154 5.

Reection and Transmission

         

5 E
H
=
1
1
1
1
E+
E
,
E+
E
=
1
2
1
1


E
H
(5.1.4)

Two useful quantities in interface problems are the wave impedance at z:

Reection and Transmission E(z)
Z(z)= (wave impedance) (5.1.5)
H(z)

and the reection coefcient at position z:

E (z)
(z)= (reection coefcient) (5.1.6)
E+ (z)

Using Eq. (5.1.3), we have:

5.1 Propagation Matrices
1 E
E (E H) Z
In this chapter, we consider uniform plane waves incident normally on material inter- = = 2 = H =
faces. Using the boundary conditions for the elds, we will relate the forward-backward E+ 1 E Z+
(E + H) +
elds on one side of the interface to those on the other side, expressing the relationship 2 H
in terms of a 22 matching matrix. Similarly, using Eq. (5.1.1) we nd:
If there are several interfaces, we will propagate our forward-backward elds from
one interface to the next with the help of a 22 propagation matrix. The combination of E
1+
a matching and a propagation matrix relating the elds across different interfaces will E E+ + E E+ 1+
Z= = = =
be referred to as a transfer or transition matrix. H 1 E 1
(E+ E ) 1
We begin by discussing propagation matrices. Consider an electric eld that is lin- E+
early polarized in the x-direction and propagating along the z-direction in a lossless
Thus, we have the relationships:
(homogeneous and isotropic) dielectric. Setting E(z)= x Ex (z)= x E(z) and H(z)=
y Hy (z)= y H(z), we have from Eq. (2.2.6): 1 + (z) Z(z)
Z(z)=  (z)= (5.1.7)
jkz jkz 1 (z) Z(z)+
E(z) = E0+ e + E0 e = E+ (z)+E (z)
1  1  (5.1.1) Using Eq. (5.1.2), we nd:
H(z) = E0+ ejkz E0 ejkz = E+ (z)E (z)

E (z) E0 ejkz
where the corresponding forward and backward electric elds at position z are: (z)= = = (0)e2jkz
E+ (z) E0+ ejkz
E+ (z)= E0+ ejkz where (0)= E0 /E0+ is the reection coefcient at z = 0. Thus,
(5.1.2)
E (z)= E0 ejkz
(z)= (0)e2jkz (propagation of ) (5.1.8)
We can also express the elds E (z) in terms of E(z), H(z). Adding and subtracting
the two equations (5.1.1), we nd: Applying (5.1.7) at z and z = 0, we have:
1 
Z(z) Z(0) 2jkz
E+ (z)= E(z)+H(z) = (z)= (0)e2jkz = e
2 Z(z)+ Z(0)+
(5.1.3)
1 
E (z)= E(z)H(z)
2 This may be solved for Z(z) in terms of Z(0), giving after some algebra:
Eqs.(5.1.1) and (5.1.3) can also be written in the convenient matrix forms:
Z(0)j tan kz
Z(z)= (propagation of Z) (5.1.9)
jZ(0)tan kz
5.1. Propagation Matrices 155 156 5. Reection and Transmission

The reason for introducing so many eld quantities is that the three quantities which gives after some algebra:
{E+ (z), E (z), (z)} have simple propagation properties, whereas {E(z), H(z), Z(z)}     
do not. On the other hand, {E(z), H(z), Z(z)} match simply across interfaces, whereas E1 cos kl j sin kl E2
= (propagation matrix) (5.1.13)
{E+ (z), E (z), (z)} do not. H1 j1 sin kl cos kl H2
Eqs. (5.1.1) and (5.1.2) relate the eld quantities at location z to the quantities at
Writing = 0 /n, where n is the refractive index of the propagation medium,
z = 0. In matching problems, it proves more convenient to be able to relate these
Eq. (5.1.13) can written in following form, which is useful in analyzing multilayer struc-
quantities at two arbitrary locations.
tures and is common in the thin-lm literature [632,634,638,649]:
Fig. 5.1.1 depicts the quantities {E(z), H(z), E+ (z), E (z), Z(z), (z)} at the two
    
locations z1 and z2 separated by a distance l = z2 z1 . Using Eq. (5.1.2), we have for E1 cos jn1 0 sin E2
= (propagation matrix) (5.1.14)
the forward eld at these two positions: H1 jn1
0 sin cos H2

E2+ = E0+ ejkz2 , E1+ = E0+ ejkz1 = E0+ ejk(z2 l) = ejkl E2+ where is the propagation phase constant, = kl = k0 nl = 2(nl)/0 , and nl the
optical length. Eqs. (5.1.13) and (5.1.5) imply for the propagation of the wave impedance:

E2
cos kl + j sin kl
E1 E2 cos kl + jH2 sin kl H2
Z1 = = =
H1 jE2 1 sin kl + H2 cos kl E2
cos kl + j sin kl
H2

which gives:

Z2 cos kl + j sin kl
Z1 = (impedance propagation) (5.1.15)
cos kl + jZ2 sin kl

Fig. 5.1.1 Field quantities propagated between two positions in space.

It can also be written in the form:

And similarly, E1 = ejkl E2 . Thus, Z2 + j tan kl

Z1 = (impedance propagation) (5.1.16)
E1+ = e jkl
E2 + , jkl
E1 = e E2 (5.1.10) + jZ2 tan kl

and in matrix form: A useful way of expressing Z1 is in terms of the reection coefcient 2 . Using (5.1.7)
     and (5.1.12), we have:
E1+ ejkl 0 E2+
= (propagation matrix) (5.1.11)
E1 0 ejkl E2 1 + 1 1 + 2 e2jkl
Z1 = = or,
1 1 1 2 e2jkl
We will refer to this as the propagation matrix for the forward and backward elds.
It follows that the reection coefcients will be related by:
1 + 2 e2jkl
Z1 = (5.1.17)
E1 E2 e jkl 1 2 e2jkl
1 = = = 2 e2jkl , or,
E1+ E2+ ejkl
We mention nally two special propagation cases: the half-wavelength and the quarter-
2jkl wavelength cases. When the propagation distance is l = /2, or any integral multiple
1 = 2 e (reection coefcient propagation) (5.1.12)
thereof, the wave impedance and reection coefcient remain unchanged. Indeed, we
Using the matrix relationships (5.1.4) and (5.1.11), we may also express the total have in this case kl = 2l/ = 2/2 = and 2kl = 2. It follows from Eq. (5.1.12)
electric and magnetic elds E1 , H1 at position z1 in terms of E2 , H2 at position z2 : that 1 = 2 and hence Z1 = Z2 .
         If on the other hand l = /4, or any odd integral multiple thereof, then kl = 2/4 =
E1 1 E1+ 1 1 1 ejkl 0 E2+ /2 and 2kl = . The reection coefcient changes sign and the wave impedance
= 1 = 1 1 1 jkl
H1 E1 0 e E2 inverts:
    
1 1 1 ejkl 0 1 E2 1 + 1 1 2 1 2
=
1
1
0 ejkl 1 H2 1 = 2 e2jkl = 2 ej = 2 Z1 = = = =
2 1 1 1 + 2 Z2 / Z2
5.2. Matching Matrices 157 158 5. Reection and Transmission

Thus, we have in the two cases: and inversely:

    

E+ 1 1  E+
l= Z1 = Z2 , 1 = 2  =  (matching matrix) (5.2.4)
2 E  1 E
(5.1.18)
2 where {, } and { ,  } are the elementary reection and transmission coefcients
l= Z1 = , 1 = 2
4 Z2 from the left and from the right of the interface, dened in terms of ,  as follows:

 2
5.2 Matching Matrices = , = (5.2.5)
 +  +
Next, we discuss the matching conditions across dielectric interfaces. We consider a
planar interface (taken to be the xy-plane at some location z) separating two dielec-  2
 = ,  = (5.2.6)
tric/conducting media with (possibly complex-valued) characteristic impedances ,  , +  + 
as shown in Fig. 5.2.1.
Writing = 0 /n and  = 0 /n , we have in terms of the refractive indices:

n n 2n
= , =
n + n n + n
(5.2.7)
n n 2n
 = ,  =
n + n n + n

These are also called the Fresnel coefcients. We note various useful relationships:

= 1 + ,  = ,  = 1 +  = 1 ,  = 1 2 (5.2.8)

Fig. 5.2.1 Fields across an interface.

In summary, the total electric and magnetic elds E, H match simply across the
interface, whereas the forward/backward elds E are related by the matching matrices
Because the normally incident elds are tangential to the interface plane, the bound- of Eqs. (5.2.3) and (5.2.4). An immediate consequence of Eq. (5.2.1) is that the wave
ary conditions require that the total electric and magnetic elds be continuous across impedance is continuous across the interface:
the two sides of the interface: E E
Z= =  = Z
H H
E = E
(continuity across interface) (5.2.1) On the other hand, the corresponding reection coefcients = E /E+ and  =
H = H  
E /E+ match in a more complicated way. Using Eq. (5.1.7) and the continuity of the
In terms of the forward and backward electric elds, Eq. (5.2.1) reads: wave impedance, we have:

  1+ 1 + 
E+ + E = E+ + E = Z = Z = 
1 1 
1  1  
 (5.2.2)
E+ E =  E+ E which can be solved to get:

Eq. (5.2.2) may be written in a matrix form relating the elds E on the left of the +   +
= and  =

interface to the elds E on the right: 1 +  1 + 
     The same relationship follows also from Eq. (5.2.3):

E+ 1 1 E+
=  (matching matrix) (5.2.3)
E 1 E 1 E

(E+ 
+ E ) +
E 
E+ + 
The arrows in this gure indicate the directions of propagation, not the direction of the eldsthe eld
= = =  =
E+ 1   E 1 + 
vectors are perpendicular to the propagation directions and parallel to the interface plane. (E + E ) 1+
+ E+
5.2. Matching Matrices 159 160 5. Reection and Transmission

To summarize, we have the matching conditions for Z and : 5.3 Reected and Transmitted Power

+   + For waves propagating in the z-direction, the time-averaged Poynting vector has only a
Z = Z  =   = (5.2.9) z-component:
1 +  1 + 
1   1
P = Re x E y H = z Re(EH )
Two special cases, illustrated in Fig. 5.2.1, are when there is only an incident wave 2 2

on the interface from the left, so that E = 0, and when the incident wave is only from A direct consequence of the continuity equations (5.2.1) is that the Poynting vector
the right, so that E+ = 0. In the rst case, we have  = E  
/E+ = 0, which implies is conserved across the interface. Indeed, we have:
    
Z = (1 + )/(1 )= . The matching conditions give then:
1 1
+  P= Re(EH )= Re(E H )= P (5.3.1)
Z = Z =  , = = 2 2
1 + 
In particular, consider the case of a wave incident from a lossless dielectric onto a
The matching matrix (5.2.3) implies in this case:
       lossy dielectric  . Then, the conservation equation (5.3.1) reads in terms of the forward
  
E+ 1 1 E+ 1 E+ and backward elds (assuming E = 0):
= = 
E 1 0 E+
1    1   2

Expressing the reected and transmitted elds E , E+ in terms of the incident eld E+ , P= |E+ |2 |E |2 = Re |E+ | = P
2 2
we have:
The left hand-side is the difference of the incident and the reected power and rep-
E = E+ resents the amount of power transmitted into the lossy dielectric per unit area. We saw
 (left-incident elds) (5.2.10) in Sec. 2.6 that this power is completely dissipated into heat inside the lossy dielectric
E+ = E+
(assuming it is innite to the right.) Using Eqs. (5.2.10), we nd:
This justies the terms reection and transmission coefcients for and . In the
1   1 
right-incident case, the condition E+ = 0 implies for Eq. (5.2.4): P= |E+ |2 1 ||2 )= Re |E+ |2 ||2 (5.3.2)
       2 2
  
E+ 1 1 0 1 E
 = = This equality requires that:
E   1 E  E
These can be rewritten in the form: 1 1
(1 ||2 )= Re  ||2 (5.3.3)


E+ =  E
 This can be proved using the denitions (5.2.5). Indeed, we have:
  (right-incident elds) (5.2.11)
E = E 
1 1 ||2 1 ||2
  = Re = =
which relates the reected and transmitted elds E+ , E to the incident eld E . In this  1+  |1 + |2 ||2
case = E /E+ = and the third of Eqs. (5.2.9) gives  = E 
/E+ 
= 1/ , which is
consistent with Eq. (5.2.11). which is equivalent to Eq. (5.3.3), if is lossless (i.e., real.) Dening the incident, re-
 ected, and transmitted powers by
When there are incident elds from both sides, that is, E+ , E , we may invoke the
linearity of Maxwells equations and add the two right-hand sides of Eqs. (5.2.10) and
 1
(5.2.11) to obtain the outgoing elds E+ , E in terms of the incident ones: Pin = |E+ |2
2

E+ = E+ +  E
 1 1
(5.2.12) Pref = |E |2 = |E+ |2 ||2 = Pin ||2
E = E+ +  E
 2 2
 1   2  1  
This gives the scattering matrix relating the outgoing elds to the incoming ones: Ptr = Re |E+ | = Re |E+ |2 ||2 = Pin Re  ||2
2 2
    

E+  E+ Then, Eq. (5.3.2) reads Ptr = Pin Pref . The power reection and transmission
= (scattering matrix) (5.2.13)
E  
E coefcients, also known as the reectance and transmittance, give the percentage of the
Using the relationships Eq. (5.2.8), it is easily veried that Eq. (5.2.13) is equivalent incident power that gets reected and transmitted:
to the matching matrix equations (5.2.3) and (5.2.4).
5.3. Reected and Transmitted Power 161 162 5. Reection and Transmission


Solution:

For a good conductor, we have 0 /  1. It follows from Eq. (2.8.4) that Rs /0 =
Pref Ptr   n  2 0 /2  1. From Eq. (2.8.2), the conductors characteristic impedance is c = Rs (1 +
= ||2 , = 1 ||2 = Re  ||2 = Re || (5.3.4)
Pin Pin n j). Thus, the quantity c /0 = (1 + j)Rs /0 is also small. The reection and transmission
If both dielectrics are lossless, then , are real-valued. In this case, if there are coefcients , can be expressed to rst-order in the quantity c /0 as follows:
incident waves from both sides of the interface, it is straightforward to show that the
2 c 2c 2 c
net power moving towards the z-direction is the same at either side of the interface: = , = 1 1 +
c + 0 0 0
1   1   2  2

P= |E+ |2 |E |2 = |E+ | |E | = P (5.3.5) Similarly, the power transmission coefcient can be approximated as
2 2
This follows from the matrix identity satised by the matching matrix of Eq. (5.2.3): 2 Re(c ) 4R s
1 ||2 = 1 | 1|2 = 1 1 ||2 + 2 Re() 2 Re()= 2 =
      0 0
1 1 1 0 1 1 0
=  (5.3.6)
2 1 0 1 1 0 1 where we neglected ||2 as it is second order in c /0 . For copper at 1 GHz, we have
 
0 /2 = 2.19105 , which gives Rs = 0 0 /2 = 3772.19105 = 0.0082 . It
If , are real, then we have with the help of this identity and Eq. (5.2.3):
follows that 1 ||2 = 4Rs /0 = 8.76105 .
  
1   1   1 0 E+ This represents only a small power loss of 8.76103 percent and the sheet acts as very
P= |E+ |2 |E |2 = E ,E
2 2 + 0 1 E good mirror at microwave frequencies.
     On the other hand, at optical frequencies, e.g., f = 600 THz corresponding to green
1     1 1 1 0 1 
E+
= E , E light with = 500 nm, the exact equations (2.6.5) yield the value for the character-
2 +
1 0 1 1 
E istic impedance of the sheet c = 6.3924 + 6.3888i and the reection coefcient
   = 0.9661 + 0.0328i. The corresponding power loss is 1 ||2 = 0.065, or 6.5 percent.
1     1 0 
E+ 1   2  2

= E , E = |E+ | |E | = P Thus, metallic mirrors are fairly lossy at optical frequencies. 

2  + 0 1 
E 2
Example 5.3.3: A uniform plane wave of frequency f is normally incident from air onto a thick
Example 5.3.1: Glasses have a refractive index of the order of n = 1.5 and dielectric constant
conductor with conductivity , and = 0 , = 0 . Determine the reected and trans-
= n2 0 = 2.25 0 . Calculate the percentages of reected and transmitted powers for
mitted electric and magnetic elds to rst-order in c /0 and in the limit of a perfect
visible light incident on a planar glass interface from air.
conductor (c = 0).
Solution: The characteristic impedance of glass will be = 0 /n. Therefore, the reection and
Solution: Using the approximations for and of the previous example and Eq. (5.2.10), we
transmission coefcients can be expressed directly in terms of n, as follows:
have for the reected, transmitted, and total electric elds at the interface:

0 n1 1 1n 2 
= = 1 = , =1+= 2 c
+ 0 n +1 1+n 1+n E = E+ = 1 + E+
0
For n = 1.5, we nd = 0.2 and = 0.8. It follows that the power reection and 2c

transmission coefcients will be E+ = E+ = E+
0
2 c 
||2 = 0.04, 1 ||2 = 0.96 E = E+ + E = E+ = E+ = E
0
That is, 4% of the incident power is reected and 96% transmitted. 

For a perfect conductor, we have and c /0 0. The corresponding total tangen-
Example 5.3.2: A uniform plane wave of frequency f is normally incident from air onto a thick tial electric eld becomes zero E = E = 0, and = 1, = 0. For the magnetic elds, we
conducting sheet with conductivity , and = 0 , = 0 . Show that the proportion need to develop similar rst-order approximations. The incident magnetic eld intensity
of power transmitted into the conductor (and then dissipated into heat) is given approxi- is H+ = E+ /0 . The reected eld becomes to rst order:
mately by 
1 1 2c

H = E = E+ = H+ = 1 H+
Ptr 4Rs 8 0 0 0 0
= =
Pin 0
Similarly, the transmitted eld is
Calculate this quantity for f = 1 GHz and copper = 5.8107 Siemens/m.
5.4. Single Dielectric Slab 163 164 5. Reection and Transmission

Let 1 , 2 be the elementary reection coefcients from the left sides of the two
 interfaces, and let 1 , 2 be the corresponding transmission coefcients:
 1  1 0 0 2c 20 c
H+ = E+ = E+ = H+ = H+ = H+ 2 1 H+
c c c c c + 0 c + 0 0 1 a b 1
1 = , 2 = , 1 = 1 + 1 , 2 = 1 + 2 (5.4.1)
1 + a b + 1
The total tangential eld at the interface will be:
To determine the reection coefcient 1 into medium a , we apply Eq. (5.2.9) to

c relate 1 to the reection coefcient 1 at the right-side of the rst interface. Then, we
 
H = H+ + H = 2 1 H+ = H+ =H
0 propagate to the left of the second interface with Eq. (5.1.12) to get:

In the perfect conductor limit, we nd H = H = 2H+ . As we saw in Sec. 2.6, the elds just 1 + 1 1 + 2 e2jk1 l1
1 =  = (5.4.2)

inside the conductor, E+ 
, H+ , will attenuate while they propagate. Assuming the interface 1 + 1 1 1 + 1 2 e2jk1 l1
is at z = 0, we have:
At the second interface, we apply Eq. (5.2.9) again to relate 2 to 2 . Because there
  z jz   z jz are no backward-moving waves in medium b , we have 2 = 0. Thus,
E+ (z)= E+ e e , H+ (z)= H+ e e
 2 + 2
where = = (1 j)/, and is the skin depth = /2. We saw in Sec. 2.6 that 2 = = 2
1 + 2 2
the effective surface current is equal in magnitude to the magnetic eld at z = 0, that is,

Js = H+ . Because of the boundary condition H = H = H+ 
, we obtain the result Js = H, We nally nd for 1 :
or vectorially, Js = H z = n H, where n = z is the outward normal to the conductor.
This result provides a justication of the boundary condition Js = n H at an interface 1 + 2 e2jk1 l1
with a perfect conductor. 
 1 = (5.4.3)
1 + 1 2 e2jk1 l1

This expression can be thought of as function of frequency. Assuming a lossless

5.4 Single Dielectric Slab medium 1 , we have 2k1 l1 = (2l1 /c1 )= T, where T = 2l1 /c1 = 2(n1 l1 )/c0 is the
two-way travel time delay through medium 1 . Thus, we can write:
Multiple interface problems can be handled in a straightforward way with the help of
the matching and propagation matrices. For example, Fig. 5.4.1 shows a two-interface
1 + 2 ejT
problem with a dielectric slab 1 separating the semi-innite media a and b . 1 ()= (5.4.4)
1 + 1 2 ejT

This can also be expressed as a z-transform. Denoting the two-way travel time delay
in the z-domain by z1 = ejT = e2jk1 l1 , we may rewrite Eq. (5.4.4) as the rst-order
digital lter transfer function:

1 + 2 z1
1 (z)= (5.4.5)
1 + 1 2 z1

An alternative way to derive Eq. (5.4.3) is working with wave impedances, which
are continuous across interfaces. The wave impedance at interface-2 is Z2 = Z2 , but
Z2 = b because there is no backward wave in medium b . Thus, Z2 = b . Using the
propagation equation for impedances, we nd:

Z2 + j1 tan k1 l1 b + j1 tan k1 l1
Fig. 5.4.1 Single dielectric slab. Z1 = Z1 = 1 = 1
1 + jZ2 tan k1 l1 1 + jb tan k1 l1

Let l1 be the width of the slab, k1 = /c1 the propagation wavenumber, and 1 = Inserting this into 1 = (Z1 a )/(Z1 + a ) gives Eq. (5.4.3). Working with wave
2/k1 the corresponding wavelength within the slab. We have 1 = 0 /n1 , where 0 is impedances is always more convenient if the interfaces are positioned at half- or quarter-
the free-space wavelength and n1 the refractive index of the slab. We assume the incident wavelength spacings.
eld is from the left medium a , and thus, in medium b there is only a forward wave.
5.4. Single Dielectric Slab 165 166 5. Reection and Transmission

If we wish to determine the overall transmission response into medium b , that is, slabs. These equations were simplied in the single-slab case because we assumed that
the quantity T = E2 + /E1+ , then we must work with the matrix formulation. Starting at there was a uniform medium to the right and that there were no backward-moving waves.
the left interface and successively applying the matching and propagation matrices, we For lossless media, energy conservation states that the energy ux into medium 1
obtain: must equal the energy ux out of it. It is equivalent to the following relationship between
         and T, which can be proved using Eq. (5.4.6):
E1+ 1 1 1 E1 + 1 1 1 ejk1 l1 0 E2+
= =
E1 1 1 1 E1 1 1 1 0 ejk1 l1 E2 1   1
1 |1 |2 = |T|2 (5.4.10)
      a b
1 1 1 ejk1 l1 0 1 1 2 E2 +
= Thus, if we call |1 |2 the reectance of the slab, representing the fraction of the
1 1 1 0 ejk1 l1 2 2 1 0
incident power that gets reected back into medium a , then the quantity
where we set E2 = 0 by assumption. Multiplying the matrix factors out, we obtain:
a nb
1 |1 |2 = |T|2 = |T|2 (5.4.11)
ejk1 l1   b na
E1+ = 1 + 1 2 e2jk1 l1 E2 +
1 2
will be the transmittance of the slab, representing the fraction of the incident power that
ejk1 l1   gets transmitted through into the right medium b . The presence of the factors a , b
E1 = 1 + 2 e2jk1 l1 E2 +
1 2 can be can be understood as follows:

These may be solved for the reection and transmission responses: 1

|E |2
Ptransmitted 2b 2+ a
E1 1 + 2 e2jk1 l1 = = |T|2
1 = = Pincident 1 b
|E1+ |2
E1 + 1 + 1 2 e2jk1 l1 2a
(5.4.6)
E2 + 1 2 ejk1 l1
T= = 5.5 Reectionless Slab
E1 + 1 + 1 2 e2jk1 l1

The transmission response has an overall delay factor of ejk1 l1 = ejT/2 , repre- The zeros of the transfer function (5.4.5) correspond to a reectionless interface. Such
senting the one-way travel time delay through medium 1 . zeros can be realized exactly only in two special cases, that is, for slabs that have either
For convenience, we summarize the match-and-propagate equations relating the eld half-wavelength or quarter-wavelength thickness. It is evident from Eq. (5.4.5) that a
quantities at the left of interface-1 to those at the left of interface-2. The forward and zero will occur if 1 + 2 z1 = 0, which gives the condition:
backward electric elds are related by the transfer matrix:
      2
z = e2jk1 l1 = (5.5.1)
E1+ 1 1 1 ejk1 l1 0 E2+ 1
= jk1 l1
E1 1 1 1 0 e E2 Because the right-hand side is real-valued and the left-hand side has unit magnitude,
     (5.4.7)
this condition can be satised only in the following two cases:
jk1 l1 jk1 l1
E1+ 1 e 1 e E2+
=
E1 1 1 ejk1 l1 ejk1 l1 E2 z = e2jk1 l1 = 1, 2 = 1 , (half-wavelength thickness)
The reection responses are related by Eq. (5.4.2): z=e 2jk1 l1
= 1, 2 = 1 , (quarter-wavelength thickness)
2jk1 l1
1 + 2 e The rst case requires that 2k1 l1 be an integral multiple of 2, that is, 2k1 l1 = 2m,
1 = (5.4.8)
1 + 1 2 e2jk1 l1 where m is an integer. This gives the half-wavelength condition l1 = m1 /2, where 1
The total electric and magnetic elds at the two interfaces are continuous across the is the wavelength in medium-1. In addition, the condition 2 = 1 requires that:
interfaces and are related by Eq. (5.1.13): b 1 a 1
     = 2 = 1 =  a = b
b + 1 a + 1
E1 cos k1 l1 j1 sin k1 l1 E2
= (5.4.9)
H1 j 1
1 sin k1 l1 cos k1 l1 H2 that is, the media to the left and right of the slab must be the same. The second pos-
Eqs. (5.4.7)(5.4.9) are valid in general, regardless of what is to the right of the second sibility requires e2jk1 l1 = 1, or that 2k1 l1 be an odd multiple of , that is, 2k1 l1 =
interface. There could be a semi-innite uniform medium or any combination of multiple
5.5. Reectionless Slab 167 168 5. Reection and Transmission

(2m + 1), which translates into the quarter-wavelength condition l1 = (2m + 1)1 /4. These expressions are periodic in l1 with period 1 /2, and periodic in with period
Furthermore, the condition 2 = 1 requires: 0 = 2/T. In DSP language, the slab acts as a digital lter with sampling frequency
0 . The maximum reectivity occurs at z = 1 and z = 1 for the half- and quarter-
b 1 1 a
= 2 = 1 =  21 = a b wavelength cases. The maximum squared responses are in either case:
b + 1 1 + a
421
To summarize, a reectionless slab, 1 = 0, can be realized only in the two cases: |1 |2max =
(1 + 21 )2
1
half-wave: l1 = m , 1 arbitrary, a = b Fig. 5.5.1 shows the magnitude responses for the three values of the reection co-
2 (5.5.2)
1 efcient: |1 | = 0.9, 0.7, and 0.5. The closer 1 is to unity, the narrower are the reec-
quarter-wave: l1 = (2m + 1) , 1 = a b , a , b arbitrary tionless notches.
4

An equivalent way of stating these conditions is to say that the optical length of
the slab must be a half or quarter of the free-space wavelength 0 . Indeed, if n1 is the
refractive index of the slab, then its optical length is n1 l1 , and in the half-wavelength
case we have n1 l1 = n1 m1 /2 = m0 /2, where we used 1 = 0 /n1 . Similarly, we have
n1 l1 = (2m + 1)0 /4 in the quarter-wavelength case. In terms of the refractive indices,
Eq. (5.5.2) reads:

0
half-wave: n1 l1 = m , n1 arbitrary, na = nb
2 (5.5.3)
0
quarter-wave: n1 l1 = (2m + 1) , n1 = na nb , na , nb arbitrary
4

The reectionless matching condition can also be derived by working with wave
impedances. For half-wavelength spacing, we have from Eq. (5.1.18) Z1 = Z2 = b . The Fig. 5.5.1 Reection responses |()|2 . (a) |1 | = 0.9, (b) |1 | = 0.7, (c) |1 | = 0.5.
condition 1 = 0 requires Z1 = a , thus, matching occurs if a = b . Similarly, for the
quarter-wavelength case, we have Z1 = 21 /Z2 = 21 /b = a . It is evident from these gures that for the same value of 1 , the half- and quarter-
We emphasize that the reectionless response 1 = 0 is obtained only at certain slab wavelength cases have the same notch widths. A standard measure for the width is
widths (half- or quarter-wavelength), or equivalently, at certain operating frequencies. the 3-dB width, which for the half-wavelength case is twice the 3-dB frequency 3 , that
These operating frequencies correspond to T = 2m, or, T = (2m + 1), that is, is, = 23 , as shown in Fig. 5.5.1 for the case |1 | = 0.5. The frequency 3 is
= 2m/T = m0 , or, = (2m + 1)0 /2, where we dened 0 = 2/T. determined by the 3-dB half-power condition:
The dependence on l1 or can be seen from Eq. (5.4.5). For the half-wavelength
case, we substitute 2 = 1 and for the quarter-wavelength case, 2 = 1 . Then, the 1
|1 (3 )|2 = |1 |2max
reection transfer functions become: 2
or, equivalently:
1 (1 z1 )
1 (z) = , (half-wave) 221 (1 cos 3 T) 1 421
1 21 z1 =
(5.5.4) 1 221 cos 3 T + 41 2 (1 + 21 )2
1 (1 + z1 )
1 (z) = , (quarter-wave)
Solving for the quantity cos 3 T = cos(T/2), we nd:
1 + 21 z1

where z = e2jk1 l1 = ejT . The magnitude-square responses then take the form:  T  221  T  1 21
cos =  tan = (5.5.6)
2 1+ 41 4 1 + 21
 
221 1 cos(2k1 l1 ) 221 (1 cos T) If 21 is very near unity, then 1 21 and become small, and we may use the
|1 |2 = 4 = , (half-wave)
2
1 21 cos(2k1 l1 )+1 1 221 cos T + 41 approximation tan x x to get:
  (5.5.5)
221 1 + cos(2k1 l1 ) 221 (1 + cos T) T 1 21 1 21
|1 |2 = 2 4 = , (quarter-wave)
1 + 21 cos(2k1 l1 )+1 1 + 221 cos T + 41 4 1+ 21 2
5.5. Reectionless Slab 169 170 5. Reection and Transmission

which gives the approximation: Example 5.5.1: Determine the reection coefcients of half- and quarter-wave slabs that do not
necessarily satisfy the impedance conditions of Eq. (5.5.2).
T = 2(1 21 ) (5.5.7)
Solution: The reection response is given in general by Eq. (5.4.6). For the half-wavelength case,
This is a standard approximation for digital lters relating the 3-dB width of a pole we have e2jk1 l1 = 1 and we obtain:
peak to the radius of the pole [49]. For any desired value of the bandwidth , Eq. (5.5.6)
1 a b 1
or (5.5.7) may be thought of as a design condition that determines 1 . +
1 + 2 1 + a b + 1 b a na n b
Fig. 5.5.2 shows the corresponding transmittances 1 |1 ()|2 of the slabs. The 1 = = 1 a b 1 = b + a = na + nb
1 + 1 2 1+
transmission response acts as a periodic bandpass lter. This is the simplest exam- 1 + a b + 1
ple of a so-called Fabry-Perot interference lter or Fabry-Perot resonator. Such lters
nd application in the spectroscopic analysis of materials. We discuss them further in This is the same as if the slab were absent. For this reason, half-wavelength slabs are
Chap. 6. sometimes referred to as absentee layers. Similarly, in the quarter-wavelength case, we
have e2jk1 l1 = 1 and nd:

1 2 21 a b na nb n21
1 = = =
1 1 2 21 + a b na nb + n21

The slab becomes reectionless if the conditions (5.5.2) are satised. 



Example 5.5.2: Antireection Coating. Determine the refractive index of a quarter-wave antire-
ection coating on a glass substrate with index 1.5.

Solution: From Eq. (5.5.3), we have with na = 1 and nb = 1.5:

n1 = na nb = 1.5 = 1.22

The closest refractive index that can be obtained is that of cryolite (Na3 AlF6 ) with n1 =
Fig. 5.5.2 Transmittance of half- and quarter-wavelength dielectric slab. 1.35 and magnesium uoride (MgF2 ) with n1 = 1.38. Magnesium uoride is usually pre-
ferred because of its durability. Such a slab will have a reection coefcient as given by
Using Eq. (5.5.5), we may express the frequency response of the half-wavelength the previous example:
transmittance lter in the following equivalent forms:
1 2 2 a b na nb n21 1.5 1.382
1 = = 12 = 2 = = 0.118
(1 21 )2 1 1 1 2 1 + a b na n b + n 1 1.5 + 1.382
1 |1 ()|2 = = (5.5.8)
1 221 cos T + 41 1 + F sin2 (T/2)
with reectance ||2 = 0.014, or 1.4 percent. This is to be compared to the 4 percent
where the F is called the nesse in the Fabry-Perot context and is dened by:
reectance of uncoated glass that we determined in Example 5.3.1.
421 Fig. 5.5.3 shows the reectance |()|2 as a function of the free-space wavelength . The
F=
(1 21 )2 reectance remains less than one or two percent in the two cases, over almost the entire
visible spectrum.
The nesse is a measure of the peak width, with larger values of F corresponding
The slabs were designed to have quarter-wavelength thickness at 0 = 550 nm, that is, the
to narrower peaks. The connection of F to the 3-dB width (5.5.6) is easily found to be:
optical length was n1 l1 = 0 /4, resulting in l1 = 112.71 nm and 99.64 nm in the two cases
of n1 = 1.22 and n1 = 1.38. Such extremely thin dielectric lms are fabricated by means
 T  1 21 1
tan = =  (5.5.9) of a thermal evaporation process [632,634].
4 1 + 21 1+F
The MATLAB code used to generate this example was as follows:
Quarter-wavelength slabs may be used to design anti-reection coatings for lenses,
so that all incident light on a lens gets through. Half-wavelength slabs, which require that n = [1, 1.22, 1.50]; L = 1/4; refractive indices and optical length
the medium be the same on either side of the slab, may be used in designing radar domes lambda = linspace(400,700,101) / 550; visible spectrum wavelengths
(radomes) protecting microwave antennas, so that the radiated signal from the antenna Gamma1 = multidiel(n, L, lambda); reection response of slab

goes through the radome wall without getting reected back towards the antenna.
5.5. Reectionless Slab 171 172 5. Reection and Transmission

Antireflection Coating on Glass where we used 1 = (1 n)/(1 + n). This explains why glass windows do not exhibit a
5
frequency-selective behavior as predicted by Eq. (5.5.5). For n = 1.5, we nd 1 |1 |2 =
nglass = 1.50 0.9231, that is, 92.31% of the incident light is transmitted through the plate.
4

| 1 ()|2 (percent)
The same expressions for the average reectance and transmittance can be obtained by
n1 = 1.22
3 n1 = 1.38 summing incoherently all the multiple reections within the slab, that is, summing the
uncoated glass multiple reections of power instead of eld amplitudes. The timing diagram for such
multiple reections is shown in Fig. 5.6.1.
2
Indeed, if we denote by pr = 21 and pt = 1 pr = 1 21 , the power reection and trans-
1 mission coefcients, then the rst reection of power will be pr . The power transmitted
through the left interface will be pt and through the second interface p2t (assuming the
0
same medium to the right.) The reected power at the second interface will be pt pr and
400 450 500 550 600 650 700
will come back and transmit through the left interface giving p2t pr .
(nm)
Similarly, after a second round trip, the reected power will be p2t p3r , while the transmitted
Fig. 5.5.3 Reectance over the visible spectrum. power to the right of the second interface will be p2t p2r , and so on. Summing up all the
reected powers to the left and those transmitted to the right, we nd:

The syntax and use of the function multidiel is discussed in Sec. 6.1. The dependence p2t pr 2pr
|1 |2 = pr + p2t pr + p2t p3r + pt2 p5r + = pr + =
of on comes through the quantity k1 l1 = 2(n1 l1 )/. Since n1 l1 = 0 /4, we have 1 p2r 1 + pr
k1 l1 = 0.50 /. 

p2t 1 pr
1 |1 |2 = p2t + p2t p2r + p2t p4r + = =
Example 5.5.3: Thick Glasses. Interference phenomena, such as those arising from the mul- 1 p2r 1 + pr
tiple reections within a slab, are not observed if the slabs are thick (compared to the
where we used pt = 1 pr . These are equivalent to Eqs. (5.5.10). 

wavelength.) For example, typical glass windows seem perfectly transparent.
If one had a glass plate of thickness, say, of l = 1.5 mm and index n = 1.5, it would have Example 5.5.4: Radomes. A radome protecting a microwave transmitter has = 4 0 and is
optical length nl = 1.51.5 = 2.25 mm = 225104 nm. At an operating wavelength designed as a half-wavelength reectionless slab at the operating frequency of 10 GHz.
of 0 = 450 nm, the glass plate would act as a half-wave transparent slab with nl = Determine its thickness.
104 (0 /2), that is, 104 half-wavelengths long.
Next, suppose that the operating frequency is 1% off its nominal value of 10 GHz. Calculate
Such plate would be very difcult to construct as it would require that l be built with the percentage of reected power back towards the transmitting antenna.
an accuracy of a few percent of 0 /2. For example, assuming n(l)= 0.01(0 /2), the
Determine the operating bandwidth as that frequency interval about the 10 GHz operating
plate should be constructed with an accuracy of one part in a million: l/l = nl/(nl)=
frequency within which the reected power remains at least 30 dB below the incident
0.01/104 = 106 . (That is why thin lms are constructed by a carefully controlled evapo-
power.
ration process.)
More realistically, a typical glass plate can be constructed with an accuracy of one part in a Solution: The free-space wavelength is 0 = c0 /f0 = 30 GHz cm/10 GHz = 3 cm. The refractive
thousand, l/l = 103 , which would mean that within the manufacturing uncertainty l, index of the slab is n = 2 and the wavelength inside it, 1 = 0 /n = 3/2 = 1.5 cm. Thus,
there would still be ten half-wavelengths, nl = 103 (nl)= 10(0 /2). the slab thickness will be the half-wavelength l1 = 1 /2 = 0.75 cm, or any other integral
multiple of this.
The overall power reection response will be obtained by averaging |1 |2 over several 0 /2
cycles, such as the above ten. Because of periodicity, the average of |1 |2 over several cycles Assume now that the operating frequency is = 0 + , where 0 = 2f0 = 2/T.
is the same as the average over one cycle, that is, Denoting = /0 , we can write = 0 (1 + ). The numerical value of is very
small, = 1% = 0.01. Therefore, we can do a rst-order calculation in . The reection
 0 coefcient 1 and reection response are:
1
|1 |2 = |1 ()|2 d
0 0
0 0.5 1 1 1 (1 z1 ) 1 (1 ejT )
1 = = = , 1 ()= 2 1 =
where 0 = 2/T and T is the two-way travel-time delay. Using either of the two expres- + 0 0.5 + 1 3 1 1 z 1 21 ejT
sions in Eq. (5.5.5), this integral can be done exactly resulting in the average reectance
and transmittance: where we used = 0 /n = 0 /2. Noting that T = 0 T(1 + )= 2(1 + ), we can
expand the delay exponential to rst-order in :
221 1 21 2n
|1 |2 = , 1 |1 |2 = = (5.5.10)
1+ 21 1 + 21 n2 + 1 z1 = ejT = e2j(1+) = e2j e2j = e2j 1 2j
5.5. Reectionless Slab 173 174 5. Reection and Transmission

Thus, the reection response becomes to rst-order in : 5.6 Time-Domain Reection Response
 
1 1 (1 2j) 1 2j 1 2j We conclude our discussion of the single slab by trying to understand its behavior in
1 =
1 21 (1 2j) 1 21 + 21 2j 1 21 the time domain. The z-domain reection transfer function of Eq. (5.4.5) incorporates
where we replaced the denominator by its zeroth-order approximation because the numer- the effect of all multiple reections that are set up within the slab as the wave bounces
ator is already rst-order in . It follows that the power reection response will be: back and forth at the left and right interfaces. Expanding Eq. (5.4.5) in a partial fraction
expansion and then in power series in z1 gives:
21 (2)2
|1 |2 = 1 + 2 z1 1 1 (1 21 )
(1 21 )2
1 (z)=
=
= 1 + (1 21 )(1 )n1 n2 zn
1 + 1 2 z 1 1 1 1 + 1 2 z 1
n=1
Evaluating this expression for = 0.01 and 1 = 1/3, we nd ||2 = 0.00049, or
0.049 percent of the incident power gets reected. Next, we nd the frequency about Using the reection coefcient from the right of the rst interface, 1 = 1 , and the
0 at which the reected power is A = 30 dB below the incident power. Writing again, transmission coefcients 1 = 1 + 1 and 1 = 1 + 1 = 1 1 , we have 1 1 = 1 21 .
= 0 + = 0 (1 + ) and assuming is small, we have the condition: Then, the above power series can be written as a function of frequency in the form:
21 (2)2 Pre 1 21


|1 |2 = = = 10A/10 = 10A/20 1 ()= 1 + 1 1 (1 )n1 n2 zn = 1 + 1 1 (1 )n1 n2 ejnT
(1 21 )2 Pinc 2|1 |
n=1 n=1
Evaluating this expression, we nd = 0.0134, or = 0.01340 . The bandwidth will
where we set z1 = ejT . It follows that the time-domain reection impulse response,
be twice that, = 2 = 0.02680 , or in Hz, f = 0.0268f0 = 268 MHz. 

that is, the inverse Fourier transform of 1 (), will be the sum of discrete impulses:
Example 5.5.5: Because of manufacturing imperfections, suppose that the actual constructed


thickness of the above radome is 1% off the desired half-wavelength thickness. Determine
the percentage of reected power in this case.
1 (t)= 1 (t)+ 1 1 (1 )n1 n2 (t nT) (5.6.1)
n=1
Solution: This is essentially the same as the previous example. Indeed, the quantity = T = This is the response of the slab to a forward-moving impulse striking the left inter-
2k1 l1 = 2l1 /c1 can change either because of or because of l1 . A simultaneous in- face at t = 0, that is, the response to the input E1+ (t)= (t). The rst term 1 (t) is the
nitesimal change (about the nominal value 0 = 0 T = 2) will give: impulse immediately reected at t = 0 with the reection coefcient 1 . The remaining
l1 terms represent the multiple reections within the slab. Fig. 5.6.1 is a timing diagram
= 2()l1 /c1 + 20 (l1 )/c1 = = + that traces the reected and transmitted impulses at the rst and second interfaces.
0 0 l1

In the previous example, we varied while keeping l1 constant. Here, we vary l1 , while
keeping constant, so that = l1 /l1 . Thus, we have = 0 = 2. The correspond-
ing delay factor becomes approximately z1 = ej = ej(2+) = 1 j = 1 2j.
The resulting expression for the power reection response is identical to the above and its
numerical value is the same if = 0.01. 


Example 5.5.6: Because of weather conditions, suppose that the characteristic impedance of
the medium outside the above radome is 1% off the impedance inside. Calculate the per-
centage of reected power in this case.

Solution: Suppose that the outside impedance changes to b = 0 + . The wave impedance
at the outer interface will be Z2 = b = 0 + . Because the slab length is still a half-
wavelength, the wave impedance at the inner interface will be Z1 = Z2 = 0 + . It
follows that the reection response will be:

Z1 0 0 + 0
1 = = =
Z1 + 0 0 + + 0 20 + 2 0
Fig. 5.6.1 Multiple reections building up the reection and transmission responses.
where we replaced the denominator by its zeroth-order approximation in . Evaluating
at /0 = 1% = 0.01, we nd 1 = 0.005, which leads to a reected power of |1 |2 = The input pulse (t) gets transmitted to the inside of the left interface and picks up
2.5105 , or, 0.0025 percent. 
 a transmission coefcient factor 1 . In T/2 seconds this pulse strikes the right interface
5.6. Time-Domain Reection Response 175 176 5. Reection and Transmission

and causes a reected wave whose amplitude is changed by the reection coefcient 2 For a causal waveform E1+ (t), the summation over n will be nite, such that at each
into 1 2 . time t 0 only the terms that have t nT 0 will be present. In a similar fashion, we
Thus, the pulse 1 2 (t T/2) gets reected backwards and will arrive at the left nd for the overall transmitted response into medium b :
interface T/2 seconds later, that is, at time t = T. A proportion 1 of it will be transmit- 

ted through to the left, and a proportion 1 will be re-reected towards the right. Thus,
E2 + (t)= T(t )E1+ (t t )dt = 1 2 (1 )n n2 E1+ (t nT T/2) (5.6.5)
at time t = T, the transmitted pulse into the left medium will be 1 1 2 (t T), and n=0
the re- reected pulse 1 1 2 (t T).
The re-reected pulse will travel forward to the right interface, arriving there at time We will use similar techniques later on to determine the transient responses of trans-
t = 3T/2 getting reected backwards picking up a factor 2 . This will arrive at the left mission lines.
at time t = 2T. The part transmitted to the left will be now 1 1 1 22 (t 2T), and
the part re-reected to the right 1 1 2 22 (t 2T). And so on, after the nth round trip,
the pulse transmitted to the left will be 1 1 (1 )n1 n
5.7 Two Dielectric Slabs
2 (t nT). The sum of all the
reected pulses will be 1 (t) of Eq. (5.6.1).
Next, we consider more than two interfaces. As we mentioned in the previous section,
In a similar way, we can derive the overall transmission response to the right. It is
Eqs. (5.4.7)(5.4.9) are general and can be applied to all successive interfaces. Fig. 5.7.1
seen in the gure that the transmitted pulse at time t = nT+(T/2) will be 1 2 (1 )n n 2.
shows three interfaces separating four media. The overall reection response can be
Thus, the overall transmission impulse response will be:
calculated by successive application of Eq. (5.4.8):


T(t)= 1 2 (1 )n n2 (t nT T/2) 1 + 2 e2jk1 l1 2 + 3 e2jk2 l2
1 = , 2 =
n=0 1 + 1 2 e2jk1 l1 1 + 2 3 e2jk2 l2

It follows that its Fourier transform will be:

T()= 1 2 (1 )n n2 ejnT ejT/2
n=0

which sums up to Eq. (5.4.6):

1 2 ejT/2 1 2 ejT/2
T()= = (5.6.2)
1 1 2 ejT 1 + 1 2 ejT
For an incident eld E1+ (t) with arbitrary time dependence, the overall reection
response of the slab is obtained by convolving the impulse response 1 (t) with E1+ (t).
This follows from the linear superposition of the reection responses of all the frequency
components of E1+ (t), that is,
 
d d Fig. 5.7.1 Two dielectric slabs.
E1 (t)= 1 ()E1+ ()ejt , where E1+ (t)= E1+ ()ejt
2 2
If there is no backward-moving wave in the right-most medium, then 3 = 0, which
Then, the convolution theorem of Fourier transforms implies that:
implies 3 = 3 . Substituting 2 into 1 and denoting z1 = e2jk1 l1 , z2 = e2jk2 l2 , we
 
d eventually nd:
E1 (t)= 1 ()E1+ ()ejt = 1 (t )E1+ (t t )dt (5.6.3)
2
1 + 2 z 1 1 1 1
1 + 1 2 3 z2 + 3 z1 z2
Inserting (5.6.1), we nd that the reected wave arises from the multiple reections 1 = (5.7.1)
1 + 1 2 z1 + 2 3 z2 + 1 3 z11 z
1 1
2
1
of E1+ (t) as it travels and bounces back and forth between the two interfaces:
The reection response 1 can alternatively be determined from the knowledge of

the wave impedance Z1 = E1 /H1 at interface-1:
E1 (t)= 1 E1+ (t)+ 1 1 (1 )n1 2n E1+ (t nT) (5.6.4)
n=1 Z1 a
1 =
Z1 + a
5.7. Two Dielectric Slabs 177 178 5. Reection and Transmission

The elds E1 , H1 are obtained by successively applying Eq. (5.4.9): The frequency dependence of Eq. (5.7.1) arises through the factors z1 , z2 , which can
     be written in the forms: z1 = ejT1 and z2 = ejT2 , where T1 = 2l1 /c1 and T2 = 2l2 /c2
E1 cos k1 l1 j1 sin k1 l1 E2
= are the two-way travel time delays through the two slabs.
H1 j1
1 sin k1 l1 cos k1 l1 H2
A case of particular interest arises when the slabs are designed to have the equal
   
cos k1 l1 j1 sin k1 l1 cos k2 l2 j2 sin k2 l2 E3 travel-time delays so that T1 = T2 T. Then, dening a common variable z = z1 =
= z2 = ejT , we can write the reection response as a second-order digital lter transfer
j 1
1 sin k1 l1 cos k1 l1 j1
2 sin k2 l2 cos k2 l2 H3
function:
But at interface-3, E3 = E3 = E3 + and H3 = Z31 E3 = 1 
b E3+ , because Z3 = b .
Therefore, we can obtain the elds E1 , H1 by the matrix multiplication: 1 + 2 (1 + 1 3 )z1 + 3 z2
1 (z)= (5.7.2)
      1 + 2 (1 + 3 )z1 + 1 3 z2
E1 cos k1 l1 j1 sin k1 l1 cos k2 l2 j2 sin k2 l2 1
= E3 + In the next chapter, we discuss further the properties of such higher-order reection
H1 j1
1 sin k1 l1 cos k1 l1 j21 sin k2 l2 cos k2 l2
b
1
transfer functions arising from multilayer dielectric slabs.
Because Z1 is the ratio of E1 and H1 , the factor E3 + cancels out and can be set equal
to unity.
5.8 Reection by a Moving Boundary
Example 5.7.1: Determine 1 if both slabs are quarter-wavelength slabs. Repeat if both slabs
are half-wavelength and when one is half- and the other quarter-wavelength. Reection and transmission by moving boundaries, such as reection from a moving
mirror, introduce Doppler shifts in the frequencies of the reected and transmitted
Solution: Because l1 = 1 /4 and l2 = 2 /4, we have 2k1 l1 = 2k2 l2 = , and it follows that
waves. Here, we look at the problem of normal incidence on a dielectric interface that
z1 = z2 = 1. Then, Eq. (5.7.1) becomes:
is moving with constant velocity v perpendicularly to the interface, that is, along the
1 2 1 2 3 + 3 z-direction as shown in Fig. 5.8.1. Additional examples may be found in [474492]. The
1 =
1 1 2 2 3 + 1 3 case of oblique incidence is discussed in Sec. 7.12.

A simpler approach is to work with wave impedances. Using Z3 = b , we have:

21 2 2 2
Z1 = = 2 1 = 12 Z3 = 12 b
Z2 2 /Z3 2 2

Inserting this into 1 = (Z1 a )/(Z1 + a ), we obtain:

21 b 22 a
1 =
21 b + 22 a

The two expressions for 1 are equivalent. The input impedance Z1 can also be obtained
by matrix multiplication. Because k1 l1 = k2 l2 = /2, we have cos k1 l1 = 0 and sin k1 l1 = 1
and the propagation matrices for E1 , H1 take the simplied form:
        Fig. 5.8.1 Reection and transmission at a moving boundary.
E1 0 j1 0 j2 1 1 21
= E3 + = E3 +
H1 j
1
1
0 j21 0
b
1
2 11
b
1

The dielectric is assumed to be non-magnetic and lossless with permittivity . The

The ratio E1 /H1 gives the same answer for Z1 as above. When both slabs are half-wavelength, left medium is free space 0 . The electric eld is assumed to be in the x-direction and
the impedances propagate unchanged: Z1 = Z2 = Z3 , but Z3 = b . thus, the magnetic eld will be in the y-direction. We consider two coordinate frames, the
If 1 is half- and 2 quarter-wavelength, then, Z1 = Z2 = 22 /Z3 = 22 /b . And, if the xed frame S with coordinates {t, x, y, z}, and the moving frame S with {t , x , y , z }.
quarter-wavelength is rst and the half-wavelength second, Z1 = 21 /Z2 = 21 /Z3 = 21 /b . The two sets of coordinates are related by the Lorentz transformation equations (K.1)
The corresponding reection coefcient 1 is in the three cases: of Appendix K.
We are interested in determining the Doppler-shifted frequencies of the reected and
b a 22 a b 21 a b transmitted waves, as well as the reection and transmission coefcients as measured
1 = , 1 = , 1 =
b + a 22 + a b 21 + a b in the xed frame S.
These expressions can also be derived by Eq. (5.7.1), or by the matrix method. 

5.8. Reection by a Moving Boundary 179 180 5. Reection and Transmission

The procedure for solving this type of problemoriginally suggested by Einstein The phase velocities of the incident, reected, and transmitted waves are:
in his 1905 special relativity paper [474]is to solve the reection and transmission
r t 1 + n
problem in the moving frame S with respect to which the boundary is at rest, and vi = = c, vr = = c, vt = =c (5.8.6)
then transform the results back to the xed frame S using the Lorentz transformation ki kr kt n+
properties of the elds. In the xed frame S, the elds to the left and right of the These can also be derived by applying Einsteins velocity addition theorem of Eq. (K.8).
interface will have the forms: For example, we have for the transmitted wave:

Ex = Ei ej(tki z) + Er ej(r t+kr z) Ex = Et ej(t tkt z)
vd + v c/n + v 1 + n
left
Hy = Hi ej(tki z) Hr ej(r t+kr z)
right
Hy = Ht ej(t tkt z)
(5.8.1) vt = = =c
1 + vd v/c2 1 + (c/n)v/c2 n+

where , r , t and ki , kr , kt are the frequencies and wavenumbers of the incident, where vd = c/n is the phase velocity within the dielectric at rest. To rst-order in
reected, and transmitted waves measured in S. Because of Lorentz invariance, the = v/c, the phase velocity within the moving dielectric becomes:
propagation phases remain unchanged in the frames S and S , that is,  
1 + n c 1
vt = c +v 1 2
i = t ki z =  t ki z = i n+ n n
r = r t + kr z =  t + kr z = r (5.8.2) The second term is known as the Fresnel drag. The quantity nt = (n+)/(1 +n)
t = t t kt z =  t kt z = t may be thought of as the effective refractive index of the moving dielectric as measured
in the xed system S.
In the frame S where the dielectric is at rest, all three frequencies are the same Next, we derive the reection and transmission coefcients. In the rest-frame S of
and set equal to  . This is a consequence of the usual tangential boundary conditions the dielectric, the elds have the usual forms derived earlier in Sections 5.1 and 5.2:
applied to the interface at rest. Note that r can be written as r = r t (kr )z   j  
 j  j
implying that the reected wave is propagating in the negative z-direction. In the rest
Ex = Ei e i + e r
Ex = Ei e t
frame S of the boundary, the wavenumbers are: left
1   j  right
1  (5.8.7)
Hy = E e i ejr Hy = Ei ejt
0 i
  
ki = , kr = kt =  0 = n
, (5.8.3)
c c c where
 0 0 1n 2
= , = = , =1+=
where c is the speed of light in vacuum and n = / 0 is the refractive index of the n + 0 1+n 1+n
dielectric at rest. The frequencies and wavenumbers in the xed frame S are related
The primed elds can be transformed to the xed frame S using the inverse of the
to those in S by applying the Lorentz transformation of Eq. (K.14) to the frequency-
Lorentz transformation equations (K.31), that is,
wavenumber four-vectors (/c, 0, 0, ki ), (r /c, 0, 0, kr ), and (t /c, 0, 0, kt ):
Ex = (Ex + cBy )= (Ex + 0 Hy )
= ( + cki )=  (1 + ) (5.8.8)
Hy = (Hy + cDx )= (Hy + c Ex )

ki = (ki +  )= (1 + )
c c where we replaced By = 0 Hy , c0 = 0 , and Dx = Ex (of course, = 0 in the left
  
r = + c(kr ) =  (1 ) medium). Using the invariance of the propagation phases, we nd for the elds at the
  (5.8.4) left side of the interface:
kr = (kr + )= (1 )
c c    
t = ( + ckt )=  (1 + n) Ex = Ei (eji +ejr )+Ei (eji ejr ) = Ei (1 +)eji +(1 )ejr (5.8.9)
  Similarly, for the right side of the interface we use the property 0 / = n to get:
kt = (kt + )= (n + )
c c
 
 Ex = Ei ejt + nEi ejt = Ei (1 + n)ejt (5.8.10)
where = v/c and = 1/ 1 2 . Eliminating the primed quantities, we obtain the
Doppler-shifted frequencies of the reected and transmitted waves:
Comparing these with Eq. (5.8.1), we nd the incident, reected, and transmitted
electric eld amplitudes:
1 1 + n
r = , t = (5.8.5)
1+ 1+ Ei = Ei (1 + ) , Er = Ei (1 ) , Et = Ei (1 + n) (5.8.11)
5.9. Problems 181 182 5. Reection and Transmission

from which we obtain the reection and transmission coefcients in the xed frame S: Moreover, using these expressions show and interpret the relationship:

1   1
Er 1 Et 1 + n 1 ||2 = |T|2
= , = (5.8.12) a b
Ei 1+ Ei 1+
5.5 A 1-GHz plane wave is incident normally onto a thick copper plate ( = 5.8107 S/m.) Can
The case of a perfect mirror is also covered by these expressions by setting = 1 the plate be considered to be a good conductor at this frequency? Calculate the percentage
and = 0. Eq. (5.8.5) is widely used in Doppler radar applications. Typically, the of the incident power that enters the plate. Calculate the attenuation coefcient within the
boundary (the target) is moving at non-relativistic speeds so that = v/c  1. In such conductor and express it in units of dB/m. What is the penetration depth in mm?
case, the rst-order approximation of (5.8.5) is adequate: 5.6 With the help of Fig. 5.5.1, argue that the 3-dB width is related to the 3-dB frequency
3 by = 23 and = 0 23 , in the cases of half- and quarter-wavelength slabs.
 v f v
fr f (1 2)= f 1 2 = 2 (5.8.13) Then, show that 3 and are given by:
c f c
 
221 T 1 21
where f = fr f is the Doppler shift. The negative sign means that fr < f if the target cos 3 T = , tan =
1+ 41 4 1 + 21
is receding away from the source of the wave, and fr > f if it is approaching the source.
5.7 A berglass ( = 4 0 ) radome protecting a microwave antenna is designed as a half-wavelength
reectionless slab at the operating frequency of 12 GHz.
As we mentioned in Sec. 2.12, if the source of the wave is moving with velocity va and
the target with velocity vb (with respect to a common xed frame, such as the ground), a. Determine three possible thicknesses (in cm) for this radome.
then one must use the relative velocity v = vb va in the above expression:
b. Determine the 15-dB and 30-dB bandwidths in GHz about the 12 GHz operating fre-
quency , dened as the widths over which the reected power is 15 or 30 dB below the
f fr f va vb incident power.
= =2 (5.8.14)
f f c
5.8 A 5 GHz wave is normally incident from air onto a dielectric slab of thickness of 1 cm and
refractive index of 1.5, as shown below. The medium to the right of the slab has an index of
5.9 Problems 2.25.

a. Write an analytical expression of the reectance |(f )|2 as a function of frequency

5.1 Fill in the details of the equivalence between Eq. (5.2.2) and (5.2.3), that is,
and sketch it versus f over the interval 0 f 15 GHz. What is the value of the

E+ + E = E+ 
+ E      reectance at 5 GHz?

E+ 1 1 E+
1  1  
  =  b. Next, the 1-cm slab is moved to the left by a distance of 3 cm, creating an air-gap
E+ E =  E + E E 1 E
between it and the rightmost dielectric. Repeat all the questions of part (a).

5.2 Fill in the details of the equivalences stated in Eq. (5.2.9), that is, c. Repeat part (a), if the slab thickness is 2 cm.

+   +
Z = Z  =   =
1 +  1 + 

Show that if there is no left-incident eld from the right, then = , and if there is no
right-incident eld from the left, then,  = 1/ . Explain the asymmetry of the two cases.
5.3 Let , be the reection and transmission coefcients from the left side of an interface and
let  ,  be those from the right, as dened in Eq. (5.2.5). One of the two media may be
lossy, and therefore, its characteristic impedance and hence , may be complex-valued.
5.9 A single-frequency plane wave is incident obliquely from air onto a planar interface with
Show and interpret the relationships:
a medium of permittivity = 2 0 , as shown below. The incident wave has the following
 2 phasor form:
1 ||2 = Re || = Re(  )


5.4 Show that the reection and transmission responses of the single dielectric slab of Fig. 5.4.1  
x + z
are given by Eq. (5.4.6), that is, E(z)= + j y ejk(zx)/ 2 (5.9.1)
2
1 + 2 e2jk1 l1 E 1 2 ejk1 l1
= , T = 2+ =
1 + 1 2 e2jk1 l1 E1+ 1 + 1 2 e2jk1 l1
5.9. Problems 183 184 5. Reection and Transmission

a. Determine the angle of incidence in degrees and decide which of the two dashed lines 5.12 Consider the slab of the previous problem. The tangential electric eld has the following
in the gure represents the incident wave. Moreover, determine the angle of refraction form in the three regions z 0, 0 z d, and z d:
 in degrees and indicate the refracted waves direction on the gure below.

jk0 z
+ ejk0 z , z0
b. Write an expression for the reected wave that is similar to Eq. (5.9.1), but also includes e
if
jkz
E(z)= Ae + Bejkz , if 0zd
the dependence on the TE and TM Fresnel reection coefcients (please evaluate these


Tejk0 (zd) ,
coefcients numerically.) Similarly, give an expression for the transmitted wave. if zd
c. Determine the polarization type (circular, elliptic, left, right, linear, etc.) of the incident
where k0 and k were dened in the previous problem.
wave and of the reected wave.
a. What are the corresponding expressions for the magnetic eld H(z)?
5.10 A uniform plane wave is incident normally on a planar interface, as shown below. The
medium to the left of the interface is air, and the medium to the right is lossy with an b. Set up and solve four equations from which the four unknowns , A, B, T may be

effective complex permittivity c , complex wavenumber k =  j = 0 c , and determined.

complex characteristic impedance c = 0 / c . The electric eld to the left and right of the c. If the slab is lossless and is designed to be a half-wave slab at the frequency , then
interface has the following form: what is the value of T?
d. If the slab is is lossy with nc = nr jni and is designed to be a half-wave slab with
respect to the real part of k, that is, d = , then, show that T is given by:

E0 ejkz + E0 ejkz , z 0
Ex = 1

E0 ejk z , T=  
z0 1 1
cosh d + nc + sinh d
2 nc

where , are the reection and transmission coefcients.
1. Determine the magnetic eld at both sides of the interface.
2. Show that the Poynting vector only has a z-component, given as follows at the two
sides of the interface:
5.13 Consider a two-layer dielectric structure as shown in Fig. 5.7.1, and let na , n1 , n2 , nb be the
refractive indices of the four media. Consider the four cases: (a) both layers are quarter-
wave, (b) both layers are half-wave, (c) layer-1 is quarter- and layer-2 half-wave, and (d) layer-1
is half- and layer-2 quarter-wave. Show that the reection coefcient at interface-1 is given
by the following expressions in the four cases:

|E0 |2   |E0 |2  2 2 z na n22 nb n21 na nb na nb n21 na nb n22

P= 1 ||2 , P = || e 1 = , 1 = , 1 = , 1 =
20 20 na n22 + nb n21 na + n b na nb + n21 na nb + n22

3. Moreover, show that P = P at the interface, (i.e., at z = 0). 5.14 Consider the lossless two-slab structure of Fig. 5.7.1. Write down all the transfer matrices
relating the elds Ei , i = 1, 2, 3 at the left sides of the three interfaces. Then, show the
5.11 Consider a lossy dielectric slab of thickness d and complex refractive index nc = nr jni at energy conservation equations:
an operating frequency , with air on both sides as shown below.
1   1   1   1
|E1+ |2 |E1 |2 = |E2+ |2 |E2 |2 = |E3+ |2 |E3 |2 = |E |2
a. Let k = j = k0 nc and c = 0 /nc be the corresponding complex wavenumber and
 a 1 2 b 3+

characteristic impedance of the slab, where k0 = 0 0 = /c0 and 0 = 0 / 0 .
Show that the transmission response of the slab may be expressed as follows: 5.15 An alternative way of representing the propagation relationship Eq. (5.1.12) is in terms of the
hyperbolic w-plane variable dened in terms of the reection coefcient , or equivalently,
the wave impedance Z as follows:

1
T=   = e2w  Z = coth(w) (5.9.2)
1 1
cos kd + j nc + sin kd
2 nc Show the equivalence of these expressions. Writing 1 = e2w1 and 2 = e2w2 , show that
Eq. (5.1.12) becomes equivalent to:
b. At the cell phone frequency of 900 MHz, the complex refractive index of concrete is
nc = 2.5 0.14j. Calculate the percentage of the transmitted power through a 20-cm w1 = w2 + jkl (propagation in w-domain) (5.9.3)
concrete wall. How is this percentage related to T and why?
This form is essentially the mathematical (as opposed to graphical) version of the Smith
c. Is there anything interesting about the choice d = 20 cm? Explain. chart and is particularly useful for numerical computations using MATLAB.
5.9. Problems 185

5.16 Plane A ying at a speed of 900 km/hr with respect to the ground is approaching plane B.
Plane As Doppler radar, operating at the X-band frequency of 10 GHz, detects a positive
Doppler shift of 2 kHz in the return frequency. Determine the speed of plane B with respect
to the ground. [Ans. 792 km/hr.]
6
5.17 The complete set of Lorentz transformations of the elds in Eq. (5.8.8) is as follows (see also
Eq. (K.31) of Appendix K): Multilayer Structures
1 1
Ex = (Ex + cBy ), Hy = (Hy + cDx ), Dx = (Dx + Hy ), By = (By + Ex )
c c

The constitutive relations in the rest frame S of the moving dielectric are the usual ones, that
is, By = Hy and Dx = Ex . By eliminating the primed quantities in terms of the unprimed
ones, show that the constitutive relations have the following form in the xed system S:

(1 2 ) Ex (n2 1)Hy /c (1 2 )Hy (n2 1)Ex /c

Dx = , By =
1 2 n 2 1 2 n2
 Higher-order transfer functions of the type of Eq. (5.7.2) can achieve broader reection-
where n is the refractive index of the moving medium, n = / 0 0 . Show that for free less notches and are used in the design of thin-lm antireection coatings, dielectric
space, the constitutive relations remain the same as in the frame S . mirrors, and optical interference lters [632694,754787], and in the design of broad-
band terminations of transmission lines [822832].
They are also used in the analysis, synthesis, and simulation of ber Bragg gratings
[788808], in the design of narrow-band transmission lters for wavelength-division
multiplexing (WDM), and in other ber-optic signal processing systems [818821].
They are used routinely in making acoustic tube models for the analysis and synthe-
sis of speech, with the layer recursions being mathematically equivalent to the Levinson
lattice recursions of linear prediction [833839]. The layer recursions are also used in
speech recognition, disguised as the Schur algorithm.
They also nd application in geophysical deconvolution and inverse scattering prob-
lems for oil exploration [840849].
The layer recursionsknown as the Schur recursions in this contextare intimately
connected to the mathematical theory of lossless bounded real functions in the z-plane
and positive real functions in the s-plane and nd application in network analysis, syn-
thesis, and stability [853867].

6.1 Multiple Dielectric Slabs

The general case of arbitrary number of dielectric slabs of arbitrary thicknesses is shown
in Fig. 6.1.1. There are M slabs, M + 1 interfaces, and M + 2 dielectric media, including
the left and right semi-innite media a and b .
The incident and reected elds are considered at the left of each interface. The
overall reection response, 1 = E1 /E1+ , can be obtained recursively in a variety of
ways, such as by the propagation matrices, the propagation of the impedances at the
interfaces, or the propagation of the reection responses.
The elementary reection coefcients i from the left of each interface are dened
in terms of the characteristic impedances or refractive indices as follows:

i i1 ni1 ni
i = = , i = 1, 2, . . . , M + 1 (6.1.1)
i + i1 ni1 + ni