams are shown; Figure 32(a) and (b) is equivalent to Figure 31(a)
2π X
W(ωs ; es ) = |M1 + M2 |2 δ(εf − εi ) (233) and (b), respectively.
h̄ In Eq. (234), the intermediate states represent to the electron-
f
hole pair in a virtual state, while the intermediates states |bi are re-
where lated to the interference state, which can be neglected for semicon-
bjs |ai ha|H
X hf |H bjl |jl ii bjs |jl bi hb|H
X hf |H bjs |ii ductors with a large enough energy gap (for example, the GaAs,
Mj = + (234) CdTe, etc.).
a
εi − εa + i0a εi − εb + i0b
b The wave functions and energy levels of the electrons and holes
and j = 1; 2 denotes electron or hole contributions, respectively. system in the nanostructure are described by using the Schrödinger
εi and εf are the energies of the initial |ii and final |f i states of equation in the envelope function approximation. This equation is
38 RIERA ET AL.

limited to the case in which the minimum of the conduction band electron-emitted photon interaction can be written as
and the maximum of the valence band are located in the point s Z
D E |e| 2π h̄
0(case of interest). In these systems, the width or the radius of con- b
f |Hjs |a = (−1) j+1 e · χ̄∗j (r)p̂χ0j (r)dV (237)
m∗j Vωs
s
finement is less than the Bohr’s exciton radius, being the confine- V
ment energy of the electron-hole pair dominant over the Coulom-
Physically, a quantum transition from the intermediate state |ai to
bic interaction energy. Therefore, in the first approach it is con-
the final state |f i is obtained. The selection rules will depend again
sidered that the electron and the hole form an uncorrelated pair.
on the form of wave functions.
In this model, it is assumed that in semiconductor nanostructures The particular type of nanostructure requires a specific knowl-
in the initial state the conduction band is totally empty and the edge of wave functions and energy in the different states.
valence band is completely full, so that the presence of impurities
does not exist. However, in the semiconductors with the structure
of the of zinc-blende, the heavy- and light-hole bands are degener- 3.2. Electron Raman Scattering in Quantum Well
ated in the point 0. Nevertheless, in nanostructures the mixture of The electron wave function in the effective-mass approximation
states of light and heavy holes diminishes when the confinement and infinite potential barrier can be written as
radius or width decreases, which coincides with the calculations r !
made by using the multiband formalism: this permits, in the first 2 N1 π
ψk1⊥ ;N1 (r) = exp(ik1⊥ · r⊥ ) sin z uc (r) (238)
approach, to consider that the light and heavy holes can be consid- V L
ered decoupled. r !
2 N2 π
Parabolic conduction and valence bands are also assumed of ψk2⊥ ;N2 (r) = exp(ik2⊥ · r⊥ ) sin z uv (r) (239)
V L
isotropic nature. The parabolicity of bands is fulfilled around the
0 point. Due to degeneration, the valence band does not fulfill the The electron energies are given by
isotropic feature. Nevertheless, good results can be obtained if the
effective mass on several directions (cubic symmetry crystals) is h̄2 k21⊥ π 2 h̄2 2
εc (k1⊥ ; N1 ) = ∗ + N1 N1 = 1; 2; : : : (240)
averaged. 2m1 2m21 L2
In a theoretical investigation about electron Raman scattering,
h̄2 k22⊥ π 2 h̄2 2
it must be taken into account that: εv (k2⊥ ; N2 ) = − − N2 − εg
∗
2m2 2m22 L2
1. The electron-photon Hamiltonian is given by Eq. (235)
and the wave function corresponding to the |ai and |ii N2 = 1; 2; : : : (241)
states are given by (in the envelope function where “c”(“v”) denotes conduction (valence) band and εg is the
approximation) ψ(r) = χ(r)u(r). gap energy for the bulk semiconductors.
2. The crystal has a volume V = N1 0 , being N1 and 0 The initial state energy is
the number of cells and the volume of the unitary cell,
εi = h̄ωl (242)
respectively.
3. The functions u(r)(χ(r)) vary quickly (slowly) in the The final state is characterized by an electron with wave vector k1⊥
volume. occupying the N1 subband of the original conduction band, a hole
4. The Bloch functions between different bands are with momentum k2⊥ in the N2 subband of the original valence
orthogonals. band and a secondary radiation photon with frequency ωs .
5. The matrix element that connects the conduction and The final energy of the system is thus given by
valence bands in the center of R the Brillouin zone, k = 0, h̄2 k21⊥ h̄2 k22⊥ π 2 h̄2 2 π 2 h̄2 2
is given by Pcv (0) = (1/0 ) u∗1 (r)p̂u2 (r)dV . εf = h̄ωs +εg + + + ∗ N2 + ∗ N1 (243)
2m∗1 2m∗2 2m L2 2m L2
6. The variation of the functions in the whole crystal 2 1
volume, allows us to write the matrix element that According to diagrams of Figure 31, the intermediate state elec-
characterizes the electron-incident photon interaction in tron energies εa and εb are
the following form, !
s
h̄2 k2⊥ h̄2 π 2 N12 N22
D E Z εa = + εg + + ∗ (244)
|e| 2π h̄ 2µ∗ 2L2 m∗1 m2
b1l |i =
a|H el · Pcv (0) χ̄0∗
1 (r)χ2 (r)dV !
m0 Vωl h̄2 k2⊥
V h̄2 π 2 N12 N22
(236) εb = h̄ωl + h̄ωs + + εg + + ∗ (245)
2µ∗ 2L2 m∗1 m2
Physically, a quantum transition from the valence band to the con-
In Eqs. (244) and (245), we explicitly assumed the momentum con-
duction band will occur. The possible selection rules will depend
servation law and k1⊥ = −k2⊥ = k⊥ .
on the explicit form of the wave functions that are involved in the
The electron-hole pair does not change its total momentum
integral. For the holes, an expression similar to Eq. (235) is ob-
during the absorption or the emission of a photon (photon mo-
tained, but the indexes in the terms under the integral sign must
mentum is neglected).
be exchanged. From Eq. (242) to (245) and using energy and momentum con-
Considering the previous 1, 2, and 3 conditions and knowing servation laws, the denominators in Eq. (234) can be evaluated
that the integral of an odd operator between states of the same
parity is equal to zero, and that the Bloch functions are normal- π 2 h̄2  2 
εi − εa = h̄ωs + ∗
N1 − N202 (246)
ized in same band, then the matrix element that characterizes the 2m1 L2
 OPTICAL PROPERTIES IN NANOSTRUCTURES 39
"
π 2 h̄2  2 02
 1
εi − εb = −h̄ωl + N − N (247) ×
2m∗1 L2 1 2 h̄ω s
(1 + β−1 ) + (N12 − N22 )
ε0
From the latter results, it can be deduced that for semiconductors #2 
with a large enough energy gap εg (e.g., for GaAs) the contribu- 1 
tion from the interference diagram (c) of Figure 31 can be ne- + (251)
h̄ωs 
glected if is compared with the contribution of diagram (a) [119]. (1 + β) − (N12 − N22 )
ε0
Similar expressions can be written for the hole intermedi-
ate state energies and analogous conclusions can be made: dia- where
gram (d) in Figure 31 is negligible compared with diagram (b). π 2 h̄2 m∗2
ε0 = β= (252)
Thus, for the determination of the cross-section, the contribu- 2µ∗ L2 m∗1
tion of just the first term in the right-hand side of Eq. (234) must
and √
be considered during the calculation of Mj .
4 2Ve4 |el · Pcv (0)|2 η(ωs )µ∗1/2 ε0
3/2
σ0 = (253)
3.2.1. Differential Cross-section 9π 2 m20 h̄4 c 4 η(ωl )ωl ωs

From Eqs. (235) and (238), (239), it can be obtained that In Eq. (251), summation is over all the polarizations of the sec-
s ( ondary radiation field, while θ is the angle between the secondary
D E |e|h̄ 2π h̄ 2iesz radiation photon wave vector and the growth direction of the
b
f |Hjs |a = es · k⊥e δN 0 Nj + (−1)j+1
m∗j Vωs j L quantum well (the z-axis). Summations over the subband labels
) (N1 and N2 in Eq. (251)) must be done under the following re-
Nj0 Nj h
Nj0 +Nj
i quirements,
× (1 − δN 0 Nj ) (−1) −1
Nj2 − Nj02 j
π 2 h̄2
h̄ωl − h̄ωs − εg − N2 ≥ 0 (254)
j = 1; 2 (248) 2µ∗ L2 1
s ( !
D E |e| δN 0 N2 for j = 1 N12 N22
bjl |i = 2π h̄ π 2 h̄2
a|H el · Pcv (0) 1 (249) h̄ωl − h̄ωs − εg − + ∗ ≥ 0 (255)
m0 Vωl δN1 N 0 for j = 2 2L2 m∗1 m2
2

where m0 is bare electron mass and Pcv (0) the interband momen- The electron Raman scattering differential cross-section for a
tum matrix element (evaluated at k = 0). semiconductor quantum well Eq. (251) presents singular peaks for
The matrix elements from Eq. (233) are given by the secondary radiation frequency ωs such that
M1 + M2 ωs = ωes (N1 ; N2 ) or ωs = ωh
s s (N1 ; N2 ) (256)
2πe2 h̄2 1 where
= (e · Pcv (0))
m0 V ωl ωs l
 π 2 h̄  2 
e · k δ ωes (N1 ; N2 ) = N − N 2 (257)
s ⊥ N1 N2 2iesz N2 N1 2m∗1 L2 2 1
× ∗
+
 h̄ωs µ L N2 − N2 π 2 h̄  2 
1 2
h i ωh
s (N1 ; N2 ) = ∗
N1 − N22 (258)
N +N 2m2 L 2
× (1 − δN1 N2 ) (−1) 1 2 −1
" These frequency values correspond to conduction electron inter-
m∗1 −1
× subband transitions (ωes (N1 ; N2 )) or hole intersubband transi-
π 2 h̄2 tions in the valence-band (ωh s (N1 ; N2 )). Intersubband transitions
h̄ωs + (N 2 − N22 )
2m∗1 L2 1 obey a selection rule requiring N1 + N2 to be odd, otherwise the
# transition is forbidden. The latter result is in close analogy with the
m∗2 −1 
case of electron Raman scattering in bulk semiconductors subject
+ (250)
π 2 h̄2  to a magnetic field [117], where singularities in the emission spec-
h̄ωs − (N12 − N22 ) tra are also observed due to transitions between Landau subbands
2m∗2 L2
satisfying the selection rule N 0 = N ±1 (with N; N 0 = 0; 1; 2; : : : ).
After substitution of Eq. (250) into Eqs. (232) and (233), we obtain In the case of a semiconductor quantum well, the frequencies
ωs must lie in the interval,
d2 σ
ddωs 0 < h̄ωs ≤ h̄ωl − εg − ε0 (259)
 " #
9 X h̄ω − ε h̄ωs From Eq. (259), it can be deduced that secondary radiation is emit-
l g
= σ − − N12 (cos2 θ + 1) ted only for h̄ωl > εg + ε0 , otherwise the electron Raman scatter-
8 0 ε0 ε0
N1 ing cross-section is equal to zero.
!2 Let us note that Eq. (251) approaches the corresponding bulk
8 h̄ωs X N22 N12
+ sin2 θ semiconductor expression [120]. Then,
π 2 ε0 (N12 − N22 )2
N1 N2 !3/2
h i2 d2σ 3 h̄ωl − εg − h̄ωs
= σ0 (260)
× (1 − δN1 N2 ) (−1)N1 +N2 − 1 d dωs 2 ε0
40 RIERA ET AL.

Fig. 33. Electron Raman scattering cross-section (in units of σ0 ) as a

function of h̄ωs /ε0 for h̄ωl − εg = 10ε0 (solid line), h̄ωl − εg = 6ε0
(dashed line), and h̄ωl − εg = 10ε0 (dash-dotted line) for a bulk semicon-
ductor.

when L → ∞. It can also be noticed that the first term in the Fig. 34. Electron Raman scattering cross-section (in units of σ0 ) as a
differential cross-section Eq. (251) is free from singularities and function of (h̄ωl − εg )ε−1
0 for three values of h̄ωs /ε0 : h̄ωs = 4ε0 (solid
is related to transitions between valence and conduction subbands line), h̄ωs = 2ε0 (dashed line), and h̄ωs = 8ε0 (cross-dashed line). Step-
with N1 = N2 . like behavior due to transitions between valence and conduction subbands
In Figure 33, the electron Raman scattering differential cross- is indicated by (N1 ; N2 ) (as in Fig. 33). The dash-dotted curve shows the
electron Raman scattering cross-section for a bulk semiconductor with
section in relative units as a function of h̄ωs /ε0 , the so-called emis-
h̄ωs = 2ε0 .
sion spectrum, was shown. The masses were chosen for the case of
bulk GaAs (m∗1 = 0:0665m0 and m∗2 = 0:45m0 (heavy hole)).
In Figure 33, θ = π/2 has been chosen. It can be observed that
for h̄ωl − εg = 10ε0 the differential cross-section displays four For increasing values of ωl , there will be more terms contribut-
ing to the double sum in Eq. (251), which will produce steplike
singularities at h̄ωh h e e
s (2; 1), h̄ωs (2; 3); h̄ωs (1; 2), and h̄ωs (3; 2), increments in the differential cross-section. A similar result was
while, for h̄ωl − εg = 6ε0 only two singularities are observed
obtained for the light absorption coefficient in a semiconductor
at h̄ωh e
s (2; 1) and h̄ωs (1; 2). It should be noted that the ωs val- quantum well [13].
ues for which singularities are found do not depend on the inci- Now let consider the first term in Eq. (251). This term weakly
dent radiation frequency ωl and only depend on the energy differ- depends on ωl and its increase can be connected with the par-
ences between the valence and conduction subbands. In general, ticipation of a higher number of subbands of the valence or con-
for the model of a quantum well with infinite potential barriers duction type. The latter dependence can be observed as the slope
the position peaks ωs (N1 ; N2 ) will change with the variations of changes in the curves of Figure 35 for h̄ωl = h̄ωs + εg +
the quantum-well width according to the law ≈ L−2 . For higher (π 2 h̄2 /2µ∗ L2 )N12 and different values of h̄ωs /ε0 .
energies h̄ωl of the incident radiation photon, a larger number To give numerical estimations of the secondary radiation in-
of singular peaks in the emission spectra as can be deduced from tensities predicted by these results, the quantum efficiency (1/V )
Eqs. (257) and (258) will be observed. For a fixed value of ωl , in dσ/d for a fixed values of ωl is computed. Therefore, it was cho-
the emission spectra a certain steplike behavior at given values of sen: εg = 1:43 eV, η(ωl ) = 0:899, η(ωs ) = 3:46 [121], and h̄ωl =
ωs can be observed. The maximum number of steps is determined h̄ωg = 1:182ε0 . By integrating Eq. (251) from h̄ωs = 0:15εg to
by Eqs. (254) and (255) when we set ωs = 0 and use is made of the h̄ωs = 0:091εg , with L = 10 nm, the result is (1/V ) dσ/d =
selection rule N1 + N2 = 2N + 1 (N = 1; 2; : : : ). The position of 2:84 × 10−6 fcvx sr−1 cm−1 , where f x = 2/(m ε )|(P ) |2 is the
cv 0 g cv x
the steps in the spectra depends on the value chosen for ωl . oscillator strength in the xa -direction. For GaAs, fcv x = 6:7599
In Figure 34, the differential cross-section as a function of [122] and (1/V) dσ/d = 1:92 × 10−5 sr−1 cm−1 .
(h̄ωl − εg )ε−10
for three values of h̄ωs /ε0 (the so-called excita- Finally, some comments should be made. Electron Raman scat-
tion spectrum) is shown. As can be observed, for increasing values tering in semiconductor quantum wells can be used for the deter-
of (h̄ωl − εg )ε−10
new steps will be appearing because new sub- mination of the subband structure in real heterostructures of this
bands will be accessible to electrons and holes in the quantum well. kind. The fundamental features of the differential cross-section,
The step positions are given by as described in this subsection, should not change very much in
! the real case. It can be easily proved that the singular peaks in
π 2 h̄2 N12 N22 the cross-section will be present irrespective of the model used for
h̄ωl = h̄ωs + εg + + ∗ (261)
2L2 m∗1 m2 the subband structure and shall be determined for values of h̄ωs
 OPTICAL PROPERTIES IN NANOSTRUCTURES 41
!
n +1/2 r
× Jne +1/2 µke Ynme e (θ; ϕ)uc (262)
e r0

for conduction electrons; Jn+1/2 (µ) is the Bessel function of

n+1/2
order n + 1/2, µk denotes the zeros of Bessel function:
n+1/2
Jn+1/2 (µk ) = 0, Y(θ; ϕ) are the spherical harmonics, uc is
the Bloch function taken at k0 = 0 in the Brillouin zone.
From Eqs. (235) and (262), the matrix element can be obtained
as,
s
D E |e| 2π h̄
b
a|Hjl |i = e
m0 Vωl l
(
δm0 m2 δn0 n2 δk0 ;k2 j = 1
× P̂cv (0) 1 1 1 (263)
δm1 m0 δn1 n0 δk1 ;k0 j = 2
2 2 2

Physically, this matrix element represents a quantum transition of

an electron (hole) from the valence (conduction) band to the of
conduction (valence) band, of such a way that only those transi-
tions between subbands indexed by the same quantum numbers
are permitted. Otherwise, the transitions are forbidden.
Fig. 35. Electron Raman scattering cross-section (in units of σ0 ) as a The selection rules for the intersubband transitions from the
function of (h̄ωl − εg )ε−1
0 for four values of the parameter h̄ωs /ε0 and
intermediate to the final states are defined by the wave functions
θ = π/2. The slope change for the transition h̄ωl − εg = h̄ωs + 4ω0 is previously used and by Eq. (237).
indicated by an arrow. The matrix element defined by H bjs is given by
s
D E i h̄|e| 2π h̄
bjs |a = (−1)j+2
f |H
equal to the energy difference between two subbands: h̄ωs
e(h)
= m∗j Vωs
e(h) e(h) e(h) e(h) ( )
εα − εβ , where εα > εβ are the respective electron kj k0j 1 kj k0j
(hole) energies in the subbands. × j In n0 Mdj + j IIn n0 Ndj (264)
j j h̄ j j
Similarly, it will have a steplike dependence in the differential
cross-section for h̄ωl = h̄ωs + εg + εα + εβ . where
A complete study of electron Raman scattering in semiconduc- Zr0 Zr0
tor quantum-well heterostructures should also be concerned with kj k0j kj k0j
j Inj n0 = fj (r)fj0 (r)r 2 dr j IInj n0 = fj (r)fj0 (r)r dr (265)
the case when phonon-assisted transitions are involved. In this j j
case, the intermediate states in Eq. (234) are real and the electron- 0 0
phonon interaction must be explicitly included in the calculations. and
This latter point requires an independent work. All the obtained
(+) (+) (−) (−) (z)
results can be extended to the case of a superlattice where the cal- Xdj = Xdj es + Xdj es + Xdj esz
culations would be concerned with the miniband structure created (266)
Xdj = Mdj ; Ndj
by superlattice superperiodicity.
with
3.3. Electron Raman Scattering in Quantum Dots kj
(±) Cnj mj  k0j ∓m0j
The differential cross-section for the electron Raman scattering Mdj = √ ±Cn0 m0 Dn0 δnj ;n0 +1 δmj ;m0 ∓1
2 j j j j j
process in semiconductor quantum dots of spherical and rectan-
k0j ∓m0j 
gular shape are calculated following Bergues et al. [123].
∓Cn0 m0 En0 δnj ;n0 −1 δmj ;m0 ∓1 (267)
j j j j j
 k0 m0j
(Z) j
3.3.1. Spherical Quantum Dot Mdj = Cn0 m0 Bn0 δnj ;n0 +1 δmj ;m0
j j j j j
We consider a complete electron confinement structure. The solu-
k0j m0j  k
tion of the Schrödinger equation in the envelope function approx- j
+Cn0 m0 Cn0 δnj ;n0 −1 δmj ;m0 Cnj mj (268)
imation leads to j j j j j
s  !
(2ne + 1)(ne − me )! h̄m0j ∓m0j ∓m0j m0j ∓1
ψc = χc uc = (±) kj
Ndj = Cnj mj  √ Dn0 ∓ An0 Bn0
2π(ne + me )!
2 j j j
" !−1
n +1/2 
× r0 Jne +3/2 µke r −1/2 k0j
e × Cn0 m0 δnj ;n0 +1 δmj ;m0 ∓1
j j j j
42 RIERA ET AL.
!
h̄m0j ∓m0j ∓m0j m0j ∓1 Substituting Eqs. (263), (264), (274), and (275) in Eq. (234) we
∓ ± √ En0 + An0 Cn0
2 j j j obtain
 A0
Mj = (−1)j+1
k0j m∗j
× Cn0 m0 δnj ;n0 −1 δmj ;m0 ∓1  (269)
j j j j k k k k k k
In11n22 Mdj + (1/h̄)(1 IIn11n22 δ1j + 2 IIn22n11 δ2j )Ndj
× (276)
(s) k (n) (s) k (n)
(ε0 gk1 (n−1) + i0)δ1j + (ε0 dk 1(n−1) + i00 )δ2j
Cnj mj  
kj 2 2
(Z) m0j −(m0j +1) −m0j m0j −1
Ndj = √ An0 En0 + An0 En0 where
2 j j j j     
k0j k1 (i) h̄ωs i+1/2 2 i+1/2 2
× Cn0 m0 δnj ;n0 −1 δmj ;m0 gk (i) = (s) + β1 µk − µk (277)
2 1 2
j j j j ε0
      
m0j −(m0j +1) −m0j m0j −1 k (i) h̄ωs i+1/2 2 i+1/2 2
− An0 Dn0 + An0 Dn0 dk 1(i) = (s) + β2 µk − µk (278)
j j j j 2 2 1
 ε0
k0j
× Cn0 m0 δnj ;n0 +1 δmj ;m0 (270) (s) h̄2 k1 k2 k2 k1 k1 k2
j j j j ε0 = 1 IIn1 n2 = 2 IIn2 n1 = In1 n2
2m∗j r02
e±
s = es · e± and esz = es · ez (271)
µ∗
and βj = j = 1; 2 (279)
r m∗j
(n ∓ m)(n ± m + 1)
A±m
n = h̄ e2 2π h̄2
2 A0 = −i √ e · Pcv (0) (280)
s m0 V ωl ωs l
(n + 1 − m)(n + 1 + m)
Bnm = Here, the following expression for the Dirac’s delta is used:
(2n + 1)(2n + 3)
s 0f
1
(n − m)(n + m) δ(εf − εi ) → (281)
Cnm = (272) π (εf − εi )2 + 02
(2n − 1)(2n + 1) f
s
(n ± m + 1)(n ± m + 2) Taking into account the expressions for initial and final energies
Dm
n = (Eqs. (273) and (274)), the Dirac’s delta can be expressed as
(2n + 1)(2n + 3)
s
1 δf
(n ∓ m)(n ∓ m − 1) δ(εf − εi ) = (282)
Enm = πε0 |P + iδf |2
(2n − 1)(2n + 1)

Equations (266)–(270) and (264) show that in the transition be- where
tween |ai and |f i states, three polarizations appear, in the first h̄ωl − εg − h̄ωs    
n +1/2 2 n +1/2 2
one the electron (hole) transition is such that mj = m0j − 1, in the P= − β1 µk1 − β2 µk2 (283)
ε0 1 2

second one mj = m0j + 1 and in the third one mj = m0j . In the

Substituting the Eqs. (276) and (282) in Eq. (233) we obtain
three cases, the requirements nj = n0j + 1 and nj = n0j − 1 must be
fulfilled. 20f |A0 |2 (s)
W = Wf (284)
h̄ε40 m∗ 2 r02
3.3.1.1. Differential Cross-section where
 2  2 
The energy of initial, final, and intermediate states is given by X     β Y J 
β1 Yj1   2 j2 k1(n)  
Wj
(s)
=   + 
εi = h̄ωl (273)   gk1 (n) + iδ   d k1 (n) + iδ0 
nmk1 k2 k2 (n−1) k2 (n−1)
 
h̄2
n +1/2 2  2  2
εf = µk1  X k1 (n)  β Y J 
2m∗1 r02 1 
×
k2 (n−1)   1 j3
 +  k (n)
k 2 (n+1) 

 P k1 (n) + iδ  g 1 + iδ 
h̄2  n2 +1/2 2 k2 (n−1) f k2 (n+1)
+ µ + h̄ωs + εg (274)  2  2 
2m∗2 r02 k2  β Y   X k1 (n)  
 2 j4   k2 (n+1) 
+    (285)
and  d k1 (n) + iδ0   P k1 (n) + iδ  
k2 (n+1) k2 (n+1) f
h̄2  n0 +1/2 2 h̄2  2
n +1/2
εa = µk10 + µk2 + εg (275) where j = 1 = esz , j = 2 = e+ −
s , j = 3 = es are the polarization
2m∗1 r02 1 2m∗2 r02 2
directions and
which correspond to the electronic contribution. For the hole, the i+1/2 i+1/2
k (i)
µk µk
expression is similar to Eq. (275), but is needed to change n2 ; k2 Xk 1(i) = 1 2
(286)
i+1/2 2 i+1/2
by n02 ; k02 and n01 ; k01 by n1 ; k1 . ) − (µk
2
(µk )
1 2
 OPTICAL PROPERTIES IN NANOSTRUCTURES 43

The terms Yjα are characterization coefficients of the differential spectra. The transitions involved in Eqs. (292) and (293) are inde-
cross-section for each polarization, pendent of ωl and correspond to intraband transitions. The last
√ a! √ a! singularities are present only in the emission spectra.
Y11 = 2q b = Y12 Y13 = 2q (a + 1) = Y14
b! b!
a! a! 3.3.2. Rectangular Quantum Dots
Y21 = q = Y32 Y22 = −q b(b − 1) = Y31 (287)
b! b! 3.3.2.1. Matrix Elements
a! a!
Y23 = −q = Y34 Y24 = q (a + 2)(a + 1) = Y33 The solution of the Schrödinger equation in the envelope function
b! b!
approximation for the rectangular quantum dot of lengths d1 , d2 ,
and d3 is given by
1 s s s
a = n−m b = n + m and q = √ (288) 2 2 2
2π 2 ψc = χc uc =
d1 d2 d3
To obtain Eq. (284), the definitions given in Eq. (287) were taken ne π me π ke π
into account. So, substituting Eq. (284) in Eq. (232) the differential × sin x sin y sin zuc (295)
d1 d2 d3
cross-section is obtained
" # From Eqs. (235) and (295), the following matrix element is ob-
d2σ X3
d2σ tained
= (289) s
dωs d dωs d
j=1 j D E |e| 2π h̄
bjl |i =
a|H e
!(s) ! m0 Vωl l
d2σ 27 h̄ωs (s) (
= σ E j )2
δf Wj x(es σ (290) δn0 n2 δm0 m2 δk0 ;k2 j = 1
dωs d 4 0 ε(s) b
j 0 ×Pcv (0) 1 1 1 (296)
δn1 n0 δm1 m0 δk1 ;k0 j = 2
2 2 2
where
√ This matrix element represents a quantum transition of an elec-
4 2Ve4 |el · Pcv |2 η(ωs )µ∗1/2 ε0
3/2
σ0 = (291) tron from the valence band to the conduction band, of such way
9π 2 m20 h̄4 c 4 η(ωl )ωl ωs that only those transitions between subbands indexed by the same
quantum numbers are permitted. Otherwise, the transitions are
In a similar way to the case of the quantum well, in Eq. (289) forbidden.
three scattering configurations appear, in which the electron Ra- The selection rules for intersubband transitions from interme-
man scattering emission spectrum in the spherical quantum dot diate to final states are defined by the wave functions previously
has maximum values for the following values of ωs , used and by the expression (237). Then, the matrix element can be
(s) determined as
β1 ε0 s
ωs → ω1 (n1 ; k1 ; k2 ) = D E |e| 2π h̄ 8
h̄ bjs |a = (−1)
f |H j+1
 2    es
n +1/2 n +1/2 2 m∗j Vωs d1 d2 d3
× µk1 − µk2 (292) 
1 2
 Z d1 Z d2
d
β2 ε0
(s) × −ih̄i ϕ∗j (x) ϕ0j (x)dx ϕ∗j (y)ϕ0j (y)dy
ωs → ω2 (n1 ; k1 ; k2 ) =  0 dx 0
h̄
     Z d3 Z d1
n2 +1/2 2 n1 +1/2 2 × ϕ∗j (z)ϕ0j (z)dz − ih̄j ϕ∗j (x)ϕ0j (x)dx
× µk − µk (293)
2 1 0 0
Z d2 Z d3
d 0
As can be observed from Eq. (292) or (293) these frequencies × ϕ∗j (y)
ϕj (y)dy ϕ∗j (z)ϕ0j (z)dz − ih̄k
correspond to electron or hole intersubband transitions in which 0 dy 0
Z d1 Z d2
the following selection rules are satisfied: n2 = n1 ± 1, the minus
sign is applied to Eq. (292) and the plus sign is applied to (293). × ϕ∗j (x)ϕ0j (x)dx ϕ∗j (y)ϕ0j (y)dy
0 0
Other singularities of Eqs. (290) and (285) occur whenever 
Z d3 
P = 0. Such singularities are mainly related to certain values of ∗ d 0
× ϕj (z) ϕj (z)dz (297)
the frequency ωl of the incident photon. For the excitation spec- 0 dz 
tra, the positions of these singularities are given as
where
1
ω(n1 ; k1 ; k2 ) = (h̄ωl − εg ) 1j π
ε0
(s) ϕj (ξ) = sin ξ (ξ = x; y; z; 1j = Mj ; Kj ; Nj ) (298)
dξ
   
h̄ωs n +1/2 2 n +1/2 2
= (s) + β1 µk1 + β2 µk2 (294) Keeping in mind the Eq. (298),
1 2
ε0
Zd
Here, the selection rules n2 = n1 ± 1 are satisfied. The peaks re- d 2 nm
ϕ∗n (ξ) ϕm (ξ)dξ = −
lated with the last singularities correspond to interband transitions dξ d n2 − m2
0
of the electron-hole pair and their positions depend on the fre- h i
quency of the incident radiation ωl for the emission and excitation × (−1)n+m − 1 (1 − δnm ) (299)
44 RIERA ET AL.

2i h i N N
therefore Eq. (297) can be expressed as + esz (−1)N1 +N2 − 1 1 2
D E d3 N12 − N22
bjs |a
f |H )
 
s × 1 − δN1 N2 δM1 M2 δK1 K2
j+1 |e|h̄ 2π h̄
= (−1)  "
m∗j Vωs 
( h̄2 π 2  2 
Mj Mj0 × h̄ωs + (−1)j+1 M1 − M22 (303)
2i  2m∗ d 2
× esx j 1
d1 Mj2 − Mj02
h̄2 π 2  2 
h i  + K1 − K22
M +Mj0
× (−1) j − 1 1 − δMj M 0 δkj k0 δNj N 0 2m∗j d22
j j j
# −1
Kj Kj0 
2i
+ esy h̄2 π 2  2 2

d2 02 + N1 − N2 + i0a j = 1; 2
Kj2 − Kj ∗
2mj d3 2 
h i 
K +K 0
× (−1) j j − 1 1 − δkj k0 δMj M 0 δNj N 0 being
j j j

Nj Nj0 2πe2 h̄2 (el · Pcv (0))

2i A0 = √ (304)
+ esz m0 V ωl ωs
d3 Nj2 − Nj02 2
) and
h i 
N +N 0 Y h i qj q  
× (−1) j j − 1 1 − δNj N 0 δMj M 0 δKj K 0 (300)
(qj ; q4 ) = (−1)qj +q4 − 1
4
j j j 1 − δqj q4 (305)
qj2 − q42
For quantum transition from the |ai to the |f i state, three exclud-
ing polarizations contribute, which will be different from zero only where, for j = 1; 2; 3, q1 = M1 , q2 = K1 , q3 = N1 , and q4 = M2 ,
if [(−1)n+m − 1]δab δcd = −2, otherwise this would be forbidden. K2 , N2 . The sum of electronic and hole contributions is
"
A0  2
M1 + M2 = ∗ i esx π(M1 ; M2 )δK1 K2 δN1 N2
3.3.2.2. Differential Cross-section µ ε0  d1
The final and intermediate states can be written as 2  
+ esy π K1 ; K2 δM1 M2 δN1 N2
h̄2 π 2 M12 h̄2 π 2 K12 h̄2 π 2 N12 h̄2 π 2 M22 d2
εf = + + + #
 
2m∗1 d12 2m∗1 d22 2m∗1 d32 2m∗2 d12 +
2
esz π N1 ; N2 δM1 M2 δN1 N2
d3
h̄2 π 2 K22 h̄2 π 2 N22
+ + + εg (301) " #

2m∗2 d22 2m∗2 d32 β1 β2
× + (306)
ε1 + iδa ε2 + iδ0a 
and

h̄2 π 2 M102 h̄2 π 2 K102 h̄2 π 2 N102 h̄2 π 2 M22 where

εa = + + + µ∗
2m∗1 d12 2m∗1 d22 2m∗1 d32 2m∗2 d12 βj = (j = 1; 2)
m∗j
h̄2 π 2 K22 h̄2 π 2 N22
+ + + εg (302) m∗1 m∗2 h̄2 π 2
2m∗2 d22 2m∗2 d32 µ∗ = ε0 = (307)
m∗1 + m∗2 2µ∗ ∂2
which correspond to the electronic contribution. For the hole con- d1 d2 d3 0a 00a
tribution, we index with primes the terms with subindex 2. ∂ = δa = δ0a = (308)
d2 d3 + d1 d3 + d1 d2 ε0 ε0
Substituting Eqs. (296), (297), (301), and (302) in Eq. (234) we
obtain h̄ωs qj2 − q42
εi = − (−1)i βi i = 1; 2; j = 1; 2; 3 (309)
2πe2 h̄2 el · Pcv (0)
ε0 ∂2j
Mj = √
m0 m∗j V ωs ωl d d1 d d
( ∂1 = 1 + 1 + ∂2 = 1 + 2 + 2
h i M M d2 d3 d1 d3
2i
× esx (−1)M1 +M2 − 1 1 2 d d3
d1 M12 − M22 ∂3 = 1 + 3 + (310)
  d1 d2
× 1 − δM1 M2 δK1 K2 δN1 N2 Taking into account Eqs. (273) and (301), the Dirac’s delta is given
2i h i K K by
+ esy (−1)K1 +K2 − 1 1 2
!
d2 K12 − K22 1 δf 0f
  δ(εf − εi ) = δf = (R) (311)
× 1 − δK1 K2 δM1 M2 δN1 N2 πE0 K 2 + δ2
0 f ε0
 OPTICAL PROPERTIES IN NANOSTRUCTURES 45

h̄ωl − εg h̄ωs M12 K12 N12

K0 = − − − −
ε0 ε0 β1 ∂21 β1 ∂22 β1 ∂23
M2 K22 N22
− 2 − − (312)
β2 ∂21 β2 ∂22 β2 ∂23
Substituting Eqs. (306) and (311) in Eq. (233) and this in Eq. (232)
we obtain " #(R)
d2σ X3
d2σ
= (313)
dωs d dωs d
j=1 j

where
" #(R) !2
d2σ 36 ∂3 h̄ωs (R)
= σ0 δf Wj E j )2
(es · σ (314)
dωs d π4 V (R)
j ε0
being
M K
W1 = WM 2K N W2 = WM 2K N
1 1 1 1 1 1
(315)
N
W3 = WM 2K N
1 1 1
X  2
q4 1  
Wq1 q2 q3 = π 2 (qj ; q4 )Z1 (qj ; q4 ) + iZ2 (qj ; q4 )
∂2j q1 q2 q3 q4 Fig. 36. Electron Raman scattering cross-section (in arbitrary units of σ0 )
as a function of h̄ωs /ε0 (emission spectrum), for GaAs spherical quantum
 2
  dots in the scattering configuration Z(el ; b Z)X, with 0f = 3 and 0a = 0b =
 1  1 meV. The solid curve corresponds to h̄ωl = 2:3 eV, the dashed curve to
×  (316)
 Koj (q1 ; q2 ; q3 ; q4 ) + iδf  h̄ωl = 2:9 eV, in both cases r0 = 4 nm. Resonant transitions are indicated
by ωe (ne ; ke ; nh ; kh ) and ωh (ne ; ke ; nh ; kh ) corresponding to electron
with or hole intersubband transitions.
β1 ε1 (qj ; q4 ) β2 ε2 (qj ; q4 )
Z1 (qj ; q4 ) = +
ε21 (qj ; q4 ) + δ2a ε22 (qj ; q4 ) + δ02
a
(317) Landau subbands satisfying the selection rules N 0 = N ± 1 (with
β1 δa β2 δ0a N; N 0 = 0; 1; 2; : : :):
Z2 (qj ; q4 ) = +
ε21 (qj ; q4 ) + δ2a ε22 (qj ; q4 ) + δ02
a h̄ωl − εg
ω(M1 ; K1 ; N1 ; M2 ) =
(R)
ε0
h̄ωl − εg h̄ωs q12 q22
K0j (q1 ; q2 ; q3 ; q4 ) = (R)
− (R) − − β1 M12 K2
ε0 ε0 β1j ∂21 β2j ∂22 h̄ωs
= (R)
+ + 1
2
∂1 ∂22
ε0
q32 β2 q42
− − (318) N2 β21 M22
β3j ∂23 ∂2j + 1 + (322)
 ∂23 ∂21
βkj = β−1
1 k=j (319)
1 k 6= j For increasing values of ωl , there will be more terms contribut-
ing to the sum in Eqs. (314)–(316), which will produce a step-
and σ0 is similar to the Eq. (291).
like change in the differential cross-section. A similar result was
For quantum dot, the electron Raman scattering emission spec-
obtained for the light absorption coefficient in a semiconductor
trum has maximum values at the following values of ωs ,
quantum well [13].
  β ε(R) m∗2 − m∗2 The calculations of the differential cross-section reported in the
ωs → ω1 m∗1 ; m∗2 =
1 0 2 1 (320) expressions (285), (289), (290) and (313)–(316), provide a trans-
h̄ ∂2 1 parent understanding of the structure of subbands of the stud-
  β ε(R) m∗2 − m∗2 ied nanostructures, information that is possible obtain quickly,
ωs → ω2 m∗1 ; m∗2 =
2 0 1 2 (321) by means of the emission and excitation spectra of the electron
h̄ ∂2 1 Raman scattering process. The physical parameters entering in
These frequency values correspond to electron intersubband tran- these formulas were taken for the GaAs case; i.e., εg = 1:43 eV,
sitions in conduction band, ω1 (m∗2 − m∗1 ), or hole intersubband m∗1 = 0:0665m0 , m∗2 = 0:45m0 (the heavy-hole band). It was also
transition in the valence band, ω2 (m∗1 − m∗2 ). Intersubband tran- chosen that 0f = 1 meV and 0a = 3 meV. This permits us to
sitions obey selection rules, which require m∗1 + m∗2 to be odd, plot the so-called emission spectra 1/σ0 [d 2 σ/(d dωs )] versus
otherwise the transition is forbidden. This result is related closely h̄ωs /ε0 and the excitation spectra 1/σ0 [d 2 σ/(d dωs )] versus
with the case of electron Raman scattering in bulk semiconduc- (h̄ωl − εg )/ε0 .
tors subject to a magnetic field [117], where the singularities in the Figures 36 and 37 show the emission spectra for the spherical
emission spectra are also observed due to transitions between the and the rectangular quantum dots, respectively.
46 RIERA ET AL.

Fig. 39. Same as in Figure 38 but for a cubical quantum dot. The solid
Fig. 37. Same as in Figure 36, but for a cubical quantum dot. The solid
curve corresponds to dx = dy = dz = 9 nm and h̄ωs = 2 eV, the medium
curve corresponds to dx = dy = dz = 12 nm, the dashed curve corre-
dashed curve corresponds to h̄ωs = 2:3 eV, and the long dashed curve
sponds to dx = dy = dz = 15 nm, in both cases h̄ωl = 2:3 eV. Resonant
corresponds to dx = dy = dz = 9 nm, and h̄ωs = 2 eV.
transitions are indicated by ωe (ne ; mh ) and ωh (ne ; mh ) corresponding to
electron or hole intersubband transitions.

Fig. 40. Same as in Figure 38 but in all cases h̄ωs = 2:3 eV. The solid
curve corresponds to r0 = 3 nm, the medium dashed curve corresponds to
r0 = 4 nm, and the short dashed curve corresponds to r0 = 5 nm.

Fig. 38. Electron Raman scattering cross-section (in arbitrary units of σ0 )

as a function of (h̄ωl − εg )ε−1
0 (excitation spectrum) for a spherical quan-
tum dot. The calculation was performed applying the same parameters and Figures 38–40 show the excitation spectra for the spherical
the same configuration as in Figure 36. The solid curve corresponds to and the rectangular quantum dots, respectively. The quantity 1/σ0
r0 = 3 nm and h̄ωs = 2 eV, the medium dashed curve corresponds to [d 2 σ/(d dωs )] is frequently called “scattering efficiency.”
h̄ωs = 2:3 eV, and the short dashed curve corresponds to r0 = 4 nm and The emission and excitation spectra were computed for a given
h̄ωs = 2 eV. polarization esz of the emitted radiation.
 OPTICAL PROPERTIES IN NANOSTRUCTURES 47

The excitation spectra for the spherical and the rectangular The resonant transitions are indicated only in one curve, because
quantum dots were computed for the scattering configuration they are the same in all curves, we should only keep the same or-
Z(el ; b
Z)X. In this case, only the first term at the right-hand side der, which is given by expression (294). It is also observed that
n +1/2
of Eqs. (289) and (313) contribute, thus, there are not any singu- a threshold for the lower values of (h̄ωl − εg )ε−1 0
when µke
e
larities, that is, for fixed values of h̄ωs and r0 or d, the expressions n +1/2
(292), (293), (320), and (321) do not contribute. Furthermore, for and µkh take their minimum values for fixed values of h̄ωs
h
certain values of ωl abrupt changes can be found in the curve slope and r0 .
which corresponds to different thresholds related to the points In Figure 39, the excitation spectra for a cubical quantum dot is
where a given subband starts to contribute to the differential cross- displayed. A similar result as in Figure 38 is observed. This is due,
section. This explains the steplike character of the curve. The low- again, to the same physical properties for both symmetries.
est admissible value of h̄ωl − εg is defined by the minimum value In Figure 40, the excitation spectra for the spherical quantum
n+1/2
of µk . For higher values of h̄ωl − εg , new subbands start to dot for three values of r0 is shown, but now it is chosen to be h̄ωs =
contribute, thus defining the other thresholds seen in the figures. 2:3 eV in all cases. In this case, only a shift in the spectra can be
Explicit indication of the points where the thresholds are present observed.
is given, specifying the involved subbands. Figures 36 and 37 show To give numerical estimates of the secondary radiation inten-
the emission spectra for the same scattering configuration. The sity, the quantum efficiency 1/V(dσ/d) was calculated. With such
incident radiation frequency was fixed as h̄ωl = 2:3 and 2.9 eV purpose, the expressions (289) and (313) are integrated in terms
for both quantum dots. Abrupt changes in the slope can be ob- of the secondary radiation energy in units of ε0 . The results were
served, thus providing a certain steplike dependence of the scat- evaluated for the GaAs considering the parameters defined previ-
tering efficiency. The points where the curve slope presents abrupt ously and are illustrated in Table III.
changes are related to threshold values of h̄ωs representing the in- The effective width reported in lines 4 and 5 corresponds to the
corporation of new subbands to the process. It should be realized cases in which d1 = d2 = d3 = 120 Å and d1 = d2 = d3 =
that, for lower values h̄ωs , more subbands could participate in the 64:28 Å, respectively. In the latter case, the cube dimensions are
n+1/2 determined upon considering the volumes of the rectangular and
emission process. The condition h̄ωl − h̄ωs − εg > ε0 (µk )
must be fulfilled to have the emission of secondary radiation pho- the spherical quantum dots equal (with radius r0 = 40 Å). The
tons. For fixed values of h̄ωl , εg , and ε0 , the threshold positions quantum efficiency values in both cases indicate that the rectangu-
n+1/2 lar quantum dot that is more like the spherical quantum dot with
are defined by µk . This is explicitly indicated in Figures 38 r0 = 40 Å is corresponding to row 4 in Table III. This is due to the
and 39.
effective width determining the electron-hole pair formation, i.e.,
In Figure 36, the emission spectra for the Z(el ; b Z)X configu-
the electron-hole pair moves in a region which feels the average of
ration for the spherical quantum dot is shown. It has been chosen
spatial confinement directions.
that r0 = 4 nm. It can be observed that for h̄ωl = 2:3 eV, the differ-
The quantum efficiency depends on the incident radiation en-
ential cross-section displays only two singularities at ωe (0; 1; 1; 1),
ergy, the confinement dimension and the interval of the secondary
and ωh (1; 1; 0; 1), while for h̄ωl = 2:9 eV four singularities
radiation energy in which it is determined. However, the results il-
are observed at ωe (0; 1; 1; 1), ωh (1; 1; 0; 1), ωe (1; 1; 2; 1), and lustrated in the Table III are the consequence of the subband struc-
ωh (2; 1; 1; 1). It should be noted that the ωs values for which
ture variation, which is due to the change in the geometry and the
we find singularities do not depend on the incident radiation fre-
electron confinement dimensionality. So, we can observe that the
quency ωl and only depend on the energy differences between the
quantum efficiency:
valence and conduction subbands for the fixed value r0 . For higher
energies h̄ωl of the incident radiation photon, it shall be observed • Increases with the increment of the confinement radius, while
that a large number of singular peaks in the emission spectra can the incident radiation energy and the interval of the secondary
be deduced from Eqs. (292) and (293). For a fixed value of ωl , radiation energy are kept constant.
a certain steplike behavior at given values of ωs can also be ob- • Increases with the increment of the incident energy being fixed
served. The maximum number of steps is determined by Eq. (294). at the confinement radius and the interval of the secondary
The position of the steps in the spectra depends on the value cho- radiation energy.
sen for ωl . • The rectangular quantum dot is greater than the spherical
In Figure 37, the emission spectra for the same configuration quantum dot.
as Figure 36 for the cubical quantum dot is shown, it is chosen that • The quantum-well wire is greater than the quantum dot.
h̄ωl = 2:3 eV and d1 = d2 = d3 = 12 nm and d1 = d2 = d3 =
15 nm. One can observe the same physical situation as Figure 36. The previous comments justify the importance of studying the
It should also be noted that the singularities due the electron tran- electron Raman scattering in different systems with several ge-
sitions are more intense than hole transitions. In conclusion, the ometries.
different symmetry of the quantum dots do not change the optical
physical properties, the main difference are given by ε0 , which, of 3.4. Electron Raman Scattering in Cylindrical
course, is different for both quantum dots. Quantum-well Wires
In Figure 38, the excitation spectra for the spherical quantum
dot for the same scattering configuration as in Figure 36 is shown. The differential cross-section for electron Raman scattering pro-
In this case, h̄ωs is fixed and h̄ωl is a variable quantity. The other cess in semiconductor quantum-well wires of a cylindrical section
parameters coincide with those of Figure 36. The same behav- is calculated following Bergues et al. [124].
ior, in the spectra for different values of r0 with the same h̄ωs , The solution of the Schrödinger equation in the envelope func-
as those for the same r0 with different values h̄ωs , is observed. tion approximation, considering an infinite potential barrier in the
48 RIERA ET AL.

Table III. Quantum Efficiency Values (Spherical Quantum Dot (SQD), Rectangular Quantum Dot (RQD), and Quantum-well Wire (QWW)

System h̄ωl (eV) r0 [∂] (Å) 1 dσ −1 h̄ωs

V d (cm ) 1 ε0

SQD 3.328 30 1:967 × 10−6 (8.2, 10.5), 6 × 10−5 (1, 2.6)

SQD 3.328 40 4:62 × 10−3 (8.2, 10.5), 3:14 × 10−3 (1, 2.6)
SQD 2.3 40 5 × 10−5 (8.2, 10.5)
RQD 2.3 [40] 6:132 × 10−3 (0.45, 0.5)
RQD 2.3 [21, 46] 7:782 × 10−6 (0.2, 26.7)
QWW 3.328 40 1:511 × 10 (8.2, 10.5), 6:372 × 10−2 (12, 12.92)
−2

QWW 3.328 30 5:25 × 10−3 (8.2, 10.5), 2:47 × 10−2 (7, 8.5)

quantum-well wire leads to Just the final results are written

! " # " # " #
h√ i−1 d2σ d2σ d2σ
m r = +
πLr0 jn0 e (µnee )
m
ψc = χc uc = jne µnee dωs d dωs d dωs d +
r0 esz es
h i " #
× exp −i(ne φ + ke z) uc (323) d2 σ
+ (327)
dωs d −
es
for the conduction electron. In Eq. (323), jn (µ) is the Bessel func-
tion of order n. µm where
n denotes the zeros of the Bessel function: " # s
jn (µm r
n ) = 0. uc is the Bloch function taken at k0 = 0 in the Bril- d2σ 9σ X
louin zone, where (by assumption) the band extrema are located. = √0 g2 + g4 + δ4f
dωs d 2 2 ne me nh mh
The complete electron confinement in the quantum-well wire im- esz
 2
plies χc |r0 = 0.  
× 1(ne − nh )1(me − mh )es · b
Z (328)
The electronic states are described by the quantum numbers:
" # !2
ne , me , ke . The eigenenergies are d2σ 9σ0 h̄ωs
 = √
!  dωs d 4 2 ε0
h̄2  2 µm e 2 e±
s 
εc (ne ; me ; ke ) = ke +
ne  !2  ∓ 2
2m∗e r0
(324) X  β Y
 ne me mh 

×  
n m n m
 1 + β  pe + iδ 
e e h h
For the holes, the analogous quantities are obtained essentially by !2
replacing suffix e by h (labeling hole quantities). 1
×1(nh − ne − 1) +
The initial energy is given by 1+β
 2 
εi = h̄ωl (325) Y∓  
 n m m 
× h h e  1(nh − nh − 1)
 ph + iδ  
The final state involves an electron-hole pair in a real state and a
secondary radiation emitted photon of energy h̄ωs . Hence, " 2 q 4 #
g + g + δ4f 1/2
 !    |es · σ̂± |2
m 2 m !2 × (329)
h̄2  µnee h̄2 µnhh g4 + δ4f
εf = + ke  +
2  + kh 
2
2m∗e r0 2m∗h r0
The following notation for the Kronecker delta is used
+ h̄ωs + εg (326) 
1 x=0
1(x) = (330)
0 x 6= 0
For the intermediate states |ai, the energies εa are easily obtained
from the preceding discussion. The matrix elements that appear σ0 is given by Eq. (291), defined in the electron Raman scattering
in the Eq. (234) are calculated in the same form as that in the in quantum dot. In the same way, the parameters β, µ∗ , and ε0 are
quantum well and quantum dot, therefore it is passed directly to the same as the parameters defined in the spherical quantum dot
the calculations of the differential cross-section. electron Raman scattering calculation.
Additionally,

3.4.1. Differential Cross-section 1 µ∗ m2 µ∗ m2

g2 = (h̄ωl − h̄ωs − ε0 ) − ∗ µnee − ∗ µnhh (331)
ε0 me mh
Using the theory depicted in previous subsections, after tedious ∗  2 
h̄ωs µ m2 m
calculations, explicit expressions for the differential cross-section pe = + ∗ µnee − µnhh
of the electron Raman scattering process can be obtained. In all ε0 me
(332)
calculations, the photon wave vector in comparison with the elec- h̄ωs µ∗  m2 m2 
ph = − ∗ µnee − µnhh
tron wave vector is neglected. Hence, in the final state ke +kh = 0. ε0 mh
 OPTICAL PROPERTIES IN NANOSTRUCTURES 49

and
µn±1;m µnl 1  
Y ± (n; l; m) = σ± = √ b X ± ib
Y (333)
µ2nl − µ2n±1;m 2

The b X; b
Y and b
Z are vectors along the corresponding Cartesian
axes.
In the deduction of Eqs. (327)–(329), it was used that
1 0f
δ(εi − εf ) → (334)
π (εf − εi )2 + 02
f
assuming a finite lifetime for the electron-hole pair in the final
state, while s
0f 0
δf = δ= (335)
ε0 ε0
Let us make some notes concerning the foregoing equations. For
a general scattering configuration, there should be three terms in
the differential cross-section, as explicitly in Eq. (327). However,
for particular choices of the scattering configuration some of these
terms could be absent. For instance, if we have a backscattering
configuration with b Z parallel to the radiation wave vector k, then
Eq. (328) will not contribute to the differential cross-section. In
particular, for the scattering configuration Z(el ; σ± )Z the contri-
bution to the differential cross-section is given by Eq. (329). In
the configuration where scattered radiation wave vector is paral-
lel to the x-axis with polarization es kẑ, i.e., Z(el ; esz )X, only the
first term on the right-hand side of Eq. (327) will be present in
the differential cross-section. In the Z(el ; σ± )Z configuration, the
emission spectrum of electron Raman scattering in a quantum-well
wire shows maxima at the following values of ωs ,
µ∗ ε h m2 m2
i
ωs → ωe (ne ; me ; mh ) = ∗ 0 µnhh − µnee (336)
me h̄
µ∗ ε h m2 m2 i
ωs → ωh (ne ; me ; mh ) = ∗ 0 µnee − µnhh (337)
mh h̄
As can be seen from Eqs. (336) or (337), these frequencies cor-
respond to electron transitions connecting the subband edges for
a process involving just the conduction or just the valence band
(i.e., intraband transitions). The following selection rule is ful- Fig. 41. Electron Raman scattering cross-section (in arbitrary units) for
filled: ne = nh ± 1; the minus sign applies to Eq. (336) and the a GaAs quantum-well wire in the scattering configuration Z(el ; b
Z)X. The
solid curve corresponds to r0 = 4 nm, the dashed curve corresponds to r0 =
plus sign applies to Eq. (337).
3 nm. We have set 0f = 1 meV and 0 = 3 meV. (a) Scattering efficiency
Other singularities of Eq. (329) occur whenever g = 0. Such
singularities are mainly related to certain values of the frequency as a function of (h̄ωl − εg )ε−1
0
(excitation spectrum). The value h̄ωs =
2:3 eV is fixed. (b) Scattering efficiency as a function of h̄ωs /ε0 (emission
ωl of the incident photon. For the excitation spectra, the positions
spectrum). The value h̄ωs = 2:8 eV is fixed. The thresholds indicating
of these singularities are given as different transitions between the valence and the conduction subbands are
1 labeled by (n; m).
ω(ne ; me ; mh ) = (h̄ωl − εg )
ε0
h̄ωs µ∗ m2 µ∗ m2
= + ∗ µnee + ∗ µnhh (338) tion were computed. The physical parameters entering the formu-
ε0 me mh
las are taken for the GaAs case which are well known and given in
Here, the ne = nh + 1 selection rule must be fulfilled. The peaks the electron Raman scattering.
related to the latter singularities correspond to interband electron- In Figure 41(a), the excitation spectrum is shown (scattering
hole pair transitions and their positions depend on the incident ra- efficiency against (h̄ωl − εg )/ε0 ) for the scattering configura-
diation frequency ωl for both excitation and emission spectra. The tion Z(el ; b
Z)X. In this case, just the first term on the right-hand
singularities involved in Eqs. (336) and (337) are independent of side of Eq. (327) contributes. Hence, we do not have singularities.
ωl and correspond to intraband transitions. The latter singularities For certain values of ωl , abrupt changes in the curve slope can
are present just in the emission spectra. be found, which correspond to different thresholds related to the
The emission and excitation spectra of the electron Raman points where a given subband starts to contribute to the differen-
scattering process for a given polarization es of the emitted radia- tial cross-section. This provides the steplike character of the curve.
50 RIERA ET AL.

Fig. 43. Scattering efficiency as a function of (h̄ωl − εg )ε−1

0 (excitation
spectrum) for a GaAs quantum-well wire for the Z(el ; bZ)X scattering con-
figuration. The calculation was performed applying the same parameters as
Fig. 42. Scattering efficiency as a function of h̄ωs /ε0 (emission spec- in Figure 41. Resonant electron-hole interband transitions are indicated by
trum) for a GaAs quantum-well wire for the Z(el ; b Z)Xscattering config- ω(ne ; me ; mh ).
uration. The calculation was performed applying the same parameters as
in Figure 41. Resonant transitions are indicated by ωe (ne ; me ; mh ) and
ωe (ne ; me ; mh ) corresponding to electron or hole intersubband transi-
those of Figure 42. The singular behavior with direct indication
tions, respectively.
of the positions of the singularities is observed. They are given
by expression (338). Again, a threshold for the lower values of
(h̄ωl − εg )ε−1
mh
when µm e
ne and µnh take their minimum values can
The lower admissible value of h̄ωl − εg is defined by the minimum 0
be seen.
value of µm n (i.e., µ01 ). For higher values of h̄ωl − εg , new sub-
The singular structure of the differential cross-section, as given
bands start to contribute, thus defining the other thresholds seen
in the figures, provides a transparent understanding of the energy
in the figure. Explicit indication of the points where the thresh-
subband structure of the quantum-well wire.
olds are present is given, specifying the involved subbands. In
Figure 41(b), the emission spectrum (scattering efficiency against
h̄ωs /ε0 ) is shown for the same scattering configuration. The inci- 3.5. One Phonon-assisted Electron Raman Scattering in
dent radiation frequency was fixed as h̄ωl = 2:8 eV. The rest of Quantum-well Wire, Free Standing Wire, and
the parameters are chosen as in Figure 41(a). In the figure, abrupt Quantum Dot
changes in the slope are observed again, thus providing a certain
steplike dependence of the scattering efficiency. The points where 3.5.1. Introduction
the curve slope presents abrupt changes are related to threshold In the previous subsections, the the electron Raman scattering in
values of h̄ωs again representing the incorporation of new sub- the nanostructures was investigated. However, a complete study of
bands to the process. It should be realized that, for lower values the electron Raman scattering in nanostructures requires consid-
of h̄ωs , more subbands could participate in the emission process. ering the phonons presence in the intersubband transitions, which
2
The condition h̄ωl − h̄ωs − εg > ε0 µm n must be fulfilled to have notably complicates the problem from mathematical point of view.
emission of secondary radiation photons. For fixed values of h̄ωl , This allows a more real description of the electron band structure
εg , and ε0 , the threshold positions are defined by µm n . This is ex- of these systems, so that clarifies the article of the electron-phonon
plicitly indicated in Figure 41(a). interaction in the differential cross-section.
In Figure 42, the emission spectra for the Z(el ; σ− )Z configu- In this subsection, the differential cross-section for electron Ra-
ration. Hence, just the term with the minus sign in Eq. (329) con- man scattering in a semiconductor quantum-well wire and a free
tributes. In this case, we are faced with a singular behavior of the standing wire of cylindrical geometry regarding phonon-assisted
scattering efficiency; the positions of the singularities are defined transition is calculated following Bergues et al. [125].
by Eqs. (336) and (337). The positions of the singular points were Raman scattering experiments are well known to provide a
indicated according to the notation of Eqs. (336) and (337). powerful tool for the investigation of different physical properties
In Figure 43, the excitation spectra for the same scattering con- of semiconductor nanostructures [126–128]. In particular, the elec-
figuration as in Figure 42 is shown. Now, h̄ωs = 2:3 eV and tronic structure of semiconductor materials and nanostructures
h̄ωl is a variable quantity. The other parameters coincide with can be thoroughly investigated considering different polarizations
 OPTICAL PROPERTIES IN NANOSTRUCTURES 51

for the incident and emitted radiation [128, 129]. In connection absorbing the photon can reach the first subbands and from the
with these kinds of experiments, the calculation of the differential latter results the contribution of the second term of the Eq. (339)
cross-section for electron Raman scattering remains a rather in- can be neglected if compared with the contribution of the first one.
teresting and fundamental issue to achieve a better understanding We thus should consider, for the determination of the differential
of the man-made semiconductor nanostructures characterized by cross-section, the contribution of just the first term in the right-
their mesoscopic dimensions [123, 124, 130–133]. hand side of Eq. (339) during the calculation of Mj .
Among the various Raman scattering processes involved in this Then, under these requirements the final state |f i involves an
kind of research electron Raman scattering seems to be a useful excited electron-hole pair in a real state, a secondary radiation
technique providing direct information on the energy band struc- emitted photon of energy h̄ωs .
ture and the optical properties of the investigated systems. Elec- For the intermediate states |ai, |bi, and |ci the energies εa , εb ,
tron Raman scattering is qualitatively explained as a three-step and εc are easily obtained from the previous discussion and from
process: in the first step the system absorbs a photon from the inci- using energy and momentum conservation laws.
dent radiation and an electron-hole pair is created in a virtual state
(after an interband electron transition); in the second step, the
3.5.3. Electron Raman Scattering in Semiconductor
electron and the hole move independently of each other and emit
Quantum-well Wires
one optical phonon performing intraband transitions. In the last
step, the electron and the hole move independently of each other The solution of the Schrödinger equation, in the envelope function
and emit photons of secondary radiation performing intraband approximation, is given, for the wave function, by Eq. (323), and
transitions [123]. In the final state, an excited electron-hole pair, for energy, by Eq. (324). The states are characterized for a set of
a phonon of frequency ωnm , and a photon of frequency ωs appear quantum numbers given by: nj , mj , and kj . The system energy in
in a real state of the system, which is thus left in an excited state. the intermediate states is
Moreover, the differential cross-section for electron Raman scat-  !2 
h̄2  n001 1
tering, in the general case, usually shows singularities related to εa = µm00 + k21 
interband and intraband transitions. The latter result strongly de- 2m∗1 1 r0

pends on the scattering configurations, namely, the structure of the  !2 

singularities is varied when the photon polarization change [134]. h̄2 1
+  µm2n
+ k22  + εg (340)
This peculiar feature of electron Raman scattering allows us to de- 2m∗2 2r
0
termine the subband structure of the system by direct inspection of  
!2
the singularity positions in the spectra. h̄2  n01 1
The present subsection is devoted to the study of the electron εb = µm0 + k21 
2m∗1 1 r0
Raman scattering in the quantum-well wire, free standing wire,
 !2 
and quantum dot, considering as an additional scattering mech-
h̄2  n2 1
anism the electron-phonon interaction. The objective is to investi- + µm2 + k22  + εg + h̄ωnm (341)
gate the same by means of the analysis of the emission spectra. For 2m∗2 r0
such a purpose, the Raman scattering differential cross-section is
calculated considering now that both the electrons and the optical and in the final state the energy is
 !2   !2 
phonon are completely confined.
h̄2  n1 1 h̄ 2 1
+ k1  +  µm + k2 
2 n 2
εf = µm1 2
2m∗1 r0 2m∗2 2r
0
3.5.2. General Model and Applied Theory
+ εg + h̄ωnm + h̄ωs (342)
The differential cross-section for electron Raman scattering per
unit solid angle for incoming light and scattered light is given In the case of the hole contribution, the energy expressions are
by Eqs. (232) and (233), where the probability amplitude in the similar to those of the electron, we only put primes to n2 and m2
Fermi’s golden rule (Eq. (233)) is now calculated in the third or- and we suppress the primes to n1 and m1 .
der of the perturbation theory in the following form, Utilizing the Eqs. (340), (341), and (342), we can evaluate the
denominators in Eq. (339),
bjs |bihb|H
X hf |H bjph |aiha|H
bjl |ii
Mj = h̄2 h n 2  n 2 i
(εi − εb + i0b )(εi − εa + i0a ) εi − εa =
j
µmj − µmj
j
+ h̄(ωs + ωnm ) (343)
a;b 2m∗j r02
bjph |cihc|H
X hf |H bjs |aiha|H
bjl |ii
+ (339) h̄2 h n 2  n0 2 i
j j
(εi − εa + i0b )(εi − εc + i0a ) εi − εb = µmj − µm0 + h̄ωs (344)
a;b 2m∗j r02 j

In Eq. (339), j = 1; 2 for the case of electrons and holes, respec-

tively, |ai, |bi, and |ci are intermediate states with energies εa , εb ,
and εc while 0a , 0b , and 0c are the corresponding lifetime width.
3.5.3.1. The Electron-phonon Interaction Hamiltonian
In this equation, we have not considered the states related to the In the frame of the present treatment, we restrict ourselves to the
“interference diagrams” [114, 119] because they involve a negligi- case of oscillations perpendicular to the wire axis, i.e., qz = 0 and
ble contribution whenever the energy gap is large enough. hence Uz = 0 where q is the phonon wave vector. We are thus
In the initial state |ii, we have a completely occupied valence considering a particular case which, however, entails a direct inter-
band, an unoccupied conduction band, and an incident photon of est for the study of certain physical processes (one phonon Raman
energy h̄ωl . Thus, εi = h̄ωl and is such that the electron after scattering configuration, see Ref. [135], for instance) and give us an
52 RIERA ET AL.

insight into the nature of the oscillations. It is shown that the corre- ω2L − ω2nm ω2T − ω2nm
q2 = Q2 =
sponding eigensolutions constitute a complete orthonormal basis β2L β2T
of eigenvectors Unm and the electron-phonon interaction Hamil-
tonian is derived as [136], and fn (x) represents a solution of the Bessel equation of order n
X h i (this function should be bounded in its domain of definition). It
bph =
H Cnm Fnm (r) exp(inθ)b̂nm + CC (345) must be noticed that ω < ωL for all the frequencies involved; thus
n;m q is always a real quantity. In contrast, Q is real for ω < ωT , but a
where pure complex quantity for ωT < ω < ωL .
"
#1/2 The eigenfrequencies of the oscillation modes for this case are
πωL ρ reported in [137] as
Cnm = r02 CF
ωnm κa∞ + κb∞ h i
s (346) fn−1 (x)fn+1 (y) + fn−1 (y)fn+1 (x)
2
2πe2 h̄ωL −1 !2
CF = (κa∞ − κ−1
a0 ) βL κa∞ 2
V + R
βT y2
ωL is the limiting (bulk) longitudinal optical frequency of the os-  !
cillations; ρ is the reduced mass density; ωnm is the frequency n κ
×fn−1 (x) + fn (x) b∞ − 1 fn+1 (y) = 0 (353)
of the nm mode, the function Fnm (r) has different forms for the x κa∞
quantum-well wire (QWW) and the free standing wire (FSW). For
a QWW, we have that and for the FSW case are given by solving the following secular
 ! !n  equation:
QWW = B  f 1 xr r !2
Fnm nm n − Sn (x)  (347)
x r0 r0 βL
2n(n − 1) R2 tn (x; y)fn+2 (y) + 2ny 2 fn+2 (y)
βT
where h i
 × xfn0 (x) − fn (x) + gn (x)
κa∞ h i
Sn (x) = 1f
κa∞ + κb∞ n n−1
(x) × 2yfn0 (y) + (y 2 − 2n2 )fn (y) = 0 (354)

! 
1 κb∞ − κa∞
+ fn (x) (348) 3.5.3.2. Calculation of the Raman Scattering Intensities in
x κa∞ QWW and FSW
while for a FSW, If we consider allowed direct electron transitions between conduc-
 tion and valence bands, the matrix element ha|H bjl |ii, can be writ-
! !n 
FSW = B  1 f xr r ten as
Fnm nm n − tn (x; y)  (349) s
x r0 r0 D E |e| 2π h̄
bjl |i =
a|H e · P̂cv (0)
m0 Vωl l
where (
" # δn00 n2 δm00 m2 δk00 ;−k2 j = 1
κa∞ × 1 1 1 (355)
tn (x; y) = y 2 fn (x)fn+2 (y) + gn (x)fn (y) δn1 n00 δm1 m00 δk1 ;−k00 j = 2
κb∞ 2 2 2

 !2 ! −1 Physically, this matrix element represents a quantum transition of

κa∞ βL an electron from the valence band to the conduction band, of such
× x2 + y 2 fn+2 (y) (350)
κb∞ βT form that only those transitions between indexed subbands with
the same quantum numbers are allowed. Otherwise, the transi-
and tions are forbidden.
β2L 2 The matrix element of the electron (hole)-phonon interaction
gn (x) = x fn (x) − 2(n + 1)xfn+1 (x) (351) is given by
β2T
D E n0 m0 m
Bnm is a normalization constant and can be determined from bjph |a = Cnm Cj I 00j j00 (r)δ 00 0
b|H (356)
n m n n ;n −n j j
Eq. (34) in Ref. [136]. In this constant, we may include the term j j

ρ from Eq. (346) and it will not appear later. κb∞ is the high fre- being
quency dielectric constant for the AlAs. βL (βT ) is a parameter  2 −1
describing the dispersion of the longitudinal (transverse) oscilla- r0 0 n0j 0 n00j
Cj = J 0 (µ 0 )J 00 (µ 00 ) (357)
tions. 2 nj mj nj mj
On the other hand, Z !
n0j m0j m 1 r0 n0j 0 r
In00 m00 n (r) = rJn0 µm0 Fnm (r)Jn00
x = qr0 y = Qr0 j j r0 0 j j r0 j
!2 !2 !
r0   βT n00j r
ω2L − ω2T = R2 = x2 − y2 (352) × µm00 dr (358)
βL βL j r0
 OPTICAL PROPERTIES IN NANOSTRUCTURES 53

with Adding both contributions, we obtain

(
QWW
Fnm (r)
Fnm (r) = (359) M1 + M2
FSW
Fnm (r) 



 C n1 m1 I n1 m1 m (r)k1 r0
Physically, the expression (356) indicates to us that intersubband A0 Cnm  n2 m02 n2 m2 n
e
= ! i
transitions assisted by phonons are possible when the relationship r0 µ∗ 
 h̄ωs
between subbands satisfies the condition n00j = n0j − n. Otherwise, 
 + iδb

ε0
the transitions are forbidden. " #
For the electron-(hole)-secondary radiation interaction matrix β1 δn1 ;n2 +n β2 δn2 ;n1 +n
× + (es · Z)
element, we have p1a + iδa p2a + iδa
s "
D E (−1)j+1 ih̄e 2π h̄ X2
β2 δn2 ;n1 +n∓1
b
f |Hjs |b = +
m∗j Vωs p2a + iδa
l=1
(
es · σ− + n2 ±1;m02 n2 ±1;m02 m
× ik0j es · b
Zδnj n0 δmj m0 − Yn m m0 δn0 ;nj +1 X Cn2 m1 In1 m1 n (r)Yn± m m0
r0 2 2 2
j j j j j j
×
) (p2b + iδb )
es · σ+ − m02
− Yn m m0 δn0 ;nj −1 δkj k0 (360)
r0 j j j j j β1 δn1 ;n2 +n∓1
−
p1a + iδa
being n1 ±1;m01 n1 ±1;m01 m
X Cn2 m2 In2 m2 n (r)Yn± m m0
1 1 1
± µn±1
m µl
n b
X ± ibY ×
Y(n;l;m) = σ± = √ (361) (p1b + iδb )
n2 (n±1)2 2 m01
µl − µm 

where b X, bY, b
Z are unitary vectors along the corresponding Carte- × (δl1 + δl2 )(e · σ∓ ) δk1 ;−k2 (363)

sian axes. On the other hand, the vector σ+(−) represents the ro-
tation to the right (left) in the plane XY .
Equation (360) shows that in the transition between the states where l = 1(2) are associated with σ− (σ+ ), pja and pjb are eval-
|bi and |f i three excluding polarizations appear: in the first one the uated with the corresponding deltas. The term A0 is determined
electron (hole) spreads freely in the same subband, in the second by
and third ones the transitions are made under the requirements 2πih̄2 e2 el · Pcv (0)
A0 = √ (364)
n0j = nj + 1, n0j = nj − 1, respectively. In all the cases, the change ε20 m0 V ωl ωs
of state is carried out keeping constant the momentum of the elec-
tron (hole). The treatment of the function δ(εf − εi ) is carried out according
to the previous subsection, and can be shown that
v q
u
3.5.3.3. Differential Cross-section X u g2 + g4 + δ2
L u f
δ(εf − εi ) = √ t
We can obtain, after cumbersome calculation, explicit expressions 2 2πr0 ε0 g + δf
4 2
for the probability amplitude, k1
s r
(365)
X y/L
2πih̄2 e2 el · Pcv (0) k21 δ(εf − εi ) = √ g2 + g4 + δ2f
Mj = √ 2 2πr 3ε
ε20 m0 V ωl ωs k1 0 0

n0 m0 n0 m0 m
enm C 00j j00 I 00j j00 (r)
C being
X nj mj nj mj n
×    
m∗j (pjb + iδb )(pja + iδa ) h̄ n 2 n 2
n00j m00j n0j m0j
g2 = (ωl − ωs − ωnm − ωg )−β1 µm11 −β2 µm22 (366)
ε0
(
Here, the photon wave vector is neglected as is compared to
× ik0j (es · b
Z)δnj n0 δmj m0 + (−1)j+2
j j the electron wave vector. Hence, in the final state we have:
" (ke + kh = 0).
es · σ− +
× Yn m m0 δn0 ;nj +1 The final result is
r0 j j j j
#) d2 σ
es · σ+ − = σ0 [Wesz + We+ + We− + We+ e− ] (367)
+ Yn m m0 δn0 ;nj −1 d dωs s s s s
r0 j j j j
 where
× δn0 ;n2 +n δn00 ;n2 δm00 ;m2 δ1j √
1 1 1
 2 2e6 |el · P̂cv |2 η(ωs )ωL ωs h̄4  −1 
σ0 = κ∞ − κ−1 (368)
+ δn0 ;n1 +n δn00 ;n1 δm00 ;m1 δ2j δk1 ;−k2 (362) ∗2 5 2 4
µ ε0 m0 c η(ωl )ωl r0 3 0
2 2 2
54 RIERA ET AL.
 !2
X  ω n m m
In21m21n (r) Let us make some notes concerning the preceding equations.
L
Wesz = For a general scattering configuration, we should have four terms
 ωnm J(n1 m1 n2 m2 )
n1 m1 n2 m2 nm in the differential cross-section, as is explicitly seen in Eq. (367).
r q However, for particular choices of the scattering configuration
g2 + g4 + δ2f some of these terms could be absent. For instance, if we have a
× !2 (369) backscattering configuration with Z parallel to the radiation wave
h̄ωs vector k, then the Eq. (369) will not contribute to the differen-
+ δ2b
ε0 tial cross-section. In particular, for the scattering configuration
 2  Z(el ; σ± )Z, the contribution to the differential cross-section is
β δ β2 δn2 ;n1 +n  
 1 n2 ;n1 −n given by (370) and (371). In the configuration where scattered
× +  |es · Z|2
 p1a + iδa p2a + iδa   radiation wave vector is parallel to the x-axis with polarization
X ωL es //Z, i.e., Z(el ; σsz )X only the first term at the right-hand side
We∓ = |Me∓ |2 of Eq. (367) will be present in the differential cross-section. This
s
n1 m1 n2 m2 nm
ωnm s
result is in close analogy with the case of electron Raman scat-
v q
u tering in cylindrical quantum-well wires [124]. In Eqs. (370) and
u g2 + g4 + δ2
u f (371), the summations over the labels (n1 ; m1 ; n2 ; m2 ; n; m) must
×t |es · σ∓ |2 (370)
g4 + δ2f be done under the following requirements:
   
h̄ n 2 n 2
and (ωl − ωs − ωnm − ωg ) − β1 µm11 − β2 µm22 ≥ 0 (378)
ε0
X ωL
We+ e− = 4 In the Z(el ; σ± )Z configuration with n 6= 0, the emission spec-
s s ω
n1 m1 n2 m2 nm nm
h i tra of electron Raman scattering in a quantum-well wire show max-
× (Re Me− s )(ReMe + ) + (Im Me− )(Im Me+ ) δ
s s s n0
ima at the following values of ωs ,
v q βj ε0 h n2 2  n1 2 i
u j
u g2 + g4 + δ2 ωs = (−1)j+1 µm2 − µm1 − ωnm (379)
u f h̄
×t (es · σ− )(es · σ+ ) (371)
g4 + δ2f and
j βj ε0 h n0j 2  nj 2 i
ωs = µm0 − µmj (380)
where h̄ j

( ± As can be seen from Eqs. (379) or (380), these frequencies cor-

X
2 X βj Ij Ynj mj m0j δn0j ;nj ±1
Me∓ = (−1)j respond to electron transitions connecting the subbands edges for
s Jj (pjb + iδb ) a process involving just the conduction or just the valence band
j=1 n0j m0j
) (i.e., intraband transitions). The values of frequencies reported in
δ1j δn2 ;−n+n1 ±1+ δ2j δn2 ;n+n1 ∓1 Eqs. (379) and (380) are associated to the emission of a phonon
× (372)
(pja + iδa ) (photon) by the electron or the hole in intersubband transitions.
The selection rules are n2 = n1 ± 1 − n for the electron transi-
being tions and n2 = n1 ∓ 1 + n for the hole transitions, both related
n0 m0 m n0 m0 m
to Eq. (379). We also observe in Eq. (380), that for n01 = n1 ± 1
I1 = In21m21n (r) I2 = In12m12n (r) (n02 = n2 ± 1) electron (hole) transitions are allowed. This selec-
(373)
J1 = J(n01 m01 n2 m2 ) J2 = J(n1 m1 n02 m02 ) tion rule corresponds to e±s polarization. The singularities involved
in Eqs. (379) and (380) are independent of ωl and correspond to
and intraband transitions.
J(n1 m1 n2 m2 ) = Jn0 1 (µm11 )Jn0 2 (µm22 )
n n
(374) Other singularities in Eqs. (370) and (371) occur whenever
g = 0. Such singularities are mainly related to certain values of
Besides, the frequency ωl of the incident photon. For the emission spectra,
h    i h̄(ω + ω ) the positions of these singularities are given as
n 2 n 2 s nm
pja = (−1)j+1 βj µm11 − µm22 + (375)
ε0 β1 ε0 n1 2
h n 2  n0 2 i h̄ω ωs (n1 ; m1 ; m2 ; n; m) = ωl − ωnm − ωg − (µm1 )
j j s h̄
pjb = βj µmj − µm0 + (376)
ε0 β ε n
− 2 0 (µm22 )2
j
(381)
h̄
where ε0 , µ∗ , and βj were previously defined.
Assuming a finite lifetime for the electron-hole pair in the final Here, the same selection rule as that of Eq. (379) must be fulfilled.
state, The peaks related to the latter singularities correspond to inter-
0f 0a 0 band electron Raman scattering transitions and their positions de-
δf = δa = δb = b (377) pend on the incident radiation frequency ωl .
ε0 ε0 ε0
In the case of scattering configuration Z(el ; σsz )X, the follow-
In Eq. (376), four terms can be observed, in contrast with Eq. (10) ing selection rules are satisfied: n2 = n1 −n for electron transitions
of Ref. [124] where only the first three exist; the last term is related while n2 = n1 +n for the hole transitions. In this case, the emission
with the confined phonons and only have sense for n = 0, where spectra of electron Raman scattering show maxima for the values
the LO and TO phonon modes are uncoupled (see Eq. (371)). of ωs given in (379).
 OPTICAL PROPERTIES IN NANOSTRUCTURES 55

Fig. 44. Scattering efficiency (SE) is plotted in σ0 units versus h̄ωs /ε0 . Fig. 45. Same as in Figure 44 but with n = 1 and two values of r0 .
The solid curve corresponds to quantum-well wire and the dashed curve
corresponds to a free standing wire for qx = 0, h̄ωl − εg = 31ε0 , and
0f = 3 meV, 0a = 0b = 1 meV. Here, the curves 1 and 3 correspond
to n = 2, while the remainder corresponds to n = 1. Here, the scattering
configuration is Z(el ; σ+ )Z and r0 = 3 nm.

We focus our attention on the optical phonon dispersion rela-

tions of the GaAs quantum-well wire and free standing wire semi-
conductors. These relations for three values of n(n = 0; n = 1,
n = 2) were illustrated previously for the same systems in Figure 1
of [136]. Our interest in analyzing the phonon modes n = 1 and 2
in the expression for the differential cross-section rests in the fact
that they are closely linked to the coupled phonon modes.
We computed the scattering efficiency (SE) against h̄ωs /ε0 , the
so-called “emission spectra” of the electron Raman scattering pro-
cess for a given polarization es of the emitted radiation. The phys-
ical parameters entering in our formulas were for the GaAs which
were defined previously. We set 0f = 3 meV, 0a = 0b = 1 meV,
and h̄ωl − εg = 31ε0 . (See Fig. 44.)
In all the plots, the solid lines represent the results for the
quantum-well wire while the dashed lines represent those for the
free standing wire. Electronic and hole singularities due to pho-
tons have the same position for both. Fig. 46. Same as in Figure 44 but with Z(el ; esz )X scattering configura-
We have shown not only the resonant electron transitions at all tion.
spectra, but also a few hole transitions. The steplike that appears in
Figure 2 of Ref. [124] is not shown for sake of clarity of the figures.
Resonant electron transitions mediated by photon (phonon) are In Figures 45–48, the peaks where the electronic transitions
indicated by ωes (n1 ; m1 ; n01 ; m01 ); (ωeph (n1 ; m1 ; n2 ; m2 )). In the mediated by phonons appear are marked with the frequency ωeph ,
hole case, we have ωh 0 0
s (n2 ; m2 ; n2 ; m2 ); where (n1 ; m1 ; n2 ; m2 ) where the last index m takes several values. For example, the peaks
represents the subband structure. When we want to indicate which 2 and 3 (Fig. 45), correspond to a group or set of peaks displaced
phonon modes produce the transitions, ωeph (n1 ; m1 ; n2 ; m2 , very slightly in the value of the ωs and by scale reasons they are ob-
n; m) may be written. served as a single peak, although in fact with a more appropriate
For a fixed energy of the incident radiation h̄ωl , if the nm scale they would be observed at all.
phonon modes increase when m increases, the differential cross- In all spectra, we can note that the differential cross-section
section tends to be insignificant, hence we chose the radius of the in quantum-well wire is much smaller than the differential cross-
quantum-well wire between 2 and 3 nm, while the anticrossing section in free standing wire for any value of n and r0 . This is a
modes are between 27 and 31.4 meV. consequence of the following fact: the phonon oscillations can be
56 RIERA ET AL.

1 and 3 correspond to n = 1 and the curves 2 and 4 to n = 2,

observing that for the transitions by photons of the electron or the
hole the change in the phonon modes only affects the intensity of
the peak, but not the frequency ωs . The first statement is due to
the fact that, when radius r0 is fixed and n increases the energy of
the phonon modes diminish (see Fig. 1 of Ref. [136]), thus increas-
ing the scattering efficiency, because SE and the frequency modes
are inversely proportional (see Eq. (370)). The second statement
is due to fact that secondary radiation frequency does not depend
on the phonon modes (see Eq. (380)).
In particular, in Figure 44 we observe a set of three peaks
corresponding to four singularities at ωh h
s (2; 2; 1; 1); ωs (1; 1; 0; 3),
e e
ωs (0; 1; 1; 1), and ωs (1; 1; 2; 1). The second is almost at the same
position, i.e., h̄ωs /ε0 = 7:7514, and h̄ωs /ε0 = 7:7530. This
displacement occurs because the frequency depends on the sub-
bands and on electron and hole effective mass; since the effec-
tive mass of the electron is lower than that of the hole, it causes
that the frequency to be slightly different. The positions of singu-
larities are defined by Eq. (380) and have the characteristics dis-
cussed previously. In general, for the model of a quantum-well
wire with infinite potential barriers the position of the singular
peaks ωs (n1 ; m1 ; n2 ; m2 ) will change with the variations of the
Fig. 47. Same as in Figure 46 but with n = 1 and two values of r0 . quantum-well wire radius according to the law ≈ r0−2 . For higher
energies h̄ωl of the incident radiation photon, we observe a large
number of singular peaks in the emission spectra as can be de-
duced from Eq. (380) because new subbands are accessible to the
electron; similar results can be observed in Ref. [124]. The sin-
gularities ωes (n1 ; m1 ; n01 ; m01 ) and ωh 0 0
s (n2 ; m2 ; n2 ; m2 ) have the
same position because we take into account an infinite potential
barrier model and according to that, the electronic behavior is the
same.
In Figure 45, we considered the same scattering configura-
tion as in Figure 44. In this figure, we can observe that, for n
fixed, when r0 increases the SE diminishes. The latter is due to
increasing the radius of the QWW (FSW). The ωnm frequency
modes increase (see Fig. 1 of Ref. [136]) and these terms are
inversely proportional to SE (see Eq. (370)). The SE displays
three sets of peaks at ωes (1; 1; 2; 1), ωeph (0; 1; 0; 2; 1; m) where
m = 1; 2; 3; 4; 5 and ωeph (0; 1; 0; 2; 1; m) with m = 1; 2: They are
slightly shifted for a QWW in comparison with FSW. The phonon
peaks for a QWW [FSW] are the following: ωeph (0; 1; 0; 2; 1; 1)
with h̄ωs /ε0 = 21:0151 [21:0160]; ωeph (0; 1; 0; 2; 1; 2), h̄ωs /E0 =
21:0193 [21:0212]; ωeph (0; 1; 0; 2; 1; 3), h̄ωs /E0 = 21:0255
[21:0282]; ωeph (0; 1; 0; 2; 1; 4), h̄ωs /ε0 = 21:0337 [21:0374]; and
ωeph (0; 1; 0; 2; 1; 5), h̄ωs /ε0 = 21:0440 [21:0453]; ωeph (0; 1; 0; 2,
Fig. 48. Same as in Figures 44–47 but with n = 1 and r0 = 2 nm. Both
scattering configurations are included. 1; 1), h̄ωs /ε0 = 21:2911 [21:2921]; and ωeph (0; 1; 0; 2; 1; 2),
h̄ωs /ε0 = 21:2955 [21:2973].
The relative positions of the singularities associated with the
considered as free on the surface S in the free standing wire and emission of a phonon depend on the radius r0 for any nm modes,
QWW
the condition σ · N = 0 for r ∈ S is required. Then the vector U is as can be seen in Eq (379); always the ωFSW
nm is smaller than ωnm
only defined in the active medium (see Ref. [136]), hence the term (although only slightly). When the nm phonon mode is fixed and
FSW  F QWW .
Fnm r0 increases the value of the ωeph diminishes. This explains why in
nm
In Figure 44, we show the SE for Z(el ; σ+ )Z scattering config- the second peak a transition is made to a lower frequency that in
uration and r0 = 3 nm, n 6= 0 modes. In this case, just the term the third one, even when the same subbands are involved. If we
e+
s in Eq. (367) contributes. The purpose is to show that the elec- compare the ωes (1; 1; 2; 1) transition of Figure 45 with the same
tronic and the hole transitions assisted by photons vary for a same transition displayed in Figure 44 (curves 2 and 4), we observe that
phonon mode (or different modes) when the polarization and the when the radius r0 increase with nm modes fixed, the SE increase
radius are fixed. Notice that in the third group of peaks the curves because the energy of the secondary radiation decreases.
 OPTICAL PROPERTIES IN NANOSTRUCTURES 57

In Figures 46 and 47, we show the SE for Z(el ; esz )X scattering 1. The relative positions of the singularities depend on the
configuration. In this case, the term esz in Eq. (367) contributes. radius r0 and the nm modes; i.e., when the radius
In Figure 46, the SE displays two sets of peaks corresponding to decreases the SE increases, while r0 increases, the values
the transitions ωeph (2; 1; 0; 2; 2; m) with m = 1; 2; 3; 4; n = 2 and of ωeph do not change.
ωeph (1; 1; 0; 2; 1; m) with m = 1; 2; 3; 4; 5; n = 1. If we keep the 2. The values of ωeph for FSW and QWW are approximately
value of the radius r0 = 3 nm fixed when the phonon mode in- equal, but they are slightly shifted.
creases, different subbands become involved in the transitions and Finally, in the present subsection we applied a simplified model
the SE also increases. Here, we also observe the slight displace- for the electronic structure of the system. In a more realistic case,
ment between transitions belonging to the FQW and to the QWW we should consider coupled band structure using a calculation
when these are through phonons (the same as in Fig. 45). In Fig- model like that of Luttinger–Kohn or the Kane model. We also
ure 47, we keep the value of n fixed and the radius is varied, observ- assumed an infinite potential barrier for the electron at the QWW
ing that the same subbands are involved in the transitions, (with interface. A calculation assuming a finite barrier would be better,
different energies, since the radius is not the same) and the SE but it is also possible to introduce a certain redefined effective
increase when the radius decreases. mass for the infinite-barrier case leading to the correct energy lev-
In Figure 48, the SE is shown with n = 1 and r0 = 2 nm els for the electrons and the holes [138]. The previously mentioned
for the Z(el ; esz )X and Z(el ; e+ )Z scattering configurations. The assumptions would lead to better results but entail more compli-
change of polarization in a same system (keeping the same oscil- cated calculations. However, within the limits of this simple model
lation modes) produces that different subbands participate in the we are able to account for the essential physical properties of the
transitions with different frequencies that are larger for the case of discussed problem. The fundamental features of the DCS, as de-
Z(el ; e+ )Z configuration. scribed here, should not change very much in the real QWW case.
The electronic or hole transitions reported in Figures 44 and It can be easily proved that the singular peaks in the DCS will be
45 correspond to the Z(el ; e+ )Z scattering configuration. These present irrespective of the model used for the subbands structure
are only allowed when the following requirements are satisfied: and may be determined for the values of h̄ωs equal to the energy
n0j = nj + 1 with j = 1; 2 and n2 = n1 + 1 − n (for the electron) difference between two subbands: h̄ωs
e(h) e(h) e(h)
= εα − εβ , where
and n2 = n1 − 1 + n (for hole), of such form that: e(h) e(h)
εα (εβ ) are the respective electron (hole) energies in the sub-
• The relaxation intrasubband is not possible because qz = 0, bands. Similarly, we shall have a steplike dependence in the DCS
therefore the transitions are vertical. for h̄ωl = h̄ωs + h̄ωnm + εg + εα + εβ .
• The transitions between subbands of a same band are not
possible since, for nj fixed, there are no transitions between
subbands with m different due to the selection rule for nj . 3.5.4. One Phonon-assisted Electron Raman Scattering in
• The transitions between subbands of different bands are Spherical Semiconductor Quantum Dots
possible with equal and different mj whenever they satisfy the The differential cross-section for an electron Raman scattering
selection rule previously quoted. process in a semiconductor quantum dot of spherical geometry
regarding phonon-assisted transitions, is calculated for T = 0 K
For the Z(el ; esz )X scattering configuration, the selection rule
[139]. We present a short description of the electron-confined
is given as n2 = n1 − n for the electronic transitions, while n2 =
phonon interaction of spherical structures embedded in another
n1 + n for hole transitions.
material, including a correct treatment of the mechanical and elec-
This latter scattering configuration constitutes a difference with
trostatic matching conditions at the surface. We illustrate the the-
respect to the differential cross-section reported in Ref. [124], the
ory for a GaA–AlAs system. Electron states are considered to be
peculiarity of presenting terms that perform electronic or hole
completely confined within the quantum dot. We also assume sin-
transitions, which is a consequence of the electron-phonon inter- gle parabolic conduction and valence bands. The emission and
action. excitation spectra are discussed for different scattering configu-
From the dispersion relation, curves as a function of the wire rations and the selection rules for the processes are also studied.
radius r0 , for different values of n corresponding to the QWW and Singularities in the spectra are found and are interpreted. The one
the FSW cases [136], within this model, we note the following as- phonon-assisted electron Raman scattering studied here can be
pects: used to provide direct information about the electron band struc-
In the case of the resonant electron transitions labeled as ture of these systems.
ωes (n1 ; m1 ; n2 ; m2 ) we can say that: Among the various Raman scattering processes involved in this
1. The relative positions of the singularities, kind of research, electron Raman scattering seems to be a use-
h̄ωes (n1 ; m1 ; n2 ; m2 )/ε0 do not depend of the radius r0 ; ful technique providing direct information about the energy band
the incident energy, and n. structure and the optical properties of the investigated systems.
2. The QWW’s singularities position is the same as that for Electron Raman scattering is qualitatively explained as a three-
the FSW. step process: in the first step the system absorbs a photon from
3. When the radius r0 increases for n fixed, so does the the incident radiation and an electron-hole pair is created in a vir-
differential cross-section. tual state (after an interband electron transition); in the second
step the electron and the hole move independently of each other
Of course, in such a statement, a compromise between the and emit one optical phonon performing intraband transitions.
value of r0 and n should be found. In the last step, the electron and the hole move independently
In the case of resonant electron transitions labeled as ωeph (n1 , of each other and emit photons of secondary radiation perform-
m1 ; n2 ; m2 ) we can say that: ing intraband transitions [123]. In the final state, an electron-hole
58 RIERA ET AL.

pair, an excited phonon of frequency ωnp kp , and a photon of fre- h̄(ωs + ωnp kp )
+ (386)
quency ωs appear in a real state of the system, which is thus left in ε0
an excited state. Moreover, the differential cross-section for elec- h n +1/2 2  n0 +1/2 2 i h̄ω
j j s
tron Raman scattering, in the general case, usually shows singu- pjb = βj µkj − µk0 + (387)
j ε0
larities related to interband and intraband transitions. The latter
result strongly depends on the scattering configurations, namely, Considering that the energy h̄ωl of the radiation incident, is such
the structure of the singularities is varied when the photon polar- that the electron upon absorbing the same can reach the first sub-
ization change [134]. This peculiar feature of electron Raman scat- bands and from the latter results the contribution of the second
tering allows us to determine the sub-band structure of the system term of Eq. (234) can be neglected if compared with the contribu-
by direct inspection of the singularity positions in the spectra. tion of the first one.
The electron-hole Coulombic interaction is neglected since the We thus should consider, for the determination of the differ-
strong confinement regime was assumed, i.e., the radius of the ential cross-section, the contribution of just the first term in the
quantum dot is smaller than the corresponding effective exciton right-hand side of Eq. (234) in the calculation of Mj .
Bohr’s radius in this material. In this context, the Coulombic inter-
action between electron and hole represents a small perturbation.
3.5.4.2. Electron-phonon Interaction Hamiltonian
In the case of a GaAs–Gax Al1−x As quantum dot, the major
band offset can be obtained when x = 0, thus justifying, approxi- It is shown that the corresponding eigensolutions constitute a com-
mately, the assumption of complete confinement made before. plete orthonormal basis of eigenvectors Unp kp and the electron-
Regarding the complex valence band of III-V and II-VI semi- phonon interaction Hamiltonian is derived using the second quan-
conductor compounds, the contribution of band mixing is small tization formalism and can be written as [141],
in the confinement regime assumed in this work and is consistent r
np
with previous treatments [63, 140]. When the excitation energy is X X X 4π
bph =
H r0 CF 8 Y m (θ; ϕ)
close to the bandgap energy, the valence band structure is very im- 3 np ;kp np
portant in the strong confinement regime and only one conduction np =0 kp m=−np
h i
band can be assumed since the electronic contribution to the pro- × b̂kp np m + b̂+
kp np m
(388)
cess is negligible, as was pointed out by Fomin et al. [73], however
this is not the case because the excitation energy considered here- where CF was defined previously (see Eq. (346)). The functions
in is much higher than the bandgap energy. 8np kp (for r < r0 because infinite barrier potential was assumed)
are given as
3.5.4.1. Model and Applied Theory 8np kp (r)
Let us briefly describe the model and fundamental theory applied √
r0
in the calculations. The quantum dot geometry is spherical of ra- =
|u|
dius r0 . As explained earlier, we consider a single conduction (va-  !
lence) band, which is split into a subband system due to com-  r
plete electron confinement within the structure. The solution of × jnp kp
 r0
Schrödinger equation, in the envelope function approximation is
!np 
given by Eq. (262). kp jn0 p (kp ) + (np + 1)κb∞ /κa∞ jnp (kp ) r 
The final state |f i involves an electron-hole pair excited in a − (389)
np + (np + 1)κb∞ /κa∞ r0 
real state, a secondary radiation emitted photon of energy h̄ωs .
Hence,
for np = 0 the values of kp are given by
h̄2  h̄2  n2 +1/2 2
n +1/2 2
εf = µk1 + µ tan kp = kp (390)
∗ 2
2m1 r0 1 2m∗2 r02 k2
while for np = 1 and phonon frequencies ω above the TO phonon
+ εg + h̄ωnp kp + h̄ωs (382)
frequency ωT one needs to solve
The electron contribution to the intermediate states is given by h i
2kp cos kp + (k2p − 2) sin kp
h̄2  n00 +1/2 2 h̄2  n2 +1/2 2 
εa = µk100 + µ + εg (383) ! ! !
∗ 
2
2m1 r0 2m∗2 r02 k2 γ 3 κ 1
× − 0 
1 b∞
−1 + 1+2 1−
 v2 v2 κa∞ k2p
h̄2  n01 +1/2 2 h̄2  n2 +1/2 2
εb = µk0 + µ  ! 
∗ 2
2m1 r0 1 2m∗2 r02 k2 
3γ 1 κ
× sin v + − 0
+ 1+2 b∞  cos v
+εg + h̄ωnm (384) v3 v κa∞ 
The difference of energy in the denominators in Eq. (234) is given  " #
h i 1 κb∞ γ0 κb∞
by = 2 sin kp − kp cos kp −1 − 2 +3 cos v
εi − εa = ε0 pja εi − εb = ε0 pjb (385) v κa∞ v2 κa∞
 ! ! 
where 
h   n00 +1/2 2 i 1 κb∞ 3 κb∞
n +1/2 2 +  γ0 1 − + 1+  sin v (391)
pja = (−1)j+1 βj µk1 − µk100 v2 κa∞ v3 κa∞ 
1 1
 OPTICAL PROPERTIES IN NANOSTRUCTURES 59
" #
with n0j
(2n0j + 1)(2np + 1)(2n00j + 1)
= (−1) S (398)
!2 4π
ω2L − ω2T kp
γ0 = ω2 = ω2L − β2L Zr0 !
β2T r0 1 n0j +1/2 r
!2 γj0 = rJn0 +1/2 µk0
r0 j j r0
v 0
ω2 = ω2T − β2T (392) !
r0 n00j +1/2 r
× φnp kp (r)Jn00 +1/2 µk00 dr (399)
and βL (βT ) is a parameter describing the dispersion, assumed to
j j r0
be parabolic for the LO (TO) phonon in the bulk. The analyti- being S the 3j symbols given by
cal expressions for the solution of the np ≥ 2 modes are some-  0  0 
nj np n00j nj np n00j
what more complicated and will not be shown here. They can be S= (400)
obtained from the general secular equation given by Eq. (30) in m0j mp m00j 0 0 0
Ref. [141]. In Eq. (389), the term |u| represents the norm of the which is different from zero for: n0j + np + n00j even and −m0j +
vibrational amplitude and is equal to
mp + m00j = 0. In Eq. (397), Jj is
|u|2 h  n +1/2   n0 +1/2 i−1
 ! ! Jj = Jn0 +1/2 µk Jn0 0 +1/2 µk
j j
Z r0  (401)
d r np + 1 r 1 j j
=  − jnp kp + pGnp v
1

0  dr r0 r r0 For the electron (hole)-secondary radiation interaction matrix el-

!np −1 2  ! ement, we have
np t r (np + 1) D E D E D E D E
−  + −jnp kp r bjs |b = f |H
f |H bjs |b b b
+ + f |Hjs |b − + f |Hjs |b (402)
r0 r0 r2 r0 es es esz

!! !np 2  where
s

+
p d
rGnp v
r
−t
r  r 2 dr (393) D E |e| 2π h̄
np dr r0 r0  bjs |b
f |H = (−1)j+1 ih̄ IJ
el r0 m∗j Vωs j j
h i
with × Mj1l δn0 ;nj +1 + Mj2l δn0 ;nj −1
j j
kp jn0 p (kp ) − np jnp (kp ) 
 δ 0 ;m +1 es · e+ l = 1
p =  m
 j j
np Gnp (v) − vG0np (v) δ · l = 2
× e
mj ;mj −1 s −
0 e (403)
0 

γ [kp jnp (kp ) + (np + 1)(κb∞ /κa∞ )jnp (kp )]  δm0 ;m es · z l=3
t = 0 (394) j j
v2 np + κb∞ /κa∞ (np + 1)
being
and the function Gnp (z) is defined as
n0j +1/2 n0j +1/2 nj +1/2 0 nj +1/2
( µk0 Jnj +1/2 (µk0 )µk Jn +1/2 (µk )
Jnp (z) z2 > 0 Ij =
j j j j j
(404)
Gnp (z) = (395) nj +1/2 2 n0j +1/2
Inp (z) z2 < 0 (µk ) − (µk0 )2
j j
Inp (z) being the modified spherical Bessel function. and
1 (nj − mj )!
Mj11 = √ = Mj21
3.5.4.3. Calculation of the Raman Scattering Intensities 2π 2 (nj + mj )!
√
The matrix element ha|H bjl |ii is the same as that calculated in the Mj13 = 2(nj − mj + 1)Mj11
previous subsection. The matrix element for intersubband transi- Mj22 = (nj + mj − 1)(nj + mj )Mj11 (405)
tions due to electron-(hole)-phonon interaction is given by
Mj12 = (nj − mj + 1)(nj − mj + 2)Mj11
D E r √
bjph |a = 4π r0 CF 00 30 γ 0
b|H (396) Mj23 = 2(nj + mj )Mj11
3 2π j j j
The two contributions to the probability amplitude can be ex-
where pressed as
" #
(n0j + m0j ) 1/2 ih̄2 e2 CF el · Pcv
00j = (2n0j + 1) Mj = (−1)j+1
(n0j − m0j ) Vm0 ε20
" # s
(n00j + m00j ) 1/2 4π X 00j γj0 30j Ij Jj
00
× (2nj + 1) 00 Jj ×
(nj − m00j )
(397) 3ωl ωs m∗j
n00j m00j k00j n0j m0j k0j
Z
m0j mp m00j (δm00 m2 δn00 n2 δk00 k2 δ1j + δm00 m1 δn00 n1 δk00 k1 δ2j )
30j = Yn0 Ynp Yn00 d
j j × 1 1 1 2 2 2

 (pja + iδa )(pjb + iδb )

60 RIERA ET AL.

m∗1 m∗2 h̄2
× (Mj11 + Mj21 )δm0 ;mj +1 es · e+ µ∗ = ε0 = (413)
j
m∗1 + m∗2 2µ∗ r02
+ (Mj12 + Mj22 )δm0 ;mj −1 es · e−
j
 Now, 01 , γ1 , 31 (02 ; γ2 ; 32 ) have the same form of Eqs. (397)–
+ (Mj13 + Mj23 )δm0 ;mj es · Z (399), only that we should change the double indexes n001 , m001 , k001
j
(n002 ; m002 ; k002 ) by n2 , m2 , k2 (n1 ; m1 ; k1 ).
× (δn0 ;nj +1 + δn0 ;nj −1 ) (406) Being
j j

The total probability amplitude is given by 11 = (p1a δ1j + p2a δ2j )pj − 0a 0b (414)
12 = (p1a δ1j + p2a δ2j )0b − 0a pj (415)
M1 + M2
X
2 X 00j γj0 30j Ij Jj Assuming a finite lifetime for the electron-hole pair in the final
= A0 (−1)j+1 state the quantities δ(εf − εi ); δf ; δa ; δb have the same way as
m∗j (pja + iδa )(pjb + iδb )
j=1 n0j m0j k0j those in the previous subsection.
h Let us make some notes concerning the foregoing equations.
× (Mj11 + Mj21 )δm0 ;mj +1 es · e+ (407) For a general scattering configuration, we should have three terms
j
in the differential cross-section, as explicitly seen in Eq. (409).
+ (Mj12 + Mj22 )δm0 ;mj −1 es · e−
j However, for particular choices of the scattering configuration
i
+ (Mj13 + Mj23 )δm0 ;mj es · Z (δn0 ;nj +1 + δn0 ;nj −1 ) some of these terms could be absent. For instance, if we have a
j j j backscattering configuration with Z parallel to the radiation wave
in which the terms with double superindexes were evaluated and vector k, then the third term of Eq. (409) will not contribute to
the term A0 is defined as the differential cross-section. In the configuration where scattered
s radiation wave vector is parallel to the x-axis with polarization
ih̄2 e2 CF el · Pcv 4π es //Z, i.e., Z(el ; esz )X only the third term of Eq. (409) will be
A0 = (408)
Vm0 ε 2 3ωl ωs present in the differential cross-section. This result is in close anal-
0
ogy with the case of electron Raman scattering in the spherical
After cumbersome calculation, explicit expressions for the differ- quantum dot [123]. In Eq. (409), the summations over the labels
ential cross-section of the electron Raman scattering process is ob- (n1 ; m1 ; k1 ; n2 ; m2 ; k2 ; np ; mp ; kp ) must be done under the fol-
tained lowing requirement:
! ! !
d2σ d2σ d2σ d2 σ h̄ωl − εg h̄ωnp kp
= + + −
h̄ωs
−
d dωs d dωs + d dωs − d dωs ε0 ε0 ε0
es es esz
X X Ml    
n1 +1/2 2 n +1/2 2
= σ0 − β1 µk − β2 µk2 ≥0 (416)
np kp mp n1 k1 m1 n2 k2 m2
g2 + δ2f 1 2

In the Z(el ; σ± )Z configuration, the emission spectra of electron

l = 1; 2; 3 (409)
Raman scattering in the quantum dot show maxima at the follow-
where ing values of ωs ,
!
1 1 ωs ωL 0f h    i h̄ωn k
σ0 = |el · P̂cv |2 − h̄ωs n +1/2 2 n +1/2 2 p p
= (−1)j+1 µk2 − µk1 − (417)
εab εa0 ε30 ε0 2 1 ε0
3Ve6 η(ωs )µ∗ and
× (410) h̄ωs h n0 +1/2 2  n +1/2 2 i
m20 π 3 c 4 η(ωl )ωl h̄2 = βj µk0
j
− µk
j
(418)
ε0 j j
and
As can be seen from Eqs. (417) or (418), these frequencies cor-
h̄ωl − εg h̄ωs h̄ωnp kp respond to electron transitions connecting the subbands edges for
g = − −
ε0 ε0 ε0 a process involving just the conduction or just the valence band
    (i.e., intraband transitions). The values of frequencies reported in
n1 +1/2 2 n +1/2 2
− β1 µk − β2 µk2 (411) Eqs. (417) and (418) are associated with the emission of a phonon
1 2
 2 (photon) by the electron or the hole in intersubband transitions.
X2  X βj Dj 1j (Mj1l δn0 ;n +1 + Mj2l δn0 ;n −1 ) 
j j j j The selection rules are such that n1 ± 1 + np + n2 should be even
Ml =
 0 0 0 (p2 δ1j + p2 δ2j + δ2a )(p2 + δ2 )  and −m01 + mp + m2 = 0, (m01 do depend on the polarization
i=1 jkj nj mj 1a 2a j b
case chosen) for the electronic transitions and n2 ± 1 + np + n1


 δm0 ;m +1 |es · e+ |2 l=1 should be even and −m02 + mp + m1 = 0 (m02 do depend on the

 j j polarization case chosen too) for the hole transitions, both related
2
× δm0j ;mj −1 |es · e− | l=2 (412) to Eq. (417). The singularities involved in Eqs. (417) and (418) are



δ 0 |e · Z|2 l=3 independent of ωl . Their values do depend on the relative differ-
mj ;mj s
ence of the subbands involved in the transitions. Moreover, in the
with first case they depend on the phonon modes np kp . As we can see,
µ∗ the values of the frequency ωs depend on both cases of the radius
Dj = 0j γj 3j Ij Jj βj =
m∗j r0 as r0−2 .
 OPTICAL PROPERTIES IN NANOSTRUCTURES 61

Fig. 49. Scattering efficiency (SE) is plotted in σ0 units versus h̄ωs /ε0
for Z(el ; σ+ )Z scattering polarization. The symbols that indicate resonant Fig. 50. This figure is an amplification of Figure 49 between h̄ωs /ε0 =
electron and hole transitions mediated by photon (phonon) are indicated 10:5 and 12:0, and was evaluated with the same parameters.
as follows: (h; ph) = ωh ph
, (e; ph) = ωeph , (h; s) = ωh e
s , (e; s) = ωs , and
(S; L) = ωSL . The solid curve corresponds to h̄ωl − εg = 34ε0 , np = 0,
r0 = 3 nm, and 0f = 3 meV, 0a = 0b = 1 meV. The dashed curve
corresponds to h̄ωl − εg = 26ε0 . On the left-hand side of the plot the
solid line is multiplied by 50 and the dashed one is multiplied by 105 . On
the right-hand side the dashed line is multiplied by 104 .

Other singularities of Eq. (409) occur whenever g = 0.

Such singularities are mainly related to certain values of the fre-
quency ωl of the incident photon and the values of the phonon
modes np kp . For the emission spectra, the positions of these sin-
gularities are given as
h̄ωs = h̄ωl − εg − h̄ωnp kp − ε0
h    i
n +1/2 2 n +1/2 2
× β1 µk1 − β2 (µk2 (419)
1 2

Here, the same selection rule as that of Eq. (417) must be fulfilled.
The peaks related to the latter singularities correspond to inter-
band electron-hole pair transitions and their position depends on
the incident radiation frequency ωl .
We computed the scattering efficiency (SE) (1/σ0 )(d 2 σ/
d dωs ) versus h̄ωs /ε0 , the so-called emission spectra of the elec-
tron Raman scattering process for a given polarization es of the
emitted radiation. The physical parameters entering in the formu-
Fig. 51. Same as in Figure 49 but with r0 = 3 nm (solid curves) and r0 =
las were taken for the GaAs case and κb∞ = 8:16 (for AlAs). We
2 nm (dashed curves).
set 0f = 3 meV, 0a = 0b = 1 meV.
In all the plots, we show the emission spectra for scattering con-
figuration Z(el ; e+ )Z. From Figures 49 to 51, solid lines represent
case, we have ωh 0 0 h
the results for a quantum dot with h̄ωl − εg = 34ε0 , np = 0 s (n2 ; k2 ; n2 ; k2 ) (ωph (n1 ; k1 ; n2 ; k2 ; np ; kp )):
phonon mode and radius r0 = 3 nm. In all cases, (n1 ; k1 ; n2 ; k2 ) are the subbands involved in the tran-
Due to scale problems, we have shown only a few transi- sitions and np kp are the phonon modes.
tions at all spectra. The step-like that appear in all figures will In all the plots, the peak positions mediated by photons do not
be represented by ωSL (n1 ; k1 ; n2 ; k2 ; np ; kp ). Resonant elec- depend on the incident radiation energy, the np kp phonon modes
tron transitions mediated by photon (phonon) are indicated neither depend on the radius r0 and only depend on the difference
by ωes (n1 ; k1 ; n01 ; k01 ) (ωeph (n1 ; k1 ; n2 ; k2 ; np ; kp )): In the hole between the subbands involved in the transition (see Eq. (418)).
62 RIERA ET AL.

In all cases, the resonant electron (hole) transitions medi- Table IV. The Emission Spectra for h̄ωl − εg = 34ε0 ∗
ated by the phonon do not depend on the incident radiation en-
h̄ωs
ergy h̄ωl , this also occurs for the same transitions mediated by ε0
the photon. They depend on the difference between subbands in- Transition r0 = 2 nm r0 = 3 nm
volved in the transitions, the radius r0 , and the frequency ωnp kp
of the phonon modes. When we are interested in the value of the ωh
ph
(2; 1; 1; 1; 0; 1) 1:458 1:182
frequency ωs , we use the expression (417). Hence, we found that ωh
s (2; 1; 1; 1) 1:677 1:677
the value ωs depends inversely on the radius and when ωnp kp in- ωeph (1; 1; 2; 1; 0; 1) 11:130 10:854
creases, ωs decreases. The behavior of the ωnp kp frequency as a ωes (1; 1; 2; 1) 11:349 11:349
function of r0 and np kp phonon modes can be seen from the dis- ωSL (1; 1; 2; 1; 0; 1) 11:913 11:637
persion relation curve of Figure (1) in Ref. [141] and Figure (3) in
Ref. [142].
On the other hand, for a steplike, these frequencies depend on *Five peaks are shown.
the incident radiation h̄ωl , the radius r0 , and the ωnp kp phonon
modes. ωs values depend on the dispersion relation curves (see
Fig. (1) of Ref. [141] and Fig. (3) in Ref. [142]) and expres-
sion (419).
In all cases, when kp takes different values we should under-
stand that it is associated to various transitions, each one corre-
sponding to the same peaks number, which, due to scale reasons,
they cannot be seen in the figures. Moreover, for a fixed energy
of the incident radiation h̄ωl , if the np kp phonon modes increase
when kp increases, the SE tend to be insignificant, because ωs in-
creases.
In Figure 49, the dashed line corresponds to h̄ωl − εg = 26ε0 .
The spectrum has the following transitions: ωh ph
(1; 1; 0; 1; 0; kp ),
ωh h e
s (1; 1; 0; 1), ωs (1; 2; 0; 2), ωSL (1; 1; 0; 2; 0; kp ), ωph (0; 1; 1,
1; 0; kp ), and ωeph (0; 1; 1; 1). The solid line corresponds to ωh
ph
(2; 1; 1; 1; 0; kp ), ωh e
s (2; 1; 1; 1), ωSL (2; 1; 1; 10; kp ), ωph (1; 1; 2,
1; 0; kp ), ωes (1; 1; 2; 1) and ωSL (1; 1; 2; 1; 0; kp ) transitions. In
all cases kp = 1; 2; 3; 4; 5.
On the left-hand side of the plot, the solid line is multiplied by
50 and the dashed one is multiplied by 105 . On the right-hand side,
the dashed line is multiplied by 104 .
We can see that when the incident energy increases new sub-
bands are accessible to the electrons but they do not appear due to Fig. 52. Same as in Figure 49 but the solid curves were multiplied by 103 .
scale problems. In the case of the spectrum for h̄ωl − εg = 34ε0 ,
we must have two resonant transitions more than the h̄ωl − εg =
26ε0 one, as shown in the spectrum. labeled with ωh ph
(2; 1; 1; 1; 0; kp ), ωh e
s (2; 1; 1; 1), ωph (1; 1; 2; 1,
When the incident radiation energy h̄ωl is increased the SE 0; kp ), ωes (1; 1; 2; 1), and ωSL (1; 1; 2; 1; 0; kp ). The main results
increases too, because the difference between h̄ωl and h̄ωs is are reported in Table IV.
greater. We would get the same results if we use the remainder values
Figure 50 shows the amplification of some Figure 49 zones to of kp .
corroborate the before statements. In Figure 50, we can see the In Figure 52, we show the emission spectra for h̄ωl −εg = 30ε0
spectra corresponding to h̄ωl − εg = 34ε0 , where we can appre- and r0 = 2 nm. The solid line corresponds to the np = 1 mode and
ciate the peaks associated to the transitions ωeph (1; 1; 2; 1; 0; kp ), the dashed line corresponds to the np = 0 mode. The values on
ωes (1; 1; 2; 1), and ωSL (1; 1; 2; 1; 0; kp ) with kp = 2; : : : ; 5. the solid line were multiplied by 103 . For np = 0, the spectrum
The peaks related to resonant electron transitions are the fol- shows six peaks labeled with ωh ph
(1; 1; 0; 1; 0; kp ), ωh
s (1; 1; 0; 1),
lowing: ωeph (1; 1; 2; 1; 0; kp ) with h̄ωs /ε0 = 10:858, ωeph (1; 1; 2,
ωSL (1; 1; 2; 2; 0; kp ), ωSL (1; 1; 2; 1; 0; kp ), ωeph (0; 1; 1; 1,
1; 0; 3) with h̄ωs /ε0 = 10:863, ωeph (1; 1; 2; 1; 0; 4) with h̄ωs /ε0 = e
0; kp ), and ωs (0; 1; 1; 1).
10:871, ωeph (1; 1; 2; 1; 0; 5) with h̄ωs /ε0 = 10:881, and for step- The features previously discussed are also satisfied here. An-
like: ωSL (1; 1; 2; 1; 0; 2) with h̄ωs /ε0 = 11:640, ωSL (1; 1; 2; 1, other peak on the solid line curve is not seen due to scale prob-
0; 3) with h̄ωs /ε0 = 11:646, ωSL (1; 1; 2; 1; 0; 4) with h̄ωs /ε0 = lems. Furthermore, we may note that the SE for np = 1 is less
11:654, and ωSL (1; 1; 2; 1; 0; 5) with h̄ωs /ε0 = 11:663. than the SE for np = 0.
In this case, we can observe directly the invariance in the peaks In Figure 53, the excitation spectra for a quantum dot are dis-
positions of the resonant electron transitions mediated by photons. played for two emission energies as a function of (h̄ωl − εg )/ε0 .
In Figure 51, the dashed line corresponds to h̄ωl − εg = 34ε0 , As can be noted, to obtain a greater emission energy, one must
np = 0, and r0 = 2 nm. The emission spectra show five peaks increase the excitation energy, as expected.
 OPTICAL PROPERTIES IN NANOSTRUCTURES 63

Fig. 54. Evolution of the shape of the impurity wave function in a quan-
tum well of decreasing well thickness for an on-center impurity (upper
part) and with the impurity position in a thick well (lower part).

The second important feature of Coulombic problems in

nanostructures is the variation of the impurity binding energy with
the characteristic dimension of the well. When the nanostructure
Fig. 53. Scattering efficiency (SE) as a function of (h̄ωl − εg )ε−1
0 . Exci- confinement dimension (confinement size) decreases, the impurity
tation spectra for a GaAs quantum dot. The parameters used are the same
as in Figure 49.
binding energy increases, as long as the penetration of the unper-
turbed nanostructure wave function in the barrier remains small.
This may seem surprising at first sight since we intuitively asso-
Table V. Quantum Efficiency for Phonon-assisted Intersubband
ciate extra kinetic energy with the localization of a particle in a
Transitions
finite region of space. Moreover, it is actually true that the en-
1 dσ h̄ωs ergy of the ground bound state of the impurity increases as the
System h̄ωl (eV) r0 (Å) (cm−1 Sr−1 ) − 1 confinement size decreases, when it is measured from some fixed
V d E0
reference. However, the onset of the impurity continuum, which
SQDph 3.328 30 1:51 × 10−9 –(8.2, 10.5) (for donors) coincides with the energy of the nanostructure bound
QWWph 3.328 40 3:41 × 10−9 –(8.2, 10.5) state, also moves up with decreasing confinement size. The bind-
QWWph 3.328 30 6:745 × 10−7 –(9.5, 10.5) ing energy finally increases since, in the bound impurity states, the
FSWph 3.328 30 3:129 × 10−6 –(9.5, 10.5) carrier is kept near to the attractive center by nanostructure walls
and thus experiences a larger potential energy than in the absence
of a nanostructure. On the other hand, the onset of the impurity
continuum does not benefit from any extra potential energy gain,
as the continuum states are hardly affected by the Coulombic po-
Table V shows the values of the quantum efficiency when the
tential.
electron-phonon interaction is present in the intersubband transi-
When the Schrödinger equation for a given quantum system
tions.
is nonseparable in nature, then some approximation techniques
Comparing these values with those shown in Table III for elec-
must be used to study its solution. In a similar situation for a con-
tron Raman scattering without phonon participation, a decrease
fined quantum system, the variational method might constitute an
on the quantum efficiency value is observed, which indicates that
economical and physical appealing approach [143]. It was shown
the electron-phonon interaction is a smaller interaction than the
[144] that the variational method is useful for studying this class
electron-photon interaction.
of systems when their symmetries are compatible with that of the
confining boundaries. In general, in the problems related with im-
purity and atoms confined in the different nanostructures, these
4. PHYSICAL EFFECTS OF IMPURITY STATES AND considerations are fulfilled.
ATOMIC SYSTEMS CONFINED IN
SEMICONDUCTOR NANOSTRUCTURES
4.1. Hydrogenic Impurity in Quasi-two-dimensional
In contrast with bulk materials, the nanostructures are character- Systems
ized by lack of translational invariance along the confinement di-
In Figure 54, we sketched the shape of the impurity wave function,
rection. Thus, the impurity binding energy explicitly depends on
keeping the in-plane
q distance between the carrier and the attrac-
the precise location of the impurity. It is important to know where
the impurity atom is placed in the nanostructure, and the bind- tive center ρ(ρ = x2 + y 2 ) equal to zero. Several quantum-well
ing energy of the donor state associated with impurity atom de- thicknesses (upper part of Fig. 54) and several impurity positions
pends on whether the impurity atom sits at the center or axis of in a thick well (lower part of Fig. 54) were considered. For thick
the nanostructure. wells (L  a∗B ), the on-center impurity wave function resembles
64 RIERA ET AL.

that of the bulk 1s states. On the other hand, the on-edge impu- where Veff is the effective in-plane Coulombic potential,
rity wave function approaches the shape of a 2pz wave function if Z
(L/a∗B )  1. This is because the barrier potential forces the impu- −e2 +∞ 1
Veff (ρ) = dz χ2ν0 (z) q (425)
rity wave function to almost vanish at z = ±L/2. If L/a∗B  1 and κ −∞
ρ2 + (z − zi )2
if the barrier height is very large, the electron z motion becomes
forced by the quantum well potential. Thus, along the z-axis the which is ν0 - and zi -dependent. A solution of Eq. (424) can be
impurity wave function looks like the ground state wave function sought variationally, the simplest choice for the ground state be-
of the well χ1 (z). ing the nodeless one-parameter trial wave function,
r
1 2
ϕ0 (ρ) = exp(−ρ/λ) (426)
4.1.1. Approximate Solutions of the Hydrogenic Impurity λ π
Problem where λ is the variational parameter. The bound state energy is
In all that follows, we consider donorlike impurities unless other- obtained through the minimization of the function,
wise specified. The conduction bands of both host materials of the h̄2 2e2
quantum well are assumed to be isotopic and parabolic in k. We εν0 (zi ; λ) = ∗
− (427)
2me λ 2 κλ
neglect the effective mass jumps at the interfaces as well as the
Z ∞ Z +∞
differences in the relative dielectric constants of the two host ma- dz χ2ν0 (z)
× xe−x r dx + εν0
terials. The impurity envelope functions are the solutions of the 0 −∞ 4
effective Hamiltonian, x2 + (z − zi )2
λ2
!
p̂2 p̂2x + p̂2y L 2 and the binding energy is
b = H
H b0 + VI = z
+ 2
+ V0 Y z −
2m∗e 2m∗e 4 εbν0 (zi ) = εν0 − Min εν0 (zi ; λ) (428)
λ
e2
− q (420) One should be aware that the decoupling procedure (Eqs.
κ ρ2 + (z − zi )2 (423)–(428)) runs into difficulties in the limits L → ∞ and L → 0.
In the former case, the energy difference εν0 +1 −εν0 becomes very
where V0 is the barrier height and Y(x) is the step function
small and many subbands become admixed by the Coulombic po-
(Y(x) = 1 if x > 0; Y(x) = 0 if x < 0). The impurity position tential. Clearly, for infinite L one cannot describe the bulk 1s hy-
along the growth axis is zi . The (x; y) origin is at the impurity site drogenic bound state by a separable wave function. In the latter
because all the impurity positions are equivalent in the layer plane, case (L → 0), one finds similar problems due to the energy prox-
ρ is the projection of the electron position vector in the layer plane imity between the ground quantum-well subband ε1 and the top
(ρ(x; y)). of the well V0 . Consequently, χ1 (z) leaks more and more heav-
In the absence of the impurity, the eigenstates of H b0 are sepa-
ily in the barrier. At L = 0, the only sensible result is to find a
rable in (x; y) and z, 1s bulk hydrogenic wave function corresponding to the barrier-
! acting material (in this model the latter has a binding energy equal
h̄2 k2⊥
b
H0 |ν; k⊥ i = εν + |ν; k⊥ i (421) to R∗y ). Once again this state cannot have a wave function like that
2m∗e of Eq. (423). If V0 is infinite, the result at L = 0 is qualitatively
different. The quantum well only has bound levels whose energy
where ν labels the quantum-well eigenstates (energy εν ), i.e., the
separation increases like L−2 when L decreases. The smaller the
quantum-well bound (εν < V0 ) and unbound states (εν > V0 ) and
well thickness, the better the separable wave function becomes. At
k⊥ = (kx ; ky ). Since the |ν; k⊥ i basis is complete we may always
L = 0, one obtains a true two-dimensional hydrogenic problem
expand the impurity wave function ψloc in the form:
whose binding energy is 4R∗y whereas λ = a∗B /2.
X To circumvent the previous difficulties and to obtain the exact
|ψloc i = c(ν; k⊥ )|ν; k⊥ i (422)
limits at L = 0 and L = ∞ for any V0 , one may use [147–149],
ν;k⊥
" q #
The Coulombic potential couples a given subband ν, as well as a 1
ψloc (r) = Nχ1 (z) exp − ρ + (z − zi )
2 2 (429)
given vector k⊥ with all others. The intersubband coupling (espe- λ
cially the one with the subbands of the quantum-well continuum) is
where N is a normalization constant, where λ is the varia-
difficult to handle. In a quasi-two-dimensional situation, we would
tional parameter, and where attention is focussed on the ground
like to set c(ν; k⊥ ) ≡ cν0 (k⊥ )δν;ν0 , i.e., to neglect intersubband
bound state attached to the ground quantum-well subband ε1 .
coupling [145, 146]. This procedure is convenient as the impurity
Calculations are less simple than with the separable wave func-
wave function displays a separable form:
tion (Eq. (423)). Comparing the binding energy deduced from
hr|ψloc i = χν0 (z)ϕ(ρ) (423) Eqs. (423)–(429) one finds, for infinite V0 , that the separable wave
function gives almost the same results as the nonseparable wave
The wave function ϕ(ρ) is the solution of the two-dimensional function if L/a∗B ≤ 3. This is the range where for most materials
Schrödinger equation, the quantum size effects are important.
" 2 # Other variational calculations were proposed [150, 151]. For ex-
p̂x + p̂2y ample, instead of using a nonlinear variational parameter one uses
+ Veff (ρ) ϕ(ρ) = (ε − εν0 )ϕ(ρ) (424)
2m∗e a finite basis set of fixed wave functions (often Gaussian ones) in
 OPTICAL PROPERTIES IN NANOSTRUCTURES 65

which Hb is numerically diagonalized. The numerical results ob- (L/a∗B ≤ 0:2). The on-center donor binding energy
tained by using a single nonlinear variational parameter compare increases from R∗y (L → ∞) to reach a maximum
favorably with these very accurate treatments. (L/a∗B ≤ 1) whose exact L-location and amplitude
depend on Vb . Finally, it decreases to the value R∗y at
4.1.2. Results for the Ground Impurity State Attached to the L = 0 [150, 151]. If V0 is infinite, the maximum is only
Ground Subband reached at L = 0 and has a value of 4R∗y [147].
(ii) Position dependence of the impurity binding energy: the
Figures 55–57 give a sample of some calculated results for the impurity binding energy monotonically decreases when
ground impurity state attached to the ground subband. Two pa- the impurity location zi moves from the center to the
rameters control the binding energy: edge of the well and finally moves deep into the barrier.
(i) Thickness dependence of the impurity binding energy: In Figure 58, it is shown that this decrease is rather slow
the dimensionless ratio L/a∗B indicates the amount of for zi > L/2; for instance, a donor placed 150 Å away
two-dimensionality of the impurity state. If L/a∗B ≥ 3 or from a GaAs–Ga0:7As Al0:3 As quantum well
L/a∗B ≤ 0:2 and V0 ≈ 3 eV in GaAs–Ga1−x Alx As the (L = 94:8 Å) still binds a state by ≈ 0:5R∗y which is
problem is almost three-dimensional. This is either ≈ 2:5 meV [148].
because the subbands are too close (L/a∗B ≥ 3) or
because the quantum-well continuum is too close
4.1.3. Excited Subbands, Continuum
The procedure followed for the bound state attached to the ε1 sub-
band can be generalized for excited subbands ε2 ; ε3 ; : : :, as well as
for the quantum-well continuum.
However, it becomes rapidly cumbersome since a correct vari-
ational procedure for excited states requires the trial wave func-
tions to be orthogonal to all the states of lower energies. For sep-
arable wave functions (Eq. (423)), this requirement is automat-
ically fulfilled and one may safely minimize εν0 (zi ; λ) to obtain
a lower bound of εbν0 (zi ). The new feature associated with the
bound states attached to excited subbands is their finite lifetime.
This is due to their degeneracy with the two-dimensional continua
of the lower lying subbands. This effect however is not very large
[145, 146] since it arises from the intersubband coupling which
is induced by the Coulombic potential: if the decoupling proce-
dure is valid the lifetime of the quasi-bound state calculated with
Eq. (423) should be long.

Fig. 55. Calculated dependence of the on-edge (crosses) and on-center

(circles) hydrogenic donor binding energy versus the well thickness L in a
quantum well with an infinite barrier height.

Fig. 56. Dependence of on-center hydrogenic donor binding energy in Fig. 57. Dependence of the on-edge hydrogenic donor binding energy
GaAs–Ga1−x Alx As quantum wells versus the GaAs slab thickness L. in GaAs–Ga1−x Alx As quantum wells versus the GaAs slab thickness L.
Vb (x) = 0:85 [εg (Ga1−x Alx As) − εg (GaAs)]. One monolayer is 2.83- m∗e = 0:067m0 , κ = 13:1, V0 = 212 meV (curve 1); 318 meV (curve 2);
Åthick. 424 meV (curve 3) and infinite (curve 4).
66 RIERA ET AL.

Fig. 58. Dependence of the hydrogenic donor binding energy in a quan-

tum well versus the impurity position zi (a) in the case of a finite barrier
well (V0 = 318 meV) and (b) in the case of an infinite barrier well [147].
There is an interface at −zi = L/2, L = a∗B = 94:8 Å. (Reprinted from
G. Bostard, Phys. Rev. B, 24, 4714, (1981).)

For impurities located in the barriers, (an important practical Fig. 59. Dependence of the on-center carbon binding energy ver-
topic with regard to the modulation-doping technique), we have sus well width in GaAs–Ga1−x Alx As quantum wells. Vb (x) = 0:15
seen that they weakly bind a state below the ε1 edge. There ex- [εg (Ga1−x Alx As) − εg (GaAs] is the assumed hole confining barrier
ists (at least) a second quasi-discrete level attached to the barrier height. The open circles are the experimental values obtained by Miller
edge, i.e., with an energy ≈ V0 − R∗y . This state is reminiscent of et al. [154]. (Reprinted from R. C. Miller et al., Phys. Rev. B, 25, 387,
(1982).)
the hydrogenic ground state of the bulk barrier. Due to the pres-
ence of the quantum-well slab, it becomes a resonant state since it
interferes with the two-dimensional continua of the quantum-well
subbands ν(εν < V0 ). Its lifetime τ can be calculated using the ex- on-center donor ground state (quasi-1s) and the excited states
pression, (quasi-2px ; 2py ) agrees with the far-infrared absorption and the
*   +2 magneto-absorption data [152]. On-edge donor levels were also
h̄ 2π X  
 −e2 


 investigated.
=  ϕ̃(r; zi 
) q  ⊥  δ
νk
2τ h̄    
ν;k⊥ κ ρ2 + (z − zi )2
! 4.1.5. Acceptor Levels in a Quantum Well
h̄2 k2⊥
× V0 − R∗y − εν − (430)
2m∗e The problem of acceptor levels in semiconductor quantum wells
is much more intricate than the equivalent donor problem. This is
where ϕ̃ is the 1s bulk hydrogenic wave function of the barrier. due to the degenerate nature of the valence bands in cubic semi-
If the distance d separating the impurity from the quantum-well conductors. In quantum wells, this degeneracy is lifted (light and
edge is much larger than a∗B , Eq. (430) simplifies into heavy holes have different confinement energies).
!3/2 s ! However, the energy separation between the heavy- and light-
h̄ ∗
X R∗y 2d V0 − εν hole subbands is seldom comparable to the bulk acceptor binding
≈ 16Ry Pν exp − ∗ (431)
2τ ν
V0 − εν aB R∗y energies. Thus, many subbands are coupled by the combined ac-
tions of the Coulombic potential and the quantum-well confining
where Pν is the total integrated probability of finding the carrier in
barrier potential. No simple decoupling procedures appear man-
the νth state (energy εν ) in any of the two barriers of the quantum
ageable.
well. One sees from Eq. (431) that the lifetime τ is strongly dom-
Masselink et al. [153] used variational calculations to estimate
inated by the escape processes to the excited subband εν whose
the binding energy of the acceptor level due to carbon (a well-
energy is nearest to V0 − R∗y . As an example, let us take a GaAs–
known residual impurity in MBE grown GaAs layers). Their re-
GaAlAs quantum well with thickness L = 50 Å, V0 = 0:2 eV sults, shown in Figure 59, agree remarkably with the experiments
and let us assume that d = 3a∗B (i.e., d ≈ 300 Å). We then get of Miller et al. [154]. One notices in Figure 59 the same trends
τ ≈ 3 × 10−6 s. The quasi-bound state can thus be considered, to versus quantum-well thickness as displayed by Coulombic donors
a reasonable approximation, as stationary (h̄/τ ≈ 4 × 10−8 R∗y ). (Figs. 56 and 57). The binding energy first increases when L de-
creases (increasing tendency to two-dimensional behavior), then
saturates and finally drops to the value of the acceptor binding en-
4.1.4. Excited Impurity Levels Attached to the Subband ergy of the bulk barrier. Finally, it should be noted that the relative
The Schrödinger equation (Eq. (420)) has several bound states increase in binding is smaller for acceptors than for donors. This
below ε1 . Their binding energies were calculated by several is because the bulk acceptor Bohr radius is much smaller than that
groups [150, 151]. The calculated energy difference between the of the bulk donor.
 OPTICAL PROPERTIES IN NANOSTRUCTURES 67

4.2. Hydrogenic Impurities within Asymmetric and must satisfy

Symmetric Quantum-well Wires
ψi |ρ=ρ0 = ψo |ρ=ρ0 (437)
In this subsection, the energy of the ground and first excited states,  
the binding energy and oscillator strengths for hydrogenic impuri- ∂ψi  ∂ψo 
 =  (438)
ties confined within a cylindrical quantum-well wire with a finite- ∂ρ  ∂ρ 
ρ=ρ0 ρ=ρ0
height-potential well are studied variationally as a function of the
wire radius and of the relative position of the nucleus within the By imposing condition (438) on the function given by Eq. (436),
quantum-well wire for different barrier-height potential [155]. The we have that
α
trial wave function is constructed as the product of the free wave β= (439)
function of hydrogen impurity and a simple auxiliary function that (1 − α)ρ0
allows the appropriate boundary conditions to be satisfied. that is, we need to find only one variational parameter to minimize
In this context, the variational method used here constitutes the ground state energy. Furthermore, when V0 → ∞, α → 1,
a useful approach to study asymmetric and symmetric quantum β → ∞, and ψ0 → 0, as expected.
systems confined by penetrable potentials. The continuity condition on the boundary, Eq. (437), relates B
and A: !
4.2.1. On-axis Hydrogenic Impurity α
B = Aρ0 (1 − α) exp (440)
1−α
If the nucleus of the hydrogen impurity is located on the symmetry
axis of the cylindrical quantum-well wire, the model Hamiltonian The suitable Hamiltonian for 2p type excited states is the same
for the electron within the quantum-well wire can be written in that given by Eq. (432), and the variational wave functions for 2px ,
atomic units (h̄ = m∗e = e = 1) as 2py , and 2pz states, when the nucleus of hydrogenic impurity is
fixed on the symmetry axis of the wire, are of the form,
b = − 1 ∇ 2 − 1 + Vb (ρ)
H (432)
2 r ψ2px
with  
0 ≤ ρ ≤ ρ0 ψi = C(ρ0 − γρ) exp(−γr)ρ cos ϕ 0 ≤ ρ ≤ ρ0
Vb (ρ) =
0 = (441)
V0 ρ0 ≤ ρ < ∞
(433) ψo = D exp(−γr) exp(−ξρ)ρ cos ϕ ρ0 ≤ ρ < ∞
q ψ2py
where r = ρ2 + z2 is the electron-nucleus distance, ρ is the cylin- 
ψi = C(ρ0 − γρ) exp(−γr)ρ sin ϕ 0 ≤ ρ ≤ ρ0
drical coordinate parallel to the xy-plane, z is the coordinate along = (442)
ψo = D exp(−γr) exp(−ξρ)ρ sin ϕ ρ0 ≤ ρ < ∞
the wire axis, ρ0 is the wire radius, and V0 is the confining potential
ψ2pz
barrier. 
The physical meaning of V0 in this context is to simulate, on the ψi = C(ρ0 − γρ) exp(−γr)z 0 ≤ ρ ≤ ρ0
= (443)
average, the effective potential step created by the composition ψo = D exp(−γr) exp(−ξρ)z ρ0 ≤ ρ < ∞
difference between the quantum-well wire and its surroundings.
where C and D are normalization constants and γ, ξ are varia-
It is well known that the Schrödinger equation for this
tional parameters to be determined and they are related by
Coulomb-type potential is not separable in cylindrical coordinates,
the natural symmetry of the wire, thus we are enforced to use an γ
ξ= (444)
approximate method to calculate the ground state energy and the (1 − γ)ρ0
first excited states for this system.
and !
To use the variational method to approximately solve the
Schrödinger equation with the Hamiltonian given by Eq. (432), we γ
D = Cρ0 (1 − γ) exp (445)
must construct a trial-wave-function ψ with the basic properties 1−γ
listed 
ψ(0) finite The results for the energy of the ground and 2pz states as a


ψ(r) → 0 as r → ∞ function of wire radius and different barrier heights are displayed
(434)


1 ∂ψ(r)
continua at ρ = ρ0
in Figure 60. For a given value of the finite barrier potential, the
ψ(r) ∂ρ ground state (excited state) energy increases from −0:5 (−0:125)
Hartrees as the wire radius is reduced. These values are character-
The function ψ contains a parameter or a set of parameters that
istic of the free hydrogen atom.
allows us to minimize the ground state energy by imposing that
The binding energy εb for the hydrogenic impurity is defined
∂ε0 as the ground state energy of the system without Coulomb interac-
=0 (435)
∂αi tion εw , minus the ground state energy in the presence of electron-
where {αi } is the set of parameters previously mentioned. nucleus interaction ε0 , i.e.,
A possible ansatz wave function for this problem is of the form, εb = εw − ε0 (446)

ψi = A(ρ0 − αρ) exp(−αr) 0 ≤ ρ ≤ ρ0
ψ1s = (436) The binding energy defined in this way is a positive quantity.
ψo = B exp(−αr) exp(−βρ) ρ0 ≤ ρ < ∞
In Figure 61(a), we display the variation of the hydrogenic im-
where A, B are normalization constants and α, β are the vari- purity binding energy εb as a function of wire radius ρ0 for several
ational parameters involved in the calculation. These functions values of the finite potential barrier.
68 RIERA ET AL.

Fig. 60. Energy of the ground and 2pz states of the hydrogenic impurity Fig. 62. Energy of the ground and the 2pz states of an off-axis hydrogenic
confined in a penetrable quantum-well wire with the nucleus on the axis as impurity enclosed within a penetrable quantum-well wire of radius ρ0 =
a function of wire radius and potential barrier heights. 1:0 Bohr as a function of the relative nucleus position and the potential
barrier height.

4.2.2. Off-axis Hydrogenic Impurity

If the nucleus of the atom is located on the x-axis, at a distance b
from the axis of the wire, the electron-nucleus distance is
q
r 0 = |r − bex | = (x − b)2 + y 2 + z2
q
= ρ2 + b2 + z2 − 2ρb cos ϕ (447)
with r the position vector of the electron relative to the origin on
the wire axis, then the suitable Hamiltonian is

b = − 1 ∇ 2 − 1 + Vb (ρ)
H (448)
2 r0
where Vb (ρ) is the same given by Eq. (433).
The trial wave function for the ground state energy can then be
written as

ψi = A(ρ0 − αρ) exp(−αr 0 ) 0 ≤ ρ ≤ ρ0
ψ1s = (449)
ψ0 = B exp(−αr 0 ) exp(−βρ) ρ0 ≤ ρ < ∞
where, as before, A and B are normalization constants and α, β
are the variational parameters involved in the calculation. Again,
α; β and A, B are, respectively, related by Eqs. (439) and (440).
When the nucleus of hydrogenic impurity is off the symmetry
axis, it is not possible to find the wave functions for the 2px and the
2py states that satisfy the orthonormality condition as is required
by the method, therefore we restrict to the 2pz state only. The
suitable Hamiltonian is the same as that given by Eq. (434), and
the variational wave function is given by

ψi = C(ρ0 − γρ) exp(−γr 0 )z 0 ≤ ρ ≤ ρ0
ψ2pz = (450)
ψ0 = D exp(−γr 0 ) exp(−ξρ)z ρ0 ≤ ρ < ∞
where, as before, C, D are normalization constants and γ, ξ are
the variational parameters involved in the calculation. Again, γ, ξ
and C, D are, respectively, related by Eqs. (444) and (445).
Fig. 61. (a) Binding energy of the hydrogenic impurity confined in a pen- In Figure 62, we show the energy of the ground and the 2pz
etrable quantum-well wire with the nucleus on the axis as a function of the states as a function of (b/ρ0 ) for ρ0 = 1:0 Bohr and several heights
wire radius and potential barrier height. (b) Binding energy of the hydro- of potential barrier.
genic impurity confined in a penetrable GaAs–Ga1−x Alx As quantum-well In Figure 63(a), we display the variation of the hydrogenic im-
wire with the nucleus on the axis as a function of the wire radius and po- purity binding energy εb as a function of (b/ρ0 ) for ρ0 = 1:0 Bohr
tential barrier height. and several heights of potential barrier.
 OPTICAL PROPERTIES IN NANOSTRUCTURES 69

Fig. 64. 1s − 2pz oscillator strength of the hydrogenic impurity confined

in a penetrable quantum-well wire with the nucleus on the axis as a function
of the wire radius and potential height.

Fig. 63. (a) Binding energy of an off-axis hydrogenic impurity enclosed

within a penetrable quantum-well wire of radius ρ0 = 1:0 Bohr as a func- Fig. 65. 1s − 2pz oscillator strength of an off-axis hydrogenic impurity
tion of the relative nucleus position and potential barrier height. (b) Bind- enclosed within a penetrable quantum-well wire of radii ρ0 = 2:0, ρ0 =
ing energy of an off-axis hydrogenic impurity enclosed within a penetrable 3:0, and ρ0 = 4:0 Bohr as a function of the relative nucleus position and
GaAs–Ga1−x Alx As quantum-well wire of radius ρ0 = 1:0 Bohr as a func- potential height.
tion of the relative nucleus position and potential barrier height.

2:0, 3:0, and 4:0 Bohrs and several potential barrier heights for the
4.2.3. Optical Properties off-axis case.
To predict the absorption peak due to 1s −2p transitions as a func- In Figure 60, we show the ground and the first excited states
tion of the wire radius and the confining degree, we calculated the energies of the hydrogenic impurity confined within the cylindrical
transition energy between these states as well as their oscillator quantum dot with the nucleus on the symmetry axis, as a function
strengths. of ρ0 for several V0 values. The ground state energy has a simi-
The f0 oscillator strength is defined as lar behavior as in Ref. [143], as well as for the hydrogen atom and
D E2 for electron systems (like H− , He, Li+ , and Be2+ ) within a pen-
  etrable spherical box [156]; that is, the energy diminishes as the
f0 = 23  1s|r · er |2p  1ε (451)
confinement box size increases (for a given value of V0 ), and for
where er is the incident light polarization vector and 1ε = ε2p − a given size of the box the energy increases as V0 increases. Simi-
ε1s is the transition energy between the 2p and 1s states. lar behavior is found by Nag and Gangopadhayay [157] who show
The 1s − 2pz oscillator strength f0 (for er //z) is shown in Fig- graphically qualitative results for heavy holes and electrons within
ure 64 as a function of wire radius for several heights of potential a quantum-well wire with cylindrical and elliptic cross-sections.
barrier, when the nucleus is fixed on the center of the wire. For They used the physical constants of the Ga0:47 In0:53 As/InP sys-
a given value of the finite barrier potential, the oscillator strength tem.
increases from 0.139, the characteristic value of the free hydrogen This energy behavior can be understood on the basis of an
atom, as the wire radius is reduced. uncertainty principle because when confinement dimensions are
Finally, in Figure 65, we show the oscillator strength f0 (for reduced the energy increases. If dimensions continue decreasing
er //z) as a function of relative nucleus position (b/ρ0 ) for ρ0 = there will be a point in which the kinetic energy will be greater than
70 RIERA ET AL.

the potential energy associated to the internal interactions of the In Figure 63(a), we can note that the value of the binding en-
system. In an extreme situation of confinement, the potential en- ergy for the asymmetric case decreases as the nucleus approaches
ergy is only a perturbation of total energy for a free particle system. the boundary because the wave functions diminish at the bound-
For attractive potentials, like Coulomb’s potential, there is a very aries and thus their contributions to the energy when the nucleus
clear competition between kinetic energy and potential energy. If is at the boundary are smaller than when the nucleus is at the cen-
ρ0 decreases, then potential energy decreases, but kinetic energy ter of the wire; hence the probability that the electron has to leave
always increases because of the localization of the wave function. the quantum-well wire is increased. A similar behavior occurs for
In Figure 61(a), we display the binding energy of a hydrogenic the binding and the ground state energy of a hydrogenic impurity
impurity in a quantum-well wire with a finite potential barrier, as a placed in a rectangular cross-section quantum-well wire of GaAs–
function of the width of the wire. The results show that, due to the Gax Al1−x As [158].
finite confining capacity of the quantum-well wire, there is a criti- In Figure 63(b), we show the same results as in Figure 63(a),
cal radius for which the electron is no longer confined. Indeed, as for the case of a hydrogenic impurity within a cylindrical quantum-
the confining barrier increases this critical radius becomes smaller, well wire of GaAs–Gax Al1−x As with ρ0 = 103:4 Å. The calcula-
as expected from simple physical considerations. For a given value tions were carried out for barrier potentials V0 = 265:0, 53.0, 10.6,
of the finite barrier potential, the binding energy increases from its 5.3 meV that correspond to aluminum concentrations x = 0:36,
bulk value as the wire radius is reduced, reaches a maximum value, 0.08, 0.016, 0.008. The quantitative comparison of Figure 62(b)
and then drops to the bulk value characteristic of the barrier mate- with the results of Ref. [158] shows the effect of the geometry on
rial as the wire radius goes to zero. This is due to the fact that as the the binding energy.
wire radius is decreased the electron wave function is compressed The binding energy for ρ0 = 1 Bohr is in agreement with the
thus leading to an enhancement of the binding energy. However, binding energy when the nucleus is on the axis of the wire (com-
below a certain value of ρ0 the leakage of the wave function into pare Fig. 63(a) with Fig. 61(a)).
the barrier region becomes more important and, the binding en- In Figure 64, the 1s − 2pz transitions of the studied states in
ergy begins to decrease until it reaches a value that is characteristic the symmetric case are shown. We note a similar behavior to that
of the barrier material as ρ0 → 0. This effect was studied for heavy found in other systems, which is confined in regions with different
and light excitons confined within quantum wires [64], for hydro- symmetries, e.g., the heavy exciton case confined in a KCl ionic
gen impurities within the quantum-well wire of GaAs1−x Alx As sphere for several radii and penetrability, in which the excitonic
[158, 159], and excitons within a quantum-well wire in the pres- transitions vanish for given sphere sizes, as a consequence of the
ence of a magnetic field [160]. In Figure 61(b), we show the same finite confining potential value and moreover, in case the absorp-
results for the quantum well wire of GaAs1−x Alx As for the bar- tion peak is shifted to high energies as the sphere radius decreases.
riers potential V0 = 265:0, 53:0, 10:6, 5.3 meV, that correspond This effect was experimentally observed in SiO2 spheres [161]. In
to the x = 0:36, 0.08, 0.016, 0.008 Al concentrations, respectively. addition, for V0 = ∞, there is a critical radius of a quantum-well
The quantitative comparison of curves in Figure 61(b) with the re- wire for which the absorption has a maximum and then returns to
sults of Refs. [64, 159, 160] show that our variational calculations bulk value as the radius continues decreasing.
lead to the same results. In Figure 65, we show the oscillator strength f0 (for el //z) as a
The peak in the binding energy occurs for the smallest value function of relative nucleus position (b/ρ0 ) for ρ0 = 2:0, 3.0, and
of ρ0 for which the probability of the electron found outside the 4.0 Bohrs, with the following potential barrier values: V0 = 0:1 and
well is not significant; that is, the enhancement in binding energy 0:2 Hartrees for the off-axis case. The maximum in the absorp-
occurs because the confining potential is forcing the electron to tion peak allows us to predict the optimum site for the location
move only in a smaller space and to spend most of its time closer of the hydrogenic impurity within the quantum-well wire. We can
to the nucleus. This strong enhancement of the binding energy note that, for the same parameter values as those previously men-
has important consequences for optical and transport properties tioned the transition intensities are in agreement with the transi-
of quantum-well wires. tion intensities of the studied states in the symmetric case (com-
In Figure 62, we show the ground and the first excited state pare Fig. 65 with Fig. 64).
energies for the asymmetric case. The ground state energy ap-
proaches the value calculated in [143] for the infinite potential
4.3. Asymmetric Confinement of Hydrogen by Hard
barrier case, as the potential barrier is increased, and when the
Spherical and Cylindrical Surfaces
size of the quantum-well wire becomes infinite and the nuclei is
close to the boundary. For a given value of the finite barrier po- The variational method is used to calculate the ground state en-
tential, the energy of the 2pz excited state is almost independent ergy of the hydrogen atom confined within hard spherical and
of the relative position of the nuclei within the quantum-well wire, cylindrical surfaces, for an atomic nucleus, which is off the cen-
as compared with the variation of the ground state energy, this is ter of symmetry of the confining boundary. It is shown that the
due to the orientation of the orbital in which the electron moves; wave function for the free hydrogen atom in spherical coordinates
a similar effect is studied for a hydrogen atom enclosed between (referred to the center of symmetry of the confining surface) can
two impenetrable parallel planes and for a heavy exciton in a CdS be used, without further assumptions, to construct the trial wave
film in [161]. function systematically. The latter is assumed to be the product of
Also, we can note that for ρ0 = 1 Bohr, the ground state ener- the free wave function and a simple cutoff function that satisfies
gies calculated for V0 = 1, 2, 10, 50 Hartrees are in exact agree- the appropriate boundary conditions.
ment with the ground state energies when the atom nucleus is on To show the advantage of the method, we present in detail
the symmetry axis as we can see in Figure 60. The same is true for two cases: the hydrogen atom confined within hard spherical and
the first excited state energies (compare Fig. 62 with Fig. 60). within cylindrical surfaces. The asymmetry of these systems is due
 OPTICAL PROPERTIES IN NANOSTRUCTURES 71

Table VI. Ground State Energy for the Off-center Hydrogen Atom
within an Impenetrable Spherical Box as a Function of the Position of the
Nucleus (Relative to the Center of the Confining Sphere) for Various
Radii of the Box∗

r0 a/r0 ε1 ε2 ε3

0.1 −0:4804 −0:483 18 −0:483 15

4.0 0.5 −0:4673 −0:481 05 −0:481 02
1.0 −0:4390 −0:473 35 −0:473 42
0.1 −0:4209 −0:423 58 −0:423 58
3.0 0.5 −0:3910 −0:413 73 −0:413 92
1.0 −0:3223 −0:377 19 −0:378 40
0.1 −0:1198 −0:122 86 −0:122 86
Fig. 66. Coordinates for the off-center hydrogen atom relative to the cen-
2.0 0.5 −0:0235 −0:066 07 −0:068 89
ter of the confining sphere.
1.0 +0:2022 +0:158 52 +0:127 51

to the nucleus of the atom being shifted off the center of symme- *(1) Results of this subsection, (2) results of Gorecki and Byers Brown
try of the confining surface (i.e., either off the center of the sphere [162] obtained by boundary perturbation theory, (3) results of Brownstein
or off the axis of the cylinder). Of course, in such cases the corre- [163] obtained by a variational method in which the trial wave function
sponding Schrödinger equation is no longer separable. does not satisfy the boundary conditions. Energy units: Hartrees, distance
units: Bohrs.

4.3.1. Applications of the Method

In this subsection, two explicit examples of using the variational
method application are described. Both involve the hydrogen atom
confined within a domain with impenetrable boundaries. In the
first example, we consider the atom within a spherical surface
where the center of the surface and the nucleus of the atom differ
by a constant distance a. Referring to Figure 66, the coordinates
of the electron of the hydrogen atom with respect to the nucleus
(r0 ) and with respect to the center of the confining sphere (r) are
related by
r0 = r − a (452)
The corresponding Hamiltonian can be now written as

b = − 1 ∇ 2 − 1 + Vb (r)
H (453)
2 |r − a|
where Vb is a confining potential defined as

+∞ r > r0
Vb (ρ) = (454)
0 r < r0
The ground state wave function for the free hydrogen atom in
spherical coordinates is given as
ψ0 = A exp(−r 0 ) (455) Fig. 67. Ground state energy for an off-center hydrogen atom enclosed
within an impenetrable spherical box as a function of the position of the
where r 0 is the electron-nucleus distance and A is a normalization nucleus (relative to center of the sphere), for various radii of the confining
constant. box. Note that when a/r0  1, and r0  1, the energy approaches −1/8
Without loss of generality, we can assume that the nucleus is Hartree which corresponds to the ground state of the hydrogen atom close
located on the z-axis. The trial wave function for the ground state to an infinite planar surface.
can be written as
χ(r) = A(r − r0 ) exp(−α|r − a|) (456)
A comparison with the results obtained in Refs. [162] and [163] is
where r0 is the radius of the spherical confining domain, A is a also shown. Figure 67 shows the ground state energy for different
normalization constant, and sizes of the confining sphere as a function of (a/r0 ).
In the case of cylindrical coordinates, it is obvious that the
|r − a| = [r 2 + a2 − 2ar cos θ]1/2 (457)
Schrödinger equation for the Coulomb potential is nonseparable,
Here, θ is the usual polar angle of the spherical coordinates. however, the application of the variational method is still possible
The results of the variational calculations of the ground state in the same context as it could be done for spherical coordinates,
energy with this trial wave function are displayed in Table VI. as we shall see.
72 RIERA ET AL.

Table VII. Ground State Energy for the Off-axis Hydrogen Atom
within an Impenetrable Cylindrical Box as a Function of the Position of
the Nucleus (Relative to the Axis of the Confining Cylinder), for Various
Radii of the Box∗

ρ0 b/ρ0 ε1 ε2

0.0 1.5854
0.2 1.6544 1.385 68
1.0 0.4 1.8111 1.553 68
0.6 2.0243 1.794 95
0.8 2.2305 2.036 988
0.99 2.3724 2.203 82
0.0 −0:1791
Fig. 68. Coordinates for the off-axis hydrogen atom relative to the axis of 0.2 −0:1363 −0:247 82
the confining cylinder. 2.0 0.4 −0:0401 −0:159 586
0.6 0.1050 0.0059 04
0.8 0.2704 0.2085 54
0.99 0.3785 0.3369 83
0.0 −0:4677
0.2 −0:4477 −0:487 374
4.0 0.4 −0:4044 −0:467 144
0.6 −0:3291 −0:398 212
0.8 −0:1671 −0:199 550
0.99 −0:0388 −0:038 415
0.0 −0:4912
0.2 −0:4921 −0:499 997
10.0 0.4 −0:4859 −0:499 943
0.6 −0:4696 −0:498 329
0.8 −0:3988 −0:456 063
0.99 −0:1183 −0:104 528

*(1) Results of this subsection, (2) results of Tsonchev and Goodfriend

[164] obtained by expanding the wave functions in a basis of functions
which depend on the polar angle θ. Energy units: Hartrees, distance
Fig. 69. Ground state energy of an off-axis hydrogen atom enclosed units: Bohrs.
within an impenetrable cylindrical box as a function of the position of the
nucleus (relative to axis of the cylinder), for various radii of the confining
box. Note that when b/ρ0  1 and ρ0  1, the energy approaches −1/8
Hartree which corresponds to the ground state of the hydrogen atom close
shown. In Figure 69, we show the ground state energy as a function
to an infinite planar surface. of (b/ρ0 ) for different values of ρ0 .
The trial wave function is accomplished by referring the
electron-nucleus distance for the free atom to the origin placed
at the center of symmetry of the confining surface, without further
If the nucleus of the atom is located at the x-axis, at a distance
assumptions.
b (as depicted in Fig. 68), the electron-nucleus distance r 0 and the
The results obtained by applying the direct variational method
position of the electron relative to the origin on the axis of the
to compute the ground state energy of the hydrogen atom en-
cylinder r are related by
closed within spherical or cylindrical surfaces show good agree-
r 0 = |r0 | = |r − b| (458) ment with more elaborated calculations as can be seen from Ta-
bles VI and VII. Furthermore, the trial wave functions are flexible
that is,
enough to describe the case when the size of the confining surface
h i1/2
r 0 = (x − b)2 + y 2 + z2 = [ρ2 + b2 − 2ρb cos φ + z2 ]1/2 becomes infinite and the nucleus is close to the boundary. The lat-
ter corresponds to the case when the hydrogen atom is close to a
(459) plane.
The trial wave function can be written as
 
χ(r) = B(ρ − ρ0 ) exp −α|r − b| (460) 4.4. Confined Electron and Hydrogenic Donor States in a
where ρ0 is the radius of the cylinder and |r − b| is given by Spherical Quantum Dot
Eq. (459). The ground state energy resulting from the variational According to hydrogenic effective mass theory, exact solutions and
calculations, using this trial wave function, are displayed in Ta- quantum level structures are presented for confined electron and
ble VII. A comparison with the results from Ref. [164] is also hydrogenic donor states in a spherical quantum dot of GaAs–
 OPTICAL PROPERTIES IN NANOSTRUCTURES 73

Ga1−x Alx As [165]. Calculated results reveal that the values of the is still reliable for the bound states in spherical quantum dots of
quantum levels of a confined electron in a quantum dot can be GaAs–Ga1−x Alx As.
quite different for cases with finite and infinite barrier heights. The Let us for definiteness consider a hydrogenic donor at the cen-
quantum level sequence and degeneracy for an electron in a quan- ter of the quantum dot of radius r0 . The potential due to the dis-
tum dot are similar to those of a superatom of GaAs–Ga1−x Alx As continuity of the band edges at the GaAs–Ga1−x Alx As interface
but different from those in a Coulombic field. There is a stronger r = r0 is as

confinement and a larger binding energy for a hydrogenic donor in V0 r ≥ r0
Vb (r) = (461)
a spherical quantum dot of GaAs–Ga1−x Alx As than in the corre- 0 r < r0
sponding quantum-well wires and two-dimensional quantum well
structures. The binding energy and its maximum of the ground where r is the electron-donor distance. The barrier height V0 is ob-
state of a donor at the center of a quantum well are found to be tained from a fixed ratio of the bandgap discontinuity. According
strongly dependent on the well dimensionality and barrier height. to hydrogenic mass theory, the Hamiltonian for the donor is
Because the transverse and longitudinal variables do not sep-
b = −∇ 2 − 2w + Vb (r)
H (462)
arate, the impurity states in two-dimensional quantum wells and r
quantum-well wires cannot be solved exactly. Therefore, approx-
It is written in a dimensionless form so that all energies are
imation methods should be used. A reasonable trial function is
measured in units of the effective Rydberg R∗y and all distances
needed to obtain a correct variational state of an impurity in two-
are measured in units of the effective Bohr radius a∗B . w is equal
dimensional and one-dimensional confined systems, and calcu-
to 1.
lated results are more accurate if the coupling effect between the
To solve the Schrödinger-like equation,
impurity and well potentials is considered using a trial function
which or a part of which has correctly both donor and well po- b
Hψ(r; θ; ϕ) = εψ(r; θ; ϕ) (463)
tential effects [166]. However, for a hydrogenic donor at the cen-
ter of the spherical quantum dots the exact solutions [167] can be the wave functions of an electron with well-defined values of the
obtained. It is interesting not only from a physical point of view orbital (l) and magnetic (m) quantum numbers in a spherical sym-
but also from a mathematical point of view to compare the so- metric potential, which is the quantum well and Coulomb poten-
lutions and the binding energies with those of two-dimensional tial, are written in the form,
and one-dimensional systems. In this subsection, the exact solu- ψlm (r; θ; ϕ) = ψ(l) (r)Ylm (θ; ϕ) (464)
tions and quantum level structures for confined electron and hy-
drogenic donor states in spherical quantum dots are reported. The Substituting Eq. (464) in Eq. (463), we find an equation for the
dependence of the quantum levels and the binding energies on the radial function:
dimensionality of quantum wells is also presented.
d 2 ψ(l) (r) dψ(l) (r)
The calculation is based on the effective mass approxima- r2 + 2r
dr 2 dr
tion. It is known to give excellent results for electronic struc- nh i o
ture of GaAs–Ga1−x Alx As two-dimensional quantum wells and + ε(l) − Vb (r) r 2 − l(l + 1) + 2wr ψ(l) (r) = 0 (465)
(AlAs)n /(GaAs)n superlattices if the well width or n is sufficiently
large. The limit is estimated to be about 30 Å (n ≈ 10) [168]. Using the method of series expansion, we can solve Eq. (465) ex-
Therefore, it should also be valid for the GaAs–Ga1−x Alx As actly. It should be noted that the zero and infinity are a regular and
quantum dots, as the size (diameter for a ball) is sufficiently large. an irregular singular point of Eq. (465), respectively. In the region
Based on the facts mentioned earlier, the limit for a ball diameter 0 < r, we have a series solution which has a finite value at r = 0,
is also estimated to be the same value and equal to 30 Å. Here, we as
∞
X
treat the cases where the diameter is larger than the critical size. It (l)
ψ(l) (r) = Ar l an r n (466)
is interesting to point out that the maximum quantum confinement
n=0
of an electron in the GaAs–Ga1−x Alx As quantum ball is already
obtained before the diameter approaches the critical value. In ad- where
dition, polarization and image charge effects can be significant if (l) (l) 1
a0 = 1 a1 = − (467)
there is a large dielectric discontinuity between the quantum ball l+1
and the surrounding medium [169]. However, this is not the case and
for the GaAs–Ga1−x Alx As quantum system; therefore we ignore h i
(l) (l) (l)
such effects. an = − 2wan−1 + ε(l)an−2 /n(n + 2l + 1)
According to hydrogenic effective mass theory, the electron n = 2; 3; 4; : : : (468)
bound states and their binding energies were found in two-
dimensional quantum wells and quantum-well wires. Normally, the A is a constant. In the region r0 < r, we can obtain a normal
effective mass equation is reliable weakly bound states, and one solution [170]. It approaches zero at r = ∞ and is found in the
might worry that the effective mass equation is inappropriate when form,
the binding energy is greatly enhanced in spherical quantum dots X
N
bn r −n
(l)
of GaAs–Ga1−x Alx As. However, the bandgap of GaAs is 1.4 eV, ψ(l) (r) = B exp(−Kl r)r ρl (469)
while R∗y = 5:3 meV. Thus, roughly a 100-fold enhancement of the n=0
binding energy is necessary before the effective mass equation be- where
comes inapplicable. This difference is much greater than the en- h i1/2 w
hancement seen in the cases considered here, so that the theory Kl = V0 − ε(l) ρl = −1 + (470)
Kl
74 RIERA ET AL.

and the Hamiltonian and the matching conditions. If the effective mass
(l) (l) difference is considered, similar formulas can be obtained.
b0 = 1 bn+1 = −(ρl − n − 1)
If there is no Coulomb potential in the Hamiltonian of
(l)
× (ρl − n + l + 1)bn /2Kl (n + 1) Eq. (462), i.e., w = 0, using the same formulas, we can obtain the
(l)
n = 0; 1; 2; : : : (471) wave function ψn (r; w = 0), and the quantum levels εn (l; w = 0)
of an electron in the quantum well. In fact, Eqs. (466) and (469)
B is a constant. The series appears suitable for numerical com- become the spherical Bessel function and the Hankel function if
putation for large r [170]. However, they are not suitable for r0 w = 0. The equation of eigenenergies ε(l; w = 0) is as
if it is small. To get the exact value at small r0 , we find a solution
of uniformly convergent Taylor series in the region r0 < r ≤ Rp , k0 + K0 tan(k0 r0 ) = 0 if l = 0 (477)
where Rp is a proper point, e.g., Rp ≥ 2a∗B , for using Eq. (469).
For the sake of using the matching conditions at r = Rp to obtain ikl hl (iKl r0 )jl−1 (kl r0 ) + Kl hl−1 (iKl r0 )jl (kl r0 ) = 0
the eigenenergy equation in the following, it is written as if l ≥ 1 (478)
∞
X ∞
X and
ψ(l) (r) = C cn (r − Rp )n + D dn (r − Rp )n (472) h i1/2 h i1/2
n=0 n=0 kl = ε(l; w = 0) Kl = V0 − ε(l; w = 0) (479)
where C and D are constants, c0 and d1 are equal to 1, and c1 and
where jl and hl are the lth-order spherical Bessel function and
d0 are equal to 0, respectively. Noting that cn and dn are equal to
Hankel functions of the first kind, respectively. Then, the same re-
0 for negative n, the other cn can be determined by the following
sults are obtained if the wave functions and quantum levels are
recurrence relation,
n calculated with the use of the Bessel and the Hankel functions.
cn = −2Rp (n − 1)2 cn−1 Once εn (l; w = 1) and εn (l; w = 0) are obtained, the binding en-
h i ergy of the corresponding donor states in the spherical quantum
+ −(n − 2)(n − 1) + l(l + 1) − 2wRp + Kl2 R2p cn−2 dot is given by
o.h i
+ 2(Kl2 Rp − w)cn−3 + Kl2 cn−4 R2p n(n − 1) (473) εnb (l) = εn (l; w = 0) − εn (l; w = 1) (480)

and dn s obey a similar recurrence relation.

4.4.1. Quantum Levels and Binding Energies
Using the matching conditions at the interface r = r0 and Rp ,
we can obtain the equation of the eigenenergies ε(l) as A numerical calculation for GaAs–Ga1−x Alx As spherical quan-
  tum dots of the r0 between 0:15a∗B and 7:0a∗B with different V0
 W11 0 W13 W14 
  was performed. In Table VIII, the quantum levels of an electron
 W21 0 W23 W24 
 =0 (474) in a spherical quantum dot with different r0 and V0 were shown.
 0 W32 1 0 
  The levels εn (l) are indicated by two symbols n and l as shown in
0 W42 0 1
Eq. (480). The n is equal to the number of the root of Eqs. (476) or
that is, (477) and (478) of increasing magnitude, i.e., n = 1; 2; 3; : : : and
W21 (W42 W14 + W32 W13 ) − W11 (W42 W24 + W32 W23 ) = 0 (475) the n−1, hence, is the radial quantum number as usual. The l is the
usual notation, i.e., s; p; d; : : : . Thus, we have 1p, 1d, 2s, 1f levels
where (states) and so on if the n and l are used as the level notation, and
∞
X ∞
X we have 1s, 2p, 3d, 2s, 4f levels, and so on, if the principal quan-
(l)
W11 = an r0n W13 = cn (r0 − Rp )n tum number, which is equal to n + l, and l is used as the notation.
n=0 n=0 It is interesting to point out that when V0 approaches infinity,
X∞ ∞
X  2
(l)
W14 = dn (r0 − Rp )n W21 = (l + n)an r0n−1 εn (l) = µln /r0 (481)
n=1 n=0
X∞ X
N where µln is the nth root of the lth-order spherical Bessel function.
bn R−n
(l)
W23 = ncn (r0 − Rp )n−1 W32 = p (476) In Table VIII, it is shown that the different values of εn (l) are ob-
n=0 n=0 tained as the r0 is equal to 1a∗B and 2:5a∗B , respectively. It is also
X∞ shown that the values of quantum levels are different between in-
W24 = ndn (r0 − Rp )n−1 finite and finite barrier heights. The differences increase as the r0
n=1 and finite V0 decrease. There are an infinite number and a finite
number of bound states for a spherical quantum dot with infinite
XN
(−n + ρl − Kl Rp )bn R−n−1
(l) and finite barrier heights, respectively. There is no bound state if
W42 = p
n=0 r0 < Rc = 0:5π/(V0 )1/2 [171]. However, the order of εn (l) is the
same for both infinite and finite barriers, i.e., the unique level se-
it can be solved numerically. Once the nth eigenenergy εn (l) is
(l)
quence 1s, 1p(2p), 1s(3d), 2s, 1f(4f ), and so on. We should note
known, the A, B, C, and D [hence ψn (r)] are known with the use that the level order is different between both cases of a spherical
(l) (l)
of the normalized condition of ψn (r). This ψn (r) depends on quantum dot and the Coulomb field, in which the level order of
the value of l, the quantum well, the Coulomb potential, and the an electron is 1s, 2s, 2p, 3s, 3p, 3d, and so on if the principal and
energy εn (l). We should point out that we neglected the difference orbital quantum numbers are used as the level notation. It is be-
of the electron effective mass between GaAs and Ga1−x Alx As in cause of the lack of the deep attractive region in the vicinity of
 OPTICAL PROPERTIES IN NANOSTRUCTURES 75

Table VIII. Quantum Levels of an Electron in an SQD of GaAs–Ga1−x Alx As∗

1s 1p 1d 2s 1f 2p
ε(l) nl (1s) (2p) (3d) (2s) (4f) (3p)

r0 = 1 V0 = ∞ 9.872 20.187 33.212 39.476 48.832 59.676

80 7.957 16.225 26.593 31.425 38.919 47.016
60 7.702 15.679 25.642 30.191 37.420 44.789
40 7.292 14.786 24.045 28.004 34.777
2.5 ∞ 1.580 3.230 5.314 6.316 7.814 9.548
80 1.446 2.958 4.866 5.781 7.151 8.735
60 1.427 2.919 4.800 5.702 7.054 8.613
40 1.396 2.854 4.692 5.572 6.892 8.409

*Effective atomic units are used.

the center of a spherical quantum dot. For the motion of an elec-

tron in a Coulomb field, the quantum levels are only dependent on
the principal quantum number np , and degenerate with respect to
both l (orbital quantum number) and m (magnetic quantum num-
ber). The total degree of degeneracy of a quantum level with the
np is equal to n2p (excluding spin degeneracy). For an electron in
a spherical quantum dot, however, the quantum levels are depen-
dent on both n and l and only degenerate with respect to the m.
The total degree of degeneracy of a quantum level with n and l
is equal to 2l + 1 (excluding spin degeneracy). It is worthwhile to
point out that the degeneracy can be lifted in the other kinds of
quantum dots. In quantum boxes with circle cross-sections, for ex-
ample, the degeneracy is lifted partly. Now, we can conclude that
the quantum-level sequence and degeneracy for an electron in a
spherical quantum dot are quite different from those in a Coulomb
field, and that this distinguishing feature of levels might cause new
phenomena in this type of GaAs–Ga1−x Alx As structure.
In Figure 70, the ground and first excited energy levels of an
electron in a spherical quantum dot as a function of r0 for an in-
finite barrier height and two finite barrier heights V0 = 40 and
80R∗y were, respectively, plotted. It is shown that the differences
of energy levels between different barrier heights increase as the
r0 is decreased, and that the difference of the first excited state
energy is larger than that of the ground state energy for a fixed
value of r0 . It is also shown that there are no bound states for a
spherical quantum dot with a finite V0 if r0 < Rc , as mentioned
previously. In Figure 71, the binding energies of the ground and
first excited states of a donor in a spherical quantum dot as a func-
tion of the r0 for three barrier heights V0 = 80, 60, and 40R∗y ,
respectively, were shown. It is readily seen that as r0 decreases
both the binding energies increase continuously until their maxima Fig. 70. Ground state energy (ε10 ) and first excited state (ε11 ) energy
and, then, decrease fast. The values of the binding energies can be levels of an electron in a spherical quantum dot versus the dot radius r0 .
much larger than those of quantum-well wire and two-dimensional The top and middle dashed curves represent the levels ε11 and ε10 of the
quantum well as r0 is smaller. It is interesting to point out the ra- dot of V0 = ∞, respectively. The solid curves a, b, c, and d represents the
tio ε1b (0)/ε1b (1) increases as r0 increases from some small value. levels ε11 and ε10 of the wells of V0 = 80 and 40R∗y , respectively.
The ε1b (0) and ε1b (1) are almost independent of V0 and, respec-
tively, equal to 1:192 and 0:576R∗y at r0 = 7:0a∗B . However, the
ratio 1:192/0:576 is still much less than 4, which is the limit value
of a three-dimensional hydrogenic donor as r0 (approaches infin- increasing of the binding energies are always much less than the in-
ity). creasing of the energies of the corresponding states of an electron
In Figures 70 and 71, it is easily seen that as the r0 decreases only confined by the spherical quantum dot, although the binding
the binding energies with respect to different states of a donor in energies can be much larger than those in the corresponding two-
a spherical quantum dot increase until their maxima, and that the dimensional quantum well and quantum-well wire. It means that
76 RIERA ET AL.

Fig. 72. Maximum binding energies εb max for the hydrogenic donor
ground (l = 0) and the first excited (l = 1) states in a spherical quantum
dot versus the barrier height V0 . Arrows indicate the relevant scales.

Fig. 71. Binding energy of the ground state (ε0B ) and the first excited
(ε1B ) states of a donor in a spherical quantum dot versus the dot radius r0 .
The curves a, ab, and b represent ε0B of the dot of V0 = 80, 60, and 40R∗y ,
and the curves c, cd, and d represent ε1B of V0 = 80, 60, and 40R∗y , respec-
tively. Arrows indicate the relevant vertical scales.

confinement effects [172, 173] are dominant in the range of r0 .

Further, it is also true for the higher excited states. Therefore, we
can know what kind of quantum-level sequence we will have if the
motion of an electron is confined by both the spherical quantum
well and the Coulomb field with a same center. The level sequence
is similar to that of a three-dimensional hydrogenic donor if r0 is
much larger and quantum confinement due to the spherical quan- Fig. 73. Maximum binding energies εb max of a donor ground state in a
tum dot is very weak. However, it is similar to that of the elec- quantum well versus the well dimensionality and the barrier height V0 . For
the quasi-one-dimensional case, the dashed lines represent the maximum
tron in the spherical quantum dot if the quantum confinement of
binding energies of quantum-well wires and the solid lines represent the
the spherical quantum dot is stronger than that of the Coulomb
mean values of the maxima of the two-dimensional quantum wells and the
potential. Based on what we mentioned earlier, we can under- spherical quantum dots.
stand why the quantum-level structure of GaAs–Ga1−x Alx As su-
peratoms [170, 174] is similar to that of an electron in a spherical
quantum dot and quite different from those of ordinary atoms,
and that the electronic structure of the superatoms is dominated tron confinement in three dimensions in the spherical quantum
by no-radial-node states of 1s, 1p(2p), 1d(3d), and so on. dot. In Figure 73, it was shown that the maximum binding energy
In Figure 72, the maximum binding energies εb max for the hy- εb max of ground states of a donor at the center of a different kind
drogenic donor ground and first excited states in a spherical quan- of quantum well depends on the well dimensionality and barrier
tum dot as a function of the barrier height V0 were plotted. It height V0 and presents quasi-two-, -one-, and -zero-dimensional
is shown that the enhancement of the maximum is greater in a features of the hydrogenic donor, respectively. It is interesting to
spherical quantum dot (quasi-zero-dimensional) than in the cor- note that the mean values of the maxima of the two-dimensional
responding quantum-well wire and two-dimensional quantum well quantum well (Ref. [166]) and spherical quantum dot are very
as V0 is increased. This is because of the enhancement of the elec- close to (slightly larger than) the maxima for the quantum-well
 OPTICAL PROPERTIES IN NANOSTRUCTURES 77

wire [159]. We should point out that the maximum binding ener- The behavior of εb under an electric field is different for
gies of higher excited states can also be used to present the dimen- quantum-well wires of rectangular and cylindrical cross-section.
sion features. While the direction of the electric field is immaterial for cylindrical
The radial Eq. (465) was solved and the exact solutions of con- wires, it is very important for wires with rectangular cross-section.
fined electron and donor states in a spherical quantum dot were It is found that the binding energy of the hydrogenic impurities
obtained. The quantum levels and binding energies of a donor is a rather sensitive function of the geometry of the wire especially
in the spherical quantum dot are calculated numerically. The nu- under the influence of an electric field.
merical results reveal that the values of the quantum levels of a It is also found that the electric field effects on εb are extremely
confined electron in a spherical quantum dot with a finite barrier sensitive to the impurity position in or outside the wire.
height are different from those with an infinite barrier height. The
differences increase as the r0 decreases. However, the quantum-
level order is the same for both infinite and finite harrier heights. 4.5.1. Theory and Calculation
It is also shown that the quantum-level sequence and degeneracy It is convenient to use the Cartesian coordinates for wires of rect-
for an electron in a spherical quantum dot are similar to those of angular cross-section and cylindrical coordinates for wires of cir-
a superatom and are different from those in a Coulomb field. The cular cross-section. The Hamiltonian for the wire of rectangular
quantum-level structure of a donor in a spherical quantum dot is cross-section, lying along the z-direction, is
similar to that of an electron only confined by the spherical quan- !
h̄2 ∂2 ∂ 2
tum dot as the quantum confinement due to the fact that the well b0 = −
H + + Vb (x; y) (482)
potential is stronger than that due to the donor potential. It is use- 2m∗e ∂x2 ∂y 2
ful for understanding the shell model [174] in microclusters.
On the basis of the calculated results, the crossover from three- where

dimensional to zero-dimensional behavior of the donor states in a  Lx Ly

 V0 x<− y<−
spherical quantum dot is shown when the radius becomes small. 
 2 2

The binding energy of a hydrogenic donor state in the well of Lx Lx Ly Ly
Vb (x; y) = 0 − ≤x≤ − ≤y≤ (483)
GaAs–Gax Al1−x As and its maximum are strongly dependent on 
 2 2 2 2


the well dimensionality and the barrier height and there is a larger 
 V0 Lx Ly
<x <y
confinement and binding energy of a donor state in a spheri- 2 2
cal quantum dot than in a quantum-well wire and quantum well. Thus, the electron is free to move along the z-direction, but
Using calculated results of a quantum well and a quantum-well constrained along the x- and the y-directions. The wire subband
wire, it was shown that the maxima of the binding energies of structure is obtained by the variational method using the following
hydrogenic donors in a quantum well, quantum-well wire, and trial wave function,
a spherical quantum dot of GaAs–Gax Al1−x As can be used to
present, respectively, quasi-two-, -one-, and -zero-dimensional fea- f0 (x; y)
 h i
tures of the hydrogenic donor states. Further, it was found that 
 exp k2 (x + y)


the maximum of the binding energy of donor ground state in a 


 x < − Lx y < − Ly
GaAs–Gax Al1−x As quantum-well wire is about half of the sum- 


 2 2
mation of the maximum binding energies in the corresponding 
 cos(k1 x) cos(k1 y)
two-dimensional quantum well and spherical quantum dot. = N0 Lx Lx Ly Ly (484)

 − ≤x≤ − ≤y≤
It should be pointed out that impurities could be located any- 
 h 2 i
2 2 2


where in a spherical quantum dot, and that the binding energies 
 exp −k (x + y)

 2
will decrease and the level ordering will change as the impurity lo- 

cation shifts to the edge or out of the spherical quantum dot. Based 
 Lx < x Ly < y
on the exact solutions obtained, the quantum levels and binding 2 2
energies of a donor located out of the center of a spherical quan- where N0 is the normalization coefficient,
tum dot can be obtained by use of a variation method. The exact q q
solutions are also useful for the calculation of excited states in a k1 = 2m∗e ε0 /h̄2 k2 = 2m∗e (V0 − ε0 )/h̄2 (485)
spherical quantum dot, which is a kind of quantum dot. It will be
For the infinite potential barrier model V0 is taken to be infinite
interesting to compare the calculated results about quantum levels
and the wave function outside the barrier to be zero. Matching the
and binding energies of impurity and exciton states in a spherical
wave function and its derivative at the boundaries yields
quantum dot with those of other kinds of quantum dots.
k21 (1 − cos k1 L)
k22 = (486)
4.5. Shallow Donors in a Quantum-well Wire: Electric 1 + cos k1 L
Field and Geometrical Effects where the cross-section is taken to be a square with sides Lx =
In this subsection, the effects of an external electric field on donor Ly = L.
binding energies in quantum-well wires with cylindrical and square Next, the effect of an electric field on the subband energies is
cross-sections are investigated. A system was chosen with a GaAs calculated by using the Hamiltonian,
quantum well surrounded by Alx Ga1−x As potential barriers in the b0 + η(x cos θ + y sin θ)
b1 = H
H (487)
x; y-plane. The electron is thus free to move in the z-direction, in
the absence of a Coulomb center binding the electron. A realistic where η = |e|E. E is the electric field strength applied in the x; y-
finite potential well model is considered [175]. plane and θ is the angle between the electric field and the x-axis.
78 RIERA ET AL.

The trial function in this case is modified to be

h i
ψ1 (x; y) = N1 ψ0 (x; y) exp −β(x cos θ + y sin θ)/L (488)

where N1 is the normalization coefficient and β is the variational

parameter.
With an impurity at (xi ; yi ; 0) the Hamiltonian becomes
2 2 e2
H b1 − h̄ ∂ − q
b2 = H (489)
2m∗e ∂z2
κ z2 + (x − xi )2 + (y − yi )2
The trial function for this problem is taken to be
ψ2 (x; y; z) = N2 ψ1 (x; y)
 q 
× exp −λ z2 + (x − xi )2 + (y − yi )2 (490)

where N2 is the normalization constant and λ is a variational

parameter. The binding energy of the electron is written with
respect to the subband energy calculated in the presence of
an applied electric field. Numerical results are found for the
GaAs/Ga1−x Alx As system where, within the finite barrier model
that is taken, it is considered that the finite barrier potential and
dielectric constant depend on the x concentration of Al as
V0 = 0:6(1:247x) eV κ = 12:5(1 − x) + 10:1x (491)
For cylindrical quantum-well wires, the Hamiltonian for the rel-
ative motion is given by
!
b h̄2 ∂2 1 ∂ 1 ∂2
H0 = − ∗ + + + Vb (r; ϕ) (492)
2me ∂r 2 r ∂r r 2 ∂ϕ2
where 
0 r ≤ r0
Vb (r; ϕ) = (493) Fig. 74. The binding energy as a function of (a) side length of the square
V0 r ≥ r0 wire and (b) radius of the cylindrical wire. For both cases, E = 0, 10, and
The wave function for the ground state becomes 20 kV/cm (from top to bottom).

 J0 (k10 r) r ≤ r0
ψ0 (r; ϕ) = N0 J0 (k10 r0 ) (494)
 K (k r) r ≥ r0 The results are in perfect agreement with previous calculations
K0 (k20 r0 ) 0 20
without the electric field. For example, the binding energies are
where s s found to be almost identical for wires of circular and square cross-
2m∗e 2m∗e sections if wire dimensions are taken to be comparable. The bind-
k10 = ε0 k20 = (V0 − ε0 ) (495) ing energies are different for finite and infinite barrier potentials.
h̄2 h̄2
For infinite barrier potentials, the binding energy εb tends to in-
With an applied electric field in the x; y-plane, the Hamiltonian finity as the diameter of the wire tends to zero. It tends to smaller
becomes finite values for finite barrier potentials.
b1 = H
H b0 + ηr cos(ϕ − θ) (496) The behavior of the binding energy under a different electric
The trial function in this case is taken to be field is shown in Figure 74(a) and (b) for wires of square and cir-
h i cular cross-sections, respectively. The electric field is taken to be
ψ1 (r; ϕ) = N1 ψ0 (r; ϕ) exp −βr cos(ϕ − θ)/r0 (497) applied along the positive axis direction with θ = 0. Thus, the elec-
tron shifts toward the negative part of the axis. It is seen from these
where β is the variational parameter. The Hamiltonian becomes
! figures that the binding energy is a more sensitive function of the
2 ∂2 e2 electric field for wires of cylindrical cross-section.
b2 = H
H b1 − h̄ − q (498)
∗
2me ∂z2 The impurity position dependence of the binding energy is
κ z2 + (r − ri )2 shown in Figure 75(a) when the impurity position is changed
√
The trial function for the bound electron is taken to be along the diagonal of the square wire (L = πa∗B ) for different
h q i electric fields. The impurity position dependence of the binding
ψ2 (r; ϕ; z) = N2 ψ1 (r; ϕ) exp −λ z2 + (r − ri )2 (499) energy is shown in Figure 75(b) when the impurity position is
changed radially for a cylindrical wire (r0 = a∗B ) for different elec-
where λ is the variational parameter. tric fields. As expected, the binding energy becomes smaller for
The calculations are carried out for the model system of impurities located at the boundary of the wires since for this posi-
GaAs/Alx Ga1−x As for which the effective Bohr radius is a∗B = tion the quantum-well walls prevent the electron from feeling the
98:7 Å and the effective Rydberg is R∗y = 5:83 meV for x = 0:3. full power of the Coulomb attraction.
 OPTICAL PROPERTIES IN NANOSTRUCTURES 79

Fig. 76. The binding energy as a function of the impurity position along
√
Fig. 75. The binding energy as a function of the impurity position along
√ the (a) x-axis of the square wire (L = π2a∗B ) and (b) radius of the cylin-
the (a) diagonal of the square wire (L = πa∗B ) and (b) radius of the ∗
drical wire (r0 = 2aB ). Same electric field values as in Figure 74.
∗
cylindrical wire (r0 = aB ). Same electric field values as in Figure 74.

The behavior of the binding energy is also checked as the wire

dimension is increased. Figure 76(a) shows the binding energy as
a function of impurity position along the x-axis of the square wire
√
(L = π2a∗B ) for different electric fields. Now, the binding en-
ergy is generally smaller since the walls are not pushing the elec-
tron closer to the Coulomb center. Figure 76(b) shows the binding
energy as a function of impurity position along the radius of the
cylindrical wire (r0 = 2a∗B ) for different electric fields. In this case
also, the binding energy becomes generally smaller for the same
reason.
The Al concentration (x) dependence of the binding energy is
shown in Figure 77. As expected, increasing the potential barrier
increases the binding energy, because the electron is better pushed
toward the Coulomb center when the walls are at a higher poten-
tial. Fig. 77. The binding energy as a function of Al concentration for hydro-
In conclusion, the variational method employed in this context genic impurities in a cylindrical wire (r0 = a∗B ). Same electric field values
is capable of giving all the correct trends for impurity binding en- as in Figure 74.
ergies as a function of the applied electric field, impurity position,
and wire geometry.
field are calculated by a variational method within the effective
mass approximation [176]. The simple hydrogenic trial wave func-
4.6. Hydrogenic Impurities in a Spherical Quantum Dot
tions used in this calculation are flexible enough to treat on-center,
in the Presence of a Magnetic Field
off-center, and edge states of impurities in a quantum dot. Over-
The ground state and binding energies for a hydrogenic impurity all results are in reasonable agreement when compared to other
in a spherical quantum dot in the presence of a uniform magnetic calculations. Interesting enough, in the case of an off-center im-
80 RIERA ET AL.

purity, a critical point from which bulk states change their nature A similar choice of the trial wave function to calculate the ground
(edge states) is also found. state energy was widely used in the case of hydrogenic impurities
Semiconductors composed of III-V materials usually have small within spherical quantum dots [143] and cylindrical quantum wires
effective masses, which means that the Bohr radius associated to [155] without a magnetic field.
the impurity is large as compared with achievable quantum dot In the case of a hydrogenic impurity displaced a constant dis-
sizes. The binding energy of the impurity increases as the size tance a from the center of symmetry of a quantum dot, the Hamil-
of the confining region becomes of the order of the Bohr ra- tonian of Eq. (500) is modified as
dius, which makes possible, for instance, the fabrication of a low- !2
threshold laser diode [177]. b 1 e e2
H= p̂ − A − + Vb (r) (505)
In this subsection, a variational study about the effect of a mag- 2m∗e c κ|r − aez |
netic field on the ground state and binding energies of a hydro-
genic impurity within a spherical quantum dot is made. The impu- where the confining potential, Vb (r), is still given by Eq. (501), and
rity can be located on (off) the center of symmetry to get a symmet- ez is a unit vector on the z-axis. Hence, the corresponding Hamil-
ric (asymmetric) situation, or near the boundary of the quantum tonian and trial wave function, Eqs. (502) and (504), are given by
dot to get surface-like states (edge states) of this system. e2
b = − 1 ∇2 −
H −
e
Bb
Lz
∗
2me κ|r − aez | 2m∗e c
4.6.1. Theory and Model e2
+ B2 r 2 sin2 θ (506)
Within the framework of the effective mass approximation, the 8m∗e c 2
Hamiltonian of a hydrogenic impurity located at the center of sym-
and
metry of a spherical quantum dot of radius r0 , in the presence of a    
magnetic field, can be written as ψ(r) = N exp −α|r − aez | exp − 18 Br 2 sin2 θ (r0 − r) (507)
!2
b= 1 e e2 respectively.
H ∗ p̂ − A − + Vb (r) (500) Notice that the wave function (Eq. (507)) satisfies the boundary
2me c κr
condition imposed by Eq. (503). Again, since m = 0 is chosen, thus
where κ is the dielectric constant of material inside the quantum the Hamiltonian (Eq. (506)) will not have linear dependence on a
dot, m∗e is the electron effective mass, A is the vector potential, and magnetic field B, just like the symmetric case studied before.
Vb (r) is the confining potential defined by As we see, the trial wave function (Eq. (507)) is flexible enough
 to describe the case when the radius of the confining sphere be-
0 0 ≤ r ≤ r0
Vb (r) = (501) comes infinite and the hydrogenic impurity is close to its boundary.
∞ r > r0
The latter would correspond to the case of a hydrogenic impurity
For a uniform magnetic field, A(r) = B × r/2 if the Coulomb close to a plane, which allows us to investigate the ground state
gauge ∇ · A = 0 is chosen. In spherical coordinates, if the magnetic edge level of the spherical quantum dot.
field is along the z-direction, B = Bez , the Hamiltonian can be The binding energy (εb ) for the hydrogenic impurity is defined
expressed as as the ground state energy of the system without Coulombic inter-
2 action (εw ) minus the ground state energy including the presence
b = − 1 ∇ 2 − e − e Bb
H Lz of the electron-nucleus interaction (ε0 ), i.e.,
∗
2me κr 2m∗e c
εb = εw − ε0 (508)
e2
+ B2 r 2 sin2 θ (502)
8m∗ c 2 The binding energy defined before is a positive quantity. The εw
term of Eq. (508) is the ground state energy of an electron confined
Here, bLz is the z-component of the angular momentum operator,
inside a spherical quantum dot of radius r0 , that is
θ is the usual polar angle in these coordinates. The magnetic field
preserves the azimuthal symmetry, i.e., m is still a good quantum π 2 h̄2
number. As only the ground state is studied, the value m = 0 is re- εw = (509)
2m∗e r02
quired, thus the Hamiltonian (Eq. (502)) will retain the quadratic
dependence on B only. The values of the physical parameters pertaining to a GaAs quan-
Because of confinement, the solution of the Schrödinger equa- tum dot used in these calculations, for the sake of comparison,
tion associated to Eq. (502), must satisfy are m∗e = 0:067m0 and κ = 12:5, thus a∗B = 98:7 Å, and
R∗y = 5:83 meV. Moreover, the linear dependence term on B, in
ψ(r = r0 ; θ; ϕ) = 0 (503)
Eq. (502), can be scaled down to
To use the variational method, the trial wave function for the cal- eh̄ m
culation of the ground state energy is chosen as γm = Bm = ∗ 0 ∗ µ0 Bm (510)
  2m∗e cR∗y me Ry
ψ(r) = N exp(−αr) exp − 18 Br 2 sin2 θ (r0 − r)
  where µ0 = 5:78838263 × 10−2 meV T−1 is the Bohr magneton;
= N exp(−αr) exp − 12 1
Br 2 (r0 − r) (504) then the value B = 6:75 T corresponds to an effective Rydberg
(γ = 1).
where N is the normalization constant and α is a variational pa- The results for the symmetric case (on-center impurity) are dis-
rameter to be determined. The last term is obtained consider- played in Figure 78, in which the binding energy of the hydrogenic
ing that in the absence of A the system holds spherical symmetry. impurity is plotted as a function of the quantum dot radius for
 OPTICAL PROPERTIES IN NANOSTRUCTURES 81

Fig. 78. The binding energy of an on-center hydrogenic impurity as a Fig. 79. The binding energy of an on-center hydrogenic impurity as a
function of the quantum dot radius for different magnetic fields. function of the magnetic field for different dot radii.

different magnetic field strengths. The qualitative and quantita-

tive behavior of the results are in good agreement with those of
Xiao et al. [178], and Branis et al. [179], who studied the magnetic
field dependence over the binding energies of hydrogenic impuri-
ties in quantum dots and quantum-well wires, respectively. In the
strong spatial confinement case, (r0 < a∗B ), the binding energy is
relatively insensitive to the magnetic field and diverges as the dot
radius r0 → 0, as in the case of the zero magnetic field. The lat-
ter reflects that the main contribution to the binding energy is due
to electron spatial confinement, which prevails over magnetic field
confinement. The latter can be estimated through the magnetic
length defined as
r
h̄c
lB = = 484:82 × B−1/2 (511)
eB
where lB is expressed in effective Bohrs, and B is given in Tes-
las. For dot radius r0 ∼ = a∗B , the different magnetic field curves Fig. 80. The ground state energy of an off-center hydrogenic impurity as
deviate from each other reaching steady values as the dot radius a function of the relative position of the impurity for different dot radii.
increases. For weak spatial confinement (r0  a∗B ), the binding
energy converges asymptotically to the corresponding bulk values.
In the limit of a large dot radius, the binding energy for B = 0
approaches its bulk value εb = 1R∗y , as expected. An increase of
the field strength decreases the magnetic length as compared with
the dot radius; thus in this case the binding energy increases due to
magnetic confinement that compels the electron to move “closer”
to the on-center impurity.
Figure 79 shows the behavior of the binding energy as a func-
tion of field strength for different dot radii of an on-center impu-
rity. The variation of the binding energy as the field increases is
less sensitive as dot radius decreases, thus enforcing the previous
comment regarding Figure 78.
The results for the asymmetric case (off-center impurity) are
plotted in Figures 80 and 81 with zero magnetic field. Figure 80
shows the ground state energy for different sizes of quantum dot
versus the relative position of the off-center impurity (a/r0 ). The
ground state energy increases from its on-center impurity value
(a/r0 = 0) as the impurity is shifted off the center; this is a general
behavior. For larger radii, the ground state energy becomes lower, Fig. 81. The binding state energy of an off-center hydrogenic impurity as
and then, for closer distances to the wall, evolves from the 1s free a function of the relative position of the impurity for different dot radii.
82 RIERA ET AL.

Fig. 82. The ground state energy of a hydrogenic impurity in a spheri- Fig. 83. The binding state energy of a hydrogenic impurity as a function
cal quantum dot of very large radius. Note that the ground state changes of the relative position of the impurity for large dot radii. Note that there
gradually from its bulk value (−1R∗y ) to the Levine’s state (−1/4R∗y ). is a point from which the edge states start to appear.

hydrogen ground state (−1/2-Hartree) to 2pz free hydrogen first (r0 = 100 effective Bohr) decreases from 1R∗y (effective Rydberg)
excited state (−1/8-Hartree). These results are in good agreement for an impurity located far away from the quantum dot surface to
with those of Brownstein [163], also with the results of Gorecki and 1/4R∗y for impurity exactly located at the surface of the quantum
Byers Brown [162], and those of Marín et al. [143]. It is clear now dot. The binding energy behavior is in reasonable agreement with
that the trial wave function as chosen, is flexible enough to describe that of Chen [180], who studied the problem of hydrogenic donors
the case when the size of the quantum dot becomes infinite and the near semiconductor surfaces. Furthermore, starting from the crit-
nucleus is close to the surface. The latter corresponds to the case ical point (near the quantum dot surface), there is an inversion
of a hydrogenic impurity close to a plane. This situation will be of binding energy for the quantum dot of different radii. This oc-
discussed later when the effect of a magnetic field on the ground curs due to the fact that the boundary has a greater effect on the
state of a hydrogenic impurity near semiconductor surface will be binding energy when the quantum dot size increases, as a result of
considered. the strong spatial limitation due to the closeness of the impurity
In Figure 81, the binding energy for different sizes of the quan- to the surface. When a magnetic field is applied, the critical point
tum dot versus the relative position of the off-center impurity is removed and the binding energy curves cross one to the other
(a/r0 ) was plotted. The binding energy decreases as the impurity at only one point. Moreover, crossing points are shifted away from
approaches to the boundary of the quantum dot. This behavior is the boundary as the magnetic field is increased, finally a moderate
due to the fact that wave functions vanish at the boundaries and field strength causes them to vanish.
thus their contributions to the binding energy of a quantum dot
with an off-center impurity are smaller than a quantum dot with an
on-center impurity. The results are similar to the case of a hydro-
4.7. Two-electron Atomic Systems Confined within
genic impurity placed in rectangular [158] or circular [155] cross-
Spheroidal Boxes
section quantum-well wires, and in a spherical quantum dot [178]. The direct variational method is used to estimate some interest-
The problem of a hydrogenic impurity near a crystal surface as ing physical properties of the He atom and the Li+ ion confined
a limit case of a hydrogenic impurity in a spherical quantum dot of within impenetrable spheroidal boxes. A comparative investiga-
very large radius is considered next. In Figure 82, the ground state tion of the ground state energy, pressure, polarizability, dipole,
energy for different radii of the quantum dot versus the relative and quadrupole moments with those of the He atom inside boxes
position of the off-center impurity (a/r0 ) is shown. For very large with paraboloidal walls is made [181, 182]. The overall results show
quantum dot radii, the Levine’s ground state energy of an atom a similar qualitative behavior. However, for Li+ there are quan-
located precisely on the surface of a semi-infinite medium is ob- titative differences on such properties due to its major nuclear
tained, that is, the ground state energy changes gradually from its charge, as expected. The trial wave function is constructed as a
bulk value (−1 Rydberg) into the Levine’s state (−1/4 Rydberg) product of two hydrogenic wave functions adapted to the geome-
as the impurity reaches the surface. try of the confining boxes.
In Figure 83, the ground state binding energy for different radii Today, there is a special interest in the investigation of atoms
of the quantum dot versus the relative position of the off-center and molecules confined in nanostructures because their physical
impurity (a/r0 ) was plotted. For too large quantum dot radii con- properties become highly dependent on the size and shape of the
sidered here, binding energy curves cross one another and define confinement volume [143, 155, 156, 182–188]. The reduction of
a point from which the edge states begin to appear. This critical space where the atom is located is equivalent to subject it to high
point corresponds to a hydrogenic impurity placed at the relative pressures and the results of this new spatial condition are, besides
distance a/r0 = 0:68 from the center of the quantum dot. The other effects, the increase in total energy of the atom and the de-
associated binding energy with the quantum dot of largest radius crease of its polarizability.
 OPTICAL PROPERTIES IN NANOSTRUCTURES 83

The “atom in a box” model consists in replacing its interaction The electron–electron repulsive potential can be written as a series
with neighboring particles by a potential wall. The simplest situa- expansion in products of associate Legendre polynomials,
tion is to assume an infinite potential wall since for this case the " #2
∞
1 X Xl
wave function must vanish at the boundary of the box. The latter 1 (l − m)!
condition simplifies the calculations extremely. When a quantum = (2l + 1) m i m cos m(ϕ − ϕ0 )
r12 R (l + m)!
system is localized in a box, the result is that the wave function of l=0 m=0
the system is constrained into that region. × Plm (η)Plm (η0 )flm (ξ; ξ0 ) (518)
The study of hydrogenic impurities trapped in solids signaled
the limitations of the model of confinement inside spherical boxes. where
 m 0 m
For example, the observed hyperfine splitting of atomic hydro- Pl (ξ )Ql (ξ) ξ > ξ0
flm (ξ; ξ0 ) = (519)
gen in α-quartz shows the presence of nonvanishing anisotropic Plm (ξ)Qlm (ξ0 ) ξ < ξ0
components [189], and moreover, the electron microscopy studies
of semiconductor clusters show microparticles whose shapes vary m is the Neumann’s factor: 0 = 1; n = 2 (n = 1; 2; 3; : : :) [193].
from quasi-spherical to pyramidal [190]. This motivated us to use If the atom is enclosed into an impenetrable prolate spheroidal
other models involving nonspherical boxes. box defined as
The helium atom inside spherical boxes was investigated for
{ξ = ξ0 ; −1 ≤ η ≤ 1; 0 ≤ ϕ ≤ 2π} (520)
modeling the effect of pressure on atoms with more than one elec-
tron [183–185]. The helium atom inside boxes with paraboloidal then the wave functions must fulfill the requirements,
walls [182] and the helium atom in a semi-infinite space limited by
a paraboloidal boundary [191] were studied. ψ(ξ = ξ0 ; η; ϕ; ξ0 ; η0 ; ϕ0 ) = 0
(521)
ψ(ξ; η; ϕ; ξ0 = ξ0 ; η0 ; ϕ0 ) = 0
4.7.1. Variational Calculation of Ground State Energy
To obtain the ground state energy of the helium atom by us-
The prolate spheroidal coordinates (ξ; η; ϕ) are defined by ing the direct variational method, the ansatz wave function is con-
(r1 + r2 ) (r1 − r2 ) structed as a product of two hydrogenic functions, namely,
ξ= η= ϕ≡ϕ (512)
2R 2R
ψ(ξ; η; ϕ; ξ0 ; η0 ; ϕ0 )
where r1 and r2 are the distances from any point to two fixed points h i
(foci) separated by a distance 2R. In these coordinates, = A exp −αR(ξ + η)
h i
{ξ = const:; −1 ≤ η ≤ 1; 0 ≤ ϕ ≤ 2π} (513) × exp −αR(ξ0 + η0 ) (ξ − ξ0 )(ξ0 − ξ0 ) (522)

defines a family of ellipsoids of revolution around the z-axis, while where A is a normalization constant, α is a variational parameter
to be determined after minimizing the energy functional with the
{1 ≤ ξ ≤ ∞; η = const., 0 ≤ ϕ ≤ 2π} (514)
additional constrictions imposed by Eq. (521). To satisfy the latter,
defines a family of hyperboloids of revolution around the z-axis. two auxiliaries functions are included which have a similar contour
In both cases, 2R is the focal distance and ϕ is the usual azimuth to the spheroidal confinement box to allow the total wave function
angle [192]. to cancel out at the boundary of the box. The kinds of wave func-
If the nucleus of the helium atom is located on one of the foci tions like those given in Eq. (522), were used to investigate one-
(z = −Re), r1 and r2 denote the location of the two electrons and two-electron atomic systems confined by spherical penetra-
relative to the fixed nucleus and r12 is the interelectronic distance, ble boxes [156], the H, H+ 2
, and HeH2+ quantum systems within
then the Hamiltonian for this system formed by two electrons and soft spheroidal boxes [188], and the helium atom inside boxes with
one nucleus of charge Z (= 2) is given by paraboloidal walls [182]. In all of these confined quantum systems,
such simple functions can describe the symmetry in a qualitative
b = − 1 ∇2 − 1 ∇2 − Z − Z + 1
H (515) way. Quantitatively, the results are close to those obtained by other
2 1 2 2 r1 r2 r12 methods [183, 184].
where it is assumed that h̄ = m∗e = e = 1 (atomic units) and the The evaluation of the matrix element for the electron–electron
mass of the nucleus is considered to be very large as compared Coulomb repulsion with the trial wave function (Eq. (522)) in-
with the electron’s mass (Born–Oppenheimer approximation). volves the integration over the azimuthal angles, which can be
In spheroidal coordinates, the Laplacian operator can be writ- done immediately and only the term with m = 0 of Eq. (518) re-
ten as mains. The latter shows the symmetry of a system under rotations
( around the z-axis. Additionally, the associate Legendre polynomi-
2 1 ∂ 2 ∂ ∂ ∂ als become ordinary Legendre polynomials. A complete evalua-
∇ = (ξ − 1) + (1 − η2 )
R (ξ − η ) ∂ξ
2 2 2 ∂ξ ∂η ∂η tion of the leading expression for the matrix element is presented
) in Subsection 4.7.3.
2
ξ −η 2 ∂2
+ (516) The ground state energy of He and Li+ for R = 0:1, 0.5, and
(ξ − 1)(1 − η ) ∂ϕ2
2 2
1:0 Bohr as a function of volume of the enclosing box is displayed
and the electron-nucleus distances are, respectively, in Figure 84, where a comparison with results of Ley-Koo and
Flores-Flores [182] for the He atom within a paraboloidal box is
r1 = R(ξ + η) r2 = R(ξ0 + η0 ) (517) also made.
84 RIERA ET AL.

Following Ley-Koo and Flores-Flores, we used an adapted ex-

pression for the polarizability of atoms to molecules [194–196],
then the parallel and perpendicular components of the polarizabil-
ity are given by
* 2 +
4 X 2 2 16 h 2 i2
αk ≡ αzz = zi − hzi = hz i − hzi2 (524)
Z Z
i=1
* 2 +2
4 X 2 16 D 2 E2
α⊥ ≡ αxx = αyy = xi = x (525)
Z Z
i=1

Due to the asymmetric position of the nucleus and the asym-

metric distribution of the electrons inside the confinement box, the
atom has an electric dipole moment, which is given by the expec-
tation value of the electron position plus the nucleus contribution,
that is
*  2  +
X  h i
 
d = −RZez − ψ ri ψ = − ZR + 2hzi ez (526)
Fig. 84. Ground state energy for the helium atom and the lithium ion
 
i=1
versus the volume of the spheroidal box and the interfocal distance. Energy
units, Hartree; volume units, Bohr3 . The components of the electric quadrupole moment tensor are
given by
* 2 +
X
2
Qzz = Z(2R ) − 2 2 2
[2zi − xi − yi ]
i=1
h i
= Z(2R2 ) − 4 hz2 i − hx2 i (527)
1
Qxx = Qyy = − Qzz
2
Figure 86(a)–(c) shows the polarizability components for He and
Li+ as a function of major semiaxis, Rξ0 , for R = 0:1, 0.5, and
1:0 Bohr, while Figure 87 shows the electric dipole moment as a
function of Rξ0 (for the same R values given previously). Figure 88
shows the electric quadrupole moment as a function of Rξ0 for the
same values of Rξ.
As can be seen in Figure 84, the qualitative behavior for a con-
fined helium atom is that the energy increases from −2:847656
Hartree (corresponding to free helium atom) to a large positive
value as a consequence of volume reduction of the confinement
box. It can be also noted that for large volumes of spheroidal
Fig. 85. Pressure of the helium atom and the lithium ion versus the vol- boxes there is a weak confinement of the atom and the eccentric-
ume of the spheroidal box and the interfocal distance. The atomic unit for ity of spheroid becomes irrelevant, but for small volumes of the
pressure is equivalent to 2:94 × 1013 Pa. spheroidal box there is a strong confinement and the eccentricity
plays a very important role in the energy value as it can be corrob-
orated in Figure 84. For R = 0:5 Bohr, the energy curve coincides
4.7.2. Pressure, Polarizability, Dipole, and Quadrupole in both boxes. The latter is an important situation because it is be-
Moments ing compared to the energy of an atom enclosed within two boxes
of different shape, where additionally the nucleus of the atom for
The computed variational energy and trial wave functions allow
a paraboloidal box is located in the common focus (origin of coor-
calculating some of the properties of the confined atom. The aver-
dinates) of paraboloids (which corresponds to a symmetric case)
age pressure exerted by the boundary on the atom is given by
while for the spheroidal box the nucleus is located out of the ori-
dε 3 dε gin, leading to an asymmetric case. It seems that for R = 0:5 Bohr
P =− =− (523)
dV 3 2
4πR (3ξ − 1) 0
dξ there is no difference for the atom to be into either of these boxes,
0
i.e., boundary shape changes the electron distribution so that both
where V = 4πR3 ξ0 (ξ02 − 1)/3 is the volume of the spheroidal box effects combine to give the same energy for the atom. Regarding
and ε is the total ground state energy of the atom. The results the same figure, the ground state energy for Li+ increases from
obtained for pressure are displayed in Figure 85 for R = 0:1, 0.5, its asymptotic value −7:222656 Hartree as the volume of the con-
and 1:0 Bohr as a function of V ; they are also compared with the finement box decreases, showing the same qualitative behavior as
results of Ref. [182]. in the helium case. For the same confinement volume, the energy
 OPTICAL PROPERTIES IN NANOSTRUCTURES 85

Fig. 87. Electric dipole moment versus major semiaxis of the spheroidal
box. The atomic unit for the dipole moment is equivalent to 2:54 Debye.

Fig. 88. Electric quadrupole moment versus major semiaxis of the

spheroidal box. The atomic unit for the quadrupole moment is equivalent
to 1:34 × 10−26 esu × cm2 .

of Li+ is smaller than the energy of He due to the Coulomb at-

traction between the nucleus of Li+ (Z = 3) and the electrons are
greater than in the helium atom (Z = 2).
In Figure 85, curves of pressure applied on the atom by the
spheroidal box walls are displayed. In this figure, we again compare
the results for the helium atom inside impenetrable paraboloidal
boxes. As can be noted, the pressure curve for R = 0:5 Bohr is
nearest to the corresponding pressure curve for helium enclosed
within a paraboloidal box. In same figure, the pressure curves for
a lithium ion are also displayed. For the same R and confinement
volume, the pressure on the Li+ ion is smaller than for the He
atom because the Coulomb attraction for the former is greater,
which forces the ion electrons to move closer to the nucleus, i.e.,
they are farther away from the confinement walls than the He elec-
Fig. 86. Polarizability tensor components versus major semiaxis of trons.
spheroidal box. Polarizability units, Bohr3 ; distance units, Bohr. (a) R = In Figure 86(a)–(c), the polarizability tensor components αxx
0:1, (b) R = 0:5, and (c) R = 1:0. and αzz are shown. When R = 0:1 Bohr, they begin to decrease
from their asymptotic common value 0:986540 (corresponding to
86 RIERA ET AL.

a free helium atom) to zero as the volume of the enclosing box × Pl (η)Pl (η0 )fl (ξ; ξ0 )
diminishes. It can also be noted that for a nearly spherical box × (ξ2 − η2 )(ξ02 − η02 ) dξ dη dξ0 dη0 (528)
no differences on values of polarizability tensor components are
found. For R = 0:5 Bohr, the parallel component of polarizability where flm (ξ; ξ0 ) (Eq. (519)) is reduced to
αzz (which always takes a value greater than the value of transver- 
Pl (ξ0 )Ql (ξ) ξ > ξ0
sal component αxx ) approaches a positive value near to zero as fl (ξ; ξ0 ) = (529)
the volume of the box diminishes. Moreover, the influence of non- Pl (ξ)Ql (ξ0 ) ξ < ξ0
sphericity of the box on values of polarizability tensor components The integration over η and η0 coordinates give the following
can also be observed. When R = 1:0 Bohr, αzz has a minimum result,
at Rξ0 ∼ = 1:52 Bohr, then increases to a maximum located at * +
Rξ0 ∼ = 1:24 Bohr, and finally diminishes approaching a positive 1
= (2π)2 A2 R5
value greater than for the other two R values considered before. r12
This effect in polarizability is also present for Li+ , where its polar- ∞ Z ξ0 Z ξ0
X
izability decreases from its asymptotic value 0.153354. The latter × (2l + 1) exp[−2αRξ] exp[−2αRξ0 ]
is a consequence of the combination of: (a) the deformation of 1 1
l=0
the spheroidal box by increasing its eccentricity (given by 1/ξ0 )
as the volume diminishes and (b) the proximity of the wall to the × (ξ − ξ0 )2 (ξ0 − ξ0 )2 fl (ξ; ξ0 )
h i
nucleus, which makes the electrons align on the z-axis, thus the × ξ2 ξ02 I02 − (ξ2 + ξ02 )I0 I2 + I22 dξ dξ0 (530)
system reaches the maximum of its polarizability. Then, its subse-
quent decreasing (not to zero) is due to the asymmetric reduction where I0 and I2 are integrals which involve spherical Bessel func-
of space where the electrons can move. tions, that is
In Figure 87, the electric dipole moment for R = 0:1, 0.5, and Z 1
1:0 Bohr, respectively, are displayed. The feature of this physical I0 = exp(−2αRη)Pl (η)dη
quantity is its decreasing behavior from zero for the He atom and −1
from an asymptotic value (different for each value of R) for the ∞ Z 1
X 0
Li+ ion to limiting negative values as ξ0 → 1. All these asymp- = (2l0 + 1)il jl0 (i2αR)Pl0 (η)Pl (η)dη
−1
totic values can immediately be determined from Eq. (526). The l0 =0
∞
nonzero value of the dipole moment is due to the asymmetric po- 2 X 0 0
sition of the nucleus inside the spheroid which makes a permanent = (2l + 1)il jl0 (i2αR)δl0 l = 2il jl (i2αR) (531)
2l + 1 0
dipole moment (−ZR) since hzi → 0 as the volume approaches l =0
zero. This behavior is qualitatively different from that of Ref. [191] and
where the electric dipole moment of a helium atom in a semi- Z 1
infinite space limited by a paraboloidal boundary was investigated. I2 = η2 exp(−2αRη)Pl (η) dη
In the latter case, the magnitude of the electric dipole moment −1
seems to increase monotonically from zero (free helium atom) to ∞ Z 1
X 0
a limiting value as the volume is reduced. These variations show = (2l0 + 1)il jl0 (i2αR)Pl0 (η)Pl (η)η2 dη (532)
−1
the shifting of the electron distribution away from the boundary as l0 =0
the latter closes in around the nucleus. With the recurrence relation,
Figure 88 displays the electric quadrupole moment of the He
atom increasing from its asymptotic value (corresponding to the k+1 k
ηPk (η) = P (η) + P (η) (533)
free atom), to a positive limiting value different for each value of 2k + 1 k+1 2k + 1 k−1
R, as ξ0 → 1. In contrast to the case studied in [182], the positive the I2 integral can be written as
sign of the quadrupole moment indicates that the electrons pref-
X∞ Z 1
erentially distribute themselves on the major axis of the spheroid. 0
I2 = (2l0 + 1)il jl0 (i2αR)
Additionally, for this kind of confinement box, due to the asym- −1
l0 =0
metric nucleus position within the spheroid, there is not a maxi- (
mum for this property as in the case of paraboloidal walls where (l0 + 1)(l + 1)
× P 0 (η)Pl+1 (η)
the nucleus is located on the symmetry center of box. (2l0 + 1)(2l + 1) l +1
l0 l
+ P 0 (η)Pl−1 (η)
4.7.3. Explicit Evaluation of Coulombic Interaction Term (2l0 + 1)(2l + 1) l −1
The evaluation of expectation value of the electron–electron re- (l0 + 1)l
+ P 0 (η)Pl−1 (η)
pulsive potential requires performing integrals of the form, (2l + 1)(2l + 1) l +1
0
)
* + l0 (l + 1)
1 + P 0 (η)Pl+1 (η) dη
= (2π)2 A2 R5 (2l0 + 1)(2l + 1) l −1
r12
(
X∞ Z ξ0 Z 1 Z ξ0 Z 1 h i (l + 1)(l + 2) l+2
= 2 i jl+2 (i2αR)
× (2l + 1) exp −2αR(ξ + η) (2l + 1)(2l + 3)
1 −1 1 −1
l=0
h i (2l2 + 2l − 1) l
× exp −2αR(ξ0 + η0 ) (ξ − ξ0 )2 (ξ0 − ξ0 )2 + i j (i2αR)
(2l − 1)(2l + 3) l
 OPTICAL PROPERTIES IN NANOSTRUCTURES 87
)
l(l − 1) The formal solutions of the Schrödinger equation for the inner
+ il−2 jl−2 (i2αR) (534)
(2l − 1)(2l + 1) (ξ ≤ ξ0 ) and the outer (ξ ≥ ξ0 ) regions must satisfy continuity
conditions at the boundary (ξ = ξ0 ) which are equivalent to
The integrals of Eq. (530) are evaluated numerically. 
1 ∂ψi 1 ∂ψ0 
| =  (541)
ψi ∂ξ ξ=ξ0 ψ0 ∂ξ 
4.8. Quantum Systems within Penetrable Spheroidal ξ=ξ0
Boxes Equation (541) allows us to find the energy of the system if an
In this subsection, the direct variational method is used to study explicit form of the wave function is known.
some physical properties of H, H+ 2 , and HeH
2+ enclosed within The V0 = ∞ case was studied exactly by Ley-Koo and
soft spheroidal boxes [188]. The ground state energy, polarizabil- Cruz [197]. An approximate study of these systems for finite values
ity, and pressure are calculated for these systems as a function of V0 can be performed using a direct variational method. In this
of the size and penetrability of the boxes. For these systems, the case, an ansatz wave function χ must be constructed. Following
ground state energy as a function of box eccentricity for different Refs. [198] and [156], we have
values of the barrier height is calculated. Furthermore, to demon-
χi (ξ; η; ϕ) = Ai ψ0i (ξ; η; ϕ; α)f(ξ; α) (542)
strate the applicability of the variational method, other properties
of physical interest directly involving the wave function, such as the and
polarizability and the pressure are calculated. χo (ξ; η; ϕ) = Ao ψ0o (ξ; η; ϕ; α)f(ξ; α) (543)
where ψ0i and ψ0o are the solutions of the Schrödinger equation for
4.8.1. H, H+
2 , and HeH
2+ within Penetrable Spheroidal Boxes the free system [199, 200], f is an auxiliary function which allows
the total wave function to satisfy the condition shown in Eq. (541),
The Hamiltonian for a one-electron molecular ion in the Born–
and α is a variational parameter to be determined once the corre-
Oppenheimer approximation can be written as
sponding energy functional is minimized with the additional con-
b = − 1 ∇ 2 − Z1 − Z2 + Z1 Z2
H (535)
strictions improved by Eq. (541).
2 Following Coulson and Robinson [199], the ansatz wave func-
r1 r2 2R
tion for the ground state is constructed as
where the units are chosen to make h̄ = e = m∗e = 1. The sub- h i
scripts 1 and 2 denote the nucleus of charges Z1 , and Z2 , respec- χi (ξ; η; ϕ) = Ai (ξ0 − αξ) exp −α(ξ + η) (ξ ≤ ξ0 ) (544)
tively.
The prolate spheroidal coordinates (ξ; η; ϕ) are defined as and
[192],
χo (ξ; η; ϕ) = Ao exp(−βξ) exp(−αη) (ξ ≥ ξ0 ) (545)
(r + r2 ) (r − r2 )
ξ= 1 η= 1 ϕ≡ϕ (536)
2R 2R where Ai and A0 are two constants related through the normal-
where 2R is the interfocal distance of the prolate spheroids of rev- ization condition on the total wave function. A relation between
olution {ξ = const.; − 1 ≤ η ≤ 1; 0 ≤ ϕ ≤ 2π}. the variational parameters α and β can be obtained through the
If the nuclei are located in the foci of this coordinate system, continuity condition at ξ = ξ0 so that the energy functional is to
the Hamiltonian of Eq. (535) can be now expressed as be minimized with respect to only one parameter, either α or β.
The energy of the ground state of H, H+ 2 , and HeH
2+ for
∗
b = − 1 ∇2 −
H
Z1
−
Z2 Z Z
+ 1 2 (537) V0 = 0, 2, 8, 20, and ∞ Ry is displayed in Figure 89 as a function
2 R(ξ + η) R(ξ − η) 2R of the size of the enclosing box ξ0 . As can be noted, good agree-
where the Laplacian operator is then ment is found with the exact results for the case of V0 = ∞. When
 " # V0 has a finite value, the calculations show the correct qualitative
1  ∂ ∂ behavior, similar to that found by Ley-Koo and Rubinstein [201]
2
∇ = 2
(ξ − 1) for hydrogen within penetrable spherical boxes. To gain some con-
R2 (ξ2 − η2 )  ∂ξ ∂ξ
 fidence in the value of the ansatz wave functions, we calculated
" # the polarizability and pressure for H+
∂ 2 ∂  2 . In the Kirkwood approx-
+ (1 − η ) imation [196], the parallel and perpendicular components of the
∂η ∂η  polarizability are given by
1 ∂2 α// ≡ αzz = 4hz2 i2 α⊥ ≡ αxx = αyy = 4hx2 i2 (546)
+ (538)
R2 (ξ2 − 1)(1 − η2 ) ∂ϕ2
Finally, the pressure on the system due to the confinement can
We now consider the system as confined within a prolate sphe- be calculated through the relationship,
roidal box of eccentricity 1/ξ0 and barrier height V0 . The corre-
dε
sponding modified Hamiltonian is then P =− (547)
dV
b = − 1 ∇ 2 + Vb (ξ; η)
H (539) where V is the volume of the spheroidal box and ε is the total
2

where ground state energy of the system calculated at the equilibrium

bond length. The results obtained for these properties are dis-
Vb (ξ; η) played in Figure 90 for V0 = 0 and ∞ R∗y as a function of ξ0 .

Z1 /R(ξ + η) − Z2 /R(ξ − η) (1 ≤ ξ ≤ ξ0 ) As can be seen, good agreement is found when our results are
= (540) compared for V0 = ∞ with the calculations reported by LeSar
V0 (ξ0 ≤ ξ < ∞)
88 RIERA ET AL.

Fig. 89. Ground state energies for H (a), H+2 (b), and HeH
2+ (c) enclosed within penetrable spheroidal boxes as a function of box size ξ . The
0
nucleus is clamped and R = 1 au. The continuous lines represent the variational calculations of this subsection. The circles are the exact results
for V0 = ∞ reported in Ref. [197]. (Reprinted from E. Ley-Koo and S. A. Cruz, J. Chem. Phys., 74, 4603, (1981).)

and Herschbach [186, 187], who used a five-term James–Coolidge rier. Additionally, some quantities of physical interest are evalu-
variational wave function. For finite values of V0 , the correct qual- ated for the case of hydrogen.
itative behavior is obtained [202]. The case of hydrogen was thoroughly analyzed by Ley-Koo and
Rubinstein [201]. Proper comparison with their results will yield
some confidence on the capability of the procedure proposed in
4.9. Confinement of One- and Two-electron Atomic this work to account for the behavior of certain physical proper-
Systems by Spherical Penetrable Boxes ties of the system as a function of box size and barrier height. Ac-
cordingly, for this system, in addition to the energy we evaluate the
We explicitly treat the case of some one- and two-electron atomic Fermi contact term, polarizability, pressure, and magnetic shield-
systems (H, H− , He, Li2+ , and Be2+ ) in their ground state within ing constant.
spherical confining barriers [144]. In all cases, a simple representa- For the rest of the systems, the estimated energies as a function
tion for the ansatz wave function is employed and the correspond- of the characteristics of the confining potential and the comparison
ing energies are given as a function of size and height of the bar- of their qualitative and quantitative behavior with available calcu-
 OPTICAL PROPERTIES IN NANOSTRUCTURES 89

Fig. 90. Physical properties of H+ 2 enclosed within penetrable spheroidal boxes as a function of box size ξ0 . The
continuous lines represent the variational calculations of this subsection. The circles are the results of Refs. [186] and
[187], while the squares are those of Ref. [202]. (Reprinted from R. LeSar and D. R. Herschbach, J. Phys. Chem., 85,
2798, (1981); 87, 5202, (1983); J. Gorecki and W. Byers-Brown, J. Chem. Phys. 89, 2138, (1988).)

lations for the impenetrable case by Ludeña and Gregori [184] are These requirements, together with the fulfillment of the virial the-
presented. orem (Fernandez and Castro [203]), constitute the key assump-
tions to be employed to construct our ansatz wave functions.
According to the direct variational method, an upper bound to
4.9.1. Model and Theory
the energy for a particular state may be found by requiring that
For penetrable walls, let the ansatz wave function in the interior of D E D E
the box be given as bi |χi + χo |H
χi |H bo |χo = minimum (550)
 0 
χi = χ0;i f (548)
where
where χ0;i denotes the wave function of the free system and f de-
notes an auxiliary function that guarantees the adequate matching bi = − 1 ∇q2 + V(q)
H 2 q∈
at the boundary with the exterior wave function, (551)
bo = − 1 ∇ 2 + V0
H 2 q q ∈ 0
χo = χ0;e (549)
such that χ0;e keeps the proper asymptotic behavior characteristic  and 0 being the interior and exterior regions, respectively, and
of the system under study. Furthermore, the choice of the auxiliary {q} represents the set of coordinates used. V(q) is the acting poten-
function f must be such that it reduces to the familiar cutoff func- tial within the interior region and V0 is the height of the confining
tion proposed previously for an infinitely high confining potential. potential.
90 RIERA ET AL.

In addition to Eqs. (550) and (551), the normalization condi-

tion, D E D E
χi |χi + χo |χo 0 = 1 (552)
 
must be satisfied together with the requirement for continuity of
the logarithmic derivative at the boundary (q = q0 ):
 
1 ∂χi  1 ∂χo 
 =  (553)
χi ∂q  χo ∂q 
q=q0 q=q0

Indeed, if we are going to consider several states, we must im-

pose the constraint of orthogonality among them, both in the in-
terior as well as in the exterior region. Note that the preceding ar-
guments are of a general character and should apply to any system
whether confined by penetrable or impenetrable walls.

4.9.2. Ground State of the Confined Hydrogen Atom

The potential energy associated with the Hamiltonian of the hy-
drogen atom confined by a spherical box with penetrable walls is
given by a Coulomb term inside the box,
Ze2 Fig. 91. Ground state energy of H as a function of confining radius r0 , for
V(r) = − r < r0 (554) different barrier heights.
r
and a constant barrier height outside the box,
V(r) = V0 r > r0 (555)
with r0 being the radius of the confining box.
For the purposes of this subsection, we consider now the 1s
state. For this case, the corresponding wave functions are chosen
as
χi (1s) = Ni (r0 − γr) exp(−αr) r < r0
(556)
χo (1s) = No r −1 exp(−βr) r > r0
with α, β, and γ variational parameters and Ni and No normaliza-
tion constants.
In contrast with the impenetrable case, we considered three
variational parameters to allow for more flexibility in the match-
ing procedure at the boundary. In fact, only two of these parame-
ters (α; γ) need to be found, the third one being defined through
Eq. (553) as
αr (1 − γ) + 2γ − 1
β= 0 (557)
r0 (1 − γ)
Following Ley-Koo and Rubinstein [201], some quantities of phys-
ical interest such as the Fermi contact term (A), the diamagnetic
screening constant (σ), the polarizability (α), and the pressure (P) Fig. 92. Fermi contact term (A) for H as a function of confining radius r0 ,
are defined, respectively, as for different barrier heights, in units of (2/3)gβgn βn = 12:690 mT.
* +
 2 2
  e 1
A = 23 gβgn βn χi (1s) σ=
r=0 3µc 2 r
comparative purposes, Eq. (558) will be employed for the evalua-
α= 4
9 a−1 2 2
B hr i (558) tion of P.
1 dε 1   Figures 91–95 show the values obtained by this method for the
P = − = 2ε − hV i energy (ε), A, σ, α, and P for the 1s state with barrier heights
4πr02 dr0 4πr03
of 0, 4, and 10 Ryd (e2 /2aB ). For comparison, the corresponding
Fernandez and Castro [204] showed that the virial theorem for exact values calculated by Ley-Koo and Rubinstein are also shown.
systems with sectionally defined potentials must be reformulated As the reader must be aware, there is a fair agreement between
as compared to the case of systems constrained by the Dirich- the energy values obtained by the method used here and the exact
let boundary conditions. Hence, the use of Eq. (558) for P is not calculations. Also, regarding the values estimated for the various
correct for confining potentials with finite barrier heights but, for physical quantities, the agreement is also fairly good.
 OPTICAL PROPERTIES IN NANOSTRUCTURES 91

Fig. 93. Diamagnetic screening constant (σ) for H as a function of con- Fig. 95. Pressure (P) for H as a function of confining radius r0 , for differ-
fining radius r0 , for different barrier heights, in units of e2 /2µa0 c 2 . ent barrier heights, in units of 106 atm.

Considering each electron in a 1s state, the interior wave func-

tion is chosen as
h i
χi (r1 ; r2 ) = A(r0 − αr1 )(r0 − αr2 ) exp −α(r1 + r2 ) (561)

and the exterior function,

exp[−β(r1 + r2 )]
χo (r1 ; r2 ) = B (562)
r1 r2
with A and B normalizing constants. Note that in this case we are
using only two variational parameters. This choice will certainly
make the trial wave function too rigid. However, as will be seen
further in the following, even with this crude approximation fair
results may be obtained.
Following the procedure outlined by Eqs. (550)–(553), the en-
ergy values for H− , He, Li+ , and Be2+ are obtained as a function
of r0 and V0 . Table IX displays the results of this exercise for some
selected barrier heights. We note first the cases of He, Li+ , and
Be2+ for which the qualitative and quantitative behavior of the
energies tend to the correct values (Ludeña and Gregori [184])
as the barrier height approaches infinity. Note, however, that for
Fig. 94. Polarizability (α) for H as a function of confining radius r0 , for relatively large box sizes the energy values obtained for the pen-
different barrier heights, in units of 10−24 cm2 . etrable case lie slightly lower to those for the impenetrable case.
This effect is attributed to the rigidity imposed on the input wave
function through the assignment of a single variational parameter
4.9.3. Ground State of Confined Two-electron Systems for the interior as well as the exterior regions.
Interestingly enough, the H− system shows a particular sensi-
The Hamiltonian for a confined atomic system with two electrons tivity to the confining effect as may be verified from the overall
is given as trend of the energy values. Unfortunately, there are no other cal-
Hb = − 1 ∇ 2 − 1 ∇ 2 + V(r1 ; r2 ) (559)
2 1 2 2 culations to compare with for this confined system. The energies
with obtained in this work for large values of r0 correspond to the value
 ε = −0:473 Hartree reported by Wu and Falicov [205] in their
 Z Z 1
variational treatment of free H− with a single exponential param-
− − + r1 ; r2 ≤ r0
V(r1 ; r2 ) = r1 r2 |r1 − r2 | (560)
 eter. Furthermore, these authors show that using two exponential
V0 r1 ; r2 > r0
parameters in their trial wave function, the ground state energy for
where Z denotes the nuclear charge; r1 ; r2 are the electron posi- H− becomes ε = −0:513 Hartree; a closer value to that of −0:527
tions relative to the nucleus and r0 the radius of the confining box. reported by Pekeris [206] using a 1078-parameter wave function.
92 RIERA ET AL.

Table IX. Energy Values and Variational Parameters as a Function of Box Size for Various Heights of the Potential Barrier∗

V0 = 0 V0 = 10 V0 = ∞

r0 (aB ) α ε α ε α ε

H− 0.4 0.5007 9.9141 0.3392 54.3573

0.6 0.5923 7.4652 0.2693 22.3108
0.8 0.4040 0.0000 0.6319 4.5285 0.2620 11.4966
1.2 0.4271 −0:0199 0.6567 1.7001 0.3163 4.1492
1.6 0.4550 −0:1160 0.6583 0.6222 0.3272 1.7765
2.0 0.4703 −0:2373 0.6507 0.1330 0.3389 0.7677
4.0 0.5040 −0:4433 0.5862 −0:4198 0.4054 −0:3295
6.0 0.5676 −0:4655 0.5725 −0:4650 0.4746 −0:4406
He† 0.5 0.5199 −0:2412 0.6218 4.6722 0.7465 22.9229
1.0 0.6081 −2:0522 0.6918 −1:4621 0.8320 1.0626
1.5 0.6592 −2:5086 0.7329 −2:4201 0.9330 −1:8456
2.0 0.8438 −2:6184 0.8465 −2:6179 1.0435 −2:5285
3.0 1.0746 −2:7579 1.0476 −2:7579 1.2428 −2:7935
4.0 1.2264 −2:8054 1.2264 −2:8054 1.3701 −2:8302
6.0 1.3855 −2:8341 1.3855 −2:8341 1.4924 −2:8426
Li+ 0.4 0.5818 −2:1084 0.6338 1.6199 1.2102 27.8079 27.6302
0.6 0.6346 −4:8914 0.6819 −3:7712 1.2979 3.9773 3.9284
0.8 0.6661 −5:8925 0.7126 −5:4793 1.3962 −2:8064 −2:8612
1.2 0.8201 −6:1460 0.8363 −6:1414 1.6171 −6:3075 −6:4047
1.6 1.0854 −6:4588 1.0854 −6:4588 1.8383 −6:9780 −7:0869
2.0 1.2347 −6:6033 1.2296 −6:6008 2.0190 −7:1370 −7:2356
4.0 1.9591 −7:1300 1.9591 −7:1300 2.3967 −7:2154 −7:2783
6.0 2.2081 −7:1979 2.2081 −7:1979 2.5037 −7:2201 −7:2784
Be2+ 0.4 0.6374 −8:1952 0.6669 −6:3628 1.7401 13.7130 13.6356
0.6 0.6835 −10:9415 0.7128 −10:4348 1.9233 −6:1782 −6:2402
0.8 0.7324 −11:1498 0.7623 −11:0389 2.1300 −11:1522 −11:2658
1.0 0.9958 −11:1894 0.9958 −11:1894 2.5486 −13:2248 −13:3701
2.0 1.4081 −11:5445 1.3810 −11:4901 3.0595 −13:5616 −13:6493
4.0 2.6901 −13:4399 2.6901 −13:4399 3.4089 −13:5931 −13:6539
6.0 3.0308 −13:5595 3.0308 −13:5595 3.5087 −13:5959 −13:6539

*Energies given in Hartrees. Values in the last column for Li+ and Be2+ correspond to self-consistent field calculations by Ludeña and Gregori [184].
†For V0 = ∞ the energy values are taken from [253].

Hence, we believe that a more flexible trial wave function in our sorption edge associated with transitions involving impurities at
treatment of two-electron confined atoms will improve the quality the center and a peak related to impurities at the edge of the dot.
of the predictions. For all sizes of the quantum dot, the peak associated with impu-
rities located next to the edge always governs the total absorption
probability.
4.10. Optical Properties of Impurity States in Oliveira and Perez-Alvarez [208], Porras-Montenegro and
Semiconductor Quantum Dots Oliveira [209], and Porras-Montenegro et al. [210] studied the op-
The optical absorption spectra associated with transitions be- tical absorption spectra associated with shallow donor impurities
tween the n = 1 valence level and the donor impurity band for both finite and infinite barrier GaAs–GaAlAs quantum wells
was calculated for spherical GaAs quantum dots with infinite po- and quantum-well wires. The main features were an absorption
tential confinement, using a variational procedure within the ef- edge associated with transitions involving impurities at the cen-
fective mass approximation, following Silva-Valencia and Porras- ter and a peak related to impurities at the edge of the wire. For
Montenegro [207]. We show results either for one impurity and quantum-well wires, the situation of the bulk material is reached
for a homogeneous distribution of impurities inside of the quan- for radii of 1000 Å and the peak associated with transitions involv-
tum dot. The interaction between the impurities was neglected. ing impurities at the center decreases as the radius of the structure
The main features found in the theoretical spectra were an ab- is diminished.
 OPTICAL PROPERTIES IN NANOSTRUCTURES 93

Porras-Montenegro et al. [211, 212] calculated the impurity

binding energies as functions of the radius, and impurity posi-
tion as well as the density of impurity states in spherical GaAs–
Ga1−x Alx As quantum dots, finding two structures associated to
impurities located at the center and at the edge of the quantum
dots, which are expected to show up in absorption and photo-
luminescence spectra associated with shallow hydrogenic impuri-
ties in quantum dots, as was the case in detailed calculations of
the impurity-related optical absorption spectra in GaAs–GaAlAs
quantum wells [208] and quantum-well wires [210, 213].
Helm et al. [214] performed a far-infrared absorption study
in lightly doped GaAs–Ga1−x Alx As superlattices, found that at
low temperatures, the absorption spectra are dominated by donor
transitions. They studied the 1s − 2pz donor transitions experi-
mentally and theoretically, obtaining excellent agreement.
Otherwise experimental progress through spectroscopic tech-
niques made possible a detailed analysis of the effects of con-
Fig. 96. Schematic of some possible absorption transitions in an infinite
finement on shallow impurities in quantum wells. Far-infrared
GaAs–(Ga,Al)As quantum dot of radius r0 = 300 Å with a donor impurity
magnetospectroscopy measurements on shallow donor impuri- band. The density of state, ρ(ε) from the positional dependent donor bind-
ties in GaAs–Ga1−x Alx As multiple quantum-well structures were ing energy is shown schematically on the left-hand side. The dependence
performed by Jarosik et al. [215], who assigned structures in of the binding energy as a function of the donor impurity position is shown
the transmission spectra to intraimpurity 1s − 2p± , transitions. on the right-hand side.
Work on magnetic field effects on shallow impurities in GaAs–
Ga1−x Alx As multiple quantum-well structures was also reported
by Yoo et al. [216].
For transitions from the first valence level to a donor impurity
level, we have for the initial and final states,
4.10.1. Model and Theory sin(k10 r)
ψi10 = ui (r) (567)
The Hamiltonian of a shallow hydrogenic impurity in a spherical (2πr0 )1/2 r
sin(k10 r)   
quantum dot of GaAs can be written in the effective mass approx- f
ψ10 = N(R; λ) exp −λ|r − R| uf (r (568)
imation as r
2 e2
b = p̂ −
H + Vb (r) (563) where ui (r), uf (r) are the periodic parts of the Bloch states for
2me ∗ κ|r − R| the initial and final states.
Taking the energy origin at the first conduction subband as de-
where Vb (r) is the confining potential which is zero for r < r0 and
picted in Figure 96, we have for the energy of the initial (first va-
infinite for r > r0 , r0 being the radius of the dot. The impurity
lence level) state,
position is denoted by R.
εi = −Eg (569)
The eigenfunction of the Hamiltonian in the absence of the im-
purity for the ground state (n = 1 and l = 0) and for the infinite where Eg is given by
potential well is [211],
h̄2 k210
h̄2 k210
sin(k10 r) Eg = εg + ∗ + (570)
ψ(r) = (564) 2me 2m∗h
(2πr0 )1/2 r
with εg being the bulk GaAs bandgap.
where r is (r; θ; ϕ). To satisfy the boundary conditions ψ(r = r0 ) =
The energy of the final state is
0, the eigenenergies corresponding to Eqs. (563) and (564) are
ε10 = h̄2 k210 /2m∗e with k10 = π/r0 . εf = −εb (r0 ; R) (571)
With the inclusion of the impurity potential, one should use a
variational approach to determine the ground state binding en- The transition probability per unit time for valence-to-donor tran-
ergy. The trial wave function considered is sitions associated with a donor impurity located at R is propor-
tional to the square of the matrix element of the electron-photon
( bel between the wave functions of the initial state (first
N(R; λ) sin(kr 10 r) exp(−λ|r − R|) r ≤ r0 interaction H
ψ(r) = (565) valence level) and the final (impurity) states, i.e.,
0 r > r0
2π XD b E2
with λ being a variational parameter and N(R; λ) being the nor- W =  f |Hel |i  δ(εf − εi − h̄ω) (572)
h̄
malization factor. The binding energy of the impurity is given by
  with Hbel = Cer · p̂, where er is the polarization vector in the di-
εb (r0 ; R) = ε10 − hT i + hV i (566) rection of the electric field of the radiation, p̂ is the momentum
operator, and C is a prefactor which contains the photon vector
where hT i and hV i are kinetic and potential energies. potential. Following the effective mass approximation, the forego-
94 RIERA ET AL.

Fig. 97. Donor binding energy as a function of the impurity position for
infinite GaAs–(Ga,Al)As quantum dots with distinct dot radii. From top to
bottom, r0 = 50, 100, 200, 1000 Å.

Fig. 99. Total optical absorption spectra (in units of W0 ), as a function of

the h̄ω − εg , for valence to donor transitions in infinite GaAs–(Ga,Al)As
quantum dots of different radii. r0 = 3000 Å (a), and r0 = 500 Å. (b) for a
quantum dot with a homogeneous distribution of impurities, where there
is no interaction between them.

ing matrix element may be written as [217],

D E
be−l |i ∼
f |H = Cer · Pfi Sfi (573)

with Z
1
Pfi = dru∗f (r)p̂ui (r) (574)
 
and Z
Sfi = drff∗ (r)fi (r) (575)

where  is the volume of the unit cell and ff (fi ) is the envelope
function for the final (initial) state. For the case of the donor im-
purity we have for Sfi = Sfi (R; λ),
!1/2 Z r0
2π
Sfi = N(R; λ) dr sin2 (k10 r)
r0 0
Z π h i
× dθ sin θ exp −λ|r − R| (576)
0
Fig. 98. Optical absorption spectra (in units of W0 ), as a function of the For an infinite GaAs quantum dot of radius r0 with one impu-
h̄ω−εg , for valence to donor transitions in infinite GaAs–(Ga,Al)As quan- rity inside, the transition probability per unit time for valence to
tum dots of different radii. r0 = 3000 Å (a), r0 = 1000 Å (full curve), and donor transitions is given by
r0 = 500 Å (dashed curve) (b). ε1 and ε01 (ε2 and ε02 ) correspond, respec- !
tively, to the onset of transitions from the first valence subband to the lower π h̄2
W(ω; R) = W0 S 2 Y(1) (577)
(upper) edge of the impurity band. 2m0 a2B fi
 OPTICAL PROPERTIES IN NANOSTRUCTURES 95

where aB is the Bohr radius and Y(1) is the step function. In this εmax
i Van Hove-like singularity is clearly a peak associated to on-
expression, we have for 1 and W0 , center donors which corresponds to the bulk limit. For quantum
dots of r0 > 1000 Å, it is observed that the peak related to on-edge
1 = h̄ω − Eg + εg (r0 ; R) (578)
donors is still significant. A situation which is very well reflected in
4m0 2 our results. On the other hand, a remarkable difference in the total
W0 = aB |C|2 |er · Pfi |2 (579)
h̄3 absorption probability, is the absence of the peak structure asso-
For a homogeneous distribution of impurities and assuming ciated with impurities located at well for quantum dots with radii
that the quantum dot radius is much larger than the lattice param- less than 500 Å, with quantum wells, and with quantum-well wires.
eter, one has for the total transition probability per unit of time,
Z
3 r0 5. NONCONVENTIONAL AND IDEALIZED
WT (ω) = dR W(ω; R)R2 (580)
r03 0 CONFINED SYSTEMS AND GEOMETRICAL
EFFECTS IN NANOSTRUCTURED MATERIALS
For the numerical computing, we used κ = 12:58, m∗e =
0:0665m0 , m∗h = 0:30m0 , where εg = 1:424 eV. 5.1. Ground State Energy of the Two-dimensional
A schematic of a GaAs quantum dot doped with a homoge- Hydrogen Atom Confined within Conical Curves
neous distribution of the donor impurities is shown in Figure 96.
The edges for optical absorption from the first valence subband to In this subsection, the variational method is used to calculate the
the donor impurity band is represented by h̄ω1 , and to the first ground state energy of the two-dimensional hydrogen atom con-
conduction subband is represented by Eg . The transition h̄ω2 cor- fined by impenetrable conical curves [218]. The confinement of the
responds to absorption to an impurity level associated with donors atom in a region limited by two intersecting parabolas as well as
at the edges of the quantum dot. At the right, we also show the within a cone, was also considered. The results show the transition
density of the states for the impurity band. of this system to a quasi-one-dimensional hydrogen atom when
In Figure 97, we display the donor binding energy as a func- the curves are open. In the case of closed curves, a transition to
tion of the donor position inside the quantum dot for an infinite a quasi-zero-dimensional hydrogen atom is observed. These lim-
potential well with different radii. The donor binding energy de- iting cases, as well as other intermediate situations, are discussed
creases as the donor position increases reaching a minimum as the within this subsection.
donor position is equal to the radii of the quantum dot. The ab- In particular, the ground state energy of these kinds of systems
sorption probability W(ω; R) for an infinite quantum dot with one is calculated for different confining situations. Moreover, the be-
single impurity as a function of h̄ω − εg is shown in Figure 98. havior of the ground state energy is analyzed, as the confining
In Figure 98(a), we present the absorption probability for an infi- region reaches an extreme value, i.e., when the system becomes
nite GaAs quantum dot of radius r0 = 3000 Å. We observe that quasi-one-dimensional (open curves) or quasi-zero-dimensional
there is a noticeable peak structure associated with a single impu- (closed curves) as well as other intermediate situations.
rity located at the center of the dot, which is much larger than the The results show a very interesting behavior of this property
structure associated with a single impurity next to the edge of the depending upon the confinement region as well as depending upon
dot, meaning that we essentially reached the bulk limit. Our results the closed or open character of the curves.
for an r0 = 1000 Å and r0 = 500 Å quantum dot are shown in Fig- The latter would represent an interesting research topic con-
ure 98(b); the structure associated with a single impurity located at cerning the study of other properties of this system such as its po-
the center of the dot is smaller than the structure associated with larizability, pressure due to the boundaries, dipole moment, etc.
a single impurity at the edge of the dot.
In Figure 99, we display the total absorption probability for an 5.1.1. The Schrödinger Equation
infinite GaAs quantum dot with a homogeneous distribution of
impurities. For a quantum dot of r0 = 3000 Å, the total absorption In atomic units (h̄ = e = m∗e = 1), the Schrödinger equation for
probability as a function of h̄ω − εg is shown in Figure 99(a). An the two-dimensional confined hydrogen atom in orthogonal curvi-
absorption edge associated with transitions involving impurities at linear coordinates can be written as
 ! !
the center of the well and a peak related with impurities next to
1 1  ∂ h2 ∂ ∂ h2 ∂
edge of the dot are observed. The peak associated with impurities −  + 
located next to the edge of the dot is much larger than the peak as- 2 h1 h2  ∂q1 h1 ∂q1 ∂q2 h1 ∂q2
sociated with impurities located next to center of the dot. This be- 

havior is quite different from that found for GaAs quantum wells
+ Ven (q1 ; q2 ) + Vb (q1 ; q2 ) ψ(q1 ; q2 ) = εψ(q1 ; q2 ) (581)
and quantum-well wires of the comparable dimensions. This is a 
consequence of the quantum confinement and the homogeneous
distribution of the impurities in the quantum dot. where (q1 ; q2 ) is the orthogonal curvilinear coordinate system in
When the radius of the quantum dot decreases, we observe that the plane and {hi } are scale factors, given by
the peak associated with impurities located next to the center di- !2 !2
minishes [see Fig. 99(b)], which may be understood by means of ∂x ∂y
h2i = + (582)
the behavior of the density of the states as a function of the bind- ∂qi ∂qi
ing energy (see Fig. (3) in Ref. [212]). In this figure, it is seen that
for small radii the density of the states is a smooth function which Ven (q1 ; q2 ) is the Coulombic interaction,
exhibits a well-defined peak at the binding energy associated to 1
on-edge donors. When the radius is very large, the structure of the Ven (q1 ; q2 ) = − (583)
r(q1 ; q2 )
96 RIERA ET AL.

while r(q1 ; q2 ) is the electron-nucleus separation, in this set of co- where (q1 ; q2 ) is the system of coordinates compatible with the
ordinates. The confining potential that it will be assumed is of the symmetry of the confining boundary, r (q1 ; q2 ) is the electron-
form,  nucleus distance in these coordinates, f (qi ; q0 ) is a geometry-
+∞ (q1 ; q2 ) ∈/D adapted auxiliary function such that f = 0 at qi = q0 ; qi is the
Vb (q1 ; q2 ) = (584)
0 (q1 ; q2 ) ∈ D coordinate associated with the geometry of the boundary, and α is
where D is a given finite (or infinite) confining domain of the a variational parameter.
plane. The solution of Eq. (581), must satisfy The geometry of the confining boundary allows choosing f in a
quite simple way, namely, similar to its contours. To exemplify the
ψ(q1 ; q2 ) = 0 q ∈ ∂D (585) latter, if the assumed confining boundary is a circle of radius r0 ,
where ∂D is the boundary of D. then f = r0 − r, i.e., this function map all circles of radius 0 ≤ r ≤
The model Hamiltonian for the confined quantum system un- r0 , that is, the allowed region for the atomic electron. Similarly,
der study, can be written, in atomic units, as if the confining boundary is an ellipse of “size” ξ0 (eccentricity
= 1/ξ0 ), then f = ξ0 − ξ and the function would map all ellipses
b = − 1 ∇q2 + Ven (q) + Vb (q)
H (586) of size consistent with 1 ≤ ξ ≤ ξ0 , thus generating the allowed
2
space for the atomic electron in the plane. The definition of f can
where Ven (q) and Vb (q) are given by Eqs. (583) and (584), and ∇q2 be done for other confining symmetries, accordingly.
is derived from Eq. (581). A description on the flexibility of so-constructed trial wave
At this point, the variational method can be implemented to functions, to deal with a variety of confining situations, can be
calculate the ground state energy of the system by constructing the found in Refs. [143, 144, 156, 188, 198]. In particular, in Refs.
functional, [156, 188], other physical properties for one- and two-electron con-
Z n o fined atoms, are calculated to test the goodness of the wave func-
ε(α) = χ∗i − 12 ∇q2 + Ven (q) χ i d 2 q (587)
D tions. The results show good agreement as compared with exact or
more elaborated approximated methods.
and minimizing it with respect to α, restricted to
Z Hence, once that coordinate system (q1 ; q2 ) is chosen, all nec-
χ∗i χi d 2 q = 1 (588) essary elements for constructing the trial wave function are defined
D and the procedure outlined in the previous subsection, can imme-
diately be used to calculate and minimize the energy functional.
as usual. The area element is given as d 2 q = h1 h2 dq1 dq2 .
Of course, the transformation equations defining the chosen coor-
At this extent, the method described before can be used to
dinate system are assumed to be known, indeed, it is usually the
study a confined system for which the interactions with the sur-
case and they can be found elsewhere.
rounding medium are not considered, i.e., enclosed within a po-
It is worth mentioning that the binding energy, defined as the
tential barrier of infinite depth. The next subsection is restricted
absolute value of the difference between the energies of the elec-
to deal with the case of a typical one-electron confined system in
tron with and without the Coulombic interaction, has only mean-
two dimensions, namely, the case of the hydrogen atom confined
ing in the case of closed curves. In the case of open curves, the
within a given region of the plane with different geometries.
electron energy belongs to the continuum spectrum and the bind-
ing energy is meaningless. Moreover, in the former situation, the
5.1.2. Variational Wave Functions for Different Geometries binding energy would behave as the total energy, that is, as the
As is well known, the Schrödinger equation for the unconfined confining region becomes smaller either grows without limit, due
two-dimensional hydrogen atom is separable in polar coordinates to the closed and impenetrable property of the confining bound-
(r; θ); its ground state wave function can be written as ary.
Figures 100(a), 101(a), 102(a), 103(a), 104(a), and 105(a), dis-
φ(r; θ) = C0 exp(−k0 r) (589) play the confining geometries considered in this subsection, while
where C0 is apnormalization constant, r is the electron-nucleus dis- Figures 100(b), 101(b), 102(b), 103(b), 104(b), and 105(b), show
tance, k0 = −2ε0 , ε0 = −2 Hartrees is its ground state energy, the behavior of their corresponding ground state energy as a func-
and the nucleus was assumed to be located at the origin. tion of confining parameter, respectively.
When the atom in the plane is restricted to a given open or The ground state energy of the two-dimensional hydrogen atom
closed domain D, φ, and ε0 must change to fit the new conditions confined within an ellipse, is displayed in Figure 100(b), as a func-
accordingly. If the boundary ∂D of D, is impenetrable for the elec- tion of parameter ξ0 (which defines the degree of confining), for
tron, then φ = 0 at ∂D. The latter means that due to confinement, a = 0:5, 1.0, and 2:0 Bohrs, respectively. Notice that for a given ξ0
new quantization rules must be found, that is, the old good quan- different of 1 and ∞, when a decreases, the major and minor semi-
tum numbers used to define the energy of the unconfined system axis also decreases, i.e., the region of confinement is smaller and
are no longer useful to characterize the energy of the now confined the energy increases. If ξ0 → 1, the two-dimensional hydrogen
system. Moreover, as the region of confinement was assumed arbi- atom becomes a quasi-zero-dimensional system and the ground
trary (in shape and size), the energy of a given state would depend state energy goes to ∞ (see Fig. 100(a)). In the case that ξ0 → ∞,
on a continuous parameter associated with the size (or shape) of the atom is allowed to occupy the whole plane for any value of a,
the domain D (or its boundary ∂D). thus recovering the case of the unconfined two-dimensional hydro-
In the case of the confined two-dimensional hydrogen atom, gen atom, as expected.
the approximate (variational) ground state energy would involve In Figure 101(b), the energy for the ground state of the two-
the use of a trial wave function given by dimensional hydrogen atom confined in a region of the plane lim-
h i ited by a hyperbola (see Fig. 101(a)), is displayed as a function
χ(q1 ; q2 ) = Af (qi ; q0 ) exp −αr (q1 ; q2 ) (590) of η0 (which defines the degree of confining), for a = 0:5, 1.0,
 OPTICAL PROPERTIES IN NANOSTRUCTURES 97

Fig. 101. (a) Region of hyperbolic confinement. Limiting case: Q-one-

dimentional. (b) Ground state energy of the two-dimensional hydrogen
atom confined within a region of the plane limited by a hyperbola, as a
function of η0 , for three values of the semidistance between foci a = 0:5,
1.0, and 2:0 Bohr, respectively.
Fig. 100. (a) Region of elliptic confinement. Limiting case: Q-one-
dimensional. (b) Ground state energy of the two-dimensional hydrogen
atom confined within an ellipse as a function of the parameter ξ0 , for three for a three-dimensional hydrogen atom limited by paraboloidal or
values of the semidistance between foci a = 0:5, 1.0, and 2:0 Bohr, respec- hyperboloidal surfaces.
tively. Figure 102(b) shows the energy for the ground state of the two-
dimensional hydrogen atom confined in a region of the plane lim-
ited by a parabola (see Fig. 102(a)) as a function of ξ0 (which
and 1:5 Bohrs, respectively. As can be observed, for η0 = 0, the defines the degree of confining). Notice that as ξ0 → ∞ the
hyperbola becomes degenerate and their two branches coincide parabola opens and the confinement region reduces until the
with the y-axis, so that the confinement region becomes a half whole plane is available, thus occurring a transition from a con-
plane for any value of a different from zero and infinite. When fined two-dimensional system to an unconfined two-dimensional
η0 and a → 0, the coalescence of the two branches of the hyper- system, (see Fig. 102(a)), whose energy is −2 Hartrees. In the
bola (η0 → 0− and η0 → 0+ ) leads to a quasi-two-dimensional same way, if ξ0 → 0 the parabola closes and the forbidden region
hydrogen atom with its nucleus at the y-axis, the ground state en- increases, confining the system in the −y-axis, thus realizing the
ergy approaches −2/9 Hartree. When a → ∞, the whole plane transition from a confined two-dimensional system to a confined
becomes available for the atom, thus leading to the unconfined quasi-one-dimensional system, whose energy approaches zero (see
two-dimensional hydrogen atom (see Fig. 101(a)). Another obser- Fig. 102(b)).
vation is that, for η0 → −1 the whole plane is forbidden, (x-axis In Figure 103(b), the ground state energy of the two-dimension-
is excluded), thus resulting in the one-dimensional hydrogen atom al hydrogen atom confined in a circular region of the plane is de-
with zero energy (see Fig. 101(b), which physically means that all picted as a function of r0 (radius of the circle in atomic units).
excited states belong to the continuous spectrum or belong to the As can be observed for r0 → ∞, the allowed space approaches
ionized atom. A similar behavior was found in Refs. [219, 220]) the whole plane, i.e., the system becomes the unconfined two-
98 RIERA ET AL.

Fig. 102. (a) Region of parabolic confinement. Limiting case: Q-one-

dimensional. (b) Ground state energy of the two-dimensional hydrogen
atom confined within a region of the plane limited by a parabola, as a func-
tion of ξ0 .

dimensional hydrogen atom whose ground state energy is −2

Fig. 103. (a) Region of circular confinement. Limiting case: Q-zero-
Hartrees. Another observation is that, when r0 → 0, the for- dimensional. (b) Ground state energy of the two-dimensional hydrogen
bidden circular region becomes a point and a transition of the atom confined within a region of the plane limited by a circle, as a function
two-dimensional system to a quasi-zero-dimensional system (see of its radius r0 .
Fig. 103(a)) with energy ε → ∞ occurs.
Figure 104(b) displays the ground state energy of the two-
dimensional hydrogen atom confined in a region of the plane lim- Figure 105(b) shows the ground state energy of the two-
ited to a cone obtained when θ = θ0 and 0 ≤ r < ∞, as a func- dimensional hydrogen atom confined in a region of the plane given
tion of θ0 (polar angle in radians). In the limit when θ0 → 0 the by the intersection of two symmetrical confocal parabolas (see
cone reduces, then occurring the transition of the confined two- Fig. 105(a)), as a function of ξ0 (root of the distance to the foci, in
dimensional system to the quasi-one-dimensional system with en- atomic units). As can be observed, when ξ0 → 0 the two parabolas
ergy ε → 0, as in the previous situations. When θ0 increases to (and their intersections) are closer to the origin. The allowed re-
2π, the allowed region grows and the system behaves as a quasi- gion is almost a point, the transition of confined two-dimensional
two-dimensional system, the edges of the cone are still a forbidden system to a quasi-zero-dimensional one occurs, accordingly, the
region (see Fig. 104(a)). The latter case corresponds to the mini- energy becomes infinite, as expected. Moreover, in the case when
mum of the curve at 6:2856 rad (2π) where the energy is ≈ −2/9 ξ0 → ∞, the available space enlarges until an unconfined two-
Hartree and coincides with the case of hyperbolic confinement in dimensional atom is recovered, i.e., the energy approaches −2
the limit η0 → 1, a → 0 (see Figs. 101(a) and 104(a)), as expected. Hartrees (see Fig. 105(b)).
 OPTICAL PROPERTIES IN NANOSTRUCTURES 99

Fig. 104. (a) Region of conical confinement. Limiting case: Q-two-

dimensional. (b) Ground state energy of the two-dimensional hydrogen
atom confined within a region of the plane limited by a cone, as a function
of the opening angle θ0 .
Fig. 105. (a) Region of confinement given by the intersection of two sym-
metrical parabolas. Limiting case: Q-zero-dimensional. (b) Ground state
energy of the two-dimensional hydrogen atom confined within a region of
The overall results show that the ground state energy of the the plane limited by the intersection of two symmetrical confocal parabo-
two-dimensional hydrogen atom confined by an impenetrable po- las, as a function of ξ0 .
tential has two extreme limiting cases, namely: (a) In the case of
closed conical curves, a transition to a quasi-zero-dimensional sys-
tem where the ground state energy becomes infinite, that is, the
than the rest mass energy of the electron, the treatment of the
confinement region approaches to a point; (b) In the case of open
atom should be done in a relativistic scheme which renders a large,
conical curves, a transition to a quasi-one-dimensional system oc-
but finite, binding energy [227].
curs in which the atom is confined to a straight line with origin
at the nucleus and whose ground state energy becomes zero. The In summary, the confining geometries studied in this subsection
case (b) is closely related to the one-dimensional hydrogen atom lead us to conclude following extreme behavior:
discussed in Refs. [221–226], except that the binding energy is Circular: r0 → 0, ε → ∞ (quasi-zero-dimensional).
found to be infinite in Ref. [225]. The main difference between Elliptic: ξ0 → 1, ε → ∞ (quasi-zero-dimensional).
the limiting case (b) and the so-called Loudon’s one-dimensional Hyperbolic:
atom [225] rests in the fact that Loudon did not take into ac- 
count the energy due to the confinement because his model is one-  η0 → 0


ε = ε(a); (a 6= 0) (quasi-two-dimensional)
dimensional and our limit case is quasi-one-dimensional. A discus- η0 → 1 ε = ε(a) (quasi-two-dimensional)
sion of this matter is made in the next subsection. Moreover, since 
 (η0 ; a) → 0 ε → −2/9 (quasi-two-dimensional)

the binding energy of Loudon’s one-dimensional atoms is greater η0 → −1 ε → 0 (quasi-one-dimensional)
100 RIERA ET AL.

Parabolic (one): η0 (ξ0 ) → 0, ε → 0 where εx = ε − εy and

(quasi-one-dimensional). Z y0 " #
Parabolic (two): η0 (ξ0 ) → 0, ε → ∞ 2 −2
Veff (x) = N q cos2 (k0 y) dy (600)
(quasi-zero-dimensional). −y0 x +y
2 2

θ0 → 0 ε → 0 (quasi-one-dimensional) A similar equation was found in Ref. [228] for a hydrogenic impu-
Cone:
θ0 → 2π ε ≈ −2/9 (quasi-two-dimensional) rity confined within a quantum-well wire.
The result of averaging the two-dimensional Coulombic attrac-
5.1.3. One-dimensional Hydrogen Atom, as a tion over the fast y-motion is such that, Eq. (599) can be inter-
Limiting Situation preted as the Schrödinger equation for the one-dimensional atom
under the potential Veff (x).
The behavior of the ground state energy of the confined two- Moreover, a careful analysis of Veff (x) in the limit y0 → 0
dimensional hydrogen atom, when one dimension becomes in- reveals that this potential has the same behavior as the one-
finitely small is analyzed in this subsection. dimensional Coulomb potential, as was also pointed out by Jan
If the nucleus of the atom is located at the origin and the con- and Lee [228] in the case of a three-dimensional hydrogen atom
fining region is a symmetrical narrow strip of width 2y0 , the Hamil- confined within a cylinder whose radius approaches zero. The en-
tonian for the system is (in atomic units), ergy εx in this limiting situation becomes the same as that ob-
2 tained by Loudon [225], that is, −∞. It can also be noted that when
b = −1 ∂ − ∂2 1
H 2
1
2 + Vb (y) − q (591) y0 → 0, εy → ∞ and thus the total energy for the confined one-
∂x2 ∂y 2
x2 + y 2 dimensional atom ε = εx + εy → 0. The latter fact supports the
results obtained in cases when the open confining curves approach
where Vb (y) is the confining potential, defined as a straight line.
n
Vb (y) = 0 −y0 ≤ y ≤ y0 (592)
∞ otherwise 5.2. Geometrical Effects on the Ground State Energy of
The corresponding Schrödinger equation can be written as Hydrogenic Impurities in Quantum Dots
( )
∂2 1 ∂2 1 The effect of nonsphericity of quantum dots on the ground state
−2
1 − +Vb (y)− q ψ(x; y) = εψ(x; y) (593) energy of hydrogenic impurities is studied in the frame of the ef-
∂x2 2 ∂y 2 x2 + y 2 fective mass approximation and the variational method [229]. The
difference in composition of the quantum dot and the host mate-
where ε is the total energy and the wave function ψ(x; y) must
rial is modeled with a potential barrier at the boundary of the dot.
satisfy the boundary condition,
To make the analysis, two symmetries are considered for the quan-
ψ(x; ±y0 ) = 0 (594) tum dot: spherical and spheroidal. In this way, the ground state
energy is calculated as a function of the volume of the quantum
As pointed out by Jan and Lee [228], as the strip becomes very dot, for different barrier heights. The results show that the ground
narrow, i.e., y0 is very small, the fast motion on the y-direction state is strongly influenced by the geometry of the dot, that is, for
allows the wave function to be written, approximately, as a given volume and barrier height, the energy is clearly different
ψ(x; y) ≈ χ(y)8(x) (595) if the dot is spherical or spheroidal in shape when the volume of
the dot is small (strong confinement regime). As the volume of the
Since ψ(x; y) must satisfy the boundary condition (594), then dot increases (weak confinement regime), the geometry becomes
irrelevant, as expected.
χ(y) ≈ cos(k0 y) (596)
The confinement of excitons in quantum dots and other mi-
with k0 = π/2y0 . crostructures such as quantum wells, quantum-well wires, etc. was
Introducing Eq. (595) into Eq. (593) and multiplying the result corroborated experimentally in past years [190, 230–233]. The
by cos(k0 y), the integration over y from −y0 to y0 yields main feature of this effect corresponds to a frequency shift of the
  first excitonic peak to higher energy, compared to the mean bulk
 d2 Z y0 
−2 electronic peak. There have been many theoretical efforts to quan-
− + N2 q cos2 (k0 y) dy 8(x) titatively explain this quantum mechanical effect [169, 234–240].
 dx2 −y0 x2 + y 2 
However, in all these works, the dot is assumed to be of spherical
= (ε − εy )8(x) (597) shape. This last assumption is not strictly true, since the electron
microscopy studies of these systems show microparticles whose
where shapes vary from quasi-spherical to pyramidal [190]. The aim of
"Z #−1
y0 this work is to quantitatively analyze the dependence of the ground
N2 = cos2 (k0 y) dy = y0−1 (598) state energy of a hydrogenic impurity on the geometry of the quan-
−y0
tum dot. The latter would be of interest since it might constitute a
is a normalization constant and εy = k20 = π 2 /4y02 is the kinetic qualitative study of the behavior of excitons confined within these
energy of the fast motion in the y-direction due to confinement. microstructures. We assume that the impurity is confined within a
Equation (597) can be rewritten as potential barrier of finite depth, which emulates, on the average,
" # the surrounding medium in which the dot is embedded. The cal-
d2 culation of the impurity ground state energy is performed using a
− + Veff (x) 8(x) = εx 8(x) (599)
dx2 variational approach, which previously was shown to be useful in
 OPTICAL PROPERTIES IN NANOSTRUCTURES 101

dealing with these types of confined systems in a fairly well fashion and
[144, 156, 188, 198]. The model Hamiltonian for the impurity is as- 1
V(q) = − (610)
sumed to be valid within the effective mass approximation and we rq
considered, for the sake of comparison, spherical and spheroidal Here, rq is the electron-nucleus separation and we use effective
dots of the same volume and confining barrier. In this way, the atomic units. The symmetry of the confining domain determines
ground state energy of the impurity is calculated as a function of the form of rq in the set of coordinates compatible with it. In the
the volume and the confining barrier heights. following, we explicitly construct the trial wave functions for the
chosen symmetries.
5.2.1. Brief Description of the Method First, if we assume that a hole is located at the center of a con-
fining sphere of radius r0 , the trial wave function in spherical co-
If {qi } denotes the orthogonal set of coordinates compatible with ordinates can be written as [144],
the symmetry of the confining boundary, the trial wave functions
can then be written as χi = A exp(−αr)(r0 − αr) r < r0 (611)

χi (q; α) = Af(q; q0 ; α)φ0i (q; α) q < q0 (601) and

χo = B exp(−βr)/r r > r0 (612)
and
χ0 (q; β) = Bφ(q; β) q > q0 (602) The continuity of the logarithmic derivative at r = r0 , given in
Eq. (603), allows one relate α and β by
where α and β are parameters to be determined, q0 is associated
with the size and symmetry of the confining boundary, A and B are αr0 (1 − α) + (2α − 1)
β= (613)
normalization constants, φ0i is the free system wave function, φ is r0 (1 − α)
a function with the proper asymptotic behavior as q0 → ∞ and f in such a way that only a variational parameter needs to be deter-
is an auxiliary function that allows the condition, mined, once that the energy functional is minimized in the stan-
1 ∂χi 1 ∂χ0 dard way.
= (603) If we now consider the exciton confined within the prolate
χi ∂q χ0 ∂q
spheroid,
at q = q0 to be satisfied.
{ξ = ξ0 ; −1 ≤ η ≤ 1; 0 ≤ ϕ ≤ 2π} (614)
The model Hamiltonian for the confined quantum system un-
der study, in the effective mass approximation, can be written, in with the nucleus located at a foci, the electron-nucleus separation
atomic units, as can then be written as
b = − 1 ∇ 2 + V(q) + Vb (q)
H (604) r1 = R(ξ − η) (615)
2 q
where V is the potential that accounts for the internal interactions where 2R is the interfocal distance of the spheroid. Following the
of the components of the system and Vb is the confining potential. solutions of Coulson and Robinson [199] for the free hydrogen
Here we assume that Vb has the form atom in this coordinate system, the trial wave function is of the
 form,
0 q < q0 h i
Vb = (605) χi = A exp −α(ξ − η) (ξ − αξ0 ) ξ < ξ0 (616)
V0 q > q0
At this point, the variational method can be implemented to cal- and
culate the energy of the ground state of the system by construing χo = B exp(−βξ − αη) ξ > ξ0 (617)
the functional,
Z once again, the boundary condition given by Eq. (603), allows one
n o
ε(α; β) = χ∗i − 12 ∇q2 + V(q) χi d 3 q to connect α and β through
Z n o αξ0 (1 − α) + α
β= (618)
+ χ∗o − 12 ∇q2 + V0 + V(q) χo d 3 q (606) ξ0 (1 − α)
As the reader must be aware, this form for the trial wave func-
and minimizing it with respect to (α; β), restricted to
Z Z tion automatically satisfies the boundary condition on its logarith-
mic derivative with respect to η, at ξ = ξ0 . As in the previous case,
χ∗i χi d 3 q + χ∗o χo d 3 q = 1 (607)
only a variational parameter needs to be determined when we use
as usual. the variational approach.
We have to point out that the auxiliary function of the form
(q0 −αq), which is used to construct the trial wave function for two
5.2.2. Geometry-adapted Trial Wave Functions symmetries, is flexible enough to describe the situation in which
In the effective mass approximation, the model Hamiltonian for a V0 → ∞, in that case α → 1 and correspondingly β → ∞ (see
hydrogenic impurity (m∗n = ∞) confined within a quantum dot, in Eqs. (613) and (618)) so that the external wave function becomes
the context of the later subsection, can be written as vanishingly small in this limit, as expected.
In Figure 106(a) and (b), the energy for the ground state of
b = − 1 ∇ 2 + V(q) + Vb (q)
H (608) the hydrogenic impurity confined within spherical and prolate
2 q
spheroidal quantum dots is displayed as a function of the volume,
where 
0 q < q0 for V0 = 2:0 and 8:0 effective Rydbergs, respectively. In both fig-
Vb = (609) ures, we assumed R = 1 effective Bohr for the case of the prolate
V0 q > q0
102 RIERA ET AL.

Fig. 106. Ground state energy for a hydrogenic impurity within quantum
dots of a spherical and prolate spheroidal shape as a function of their vol-
ume for a fixed value of the barrier height. (a) V0 = 2:0R∗y , (b) V0 = 8:0R∗y . Fig. 107. Ground state energy for a hydrogenic impurity within quantum
dots of: (a) spherical symmetry as a function of the dot radius and barrier
height, (b) prolate spheroidal symmetry as a function of the eccentricity
and barrier height.
spheroidal dot. The effect of considering a different geometry can
be noticed immediately from these figures, which supports that,
for a given volume of the dot and barrier height, the ground state latter is due to the fact that confining capability of the box de-
energy shows a dependence on the shape of the confining box that creases and correspondingly, the probability of finding the electron
is particularly stronger for smaller sizes of the dot. The effect of in the exterior of the quantum dot is increased (see Fig. 106(a)). In
including the internal interaction potential (Coulomb term) of the Figure 106(b), this effect is smaller. In Figure 106(a), we can also
components of the system (electron-nucleus) in both regions (in- observe a stronger dependence of including the Coulomb term
side and outside the quantum dot), is also shown in the figures. in the exterior of the quantum dot, for the spheroidal symmetry
When the Coulomb term is considered in the exterior of the quan- than for the spherical one. Another observation is that, for the
tum dot, a remarkable difference can be noticed for the smaller spheroidal symmetry, we considered the nucleus of the impurity
barrier height potential and smaller sizes of the quantum dot. The placed in a focus not in the origin, but in the later case, keeping
 OPTICAL PROPERTIES IN NANOSTRUCTURES 103

Fig. 109. Potential energy curve for a harmonic oscillator between two
Fig. 108. Ground state energy for a hydrogenic impurity as a function of infinite potential walls. Only the z-dependence is drawn.
the interfocal distance and the eccentricity for a fixed value of the barrier
height.

5.3. A Harmonic Oscillator Confined within an Infinite

Potential Well
in mind the work of Marin et al. [143], the resulting geometrical The exact solution to the Schrödinger equation for a three-
effect would be stronger. dimensional harmonic oscillator confined by two impenetrable
In Figure 107(a), we show the energy for the ground state of walls is presented [241]. The energy levels of this system are ob-
the hydrogenic impurity confined in a spherical quantum dot as a tained as a function of wall separation as well as distance of the
function of the volume, for V0 = 8:0 and 2:0 effective Rydbergs. In center of the oscillator to the walls. The force exerted by the walls
the solid curves, the Coulomb term is not included in the external on the oscillator is also evaluated, showing a classical behavior.
region of the quantum dot while in the dashed ones we consider The harmonic oscillator is perhaps one of the simplest systems
it. Figure 107(b) shows the same curves as in Figure 107(a) for the that is extensively studied both classically as well as quantumly. At
prolate spheroidal quantum dot with R = 1:0 effective Bohr. A dif- the undergraduate level, the student learns that the quantum oscil-
ference between the curves with and without the Coulomb term in lator problem allows for exact solutions to the Schrödinger equa-
tion, providing us with a complete set of basis functions useful in
the exterior of the quantum dot can be noticed. This difference is
the treatment of a great variety of problems in modern physics
greater for the smaller barrier height potential and in the region of
[242]. In passing, we should mention that the study of bounded
strong confinement regime (small sizes of the quantum dot). This
quantum systems has become increasingly important, mainly to
effect is similar to that discussed earlier for Figure 106.
understand the behavior of real systems, such as atoms under high
In Figure 108, we make a comparison of the ground state
pressure, electrons trapped in vacancies of crystals, tunneling of
energy of the hydrogenic impurity confined within a prolate
electrons, and electron-hole pairs through multilayered crystalline
spheroidal quantum dot, for V0 = 2:0 effective Rydbergs and dif-
structures, etc.
ferent interfocal distances (2R). The solid curves correspond to
R = 0:5 and R = 1:0 effective Rydbergs without the Coulomb
term, while the dashed ones are for the same values of R but the 5.3.1. Separation of the Schrödinger Equation
Coulomb term in the exterior of the quantum dot is included. The
Let us assume a three-dimensional harmonic oscillator bounded
latter figure confirms that there is a significant variation of the en-
by two impenetrable walls, as depicted in Figure 109. The bound-
ergy for different interfocal distances of the spheroidal quantum
ary condition makes the problem in spherical coordinates difficult
dot and that the difference is greater for the lower value of Rand
to treat, as would be the case for the free-oscillator. Making use of
even more significant when the Coulomb term in the exterior of
cylindrical coordinates (ρ; θ; z), the Schrödinger equation may be
the quantum dot is not included.
written as
Thus, the previous results show that a quantitative fitting of the ( )
energy-size curves for the impurity (or, qualitatively, for the ex- h̄2 2
− ∇ + V(ρ; z) − ε ψ(ρ; θ; z) = 0 (619)
citon) to experimental results (for a given value of V0 ) must be 2M
taken with caution, independently of the model used to make the
comparison. This is the case in most of the theoretical approaches where

dealing with semiconductor crystallites, since a spherical shape is (K/2)[ρ2 + (z − a)2 ] 0 < z < d
assumed to fit the experimental results. V(ρ; z) = (620)
0 otherwise
104 RIERA ET AL.

|m|
where d is the distance between the walls and a is the position of where Ln (w) corresponds to an associated Laguerre polynomial.
the center of the oscillator well relative to one of the walls (see Returning to the original variable ρ, the normalized radial wave
Fig. 109). The Laplacian operator in Eq. (619) in cylindrical coor- function results as
dinates reads explicitly as h i1/2
|m|
! Rnm (ρ) = 2n!/(|m| + n)! exp(−ρ2 /2)ρ|m| Ln (ρ2 ) (634)
2 1 ∂ ∂ 1 ∂2 ∂2
∇ ≡ ρ + + (621)
ρ ∂ρ ∂ρ 2
ρ ∂θ 2 ∂z2
5.3.2. Solution of the z-equation
Setting M = h̄ = K = 1 and writing ψ(ρ; θ; z) as the product,
Due to the boundary conditions, it is clear that the solution of
ψ(ρ; θ; z) = R(ρ)F(θ)G(z) (622) Eq. (625) contains the relevant information on the effect of con-
finement on the energy
√ levels of the system. Let us define the aux-
after separating the variables, the following set of equations is ob-
iliary variable y = 2(z − a). Equation (625) may then be written
tained
  as  
R00 (ρ) + ρ−1 R0 (ρ) + 2ε1 − m2 /ρ2 − ρ2 R(ρ) = 0 (623) G00 (y) − y 2 /4 − ε2 G(y) = 0 (635)
F 00 (θ) + m2 F(θ) = 0 (624) This equation is satisfied by the parabolic cylinder function [243]:
h i
G00 (z) + 2ε2 − (z − a)2 G(z) = 0 (625) U(−ε2 ; y) and U(−ε2 ; −y). Therefore, the most general solution
will be
where ε1 , ε2 , and m are separation constants, with the require- G(y) = AU(−ε2 ; y) + BU(−ε2 ; −y) (636)
ment that the total energy ε be given as
with A and B two constants to be determined by imposing the
ε = ε1 + ε2 (626) boundary conditions at
√ h √ i
The primes in Eqs. (623)–(626) indicate differentiation with re- z = 0 (y = −a 2) and z = d y = 2(d − a)
spect to the argument of the corresponding function. h√ i
√
Solution of Eq. (624) immediately leads to G(−a 2) = G 2(d − a) = 0 (637)
F(θ) = (2π)−1/2 exp(imθ) (627) plus the normalization condition,
where the normalization factor was obtained through the peri- Z d Z √2(d−a)
1
odicity requirement F(θ) = F(θ + 2π), characteristic of the az- G2 (z)dz = √ √ G2 (y) dy = 1 (638)
imuthal solution. The quantity m in Eq. (627) is recognized as 0 2 −a 2
the magnetic quantum number and is restricted to the values Using Eqs. (636)–(638), we obtain
m = 0; ±1; ±2; : : : . √ √
To solve Eq. (623), we first write the radial function as AU(−ε2 ; −a 2) + BU(−ε2 ; a 2) = 0 (639)
h √ i h √ i
R(ρ) = w−1/2 H(w) (628) AU −ε2 ; 2(d − a) + BU −ε2 ; 2(a − d) = 0 (640)

where we defined w = ρ2 . After introducing R(ρ) as given by A nontrivial solution to Eqs. (639) and (640) exists only if the sec-
Eq. (628) into Eq. (623), it may be easily shown that H(w) satisfies ular determinant is zero, i.e.,
h i √ h √ i
H 00 (w) + − 14 + (ε1 /2w) + (1 − m2 )/(4w2 ) H(w) = 0 (629) U(−ε2 ; −a 2)U −ε2 ; 2(a − d)
√ h √ i
This is precisely Whittaker’s equation, whose solution may be ex- −U(−ε2 ; a 2)U −ε2 ; 2(d − a) = 0 (641)
pressed as [243]
  This last equation furnishes the quantization condition that al-
H(w) = D exp(−w/2)wµ+1/2 M µ −  + 12 ; 2µ + 1; w (630) lows us to find ε2 as a function of the distance d between the walls
and the position a of the oscillator center relative to one of the
where  = ε1 /2, µ = |m|/2, and M(α; β; w) is the Kummer func- walls.
tion [243] with α = µ−+ 12 , β = 2µ+1, and D is a normalization (s)
Denoting by ε2 (a; d) the sth root of Eq. (641), with ε2 <
(1)
constant. (2) (3)
ε2 < ε2 ; : : :, the total energy of the system then becomes
To guarantee proper behavior of H(w) as w → ∞, one must
have (s)
 εnms = 2n + |m| + ε2 (a; d) + 1 (642)
µ −  + 12 = −n = |m| − ε1 + 1)/2 n = 0; 1; 2; : : : (631)
where n = 0; 1; 2; : : :; m = 0; ±1; ±2; : : :, and s = 1; 2; 3; : : : .
therefore, the energy ε1 , is given as Note that, just as in the case of the free oscillator, an accidental
degeneracy is observed. For a given value of s, we have a manifold
ε1 = 2n + |m| + 1 (632) of values for n and m, yielding the same result for 2n + |m| + 1. For
instance, n = 1, |m| = 0 (s state), and n = 0, |m| = 2 (d state) give
Substituting Eqs. (631) and (632) into Eq. (630) yields
h i the same energy for fixed s. Also, as expected, the ground state is
H(w) = D n!/|m0 |!(|m| + n)! nondegenerate.
Before analyzing the results of the calculations indicated pre-
|m|
× exp(−w/2)w(|m|+1)/2 Ln (w) (633) viously, it is interesting to evaluate the force exerted by the walls
 OPTICAL PROPERTIES IN NANOSTRUCTURES 105

Fig. 110. Ground state (A), and first two excited states (B) and (C) for
Fig. 111. Force exerted by the walls on the oscillator calculated according
different wall separation (d) as a function of relative position (a) of the
to Eq. (643) as a function of parameter (a) for d = 2 and d = ∞.
oscillator center from one of walls (see text).

that the force is zero for a = d/2, which means that the two walls
on the oscillator. This may be done through use the Hellmann–
are exerting equal and opposite forces at the center. The oscillator
Feynman theorem [244],
*  + will bounce back and forth between the walls, driven by a restoring
 ∂V force proportional to its displacement relative to the central region
∂εgr ∂ε001 
Fe = − =− = ψ |ψ of confinement. This is just what we would expect classically, too.
∂a ∂a  ∂a
Z ∞ Z 2π Z d When one of the walls is brought to infinity, we see that for dis-
tances to the wall a > 3, the oscillator is practically unperturbed.
= (z − a)G2 (z)R2 (ρ)F 2 (θ)ρ dρ dθ dz
0 0 0 Examination of Figure 110 supports this observation.
Z √2(d−a)
1
= √ yG2 (y) dy (643)
2 −a 2 5.4. Hydrogen Atom and Harmonic Oscillator Confined by
Impenetrable Spherical Boxes
with G(y) as defined by Eq. (636).
The roots of Eq. (641) were found numerically for different val- The direct variational method is used to study two simple confined
ues of the parameters a and d with a precision of 10−6 . Figure 110 systems, namely, the hydrogen atom and the harmonic oscillator
shows the values of εnms for the ground state (ε001 ) and the first within impenetrable spherical boxes. The trial wave functions were
two excited states {ε011 ; (ε101 ; ε021 )}, as a function of a and for assumed as the product of the free solutions of the corresponding
d = 2, 4, and 6, respectively. For clarity of presentation, we la- Schrödinger equation and a simple function that satisfies the re-
beled the curves for each value of d by A, B, and C, indicating spective boundary conditions. The energy levels obtained in this
ε001 , ε002 , and ε003 , respectively. way are extremely close to the exact ones, thus proving the utility
We first note the symmetry shown by the energy curves around of the proposed method.
a = d/2 for all the states. This is not surprising, due to the symme-
try of the problem. Note also the shift of the energy levels toward
5.4.1. Direct Variational Approach
higher values as the distance between the walls is reduced. This
energy shift is larger for the excited states, showing a slower ten- The exact solution for the free system can be found in any text of
dency to the unperturbed situation as one of the walls is taken to quantum mechanics [171]. Indeed, the corresponding energy and
infinity. In fact, in the latter case, the problem is reduced to that wave functions are given as
of a three-dimensional oscillator in front of a wall. As a → 0 and 1
d → ∞, U(−ε; y) approaches the Hermite polynomial Hs (y) and εn = − n = 1; 2; : : : (644)
n2
εs → s + 1/2, (s = 1; 3; 5; : : :). Hence, according to Eq. (642),
εnms → 2n + |m| + s + 3/2; i.e., only the odd states will satisfy ψnlm (r; θ; ϕ) = Nnl (2r/n)l F(−n + l + 1; 2l + 2; 2r/n)
!
the boundary condition. This behavior appears due to the lack of r
nodes of the even y-functions at y = a = 0. Conversely, if a, × exp − Ylm (θ; ϕ) (645)
n
d → ∞, we recover the energy spectrum of the free oscillator,
as expected. where h̄ = m = 1, Ylm (θ; ϕ) are the spherical harmonics,
With regard to the force on the oscillator, Figure 111 shows F(a; b; z) is the confluent hypergeometric function [243], and Nnl
the corresponding values obtained through Eq. (643). We observe is a normalization constant.
106 RIERA ET AL.

When we impose confinement on this system, the Hamiltonian

is slightly modified and can be written as

b0 = − 1 ∇ 2 − 1 + V 0 (r)
H (646)
2
r
where 
+∞ r > r0
V 0 (r) = (647)
0 r ≤ r0
and r0 is the radius of the confining spherical box.
The corresponding Schrödinger equation is still separable, but
the resulting radial equation,
" #
d2 2 d l(l + 1) 2
+ − + + 2ε R(r) = 0 (648)
dr 2 r dr r2 r

must be solved with the following boundary condition:

R(r0 ) = 0 (649)
This means that, to obtain the energy spectrum, we must find the
roots of Eq. (649). The new situation must be tackled in a more
complicated way since we must construct a convergent series rep-
Fig. 112. Energy levels of the enclosed hydrogen atom as a function of
resentation of R(r) and then solve it numerically.
the radius of the box r0 . Exact and variational calculations are compared
Alternatively, we solve the same problem approximately, with (see text).
the aid of the modified variational method discussed previously.
We exploit the fact that the solutions of the free hydrogen atom
are known [see Eq. (645)] to choose the trial wave functions as
Fernandez and Castro [246] who used a method based on both hy-
χ0nlm (r; θ; ϕ) = N(r0 − r)(2αr)l F(−n + l + 1; 2l + 2; 2αr) pervirial theorems and perturbation theory [247]. A remarkable
agreement can be noticed, showing that the problem can be tack-
× exp(−αr)Ylm (θ; ϕ) (650)
led in the simpler fashion proposed here. Note that, in contrast
where N is a normalization constant that depends on r0 and α as with the hypervirial treatment, the energy values obtained by the
well as on n and l. variational method always lie as an upper bound to the exact re-
We can note that 1/n is replaced by α in this choice for χ0 . sults for all box sizes, as expected.
The reason is that, as a result of confinement, the number n is no Incidentally, Fernandez and Castro [245] were the first to ob-
longer a good quantum number to specify the state of the system tain the ground state energies shown in Figure 112 variationally.
(l is still a good quantum number since symmetry is not broken). These authors used a trial wave function identical to Eq. (650)
This ansatz gives more flexibility to the variational wave functions, while analyzing the fulfillment of the quantum virial theorem for
allowing for the calculation not only of the ground state energy, but enclosed systems by approximate wave functions.
also of the excited states, with only one variational parameter. Fur- We now turn our attention to the case of the enclosed harmonic
thermore, the quantum virial theorem for enclosed systems [245] oscillator following a procedure similar to the case of the enclosed
is satisfied by these functions. hydrogen atom.
To show the adequacy of the method, in the following we re- Once again, the problem of the free oscillator is exactly solved
strict ourselves to states described by nodeless wave functions in- and the energy and wave functions are given by [171],
volving different symmetries.
εnl = 2n + l + 3/2 n = 0; 1; 2; : : : (653)
When we use Eq. (650) as a trial wave function, together with
the Hamiltonian given by Eq. (646), the energy can be readily ψnl (r; θ; ϕ) = Nr l F(−n; l + 3/2; r 2 )
found by minimizing the functional, × exp(−r 2 /2)Ylm (θ; ϕ) (654)
Z
χ0 ∗ Hχ
b 0 dτ (651) where we set h̄ = m = ω = 1, Ylm (θ; ϕ) are the spherical har-
 monics, F(a; b; z) is the hypergeometric function, and N is a nor-
with respect to α, within the bounded volume , restricted to sat- malization constant.
isfy the constraints implied by As in the case of the hydrogen atom, when we impose the con-
Z finement, the Hamiltonian is modified to give
χ0n ∗ χ0m dτ = δnm (652) 2
 b0 = − 1 ∇ 2 + r + V 0 (r)
H (655)
2
2
This procedure is straightforward and can be done through direct
algebraic manipulation. where 
+∞ r > r0
Figure 112 shows the results obtained after minimizing V 0 (r) = (656)
0 r ≤ r0
Eq. (651) as compared to the “exact” values obtained by numeri-
cally solving Eq. (649), [201] for the ground state and the 2p and r0 being the radius of the confining spherical box in units of
3d excited states. Also shown in this figure are the results due to (h̄/mω)1/2 .
 OPTICAL PROPERTIES IN NANOSTRUCTURES 107

Table X. Energy Levels of the Enclosed Harmonic Oscillator as a Function of the Radius of the Box∗

Ground state First excited state

(n = 0, l = 0) (n = 0, l = 1)

r0 α εHV † εvar εexact α εvar εexact

1.0 0.1310 5.0756 5.1313 5.0755 0.6339 10.3188 10.2822

1.5 0.1073 2.5050 5.5265 2.5050 0.3217 4.9169 4.9036
2.0 0.1365 1.7648 1.7739 1.7648 0.2402 3.2514 3.2469
2.5 0.1935 1.5517 1.5567 1.5514 0.2385 2.6901 2.6881
3.0 0.2606 1.5105 1.5061 0.2769 2.5337 2.5313
4.0 0.3530 1.5033 1.5000 0.3645 2.5015 2.5001
5.0 0.4085 1.5025 1.5000 0.4075 2.5012 2.5000

*Exact and variational calculations are compared. Energies in units of h̄ω. Radii in units of (h̄/mω)1/2 .
†Ref. [195].

Once again, if we were to solve the problem exactly, the 5.5. Sublevels and Excitons in GaAs–Ga1−xAlx As
Schrödinger equation is separable and the resulting radial equa- Parabolic-Quantum-well Structures
tion, In this subsection, energy levels and wave functions of electrons
" # and holes in parabolic-quantum-well structures are calculated ex-
d2 2 d l(l + 1) actly by using the envelope function approach and the usual con-
+ − − r 2 + 2ε R(r) = 0 (657)
dr 2 r dr r2 nection rules. They show harmonic oscillator-like behavior even
under an electric field. Variational calculations are presented for
should be solved with the exciton binding energies and their Stark effect on the parabolic
quantum-well structures, which show stronger confinement than in
R(r0 ) = 0 (658)
a single rectangular quantum well [252].
The exciton 1s ground state binding energy is estimated to be
which would force us to use a numerical treatment. In the modified about 9 meV and those involving heavy and light holes differ by
variational method, (as in the case of the confined hydrogen atom) typically 1–2 meV for a GaAs single rectangular quantum well of
the symmetry of the enclosed harmonic oscillator is not broken width L ≈ 100 Å [25, 253]. Also, the exciton binding energies for
by the confinement and l is still a good quantum number but n the excited states of the electrons and holes were assumed equal to
ceases to be so. To construct the trial wave functions, we simply those of the ground state. This assumption may not be as good as
replace r 2 by αr 2 in the argument of the functions in Eq. (654) for the single rectangular quantum well because the extent of the
and we proceed, as in our previous example, to find the energy of wave functions differs more strongly for the levels of a parabolic
the system for each box size. potential [15, 254]. The first successful fabrication of layered struc-
In Table X, we compare the exact results obtained from tures of GaAs and Ga1−x Alx As to simulate a parabolic band pro-
Eq. (658) [201] and those obtained by the application of the vari- file between the confining Ga1−x Alx As layers was reported. By
ational procedure, for the ground state and the first excited state. varying the layer thickness quadratically with the distance from
For completeness, we also show the results for the ground state, the center of the sample, Miller et al. [15] showed that the sub-
obtained through the hypervirial perturbational method using an levels were almost uniformly spaced in energy. They concluded the
11-term perturbation expansion [248]. Once again, a very good value of the band offset parameter Qe , to be 0:51, and later to be
agreement is observed. 0:57, from the exciton spectra in the parabolic-like multiple quan-
An important criterion for this selection is the fulfillment of the tum wells (MQW) without considering the differences of exciton
virial theorem for these systems [245]. At this stage it is important binding energies [14, 15]. The band offset at the interfaces is still
causing problems; these are actually important and require further
to note that there are several articles in the literature dealing with
investigation [255, 256]. Chan studied the effects of the hole sub-
different techniques to tackle these problems [202].
band mixing on the energies of excitons in a simple rectangular
In this connection, Fernandez and co-workers [245, 249–251]
quantum well [257]. Here, we use the envelope function approach
have done a thorough study of the use of the virial theorem for and the usual connection rules [258] to calculate the sublevel en-
quantum systems subject to Dirichilet and/or Neumann boundary ergies and the wave functions of electrons and holes in the MQW
conditions. As we mentioned before, in both examples, the agree- shown in Figure 113. The effective potential may be divided into
ment of the results obtained by the proposed modification of the many regions (21 wells of GaAs and 20 barriers of Ga1−x Alx As
direct variational approach is remarkable, in spite of the simplicity as well as two confining barriers of Ga1−x Alx As, each of which
of the method. is a square well or barrier. By connecting the wave functions at
We note that the symmetry of the systems and their confine- the interfaces of these regions, we can obtain the eigenstates en-
ment were intentionally chosen to be compatible, i.e., both being ergies and wave functions. Numerical results of sublevel energies
of spherical symmetry. and wave functions as well as their Stark effect, for electrons as an
108 RIERA ET AL.

example, are shown in Table XI and Figure 114. The origins of dis-
tance and of electrostatic potential are chosen at the center of the
well. In this calculation, the bandgap discontinuity of GaAs and
Ga1−x Alx As is, 1εg (x) = 1:42x − 0:90x2 +1:1x3 eV; the electron
effective mass is m∗e = 0:0665 + 0:0835x [259]. The wave functions
in Figure 114 are very similar to those of a harmonic oscillator,
even under an electric field, although there are some additional
wiggles in the curves, especially for higher states. Therefore, we
confirm that the MQW shown in Figure 113 can be treated as a
finite parabolic quantum well (PQW) defined by

1 Kz 2 |z| ≤ Lz /2
Vb (z) = 2 (659)
V0 |z| ≥ Lz /2
with
8V0
K= (660)
L2z
where Lz is the well width at the top and V0 is the height of the
confining barriers representing 1εc , and 1εv , for electrons and
holes, respectively. When an electric field E is applied, the quasi-
bound condition for the ground state is [260],
eEβ  V0 − ε1 (1/β)2 = 2m∗2 (V0 − ε1 )/h̄2 (661)
where β is the characteristic decay length of the unperturbed (E = Fig. 113. Effective potential for electrons or holes in the multiple quan-
0) ground state wave function inside the finite barrier, ε1 is the tum wells. V0 represents 1εc or 1εv , respectively.
ground state energy, and m∗2 is the effective mass of the electron
or hole in the Ga1−x Alx As barriers. With H b0 = Vb (z) − Kz2 /2
[−∞ < z < +∞], Vb (z) is shown in Eq. (659) as the perturbation
Hamiltonian, the first-order perturbation shows that the energy
differences between finite and infinite PQWs are very small and
negligible for the not very narrow well width, e.g., Lz  10 Å for
the ground state. So, if the weak field condition of Eq. (661) is
satisfied and the well depth Lz is not very narrow, the behavior of
the 1s ground excitons in the MQW may be approximated by that
in the infinite PQW.
To calculate the 1s ground state binding energies of heavy and
light excitons and their Stark effect in an infinite PQW, we consider
the Hamiltonian for the envelope functions of electrons and holes
within the effective mass approximation,
b=H
H bez + H
bhz + H
beh (662)
where
2 ∂2
bez = − h̄
H + Vbe (ze ) + eEze
∗
2me⊥ ∂ze2
2 ∂2 Fig. 114. Exact solutions of wave functions of electrons in the multiple
bhz = − h̄
H + Vbh (zh ) − eEzh
2m∗h⊥ ∂z2 quantum wells (Fig. 108) with x = 0:3, 1εc = 0:651εg (i.e., Qe = 0:65),
h and (a) Lz = 510 Å, (b) Lz = 100 Å. Solid curve E = 0, dashed curve
 ! 
2 2 eELz = 1εc .
beh = − h̄  1 ∂ ∂ 1 ∂ 
H r − (663)
2µ∗ r ∂r ∂r r 2 ∂θ2

e2
− Table XI. Sublevel Energies of Electrons in the Multiple Quantum
κ[(ze − zh )2 + r 2 ]1/2
Well Shown in Figure 116 with x = 0:3, Qe = 0:65, and Ly = 510 Å.
Here, we neglect the center of mass kinetic energy of the ex-
citon in the (x; y) plane; ze (zh ) is the z-coordinate of the elec- ε1 ε2 − ε1 ε3 − ε2 ε4 − ε3 ε5 − ε4 ε6 − ε5
tron (hole); m∗e⊥ (m∗h⊥ ) is the effective mass of the electron (hole) (meV)
in the z-direction; r(r; θ) is the relative position of the electron
and the hole in the plane; and µ∗ = m∗e// m∗h// /(m∗e// + m∗h// ) is E=0 14.3 28.7 28.2 27.9 27.7 27.6
the reduced effective mass in the plane of layers (m∗e// (m∗h// ) is eELz = 1εc −1:4 28.1 28.1 28.0 26.8
the electron (hole) effective mass in the plane). Vbe (ze )[Vbh (zh )]
 OPTICAL PROPERTIES IN NANOSTRUCTURES 109

is the effective parabolic potential defined by Vb (z) = Kz2 /2

(−∞ < z < +∞), where K is shown in Eq. (659). E is the elec-
tric field satisfying condition (661). We choose a separable wave
function for the exciton for simplicity,
8(ze ; zh ; r) = ψe (ze )ψh (zh )φeh (r)
!1/2 " #
αe (ze + ze0 )2
ψe (ze ) = √ exp −α2e
π 2
!1/2 #
αh 2
(zh + zh0 )2
ψh (zh ) = √ exp[−αh (664)
π 2
!1/2
2  
φeh (r) = exp −r/λ /λ
π

where Fig. 115. Exciton binding energies of ε1h and ε1l versus the well width
Lz for x = 0:3 and various band-offset parameter Qe in the parabolic
ze0 = eE/Ke zh0 = −eE/Kh quantum well.
αe = (Ke m∗e⊥ /h̄2 )1/4 αh = (Kh m∗h⊥ /h̄2 )1/4 (665)
Ke = 8Qe 1εg /L2z Kh = 8(1 − Qe )1εg /L2z

Here, ψe (ze )[ψh (zh )] is the eigenfunction of H bez (Hbhz );

φeh (r) is a 1s-like Bohr orbital with variational parameter λ. With
the separable wave function approximation, we can obtain the to-
tal energy of the 1s ground state exciton by εex = εez + εhz + εexc ,
where εez and εhz are the eigenvalues of H bez and H
bhz , respec-
b
tively; εexc = h8|Heh |8i is variationally minimized, i.e., the exci-
ton binding energy. From Eq. (664), we have
D E 2 Z +∞ Z +∞ h i
beh |8 = h̄ 2e2
8|H − exp −(x2 + y 2 ) dx dy
2µ∗ λ2 πκλ −∞ −∞
Z ∞ h i
× exp −(t 2 + 2at)1/2 dt (666)
0
where a = 2(x/αe − y/αh − ze0 + zh0 )/λ. The parameters in the
calculation are the electron effective mass, 0.0665; heavy- (light-)
hole effective mass perpendicular to the plane, 0.34, (0.094) (hole
effective masses parallel to the plane were calculated as speci-
fied by Miller et al. [25]); and the appropriate dielectric constant
κ = 12:15 [261]. It is clear, from Figure 115, that Qe has no
significant influence on exciton binding energies. We may take a
value Qe = 0:65 to estimate the exciton binding energies, although
the band offset still remains an unsolved problem [256, 262]. Fig-
ure 116 shows the binding energies of ε1h and ε1l , excitons as a Fig. 116. Exciton binding energies of ε1h and ε1l versus the well width
function of the well width Lz , for different Al concentrations x. Lz for Qe = 0:65 and various x in the parabolic quantum well.
For Lz ≈ 100 Å, the 1s ground state binding energy in the PQW is
about 10 meV, as shown in Figures 115 and 116. The Stark effect
on the exciton binding energy is demonstrated in Figure 117, which
should be considered to obtain the field induced exciton peak shift 1/2
in the PQW. To compare the behavior of ground excitons in a PQW For example, Leff = 8:4Lz Å (Lz in angstroms) with x = 0:3,
with those in a SQW, it is best to use the effective SQW width Leff Qe = 0:65, and m∗e = 0:0665. The binding energy of about 10 meV
relating to the ground state in the PQW of width Lz at the top, i.e., at Lz = 100 Å L, from Figure 116 is a little larger than 9 meV
approximately, of a SQW at Leff = 84 Å which shows stronger confinement in
!1/2 PQW than in SQW. It should be pointed out that the calculation of
π 2 h̄2 h̄ 8V0 binding energies in the preceding text is based on the infinite PQW
= 21 (667)
2m∗ L2 Lz m∗ approximation. However, for the ground excitons in a finite PQW
eff
with a not very narrow well width at the top, the results, we think,
for an electron, are exact enough and can be regarded as those in the MQW for
L2eff π 2 h̄ the reasons we mentioned earlier. Effects of valence band mixing
= (668)
Lz (8m∗e Qe 1εg )1/2 are not considered.
110 RIERA ET AL.

5.6.1. Theory and Model

Within the effective mass approximation, the Hamiltonian for
an interacting pair of electrons confined in a quantum dot by
parabolic potential of the form m∗e ω0 r 2 /2 in a magnetic field ap-
plied parallel to the z-axis (and perpendicular to the plane where
the electrons are restricted to move) in the symmetric gauge is
written as
( )
X 2
h̄2 ∇i2 1 h̄ω e2
b=
H − ∗ 2
+ me ω ri +2 c z
Li + (669)
2me ∗ 2 2 κ|r1 − r2 |
i=1
where the two-dimensional vectors r1 and r2 describe the positions
of the first and the second electron in the (x; y)-plane, respectively.
Lzi stands for the z-component of the orbital angular momentum
for each electron and ωc = eB/m∗e c, m∗e , and κ are the cyclotron
frequency, effective mass, and dielectric constant of the medium,
respectively. The frequency ω depends on both the magnetic field
Fig. 117. The Stark effect on exciton binding energies of ε1h and ε1l in B and the confinement frequency ω0 and is given by
the parabolic quantum well with Qe = 0:65, x = 0:3, and Lz = 510 Å.
!1/2
2 ω2c
ω = ω0 + (670)
4
5.6. Energy States of Two Electrons in a Parabolic
The natural units of length and energy to be used are the ef-
Quantum Dot in Magnetic Field
fective Bohr radius a∗B = κh̄2 /m∗e e2 and the effective Rydberg
The energy spectra of two interacting electrons in a quantum dot
R∗y = h̄2 /2m∗e a∗2
B . The dimensionless constant γ = h̄ωc /2Ry
∗
confined by a parabolic potential in an applied magnetic field of plays the role of an effective magnetic field strength.
arbitrary strength are obtained in this subsection. The shifted 1/N √
expansion method is used to solve the effective mass Hamilto-
Upon introducing the center of mass √ R = (r1 + r2 )/ 2 and
the relative coordinates r = (r1 − r2 )/ 2, the Hamiltonian [271]
nian [263]. The influence of the electron–electron interaction on in Eq. (669) can be written as a sum of two separable parts that
the ground state energy and its significant effect on the energy represent the center of mass motion Hamiltonian,
level crossings in states with different angular momenta is shown.
2 ∗
The dependence of the ground state energy on the magnetic field bR = − h̄ ∇ 2 + me ω2 R2 + h̄ωc LR
H (671)
strength for various confinement energies is presented. 2me ∗ R 2 2 z
The magnetic field dependence plays a useful role in identi-
and the relative motion Hamiltonian,
fying the absorption features. The effects of the magnetic field on
2 ∗ 2
the state of the impurity [264] and the excitons [172, 236, 238, 265– br = − h̄ ∇r2 + me ω2 r 2 + h̄ωc Lrz + e
H (672)
267] confined in a quantum dot were extensively studied. Kumar 2me ∗ 2 2 r
et al. [268] self-consistently solved the Poisson and Schrödinger
Equation (671) describes the Hamiltonian of the harmonic oscil-
equations and obtained the electron states in GaAs–GaAlAs for
lator with the well-known eigenenergies,
both cases: in zero and for magnetic fields applied perpendicu-
  h̄ωc
lar to the heterojunctions. The results of their work [268] indi-
εncm ;mcm = 2ncm + |mcm | + 1 h̄ω + mcm (673)
cated that the confinement potential can be approximated by a 2
simple one-parameter adjustable parabolic potential. Merkt et al. labeled by the radial (ncm = 0; 1; 2; : : : and the azimuthal (mcm =
[269] presented a study of quantum dots in which both the mag- 0; ±1; ±2; ±3; : : :) quantum numbers. The problem is reduced to
netic field and the electron–electron interaction terms were taken obtaining eigenenergies εnr ;m of the relative motion Hamiltonian.
into account. Pfannkuche and Gerhardts [270] devoted a theoreti- The energy states of the total Hamiltonian are labeled by the cm
cal study to the magneto-optical response to far-infrared radiation and the relative quantum numbers, |ncm mcm ; nr mi. The coexis-
(far-IR) of quantum dot helium, accounting for deviations from tence of the electron–electron and the oscillator terms make the
the parabolic confinement. De Groote et al. [271] investigated the exact analytic solution with the present special functions not pos-
thermodynamic properties of quantum dots taking into consider- sible.
ation the spin effect, in addition to the electron–electron interac-
tion and magnetic field terms. The purpose of this subsection is to
show the effect of the electron–electron interaction on the spectra 5.6.2. The Shifted 1/N Expansion Method
of the quantum dot states with nonvanishing azimuthal quantum The shifted 1/N expansion method, N being the spatial dimen-
numbers and the transitions in the ground state of the system as sions, is a pseudo-perturbative technique in the sense that it pro-
the magnetic field strength increases. poses a perturbation parameter that is not directly related to the
Here we use the shifted 1/N expansion method to obtain an en- coupling constant [272–275]. The aspect of this method was clearly
ergy expression for the spectra of two confined electrons in a quan- stated by Imbo et al. [272–274] who displayed step-by-step calcula-
tum dot by solving the effective mass Hamiltonian including the tions relevant to this method. Following their work, here only the
following terms: the electron–electron interaction, applied field, analytic expressions which are required to determine the energy
and the parabolic confinement potential. states are presented.
 OPTICAL PROPERTIES IN NANOSTRUCTURES 111

The method starts by writing the radial Schrödinger equa-

tion, for an arbitrary cylindrically symmetric potential, in an
N-dimensional space as
( )
d2 (k − 1)(k − 3)
− + + V(r) ψ(r) = εr ψ(r) (674)
dr 2 4r 2
where k = N + 2m.
To get useful results from the 1/k̄ expansion, where k̄ = k − a
and a is a suitable shift parameter, the large k̄-limit of the potential
must be suitably defined [276]. Since the angular momentum bar-
rier term behaves like k̄2 at large k̄, so the potential should behave
similarly. This gives rise to an effective potential, which does not
vary with k̄ at large values of k̄ resulting in a sensible zeroth-order
classical result. Hence, Eq. (674) in terms of the shift parameter Fig. 118. The relative ground state energy |00i for the electrons in a quan-
becomes tum dot as a function of confinement length `0 = (h̄/m∗e ω0 )1/2 for the
( ) zero magnetic field: this subsection calculations, closed circles; Ref. [269]
d2 k̄2 [1 − (1 − a)/k̄] [1 − (3 − a)/k̄] V(r) solid line. (Reprinted from U. Merkt et al., Phys. Rev. B, 43, 7320, (1991).)
− + + ψ(r)
dr 2 4r 2 Q
= εr ψ(r) (675)
where
2 1 2 2 ωc
V(r) = + 4 ω r +m (676)
r 2
and Q is a scaling constant to be specified from Eq. (678). The
shifted 1/N-expansion method consists in solving Eq. (675) sys-
tematically in terms of the expansion parameter 1/k. The leading
contribution term to the energy comes from
2
!
k̄2 1 r0 V(r0 )
k̄2 Veff (r) = + (677)
r2 40
Q

where r0 is the minimum of the effective potential, given by

2r03 V 0 (r0 ) = Q (678)

It is convenient to shift the origin to r0 by the definition,
Fig. 119. The low-lying relative states |00i, |10i, and |20i for two electrons
x = k̄1/2 (r − r0 )/r0 (679) in a quantum dot made of InSb as a function of confinement frequency ω.
and expanding Eq. (675) about x = 0 in powers of x. Compar-
ing the coefficients of powers of x in the series with the corre-
sponding ones of the same order in the Schrödinger equation for a by requiring an agreement between the 1/k̄ expansion and the
one-dimensional anharmonic oscillator, we determine the anhar- exact analytic results for the harmonic and Coulomb potentials.
monic oscillator frequency, the energy eigenvalue, and the scaling From Eq. (682), we obtain
constant in terms of k̄, Q, r0 , and the potential derivatives. The a = 2 − (2nr + 1)ω (683)
anharmonic frequency parameter is
" #1/2 where nr is the radial quantum number related to the principal n
V 00 (r0 ) and magnetic m quantum numbers by the relation nr = n−|m|−1.
ω= 3+ 0 (680)
V (r0 ) Energies and lengths in Eqs. (674)–(683) are expressed in units of
R∗y and a∗B , respectively.
and the energy eigenvalues in powers of 1/k̄ (up to third order) For the two-dimensional case, N = 2, Eq. (678) takes the fol-
read as lowing form,
" # p
k̄2 1 (1 − a)(3 − a) γ 2r0 V 0 (r0 ) = 2 + 2m − a = Q1/2 (684)
εnr ;m = V(r0 ) + + + γ1 + 2 (681)
4r0 r3 0
4 k̄r 2 0 Once r0 (for a particular quantum state and confining fre-
The explicit forms of γ1 and γ2 are given in Subsection 5.6.3. quency) is determined, the task of computing the energy is rela-
The shift parameter a, which introduces an additional degree of tively simple.
freedom, is chosen so as to make the first term in the energy series The results are presented in Figures 118–122 and Tables XII
of order k̄ vanish, namely, and XIII. The relative ground state energy |00i of the relative
" # motion, for the zero magnetic field case, against the confinement
k̄   (2 − a) length is displayed in Figure 118. The present results (black dots)
nr + 2 ω −
1
=0 (682)
r20
2 clearly show an excellent agreement with the numerical results of
112 RIERA ET AL.

Table XII. The Roots r0 Determined by Eq. (684) for Quantum Dot
States with Nonvanishing Azimuthal Quantum Number (m) against the
Ratio ωc /ω0

m
ωc /ω0 0 −1 −2 −3 −4 −5

1 4.262 4.827 5.457 6.075 6.659 7.209

2 3.692 3.682 4.212 4.719 5.193 5.635
3 3.188 2.943 3.398 3.825 4.229 4.586

Table XIII. The Ground State Energies (in Atomic Units) of the
Relative Hamiltonian Calculated by 1/N-Expansion at Different
Frequencies, Compared with the Results of Taut [278]

1/ω 1/N-Expansion Taut

Fig. 120. The total ground state energy |00; 00i for two electrons in a
quantum dot as a function of the ratio ωc /ω0 . For independent (solid line) 4 0.4220 0.6250
and interacting (dashed line) electrons. 20 0.1305 0.1750
54.7386 0.0635 0.0822
115.299 0.0375 0.0477
523.102 0.0131 0.0162
1054.54 0.0081 0.0100
1419.47 0.0067 0.0081

Source: M. Taut, Phys. Rev. A, 48, 3561, (1993).

Ref. [269] (dashed line). In Figure 119, the first low energy levels,
|00i; |10i, and |20i of the relative Hamiltonian are presented as a
function of the effective confinement frequency ω, using parame-
ters appropriate to InSb, where the dielectric constant κ = 17:88,
electron effective mass m∗e = 0:014m0 , and confinement energy
h̄ω0 = 7:5 meV [271]. The energy levels obviously show a lin-
ear dependence on the effective frequency. As the effective fre-
quency ω increases, the confining energy term dominates the in-
Fig. 121. The total eigenenergies of the states |00; 0mi, m = teraction energy term and thus the linear relationship between the
0; −1; −2; : : : ; −5, for two interacting electrons parabolically confined in energy and the frequency is maintained. This result is consistent
the quantum dot of size `0 = 3a∗B as a function of the ratio ωc /ω0 . with Ref. [271].
To investigate the effect of the electron–electron interaction
on the energy spectra of the quantum dot, we plotted in Fig-
ure 120 the total ground state energy |00; 00i of the full Hamil-
tonian for independent (solid line) and interacting (dashed line)
electron as a function of the ratio ωc /ω0 . The figure shows, as
we expect, a significant energy enhancement when the electron–
electron Coulombic interaction term is turned on. Furthermore,
as the magnetic field increases, the electrons are further squeezed
in the quantum dot, resulting in an increase of the repulsive
electron–electron Coulombic energy and in effect the energy lev-
els.
The energy level crossings are shown in Figure 121. We dis-
played the eigenenergies of the states |00; 0mi; m = 0; −1; −2,
: : : ; −5, for two interacting electrons parabolically confined in
the quantum dot of size `0 = 3a∗B as a function of the ratio
ωc /ω0 . As the magnetic field strength increases, the energy of
the state m = 0 increases while the energy of the states with
nonvanishing quantum number m decreases, thus leading to a se-
Fig. 122. The relative ground state |00ienergy versus the magnetic field quence of different ground states, as reported in Ref. [276]. In
strength for two different confinement energies. the interacting system, the interaction energy is lower the higher
 OPTICAL PROPERTIES IN NANOSTRUCTURES 113

the angular momentum of the relative motion. This is caused by of an electron system in a parabolic potential is independent of
the structure of the relative wave function: The larger the an- electron–electron interactions and thus the actual number of elec-
gular momentum the larger the spatial extent and therefore, the trons in the well, as reported by Wixforth et al. in a review arti-
larger is the distance between the electrons [277]. To confirm cle [279].
this numerically, we list in Table XII the roots r0 of the poten- In conclusion, we obtained the energy spectra of two interacting
tial for the interacting electrons in the quantum dot, for states electrons as a function of confinement energies and magnetic field
with different angular momentum. At particular values of the ra- strength. The method showed good agreement with the numerical
tio ωc /ω0 , as the azimuthal quantum number |m| increases, the results of Merkt et al. [269], Taut [278], and Wagner et al. [276].
root r0 also increases and thus the electron–electron interaction Our calculations also showed the effect of the electron–electron
Vee (r) = 2/r0 , in the leading term of the energy series expression, interaction term on the ground state energy and its significance on
decreases. the energy level crossings in states with different azimuthal quan-
In Figure 122, we showed the dependence of the ground tum numbers. The shifted 1/N-expansion method yields quick re-
state energy on the magnetic field strength for confinement en- sults without putting restrictions on the Hamiltonian of the sys-
ergies: h̄ω0 = 6 and 12 meV. For constant value of the mag- tem.
netic field, the larger the confinement energy, the greater the
energy of the interacting electrons in the quantum dot. The
5.6.3. Calculation of Parameters γ1 and γ2
spin effect can be included in the Hamiltonian Eq. (669) added
to the center of mass part as a space independent term, and The explicit forms of the parameters γ1 and γ2 are given in the
Eq. (671) is still an analytically solvable harmonic oscillator Hamil- following. Here, R∗y and a∗B are used as units of energy and length,
tonian [271]. respectively,
We compared, in Table IX, the calculated results for the ground h i
state energies |00i of the relative Hamiltonian at different con- γ1 = c1 e2 + 3c2 e4 − ω−1 e21 + 6c1 e1 e3 + c4 e32 (686)
fining frequencies with the results of Taut [278]. Taut reported a
and
particular analytical solution of the Schrödinger equation for two
γ2 = T7 + T12 + T16 (687)
interacting electrons in an external harmonic potential. The ta-
ble shows as 1/ω increases the difference between both results where
noticeably decreases until it becomes ≈ 1:4 × 10−3 at 1/ω =
1419:47. T7 = T1 − ω−1 [T2 + T3 + T4 + T5 + T6 ]
Quantum dots with more than two electrons can also be stud- T12 = ω−2 [T8 + T9 + T10 + T11 ] (688)
ied. The Hamiltonian for ne -interacting electrons, provided that T16 = ω−2 [T13 + T14 + T15 ]
the electron–electron interaction term depends only on the rel-
ative coordinates between electrons V(|ri − rj |) = e2 /κ|rij |, and with
parabolically confined in the quantum dot, is separable into a cm
T1 = c1 d2 + 3c2 d4 + c3 d6 T2 = c1 e22 + 12c2 e2 e4
and a relative Hamiltonian. The parabolic potential form V(ri ) =
mi ω20 ri2 /2, i = 1; 2; 3; : : : ; ne is the only potential which leads T3 = 2e1 d1 + 2c5 e24 T4 = 6c1 e1 d3 + 30c2 e1 d5
to a separable Hamiltonian. The cm motion part is described by T5 = 6c1 e1 d3 + 2c4 e3 d3 T6 = 10c6 e3 d5
the one-particle Hamiltonian, Eq. (671), with the electron mass 2
T8 = 4e1 e2 + 36c1 e1 e2 e3 T9 = 8c4 e2 e23 (689)
replaced by the total mass M = ne m∗e and the electron charge
replaced by the total charge Q = ne e. The relative Hamilto- T10 = 24ce21 e4 + 8c7 e1 e3 e4 T11 = 12c8 e23 e4
nian part, which involves only the relative coordinates and mo- T13 = 8e1 e3 + 108c1 e1 e3 T14 = 48c4 e1 e3
menta, has a cylindrically symmetric potential and can be han-
dled by the 1/N-expansion technique. When the confining poten- T15 = 30c9 e3
tial is quadratic, far-IR spectroscopy is insensitive to the interac- Where cs, ds and es are parameters given as
tion effects because of Pthe cm and relative motions. The radia-
tion dipole operator i ei ri = QR, being a pure cm variable, c1 = 1 + 2nr c2 = 1 + 2nr + 2n2r
does not couple to H br which contains all the electron–electron in- c3 = 3 + 8nr + 6n2r + 4n3r c4 = 11 + 30nr + 30n2r
teractions. The dipole operator then induces transitions between
the states of the cm but does not affect the states of the rel- c5 = 21 + 59nr + 51n2r + 34n3r
ative Hamiltonian. The eigenenergies for the cm Hamiltonian, c6 = 13 + 40nr + 42n2r + 28n3r (690)
Eq. (673), does not change because ωc , in the energy expression
c7 = 31 + 78n2r + 78n3r
remains the same, namely, QB/Mc = eB/m∗e c. Consequently,
the far-IR absorption experiments see only the feature of the sin- c8 = 57 + 189nr + 225n2r + 150n3r
gle electron energies. There are only two allowed dipole transi- c9 = 31 + 109nr + 141n2r + 94n3r
tions (1m = ±1) and the far-IR resonance occurs at frequen-
cies, ej = κj /ωj/2 di = δi /2
v
u ! where j = 1; 2; 3; 4 and i = 1; 2; 3; 4; 5; 6,
u ω 2
ω± = t c
+ ω20 ±
ωc
(685)
2 2 3(2 − a)
κ1 = (2 − a) κ2 = −
2
Many different experiments on quantum dots prove the valid- 2r0 2r
ity of Kohn’s theorem and that the observed resonance frequency κ3 = −1 − κ4 = 4 + 0
5 (691)
Q Q
114 RIERA ET AL.

(1 − a)(3 − a) 3(1 − a)(3 − a)

δ1 = − δ2 = − 32. J. Shah, A. Pinczuk, H. L. Stormer, A. C. Gossard, and W. Wiegmann,
2 4 Appl. Phys. Lett. 42, 55 (1983).
5(2 − a) 33. D. Bimberg, J. Christen, A. Steckenborn, G. Weimann, and
δ3 = 2(2 − a) δ4 = − W. Schlapp, J. Lumin. 30, 562 (1985).
2
2r0 2r 34. P. Dawson, G. Duggan, H. I. Ralph, and K. Woodbridge, Phys. Rev. B
δ5 = − 32 − δ6 = 4 + 0
7
28, 7381 (1983).
Q Q
35. P. Voisin, Thése de Doctorat d’Etat, Paris, 1983, unpublished.
36. P. Voisin, G. Bastard, and M. Voos, Phys. Rev. B 29, 935 (1984).
37. A. Pinczuk, J. Shah, H. L. Stormer, R. C. Miller, A. C. Gossard, and
REFERENCES W. Wiegmann, Surf. Sci. 142, 492 (1984).
38. T. M. Rice, “Solid State Physics” (H. Ehrenreich, F. Seitz, and
1. C. Weisbuch and B. Vinter, “Quantum Semiconductor Structures: D. Turnbull, Eds.), Vol. 32. Academic Press, New York, 1977.
Fundamentals and Applications,” Academic Press, Boston, 1991. 39. C. Weisbuch, R. C. Miller, R. Dingle, A. C. Gossard, and W. Wieg-
2. R. A. Stradling and P. C. Klipstein, “Growth and Characterisation of mann, Solid State Commun. 37, 219 (1981).
Semiconductors,” Hilger, Bristol, U.K., 1990. 40. G. Bastard, C. Delalande, M. H. Meynadier, P. M. Frijlink, and
3. G. Fasol, A. Fasolino, and P. Lugli, Eds., “Spectroscopy of Semicon- M. Voos, Phys. Rev. B 29, 7042 (1984).
ductor Microstructures,” Plenum, New York, 1989. 41. C. Delalande, M. H. Meynadier, and M. Voos, Phys. Rev. B 31, 2497
4. F. H. Pollak and O. J. Glembocki, SPIE 946, 2 (1988). (1985).
5. R. R. Alfano, SPIE 793, (1986); 942, (1988); 1282, (1990). 42. B. Deveaud, J. Y. Emery, A. Chomette, B. Lambert, and M. Baudet,
6. D. Heitmann, in “Physics of Nanostructures.” Proceedings of the Superlattices Microstruct. 1, 205 (1985).
Thirty-Eighth Scottish Universities Summer School in Physics, St. 43. M. D. Sturge, J. Hegarty, and L. Goldner, in “Proceedings of the 17th
Andrews (J. H. Davies and A. R. Long, Eds.), Institute of Physics International Conference on the Physics of Semiconductors,” San
Publishing, Bristol, U.K., 1992. Francisco, 1984, (J. D. Chadi and W. A. Harrison, Eds.), Springer-
7. A. Pinczuk, J. P. Valladares, D. Heiman, L. N. Pfeiffer, and Verlag, New York, 1985.
K. W. West, Surf. Sci. 229, 384 (1990). 44. J. Hegarty, M. D. Sturge, C. Weisbuch, A. C. Gossard, and W. Wieg-
8. M. Born and E. Wolf, “Principles of Optics,” Pergamon, Oxford, mann, Phys. Rev. Lett. 49, 930 (1982).
U.K., 1964. 45. J. Singh, K. K. Bajaj, and S. Chaudhuri, Appl. Phys. Lett. 44, 805
9. J. Pankove, “Optical Processes in Semiconductors,” Dover, New (1984).
York, 1975. 46. E. O. Gobel, J. Kuhl, and R. Hoger, J. Lumin. 30, 541 (1985).
10. H. N. Spector, Phys. Rev. B 28, 971 (1983). 47. D. Bimberg, J. Christen, A. Steckenborn, G. Weimann, and
11. G. D. Sanders and Y.-C. Chang, J. Vac. Sci. Technol. B 3, 1285 (1985). W. Schlapp, Appl. Phys. Lett. 44, 84 (1984).
12. J. N. Schulman and Y.-C. Chang, Phys. Rev. B 31, 2056 (1985). 48. U. Bockelmann and G. Bastard, Phys. Rev. B 45, 1688 (1992).
13. R. Dingle, in “Festkörperprobleme.” Advances in Solid State Physics 49. J. Cibert, P. M. Petroff, G. J. Dolan, S. J. Pearton, A. C. Gossard, and
(H. J. Queisser, Ed.), Vol. 15. Pergamon/Vieweg, Braunschweig, J. H. English, Appl. Phys. Lett. 49, 1275 (1986).
1975. 50. Y. Hirayama, S. Tarucha, Y. Suzuki, and H. Okamoto, Phys. Rev. B
14. R. C. Miller, D. A. Kleinman, and A. C. Gossard, Phys. Rev. B 29, 37, 2774 (1988).
7085 (1984). 51. D. Gershoni, H. Temkin, G. J. Dolan, J. Dunsmuir, S. N. G. Chu, and
15. R. C. Miller, A. C. Gossard, D. A. Kleinman, and O. Munteanu, Phys. M. B. Panish, Appl. Phys. Lett. 53, 995 (1988).
Rev. B 29, 3740 (1984). 52. E. Kapon, S. Simhony, R. Bhat, and D. M. Hwang, Appl. Phys. Lett.
16. M. H. Meynadier, C. Delalande, G. Bastard, M. Voos, F. Alexandre, 55, 2715 (1989).
and J. L. Lievin, Phys. Rev. B 31, 5539 (1985). 53. M. Tsuchiya, J. M. Gaines, R. H. Yan, R. J. Simes, P. O. Holtz,
17. G. Duggan, J. Vac. Sci. Technol. B 3, 1224 (1985). L. A. Coldren, and P. M. Petroff, Phys. Rev. Lett. 62, 466 (1989).
18. H. Kroemer, Surf. Sci. 174, 299 (1986). 54. M. Tanaka, J. Motohisa, and H. Sakaki, Surf. Sci. 228, 408 (1990).
19. P. Voisin, M. Voos, and M. Razeghi, 1984, unpublished. 55. M. Kohl, D. Heitmann, P. Grambow, and K. Ploog, Phys. Rev. Lett.
20. A. F. S. Penna, J. Shah, A. Pinczuk, D. Sivco, and A. Y. Cho, Appl. 63, 2124 (1989).
Phys. Lett. 46, 184 (1985). 56. M. Kohl, D. Heitmann, W. W. Rühle, P. Grambow, and K. Ploog,
21. P. Voisin, C. Delalande, M. Voos, L. L. Chang, A. Segmuller, Phys. Rev. B 41, 12,338 (1990).
C. A. Chang, and L. Esaki, Phys. Rev. B 30, 2276 (1984). 57. J. S. Weiner, G. Danan, A. Pinczuk, J. Valladares, L. N. Pfeiffer, and
22. J. Y. Marzin, M. Quillec, E. V. K. Rao, G. Leroux, and L. Goldstein, K. West, Phys. Rev. Lett. 63, 1641 (1989).
Surf. Sci. 142, 509 (1984). 58. T. Egeler, G. Abstreiter, G. Weimann, T. Demel, D. Heitmann,
23. J. D. Dow, in “Optical Properties of Solids, New Developments” P. Grambow, and W. Schlapp, Phys. Rev. Lett. 65, 1804 (1990).
(B. O. Seraphin, Ed.), North-Holland, Amsterdam, 1976. 59. J. A. Brum and G. Bastard, Superlattices Microstruct. 4, 443 (1988).
24. R. S. Knox, “Theory of Excitons, Solid State Phys. Suppl. 5,” Aca- 60. M. Sweeny, J. Xu, and M. Shur, Superlattices Microstruct. 4, 623
demic Press, New York, 1963. (1988).
25. R. C. Miller, D. A. Kleinman, W. T. Tsang, and A. C. Gossard, Phys. 61. D. S. Citrin and Y. C. Chang, Phys. Rev. B 40, 5507 (1989).
Rev. B 24, 1134 (1981). 62. D. S. Citrin and Y. C. Chang, J. Appl. Phys. 68, 161 (1990).
26. R. L. Greene and K. K. Bajaj, Solid State Commun. 45, 831 (1983). 63. P. C. Sercel and K. J. Vahala, Phys. Rev. B 42, 3690 (1990).
27. A. Messiah, “Mécanique Quantique,” Dunod, Paris, 1959. 64. J. W. Brown and H. N. Spector, Phys. Rev. B 35, 3009 (1987).
28. F. L. Lederman and J. D. Dow, Phys. Rev. B 13, 1633 (1976). 65. M. Asada, Y. Miyamoto, and Y. Suematsu, Jpn. J. Appl. Phys. 24, L95
29. C. Weisbuch, R. Dingle, A. C. Gossard, and W. Wiegmann, Solid State (1985).
Commun. 38, 709 (1981). 66. D. A. B. Miller, D. S. Chemla, and S. Schmitt-Rink, Appl. Phys. Lett.
30. Y. Masumoto, M. Matsuura, S. Tarucha, and H. Okamoto, Phys. Rev. 52, 2154 (1988).
B 32, 4275 (1985). 67. U. Bockelmann and G. Bastard, Europhys. Lett. 15, 215 (1991).
31. H. B. Bebb and E. W. Williams, in “Semiconductors and Semimetals” 68. E. O. Kane, J. Phys. Chem. Solids 1, 249 (1957).
(R. K. Willardson and A. C. Beer, Eds.), Vol. 8. Academic Press, New 69. J. M. Luttinger and W. Kohn, Phys. Rev. 97, 869 (1955).
York, 1972. 70. J. M. Luttinger, Phys. Rev. 102, 1030 (1956).
 OPTICAL PROPERTIES IN NANOSTRUCTURES 115

71. M. Altarelli, in “Semiconductor Superlattices and Heterojunctions” 106. Yu. E. Perlin, Fiz. Tverd. Tela (Leningrad) 10, 1941 (1968).
(G. Allan, G. Bastard, N. Boccara, M. Lannoo, and M. Voos, Eds.), 107. S. N. Klimin, E. P. Pokatilov, and V. M. Fomin, Phys. Status Solidi B,
Springer-Verlag, Berlin, 1986. 184, 373 (1994).
72. R. Ferreira and G. Bastard, Phys. Rev. B 43, 9687 (1991). 108. T. Takagahara, Phys. Rev. B, 47, 4569 (1993).
73. V. M. Fomin, V. N. Gladilin, J. T. Devreese, E. P. Pokatilov, S. N. Bal- 109. M. Nirmal, D. J. Norris, M. Kuno, M. G. Bawendi, Al. L. Efros, and
aban, and S. N. Klimin, Phys. Rev. B 57, 2415 (1998). M. Rosen, Phys. Rev. Lett. 75, 3728 (1995).
74. M. G. Bawendi, W. L. Wilson, L. Rothberg, P. J. Carroll, T. M. Jedju, 110. G. L. Bir and G. E. Pikus, “Symmetry and Strain-Induced Effects in
M. L. Steigerwald, and L. E. Brus, Phys. Rev. Lett. 65, 1623 (1990). Semiconductors,” Wiley, New York, 1975.
75. M. Nirmal, C. B. Murray, D. J. Norris, and M. G. Bawendi, Z. Phys. 111. A. L. Efros, Phys. Rev. B 46, 7448 (1992).
D 26, 361 (1993). 112. D. J. Norris and M. G. Bawendi, Phys. Rev. B 53, 16,338 (1996).
76. V. Jungnickel, F. Henneberger, and J. Puls, in “22nd International 113. “Semiconductors. Physics of II-VI and I-VII Compounds, Semimag-
Conference on the Physics of Semiconductors” (D. J. Lockwood, netic Semiconductors,” New Series, Group III, (K. H. Hellwege,
Ed.), World Scientific, Singapore, 1994. X. X. Landolt-Börnstein, Eds.), Vol. 17, Pt. B. Springer-Verlag,
77. V. Jungnickel and F. Henneberger, J. Lumin. 70, 238 (1996). Berlin, 1982.
78. D. J. Norris, Al. L. Efros, M. Rosen, and M. G. Bawendi, Phys. Rev. 114. R. Riera, F. Comas, C. Trallero-Giner, and S. T. Pavlov, Phys. Status
B 53, 16,347 (1996). Solidi B 148, 533 (1988).
79. M. C. Klein, F. Hache, D. Ricard, and C. Flytzanis, Phys. Rev. B 42, 115. R. F. Wallis and D. L. Mills, Phys. Rev. B 2, 3312 (1970).
11,123 (1990). 116. F. Comas, C. Trallero-Giner, I. G. Lang, and S. T. Pavlov, Fiz. Tverd.
80. A. P. Alivisatos, T. D. Harris, P. J. Carroll, M. L. Steigerwald, and Tela 25, 57 (1985).
L. E. Brus, J. Chem. Phys. 90, 3463 (1989). 117. A. V. Goltsev, I. G. Lang, and S. T. Pavlov, Phys. Status Solidi B 94, 37
81. Al. L. Efros, A. I. Ekimov, F. Kozlowski, V. Petrova-Koch, H. Schmid- (1979).
baur, and S. Shulimov, Solid State Commun. 78, 853 (1991). 118. E. L. Ivchenko, I. G. Lang, and S. T. Pavlov, Fiz, Tverd. Tela 19, 2751
82. J. J. Shiang, S. H. Risbud, and A. P. Alivisatos, J. Chem. Phys. 98, 8432 (1977).
(1993). 119. F. Comas, C. Trallero-Giner, and R. Perez-Alvarez, J. Phys. C 19, 6479
83. A. Mlayah, A. M. Brugman, R. Carles, J. B. Renucci, M. Ya. Valakh, (1986).
and A. V. Pogorelov, Solid State Commun. 90, 567 (1994).
120. A. V. Goltsev, I. G. Lang, and S. T. Pavlov, Fiz. Tverd. Tela 20, 2542
84. G. Scamarcio, V. Spagnolo, G. Ventruti, M. Lugará, and G. C. Righ-
(1978).
ini, Phys. Rev. B 53, R10,489 (1996).
121. B. O. Seraphin and H. E. Bennet, “Semiconductor and Semimetal”
85. S. V. Gaponenko, Fiz. Tverd. Tela (Leningrad) 30, 577 (1996).
(R. K. Willardson and A. C. Beer, Eds.), Vol. 3. Academic Press, New
86. F. Henneberger and J. Puls, in “Optics of Semiconductor Nanocrys-
York, 1967.
tals” (F. Henneberger, S. Schmitt Rink, and E. O. Göbel, Eds.),
122. H. C. Weisbugh, Phys. Rev. B 15, 823 (1977).
Akademie-Verlag, Berlin, 1993.
123. J. M. Bergues, R. Betancourt-Riera, J. L. Marin, and R. Riera, Phys.
87. S. Nomura and T. Kobayashi, Phys. Rev. B 45, 1305 (1992).
Low-Dim. Struct. 7–8, 81 (1996).
88. Y. Chen, S. Huang, J. Yu, and Y. Chen, J. Lumin. 60–61, 786 (1994).
124. J. M. Bergues, R. Riera, F. Comas, and C. Trallero-Giner, J. Phys.:
89. S. I. Pekar, Zh. Eksp. Teor. Fiz. 20, 267 (1950).
Condens. Matter. 7, 7273 (1995).
90. K. Huang and A. Rhys, Proc. R. Soc. London, Ser. A 204, 406 (1950).
125. J. M. Bergues, R. Betancourt-Riera, R. Riera, and J. L. Marín, to be
91. A. Baldereschi and N. O. Lipari, Phys. Rev. B 8, 2697 (1973).
published.
92. N. O. Lipari and A. Baldereschi, Phys. Rev. Lett. 42, 1660 (1970).
126. M. V. Klein, IEEE J. Quantum Electron. QE-22, 1760 (1986).
93. J. B. Xia, Phys. Rev. B 40, 8500 (1989).
94. D. M. Mittleman, R. W. Schenlein, J. J. Shiang, V. L. Colvin, 127. M. Cardona, Superlattices Microstruct. 7, 183 (1990).
A. P. Alivisatos, and C. V. Shank, Phys Rev. B 49, 14435 (1994). 128. “Light Scattering in Solids V.” Topics of Applied Physics, (M. Car-
95. R. Englman, “The Jahn–Teller Effect in Molecules and Crystals,” Wi- dona and G. Güntherodt, Eds.), Vol. 66. Springer-Verlag, Heidelberg,
ley, New York, 1972. 1989.
96. A. M. Stoneham, “Theory of Defects in Solids: Electronic Structure 129. A. Pinczuk and E. Burstein, in “Light Scattering in Solids I.” Topics
of Defects in Insulators and Semiconductors,” Clarendon, Oxford, in Applied Physics, (M. Cardona, Ed.), Springer-Verlag, Heidelberg,
U.K., 1975. 1983.
97. Yu. E. Perlin and B. S. Tsukerblat, “The Effects of Electron- 130. C. Colvard, T. A. Gant, M. V. Klein, R. Merlin, R. Fisher, H. Morkoc,
Vibrational Interaction on the Optical Spectra of Paramagnetic Im- and A. G. Gossard, Phys. Rev. B 31, 2080 (1985).
purity Ions,” Shtiintsa, Kishinev, 1974. 131. A. Cros, A. Cantarero, C. Trallero-Giner, and M. Cardona, Phys. Rev.
98. Yu. E. Perlin and B. S. Tsukerblat, in “The Dynamical Jahn–Teller B 46, 12627 (1992).
Effect in Localized Systems” (Yu. E. Perlin and M. Wagner, Eds.), 132. A. J. Shield, C. Trallero-Giner, M. Cardona, H. T. Grahn, K. Ploog,
Elsevier, Amsterdam, 1984. J. A. Hoisler, D. A. Tenne, N. T. Moshegor, and A. Toropov, Phys.
99. I. B. Bersuker and V. Z. Polinger, “Vibronic Interactons in Molecules Rev. B 46, 6990 (1992).
and Crystals,” Springer-Verlag, Berlin, 1989. 133. A. K. Sood, J. Menéndez, M. Cardona, and K. Ploog, Phys. Rev. Lett.
100. H. A. Jahn and E. Teller, Proc. R. Soc. London, Ser. A 161, 220 (1937). 54, 2111 (1985).
101. V. M. Fomin, E. P. Pokatilov, J. T. Devreese, S. N. Klimin, S. N. Bal- 134. M. Cardona and G. Güntherodt, Eds., “Light Scattering in Solids II.”
aban, and V. N. Gladilin, in “Proceedings of the 23rd International Topics in Applied Physics, Vol. 50. Springer-Verlag, Berlin, 1982.
Conference on the Physics of Semiconductors” (M. Scheffler and 135. B. Jusserand and M. Cardona, in “Light Scattering in Solids V.”
R. Zimmermann, Eds.), World Scientific, Singapore, 1996. Springer Topics in Applied Physics (M. Cardona and G. Güntherodt,
102. R. Kubo, J. Phys. Soc. Jpn. 12, 570 (1957). Eds.), Vol. 66. Springer-Verlag, Heidelberg, 1989.
103. V. B. Berestetskii, E. M. Lifshitz, and L. P. Pitaevskii, “Relativistic 136. F. Comas, A. Cantarero, C. Trallero-Giner, and M. Moshinsky,
Quantum Theory,” Pergamon, Oxford, U.K., 1971. J. Phys.: Condens. Matter. 7, 1789 (1995).
104. R. P. Feynman, Phys. Rev. 84, 108 (1951). 137. F. Comas, C. Trallero-Giner, and A. Cantarero, Phys. Rev. B 47, 7602
105. M. Grundmann, J. Christen, N. N. Ledentsov, J. Böhre, D. Bim- (1993).
berg, S. S. Ruvimov, P. Werner, U. Richter, U. Gösele, J. Heyden- 138. C. Trallero-Giner and J. López-Gondar, Phys. B 138, 287 (1986).
reich, V. M. Ustinov, A. Yu. Egorov, A. E. Zhukov, P. S. Kop’ev, and 139. R. Betancourt-Riera, J. M. Bergues, R. Riera, and J. L. Marín, to be
Zh. I. Alferov, Phys. Rev. Lett. 74, 4043 (1995). published.
116 RIERA ET AL.

140. P. M. Chamberlain, C. Trallero-Giner, and M. Cardona, Phys. Rev. B 188. J. L. Marín and G. Muñoz, J. Mol. Struct. (Teochem) 287, 281 (1993).
51, 1680 (1995). 189. B. D. Perlson and J. A. Weil, J. Magn. Reson. 15, 594 (1974).
141. E. Roca, C. Trallero-Giner, and M. Cardona, Phys. Rev. B 49, 13,704 190. Y. Wang and N. Herron, J. Phys. Chem. 95, 525 (1991).
(1994). 191. E. Ley-Koo and K. P. Volke-Sepúlveda, Int. J. Quantum Chem. 65,
142. C. Trallero-Giner and F. Comas, Philos. Mag. B 70, 583 (1994). 269 (1997).
143. J. L. Marín, R. Rosas, and A. Uribe, Am. J. Phys. 63, 460 (1995). 192. G. Arfken, “Mathematical Methods for Physicists,” Academic Press,
144. J. L. Marín and S. A. Cruz, Am. J. Phys. 59, 931 (1991). New York, 1971.
145. C. Priester, G. Allan, and M. Lannoo, Phys. Rev. B 28, 7194 (1983). 193. P. M. Morse and H. Feshbach, “Methods of Theoretical Physics,”
146. C. Priestier, G. Allan, and M. Lannoo, Phys. Rev. B 29, 3408 (1984). McGraw-Hill, New York, 1953.
147. G. Bastard, Phys. Rev. B 24, 4714 (1981). 194. J. O. Hirschfelder, C. F. Curtis, and R. B. Bird, “Molecular Theory of
148. K. Tanaka, M. Nagaoka, and T. Yamabe, Phys. Rev. B 28, 7068 (1983). Gases and Liquids,” Wiley, New York, 1954.
149. S. Chaudhuri, Phys. Rev. B 28, 4480 (1983). 195. E. A. Hylleraas, Z. Phys. 65, 209 (1930).
150. C. Mailhiot, Y.-C. Chang, and T. C. McGill, Phys. Rev. B 26, 4449 196. J. G. Kirwood, Phys. Z 33, 57 (1932).
(1982). 197. E. Ley-Koo and S. A. Cruz, J. Chem. Phys. 74, 4603 (1981).
151. R. L. Greene and K. K. Bajaj, Solid State Commun. 45, 825 (1983). 198. J. L. Marín and S. A. Cruz, J. Phys. B: At. Mol. Opt. Phys. 24, 2899
152. R. J. Wagner, B. V. Shanabrook, J. E. Furneaux, J. Comas, (1991).
N. C. Jarosik, and B. D. McCombe, in “GaAs and Related Com- 199. C. A. Coulson and P. D. Robinson, Proc. R. Soc. London 71, 815
pounds 1984.” Institute of Physics Conference Series 74, (B. De Cré- (1958).
moux, Ed.), Hilger, Bristol, 1985. 200. D. R. Bates, K. Ledsham, and A. L. Stewart, Philos. Trans. R. Soc.
153. W. T. Masselink, Y.-C. Chang, and H. Morkoc, Phys. Rev. B 28, 7373 London, Ser. A 246, 215 81953).
(1983). 201. E. Ley-Koo and S. Rubinstein, J. Chem. Phys. 71, 351 (1979).
154. R. C. Miller, A. C. Gossard, W. T. Tsang, and O. Munteanu, Phys. Rev. 202. J. Gorecki and W. Byers-Brown, J. Chem. Phys. 89, 2138 (1988).
B 25, 3871 (1982). 203. F. M. Fernandez and E. A. Castro, Int. J. Quantum Chem. 19, 533
155. A. Corella-Madueño, R. Rosas, J. L. Marín, and R. Riera, Phys. Low- (1981).
Dim. Struct. 5–6, 75 (1999). 204. F. M. Fernandez and E. A. Castro, J. Chem. Phys. 75, 2908 (1981).
156. J. L. Marín and S. A. Cruz, J. Phys. B: At. Mol. Opt. Phys. 25, 4365 205. Y. Wu and L. M. Falicov, Phys. Rev. B 29, 3671 (1984).
(1992). 206. C. L. Pekeris, Phys. Rev. 112, 1649 (1958).
157. B. R. Nag and S. Gangopadhyay, Phys. Status Solidi A 179, 463 (1993). 207. J. Silva-Valencia and N. Porras-Montenegro, J. Appl. Phys. 81, 901
158. F. A. P. Osório, M. H. Degani, and O. Hipólito, Phys. Rev. B 37, 1402 (1997).
(1988). 208. L. E. Oliveira and R. Perez-Alvarez, Phys. Rev. B 40, 10, 460 (1989).
159. G. W. Bryant, Phys. Rev. B 29, 6632 (1987). 209. N. Porras-Montenegro and L. E. Oliveira, Solid State Commun. 76,
160. G. Li, S. V. Branis, and K. K. Bajaj, J. Appl. Phys. 77, 1097 (1995). 275 (1990).
161. J. L. Marín, Ph. D. Thesis, Facultad de Ciencias, UNAM, 1992. 210. N. Porras-Montenegro, A. Latgé, and L. E. Oliveira, J. Appl. Phys. 70,
162. J. Gorecki and W. Byers Brown, J. Phys. B 22, 2659 (1989). (1991).
163. K. R. Brownstein, Pys. Rev. Lett. 71, 1427 (1993). 211. N. Porras-Montenegro and S. T. Perez-Merchancano, Phys. Rev. B 46,
164. S. I. Tsonchev and P. L. Goodfriend, J. Phys. B 25, 4685 (1992). 9780 (1992).
165. J.-L. Zhu, J.-J. Xiong, and B.-L. Gu, Phys. Rev. B 41, 6001 (1990). 212. N. Porras-Montenegro, S. T. Perez-Merchancano, and A. Latgé,
166. J.-L. Zhu, J. Phys. Condens. Matter 1, 1539 (1989). J. Appl. Phys. 74, 7624 (1993).
167. J.-L. Zhu, Phys. Rev. B 39, 8780 (1989). 213. A. Latgé, N. Porras-Montenegro, and L. E. Oliveira, Phys. Rev. B 51,
168. A. Ishibashi, Y. Mori, M. Itabashi, and N. Watanabe, J. Appl. Phys. 13,344 (1995).
58, 2691 (1985). 214. M. Helm, F. M. Peeters, F. DeRosa, E. Colas, J. P. Harbison, and L.
169. L. E. Brus, J. Chem. Phys. 80, 4403 (1984). T. Florez, Phys. Rev. B 43, 13,983 (1991).
170. A. Erdelyi, “Asymptotic Expansions,” Dover, New York, 1956. 215. N. C. Jarosik, B. D. McCombe, B. V. Shanabrook, J. Comas, J. Ral-
171. A. S. Davydov, “Quantum Mechanics,” NEO Press, Peaks Island, ston, and G. Wicks, Phys. Rev. Lett. 54, 1283 (1985).
ME, 1966. 216. B. Yoo, B. D. McCombe, and W. Schaff, Phys. Rev. B 44, 13,152
172. G. W. Bryant, Phys. Rev. B 37, 8763 (1988). (1991).
173. G. W. Bryant, Surf. Sci. 196, 596 (1988). 217. F. Bassani and G. Parravicini, in “Electronic States and Optical Tran-
174. T. Inoshita and H. Watanabe, in “Microstructures” (S. Sugano, sitions in Solids” (R. A. Ballinger, Ed.), Pergamon, Oxford, U.K.,
Y. Nishina, and S. Ohnishi, Eds.), Springer-Verlag, Berlin, 1986. 1975.
175. M. Ulas, H. Akbas, and M. Tomak, Phys. Status Solidi B 200, 67 218. R. A. Rosas, J. L. Marín, R. Riera, and R. Núñez, Phys. Low-Dim.
(1997). Struct. 5–6, 145 (1999).
176. A. Corella-Madueño, R. Rosas, J. L. Marín, and R. Riera, Int. J. 219. E. Ley-Koo and R. M. G. García-Castelán, J. Phys. A: Math. Gen. 24,
Quantum Chem., in press. 1481 (1991).
177. P. Ramvall, S. Tanaka, S. Nomura, P. Riblet, and Y. Aoyagi, Appl. 220. E. Ley-Koo and S. Mateos-Cortés, Int. J. Quantum Chem. 46, 609
Phys. Lett. 73, 1104 (1998). (1993).
178. Z. Xiao, J. Zhu, and F. He, J. Appl. Phys. 79, 9181 (1966). 221. R. E. Moss, Am. J. Phys. 55, 397 (1987).
179. S. V. Branis, G. Li, and K. K. Bajaj, Phys. Rev. B 47, 1316 (1993). 222. M. Andrews, Am. J. Phys. 44, 1064 (1976).
180. Y. Chen, Phys. Lett. A 143, 152 (1990). 223. M. Andrews, Am. J. Phys. 34, 1194 (1966).
181. A. Corella-Madueño, R. A. Rosas, J. L. Marín, and R. Riera, in press. 224. M. Martin, Am. J. Phys. 47, 1067 (1979).
182. E. Ley-Koo and A. Flores-Flores, Int. J. Quantum Chem. 66, 123 225. R. Loudon, Am. J. Phys. 27, 649 (1959).
(1998). 226. L. S. Davtyan, G. S. Pogosyan, A. N. Sissakian, and V. M. Ter-
183. E. V. Ludeña, J. Chem. Phys. 69, 1770 (1978). Antonyan, J. Phys. A: Math. Gen. 20, 2765 (1987).
184. E. V. Ludeña and M. Gregori, J. Chem. Phys. 71, 2235 (1979). 227. H. N. Spector and J. Lee, Am. J. Phys. 53, 248 (1985).
185. J. Gorecki and W. Byers Brown, J. Phys. B: At. Mol. Opt. Phys. 21, 403 228. J. F. Jan and Y. C. Lee, Phys. Rev. B 50, 14,647 (1994).
(1988). 229. J. L. Marín, R. Riera, R. Rosas, and A. Uribe, Phys. Low-Dim. Struct.
186. R. LeSar and D. R. Herschbach, J. Phys. Chem. 85, 2798 (1981). 3–4, 73 (1998).
187. R. LeSar and D. R. Herschbach, J. Phys. Chem. 87, 5202 (1983). 230. L. E. Brus, J. Phys. Chem. 90, 2555 (1986).
 OPTICAL PROPERTIES IN NANOSTRUCTURES 117

231. L. E. Brus, IEEE J. Quantum Electron. QE-22, 1909 (1986). 256. M. Voos, Surf. Sci. 168, 852 (1986).
232. R. A. Morgan, S. H. Park, S. W. Koch, and N. Peyghambarian, Semi- 257. K. S. Chan, J. Phys. C 19, L125 (1986).
cond. Sci. Technol. 5, 544 (1990). 258. G. Bastard, Phys. Rev. B 24, 5693 (1981).
233. M. Yamamoto, R. Hayashi, K. Tsunetomo, K. Kohno, and Y. Osaka, 259. D. A. Kleinman and R. C. Miller, Phys. Rev. B 32, 2266 (1985).
Jpn. J. Appl. Phys. 30, 136 (1991). 260. G. Bastard, E. E. Mendez, L. L. Chang, and L. Esaki, Phys. Rev. B 28,
234. L. E. Brus, J. Chem. Phys. 79, 5566 (1983). 3241 (1983).
235. Y. Kayanuma, Phys. Rev. B 38, 9797 (1988). 261. D. A. B. Miller, D. S. Chemla, T. C. Damen, A. C. Gossard, W. Wieg-
236. Y. Kayanuma, Phys. Rev. B 41, 10,261 (1990). mann, T. H. Wood, and C. A. Burrus, Phys. Rev. Lett. 53, 2173 (1984).
237. Y. Wang and N. Herron, Phys. Rev. B 42, 7253 (1990). 262. W. Tan and C. Yang, J. Phys. C 21, 1935(1988).
238. G. T. Einevoll, Phys. Rev. B 45, 3410 (1992). 263. M. El-Said, J. Phys. I France 5, 1027 (1995).
239. P. E. Lippens and M. Lannoo, Phys. Rev. B 39, 10,935 (1989). 264. K. D. Zhu and S. W. Gu, Phys. Lett. A 172, 296 (1993).
240. H. M. Schmidt and H. Weller, Chem. Phys. Lett. 129, 615 (1986). 265. W. Que, Phys. Rev. B 45, 11,036 (1992).
241. J. L. Marín and S. A. Cruz, Am. J. Phys. 56, 1134 (1988). 266. V. Halonen, T. Chakraborty, and M. Pietalainen, Phys. Rev. B 45, 5980
242. M. Moshinsky, “The Harmonic Oscillator in Modern Physics: From (1992).
Atoms to Quarks,” Gordon & Breach, New York, 1969. 267. G. W. Bryant, Phys. Rev. Lett. 59, 1140 (1987).
243. M. Abramowitz and I. A. Stegun, “Handbook of Mathematical Func- 268. A. Kumar, S. E. Laux, and F. Stern, Phys. Rev. B 42, 5166 (1990).
tions,” Dover, New York, 1970. 269. U. Merkt, J. Huser, and M. Wagner, Phys. Rev. B 43, 7320 (1991).
244. I. N. Levine, “Quantum Chemistry,” Allyn & Bacon, Boston, 1974. 270. D. Pfannkuche and R. R. Gerhardts, Phys. Rev. B 44, 13,132 (1991).
245. F. M. Fernandez and E. A. Castro, Int. J. Quantum Chem. 21, 741 271. J. J. S. De Groote, J. E. M. Honos, and A. V. Chaplik, Phys. Rev. B 46,
(1982). 12,773 (1992).
246. F. M. Fernandez and E. A. Castro, J. Math. Phys. 23, 1103 (1982). 272. T. Imbo, A. Pagnamento, and U. Sukhatme, Phys. Rev. D 29, 8763
247. F. M. Fernandez and E. A. Castro, Int. J. Quantum Chem. 22, 623 (1984).
(1981). 273. T. Imbo and U. Sukhatme, Phys. Rev. D 28, 418 (1983).
248. F. M. Fernandez and E. A. Castro, Phys. Rev. A 5, 2883 (1981). 274. T. Imbo and U. Sukhatme, Phys. Rev. D 31, 2655 (1985).
249. G. A. Arteca, F. M. Fernandez, and E. A. Castro, J. Chem. Phys. 80, 275. R. Dutt, Mukherji, and Y. P. Varshni, J. Phys. B 19, 3411 (1986).
1569 (1984). 276. M. Wagner, U. Merkt, and A. V. Chaplik, Phys. Rev. B 45, 1951 (1992).
250. F. M. Fernandez and E. A. Castro, Am. J. Phys. 52, 453 (1984). 277. D. Pfannkuche and R. R. Gerhardts, Phys. B 189, 6 (1994).
251. F. M. Fernandez and E. A. Castro, Am. J. Phys. 50, 921 (1982). 278. M. Taut, Phys. Rev. A 48, 3561 (1993).
252. C. Yang and Q. Yang, Phys. Rev. B 37, 1364 (1988). 279. A. Wixforth, M. Kaloudis, C. Rocke, K. Ensslin, M. Sundaram,
253. R. L. Greene, K. K. Bajaj, and D. E. Phelps, Phys. Rev. B 29, 1807 J. H. English, and A. C. Gossard, Semicond. Sci. Technol. 9, 215
(1984). (1994).
254. W. Pötz, W. Porod, and D. K. Ferry, Phys. Rev. B 32, 3868 (1985).
255. G. Margaritondo, Surf. Sci. 168, 439 (1986).