1

Nikolai A. Sobolev,
Svetlana P. Kobeleva
Национальный
Physics and technology исследовательский
технологический
Universidade de
Aveiro
of nanosize structures университет «МИСиС»
National University of
Departamento de 2014/2015 Science and Technology
Física "MISIS"

4. QUANTUM DOTS

Recent breakthroughs in solid state physics are related to the fast development
of the new classes of structures – self-organized quantum dots (QDs). These unique
objects combine advantages of a bulk semiconductor with those of single atoms.
Their physical properties resemble an atom in a cage.

It is possible to touch a semiconductor quantum dot with a tip of a scanning

tunneling microscope, inject charged carriers and monitor emission from a single QD,
and realize many other unique experiments. Interaction of carriers in QDs is defined
completely by the laws of quantum mechanics and no “motion” of carriers is possible
in a traditional sense. The main mechanism of the information transfer between QDs
becomes single electron tunneling and Coulomb interaction of carriers in neighboring
dots. These give rise to novel directions in microelectronics: single electronics and
quantum dot cellular automata computing. QDs allow significant improvements of
characteristics of many modern devices: e.g. infrared detectors, light-emitting diodes,
lasers, and solar cells.
 2

1
Another important type of QDs are nanocrystal (NC) QDs which are also
known as colloidal QDs or nanoparticles (NPs). These are crystals whose nanometre-
scale size is similar to or less than that of the Bohr radius of an exciton in the
corresponding bulk material. The spatial extent of the electron-hole pair (EHP) is thus
determined by the dimensions of the NC, some of the important consequences of
which are outlined in this chapter.

Both the conduction and valence band in a QD separate into a set of discrete
quantised states for electrons and holes, respectively, the energy of which depends on
the QD size (Fig. 4-1). Each electron state can be described by two quantum
numbers: L, which determines the angular momentum of the envelope wavefunction,
and n, which is analogous to the principal quantum number of an atom; the notation
for these states is usually written in the form, nX, (where X = S, P, D, ... for L = 0, 1,
2, ...). However, because of sub-band mixing, the hole states are better described by
the quantum numbers n and F, where F determines the total angular momentum (i.e.
the sum of the envelope wavefunction and Bloch function angular momenta); the
notation for these states is nXF.

Fig. 4-1.
Energy band structure of Si
nanoparticles (NPs) as a
function of the radius.

1
A more general and also frequently used term is nanoparticle (NP).
 3

4.1. BASIC REQUIREMENTS FOR QDs IN ROOM TEMPERATURE

DEVICES

QDs should fulfill the following requirements in order to make them useful for
devices at room temperature:

(a) Sufficiently deep localizing potential and small QD size (a prerequisite for
observation and utilization of zero-dimensional confinement effects;

(b) QD ensembles should show high uniformity and high volume filling factor;

(c) The material should be coherent without defects like dislocations.

4.1.1. Size

The lower size limit of a QD is given by the condition that at least one energy
level of an electron or hole or both is present. The critical diameter (for a spherical
QD) Dmin depends strongly on the band offset of the corresponding bands in the
materials system used. An electron level exists in a spherical QD if the confinement
potential, defined by the conduction band offset (in type-I heterostructure) ΔEC,
exceeds the value (see Sec. 4.2.1)

2
2   
Eс   EQW , (1)
*
2me  Dmin  1

where me* is the effective electron mass and E QW the first level in a rectangular QW
1
with infinitely high barriers and the width Dmin (cf. Chap. 2). Assuming a conduction
band offset of ~0.3 eV for GaAs/Al0.4Ga0.6As heterostructures, the diameter of the
QD should be larger than 4 nm. This is the lower limit of the QD size. In QDs of this
or slightly larger size the separation between the electron level and the barrier energy
is very small, and at finite temperatures thermal escape of carriers from QDs will
result in their depletion. For the InAs/AlGaAs system, the conduction band offset is

much larger while the electron effective mass is smaller, so the product of Eс  me* is
 4

comparable, and the critical size is also about 3-5 nm depending on the non-
parabolicity effects in the InAs conduction band.

There also exists a limit for the maximum size of a QD. A thermal population
of higher-lying levels is undesirable for particular devices like, e.g., lasers. The
condition to limit the thermal population of higher-lying levels to 5% ≈ e–3 can be
written as

 QD QD   3kT .
 E2  E1  (2)
 

Here E QD and E QD are the energies of the first and second levels in the QD,
1 2
respectively. This equation establishes an upper limit at RT for the size of
GaAs/AlGaAs QDs of ~12 nm, and of ~20 nm for InAs/AlGaAs QDs if electron
levels are considered. (The maximum is, of course, a function of the operation
temperature.)

Consideration of hole quantization in QDs leads to even smaller maximum

size, namely, 5-6 nm for both material systems discussed above. This means that QDs
with sufficient sublevel separation for both electrons and holes are difficult to realize
in most III-V material systems like GaAs/AlGaAs and InAs/AlGaAs because of the
large electron-hole mass ratio.

NB! (a) For electrical applications, where confinement has to be realized only
for one sort of carriers, the above mentioned problem does not exist.

(b) Strong localization of both carrier types might be possible in other

materials having more similar electron and hole masses (e.g. group III nitrides and II-
VI compounds).

4.1.2. Uniformity and density

In principle, all structural parameters of a QD such as size, shape, and chemical

composition are subject to random fluctuations, even in the presence of ordering
 5

mechanisms. For optoelectronic applications, a dense array of equisized and

equishaped QDs is desired. A high density of QDs is necessary to obtain a high
modal gain in the lasers. The main impact of size fluctuations is a variation of the
energy position of electronic levels. For a laser, the inhomogeneous energetic
broadening should be as small as possible. In order to realize a reasonable width of a
QD ensemble luminescence spectrum, comparable to that of a QW at RT (20-30
meV) for a GaAs QD size of about 10 nm, one needs to have a size dispersion lower
than 1 nm.

4.1.3. Material quality

For the use in the optoelectronic devices, the density of dislocations and point
defects in the QD material and the surface recombination rate at all heterointerfaces
should be as low as possible, on the level of in situ grown QWs and interfaces used in
the state-of-the-art devices. This requirement makes QD fabrication using self-
organized growth favorable since all interfaces are formed in situ during crystal
growth.
 6

4.2. ELECTRONIC SPECTRUM OF A QUANTUM DOT

To understand the reasons for the unique properties of QDs let us first refer to
the density of states (DOS) of 3D, 2D, 1D and 0D (“pseudoatom”) solids, Fig. 4-2.

Fig. 4-2.

The electronic DOS profiles for systems of different dimensionality (3D, 2D,
1D, and 0D) are very different from one another, as shown in Fig. 4-2. The typical
DOS dependence on energy near an energy band extremum, D(E) is given by
  / 2  1
   E  E0   , where Δ is an integer, denoting the spatial dimension and Δ

assumes the values 1, 2, and 3, respectively, for 1D, 2D, and 3D systems. The
parameter E0 appearing in the density of states D(E) denotes the energy band
minimum (or maximum) for the conduction (valence) energy bands. For a 1D system,
E0 would correspond to the energy of a van Hove singularities (vHS) in the DOS
occurring at each subband edge, where the magnitude of the DOS becomes very
large. One can see from Fig. 4-2 that 1D systems exhibit DOS profiles which have
some similarity to the case of 0D systems, with both 0D and 1D systems having very
sharp maxima at certain energies, in contrast to the DOS profiles for 2D and 3D
systems, which show a more monotonic increase with energy (see Fig. 4-2).
However, the 1D DOS is different from the 0D DOS (δ-function at each discrete
 7

energy level) in that the 1D DOS has a sharp threshold and a decaying tail for each
cutting line, so that the 1D DOS does not go to zero between the sharp maxima, as
the 0D DOS does (see Fig. 4-2). This is even true for semiconducting nanotubes
which have a finite band gap and no occupied states between the first cutting lines in
the valence and conduction bands.

When an atom is excited, the electron goes to the higher energy level, and
when it relaxes back to the ground state, a photon with strictly defined energy is
emitted. The width of the emission or absorption line (ΔE) is defined by the
fundamental relation with the lifetime of the electron in the upper state. The
uncertainty in the emitted energy is:

τ ΔE ≥ ћ, (3)

where τ is the total lifetime of the electron in the excited state.

In contrast to the case of a diluted gas of atoms, the atoms in crystals are
strongly bound to each other. Their high density in crystals plays a very important
role in modern solid-state devices allowing high absorption or (in case of population
inversion) gain coefficients, providing high conductivity and making possible high
density flows of charged carries through the crystal. Due to this, a modern
semiconductor laser having the length of 1 mm and the cross-section of 10–4 mm2 can
emit continuous light with power of several watts, while a corresponding gas laser
has several meter long size.

At the same time, small separations between atoms make interaction of their
electron levels unavoidable. This interaction results in formation of wide bands of
allowed states in contrast to the discrete (δ-function-like) energy spectrum of single
atoms. In semiconductors, the last filled band of allowed states is called ”valence
band” and the next empty band is called ”conduction band”. Due to the broad
spectrum of allowed states in these bands, a wide range of transition energies
between electrons from the filled valence band to empty states in the conduction band
is possible. The absorption band becomes rather broad of the order of several
 8

electronvolts, in marked difference with sharp line absorption spectrum of single

atom. The excited electrons in the conduction band, as well as the empty states in the
valence band (”holes”) can move in the crystal via tunneling between sites of the
crystal lattice. As the atom potential profile in a crystal is periodic, electrons and
holes can move freely through the crystal, as free carriers do in vacuum. However,
the motion of charged carriers in crystals is described by a mass different of that of
free electrons, defined by the crystal field. The carriers are thus called
“quasiparticles”. In the widely used optoelectronic III–V materials (gallium arsenide
– GaAs; indium arsenide – InAs, etc.), the electrons effective masses are in the range
of 0.01 – 0.1 of the electron mass in vacuum.

Wide bands of allowed states in the crystal provide a lot of possibilities for
scattering of electrons and holes. Lattice vibrations easily stimulate transitions of
charge carriers in the energy range defined by the lattice temperature and / or scatter
their direction. The tails of the carrier distribution near the bottom of the conduction
band and the top of the valence band increase remarkably with temperature. Thus, the
concentration of carriers per energy interval near the band-edge drops. For the same
concentration of injected carriers broadening of their energy spectra results, among
other disadvantages, leading to the decrease in maximum gain and degradation of
laser performance.

The situation changes remarkably, if the motion of the charged carrier in the
crystal is limited to a very small volume, e. g. in a three-dimensional rectangular box.
Localization of carriers can be provided by a surrounding (matrix) material. In the
laser case it is important that the matrix material has a larger bandgap than the box
material and that the potential wells are attractive both for electrons and holes. As an
electron shows properties of both a particle and a wave, if the size of the box is small,
the electron energy spectrum is quantized similar to the case of the electron
quantization in the attracting Coulomb potential of a nucleus. In a simplified case of
infinite barriers at the (rectangular) box-matrix interface, the size quantization energy
is described by:
 9

 n
2

2  i 
E   , (4)
x, y, z 2m*   L
e i  x, y, z  i 


where me* is the electron effective mass, Ex,y,z is the size quantization energy due to
electron localization in the box with dimensions Lx, Ly, and Lz, respectively, and n =
1, 2, 3, ….

Electrons in crystals usually have rather small effective mass, and, even a
rather large size of the box of about 10 nm can result in a large energy separation
between electron sublevels (about 100 meV for GaAs QD). The latter value
significantly exceeds the thermal energy at room temperature (26 meV), and, thus,
population of excited states can be avoided. In this sense, there will be no
temperature dependence of the optical spectrum of such box in a wide temperature
range, and the realization of temperature insensitive devices becomes possible.

The problem of a particle in a finite potential is of course more difficult.

Moreover, the dots often have a non-spherical shape.

A schematics of the potential profile, confined energy levels and optical

transitions in a heteroepitxial pyramidal InAs/GaAs QD as well as the results of the
calculation for an especial case are shown in Figs. 4-3, 4.4.
 10

Fig. 4-3 (left) Schematics of the

potential profile, confined energy
levels and optical transitions in a
heteroepitaxial InAs/GaAs QD
(note the existence of a 2D wetting
layer). Electron energy level
splitting is 20-70meV, hole levels
are spaced by ~10meV.

Fig. 4-4 (below) Results of the

calculation for a InAs/GaAs
pyramid.
 11

4.2.1. Coulomb interaction

Confined charge carriers, electrons and holes interact via Coulomb interaction.

Each pair (i,j) of particles with charge q at position r contributes the Coulomb
interaction energy Wij to the system:

  1 qi q j
Wij (ri , r j )    . (5)
4 r  | ri  r j |
0

For qi = – qj (electron and hole) the energy of the system is lowered and an exciton is
formed. The correlation of the carrier motion by the Coulomb interaction not only
changes the energy spectrum of the system but also alters the oscillator strength of
transitions, e.g. the lifetime of the energy levels, which is of particular importance for
optoelectronic applications.

For Eq. (5), a uniform relative dielectric constant εr has been assumed
throughout the structure. For a non-uniform dielectric constant, the surface
polarization and image charges have to be taken into account. The correction Wij for
ε2 ≠ ε1 is caused by the electric field extending generally much further into the barrier
than the wavefunctions. Since in semiconductor heterostructures the barrier material
with larger bandgap has usually the smaller dielectric constant, the electrostatic
interaction is increased by the surface polarization compared to a calculation with the
same dielectric constant for the barrier and the dot region. A similar, more extreme
situation arises for semiconductor quantum dots in glass or solution, where 1 ≈ ε2 <<
ε1. The effect of image charges is enhanced for strongly anisotropic structures (the
extreme case being the quantum well), because then a large fraction of the field lines
penetrates the barrier.
 12

4.2.2. Coulomb blockade and single-electron transistor

Fig. 4-5. The single-island,

double tunnel junction
system.
(a) Schematic diagram, where
the island and electrode regions
(shaded) are formed
byconducting materials. The
gaps between these regions
form tunnel barriers.
(b) The potential energy across
the system, at zero bias.

4.2.3. Shallow impurities and excitons

When the ground states for electrons and holes in a QD are populated by one
carrier each, automatically an exciton is formed. In a bulk crystal, an exciton can
dissociate into a pair of free carriers in the conduction and valence bands, and in
principle, an exciton binding energy is measurable. The definition of an exciton
binding energy in a quantum dot as the difference of the eigenenergies of a
Hamiltonian with and without Coulomb interaction is mainly a theoretical concept
since an uncorrelated ground state is nonexistent. The exciton binding energy could
be measured in principle if the single-particle energies of electrons and holes are
determined independently, e.g. by inter-subband transitions in n- and p-doped QDs,
respectively.

Three regimes can be defined to compare the importance and role of Coulomb
effects with the effects due to the quantization of the kinetic energy:
 13

(a) Strong confinement regime. The Coulomb effects are only a small correction to
the dominant quantization of the kinetic energy, i.e. ΔEe >> EC and ΔEh >> EC,
where ΔEe and ΔEh are the electron and hole sublevel separations and EC is the
Coulomb interaction energy. Electron and hole wavefunctions are largely
uncorrelated. An example is a small QD where the dot radius is smaller than the
bulk exciton radius.

(b) Weak confinement regime, where the exciton radius is smaller than the dot
radius: the opposite is true. In this case, the electrons and holes form pairs whose
center of mass motion is quantized by the confinement potential. ΔEe and ΔEh
<≈ EC.

(c) Intermediate confinement regime: ΔEe >> EC and ΔEh << EC. This situation can
arise due to the different masses of electrons and holes. The hole energy is then
quantized by the electrostatic potential of the electron orbital.

Since the Coulomb energy depends largely on the value of the dielectric
constant, dots of the same size can belong to different regimes in different materials.
Typically, in III-V compounds, due to the relatively small dielectric constants, the
bulk exciton radius is > 10 nm, causing a structural QD of similar dimension and
sufficiently deep potential to be in the strong confinement regime. However, in II-VI
compounds the bulk exciton radius is smaller, and typically only a few nm, making it
much harder to obtain strong confinement in structural QDs.

The lifetime of the optically allowed states of excitons in QDs also depends on
the confinement regime (dot size). Calculations for quantum disks of fixed height and
varying radius gave the following results:

(a) Strong confinement regime (lifetime is independent of dot size, τ ≈ 1 ns);

 14

(b) Weak confinement regime (lifetime decreases proportionally to the volume, τ–

1
 V);

(c) Ideal QW (lifetime of free excitons, τ ≈ 1 ps).

For increasingly weaker confinement, the lifetime does not decrease anymore
because the expression yielding τ–1  V is obtained using the electric dipole
approximation, which starts to break down for disk radii of several tens of nm.

NB! The 2D limit of the QW exciton has to be compared with experimental

determinations of the lifetime of a true free QW exciton! Low-temperature
experiments on QWs mostly investigate excitons that are more ore less weakly bound
to lateral potential fluctuations, which results in radiative lifetimes in the range of
several hundreds of ps.

The ground and first excited energy state energy of an electron-hole pair in a
spherical QD with finite potential barriers can be calculated using a variational
approach:

In the limit R0 >> aB the calculation yields the correct values for bulk material.
For small radii a significant deviation from the result for infinite barriers occurs for
R0 < 3aB.

For the InAs/GaAs pyramids observed in the experiment, the exciton binding
energy has been calculated in the first-order perturbation theory:
 15

Fig. 4-6. (a) The ground state

wave function of a pyramidal
InAs/GaAs QD.
(b) Exciton binding energy in
InAs/GaAs QD calculated as a
function of the width of the base of
the pyramid.

A value of 20 meV has been obtained for a pyramid size of 12 nm. This value is large
compared to the InAs and GaAs bulk exciton binding energies of about 1 and 4 meV,
respectively.

4.2.4. Optical transitions

The possible inter-band optical transition in a single QD consists of a series of

δ-function lines, whose energy positions depend on the confined levels. In a real QD
ensemble, each individual dot has a slightly different size, shape, strain etc.;
 16

therefore, the energy spectrum varies from dot to dot. This leads to the observation of
inhomogeneous broadening of optical spectra.

The lifetime of a particular electron and hole (excitonic) state is determined by

the oscillator strength of the transition. The radiative lifetime sets a lower bound for
the linewidth Γ:

Γ  ħ/τ, Γ(meV)  0.65822/τ(ps) (6)

For a typical radiative lifetime τ = 300 ps, the broadening is very small, Γ  2.2 μeV.

Another consequence of the three-dimensional confinement is the

disappearance of the temperature dependence of the linewidth observed for three-,
two-, and one-dimensional systems. No continuum states can be populated, as in the
case of a bulk crystal or a QW, where the width of the band-to-band recombination
spectrum is given by FWHM = 1.8 kT and FWHM = ln(2) kT, respectively (see
Fig. 4-7).

Fig. 4-7.
Theoretical emission spectrum
of a semiconductor due to band-
to-band recombination. The full
width at half maximum (FWHM)
of an emission line is 1.8 kT.

The ionization of excitons by absorption of acoustic phonons to create free

excitons is absent; there is no final state for this process.
 17

However, the broadening can be measurable at a sufficiently high temperature

when the total lifetime of excitons in the ground state becomes shortened by phonon-
activated transitions between different excited and barrier states.

The simplest optical transitions in a spherical QD are shown in Fig. 4-8. An

empty QD can absorb one photon of energy EX to create an exciton (a). After being
populated with exciton, the QD cannot absorb another photon of this energy EX but a
photon of energy EXX – EX (b) and is then populated with a biexciton. In reverse, the
biexciton state can decay into a photon of energy EXX - EX which subsequently
decays into a photon of energy EX. The electrons and holes making up the biexciton
have opposite spins. With a nonlinear two-photon absorption process a biexciton can
be created by absorbing two photons of energy (EXX + EX)/2 (c). Since two-photon
transitions involve changes of the total angular momentum of 0 and 2, the creation of
a |1p, 1s3/2> (d) or |1s, 1p3/2> exciton (e) is possible.

Fig. 4-8.
 18

In a QD ensemble (as probed by nonspatially resolved experimental

techniques) each individual dot has slightly different properties induced by
fluctuations in size, shape, strain etc., leading to a variation of energy levels and
subsequently to an inhomogeneous broadening of the ensemble properties, e.g. the
spectrum. Typical inhomogeneously broadened linewidth of QD ensemble spectra are
10-100 meV.

4.3. GROWTH OF HETEROEPITAXIAL QDs

4.3.1. Lithographic techniques

Patterning of QWs is the most straightforward way of QD fabrication. It has

the following advantages:

(a) QDs of almost arbitrary lateral shape, size and arrangement can be realized.

(b) A variety of process techniques, like dry and wet etching, are at the disposal.

(c) It is generally compatible with modern VLSI semiconductor technology.

NB! On the other hand, dry etching (e.g., reactive ion etching) usually results
in the formation of a highly damaged or even amorphized layer at the surface. In such
a layer, the nonradiative recombination is predominant. Another detrimental effect is
a possible loss of the stoichiometry of compound targets.

Lithographic techniques comprise:

(a) Optical lithography and holography. Optical UV-lithography presently

provides resolution below 0.2 μm. Improvements of resolution below 100 nm
have been predicted. Nevertheless, the resolution will not be sufficient to
fabricate structures of lateral dimensions of 20 nm or less.

(b) X-ray lithography has the advantage of much shorter wavelengths and can be
used for nanostructure fabrication.
 19

(c) Electron beam lithography. Resolution in the 10-20 nm range has already been
demonstrated at the beginning of the 80s. Peculiarity of the method: to focus an
electron beam in the nanometer range a short working distance between the
final lens and the substrate is required. The final resolution of the process of
around 10 nm is limited by the resists due to finite length of the organic
molecules and the grain size.

A novel way to improve resolution is mask formation without resists. These

techniques, however, are not sufficiently developed.

Periodic nanostructures can also be fabricated using an electron beam

interference technique.

Fig. 4-9. Micrograph of quantum

dots made using electron beam
lithography and etching. This type
of quantum dot can be shaped and
positioned more reliably than dots
made with conventional crystal
growth methods.

(d) Focused ion beam lithography. At the present, the minimum beam diameter of
ca. 30 nm is somewhat larger than that of an electron beam. It is limited by the
ion energy spread and by chromatic aberration of electrostatic lenses. FIBL can
be used for:

(i) Maskless etching;

 20

(ii) Maskless implantation of dopants. If the lateral resolution further improves,

focused ion beams might become a basis for fabrication of QD arrays, e.g., by
focused implantation of In ions into GaAs/AlGaAs QWs.

(iii) Deposition of metallic structures;

(iv) Patterning of resists with strongly reduced proximity effect.

The main disadvantage of the ion beam methods are the introduced radiation
defects. The annealing is usually made by high-temperature treatment that
induces an interdiffusion of components.

(e) Scanning tunneling microscopy.

4.3.2. Strain-induced QDs

We have seen this method in connection with the fabrication of QWRs. The
lateral resolution of the method is only limited by the possibilities of electron
lithography, and very small mesas of stressors can be potentially formed. There are
no problems with the surface and defect states of the induced QDs. However, the
confinement energies are rather small.

4.3.3. Self-organized QDs

There are three well-known modes of heteroepitaxial growth (Fig. 4-10):

ESF(epi)+EIF ESF(epi)+EIF
< ESF(subs) > ESF(subs)

Fig. 4-10. Modes of

heteroepitaxial growth.

Frank - Volmer- Stranski-

van der Merwe Weber Krastanow
 21

 Frank-van der Merwe (layer-by-layer growth, 2D);

 Volmer-Weber (island growth, 3D);

 Stranski-Krastanow (wetting layer plus islands).

The particular growth mode for a given system depends on the interface
energies and on the lattice mismatch.

In lattice-matched systems, the growth mode is solely governed by the

interface and surface energies.

Experimental studies of coherent islands of InGaAs/GaAs (001) and

InAs/GaAs (001) systems have revealed a surprisingly narrow size distribution of
the islands which does not follow from the SK growth mode itself. Beside that,
coherent islands of InAs form, under certain conditions, a quasi-periodic square
lattice on the GaAs (001) surface. It is a new class of self-ordered nanostructures,
namely ordered arrays of coherently strained three-dimensional islands.

Fig. 4-11. Plan view TEM image of an InGaAs/GaAs

QD array (single sheet) grown at optimized
conditions TG = 500°C with a lateral QD density of
1.31011 cm–2.

Optimization of growth parameters to realize equilibrium arrays is a difficult

task to be solved for each material system separately. If islands uniform in size and in
shape are formed, one speaks of “self-organized quantum dots”, as this system
represents a clear example of spontaneous formation of macroscopic order from
initially random distributions. In the case of dense arrays of QDs, their interaction via
 22

the strained substrate makes their lateral ordering favorable. Growth on patterned
surfaces can also lead to ordered QDs. For multi-stack QD deposition, vertically
correlated growth of QDs has been demonstrated and, thus, quasicrystals composed
of quantum dots either in two or in all three dimensions can be fabricated. For islands
having a two-dimensional shape, either correlated or anticorrelated growth is possible
depending on the relative thickness of the spacer layer.

4.4. APPLICATIONS OF HETEROEPITAXIAL QUANTUM DOTS

4.4.1. Lasers

4.4.1.1. The ideal QD laser

In the 1970s Dingle and Henry recognized that energy quantization in quantum
wells could be exploited for semiconductor lasers to lower the lasing threshold and to
provide a tunable wavelength. Size quantization leads to an increase of the density of
states (DOS) near the band edge in QWs (see Chap. 2). When the dimensionality of
the active layer is further reduced, injected charge carriers concentrate in an
increasingly narrow energy range near the band edge. Consequently the maximum
material gain increases and the temperature dependence of laser parameters is
reduced. These effects were predicted theoretically in the 80s for lattice-matched
heterostructures like GaAs/AlGaAs. By considering dots with a cubic shape,
containing just one confined electron and hole state, a gain increase of an order of
magnitude and a corresponding reduction of the threshold current was predicted,
along with an increase of temperature stability. The final point is a beautiful
illustration of the impact of a discrete DOS. While in bulk materials charge carriers
are continuously distributed to higher energy states as the temperature is raised, no
such redistribution can occur in zero-dimensional structures, if the excited states lie
sufficiently above the ground state. All carriers then contribute to the inversion at
lasing energy, independent of temperature.
 23

The gain spectra of bulk, QW, QWR and QD laser based on In0.53Ga0.47As/InP
are compared in Fig. 4-12. For this calculation a homogeneous (relaxation)
broadening with τin = 0.1 ps (Γ = 6.6 meV) has been used, and no inhomogeneous
broadening of the two-, one-, and zero-dimensional system is considered.

Fig. 4-12. Gain spectra for

In0.53Ga0.47As/InP quantum dot, wire,
well and bulk (double heterostructure) at
T = 300 K, electron density 31018 cm–3
and in = 0.1 ps. Confined lateral
dimensions are 10 nm each.

The material gain spectrum for QDs is much narrower than for higher-
dimensional systems and the maximum gain is larger.

For a given carrier density of 3  1018 cm–3 the calculations yield that the gain
of QDs is about 20 times higher than for bulk crystals and 10 times higher than for
QWs.
 24

Fig. 4-13. Theoretical prediction

of the threshold current in double
heterostructure, quantum well,
quantum wire and quantum dot
lasers. The threshold current as a
function of temperature, T, is
given by: Jth(T) = Jth(0)exp (T/T0),
where T0 is the laser
characteristic temperature.

In an ideal QD laser diode with no inhomogeneous broadening, the density of

electronic states becomes a delta function. The electron distribution among the levels
then becomes temperature independent, and the common belief is that, assuming
infinite barriers for the carrier confinement, the threshold current becomes
independent of the temperature, T0 = ∞ (Fig. 4-13). This belief is a carry over from
bulk and planar QW laser diodes, for which the temperature dependence of the lasing
threshold is approximately set by the electron distribution in the QW laser or bulk
laser. However, these assumptions are wrong for the QD laser diode that is a much
different quantum system, and leads to the erroneous conclusion that the ideal QD
laser diode can achieve a T0 = ∞. The mistake is made by following the same physics
as for bulk and planar QW laser diodes. However, the QD laser diode is a much
different system, and the same approximations for temperature dependence become
invalid for a delta function density of electronic levels. In fact, for the ideal QD laser
diode we should expect a finite T0 (see below). Still, despite the temperature
dependence in its threshold, the QD laser diode offers dramatic benefits in its
threshold current density.
 25

4.4.1.2. The real QD laser

Fig. 4-14. Design of a vertically

emitting microcavity laser with
vertically coupled InGaAs quantum
dots. DBR stands for Distributed
Bragg Reflector.

Real QDs usually have several excited states which may be thermally
occupied. Stranski-Krastanow growth leads, moreover, to the formation of a wetting
layer of the QD material underneath the dots, influencing carrier capture and
reemission processes and thus the dynamics of lasers. Self-organized growth induces
some variation of the size and shape of the dots, leading to a spread of localization
energy within a dot ensemble. The lineshape of the ensemble luminescence, which is
composed of a superposition of numerous Lorentzian single-dot lineshapes, is then
inhomogeneously broadened (typically ≥ 50 meV). Thus, the carrier distribution
function of a real dot ensemble progressively changes from a non-thermal distribution
to a Fermi function at high temperatures (typically ≥ 300 K), also found for the bulk
material if the barriers are not too high.

For typical laser dimensions of several μm in stripe width and several hundred
μm in cavity length and QD densities of several 1010 cm–2, there are about 106 QDs in
the active zone. Depending on the broadening, the peak material gain can be very
high. However, the optical confinement factor for QD arrays is very small. The
resulting modal gain might be small. In real lasers all confinement potentials are
finite, making the escape of carriers into the barrier possible. Also, the energy
separation of a quantized state becomes finite, enabling a thermal population of
excited states.
 26

At least two schemes to prevent gain saturation are feasible:

(a) Increase of the modal gain by stacking layers of QDs;

(b) Decrease of mirror losses by high reflection coating of the facets.

The main part of Fig. 4-15 shows the typical Jth temperature dependence
observed for a QD laser. Below 250 K, Jth is relatively temperature insensitive and
for low temperatures ( 200 K) even decreases with increasing temperature.
Although the predicted temperature insensitivity of Jth is observed below 250 K, at
higher temperatures Jth is typically found to increase relatively rapidly. For the device
of Fig. 4-15, T0 has a value of ~50°C at RT, although similar devices can exhibit
values as high as ~100°C.

Fig. 4-15. Temperature

dependence of the threshold
current density for an InAs dot-in-
a-well (DWELL) laser. The device
contains five layers of QDs. The
right-hand inset shows emission
spectra below and above
threshold. The left hand inset
compares an undoped and p-
type modulation doped device.
Both devices contain five QD.

QD lasers today have the lowest threshold current density of any

semiconductor lasers. They are far superior to QWs as amplifiers, and their nonlinear
optical applications such as cross-gain modulation in local area networks, present the
basis for novel types of solar cells, nanoflash memories, single q-bit emitters for
quantum cryptography, etc.
 27

4.4.2. Solar cells

Fig. 4-16. Schematic of band

structure in the i-region of the
strain balanced QD p-i-n solar
cell.

4.4.3. QD memory: Optical writing – electrical erasing

Photoluminescence (PL) investigations on large ensembles of self-organized

QDs usually reveal a strongly inhomogeneously broadened PL linewidth (often Γi >
50 meV) which arises from naturally occurring size fluctuations. This contrasts
strongly with the much narrower (typically Γh < 150 µeV), homogeneously
broadened emission observed from single dots which reflects the d-function like
density of states expected for 0D systems.

Therefore, illumination of the QD ensemble with highly monochromatic light

(FWHM << Γi) is expected to enable selective excitation of the QD ensemble with
potential applications such as wavelength sensitive optical data storage media. In
order to develop such technology, device concepts are required which enable the
efficient electrical detection of optically written stored charge.

Charge storage effects in quantum dots have been investigated using a

transistor-like structure which incorporates a modulation doped 2D electron channel
 28

which is spatially separated from a parallel layer of self assembled InAs quantum
dots.

The quantum dot layer is embedded within the intrinsic region of a p-i-n
junction and spatially separated from the 2D electron channel by an asymmetric
barrier. Separate electrical contacts are established to both the lateral n-channel and
the back p-contact. Fig. 4-17 shows the band structure and illustrates the operating
principles of the device.

Fig. 4-17.
 29

Operating Principles:

 Following selective optical excitation of the dots (Fig. 4-17(a)) the

photogenerated holes are removed by the electric field in the p-i-n junction.

 The electrons remain stored since their motion in the z-direction is prevented
by the wide asymmetric tunnel barrier (Fig. 4-17(b)).

 The stored electrons in the quantum dot layer then selectively deplete the 2D
electron channel. This can be detected electrically by monitoring the channel
lateral resistivity (R).

 The stored electrons are removed and the device reset by injecting holes into
the quantum dot layer from the p-contact where they become captured by the
quantum dots recombining with the stored electrons (Fig. 4-17(c)).

 The system then reverts to its uncharged state (Fig. 4-17(a)) and a new write,
read and erase cycle can be performed.

Results:

Fig. 4-18 shows the temporal dependence of the channel resistance for such an
illumination, recovery and reset cycle.

Fig. 4-18.
 30

 Following illumination with photon energies > 1.02eV a rapid increase of the
channel resistance (ΔR) is observed.

 This effect is not observed in a separate reference sample which does not
contain quantum dots or following excitation at energies below the quantum
dot photoluminescence emission. This confirms that the effect arises from
electron storage in the quantum dot layer.

 Upon switching off the illumination (D), the photoeffect is persistent over
times cales longer than 8 hours at 140K.

 After the sample reset (R), the channel resistance returns to the pre-
illumination value → quantum dots are emptied.

Fig. 4-19 (dotted curve) shows the percentage photo-induced change of the
channel resistance (%ΔR/R) as a function of the excitation energy.

Fig. 4-19.

 The magnitude of the induced photoeffect increases when the quantum dot
containing p-i-n junction is reverse biased and the n-channel becomes
selectively depleted.
 31

 Resonances are observed in the ΔR/R excitation spectrum (A0, A1 and A2)
which are shifted from the quantum dot ground state PL emission - see Fig.
d29 (solid curve).

Resonances reflect LO phonon assisted excitation of the QD ensemble device is

sensitive to selective optical excitation.

4.5. APPLICATIONS OF COLLOIDAL QUANTUM DOTS

4.5.1. Introduction

As we already know, squashing the exciton into a smaller space makes it more
energetic through what is known as quantum confinement, so with smaller and
smaller nanocrystals, higher and higher energy excitons are produced. This
phenomenon is illustrated in Fig. 4-20, which shows a rainbow of fluorescent colors
all produced by QDs of various sizes. The energy level quantization at the band edge
similar is shown by the distinct sharp peaks in the absorption spectrum shown in
Fig. 4-21.

Fig. 4-20. (a) Colloidal CdSe QDs with a

range of fluorescent colors based only on
variations in particle size, excited by an
ultraviolet lamp. The QD size decreases
from the red (about 6 nm in diameter) to the
blue QDs (about 2 nm).

(b) Luminescence from solutions of InP

QDs of different sizes.
 32

The quantum confinement modifies also the properties of the impurity

wavefunctions, as it has been discussed in the case of QWs in Chap. 2.

Fig. 4-21. Absorption spectrum of CdSe

QDs (red) and the organic dye Rhodamine
640 (blue). The QD exhibits well defined first
and second exciton peaks around 525 and
425 nm, respectively. The individual
“molecular” like QD energy levels that
contribute to the first exciton peak are
shown as bars.

In several respects, QDs can potentially combine the best properties of small
molecule chromophores and bulk semiconductors. As fluorophores, high quality QDs
are extremely bright, with fluorescence efficiencies or quantum yields (QYs)
approaching unity, meaning that nearly every absorbed photon results in an emitted
fluorescence photon. In terms of photostability, individual organic molecules tend to
photobleach on the timescale of seconds or minutes while QDs can last significantly
longer, up to hours under constant excitation. To draw another parallel to organic
dyes, compare the absorption spectrum of CdSe QDs to that of Rhodamine 640 (a
high QY dye) also shown in Fig. 4-21. Any given dye only absorbs light in a very
narrow spectral window, while QDs absorb any photons higher in energy than their
bandgap. Dyes also tend to have modest Stokes shifts, meaning that the wavelengths
they emit are spectrally close to where they absorb, requiring high quality spectral
filters to remove excitation light in applications such as fluorescence microscopy.
While QDs also have small Stokes shifts, they can be excited at energies far above
their bandgap, which makes them much more versatile for some applications.
 33

Arguably even more important is that multiple QD emission wavelengths can be

collected from the same excitation wavelength. For someone with a mind to make a
QD sensor, QDs allow the possibility of simultaneous multiple analyte detection with
one excitation wavelength. In the last two decades, colloidal QDs have been made
from a variety of semiconductor materials and are optically active from the ultraviolet
to the near infrared portions of the electromagnetic spectrum. CdSe and CdTe, QDs,
which emit largely in the visible region, are joined by CdS, ZnSe, and ZnS QDs in
the ultraviolet, PbSe, PbTe, and PbS QDs in the near infrared. Cadmium and lead-
free visible and near infrared emitting QDs are also being studied, such as InP, InAs,
(though the tradeoff is lower toxicity for higher expense considering the rarity of
indium), and GaP. Generally, III-V colloidal particles are challenging to synthesize
and less extensively studied than II-VI or IV-VI materials. Despite their indirect
bandgaps in the bulk material, Si and Ge nanoparticles exhibit reasonably strong light
emission as well. Si QDs emit in the visible (QYs up to 20%), and Ge QDs in the
near infrared (QYs under 10%), although the mechanism for light emission in these
nanoparticles is still not well understood.

One common method for growing QDs involves combining highly reactive and
air-sensitive organometallic precursors at high temperatures. QD size is dictated by
how long the reaction mixture is heated (the longer the heating time, the larger the
QDs), so a range of sizes can be produced in one synthesis by pulling successive
aliquots out of the reaction vessel. Reactant to product chemical conversion yields
tend to be low, though recent work has shown significant improvement in that metric.
The final product of the synthesis reaction is a liquid, but QDs are actually in a
colloidal suspension rather than a solution, much like the protein and fat molecules in
a glass of milk. The outermost atoms of the nanocrystal are coordinated to organic
ligand molecules, which allow the QDs to interact with the solvent without
precipitating. However, new progress has been made on development of inorganic
capping ligands as well. With the right chemistry, ligands can be changed, like a set
of clothes, to suit the environment. If one would like QDs in water rather than an
 34

organic solvent, protocols exist to exchange from hydrophobic to hydrophilic ligands

(Fig. 4-22). Ligands also allow one to perform chemistry with the QD, as chemical
reactions on the outer ligand chemical groups can facilitate attachment of various
chemical and biochemical recognition elements such as antibodies or fluorescent
species.

In the last decade, the simple core QD particle has given way to studies of
more highly engineered structures. QDs are frequently overcoated with an outer
layer, or shell, of another semiconductor material, a few atoms thick. This shell
serves to passivate any dangling bonds on the core QD surface and prevents the inner
QD core from photodegradation processes such as reaction with oxygen. Using a
higher bandgap semiconductor as the shell in core/shell particles can also improve the
optical properties of a QD by causing tighter confinement and localization of the
exciton in the QD core, offering dramatically increased QY and improved
photostability. Alternately, “type-II” core/shell structures promote carrier separation
by engineering band offsets to necessitate charge transfer at the core/shell interface.
This structure creates tunable infrared emission even when both components are
visible emitting materials, while the ability to direct carriers is beneficial for device
applications.
 35

Fig. 4-22.
(top) A typical QD with hydrophobic surface
capping ligands (tri-octylphosphine oxide)
and
(bottom) a QD after a ligand exchange to a
hydrophilic ligand.

4.5.2. Applications in biology

QDs have been used for a wide variety of biological applications, such as
staining and lighting up cells for visualization or detecting a specific analyte. QD
bioconjugates can specifically bind to cancer cells or tag bacteria such as E. coli and
accurately differentiate between pathogenic and harmless strains. QDs are available
for purchase, suspended in water and capped in several kinds of ligand chemistries,
ready for functionalization. QDs can even come already coated with proteins such as
streptavidin, or some common antibodies, enabling attachment to a protein coated
substrate, an antigen, or simple conjugation to a biomolecule of your choice.
 36

For the short term, QDs will be found only under the microscope or on the
bench-top rather than in the patient. In vivo imaging is being vigorously pursued,
especially for particles with near infrared emission wavelengths which have good
transmission through tissue and are well separated spectrally from background
autofluorescence. However, questions of cytotoxicity still remain unresolved, and
conclusions about the relative harmful effects of QDs vary greatly depending on
specific QD surface coatings. Studies with animal models suggest that particles
around 5 nm in diameter (and smaller) are able to pass quickly out of the body in the
urine. But most QDs suspended in water, considering their ligands and protective
inorganic shells (which help prevent the leeching of QD core metals into the body in
addition to protecting the QD), tend to be bulkier than this. The ideal particle for in
vivo applications, therefore, is extremely small, emits between 700 and 1000 nm, and
is either exceptionally stable or made of non-toxic elements. For example,
InAs/ZnSe, core/shell particles, InAs/InP/ZnSe core/shell/shell particles, PbSe
nanoclusters and CuInSe ternary particles show promise in fulfilling some of these
parameters.

A major breakthrough in biomedical imaging would come from the ability to

follow single proteins inside a live cell in real time, for minutes to hours. In principle,
QD fluorescence is bright and robust enough to potentially make this long standing
biological “holy grail” a reality. However, individual QDs exhibit fluorescence
intermittency, or “blinking,” a feature common to most single fluorophore molecules.
Under constant excitation, a single QD will switch between a bright “on” state and a
dark “off” period.

Early on, it was hypothesized that trapped surface charges could play a role in
the blinking. Indeed, single charges on QDs as observed through electrostatic force
microscopy also exhibited blinking behavior, suggesting that surface charge traps
may play a role in creating the off state. However, the observation of fluorescence
from charged QDs complicates matters, and the complexity of blinking statistics
necessitates dynamic models to approximate experimental observations. While
 37

blinking is a significant hurdle to overcome in order for QDs to be useful as single

photon emitters and single particle trackers, recent results show that it can be
suppressed both by growing an extremely thick CdS shell on CdSe, and by creating
“alloy” QDs, where the core and shell materials make a gradual transition rather than
having an abrupt boundary. However, these steady emitting particles are still the
exception rather than the rule.

4.5.3. Applications in devices

4.5.3.1. Solar cells

The theoretical upper limit (i.e., Shockley-Queisser limit) for the energy
conversion efficiency of a Si photovoltaic cell is about 30%. In practice, the best Si
devices have efficiencies around 25%. Additional efficiency can be gained by using
layers of different materials tailored to absorb different parts of the solar spectrum in
multi-junction cells, with three junction cells exceeding 40% efficiency. So why
haven’t solar cells sprouted up on every rooftop yet? In part, it comes down to
expense. Semiconductor photovoltaics capable of those high efficiencies are fairly
costly to produce and install. A pocket-sized solar cell that can charge a cell phone or
MP3 player in a few hours retails for around $40. Considering the current cost of
electricity, it would take over 4000 charges to break even compared to plugging in!

Fortunately, there are potentially more inexpensive options in development,

like a dye-sensitized TiO2 nanoparticle based Graetzel cell which boasts a very
respectable 11% energy conversion efficiency and easy, inexpensive solution-phase
processing. Since colloidal QDs share the advantage of easy processing with dyes,
their broadband absorption properties (recall that instead of the tiny window of the
solar spectrum accessible to a dye, QDs gobble up any photons higher in energy than
their bandgap) can be harnessed with Graetzel cell-type architectures. “Rainbow
cells” use a variety of QD sizes, taking advantage of the fact that while larger QDs
have a better match to the solar spectrum and can collect more light, for smaller QDs
 38

energy bands line up better to transfer that energy to the wide bandgap TiO2. Close
packed films of QDs can be used as the basis for a solar cell on their own, though
overall device efficiencies remain quite low. Additionally, photovoltaics can be made
by incorporating QDs into conducting polymer matrices. Quantum rods, the
elongated relatives of the QD, may prove to be a superior choice for these types of
devices due to improved charge separation along the long axis of the particle.

A factor that could help overcome the fundamental Shockley-Queisser limit is

the phenomenon of multiple exciton generation (MEG), or carrier multiplication, see
Fig. 4-23, resulting in a higher photocurrent, and giving the overall device efficiency
quite a boost. It remains a matter of some debate whether quantum confinement
effects cause an enhancement of MEG efficiency such that it will have considerable
impact on practical applications. However, recent prospects look fairly positive since
a PbS-sensitized photovoltaic device has been shown to produce photocurrents
corresponding to greater than one electron created per high energy photon absorbed.
At any rate, even those skeptical of MEG can agree that QDs have the potential to
improve solar cells, at least by the ability to better match the solar spectrum due to
QD bandgap tunability.
 39

Fig. 4-23.
(a) A high energy photon is
absorbed by a QD, creating one
exciton while excess energy is
lost to fast phonon relaxation as
heat.
(b) A high energy photon is
absorbed by a QD, creating two
excitons through the MEG
process.

4.5.3.2. Light-emitting diodes

Light-emitting diodes (LED) are superior to both incandescent and fluorescent

bulbs in terms of energy efficiency, lifetime, and robustness. Since LEDs are
semiconductor devices, the color of light produced is characteristic of the bandgap of
the material, and (similar to QDs) spectrally narrow. Much in the same way that
fluorescent light bulbs convert ultraviolet radiation into visible light, blue or
ultraviolet LED bulbs can be coated with phosphors for white light emission. White
LED lamps and bulbs currently on the market have at least four times the luminous
efficacy (lumens/Watt) of incandescent bulbs. However, the clear advantage in the
efficiency arena is counterbalanced by a deficiency in the color quality of most white
LEDs. Based on the color rendering index (CRI) scale, typical white light LEDs only
score about 75 out of 10065; and their light appears cold and harsh. One method to
improve the CRI score employs multiple sizes of QDs as phosphors to produce white
light. These downconversion (high energy light to lower energy light) devices have
even made it out of the lab: currently there are QD-LED lamps for sale commercially
 40

with a CRI of 9065. Similar to QD photovoltaics, QD films for downconversion

devices are easily made with solution processing.

While direct electroluminescence from QDs is certainly possible and a hot area
for research, device efficiencies have not yet reached a point where they are viably
competitive with existing bulk semiconductor or organic LED (OLED) technologies.
In addition to efficient solid-state lighting, QD emission can be applied to the
production of QD-LED displays. For these full color organic/nanoparticle hybrid
devices, QD films can be created by inkjet printing or contact printing (Fig. 4-24) for
a downconversion device, or by solvent free ink/stamp transfer for direct QD
electroluminescence with a pixel size comparable to existing high definition displays.
Both displays have been successfully created on flexible substrates. The former
features color saturation and purity comparable to the International
Telecommunications Union HDTV standard, though QDs have the potential to
exceed that standard.

Fig. 4-24.

(A) The schematic of the contact

printing process.

(B) (a) Electroluminescent red and

green QD pixels, fabricated on the
same substrate. A blue pixel is a
result of TPD emission in the area
without QDs.

(b) QD-LED pixel, patterned with

25 μm wide stamp features.
 41

As other electrically driven LEDs, NP-LEDs can be as well divided into those
operated at direct (DC) and at alternating (AC) currents. Typically, the light emission
processes differ in these two cases demanding different device architecture types and
involved material systems. The research on these new device concepts and their
development started in 1994 and rapidly proceeded since then. First LEDs based on
nanoparticles and operated under DC were presented historically earlier than AC
devices, mainly pushed by the OLED technology and the development of QDs with
high room temperature quantum efficiency (QE) and the availability of processible
stable dispersions.
 42

4.6. SUMMARY OF THE PROPERTIES OF HETEROSTRUCTURE

QUANTUM WIRES AND QUANTUM DOTS

4.6.1. Basic physical phenomena

 1D electron gas (wires)

 Density-of-state function with sharp maximums (wires)

 0D electron gas (dots)

 δ-function type of density-of-state function (dots)

 Increasing binding energy of excitons.

4.6.2. Important applications in electronics

 Reduced lasing threshold current and larger differential gain

 Reduced temperature dependence of threshold current (wires)

 Temperature stability of the threshold current (dots)

 Discrete amplification spectrum and a possibility of obtaining performance

characteristics similar to those of solid-state or gas lasers (dots)

 Higher modulation factor in electro-optical modulators

 Possibility of creating ‘‘single-electron’’ devices

 A new possibility for the development of field-effect transistors.

4.6.3. Important technological peculiarities

 The application of self-organization effects for growth

 Epitaxial growth in V grooves (wires)

 High-resolution lithography of heterostructure quantum well lasers.