File PDF

Table of contents
List of Figures
iv
List of Tables
viii
List of Publications
ix
Acknowledgements
1 Introduction
2 Negative refraction of light at the edges of one and two-dimensional

photonic crystals.
2.1 Review of negative refraction. . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Scattering of light at the side-edge of a Bragg reector. . . . . . . . . . . .
2.2.1 Calculation method . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.2 Comparison of the analytical theory and numerical modelling results
2.2.3 Results and conclusions . . . . . . . . . . . . . . . . . . . . . . . .
2.3 Negative refraction of THz radiation by a prism made of a two-dimensional
metallic photonic crystal . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1 Calculational method . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.2 Numerical modelling results . . . . . . . . . . . . . . . . . . . . . .
2.3.3 Comparison of theory and experiment . . . . . . . . . . . . . . . .
2.4 Summary of key original results . . . . . . . . . . . . . . . . . . . . . . . .
3 Eigenmode analysis of resonant photonic nanostructures
3.1 Background theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.1 Transfer matrix method . . . . . . . . . . . . . . . . . . . . . . .
3.1.2 Tamm plasmon eigenfrequencies . . . . . . . . . . . . . . . . . .
3.1.3 Reection coecient of an exciton quantum well . . . . . . . . .
3.2 Parabolic polarization splitting of Tamm states in a metal-organic microcavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1 Structure and experimental set up . . . . . . . . . . . . . . . . .
3.2.2 Comparison of results of numerical modelling with experiment .
3.2.3 Analytical expressions for the system eigenfrequencies . . . . . .
3.3 Eigenmode structure of periodic array of quantum wells buried in layers
of one-dimensional photonic crystal. . . . . . . . . . . . . . . . . . . . .
6
7
10
10
17
19
21
21
24
26
30
.
.
.
.
31
33
33
36
38
.
.
.
.
40
40
42
43
. 48
Contents
3.4
3.5
Parametric amplication by a periodic array of quantum wells

in layers of a one-dimensional photonic crystal. . . . . . . . .
3.4.1 Parametric amplication of exciton-polaritons . . . . .
3.4.2 Experimental measurements and results . . . . . . . .
3.4.3 Three-level model of parametric amplication . . . . .
Brief summary . . . . . . . . . . . . . . . . . . . . . . . . . .
ii
embedded
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
.
.
.
.
.
4 Kinetics of exciton-polariton condensates and the modelling of polariton lasers

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Background theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.1 Bose-Einstein condensation of exciton-polaritons . . . . . . . . . .
4.2.2 Polariton lasing . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3 Designs of the polariton lasers considered . . . . . . . . . . . . . . . . . .
4.4 Modelling formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4.1 Semiclassical Boltzmann equations . . . . . . . . . . . . . . . . . .
4.4.2 Scattering rates in polariton relaxation processes . . . . . . . . . .
4.5 Numerical results and discussion . . . . . . . . . . . . . . . . . . . . . . .
4.5.1 Simplied rate equation modeling: steady-state solutions . . . . .
4.5.2 Simplied rate equation modelling: high-speed current modulation treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
56
57
57
61
63
63
64
64
66
68
70
70
73
80
83
85
91
5 Summary, conclusions and further work
92
References
96
List of Figures
2.1
2.2
2.3
2.4
2.5
A schematic diagram of the refraction of light at the side edge of a Bragg

reector and the Cartesian coordinate system used in the theory. The
structure has alternate layers with thicknesses d1 and d2 and respective
refractive indices n1 and n2 . Light is shown incident from vacuum at angle
with wavevector K0 parallel to the interface. Also shown is a refracted
wave with wavevector components K0 and , respectively, parallel and
normal to the interface. . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dependence of the norm of the transmitted amplitudes vector on the
number of equations. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
(a) Amplitudes of the eigenmodes for the case of a photonic crystal with
refractive indices 1.5 and 1.7. The incident light frequency is 1.5 eV, and
the angle of incidence is 60 degrees. (b) C1 /C2 ratio vs relative contrast
of the photonic crystal. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
(a) Isofrequency contours for the photonic crystal with refractive indices
1.5 and 1.7 for a frequency of 1.6 eV. Branches which correspond to negative values of Sz K0 are depicted by blue lines and to positive values by
red lines. Dashed arrows show the directions of the negatively refracted
modes, and solid arrows the directions of the diracted modes. (b) Isofrequency contours for the photonic crystal with refractive indices 1.1 and
2.1 at a frequency 1.6 eV. In each case isofrequency contours for isotropic
media with the eective refractive indices of the photonic crystal are depicted by dashed semicircles. The direction and absolute of the tangential
component of the wavevector for the incident wave is shown with red arrow for all the cases. The directions of group velocity and energy ow are
shown with green arrows for all the cases. For all the cases the dotted
lines show the value of the in-plane component of the incident wave (upper line) and the value of the in-plane wavevector of the diracted wave
(lower line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
FDTD simulation results for waves scattered at the interface of high
(n1 = 2.1, n2 = 1.1) (a,b) and low (n1 = 1.4, n2 = 1.8) (c,d) contrast
photonic crystal. (a) - Spatial distribution of the square of electric eld
for the case of negative refraction (~ = 1.6eV, angle of incidence = 55
degrees);(b) Spatial distribution of the square of electric eld for the case
of normal propagation (~ = 2.0 eV, = 55 degrees); (c) Spatial distribution of the absolute value of the square of electric eld for the case of
spatial oscillations of the Poynting vector (~ = 2.0 eV, = 55 degrees);
(d) Spatial distribution of the real part of the electric eld for the case
described in (c). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iii
. 11
. 14
. 14
. 16
. 18
List of Figures
(a) Band diagram of a hexagonal photonic crystal formed by gold cylindrical rods with diameter 80 m. Period of the structure is 200m. fp is
an eective plasma frequency, which is a cut-o frequency for the propagating solutions. The group velocity is collinear with the wavevector in
the rst band (red), which corresponds to positive refraction. The second band (blue), where the group velocity has opposite direction to the
wavevector is the source of negative refraction. (b) Isofrequency contours
for the frequencies close to high frequency edge of the second band. . . .
2.7 Squared electric eld for waves refracted on a photonic crystal prism.
(a) Wave frequency f = 0.8 THz, positive refraction. (b) f = 1.2 THz,
negative refraction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.8 The initial design of the photonic crystal prism. The separation of the
rods on the right-hand facet is dierent from that in the bulk. . . . . .
2.9 Micrograph of the photonic crystal prism studied in the experiment. Arrows have been added to show the directions of the negatively (blue) and
positively (red) refracted beams. . . . . . . . . . . . . . . . . . . . . . .
2.10 (a) Temporal dependence of the reference signal. Inset shows the Fourier
transform. (b) Temporal dependence of the negatively refracted beam.
2.11 (a) Fourier transforms of the refracted signals. Horizontal lines show the
widths of the rst and second bands, calculated by the complex band
structure method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iv
2.6
3.1
3.2
3.3
3.4
3.5
3.6
. 24
. 25
. 25
. 26
. 27
. 28
The simplest realization of the structure where a Tamm plasmon mode

can exist. A semi-innite Bragg reector (grey layers of dielectric permittivity 1 and 2 ) adjacent to a semi-innite metal layer (yellow) with
dielectric permittivity M . The thickness of the dielectric layer adjacent
to the metal may be varied to tune the frequency of the mode. . . . . . .
Design of the structure investigated. Two Bragg reectors enclose the
cavity layer and the silver layer of variable thickness (gradient from 0
to 40 nm). The thicknesses of the layers being most sensitive to the
eigenmodes are denoted by di . . . . . . . . . . . . . . . . . . . . . . . . . .
Experimentally observed angle-resolved emission spectra (left panels) along
with numerically calculated spectra (right panels). Deviations of the
shape of the parabolae are due to imperfect refractive indices and their
dispersions, on which the simulations are based. . . . . . . . . . . . . . . .
(a) For the special case of a 40 nm silver layer inside the cavity, analytical
results (black solid and red dashed line) showing good agreement with the
numerical results. (b) Parabolic dependence of the splitting between the
TE- and TM-modes of the shifted cavity resonance and Tamm plasmonpolariton as a function of the angle of incidence. . . . . . . . . . . . . . .
(a) Geometry of the structure under consideration. (b) Experimental
reection spectra. (c) Reection spectra calculated with transfer matrix
method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Band structure of the photonic crystal without quantum wells. Red lines
show the dispersions of the propagating photonic modes a
(q),
a (q),
black lines show the dispersions of the Bloch modes U ,L. Right picture
shows the reection spectrum from the nite structure. Reection coecient minima correspond to discrete Bloch modes, which are characterized
by frequencies U 1 , U 2 , etc. . . . . . . . . . . . . . . . . . . . . . . . . .
37
41
43
47
49
53
List of Figures
Dependence of the system eigenfrequencies on the detuning between the
exciton energy and photonic band edge, obtained from experiment, transfer matrix method and with the analytical model. . . . . . . . . . . . .
3.8 Eigenmode dispersion in the case of tuning of the exciton mode to the
upper edge of the second photonic band. Exciton frequency - 1.96 eV.
Refractive indices of the layers n1 = 3.5, n2 = 3.2, and thicknesses of the
layers d1 = 133 nm, d2 = 43 nm. . . . . . . . . . . . . . . . . . . . . . .
3.9 Dispersion of the exciton-polariton. Arrows show how the two pump
polaritons p scatter elastically into a ground state s and an idler state i.
3.10 An illustration of parametric amplication. (a) Reection spectrum in the
case of a single probe pulse (violet line). The green line corresponds to a
50 mW pump pulse, which is tuned to the ground polaritonic branch with
wavevector equal to the magic wavevector (corresponding to an incidence
angle 20 degrees). In this case there is strong amplication of a probe
pulse (red line). (b) Colour map showing the dependence of the reection
on the frequency and delay time. (c) Dependence of the probe pulse
enhancement on the delay time. . . . . . . . . . . . . . . . . . . . . . . .
3.11 Numerical modelling of the experiment, performed by the Cambridge
group, with the sample considered in section 3.3 demonstrating parametric amplication in the system. (a) Time dependence of the pump state
(red line) and signal state (blue line) occupancies in the case when the
pump and probe pulse arrive at the same time t = 3 ps. Blue dashed line
shows the evolution of the signal state in the absence of the pump pulse.
All the values are normalized to the maximum occupancy of the signal
state in the absence of a pump pulse. t shows the nonlinear response
time of the system. (b) Dependence of the probe pulse amplication on
the pump-probe delay time. . . . . . . . . . . . . . . . . . . . . . . . .
3.7
4.1
4.2
4.3
4.4
. 54
. 54
56
. 58
. 61
(a) Schematic diagram of a polariton laser. (b) Schematic illustration of

the kinetic processes in the system: pump to the excited states, relaxation
to and from the ground state, and coherent emission from the ground state
due to the nite polariton lifetime. . . . . . . . . . . . . . . . . . . . . . 67
Schematic 3D cross-section of an InGaN/GaN MQW polariton LD based
on an intracavity pumping geometry. The structure is sandwiched between bottom (shown in dark blue/light blue) and top (shown in green/white)
DBRs. The polaritons are formed in the QWs-2 region (shown in orange/light blue). In the case of intracavity pumping, the QWs-2 are pumped
with the optical emission from the electrically pumped QWs-1. . . . . . . 70
Results of a transfer matrix simulation of the eld intensity prole of
a polariton LD shown in Fig.4.2 centred at = 415 nm along with the
corresponding refractive index prole. . . . . . . . . . . . . . . . . . . . . 71
(a)-(c) Occupancy of the polariton ground state vs pump current density
calculated for the two pumping geometries at various temperatures and
detunings (see text for details). . . . . . . . . . . . . . . . . . . . . . . . . 81
List of Figures
4.5
4.6
4.7
4.8
Plots of Jthr vs detuning and temperature for (a) the direct electrical and
(b) the intracavity optical pumping schemes. The red dashed line on each
plot corresponds to the evolution of Jthr,min as a function of temperature.
(c) Evolution of the optimum detuning (black solid and black dashed lines)
and condensation threshold current density (red solid and red dashed
lines) at the optimum detuning as a function of lattice temperature for the
direct electrical (solid lines) and the intracavity (dashed lines) pumping
schemes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Averaged scattering rates a, b, and c as a function of detuning obtained
by tting the full semi-classical Boltzmann system of equations at various
temperatures for the two pumping schemes: electrical (solid lines) and
intracavity (dashed lines) pumping. . . . . . . . . . . . . . . . . . . . . .
Left-hand side vertical scale: evolution of nx as a function of temperature calculated at the optimum detuning using the exact expressions for
the electrical (connected black dots) and the intracavity (connected black
circles) pumping geometries (see text for details). Right-hand side vertical scale: relative deviation between the exact and the approximated
expressions for the electrical (connected red dots) and the intracavity
(connected red circles) pumping geometries. . . . . . . . . . . . . . . . .
(a) Polariton condensate occupation number vs current density for the
electrical (red line) and intracavity (black line) pumping geometries determined at 320 K and at the optimum detuning. (b) Frequency dependence
( = /2) of the square modulus of the modulation transfer function,
|H()|2 . Each curve corresponds to one of the steady state solutions
indicated in Fig. 4.8(a). . . . . . . . . . . . . . . . . . . . . . . . . . . .
vi
. 81
. 84
. 86
. 88
List of Tables
4.1
Parameters of the modelling of polariton laser . . . . . . . . . . . . . . . . 80
vii
List of Publications
1. I.V. Iorsh, M.A. Kaliteevski, S. Brand, R.A. Abram, N.A. Kaliteevskaya. An
electromagnetic wiggler originating from refraction of waves at the side edge of a
Bragg reector. Journal of Modern Optics 58, 8, 683-693 (2011).
2. G. P. Swift, A. J. Gallant, N. Kaliteevskaya, M. A. Kaliteevski, S. Brand, D. Dai,
A. J. Baragwanath, I. Iorsh, R. A. Abram, and J. M. Chamberlain. Negative
refraction and the spectral ltering of terahertz radiation by a photonic crystal
prism, Optics Letters 36, 9, 1641-1643 (2011).
3. A. Askitopoulos, L. Mouchliadis, I. Iorsh, G. Christmann, J.J. Baumberg, M.A.
Kaliteevski, J. Hatzopoulos, P.G. Savvidis. Bragg polaritons: strong coupling and
amplication in an unfolded microcavity, Physical Review Letters 18, 107, 076401
(2011).
4. R. Bruckner, M. Sudzius, S. I. Hintschich, H. Frob, V. G. Lyssenko, M. A. Kaliteevski, I. Iorsh, R. A. Abram, A. V. Kavokin, and K. Leo. Parabolic polarization
splitting of Tamm states in a metal-organic microcavity, Applied Physics Letters,
100, 062101 (2012).
5. I. Iorsh, M. Glauser, G. Rossbach, J. Levrat, M. Cobet, R. Butte, N. Grandjean,
M. A. Kaliteevski, R. A. Abram, and A. V. Kavokin. Generic picture of the
emission properties of III-nitride polariton laser diodes: Steady state and current
modulation response, Physical Review B, 86, 125308 (2012).
viii
Acknowledgements
I would like to express my deepest appreciation to my supervisors, Professor Richard
Abram and Doctor Stuart Brand. Without their persistent guidance and help this book
would not have been possible.
I would like to thank Professor Alexey Kavokin, who is a coordinator of Clermont4
project, which provided nancial support of my research.
In addition it gives me pleasure in acknowledging the support of Professor Mikhail
Kaliteevski, who continually and convincingly conveyed a spirit of adventure in regard
to research and scholarship.
I share a credit of my work with our collaborators from University of Dresden, Cambridge
University, University of Crete, and Ecole

Politechnique Federale de Lausanne.
Finally, I am indebted to my wife, my parents, and friends for their help and encouragement.
ix
Chapter 1
Introduction
A key trend in information technology over the last fty years has been miniaturisation.
The characteristic size of devices has been decreasing constantly as recognized by the
famous empirical Moores law, which states that the number of transistors on integrated
circuits doubles approximately every two years, and which has been valid with only minor
variations for the last forty years. However, it is quite evident, that the characteristic
sizes of the elements of conventional electronic integrated circuits have almost reached
their physical limit, and that further reduction is precluded not by fabrication technology
but rather by fundamental physical limitations. In order to continue the speed of growth
of computational power of modern computers, new concepts of the computation must
be employed. Such concepts include quantum computers, biological computers and,
most relevant to this book, optical information processing. For the physicist, the main
challenge posed by this trend is how best to describe systems and model devices operating
at the boundary of classical and quantum domains. It is in this context that this book
addresses some topical problems in optical structures and devices that are concerned
with the control of light and light-matter interaction.
We begin by describing in rather general terms some concepts in the electromagnetic
theory of optical structures and in the physics of the light-matter interaction that are
relevant to the research reported in the book. We also give a very brief introduction to
the research problems that have been addressed and describe how the book is structured.
A more detailed introduction to each research topic is given in the appropriate chapter.
Electromagnetic radiation obeys Maxwell equations:
D = 4,
B = 0,
1 B
,
c t
1 D 4
H=
+
J,
c t
c
E=
(1.1)
Chapter 1. Introduction
where respectively E and H are the electric and magnetic elds; D and B are the
electric displacement eld and magnetic inductance; and J are the charge and current
densities, and c is the speed of light.
There are no external charges or currents in the optical structures considered in the book
and it is assumed that the structures do not have any magnetic response. Consequently,
Maxwell equations can be reduced to
E=
1 2D
,
c2 t2
(1.2)
which governs the optical properties of all the structures considered in the book. However, depending on the behaviour of D, two general types of structure can be distinguished in the work reported. In one type of structure the electric displacement is dened
only by a spatially-varying, frequency-independent dielectric permittivity through the
relation D(r, t) = (r)E(r, t). In this case we can look for solutions in the form E eit
and equation (1.2) can be written as
E(r) =
2
(r)E(r).
c2
(1.3)
The solutions of this equation obey a simple scaling law: if we decrease all the dimensions
of the system by a factor of then to get the same distribution of the elds we must
increase the frequency by a factor of .
The idea that periodic modulation of the dielectric permittivity can determine certain
optical properties of a structure was recognized by Bragg [1] when he explained the
resonant peaks in the X-ray reection spectra of crystalline solids. In this work Bragg
modelled the crystal as a set of parallel planes separated by a distance d and explained
the reection peaks by the constructive interference of X-rays reected from those planes.
The condition for constructive interference is the well known Braggs law
n = 2d sin ,
(1.4)
where is the wavelength, is the incidence angle and n is an integer. In such Bragg
diraction, the X-ray wavelength is comparable with the crystal interplane distance
d.
The scaling of Maxwells equations discussed earlier suggests that in order to obtain
Bragg diraction in the optical and infra-red frequency range, the period of the structure should be of the order of 100 nm. Although such structures have been studied
in one form or another since 1887 [2], the term photonic crystal was only used for the
rst time some 100 years later, when Eli Yablonovitch and Sajeev John published two
milestone papers on periodic dielectric structures in 1987 [3, 4]. The long delay between
the initial work and the advent of the rapid emergence of the new eld of solid state
physics dedicated to photonic crystals can be explained by the fact that the fabrication of a wide range of suitable structures only became possible in the 1980s following
substantial progress in semiconductor technology. Moreover, the development of semiconductor optoelectronic devices such as light-emitting diodes (LEDs) and lasers, and
a growing interest in integrated optics led to the need for new ways to control light on
a microscopic scale, and photonic crystals provided such a possibility. The principal
property of photonic crystals is their ability to modify the propagation of light. The
propagating modes in photonic crystal can be described in reciprocal space in much the
same way as the electronic Bloch states of a semiconductor. The frequency dispersion of
the modes can be substantially dierent from a homogeneous dielectric. Furthermore,
there can be frequency gaps in the band structure for which light propagation is not
possible. Photonic crystals are now used in a wide range of optoelectronic applications
from the anti-reection coatings for solar-cells to increasing the eciency LEDs [5].
Chapter 2 of the book is dedicated to studies of the scattering of light at the interfaces
of one- and two-dimensional photonic crystals. A new eect of Poynting vector beating
of the light refracted at the side edge of one-dimensional photonic crystal is presented
and also a design for a spectral lter for the terahertz frequency range based on a twodimensional metallic photonic crystal prism is developed.
In deriving of equation (1.3) and the corresponding scaling laws from equation (1.2)
it was assumed that the dielectric permittivity is frequency-independent. However, if
we now introduce an additional polarization P(r, t) for which this assumption does not
hold, equation (1.2) becomes
E(r, t) +
1 2 P(r, t)
(r) 2 E(r, t)
=
.
c2
t2
c2
t2
(1.5)
It is clear that we cannot solve equation (1.5) without additional information on the behaviour of the polarization P. In the book only one special type of additional polarization
is considered, namely that of a quantum well exciton. Excitons are Coulomb-correlated
electron - hole pairs that can take dierent forms depending on the properties of the
host medium. A broad distinction can be made based on spatial size and binding energy,
and here we consider Wannier-Mott excitons, characteristic of the common semiconductors and having relatively large size (several nanometers) and relatively small binding
energy (a few tens of meV) [6]. Excitons can be observed experimentally by sharp peaks
in the absorption spectrum of the semiconductors close to the fundamental absorption
edge. The energy spectrum of excitons in bulk semiconductors is like that of a hydrogen
atom with the following renormalization: m0 , and e2 e2 /, where m0 is the
free electron mass, and is the reduced exciton mass. As a result of spatial connement, excitons in quantum wells, wires and dots have properties rather dierent from
bulk excitons, including qualitatively dierent energy spectra, larger binding energies
and smaller eective localization lengths. The polarization induced by a quantum well
exciton can be written in the frequency domain as [7]:
P(, z) =
(, z, z )E(, z )dz .
(1.6)
Here z is the direction perpendicular to the quantum well plane, and (, z, z ) is the
nonlocal susceptibility given by:
=
Q
(z)(z )
0 i
(1.7)
where is the exciton wavefunction in the case when the electron and hole are at the
same position in the well plane, Q is a quantity dened by the material properties of
the quantum well, 0 is the exciton frequency, and is the exciton line broadening.
Substitution of the polarization into equation (1.5) shows that the eigenfrequencies of
the structures are no longer only dened by the geometry and dimensions of the photonic
structures but also by the exciton resonance frequency 0 , precluding the simple scaling
relation described earlier for non-resonant photonic crystals.
Placing the quantum well in the centre of an optical cavity and tuning a cavity resonant
mode to the exciton resonant frequency can lead to the formation of a new type of
eigenmode, an exciton-polariton, which in some circumstances can be considered to be
a composite light-matter quasiparticle with bosonic properties [8].
The fabrication of the structures where the exciton-polaritons can be excited is an
extremely technologically challenging process, since it requires both very high quality Bragg mirrors, providing long cavity photon lifetimes, and high quality quantum
wells with small exciton broadening. As a result, exciton-polaritons have hitherto been
experimentally studied by only a quite limited number of scientic groups, despite their
fascinating fundamental properties and possible applications. A particular attraction
of exciton-polaritons is their bosonic nature combined with an extremely small eective mass, which makes possible their condensation at relatively high temperatures, and
the observation of related phenomena such as superuidity. A condensate of excitonpolaritons also emits coherent light and there are intensive eorts to develop an electrically pumped polariton laser [9] based on the eect. However there are other potential
applications based on the relatively strong nonlinear eects resulting from the strong
interactions of the excitonic component which lead to the possibility of new types of
photonic devices such as parametric ampliers [10] and optical integrated circuits [11].
Also as a result of their composite light-matter nature, exciton-polaritons facilitate the
ecient control of quantum processes in semiconductor structures with light, including
the recently proposed possibility of light-mediated room-temperature superconductivity
[12] and quantum information processing [13].
Chapters 3 and 4 of the book are dedicated to the study of some optical properties of
exciton-polaritons. Chapter 3 considers the eigenmode structure and parametric amplication of exciton-polaritons in a one-dimensional photonic crystal with the quantum
wells embedded into its layers. The theory and calculations support the reectivity measurements and pump-probe experiments on these structures that have been performed
by collaborators. Moreover, it is shown that a slight change in the design of the structures can lead to the formation of a new type of exciton-polariton, dened by a negative
eective mass and group velocity.
The work reported in chapter 4 was largely motivated by our experimental collaborators
at the Ecole
Polytechnique Federale de Lausanne, who are working on the realization of
the rst room-temperature electrically-driven polariton laser. Such a laser is expected
to have a much lower threshold compared to the current conventional vertical cavity surface emitting laser diodes (VCSELs). Numerical modelling of the kinetics of prototype
polariton lasers has been performed and predictions have been made of designs providing
the best operational characteristics and of parameters such as such as threshold current
density and current modulation frequency.
Currently the study of photonic crystals and exciton-polariton physics is largely driven
by the technologists and experimentalists and most of the work reported in the book has
been done in collaboration with experimental groups. Where appropriate, experimental
details have been given in the relevant chapters and further information can be obtained
in the references provided. It is hoped, that the close link with experiment will give the
theory described in the book a good chance to benet future technology.
Chapter 2
Negative refraction of light at the

edges of one and two-dimensional
photonic crystals.
This chapter is dedicated to the study of certain aspects of negative refraction at the
edge of photonic crystals. It begins with a general review of the negative refraction
eect and specically negative refraction in photonic crystals. The rst investigation
reported concerns the study of the scattering of light at the side edge of an innite Bragg
reector and is presented in section 2. A semi-analytical approach based on the method
of moments is used to calculate the scattering. It is shown that the results of the semianalytical model predict two new eects: spatial oscillations of the Poynting vector for
the light refracted into the photonic crystal for the case of low-contrast refractive index
and normal propagation of the refracted beam for the case of high-contrast refractive
index. The results of numerical nite-dierence-time-domain (FDTD) modelling which
have been performed to verify the semi-analytical model are also presented 1 .
The third section of this chapter considers the refraction of terahertz radiation by a
prism made from a metallic photonic crystal and has been published in [15]. It has
been shown that due to the negative refraction eect, this structure can be used as an
eective terahertz low-pass or high-pass lter. The theoretical results based on band
structure calculation and (FDTD) modelling have been veried experimentally 2 .
The work described in section 2 has been published in [14] in co-authorship with a number of
colleagues. My contribution to the publication was in applying an established theory and in obtaining
the results from the semi-analytical and numerical modelling.
2
The work was carried out in collaboration with colleagues working on microfabrication and terahertz
technology who made the photonic crystal prisms and performed all the experiments. My contribution
to the work was in performing the numerical FDTD modelling in order to obtain the optimal photonic
crystal prism design.
Chapter 2. Negative refraction of light at the interface of one and two dimensional
photonic crystals.
2.1
Review of negative refraction.
The systematic study of the optical properties of media with simultaneously negative
dielectric permittivity and magnetic permeability was rst reported in [16]. In this paper
Veselago proposed the term left-handed medium for such materials. This is because if we
substitute the plane wave solution E, H exp[ikr it] into Maxwells curl equations,
we get:
k E = H,
c
k H = E.
c
(2.1)
(2.2)
In the case of positive and , the vectors E, H, and k form a right-handed basis, and a
left-handed basis in the case of negative and . However, Poyntings vector S = E H
always forms a right-handed basis with the vectors E and H; the group velocity is always collinear with Poyntings vector and the phase velocity is always collinear with the
wavevector. Thus, in the case of isotropic media, the group and phase velocities are
oppositely directed if and are negative. Such media are also usually characterized
by a negative refractive index n = , and as a consequence, with negative phase

velocity. It should be noted, however that Maxwell equations do not contain the refractive index per se, and thus the right choice of the sign of the refractive index could be
subject to some discussion [17]. Pendry [18] presents the following argument in favour of
choosing the negative sign: in realistic physical media losses are always present, which
means that both the dielectric permittivity and magnetic permeability have nite imaginary parts: = + i ; = + i . The refractive index in this case is given by:
n = ( ) i| + |. However, in order to comply with the causality

principle, the imaginary part of the refractive index should be positive, and this condition corresponds to the negative sign of the real part of the refractive index. Thus, the
presence of losses is a necessary condition for the right choice of the sign of the refractive
index [19]. However, we should note that the arguments provided by Pendry are only
valid if we dene the refractive index as n2 = , and there are several alternative methods of introducing the refractive index, such as utilizing Snells law n = sin 0 / sin ,
where 0 are are the angles between the wavevector and the normal to the interface in
vacuum and in the medium respectively. We should also be careful substituting negative refractive index into known optical relations, since many of them have been derived
within the assumption of a non-magnetic medium [20]. For example, in the well known
expression for the Brewster angle, tan = n2 /n1 , the magnetic permeability of both
media is set to unity by default.
Frequency dispersion of the dielectric permittivity and magnetic permeability is a necessary condition for a negative refractive index [21]. Indeed, if the dielectric permittivity
and magnetic permeability were frequency-independent, the electromagnetic eld energy
density W = (E 2 + H 2 )/2 would be negative. However, in the presence of frequency
photonic crystals.
dispersion, W is given by [22]:

(
W =
)
() 2 () 2
E +
H /2,
(2.3)
and for certain type of frequency dispersion, W can be positive even for negative dielectric permittivity and magnetic permeability. For example, the dielectric permittivity of
a metal within the simplied Drude model is given by:
=1
p2
,
2
(2.4)
where p is the metal plasma frequency, and

p2
()
= 1 + 2,
(2.5)
which is always positive.

The above arguments are based on the presence of both losses and frequency dispersion
of the dielectric permittivity and permeability to obtain a negative refractive index.
However, we should remember that the Kramers-Kronig relations show that the presence
of losses implies frequency dispersion of the permittivity and vice versa.
Media with negative dielectric permittivity and magnetic permeability have been achieved
in metamaterials. A metamaterial is an ordered array of identical metallic elements, essentially artificial atoms. Just as in a conventional material where the dielectric permittivity and magnetic permeability are an averaged electromagnetic response is determined
by both the optical properties of the individual atoms and by the geometrical properties
of the lattice, and of a metamaterial are dened by the electromagnetic properties
of the individual elements and by their spatial arrangement. A key dierence is that
in the case of a metamaterial we can engineer the electromagnetic properties of the
constitutive elements by changing their internal structure. In particular, the metamaterial where both negative and were achieved for the rst time [23, 24] was a three
dimensional cubic lattice of metallic split rings. The electric response of this medium
was controlled by tailoring the lling factor of the metal and as a consequence changing
the eective plasma frequency [25]. The magnetic response was engineered by using
a resonant LC circuit in each element. In the vicinity of the resonance frequency, the
magnetic permeability of the medium was negative.
Negative refraction is the term used to describe the eect when an electromagnetic
wave, refracted at the interface between two media is transmitted on the same side of the
normal as the incident one and will occur when the refractive indices are of opposite sign.
However, a sucient condition for negative refraction is that the Poyntings vector and
wavevector are in opposite directions in one plane of one of the media and negative and
photonic crystals.
are not essential for negative refraction. In fact, negative refraction has been predicted
and observed in a wide variety of systems, including photonic crystals, metamaterials
and plasmonic structures [23, 24, 2629]. It leads to, for example, the possibility of
realizing a superlens, which could overcome the diraction limit [30], the possibility of
electromagnetic cloaking [31] and the formation of non-diracting beams [32]. However,
negative refraction will only lead to any of the above-mentioned eects if it is present
for a wide range of angles of incidence at the relevant frequency.
As noted, negative refraction has been predicted and observed in photonic crystals [33],
but there are certain dierences between negative refraction in metamaterials and photonic crystals. First of all, photonic crystals are usually strongly anisotropic media,
where it is dicult to introduce the concept of eective refractive index. This also
means that the vectors of the phase and group velocities are usually not parallel as in
the case of isotropic metamaterials. Moreover, even when a photonic crystal can be
treated as an isotropic medium (as to a good approximation in the case of a hexagonal
photonic crystal for the frequency ranges close to band edges), negative refraction is
realized not because of a negative refractive index, and thus a negative phase velocity,
but because of a negative group velocity as, for example , in the second band of a twodimensional photonic crystal. On the other hand, it is clear that the eect of negative
group velocity cannot be the result of simple folding of the photonic bands in the rst
Brillouin zone since negative group velocity could then be observed even in a photonic
crystal with an empty elementary cell, i.e. in a homogeneous dielectric or even in vacuum. Thus, negative group velocity and negative refraction require the existence of a
photonic band gap. Finally we note that another mechanism for negative refraction is
the diraction of light at a crystal interface.
photonic crystals.
10
2.2
Scattering of light at the side-edge of a Bragg reector.
Experiments demonstrating negative refraction at the side-edge of one-dimensional photonic crystals have been reported in [34] and the concept of a superlens [30] consisting
of two photonic crystals has been proposed in [35].
In this section, we provide a detailed theoretical study of negative refraction at the side
edge of a one-dimensional photonic crystal and consider whether such structures are
capable of providing useful optical eects.
2.2.1
Calculation method
The geometry of the problem under consideration is depicted in Fig. 2.1. A plane wave
in vacuum with wavevector of magnitude k0 is incident at angle on the side-edge of a
photonic crystal.
A good method of calculating the reected and transmitted electromagnetic waves for
this geometry has been presented in [36] and here we just describe the key steps of the
approach before applying it to the specic systems of interest.
We consider the TE polarisation case, for which electric eld vector is orthogonal to the
plane of incidence.
In order to obtain the eld distribution of the reected and refracted waves, we decompose the tangential components of the electric and magnetic elds into series of
modes both in vacuum and in the crystal and then apply the continuity conditions at
the interface.
Electric and magnetic elds in vacuum

The ycomponent of the electric eld in vacuum can be decomposed into the incident
plane wave and a series of reected plane waves:
Ey (x, z) =
p=
[exp(i0 x)p0 + Rp exp(ip x)] exp(iKp z),
(2.6)
p=
where K0 = k0 sin , Kp = K0 + 2p/D, p =
k02 Kp2 , the Rp are the amplitudes
of the reected waves, and p0 is a Kronecker delta. We can obtain the magnetic eld
photonic crystals.
11
k0
K0
K0
Fig. 2.1: A schematic diagram of the refraction of light at the side edge of a Bragg
reector and the Cartesian coordinate system used in the theory. The structure has
alternate layers with thicknesses d1 and d2 and respective refractive indices n1 and
n2 . Light is shown incident from vacuum at angle with wavevector K0 parallel to
the interface. Also shown is a refracted wave with wavevector components K0 and ,
respectively, parallel and normal to the interface.
component Hz (x, z) from the Maxwell equation E = (H/t)/c = ik0 H:

p=
p
1 Ey
Hz (x, z) =
=
[ exp(i0 x)p0 + Rp exp(ip x)] exp(iKp z).
ik0 x
k
p= 0
(2.7)
Electric and magnetic elds in the photonic crystal

The electric eld inside the photonic crystal can be written as the series:
Ey (x, z) =
Cm um (z) exp(im x),
(2.8)
where the um satisfy the continuity condition at the interfaces of the layers of the
photonic crystals and the Bloch theorem: um (z + D) = exp[iK0 D]um (z). Substitution
of Ey (x, z) into wave equation shows that the um (z) are solutions of the equation
d2 um (z)
2
+ k02 n2 (z) = m
um (z),
dz 2
(2.9)
where n(z) is the refractive index. m can then be obtained from the well known
dispersion equation for the eigenmodes of a photonic crystal:
1
cos kz1 d1 cos kz2 d2
2
kz1 kz2
+
kz2 kz1
)
sin kz1 d1 sin kz2 d2 = cos(K0 D),
(2.10)
photonic crystals.
12
where kz1 =
1 k02 2 , kz2 =
2 k02 2 . It has been shown that these equation
has an innite number of real solutions 2 , but only a nite number are positive and
thus correspond to the propagating solutions (real [36]). The corresponding u can be
written as:
u=
(d1 )
(x) + (x),
exp(iK0 D) (d1 )
(2.11)
where the functions (z), (z) are given by

{
=
cos(kz1 z), 0 z d1
cos(kz1 d1 ) cos(kz2 (z d1 ))
kz1
kz2
sin(kz1 d1 ) sin(kz2 (z d1 )), d1 z D

(2.12)
{
=
1
kz1
1
kz1
sin(kz1 z), 0 z d1
sin(kz1 d1 ) cos(kz2 (z d1 )) +
1
kz2
sin(kz1 d1 ) sin(kz2 (z d1 )), d1 z D

(2.13)
Applying the continuity conditions at the interface

The continuity of the electric eld at the side edge of the structure (x = 0) requires
p=
[p0 + Rp ] exp iKp z =
Cm um (z).
(2.14)
p=
Multiplying this equation by un and integrating over the period of the structure, and
using the orthogonality of the um we get:
Cn =
p=
[p0 + Rp ] Jpn
,
(2.15)
p=
where
Jpn
1
=
D
D
exp(iKp z)un (z).
(2.16)
The proof of the orthogonality of the set um is presented in [36] and is based on the fact
that the dierential operator governing the set um is Hermitian. Equation (2.15) can be
rewritten in the matrix form:
C = JH (R + F),
(2.17)
photonic crystals.
13
where C = [Cn ], J = [Jpn ], R = [Rp ] and F = p0 , and the superscript H denotes the
Hermitian conjugate. The continuity equation for the magnetic eld is
p=
p [p0 + Rp ] exp(iKp z) =
m Cm um (z).
(2.18)
p=
Multiplying this equation by 1/ D exp(iKq z) and integrating over the period gives:
q [Rq q0 ] =
m Cm um (z)Jqm ,
(2.19)
which can be written in matrix notation as

R = F 1 JCm ,
(2.20)
where = diag[q ], = diag[m ]. Then substituting R into (2.17),then gives

(
)
I + JH 1 J C = 2JH F,
(2.21)
where I - is the unit matrix. The resulting system of equations can be solved to obtain
the amplitudes of the transmitted eigenmodes. The derivation for the TM polarization
is carried out in a similar way.
The matrix equation (2.21) is an innite system of linear equations and a numerical
solution can only be obtained by truncating it. To dene the number of equations n to
n
be included, we rst introduce the sum sn =

|Ci |2 . The Ci set is ordered such that C1
i=1
corresponds to the eigenmode with the largest value of 2 . We assume that sn converges
to a constant value s, as n . The demonstration of the convergence of sn is shown
in Fig.2.2, where it can be seen that as the number of equations n is increased, sn
rapidly approaches some constant value. Assuming the convergence of sn , it is justied
to dene the truncation error in the following way: if Ci is the vector of solutions for
n eigenmodes, and Ci is the vector for n + 1 eigenmodes, then the truncation error
n+1
n
n can be dened as n =
|Ci |2
|Ci |2 . We then dene the maximum allowed
i=1
i=1
truncation error, which in our calculations has been set to 106 . It has been shown
that decreasing the maximum truncation error by an order of magnitude changes the
results in an essentially negligible way. The number of equations n
to be considered was
then dened as n
= minn (n : n < 106 ). The same condition has been applied to the
truncation of the reected amplitudes R.
In the numerical calculations we have considered quarter-wavelength Bragg reectors
with a stop-band frequency centre tuned to 1 eV. The values of |Cn |2 of some of the
photonic crystals.
14
1,0
i=1
|c |
0,9
0,8
0,7
0,6
11
16
21
26
31
Fig. 2.2: Dependence of the norm of the transmitted amplitudes vector on the number
of equations.
eigenmodes for a Bragg reector with the refractive indices of the layers equal to 1.7
and 1.5 are shown in Fig. 2.3.
Fig. 2.3: (a) Amplitudes of the eigenmodes for the case of a photonic crystal with
refractive indices 1.5 and 1.7. The incident light frequency is 1.5 eV, and the angle of
incidence is 60 degrees. (b) C1 /C2 ratio vs relative contrast of the photonic crystal.
The frequency of the incident light is 1.5 eV which corresponds to the second photonic
band of the crystal, and the angle of incidence is 60 degrees. In Fig.2.3(a) we can
see that there are two solutions which correspond to bulk waves propagating inside
the crystal ( 2 > 0) and other solutions corresponding to surface waves, which decay
exponentially in the direction normal to the interface ( 2 < 0). Note that the ratio
s=(
i:i2 <0
|Ci |2 )/(
|Ci |2 ) 1,
(2.22)
i:i2 >0
which means that the contribution of the surface waves to the total energy of the transmitted electromagnetic eld is negligibly small. The number of the bulk waves is determined by the number of photonic band NB , which the frequency of the incident light
corresponds to and by the angle of incidence. Depending on the angle of incidence, the
photonic crystals.
15
number of bulk waves is either exactly equal to the number of photonic bands NB , or
to NB 1.
It is also considering the dependence of the ratio of amplitudes of the bulk modes on
the relative contrast n of the photonic crystal dened as |n1 n2 |/(n1 + n2 ), which
is illustrated in Fig. 2.3(b). As the relative contrast is increased, one of the mode
coecients becomes dominant, so most of the energy of the electromagnetic eld is
transferred to a single eigenmode. We should note, that this property depends on the
choice of the basis of the photonic crystal eigenmodes. For example, instead of using
the basis un discussed above, we could describe the eld in a plane wave basis:
E(x, z) =
Cp exp[iKp z + i
k02 Kp2 x],
(2.23)
p=
where Kp = K0 + 2p/D. However, in this case, as the relative contrast of the photonic
crystal is increased, the Cp becomes less convergent, which means that the electromagnetic energy exists in a large number of eigenmodes. In contrast, for the basis un most of
the electromagnetic energy is transferred into a single eigenmode as the relative refractive index contrast is increased, which facilitates the analytical study of the system. For
example, if we include only the zero-order diraction of the reected waves, i.e. set Rp=0
to zero, we can analytically derive an expression for the ratio of bulk mode amplitudes
(see Fig.2.3(b)):
|C1 |2
|J01 |2
=
,
|C2 |2
|J02 |2
(2.24)
which by the denition of the matrix J is the ratio of the overlap integrals of the incident
light and each of the two bulk modes.
To dene the direction of propagation of the transmitted light, we calculate the component of the Poynting vector tangential to the Bragg reector side edge averaged over
the crystal period:
1
Sz =
2D
1
Re [Ey Hx ] dz =
2k0 D
[
Im Ey
Ey
z
) ]
dz.
(2.25)
In the case of high refractive index contrast, it is a good approximation to retain only
one mode in the series, and using the expression (2.11), we can evaluate the integral
(2.25) analytically to obtain:
Sz =
sin K0 D
1
|Cm |2
.
2k0 D
kz1 cos kz2 d2 sin kz1 d1 + kz2 cos kz1 d1 sin kz2 d2
(2.26)
The z-component of the group velocity and energy ow of the incident light has the
same sign as K0 . Therefore if we multiply Eq. (2.26) by K0 , we will get a value that is
photonic crystals.
16
negative if the zcomponents of the Poynting vectors in the crystal and in the vacuum
are antiparallel, which manifests itself physically as negative refraction. In the case of
the low contrast photonic crystal we include both bulk modes in the Poynting vector
calculation. Then the nal expression for the Poynting vector tangential components
will contain two terms. corresponding to (2.26) for each of the bulk modes and an
additional interference term:
)
D (
u
1
u
Im exp(i(m n )x) dz Cm Cn um n Cn Cm
un m .
2k0 D
z
z
(2.27)
We note that this additional term contains a factor, which oscillates in the direction
perpendicular to the interface.
Fig. 2.4: (a) Isofrequency contours for the photonic crystal with refractive indices
1.5 and 1.7 for a frequency of 1.6 eV. Branches which correspond to negative values
of Sz K0 are depicted by blue lines and to positive values by red lines. Dashed arrows
show the directions of the negatively refracted modes, and solid arrows the directions of
the diracted modes. (b) Isofrequency contours for the photonic crystal with refractive
indices 1.1 and 2.1 at a frequency 1.6 eV. In each case isofrequency contours for isotropic
media with the eective refractive indices of the photonic crystal are depicted by dashed
semicircles. The direction and absolute of the tangential component of the wavevector
for the incident wave is shown with red arrow for all the cases. The directions of group
velocity and energy ow are shown with green arrows for all the cases. For all the cases
the dotted lines show the value of the in-plane component of the incident wave (upper
line) and the value of the in-plane wavevector of the diracted wave (lower line).
In order to illustrate the transmission through the photonic crystal interface and the
photonic crystals.
17
connection between negative refraction and the diraction of modes, it is instructive to
consider the isofrequency contours of the bands of the photonic crystal. We consider
waves incident on the photonic crystal with a given frequency but varying angle of
incidence and hence tangential wavevector K0 . The transmitted wave must have the
same frequency and a tangential wavevector that is either K0 or diers from it by a
reciprocal lattice vector (2n/D, where n is an integer). In the example shown in
Fig. 2.4 the blue lines correspond to the isofrequency branches, for which the product
Sz K0 is negative, and the red lines to the branches for which Sz K0 is positive. The group
velocity of each mode is normal to the isofrequency contour and is shown with a green
arrow. Since the directions of the group velocities for the points 1 and 2 are parallel,
a negatively refracted beam corresponding to arrow 1 is indistinguishable from the 1
diracted beam shown with arrow 2. The same arguments are valid for the directions 3
and 4. In the case of low contrast (Fig.2.4(a)) the wavevector is almost collinear with
the group velocity. However, in the high contrast case (Fig.2.4(b)) the wavevector and
group velocity are in markedly dierent directions, which is a consequence of strong
anisotropy of photonic crystal, which increases with increasing relative refractive index
contrast. However, even in the case of high contrast regime, the diracted beams 2 and
4 are indistinguishable from the refracted beams 1 and 3, since their group velocities
are parallel. Physically this means, that there is no way to distinguish the negatively
refracted beam from the diracted one.
It is interesting to consider whether superlensing could be realized in a suitable structure.
The electromagnetic eld produced by a point source placed close to the edge of the
structure can be represented as an integral of plane waves with components of wavevector
parallel to the edge of the structure varying continuously from minus innity to plus
innity. Each of these plane wave components will be refracted by the structure in a
distinct way. If we consider the isofrequency contours shown in Figs.2.4(a),(b) we can
see that there is always a range of incident tangential wavevectors K0 for which negative
refraction will not occur. As a result, some of the radiation from a point source will
always be refracted positively and only some will be refracted negatively, which means
that eective subwavelength imaging can not be achieved.
2.2.2
Comparison of the analytical theory and numerical modelling

results
The analytical theory and associated results presented in the previous subsection have
been veried by numerical modelling using the nite dierence-time-domain-technique
(FDTD). Figures 2.5(a) and 2.5(b) show the distribution of the square of electric eld for
waves scattered at the interface of the high contrast photonic crystal (n1 = 2.1, n2 = 1.1).
Figure 2.5(a) shows negative refraction with the frequency of the incident wave lying
in the second photonic band of the photonic crystal. Figure 2.5(b) corresponds to the
photonic crystals.
18
case when sin(K0 D) = 0, resulting in the tangential component of the Poynting vector
being equal to zero and the propagation of the wave being normal to the interface.
(a)b
(b)
(c)
(d)
Fig. 2.5: FDTD simulation results for waves scattered at the interface of high (n1 =
2.1, n2 = 1.1) (a,b) and low (n1 = 1.4, n2 = 1.8) (c,d) contrast photonic crystal. (a)
- Spatial distribution of the square of electric eld for the case of negative refraction
(~ = 1.6eV, angle of incidence = 55 degrees);(b) Spatial distribution of the square
of electric eld for the case of normal propagation (~ = 2.0 eV, = 55 degrees); (c)
Spatial distribution of the absolute value of the square of electric eld for the case of
spatial oscillations of the Poynting vector (~ = 2.0 eV, = 55 degrees); (d) Spatial
distribution of the real part of the electric eld for the case described in (c).
Figure 2.5(c) shows the spatial distribution of the square of the electric eld in a
low contrast photonic crystal(n1 = 1.8, n2 = 1.4), when the frequency is again chosen
to make sin(K0 D) = 0. In this case, in the analytical expression for the tangential
component of the Poynting vector, only the oscillating term in Eq. (2.27) is non-zero.
As predicted by the analytical theory, the numerical results show spatial oscillations of
the electromagnetic eld in the direction normal to the interface. Figure 2.5(d) shows
the distribution of the real part of electric eld for the same case as in Fig. 2.5(c). For
the low contrast photonic crystal, in the case when sin(K0 D) = 0, which is not shown
here, we observe a splitting of the incident beam into positively and negatively refracted
beams as also predicted by the analytical theory.
photonic crystals.
19
2.2.3
Results and conclusions
It should be noted that the analytical model presented assumes an incident plane-wave
with an innite wave front. In an experiment, however, we would be using a light beam
of nite size. The numerical calculations show that there exists a critical source size
above which the refraction of wave at the interface can be described by the analytical
theory. We have derived an estimate of the critical source diameter which is given by
dthr D/n, where n is the relative contrast of the photonic crystal. That formula is
similar to the expression obtained in [37], which denes the minimum size of the Bragg
reector for the formation of a photonic stop-band suggesting that the origin of the
critical source size is the existence of a minimum photonic crystal size for the formation
of the photonic band structure. When the source size is signicantly less than the critical
value, the refraction of the wave at the side edge of the photonic crystal is determined
by the eective refractive index nef f = (n1 d1 + n2 d2 )/(n1 + n2 ).
The propagation of the refracted beam normal to the interface can be considered to
be the result of an excitation of an array of coupled waveguide modes, where the high
refractive index layers play the role of the waveguide core and coupling is achieved
via evanescent waves inside the lower refractive index layers that play the role of a
waveguide cladding. However, the case of normal propagation is not entirely equivalent
to the excitation of waveguide modes since in general it can occur even in the case when
the light propagates in both the high and low refractive index layers.
The spatial oscillations of the Poynting vector resemble the Bloch oscillations which
have been predicted and observed both in superlattices, for electrons in an external
electric eld, and for the electromagnetic eld in wedge-shaped photonic crystals[38].
In the latter case the gradient of the photonic crystal slab leads to a gradient of the
quantized lateral photon momentum which can be considered as an eective potential
or force acting on the photon propagating in the structure. Bloch oscillations occurs
when an electron (or photon) propagates in a periodic potential with an external force
applied. The wavevector of the particle is changed by an external force, and when it
reaches the edge of the Brillouin zone, it changes its wavevector by 2/D, where D
is the period of the structure. The main dierence of the oscillations in our case is
the that the oscillations take place in the absence of an external eld. However, the
wavevector lies exactly at the edge of the Brilloin zone since the condition sin(K0 D) = 0
applies. Also, the oscillations are only observed in the case of a quarter wavelength Bragg
reector, for which the second band gap of the photonic crystal becomes degenerate. The
observed eect of the Poynting vector oscillations can thus be regarded as a result of the
interference of two degenerate Bloch modes existing at a photonic crystal band edge.
To conclude, in this section we have developed a theoretical model, which facilitates the
description of refraction of electromagnetic wave at the side edge of a one-dimensional
photonic crystal. Using the model we have predicted and veried two new eects: normal
photonic crystals.
20
propagation of the electromagnetic wave and spatial oscillations of the Poynting vector
inside the photonic crystal.
photonic crystals.
21
2.3
Negative refraction of THz radiation by a prism made

of a two-dimensional metallic photonic crystal
In the last few years, terahertz (THz) technology [39] has been applied to a large variety of areas including art conservation [40], medicine [41] and security systems. As
a consequence, the demand for new waveguides, lters and other components for the
terahertz frequency range has emerged. In this section, the concept of a new lter for
the terahertz range is presented. The lter is a prism made from a two-dimensional
hexagonal metallic photonic crystal and operates on the basis of obtaining both positive
and negative refraction, which can lead to a large directional dispersion for incident light
of dierent frequencies.
2.3.1
Calculational method
We consider a two-dimensional hexagonal photonic crystal formed by a periodic array of

metallic nanorods. To calculate the eigenmodes and band structure of such a photonic
crystal, we cannot use the conventional plane-wave band structure method since the
dielectric permittivity of the metal depends on frequency. However a complex band
structure method [42] provides a suitable approach. The method has poor convergence
at frequencies close to the metal plasma frequency, but here we use it for the terahertz
frequency range, which is three orders of magnitudes less then the plasma frequencies
of the noble metals.
We consider the metallic rods to be aligned parallel to the z-axis and begin by studying
the TM modes for which the electric eld is parallel to the rods. The wave equation for
the electric eld is then
2 Ez (r) =
2
(r)Ez (r),
c2
(2.28)
where r = (x, y). The electric eld can be written as the series:
Ez (r) =
Eg ei(k+g)r ,
(2.29)
which substituted into Eq. 2.28 gives:
Eg (k + g)2 ei(k+g)r =
2
(r)
Eg ei(k+g)r .
c2
g
(2.30)
The spatially-dependent dielectric permittivity (r) of the structure can be written as

(r) = b + [() b ] S(r),
(2.31)
photonic crystals.
22
where b is the background dielectric permittivity, () is the permittivity of the metal,
and S(r) is equal to unity inside the metallic elements and zero outside. We can write
S(r) as a series: S(r) =

Sg ei(k+g )r and if we substitute this expression into 2.30,
g
we get
[
Eg
]
2 b i(k+g)r 2 (() b )
Eg Sg ei(k+g+g )r .
(k + g) 2 e
=
2
c
c
(2.32)
g,g
Multiplying this by ei(k+g)r , where g is a reciprocal lattice vector, and integrating over
the photonic crystal unit cell gives
[
Eg
]
2 b
2 (() b )
(k + g) 2 =
Sgg Eg .
c
c2
g
2
(2.33)
Writing the wavevector k as k

u, where u
is a unit vector that denes the direction of k,
equation (2.33) can be rewritten as
k 2 IE + kAE = (C B)E = DE,
(2.34)
where I is the identity matrix, A is a diagonal matrix with elements: Ag,g = 2g

u, C is a
matrix with elements Cg,g == ( 2 (() b ))/c2 Sgg , and B is a diagonal matrix with
elements: Bg,g = g 2 2 b /c2 . Equation 2.34 can be reformulated as the eigenvalue
problem
[
](
D A
E
kE
(
=k
E
kE
)
,
(2.35)
from which, after choosing the frequency and direction u

, we can calculate k, which
is generally complex. Real k corresponds to propagating solutions, and complex k to
decaying solutions.
In the case of the TE polarisation for which the magnetic eld is parallel to the rods,
the wave equation is
[
]
1
2
H(r, t) = 2 H(r, t),

(r)
c
(2.36)
and the inverse dielectric permittivity can be written as:

1
1
=
(r)
b
(
)
() b
1
S(r) .
()
(2.37)
photonic crystals.
23
If we then substitute (2.37) into (2.36) and take the Fourier transform of S(r), we get
Hg (k + g)2 ei(k+g)r
() b
Sg,g Hg (k + g + g )ei(k+g+g )r =
()
(2.38)
g,g
2 b
Hg Sg ei(k+g)r .
c2 g
Multiplying (2.38) by e(k+g)r and integrating over the unit cell gives:
Hg (k + g)2
() b
2 b
H
=
Sgg Hg (k + g)(k + g.),
g
c2
()
g
(2.39)
which can be rewritten as

k2 (I P)H + k(A Q)H = (R B)H,
(2.40)
where the matrices A, B, I are dened in the same way as in the TE case, and matrices
P, Q, and R are given by
() b
Sgg ,
()
() b
(g + g),
=
Sgg u
()
() b
Sgg gg.
=
()
Pg,g =
(2.41)
Qg,g
(2.42)
Rg,g
(2.43)
Multiplying (2.40) by (I P)1 , gives

k2 IH + kA H = D H,
(2.44)
where A = (I P)1 (A Q) and D = (I P)1 (R B). Equation (2.44) has the same
structure as (2.34), and can also be rewritten as a similar eigenvalue equation:
[
D A
](
H
kH
(
=k
H
kH
)
.
(2.45)
The theory presented above has been used to calculate the band structure of a hexagonal
photonic crystal with a unit cell consisting of a gold cylinder with radius equal to 80
m and a period of 200 m. The structure of the rst two photonic bands of the
crystal is shown in Fig.2.6(a). The cut-o frequency fp , below which no propagating
solutions exist, is equal to 0.76 THz and corresponds to the eective plasma frequency
of the photonic crystal. The rst photonic band is characterized by a positive group
velocity and its upper bound is equal to fJ = 0.96 THz. The second photonic band,
lying in the frequency range from fJ to 1.45 THz, is characterized by a negative group
photonic crystals.
24
2nd band
1st band
Fig. 2.6: (a) Band diagram of a hexagonal photonic crystal formed by gold cylindrical
rods with diameter 80 m. Period of the structure is 200m. fp is an eective plasma
frequency, which is a cut-o frequency for the propagating solutions. The group velocity
is collinear with the wavevector in the rst band (red), which corresponds to positive
refraction. The second band (blue), where the group velocity has opposite direction to
the wavevector is the source of negative refraction. (b) Isofrequency contours for the
frequencies close to high frequency edge of the second band.
velocity. The isofrequency contours near the top of the second band are shown inS
Fig.2.6(b). They are essentially circular over a wide frequency range, which means an
eective dielectric permittivity can be introduced. The refractive index is then dened
as nef f = (d/dk) , where is the eective dielectric permittivity. Therefore, the

structure is characterized by a positive refractive index for the frequencies in the rst
photonic band and by a negative eective refractive index for the frequencies in the
second photonic band. Due to the fact that the refracted wave should propagate away
from the interface, a wave incident on the crystal from a homogeneous medium with
frequency lying in the rst photonic band will refract positively, while a wave with
frequency lying in the second band will refract negatively.
2.3.2
Numerical modelling results
Figure 2.7 shows numerical results from the-nite-dierence-time-domain modelling of

the refraction of a plane wave by a prism made of the photonic crystal having facets
of length of 5 mm. The angle of incidence is 15 degrees with respect to the normal
to the surface of the prism. Figure 2.7(a) corresponds to a frequency f = 0.8 THz
lying in the rst photonic band and positive refraction is clearly observed. Figure 2.7(b)
corresponds to a frequency f = 1.2 THz lying in the second band and shows a distinct
negative refraction eect.
photonic crystals.
25
SOURCE
IF
PRF
(a)
NRF
(a)
(b)
Fig. 2.7: Squared electric eld for waves refracted on a photonic crystal prism. (a)
Wave frequency f = 0.8 THz, positive refraction. (b) f = 1.2 THz, negative refraction.
Fig. 2.8: The initial design of the photonic crystal prism. The separation of the rods
on the right-hand facet is dierent from that in the bulk.
The prism shown in Fig.2.7 has the shape of an equilateral triangle. Initially, a prism
in the form of a right-angled triangle (see Fig.2.8) was used but the experiments were
unsuccessful because no refracted beam was detected in the direction predicted by the
theory. However, during the numerical modelling it was realized that a facet of the
photonic crystal which does not coincide with one of the main crystallographic axes,
acts as an eective diraction grating for the incident light, destroying the negative
refraction eect through additional diraction at the interface. As a result, most of the
radiation has been diracted from this eective diraction grating and did not reach the
receivers.
photonic crystals.
26
2.3.3
Comparison of theory and experiment
Experiments were performed by collaborators guided by the results of the numerical

modelling and a micrograph of the sample used is shown on Fig.2.9.
Fig. 2.9: Micrograph of the photonic crystal prism studied in the experiment. Arrows
have been added to show the directions of the negatively (blue) and positively (red)
refracted beams.
The details of the prism fabrication methods employed can be found in [43]. Briey,
high-aspect-ratio polymer pillars were formed by backside UV exposure of SU-8 50 on a
glass substrate, before being sputtered with gold. Because the gold thickness is greater
than the electromagnetic radiation skin depth, the pillars behave in eect like solid
metallic rods.
The refraction of the THz light by a photonic crystal can be experimentally measured
using a broadband spectrometer. The THz radiation was generated using a photoconductive switch, fabricated on low-temperature-grown GaAs and was detected electrooptically. After generation, the THz radiation was collected and focused onto the sample
by a pair of parabolic mirrors. For both the cases of negative and positive refraction,
a plane mirror placed at the THz beam focus allowed reference measurements to be
undertaken to check the system alignment and performance. Another pair of parabolic
mirrors, placed in an appropriate angular position, collected the radiation from the sample, which was then focused onto the detector. It should be noted that the parabolic
mirrors focused the broadband THz beam onto, and collected the narrowband refracted
photonic crystals.
27
beam from, the sample over an angular range of approximately 10 degrees which is much
smaller than the measured angular separation of the refracted beams.
Figure 2.10(a) shows the typical temporal dependence of the reference signal obtained
from the spectrometer. The Fourier transform, which is shown in the inset, indicates a
usable bandwidth of approximately 3 THz with maximum power being output around 1
THz, which corresponds to the operating frequency of the device. Figure 2.10(b) shows
the pulse that emerges from the negative refraction facet.
Amp (a.u.)
60
40
20
f (THz)
0
-20
-40
-60
0,6
0,4
0,2
0,0
-0,2
-0,4
-0,6
0
20
40
time (ps)
60
Fig. 2.10: (a) Temporal dependence of the reference signal. Inset shows the Fourier
transform. (b) Temporal dependence of the negatively refracted beam.
Even though the amplitude is much reduced in the time domain, the negatively refracted
pulse contains numerous oscillations. This pulse ringing is clear initial evidence that only
a limited range of frequency components are steered by the device to emerge from the
negative refraction facet. The THz beam emerging from the positive refraction facet is
of a similar structure. The Fourier transforms of the pulses (Fig.2.11) show the spectral
content contained in both the negatively and positively refracted components of the
original, incident THz beam, the rest being ltered away by the device.
It is clear that while there is some overlap of the tails of the spectral peaks, there are
distinct pass-bands for the positively and negatively refracted waves, which correspond
to the photonic bands having a positive and negative group velocities respectively. The
positive refraction band extends from 0.73 to 0.91 THz, while the negatively refracted
Intensity
photonic crystals.
28
positive
refraction
negative
refraction
f (THz)
Fig. 2.11: (a) Fourier transforms of the refracted signals. Horizontal lines show the
widths of the rst and second bands, calculated by the complex band structure method.
band covers the spectral range of 0.91 to 1.24 THz, when dened by their FWHM. In
both bands, approximately 10 % of the incident beam power is refracted. The general
form of these results is consistent with our theoretical band structure analysis, which
should be recognized as being for a system consisting of an innite array of pillars using
a nite number of plane waves, whereas in the experiment a focused beam is incident
on a nite structure. When losses are included, the value of fP is no longer clear-cut
and losses also reduce the value of fJ , and both these trends are consistent with the
experimental observations. Also, limitations on the micromachining processes used to
construct the prism mean that the pillar diameters are somewhat more narrow toward
their tops, and this will further reduce the predicted frequencies.
In summary, we have demonstrated, for the rst time to our knowledge, the ability of a
metallic photonic crystal prism to lter a broadband THz beam in such a way that two
angularly resolved beams at dierent frequencies emerge. Essentially, the device works
in the way predicted by our theory: by positively refracting the frequency components in
one interval and negatively refracting the frequency components in an adjacent interval
by making use of the properties of the two lowest photonic bands of the crystal. It should
also be noted that the top-hat shape of the transmission spectrum for the negatively
refracted wave makes such a device attractive for use as a spectral lter.
We believe the proposed lter is useful advance in THz technology, particularly as currently there is a lack in practical realizations of interference or absorbing terahertz
spectral lters - the approaches most commonly used in the optical range. Most of the
existing terahertz spectral lters are based on periodic or quasiperiodic metal dielectric
photonic crystals.
29
structures [44, 45] exhibiting resonant transmission (or absorption) due to the excitation of surface plasmon polariton modes at metal-dielectric interfaces. Our approach
however is advantageous due to the fact that the eld does not penetrate in the metal
signicantly, thus reducing the losses as compared to the metamaterial-based lters.
photonic crystals.
30
2.4
Summary of key original results
Main results obtained in Chapter 2.

Making use of an established theoretical method to describe the scattering of
plane wave from the side edge of one-dimensional photonic crystal, two new optical
eects have been predicted analytically and also observed in numerical simulations
- propagation normal to the interface of the transmitted electromagnetic eld and
spatial oscillations of the Poynting vector.
A new type of spectral lter for the terahertz frequency range has been proposed.
The design of the lter has been developed and rened using numerical modelling
and the results compared with experimental measurements on a prototype structure.
Chapter 3
Eigenmode analysis of resonant

photonic nanostructures
This chapter is dedicated to the study of eigenmode structure and nonlinear eects in
certain resonant photonic nanostructures.
One investigation concerns the analysis of the eigenmode structure of a microcavity on
organic cavity layer adjacent to a metal lm. This work was done in collaboration with
an experimental group from the University of Dresden [46]. An experimental sample was
fabricated by our collaborators and a series of the reectivity measurement experiments
were made1
Another study is of the eigenmode structure of a periodic array of semiconductor quantum wells buried in the layers of a distributed Bragg reector [47]. This work was done
in collaboration with experimental groups from the University of Crete and the University of Cambridge. The sample was fabricated by the Crete group and a series of the
reectivity spectra measurements were performed. The Cambridge group performed the
pump probe experiments revealing parametric amplication occurring in the structure2
The last section of the chapter presents unpublished results on the modelling of the parametric amplication that has been experimentally observed in the structure described
above
However, before describing the three studies in details we present some relevant theoretical background material starting with a short overview of the transfer matrix method
1
I performed the numerical modelling of the reflectivity spectra and derived analytical expressions
for the eigenfrequencies of the structure.
2
I performed the numerical modelling of the reflectivity spectra and derived the approximate analytical expressions for the structure eigenfrequencies.
3
I implemented a three-level formalism, which had been developed in [48], for the periodic quantum
well structure and show that the simulation results agree well with the experimental data obtained by
the Cambridge group.
31
Chapter 3. Eigenmode analysis of resonant photonic nanostructures
32
which is extensively used throughout the chapter. It is followed by the calculation of the
eigenfrequency of a specic type of surface optical mode occurring at the interface of a
metal lm and a distributed Bragg reector, a Tamm plasmon polariton. Finally, the
derivation of the expression for the reection coecient from a semiconductor quantum
well is presented.
3.1
3.1.1
33
Background theory
Transfer matrix method
A detailed description of the transfer matrix method can be found in [49] and we present
only the key features that are relevant to our work here. We consider an arbitrary
multilayered structure consisting of layers of isotropic material grown in the z-direction.
The electromagnetic eld can be decomposed into two polarizations: the TE polarisation,
for which the electric eld is orthogonal to the plane formed by the wavevector and the
normal to the layers, and the TM polarisation for which the magnetic eld is orthogonal
to that plane.
= (0, Ey , 0)
In the case of the TE polarization we write the the electric eld vector as E
= (Hx , 0, Hz ). We take the elds to have the
and the magnetic eld vector as H
harmonic time dependence exp[it] and without the loss of generality we can set
the y component of the wavevector to zero. The spatial variation of the elds parallel to
the structure is then exp[ikx x], where kx is the same for all the layers. Then Maxwell
equations can be reduced to the Helmholtz equation for the spatial dependence of the
electric eld Ey in each layer:
2 Ey
+ kz2 Ey = 0,
z 2
(3.1)
where kz2 = (/c)2 kx2 , and is the layer dielectric permittivity. The general solution
of the equation (3.1) is:
Ey = A exp[ikz z] + B exp[ikz z],
(3.2)
where A is the amplitude of the forward propagating wave and B - that of the backward
propagating wave. Now, in a homogeneous medium, if we know the amplitudes in the
plane z = z0 , say, (A0 , B0 ), the amplitudes (A1 , B1 ) of the eld in the plane z = z1 are
simply A1 = A0 exp[ikz (z1 z0 )], B1 = B0 exp[ikz (z1 z0 )]. It is convenient to write
this result in matrix form:
( )
A1
B1
=M
A0
B0
(
;
=
M
eikz d
eikz d
)
,
(3.3)
is conventionally referred to as a transfer matrix. Here M
where d = z1 z0 , and M
is expressed in the basis of forward and backward propagating waves, but it is possible
to transform to the basis of the tangential eld components Ey and Hx . The tangential
34
components corresponding to Eq.(3.1) are:

Ey = A exp[ikz z] + B exp[ikz z],
(3.4)
kz
kz
A exp[ikz z] + B exp[ikz z],
k0
k0
(3.5)
Hx =
where k0 = /c. In matrix form:

(
Ey
Hx
( )
A
= S
,
B
(3.6)
where
(
S =
kz /k0 kz /k0
(3.7)
S1
It follows that the transfer matrix in the basis of tangential components is T = SM
and
(
Ey1
Hx1
(
= T
Ey0
Hx0
)
(3.8)
where
(
T =
)
i(k0 /kz ) sin(kz d)
cos(kz d)
i(kz /k0 ) sin(kz d)
cos kz d
(3.9)
The matrix transferring the tangential elds across the interface of two layers with
dielectric permittivities 1 and 2 is the unit matrix due to the electromagnetic boundary
across an interface in
conditions. This fact can be used to derive the transfer matrix L
the case of the forward and backward propagating basis:
=
L
S21 S1
kz2
=
2kz1
1 + kz1 /kz2 1 kz1 /kz2
)
.
1 kz1 /kz2 1 + kz1 /kz2
(3.10)
The corresponding results for the TM polarization can be derived in a similar manner,
and in that case:
(
T =
= 1 kz2
L
22 kz1
cos(kz d)
ikz /(k0 ) sin(kz d)
ik0 /kz sin(kz d)
cos kz d
)
.
)
1 + (2 kz1 )/(1 kz2 ) 1 (2 kz1 )/(1 kz2 )
1 (2 kz1 )/(1 kz2 ) 1 + (2 kz1 )/(1 kz2 )
(3.11)
(3.12)
Thus, we now have expressions for all the matrix components in both bases and for
35
both polarizations. By forming product of the appropriate layer and interface transfer
matrices, we can obtain the transfer matrix of an arbitrary multilayered structure.
If we have the transfer matrix of a structure, we can calculate its transmission and
reection coecients. In the basis of forward and backward propagating components, if
we assume that the amplitude of the incident light is unity we can write:
( )
t
0
( )
1 ,
=M
r
(3.13)
is the
where r, t are the reection and transmission coecients respectively and M
transfer matrix. It then follows that:
21
M
,
22
M
1
t=
,
22
M
r=
(3.14)
(3.15)
where the derivation of the expression for t has used the fact that all the transfer
matrices for the structures placed in the symmetric environment, i.e. if the refractive
indices of the surrounding media from the both sides of the structures are equal, since
the determinant of such structure is equal to unity [50].
We can also derive the components of the transfer matrix from the reection and transmission coecients of the structure. Imagine an arbitrary layered structure bound with
half-space of refractive index n0 to the left and half-space of refractive index nf to the
right. We assume that we know the reection and transmission coecients for the light
travelling at normal incidence from the left, denoted as rl , tl , and the reection and
transmission coecients for the light travelling from the right, denoted as rr , tr . Our
aim is to construct the transfer matrix of the layered structure T corresponding to the
transfer of the electromagnetic eld from the left side of the structure to the right side.
We can write down two matrix equations:
( ) ( )
1
tl
,
T
=
0
rl
(3.16)
and
( ) ( )
rr
0
=
,
T
1
tr
(3.17)
Since the system is not symmetric, it can be shown that the determinant of T in the
case of normal incidence is equal to n0 /nf . Using this relation and the four equations
36
obtained from equation (3.16) and (3.17) we can evaluate the components of matrix T:
1
T =
tr
(
)
(tl tr rl rr )/tr rr
rl
(3.18)
For the case of normal incidence the following relations holds between tl and tr : tl =
(n0 /nf )tr . The picture simplies drastically if n0 = nf . In this case rl = rr = r, tl =
tr = t and
1
T =
t
(
)
(t2 r2 )/t r
r
(3.19)
Using the transfer matrix approach it is also possible to derive the eigenvalues and
eigenmodes of an innite periodic structure. Applying Blochs theorem for a periodic
structure with period D:
(
Ey (z = D)
Hx (z = D)
=M
Ey (z = 0)
Hx (z = 0)
(
= eiKD
Ey (z = 0)
Hx (z = 0)
)
,
(3.20)
is the transfer matrix over the period of the structure and K is the Bloch
where M
vector. It follows that:
)/2.
cos(KD) = T r(M
3.1.2
(3.21)
Tamm plasmon eigenfrequencies
In section 3.2 we consider the coupling of Tamm plasmon modes in the organic microcavities and obtain the analytical expressions for the resulting eigenmodes. In preparation,
here we provide a denition of the Tamm plasmon polaritons and show how to obtain
the dispersion relation for a single Tamm plasmon mode.
The Tamm plasmon polariton is a special type of cavity mode that can exist in a cavity
with mirrors made of a metallic layer on one side and a Bragg reector on the other side.
Due to the use of a metallic mirror on one side, the cavity length needed to support the
eigenmode at a xed wavelength is smaller than that of a conventional Bragg reector
cavity for the same wavelength. At the same time, due to the fact that the mode is still a
cavity mode, it is characterized with a parabolic in-plane dispersion relation. Therefore,
a Tamm plasmon mode can be used as a realization of a spatially compact (because of
the small cavity length) and slow light (because of the parabolic dispersion and thus
small group velocity for the small in-plane wavevectors). Despite the simplicity of the
concept, Tamm plasmon polaritons were only proposed recently but have since attracted
a lot of attention. Some of the many applications of Tamm plasmon polaritons include
37
the realization of polariton logic devices [51], and the control of the radiative decay of
the quantum dots [52].
In order to derive the dispersion relation for a Tamm plasmon polariton, we consider
the structure shown in Fig.3.1 and look for optical modes localized close to the metal
e1
eM
e2
x
z
d0
d1 d2
Fig. 3.1: The simplest realization of the structure where a Tamm plasmon mode
can exist. A semi-innite Bragg reector (grey layers of dielectric permittivity 1 and
2 ) adjacent to a semi-innite metal layer (yellow) with dielectric permittivity M .
The thickness of the dielectric layer adjacent to the metal may be varied to tune the
frequency of the mode.
layer. To nd the eigenfrequencies of such modes, we can utilize the well known general
gcardinal condition for the eigenmodes of a general cavity:
rL rR e2i = 1,
(3.22)
where rL , rR are the reection coecients of the left and right mirrors and is the phase
change across the width of the cavity.
In the case of normal incidence the reection coecient of the metal layer is:
rM
1 M
n1 nM
=
=
.
1 + M
n1 + nM
(3.23)
In the simplest Drude model:

M = 1 p2 / 2 ,
(3.24)
where p is the plasma frequency of the metal. In the application considered here the
eigenfrequencies of the modes are around 1 eV and are much smaller than the plasma
frequency which for the common noble metals is about 10 eV [53]. Therefore, the
38
reection coecient of the metal half-space can be approximated as

rM
[ (
)]
2 1
exp i +
.
p
(3.25)
The reection coecient of the Bragg reector calculated at the interface of layers with
refractive indices n1 = 1 and n2 = 2 for the frequency in the vicinity of the stop
band centre depends on the relation between n1 and n2 . If n1 > n2 then
(
rBr = exp
in2 0
|n1 n2 | 0
)
,
(3.26)
and if n1 < n2 then

(
rBr = exp
in1 0
|n1 n2 | 0
)
,
(3.27)
where 0 is the Bragg reector stop band centre frequency. Finally, the phase change
in the vicinity of the stop band centre is given by:
(
)
0
1+
.
0
(3.28)
For the case n1 > n2 the eigenmode condition of equation (3.22) requires:
+
0
n2 0
2n1
++
+
= 2.
p
0
n1 n2 0
(3.29)
0
.
1 + 20 /(p )|n1 n2 |
(3.30)
which has the solution

=
Equation (3.30) refers to the case when the thickness of the layer adjacent to the metal
layer is equal to that of the other layers of the same type forming the Bragg reector.
However, by varying the thickness it is possible to tune the Tamm plasmon frequency
exactly to the Bragg frequency 0 .
It should be noted that when n2 > n1 there will be an additional contribution of to
the overall phase at the left side of the equation (3.29) and there will be no solutions in
the vicinity of the Bragg frequency.
3.1.3
Reection coecient of an exciton quantum well
In section 3.3 of this chapter we study a particular type of resonant periodic nanostructures in which light couples to excitons excited in quantum wells embedded in the
layers of a Bragg reector. To calculate the eigenfrequencies of the nanostructures using
transfer matrix theory we need to derive the reection coecient for a single quantum
39
well with a view to incorporating the eect of the quantum well excitons in the previously discussed transfer matrix theory. The detailed derivation of the quantum well
reection coecient can be found in [54] and only the main steps are described here.
Also, for simplicity we deal only with the normal incidence of electromagnetic waves on
the quantum well.
In the case of normal incidence the electric eld of the system is governed by the equation:
2E
+ k02 E = k02 4Pexc ,
z 2
(3.31)
where is the background dielectric permittivity of the quantum well, k0 = /c, and
Pexc is the excitonic polarization given by
4Pexc = G(z)
(z )E(z )dz .
(3.32)
Here (z) = exc (R = 0, = 0, ze = zh = z) is the exciton wavefunction taken with

electron and hole having the same position in space,
G=
Q
,
0 i
(3.33)
where
Q = a3B LT ,
(3.34)
in which aB is the exciton eective Bohr radius, LT is the longitudinal-transverse

splitting, 0 is the exciton frequency and is the nonradiative damping. (z) can be
chosen to be real when the origin of z coincides with the centre of the quantum well.
The general solution of an equation of the type E (x) + k 2 E = F (x) can be written
in the form:
i
E = E1 exp[ikx] + E2 exp[ikx] +
2k
dx exp[ik(x x )]F (x ).
(3.35)
Thus, when a wave with amplitude E0 is incident from z < 0 on the quantum well, the
solution to Eq. (3.31) can be written as:
ik 2
E(z) = E0 exp[ikz] + 0 G
2k
dz exp[ik(z z )](z )
(z )E(z )dz .
(3.36)
Multiplying this equation by (z) and integrating over z gives the following algebraic
equation for = E(z)(z)dz:

k2
= 0 + i 0 G()
k
dzdz exp[ik(z z )](z)(z ),
(3.37)

where 0 = E0
dz(z)E(z). The solution for is then:
40
E(z)(z)dz =
(ik02 )/(k)G()
0
,
dzdz exp[ik(z z )](z)(z )
(3.38)
which can be substituted into Eq. (3.36) to give:
E0 (ik02 )/(2k)Q dz exp[ikz ](z ) dz exp[ik(z z )](z )

E(z) = E0 exp[ikz] +
.
k2
0 i 1 i k0 Q
dz dz exp[ik(z z )](z )(z )
(3.39)
The reection and transmission coecients are dened as:
r=
E(z) E0 (z)eikz
;
E0 eikz
(z)
t=
E(z)
.
E0 (z)eikz (z)
(3.40)
And using the fact that (z) is an even function with respect to z, it follows that:
r=
i0
;
0 i( + 0 )
t = 1 + r,
(3.41)
where
Qk02
0 =
2k
]2
(z) cos(kz)dz
(3.42)
is called the exciton radiative broadening and

Qk02
0 = 0 +
2k
dzdz (z)(z ) sin(k|z z |)
(3.43)
is the renormalisation of the exciton resonance frequency due to the polariton eect. As
soon as we know the reection and transmission coecients of the quantum well we can
easily reconstruct its transfer matrix using the expression (3.19) from section 3.1.1.
3.2
Parabolic polarization splitting of Tamm states in a

metal-organic microcavity
In this section we study the optical properties of the structure shown in Fig.3.2 which
consists of an a metallic layer and adjacent organic cavity layer sandwiched between two
Bragg reectors. The structure was grown and studied experimentally by our collaborators in Dresden.
3.2.1
Structure and experimental set up
The sample shown in Fig.3.2 consists of a microcavity structure composed of two dis-
41
Ag
Alq3
+
DCM
dL dm dR
e1 e2
d1 d2
Fig. 3.2: Design of the structure investigated. Two Bragg reectors enclose the cavity layer and the silver layer of variable thickness (gradient from 0 to 40 nm). The
thicknesses of the layers being most sensitive to the eigenmodes are denoted by di .
tributed Bragg reectors (DBRs) with two embedded layers: a thin silver layer and an
organic half- cavity layer made of the host-guest system tris-8-hydroxyquinoline aluminium (Alq3) doped with 2 wt.percent of 4-dicyanomethylene-2 -methyl-6-p-imethylaminostyryl-4 H-pyran (DCM). The silver layer was deposited with a thickness graded
from 0 to 40 nm on top of the rst DBR (left hand side of Fig. 3.2) and then enclosed
by the organic cavity layer and the top DBR. The optical thicknesses of both the cavity and the 21 alternating /4 layers of TiO2 /SiO2 forming each DBR were chosen to
correspond to the maximum of the DCM emission at 630 nm, so the structure could
be used in future to demonstrate direct coupling between exciton and Tamm plasmon
states. However, the current structure used in the initial experiment discussed here
was for proof-of-concept only, and the concentration of DCM was actually too small to
achieve considerable exciton absorption. Due to the wedge shape, the silver layer only
partly covers the rst mirror resulting in a high quality, all dielectric microcavity as
well as two microcavities separated by a varying metal thickness up to 40 nm and each
capable of supporting Tamm plasmon polaritons that can couple through the thin silver
lm. A -photoluminescence (-PL) microscope setup was utilized to investigate the
angle-resolved emission spectra of the structure at room temperature. By focusing the
beam of a 405 nm cw laser with a microscope objective (25, NA 0.5) to a spot diameter
of less than 2 m, areas either with or without metal could be selectively excited. Such
a small spot size was necessary in order to avoid collecting PL-signal from areas with
dierent thicknesses of the silver layer and resulting in dierent spectral features. The
42
high aperture of the objective (63, NA 0.8) collecting the sample emission covered a
large angular range of around 55o in the far eld geometry. A second lens was used to
map the Fourier plane of the rst collecting objective, enabling the observation of the
dependence E(k). A polarization lter installed in front of the spectrometer was utilized
to characterize the polarization properties of the sample emission. The spectrograph was
equipped with a cooled charge-coupled device to record the spectrally, angularly, and
polarization-resolved PL signal.
3.2.2
Comparison of results of numerical modelling with experiment
In Fig.3.3, the experimental angle-resolved emission spectra of the metal organic microcavity are shown for dierent metal thicknesses (left panels of Figs.3.3(a-d). These are
compared to the results of numerical calculations based on the transfer matrix method
(right panels). In the numeric simulation the full transfer matrix has been calculated
based on the formalism presented in section 3.1.1. Transmission spectra were then calculated using (3.15) for dierent incidence angles resulting in the transmission coecient
map. Due to the small concentration of the DCM, the transmission spectra was almost identical to the experimental emission spectra. Three phenomena are apparent:
First, because of the graded thickness of the metal layer placed inside the cavity and
the small excitation spot used, an increasing metal thickness means moving away from
metal-free areas. Thus, any scatter of the emission into metal free areas is reduced,
leading to a weaker coupling into the original cavity mode at 632 nm. This behaviour
is not accounted for in the numerical calculations, since we assume a constant metal
thickness over large areas (compare left and right panels in Figs.3.3(a-d) at 632 nm).
Next, depending on the metal thickness, two modes arise. The spectrally broad emission above 700 nm corresponds to emission into the long wavelength sideband of the
DBRs. At silver thicknesses of 25 nm, an initially broad mode emerges from the long
wavelength sideband. This is a so-called Tamm plasmon-polariton, which has its origin
in the localization of the electromagnetic eld next to the metal layer. In addition to
this, a high energy resonance is observed, which at zero metal thickness converges to the
original cavity mode. As metal is introduced, it starts shifting to the red. The reason
for this shift is the coupling of this mode to the Tamm plasmon polariton evidenced
by a clear anticrossing behaviour of the modes involved. Finally, at oblique angles and
certain thicknesses of the embedded metal layer, a splitting of each mode is observed.
Moreover, this splitting increases towards large angles of incidence. We observe that
the polarizations of both branches are orthogonal to each other, characteristic of a splitting between TE- (lower branch) and TM-polarization (upper branch) modes. With
increasing metal thickness, the modes become increasingly detuned with respect to the
centre of the DBR stop bands. The most sensitive way to control these phase shifts
is by varying the thickness of the metal layer. In addition to the phase shift at the
dielectric/metal interface, the reectance and transmittance of this layer can be varied,
43
Fig. 3.3: Experimentally observed angle-resolved emission spectra (left panels) along
with numerically calculated spectra (right panels). Deviations of the shape of the
parabolae are due to imperfect refractive indices and their dispersions, on which the
simulations are based.
which aects the coupling of the resonant modes. The observed splitting is strongly
pronounced, reaching values of 37 meV for the shifted cavity mode and 45 meV for the
Tamm plasmon polariton state at angles 55o and a silver layer thickness of 40 nm.
3.2.3
Analytical expressions for the system eigenfrequencies
We start the analytical treatment of the system from writing down its transfer matrix:
(
A
1
rR
(
=
eiR
eiR
)(
(t t+ r r+ )/t r /t
r+ /t
1/t
)(
eiL
eiL
)(
rL
1
)
, (3.44)
where A is a constant, rL , rR are respectively the reection coecients from the left and
the right Bragg mirrors, L , R are respectively the phases gained in the left and the
right parts of the cavity, and r , t are the reection and transmission coecients of
44
the metal layer (reection (r) and transmission (t) from the left cavity to the right one
corresponds to a plus sign and the opposite direction to a minus sign).
Eliminating A from Eq.(3.44) gives the equation for the eigenfrequencies:
(
1
1
rL r+ e2iL
)(
1
1
rR r e2iR
)
=
t t+
.
r r+
(3.45)
Comparison with the Eq.(3.22) shows that the left side of the Eq.(3.45) contains two
factors, whose roots correspond to the individual Tamm plasmons at the left and the
right metal boundaries respectively. The right-hand side of Eq.(3.45) describes the
coupling between the two Tamm states. It is now necessary to express Eq.(3.45) in
terms of the parameters of the structure. We begin by looking at the right-hand side of
the equation.
For the TE polarization
t t+
=
r r+
(3.46)
4kzR kzL 2
,
cosh2 (dm )(kzL kzR )2 + sinh2 (dm )(2 + kzR kzL )2
where kzm =
m k02 kx2 , dm is the thickness of the metal layer, kzR,L =
R,L k02 kx2 ,
L = n2L , R = n2R , m = n2m are the dielectric permittivities of left cavity, right cavity
and the metal layer respectively, and = ikzm . There is analogous formula for the TM
polarization. We assume that the transmission through the metal layer is very small,
and thus the expression at the right-hand side of Eq.(3.47) is small, which will lead to the
small splitting of the Tamm plasmon states compared to the frequencies of the Tamm
plasmons. We then neglect the frequency and incidence angle dispersion of this coupling
term, since the frequency and wavvector shift corrections would be proportional to a
higher order of a small quantity. If we set = 0 and kx = 0 in Eq.(3.47), where 0 is
the Bragg reector stop band centre frequency, we get the expression which is the same
for both polarizations:
t t+
4nR nL |m |
=
.
2
r r+
cosh (|nm |0 /cdm )(nL |nm | |nm |nR )2 + sinh2 (|nm |0 /cdm )(|m | + nL nR )
(3.47)
We now consider the left-hand side of the Eq.(3.45), beginning with the coecients of
reection for the metal layer. When light is incident on the metal layer from the left
cavity in the TE polarization, the reection coecient is given by
rT E+ =
i(kzL kzR ) + tanh(dm )(2 kzR kzL )

.
i(kzL + kzR ) + tanh(dm )(2 kzR kzL )
(3.48)
45
At frequencies well below the plasma frequency, dm 1 and tanh(dm ) 1, and

rT E+
1 ikzL /
,
1 + ikzL /
(3.49)
which can be written in terms of low frequency and small angle of incidence as
( [
])
rT E+ exp i + 2nL /p b /nL /p / b 2 ,
(3.50)
where p is the metal plasma frequency, and b is the metal background dielectric constant. The corresponding results for the TM polarization are:
rT M + =
im (kzR nL /nR kzL nR /nL ) tanh(dm )(2 nL nR kzR kzL /nR /nL 2m )
,
im (kzR nL /nR + kzL nR /nL ) tanh(dm )(2 nL nR kzR kzL /nR /nL 2m )
(3.51)
rT M +
1 i(L )/(m kzL )

,
1 + i(L )/(m kzL )
(3.52)
( [
])
rT M + exp i + 2nL /p b + /nL /p / b 2 .
(3.53)
The expressions for rT E and rT M are obtained from Eqs. (3.50),(3.53) by changing
nL to nR .
To calculate the reection coecients for the Bragg reectors, that appear on the lefthand side of Eq.(3.45) it is instructive to recall that for a light in a medium A incident
on an interface with medium B the reection coecient can be expressed in terms of
the wave impedances ZA and ZB given by the ratio between the tangential components
of the magnetic and electric elds Zi = Hi /Ei at the interface:
rAB =
ZA ZB
.
ZA + ZB
(3.54)
The impedances in the cavities for both polarizations are given by simple expressions:
kz(L,R)
,
k0
n2L,R k0
.
=
kz(L,R)
TE
ZL,R
=
TM
ZL,R
(3.55)
(3.56)
The impedances at the internal interfaces of the Bragg reectors can be obtained using
the Bloch theorem. We assume that z = 0 coincides with an interface of the reector
and use the Bloch theorem:
(
)
E(z = D)
H(z = D)
(
= T
E(z = 0)
H(z = 0)
(
= eiKD
E(z = 0)
H(z = 0)
)
,
(3.57)
46
where D is the period of the structure, T is the transfer matrix for the period of the
Bragg reector, and K is the Bloch wavevector which can be obtained using Eq.(3.21).
The expression for the wave impedance then is
ZBr = H/E =
exp[iKD] T11
.
T12
(3.58)
T E,T M
The general expressions for rL,R
then from Eq.(3.54):
TE
rL,R
TM
rL,R
=
TE
kz(L,R) /k0 ZBr
TE
kz(L,R) /k0 + ZBr
TM
n2L,R k0 /kz(L,R) ZBr
TM
n2L,R k0 /kz(L,R) + ZBr
(3.59)
(3.60)
At frequencies in the vicinity of the Bragg reector stop band centre and at small
incidence angle reection coecients can be approximated as:
TE
rL,R
TM
rL,R
[ (
(
))]
nL,R ( 0 )
2
exp i
1
,
|n1 n2 |0
2n1 n2
[ (
(
))]
nL,R ( 0 )
2
exp i
1+
,
|n1 n2 |0
2n1 n2
(3.61)
(3.62)
where n1 , n2 are the refractive indices of the Bragg reector layers. The quantities on
the left-hand side of Eq.(3.45) that remain to be evaluated are the cavity transit phase
changes. L and R in the approximation of frequencies in the vicinity of the Bragg
frequency, for the small angles of incidence and for cavity optical thicknesses equal to a
quarter of the Bragg wavelength are given by:
(
2L,R
2
0
2
= 1+
0
2nL,R
)
.
(3.63)
Finally, the equation for the eigenfrequencies of the structure can be written in the form:
( T P,L ())( T P,R ()) = 2 =
=
(3.64)
n1
,
(2nL 0 /p + (n2 + nL n1 )/(n2 n1 ))(2nR 0 /p + (n2 + nR n1 )/(n2 n1 ))
02 (n2
)2
where is the expression at the right hand side of Eq.(3.47), and T P,(L,R) are the
angle-dependent eigenfrequencies of the Tamm plasmon polaritons in each cavity. The
eigenfrequencies of the structure for the two polarization are then given by:
1,2
.
T P,L + T P,R
=
(T P,L T P,R )2
+ 2 .
4
(3.65)
47
Fig. 3.4: (a) For the special case of a 40 nm silver layer inside the cavity, analytical
results (black solid and red dashed line) showing good agreement with the numerical
results. (b) Parabolic dependence of the splitting between the TE- and TM-modes of
the shifted cavity resonance and Tamm plasmon-polariton as a function of the angle of
incidence.
To verify the analytical formalism we have compared it with the experimental results
and numerical results. Fig.3.4(a) shows the good agreement between the analytical approximations and the numerical results. Figure 3.4(b) illustrates large TE-TM splitting
of Tamm plasmon modes.
In summary, the photon density of states of a microcavity changes substantially when
48
a silver layer is embedded. The cavity mode is red-shifted due to coupling with the
Tamm plasmon-polariton, and we are able to observe the associated spectral changes by
using the emission spectrum of an organic semiconductor, which provides large oscillator
strengths over a wide spectral range.
3.3
Eigenmode structure of periodic array of quantum wells

buried in layers of one-dimensional photonic crystal.
In this section we study the linear optical properties of the Bragg structure shown in
Fig.3.5(a), which comprises a periodic array of quarter-wavelength layers of GaAs with
refractive index n1 = nGaAs = 3.5 and pseudo-layers AlAs/GaAs/InGaAs/GaAs/AlAs
with eective refractive index n2 = nef f = 3.2 and eective optical length equal to
/4. The pseudo-layers have at their centre 10 nm wide In0.1 Ga0.9 As quantum wells
which can support excitons with a ground state energy of 1.418 eV. Fig.3.5(b) shows
experimental reection spectra at 32 K which have been obtained by our collaborators
in Crete. The structure has a wedge-like form, so that the eective layer thicknesses and
hence the photonic stop-band frequency changes along the structure. In particular, the
high-frequency edge of the stop-band, which is close to the exciton energy, varies from
1.38 to 1.44 eV.
The sharp extrema in the reection spectra correspond to the eigenmodes of the system.
It should be noted, that the exciton eigenfrequency can correspond to both the dips in
the reection spectra when the electric eld antinodes are in the vicinity of the quantum
wells and the sharp maxima in the reection when the electric eld nodes are in the
vicinity of the quantum wells.
The dispersion of the eigenmodes can be extracted from the reection spectra. The
anticrossing behaviour of the exciton mode and a Bloch photonic mode positioned at
the band-gap edge can be observed in the spectra. Moreover, in the case of resonance
between the photonic band-gap edge and the exciton frequency, a triplet structure is
observed in the spectra. The spectra have been calculated the spectra numerically using
the transfer matrix method and are depicted in Fig.3.5(c). It can be seen, that the
numerical model agrees well with the experimental data. However it is also instructive
to derive some analytical expressions for the eigenmode frequencies.
A thorough analysis of the similar structures was performed by Ivchenko et al. in
2004 [54]. Using that theory, it is straightforward to show that in the case of the
quarter-wavelength photonic crystal with quantum wells, the mode structure is dened
49
30 periods
Experiment
TMM modelling
eV
Fig. 3.5: (a) Geometry of the structure under consideration. (b) Experimental reection spectra. (c) Reection spectra calculated with transfer matrix method.
by the equation
(
)
n2 + n22
cos2 () 1
sin2 ()
2n1 n2
(
)
(n1 + n2 )2
n1 n2
0
sin(2) 2
sin()
X i 2n1 n2
n1 + n2
cos(KD) =
(3.66)
for the Bloch wavevector K, where = /(2B ). Here B is the stop-band centre frequency, - is the non-radiative exciton decay rate, 0 is the radiative exciton broadening
given by Eq.(3.42), X is the exciton frequency, and D is the period of the structure. In
what follows we will neglect the non-radiative exciton decay rate , which is negligible
for the cryogenic temperatures considered. We note that rst terms enclosed in the rst
parentheses on the right hand-side coincide with the dispersion relation for the Bragg
reector without quantum wells and thus correspond to the non-resonant contribution
50
to the dispersion. The terms enclosed in the last parentheses correspond to the exciton contribution to the dispersion. For the analysis it is convenient to rewrite equation
(3.66) in two equivalent forms:
(
sin
KD
2
(
(
))
(n1 + n2 )2
0
n1 n2
=
sin sin
cos
,
4n1 n2
X
n1 + n2
(
cos
KD
2
(
)
0
(n1 + n2 )2
cos cos +
=
sin .
4n1 n2
X
(3.67)
(3.68)
Equation (3.67) is convenient to obtain the mode frequencies of the band edge corresponding to K = 0, and Eq.(3.68) for the frequencies of the band edge, corresponding
to K = /D. We can then expand the sine and cosine functions for frequencies close
to the stop band centre and obtain expressions for the band edges in closed form. The
details can be found in [54].
Here we describe an alternative way to obtain the expressions for the eigenmode frequencies in a closed form, which also has the advantage of illustrating the mechanism
of formation of the spectral triplet structure seen in gure 3.5. The formalism is mainly
based on the model presented in [55] for the bulk excitons in photonic crystals.
We rst write down the energy density of the electromagnetic eld in the photonic crystal
as
Eph
1
=
2
[
]
dr (z)E2 + H2 .
(3.69)
The dielectric permittivity (z) is a periodic function of z:

(a) = + F (z),
(3.70)
F (z + D) = F (z),
where is the average dielectric permittivity and the periodic function F (z) has unit
amplitude. If the amplitude of the dielectric permittivity modulation is relatively weak,
that is /
1, then we can make an approximation of only considering the rst
Fourier component of the periodic function F : F = cos(Gz + 0 ), where G = 2/D is
the smallest reciprocal lattice vector, and 0 is a modulation phase, which is zero for a
suitable choice of the coordinate system. The Fourier components of couple the terms
with k and kG and if we consider only the frequencies close to the rst Bragg resonance
at k G/2 = /D, we can need only consider the rst diraction order G k G/2.
51
Then the full photon Hamiltonian may be written as:

Hph =
[
]
(q)
a+
a (q) + (q)
a+
a (q) +
(q)
(q)
(3.71)
[ +
]
a (q) .
a (q) + a
+
+ B a
(q)
(q)
Here q = kG/2, a
(q) = a
k , a
(q) = a
kG , where a
k is a photon annihilation operator
and B is the band gap half-width which can be estimated by photonic crystal theory
to be: B B (n1 n2 )/(n1 + n2 ). (q) can be written as (q) = B (1 + 2q/G).
The exciton kinetic energy and exciton-photon interaction Hamiltonians are
HX = X
b+ (k)b(k)+
(3.72)
]
[
b+ (k)
a(k) + a
+ (k)b(k) ,
+ R
k
where X is the exciton frequency, b(k) and b+ (k) are respectively exciton annihilation
and creation operators and R is half the Rabi splitting, which describes the coupling
of the excitons with the electromagnetic eld. We should note that in the case of a
periodic array of quantum wells, and contrary to the case considered in [55], k is not a
true wavevector for the exciton since the component of the wavevector perpendicular to
the interface is not dened for the exciton in a quantum well, being a Bloch wavevector
of the periodic quantum well array. We neglect the energy dispersion of the exciton due
to its relatively large eective mass. In the case of the excitons with the wavevectors
close to the photonic band gap, approximations similar to those used for Hph can be
employed to obtain the full system Hamiltonian:
H=
+
]
(q)
a+
a (q) + (q)
a+
a (q) +
(q)
(q)
(3.73)
]
[
[ +
]
b+ (q)b (q) + b+ (q)b (q) +
(q)
a (q) + a
+
(q)
a
(q)
+
b a
]
[
+
+
b+ (q)
b (q) + b+ (q)
b (q) .
+ R
a
(q)
+
a
(q)
a
(q)
+
a
(q)
Intoducing the vector operator v = [

a (q), a
(q), b (q), b (q)], the full Hamiltonian
can be written as: H = v+ H
v , where the H is the matrix:
(q)
B
H=
R
0
(q)
R
.
0
(3.74)
The eigenvalues of H are the system eigenmode frequencies. But to obtain the eigenvalues
it is convenient to simplify H rst. We rst change the basis from v to basis v =
52
, L,
X
s, X
a ], where U
and L
are the annihilation operators of the lower and upper Bloch
[U
s and X
a are the annihilation operators for symmetric and antisymmetric
modes, and X
exciton modes. If v = M
v , then:
M=
1
1+((q)U )2 /2B
1+((q)L )2 /2B
1
1+2B /((q)U )2
1
1+2B /((q)L )2
0
0
1
2
1
2
0
,
1
2
12
(3.75)
where U,L are the eigenfrequencies of the Bloch modes, given by:
(
U,L = B
2
q2
4 2 + 2B
G
B
)
.
The Hamiltonian H in the new basis is H = M1 HM and is:
2R
U
0
0
2
0
L
R
.
H =
0 2
X
0
R
R
0
0
X
(3.76)
(3.77)
Here we should recall that the exciton frequency in the experiment is close to the upper
edge of the photonic band gap, and furthermore the Rabi frequency R is much smaller
than the photonic band gap width B . As a result, the lower Bloch mode is remote in
frequency from the other three modes and we can assume its dispersion is essentially
unperturbed and that it does not perturb the other modes signicantly. Excluding the
from the Hamiltonian gives:
column and row corresponding to L
H = 0
X
2R 0
2R
0 ,
X
(3.78)
and introducing = U X , we nd the system eigenfrequencies to be

1 = X ; 2,3
(
)
2
2
= X + + 8R /2.
(3.79)
In the case of a nite structure the Bloch vector K has a set of discrete values, which
correspond to discrete values of U and we should change U to U 1 in our analysis,
where U 1 is the frequency of the discrete Bloch mode closest to the upper band edge
as illustrated in Fig.3.6).
Figure 3.7 shows a comparison of the experimental results with the predictions of the
53
R
U2
U1
/D
Fig. 3.6: Band structure of the photonic crystal without quantum wells. Red lines show
the dispersions of the propagating photonic modes a
(q),
a (q), black lines show the
,L.
Right picture shows the reection spectrum from
dispersions of the Bloch modes U
the nite structure. Reection coecient minima correspond to discrete Bloch modes,
which are characterized by frequencies U 1 , U 2 , etc.
analytic model. It is seen that the analytic results for the mode frequencies as a function of the dierence between the photonic band edge and the exciton energy generally
agree well with the experimental data. The theory does not predict any dispersion of
the middle branch 1 = X as a result of excluding the lower Bloch mode from the
Hamiltonian. However, the experimental results conrmed that the dispersion is very
weak.
In the experiment the exciton frequency was tuned to the low frequency edge of the
second photonic band. However, it is also possible to tune the frequency to the high
frequency edge. In this case the Bloch wavevector is close to zero, and we can introduce
an eective mass for the Bloch mode, which will be negative close to the photonic band
edge. We have carried out numerical modelling of this case using the same refractive
indices as in the experiments and layers giving a Bragg frequency equal to 1 eV.
eV
1.44
Transfer matrix method

Experiment
Analytical expression
Line bandwidth
54
Photonic
Bloch mode
1.43
Exciton mode
1.42
1.41
Photonic band gap
meV
eV
Fig. 3.7: Dependence of the system eigenfrequencies on the detuning between the
exciton energy and photonic band edge, obtained from experiment, transfer matrix
method and with the analytical model.
Fig. 3.8: Eigenmode dispersion in the case of tuning of the exciton mode to the upper
edge of the second photonic band. Exciton frequency - 1.96 eV. Refractive indices of
the layers n1 = 3.5, n2 = 3.2, and thicknesses of the layers d1 = 133 nm, d2 = 43 nm.
55
The second band gap vanishes in a quarter-wavelength Bragg reector, so the layer
thicknesses and refractive indices have been chosen so that: n1 d1 = (3/8)B ; n2 d2 =
(1/8)B , where B is a Bragg wavelength. Transfer matrix calculation results given in
Fig.3.8 show that in the case of the exciton frequency being in resonance with the upper
edge of the second band at 1.96 eV, two polaritonic states emerge, and one of those
states is characterized by negative eective mass and group velocity in the direction
perpendicular to the layers interfaces. Thus, the eective mass tensor for this mode has
two positive and one negative components.
To conclude, in this section we have demonstrated a simplied analytical model, which
illustrates the mechanism of triplet structure formation in the reection spectra. It has
been shown that the eigenfrequencies measured experimentally agree well with the predictions of the analytical model. It has also been demonstrated that the strong coupling
of the photonic crystal eigenmodes and excitons can lead under some circumstances to
the formation of polaritonic modes characterized with negative eective mass and group
velocity.
3.4
56
Parametric amplication by a periodic array of quantum wells embedded in layers of a one-dimensional

photonic crystal.
3.4.1
Parametric amplication of exciton-polaritons
The eigenmodes of the system, discussed in section 3.3 are mixed exciton-photon modes
- exciton-polaritons. Here we consider the phenomenon of parametric amplication of
exciton-polaritons which occurs due to the scattering of exciton-polaritons and is one of
the two main nonlinear eects occurring in the polaritonic systems the other being a blue
shift of the ground polaritonic state [48]. The mechanism of the parametric amplication
can be understood by looking at the lower polariton dispersion curve shown in Fig.3.9.
The non-parabolicity of the ground polaritonic branch leads to the existence of a so-
Fig. 3.9: Dispersion of the exciton-polariton. Arrows show how the two pump polaritons p scatter elastically into a ground state s and an idler state i.
called magic wavevector m , where two polaritons with such a wavevector can scatter
elastically, with one scattering into the ground state and the other into the state with
wavevector 2m . Thus, if we pump such a system resonantly at the magic wavevector and
57
frequency, it is possible to observe a strong luminescence enhancement from the ground

state. Studies of polariton-polariton scattering features, which lead to the parametric
amplication have been reported in a number of theoretical and experimental papers see for example [56, 57].
A series of pump probe experiments have been performed by the Cambridge group
with the sample discussed in section 3.3, which have revealed parametric amplication
occurring in the structure. The main experimental results have been presented in [47].
We have been able to show by numerical simulation of these results, that the theory of
parametric amplication in conventional microcavity [48] can be successfully applied to
the case of a periodic quantum well array.
3.4.2
Experimental measurements and results
In the experiments, which were performed by the Cambridge group, the system was
pumped with pulses of 1ps length and 3.5 meV linewidth, at an angle of incidence
corresponding to the magic wavevector. A broadband 150 fs probe pulse was incident
normal to the interface of the structure and time-dependent reection spectrum were
measured. The violet line in Fig. 3.10(a) shows the reection spectra in the absence of
a pump pulse including dips at the frequencies corresponding to the three polaritonic
modes, which were studied in section 3.3.
If a pump pulse is applied with zero delay time relative to the probe pulse, the reection
spectrum changes dramatically, and there is a peak in the reection spectrum at a
frequency which is 3 meV higher than the ground polaritonic state. The peak height
corresponds to the enhancement of the probe pulse by a factor of 6.4 (see Fig. 3.10(c)).
The time dependence of the normalized reection spectra in Fig.3.10(b) indicates a
nonlinear response with a response time of about 1.8 ps, which is close to the values
typically obtained for the microcavity polaritons [58]. However, the polariton lifetime
in this system is less than for conventional microcavity polaritons resulting in a wider
band of polaritonic states and thus a wider enhancement peak (about 2.5 meV).
3.4.3
Three-level model of parametric amplication
It has been shown in [48], that the process of parametric scattering of two polaritons p
into the signal s and idler modes i is described by the Hamiltonian,
(
)
(3.80)
H = ~p a
+
p + ~s a
+
s + ~i a
+
i + ~M a
p a
p a
+
+
s a
i a
+
+
pa
sa
sa
pa
p +
i a
i +a
(
)
(
)
+ ~U |Xs |2 a
+
s + |Xp |2 a
+
p + |Xi |2 a
+
i |Xs |2 a
+
s + |Xp |2 a
+
p + |Xi |2 a
+
i ,
sa
pa
sa
pa
i a
i a
where s , i , p are the frequencies of the signal, idle and pump modes respectively, a
are the annihilation operators, and X are the Hopeld coecients [59], which dene
58
delay time (ps)
eV
probe
pump
probe gain
gain
incidence angle
delay time (ps)
eV
Fig. 3.10: An illustration of parametric amplication. (a) Reection spectrum in

the case of a single probe pulse (violet line). The green line corresponds to a 50 mW
pump pulse, which is tuned to the ground polaritonic branch with wavevector equal to
the magic wavevector (corresponding to an incidence angle 20 degrees). In this case
there is strong amplication of a probe pulse (red line). (b) Colour map showing the
dependence of the reection on the frequency and delay time. (c) Dependence of the
probe pulse enhancement on the delay time.
the excitonic part of a polaritonic state . The rst three terms correspond to the
kinetic energy of the polaritons in the three state, the fourth term denes the energy of
the parametric processes, M being the matrix element of these processes, and the last
term, proportional to U , corresponds to the blue-shift of exciton polaritons caused by
the exciton-exciton interaction. In what follows we will neglect the last term to simplify
the formalism, taking advantage of the fact that the eect of blue shift was weak in the
experiment. Reference [60] gives the following expression for M
M
12 a2B 2
E |Xs |2 |Xi |2 |Xp |4 ,
~ S b
(3.81)
where aB is an exciton eective Bohr radius, S is the sample area, and Eb is the exciton
binding energy.
To derive the equations for the dynamics of the occupation numbers of the three states
considered we rst introduce the density matrix of the system:
=
j=s,i,p
wj |j j |,
(3.82)
59
where j is the eigenfunction of each state, and wj is the probability of the polariton to
be in one of the states. The density operator obeys the Liouville-von Neumann equation:
i~
d
= [H, ].
dt
(3.83)
j of the number operator N

j = a
The expectation value N
+
j can be obtained as a
j a
trace over the density matrix:
Nj = Tr(
a+
j ).
j a
(3.84)
To show this, it is instructive to recall that for an operator A in a nite-dimensional

vector space with an orthonormal basis set |j the trace of operator A is dened as
=
Tr(A)
j=n
= Ajj ,
j|A|j
(3.85)
j=1
where the summation is performed over the repeated index. We can then expand Tr(A)
using equations (3.82) and (3.85) as:
=
Tr(A)
=
wk j|k k |A|j
wk k |A|jj|
k =
(3.86)
k ,
wk k |A|
since
|jj| is an identity matrix. Therefore we can can write down the equation for
the expectation value of the occupation number of state k in the form:

dNk
i
= Tr(
a+
k [H, ]).
ka
dt
~
(3.87)
Using the commutation rules for the operators we obtain:

dNs
= 2Im[M Tr(
a+
+
p a
p )],
sa
i a
dt
dNi
= 2Im[M Tr(
a+
+
p a
p )],
sa
i a
dt
dNp
= 4~Im[M Tr(
a+
+
p a
p )].
sa
i a
dt
(3.88)
(3.89)
(3.90)
+
p a
p = C is a fourth-order correlator denoting the correlations beThe operator a
+
sa
i a
tween dierent states, in the classical limit, this correlator is equal to zero since no
correlations are accounted for. To derive the dynamics of this correlator we use the
same approach taking the appropriate trace:
(
)
dC
i
= Tr(
a+
+
p a
p [H, ]) = iM Np2 (Ns + Ni + 1) 4Ns Ni (Np + 1) .
sa
i a
dt
~
(3.91)
60
We should also add phenomenological terms corresponding to the polariton decay, Ns,i,p /s,i,p
where the s,i,p are the polariton decay times in the respective states:
(
=
|X |2 |C |2
+
X
c
)1
,
(3.92)
where c , X - are the photonic and excitonic decay times. The Hopeld coecients
for the states X , C are the components of the normalized eigenvectors of the matrix (3.78), corresponding to the lower polariton branch. The photon lifetime is the
inverse imaginary part of the eigenfrequency of the system in the absence of excitons
and can be calculated by the transfer matrix method. The approximate values of the
excitonic lifetime are known from the experiment and are about 1 ns [61].
We should also add the terms corresponding to the pump. In the experiment, there
are two pumping mechanisms: resonant pumping to the pump state and nonresonant
pumping with a probe pulse, which pumps all the three states. Therefore, We introduce
three dierent pumping terms Pp (t), Ps (t), Pi (t) corresponding to the time-dependent
pump rates in each of the states. Furthermore, initial conditions should be imposed.
Here we choose for these to be an initial absence of polaritons. As a result we get a
system of four rst-order nonlinear dierential equations with zero initial conditions:
N p = Im(M C) Np /p + Pp (t),
(3.93)
N s = 2Im(M C) Ns /s + Ps (t),
N i = 2Im(M C) Ni /i + Pi (t),
)
(
( 2
)
1
1
1
+
+
C,
C = iM Np (Ns + Ni + 1) 4Ns Ni (Np + 1)
2s 2i p
S|t=0 = ns |t=0 = np |t=0 = ni |t=0 = 0.
Some Numerical modelling results are shown in Fig.3.11. Figure 3.11(a) shows the
evolution of the occupancy numbers for the pump and signal states in the case when the
both pulses arrive at the same time. We can see that the nonlinear response time t is
about 0.75 ps, which agrees well with the experimental result (1.5 ps, see Fig.3.10(b)).
Figure 3.11(b) shows a plot of maximum amplication versus the pump-probe delay
time. We observe that the maximum amplication is 10 which is fty percent larger
than the experimentally observed value (see Fig.3.10(c)).
61
np (a.u.)
ns (a.u.)
gain (a.u.)
time (ps)
delay time (ps)

Fig. 3.11: Numerical modelling of the experiment, performed by the Cambridge group,
with the sample considered in section 3.3 demonstrating parametric amplication in the
system. (a) Time dependence of the pump state (red line) and signal state (blue line)
occupancies in the case when the pump and probe pulse arrive at the same time t = 3
ps. Blue dashed line shows the evolution of the signal state in the absence of the pump
pulse. All the values are normalized to the maximum occupancy of the signal state in
the absence of a pump pulse. t shows the nonlinear response time of the system. (b)
Dependence of the probe pulse amplication on the pump-probe delay time.
3.5
Brief summary
We have obtained following main results in this chapter:

We have derived analytical expressions for the optical properties of a system consisting of adjacent metal and organic layers in a microcavity. It has been demonstrated that sharp spectral features occur in the reectivity specta of the structure
62
due to the strong coupling of Tamm plasmon polariton states at each of the metal
interfaces.
We have developed a simplied theory of the interaction of a periodic array of
quantum well excitons with photonic crystal optical eigenmodes. We have calculated theoretical reection spectra for these structures which agree well with the
experimental data. We have also predicted a new type of polaritonic mode.
We have shown that Bragg polaritons can be characterized by negative group
velocity and eective mass in the case when the exciton frequency is tuned to the
upper edge of the second photonic band.
We have performed modelling of the parametric amplication of these polaritons
within a three level formalism. The values of the nonlinear response time and
maximum signal amplication obtained agree well with the experimental results.
Chapter 4
Kinetics of exciton-polariton
condensates and the modelling of
polariton lasers
4.1
Introduction
This chapter describes work on modelling of the kinetic processes in polariton laser
diodes [62], which was motivated by and performed in collaboration with an experimental
group from the Ecole

Politechnique Federale de Lausanne (EPFL). The work covered
the modelling of both the steady state properties of polariton lasing, including the
temperature and detuning dependence of the lasing threshold, and the small signal
response of certain polariton laser diodes, where the device transfer function has been
calculated1 .
The chapter is organized as follows. In the rst section the basic concepts of polariton
condensation and polariton lasing are discussed and a brief historical overview of the
eld is presented. The next section is dedicated to a discussion of possible designs of
the polariton laser diode and is largely based on the original ideas of our experimental
collaborators. The third section describes the kinetic processes occurring in the laser
and presents the semi-classical Boltzmann equations formalism which is the basis of the
modelling carried out. That is followed by a presentation and discussion of the numerical
results obtained from the modelling. A brief summary of the chapter is presented in the
last section.
My contribution to the work involved developing and performing the modelling based on semiclassical Boltzmann equations using parameters provided by our experimental collaborators. The presentation of the work closely follows reference [62], for which I was first author.
63
Chapter 4. Kinetics of exciton-polariton condensates
4.2
4.2.1
64
Background theory
Bose-Einstein condensation of exciton-polaritons
Consider N non-interacting bosons in a system of linear dimensions L and volume Ld ,

where d is the dimensionality of the system. If the system is in thermal equilibrium
at temperature T , the bosons are distributed in energy according to a Bose-Einstein
distribution:
[
f (k, T, ) =
exp
1
E(k)
kB T
(4.1)
where k is the boson wavevector, E(k) is the energy of state k, and kB is the Boltzmann
constant. In what follows we assume that the ground state k = 0 corresponds to the
zero energy. is a chemical potential, which denotes the energy needed to add a particle
to the system, and for bosons is always less than or equal to zero. The total number of
the particles in the system can be written as
N (T, ) =
(4.2)
f (k, T, ).
To proceed it is instructive to split the above sum into the number of particles in the
ground state k = 0 and in all the excited states:
N (T, ) =
1
f (k, T, ).
+
exp [/kB T ] 1
(4.3)
k=0
Moving to the thermodynamic limit,(increasing the number of the particles and the
system size indenitely in such a way, that the density of the particles n = N/Ld
remains constant) the particles density is:
1
1
1
n(T, ) = lim
+
d
L+ L exp [/kB T ] 1
(2)d
f (k, T, )dk.
(4.4)
If is nonzero, then the rst term vanishes. On the other hand, the second term, which
denotes the density of bosons in the excited states, is an increasing function of , and
we can dene the maximum possible density of bosons in the excited states nc as:
1
nc (T ) = lim
0 (2)d
f (k, T, )dk.
(4.5)
If this integral converges, it means that there exists a maximum density of bosons nc
that can be accommodated in the excited states and if the total density n is larger than
65
nc then is zero and there is a density n0 = n nc in the ground state resulting in a

macroscopic occupation of a single quantum state.
It has been shown that the integral (4.5) converges for d > 2. For the case of three
dimensional system and parabolic dispersion the integral can be evaluated analytically
as:
1
nc =
2
)
( )(
mkB T 3/2
3
,
2
~2
(4.6)
where (x) is the Riemann zeta function. Note, that the critical concentration of the
particles is proportional to (mT )3/2 , and thus increases with temperature and boson
mass. Let us consider the xed total concentration of bosons n and consider the system in
thermal equilibrium at temperature T . If for this temperature the critical concentration
nc given by (4.6) is larger than n, there is no condensation in the system. If we start to
decrease the temperature adiabatically, at some nite temperature Tc , nc becomes equal
to n. For temperatures lower then Tc , the macroscopic occupation of the ground state
would be observed. Tc is called the Bose-condensation temperature, and is inversely
proportional to the boson mass.
The rst clear experimental verication of Bose condensation in a weekly interacting
system was achieved by Anderson et al. in 1995 using a very dilute gas of alkali atoms
at a temperature of hundreds nanokelvin and was rewarded by the Nobel Prize for
Physics in 2001.
There are some specic features about the condensation of exciton-polaritons to consider.
The rst is the fact that the exciton-polaritons are 2-dimensional particles, and the
integral for nc in equation (4.6) does not converge for n = 2, suggesting that there can
be no thermodynamic Bose condensation. However, this restriction can be overcome
if we recall that the polariton condensation usually takes place in a nite area and
that the polaritons have nite lifetime. It means, that there can be no thermodynamic
equilibrium for the polaritonic system. Another issue is that the exciton is actually a
composite particle, an electron-hole pair, which can be considered as a boson only in
the low-density regime. It is easy to estimate an upper limit for the concentration for
which the approximation of a boson gas is valid. It is required that the average distance
between the excitons a in the gas is much larger then the exciton Bohr radius aB :
a n1/2 aB .
(4.7)
The corresponding condition for the exciton concentration n is n 1/(aB )2 . The
condition for the exciton-polaritons can be written down as:

nk |X|2k 1/(aB )2 ,
where nk is the concentration of polaritons in state k and Xk is the Hopeld coecient
showing the exciton fraction in the polariton state k.
66
Another important feature of the polariton condensation is the very high temperatures
at which the condensation can take place which is due to the very small eective mass
of polaritons close to the ground state. This property makes possible the observation of
Bose-condensation and related phenomena in semiconductor systems at relatively high
temperatures and even up to room temperature in certain systems. As a consequence,
the physics of polariton condensates has become one of the most rapidly expanding and
exciting elds of condensed matter physics. The observation of the Bose-condensation
of polaritons occurring in CdTe-based planar microcavities at cryogenic temperatures
was rst reported in [63] and triggered the further research on the phenomenon. Bosecondensation was later reported at room temperature in wide bandgap organic and
inorganic semiconductors which exhibit highly stable excitons [6466]. Some of the
exotic phenomena occurring in polariton condensates include hints of superuidity such
as integer and half-quantized vortices [67, 68] and the ballistic motion of condensates
[69, 70]; and also spontaneous coherent oscillations between adjacent condensates that
share similarities with the ac Josephson eect [71].
4.2.2
Polariton lasing
Our work has been dedicated to the study of one of the most attractive of polariton
condensation phenomena with regard to potential applications - namely polariton lasing
[9]. A schematic diagram of the polariton laser is shown in g.4.1(a). To understand
the basic concept of polariton lasing, consider a polaritonic system which is being continuously pumped nonresonantly, as shown in g.4.1(b), creating polaritons in high k
states. Polaritons then either relax into the ground state or decay either radiatively
or non-radiatively. If we assume that all the polaritons excited have innite lifetime,
then according to the theory presented previously, after some certain density of the
excited states is reached, all other polaritons would fall into the ground state, where
they would decay radiatively producing coherent emission of light (laser-like light) from
the microcavity. This coherent emission needs no carrier population inversion since the
emission is spontaneous, not stimulated, and therefore should have a signicantly lower
threshold pumping requirement than in a conventional laser. In reality polaritons have
a nite lifetime and it is essential that the relaxation time of the excited polaritons to
the ground state is less then the polariton lifetime. However, due to the bosonic nature
of the polaritons, the relaxation rates are proportional to both the initial state occupation and the nal state occupation, the condition on the relaxation time is equivalent to
the condition of the critical polariton density in the system, and thus to a xed nite
threshold pump power.
Electrically pumped polariton light emitting diodes (LEDs) have been reported in [72]
to operate in a wide range between cryogenic and room temperatures. These structures,
obviously operating in the strong coupling regime (SCR), have a design relatively close
67
(a)
coherent
emission
(b)
pump
coherent emission
DBR
p contact
Polaritonic
QWs
n contact
p contact
pump
relaxation to
the ground
state
depletion of
the ground
state
n contact
DBR
0
Fig. 4.1: (a) Schematic diagram of a polariton laser. (b) Schematic illustration of the
kinetic processes in the system: pump to the excited states, relaxation to and from the
ground state, and coherent emission from the ground state due to the nite polariton
lifetime.
to that of their near-infrared vertical cavity surface emitting laser (VCSEL) counterparts
(which are based on the same material system) operating in the weak coupling regime.
However, the realization of room-temperature (RT) GaAs-based polariton laser diodes
B ) in this
(LDs) is probably precluded by the value of the exciton binding energy (EX
material system (usually less than 10 meV for the quantum wells (QWs) of interest)
[73], which is small compared to the thermal energy at 300 K. Indeed, it has been preB is the main limiting factor for the high-temperature
viously shown that the value of EX
observation of optical nonlinearities of polaritonic origin under resonant excitation [74].

In the search of the material systems dened by the larger values of the exciton binding
energy, a strong interested has been directed to the III-nitride compounds [75], which
have a number of advantages compared to the other wide bandgap semiconductor systems such as organic compounds or ZnO. It is suggested that III-nitride systems could
have threshold current densities (Jthr ) as low as 10 Acm2 at room temperature,
which is two orders of magnitude less than state of the art GaN based edge-emitting
diodes. The reason for such a low threshold is because that the condensation process is
expected to occur at medium polariton densities - lower then those causing signicant
screening and phase space lling eects that would lead to the transition of polaritons
and excitons toward an electron-hole plasma - because of the low eective density of
states of polaritons resulting from their very light mass near the center of the Brillouin
zone.
4.3
68
Designs of the polariton lasers considered
To begin, it is useful to consider some aspects of III-nitride structures relevant to the

implementation of an electrical injection scheme suitable for polariton LDs [75] 2 . A
recent theoretical study of the characteristics of a bulk GaN polariton light emitting
diode LDs predicted a Jthr value of 50 Acm2 at room temperature (RT), thereby
indicating the potential of such devices as low-threshold coherent light emitters [76].
Although such a structure has the advantage of simplicity it should be noted that a
realistic design for RT operation will most likely rely on a multiple QW (MQW) active
region because of the improved carrier connement and the higher exciton binding energy
over bulk in such two-dimensional heterostructures.
The number of quantum wells (NQW ) should be large enough to achieve a well dened
strong coupling regime signature at room temperature i.e. to get a suciently large Rabi
splitting (V RS ) to polariton linewidth ratio [77]. The latter is due to the detrimental
impact of the inhomogeneous linewidth broadening of III-nitride QWs, which can blur
the signature of polaritons if it is too large for a small NQW value (note also that the homogeneous broadening is usually non-negligible at 300 K) [78]. To get ecient electrical
injection, the most appropriate choice of active region would be obtained by switching
from the usual GaN/AlGaN MQW system to InGaN/GaN MQWs [65, 79] since good
p-type conductivity becomes progressively more dicult to obtain when increasing the
Al content of the AlGaN layers due to the signicant rise of the activation energy of
the deep Mg acceptor level [80]. At rst sight, InGaN/GaN MQWs might seem an unpromising choice to achieve strong coupling because of their large inhomogeneous QW
linewidth. But low indium content (7% < x < 10%) Inx Ga1x N/GaN MQW microcavity structures with a design close to that of GaN/AlGaN MQW microcavity samples
[65, 77] are expected to exhibit clear strong coupling features and polariton lasing under
optical pumping at RT. Such an approach is also motivated by the recent development
of crack-free, lattice-matched AlInN/GaN distributed Bragg reectors (DBRs) grown on
freestanding GaN substrates that combine low dislocation density ( 106 cm2 ), smooth
rms surface roughness ( 0.26 nm for a 4 4 m2 area) and high peak reectivity (R
99.6 %) [81]. Microcavity structures grown on such a DBR would certainly exhibit a
much improved optical quality, i.e., a reduced in-plane cavity disorder compared with
similar structures grown on c-plane sapphire substrates [82].
With such microcavity samples, another matter to be addressed is the achievement of
a uniform injection of electrons and holes into the active region, which is known to be
an issue in III-nitride optoelectronic devices [83]. A way to circumvent the conicting
2
like much of this chapter the presentation in this subsection closely follows reference [62], of which
I am first author. However, the discussion in this subsection is mainly based on work carried out largely
by the authors of [62] at EPFL and is very similar to work previously published by them in [75]. It does,
however, provide important practical background information to theory provided here and is included
for that purpose.
69
requirements presented by a structure where NQW should be large for the realization of
strong coupling and small for electrical injection would be to use an intracavity pumping
geometry as illustrated in Fig.4.2 [84]. Within such a pumping scheme, a small number
of QWs (QWs-1) sandwiched in the intrinsic region of a p-i-n diode would be electrically
pumped. Polaritons would not be formed in these QWs but they would emit photons at
an energy greater than the absorption edge of a MQW region (second QW subset, QWs2) located underneath, which would be in the strong coupling regime when pumped by
the emission from QWs-1. Other specic constraints regarding the design of polariton
LDs have been described in Ref. [75] and are included in the three-dimensional (3D)
cross-section of the InGaN/GaN MQW polariton LD displayed in Fig.4.2 An intracavity
contact scheme is used, as described in the case of III-nitride VCSEL structures [85].
To compensate for the relatively poor lateral hole spreading into the p-type GaN layer
that is mainly due to current crowding - a detrimental eect whose impact is enhanced
by the annular contact geometry - and thus to get light emission from the active region
sandwiched between the DBRs, various approaches can be implemented. First a buried,
oxidized AlInN interlayer can be inserted on the n-side underneath QWs-1 in a similar
fashion to that implemented for micro-LEDs [86] to conne the electron current ow in
the central part of the device. A transparent conductive oxide (TCO), like indium tin
oxide or ZnO, sandwiched between the p-type GaN layer and the top dielectric DBR [87]
( Fig.4.2) also might be used to improve the lateral spreading of the hole current. Then
the use of an electron-blocking layer (EBL), located on top of the electrically-pumped
region can act to avoid an excess of electrons on the p-type side and thus limit unwanted
electron-hole recombination.[88]
Beyond issues related to electrical pumping, another critical parameter when considering
the above-mentioned geometry is the signicant increase in the eective cavity length
(Lef f ). Indeed, it is known that as a rst approximation V RS scales like 1/ Lef f

[89]. However, Lef f also includes a contribution associated with the penetration depth
of the electromagnetic eld into the DBRs and one way to mitigate this eect is to use
a top dielectric DBR with a short penetration depth. This can be achieved by adopting
the SiO2 /TiO2 bilayer system since it exhibits a very large refractive index contrast
(nT iO2 ( = 415 nm) = 2.6 vs nSiO2 ( = 415 nm) = 1.495) [75]. As an illustration, the
penetration depth obtained at = 415 nm only amounts to 0.83/2 for a SiO2 /TiO2 DBR
against 2.47/2 for the more conventional SiO2 /Si3 N4 structure, where nSi3 N4 ( = 415
nm) = 1.83 [90]. Note, however, that the SiO2 /TiO2 DBR is not suited to a nonresonant
optical pumping due to the strong rise in the absorption in TiO2 layers for wavelengths
shorter than 375 nm. However, we have used this structure in the simulations. The
electric eld prole at = 415 nm of the complete structure shown in Fig.4.2, derived
from a transfer matrix simulation, along with the corresponding refractive index prole
are displayed in Fig.4.3. The refractive indices for the III-nitride compounds were taken
from the work of Brunner et al.[91] (AlGaN alloys), Carlin et al.[92, 93] (AlInN alloys)
70
and Bergmann and Casey [94] (InGaN alloys), respectively, and the refractive index
value for ZnO was taken from the work of Schmidt et al. [95].
TCO hole spreading layer
top dielectric DBR
p-type AlGaN EBL
p-type annular contact

p-type GaN interlayer
n-type annular contact

SiO2 insulating layer
InGaN/GaN QWs (QWs-1)

n-type GaN interlayer
InGaN/GaN MQWs (QWs-2)
bottom AlInN/GaN DBR
nid-GaN template
oxidized AlInN
current aperture
nid-GaN interlayer
c-plane sapphire or FS-GaN substrate
Fig. 4.2: Schematic 3D cross-section of an InGaN/GaN MQW polariton LD based

on an intracavity pumping geometry. The structure is sandwiched between bottom
(shown in dark blue/light blue) and top (shown in green/white) DBRs. The polaritons
are formed in the QWs-2 region (shown in orange/light blue). In the case of intracavity
pumping, the QWs-2 are pumped with the optical emission from the electrically pumped
QWs-1.
We have considered two types of pumping for the polariton laser illustrated in Fig.4.2:
Electrons and holes are injected into QWs-2 and QWs-1 are ignored. This pumping
scheme is further referred to as electrical pumping.
QWs-1 are electrically pumped and their optical emission pumps QWs-2. This
pumping scheme is further referred to as intracavity pumping.
The two schemes are discussed in more details in section 4.4.2
4.4
4.4.1
Modelling formalism
Semiclassical Boltzmann equations
Semiclassical Boltzmann equations have been used to describe the polariton laser kinetics. Here we briey outline how those equation can be derived from the quantum
mechanical density matrix formalism by a series of successive approximations.
71
2.6
2.4
2.2
2.0
1.8
Refractive index
Electrical field (arb. units)
2.8
1.6
1.4
0
Position ( m)
Fig. 4.3: Results of a transfer matrix simulation of the eld intensity prole of a
polariton LD shown in Fig.4.2 centred at = 415 nm along with the corresponding
refractive index prole.
Let | be the eld operator for all the quasiparticles in the system, specically polaritons, photons, phonons and electrons, which obeys the Schrodinger equation:
i~
| = H|,
t
(4.8)
is the full Hamiltonian of the system. The holes are not considered in the
where H
simulation since we assume that the active region is n doped and most of the free
carriers are electrons. Now if we introduce the density operator = ||, then from
the Schrodinger equation we can derive the so called Liouville equation for
i~

].
= [H,
t
(4.9)
To proceed further it is necessary to make approximations. We rst apply the Born

approximation, which neglects the correlations between the polariton, phonon, electron
and photon elds in the system. In this case we express the full density operator as a
direct product of density operators of the subsystems:
(t) = pol (t) phon el
(4.10)
Where phon , el , correspond to the density operators of the phonon, free electron and
photon subsystems, and pol (t) is the operator for the polariton subsystem. Moreover,
72
we assume that only the polaritons are described by a time-dependent density operator
and all the other operators correspond to thermal equilibrium. In order to eliminate
the phonon, electron and photon density operators we take the trace over all except
the polariton subsystem, which gives us the following equation for the polariton density
operator which will in what follows be denoted by .
= Lpol + Lphon + Lel + Llif etime + Lpump .
(4.11)
Here each L denotes a so called Liouvillian describing the interaction of the polaritonic
system with subsystem . Each Liouvillian contains only the annihilation (
a(k)) and
creation (
a (k)) operators of the polaritons. In order to obtain the semiclassical Boltzmann equation we should apply the Born approximation for the polariton system itself,
i.e. assume that there is no coherence between polariton states with dierent momenta,
and that the polariton density operator can be decomposed into a direct product:
(t) = 0 (t) k1 (t) . . . k (t) . . . .
(4.12)
If we now take the trace over the polariton states, to get the equations for the polariton
numbers in dierent states Nk =
a (k)
a(k), they contain fourth order correlators of the
form
a (k)
a(k)
a (k )
a(k ) and eighth order correlators of a similar form, like the fourth
order correlator C which was considered in subsection 3.4.3. However using the BornMarkov approximation we can decompose these operators as
a (k)
a(k)
a (k )
a(k ) =
a (k)
a(k)
a (k )
a(k ) = Nk Nk . As a result, the equations contain only the occupation
numbers of dierent states, resulting in a set of semiclassical Boltzmann equations for
the polariton occupation numbers Nk
dNk
Nk
Wkk Nk (1 + Nk ) +
Wk k Nk (1 + Nk ),
= Pk
dt
k
(4.13)
where for each state k, Pk denotes the pumping rate and k , and the W terms describe
the scattering rates between dierent polariton states. Each scattering rate contains
three contributions:
W = Wpolphon + Wpolel + Wpol ,
(4.14)
relating to the polariton-phonon, polariton-electron and polariton-polariton interactions.

The derivation of the scattering rates can be found in [96, 97] and the relevant formulae
are presented in the next subsection.
The values of wavevector allowed in (4.13) should be considered. To solve the system of
equations (4.13) numerically it is necessary to set up a grid of discrete values of wavevector, and the nite size of the device suggests a natural quantization step k = 2/L,
where L is the characteristic size of polariton excitation spot. The upper limit of the
allowed values of k should be chosen as kmax = 2/aB since excitons with wavevectors
73
larger than the inverse Bohr radius cannot be treated as bosons because of the uncertainty principle. Furthermore, we should account for the two-dimensional nature of the
polaritons with states described by k = [kx , ky ], so the quantization should be performed
for both wavevector components. Hence, the estimated number of wavevector values and
equations N represented by equation (4.13) is:
(
N=
4.4.2
L
aB
)2
.
(4.15)
Scattering rates in polariton relaxation processes
Pumping schemes
In the simulations two dierent pumping schemes have been considered. In the rst
scheme, it is assumed that electrically-pumped electrons and holes are uniformly injected
into the set of strongly-coupled QWs, emitting at 415 nm (i.e., the QW subset QWs1 described in section 4.3 is ignored). Such a scheme is not well suited to III-nitride
structures with a large NQW but it is adopted as a simplied approach, and one that
is potentially applicable to low temperature GaAs-based polariton LDs. After some
time, charge carriers are either removed due to nonradiative recombination occurring at
dislocations, trapping or Auger recombination, or they bind into pairs forming excitons.
Those excitons may be characterized by various energies and in-plane wavevectors (k)
and their ensemble is considered as an incoherent reservoir pumped from the electronhole plasma. This exciton reservoir then feeds the condensate of exciton-polaritons,
which is a coherent multiparticle state responsible for polariton lasing.
The second scheme concerns the design shown in Fig.4.2, where the region QWs-1, ideally
emitting in the 390-400 nm range, is uniformly electrically-pumped by charge carriers
and high energy photons subsequently emitted by those QWs are then absorbed by
QWs-2, [75] which is the set of strongly-coupled QWs, that will lead to the formation of
an incoherent exciton reservoir. This intracavity optical pumping geometry essentially
diers from the rst one by (i) the fact that overall, for a given current density, the
population of excitons in the reservoir feeding the condensate will be smaller due to
the internal quantum eciency (IQE), which is less than 100%. This latter quantity is
taken to be 90%, which which is typical of state of the art LEDs. (ii) The density of free
carriers in the region QWs-2 is constant since they are located in the ntype region (cf.
Fig. 4.2), whereas this density is current-dependent in the rst scheme and will have an
impact on the injection-dependence of the exciton-free carrier scattering term. (iii) While
one can assume that in the intracavity optical pumping scheme, free electrons in the ndoped region are thermalized and obey Fermi statistics with an eective temperature
(Tef f ) close to the lattice temperature (Tlatt ), in the direct electrical pumping schem,e
74
electrons are not thermalized with the lattice. In this latter situation, we consider a
Boltzmann carrier distribution with Tef f > Tlatt .
We now present the expressions for the rates of all the kinetic processes in the system
pointing out the dierences between the two pumping schemes.
Pumping of the polariton laser

In order to evaluate the pumping term in the case of electrical pumping we can write
down a simplied rate equation for the number of electron-hole pairs neh injected in the
microcavity:
dneh
J
neh
=
W neh ,
dt
e
eh
(4.16)
where J is the electrical pumping rate, e is the elementary charge, eh is the decay rate
of the electron-hole plasma, and W is the exciton formation rate from the electron-hole
plasma.
Eq. ((4.16)) can be solved analytically to yield the time dependence of neh (t):
[
(
)]
J
eh
t
neh (t) =
1 exp W t
,
e 1 + W eh
eh
(4.17)
assuming an electron-hole plasma density equal to zero at t = 0. We then assume that

the excitons that are formed from the electron-hole plasma, have an additional energy
compared to the ground state exciton energy Ex , which is equal to the exciton binding
energy and thus the excitons occupy only the relatively high k exciton states with energy
E(k) which are not coupled to the microcavity optical mode. Moreover, we assume that
those states are all pumped at an equal rate. Thus, we can write down the expression
for the pump term in a simple form:
Pk = 0,
Ek Ex < ,
(4.18)
and,
Pk =
W neh
,
e
N
Ek Ex > .
(4.19)
e is the number of states in our quantization lattice which have the energy E(k)
Here N
satisfying the condition (4.19).
For the intracavity optical pumping geometry we consider that the strongly-coupled
quantum wells are indirectly electrically pumped with an energy-dependent optical pump
75
intensity given by:

(
)
(Ek Epump )2
int J(Ek )
Pk =
exp
.
2(E)2
2(eE)
(4.20)
Here J is the electrical current in QWs-1, int is the internal quantum eciency of the
pumping LED, Epump is the central photon emission energy of the LED which is set
equal to the exciton energy of QWs-1, E is the linewidth of the LED, which is set to 90
meV and is considered temperature-independent as a rst approximation, and Ek is
the energy range associated with state k, which is the area of the grid cell of the state on
the energy grid. Note here that for the sake of illustration we consider the limiting case
where the pumping QW set (QWs-1) is resonant with the exciton energy of the second
QW set (QWs-2) that leads to the formation of polaritons, to make a clear distinction
between the performance of the two possible pumping geometries. However, in practice,
a nonresonant intracavity pumping scheme would probably apply.
Polariton lifetime decay

The polariton lifetime k is calculated in the same way for both pumping schemes and
can be obtained by relation:
(
k =
|Xk |
|Ck |
+
x
c (k)
)1
,
(4.21)
where x and c are the lifetimes of the exciton and cavity mode respectively and Xk
and Ck are the Hopeld coecients which dene the polariton components. While the
exciton lifetime is assumed wavevector-independent and has been chosen on the basis of
the experimental results provided by our collaborators, the cavity mode lifetime depends
on the in-plane wavevector and can be calculated exactly in the approximation of the
laterally unbound microcavity by the transfer matrix method, and is just the imaginary
part of the eigenfrequency of the structure. The Hopeld coecients can be calculated
using the coupled-oscillator approach in the following way. The Hamiltonian of the
system can be expressed in matrix form in the basis of the cavity and exciton modes:
(
=
H
Ex (k) V RS
V RS
)
(4.22)
Ec (k)
The eigenvalues of this matrix are the energies of the lower and upper polariton states:
Ex (k) + Ec (k) +
Eup (k) =
Elow (k) =
Ex (k) + Ec (k)
(Ex (k) Ec (k))2 + 42V RS

2
(Ex (k) Ec (k))2 + 42V RS
2
(4.23)
(4.24)
76
The Hopeld coecients for the lower polariton branch are the components of the normalized eigenvector corresponding to the lower polariton eigenvalue and are given by:
Xk =
1 + (Elow (k) Ec (k))/(V RS)
Ck =
]2 ,
(4.25)
(4.26)
1
1 + [V RS /(Elow (k) Ec (k))]2
Polariton-phonon scattering
The expressions for the scattering process involving acoustic phonons are equivalent for
both pumping schemes. Two opposite processes take place in the system: absorption
of an acoustic phonon by a polariton with scattering to a higher k state, and emission
of a phonon with relaxation of the polariton to a lower k state. The exciton-phonon
interaction matrix element has been calculated in [98] as:
excph
Wkk
2 (q)
=
~0 cs S 2
)
(
1 1 q2
(q qz ),
Nphon (q) +
2
qz
(4.27)
where the plus sign correponds to the emission of a phonon, and minus to the absorption
of a phonon. q is the magnitude of the phonon wavevector and can be obtained from
energy conservation and the phonon dispersion to give q = |Ep (k) Ep (k)|/cs , where
cs is the speed of sound in the material, and Ep (k) is the energy of a polariton with
wavevector k. Since the excitons are two-dimensional quasiparticles and the phonons
are three dimensional, only the in-plane component of the phonon wavevector q must
meet the momentum conservation

law: q = |k k |. The orthogonal component of the
phonon wavevector is qz = q 2 q2 . 0 is the mass density of the material and S is

the area of the cavity. Nphon is a Bose-Einstein distribution, since we assume that the
phonon subsystem remains in the thermal equilibrium. is a Heaviside step function
and the function can be written as:
= [
De Ze (qz )
Dh Zh (qz )
]3/2 + [
]3/2 ,
1 + mh q aB /(2mx )
1 + mh q aB /(2mx )
(4.28)
where De , Dh are the deformation potentials for electrons and holes, and me , mh , mx are
the electron, hole and exciton eective masses. Ze,h are the overlap integrals between
the phonon and electron (hole) wavefunctions:
z=+
dz2e,h (z)eiqz z ,
Ze,h (qz ) =
(4.29)
z=
where e,h are the normalized one-dimensional wavefunctions for electrons and holes.
In order to obtain the polariton-phonon scattering rate, we should multiply the exciton
77
phonon scattering rate by the Hopeld coecients for the initial and nal states:
polph
Wkk
= W excph |Xk ||Xk |.
(4.30)
The interaction with acoustic phonons is one of the main mechanisms of relaxation for
the excitons. However, it becomes inecient for polaritons because of the region of
very large group velocity, the so called bottleneck region, which is situated close to the
inection point point of the lower polariton band. The polariton-phonon relaxation
is suppressed in this region, because there is a lack of phonons with the appropriate
momentum and energy. Thus, when only the polariton-phonon relaxation exists, the
polaritons tend to accumulate in the bottleneck region. However, the accumulation
of the polaritons in this region is reduced by the other relaxation processes that occur,
such as polariton-polariton and polariton-free electron scattering, which will be discussed
later in this section. Polariton-phonon scattering also aects the depletion of the ground
state, since the polaritons absorb phonons and are scattered to the higher energy states
[99, 100].
Polariton-optical phonon scattering has not been considered in the modelling, because
optical phonons in the III-nitride semiconductors have an almost at dispersion throughout the Brilloin zone and have the energy of the order of 130 meV, which is larger then
the energy mismatch between the ground polariton state and the highest polariton state
considered in the simulation. Thus, optical phonons cannot provide the scattering from
one polariton state to another. However, optical phonon scattering has been accounted
for by the modication to the exciton lifetime, a value for which was provided by our
experimental collaborators.
Polariton-polariton scattering
The polariton-polariton scattering rate is the same for the both pumping schemes and
can be calculated in the following way. A derivation of the formula for the rate can be
found in [97] (and references therein) which gives:
Wkk =
2
|M ex |2 |Xk ||Xk ||Xq ||Xk+qk |nq (1 + nk+qk ) F (k, k , q),
~ q
(4.31)
where
M ex = 6
a2B B
E
S X
(4.32)
78
and the function F (k, k, q) corresponds to the energy conservation condition taking
into account the nite polariton lifetime:
F (k, k , q) =
(E(k) + E(q)
E(k )
~/k
.
E(|k + q k |))2 (~/k )2
(4.33)
A few computational considerations should be mentioned here. The rst one is the
fact that the polariton-polariton scattering rate contains the a priori unknown polariton occupation numbers. Thus, in order to calculate the polarion-polariton scattering
dynamics, the scattering rates should be recalculated at each time step, which makes
the computation extremely time-consuming. It is also worth mentioning that due to
the uniform wavevector quantization lattice it is always possible to nd a lattice node
corresponding to the state k + q k , which means that we can also full the momentum
quantization rule. However, we could consider another quantization scheme for which
the energies of the quantized states are uniformly distributed as in [96]. In this case,
we could automatically full the energy conservation condition , but we would also be
obliged to introduce an additional function describing the momentum mismatch. Unlike
the energy broadening, which occurs due to the nite polariton lifetime, the broadening
of the momentum is due to the nite area of the sample and is related to the inverse
length of the sample for all the states. It is clear that in a real experimental set up there
would be broadening both in energy and in momentum, which cannot be accounted for
in the simulation due to the computational diculty. The nal comment that should be
made is that in order to simplify the computation it is instructive to set the terms of the
sum in the Eq.(4.31) to zero in the case of large energy mismatch. If we do this, the nal
Jacobian, which is calculated during the numerical solution of the system of dierential
equations, becomes sparse and makes the computation far more time-ecient.
Polariton-free electron scattering

This mechanism has been shown to play a key role in polariton condensation in the case
of non-resonant pumping [101]. The expression for the polariton-free electron scattering
rate has been derived in [102]. Two process may occur during the exciton-electron
scattering: elastic scattering of the exciton from a free electron and the dissociation of
the exciton into an electron and hole. In our simulations we account explicitly only for
the elastic process. The inelastic dissociation process has been accounted for implicitly
by the corrections to the exciton non-radiative lifetime. The general expression for the
polariton-electron scattering rate is given by:
polel
Wkk
S
= |Xk ||X |
2~
k
d 2 ke [
|Vexc + Vdir |2 (ke ke )

]
d(Ee (ke )Ee (|ke +kk |))

dke
ke =ke
f (ke ) (1 f (|ke + k k |)).
(4.34)
79
Here ke is the electron wavevector which satises the energy conservation condition:
Epol (k) + Ee (ke ) = Epol (k ) + Ee (|ke + k k |).
(4.35)
The function f is the electron momentum distribution function, which diers signicantly for the two pumping schemes. In the case of intracavity pumping we assume
that the electrons are in thermal equilibrium with the system and are described by a
Fermi-Dirac distribution: f = [exp((Ee )/kT ) + 1]1 with temperature equal to the
lattice temperature and chemical potential :
(
[ 2 ]
)
~ ne
= kT log exp
1 ,
me kT
(4.36)
where ne is the concentration of the free electrons in the system. For the case of the
electrical pumping we also assume a Fermi Dirac distribution of the electrons, but in this
case the temperature of the electron gas is much higher than the lattice temperature
and does not depend on it, and the total concentration of the electrons depends on
the pumping current and is proportional to the total number of electrons obtained
from equation Eq.(4.17). Vexc and Vdir are the matrix elements of the exchange and
direct scattering mechanisms respectively. The matrix element of the direct Coulomb
interaction is given by:
Vdir (q) =
4e2 a3b
h(aB q) (g(ab q/2) g(ab q/2)) ,
0 SDqw
(4.37)
where q = |k k|, = me /mx , = mh /mx , and functions g, h have the forms:

g(aB q) = [1 + (aB q)2 ]3/2 ,
(4.38)
2 q2
DQW
.
2(aB q)3
(4.39)
h(aB q)
The exchange interaction term is:
8e2 a5B
Vexc (k, q) =
2
0 SDQW
d2 k1 h(aB k1 )g(|k1 + q k|)
(4.40)
(g(|k1 + q k|) g(|q k|)),

where k = ke k. This integral should be calculated numerically. Despite the
considerable diculty which arise in the calculation of the exchange term, it cannot be
omitted in the modelling since it is dominant in the polariton-free electron scattering.
4.5
80
Numerical results and discussion
In the numerical calculations we have used the parameters, presented in table 4.1 with
the right column showing the source in the literature for the value quoted. When no
source is shown the value was provided by our experimental collaborators.
Table 4.1: Parameters of the modelling of polariton laser
Parameter
Value
cavity photon lifetime c
1 [ps]
exciton lifetime x
1 [ns]
[61]
50 [m]
[61]
5 [ns]
[103]
exciton formation rate W
0.01 [ps1 ]
[103]
exciton binding energy Eb
45 [meV]
Rabi-splitting R
45 [meV]
lateral size of the system L

electron-hole pair lifetime eh
number of quantum wells NQW (QWs-2)

exciton inhomogeneous broadening ex
65
Source
device design
35 [meV]
eective electron mass me
0.2 free electron mass
[104]
eective hole mass mh
1.1 free electron mass
[104]
11.1 [eV]
[104]
deformation potential D
mass density 0
6150
speed of sound cs
[kg/m3 ]
7960 [m/s]
internal quantum eciency int
[105]
0.9
eective electron temperature

800 K
Values of the parameters used in the modelling. Right column shows the source in the
literature for the value quoted. When no source is shown the value was provided by our
experimental collaborators.
Figure 4.4 shows the evolution of the ground state polariton occupation number vs pump
current density for dierent temperatures and detunings for the two pumping geometries.
In a realistic set up, changing the cavity mode frequency would inevitably change other
parameters such as Rabi-splitting for example. However, in the modelling we assume
the detuning to be decoupled from the other parameters of the system, so that it can
be changed independently. The threshold current density for condensation (Jthr ) can be
clearly identied in each case. The detuning is dened as dierence between the ground
exciton and cavity mode energy, which is in its turn dened by the cavity length and
the refractive index.
We have also calculated the Jthr dependence as a function of temperature and detuning
which is shown in Fig. 4.5. In this way one obtains the polariton condensation phase
diagram under electrical pumping, which is analogous to that derived under optical
pumping for GaN/AlGaN MQW microcavities [79, 100]. One can see that at room
Fig. 4.4: (a)-(c) Occupancy of the polariton ground state vs pump current density
calculated for the two pumping geometries at various temperatures and detunings (see
text for details).
Fig. 4.5: Plots of Jthr vs detuning and temperature for (a) the direct electrical and
(b) the intracavity optical pumping schemes. The red dashed line on each plot corresponds to the evolution of Jthr,min as a function of temperature. (c) Evolution of
the optimum detuning (black solid and black dashed lines) and condensation threshold
current density (red solid and red dashed lines) at the optimum detuning as a function
of lattice temperature for the direct electrical (solid lines) and the intracavity (dashed
lines) pumping schemes.
81
82
temperature the lowest threshold current density (Jthr,min ) is obtained for a negative
detuning of 19 meV and amounts to 5 Acm2 for the direct electrical pumping
geometry while a Jthr,min value of 6 Acm2 at a detuning of 32 meV is derived for the
intracavity pumping geometry. The values are in good agreement with those predicted in
previous work [75]. The optimum detuning (opt ) corresponds to the case where the mean
polariton relaxation time equals the mean polariton lifetime [100, 106]. The system then
switches from the kinetic to the thermodynamic regime. In other words, it undergoes a
crossover from a regime where Jthr decreases with increasing detuning (decreasing of
absolute value), because of the enhancement of the total scattering rate to the ground
state, to a regime where Jthr increases concomitantly with due to the combined eects
of the increasing polariton eective mass (which leads to a larger value of the critical
density for polariton condensation, n2D,crit ) and thermal detrapping from the ground
state [100]. It is also noticeable that the temperature dependence of opt signicantly
diers from one geometry to the other (Figs. 4.5(a) to 4.5(c)). This is attributed to
the dierence in the eciency of the free electron scattering mechanism as a function of
temperature [101]. For the direct electrical pumping geometry, the electron distribution
does not depend on Tlatt , since electrons are not thermalized, thus leading only to slight
changes in the free electron scattering rate with temperature and thereby explaining the
weak opt (T ) variation displayed in Figs. 4.5(a) and 4.5(c). While for the intracavity
optical pumping geometry, electrons are thermalized to Tlatt , which leads to a behaviour
closer to that reported for GaN/AlGaN MQW microcavities under nonresonant optical
pumping (Figs. 4.5(b) and 4.5(c)) [100]. However, it should be recognized that assuming
an electron temperature equal to Tlatt is a crude approximation. We should also point
out that, for the sake of simplicity, we did not account for the large activation energy
of the Mg acceptor in GaN compounds, which would probably degrade the electrical
characteristics and thus lead to an increase in Jthr at low temperatures.
The low threshold values reported for the intracavity pumping geometry can probably
be explained by the broad spectral distribution of the pump Pk . In the direct electrical injection geometry, excitons characterized by high energies and large in-plane wave
vectors are created from the electron-hole plasma, which requires a comparatively long
time to relax to the k = 0 state. However, in the case of intracavity optical pumping a
broad distribution of excitons centered on the QW-1 exciton energy, but also covering
lower energy states, is created. Those excitons which occupy lower energy and lower k
states compared with the direct electrical pumping geometry quickly relax to the lower
polariton branch ground state and enhance polariton relaxation, which results in the
lower threshold for this geometry. However, if the intracavity emission line is strongly
blueshifted from the polariton modes, the ratio between the polariton lasing thresholds
for the two pumping congurations is expected to be modied. One would then expect a
higher threshold for the intracavity pumping geometry owing to the IQE of the internal
pump, which is less than 100%.
4.5.1
83
Simplied rate equation modeling: steady-state solutions
In order to obtain a qualitative understanding of the functionality of polariton laser

diodes we compare the modelling results obtained with the full set of semi-classical
Boltzmann equations with a simplied quasi-analytical model. For the direct electrical
pumping geometry, the model is based on three rate equations describing the electronhole plasma (namely Eq. (4) introduced in section III.A), the exciton reservoir and
ground state polaritons, respectively:
dnx
nx
=
+ W neh anx (np + 1) + aeesc np nx bn2x (np + 1) cneh nx (np + 1) ,
dt
x
(4.41)
np
dnp
= + anx (np + 1) aeesc np nx + bn2x (np + 1) + cneh nx (np + 1) ,
dt
p
(4.42)
where nx , and np are the concentrations of excitons and exciton-polaritons, respectively.

p is the lifetime of exciton-polaritons in the ground state, a accounts for the acoustic
phonon relaxation rates, = 1/kB T , and esc is the characteristic energy splitting
between the bottom of the LPB and states beyond the inection point of the LPB
where zero in-plane wavevector polaritons are scattered, which is a quantity sensitive to
the detuning [79, 99, 100]. b is the exciton-exciton scattering rate and c is the rate of
exciton relaxation mediated by free carriers. The solid line curves in Fig. 4.6 show the
detuning dependence of the tting parameters a, b, and c (for dierent temperatures)
for them to yield the same threshold pumping current density for the direct electrical
pumping scheme as the full semi-classical Boltzmann model predicts.
If now the more realistic intracavity optical pumping geometry is considered, the previous
quasi-analytical three-level model can be simplied further to a two-level model since
the free carrier dynamics does not have to be accounted for explicitly. Then:
dnx
nx
= Px
anx (np + 1) + aeesc np nx bn2x (np + 1) cnd nx (np + 1) , (4.43)
dt
x
np
dnp
= + anx (np + 1) aeesc np nx + bn2x (np + 1) + cnd nx (np + 1) ,
dt
p
(4.44)
where nd is the concentration of free-carriers obtained from the doping level in the
region QWs-2, and taken equal to 2 1012 cm2 . In the simplest approximation Px
-1
Scattering rate (ps )
84
-9
-10
10
-17
10
-10
10
-11
10
10
0K
= 2
-18
10
K
20
= 1
-19
10
-11
10
-12
10
-12
10
K
20
= 2
-20
10
0K
electrical
-13
10
optical
T = 32
-21
10
-13
10
-80
-60
-40
-20
-80
-60
-40
-20
-80
-60
-40
-20
Detuning (meV)
Fig. 4.6: Averaged scattering rates a, b, and c as a function of detuning obtained by

tting the full semi-classical Boltzmann system of equations at various temperatures
for the two pumping schemes: electrical (solid lines) and intracavity (dashed lines)
pumping.
can be taken to be Px = int J/q, where int is the internal quantum eciency of the
electrically-pumped region QWs-1. The dashed lines in Fig. 4.6 show the values of the
tting parameters a, b, and c which yield the same threshold pumping current density
for the intracavity optical pumping scheme as the full semi-classical Boltzmann model
for dierent temperatures and detunings.
One can see in this latter gure that - overall - all the scattering mechanisms that
contribute to populate the polariton lasing mode become more ecient with decreasing
negative detuning (in absolute value). It could be explained by the reduction of the
exciton fraction of the polariton states at large negative detuning, so that all interactions
involving polaritons become weaker than at zero or positive detuning. In some cases,
the detuning dependence of the scattering rates is non-monotonic and it is also sensitive
to the pumping scheme. This is due to the complexity of the lower polariton branch
dispersion. The average scattering rates are therefore sensitive to both the shape of the
polariton dispersion and the excitation spectrum prole.
Above threshold the (1 + np ) and (1 + nx ) terms appearing in the rate equations can be
approximated as np and nx , respectively. After some algebra the steady-state solutions
(t = +) for the electron-hole pair, exciton and polariton populations for the direct
electrical pumping geometry are obtained as:
neh =
Jeh
,
q(1 + eh W )
(4.45)
nx =
cneh + a(eesc 1) +
(cneh + a(eesc 1))2 +
85
4b
p
2b
np = p (W neh
nx
).
x
(4.46)
(4.47)
Note that for the intracavity optical pumping geometry, slight changes occur since neh
in Eq. (4.46) has to be replaced by nd , and W neh in Eq. (4.47) has to be replaced
by Px . We should emphasize that for this latter geometry the carrier population, which
acts as a reservoir for the stimulated relaxation process (here the excitons), gets clamped
once it crosses the condensation threshold, which is expected owing to the similarities
of the above-mentioned rate equations with those describing conventional laser diodes
[107]. A similar treatment can be used for the direct electrical pumping geometry since
4b/p (cneh + a(eesc 1))2 . Thus, for both cases, we obtain nx 1/ bp .

The evolution of nx as a function of temperature at the optimum detuning is shown
in Fig.4.7 using both the exact expression and the approximate one for the two pumping geometries. The validity of the approximation for nx is conrmed by the close
correspondence between the two quantities whatever the temperature. Considering its
approximate expression, the increase with temperature of nx can be directly inferred
from the results displayed in Fig. 4.6, which show a decrease in the exciton-exciton scattering rate b with increasing Tlatt , and is also fully consistent with the overall temperature
dependence of the analytical expression for b derived by Tassone and Yamamoto [96].
Note that one also expects a decrease in nx with increasing detuning probably due to
the concomitant increase in the relaxation process from the excitonic reservoir (which
coincides with the decrease or even the disappearance of the relaxation bottleneck) and
that of p [96]. Because low Jthr values in the condensation phase diagram are essentially
reported for negative values (cf. Figs. 4.5(a) and 4.5(b)), one should keep in mind
that the present situation corresponds to a decrease of in absolute value.
4.5.2
Simplied rate equation modelling: high-speed current modulation treatment
In this section, the dynamical response of polariton laser diodes (LDs) to a small perturbation, such as a modulation of the current above threshold, is investigated. Since exact
analytical solutions to the full rate equations cannot be obtained, a dierential analysis of
the simplied rate equations given in the previous section using the approach described
by Coldren and Corzine for the case of conventional diodes is considered [107]. The resulting small-signal responses are derived by taking the dierential of the rate equations
which can be written in compact matrix form for the two pumping geometries as follows:
86
(x10 )
2.5
2.0
1E-4
electrical
1.5
nx
1.0
1E-5
optical
0.5
0.0
50
100
150
200
250
Relative deviation
1E-3
3.0
300
Temperature (K)
Fig. 4.7: Left-hand side vertical scale: evolution of nx as a function of temperature

calculated at the optimum detuning using the exact expressions for the electrical (connected black dots) and the intracavity (connected black circles) pumping geometries
(see text for details). Right-hand side vertical scale: relative deviation between the
exact and the approximated expressions for the electrical (connected red dots) and the
intracavity (connected red circles) pumping geometries.
Electrical pumping geometry
d
dt
dnx
dnp
xx xp
px
pp
dnx
dnp
[
+
dneh (W cnx np )
0
]
(4.48)
where
1
+ anp + 2bnx np + cneh np anp eesc ,
x
1
pp =
anx bn2x cneh nx + anx eesc ,
p
xx =
(4.49)
(4.50)
xp = anx + bn2x + cneh nx anx eesc ,
(4.51)
px = anp + 2bnx np + cneh np anp eesc .
(4.52)
Intracavity pumping geometry
d
dt
where
dnx
dnp
xx xp
px
pp
dnx
dnp
[
+
int
q dJ
]
(4.53)
1
+ anp + 2bnx np + cnd np anp eesc ,
x
1
pp =
anx bn2x cnd nx + anx eesc ,
p
xx =
87
(4.54)
(4.55)
xp = anx + bn2x + cnd nx anx eesc ,
(4.56)
px = anp + 2bnx np + cnd np anp eesc .
(4.57)
To obtain the small-signal response of the exciton (dnx ) and the polariton (dnp ) populations to a sinusoidal current modulation dJ, we assume solutions of the form dJ =
J1 exp(it), dneh = neh1 exp(it), dnx = nx1 exp(it), and dnp = np1 exp(it). The
linear systems ((4.48)) and ((4.53)) can then be solved for the small-signal polariton
population by simply applying Cramers rule. The small-signal solutions after expansion of the determinants can be written as:
Electrical pumping geometry
np1 () =
px [W cnx np ] + (i + xx )cnx np
J1 /q
= np1 (0)H(),
2
(px /p + ixx )
i + W + 1/eh
(4.58)
where H() is the modulation transfer function given by:
H() =
px /p (W + 1/eh )
px [W cnx np ] + (i + xx )cnx np
.
px [W cnx np ] + xx cnx np (px /p 2 + ixx )(i + 1/eh + W )
(4.59)
Intracavity pumping geometry
np1 () =
px int J1 /q
= np1 (0)H()
(px /p 2 + ixx )
where H() is the modulation transfer function given by:
(4.60)
H() =
88
px /p
.
(px /p 2 + ixx )
(4.61)
10
10
np
4 1
10
10
2
-1
10
10
1
10
-2
10
= -35 meV
T = 320 K
-1
10
= -22 meV
)|
10
(b)
|H(
(a)
3
10
10
10
2
J (A/cm )
-1
10
10
10
10
Frequency (GHz)
Fig. 4.8: (a) Polariton condensate occupation number vs current density for the electrical (red line) and intracavity (black line) pumping geometries determined at 320 K
and at the optimum detuning. (b) Frequency dependence ( = /2) of the square
modulus of the modulation transfer function, |H()|2 . Each curve corresponds to one
of the steady state solutions indicated in Fig. 4.8(a).
For both pumping geometries a relaxation resonance frequency equal to R,polLD =
px /p can be dened while the xx term can be readily identied with a damping
factor.
The expression for px derived for the two geometries (Eqs. (4.52) and (4.57)) can
be greatly simplied since the term describing the exciton-exciton interaction ( b)
dominates over the phonon-exciton and the free carrier-exciton scattering terms ( a
and c, respectively) whatever the exciton-photon detuning and the temperature (cf. Fig.
4.7). Consequently, the square of the resonance frequency reduces to:
2
R,polLD
2bnx np
2np
p
b
.
p3
(4.62)
Therefore, within this theoretical framework, the resonance frequency for polariton LDs
is directly proportional to the square root of the polariton population in the condensate
and inversely proportional to the square root of the polariton lifetime (keeping in mind
that the exciton population of the reservoir nx is clamped above threshold). In this
respect, such a dependence is similar to the dependence of R in conventional LDs above
threshold, since in this latter case:
2
R
vg adif f Np
,
cav
89
(4.63)
where vg is the group velocity, adif f is the dierential gain, Np is the average photon
density in the cavity, and cav is the cavity photon lifetime already dened in section
III.A [107].
From Eqs. (4.49) and (4.54), the damping factor can be rewritten as:
xx =
1
1
2
+ px =
+ R
p .
x
x
(4.64)
It is thus seen that for large resonance frequencies, the damping of the response is ruled
by the polariton lifetime. On the other hand, the inverse of the exciton lifetime acts as
a damping factor oset, which is important for small polariton condensate populations
where the resonance frequency is small.
At this stage, we should point out that the validity of the previous treatment for the
intracavity pumping scheme might be limited by the actual response of the pumping
LED. Therefore it is necessary to determine the LED cuto frequency 3dB,LED , i.e.,
the frequency at which the electrical power response drops to half its dc value, and
compare it to R,polLD .
Following the theoretical approach described in previous sections, the rate equation
governing the emission of the LED is given by:
dnx
nx
=
+ W neh ,
dt
x
(4.65)
from which one can deduce the modulation transfer function HLED () using harmonic
analysis:
HLED () =
1/x (1/eh + W )
.
(i + 1/x ) (i + 1/eh + W )
(4.66)
The corresponding relaxation resonance frequency R,LED and the damping factor LED
can be written as:
R,LED =
1
x
1
+W
eh
20.1 ns1 ,
(4.67)
90
which corresponds to 3.2 GHz, and
LED =
1
1
+
+ W = 70.4 ns1 ,
x eh
(4.68)
which corresponds to 11.2 GHz, respectively. Additional features of the LED response
can be derived from the frequency dependence of the square modulus of HLED (). On
a general basis, such a response is characterized by a second-order, low-pass-band lter
behaviour with a damped resonance appearing near the cuto frequency LED,3dB . For
|HLED ()|2 , the expression for the peak frequency of the resonance P,LED is given by:
[
2
2
P,LED
= R,LED
1
1
2
LED
R,LED
)2 ]
,
(4.69)
and that of the cut-o frequency LED,3dB by:
2
2
3dB,LED
= P,LED
+
4
4
P,LED
+ R,LED
.
(4.70)
In the present case, since LED > R,LED , the two previous equations reduce to:
P,LED = 0,
(4.71)
3dB,LED = R,LED 20.1 ns1 .
(4.72)
and
To evaluate the impact of the LED response on the intracavity pumping geometry, one
should rst consider the evolution of |H()|2 at room temperature and at the optimum
detuning opt for dierent values of the input current J using Eqs. (4.59) and (4.61)
(Figs. 4.8(a) and 4.8(b)). Hence it is seen that for the electrical pumping case (Eq.
(4.59)) with current densities in the range 10 to 20 Acm2 , the peak frequency P lies
in the range 612 GHz and the cuto frequency 3dB is expected to be 19 GHz (Fig.
4.8(b)). Much larger values are predicted for the intracavity pumping geometry when
using Eq. (4.61) (3dB 400 GHz). However, in this regime the modulation transfer
function of the device is limited by the frequency response of the pumping LED, which
has a cuto frequency given by Eq. (4.69). Finally, we have to point out that the abovementioned values are upper bounds since they do not account for example, for possible
electrical parasitic eects that could potentially aect the transmission line impedance.
4.6
91
Conclusions
In summary, we have theoretically investigated some relevant electrical features of realistic III-nitride polariton diodes. First a formalism relying on coupled semi-classical
Boltzmann equations adapted for an electrically-driven exciton-polariton device was
used to derive the evolution of the occupation number of the polariton ground state vs
pump current density for two pumping geometries: namely the direct electrical and the
intracavity optical pumping schemes. The corresponding condensation phase diagrams
under electrical injection, i.e., plots of Jthr vs detuning and temperature, were also extracted. It led to the determination of the minimum threshold current density Jthr,min
as a function of lattice temperature for the two pumping schemes. Jthr,min values of
5 Acm2 and 6 Acm2 at room temperature has been derived for the direct electrical
and the intracavity optical pumping geometries, respectively, which are close to previous estimates. Then a simplied rate equation modeling treatment was introduced to
derive both steady-state and high-speed current modulation solutions. This simplied
analysis made it possible to show that the carrier population which, acts as a reservoir
for the stimulated relaxation process, namely that of the excitons, gets clamped once
it crosses the condensation threshold, which is a direct consequence of the similarities
of the simplied rate equations with those describing conventional laser diodes. The
analysis of the modulation transfer function, derived from the dynamical response of
polariton LDs to a small modulation of the current above threshold, demonstrates the
interesting potential of the direct electrical pumping scheme, since a cuto frequency
3dB up to 19 GHz is predicted, whereas for the intracavity optical pumping scheme,
the cuto frequency is shown to be limited by the frequency response of the pumping
LED, for which 3dB 3.2 GHz.
Chapter 5
Summary, conclusions and

further work
The book is concerned with three general areas of study: (i) the negative refraction of
certain dielectric and metallic photonic crystals, (ii) the optical properties of periodic
dielectric structures with resonant inclusions such as semiconductor quantum wells, (iii)
the kinetics of two types of microcavity-based polariton laser.
Negative refraction by photonic crystals is considered in Chapter 2. Negative refraction
has been attracting considerable interest in recent years due to the possibility of using the
eect to realize optical components such as superlenses and invisibility cloaks. Section
2.2 is concerned with negative refraction at the side-edge of a one-dimensional dielectric
photonic crystal. It has been shown that the eect of negative refraction in this case
is equivalent to 1 order diraction and is very dependent on the angle of incidence
and thus cannot be used to produce a practical superlens. However, a new eect spatial oscillations of the Poynting vector of the refracted radiation - has been predicted
analytically and observed in nite-dierence time-domain numerical simulations.
Section 2.3 is dedicated to the modelling of a new type of spectral lter for terahertz
radiation using a metallic photonic crystal prism. The ltering is based on the negative
refraction of electromagnetic waves with frequencies lying in the second photonic band
of the metallic photonic crystal. Thus, compared to a conventional prism, there is a
large angular dispersion between the frequency components lying in the second photonic
band and other frequencies. The optimal design of the prism was obtained using nitedierence time-domain simulations. Subsequently colleagues fabricated a prism, based
on the proposed design, and their experimental measurements have shown that the
proposed device can indeed act as an ecient spectral lter for the terahertz frequency
range.
92
Chapter 5. Summary, conclusions and further work
93
Chapter 3 reported studies of the optical properties of the dielectric photonic structures with resonant inclusions such as optical cavities adjacent to thin metal lms or
quantum wells the can support excitons. Section 3.2 describes theoretical calculations
of the eigenmode structure of a metal layer sandwiched between two Bragg reectors.
It is shown that Tamm plasmon polariton states exist on both sides of the metal layer
and are coupled via the evanescent waves inside the metal lm. The coupling of the
Tamm plasmon modes leads to eigenfrequency splitting and thus to distinctive features
in the band structure of the system. Analytical formulae for the system eigefrequencies have been derived and are shown to agree well with numerical modelling and with
measurements on the structures studied by collaborators.
Section 3.3 is concerned with the eigenmodes of a periodic array of quantum wells
embedded in the layers of a dielectric Bragg mirror. Analytical expressions for the mode
eigenfrequencies have been derived and used to explain the origin of the triplet structure
seen in the experimental reection spectra. Section 3.4 describes the modelling of the
parametric amplication which has also been experimentally observed in the system.
Modelling results, which have been obtained using a three-level formalism, are shown to
agree well with the experimental results.
Chapter 4 concerns modelling of the kinetics of GaN-based polariton lasers. The modelling was performed using a semiclassical Boltzmann equation approach. It is shown
that a threshold pump of the order of 50 Acm2 can be reached at room temperature
if the optimal detuning between the exciton and cavity mode is chosen. Studies of the
small signal response of the polariton laser are also presented. It has been shown that
the polariton laser operates as a low-pass-band lter with a cut-o frequency about 3
GHz. The relatively low threshold pump powers and high cut-o frequencies are very
promising for the realization of new types of optoelectronic devices based on polariton
lasers.
We now comment on the impact of the work presented in the book and possible future
work on the structures considered. We believe that the work on negative refraction
at the side-edge of one-dimensional photonic crystals has been taken as far as it is
reasonable to do so, particularly since it has been shown recently that layered metaldielectric nanostructures can exhibit all-angle negative refraction in a wide frequency
range [108], making them a far more promising basis for the realization of superlenses
and invisibility cloaks. However, the work on the terahertz spectral lter based on
a metallic photonic crystal prism, while lacking opportunities for the further original
theoretical investigations could play a signicant role in modern terahertz technology,
since eective spectral band pass lters for the terahertz range are not commonplace,
and the device proposed has the additional advantage that it could be deposited on a
chip.
94
The impact of the work on Tamm states in metal-organic cavities is illustrated by the
fact, that the subsequent work by our experimental collaborators dedicated to the same
topic was published in Nature Photonics [109]. The authors produced a lateral patterning of the metal-organic microcavity, introducing a set of metallic stripes rather than a
uniform metal lm inside the cavity. Moreover, they have shown that the introduction
of the metal inside the cavity does not signicantly degrade the coherence properties of
the optical cavity mode. Such structures could be used to organize exciton-polariton
condensates in a periodic potential and to do so at room-temperature due to the large
binding energy of the excitons in the organic systems.
We believe, the work on exciton-polaritons in photonic crystals with embedded quantum wells is important because it showed both theoretically and experimentally that it
is possible to realize strong coupling between the photonic crystal modes and the excitons. The resulting exciton-polaritons exhibit strong nonlinear dynamics, such as the
parametric amplication that is typical for the exciton-polaritons in conventional microcavities. However, photonic crystal modes provide larger exibility of the dispersion
properties of the exciton-polaritons. In particular, it has been predicted that for certain
structures, exciton-polaritons with negative group velocity and eective mass could be
observed. Such polaritons would have peculiar nonlinear dynamics since their ground
state, dened by the zero in-plane wavevector, would at the same time have the largest
energy. This property should also lead to an unusual characteristics of polariton-phonon
and polariton-electron scattering.
With regard to further work, the theoretical formalism, such as the transfer matrix
method and other theory used to describe the optical properties of periodic dielectric
nanostructures presented in chapter 3, can also be applied to the case of the periodic
layered structures containing metal layers. Consider a periodic layered metal-dielectric
nanostructure comprising alternating metal and dielectric layers with dielectric permittivitties m , d and thicknesses dm , dd , respectively. If the period of the structure is
much less than the electromagnetic wavelength then the eective medium [110] approximation can be used and the structure treated as a uniaxial, anisotropic medium with
components of dielectric permittivity tensor given by:
m dm + d dd
=
;
dd + dm
(
=
dd dm
+
d
m
)1
,
(5.1)
where the symbols and correspond to the in-plane and axial components of the dielectric permittivity tensor respectively. It follows that the dispersion relation between
2 / . It is
frequency and the wavevector components k , k is 2 /c2 = k2 / + k
clear that if the dielectric permittivity tensor components have dierent signs as occurs
in this case due to the negative permittivity of the metal layers, the isofrequency contours obtained from the dispersion relations will have a hyperbolic shape, in contrast
to the elliptical shape of conventional anisotropic media. The characteristic shape of
95
the isofrequency contours of such hyperbolic media leads to the possibility of realizing a
number of useful eects, including all-angle negative refraction [111], and superlensing
[112]. However, one of the most intriguing properties of the hyperbolic media is the
divergent density of states caused by the unbound isofrequency contours [113]. The
divergent density of states directly leads to a great enhancement in the spontaneous
emission rate (so called Purcell factor ) for point-like emitters placed inside such media.
The decease of the spontaneous emission lifetime of sources placed inside hyperbolic
media has been extensively studied both theoretically [114] and experimentally [115] in
recent years. However, I propose to study the Forster resonant energy transfer (FRET)
between two point emitters placed inside hyperbolic media. FRET is a non-radiative
resonant process of energy transfer between two two-level systems, which can be considered to be carried out by virtual photons [116]. The eciency of the FRET process
can be calculated using second-order perturbation theory and is inversely proportional
to the sixth power of the distance between the two systems, which makes it a powerful
tool for precise measurement of the distance between the chromophores in the biological
systems, for example. The picture is expected to change dramatically when the emitters
are placed inside a hyperbolic medium due to the huge density of the photonic states
[117]. Furthermore, I am also planning to study the optical properties of hyperbolic
media when the components of the dielectric permittivity tensors are characterized by a
nonlinear (Kerr) dielectric permittivity. In particular, it is planned to study the modication to the spontaneous emission lifetime of point sources placed inside such nonlinear
hyperbolic media.
Finally, the formalism for modelling GaN-based polariton laser, described in the chapter 4 together with the associated software developed constitute a powerful tool for
modelling the dynamics of polariton lasers in general. To enhance its performance, the
software should be merged with a nite element simulation code, which would calculate
the dispersion of the eigenfrequencies of device with realistic 3D geometries. Furthermore, this work could be extended to the case when a strong magnetic eld of about
1 T is applied to the microcavity. A series of recent experiments have shown that the
polariton lasing threshold can be signicantly decreased by the application of magnetic
eld. The modelling could be undertaken in the similar way to that presented in chapter
4. The magnetic eld would aect the dispersion of the low polaritonic branch (there
would actually be 4 separate polaritonic branches in this case) and the density of polaritonic states. Therefore the additional mechanism of spin-ipping resulting in the
scattering of the polaritons between dierent branches would have to be considered in
the modelling. Nevertheless, these eects can be accounted for by a relatively small
changes to the modelling software.
References
[1] W.L. Bragg. The diraction of short electromagnetic waves by a crystal. Proceedings of the Cambridge Philosophical Society, 17:4357, 1913.
[2] J. W. S. Rayleigh. On the remarkable phenomenon of crystalline reection described by Prof. Stokes. Philosophical Magazine, 26:256265, 1888.
[3] E. Yablonovitch. Inhibited spontaneous emission in solid-state physics and electronics. Phys. Rev. Lett., 58(20):20592062, 1987.
[4] S. John. Strong localization of photons in certain disordered dielectric superlattices. Phys. Rev. Lett., 58(23):24862489, 1987.
[5] J. D. Joannopoulos, S. G. Johnson, J. N. Winn, and R. D. Meade. Photonic
crystals. Molding the flow of light. Princeton University Press, 2008.
[6] G. Wannier. The structure of electronic excitation levels in insulating crystals.
Phys. Rev., 52:191, 1937.
[7] E.L. Ivchenko, A.I. Nesvizhskii, and S. Jorda. Bragg reection of light from
quantum-well structures. Phys. Solid State, 36:11561161, 1994.
[8] V.M. Agranovich and O.A. Dubovskii. Eect of retarded interaction on the exciton
spectrum in one-dimensional and two-dimensional crystals. JETP Letters, 3:233,
1966.
[9] A. Imamoglu, R. J. Ram, S. Pau, and Y. Yamamoto. Nonequilibrium condensates
and lasers without inversion: Exciton-polariton lasers. Phys. Rev. A, 53:4250,
1996.
[10] S. Kundermann, M. Saba, C. Ciuti, T. Guillet, U. Oesterle, J.L. Staehli, and
B. Deveaud. Coherent control of polariton parametric scattering in semiconductor
microcavities. Phys. Rev. Lett., 91(10):107402, 2003.
[11] T. C. Liew, A. V. Kavokin, T. Ostatnicky, M. A. Kaliteevski, I. A. Shelykh, and
R. A. Abram. Exciton-polariton integrated circuits. Phys. Rev. B, 82(3):033302,
2010.
96
Bibliography
97
[12] F. P. Laussy, A. V. Kavokin, and I. A. Shelykh. Exciton-polariton mediated

superconductivity. Phys. Rev. Lett., 104:106402, 2010.
[13] I.-H. Chen, Y. Y. Lin, Y.-C. Lai, E. S. Sedov, A. P. Alodjants, S. M. Arakelian,
and R.-K. Lee. Solitons in cavity-qed arrays containing interacting qubits. Phys.
Rev. A, 86:023829, 2012.
[14] I. V. Iorsh, M. A. Kaliteevski, S. Brand, R. A. Abram, and N. A. Kaliteevskaya.
An electromagnetic wiggler originating from refraction of waves at the side edge
of a Bragg reector. Journal of Modern Optics, 58:686693, 2011. doi: 10.1080/
09500340.2011.565887.
[15] G. P. Swift, A. J. Gallant, N. Kaliteevskaya, M. A. Kaliteevski, S. Brand, D. Dai,
A. J. Baragwanath, I. Iorsh, R. A. Abram, and J. M. Chamberlain. Negative
refraction and the spectral ltering of terahertz radiation by a photonic crystal
prism. Optics Letters, 36:16411643, 2011.
[16] V. G. Veselago. Electrodynamics of media with simultaneously negative electric
permittivity and magnetic permeability. UFN, 92(3):517526, 1967.
[17] P. M. Valanju, R. M. Walser, and A. P. Valanju. Wave refraction in negativeindex media: Always positive and very inhomogeneous. Phys. Rev. Lett., 88:
187401, 2002.
[18] J. Pendry. Negative refraction. Contemporary Physics, 45(3):191, 2004.
[19] M.I. Stockman. Criterion for negative refraction with low optical losses from a
fundamental principle of causality. Phys. Rev. Lett., 98:177404, 2007.
[20] V. Shklover C. Hafner V. Veselago, L. Braginsky. Negative refraction index materials. Journal of Computational and Theoretical Nanoscience, 3:130, 2006.
[21] D. R. Smith and N. Kroll. Negative refractive index in left-handed materials. Phys.
Rev. Lett., 85:29332936, 2000.
[22] L.D. Landau and E.M. Lifshitz. Electrodynamics of Continuous Media. Pergamon,
New York, 1974.
[23] R.A. Shelby, D.R. Smith, and S. Schultz. Experimental verication of a negative
index of refraction. Science, 292(5514):7779, 2001.
[24] D.R. Smith, S. Schultz, P. Markos, and C.M. Soukoulis. Determination of eective
permittivity and permeability of metamaterials from reection and transmission
coecients. Phys. Rev. B, 65(19), 2002.
[25] J. B. Pendry, A. J. Holden, W. J. Stewart, and I. Youngs. Extremely Low Frequency Plasmons in Metallic Mesostructures. Phys. Rev. Lett., 76(25):47734776,
1996.
Bibliography
98
[26] D.F. Sievenpiper, M.E. Sickmiller, and E. Yablonovitch. 3D wire mesh photonic
crystals. Phys. Rev. Lett, 76(14):24802483, 1996.
[27] P.V. Parimi, W.T. Lu, P. Vodo, J. Sokolo, J.S. Derov, and S. Sridhar. Negative
refraction and left-handed electromagnetism in microwave photonic crystals. Phys.
Rev. Lett, 92(12):127401, 2004.
[28] P. Vodo, W. T. Lu, Y. Huang, and S. Sridhar. Negative refraction and planoconcave lens focusing in one-dimensional photonic crystals. Appl. Phys. Lett., 89
(8):084104, 2006.
[29] D.O.S. Melville, R.J. Blaikie, and C.R. Wolf. Submicron imaging with a planar
silver lens. Appl. Phys. Lett., 84(22):44034405, 2004.
[30] J. B. Pendry. Negative refraction makes a perfect lens. Phys. Rev. Lett., 85(18):
39663969, 2000.
[31] W. Cai, U.K. Chettiar, A.V. Kildishev, and V.M. Shalaev. Designs for optical
cloaking with high-order transformations. Opt. Express, 16(8):54445452, 2008.
[32] M. Kaliteevski, S. Brand, R. A. Abram, A. J. Gallant, and J. M. Chamberlain.
Negative refraction can make non-diracting beams. Opt. Express, 16(19):14582
14587, 2008.
[33] C. Luo, S.G. Johnson, J.D. Joannopoulos, and J.B. Pendry. All-angle negative
refraction without negative eective index. Phys. Rev. B, 65(20):201104, 2002.
[34] J. E. Lugo, B. de la Mora, R. Doti, R. Nava, J. Tague na, A. del Rio, and
J. Faubert. Multiband negative refraction in one-dimensional photonic crystals.
Opt. Express, 17(5):30363041, 2009.
[35] A.V. Kavokin, G. Malpuech, and I.A. Shelykh. Negative refraction of light in
Bragg mirrors made of porous silicon. Physics Letters A, 339(3-5):387392, 2005.
[36] I.C. Botten, M.S. Craig, L.V. Adams, and J.R. Andrewarth. The dielectric lamillar
diraction grating. Journal of Modern Optics, 28(3):413428, 1981.
[37] M.A. Kaliteevski, D.M. Beggs, S. Brand, R.A. Abram, and V.V. Nikolaev. Stability of the photonic band gap in the presence of disorder. Phys. Rev. B, 73(3):
033106, 2006.
[38] A. Kavokin, G. Malpuech, A. Di Carlo, P. Lugli, and F. Rossi. Photonic Bloch
oscillations in laterally conned Bragg mirrors. Phys. Rev. B, 61:44134416, 2000.
[39] J.M. Chamberlain. Where optics meets electronics: recent progress in decreasing
the terahertz gap. Philosophical Transactions of the Royal Society A.P, 362:1815,
2004.
Bibliography
99
[40] J. B. Jackson, M. Mourou, J. F. Whitaker, I. N. Duling, S. L. Williamson,

M. Menu, and G. A. Mourou. Terahertz imaging for non-destructive evaluation of
mural paintings. Optics Letters, 281:527532, 2008.
[41] E. Pickwell and V. P. Wallace. Biomedical applications of terahertz technology.
Journal of Physics D, 39:301, 2006.
[42] S. Brand, R. A. Abram, and M. A. Kaliteevski. Complex photonic band structure
and eective plasma frequency of a two-dimensional array of metal rods. Phys.
Rev. B, 75(3):035102, 2007.
[43] A. J. Gallant, M. A. Kaliteevski, S. Brand, D. Wood, M. Petty, R. A. Abram,
and J. M. Chamberlain. Terahertz frequency bandpass lters. Journal of Applied
Physics, 102(2):023102, 2007.
[44] T. Matsui, A. Agrawal, A. Nahata, and Z.V. Vardeny. Transmission resonances
through aperiodic arrays of subwavelength apertures. Nature, 446:517521, 2007.
[45] B. Voisiat, A. Biinas, I. Kaalynas, and G. Raiukaitis. Band-pass lters for thz
spectral range fabricated by laser ablation. Applied Physics A, 104:953958, 2011.
[46] R. Br
uckner, M. Sudzius, S. I. Hintschich, H. Frob, V. G. Lyssenko, M. A. Kaliteevski, I. Iorsh, R. A. Abram, A. V. Kavokin, and K. Leo. Parabolic polarization
splitting of Tamm states in a metal-organic microcavity. Appl. Phys. Lett., 100
(6):062101, 2012.
[47] A. Askitopoulos, L. Mouchliadis, I. Iorsh, G. Christmann, J. J. Baumberg, M. A.
Kaliteevski, Z. Hatzopoulos, and P. G. Savvidis. Bragg polaritons: Strong coupling
and amplication in an unfolded microcavity. Phys. Rev. Lett., 106:076401, 2011.
[48] I. A. Shelykh, R. Johne, D. D. Solnyshkov, A. V. Kavokin, N. A. Gippius, and
G. Malpuech. Quantum kinetic equations for interacting bosons and their application for polariton parametric oscillators. Phys. Rev. B, 76:155308, 2007.
[49] P. Yeh, A. Yariv, and A. Cho. Optical surface waves in periodic layered media.
Appl. Phys. Lett., 32(2):104105, 1978.
[50] A. Yariv and P. Yeh. Optical waves in crystals: propagation and control of laser
radiation. Wiley, New York, 2002.
[51] T. C. H. Liew, A. V. Kavokin, T. Ostatnick
y, M. Kaliteevski, I. A. Shelykh, and
R. A. Abram. Exciton-polariton integrated circuits. Phys. Rev. B, 82:033302,
2010.
[52] O. Gazzano, S. Michaelis de Vasconcellos, K. Gauthron, C. Symonds, J. Bloch,
P. Voisin, J. Bellessa, A. Lematre, and P. Senellart. Evidence for conned Tamm
plasmon modes under metallic microdisks and application to the control of spontaneous optical emission. Phys. Rev. Lett., 107:247402, 2011.
Bibliography
100
[53] P. B. Johnson and R. W. Christy. Optical constants of the noble metals. Phys.
Rev. B, 6(12):43704379, 1972.
[54] E.L. Ivchenko, M.M Voronov, M.V. Erementchouk, L.I. Deych, and Lisyansky
A.A. Multiple-quantum-well-based photonic crystals with simple and compound
elementary supercells. Phys. Rev. B, 70(19):195106, 2004.
[55] A.Y. Sivachenko, M.E. Raikh, and Z.V. Vardeny. Excitations in photonic crystals
inltrated with polarizable media. Phys. Rev. A., 64(1):138091138098, 2001.
[56] C. Ciuti, P. Schwendimann, and A. Quattropani. Parametric luminescence of
microcavity polaritons. Phys. Rev. B, 63:041303, 2001.
[57] R. M. Stevenson, V. N. Astratov, M. S. Skolnick, D. M. Whittaker, M. EmamIsmail, A. I. Tartakovskii, P. G. Savvidis, J. J. Baumberg, and J. S. Roberts.
Continuous wave observation of massive polariton redistribution by stimulated
scattering in semiconductor microcavities. Phys. Rev. Lett., 85:36803683, 2000.
[58] P.G. Savvidis, J.J. Baumberg, R.M. Stevenson, M.S. Skolnick, D.M. Whittaker,
and J.S. Roberts. Angle-resonant stimulated polariton amplier. Phys. Rev. Lett,
84(7):15471550, 2000.
[59] J. J. Hopeld and D. G. Thomas. Theoretical and experimental eects of spatial
dispersion on the optical properties of crystals. Phys. Rev., 132(2):563572, Oct
1963. doi: 10.1103/PhysRev.132.563.
[60] Y. Yamamoto, F. Tassone, and H. Cao. Semiconductor Cavity Quantum Electrodynamics. Springer, New York, 2000.
[61] Y. Narukawa, Y. Kawakami, S. Fujita, and S. Nakamura. Dimensionality of excitons in laser-diode structures composed of inx ga1x N multiple quantum wells.
Phys. Rev. B, 59:1028310288, 1999.
[62] I. Iorsh, M. Glauser, G. Rossbach, J. Levrat, M. Cobet, R. Butte, N. Grandjean,
M. A. Kaliteevski, R. A. Abram, and A. V. Kavokin. Generic picture of the
emission properties of III-nitride polariton laser diodes: Steady state and current
modulation response. Phys. Rev. B, 86:125308, 2012.
[63] J. Kasprzak, M. Richard, S. Kundermann, A. Baas, P. Jeambrun, J. M. J. Keeling,
F. M. Marchetti, R. Andre M. H. Szyma
nska, J. L. Staehli, V. Savona, P. B.
Littlewood, B. Deveaud, and Le Si Dang. Bose Einstein condensation of exciton
polaritons. Nature (London), 443:409, 2006.
[64] S. Christopoulos, G. Baldassarri Hoger von Hogersthal, A. J. D. Grundy, P. G.
Lagoudakis, A. V. Kavokin, J. J. Baumberg, G. Christmann, R. Butte, E. Feltin,
J.-F. Carlin, and N. Grandjean. Room-temperature polariton lasing in semiconductor microcavities. Phys. Rev. Lett., 98:126405, 2007.
Bibliography
101
[65] G. Christmann, R. Butte, E. Feltin, J.-F. Carlin, and N. Grandjean. Room temperature polariton lasing in a GaN/AlGaN multiple quantum well microcavity.
Appl. Phys. Lett., 93:051102, 2008.
[66] S. Kena-Cohen and S. R. Forrest. Room-temperature polariton lasing in an organic
single-crystal microcavity. Nature Phot., 4:371, 2010.
[67] K. G. Lagoudakis, M. Wouters, M. Richard, A. Baas, I. Carusotto, R. Andre,
L. S. Dang, and B. Deveaud-Pledran. Quantized vortices in an exciton polariton
condensate. Nature Physics, 4:706, 2008.
[68] A. V. Kavokin Y. G. Rubo R. Andre K. G. Lagoudakis, T. Ostatnick
y and
B. Deveaud-Pledran. Observation of half-quantum vortices in an exciton-polariton
condensate. Science, 326:974, 2009.
[69] A. Amo, D. Sanvitto, F. P. Laussy, D. Ballarini, E. del Valle, M. D. Martin,
A. Lematre, J. Bloch, D. N. Krizhanovskii, M. S. Skolnick, C. Tejedor, and L. Vi
na. Collective uid dynamics of a polariton condensate in a semiconductor microcavity. Nature (London), 457:291, 2009.
[70] A. Amo, J. Lefr`ere, S. Pigeon, C. Adrados, C. Ciuti, I. Carusotto, R. Houdre,
E. Giacobino, and A. Bramati. Superuidity of polaritons in semiconductor microcavities. Nature Physics, 5:805, 2009.
[71] K.G. Lagoudakis, B. Pietka, M. Wouters, R. Andre, and B. Deveaud-Pledran.
Coherent oscillations in an exciton-polariton Josephson junction. Phys .Rev. Lett.,
105:120403, 2010.
[72] A.A. Khalifa, A.P.D. Love, D.N. Krizhanovskii, M.S. Skolnick, and J.S. Roberts.
Electroluminescence emission from polariton states in GaAs-based semiconductor
microcavities. Appl. Phys. Lett, 92:061107, 2008.
[73] H. Mathieu, P. Lefebre, and P. Christol. Simple analytical method for calculating
exciton binding energies in semiconductor quantum wells. Phys. Rev. B, 46:4092,
1992.
[74] J. Bloch V. Thierry-Mieg R. Andre Le Si Dang S. Kundermann A. Mura G. Bongiovanni J. L. Staehli M. Saba, C. Ciuti and B. Deveaud. High-temperature ultrafast polariton parametric amplication in semiconductor microcavities. Nature
(London), 414:731, 2001.
[75] R. Butte and N. Grandjean. A novel class of coherent light emitters: polariton
lasers. Semicond. Sci. Technol., 26:014030, 2011.
[76] D. Solnyshkov, E. Petrolati, A. Di Carlo, and G. Malpuech. Theory of an electrically injected bulk polariton laser. Appl. Phys. Lett., 94:011110, 2009.
Bibliography
102
[77] G. Christmann, R. Butte, E. Feltin, A. Mouti, P. A. Stadelmann, A. Castiglia, J.F. Carlin, and N. Grandjean. Large vacuum Rabi splitting in a multiple quantum
well GaN-based microcavity in the strong-coupling regime. Phys. Rev. B, 77:
085310, 2008.
[78] G. Christmann, R. Butte, E. Feltin, J.-F. Carlin, and N. Grandjean. Impact
of inhomogeneous excitonic broadening on the strong exciton-photon coupling in
quantum well nitride microcavities. Phys. Rev. B, 73:153305, 2006.
[79] R. Butte, J. Levrat, G. Christmann, E. Feltin, J.-F. Carlin, and N. Grandjean.
Phase diagram of a polariton laser from cryogenic to room temperature. Phys.
Rev. B, 80:233301, 2009.
[80] M. Suzuki, J. Nishio, M. Onomura, and C. Hongo. GaN polariton LEDs. J. Cryst.
Growth, 189/190:511, 1998.
[81] G. Cosendey, J.-F. Carlin, N.I.K Kaufmann, R. Butte, and N. Grandjean. Strain
compensation in AlInN/GaN multilayers on GaN substrates: Application to the
realization of defect-free Bragg reectors. Appl. Phys. Lett., 98(18):181111, 2011.
[82] G. Christmann, D. Simeonov, R. Butte, E. Feltin, J.-F. Carlin, and N. Grandjean.
Impact of disorder on high quality factor III V nitride microcavities. Appl. Phys.
Lett., 89(26):261101, 2006.
[83] A. David, M.J. Grundmann, J.F. Kaeding, N.F. Gardner, T.G. Mihopoulos, and
M.R. Krames. Carrier distribution in (0001)InGaN/GaN multiple quantum well
light-emitting diodes. Appl. Phys. Lett., 92(5):053502, 2008.
[84] G. Malpuech, D. Solnyshkov, A. Di Carlo, and E. Petrolati. patent pending
2.938.384, bulletin ociel de la propriete industrielle. Institut National de la Propriete Industrielle, 10/19:132, 2010.
[85] R. Butte, G. Christmann, E. Feltin, A. Castiglia, J. Levrat, G. Cosendey, A. Altoukhov, J.-F. Carlin, and N. Grandjean. Room temperature polariton lasing in
III-nitride microcavities: a comparison with blue GaN-based vertical cavity surface
emitting lasers. Proc. SPIE, 7216:721619, 2009.
[86] A. Castiglia, D. Simeonov, H. J. Buehlmann, J.-F. Carlin, E. Feltin, J. Dorsaz,
R. Butte, and N. Grandjean. Ecient current injection scheme for nitride vertical
cavity surface emitting lasers. Appl. Phys. Lett., 90(3):033514, 2007.
[87] T.-C. Lu, C.-C. Kao, H.-C. Kuo, G.-H. Huang, and S.-C. Wang. Cw lasing of
current injection blue GaN-based vertical cavity surface emitting laser. Appl.
Phys. Lett., 92(14):141102, 2008.
[88] S. Nakamura, G. Fasol, and S. J. Pearton. The Blue Laser Diode: The Complete
Story. Springer, Berlin, 2000.
Bibliography
103
[89] V. Savona, L. C. Andreani, P. Schwendimann, and A. Quattropani. Quantum well

excitons in semiconductor microcavities: Unied treatment of weak and strong
coupling regimes. Solid State Comm., 93:733, 1995.
[90] H. A. McLeod. Thin-Film Optical Filters, 2nd ed. Adam Hilger, Bristol, U.K.,
1996.
[91] D. Brunner, H. Angerer, E. Bustarret, F. Freudenberg, R. Hopler, R. Dimitrov,
O. Ambacher, and M. Stutzmann. Optical constants of epitaxial AlGaN lms and
their temperature dependence. J. Appl. Phys, 82:5090, 97.
[92] J.-F. Carlin and M. Ilegems. High-quality AlInN for high index contrast Bragg
mirrors lattice matched to GaN. Appl. Phys. Lett., 83(4):668670, 2003.
[93] J.-F. Carlin, C. Zellweger, J. Dorsaz, S. Nicolay, G. Christmann, E. Feltin, R. Butt,
and N. Grandjean. Progresses in III-nitride distributed Bragg reectors and microcavities using AlInN/GaN materials. physica status solidi B, 242(11):23262344,
2005.
[94] M. J. Bergmann and H. C. Casey. Optical-eld calculations for lossy multiple-layer
Alx Ga1x N/Inx Ga1x N laser diodes. Journal of Applied Physics, 84:1196, 1998.
[95] R. Schmidt, B. Rheinlander, M. Schubert, D. Spemann, T. Butz, J. Lenzner,
E. M. Kaidashev, M. Lorenz, A. Rahm, H. C. Semmelhack, and M. Grundmann.
Dielectric functions (1 to 5 ev) of wurtzite Mgx Zn1x O(x < 0.29) thin lms. Appl.
Phys. Lett., 82(14):22602262, 2003.
[96] F. Tassone and Y. Yamamoto. Exciton-exciton scattering dynamics in a semiconductor microcavity and stimulated scattering into polaritons. Phys. Rev. B, 59:
1083010842, 1999.
[97] A. Kavokin, J.J. Baumberg, G. Malpuech, and F.P. Laussy. Microcavities. Clarendon Press, Oxford, 2006.
[98] L. E. Golub, A. V. Scherbakov, and A. V. Akimov. Energy distributions of 2D
excitons in the presence of nonequilibrium phonons. J. Phys. : Condensed Matter,
8:2163, 1996.
[99] J. Levrat, R. Butte, G. Christmann, E. Feltin, J.-F. Carlin, and N. Grandjean.
Tailoring the strong coupling regime in III-nitride based microcavities for room
temperature polariton laser applications. Phys. Status Solidi (c), 6:2820, 2009.
[100] J. Levrat, R. Butte, E. Feltin, J.-F. Carlin, N. Grandjean, D. Solnyshkov, and
G. Malpuech. Condensation phase diagram of cavity polaritons in GaN-based
microcavities: Experiment and theory. Phys. Rev. B, 81:125305, 2010.
Bibliography
104
[101] G. Malpuech, A. Kavokin, A. Di Carlo, and J. J. Baumberg. Polariton lasing

by exciton-electron scattering in semiconductor microcavities. Phys. Rev. B, 65:
153310, 2002.
[102] G. Ramon, A. Mann, and E. Cohen. Theory of neutral and charged exciton scattering with electrons in semiconductor quantum wells. Phys. Rev. B, 67:045323,
2003.
[103] W. G. Scheibenzuber, U. T. Schwarz, L. Sulmoni, J. Dorsaz, J.-F. Carlin, and
N. Grandjean. Recombination coecients of GaN-based laser diodes. Journal of
Applied Physics, 109(9):093106, 2011.
[104] M. Suzuki, T. Uenoyama, and A. Yanase. First-principles calculations of eectivemass parameters of AlN and GaN. Phys. Rev. B, 52:81328139, 1995.
[105] R. Truell, C. Elbaum, and B. B. Chick. Ultrasonic Methods in Solid State Physics.
Academic Press, New York, 1969.
[106] J. Kasprzak, D. D. Solnyshkov, R. Andre, Le Si Dang, and G. Malpuech. Formation
of an exciton polariton condensate: Thermodynamic versus kinetic regimes. Phys.
Rev. Lett., 101:146404, 2008.
[107] L. A. Coldren and S. W. Corzine. Diode Lasers and Photonic Integrated Circuits.
Wiley, New York, 1995.
[108] X. Fan, G.P. Wang, J.C.W. Lee, and C. T. Chan. All-angle broadband negative
refraction of metal waveguide arrays in the visible range: Theoretical analysis and
numerical demonstration. Phys. Rev. Lett., 97:073901, 2006.
[109] R. Scholz M. Sudzius S. I. Hintschich H. Frob V. G. Lyssenko R. Bruckner, A.
A. Zakhidov and K. Leo. Phase-locked coherent modes in a patterned metalorganic
microcavity. Nature Phot., 6:322, 2012.
[110] V. M. Agranovich and V. E. Kravtsov. Notes on crystal optics of superlattices.
Solid State Communications, 55(1):85, 1985.
[111] D. R. Smith, P. Kolinko, and D. Schurig. Negative refraction in indenite media.
J. Opt. Soc. Am. B, 21(5):10321043, 2004.
[112] Z. Jacob, L.V. Alekseyev, and E. Narimanov. Optical hyperlens: Far-eld imaging
beyond the diraction limit. Opt. Express, 14(18):82478256, 2006.
[113] Z. Jacob, I. Smolyaninov, and E. Narimanov. Broadband Purcell eect: Radiative
decay engineering with metamaterials. ArXiv e-prints, page 0910.3981, October
2009.
Bibliography
105
[114] A. N. Poddubny, P. A. Belov, P. Ginzburg, A. V. Zayats, and Y. S. Kivshar.

Microscopic model of Purcell enhancement in hyperbolic metamaterials. Phys.
Rev. B, 86:035148, 2012.
[115] M. A. Noginov, H. Li, Yu. A. Barnakov, D. Dryden, G. Nataraj, G. Zhu, C. E. Bonner, M. Mayy, Z. Jacob, and E. E. Narimanov. Controlling spontaneous emission
with metamaterials. Opt. Letters, 35(11):18631865, 2010.
[116] Volkhard Helms. Principles of Computational Cell Biology. Weinheim: WileyVCH, 2008.
[117] C. L. Cortes, W. Newman, S. Molesky, and Z. Jacob. Quantum nanophotonics
using hyperbolic metamaterials. Journal of Optics, 14:063001, 2012.

File PDF

Uploaded by

Copyright:

Available Formats

File PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

File PDF

Uploaded by

Copyright:

Available Formats

Table of contents

2 Negative refraction of light at the edges of one and two-dimensional

Parametric amplication by a periodic array of quantum wells

4 Kinetics of exciton-polariton condensates and the modelling of polariton lasers

5 Summary, conclusions and further work

A schematic diagram of the refraction of light at the side edge of a Bragg

The simplest realization of the structure where a Tamm plasmon mode

(a) Schematic diagram of a polariton laser. (b) Schematic illustration of

Parameters of the modelling of polariton laser . . . . . . . . . . . . . . . . 80

University, University of Crete, and Ecole

exciton can be written in the frequency domain as [7]:

Negative refraction of light at the

Review of negative refraction.

by a negative refractive index n = , and as a consequence, with negative phase

n = ( ) i| + |. However, in order to comply with the causality

dispersion, W is given by [22]:

where p is the metal plasma frequency, and

which is always positive.

Scattering of light at the side-edge of a Bragg reector.

Electric and magnetic elds in vacuum

[exp(i0 x)p0 + Rp exp(ip x)] exp(iKp z),

where K0 = k0 sin , Kp = K0 + 2p/D, p =

k02 Kp2 , the Rp are the amplitudes

component Hz (x, z) from the Maxwell equation E = (H/t)/c = ik0 H:

Electric and magnetic elds in the photonic crystal

Cm um (z) exp(im x),

2 k02 2 . It has been shown that these equation

where the functions (z), (z) are given by

sin(kz1 d1 ) sin(kz2 (z d1 )), d1 z D

sin(kz1 d1 ) sin(kz2 (z d1 )), d1 z D

Applying the continuity conditions at the interface

[p0 + Rp ] exp iKp z =

which can be written in matrix notation as

where = diag[q ], = diag[m ]. Then substituting R into (2.17),then gives

be included, we rst introduce the sum sn =

k02 Kp2 x],

Comparison of the analytical theory and numerical modelling

Results and conclusions

Negative refraction of THz radiation by a prism made

We consider a two-dimensional hexagonal photonic crystal formed by a periodic array of

which substituted into Eq. 2.28 gives:

The spatially-dependent dielectric permittivity (r) of the structure can be written as

S(r) as a series: S(r) =

Writing the wavevector k as k

where I is the identity matrix, A is a diagonal matrix with elements: Ag,g = 2g

from which, after choosing the frequency and direction u

H(r, t) = 2 H(r, t),

and the inverse dielectric permittivity can be written as:

which can be rewritten as

Multiplying (2.40) by (I P)1 , gives

as nef f = (d/dk) , where is the eective dielectric permittivity. Therefore, the

Numerical modelling results

Figure 2.7 shows numerical results from the-nite-dierence-time-domain modelling of

Comparison of theory and experiment

Experiments were performed by collaborators guided by the results of the numerical

Summary of key original results

Main results obtained in Chapter 2.

Eigenmode analysis of resonant

Chapter 3. Eigenmode analysis of resonant photonic nanostructures