You are on page 1of 23

Chapter 3

Photonic crystals

3.1 Introduction
3.1.1 What are photonic crystals
Photonic (or electromagnetic) crystals are artificial periodical structures op-
erating at frequencies where the wavelength is comparable with the character-
istic period of the structure [1]. The term electromagnetic crystals is usable
when the structure operates at radio frequencies [1–3]. When such structures
operate at optical frequencies (from the far infrared range to the visible light)
they are usually called photonic crystals [1–4]. Since building blocks of pho-
tonic crystals are made of natural materials, and natural materials in the
optical range lose magnetic properties, the internal periodicity of photonic
crystals is usually referred as the periodic coordinate dependence of the per-
mittivity ε(r) = ε(r + a), where a = (ax , ay , az ) is the lattice unit vector.
Since for isotropic dielectric media the permittivity ε is simply n2 , where n is
the medium refractive index, another known definition of photonic crystals is
artificial media with a periodic contrast of refractive index [2]. Notice, how-
ever, that this definition is not suitable for photonic crystals performed from
anisotropic constituents. Among them, so-called magneto-photonic crystals
(usually performed from optically transparent ferrites) are very important
(see e.g. [5, 6]). Complex permittivity of a transparent ferrite is a tensor
value. Moreover, in presence of the biasing magnetic field this tensor has off-
diagonal components, responsible for non-reciprocal optical effects in such
media. Therefore, the refractive index in such media is a value, depending
not only on frequency, but also on the propagation direction. The definition

28
CHAPTER 3. PHOTONIC CRYSTALS 29

of [2] for such arrays is meaningless. The definition of photonic crystals via
the spatially periodic refractive index n is also unsuitable for metallic pho-
tonic crystals, though metals are usually reciprocal and isotropic materials
(see e.g. [7]). At optical frequencies metals can be described via the isotropic
complex permittivity εm , whose imaginary part, responsible for optical losses,
can be relatively small. However, metals do not possess refractive properties
because Re(εm ) < 0 and even in absence of losses their ”refractive index” is
imaginary.

Figure 3.1: Schematic illustration of the idea of 1D (left panel), 2D (central


panel) and 3D (right panel) photonic crystals.

Let us define a photonic crystal as an inhomogeneous medium whose per-


mittivity tensor is spatially periodic. This definition covers all known cases
of photonic crystals, including gyrotropic (ferrite) and plasmonic (metal)
lattices. Below, we consider photonic crystals having in mind this defini-
tion. Unlike the first definition, it also covers the low-frequency region in
which photonic crystals can be homogenized and behave either as artificial
dielectrics or as metamaterials (see chapter Metamaterials). However, in
this lecture we do not concentrate on the dielectric or metamaterial regimes
of photonic crystals. The low-frequency region is included into consideration
only in order to analyze the so-called dispersion diagram of a photonic crystal
as a whole.
Geometrically, photonic crystals split onto three classes: 1D, 2D, and
3D, as it is shown in Fig. 3.1. Notice, that 3D photonic crystals can be
implemented as lattices of air voids. Then the matrix should be transparent
CHAPTER 3. PHOTONIC CRYSTALS 30

– a dielectric or a semiconductor medium. In fact, 1D and even 2D pho-


tonic crystals had been known before this terminology has appeared. The
term photonic crystal arose 30 years ago when the interest to spatially pe-
riodic structures sharply grew. The reason of this keen interest was the
publication of two pioneering works in the same month of 1987: that by E.
Yablonovitch [8] and that by S. John [9]. These experimental papers reported
two exciting effects: an unprecedent spatial concentration of the electromag-
netic field in 3D photonic crystals and inhibited spontaneous emission of
quantum emitters in them. The key point was the fabrication of 3D pho-
tonic crystals which allowed the experimental studies. Both reported effects
– 3D field concentration and control of the spontaneous emission (which can
be either inhibited or, on the contrary, enhanced in 3D dielectric lattices) –
were theoretically predicted by V. Bykov 15 years earlier in works [10, 11].

3.1.2 Natural photonic crystals

Figure 3.2: Left panel: Sea mouse (general view); scanning electron micro-
scope picture of the cross section of its hair; its hair under normally (top)
and obliquely (bottom) incident light. Right panel: Month eye (general view)
and its nanostructure, obtained by transmission electron microscope.

Among natural crystals an only known photonic crystal is opal. Opal has
a very large unit cell compared all other solids and, though in the visible and
IR ranges of frequencies it is a continuous medium, it becomes a photonic
crystal in the ultraviolet light [10]. Photonic crystals operating in the visible
CHAPTER 3. PHOTONIC CRYSTALS 31

range were found as parts of living systems – in animals. Two most famous
examples of living photonic crystals are illustrated by Fig. 3.2. The first one
is the sea mouse hair in Fig. 3.2, left panel. Sea mouse is a marine worm
with the 15-20 cm long body living on the sea bottom at depths 1-200 m.
It is covered with the hairs also called tubular spines. Each hair is a 2D
photonic crystal of submicron cylindrical holes located along the hair axis
in the matrix of tissue whose refractive index is close to 1.5. When light is
incident perpendicular to the axis of a hair it exhibits a red colouration. For
off-axis incidences greens and blues shades become seen. This optical effect
is due to the so-called directional photonic band gap, a phenomenon which is
discussed below in details. For the normal incidence the bandgap corresponds
to the red light. This means that the red light cannot propagate inside the
hair and is totally reflected. The directional bandgap means that the band of
frequencies for which the total reflection holds varies versus the incident light
direction. When the incidence angle increases the band of totally reflected
light becomes yellow, green and finally – for larger angles – blue. This effect
visually manifests in the angle-dependent coloration.
Another example of a photonic crystal in a living system is the moth
eye. In this case, the photonic crystal (a hexagonal lattice) is formed by
identical dielectric protrusions with lengths nearly 600 nm and lattice step
nearly 400 nm (Fig. 3.2, right panel). It operates as a superprism – a device
with extremely high dispersion of light 1 . This is an important nanophotonic
device and, we discuss it below in details. Here, we only mention that the
superprism in the eye allows a butterfly to distinguish much more colors and
shades that humans can do. Some other examples of living photonic crystals
can be found in [12].
1
Superprism operation corresponds to the light propagation across the nanorods. This
is achieved in the moth eye because the lattice of nanorods covers a spherical surface,
and these spherical nanostructured surfaces are triply repeated in the moth eye being
symmetrically tilted. Therefore for whatever direction of incident light, there is obviously
a part of the moth eye operating as a superprism.
CHAPTER 3. PHOTONIC CRYSTALS 32

3.2 Photonic band structures


3.2.1 Bragg scattering as a reason of photonic bandgap
The key phenomenon inherent to every photonic crystal is the Bragg scat-
tering. This scattering is in fact destructive interference of a set of waves,
reflected from the set of crystal planes. It was explained theoretically by
Lord Rayleigh in [13] 4 years after two brothers – W.L Bragg and W.H.
Bragg – had revealed this effect for X-rays in natural crystals [14]. The
theory by Rayleigh not only predicted this phenomenon for any periodic ar-
rays of electromagnetic scatterers, it further was extended to the solid-state
physics where it was developed by F. Bloch and L. Brillouin for electron wave
functions in dielectrics and semiconductors.

Figure 3.3: Illustration to the Bragg phenomenon: the wave reflected from
the reference crystal plane is a sum of all waves reflected by an infinite set
of crystal planes. This sum diverged on the Bragg condition.

Imagine a plane wave propagating through the set of partially reflecting


sheets. Let these sheets (crystal planes) are planar arrays of natural or
artificial atoms. In natural crystals, where the electron propagates as a wave
function they are real atoms. In photonic crystals they are inclusions in the
host matrix. In Fig. 3.3 for the sake of vertical space we show only one unit
cell of these planar arrays. Let us consider a sum of the multiple reflections
that occur with coefficient r on every plane. Here we have arbitrarily chosen a
reference crystal plane to which we refer our reflection coefficient. We simply
do not consider what happens at the left of the reference plane, assuming only
that in the crystal there exists a plane wave impinging it and producing a
backward wave due to partial reflections by crystal planes. Denoting the gap
between crystal planes as a and the wave number of the propagating light
as k we can write for the total reflection coefficient R a simple geometric
CHAPTER 3. PHOTONIC CRYSTALS 33

progression:
1
R = re−jkx + re−jkx e−j2ka + re−jkx e−j4ka + ... = re−jkx . (3.1)
1 − e−j2ka
In (3.1) we took in account that the path of the wave reflected from the next
crystal plane is evidently larger by 2a than the path of the wave reflected from
the previous plane. The right-hand side of (3.1) diverges when exp(−2jka) =
1 i.e. when k ≡ ω/v = mπ/a, where m is an integer number and v is the
speed of light in the host medium. This physically meaningless divergence
(whereas |R| cannot exceed unity due to power balance condition) implies
that the normal propagation of the wave at frequencies ωm = mvπ/a, we
have assumed in the beginning, is impossible. For the oblique propagation
the differential path 2a replaces by 2a sin θ and forbidden frequencies will
change respectively. It is called the Bragg condition2 .
In (3.1) crystal planes are modelled as infinitesimally thin sheets whose
reflectance r is not frequency dispersive. These two approximations are ad-
equate for X-rays in minerals and for electron wave functions in metals. For
electron wave functions in semiconductors and solid dielectrics r strongly de-
pends on frequency, the lattices comprise different types of scatterers and the
scatterer size can be neglected rarely. Therefore, the analysis becomes more
involved and the Bragg phenomenon results in the bandgap structure [1, 2].
The bandgap structure is a set of bandgaps – intervals of prohibited fre-
quencies. One distinguishes so-called complete (or photonic) and directional
bandgaps (or stopbands). If the propagation is forbidden for any possible di-
rection it is called complete bandgap. If the propagation is forbidden for some
directions and allowed for others it is called directional bandgap. Complete
bandgaps are especially important. If the spectrum of an embedded emitter
is within the complete bandgap the radiation is either confined around the
emitter like in a cavity or is suppressed. Which regime is implemented – that
of a cavity excitation or that of inhibited emission – depends on different fac-
tors [8–11]. Both regimes are important for optical sensing. In the most part
of 3D photonic crystals bandgaps are only directional. Complete bandgaps
in conventional 3D photonic crystals are known only only for diamond-like
lattices and for inverse opal lattices. However, if a photonic crystal is imple-
2
It makes sense if the period a is much larger than the other periods of the lattice. If
they are comparable an additional set of prohibited frequencies arises due to the reflections
from other crystallographic planes.
CHAPTER 3. PHOTONIC CRYSTALS 34

mented of resonant inclusions, e.g. plasmonic nanoparticles, it exhibits the


complete bandgap even being a simple cubic lattice [2].
As to 2D photonic crystals – lattices of cylindrical inclusions assumed to
be infinitely long – a complete bandgap does not mean that the propagation
is forbidden along the inclusions [2]. As a rule this propagation is allowed
in these crystals. For 2D crystals one considers only the propagation in the
plane perpendicular to the optical axis. This is so because the axial propa-
gation is not very important. The oblique propagation is also excluded from
consideration of 2D photonic crystals except the theory of so-called photonic-
crystal fibers (see below). In practice, a 2D photonic crystal represents a
layer of comparatively long dielectric or semiconductor nanorods whose axis
is orthogonal to the layer interfaces3 . The layer is placed on a dielectric or
semiconductor substrate. The electromagnetic sources exciting the photonic
crystal layer are located either in between the rods or in the substrate. In
both these cases the exciting oscillations are polarized normally to the layer
i.e. along the rods [1, 2]. Therefore, the most important modes excited in
the photonic crystal layer are modes confined in it and propagating in the
plane transverse to the rods. Therefore, for a 2D photonic crystal there is
a convention that the complete bandgap means that the in-plane prohibited
propagation, and the directional bandgap corresponds to some directions in
this plane for which the propagation is prohibited.

3.2.2 Analogy with the solid-state bandgap structure


Quantum mechanics applied to the solid state answered a keen question that
was formulated by P.P. Drude in 19094 . Why do electrons of a metal (or
a highly-doped semiconductor) move in their high-frequency spatial oscil-
lations like in a free non-charged plasma? Electrons considered as small
particles should have scattered on the lattice ions that should have resulted
in the usual Ohm’s conductivity for whatever frequency. However, in the op-
3
A similar structure of metal nanorods is more interesting in the low-frequency range
where it is called wire medium, than at higher frequencies, were it is a bandgap material.
Wire media (see e.g. in [15]) deserve a separate study, and we do not concern them in
this lecture. As to metal photonic crystals as bandgap materials, they are applied at
microwaves and even in the range of THz frequencies (see e.g. in [16]). In nanophotonics
they are not advantageous and rarely applied.
4
The author of the first – classical – electromagnetic theory of metals unable to answer
this question.
CHAPTER 3. PHOTONIC CRYSTALS 35

Figure 3.4: Bangap structures of electron states in solid conductors, semicon-


ductors and insulators. Potential energy Wc of an electron in the conductive
state is related to the frequency of optical transition ω as Wc = h̄ω + Wv ,
where Wv is the electron potential energy in the valence zone.

tical range many metals are weakly conductive and, on the contrary, manifest
strong plasma oscillations. Adopting this experimental fact as a postulate,
Drude has developed the theory of metal electron gas. The answer to Drude’s
question was given in 1928 by A. Sommerfeld. It is as follows. In all solids the
electrons obey to the Fermi-Dirac statistics for which the potential energy of
electrons is crucial. Electrons having sufficient potential energies (Wp > 0)
propagate as packages of plane waves. Plane waves cannot experience scatter-
ing losses in a regular lattice5 . Sommerfeld’s development of Drude’s theory
of metals is based on the electron wave functions under spatially periodic
potential [17]. Since the work by F. Bloch [18], who extended Sommerfeld’s
approach to solid dielectrics and semiconductors, it has become common for
the solid state physics.
The Sommerfeld-Bloch approach resulted in the energetic zone structures
for all three solid states – conducting, semiconducting and insulating, which
are illustrated by Fig. 3.4. This figure shows only the main bandgap – that
between the valence and the conduction zones – in the logarithmic scale.
When the electron transits from the valence zone to the conduction zone (or
vice versa) the electron wave-function frequency increases (or decreases) by
the value ω which is equal to the frequency of a photon with energy Wp = h̄ω
absorbed (or emitted) during this transition. In the valence zone the potential
5
Scattering loss is a stochastic non-coherent process impossible in an infinite regular
lattice.
CHAPTER 3. PHOTONIC CRYSTALS 36

energy of electrons is negative Wv < 0 i.e. insufficient to let them move


through the crystal lattice. Electrons of the conductive zone are almost free
and can move. In metals the electron bandgap is so narrow that can be
neglected in presence of the thermal movement. Therefore, the valence zone
is not separated from the conductivity zone. In semiconductors the electronic
bandgap is noticeable. Electrons with low positive potential energies can be
neither immovable (their potential energy is above the threshold W = 0
and must transit to the kinetic energy) nor propagating due to the Bragg
phenomenon. Therefore, corresponding electron states cannot exist: there
is an electron bandgap. As to insulators (dielectric media), their electron
bandgaps are wider than that of a typical semiconductor by an order of
magnitude. Therefore, the visible and even ultraviolet light in the linear
regime cannot transfer electrons to the conduction zone6 .

Figure 3.5: Bangap structure of a typical 2D photonic crystal with honey-


comb lattice (of holes in the dielectric matrix). The bandgap is shown in the
dispersion diagram.

For electron waves in natural crystals the complete bandgap whose origin
is explained above is the most important. Therefore, one are often happy
6
Optical transitions in solid insulators (called impact ionization) are possible in the
high-intensity visible or UV light due to nonlinear effects. Linear transitions arise in the
high-energy radiation (X-rays and γ-rays) and usually result in the lattice destruction.
CHAPTER 3. PHOTONIC CRYSTALS 37

with a one-dimensional zone diagram shown in Fig. 3.4, which does not
allow one to analyze the dispersion of the wave functions and to find di-
rectional bandgaps. As to photonic crystals, one needs to obviously analyze
these properties, because applications of photonic crystals are based on them.
Therefore, band structures of photonic crystals are never shown as simply
as in Fig. 3.4. They are usually represented in the form of dispersion dia-
grams – plots of the frequency (or photon energy) versus the wave vector of
the propagating eigenmode. An example of the dispersion diagram is shown
in Fig. 3.5. Specialists in nanophotonics must understand the dispersion
diagrams very well. Therefore, let us learn this topic in details.

3.2.3 Bloch’s theorem


In order to understand the dispersion diagram of an infinite periodic struc-
ture we need to familiarize with a very important theorem on eigenwaves of
periodic structures. It was proved by F. Bloch for wave functions of elec-
trons in a natural crystal and further expanded to all periodic structures
supporting electromagnetic waves, acoustic waves, plasma waves, etc.7 .
For simplicity of writing let us consider a 1D photonic crystal with period
a as depicted in Fig. 3.6. In accordance to Bloch’s theorem each eigenmode
of this structure (in our case the eigenmode is formed by electric E and
magnetic H fields) can be presented as the discrete sum of plane waves,
which for time dependence of monochromatic fields exp(jωt) takes form:
  
E(r) E0 −jqz X En −j(qz+Gn ·r)
= e + e , (3.2)
H(r) H0 Hn
n6=0

where Gn is the vector with (in the present case an only) component Gn =
2πn/a. Bloch’s theorem evidently generalizes to 2D and 3D photonic crys-
tals. In these cases n becomes a vector. For a 3D photonic crystal with
periods (ax , ay , az ) we have n = (nx , ny , nz ) and Gn ≡ (Gx , Gy , Gz ) =
(2πnx /ax , 2πny /ay , 2πnz /az ).
In formula (3.2) the first term called the Bloch wave is shared out from
the plane-wave expansion. The Bloch wave is a plane wave with wave vector
q, whose – in the present 1D case – only component is restricted by condition
7
This is so because electron wave functions obey to the Schrödinger equation which
in the linear regime reduces to the wave equation like Maxwell’s equations do. The same
refers to governing equations of the gas, plasma and liquid dynamics.
CHAPTER 3. PHOTONIC CRYSTALS 38

Figure 3.6: Let point B in an infinite photonic crystal with period c in the
z-direction be shifted with respect to point A by Lz periods. The electro-
magnetic fields of the eigenmode at points A and B differ only by the phase
shift qLz c, where q is Bloch’s wave vector of the mode.

−π/a < q < π/a. It is the fundamental plane wave of the eigenmode expan-
sion because higher spatial harmonics of the expansion (with n 6= 0) oscillate
with smaller spatial periods. If we assume that in the Bloch expansion the
absolute value of the Bloch wave vector is higher, e.g. π/a < q < 2π/a this
wave vector will be not distinguishable from the wave vector (q −2π/a) which
enters the second term of expression (3.2) – series of high-order harmonics.
Then, the replacement qnew → (q −2π/a) will return the correct Bloch vector
with −π/a < qnew < 0 < π/a.
Bloch’s theorem has two important implications. First, the problem of
the infinite lattice can be reduced to the so-called cell problem. Really, from
(3.2) the quasi-periodicity condition follows:
 
E(r + Lz a) E(r) −jLz q·a
= e , (3.3)
H(r + Lz a) H(r)
This condition naturally generalizes to 2D and 3D photonic crystals, in the
last case Lz q · a replaces by Lx qx ax + Ly qy ay + Lz qz az . The special case
Lx,y,z = 1 shows that the fields at the opposite facets of the unit cell differ only
CHAPTER 3. PHOTONIC CRYSTALS 39

by the phase shift q · a. The boundary problem of electrodynamics when the


fields at the boundary of a finite spatial domain (cell) are a priori unknown,
but it is known how fields in some parts of the boundary are related to fields
in the opposite parts is called the cell problem of mathematical physics.
In mathematical physics there is a chapter treating the cell problems (not
only for electrodynamics). In the general case, the cell problem is called
Hamilton’s eigenvalue problem [20]. For the lattice cell with only the phase
shift difference between the electric and magnetic fields at the opposite sides
of the cell, the Hamilton theory claims the existence of a discrete set of
eigenfrequencies ω provided the phase shift qx,y,z ax,y,z between the opposite
sides of the cell with sizes (ax , ay , az ) is known. These solutions are real for
a lossless cell and exist obviously.
0.8
q → k − 2π/a
0.7

0.6

0.5
ωa/2πc

q → 2π/a − k
0.4

0.3
Photonic Band Gap
0.2

0.1

0
-0.5 0 0.5 1 1.5
ka/2π

Figure 3.7: Brillouin’s band diagram of an infinite stack of dielectric bilayers


along axis z. Layers of equal thickness have relative permittivities 1 and 7,
respectively. In the first passband the wave vector k = kz ≡ k is equal to
q. In the 2d passband k is replaced by q = 2π/a − k. In the 3d passband
q = k − 2π/a.

The second implication of Bloch’s theorem is the possibility to restrict the


wave vector axis in the dispersion diagram by only the Bloch wave vector.
This allowed L. Brillouin to replace the dispersion plot ω(k), where k is the
wave vector with arbitrary absolute value (−∞ < kz < ∞) inherent to the
CHAPTER 3. PHOTONIC CRYSTALS 40

unbounded space or aperiodic structures by the so-called Brillouin’s disper-


sion diagram [19]. This was initially done for the dispersion of electron waves
in the solid-state physics. The idea by Brillouin is illustrated by Fig. 3.7 for
a 1D crystal. In this figure the Brillouin diagram is shown for a photonic
crystal formed by two dielectric layers of thickness a/2 with permittivities
ε1 =1 and ε2 =7.
At frequencies of the 1st passband the Bloch wave is the main term in the
expansion (3.2) of the eigenmode. Within high-order passbands the Bloch
wave is not anymore the main term in this expansion. In the 2d passband
the dominant term has the wave vector with the z-component π/a < k <
2π/a. Therefore, the usual (non-squeezed) dispersion branch shown in Fig.
3.7 as a red dashed line corresponds to π/a < k < 2π/a. However, the
Bloch wave vector q is uniquely related to this k, and its z-component is
equal to q = 2π/a − k. Therefore, the corresponding dispersion branch
ω(k) describing the dispersion of the eigenmode can be transposed to the
Brillouin zone −π/a < q < π/a as it is shown in the Figure. In the 3d
passband 2π/a < k < 3π/a and Bloch’s wave vector q has the z-component
q = k − 2π/a. Again, the corresponding dispersion branch can be transposed
to the Brillouin zone. As a result, in the Brillouin dispersion diagram all
eigenfrequencies are located in the Brillouin zone, where the dispersion curves
corresponding to even-numbered passbands are inverted.
In the standard Brillouin diagram the coordinate axes are obviously nor-
malized. Instead of −π/a < q < π/a the abscissa is −1/2 < (qa/2π) < 1/2.
The normalized frequency is ωa/2πc and is also dimensionless.

3.2.4 Brillouin diagrams of 2D and 3D photonic crys-


tals
A reciprocal lattice with coordinate axes (kx , ky , kz ) which are components
of the wave vector can be introduced for a real 3D lattice with periods
(ax , ay , az ). In this lattice one may share the Brillouin zone −π/ax,y,z <
qx,y,z, < π/ax,y,z . As above, for k within the Brillouin zone we use the no-
tation k ≡ q. The Brillouin zone of a 2D square (ax = ay = a) lattice is
illustrated by Fig. 3.8. Since the Brillouin diagram is a plot ω(q), where q is a
vector, the horizontal axis splits onto intervals in which different components
of the Bloch vector vary separately. In order to make this diagram compact
and easy for analysis Brillouin has suggested the following algorithm.
CHAPTER 3. PHOTONIC CRYSTALS 41

* (&" = '/4, &! = '/4)


!
-!
" -"
$ % (&" = '/4, &! = 0)
# +

Figure 3.8: Real lattice unit cell, where the eigenmode with Bloch’s vector q
propagates, and the Brillouin zone of the reciprocal lattice (right). Brillouin’s
dispersion diagram is obtained by the variation of q between points Γ, X and
M.

First, for lattices with a symmetric unit cell (recall Fig. 3.7) the negative
part of the Brillouin zone −π/a < qx,y < 0 contains no new information,
and one may get rid of it. In fact only one quadrant of the Brillouin zone
is relevant for symmetric lattices. Moreover, there is no need to find all ω
for all possible value of the components of q within this quadrant. It is
enough to vary this vector over the triangle formed by points Γ, X and M .
Wave vectors varying over any other triangle of those onto which the whole
Brillouin zone may split will have the same eigenfrequencies.
In the first part of the band diagram of a 2D crystal with rectangular
unit cell the wave vector q is directed along x and its only x-component
varies from 0 (origin of the reciprocal space is the center of the Brillouin
zone called point Γ) to π/ax (called point X). In the second part, one keeps
qx = π/ax and varies qy from 0 to π/ay (point M ) increasing the tilt of
the wave vector with respect to the axis x until the wave vector acquires
the diagonal direction. Finally, keeping the same p diagonal direction of the
wave vector one decreases its absolute value from (π/ax )2 + (π/ay )2 to 0,
returning in this way vector q to Γ-point. Such the diagram for a lattice
of dielectric cylinders of thickness a/2 with permittivity ε = 10 located in
free space is shown in Fig. 3.9. The sufficiency of the triangle Γ − X − M
results from the so-called group symmetry in the rectangular cell problem
with symmetric inhomogeneities [1, 2, 20].
On the insets of Fig. 3.9 the color maps of the simulated electric field are
shown for two eigenmodes – that with qx = π/a, qy = 0 (corresponding to
point X) and that with qx = qy = π/a (point M ). At these edge points of the
CHAPTER 3. PHOTONIC CRYSTALS 42

Figure 3.9: Brillouin’s band diagram of a square array of cylinders with


relative permittivity ε = 10 and diameter a/2 located in free space.

Brillouin zone two inclusions, adjacent along the propagation direction, are
polarized with the opposite phases. Strictly speaking, such the propagating
wave cannot exist because for it the Bragg condition (3.1) is satisfied. The
Bloch mode whose q corresponds to X and M points is called the Bragg
mode. It is seen on the dispersion diagram that the Bragg mode has zero
group velocity ∂ω/∂q. Group velocity in the course of electrodynamics is
defined as vg = ∂ω/∂k, and inside the Brillouin zone q ≡ k. At both X and
M points the dispersion curve has zero slope, i.e. vg = 0. The wave with zero
group velocity cannot propagate, and this fits the Rayleigh theory explaining
in another way why the Bragg mode cannot be excited. If so, what is wrong
with the numerical simulations shown in Fig. 3.9?
In fact, the Bragg mode is truly prohibited only in a purely lossless crystal.
Due to finite (whatever small) optical losses the group velocity in the photonic
crystal is not exactly equal to ∂ω/∂q because Bloch’s vector q depicted in
the horizontal axis of the Brillouin diagram is only the real part of the true
wave vector k which is complex. It has an imaginary part, responsible for
the wave decay. Therefore, in realistic photonic crystals the Bragg mode can
CHAPTER 3. PHOTONIC CRYSTALS 43

be excited. Usually, it is excited by an incident wave at the crystal boundary


and propagates to a certain distance from the interface. This distance is
called the Bragg decay length. This decay length depends on the optical
losses in the medium. In the aforementioned simulations Imε was taken very
small but nonzero (of the order of 10−8 ), and in this case the Bragg decay
length equals to several dozens of lattice periods.

Figure 3.10: Brillouin’s band diagrams for an opal (a), an inverse opal (b)
and an artificial diamond (c).

Modes which have vg = 0 inside the Brillouin zone (see the 2d passband
in Fig. 3.9 between M and Γ points) can exist, too. Such modes in solid-
state physics are called polaritons [2]. Similarly, in realistic photonic crystals
polaritons can be excited as well as the Bragg modes either by an embedded
dipole source or (usually) by a plane wave at the interface of the crystal.
Many polaritons are characterized by a decay length larger than the lattice
period.
Besides orthorhombic (3D) or rectangular (2D) lattices photonic crystals
may have complex-shaped unit cell. For example, a honeycomb lattice has a
hexagonal unit cell whose maximal size is 2a if the side of the cell has length
a. Bloch’s quasi-periodicity holds in this case as well, and at two √ opposite
sides
√ of the hexagonal cell the fields differ by the phase shift ql a 3, where
a 3 is the distance between these opposite sides and ql is the projection of
q onto the line connecting them.
CHAPTER 3. PHOTONIC CRYSTALS 44

For 3D photonic crystals the band dispersion diagrams correspond to


bulk Brillouin zones. The geometry of the unit cell and the Brillouin zone
for 3D crystals is a very difficult question [1, 2]. In the theory of 3D crystals
it is necessary to distinguish the lattice unit cell from the so-called lattice
primitive or Wiegner-Seitz cell introduced into crystallography by E. Wieg-
ner and F. Seitz in work [21]8 . Notice, however, that the concept of the
primitive lattice cell was introduced in the end of XIX-th century by G.
Voronoy (also transliterated as Voronoi). The primitive cell was suggested in
work [22] as a relevant concept for the theory of mathematical groups. There-
fore, in crystallography a primitive cell of a 3D lattice is sometimes called
the Voronoy cell. This name was suggested by E. Wiegner in his book [23].
However, the contribution of Wiegner and Seitz into the theory of 3D lattices
was so important that even in Russian literature the primitive cell is called
Wiegner-Seitz-Voronoy cell. This name is, however, too long, we prefer to
use the initial name primitive cell, introduced by Voronoy. The primitive cell
is obtained by drawing planes normal to the segments joining nearest lattice
nodes to a particular lattice point (called the cell center), through the mid-
points of such segments. Usually, the center of one of the lattice inclusions
is chosen as the primitive cell center. However, the primitive cell (as well as
the unit cell) stands any parallel shift – the cell shape and the cell problem
solution do not change.
For simple (also called orthorhombic) lattices (whose unit cell is a paral-
lelepiped with one of several complete inclusions) the primitive cell and the
unit cell do not differ. However, many 3D lattices are not simple. For exam-
ple, the lattice of the opal is a body-centered cubic lattice, a 3D analogue of a
honeycomb lattice. Its cubic unit cell includes besides one entire particle also
eighth parts of eight adjacent particles at every cell corner. Therefore, the
primitive cell for such the lattice is not cubic. It has 18 faces and is called
rhombohedron. The angle between adjacent edges of the rhombohedron is
equal 109.3◦ . The face-centered cubic lattice and diamond lattice are even
more complex. For complex lattices Wiegner and Seitz in their work [21]
proved that the Brillouin zone is reciprocal not to the unit cell, but to the
8
It s interesting that F. Seitz contributed also into the creation of transistors, especially
field transistors, where his contribution was principal. Unfortunately, he is more known
in the mass media not as one of the best solid-state physicists and electronic engineers
of XX-th century, but as an author of the doubtful theory of so-called global warming
and the driving force of so-called Kyoto treaties signed in XX-th century by all developed
countries except USA.
CHAPTER 3. PHOTONIC CRYSTALS 45

primitive one of the lattice. Fortunately, in this short course we cannot stay
on this interesting question and all we need to know is that the Brillouin
zone for complex 3D lattices does not repeat the unit cell and the amount of
the edge points on the Brillouin diagram can be substantial.
In Fig. 3.10 we show few typical band diagrams for some 3D photonic
crystals. The effect of the directional bandgap is seen in the diagram of the
opal. Between points Γ and X the light is propagating along the x-axis.
Between X and W qx is fixed (qx = π/a) and qy grows from zero, i.e. the
light is obliquely propagating with growing incidence angle in the (x − y)
plane. The central frequency of the 1st directional bandgap (between the
1st and 2d passbands) clearly increases versus the angle of incidence. For a
sea mouse hair this prohibited frequency corresponds to red light at point X
and to blue light at point W . Similarly, the directional bandgap effect holds
when the incidence angle changes in other incidence planes – between other
characteristic points of the diagram.

3.3 Governing equation of photonic crystals,


their scalability and numerical simulations
If the building block of a photonic crystal is a particle of isotropic medium
of permittivity ε, the photonic crystal is determined by a scalar periodic
function ε(r). Then, from Maxwell’s equations

∇ × H(r) + jωε(r)ε0 E(r) = 0,

∇ × E(r) − jωµ0 H(r) = 0,


we obtain the governing equation of a photonic crystal:

∇ × ε−1 (r)∇ × H(r) = k02 H(r).


 
(3.4)

For a simple lattice ε(x, y, z) = ε(x + ax , y + ay , z + az ) we have in accordance


to F. Bloch (see above):
 
−j (qx + 2πm x)−j qy + 2πn y −j (qz + 2πp z)
X
H= Hmnp e a
x ay a
z . (3.5)
mnp

It is hopeless to find analytically all coefficients Hmnp entering (3.5). For


a simplest geometry – a 1D photonic crystal of bilayers one managed to
CHAPTER 3. PHOTONIC CRYSTALS 46

find exactly the fundamental coefficients H000 and E000 in Rytov’s work [24].
However, we have already mentioned that the fundamental Bloch’s wave
dominates in a photonic crystal eigenmode only at low frequencies. Below
the 1st bandgap, where the approximation of the Bloch wave is especially
accurate (all terms with (m, p) 6= 0 can be neglected) the periodical struc-
ture is an effectively homogeneous medium and can be described by averaged
effective permittivity (in this lecture we do not explain how to calculate this
effective permittivity). At higher frequencies, other terms of the expansion
become dominating. Therefore, the attempts to analytically solve the gov-
erning equation (3.4) have no big practical importance. One has to solve the
problem numerically.
Equation (3.4) after substitution of (3.5) results in the infinite system of
equations for Hmnp which is truncated. The correct truncation procedure was
elaborated in the theory of crystals [1]. Alternatively, one may use the finite-
element method for Maxwell’s equations together with the quasi-periodicity
conditions [25] or use the FDTD approach [26]. To discuss these methods is
not our purpose. Here we concentrate on the scalability of photonic crystals
i.e. the invariance with respect to the multiplication of the unit cell and the
wavelength by the same arbitrary number s.
For the dispersion this scalability follows from the fact that both axes
in the dispersion diagram are dimensionless ωa/2πc and qa/2π. For the
electromagnetic field the scalability is not evident. Let us prove it. Denote
the radius-vector in the enlarged photonic crystal as R and in the original
one let it be r. Then ∇r = s∇R and we may rewrite (3.4) as
s2 ∇R × ε−1 (R)∇R × H(R) = k02 H(R).
 

Dividing both parts of this equation by s2 we obtain k0new = k0 /s i.e λnew =


sλ. This means that optical phenomena of nanostructured photonic crystals
can be modelled at radio frequencies where the unit cell is scaled by mil-
limeters or even centimeters. Scalability offers a great tool for experimental
study and optimization of photonic crystals9 .
Numerical simulations of the band diagram and isofrequencies of photonic
crystals can be done using reliable commercial softwares, such as HFSS,
9
The main restriction of this approach is the absence of solid low-loss media with
negative permittvity at microwave which would emulate metals in the optical range. Only
plasma which has negative permittivity at microwaves.However, it is difficult to shape
vessels with plasma and to make plasmas in them identical. Therefore, optical photonic
crystals with metal inclusions are experimentally studied as such.
CHAPTER 3. PHOTONIC CRYSTALS 47

CST Studio and COMSOL Multiphysics. These softwares comprise solvers


of the cell problem using Bloch’s quasi-periodicity conditions. For example,
in the HFSS solver the option for solving the rectangular or orthorhombic
cell problem is called Optemetric Analysis. Choosing this option one draws
the unit cell with its internal geometry, defines the material parameters of
its parts and selects the type of solution – eigenmode solution. Then one
has to suggest the phase shift of the eigenwave travelling across the unit cell,
determining in the solver its master and slave faces. Master face is one of
the cell sides illuminated by eigenwave one wants to study. Slave face is
opposite to the master face. For an orthorhombic lattice there are 3 pairs
of master-slave faces. Inserting the phase shifts δφx,y,z = qx,y,z ax,y,z for these
3 pairs and drawing a unit cell with predefined internal geometry you will
obtain the set of eigenfrequencies corresponding to given q. If there are
optical losses (i.e. the permittivity of the inclusion or the host medium is
complex), these eigenfrequencies will be complex numbers. Only real parts
of these complex numbers make sense and only if the imaginary parts are
sufficiently small. Really, a useful photonic crystal is sufficiently transparent
beyond its bandgaps. Otherwise there is no difference between the passbands
and bangaps (of course, photonic crystals are never prepared of very lossy
constituents). For a reasonably chosen material parameters eigenfrequencies
must be sufficiently close to real values, otherwise, the solution is incorrect.
If the solution is correct, imaginary parts of simulated eigenfrequencies are
not very relevant, where the set of real eigenfrequencies together with the
corresponding wave vectors is used for plotting the dispersion diagram. In
this course we do not discuss the possible sources of errors and do not concern
the convergence of these numerical solutions.
Bibliography

[1] J.D. Joannopoulos, R.D. Mead, and J.N. Winn, Photonic crystals: mold-
ing the flow of light, Princeton University Press, NJ, 1995

[2] K. Sakoda, Optical properties of photonic crystals, Springer-Verlag,


Berlin, 2005

[3] Special Volume on Electromagnetic Applications of Photonic Band Gap


Materials and Structures, Progress In Electromagnetic Research (PIER)
41 (2003)

[4] Focus Issue: Photonic Bandgap Calculations, Optics Express 8, issue 3


(2001)

[5] A.P. Vinogradov, A.V. Dorofeenko, S.G. Erokhin, M. Inoue, A.A.


Lisyansky, A.M. Merzlikin, A.B. Granovsky, Surface-state peculiarities
in 1D photonic crystal interfaces, Phys. Rev. B 74, 045128 (2006).

[6] D.P. Belozorov, M.K. Khodzitsky, S.I. Tarapov, Tamm states in mag-
netophotonic crystals and permittivity of the wire medium, J. Phys. D:
Applied Physics 42, 055003 (2009)

[7] C. Luo, S.G. Johnson and J.D. Joannopoulos, Negative refraction with-
out negative index in metallic photonic crystals, Optics Express 11,
746-754 (2003)

[8] E. Yablonovich, Inhibited spontaneous emission in solid-state physics


and electronics, Phys. Rev. Lett. 58, 2059-2062 (1987)

[9] S. John, Strong localization of photons in certain disordered dielectric


superlattices, Phys. Rev. Lett. 58, 2486-2489 (1987)

48
BIBLIOGRAPHY 49

[10] V.P. Bykov, Spontaneous emission in a periodic structure, Sov. Phys.


JETP 35, 269-273 (1972)

[11] V.P. Bykov, Spontaneous emission from a medium with a band spec-
trum, Sov. J. Quant. Electron 4, 861-871 (1975)

[12] P. Fratzl, J.W.C. Dunlop and R. Weinkamer, Materials design inspired


by Nature, RSC Publishing, Dorchester, UK, 2013

[13] Lord Rayleigh, On the reflection of light from a regularly stratified


medium, Proc. Roy. Soc. London 93, 565-577 (1917)

[14] W.H. Bragg and W.L. Bragg, The reflection of X-rays by crystals, Proc.
Roy. Soc. London 88, 428-438 (1913)

[15] C.R. Simovski, P.A. Belov, A.V. Atraschenko, and Y.S. Kivshar, Wire
metamaterials: Physics and applications, Advanced Materials 24, 4229-
4248 (2012)

[16] C. Jin, B. Cheng, B. Man, D. Zhang, S. Ban, B. Sun, L. Li, X. Zhang


and Z. Zhang, Two-dimensional metallodielectric photonic crystal with
a large band gap, Appl. Phys. Lett. 75, 1201-1205 (1999)

[17] A. Sommerfeld, Elektronentheorie der Metalle, Zeitschrift für Physik


47, 1-60 (1928)

[18] F. Bloch, Uber die Quantenmechanik der Electronen in Kristallgittern,


Zeitschrift für Physik 52, 555-600 (1928)

[19] L. Brillouin, Les électrons dans les métaux et le classement des on-
des de de-Broglie correspondantes, Comptes Rendus Hebdomadaires des
Séances de l’Académie des Sciences 191 292-299 (1930).

[20] H. Sagan, Boundary and eigenvalue problems in mathematical physics,


Dover Publications Inc., NY, 2d Ed. (1989)

[21] E. Wiegner and F. Seitz. On the constitution of metallic sodium, Phys.


Rev. 43, 804-821 (1933)

[22] G.F. Voronoi, Nouvelles applications des paramètres continus à la


théorie de formes quadratiques, Journal für die reine und angewandte
Mathematik 134, 198287 (1908)
BIBLIOGRAPHY 50

[23] E. Wiegner, Group theory and its application to the quantum mechanics
of atomic spectra, Academic Press, NY (1959)

[24] S.M. Rytov, Electromagnetic properties of a finely stratified medium,


Soviet Physics (JETP) 2, 466-475 (1956)

[25] J.B. Pendry, Calculating photonic band structure, J. Phys: Cond. Mat.
8, 1085-1108 (1996)

[26] U. Andersson, M. Qiu and Z. Zhang, Parallel-power computation for


photonic crystal devices, Methods and Applications of Analysis 13, 149-
156 (2006)

You might also like