Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

3. WAVEGUIDES

We move on in these notes to examine the possibility of guiding light in a desired direction by its
confinement in a designed structure. The first means that springs to mind is to use the
phenomenon of total internal reflection that was discovered in the previous section where it is
recalled, light sourced in a medium of higher refractive index, nS , will, at the boundary with a
dielectric of lower refractive index, nU , undergo total internal reflection if the angle of incidence
n 
at the interface is above the critical angle, C  sin1 U  . To do this a symmetric, dielectric
 nS 
slab waveguide is created with a high refractive index, dielectric slab, known as the guide,
bounded on two sides by dielectrics of equal and lower refractive index known as the cladding
dielectric.

The Symmetric Slab Waveguide

Physical Structure

nC, C
x = +d/2

nG, G z

x = -d/2
nC, C

The simple structure outlined above will act as a waveguide, if nG > nC, for waves
travelling at an angle wrt the normal to the interface greater that the critical angle. It is
however the case that not any and all angles greater than the critical angle will correspond
to a guided wave and the possible guided electromagnetic fields are limited as modes

60
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

(and superpositions of modes) of the guide. The following notes use the fact that the
boundary conditions at a dielectric interface require that the tangential E and H fields are
continuous across the boundaries involved in order to analyse this problem. For
simplicity we will look at the transverse electric, TE, mode with its polarisation in the y
direction and therefore completely tangential to the interface. With Ey there is an
orthogonal H field with x and z components whose z component will be the tangential
component.
Using Maxwells curl equation for E
H H
  E  0   0  j 0 H (3.1)
t t

where a time dependence for H of the form H = H0 exp(jt) has been assumed to obtain
the time derivative. We find Hz from the z component of the curl

   
y z x y

Ey Ez Ex Ey

1     1 
Hz   E y ( x, z )  E x ( x, z )   E y ( x, z ) (3.2)
j 0  x y  j 0 x

a) Field variation in the x direction within the cladding.

We have seen previously that due to the requirement that kz (=  ) matches at the two
interfaces, kx in regions 1 and 3 must be entirely imaginary for guided modes (   C ).

kCx   j  2  kC2   j kG
2
sin2   kC2   jk0 nG
2
sin2   nC
2
  j (3.3)

and the fields are therefore of the form A1exp(-x) in cladding region 1 and A1exp(x) in
cladding region 3. This is the evanescent field just outside the guide exponentially
decaying over a length scale represented by -1.

61
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

b) Field variation in the x direction within the guide.

Within the guide there are positive and negative travelling waves in the x direction giving
rise to standing waves described by cosinusoidal functions, A2cos(kGxx) for symmetric
modes (even parity) and A2sin(kGxx) for anti-symmetric modes (odd parity).

2
kGx  kG  2 (3.4)

c) Field variation in the propagation direction , z.

The fields will propagate as exp(jz) in the z direction everywhere and from now on we

write kz as , the propagation constant.

62
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Electric Fields in Guide

The electric fields may then be written, following the above prescriptions for the three
regions, as follows

A1exp(-x) x > d/2

cos(kGxx)
Ey(x,z) = A2 exp(jz) d/2 <x < -d/2 (3.5)
sin(kGxx)

A1exp(x) x < -d/2

NB the choice of a cosine or sine variation for the field within the guide represents
the possibility of even/odd parity (symmetric/anti-symmetric) modes respectively.

Magnetic Fields in Guide

1 
The tangential magnetic fields are found by using H z  E y ( x, z ) and may be
j x
written, following the above prescriptions for the three regions and taking the derivative
wrt x of the electric fields given above, as follows

-A1exp(-x) x > d/2

sin(kGxx)
1
Hz(x,z) =   kGx A2 exp(jz) d/2 < x < -d/2 (3.6)
j 0

cos(kGxx)

A1exp(x) x < -d/2

63
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Note that the even/odd parity electric fields of 3.5 are represented by the sine/cosine
variation in the magnetic fields in 3.6 after the differentiation wrt x.

Guidance Conditions

To obtain the guidance conditions for the guide, in order to find out which

values of  represent a valid guided mode for the particular guide

under consideration, both Ey and Hz must be matched at x = +d/2 and –d/2

Symmetric guided modes.

d
For symmetric E fields the field matching at x   gives
2

k d    d 
A2 cos Gx   A1 exp  (3.7a)
 2   2 

d
For symmetric H fields the field matching at  gives
2

k d    d 
 kGx A2 sin Gx   A1 exp  (3.7b)
 2   2 

Dividing equation 3.7b by 3.7a to eliminate the amplitudes we have

k d 
  k Gx tan Gx  (3.7c)
 2 

This is the guidance condition for symmetric modes of the guide in its basic form. We
shall simplify it in order to make it more tractable after obtaining the guidance condition
for anti-symmetric modes of the guide.

64
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Anti-symmetric guided modes

d
For anti-symmetric E fields matched at x  
2

k d    d 
A2 sin Gx   A1 exp  (3.8a)
 2   2 

d
For anti-symmetric H fields matched at x  
2

k d    d 
 kGx A2 cos Gx   A1 exp  (3.8b)
 2   2 

Dividing (3.8b) by (3.8a) we have

k d 
  kGx cot  Gx  (3.8c)
 2 

This is the guidance condition for anti-symmetric modes of the guide in its basic form.

We now seek to write these two conditions, 3.7c and 3.8c, in a simpler form in order that
a graphical representation of the conditions may be easily obtained. To do this we first
need to define normalised transverse wavevectors.
d
NB we have no need to apply the boundary condition at the  boundary as this
2
would bring no new information. This is because the guide is symmetric and we go
d d
from  to  by inverting the symmetric guide with no actual physical change
2 2
in the problem being considered. Anti-symmetric guides also exist where nC1 
nC3 and nG > nC1 , nC3. In that case both boundary conditions would be used. This
does not concern us here.

65
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Normalised Transverse Wavevectors, U, W and V and

Characteristic Equation of Slab Guide

Another way to represent the guidance conditions graphically is to rewrite them

d
multiplying both sides by as follows;
2

The condition for symmetric modes

k d 
  k Gx tan Gx  (3.7c)
 2 
becomes
d kGx d k d 
 tan Gx  (3.9a)
2 2  2 

And the condition for antisymmetric modes

 k d 
 kGx cot  Gx  (3.8c)
kGx  2 

becomes

d k d k d 
  Gx cot Gx  (3.9b)
2 2  2 

These new formulations of the guidance conditions lead us to define some new
parameters of the waveguide, namely the normalised transverse wavevectors.

66
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Normalised Transverse Wavevectors

With the new formulation for the guidance conditions equs 3.9, it becomes useful to
define normalised transverse wavevectors in cladding and guide regions as follows:

(i) By transverse we indicate that it is the wavevector transverse to the guide

direction, ie. kx with which we are concerned
(ii) We normalise them (make them dimensionless) by multiplying by a natural
d
length of the guide, , to give:
2

a) The normalised exterior transverse wavevector , W , defined as

d d  2 d
W  jkCx      kC2  (3.10a)
2 2  2

b) The normalised interior transverse wavevector, U defined as

d  2
  kG   2 
d
U  kGx (3.10b)
2  2

The guidance equations may then be written

W  U tanU (3.11a)

and

W  U cot U (3.11b)

We can square and add equations 3.11a and 3.11b to give

67
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

 
2 2
d  2 2 d 
U 2  W 2  kG
2
 kC2    (nG
2
 nC )k 0    V 2 (3.12)
2 2

where V is an extremely important parameter in the description of waveguides and fibres

known as the normalised frequency. Note that V is a mixture of the guided frequency
through k0 and the guide properties, nG , nC and d.

d  2
V    kG
2
  d 
 kC2  k 0   (nG
2
2 2
 nC )

The equation

U2 W 2 V 2 (3.13)

is known as the characteristic equation of the guide (or step index

fibre). It is an easy equation to examine graphically.
To examine the guidance conditions graphically we plot a graph of W vs U and on this
graph we plot
W  U tanU even modes

W  U cot U odd modes

along with the characteristic equation

W 2 U2 V 2 .

This is the equation of a circle of radius V shown schematically on the next page along
with the guidance conditions.

68
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

p=1 p=2 p=3 p=4

even odd even odd
UtanU
W V=5 (even mode)

V=4
-UcotU
(odd mode)
V = 3.35
V=3

U2 + W2 = V2
V=2

V=1

U
/2 2/2 3/2 4/2

Graphical solution of guidance conditions

Where the UtanU (or –UcotU curve) and W  V 2  U 2 curves (quarter circles) cross
one another the values of W and U represent those for an allowed mode at that frequency,
V, ie an allowed mode of the particular guide. Remember that V represents a particular
d
value of frequency for a given guide (defined through nC , nG and ).
2

Single Mode Operation.

From the graph we can see that as V is increased (the quarter circles have radius V), each

time it increases by another TE mode is allowed (there is an equivalent diagram for
2
TM modes) and therefore the number of TE modes propagating, NP will be given by
 
V   2V 
NP  MOD   1  MOD   1 . (3.14a)
    
 2 

69
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Including TM modes then we have a total of

4V
NP  2 (3.14b)

for a multimode guide.

Generally the spatial distribution of light across the guide will be describable as a
superposition of modes. It is frequently desirable to allow only one mode to propagate.
Notably, if we desire to transmit a light pulse , necessarily made up of many frequencies
or Fourier components, because phase and group velocities differ for different modes and
for different frequencies, we reduce the temporal spreading of the pulse by allowing only
one mode to propagate. From the graphical solution in the diagram, if V is less than /2
only the first even mode will propagate. Ie when the condition

2 2 d
V  k 0 nG  nC  /2 (3.15)
2

is satisfied only one mode will propagate. Thus

i) For a given frequency  = ck0 we will have single mode propagation when

c
d SM  (3.16)
2 2
 nG  nC

ii) and for a given thickness, d, will have single mode propagation when

c
CO  (3.17
2 2
d nG  nC

70
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Modal Behaviour
The crossing points of the two curves in the above figure represent allowed modes. If the
UtanU curve is crossed an even, symmetric, mode ( E  A cos kGx x  ) is represented and

an odd, anti-symmetric, mode ( E  A sinkGx x  ) if the –UcotU curve is crossed. The

modes describe the distribution of the electric field in the x direction across the guide
with exponentially decaying fields into the cladding.

E E

TE00 x TE10
z

E E

TE20 TE30

The above diagrams show the variation of Ey with x for the lowest four TE modes, two
even and two odd.
NB the mode is described by two suffices, TEmn , and the two suffices are to describe
the field variation in the x and the y direction. We have assumed the extent of the
guiding and cladding regions to be infinite in the y direction in order to simplify the
problem. An actual guide would have a more complicated structure in order to
contain the electromagnetic wave in the x and y dimensions. Specification of a mode
would then require the two mode numbers m and n to be given. To find the fields
and allowed modes for this problem is beyond the scope of the course.

We are now in a position to examine more closely the modal behaviour in two limits.

71
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

P=1 P=2

W
Vf

Wf
Vn

Wn
Un Uf
U

Focussing on just one of the modes using the graphical solution diagram, say p = 2, there
are two limits that may be examined in some detail;

i) Near to cut off where Vn is low such that the mode is close to cut off.
ii) Far from cut off where Vf is large and the mode is well guided.

Limit 1. Low Frequency (Near Cutoff)

Looking at the graphical solution we see that the pth mode has a value Un that lies
between;

p  1   UP p

(3.18)
2 2

72
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan


At the lower end of this range UP approaches p  1 , V  UP and W approaches 0. If
2
U becomes equal to or falls below this value then the mode is lost, ie it is not guided.

From the characteristic equation, U 2  W 2  V 2 , we see that as this cutoff limit V 

UP is approached then we also require that W  0 as noted noted from the graphical
solution.
Recalling the definition of W (the exterior normalised transverse wavevector)
d
W  0   0
2
The significance of this limit is realised when we recall that the evanescent fields extend
beyond the guide a characteristic distance -1 and we see that this field is extending
further for the pth mode as the cutoff condition is approached and   0 .
The first two even modes are illustrated below in this limit.

The first two even modes near to cutoff with extensive leakage into cladding region
From the graphical solution of the guidance equations it is also seen that as this limit is

approached and kGx (= 2 U ) is reduced, recalling the relation between kGx and
d
frequency;

kGx  nG k 0 cos   nG cos 
c

73
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

the frequency of the guided mode (or V) is also being reduced and there exists a
frequency at which the mode is lost, referred to as the cut-off frequency for that mode.
This cut-off frequency may be found as follows;

UP  VPCO  p  1 (3.19)
CO 2
and

VPCO   nG 2 
k 0CO  p  1
2 d
 nC (3.20)
  2 2

k 0CO  nG
2 2  CO 
 nC 2 2  p  1
  nG  nC   (3.21)
  c   d
or

CO  CO  p  1
c 1
(3.22)
2 2d 2
nG 2
 nC

Note that the p = 1 mode, the zeroth order symmetric mode has no
cutoff and will propagate at all frequencies down to zero.

74
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Limit 2. High Frequency (Far from Cutoff)

We now look at the other, high frequency limit for the pth mode.
 d 
U P   kGx P   p (3.23)
 2 2
As from the guidance conditions WP  U P tan U P and the tangent and therefore WP
tends asymptotically to infinity at this limit

d
WP   P  (3.24)
2

As  P   the evanescent field in the cladding becomes more and more closely
1
confined to the interface (  P  0)

The first two even modes at this limit (far from cutoff ) and tightly confined to guiding
region are shown below.

75
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Geometric Picture.
We can examine the high/low frequency limits within a geometric ray representation. The
pth mode propagating at a given frequency in a given guide (ie dimension d, nC and nG)
will have a unique propagation constant P and kGxP and it will travel at some angle P
such that

p
tan  p  (3.25)
kGx p

p

kG kGxP

p

The angle at which mode p propagates is frequency dependent and will be determined
from the guidance condition.

i) Low frequency (V) limit.

As the frequency (or V) is reduced toward cutoff,  approaches zero and from the
graphical solution we see that;
d 2 2 2
U P  kGx V  kGx  V  nG  nC k0
2 d

 
And as W 2   2  kC2  0 in this limit

P  kC  nC k 0

Therefore we have

76
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

P kC nC
tan P    (3.26)
(kGx )P 2
nG 2
 nC k0 2
nG 2
 nC

alternatively written (form the right angle!)

n
sin P  C (3.27)
nG

and P  C the critical angle.

This is as we expect of course i.e. that as cutoff is approached for a given mode and guide
by lowering the frequency the ray angle decreases until at cutoff it is equal to the critical
angle. Any further decrease and no guiding occurs.

ii) High frequency (V) limit.

Propagating in a given mode, p, and increasing frequency (or equivalently with the
graphical solutions in mind, V) we have seen that UP or kGxP tend to asymptotic limits

 d 
U P   kGx P   p as V is increased
 2 2

whereas  carries on increasing and approaches kG therefore as  is increased tanP


increases more rapidly and P approaches . ie at higher frequencies the mode is
2
travelling at increasingly grazing angles to the interface.

Dispersion relations (  vs )
A major interest in waveguides is of course their use in transmitting information in the
form of streams of light pulses. Of great importance in this application is the propensity
of the individual pulses to spread in time as they propagate, a tendency which it is
necessary to avoid in order that information is not degraded or lost altogether. The degree
to which this occurs is termed dispersion and will depend on several possible mechanisms
which we shall look at presently. For now it is sufficient to note that a pulse is made up of
a Fourier superposition of cosinusoidal waves and to understand the dispersion we need

77
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

to look at the phase and group velocities of the waves in any particular mode. The
variation with frequency of the velocity of the wave in any given mode is known as
waveguide dispersion. To find the phase and group velocities of light in any particular
mode it is necessary to know the relationship between  and , () for that mode. Each
mode will have a different dispersion relation, P(P). and in general to find this will
involve numerical or graphical solution. It is useful however look at the two limiting
situations again.

Limit 1. Low Frequency (Near Cutoff)

Near cut-off as is clear from the graphical solution, W (and )  0

 
2 2
d  d 
From the equation W 2   2     2  kC2   and with   0 near cutoff as we
2 2
saw earlier,

 2  kC2 (3.28)

and using kC  nC k 0 and k 0  

c
Giving the dispersion relation in the far from cutoff limit;
n  c
  kC  nC k 0  C   (3.29a)
c nC
The mode number, p doesn’t appear in the dispersion relation in this limit, thus, the
dispersion relation is the same for all modes and the phase velocity of the modes
c
approaches , in the limit far from cutoff ie. it tends to the velocity of light in the
nC
cladding.
In this limit the phase and group velocity are;
c  c
vP  vG  
nC k nC

It is easily understood why this should be the velocity of light in the cladding as the field
extends a long way into the cladding when   0 and a larger fraction of the modal
power is contained in the cladding regions.

78
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Limit 2. High Frequency (Far from Cutoff)

For sufficiently large frequency far from cut off U  p  although V can become ever
2

larger. This means that the characteristic equation U 2  W 2  V 2 becomes W 2  V 2

Given the definitions of W and V this approximation to the characteristic equation far
from cut off may be written as;

     
2 2 2
d  d  2 d 
  kC2  
2
   nC k 02  
2 2
 nG  nC  
2 2 2
or

 2  nG
2 2
k0    nG
c

c
  (3.29b)
nG

Again in this limit the phase and group velocity

c  c
vP  vG  
nG k nG
This is again as expected as at high frequencies    and consequently most of the
c
mode is confined to the guiding layer and will travel with a velocity .
nG
c
In between these two limits the ,  curve will lie between the two lines of slope
nG

. To obtain the precise form of  is difficult and beyond the scope of these
c
and
nC
lectures.

79
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

 c/nC p=3

2CO p=2

1CO

p=1

c/nG


The dispersion relation for the first three modes showing
the cutoff frequencies.
The dashed lines are straight lines with slope c/n2 and c/n1.

80
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Numerical Aperture.
Up until now we have assumed the guide region to be the source region ignoring one
small matter, how did the light get there? If light is to be guided within a waveguide
(dielectric slab or fibre) it has to be incident at the interface between the high index guide
and the low index cladding at an angle greater than the critical angle when it will undergo
total internal reflection.
This immediately presents a problem as all rays are equally good if drawn running
backwards (Fermats Principle). It is clear that there is no ray that will allow light from
outside the guide to enter the guide AND be incident at an angle greater than the critical
angle when it next arrives at the interface so how did a guided wave come to be in the
high index region in the first place?
There are at least three ways around the problem of coupling light to a guide from an
external source
(i) Light may be coupled from one guide or fibre to another using the evanescent
fields that penetrate into the cladding.
(ii) By using the evanescent fields associated with total internal reflection from
prisms.
(iii) By use of a diffraction grating etched on the surface of the slab guide.
The first (and most obvious?) technique however, and one which is applicable to
waveguide and fibre, is to focus light into the end of a guide/fibre. Indeed with laser
diodes which tend to have a divergent output (broad angular distribution) compared with
other lasers, a lens for focusing may not be necessary.

To choose an appropriate lens for direct focus coupling it is important to consider the
collecting efficiency of the guide/fibre. A quantity that is used to express this is the
numerical aperture (NA) of the fibre. Indeed other optical systems from monochromators
to cameras use the concept and the NA is related to the inverse of the f-number of a
camera. We are interested in the NA in the context of guides.
Consider the diagram below where an external ray of light in a medium of refractive
index n0 (usually air) is introduced into a guide such that it is subsequently guided at the
critical angle. The angle Max that the ray makes with the normal to the guide is the

81
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

largest angle compatible with guiding in the guide. Were it any larger then the internal
angle would be less than the critical angle. The numerical aperture is defined such that
the inverse sin of the NA is this maximum angle, sin-1(NA) = Max.

nC

C
n0 
Max
nG

nC

The NA can be found by application of Snells law at the two interfaces of the diagram:

n0sinMax = nGsin (3.30)

nGsinC = nC (3.31)

Further

n2
sin   cos  C  1  sin 2  C  1  C (3.32)
2
nG

Then using 3.32 in 3.30

n n2
sin  Max  G 1  C  NA (3.33)
n0 2
nG

82
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

Simplifying
2 2
nG  nC
NA  (3.34)
n0

Often the external medium is air and n0 = 1, in this case the numerical aperture of the
guide/fibre is given by ;
2 2
NA  nG  nC (3.35)

It should be noted that the dimensions of the guide/fibre do not affect the numerical
aperture.

Example.
For a highly collimated laser beam (parallel beam) with a beam diameter, D, of 2mm
what choice of focal length lens, F, may be appropriate for a guide with numerical
aperture 0.15?
The lens is required to produce a cone of light whose half angle, , is sin-10.15 = 8.60 (see
diagram below). At the focus of the laser beam a distance F from the lens (we have a
D
parallel beam into the lens) the sin of the half angle would be approximately . To
2F
obtain optimum matching to the fibre a half angle of 8.60 would be appropriate. In other
D D 2mm
words  NA  0.15 or F    6.66mm .
2F 2  0.15 0.3

Max
D

F
Lens

83
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

D
The system of laser beam plus lens can be assigned a numerical aperture and the
2F
numerical aperture of the lens system has been matched to that of the fibre.
A diode laser will emit a diverging beam with a divergence angle Div and it could thus
be assigned an NA = sin-1Div . If this does not match the NA of the fibre that it is to be
coupled to, then an intervening lens would be required in order to maximise the coupling.

84
Electromagnetic Waves & Optics: Lecture Notes ©Kevin Donovan

In summary for slab waveguides;

The field internal to the guide is cosinusoidal with solutions separable into
E  exp  jz cos k x x  Symmetric Modes

E  exp  jz  sink x x  Asymmetric Modes

The field external to the guide (in the cladding) is exponentially decaying
E  E Int exp  x 

The guidance conditions for achieving a guided mode are

W  U tan U Symmetric Modes
W  U cot U Asymmetric Modes

There is also a characteristic equation

W 2 U 2  V 2

Which used with the guidance conditions allows graphical solution of the guidance
problem.
U and W are the Internal Normalised Transverse Wavevector and the External
Normalised Transverse Wavevector respectively.

d d d d d
U  kGx 2
 kG 2 W  jk Cx     2  kC2
2 2 2 2 2
2 d
V  nG  nC2 k 0
2

The guide will become single mode when V    1.57

2
The guide can be characterised by a numerical aperture
2
NA  nG  nC2

85