242 7.

Oblique Incidence

perpendicular to that plane (along the y-direction) and transverse to the z-direction.

7 In perpendicular polarization, also known as s-polarization, -polarization, or TE

polarization, the electric elds are perpendicular to the plane of incidence (along the
y-direction) and transverse to the z-direction, and the magnetic elds lie on that plane.

Oblique Incidence The gure shows the angles of incidence and reection to be the same on either side.
This is Snels law of reection and is a consequence of the boundary conditions.
The gure also implies that the two planes of incidence and two planes of reection
all coincide with the xz-plane. This is also a consequence of the boundary conditions.
Starting with arbitrary wavevectors k = x kx + y ky + z kz and similarly for k ,
the incident and reected electric elds at the two sides will have the general forms:
 
E+ ej k+ r , E ej k r , E+ ej k+ r , E ej k r

The boundary conditions state that the net transverse (tangential) component of the
electric eld must be continuous across the interface. Assuming that the interface is at
7.1 Oblique Incidence and Snels Laws z = 0, we can write this condition in a form that applies to both polarizations:
 
With some redenitions, the formalism of transfer matrices and wave impedances for E T+ ej k+ r + E T ej k r = ET+ ej k+ r + ET ej k r , at z = 0 (7.1.1)
normal incidence translates almost verbatim to the case of oblique incidence.
By separating the elds into transverse and longitudinal components with respect where the subscript T denotes the transverse (with respect to z) part of a vector, that is,
to the direction the dielectrics are stacked (the z-direction), we show that the transverse ET = z (E z)= E z Ez . Setting z = 0 in the propagation phase factors, we obtain:
components satisfy the identical transfer matrix relationships as in the case of normal
   
incidence, provided we replace the media impedances by the transverse impedances E T+ ej(kx+ x+ky+ y) + E T ej(kx x+ky y) = ET+ ej(kx+ x+ky+ y) + ET ej(kx x+ky y) (7.1.2)
T dened below.
Fig. 7.1.1 depicts plane waves incident from both sides onto a planar interface sepa- For the two sides to match at all points on the interface, the phase factors must be
rating two media ,  . Both cases of parallel and perpendicular polarizations are shown. equal to each other for all x and y:
In parallel polarization, also known as p-polarization, -polarization, or TM po-    
larization, the electric elds lie on the plane of incidence and the magnetic elds are ej(kx+ x+ky+ y) = ej(kx x+ky y) = ej(kx+ x+ky+ y) = ej(kx x+ky y) (phase matching)

and this requires the x- and y-components of the wave vectors to be equal:

kx+ = kx = kx+ = kx

(7.1.3)
ky+ = ky = ky+ = ky

If the left plane of incidence is the xz-plane, so that ky+ = 0, then all y-components
of the wavevectors will be zero, implying that all planes of incidence and reection will
coincide with the xz-plane. In terms of the incident and reected angles ,  , the
conditions on the x-components read:

k sin + = k sin = k sin + = k sin  (7.1.4)

These imply Snels law of reection:

+ =
(Snels law of reection) (7.1.5)
+ =  

from
Fig. 7.1.1 Oblique incidence for TM- and TE-polarized waves. the German word senkrecht for perpendicular.
named after Willebrord Snel, b.1580, almost universally misspelled as Snell.
7.2. Transverse Impedance 243 244 7. Oblique Incidence

And also Snels law of refraction, that is, k sin = k sin  . Setting k = nk0 , k = n k0 , Similarly, the wave reected back into the left medium will have the form:
and k0 = /c0 , we have:  
E (r) = (x cos + z sin )A + y B ej k r
sin n 1  (7.2.3)
n sin = n sin  = (Snels law of refraction) (7.1.6)
H (r) = y A + (x cos + z sin )B ej k r
sin  n

It follows that the wave vectors shown in Fig. 7.1.1 will be explicitly: with corresponding transverse parts:
 
k = k+ = kx x + kz z = k sin x + k cos z E T (x, z) = x A cos + y B ej(kx xkz z)

1  (7.2.4)
k = kx x kz z = k sin x k cos z
H T (x, z) = y A + x B cos ej(kx xkz z)
(7.1.7)
k = k+ = kx x + kz z = k sin  x + k cos  z
Dening the transverse amplitudes and transverse impedances by:
k = kx x kz z = k sin  x k cos  z
AT = A cos , BT = B
The net transverse electric elds at arbitrary locations on either side of the interface
(7.2.5)
are given by Eq. (7.1.1). Using Eq. (7.1.7), we have: TM = cos , TE =
cos
 
E T (x, z)= E T+ ej k+ r + E T ej k r = E T+ ejkz z + E T ejkz z ejkx x and noting that AT /TM = A / and BT /TE = B cos /, we may write Eq. (7.2.2)
  (7.1.8) in terms of the transverse quantities as follows:
j k+ r j k r jkz z jkz z jkx x
ET (x, z)= ET+ e + ET e = ET+ e + ET e e
 
E T+ (x, z) = x AT+ + y BT+ ej(kx x+kz z)
In analyzing multilayer dielectrics stacked along the z-direction, the phase factor

ejkx x = ejkx x will be common at all interfaces, and therefore, we can ignore it and  AT+ BT+  j(kx x+kz z) (7.2.6)
H T+ (x, z) = y x e
restore it at the end of the calculations, if so desired. Thus, we write Eq. (7.1.8) as: TM TE
E T (z)= E T+ ejkz z + E T ejkz z Similarly, Eq. (7.2.4) is expressed as:
(7.1.9)
   
ET (z)= ET+ ejkz z + ET ejkz z E T (x, z) = x AT + y BT ej(kx xkz z)
In the next section, we work out explicit expressions for Eq. (7.1.9)  AT BT  j(kx xkz z) (7.2.7)
H T (x, z) = y + x e
TM TE

7.2 Transverse Impedance Adding up Eqs. (7.2.6) and (7.2.7) and ignoring the common factor ejkx x , we nd for
the net transverse elds on the left side:
The transverse components of the electric elds are dened differently in the two po-
larization cases. We recall from Sec. 2.10 that an obliquely-moving wave will have, in E T (z) = x ETM (z) + y ETE (z)
general, both TM and TE components. For example, according to Eq. (2.10.9), the wave (7.2.8)
H T (z) = y HTM (z) x HTE (z)
incident on the interface from the left will be given by:
  where the TM and TE components have the same structure provided one uses the ap-
E+ (r) = (x cos z sin )A+ + y B+ ej k+ r propriate transverse impedance:
1  (7.2.1)
H+ (r) = y A+ (x cos z sin )B+ ej k+ r ETM (z) = AT+ ejkz z + AT ejkz z

1   (7.2.9)
where the A+ and B+ terms represent the TM and TE components, respectively. Thus, HTM (z) = AT+ ejkz z AT ejkz z
the transverse components are: TM
  ETE (z) = BT+ ejkz z + BT ejkz z
E T+ (x, z) = x A+ cos + y B+ ej(kx x+kz z)
1   (7.2.10)
1  (7.2.2) HTE (z) = BT+ ejkz z BT ejkz z
H T+ (x, z) = y A+ x B+ cos ej(kx x+kz z) TE

7.2. Transverse Impedance 245 246 7. Oblique Incidence

We summarize these in the compact form, where ET stands for either ETM or ETE : The transverse parts of these are the same as those given in Eqs. (7.2.9) and (7.2.10).
On the right side of the interface, we have:
ET (z) = ET+ ejkz z + ET ejkz z  
E  (r) = ETM (r)+ETE (r)
1   (7.2.11)  
(7.2.19)
HT (z) = ET+ ejkz z ET ejkz z H  (r)= HTM (r)+HTE (r)
T
The transverse impedance T stands for either TM or TE :  

ETM (r) = (x cos  z sin  )A+ ej k+ r + (x cos  + z sin  )A ej k r

cos , TM, parallel, p-polarization  1   j k+ r  
T = (7.2.12) HTM (r) = y A+ e A ej k r
, TE, perpendicular, s-polarization 
cos    

ETE (r) = y B+ ej k+ r + B ej k r
Because = o /n, it is convenient to dene also a transverse refractive index
through the relationship T = 0 /nT . Thus, we have:  1   
HTE (r) = (x cos  z sin  )B+ ej k+ r + (x cos  + z sin  )B ej k r

n (7.2.20)
, TM, parallel, p-polarization
nT = cos (7.2.13)

n cos , TE, perpendicular, s-polarization
7.3 Propagation and Matching of Transverse Fields
For the right side of the interface, we obtain similar expressions:
     Eq. (7.2.11) has the identical form of Eq. (5.1.1) of the normal incidence case, but with
ET (z) = ET+ ejkz z + ET ejkz z
the substitutions:
 1       (7.2.14)
HT (z) = ET+ ejkz z ET ejkz z
T T , ejkz ejkz z = ejkz cos (7.3.1)


cos ,

TM, parallel, p-polarization Every denition and concept of Chap. 5 translates into the oblique case. For example,

T =  (7.2.15) we can dene the transverse wave impedance at position z by:

, TE, perpendicular, s-polarization
cos 
ET (z) ET+ ejkz z + ET ejkz z

n ZT (z)= = T (7.3.2)
 , TM, parallel, p-polarization HT (z) ET+ ejkz z ET ejkz z
nT = cos  (7.2.16)

n cos  , TE, perpendicular, s-polarization and the transverse reection coefcient at position z:

where ET stands for AT
= A
cos or 
= BT B .
ET (z) ET ejkz z
For completeness, we give below the complete expressions for the elds on both T (z)= = = T (0)e2jkz z (7.3.3)
ET+ (z) ET+ ejkz z
sides of the interface obtained by adding Eqs. (7.2.1) and (7.2.3), with all the propagation
factors restored. On the left side, we have: They are related as in Eq. (5.1.7):

E(r) = ETM (r)+ETE (r) 1 + T (z) ZT (z)T

(7.2.17) ZT (z)= T  T (z)= (7.3.4)
H(r)= HTM (r)+HTE (r) 1 T (z) ZT (z)+T
The propagation matrices, Eqs. (5.1.11) and (5.1.13), relating the elds at two posi-
where
tions z1 , z2 within the same medium, read now:
ETM (r) = (x cos z sin )A+ ej k+ r + (x cos + z sin )A ej k r



ET1+ ejkz l 0 ET2+
1  = (propagation matrix) (7.3.5)
HTM (r) = y A+ ej k+ r A ej k r ET1 0 ejkz l ET2

  (7.2.18)
ETE (r) = y B+ ej k+ r + B ej k r



ET1 cos kz l jT sin kz l ET2
1  = (propagation matrix) (7.3.6)
HT 1 j1
T sin kz l cos kz l HT 2
HTE (r) = (x cos z sin )B+ ej k+ r + (x cos + z sin )B ej k r

7.3. Propagation and Matching of Transverse Fields 247 248 7. Oblique Incidence

where l = z2 z1 . Similarly, the reection coefcients and wave impedances propagate


as: ET ET+
T = , T = (7.3.15)
ET+ ET+
ZT2 + jT tan kz l
T1 = T2 e2jkz l , ZT1 = T (7.3.7) The relationship of these coefcients to the reection and transmission coefcients
T + jZT2 tan kz l of the total eld amplitudes depends on the polarization. For TM, we have ET =
 
The phase thickness = kl = 2(nl)/ of the normal incidence case, where is A cos and ET = A cos  , and for TE, ET = B and ET = B . For both cases,
the free-space wavelength, is replaced now by: it follows that the reection coefcient T measures also the reection of the total
amplitudes, that is,
2
z = kz l = kl cos = nl cos (7.3.8) A cos A B
TM = = , TE =
A+ cos A+ B+
At the interface z = 0, the boundary conditions for the tangential electric and mag-
whereas for the transmission coefcients, we have:
netic elds give rise to the same conditions as Eqs. (5.2.1) and (5.2.2):
A+ cos  cos  A+ B+
  TM = = , TE =
ET = ET , HT = HT (7.3.9) A+ cos cos A+ B+
and in terms of the forward/backward elds: In addition to the boundary conditions of the transverse eld components, there are
also applicable boundary conditions for the longitudinal components. For example, in
 
ET+ + ET = ET+ + ET the TM case, the component Ez is normal to the surface and therefore, we must have
1   1    (7.3.10) the continuity condition Dz = Dz , or Ez =  Ez . Similarly, in the TE case, we must

ET+ ET =  ET+ ET have Bz = Bz . It can be veried that these conditions are automatically satised due to
T T
Snels law (7.1.6).
which can be solved to give the matching matrix:
The elds carry energy towards the z-direction, as well as the transverse x-direction.




ET+ 1 1 T 
ET+ The energy ux along the z-direction must be conserved across the interface. The cor-
=  (matching matrix) (7.3.11) responding components of the Poynting vector are:
ET T T 1 ET
where T , T are transverse reection coefcients, replacing Eq. (5.2.5): 1   1  
Pz = Re Ex Hy Ey Hx , Px = Re Ey Hz Ez Hy
2 2
T T nT nT For TM, we have Pz = Re[Ex Hy ]/2 and for TE, Pz = Re[Ey Hx ]/2. Using the
T =  =
T + T nT + nT above equations for the elds, we nd that Pz is given by the same expression for both
(Fresnel coefcients) (7.3.12) TM and TE polarizations:
2T 2nT
T = =
T + T nT + nT cos   cos  
Pz = |A+ |2 |A |2 , or, |B+ |2 |B |2 (7.3.16)
2 2
where T = 1 + T . We may also dene the reection coefcients from the right side
of the interface: T = T and T = 1 + T = 1 T . Eqs. (7.3.12) are known as the Using the appropriate denitions for ET and T , Eq. (7.3.16) can be written in terms
Fresnel reection and transmission coefcients. of the transverse components for either polarization:
The matching conditions for the transverse elds translate into corresponding match-
1  
ing conditions for the wave impedances and reection responses: Pz = |ET+ |2 |ET |2 (7.3.17)
2T

 T + T T + T As in the normal incidence case, the structure of the matching matrix (7.3.11) implies
ZT = ZT  T =  T = (7.3.13)
1 + T T 1 + T T that (7.3.17) is conserved across the interface.

If there is no left-incident wave from the right, that is, E = 0, then, Eq. (7.3.11) takes
the specialized form: 7.4 Fresnel Reection Coefcients





ET+ 1 1 T ET+
= (7.3.14) We look now at the specics of the Fresnel coefcients (7.3.12) for the two polarization
ET T T 1 0
cases. Inserting the two possible denitions (7.2.13) for the transverse refractive indices,
which explains the meaning of the transverse reection and transmission coefcients: we can express T in terms of the incident and refracted angles:
7.4. Fresnel Reection Coefcients 249 250 7. Oblique Incidence

 
n n n2d sin2 n2d cos cos n2d sin2
 
TM = cos cos  = n cos n cos TM =  , TE =  (7.4.4)
n n n cos + n cos
 nd2 sin2 + n2d cos cos + n2d sin2
+ (7.4.1)
cos cos 
If the incident wave is from inside the dielectric, then we set n = nd and n = 1:
n cos n cos 
TE =
n cos + n cos 
 
We note that for normal incidence, =  = 0, they both reduce to the usual n2 2
d sin nd cos
2
cos n2
d sin
2

reection coefcient = (n n )/(n + n ). Using Snels law, n sin = n sin  , and TM =  , TE =  (7.4.5)
n2 2
d sin + nd cos
2
cos + n2
d sin
2
some trigonometric identities, we may write Eqs. (7.4.1) in a number of equivalent ways.
In terms of the angle of incidence only, we have:

 2 2
n n
sin2 cos
n n
TM =    2
n 2 n
sin2 + cos
n n
  (7.4.2)
n 2
cos sin2
n
TE =  
n 2
cos + sin2
n

Note that at grazing angles of incidence, 90o , the reection coefcients tend to
Fig. 7.4.1 Air-dielectric interfaces.
TM 1 and TE 1, regardless of the refractive indices n, n . One consequence of
this property is in wireless communications where the effect of the ground reections The MATLAB function fresnel calculates the expressions (7.4.2) for any range of
causes the power of the propagating radio wave to attenuate with the fourth (instead values of . Its usage is as follows:
of the second) power of the distance, thus, limiting the propagation range (see Exam-
ple 22.3.5.) [rtm,rte] = fresnel(na,nb,theta); % Fresnel reection coefcients
We note also that Eqs. (7.4.1) and (7.4.2) remain valid when one or both of the media
are lossy. For example, if the right medium is lossy with complex refractive index nc =
nr jni , then, Snels law, n sin = nc sin  , is still valid but with a complex-valued 
and (7.4.2) remains the same with the replacement n nc . The third way of expressing
7.5 Maximum Angle and Critical Angle
the s is in terms of ,  only, without the n, n :
As the incident angle varies over 0 90o , the angle of refraction  will have
a corresponding range of variation. It can be determined by solving for  from Snels
sin 2 sin 2 tan( )
TM = = law, n sin = n sin  :

sin 2 + sin 2 tan( + )
(7.4.3) n
sin( ) sin  = sin (7.5.1)
TE = n
sin( + )
If n < n (we assume lossless dielectrics here,) then Eq. (7.5.1) implies that sin  =
Fig. 7.4.1 shows the special case of an air-dielectric interface. If the incident wave is (n/n )sin < sin , or  < . Thus, if the incident wave is from a lighter to a denser
from the air side, then Eq. (7.4.2) gives with n = 1, n = nd , where nd is the (possibly medium, the refracted angle is always smaller than the incident angle. The maximum
complex-valued) refractive index of the dielectric: value of  , denoted here by c , is obtained when has its maximum, = 90o :
Some references dene
TM with the opposite sign. Our convention was chosen because it has the
n
expected limit at normal incidence. sin c = (maximum angle of refraction) (7.5.2)
n
7.5. Maximum Angle and Critical Angle 251 252 7. Oblique Incidence

Thus, the angle ranges are 0
as well as the case n > n .
90o and 0  c . Fig. 7.5.1 depicts this case,
Both expressions for T are the ratios of a complex number and its conjugate, and
therefore, they are unimodular, |TM | = |TE | = 1, for all values of > c . The interface
becomes a perfect mirror, with zero transmittance into the lighter medium.
When > c , the elds on the right side of the interface are not zero, but do not
propagate away to the right. Instead, they decay exponentially with the distance z. There
is no transfer of power (on the average) to the right. To understand this behavior of the
elds, we consider the solutions given in Eqs. (7.2.18) and (7.2.20), with no incident eld
from the right, that is, with A = B = 0.
The longitudinal wavenumber in the right medium, kz , can be expressed in terms of
the angle of incidence as follows. We have from Eq. (7.1.7):

k2z + k2x = k2 = n2 k20

kz 2 + kx 2 = k2 = n2 k20

Because, kx = kx = k sin = nk0 sin , we may solve for kz to get:
Fig. 7.5.1 Maximum angle of refraction and critical angle of incidence.
kz2 = n2 k20 kx2 = n2 k20 k2x = n2 k20 n2 k20 sin2 = k20 (n2 n2 sin2 )

On the other hand, if n > n , and the incident wave is from a denser onto a lighter
or, replacing n = n sin c , we nd:
medium, then sin  = (n/n )sin > sin , or  > . Therefore,  will reach the
maximum value of 90o before does. The corresponding maximum value of satises
Snels law, n sin c = n sin(/2)= n , or, kz2 = n2 k20 (sin2 c sin2 ) (7.5.4)

If c , the wavenumber kz is real-valued and corresponds to ordinary propa-

n
sin c = (critical angle of incidence) (7.5.3) gating elds that represent the refracted wave. But if > c , we have kz2 < 0 and kz
n becomes pure imaginary, say kz = jz . The z-dependence of the elds on the right of
This angle is called the critical angle of incidence. If the incident wave were from the the interface will be:
right, c would be the maximum angle of refraction according to the above discussion. 
 
If c , there is normal refraction into the lighter medium. But, if exceeds c , ejkz z = ez z , z = nk0 sin2 sin2 c
the incident wave cannot be refracted and gets completely reected back into the denser
medium. This phenomenon is called total internal reection. Because n /n = sin c , we Such exponentially decaying elds are called evanescent waves because they are
may rewrite the reection coefcients (7.4.2) in the form: effectively conned to within a few multiples of the distance z = 1/z (the penetration
length) from the interface.
 
sin2 c sin2 sin2 c cos cos sin2 c sin2 The maximum value of z , or equivalently, the smallest penetration length 1/z , is
TM =  , TE =  achieved when = 90o , resulting in:
sin2 c sin2 + sin2 c cos cos + sin2 c sin2
 
max = nk0 1 sin2 c = nk0 cos c = k0 n2 n2
When < c , the reection coefcients are real-valued. At = c , they have the
values, TM = 1 and TE = 1. And, when > c , they become complex-valued with
Inspecting Eqs. (7.2.20), we note that the factor cos  becomes pure imaginary be-
unit magnitude. Indeed, switching the sign under the square roots, we have in this case:
cause cos2  = 1 sin2  = 1 (n/n )2 sin2 = 1 sin2 / sin2 c 0, for c .
  Therefore for either the TE or TM case, the transverse components ET and HT will
j sin2 sin2 c sin2 c cos cos + j sin2 sin2 c
TM =  , TE =  have a 90o phase difference, which will make the time-average power ow into the right

j sin2 sin2 c + sin2 c cos cos j sin2 sin2 c medium zero: Pz = Re(ET HT )/2 = 0.

where we used the evanescent denition of the square root as discussed in Eqs. (7.7.9) Example 7.5.1: Determine the maximum angle of refraction and critical angle of reection for
and (7.7.10), that is, we made the replacement (a) an air-glass interface and (b) an air-water interface. The refractive indices of glass and
  water at optical frequencies are: nglass = 1.5 and nwater = 1.333.
sin2 c sin2 j sin2 sin2 c , for c
7.5. Maximum Angle and Critical Angle 253 254 7. Oblique Incidence

Solution: There is really only one angle to determine, because if n = 1 and n = nglass , then Example 7.5.4: Apparent Depth. Underwater objects viewed from the outside appear to be
sin(c )= n/n = 1/nglass , and if n = nglass and n = 1, then, sin(c )= n /n = 1/nglass . closer to the surface than they really are. The apparent depth of the object depends on
Thus, c = c : our viewing angle. Fig. 7.5.4 shows the geometry of the incident and refracted rays.

1
c = asin = 41.8o
1.5

For the air-water case, we have:


1
c = asin = 48.6o
1.333

The refractive index of water at radio frequencies and below is nwater = 9 approximately.
The corresponding critical angle is c = 6.4o .

Example 7.5.2: Prisms. Glass prisms with 45o angles are widely used in optical instrumentation
for bending light beams without the use of metallic mirrors. Fig. 7.5.2 shows two examples. Fig. 7.5.4 Apparent depth of underwater object.

Let be the viewing angle and let z and z be the actual and apparent depths. Our perceived
depth corresponds to the extension of the incident ray at angle . From the gure, we have:
z = x cot  and z = x cot . It follows that:

cot sin  cos

z = z= z
cot  sin cos 

Using Snels law sin / sin  = n /n = nwater , we eventually nd:

Fig. 7.5.2 Prisms using total internal reection. cos

z =  z
n2water sin2
In both cases, the incident beam hits an internal prism side at an angle of 45o , which is
greater than the air-glass critical angle of 41.8o . Thus, total internal reection takes place At normal incidence, we have z = z/nwater = z/1.333 = 0.75z.
and the prism side acts as a perfect mirror.

Reection and refraction phenomena are very common in nature. They are responsible for
Example 7.5.3: Optical Manhole. Because the air-water interface has c = 48.6o , if we were to the twinkling and aberration of stars, the attening of the setting sun and moon, mirages,
view a water surface from above the water, we could only see inside the water within the rainbows, and countless other natural phenomena. Four wonderful expositions of such
cone dened by the maximum angle of refraction. effects are in Refs. [5053]. See also the web page [1827].

Conversely, were we to view the surface of the water from underneath, we would see the Example 7.5.5: Optical Fibers. Total internal reection is the mechanism by which light is
air side only within the critical angle cone, as shown in Fig. 7.5.3. The angle subtended by guided along an optical ber. Fig. 7.5.5 shows a step-index ber with refractive index
this cone is 248.6 = 97.2o . nf surrounded by cladding material of index nc < nf .

Fig. 7.5.3 Underwater view of the outside world.

Fig. 7.5.5 Launching a beam into an optical ber.
The rays arriving from below the surface at an angle greater than c get totally reected.
But because they are weak, the body of water outside the critical cone will appear dark. If the angle of incidence on the ber-cladding interface is greater than the critical angle,
The critical cone is known as the optical manhole [50].

then total internal reection will take place. The gure shows a beam launched into the
7.5. Maximum Angle and Critical Angle 255 256 7. Oblique Incidence

ber from the air side. The maximum angle of incidence a must be made to correspond to The relative phase change between the TE and TM polarizations will be:
the critical angle c of the ber-cladding interface. Using Snels laws at the two interfaces,
TM
we have: = e2jTM 2jTE +j
TE
nf nc
sin a = sin b , sin c = It is enough to require that TM TE = /8 because then, after two reections, we will
na nf
have a 90o change:
Noting that b = 90o c , we nd:  2
TM TM
 = ej/4+j = ej/2+2j = ej/2
 TE TE
nf nf n2f n2c
sin a = cos c = 1 sin2 c =
na na na From the design condition TM TE = /8, we obtain the required value of x and then
of . Using a trigonometric identity, we have:
For example, with na = 1, nf = 1.49, and nc = 1.48, we nd c = 83.4o and a = 9.9o . The

angle a is called the acceptance angle, and the quantity NA = n2f n2c , the numerical tan TM tan TE xn2 x 
tan(TM TE )= = = tan
aperture of the ber. 1 + tan TM tan TE 1 + n2 x2 8

Besides its use in optical bers, total internal reection has several other applications [556 This gives the quadratic equation for x:
592], such as internal reection spectroscopy, chemical and biological sensors, ngerprint
identication, surface plasmon resonance, and high resolution microscopy.


1 1  1 cos2 c
x2 1 2 x + 2 = x2 x + sin2 c = 0 (7.5.6)
tan(/8) n n tan(/8)
Example 7.5.6: Fresnel Rhomb. The Fresnel rhomb is a glass prism depicted in Fig. 7.5.6 that
acts as a 90o retarder. It converts linear polarization into circular. Its advantage over the Inserting the two solutions of (7.5.6) into Eq. (7.5.5), we may solve for sin , obtaining two
birefringent retarders discussed in Sec. 4.1 is that it is frequency-independent or achro- possible solutions for :
matic.

x2 + sin2 c
sin = (7.5.7)
x2 + 1
We may also eliminate x and express the design condition directly in terms of :


cos sin2 sin2 c 
Fig. 7.5.6 Fresnel rhomb. = tan (7.5.8)
sin2 8
However, the two-step process is computationally more convenient. For n = 1.51, we nd
Assuming a refractive index n = 1.51, the critical angle is c = 41.47o . The angle of the
the two roots of Eq. (7.5.6): x = 0.822 and x = 0.534. Then, (7.5.7) gives the two values
rhomb, = 54.6o , is also the angle of incidence on the internal side. This angle has been
= 54.623o and = 48.624o . The rhomb could just as easily be designed with the second
chosen such that, at each total internal reection, the relative phase between the TE and
value of .
TM polarizations changes by 45o , so that after two reections it changes by 90o .
For n = 1.50, we nd the angles = 53.258o and 50.229o . For n = 1.52, we have
The angle of the rhomb can be determined as follows. For c , the reection coefcients
= 55.458o and 47.553o . See Problem 7.5 for an equivalent approach.

can be written as the unimodular complex numbers:

 Example 7.5.7: Goos-Hanchen Effect. When a beam of light is reected obliquely from a denser-
1 + jx 1 + jxn2 sin2 sin2 c to-rarer interface at an angle greater than the TIR angle, it suffers a lateral displacement,
TE = , TM = , x= (7.5.5) relative to the ordinary reected ray, known as the Goos-Hanchen shift, as shown Fig. 7.5.7.
1 jx 1 jxn2 cos
Let n, n be the refractive indices of the two media with n > n , and consider rst the case
where sin c = 1/n. It follows that:
of ordinary reection at an incident angle 0 < c . For a plane wave with a free-space
wavenumber k0 = /c0 and wavenumber components kx = k0 n sin 0 , kz = k0 n cos 0 ,
TE = e2jTE , TM = ej+2jTM the corresponding incident, reected, and transmitted transverse electric elds will be:

Ei (x, z) = ejkx x ejkz z

where TE , TM are the phase angles of the numerators, that is,
Er (x, z) = (kx )ejkx x e+jkz z
tan TE = x , tan TM = xn2 

Et (x, z) = (kx )ejkx x ejkz z , kz = k20 n2 k2x
7.5. Maximum Angle and Critical Angle 257 258 7. Oblique Incidence

follows that the two terms in the reected wave Er (x, z) will differ by a small amplitude
change and therefore we can set (kx + kx ) (kx ). Similarly, in the transmitted eld
we may set (kx + kx ) (kx ). Thus, when 0 < c , Eq. (7.5.9) reads approximately,
 
Ei (x, z) = ejkx x ejkz z 1 + ejkx (xz tan 0 )
 
Er (x, z) = (kx )ejkx x e+jkz z 1 + ejkx (x+z tan 0 ) (7.5.10)
jkx x jkz z
 jkx (xz tan 0 )

Et (x, z) = (kx )e e 1+e
 
Noting that 1 + ejkx (xz tan 0 )  2, with equality achieved when x z tan 0 = 0, it
follows that the intensities of these waves are maximized along the ordinary geometric
rays dened by the beam angles 0 and 0 , that is, along the straight lines:

x z tan 0 = 0 , incident ray

x + z tan 0 = 0 , reected ray (7.5.11)
 x z tan 0 = 0 , transmitted ray
Fig. 7.5.7 Goos-Hanchen shift, with n > n and 0 > c .

On the other hand, if 0 > c and 0 + > c , the reection coefcients become
where (kx ) and (kx )= 1 + (kx ) are the transverse reection and transmission coef- unimodular complex numbers, as in Eq. (7.5.5). Writing (kx )= ej(kx ) , Eq. (7.5.9) gives:
cients, viewed as functions of kx . For TE and TM polarizations, (kx ) is given by  
Er (x, z)= ejkx x e+jkz z ej(kx ) + ej(kx +kx ) ejkx (x+z tan 0 ) (7.5.12)
kz kz k n2 kz n2
TE (kx )= , TM (kx )= z 2 Introducing the Taylor series expansion, (kx + kx ) (kx )+kx  (kx ), we obtain:
kz + kz kz n + kz n2
  
Er (x, z)= ej(kx ) ejkx x e+jkz z 1 + ejkx (kx ) ejkx (x+z tan 0 )
A beam can be made up by forming a linear combination of such plane waves having a small
spread of angles about 0 . For example, consider a second plane wave with wavenumber Setting x0 =  (kx ), we have:
components kx + kx and kz + kz . These must satisfy (kx + kx )2 +(kz + kz )2 =  
k2x + k2z = k20 n2 , or to lowest order in kx , Er (x, z)= ej(kx ) ejkx x e+jkz z 1 + ejkx (xx0 +z tan 0 ) (7.5.13)

kx This implies that the maximum intensity of the reected beam will now be along the shifted
kx kx + kz kz = 0 kz = kx = kx tan 0
kz ray dened by:
x x0 + z tan 0 = 0 , shifted reected ray (7.5.14)
Similarly, we have for the transmitted wavenumber kz = kx tan 0 , where 0 is given
by Snels law, n sin 0 = n sin 0 . The incident, reected, and transmitted elds will be Thus, the origin of the Goos-Hanchen shift can be traced to the relative phase shifts arising
given by the sum of the two plane waves: from the reection coefcients in the plane-wave components making up the beam. The
parallel displacement, denoted by D in Fig. 7.5.7, is related to x0 by D = x0 cos 0 . Noting
Ei (x, z) = ejkx x ejkz z + ej(kx +kx )x ej(kz +kz )z that dkx = k0 n cos d, we obtain

Er (x, z) = (kx )ejkx x e+jkz z + (kx + kx )ej(kx +kx )x e+j(kz +kz )z 

d 1 d  
D = cos 0 = (Goos-Hanchen shift) (7.5.15)
 
Et (x, z) = (kx )ejkx x ejkz z + (kx + kx )ej(kx +kx )x ej(kz +kz )z
 dkx k0 n d 0

Replacing kz = kx tan 0 and kz = kx tan 0 , we obtain: Using Eq. (7.5.5), we obtain the shifts for the TE and TM cases:

  2 sin 0 n 2
Ei (x, z) = ejkx x ejkz z
1 + ejkx (xz tan 0 )
DTE =  , DTM = DTE (7.5.16)
k0 n sin 0 sin c
2 2 (n2 + n2 )sin2 0 n2
jkx x +jkz z
 
Er (x, z) = e e (kx )+(kx + kx )ejkx (x+z tan 0 ) (7.5.9)
   These expressions are not valid near the critical angle 0  c because then the Taylor
jkz z
Et (x, z) = ejkx x e (kx )+(kx + kx )ejkx (xz tan 0 ) series expansion for (kx ) cannot be justied. Since geometrically, z0 = D/(2 sin 0 ), it
follows from (7.5.16) that the effective penetration depth into the n medium is given by:
The incidence angle of the second wave is 0 + , where is obtained by expanding
kx + kx = k0 n sin(0 + ) to rst order, or, kx = k0 n cos 0 . If we assume that 1 1 1 n2
zTE =  = , zTM = (7.5.17)
0 < c , as well as 0 + < c , then (kx ) and (kx + kx ) are both real-valued. It k0 n sin 0 sin c
2 2 z z (n2 + n2 )sin2 0 n2
7.6. Brewster Angle 259 260 7. Oblique Incidence

  
where z = k2x k20 n2 = k0 n2 sin2 0 n2 = k0 n sin2 0 sin2 c . These expres- The angle B is related to B by Snels law, n sin B = n sin B , and corresponds
sions are consistent with the eld dependence e =ejkz z z z 
inside the n medium, which to zero reection from that side, TM = TM = 0. A consequence of Eq. (7.6.2) is that
shows that the effective penetration length is of the order of 1/z .

B = 90o B , or, B + B = 90o . Indeed,

sin B n sin B
= tan B = =
7.6 Brewster Angle cos B n sin B

which implies cos B = sin B , or B = 90o B . The same conclusion can be reached
The Brewster angle is that angle of incidence at which the TM Fresnel reection coef-
immediately from Eq. (7.4.3). Because, B B = 0, the only way for the ratio of the
cient vanishes, TM = 0. The TE coefcient TE cannot vanish for any angle , for
two tangents to vanish is for the denominator to be innity, that is, tan(B + B )= ,
non-magnetic materials. A scattering model of Brewsters law is discussed in [693].
or, B + B = 90o .
Fig. 7.6.1 depicts the Brewster angles from either side of an interface.
As shown in Fig. 7.6.1, the angle of the refracted ray with the would-be reected ray
The Brewster angle is also called the polarizing angle because if a mixture of TM
is 90o . Indeed, this angle is 180o (B + B )= 180o 90o = 90o .
and TE waves are incident on a dielectric interface at that angle, only the TE or perpen-
The TE reection coefcient at B can be calculated very simply by using Eq. (7.6.1)
dicularly polarized waves will be reected. This is not necessarily a good method of
into (7.4.2). After canceling a common factor of cos B , we nd:
generating polarized waves because even though TE is non-zero, it may be too small
to provide a useful amount of reected power. Better polarization methods are based 
n 2
on using (a) multilayer structures with alternating low/high refractive indices and (b) 1
n n2 n2
birefringent and dichroic materials, such as calcite and polaroids. TE (B )=  2 = 2 (7.6.4)
n n + n2
1+
n
Example 7.6.1: Brewster angles for water. The Brewster angles from the air and the water sides
of an air-water interface are:
 
1.333 1
B = atan = 53.1o , B = atan = 36.9o
1 1.333

We note that B +B = 90o . At RF, the refractive index is nwater = 9 and we nd B = 83.7o
and B = 6.3o . We also nd TE (B )= 0.2798 and |TE (B )|2 = 0.0783/ Thus, for TE
waves, only 7.83% of the incident power gets reected at the Brewster angle.

Example 7.6.2: Brewster Angles for Glass. The Brewster angles for the two sides of an air-glass
interface are:
Fig. 7.6.1 Brewster angles.  
1.5 1
B = atan = 56.3o , B = atan = 33.7o
1 1.5
The Brewster angle B is determined by the condition, TM = 0, in Eq. (7.4.2). Setting
the numerator of that expression to zero, we have: Fig. 7.6.2 shows the reection coefcients |TM ()|, |TE ()| as functions of the angle of
 incidence from the air side, calculated with the MATLAB function fresnel.
2 2
n n Both coefcients start at their normal-incidence value || = |(1 1.5)/(1 + 1.5)| = 0.2
sin2 B = cos B (7.6.1)
n n and tend to unity at grazing angle = 90o . The TM coefcient vanishes at the Brewster
angle B = 56.3o .
After some algebra, we obtain the alternative expressions:
The right graph in the gure depicts the reection coefcients |TM ( )|, |TE ( )| as
n 
n  functions of the incidence angle  from the glass side. Again, the TM coefcient vanishes
sin B =  tan B = (Brewster angle) (7.6.2) at the Brewster angle B = 33.7o . The typical MATLAB code for generating this graph was:
n2 + n2 n

Similarly, the Brewster angle B from the other side of the interface is: na = 1; nb = 1.5;
[thb,thc] = brewster(na,nb); % calculate Brewster angle
th = linspace(0,90,901); % equally-spaced angles at 0.1o intervals
n n
sin B =  tan B =  (Brewster angle) (7.6.3) [rte,rtm] = fresnel(na,nb,th); % Fresnel reection coefcients
n2 + n2 n plot(th,abs(rtm), th,abs(rte));
7.7. Complex Waves 261 262 7. Oblique Incidence

Air to Glass Glass to Air Lossy Dielectric Lossy Dielectric

1 1 1 nd = 1.50 0.15 j 1 nd = 1.50 0.30 j

TM TM TM
TE TE TE
0.8 0.8 0.8 0.8
lossless lossless
TM

|T ( )|
|T ()|

|T ()|

|T ()|
0.6 0.6 TE 0.6 0.6

0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2

B B c

0 0 0 0
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90

Fig. 7.6.2 TM and TE reection coefcients versus angle of incidence. Fig. 7.6.3 TM and TE reection coefcients for lossy dielectric.

The critical angle of reection is in this case c = asin(1/1.5)= 41.8o . As soon as  unnecessary complex algebra, it proves convenient to recast impedances, reection co-
exceeds c , both coefcients become complex-valued with unit magnitude. efcients, and eld expressions in terms of wavenumbers. This can be accomplished by
The value of the TE reection coefcient at the Brewster angle is TE = TE = 0.38, making substitutions such as cos = kz /k and sin = kx /k.
and the TE reectance |TE |2 = 0.144, or 14.4 percent. This is too small to be useful for Using the relationships k = and k/ = , we may rewrite the TE and TM
generating TE polarized waves by reection. transverse impedances in the forms:
Two properties are evident from Fig. 7.6.2. One is that |TM | |TE | for all angles of
incidence. The other is that B c . Both properties can be proved in general.

k kz kz
TE = = = , TM = cos = = (7.7.1)
cos kz kz k 
Example 7.6.3: Lossy dielectrics. The Brewster angle loses its meaning if one of the media is
lossy. For example, assuming a complex refractive index for the dielectric, nd = nr jni , We consider an interface geometry as shown in Fig. 7.1.1 and assume that there are
we may still calculate the reection coefcients from Eq. (7.4.4). It follows from Eq. (7.6.2) no incident elds from the right of the interface. Snels law implies that kx = kx , where
that the Brewster angle B will be complex-valued.
kx = k sin = 0  sin , if the incident angle is real-valued.
Fig. 7.6.3 shows the TE and TM reection coefcients versus the angle of incidence (from Assuming non-magnetic media from both sides of an interface ( =  = 0 ), the TE
air) for the two cases nd = 1.50 0.15j and nd = 1.50 0.30j and compares them with and TM transverse reection coefcients will take the forms:
the lossless case of nd = 1.5. (The values for ni were chosen only for plotting purposes
and have no physical signicance.) TE TE kz kz TM TM k  kz 
TE = = , TM = = z (7.7.2)
The curves retain much of their lossless shape, with the TM coefcient having a minimum TE + TE kz + kz TM + TM kz  + kz 
near the lossless Brewster angle. The larger the extinction coefcient ni , the larger the
deviation from the lossless case. In the next section, we discuss reection from lossy The corresponding transmission coefcients will be:
media in more detail.

2kz 2kz 
TE = 1 + TE = , TM = 1 + TM = (7.7.3)
kz + kz kz  + kz 
7.7 Complex Waves
We can now rewrite Eqs. (7.2.18) and (7.2.20) in terms of transverse amplitudes and
In this section, we discuss some examples of complex waves that appear in oblique transverse reection and transmission coefcients. Dening E0 = A+ cos or E0 = B+
incidence problems. We consider the cases of (a) total internal reection, (b) reection in the TM or TE cases and replacing tan = kx /kz , tan  = kx /kz = kx /kz , we have for
from and refraction into a lossy medium, (c) the Zenneck surface wave, and (d) surface the TE case for the elds at the left and right sides of the interface:
plasmons. Further details may be found in [902909] and [1293].
Because the wave numbers become complex-valued, e.g., k = j , the angle of
refraction and possibly the angle of incidence may become complex-valued. To avoid
7.7. Complex Waves 263 264 7. Oblique Incidence

 
E(r) = y E0 ejkz z + TE ejkz z ejkx x
   
E0 kx kx
H(r) = x + z ejkz z + TE x + z ejkz z ejkx x
TE kz kz

(TE) (7.7.4)
E  (r) = y TE E0 ejkz z ejkx x

TE E0 kx jkz z jkx x
H  (r) = x +  z e e
TE kz

and for the TM case:

    Fig. 7.7.1 Constant-phase and constant-amplitude planes for the transmitted wave.
kx kx
E(r) = E0 x z ejkz z + TM x + z ejkz z ejkx x
kz kz
The wave numbers kz , kz are related to kx through
E0  jkz z 
H(r) = y e TM ejkz z ejkx x
TM
k2z = 2  k2x , kz2 = 2  k2x
 (TM) (7.7.5)
kx 
E  (r) = TM E0 x  z ejkz z ejkx x In calculating kz and kz by taking square roots of the above expressions, it is neces-
kz
sary, in complex-waves problems, to get the correct signs of their imaginary parts, such
TM E0 jkz z jkx x
H  (r) = y e e that evanescent waves are described correctly. This leads us to dene an evanescent
TM
square root as follows. Let  = R jI with I > 0 for an absorbing medium, then

Equations (7.7.4) and (7.7.5) are dual to each other, as are Eqs. (7.7.1). They transform

  2 (R jI )k2x , if I =
0
into each other under the duality transformation E H, H E,  , and .
kz = sqrte 2 (R jI )kx 2
=  (7.7.9)
See Sec. 18.2 for more on the concept of duality.

j k2x 2 R , if I = 0
In all of our complex-wave examples, the transmitted wave will be complex with
k = kx x + kz z =  j
 = (x jx )x + (z jz )z. This must satisfy the constraint
 
If I = 0 and 2 R k2x > 0, then the two expressions give the same answer. But if

k k = 0  . Thus, the space dependence of the transmitted elds will have the
2
I = 0 and 2 R k2x < 0, then kz is correctly calculated from the second expression.
general form:
The MATLAB function sqrte.m implements the above denition. It is dened by
     
ejkz z ejkx x = ej(z jz )z ej(x jx )x = e(z z+x x) ej(z z+x x) (7.7.6) j |z| , if Re(z)< 0 and Im(z)= 0
y = sqrte(z)= (evanescent SQRT) (7.7.10)
z, otherwise
For the wave to attenuate at large distances into the right medium, it is required that
z > 0. Except for the Zenneck-wave case, which has x > 0, all other examples will
have x = 0, corresponding to a real-valued wavenumber kx = kx = x . Fig. 7.7.1 shows Some examples of the issues that arise in taking such square roots are elaborated in
the constant-amplitude and constant-phase planes within the transmitted medium de- the next few sections.
ned, respectively, by:

z z + x x = const. , z z + x x = const. (7.7.7) 7.8 Total Internal Reection

As shown in the gure, the corresponding angles and that the vectors  and We already discussed this case in Sec. 7.5. Here, we look at it from the point of view of

form with the z-axis are given by: complex-waves. Both media are assumed to be lossless, but with  >  . The angle of

incidence will be real, so that kx = kx = k sin and kz = k cos , with k = 0 .
x x Setting kz = z jz , we have the constraint equation:
tan = , tan = (7.7.8)
z z
kx2 + kz2 = k2 kz2 = (z jz )2 = 2 0  k2x = 2 0 (  sin2 )
7.8. Total Internal Reection 265 266 7. Oblique Incidence

which separates into the real and imaginary parts: 7.9 Oblique Incidence on a Lossy Medium
z2 z2 = 2 0 (  sin2 )= k2 (sin2 c sin2 ) Here, we assume a lossless medium on the left side of the interface and a lossy one, such
(7.8.1) as a conductor, on the right. The effective dielectric constant  of the lossy medium is
z z = 0
specied by its real and imaginary parts, as in Eq. (2.6.2):
where we set sin2 c =  / and k2 = 2 0 . This has two solutions: (a) z = 0 and 

z2 = k2 (sin2 c sin2 ), valid when c , and (b) z = 0 and z2 = k2 (sin2  = d j 
d + = R jI (7.9.1)

sin2 c ), valid when c .
Equivalently, we may characterize the lossy medium by the real and imaginary parts
Case (a) corresponds to ordinary refraction into the right medium, and case (b), to
of the wavenumber k , using Eq. (2.6.12):
total internal reection. In the latter case, we have kz = jz and the TE and TM
reection coefcients (7.7.2) become unimodular complex numbers:  
k =  j = 0  = 0 (R jI ) (7.9.2)
kz kz kz + jz k  k z  kz  + jz 
TE = = , TM = z = (7.8.2) In the left medium, the wavenumber is real with components kx = k sin , kz =
kz + kz kz jz kz  + k z  kz  jz 
k cos , with k = 0 . In the lossy medium, the wavenumber is complex-valued with
The complete expressions for the elds are given by Eqs. (7.7.4) or (7.7.5). The prop- components kx = kx and kz = z jz . Using Eq. (7.9.2) in the condition k k = k2 ,


agation phase factor in the right medium will be in case (b): we obtain:

jkz z z z
e ejkx x = e ejkx x kx2 + kz2 = k2 k2x + (z jz )2 = ( j )2 = 2 0 (R jI ) (7.9.3)

Thus, the constant-phase planes are the constant-x planes ( = 90o ), or, the yz- which separates into its real and imaginary parts:
planes. The constant-amplitude planes are the constant-z planes ( = 0o ), or, the xy-
planes, as shown in Fig. 7.8.1. z2 z2 = 2 2 k2x = 2 0 R k2x = 2 0 (R  sin2 ) DR
(7.9.4)
2z z = 2  = 2 0 I DI

where we replaced k2x = k2 sin2 = 2 0  sin2 . The solutions of Eqs. (7.9.4) leading
to a non-negative z are:

 1/2  1/2
D2R + D2I + DR D2R + D2I DR
z = , z = (7.9.5)
2 2

For MATLAB implementation, it is simpler to solve Eq. (7.9.3) directly as a complex

square root (but see also Eq. (7.9.10)):
  
kz = z jz = k2 k2x = 2 0 (R jI )k2x = DR jDI (7.9.6)
Fig. 7.8.1 Constant-phase and constant-amplitude planes for total internal reection ( c ).
Eqs. (7.9.5) dene completely the reection coefcients (7.7.2) and the eld solutions
It follows from Eq. (7.8.2) that in case (b) the phases of the reection coefcients are: for both TE and TM waves given by Eqs. (7.7.4) and (7.7.5). Within the lossy medium the
transmitted elds will have space-dependence:
 
z k2x k20 n2 sin2 sin2 c   
TE = e 2jTE
, tan TE = =  = ejkz z ejkx x = ez z ej(z z+kx x)
kz k20 n2 k2x cos
  (7.8.3) The elds attenuate exponentially with distance z. The constant phase and ampli-
n2 z n2 k2x k20 n2 n2 sin2 sin2 c tude planes are shown in Fig. 7.9.1.
TM = ej+2jTM , tan TM = 2 =  =
n kz n2 k20 n2 k2x n2 cos For the reected elds, the TE and TM reection coefcients are given by Eqs. (7.7.2).
If the incident wave is linearly polarized having both TE and TM components, the corre-

where k0 = 0 is the free-space wave number. sponding reected wave will be elliptically polarized because the ratio TM /TE is now
7.9. Oblique Incidence on a Lossy Medium 267 268 7. Oblique Incidence

AirWater at 1 GHz AirWater at 100 MHz

1 1

0.8 0.8

|T ()|

|T ()|
0.6 TM 0.6 TM
TE TE

0.4 0.4

0.2 0.2

0 0
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
Fig. 7.9.1 Constant-phase and constant-amplitude planes for refracted wave.

Fig. 7.9.2 TM and TE reection coefcients for air-water interface.

complex-valued. Indeed, using the relationships k2x +k2z = 2 0  and k2x +kz2 = 2 0 
in TM of Eq. (7.7.2), it can be shown that (see Problem 7.5):
Example 7.9.1: Fig. 7.9.2 shows the TM and TE reection coefcients as functions of the inci-
TM kz kz k2x kz k sin tan  jz k sin tan dent angle , for an air-sea water interface at 100 MHz and 1 GHz. For the air side we
= 2 =  = z (7.9.7)
TE 
kz kz + kx kz + k sin tan z jz + k sin tan have  = 0 and for the water side:  = 810 j/, with = 4 S/m, which gives
 = (81 71.9j)0 at 1 GHz and  = (81 719j)0 at 100 MHz.
In the case of a lossless medium,  = R and I = 0, Eq. (7.9.5) gives: 
At 1 GHz, we calculate k = 0  =  j = 203.90 77.45j rad/m and k =
   j = 42.04 37.57j rad/m at 100 MHz. The following MATLAB code was used to
|DR | + DR |DR | DR
z = , z = (7.9.8) carry out the calculations, using the formulation of this section:
2 2
ep0 = 8.854e-12; mu0 = 4*pi*1e-7;
If R > , then DR = 2 0 (R  sin2
 ) is positive for all angles , and (7.9.8) sigma = 4; f = 1e9; w = 2*pi*f;

gives the expected result z = DR = 0 (R  sin2 ) and z = 0. ep1 = ep0; ep2 = 81*ep0 - j*sigma/w;

On the other hand, in the case of total internal reection, that is, when R < , the
k1 = w*sqrt(mu0*ep1); k2 = w*sqrt(mu0*ep2); % Eq. (7.9.2)
quantity DR is positive for angles < c , and negative for > c , where the critical
angle is dened through R =  sin2 c sothat DR = 2 0 (sin2 c sin2 ). Eqs. (7.9.8) th = linspace(0,90,901); thr = pi*th/180;
  
 the right answers, that is, z = |DR | and z = 0, if c , and z = 0 and
still give
k1x = k1*sin(thr); k1z = k1*cos(thr);
z = |DR |, if > c . k2z = sqrt(w^2*mu0*ep2 - k1x.^2); % Eq. (7.9.6)
For the case of a very good conductor, we have I  R , or DI  |DR |, and

Eqs. (7.9.5) give z  z  DI /2, or rte = abs((k1z - k2z)./(k1z + k2z)); % Eq. (7.7.2)
rtm = abs((k2z*ep1 - k1z*ep2)./(k2z*ep1 + k1z*ep2));

0
z  z      , provided 1 (7.9.9) plot(th,rtm, th,rte);
2 

In this case, the angle of refraction for the phase vector  becomes almost zero The TM reection coefcient reaches a minimum at the pseudo-Brewster angles 84.5o and
so that, regardless of the incidence angle , the phase planes are almost parallel to the 87.9o , respectively for 1 GHz and 100 MHz.
constant-z amplitude planes. Using Eq. (7.9.9), we have: The reection coefcients TM and TE can just as well be calculated from Eq. (7.4.2), with
 
 n = 1 and n =  /0 , where for 1 GHz we have n = 81 71.9j = 9.73 3.69j, and for

kx 0  sin 2 100 MHz, n = 81 719j = 20.06 17.92j.

tan =  =
 = sin
z 0 /2
In computing the complex square roots in Eq. (7.9.6), MATLAB usually gets the right
which is very small regardless of . For example, for copper ( = 5.7107 S/m) at 10 answer, that is, z 0 and z 0.

GHz, and air on the left side ( = 0 ), we nd 2/ = 1.4104 . If R > , then DR = 2 0 (R  sin2 ) is positive for all angles , and (7.9.6) may
be used without modication for any value of I .
7.9. Oblique Incidence on a Lossy Medium 269 270 7. Oblique Incidence

If R <  and I > 0, then Eq. (7.9.6) still gives the correct algebraic signs for any total internal reection, we have z = 0, which gives |TE | = 1. Similar conclusions can

angle . But when I = 0, that is, for a lossless medium, then  DI = 0 and kz = DR . be reached for the TM case of Eq. (7.7.5). The matching condition at the interface is now:

For > c we have DR < 0 and MATLAB gives kz = DR = |DR | = j |DR |, which 
has the wrong sign for z (we saw that Eqs. (7.9.5) work correctly in this case.)       +  
1 |TM |2 = Re  |TM |2 = R z  2 I z |TM |2 (7.9.13)
In order to coax MATLAB to produce the right algebraic sign for z in all cases, we kz kz |kz |
may redene Eq. (7.9.6) by using double conjugation:
Using the constraint 2 o I = 2z z , it follows that the right-hand side will again

  be proportional to z (with a positive proportionality coefcient.) Thus, the non-negative
j |DR | , if DI = 0 and DR < 0
kz = z jz = (DR jDI ) =  (7.9.10) sign of z implies that |TM | 1.

DR jDI , otherwise

One word of caution, however, is that current versions of MATLAB (ver. 7.0) may
produce inconsistent results for (7.9.10) depending on whether DI is a scalar or a vector
passing through zero. Compare, for example, the outputs from the statements:
DI = 0; kz = conj(sqrt(conj(-1 - j*DI)));
DI = -1:1; kz = conj(sqrt(conj(-1 - j*DI)));
Note, however, that Eq. (7.9.10) does work correctly when DI is a single scalar with
DR being a vector of values, e.g., arising from a vector of angles .
Another possible alternative calculation is to add a small negative imaginary part to
the argument of the square root, for example with the MATLAB code:
7.10 Zenneck Surface Wave
For a lossy medium  , the TM reection coefcient cannot vanish for any
 real incident
angle because the Brewster angle is complex valued: tan B =  / = (R jI )/.
However, TM can vanish if we allow a complex-valued , or equivalently, a complex-
valued incident wavevector k = j , even though the left medium is lossless. This
leads to the so-called Zenneck surface wave [32,902,903,909,1293].
The corresponding constant phase and amplitude planes in both media are shown
in Fig. 7.10.1. On the lossless side, the vectors and are necessarily orthogonal to
each other, as discussed in Sec. 2.11.

kz = sqrt(DR-j*DI-j*realmin);

where realmin is MATLABs smallest positive oating point number (typically, equal
to 2.2251 10308 ). This works well for all cases. Yet, a third alternative is to use
Eq. (7.9.6) and then reverse the signs whenever DI = 0 and DR < 0, for example:

kz = sqrt(DR-j*DI);
kz(DI==0 & DR<0) = -kz(DI==0 & DR<0);

Next, we discuss briey the energy ux into the lossy medium. It is given by the z-
component of the Poynting vector, Pz = 2 z Re(E H ). For the TE case of Eq. (7.7.4),
1

we nd at the two sides of the interface:

|E0 |2   |E0 |2   Fig. 7.10.1 Constant-phase and constant-amplitude planes for the Zenneck wave.
Pz = kz 1 |TE |2 , Pz = |TE |2 e2z z (7.9.11)
20 20 z
We note that the TE reection coefcient can never vanish (unless =  ) because
where we replaced TE = 0 /kz and TE = 0 /kz . Thus, the transmitted power
this would require that kz = kz , which together with Snels law kx = kx , would imply
attenuates with distance as the wave propagates into the lossy medium.
that k = k , which is impossible for distinct media.
The two expressions match at the interface, expressing energy conservation, that is,
For the TM case, the elds are given by Eq. (7.7.5) with TM = 0 and TM = 1. The
at z = 0, we have Pz = Pz , which follows from the condition (see Problem 7.7):
condition TM = 0 requires that kz  = kz  , which may be written in the equivalent form
  kz k2 = kz k2 . Together with k2x + k2z = k2 and k2x + kz2 = k2 , we have three equations
kz 1 |TE |2 = z |TE |2 (7.9.12)
in the three complex unknowns kx , kz , kz . The solution is easily found to be:
Because the net energy ow is to the right in the transmitted medium, we must have
kk k2 k2
z 0. Because also kz > 0, then Eq. (7.9.12) implies that |TE | 1. For the case of kx = 2 , kz = 2 , kz = (7.10.1)
k + k2 k + k2 k2+ k2
this has been xed in versions > v7.0.
7.10. Zenneck Surface Wave 271 272 7. Oblique Incidence


where k = 0  and k =  j = 0  . These may be written in the form: If both media are lossless, then both k and k are real and Eqs. (7.10.1) yield the
 usual Brewster angle formulas, that is,
  
kx = 0 , kz = 0 , kz = 0 (7.10.2) kx k  kx k 
 +   +   +  tan B = = = , tan B = = =
kz k  kz k 
Using kx = kx , the space-dependence of the elds at the two sides is as follows:
Example 7.10.1: For the data of the air-water interface of Example 7.9.1, we calculate the fol-
ej(kx x+kz z) = e(x x+z z) ej(x x+z z) , for z 0 lowing Zenneck wavenumbers at 1 GHz and 100 MHz using Eq. (7.10.2):
j(kx x+kz z) (x x+z z) j(x x+z z)
e =e e , for z 0 f = 1 GHz f = 100 MHz
kx = x jx = 20.89 0.064j kx = x jx = 2.1 0.001j
Thus, in order for the elds not to grow exponentially with distance and to be con- kz = z jz = 1.88 + 0.71j kz = z jz = 0.06 + 0.05j
ned near the interface surface, it is required that: kz = z jz = 202.97 77.80j kz = z jz = 42.01 37.59j

x > 0 , z < 0 , z > 0 (7.10.3)

The units are in rads/m. As required, z is negative. We observe that x  |z | and that
the attenuations are much more severe within the lossy medium.

These conditions are guaranteed with the sign choices of Eq. (7.10.2). This can be
veried by writing

 = | |ej 7.11 Surface Plasmons

 +  = | +  |ej1 Consider an interface between two non-magnetic semi-innite media 1 and 2 , as shown
  in Fig. 7.11.1 The wavevectors k1 = x kx + z kz1 and k2 = x kx + z kz2 at the two sides
    j( )
=
e 1 must have a common kx component, as required by Snels law, and their z-components
+  +  
must satisfy:
and noting that 2 = 1 > 0, as follows by inspecting the triangle formed by the k2z1 = k20 1 k2x , k2z2 = k20 2 k2x (7.11.1)
three vectors ,  , and  +  . Then, the phase angles of kx , kz , kz are 2 /2, 1 /2,
where we dened the relative dielectric constants 1 = 1 /0 , 2 = 2 /0 , and the free-
and (2 + 1 /2), respectively, thus, implying the condition (7.10.3). In drawing this
space wavenumber k0 = 0 0 = /c0 . The TM reection coefcient is given by:
triangle, we made the implicit assumption that R > 0, which is valid for typical lossy
dielectrics. In the next section, we discuss surface plasmons for which R < 0. kz2 1 kz1 2
Although the Zenneck wave attenuates both along the x- and z-directions, the atten-
TM =
kz2 1 + kz1 2
uation constant along x tends to be much smaller than that along z. For example, in the
weakly lossy approximation, we may write  = R (1 j), where = I /R  1 is the
loss tangent of  . Then, we have the following rst-order approximations in :
   
1 1 R
 = R 1 j , =  1+j
2  +   + R 2  + R

These lead to the rst-order approximations for kx and kz :

    
  R

  R
kx = 0 1 j , kz = 0  1+j
 + R 2  + R  + R 2  + R

It follows that:
 
 
 R   R x 
x = 0 , z = 0  =
 + R 2  + R  2  + 
 + R R |z | R
Fig. 7.11.1 Brewster-Zenneck (TM = 0) and surface plasmon (TM = ) cases.
Typically, R > , implying that x < |z |. For example, for an air-water interface
we have at microwave frequencies R / = 81, and for an air-ground interface, R / = 6.
7.11. Surface Plasmons 273 274 7. Oblique Incidence

Both the Brewster case for lossless dielectrics and the Zenneck case were charac- It can be veried easily that these are solutions of Maxwells equations provided
terized by the condition TM = 0, or, kz2 1 = kz1 2 . This condition together with that Eqs. (7.11.1) are satised. The boundary conditions are also satised. Indeed, the
Eqs. (7.11.1) leads to the solution (7.10.2), which is the same in both cases: Ex components are the same from both sides, and the conditions 1 Ez1 = 2 Ez2 and
 Hy1 = Hy2 are both equivalent to the pole condition kz2 1 = kz1 2 .
1 2 k0 1 k0 2
kx = k0 , kz1 = , kz2 = (7.11.2) The conditions (7.11.3) guarantee  that kx is real and kz1 , kz2 , pure imaginary.
 Setting
1 + 2 1 + 2 1 + 2
2 = 2r with 2r > 1 , we have 1 + 2 = 1 2r = j 2r 1 , and 1 2 =

Surface plasmons or polaritons are waves that are propagating along the interface 1 2r = j 1 2r . Then, Eqs. (7.11.5) read
and attenuate exponentially perpendicularly to the interface in both media. They are 
characterized by a pole of the reection coefcient, that is, TM = . For such waves to 1 2r k0 1 k0 2r
k x = k0 , kz1 = j , kz2 = j (7.11.7)
exist, it is necessary to have the conditions: 2r 1 2r 1 2r 1

1 2 < 0 and 1 + 2 < 0 (7.11.3) Setting kz1 = jz1 and kz2 = jz2 , with both s positive, the z-dependence at
at least for the real-parts of these quantities, assuming their imaginary parts are small. both sides of the interface at z = 0 will be:
If the left medium is an ordinary lossless dielectric 1 > 0, such as air, then we must 
ejkz1 z = ez1 z  ejkz2 z = ez2 z
have 2 < 0 and more strongly 2 < 1 . Conductors, such as silver and gold, have this
property for frequencies typically up to ultraviolet. Indeed, using the simple conductiv-
that is, exponentially decaying for both z < 0 and z > 0. Inserting 2r = 2p /2 1
ity model (1.12.3), we have for the dielectric constant of a metal:
into kx gives the so-called plasmon dispersion relationship, For example, if 1 = 1,
0 2p 2p
()= 0 + = 0 + ()= 1 (7.11.4) 2 p
2 2
j j(j + ) 2 j k2x = 2 2
c0 p 22
Ignoring the imaginary part for the moment, we have
2p Dening the normalized variables = /p and k = kx /kp , where kp = p /c0 ,
()= 1 we may rewrite the above relationship as,
2
which is negative for < p . The plasma frequency is of the order of 10002000 THz, 1 2
and falls in the ultraviolet range. Thus, the condition (7.11.3) is easily met for optical k2 = 2
1 22
frequencies. If 1 = 1, then, the condition 2 < 1 requires further that
with solution  
2p p 
2 = 1 < 1 < 
 2 1 1
2 2 = k + k4 + (7.11.8)
2 4
and more generally, < p / 1 + 1 . The condition TM = means that there is only
a reected wave, while the incident eld is zero. Indeed, it follows from Ere = TM Einc , It is depicted in Fig. 7.11.2. In the large kx limit, it converges to the horizontal line

or Einc = Ere /TM , that Einc will tend to zero for nite Ere and TM . = p / 2. For small kx , it becomes the dispersion relationship in vacuum, = c0 kx ,
The condition TM = is equivalent to the vanishing of the denominator of TM , which is also depicted in this gure.
that is, kz2 1 = kz1 2 , which together with Eqs. (7.11.1) leads to a similar solution as Because the curve stays to the right of the vacuum line = c0 kx , that is, kx > /c0 ,
(7.10.2), but with a change in sign for kz2 : such surface plasmon waves cannot be excited by an impinging plane wave on the inter-
 face. However, they can be excited with the help of frustrated total internal reection,
1 2 k 0 1 k0 2 which increases kx beyond its vacuum value and can match the value of Eq. (7.11.7) re-
kx = k0 , kz1 = , kz2 = (7.11.5)
1 + 2 1 + 2 1 + 2 sulting into a so-called surface plasmon resonance. We discuss this further in Sec. 8.5.
The elds at the two sides of the interface are given by Eqs. (7.7.5) by taking the limit In fact, the excitation of such plasmon resonance can only take place if the metal
TM and TM = 1 + TM , which effectively amounts to keeping only the terms side is slightly lossy, that is, when 2 = 2r j2i , with 0 < 2i  2r . In this case, the
that involve TM . The elds have a z-dependence ejkz1 z on the left and ejkz2 z on the wavenumber kx acquires a small imaginary part which causes the gradual attenuation
right, and a common x-dependence ejkx x : of the wave along the surface, and similarly, kz1 , kz2 , acquire small real parts. Replacing
   2r by 2r + j2i in (7.11.7), we now have:
kx  kx
E1 = E0 x + z ejkz1 z ejkx x 
 E2 = E0 x z ejkz2 z ejkx x 
kz1  kz2 1 (2r + j2i ) jk0 1 jk0 (2r + j2i )
 (7.11.6)
1 jkz1 z jkx x  2 jkz2 z jkx x kx = k0 , kz1 =  , kz2 =  (7.11.9)
H1 = y E0 e e  H2 = y E0 e e 2r + j2i 1 2r + j2i 1 2r + j2i 1
kz1  kz2
7.12. Oblique Reection from a Moving Boundary 275 276 7. Oblique Incidence

plasmon dispersion relation

1/
2

/ p

0 1 2 3
kx / kp

Fig. 7.11.2 Surface plasmon dispersion relationship. Fig. 7.12.1 Oblique reection from a moving boundary.

Expanding kx to rst-order in 2i , we obtain the approximations: assumed to be moving with velocity v perpendicularly to the interface, that is, in the
 3/2 z-direction as shown in Fig. 7.12.1. Other geometries may be found in [474492].
1 2r 1 2 r 2i Let S and S be the stationary and the moving coordinate frames, whose coordinates
kx = x jx , x = k0 , x = k0 (7.11.10)
2r 1 2r 1 222r {t, x, y, z} and {t , x , y , z } are related by the Lorentz transformation of Eq. (K.1) of
Appendix K.
Example 7.11.1: Using the value 2 = 16 0.5j for silver at 0 = 632 nm, and air 1 = 1, We assume a TE plane wave of frequency incident obliquely at the moving inter-
we have k0 = 2/0 = 9.94 rad/m and Eqs. (7.11.9) give the following values for the
face at an angle , as measured in the stationary coordinate frame S. Let r , t be
wavenumbers and the corresponding effective propagation length and penetration depths:
the Doppler-shifted frequencies, and r , t , the angles of the reected and transmitted
1 waves. Because of the motion, these angles no longer satisfy the usual Snel laws of
kx = x jx = 10.27 0.0107j rad/m, x = = 93.6 m reection and refractionhowever, the do satisfy modied versions of these laws.
x
1
In the moving frame S with respect to which the dielectric is at rest, we have an
kz1 = z1 jz1 = 0.043 2.57j rad/m, z1 = = 390 nm ordinary TE oblique incidence problem, solved for example by Eq. (7.7.4), and therefore,
z1
all three frequencies will be the same,  = r = t , and the corresponding angles
1
kz2 = z2 jz2 = 0.601 41.12j rad/m, z2 = = 24 nm  , r , t will satisfy the ordinary Snel laws: r =  and sin  = n sin t , where
z2
n = /0 and the left medium is assumed to be free space.
Thus, the elds extend more into the dielectric than the metal, but at either side they are The electric eld has only a y-component and will have the following form at the left
conned to distances that are less than their free-space wavelength.

and right sides of the interface, in the frame S and in the frame S :

Surface plasmons, and the emerging eld of plasmonics, are currently active areas Ey = Ei eji + Er ejr , Ey = Et ejt
(7.12.1)
of study [593631] holding promise for the development of nanophotonic devices and   
Ey = Ei eji + Er ejr , Ey = Et ejt
circuits that take advantage of the fact that plasmons are conned to smaller spaces
than their free-space wavelength and can propagate at decent distances in the nanoscale where Er = TE Ei and Et = TE Ei , and from Eq. (7.7.2),
regime (i.e., tens of m compared to nm scales.) They are also currently used in chemical
and biological sensor technologies, and have other potential medical applications, such kiz ktz cos  n cos t 2 cos 
TE =   = , TE = 1 + TE = (7.12.2)
as cancer treatments. kiz + ktz cos  + n cos t cos  + n cos t

The propagation phases are Lorentz invariant in the two frames and are given by:
7.12 Oblique Reection from a Moving Boundary
i = t kiz z kix x =  t kiz z kix x = i
In Sec. 5.8 we discussed reection and transmission from a moving interface at nor- r = r t + krz z krx x =  t + krz z krx x = r (7.12.3)
mal incidence. Here, we present the oblique incidence case. The dielectric medium is
t = t t ktz z ktx x =  t ktz z ktx x = t
7.12. Oblique Reection from a Moving Boundary 277 278 7. Oblique Incidence

with incident, reected, and transmitted wavenumbers given in the frame S by: For a stationary interface, all the frequency factors drop out and we obtain the or-
dinary Snel laws. The reected and transmitted frequencies are -dependent and are
kiz = krz = ki cos  , ktz = kt cos t
(7.12.4) obtained from (7.12.5) by eliminating  :
kix = krx = ki sin  , ktx = kt sin t
1 cos  1 + n cos t
r = , t = (7.12.10)
where ki = kr =  /c and kt =  0 = n /c. The relationships between 1 + cos  1 + cos 
the primed and unprimed frequencies and wavenumbers are obtained by applying the
Lorentz transformation (K.14) to the four-vectors (/c, kix , 0, kiz ), (r , krx , 0, krz ), Replacing kiz = ki cos = (/c)cos and krz = kr cos r = (r /c)cos r in
and (t /c, ktx , 0, ktz ): Eq. (7.12.5), we obtain the relationship of the angles , r to the angle  :

= ( + ckiz )=  (1 + cos  ) cos  + cos 

cos = , cos r = (7.12.11)
 1 + cos  1 cos 
kiz = (kiz
+  )= (cos  + )
c c which can also be written as:
r = ( ckrz )=  (1 cos  )
cos cos r +
  (7.12.5) cos  = = (7.12.12)
krz = (krz + )= (cos  ) 1 cos 1 + cos r
c c
t = ( + cktz )=  (1 + n cos t ) Solving for r in terms of , we obtain:

ktz = (ktz +  )= (n cos t + )
c c (1 + 2 )cos 2
 cos r = (7.12.13)
1 2 cos + 2
where = v/c and = 1/ 1 2 . Combining Snels laws for the system S with the
invariance of the x-components of the wavevector under the Lorentz transformation
(K.14), we have also: Inserting cos  in Eq. (7.12.10), we nd the reected frequency in terms of :

kix = krx = ktx = kix = krx = ktx 1 2 cos + 2

r = (7.12.14)
   (7.12.6) 1 2
ki sin = kr sin r = kt sin t = sin  = sin r = n sin t
c c c
Because the incident and reected waves are propagating in free space, their wavenum- Eqs. (7.12.13) and (7.12.14) were originally derived by Einstein in his 1905 paper on
bers will be ki = /c and kr = r /c. This also follows from the invariance of the scalar special relativity [474]. The quantity n cos t can also be written in terms of . Using
(/c)2 k2 under Lorentz transformations. Indeed, because ki = kr =  /c in the S Snels law and Eq. (7.12.12), we have:
system, we will have: 
   2
  cos
 2   
n cos t = n sin = n 1 + cos = n 1 +
2 2 2r 2 2 2 2 2 , or,
2
k2i = 2 ki 2 = 0 , 2
k2r = 2 kr2 = 0 1 cos
c c c c

For the transmitted wavenumber kt , we nd from Eqs. (7.12.5) and (7.12.6): (n2 1)(1 cos )2 +(cos )2 Q
  n cos t = (7.12.15)
 1 cos 1 cos
kt = k2tz + k2tx = 2 (n cos t + )2 +n2 sin2 t (7.12.7)
c Using (7.12.15) and the identity (1 + cos  )(1 cos )= 1 2 , we nd for the
Setting vt = t /kt = c/nt , we obtain the effective refractive index nt within the transmitted frequency:
moving dielectric medium:
 1 + n cos t 1 cos + Q
t = = (7.12.16)
c ckt 2 (n cos t + )2 +n2 sin2 t 1 + cos  1 2
nt = = = (7.12.8)
vt t (1 + n cos t )
The TE reection coefcient (7.12.2) may also be expressed in terms of :
At normal incidence, this is equivalent to Eq. (5.8.6). Replacing ki = /c, kr = r /c,
and kt = t nt /c in Eq. (7.12.6), we obtain the generalization of Snels laws: cos  n cos t cos Q
TE = = (7.12.17)
cos  + n cos t cos + Q
sin = r sin r = t nt sin t =  sin  =  sin r =  n sin t (7.12.9)
7.13. Geometrical Optics 279 280 7. Oblique Incidence


Next, we determine the reected and transmitted elds in the frame S. The simplest where n is the refractive index of the medium n = /0 , k0 and 0 are the free-space
approach is to apply the Lorentz transformation (K.30) separately to the incident, re- wavenumber and impedance, and k, the unit-vector in the direction of propagation.
ected, and transmitted waves. In the S frame, a plane wave propagating along the unit The wavefronts are dened to be the constant-phase plane surfaces S(r)= const.,

vector k has magnetic eld: where S(r)= n k r. The perpendiculars to the wavefronts are the optical rays.
In an inhomogeneous medium with a space-dependent refractive index n(r), the
1  0  
H = k E  cB  = c0 H  = k E  = n k E  (7.12.18) wavefronts and their perpendicular rays become curved, and can be derived by consid-
ering the high-frequency limit of Maxwells equations. By analogy with Eqs. (7.13.1), we
where n = 1 for the incident and reected waves. Because we assumed a TE wave and look for solutions of the form:
the motion is along the z-direction, the electric eld will be perpendicular to the velocity,
E(r)= E0 (r) ejk0 S(r) , H(r)= H0 (r) ejk0 S(r) (7.13.2)
that is, E  = 0. Using the BAC-CAB rule, Eq. (K.30) then gives:
   where we will assume that k0 is large and that E0 , H0 are slowly-varying functions of r.
E = E = (E cB )= (E  cB  )= E  (n k E  )
(7.12.19) This means that their space-derivatives are small compared to k0 or to 1/. For example,
     E0 |  k0 .
|
= E  n(
E )k + n(
k )E = E  (1 + n k )
Inserting these expressions into Maxwells equations and assuming = 0 and  =
Applying this result to the incident, reected, and transmitted elds, we nd: n2 0 , we obtain:
 
Ei = Ei (1 + cos  ) E = ejk0 S E0 jk0 S E0 = j0 H0 ejk0 S
Er = Er (1 cos  )= TE Ei (1 cos  ) (7.12.20)  
H = ejk0 S H0 jk0 S H0 = jn2 0 E0 ejk0 S
Et = Et (1 + n cos t )= TE Ei (1 + n cos t )
Assuming | E0 |  |k0 S E0 |, and similarly for H0 , and dropping the common
It follows that the reection and transmission coefcients will be: phase factor ejk0 S , we obtain the high-frequency approximations:
Er 1 cos  r Et 1 + n cos t t
= TE = TE , = TE = TE (7.12.21) jk0 S E0 = j0 H0
Ei 1 + cos  Ei 1 + cos 
jk0 S H0 = jn2 0 E0
The case of a perfect mirror corresponds to TE = 1 and TE = 0. To be interpretable
as a reection angle, r must be in the range 0 r 90o , or, cos r > 0. This requires 1
Replacing k0 = 0 0 , and dening the vector k = S, we nd:
that the numerator of (7.12.13) be positive, or, n

2  2  n 0
(1 + 2 )cos 2 0  cos  acos (7.12.22) H0 = k E0 , E0 = k H0 (7.13.3)
1 + 2 1 + 2 0 n

Because 2/(1 + 2 )> , (7.12.22) also implies that cos > , or, v < cz = c cos .
These imply the transversality conditions k E0 = k H0 = 0. The consistency of the
Thus, the z-component of the phase velocity of the incident wave can catch up with the
equations (7.13.3) requires that k be a unit vector. Indeed, using the BAC-CAB rule, we
receding interface. At the maximum allowed , the angle r reaches 90o . In the above,
have:
we assumed that > 0. For negative , there are no restrictions on the range of . 0
k (k E0 )= k(k E0 )E0 (k k)= E0 (k k)= k H0 = E0
n
Thus, we obtain the unit-vector condition, known as the eikonal equation:
7.13 Geometrical Optics
k k = 1 S|2 = n2
| (eikonal equation) (7.13.4)
Geometrical optics and the concepts of wavefronts and rays can be derived from Maxwells
equations in the short-wavelength or high-frequency limit. This equation determines the wavefront phase function S(r). The rays are the per-
We saw in Chap. 2 that a uniform plane wave propagating in a lossless isotropic pendiculars to the constant-phase surfaces S(r)= const., so that they are in the direction
dielectric in the direction of a wave vector k = k k = nk0 k is given by: of S or k. Fig. 7.13.1 depicts these wavefronts and rays.
n The ray passing through a point r on the surface S(r)= SA , will move ahead by a
E(r)= E0 ejnk0 kr , H(r)= H0 ejnk0 kr , k E0 = 0 , H0 = k E0 (7.13.1) distance dr in the direction of the gradient S. The length of dr is dl = (dr dr)1/2 .
0
7.13. Geometrical Optics 281 282 7. Oblique Incidence

Thus, the energy transport velocity is:

P c0
v= = k (7.13.9)
w n

The velocity v depends on r, because n and k do.

7.14 Fermats Principle

An innitesimal movement by dl along a ray will change the wavefront phase function
by dS = ndl. Indeed, using Eq. (7.13.6) and the eikonal equation we nd:
Fig. 7.13.1 Wavefront surfaces and rays.
dS dr 1 1
= S = S S = n2 = n (7.14.1)
The vector dr/dl is a unit vector in the direction of S and, therefore, it must be equal dl dl n n
to k. Thus, we obtain the dening equation for the rays:
Integrating along a ray path from a point A on wavefront S(r)= SA to a point B on
dr dr 1 dr wavefront S(r)= SB , as shown in Fig. 7.13.1, gives rise to the net phase change:
= k = S n = S (7.13.5)
dl dl n dl B B
The eikonal equation determines S, which in turn determines the rays. The ray SB SA = dS = ndl (7.14.2)
A A
equation can be expressed directly in terms of the refractive index by eliminating S.
Indeed, differentiating (7.13.5), we have: The right-hand side is recognized as the optical path length from A to B. It is pro-
 
d dr d dr 1  portional to the travel time of moving from A to B with the ray velocity v given by
n = S)=
( S = S S
dl dl dl dl n Eq. (7.13.9). Indeed, we have dl = v k dt = c0 dt/n, or, dS = ndl = c0 dt. Thus,
where, in differentiating along a ray, we used the expression for d/dl: B  tB
d dr SB SA = ndl = c0 dt = c0 (tB tA ) (7.14.3)
= (7.13.6) A tA
dl dl
   
But, S S = 2 S S, which follows from the differential identity Fermats Principle states that among all possible paths connecting the two points A
Eq. (C.16) of the Appendix. Therefore, and B, the geometrical optics ray path is the one the minimizes the optical path length

d dr 1  1   1 1 (7.14.3), or equivalently, the travel time between the two points. The solution to this
n = S S = S S = (n2 )= n , or,
2n
dl dl n 2n 2n 2n minimization problem is the ray equation (7.13.7).
 Any path connecting the points A and B may be specied parametrically by the curve
d dr r(), where the parameter varies over an interval A B . The length dl may be
n = n (ray equation) (7.13.7)
dl dl written as:
 1/2  1/2 dr
The vectors E0 , H0 , k form a right-handed system as in the uniform plane-wave case. dl = dr dr = r r d , where r = (7.14.4)
d
The energy density and ux are:
Then, the functional to be minimized is:
1 1  1 B  B
we = Re  E E = 0 n2 |E0 |2  1/2
2 2 4 ndl = L(r, r) d , where L(r, r)= n(r) r r (7.14.5)
A A
1 1 n2 1
wm = 0 |H0 |2 = 0 2 |E0 |2 = 0 n2 |E0 |2 = we The minimization of Eq. (7.14.5) may be viewed as a problem in variational calculus
4 4 0 4
(7.13.8) with Lagrangian function L. Its solution is obtained from the Euler-Lagrange equations:
1
w = we + wm = 0 n2 |E0 |2 
2 d L L
= (7.14.6)
1   n d r r
P = Re E H = k |E0 |2
2 20
7.14. Fermats Principle 283 284 7. Oblique Incidence

See [868870] for a review of such methods. The required partial derivatives are:

L n  1/2 L  1/2 dr  1/2

= r r , = n r r r =n r r
r r r d
The Euler-Lagrange equations are then:

d dr  1/2 n  1/2
n r r = r r or,
d d r

 1/2 d dr  1/2 n
r r n r r = (7.14.7)
d d r
 1/2
Using dl = r r d, we may rewrite these in terms of the length variable dl,
resulting in the same equations as (7.13.7), that is,

d dr n
n = (7.14.8)
dl dl r
Fig. 7.14.1 Snels laws of reection and refraction.
A variation of Fermats principle states that the phase change between two wave-
front surfaces is independent of the choice of the ray path taken between the surfaces.
Following a different ray between points A and B , as shown in Fig. 7.13.1, gives the Thus, we obtain Snels law of refraction:
same value for the net phase change as between the points A and B:
La sin a nb
B  B = = na sin a = nb sin b
Lb sin b na

SB SA = ndl = ndl (7.14.9)
A A
7.15 Ray Tracing
This form is useful for deriving the shapes of parabolic reector and hyperbolic lens
antennas discussed in Chap. 21. In this section, we apply Fermats principle of least optical path to derive the ray curves
It can also be used to derive Snels law of reection and refraction. Fig. 7.14.1 shows in several integrable examples of inhomogeneous media.
the three families of incident, reected, and refracted plane wavefronts on a horizontal As a special case of Eq. (7.14.8), we consider a stratied half-space z 0, shown in
interface between media na and nb , such that the incident, reected, and refracted rays Fig. 7.15.1, in which the refractive index is a function of z, but not of x.
are perpendicular to their corresponding wavefronts.
For the reection problem, we consider the ray paths between the wavefront surfaces
A0 A1 and A1 A2 . Fermats principle implies that the optical path length of the rays
AOA , A0 A0 , and A2 A2 will be the same. This gives the condition:

na (la + la )= na L = na L L = L

where L and L are the lengths of the rays A0 A0 and A2 A2 . It follows that the two
triangles A0 A2 A2 and A0 A0 A2 will be congruent. and therefore, their angles at the
vertices A0 and A2 will be equal. Thus, a = a .
For the refraction problem, we consider the ray paths AOB, A0 B0 , and A1 B1 between
the wavefronts A0 A1 and B0 B1 . The equality of the optical lengths gives now:

La nb
na la + nb lb = nb Lb = na La =
Lb na
But, the triangles A0 A1 B1 and A0 B0 B1 have a common base A0 B1 . Therefore,
Fig. 7.15.1 Rays in an inhomogeneous medium.
La sin a
=
Lb sin b
7.15. Ray Tracing 285 286 7. Oblique Incidence

Let be the angle formed by the tangent on the ray at point (x, z) and the vertical. The length za can be greater or less than z. For example, if the upper half-space is
Then, we have from the gure dx = dl sin and dz = dl cos . Because n/x = 0, the homogeneous with nb < na , then za > z. If nb > na , then za < z, as was the case in
ray equation (7.14.8) applied to the x-coordinate reads: Example 7.5.4.
 Next, we discuss a number of examples in which the integral (7.15.6) can be done
d dx dx
n =0 n = const. n sin = const. (7.15.1) explicitly to derive the ray curves.
dl dl dl
This is the generalization of Snels law to an inhomogeneous medium. The constant Example 7.15.1: Ionospheric Refraction. Radio waves of frequencies typically in the range of
may be determined by evaluating it at the entry point z = 0 and x = 0. We take the about 440 MHz can be propagated at large distances such as 20004000 km by bouncing
constant to be na sin a . Thus, we write (7.15.2) as: off the ionosphere. Fig. 7.15.2 depicts the case of a at ground.

n(z)sin (z)= na sin a (generalized Snels law) (7.15.2)

The z-component of the ray equation is, using dz = dl cos :


d dz dn d dn
n = cos (n cos ) = (7.15.3)
dl dl dz dz dz
This is a consequence of Eq. (7.15.2). To see this, we write:
 
n cos = n2 n2 sin2 = n2 n2a sin2 a (7.15.4)

Differentiating it with respect to z, we obtain Eq. (7.15.3). The ray in the left Fig. 7.15.1
is bending away from the z-axis with an increasing angle (z). This requires that n(z) Fig. 7.15.2 Ionospheric refraction.
be a decreasing function of z. Conversely, if n(z) is increasing as in the right gure,
then (z) will be decreasing and the ray will curve towards the z-axis. The atmosphere has a typical extent of 600 km and is divided in layers: the troposphere up
Thus, we obtain the rule that a ray always bends in the direction of increasing n(z) to 10 km, the stratosphere at 1050 km, and the ionosphere at 50600 km. The ionosphere
and away from the direction of decreasing n(z). is further divided in sublayers, such as the D, E, F1 , and F2 layers at 50100 km, 100150
The constants na and a may be taken to be the launch values at the origin, that km, 150250 km, and 250400 km, respectively.
is, n(0) and (0). Alternatively, if there is a discontinuous change between the lower The ionosphere consists mostly of ionized nitrogen and oxygen at low pressure. The
and upper half-spaces, we may take na , a to be the refractive index and incident angle ionization is due to solar radiation and therefore it varies between night and day. We
from below. recall from Sec. 1.15 that a collisionless plasma has an effective refractive index:
The ray curves can be determined by relating x and z. From Fig. 7.15.1, we have
dx = dz tan , which in conjunction with Eqs. (7.15.2) and (7.15.4) gives: () 2p Ne2
n2 = =1 2 , 2p = (7.15.8)
0 0 m
dx n sin na sin a
= tan = =  (7.15.5) The electron density N varies with the time of day and with height. Typically, N increases
dz n cos n2 (z)n2a sin2 a
through the D and E layers and reaches a maximum value in the F layer, and then decreases
Integrating, we obtain: after that because, even though the solar radiation is more intense, there are fewer gas
atoms to be ionized.
z
na sin a Thus, the ionosphere acts as a stratied medium in which n(z) rst decreases with height
x=  dz (ray curve) (7.15.6)
from its vacuum value of unity and then it increases back up to unity. We will indicate the
0 n2 (z )n2a
2
sin a
dependence on height by rewriting Eq. (7.15.8) in the form:

An object at the point (x, z) will appear to an observer sitting at the entry point O
fp2 (z) N(z)e2
as though it is at the apparent location (x, za ), as shown in Fig. 7.15.1. The apparent or n2 (z)= 1 , fp2 (z)= (7.15.9)
f2 42 0 m
virtual height will be za = x cot a , which can be combined with Eq. (7.15.6) to give:
If the wave is launched straight up and its frequency f is larger than the largest fp , then
z it will penetrate through the ionosphere and be lost. But, if there is a height such that
na cos a 
za =  dz (virtual height) (7.15.7) f = fp (z), then at that height n(z)= 0 and the wave will be reected back down.
0 n2 (z )n2a sin2 a
7.15. Ray Tracing 287 288 7. Oblique Incidence

If the wave is launched at an angle a , then it follows from Snels law that while the
1
refractive index n(z) is decreasing, the angle of refraction (z) will be increasing and the z z0 = (x x0 )2 (7.15.16)
ray path will bend more and more away from z-axis as shown on the left of Fig. 7.15.1. 4F
Below the ionosphere, we may assume that the atmosphere has refractive index na = 1. where
2zmax sin a cos a zmax sin2 a
Then, the angle (z) may be written as: x0 = , F=
a2 a2

n2a sin2 a sin2 a Therefore, the ray follows a downward parabolic path with vertex at (x0 , z0 ) and focal
sin2 (z)= = (7.15.10) length F, as shown in Fig. 7.15.3.

n2 (z) fp2 (z)
1
f2

Because sin (z) is required to be less than unity, we obtain the restriction:

fp2 (z)
sin2 a 1 fp (z) f cos a (7.15.11)
f2

If there is a height, say zmax , at which this becomes an equality, fp (zmax )= f cos a , then
Eq. (7.15.10) would imply that sin (zmax )= 1, or that (zmax )= 90o . At that height, the
ray is horizontal and it will proceed to bend downwards, effectively getting reected from
the ionosphere.
If f is so large that Eq. (7.15.11) is satised only as a strict inequality, then the wave will
escape through all the layers of the ionosphere. Thus, there is a maximum frequency, the Fig. 7.15.3 Parabolic ray.
so called maximum usable frequency (MUF), that will guarantee a reection. There is also a
lowest usable frequency (LUF) below which there is too much absorption of the wave, such
as in the D layer, to be reected at sufcient strength for reception. Example 7.15.2: Mirages. Temperature gradients can cause several types of mirage effects that
are similar to ionospheric refraction. On a hot day, the ground is warmer than the air above
As an oversimplied, but analytically tractable, model of the ionosphere we assume that
it and therefore, the refractive index of the air is lower at the ground than a short distance
the electron density increases linearly with height, up to a maximal height zmax . Thus, the
above. (Normally, the air pressure causes the refractive index to be highest at the ground,
quantities fp2 (z) and n2 (z) will also depend linearly on height:
decreasing with height.)

z 2
fmax z Because n(z) decreases downwards, a horizontal ray from an object near the ground will
fp2 (z)= fmax
2
, n2 (z)= 1 , for 0 z zmax (7.15.12) initially be refracted downwards, but then it will bend upwards again and may arrive at an
zmax f 2 zmax
observer as though it were coming from below the ground, causing a mirage. Fig. 7.15.4
Over the assumed height range 0 z zmax , the condition (7.15.11) must also be satised. depicts a typical case. The ray path is like the ionospheric case, but inverted.
This restricts further the range of z. We have:
Such mirages are seen in the desert and on highways, which appear wet at far distances.
Various types of mirages are discussed in [5052,1827].
z z f 2 cos2 a
fp2 (z)= fmax
2
f 2 cos2 a 2
(7.15.13)
As a simple integrable model, we may assume that n(z) increases linearly with height z,
zmax zmax fmax
that is, n(z)= n0 + z, where is the rate of increase per meter. For heights near the
If the right-hand side is greater than unity, so that f cos a > fmax , then there is no height ground, this implies that n2 (z) will also increase linearly:
z at which (7.15.11) achieves an equality, and the wave will escape. But, if f cos a fmax ,
then there is height, say z0 , at which the ray bends horizontally, that is,
n(z)= n0 + z n2 (z)= n20 + 2n0 z (7.15.17)
2 2 2 2
z0 f cos a zmax f cos a
= 2
z0 = 2
(7.15.14) We consider a ray launched at a downward angle a from an object with (x, z) coordinates
zmax fmax fmax
(0, h), as shown. Let n2a = n20 + 2n0 h be the refractive index at the launch height. For
The condition f cos a fmax can be written as f fMUF , where the MUF is in this case, convenience, we assume that the observer is also at height h. Because the ray will travel
fMUF = fmax / cos a . The integral (7.15.6) can be done explicitly resulting in: downward to points z < h, and then bend upwards, we integrate the ray equation over the
limits [z, h] and nd:

 
2zmax sin2 a z
x= cos a cos2 a a2 (7.15.15) h   
a2 zmax na sin a na sin a
x=  dz = na cos a n2a cos2 a + 2n0 (z h)
z n2 (z )na sin a
2 2 n0
where we dened a = fmax /f . Solving for z in terms of x, we obtain:
7.15. Ray Tracing 289 290 7. Oblique Incidence

where we used the approximation n2 (z)= n20 + 2n0 z in the integral. Solving for z in
terms of x, we obtain the parabolic ray:

x(x 2x0 ) d n2a sin a cos a n2a sin2 a

z=h+ , x0 = = , F=
4F 2 n0 2n0

where d is the distance to the observer and F is the focal length. The apex of the parabola
is at x = x0 = d/2 at a height z0 given by:

x20 1
z0 = h z z0 = (x x0 )2
4F 4F

Fig. 7.15.5 Atmospheric refraction.

Fig. 7.15.4 Mirage due to a temperature gradient.
The launch angle that results in the ray being tangential to ground is obtained by setting
the apex height to zero, z0 = 0. This gives a condition that may be solved for a :
The bending of the light rays as they pass through the atmosphere cause the apparent
displacement of a distant object, such as a star, the sun, or a geosynchronous satellite.
Fig. 7.15.5 illustrates this effect. The object appears to be closer to the zenith.
The look-angle 0 at the ground and the true angle of the object 1 are related by Snels
law n1 sin 1 = n0 sin 0 . But at large distances (many multiples of hc ), we have n1 = 1.
Therefore,
sin 1 = n0 sin 0 (7.15.20)
The refraction angle is r = 1 0 . Assuming a small r , we may use the approximation
sin(0 + r)= sin 0 + r cos 0 . Then, Eq. (7.15.20) gives the approximate expression:
r = (n0 1)tan 0

 n0 n0 2hn0 The maximum viewing angle in this model is such that n0 sin 0 = sin 1 = 1, correspond-
x0 = 4Fh sin a = F= x0 = (7.15.18) ing to 1 = 90o and 0 = asin(1/n0 )= 88.6o , for n0 = 1.0003.
na 2
The model assumes a at Earth. When the curvature of the Earth is taken into account, the
The corresponding d = 2x0 is the maximum distance of the observer from the object for total atmospheric refraction near the horizon, that is, near 0 = 90o , is about 0.65o for a
which a ray can just touch the ground.

sea-level observer [50]. The setting sun subtends an angle of about 0.5o . Therefore, when
it appears about to set and its lower edge is touching the horizon, it has already moved
Example 7.15.3: Atmospheric Refraction [5052]. Because of the compression of gravity, the below the horizon.
density of the atmosphere and its refractive index n are highest near the ground and
decrease exponentially with height. A simplied model [721], which assumes a uniform The model of Eq. (7.15.19) may be integrated exactly. The ray curves are obtained from
temperature and constant acceleration of gravity, is as follows: Eq. (7.15.6). Setting na = n0 , a = 0 and using the denition (7.15.20), we obtain:


 

A A0 A+B
n(z)= 1 + (n0 1)ez/hc (7.15.19) x = hc tan 1 atanh atanh = tan 1 z + hc ln (7.15.21)
B B0 A0 + B0

The refractive index on the ground is approximately n0 = 1.0003 (it also has some de- where the quantities A, B, A0 , B0 are dened as follows:
pendence on wavelength, which we ignore here.) The characteristic height hc is given by
hc = RT/Mg, where R, T, M, g are the universal gas constant, temperature in absolute A = n(z) sin2 1 , A0 = n0 sin2 1
 
units, molecular mass of the atmosphere and acceleration of gravity: B = cos 1 n2 (z) sin2 1 , B0 = cos 1 n20 sin2 1
J kg m
R = 8.31 , M = 0.029 , g = 9 .8 Thus, A0 , B0 are the values of A, B at z = 0. It can be shown that A > B and therefore, the
K mole mole s2
hyperbolic arc-tangents will be complex-valued. However, the difference of the two atanh
For a temperature of T = 303K, or 30 o C, we nd a height of hc = 8.86 km. At a height of The troposphere and some of the stratosphere, consisting mostly of molecular nitrogen and oxygen.
a few hc , the refractive index becomes unity.
7.15. Ray Tracing 291 292 7. Oblique Incidence

terms is real and can be transformed into the second expression in (7.15.21) with the help Combining with Snels law, we obtain:
of the result A2 B2 = (A20 B20 )e2z/hc .
In the limit of z  hc , the quantities A, B tend to A1 = B1 = cos2 1 . and the ray equation r2 n2 sin 2 = r3 n2 sin 2 = r3 n3 sin 3
becomes the straight line with a slope of tan 1 :
 Thus, the product ri ni sin i is the same for all i = 0, 1, 2, . . . . Dening an effective refrac-
A1 + B 1
x = (z + z1 )tan 1 , z1 = hc ln (7.15.22) tive index by neff (r)= n(r)r/r0 , Bouguers law may be written as Snels law:
A0 + B 0

This asymptotic line is depicted in Fig. 7.15.5, intercepting the z-axis at an angle of 1 .

neff (r)sin (r)= n0 sin 0

Example 7.15.4: Bouguers Law. The previous example assumed a at Earth. For a spherical where we have the initial value neff (r0 )= n0 r0 /r0 = n0 .

Earth in which the refractive index is a function of the radial distance r only, that is, n(r),
the ray tracing procedure must be modied. Example 7.15.5: Standard Atmosphere over Flat Earth. For radiowave propagation over ground,
the International Telecommunication Union (ITU) [877,878] denes a standard atmo-
Snels law n(z)sin (z)= n0 sin 0 must be replaced by Bouguers law [638], which states
sphere with the values n0 = 1.000315 and hc = 7.35 km, in Eq. (7.15.19).
that the quantity rn(r)sin remain constant:
For heights of about one kilometer, such that z  hc , we may linearize the exponential,
ez/hc = 1 z/hc , and obtain the refractive index for the standard atmosphere:
rn(r)sin (r)= r0 n(r0 )sin 0 (Bouguers law) (7.15.23)

where (r) is the angle of the tangent to the ray and the radial vector. This law can be n0 1 315 106
n(z)= n0 z , = = = 4.2857 108 m1 (7.15.24)
derived formally by considering the ray equations in spherical coordinates and assuming hc 7.35 103
that n(r) depends only on r [869]. This is similar to Eq. (7.15.17), with the replacement . Therefore, we expect the
A simpler derivation is to divide the atmosphere in equal-width spherical layers and assume rays to be parabolic bending downwards as in the case of the ionosphere. A typical ray
that the refractive index is homogeneous in each layer. In Fig. 7.15.6, the layers are dened between two antennas at height h and distance d is shown in Fig. 7.15.7.
by the radial distances and refractive indices ri , ni , i = 0, 1, 2, . . . .

Fig. 7.15.7 Rays in standard atmosphere over a at Earth.

Assuming an upward launch angle a and dening the refractive index na at height h
through n2a = n20 2n0 h, we obtain the ray equations by integrating over [h, z]:

z   
na sin a na sin a
x=  dz = na cos a n2a cos2 a 2n0 (z h)
h 2  2 2
n (z )na sin a n0

Fig. 7.15.6 Ray tracing in spherically stratied medium.

where we used n2 (z)= n20 2n0 z. Solving for z, we obtain the parabola:
For sufciently small layer widths, the ray segments between the points A0 , A1 , A2 , . . .
are tangential to the radial circles. At the interface point A3 , Snels law gives n2 sin 2 = x(x 2x0 ) d n2a sin a cos a n2a sin2 a
z=h , x0 = = , F=
n3 sin 3 . On the other hand, from the triangle OA2 A3 , we have the law of sines: 4F 2 n0 2n0

r2 r3 r3 where d is the distance to the observer and F is the focal length. The apex of the parabola
= = r2 sin 2 = r3 sin 2
sin 2 sin( 2 ) sin 2 is at x = x0 = d/2 at a height z0 given by:
7.15. Ray Tracing 293 294 7. Oblique Incidence

x20 1
z0 = h + z z0 = (x x0 )2
4F 4F

The minus sign in the right-hand side corresponds to a downward parabola with apex at
the point (x0 , z0 ).

Example 7.15.6: Standard Atmosphere over Spherical Earth. We saw in Example 7.15.4 that
in Bouguers law the refractive index n(r) may be replaced by an effective index ne (r)=
n(r)r/r0 . Applying this to the case of the Earth with r0 = R and r = R + z, where R is
the Earth radius and z the height above the surface, we have ne (z)= n(z)(R + z)/R, or, Fig. 7.15.8 Rays over a spherical Earth.
 
z z
ne (z)= n(z) 1 + = (n0 z) 1 +
R R On the other hand, because h  Re the arc length x0 = (OB) may be taken to be a straight
line in Fig. 7.15.8. Applying the Pythagorean theorem to the two orthogonal triangles OAB
Thus, the spherical Earth introduces the factor (1 + z/R), which increases with height and
and CAB we nd that:
counteracts the decreasing n(z). Keeping only linear terms in z, we nd:

n0 x20 + h2 = d2 = (h + Re )2 R2e = h2 + 2hRe x20 = 2hRe

ne (z)= n0 + e z , e = (7.15.25)
R
which is the same as Eq. (7.15.28).

For the average Earth radius R = 6370 km and the ITU values of n0 and given in
Eq. (7.15.24), we nd that the effective e is positive: Example 7.15.7: Graded-Index Optical Fibers. In Example 7.5.5, we considered a step-index
optical ber in which the rays propagate by undergoing total internal reection bouncing
e = 1.1418 107 m1 (7.15.26) off the cladding walls. Here, we consider a graded-index ber in which the refractive index
of the core varies radially from the center value nf to the cladding value nc at the edge of
Making the approximation n2 (z)= n20 + 2n0 e z will result in parabolic rays bending up- the core. Fig. 7.15.9 shows the geometry.
wards as in Example 7.15.2.
Often, an equivalent Earth radius is dened by e = n0 /Re so that the effective refractive
index may be assumed to arise only from the curvature of the equivalent Earth:

z
ne (z)= n0 + e z = n0 1 +
Re

In units of R, we have:
Fig. 7.15.9 Graded-index optical ber.

Re n0 n0 As a simple model, we assume a parabolic dependence on the radial distance. We may

= = = 1.3673 (7.15.27)
R e R n0 R write in cylindrical coordinates, where a is the radius of the core:

which is usually replaced by Re = 4R/3. In this model, the refractive index is assumed to  
2 n2f n2c
be uniform above the surface of the equivalent Earth, n(z)= n0 . n2 ()= n2f 1 2 , 2 = (7.15.29)
a2 n2f
The ray paths are determined by considering only the geometrical effect of the spherical
surface. For example, to determine the maximum distance x0 at which a ray from a trans- Inserting this expression into Eq. (7.15.6), and changing variables from z, x to , z, the
mitter at height h just grazes the ground, we may either use the results of Eq. (7.15.18), or integral can be done explicitly resulting in:
consider a straight path that is tangential to the equivalent Earth, as shown in Fig. 7.15.8.

Setting e = n0 /Re in Eq. (7.15.18), we obtain: a sin a
z= asin (7.15.30)
a cos a

2n0 h
 Inverting the arc-sine, we may solve for in terms of z obtaining the following sinusoidal
x0 = = 2hRe (7.15.28) variation of the radial coordinate, where we also changed from the incident angle a to
e
the initial launch angle 0 = 90o a :
7.16. Snels Law in Negative-Index Media 295 296 7. Oblique Incidence

The TE and TM wave solutions at both sides of the interface are still given by Eqs. (7.7.4)
tan 0 and (7.7.5), and reproduced below (with ejt suppressed):
= sin(z) , = (7.15.31)
a cos 0
 
For small launch angles 0 , the oscillation frequency becomes independent of 0 , that is, E(r) = y E0 ejkz z + TE ejkz z ejkx x
= /(a cos 0 ) /a. The rays described by Eq. (7.15.31) are meridional rays, that is,    
E0 kx kx
they lie on a plane through the ber axis, such as the xz- or yz-plane. H(r) = x + z ejkz z + TE x + z ejkz z ejkx x
TE kz kz
There exist more general ray paths that have nontrivial azimuthal dependence and prop-

(TE) (7.16.2)
agate in a helical fashion down the guide [871876].

E  (r) = y TE E0 ejkz z ejkx x


TE E0 kx jkz z jkx x
H  (r) = x +  z e e
7.16 Snels Law in Negative-Index Media TE kz

Consider the planar interface between a normal (i.e., positive-index) lossless medium where, allowing for magnetic media, we have
, and a lossless negative-index medium [391]  ,  with negative permittivity and
   TE kz  kz
permeability,  < 0 and  < 0, and negative refractive index n =   /0 0 . The TE = , TE =  , TE = TE = , TE = 1 + TE (7.16.3)
 kz kz 
TE + TE kz  + kz
refractive index of the left medium is as usual n = /0 0 . A TE or TM plane wave
is incident on the interface at an angle , as shown in Fig. 7.16.1. For the TM case we have:
   
kx kx
E(r) = E0 x z ejkz z + TM x + z ejkz z ejkx x
kz kz
E0  jkz z 
H(r) = y e TM ejkz z ejkx x
TM
 (TM) (7.16.4)
kx 
E  (r) = TM E0 x  z ejkz z ejkx x
kz
TM E0 jkz z jkx x
H  (r) = y e e
TM

with
Fig. 7.16.1 Refraction into a negative-index medium. kz k  TM k  kz 
TM = , TM = z  , TM = TM = z , TM = 1 + TM (7.16.5)
  TM + TM kz  + kz 
Because n < 0, Snels law implies that the refracted ray will bend in the opposite
direction (e.g., with a negative refraction angle) than in the normal refraction case. This One can verify easily that in both cases the above expressions satisfy Maxwells equa-
follows from: tions and the boundary conditions at the interface, provided that
n sin = n sin  = |n | sin  = |n | sin( ) (7.16.1)
k2x + k2z = 2  = n2 k20
As a result, the wave vector k of the refracted wave will point towards the interface, (7.16.6)
instead of away from it. Its x-component matches that of the incident wave vector k, k2x + kz2 = 2   = n2 k20
that is, kx = kx , which is equivalent to Snels law (7.16.1), while its z-component points
towards the interface or the negative z-direction in the above gure. In fact, Eqs. (7.16.2)(7.16.6) describe the most general case of arbitrary, homoge-
Formally, we have k = k s , where s is the unit vector in the direction of the re- neous, isotropic, positive- or negative-index, and possibly lossy, media on the left and

fracted ray pointing away from the interface, and k =   = n k0 , with k0 the right and for either propagating or evanescent waves. We concentrate, next, on the case

free-space wavenumber k0 = 0 0 = /c0 . As we see below, the energy ux Poynt- when the left medium is a positive-index lossless medium, > 0 and  > 0, and the
ing vector P  of the refracted wave is opposite k and points in the direction of s , and right one is lossless with  < 0 and  < 0, and consider a propagating incident wave
therefore, carries energy away from the interface. Thus, component-wise we have: with kx = nk0 sin and kz = nk0 cos and assume, for now, that n |n | to avoid
evanescent waves into the right medium. The Poynting vector P  in the right medium
kx = n k0 sin  = kx = nk0 sin , kz = n k0 cos  = |n |k0 cos  < 0
7.16. Snels Law in Negative-Index Media 297 298 7. Oblique Incidence

can be calculated from Eqs. (7.16.2) and (7.16.4): with the TE and TM cases being obtained from each other by the duality transformations

  x y and y x. It is straightforward to verify that the ranges of the x, y parameters
1 1 kz kx for which a Brewster angle exists are as follows:
(TE): P  = Re(E  H  )= |TE |2 |E0 |2 z Re + x Re
2 2  
1 1
   (7.16.7) TE case: x > 1, y < x, y > , or, x < 1, y > x, y <
1 1   kx x x
(TM): P = Re(E  H  )= |TM |2 |E0 |2 z Re

+ x Re (7.16.11)
2 2 kz |kz |2 1 1
TM case: y > 1, x < y, y > , or, y < 1, x > y, y <
Because  < 0 and  < 0, and kz is real, the requirement of positive energy ux x x
away from the interface, Pz > 0, requires that kz < 0 in both cases. Similarly, because These regions [697], which are bounded by the curves y = x and y = 1/x, are shown
kx > 0, the x-component of P  will be negative, Px < 0. Thus, the vector P  has the in Fig. 7.16.2. We note, in particular, that the TE and TM regions are non-overlapping.
direction shown in Fig. 7.16.1. We note also that the z-component is preserved across
the interface, Pz = Pz . This follows from the relationships:

1 kz   1 kz
Pz = |E0 |2 1 |TE |2 = |E0 |2 |TE |2 Re = Pz
2 2 
 (7.16.8)
1    1 
Pz = |E0 |2 1 |TM |2 = |E0 |2 |TM |2 Re = Pz
2 kz 2 kz

If n > |n |, the possibility of total internal reection arises. When sin > |n |/n,
then kz2 = n2 k20 k2x = k20 (n2 n2 sin2 ) is negative and kz becomes pure imaginary.
In this case, the real-parts in the right-hand side of Eq. (7.16.8) are zero, showing that
|TE | = |TM | = 1 and there is no (time-averaged) power ow into the right medium. Fig. 7.16.2 Brewster angle regions.
For magnetic media, including negative-index media, the Brewster angle may also
exist for TE polarization, corresponding to TE = 0. This condition is equivalent to The unusual property of Snels law in negative-index media that the refracted ray
kz = kz  . Similarly TM = 0 is equivalent to kz  = kz  . These two conditions imply bends in the opposite direction than in the normal case has been veried experimentally
the following relationship for the Brewster angles: in articial metamaterials constructed by arrays of wires and split-ring resonators [397],
and by transmission line elements [430432,452,465]. Another consequence of Snels
  2
TE = 0 kz = kz  (2 2 )sin2 B = 2 law is the possibility of a perfect lens [398] in the case n = 1. We discuss this in

  (7.16.9) Sec. 8.6.
 2
TM = 0 kz  = kz  (2 2 )sin2 B = 2 

7.17 Problems
Clearly, these may or may not have a solution, such that 0 < sin2 B < 1, depending
on the relative values of the constitutive parameters. For non-magnetic media, =  = 7.1 The matching of the tangential components of the electric and magnetic elds resulted in
0 , the TE case has no solution and the TM case reduces to the usual expression: Snels laws and the matching matrix Eq. (7.3.11). In both the TE and TM polarization cases,
show that the remaining boundary conditions Bz = Bz and Dz = Dz are also satised.
2    n2
sin2 B = 
=  = 2 7.2 Show that the Fresnel coefcients (7.4.2) may be expressed in the forms:
 
2 2  + n + n2
sin 2 sin 2 tan( ) sin( )
Assuming that , and  ,  have the same sign (positive or negative), we may re- TM = = , TE =
sin 2 + sin 2 tan( + ) sin( + )
place these quantities with their absolute values in Eq. (7.16.9). Dening the parameters
7.3 Show that the refractive index ratio n /n can be expressed in terms of the ratio r = TM /TE
x = | /| and y = | /|, we may rewrite (7.16.9) in the form:
and the incident angle by:
 
1 y
 1/2
TE case: 1 sin2 B = 1 n 1+r 2
x2 x = sin 1 + tan2
    (7.16.10) n 1r
1 2 x
TM case: 1 sin B = 1 This provides a convenient way of measuring the refractive index n from measurements of
y2 y
the Fresnel coefcients [716]. It is valid also for complex n .
7.17. Problems 299 300 7. Oblique Incidence

7.4 It is desired to design a Fresnel rhomb such that the exiting ray will be elliptically polarized 7.8 A light ray enters a glass block from one side, suffers a total internal reection from the top
with relative phase difference between its TE and TM components. Let sin c = 1/n be side, and exits from the opposite side, as shown below. The glass refractive index is n = 1.5.
the critical angle within the rhomb. Show that the rhomb angle replacing the 54.6o angle in
Fig. 7.5.6 can be obtained from:

cos2 c cos4 c 4 sin2 c tan2 (/4)
sin =
2 
2 tan2 (/4)+ cos2 c cos4 c 4 sin2 c tan2 (/4)

Show is required to satisfy tan(/4) (n n1 )/2.

7.5 Show the relationship (7.9.7) for the ratio TM /TE by rst proving and then using the fol-
lowing identities in the notation of Eq. (7.7.4): a. How is the exit angle b related to the entry angle a ? Explain.
b. Show that all rays, regardless of the entry angle a , will suffer total internal reection
(kz kz )(k2x kz kz )= k2 kz k2 kz at the top side.
c. Suppose that the glass block is replaced by another dielectric with refractive index n.
Using (7.9.7), show that when both media are lossless, the ratio TM /TE can be expressed
What is the minimum value of n in order that all entering rays will suffer total internal
directly in terms of the angles of incidence and refraction, and  :
reection at the top side?
TM cos( +  )
= 7.9 An underwater object is viewed from air at an angle through a glass plate, as shown below.
TE cos(  )
Let z = z1 +z2 be the actual depth of the object from the air surface, where z1 is the thickness
of the glass plate, and let n1 , n2 be the refractive indices of the glass and water. Show that
Using this result argue that |TM | |TE | at all angles . Argue also that B + B = 90o ,
the apparent depth of the object is given by:
for the Brewster angles. Finally, show that for lossless media with  >  , and angles of

incidence c , where sin c =  /, we have: z1 cos z2 cos
z =  +
 n21 sin2 n22 sin2
TM j sin2 sin2 c + sin tan
= 
TE j sin2 sin2 c sin tan

Explain how this leads to the design equation (7.5.8) of the Fresnel rhomb.
7.6 Let the incident, reected, and transmitted waves at an interface be:

 r
E+ (r)= E+ ej k+ r , E (r)= E ej k r , E (r)= E0 ej k

where k = kx x kz z and k = kx x + kz z. Show that the reection and transmission

coefcients dened in Eqs. (7.7.1)(7.7.5) can be summarized compactly by the following
vectorial relationships, which are valid for both the TE and TM cases:

k (E0 k ) 2k z
= E
k2 kz kz 7.10 An underwater object is viewed from air at an angle through two glass plates of refractive
indices n1 , n2 and thicknesses z1 , z2 , as shown below. Let z3 be the depth of the object
7.7 Using Eqs. (7.7.4), derive the expressions (7.9.11) for the Poynting vectors. Derive similar
within the water.
expressions for the TM case.
Using the denitions in Eqs. (7.3.12), show that if the left medium is lossless and the right
one lossy, the following relationship holds:
 
1   1
1 |T |2 = Re |T |2
T T

Then, show that Eqs. (7.9.12) and (7.9.13) are special cases of this result, specialized to the
TE and TM cases.
7.17. Problems
a. Express the apparent depth z of the object in terms of the quantities , n0 , n1 , n2 , n3
and z1 , z2 , z3 .
b. Generalize the results of the previous two problems to an arbitrary number of layers.
c. Consider also the continuous limit in which the body of water is inhomogeneous with
a refractive index n(z) given as a function of the depth z.
7.11 As shown below, light must be launched from air into an optical ber at an angle a in
order to propagate by total internal reection.
301 302 7. Oblique Incidence
7.12 You are walking along the hallway in your classroom building wearing polaroid sunglasses
and looking at the reection of a light xture on the waxed oor. Suddenly, at a distance d
from the light xture, the reected image momentarily disappears. Show that the refractive
index of the reecting oor can be determined from the ratio of distances:
d
n=
h 1 + h2
where h1 is your height and h2 that of the light xture. You may assume that light from
the xture is unpolarized, that is, a mixture of 50% TE and 50% TM, and that the polaroid
sunglasses are designed to lter out horizontally polarized light. Explain your reasoning.
7.13 Prove the effective depth formulas (7.5.17) of the Goos-Hanchen effect by directly differen-
tiating the reection coefcient phases (7.8.3) with respect to kx , noting that the lateral shift
is x0 = 2d/dkx where is either TE or TM .
7.14 First, prove Eq. (7.12.13) from Eqs. (7.12.11). Then, show the following relationships among
the angles , r ,  :
 
tan(/2) 1 tan(r /2) 1+ tan(r /2) 1
= , = , =
tan( /2) 1+ tan( /2) 1 tan(/2) 1+

a. Show that the acceptance angle is given by:

7.15 A TM plane wave is incident obliquely on a moving interface as shown in Fig. 7.12.1. Show
 that the Doppler-shifted frequencies of the reected and transmitted waves are still given
n2f n2c
sin a = by Eqs. (7.12.14) and (7.12.16). Moreover, show that the Brewster angle is given by:
na
1 + n2 + 1
b. For a ber of length l, show that the exiting ray, at the opposite end, is exiting at the cos B =
+ n2 + 1
same angle as the incidence angle.
c. Show that the propagation delay time through this ber, for a ray entering at an angle
, is given as follows, where t0 = l/c0 :

t0 n2f
t()= 
nf n2a sin2
2

d. What angles correspond to the maximum and minimum delay times? Show that the
difference between the maximum and minimum delay times is given by:

t0 nf (nf nc )
t = tmax tmin =
nc

Such travel time delays cause modal dispersion, that can limit the rate at which digital
data may be transmitted (typically, the data rate must be fbps 1/(2t) ). See, H. A. Smith, Measuring Brewsters Angle Between Classes, Physics Teacher, Febr. 1979, p.109.