(答案) 官方ExerciseSolutionsManual

Saleh & Teich Fundamentals of Photonics, Third Edition: Exercise Solutions ©2019 page i
FUNDAMENTALS OF
PHOTONICS
THIRD EDITION
SOLUTIONS MANUAL
FOR EXERCISES†
† A solutions manual is not available for the end-of-chapter problems
FEBRUARY 20, 2019
BAHAA E. A. SALEH
University of Central Florida
Boston University
MALVIN CARL TEICH

Boston University
Columbia University
JOHN WILEY & SONS, INC.

Saleh & Teich Fundamentals of Photonics, Third Edition: Exercise Solutions ©2019 page ii
Saleh & Teich Fundamentals of Photonics, Third Edition: Exercise Solutions ©2019 page 1
C H A P T E R
1
RAY OPTICS
1.1 POSTULATES OF RAY OPTICS
EXERCISE 1.1-1
Proof of Snell’s Law
The pathlength is given by n1 d 1 sec θ1 + n2 d 2 sec θ2 . (1)
The pathlength is a function of θ1 and θ2 , which are related by
d 1 tan θ1 + d 2 tan θ2 = d. (2)
The pathlength is minimized when ∂ [n1 d 1 sec θ1 + n2 d 2 sec θ2 ] = 0,

∂θ1
i.e., when n1 d 1 sec θ1 tan θ1 + n2 d 2 sec θ2 tan θ2 (∂θ2 /∂θ1 ) = 0. (3)
From (2), we have ∂ [d tan θ + d tan θ ] = 0,

1 1 2 2
∂θ1
∂θ2 d 1 sec2 θ1
so that d 1 sec2 θ1 + d 2 sec2 θ2 (∂θ2 /∂θ1 ) = 0 and =− .
∂θ1 d 2 sec2 θ2
d 1 sec2 θ1 tan θ2
Substituting into (3), we have n1 d 1 sec θ1 tan θ1 − n2 = 0,
sec θ2
whereupon n1 tan θ1 = n2 sec θ1 sin θ2 , from which n1 sin θ1 = n2 sin θ2 , which is Snell’s
law.
1.2 SIMPLE OPTICAL COMPONENTS
EXERCISE 1.2-1
Image Formation by a Spherical Mirror
ϕ
(y1, z1) θ1
ψ ϕ y
θ2
z
(y2, z2)
A ray originating at P1 = (y1 , z1 ) at angle θ1 meets the mirror at height

y ≈ y1 + θ1 z1 . (1)
y
The angle of incidence at the mirror is φ = ψ − θ1 ≈ − θ1 .
−R
The reflected ray makes angle θ2 with the z axis:

y 2y 2(y1 + θ1 z1 )
θ2 = 2φ + θ1 = 2 − θ1 + θ1 = − θ1 = − θ1 .
−R −R −R
y + θ1 z1
Substituting f = −R , we have θ2 = 1 − θ1 . (2)
2 f
y + (−y2 )
The height y2 can be determined from z2 ≈ θ2 . (3)
1

y1 + θ1 z1
Substituting from (1) and (2) into (3), we have y1 + θ1 z1 − y2 = z2 − θ1
f
z2 y1 z1 z2
and y2 = y1 − + θ1 z 1 − + z2 .
f f
z1 z2
If z1 − + z2 = 0, or z11 + z12 = 1 , we have
f f
z2
y2 = y1 1 − , (4)
f
which is independent of θ1 .
z2 y2 z2
From (4) it is clear that =1− , so that y2 = − y1 .
f y1 z1
EXERCISE 1.2-2
Image Formation
ϕ
θ1 −θ2
y
ϕ
P1 R P2
n1 n 2
a) From Snell’s law, we have n1 sin (θ1 + φ) = n2 sin [φ − (−θ2 )]. Since all angles are
small, the paraxial version of Snell’s Law is n1 (θ1 + φ) ≈ n2 (φ + θ2 ), or
θ2 ≈ (n1 /n2 )θ1 + [(n1 − n2 )/n2 ]φ.
n1 n2 − n1
Because φ ≈ y/R, we obtain θ2 ≈ θ1 − y, which is (1.2-8).
n2 n2 R
b) Substituting θ1 ≈ y/z1 and (−θ2 ) ≈ y/z2 into (1.2-8),
(n1 /n2 ) y n2 − n1
we have −y/z2 ≈ − y, from which (1.2-9) follows.
z1 n2 R
c) With reference to Fig. 1.2-13(b), for the ray passing through the origin 0, we have
angles of incidence and refraction given by y1 /z1 and −y2 /z2 , respectively, so that
the paraxial Snell’s Law leads to (1.2-10). Rays at other angles are also directed
from P1 to P2 , as can be shown using a derivation similar to that followed in Exer-
cise 1.2-1.
EXERCISE 1.2-3
Aberration-Free Imaging Surface In accordance with Fermat’s principle, we require
y
z
P
d1 d2
P1 y P2
z1 z2 z
n1 n 2
that the optical path length obey n1 d 1 +n2 d 2 = constant = n1 z1 +n2 z2 . This constitutes
anp
equation defining thepsurface, which can be written in Cartesian coordinates as
n1 (z + z1 )2 + y 2 + n2 (z2 − z)2 + y 2 = n1 z1 + n2 z2 . (1)
Given z1 and z2 , (1) relates y to z and thus defines the surface.
EXERCISE 1.2-4
Proof of the Thin Lens Formulas
A ray at angle θ1 and height y refracts at the first surface in accordance with (1.2-8)
and its angle is altered to θ = θ1 − n − 1 y , (1)
n nR1
where R1 is the radius of the first surface (R1 < 0).
At the second surface, the angle is altered again to θ2 = nθ − 1 − n y , (2)

R2
where R2 is the radius of the second surface (R2 > 0). We have assumed that the ray
height is not altered since the lens is thin.
Substituting (1) into (2) we obtain:

θ1
θ2 = n − n − 1 y − 1 − n y = θ1 − (n − 1) y 1 − 1 .
n nR1 R2 R1 R2
Using (1.2-11), we invoke θ2 = θ1 − (y/f ). (3)
If θ1 = 0, then θ2 = (−y/f ), and z2 ≈ (y/− θ2 ) = f , where f is the focal length. In

y y −y y y
general θ1 ≈ and −θ2 = . Therefore from (3), = − , from which (1.2-
z1 z2 z2 z1 f
13) follows. Equation (1.2-14) can be proved by use of an approach similar to that used
in Exercise 1.2-1.
EXERCISE 1.2-5
Numerical Aperture and Angle of Acceptance of an Optical Fiber
n2
1
n1
θc
Applying Snell’s law at the air/core surface: θc
sin θa = n1 sin θc = n1 cos θc (1) θa
p
Since sin θc = n2 /n1 , cos θc = 1 − (n2 /n1 )2 .
p p
Therefore, from (1), NA ≡ sin θa = n1 1 − (n2 /n1 )2 = n21 − n22 .
For silica glass with n1 = 1.475 and n2 = 1.460, the numerical aperture NA = 0.21 and
the acceptance angle θa = 12.1◦ .
EXERCISE 1.2-6
Light Trapped in a Light-Emitting Diode
a) The rays within the six cones of half angle θc = θc

sin−1 (1/n) ( = 16.1◦ for GaAs) are refracted into air
in all directions, as shown in the illustration. The rays
outside these six cones are internally reflected. Since
θc < 45◦ , the cones do not overlap and the reflected
rays remain outside the cones and continue to reflect
internally without refraction. These are the trapped rays.
Rθ
b) The area of the spherical cap atop one of these cones is A = 0 c 2πr sin θ r dθ =
2πr2 (1 − cos θc ), while the area of the entire sphere is 4πr2 . Thus, the fraction of
the emitted light that lies within the solid angle subtended by one of these cones is
A/4πr2 = 12 (1 − cos θc ) (see Sec. 18.1B). Thus, the ratio of the extracted light to the
total light is 6 × 12 (1 − cos θc ) = 3(1 − cos θc ) ( = 0.118 for GaAs). Thus, 11.8% of the
light is extracted for GaAs. √
Note that this derivation is valid only for θc < 45◦ or n > 2.
1.3 GRADED-INDEX OPTICS
EXERCISE 1.3-1
The GRIN Slab as a Lens
Using (1.3-11) and (1.3-12), with θ0 = 0 and z = d, we have y(d) = y0 cos (αd) and
θ(d) = −y0 α sin (αd). Rays refract into air at an angle θ 0 ≈ n0 |θ(d)| = n0 y0 α sin (αd).
y(d) y0 cos (αd) 1
Therefore, AF ≈ = = and
θ0 no y0 α sin (αd) n0 α tan (αd)
y0 1
f= 0 = , so that
θ n0 α sin (αd)

1 − cos (αd)
AH = f − AF = 1 1 − 1 = 1
n0 α sin (αd) tan (αd) n0 α sin (αd)
1 2 sin2 (αd/2) 1
= = tan (αd/2).
n0 α 2 sin (αd/2) cos (αd/2) n0 α
Trajectories:
y0
θ'
H A F d = π/α d = π/2α
f
EXERCISE 1.3-2
Numerical Aperture of the Graded-Index Fiber
Using (1.3-11) with y0 = 0, we obtain y(z) = (θ0 /α) sin(αz). The ray traces a sinusoidal
trajectory with amplitude θ0 /α that must not exceed the radius a. Thus θ0 /α = a.
The acceptance angle is therefore θa ≈ n0 θ0 = n0 αa.
For a step-index fiber

p(Exercise 1.2-5),
p
θa = n21 − n22p= (n1 + n2 )(n1 − n2 ). a
θ0
If n1 ≈ n2 , θa ≈ 2n1 (n1 − n2 ). θa
2 2
pn0 and n2 = n0 (1 − α a /2),
If n1 =
θa ≈ 2n0 (α2 a2 n0 /2) = α a n0 , which is the
same acceptance angle as for the graded-index fiber.
1.4 MATRIX OPTICS
EXERCISE 1.4-1
Special Forms of the Ray-Transfer Matrix
Using the basic equations
y2 = Ay1 + B θ1 and θ2 = C y1 + D θ1 , we obtain:
• If A = 0, then y2 = B θ1 , i.e., for a given θ1 , we

see that y2 is the same regardless of y1 .
This is a focusing system. y2
θ1
y1
• If B = 0, then y2 = A y1 , i.e., for a given y1 , we y2
see that y2 is the same regardless of θ1 .
This is an imaging system.
• If C = 0, then θ2 = D θ1 , i.e., we see that all θ1 θ2

parallel rays remain parallel.
• If D = 0, then θ2 = C y1 , i.e., we see that all rays y1

originating from a point become parallel. θ2
EXERCISE 1.4-2
A Set of Parallel Transparent Plates

1 d1 1 0 1 d 1 /n1
The first plate has ray transfer matrix: = .
0 1 0 1/n1 0 1/n1

The second plate has ray transfer ma- 1 d2 1 0 1 d 2 n1 /n2
= .
trix: 0 1 0 n1 /n2 0 n1 /n2
The first and second plates together have a ray transfer matrix:

1 d 2 n1 /n2 1 d 1 /n1 1 d 1 /n1 + d 2 /n2
= .
0 n1 /n2 0 1/n1 0 1/n2
P
Similarly N plates have a ray transfer 1 i d i /ni .
matrix: 0 1/nN
Including the interface between the N th plate and air, the overall ray transfer matrix
becomes:
P P
1 0 1 i d i /ni = 1 i d i /ni .
0 nN 0 1/nN 0 1
The ray transfer matrix of an inhomogeneous plate with refractive index n(z) and width
d is:
 
Rd
1 dz/n(z).
0
0 1
EXERCISE 1.4-3
A Gap Followed by a Thin Lens

1 0 1 d 1 d
M= = .
−1/f 1 0 1 −1/f 1 − d/f
EXERCISE 1.4-4
Imaging with a Thin Lens

1 d2 1 d1 1 − d 2 /f d 1 + d 2 (1 − d 1 /f )
M= = .
0 1 −1/f 1 − d 1 /f −1/f 1 − d 1 /f
For imaging, the matrix element B must vanish (see Exercise 1.4-1),
so that d 1 + d 2 (1 − d 1 /f ) = 0. Dividing this by d 1 d 2 yields 1/d 2 + 1/d 1 − 1/f = 0.
For all parallel rays to be focused onto a single point, the matrix element A must vanish
(see Exercise 1.4-1), so that 1 − d 2 /f = 0 or d 2 = f .
EXERCISE 1.4-5
Imaging with a Thick Lens
a) This system is composed of 5 subsystems:

1) A distance d 1 in air, followed by
2) An air/glass refracting surface, followed by
3) A distance d in glass, followed by
4) An glass/air refracting surface, followed by
5) A distance d 2 in air.
The ray transfer matrix of subsystem 2) is:

1 0 1 0
= , where f1 = R/(n − 1).
−(n − 1)/nR 1/n −1/nf1 1/n
The ray transfer matrix of subsystems 2) and 3) is:

1 d 1 0 1 − d/nf1 d/n
= .
0 1 −1/nf1 1/n −1/nf1 1/n
The ray transfer matrix of subsystems 2), 3), and 4) (the lens) is:

1 0 1 − d/nf1 d/n 1 − d/nf1 d/n
= .
−(n − 1)/R n −1/nf1 1/n −(1 − d/nf1 )/f1 − 1/f1 −d/nf1 + 1
The ray transfer matrix of the entire system is:

1 d2 1 − d/nf1 d/n 1 d1
.
0 1 −2/f1 + d/nf12 1 − d/nf1 0 1
For this system to be an imaging system, the B element of its ray transfer matrix must
vanish, i.e., B = d 1 (1 − d/nf1 ) + d/n + d 2 [d 1 (−2/f1 + d/nf12 ) + (1 − d/nf1 )] = 0.
Grouping together the terms proportional to d 1 , d 2 , and d 1 d 2 , we have

(d 1 + d 2 )(1 − d/nf1 ) − d 1 d 2 (2/f0 − d/nf12 ) + d/n = 0. (1)
Using the definitions

1/f = 2/f1 − d/nf12 (2)
and h = (f d/nf1 ), (3)
(1) becomes: (d 1 + d 2 )(1 − h/f ) − d 1 d 2 /f + d/n = 0. (4)
We now rewrite (4) in terms of z1 and z2 by substituting d 1 = z1 − h and d 2 = z2 − h.

The results is: z1 + z2 − z1 z2 /f + b = 0, (5)
b = d/n − h2 /f − 2h(1 − h/f ) = d/n + h2 /f − 2h

where (6)
= d/n + (h/f )(h − 2f ).
If b = 0, (5) gives the desired result, 1/z1 + 1/z2 = 1/f . To prove that b = 0, we use
(2) and (3) to write 1/f = (2f − h)/f1 f , from which 2f − h = f1 . Substituting this
into (6), we obtain b = d/n − hf1 /f . We now use (3) to write d/n = hf1 /f , so that
b = hf1 /f − hf1 /f = 0, as promised.
b) We show below that a ray parallel to the optical axis at height y1 must pass through
the point F2 , a distance f − h from the right surface of the lens, regardless of the
height y1 . This can be easily shown if we consider the ray transfer matrix of the system
composed of the thick lens (subsystems 2, 3, and 4 above) followed by a distance
f − h in air. This composite system has ray transfer matrix

1 f −h 1 − d/nf1 d/n
.
0 1 −2/f1 + d/nf12 1 − d/nf1
If the element A = 0, then y2 = B θ1 so that for θ1 = 0 (for rays parallel to the optical
axis), we have y2 = 0, i.e., the rays pass through the point F2 .
We now examine A = (1 − d/nf1 ) + (f − h)(−2/f1 + d/nf12 ), and show that it is 0.

Using (2), we have A = (1−h/f )+(f − h)(−2 + h/f )/f1 . Using the relation 2f −h = f1 ,
we obtain A = (1 − h/f ) + (f − h)/(−f ) = 0, as promised.
EXERCISE 1.4-6
A Periodic Set of Pairs of Different Lenses
Here, the unit cell is composed of 2 subsystems, each comprising a distance d of free
space followed by a lens. The ray transfer matrix of the unit cell is therefore given by
the product

1 d 1 d
.
−1/f2 1 − d/f2 −1/f1 1 − d/f1
The A and D elements of this product are:

A = 1 − d/f1 , D = −d/f2 + (1 − d/f2 )(1 − d/f1 )
so that
b = (A + D)/2 = 1 − d/f1 − d/f2 + d 2 /2f1 f2 = 2(1 − d/2f1 )(1 − d/2f2 ) − 1.
The condition |b| ≤ 1 is equivalent to −1 ≤ b ≤ 1 or 0 ≤ b + 1 ≤ 2, which leads to the
desired condition
0 ≤ (1 − d/2f1 )(1 − d/2f2 ) ≤ 1.
EXERCISE 1.4-7
An Optical Resonator
The resonator may be regarded as a periodic system whose unit system is a single
round trip between the pair of mirrors. In a resonator of length d, a paraxial ray starting
at the position y0 travels a distance d in free space, is reflected from the mirror 2,
then travels again backward through the same distance of free space, and finally
is reflected from the mirror 1 at position y1 . The process is repeated periodically.
The unit cell therefore consists of a cascade of two subsystems, each comprising
propagation in free space followed by reflection from a mirror. The condition of stability
may determined by writing the ray transfer matrix of the unit cell, as in the previous
exercise. Since a mirror with radius of curvature R has the same ray transfer matrix
as a lens with focal length f , if f = −R/2, the stability condition determined for the
periodic set of pairs of lenses considered in the previous exercise may be directly used
to obtain:
0 ≤ (1 + d/R1 )(1 + d/R2 ) ≤ 1.
The same result is set forth in (11.2-5).
C H A P T E R
2
WAVE OPTICS
2.2 MONOCHROMATIC WAVES
EXERCISE 2.2-1
Validity of the Fresnel Approximation
Given: λ = 633 × 10−9 m, d = 1 m.

2
NF θm 2
The Fresnel approximation is valid when 1, where NF = a and θm = a .
4 1/4 λd d
The condition is, therefore,a4 1 or a 4λd 3
= 0.04 m. Thus the radius a
4λd 3
2
must be much smaller than 4 cm. When a = 4 cm, NF = a = 2514 and θm = a =
λd d
0.04 rad.
EXERCISE 2.2-2
The Paraboloidal Wave and the Gaussian Beam
A = (A0 /z) exp[−jk(x2 + y 2 )/2z], (1)

∂A = −jxA k ,
∂x z
2
∂ 2 A = −j k x ∂A + A = −j k −jx2 A k + A = −jA k − k x 2 A.

∂x 2 z ∂x z z z z
2
2
Similarly, ∂ A = −jA k − k y 2 A,
∂y 2 z z
2
2
so that ∇T A = −j2A − k k (x2 + y 2 ) A. (2)
z z
Now,
2 2 2 2
∂A = − A0 exp −jk (x + y ) + A0 jk (x2 + y 2 ) exp −jk (x + y )
h i
∂z z2 2z z 2z 2 2z
−A jk
= + 2 (x2 + y 2 ) A. (3)
z 2z
Substituting (2) and (3) into the paraxial Helmholtz equation, we see that
∇2T A − j2k ∂A = 0, so that (1) does indeed satisfy this equation.
∂z
Replacing z by q(z) = z + jz0 in (1) does not alter the validity of the paraxial Helmholtz
equation since jz0 is a constant and therefore [∂/∂q ](·) = [∂/∂z ](·).

−k (x2 + y 2 )
At z = 0, we have q = jz0 , whereupon (1) gives: A(r) = A0 exp ,
jz0 2 z0
−k (x2 + y 2 )
2
whence the intensity is written as |A(r)|2 = A0 exp .
z0 z0
This is a Gaussian function of p
x and y that has its peak at x = y = 0 and that decreases
2 2 2
as the
p radial coordinate ρ = x + y increases. It reaches 1/e of its peak value at
ρ = λz0 /π [see (3.1-11)].
9
2.4 SIMPLE OPTICAL COMPONENTS
EXERCISE 2.4-1
Transmission Through a Prism
Substituting d(x, y) ≈ αx into (2.4-5) leads to the desired result.
EXERCISE 2.4-2
Double-Convex Lens
jko (x2 + y 2 ) jko (x2 + y 2 )

t(x, y) = t1 (x, y) t2 (x, y) = h 01 exp h 02 exp where
2f1 2f2
R1 −R2
f1 = and f2 = and h 01 and h 02 are constants.
n−1 n−1
jko (x2 + y 2 )

Thus t(x, y) = h0 exp , where 1 = 1 + 1 =
2f f f1 f2
(n − 1)(1/R1 − 1/R2 ) and h0 = h 01 h 02 is a constant. Note that R2 is negative.
EXERCISE 2.4-3
Focusing of a Plane Wave by a Thin Lens
U1 (x, y) = exp (−jkz), and t(x, y) = h0 exp [jk (x2 + y 2 ) /2f ].

Therefore, U2 (x, y) = U1 (x, y) t(x, y) = h0 exp {−jk [z − (x2 + y 2 ) /2f ]}.
The wavefronts of this wave are paraboloids of revolution, defined by z−(x2 + y 2 ) /2f =
constant, with radius of curvature −f , i.e., they approximate a spherical wave focused
at a point a distance f to the right of the lens.
If the incident wave is a plane wave at a small angle θ, U1 (x, y) ≈ exp [−jk(z + θx)],
then
U2 (x, y) = U1 (x, y) t(x, y) ≈ h0 exp −jk z + θx − x2 + y 2 /2f .

= h0 exp −jk z − x2 − 2f θx + y 2 /2f .

n h io
2
= h0 exp −jk z − (x − f θ) + y 2 /2f .
This is a paraboloidal wave centered about the point (f θ, 0, f ), as illustrated below.
θf
θ
f
EXERCISE 2.4-4
Imaging Property of a Lens
Choose a coordinate system with z = 0 at the lens. The incident wave is a spherical
wave centered at z = −z1 , i.e., U1 (x, y) ≈ exp [−jk (x2 + y 2 ) /2z1 ] so that
U2 (x, y) ≈ exp −jk x2 + y 2 /2z1 exp jk x2 + y 2 /2f

≈ exp jk x2 + y 2 /2z2 ,

where 1 = 1 − 1 or 1 + 1 = 1 .
z2 f z1 z1 z2 f
The transmitted wave is, therefore, a spherical wave centered at the point z = z2 .
EXERCISE 2.4-5
Transmission Through a Diffraction Grating

d0 2πx
a) d(x) = 1 + cos
2 Λ
t(x) = exp (−jko d 0 ) exp [−j (n − 1) ko d(x)]
= h0 exp [−j (n − 1) (ko d 0 /2) cos (2πx/Λ)], where
h0 = exp [−j (n + 1) (ko d 0 /2)].
b) Since t(x) is a P
periodic function of x with period Λ, it can be expanded in a Fourier
series: t(x) = q Cq exp (−jq2πx/Λ), where Cq are the Fourier coefficients. If the
incident wave is a plane wave at a small angle θi , i.e., U1 (x) = exp [−jko (z + θi x)],
the transmitted wave has amplitude:
U2 (x) = t(x)U (x)
= exp [−j (ko z + ko θi x + q2πx/Λ)] = exp [−jko (z + θq x)],
where θq = θi + q2π/ko Λ = θi + qλ/Λ. Thus the transmitted wave is composed of
plane waves at angles θq .
EXERCISE 2.4-6
Graded-Index Lens
Substituting n = n0 [1 − α2 (x2 + y 2 ) /2] into (2.4-14), we obtain
t = exp (−jnko d 0 ) = h0 exp [jn0 α2 ko d 0 (x2 + y 2 ) /2], with h0 = exp (−jn0 ko d 0 ).

Thus, t = h0 exp [jko (x2 + y 2 ) /2f ], where 1/2f = n0 α2 d 0 /2 so that f = 1/n0 α2 d 0 .
This is the amplitude transmittance of a lens of focal length f .

2.5 INTERFERENCE
EXERCISE 2.5-1
Interference of a Plane Wave and a Spherical Wave
√
I = I1 + I2 + 2 I1 I2 cos ϕ, where I1 = |A1 |2 , I2 = |A2 |2 , and
ϕ = k (x2 + y 2 ) /2z = π (x2 + y 2 ) /λd.

√
Therefore I(x, y, d) = I1 + I2 + 2 I1 I2 cos [π (x2 + y 2 ) /λd].
This locus of constant I are circles (x2 + y 2 ) = constant. The function cos (πx2 ) is
plotted in Table A.1-1. It is a sinusoidal function, called the chirp function, whose
frequency increases as x increases. This is why the rings in the interference pattern
become closer and closer as x2 + y 2 increases.
EXERCISE 2.5-2
Interference of Two Spherical Waves
n h i o
U1 = A exp {−jkz} exp −jk (x − a) + y 2 /2z and
2
z n h i o
U2 = A exp {−jkz} exp −jk (x + a) + y 2 /2z .
2
z
2
At z = d, I = i where
nh2I0 + 2I0 cos ϕ, h I0 = |A/d|
io and
2 2
ϕ = (k/2d) (x + a) + y 2 − (x − a) + y 2
= (π/λd) (4ax) = 4πax/λd.
Therefore, I = 2I0 [1 + cos (2πxθ/λ)], where θ = 2a/d.
EXERCISE 2.5-3
Bragg Reflection
The phase difference between two reflections is ϕ = k (Λ2 − Λ1 ).

But Λ2 = Λ/ sin θ and Λ1 = Λ2 cos 2θ = Λ cos 2θ/ sin θ.
Therefore, ϕ = k (Λ/ sin θ) (1 − cos 2θ) = k (Λ/ sin θ) 2 sin2 θ = k (2Λ sin θ).
For ϕ = 2π, we have kΛ sin θ = π so that 2Λ sin θ/λ = 1, or equivalently,
sin θ = λ/2Λ.
Λ1
θ
θ
Λ Λ2
θ θ
2.6 POLYCHROMATIC AND PULSED LIGHT
EXERCISE 2.6-1
Optical Doppler Radar
a) The two waves have a phase shift ϕ = 2πν1 t − 2πν2 √ t = 2π (ν1 − ν2 ) t = 2π (2v /c) νt.
The intensity of their superposition is I = I1 + I2 + 2 I1 I2 cos [2π (2v /c) νt]. This is a
sinusoidal function of time with frequency 2 (v /c) ν. The velocity v can be observed
by monitoring I as a function of time.
b) ϕ = k (z2 − z1 ) = k (2v t) = (2πν/c) 2v t = 2π 2 vc ν t, so that the beat frequency is

v
±2 c ν.
C H A P T E R
3
BEAM OPTICS
3.1 THE GAUSSIAN BEAM
EXERCISE 3.1-1
Parameters of a Gaussian Laser Beam
Given: λ = 633 nm = 633 × 10−9 m; P = 10−3 W; W0 = 0.05 × 10−3 m
a) θ0 = λ/ (πW0 ) = 4.03 × 10−3 rad = 4.03 mrad.

z0 = W0 /θ0 = 0.012.
Depth of focus = 2z0 = 0.025 m = 2.5 cm. q
2
At z = 3.5 × 105 km = 3.5 × 108 m, W (z) = W0 1 + (z/z0 ) = 1.41 × 106 m.
Diameter = 2821 km.
b) At z = 0, R = ∞.
At z = z0 , R = 2z0h = 2.5 cm. i
2
At z = 2z0 , R = z 1 + (z0 /z) = 0.031 m = 3.1 cm.
c) At beam center, I = I0 = 2P/πW02 = 2.546 × 105 W/m2 = 25.46 W/cm2 .

2
On beam axis at z = z0 , I = I0 [W0 /W (z0 )] = I0 /2 = 12.73 W/cm2 .
A spherical wave of power P = 100 W at z = z0 = 2.5 cm has intensity I =
P/ (4πz 2 ) = 5.169 × 104 = 5.169 W/cm2 .
EXERCISE 3.1-2
Validity of the Paraxial Approximation for a Gaussian Beam
The condition (2.2-21) is ∂A/∂z kA.

−jkρ2
In accordance with (3.1-4), A = A1 exp where q = z + jz0 . Therefore,
q 2q
−jkρ2 jkρ2 q 0 −jkρ2

∂A A1 0 A1
=− q exp + exp
∂z q2 2q q 2q 2 2q
0 2 0

q A jkρ q
=− +A
q 2q 2
∂q
where q 0 = = 1.
∂z
The condition ∂A/∂z kA is therefore equivalent to
−A/q + [jkρ2 /2q 2 ] A kA, or −1/kq + [jρ2 /2q 2 ] 1.
Substituting 1/q = 1/R − j2/kW 2 , we then have
(1/kR) [1 + 2ρ2 /W 2 ] + j [− (2/k 2 W 2 ) (1 + ρ2 /W 2 ) + (ρ2 /W 2 ) / (2R2 /W 2 )] 1.
Assuming that ρ is not much greater than W , i.e., for points not far outside the beam
width, this condition is satisfied if
a) kR 1;
b) kW 1; and
c) R W .
14
Condition a) signifies that the radius of curvature R λ. Because the minimum radius
of curvature is z0 , condition a) is satisfied if z0 λ. Similarly, condition b) is satisfied
if W0 λ, or θ0 = λ/πW0 1. However, condition √ c) is also satisfied if θ0 1: for
small z, R W0 ; for z = z0 , R = z0 W = 2W0 because θ0 = W0 /z0 1; for
large z, R ≈ z and W = θ0 z so that R/W = 1/θ0 1.
In summary, the conditions z0 λ, W0 λ, and θ0 1 guarantee that ∂A/∂z kA

and, therefore, that the paraxial approximation is satisfied.
EXERCISE 3.1-3
Determination of a Beam with Given Width and Curvature
Use h i
2
W 2 = W02 1 + (z/z0 ) (1)
h i
2
R = z 1 + (z0 /z) (2)
to obtain W 2 /R = (z/z0 ) W02 /z0 = (z/z0 ) (λ/π), from which

(z/z0 ) = (π/λ) W 2 /R. (3)
Substituting (3) into (1) and (2) we obtain (3.1-26) and (3.1-25).
EXERCISE 3.1-4
Determination of the Width and Curvature at One Point Given the Width and
Curvature at Another Point
Given: λ = 10−6 m; At position 1, R1 = 1 m and W1 = 10−3 m.

Find: R2 and W2 at position 2, z2 = z1 + d, d = 0.1 m.
We use the relations: q2 = q1 + d

1/q1 = 1/R1 − jλ/πW12
1/q2 = 1/R2 − jλ/πW22 .
Thus, 1/q1 = 1 − j0.32 and q1 = 0.91 + j0.29.
Therefore q2 = 1.01 + j0.29 and 1/q2 = 0.92 − j0.26, so that R2 = 1/0.92 = 1.09 m
and λ/πW22 = 0.26, from which W2 = 1.11 × 10−3 m = 1.11 mm.
EXERCISE 3.1-5
Identification of a Beam with Known Curvatures at Two Points
h i
2
Using (3.1-9) and z2 = z1 + d, we obtain R1 = z1 1 + (z0 /z1 ) ,
from which z12 − R1 z1 + z02 = 0. (1)
n o
2
We also obtain R2 = (z1 + d) 1 + [z0 / (z1 + d)] ,
2 2
from which (z1 + d) − R2 (z1 + d) + z02 = 0. (2)
Equations (1) and (2) form a pair of equations in two unknowns: z0 and z1 , that can be
manipulated algebraically to obtain (3.1-27) and (3.1-28).
3.2 TRANSMISSION THROUGH OPTICAL COMPONENTS
EXERCISE 3.2-1
Beam Relaying
Considering a lens and substituting z = z 0 = d/2 in (3.2-6) we obtain M = 1.

From (3.2-9a), r = z0 /(d/2 − f ) and Mr = |f / (d/2 − f ) |.
Inserting M = 1 into (3.2-9) we obtain Mr2 = 1 + r2 , so that

2 2
f 2 / (d/2 − f ) = 1 + [z0 / (d/2 − f )] ,
2 2 2
from which f 2 = (d/2 − f ) + z02 , or z02 = f 2 − (d/2 − f ) = f d − (d/2)
i.e., z02 = d (f − d/4).
Since z0 is real, this last equation requires that f ≥ d/4 or d ≤ 4f .
EXERCISE 3.2-2
Beam Collimation
a) Substituting (3.2-9) and (3.2-9a) into (3.2-6), we obtain
2
(z − f ) [f / (z − f )]
(z 0 − f ) = h i
2
1 + z02 / (z − f )
(z − f ) f 2
= h i,
2
(z − f ) + z02
z0 z/f − 1
from which −1= follows. (1)
f (z/f − 1)2 + (z0 /f )2
b) Let a = z0 /f , x = z/f − 1, and y = z 0 /f − 1.
Then (1) becomes y = x/ [x2 + a2 ].
For a fixed value of a and allowing x to vary, y achieves its maximum value if
dy 1 2x2
= 2 − = 0.
dx [x + a ] [x2 + a2 ]2
2
This occurs at [x2 + a2 ] = 2x2 or x = a,

i.e., if z/f − 1 = z0 /f or z = f + z0 .
c) z0 = 1 cm, f = 50 cm, a = z0 /f = 0.02.
Optimum z = f + z0 = 51 cm,
Distance z 0 :
x = z/f − 1 = 51/50 − 1 = 0.02 = a.
y = x/ [x2 + a2 ] = 1/2x = 25.
But y = z 0 /f − 1.
Therefore, z 0 = f (1 + y) = 50 × 26 = 1300 cm.
Magnification: Mr = f / (z − f ) = 1/x = 50.

r = z0 / (z − f ) = a/x = 1.
√ √ √
M = Mr / 1 + r2 = Mr / 2 = 50/ 2 = 35.4.
Width: W00 = p
M W0 = 35.4W0 .
W0 = λz0 /π ≈ 56 µm, W00 ≈ 2 mm.
EXERCISE 3.2-3
Beam Expansion
Imaging at the first lens:

Since z f1 and z − f1 z0 , applying (3.2-11) and (3.2-12) we obtain:
M1 = f1 / (z − f1 ) ≈ f1 /z,
W000 = [f1 / (z − f1 )] W0 ≈ (f1 /z) W0 ,
2
z000 = M12 z0 ≈ (f1 /z) z0 ,
z 1 ≈ f1 .
Imaging at the second lens:
Based on the results of Exercise 3.2-2, the optimal distance is
2
z2 = z000 + f2 , so that d = z1 + z2 = z1 + z000 + f2 ≈ f1 + (f1 /z) z0 + f2 .
Also the magnification√at this optimal
√ distance is√
M2 = [f2 / (z2 − f2 )] / 2 = f2 /z000 2 = f2 /M12 z0 2.
The overall magnification
√ of the system is
√
M = M2 M1 = f2 /M1 z0 2 = (f2 /f1 ) z/ 2z0 .
This is a large magnification since f2 f1 and z z0 .
EXERCISE 3.2-4
Variable-Reflectance Mirrors
The complex amplitude reflectance of this mirror is exp (−jkρ2 /R) exp (−ρ2 /Wm2 ).
Therefore, upon reflection, the phase of a Gaussian beam increases by −kρ2 /R, so
that the radius of curvature becomes R2 where 1/R2 = 1/R1 + 2/R.
In addition, the amplitude of the beam is multiplied by the factor exp (−ρ2 /Wm2 ) and
becomes exp (−ρ2 /W22 ), where 1/W2 = 1/W1 + 1/Wm .
The reflected beam remains Gaussian and has width W2 and radius of curvature R2 ,
as provided by the above equations.
EXERCISE 3.2-5
Transmission of a Gaussian Beam Through a Transparent Plate
From (1.4-11), the elements of the ABCD matrix of the plate are: A = 1, B = d/n,
C = 0, D = 1. Therefore, q2 = (Aq1 + B) / (Cq1 + D) = q1 +d/n, from which z2 +jz02 =
z1 + jz01 + d/n so that z02 = z01 and z2 = z1 + d/n. It follows that the transmitted
beam has the same depth of focus and its center is displaced by a distance d/n, as
illustrated in the figure.
d d
n n
d d
3.4 LAGUERRE–GAUSSIAN BEAMS
EXERCISE 3.4-1
Laguerre–Gaussian Beam as a Superposition of Hermite–Gaussian Beams
The Laguerre–Gaussian beam LG10 is identical to the superposed Hermite–Gaussian

beams √12 (HG01 + jHG10 ), as is ascertained from the absolute square of (3.4-1) [see
also the illustration in Fig. 3.4-1(a)].
At the beam waist, the Hermite–Gaussian beams may be expressed as
√ √
I1, 0 = |A1, 0 | 2 G21 2x/W0 G20 2y/W0
√ √
I0, 1 = |A0,1 | 2 G20 2x/W0 G21 2y/W0 ,
where G20 (u) = exp (−u2 ) and G21 (u) = 4u2 exp (−u2 ).
In the absence of interference, and if |A1,0 |2 = |A0,1 |2 = I0 , the total intensity is the
sum of the intensities:
I = 8I0 x2 + y 2 /W02 exp −2 x2 + y 2 /W02 = 8I0 ρ2 /W02 exp −2ρ2 /W02 ,

where ρ2 = x2 + y 2 .
√
The peak intensity occurs at the value of ρ for which dI/dρ = 0, i.e., at ρ = W0 / 2
or ρ W0 ≈ 0.707. The intensity is√0 at ρ = 0, as shown in the
√ figure below, and the
1/e2 points occur at ρ ≈ 0.23/ 2 W0 and at ρ ≈ 2.12/ 2 W0 . Since the beam
is circularly symmetric, it takes the form of a “donut” and hence is often colloquially
referred to as the “donut beam.”
1.0
0.8
I/Imax
0.6
0.4
0.2
0.0
0 1 2 3
ρ/W0
C H A P T E R
4
FOURIER OPTICS
4.1 PROPAGATION OF LIGHT IN FREE SPACE
EXERCISE 4.1-1
Binary-Plate Cylindrical Lens
Near the position x, cos (πx2 /λf ) is approximately a harmonic function with lo-
cal frequency νx = (1/2π) (∂/∂x) (πx2 /λf ) = x/λf . Its rectified version, f (x) =
U[cos(πx2 /λf )], is approximately a periodic function with local frequency λf /x near
the position x. The periodic function f (x) can be analyzed as a sum of harmonic
functions with spatial frequencies νx = qx/λf , where q = 0, ±1, ±3, ±5 . . . . This
structure therefore acts as a diffraction grating that bends the light by the approximate
angles λνx = λ (qx/λf ) = x/ (f /q). All rays deflected by the approximate angle
x/ (f /q) meet at the position f /q. Thus, the transparency acts as a cylindrical lens with
focal lengths ∞, ±f, ±f /3, ±f /5, . . . .
EXERCISE 4.1-2
Gaussian Beams Revisited
Given: U (x, y, 0) = f (x, y) = A exp [− (x2 + y 2 ) /W02 ] at the input (z = 0) plane,

Find: U (x, y, z) = g(x, y) at the distance z.
We shall use the Fourier-domain method.

The Fourier transform of f (x, y) is obtained by using the fact that the Fourier transform
of exp (−πt2 ) is exp (−πν 2 ) (see Table A.1-1) and the scaling property of the Fourier
transform (see Appendix A). Thus:
F (νx , νy ) = AπW02 exp −π 2 W02 νx2 + νy2 .

G(νx , νy ) = F (νx , νy )H(νx , νy )

where H(νx , νy ) = H 0 exp jπλz νx2 + νy2 , H 0 = exp (−jkz)

G(νx , νy ) = AπW02 exp −π 2 W02 νx2 + νy2 · exp (−jkz) · exp jπλz νx2 + νy2

= B exp −π 2 Q2 νx2 + νy2 ,

B = AπW02 exp (−jkz), where π 2 Q2 = π 2 W02 − jπλz.
The inverse Fourier transform is g(x, y) = (B/πQ2 ) exp [− (x2 + y 2 ) /Q2 ].
Defining 1/Q2 = jk/2q = jπ/λq, we write

g(x, y) = B (j/λq) exp −jk x2 + y 2 /2q

= A jπW02 /λq exp (−jkz) exp −jk x2 + y 2 /2q .

The parameter q = (jπ/λ) Q2 = (j/πλ) (π 2 W02 − jπλz) =

jπW02 /λ + z = z + jz0 where z0 = πW02 /λ. Substituting, we obtain
g(x, y) = A (jz0 /q) exp (−jkz) exp [−jk (x2 + y 2 ) /2q], where q = z + jz0 .
This is the equation of the Gaussian beam.
19
4.2 OPTICAL FOURIER TRANSFORM
EXERCISE 4.2-1
Conditions of Validity of the Fresnel and Fraunhofer Approximations:
A Comparison
Givens: λ = 0.5 µm = 0.5 × 10−6 m, , a = 2 × 10−2 m, b = 10−2 m .

2
NF θm
As shown in (4.1-13), the validity condition for the Fresnel approximation is 1,
4
3 1/3
where NF = a2 /λd and θm 4 4
= a/d, so that a /4λd 1 or d (a /4λ) = 0.43 m.
As shown in (4.2-2), the validity conditions for the Fraunhofer approximation are
NF 1 or a2 /λd 1 or d a2 /λ = 800 m; AND NF0 1 or b2 /λd 1 or
d b2 /λ = 200 m.
Thus, the Fresnel approximation is applicable for distances much greater than 43 cm;
and the Fraunhofer approximation is applicable for distances much greater than 800 m.
EXERCISE 4.2-2
The Inverse Fourier Transform
By examining (A.3-1) and (A.3-2) of Appendix A, we see that if F (νx , νy ) is the Fourier
transform of f (x, y), then F (−νx , −νy ) is the inverse Fourier transform of f (x, y). Thus
reversal of the coordinate system replaces the Fourier transform with the inverse
Fourier transform.
4.3 DIFFRACTION OF LIGHT
EXERCISE 4.3-1
Fraunhofer Diffraction from a Rectangular Aperture
Using Table A.1-1 and the scaling property of the Fourier transform, the Fourier
transform of the aperture function p(x, y) = rect (x/Dx ) rect (y/Dy ) is P (νx , νy ) =
Dx Dy sinc(Dx νx ) sinc(Dy y). Substituting into (4.3-5) we obtain (4.3-6). The first zero
of the function sinc(·) occur when its argument is ±1, i.e., at x = ±λd/Dx and
y = ±λd/Dy .
EXERCISE 4.3-2
Fraunhofer Diffraction from a Circular Aperture
Using (A.3-5), the Fourier transform of an aperture function in the form of a circle of
radius 1 is P (νx , νy ) = J1 (2πνρ ) /νρ .
2 J (2πν D/2) J (πν D)
For a radius D , P (νx , νy ) = D = D
1 ρ 1 ρ
.
2 2 νρ D/2 2 νρ
Substituting into (4.3-5) we obtain (4.3-8).
EXERCISE 4.3-3
Spot Size of a Focused Optical Beam
Equation (4.3-10) can be obtained by using the Fourier transform property of the lens,
given in (4.2-7). Because (4.2-7) is identical to (4.3-5) with d = f , the focused beam
has intensity given by (4.3-8) with d = f .
In accordance with (3.1-12) and (3.2-15) the focused Gaussian beam has intensity
distribution I(x, y) = I0 exp (−2π 2 W02 ρ2 /λ2 f 2 ) , where W0 is the waist radius of the
incident beam. To compare this distribution with that in (4.3-10), we take 2W0 = D,
2
assume that πD/λf = 1, and plot the two functions exp (−ρ2 /2) and [2J1 (ρ)/ρ] :
1
exp(-ρ 2/2)
[2J1(ρ)/ρ]2
0
0 ρ 10
C H A P T E R
5
ELECTROMAGNETIC OPTICS
5.5 ABSORPTION AND DISPERSION
EXERCISE 5.5-1
Dilute Absorbing Medium
Let χ0 be the susceptibility of the host medium so that n20 = 1+χ0 . When impurities are
present, the susceptibility of the host medium together with its suspension of impurities
is characterized by χ = χ0 + χ0 + jχ0 0 , with χ0 1 and χ00 1. The overall refractive
index and absorption coefficient are thus given by [see (5.5-5)]:
1/2
χ0 + jχ0 0

jα p
n− = 1 + χ0 + χ0 + jχ0 0 = (1 + χ0 ) 1 +
2ko 1 + χ0
χ0 + jχ0 0 χ0 + jχ0 0

≈ n0 1 + = n0 1 +
2 (1 + χ0 ) 2n20
χ0 −ko χ0 0
so that n = n0 + and α = .
2n0 n0
22
C H A P T E R
6
POLARIZATION OPTICS
6.1 POLARIZATION OF LIGHT
EXERCISE 6.1-1
Measurement of the Stokes Parameters
The expressions for S0 and S1 follow directly from the definition. The expression for S2
is verified by substituting for A45 and A135 from (6.1-12). Similarly, the expression for
S3 is verified by substituting for AR and AL from (6.1-13).
The Stokes parameters can be measured if the absolute values (or the intensities) of
components of the Jones vector are measured in three bases: the linearly polarized
basis in the (x, y) directions, the linearly polarized basis in the (45◦ , 135◦ ) directions,
and the circularly polarized basis (R, L). All six measurements are intensity measure-
ments.
EXERCISE 6.1-2
Cascaded Wave Retarders
a) Parallel fast axes

1 0 1 0 1 0
T= = = A half-wave retarder
0 e−jπ/2 0 e−jπ/2 0 e−jπ
b) Orthogonal fast axes

1 0 e−jπ/2 0 1 0
T= −jπ/2 = e−jπ/2 = A phase shifter
0 e 0 1 0 1
EXERCISE 6.1-3
Jones Matrix of a Rotated Half-Wave Retarder
1 0
The Jones matrix of a half-wave retarder at angle 0 is T = . The Jones matrix
0 −1
of a half-wave retarder at angle θ is

cos θ − sin θ 1 0 cos θ sin θ
T= , (6.1-1)
sin θ cos θ 0 −1 − sin θ cos θ
which gives rise to

cos θ sin θ
T= , (6.1-2)
− sin θ cos θ
from which

cos 2θ sin 2θ
T= . (6.1-3)
sin 2θ − cos 2θ
If θ = 22.5◦ , then T can be written as

1 1
T= , (6.1-4)
1 −1
so that the output waves are proportional to the sum and difference of the input waves.
23
EXERCISE 6.1-4
Normal Modes of Simple Polarization Systems

1 0 1 0
a) T = . Eigenvectors are and ;
0 0 0 1
Eigenvalues are 1 and 0.

1 0 1 0
b) T = . Eigenvectors are and ;
0 e−jΓ 0 1
Eigenvalues are 1 and e−jΓ .

cos θ − sin θ 1 0
c) T = . Eigenvectors are and ;
sin θ cos θ j −j
Eigenvalues are e−jθ and ejθ .
6.2 REFLECTION AND REFRACTION
EXERCISE 6.2-1
Brewster Windows
Reflection does not occur at the first surface

when θ1 is the Brewster angle, θ1 = tan−1 n. θ2 θ1
sin θ2 =
Snell’s law√provides √ (1/n) sin θ1 =
(1/n) n/ 1 + n2 = 1/ 1 + n2 , so that θ1 θ2
tan θ2 = 1/n, i.e., θ2 is also a Brewster angle
for the second surface.
For n = 1.5, we have θ1 = tan−1 n = 56.3◦ .
6.4 OPTICAL ACTIVITY AND MAGNETO-OPTICS
EXERCISE 6.4-1
Rotatory Power of an Optically Active Medium
p p
If G n0 , n± = n20 ± G = n0 1 ± G/n20 ≈ n0 ± G/2n0 .
Therefore, ρ = π (n− − n+ ) /λo = −πG/λo n0 .

C H A P T E R
7
PHOTONIC-CRYSTAL OPTICS
7.1 OPTICS OF DIELECTRIC LAYERED MEDIA
EXERCISE 7.1-1
Quarter-Wave Film as an Anti-Reflection Coating
The M matrix for the problem at hand is readily obtained by cascading the M matrix
for a single dielectric boundary (see Example 7.1-2) and the M matrix for propagation
followed by a boundary, in reverse order as usual. The result is:
1 (n3 + n2 ) e−jϕ (n3 − n2 ) ejϕ 1 n2 + n1 n2 − n1

M= ,
2n3 (n3 − n2 ) e−jϕ (n3 + n2 ) ejϕ 2n2 n2 − n1 n2 + n1
with ϕ = n2 ko d = 2πd/λ and λ = λo /n2 .
The B element of this matrix is

1
(n3 + n2 )(n2 − n1 ) e−jϕ + (n3 − n2 )(n2 + n1 ) ejϕ .

B=
4n2 n3
The reflection coefficient can be made to vanish if B = 0, i.e., if
(n3 + n2 )(n2 − n1 ) + (n3 − n2 )(n2 + n1 ) ej2ϕ = 0.
This requires that ej2ϕ be real, i.e., that 2ϕ be an integer multiple of π.
The value 2ϕ = 4πd/λ = π leads to d = λ/4 and
(n3 + n2 )(n2 − n1 ) − (n3 − n2 )(n2 + n1 ) = 0,

√
whereupon we obtain n22 = n1 n3 or n2 = n1 n3 .
The choice 2ϕ = 2π, or any even multiple of π, leads to
(n3 + n2 )(n2 − n1 ) + (n3 − n2 )(n2 + n1 ) = 0,
which gives the trivial solution n1 = n3 .
25
C H A P T E R
9
GUIDED-WAVE OPTICS
9.1 PLANAR-MIRROR WAVEGUIDES
EXERCISE 9.1-1
Optical Power
In accordance with (5.3-10), the power flow is determined by the Poynting vector
S = 12 E × H∗ . For the TE mode, we have Ey = Ez = Hx = 0. The component of
S in the z direction is therefore Sz = 12 Ex Hy∗ . Also, from Maxwell’s equation (5.3-13),
∇ × E = −jωµo H, we have −jωµo Hy = ∂Ex /∂z, so that Sz = (1/2jωµo ) Ex ∂Ex∗ /∂z.
Substituting Ex = am um (y) exp(−jβm z), we obtain Sz = (βm /2ωµo )|am |2 |um (y)|2 .
The total power flow in the z direction is the integral of Sz with respect to y. Since the
integral of |um (y)|2 is unity, the power flow is (βm /2ωµo )|am |2 . Furthermore, because
βm = k cos θm = (ω/c) cos θm , we can write the power flow as (1/2µo c)|am |2 cos θm =
(1/2η)|am |2 cos θm .
EXERCISE 9.1-2
Optical Power in a Multimode Field
In accordance with Exercise 9.1-1, the power flow in the z direction is the integral
of Sz P= (1/2jωµo ) Ex ∂Ex∗ /∂z with respect to y. Making use of the substitution
Ex = m am um (y) exp (−jβm z), we obtain
Sz = (βm /2ωµo ) m am um (y) exp (−jβm z) n a∗n u∗n (y) exp (jβn z).
P P
Because the integral of um (y)u∗n (y) with respect to y is unity for n = m and zero
otherwise, the total power Pis
2 2
P
m (β m /2ωµo ) |a m | = m (1/2η) |am | cos θm .
9.2 PLANAR DIELECTRIC WAVEGUIDES
EXERCISE 9.2-1
Confinement Factor
Since the waveguide is symmetric we consider confinement only for y > 0.

For y < d/2, um (y) = Am cos (k sin θm y), m even (1a)
= Am sin (k sin θm y), m odd.
For y > d/2, um (y) = Bm exp (−γm y), (1b)
q
2
where γm = n2 ko (n1 /n2 ) cos2 θm − 1 . (2)
26
Because um (y) must be continuous at y = d/2 ,

Am cos (k sin θm d/2) = Bm exp (−γm d/2), m even (3a)
Am sin (k sin θm d/2) = Bm exp (−γm d/2), m odd. (3b)
R d /2
The power in the region y < d/2 is P1 = 0 u2m (y)dy .
Substituting from (1a) and integrating, we have
P1 = A2m (d/4) [1 + (−1)m sin(kd sin θm )/kd sin θm ] . (4)
Similarly, the power in the region y > d/2 is:

2
P2 = Bm (1/2γm ) exp (−γm d). (5)
P1 1
The confinement ratio Γm = = (6)
P1 + P2 1 + P2 /P1
can be obtained by substituting from (4) and (5) and using (3) to substitute for Bm /Am :
P2 (1/γm d) [1 + (−1)m cos(kd sin θm )]

= . (7)
P1 1 + (−1)m sin(kd sin θm )/kd sin θm
sin θc
It is convenient to write the result in terms of the variable M = (8)
λ/2d
by writing kd = 2πd/λ = πM/ sin θc , (9)
q
2
γm d = kd (n2 /n1 ) (n1 /n2 ) cos2 θm + 1
q q
= kd cos2 θm − cos2 θc = kd sin2 θc − sin2 θm
q
= πM 1 − sin2 θm / sin2 θc . (10)
It is also convenient to
√define the ratio: sm = sin θm / sin θc
and write γm d = πM 1 − sm . (11)
Using (10) and (11) in (7) then leads to

P2 sm 1 + (−1)m cos (πM sm )
= p . (12)
P1 1 − s2m πM sm + (−1)m sin (πM sm )
This provides an expression for the confinement ratio Γm = 1/ (1 + P2 /P1 ) as a
function of the parameter M , which represents the number of modes, and the
parameter sm = sin θm / sin θc , which is determined by the normalized angles of
the modes.
As an example, consider the case M = 8. The parameters sm are determined from the
characteristic equation (9.2-4), which can be written in terms of M and sm as:
p
tan (M sm π/2 − mπ/2) = 1/s2m − 1 .
Solutions of this equation are displayed in Fig. 9.2-2 for M = 8. For m = 0, the
first intersection point occurs at sin θ0 = 0.933(λ/2d), or s0 ≈ 0.933/M . Similarly,
s1 ≈ 1.86/M ; s2 ≈ 2.778/M ; and so on.
Substituting these values into (12) and (6) leads to the following confinement ratios:
Γ0 ≈ 0.999; Γ1 ≈ 0.996; and Γ2 ≈ 0.990. The lowest-order mode therefore has the
highest power confinement factor, as promised.
EXERCISE 9.2-2
The Asymmetric Planar Waveguide
Let the complements of the critical angles for reflection from the substrate and the cover
be θc2 = cos−1 (n2 /n1 ) and θc3 = cos−1 (n3 /n1 ), respectively. Since n2 > n3 , θc2 < θc3 .
Therefore, a guided ray must be inclined at an angle θ smaller than the smaller of θc2
and θc3 , i.e., θ < θc2 .
p
a) Since the numerical aperture is governed by θc2 , NA = n21 − n22 .
b) The self-consistency condition in the symmetric waveguide (9.2-1) is thus modified
to:
2π
2d sin θ − ϕr2 − ϕr3 = 2πm , m = 0, 1, 2, . . . ,
λ
where ϕr2 and ϕr3 are, respectively, the phase shifts introduced by total internal
reflection at the substrate and cover boundaries. These phases are given by the
general formula in (9.2-3), making use of the appropriate critical angles θc2 and θc3 .
c) The number of modes is governed by the critical angle pof reflection at the substrate.
.
It is therefore given by M = (2d/λo ) NA, where NA = n21 − n22 .
C H A P T E R
10
FIBER OPTICS
10.3 ATTENUATION AND DISPERSION
EXERCISE 10.3-1
Optimal Grade Profile Parameter
−1 s
The group velocities are vq = (dβq /dω) , where βq = n1 ko [1 − (q/M ) ∆];
M = sn21 ko2 a2 ∆; s = p/(p + 2); and ko = ω/co .
To simplify the process of taking the derivative we write

s
βq = (n1 ω/co ) [1 − xq ] where xq = (q/M ) ∆.
dβq /dω = (1/co ) [d (n1 ω) /dω] (1 − xq ) − n1 (ω/co ) dxq /dω
= (N1 /co ) (1 − xq ) n1 (ω/co ) dxq /dω
= (N1 /co ) [1 + xq φ] ,
where N1 = d (n1 ω) /dω is the group refractive index and

φ = −1 − (n1 /N1 ) (ω/xq ) dxq /dω. (1)
If φxq is small, the group velocity is

−1 −1
vq = (dβq /dω) = (co /N1 ) [1 + xq φ] ≈ (co /N1 ) [1 − xq φ]. (2)
We now proceed to determine φ:
dxq /dω = s(q/M )s−1 q −1/M 2 dM/dω ∆ + (q/M )s d∆/dω

= −sxq (1/M )dM/dω + xq (1/∆)d∆/dω ; (3)
2
dM/dω = 2sn1 ko [d (n1 ko ) /dω] a2 ∆ + s (n1 ko ) a2 d∆/dω
= 2M (1/n1 ko ) d (n1 ko ) /dω + M (1/∆) d∆/dω
= M [(2/n1 ko ) N1 /co + (1/∆) d∆/dω] . (4)
Substituting into (3), we have:

(1/xq ) dxq /dω = −s [(2/n1 ko ) N1 /co + (1/∆)d∆/dω] + (1/∆)d∆/dω
= −2sN1 /n1 ω + (1 − s)(1/∆)d∆/dω.
We now use (1) to obtain φ = −1 + 2s − (1 − s)ps /2 with ps = 2 (n1 /N1 ) (ω/∆) d∆/dω.
Thus φ = −1 + 2p/(p + 2) − ps /(p + 2) = (p − 2 − ps )/(p + 2), which, when substituted

into (2), gives (10.3-10).
29
EXERCISE 10.3-2
Differential Group Delay in a Two-Segment Fiber
a) If L = 500 m is the length of a segment, then the group delays of the x and y
components at the end of the first segment are:
Tx = LNx /c = 2.4367 µs and Ty = LNy /c = 2.4383 µs.
Each of these components can be analyzed into two components of equal magni-
tudes along the principal axes x0 and y 0 of the second segment. These components
travel to the end of the second segment with group delays Tx0 and Ty0 . The
overall delay may therefore take four values: Tx + Tx0 , Ty + Ty0 , Tx + Ty0 , and
Ty + Tx0 . Since Tx = Tx0 and Ty = Ty0 , we actually have three possible delays:
2Tx = 4.8733 µs, 2Ty = 4.8767 µs, and Tx + Ty = 4.873 µs. Since the pulse with
the delay Tx + Ty results from two possibilities, its amplitude depends on the phase
shifts encountered, which are sensitive to the phase velocities and the exact lengths
of the fiber segments, and is sensitive to any slight disturbance in the system. This
middle pulse will therefore have random polarization.
The differential delays between the fastest pulse and the slowest pulse is 2Ty −
2Tx = 3.4 ns. To determine whether this differential delay will be visible, we examine
the pulse broadening due to GVD. For a single segment, the GVD broadening is
Dσλ L = 20 × 50 × 0.5 = 500 ps, so that the width of each pulse is broadened from
an initial value of 100 ps to a value of 1 ns. The shape of the received pulses will
therefore appear as shown below:
3.4 ns
1 ns 1 ns
t
b) The two fiber segments are equivalent to two cascaded identical retarders with their
principal axes rotated by 45◦ . The Jones matrix of this system is the product of the
matrices

cos θ − sin θ 1 0 cos θ sin θ 1 0
T= −jϕ −jϕ ,
sin θ cos θ 0 e − sin θ cos θ 0 e
where ϕ = (Nx − Ny )2πL/λ is the retardation introduced by a segment and θ is the

angle of rotation. Since θ = 45◦ ,

1 1 −1 1 0 1 1 1 1 0
T= √ −jϕ √ −jϕ
2 1 1 0 e 2 −1 1 0 e
and therefore

1 1 + e−jϕ e−jϕ (1 − e−jϕ )
T= .
2 1 − e−jϕ e −jϕ
(1 + e −jϕ
)
The eigenvalues and eigenvectors of this matrix may be determined for any value
of ϕ. Since the matrix is unitary, the eigenvalues will always be phase factors.
For
1
example, if ϕ = π then the eigenvalues are ±j and the eigenvectors are , rep-
∓j
resenting circularly polarized modes. In any case, a pulse in one of the polarization
modes travels with a single group velocity so that it arrives as a single pulse instead
of two.
C H A P T E R
11
RESONATOR OPTICS
11.1 PLANAR-MIRROR RESONATORS
EXERCISE 11.1-1
Resonance Frequencies of a Traveling-Wave Resonator
a) Three-mirror ring resonator : At resonance, the round trip phase shift, 3kd + 3π,
is equal to a multiple of 2π. Thus, 3kd + π = q2π, where q is an integer, so that
3kd = (2q − 1)π or 3(2πν/c)d = (2q − 1)π. Consequently νq = (2q − 1)(c/6d) so
that the allowed frequencies are odd multiples of c/6d. Two consecutive resonances
are therefore separated by a frequency νF = 2(c/6d) = c/3d.
b) Four-mirror
√ bow-tie resonator : At resonance, the round
√ trip phase shift, (4 +
2 5)kd + 4π, is equal to a√multiple of 2π, i.e., (4√+ 2 5)kd + 4π = q2π, where
q is an integer. Thus,
√ (4 + 2 5)kd = q2π, or (4 + 2 5)(2πν/c)d = q2π, from which
νq = q[c/(4 + 2 5)d ]. Two√ consecutive resonances are therefore separated by a
frequency νF = [c/(4 + 2 5)d ].
EXERCISE 11.1-2
Resonator Modes and Spectral Width
Given: R1 = 0.98, R2 = 0.99, d = 1 m, n = 1, c = co /n = co = 3 × 108 m/s.

Frequency spacing between modes is νF = c/2d = 1.5 × 108 Hz = 150 MHz.
Loss coefficient αr = (1/2d) ln(1/R1 R2 ) = 0.015.
Using (11.1-28), the Finesse F ≈ π/αr d = 207.7.
The spectral linewidth is δν = νF /F = 7.22 × 105 Hz = 722 kHz.
This approximation is appropriate since αr d = 0.015 1.
11.2 SPHERICAL-MIRROR RESONATORS
EXERCISE 11.2-1
Maximum Resonator Length for Confined Rays
The confinement condition is 0 ≤ (1 + d/R1 ) (1 + d/R2 ) ≤ 1. Substituting R1 = −0.5 m

and R2 = −1 m, we obtain 0 ≤ (1 − 2d) (1 − d) ≤ 1. Letting x = (1 − 2d)(1 − d), the
confinement condition becomes 0 ≤ x ≤ 1. The figure below shows a plot of x versus
d. Based on this figure, the maximum value of d for which the resonator is stable is
d = 1.5 m.
1
x
0
0 0.5 1 1.5 d
31
EXERCISE 11.2-2
A Plano-Concave Resonator
For a plano-concave resonator, R1 = ∞ and R2 = −|R|. Substituting z1 = 0 and z2 = d

into (11.2-13) we have |R| = d + z02 /d, from which z02 = d(|R| − d). For confinement,
z02 > 0 so that |R| > d.
From (11.2-10), we have

1/2
W02 = λz0 /π = (λ/π) [d(|R| − d)] and
1/2 1/4
W1 = W0 = (λd/π) (|R|/d − 1) . Using (11.2-16), this leads to
2
W22 = W02 (1 +d /z02 ) = W02 [1 + d/(|R| − d)] = W02 |R|/(|R| − d),
from which 1/4
1/2 1/4 1/2
(|R|/d)2

(λd/π) (|R|/d − 1) (|R|/d) 1/2
W2 = 1/2
= (λd/π) .
(|R|/d − 1) (|R|/d − 1)
The quantities W1 and W2 are plotted versus d/|R| below:
2 λd
π W2
λd
π W1
0 d 1
|R|
EXERCISE 11.2-3
Resonance Frequencies of a Confocal Resonator
Given: d = 30 cm = 0.3 m; c = co /n = co .
z1 = −z0 and z2 = z0 .
νF = c/2d = 5 × 108 Hz = 500 MHz.
∆ζ = tan−1 (z2 /z0 ) − tan−1 (z1 /z0 )
= tan−1 (1) − tan−1 (−1)
= π/4 − (−π/4)
= π/2.
(∆ζ/π)νF = ∆νF /2 = 250 MHz.
νq = qνF + νF /2 = (q + 1/2)νF .
At the central frequency q ≈ ν/νF = (5 × 1014 )/(5 × 108 ) = 106 ,

for
q = 106 : νq = 5 × 1014 + 2.5 × 108 Hz
q = 10 + 1: νq = 5 × 1014 + 7.5 × 108 Hz
6
q = 106 + 2: νq = 5 × 1014 + 12.5 × 108 Hz

q = 106 + 3: νq = 5 × 1014 + 17.5 × 108 Hz
q = 106 − 1: νq = 5 × 1014 − 2.5 × 108 Hz

q = 106 − 2: νq = 5 × 1014 − 7.5 × 108 Hz
q = 106 − 3: νq = 5 × 1014 − 12.5 × 108 Hz
q = 106 − 4: νq = 5 × 1014 − 17.5 × 108 Hz
Thus, there are 8 modes within the band 5 × 1014 ± 2 × 109 Hz.
EXERCISE 11.2-4
Resonance Frequencies of the Symmetrical Confocal Resonator
For confocal symmetric resonators, we have (∆ζ/π)νF = νF /2.

From (11.2-33), we see that ν l, m, q = [q + (l + m + 1)/2] νF .
The set of modes for which l + m + 1 is even are spaced at frequency intervals νF .
Modes for which l + m + 1 is odd are also spaced at frequency intervals νF , but are
displaced from the even modes by frequency νF /2.
11.3 TWO- AND THREE-DIMENSIONAL RESONATORS
EXERCISE 11.3-1
Density of Modes in a Two-Dimensional Resonator
a) The number of modes with frequency between 0 and ν is the same as the number
of modes with wavenumber between 0 and k = 2πν/c.
In accordance with Fig. 11.3-2, this number is approximated by the area of a

2
quadrant in k space (πk 2 /4) divided by the area per mode (π/d ) , and then
multiplied by a factor of two to account for the two polarizations per mode. This
2 2
number is thus 2 (πk 2 /4) / (π/d ) = k 2 d 2 /2π = (2πν/c) d 2 /2π = 2πν 2 d 2 /c2 .
Consequently, the number of modes per unit area, in the frequency band 0 to ν, is
Nν = 2πν 2 /c2 .
b) The density of modes per unit area per unit frequency interval is therefore
M(ν) = dNν /dν = 4πν/c2 .
C H A P T E R
12
STATISTICAL OPTICS
12.1 STATISTICAL PROPERTIES OF RANDOM LIGHT
EXERCISE 12.1-1
Coherence Time

R∞ 2 −2|τ |
R∞ R∞ −2τ
a) Coherence time = −∞ |g(τ )| dτ = −∞exp dτ = 2 0 exp dτ = τc .
τc τc
|g(τ )| decreases by a factor 1/e = 0.368 at τ = τc .
−πτ 2

R∞ R∞
b) Coherence time = −∞
|g(τ )|2 dτ = −∞
exp dτ = τc .
τc2
|g(τ )| decreases by a factor of exp (−π) = 0.043 at τ = τc .
EXERCISE 12.1-2
Relation Between Spectral Width and Coherence Time
Since
R ∞ S(ν) is the Fourier transform of G(τ ), we have
0
S(ν)dν = G(0). (1)
From
R ∞ 2Parseval’sR ∞
theorem, we write
0
S (ν)dν = −∞
|G(τ )|2 dτ . (2)
Squaring both sides of (1) and dividing by the two sides of (2), while making use of
the definitions of
R ∆νc , τc , and g(τR), we obtain
∆νc = |G(0)|2 / |G(τ )|2 dτ = 1/ |g(τ )|2 dτ = 1/τc .
EXERCISE 12.1-3
Differential Equations Governing the Mutual Coherence Function
G = hU ∗ (r1 , t)U (r2 , t + τ )i.

Therefore, ∇21 G = h[∇21 U ∗ (r1 , t)] U (r2 , t + τ )i.
Since U obeys the wave equation, ∇2 U = (1/c2 )∂ 2 U/∂t2 , and

∇21 G = (1/c2 )h[(∂ 2 /∂t2 ) U ∗ (r1 , t)] U (r2 , t + τ )i. (1)
We now proceed to prove that h[(∂ 2 /∂t2 ) U ∗ (r1 , t)] U (r2 , t + τ )i = (∂ 2 /∂τ 2 ) G,
so that ∇21 G = (∂ 2 /∂τ 2 ) G:
Proof: h[(∂/∂t)U ∗ (r1 , t)] U (r2 , t + τ )i

= h lim (1/∆t) [U ∗ (r1 , t + ∆t) − U ∗ (r1 , t)] U (r2 , t + τ )i
∆t→0
= lim (1/∆t) [G(r1 , r2 , τ − ∆t) − G(r1 , r2 , τ )]

∆t→0
= −(∂/∂τ )G(r1 , r2 , τ )
Similarly, h[(∂ 2 /∂t2 )U ∗ (r1 , t)] U (r2 , t + τ )i = (∂ 2 /∂τ 2 ) G.
34
12.4 PARTIAL POLARIZATION
EXERCISE 12.4-1
Partially Polarized Light
The coherency matrix for the superposition of unpolarized light whose intensity is given
by (Ix + Iy )(1 − P), and linearly polarized light of intensity (Ix + Iy )P at angle θ, is

Ix + Iy 1 0 cos2 θ cos θ sin θ
G = (1 − P) + P (Ix + Iy ) .
2 0 1 cos θ sin θ sin2 θ
The four elements of this matrix are
Gxx = (Ix + Iy ) (1 − P)/2 + (Ix + Iy ) P cos2 θ, (1)

2
Gyy = (Ix + Iy ) (1 − P)/2 + (Ix + Iy ) P sin θ, (2)
Gxy = Gyx = (Ix + Iy ) P sin θ cos θ. (3)
We wish to show that for some θ,

Gxx = Ix , (4)
Gyy = Iy , (5)
2
Gxy = Gyx = (Ix Iy ) |gxy | . (6)
From (4) and (1) we have

Ix − (Ix + Iy ) (1 − P)/2
cos2 θ = , (7)
(Ix + Iy )P
while from (5) and (2) we have
Iy − (Ix + Iy ) (1 − P)/2
sin2 θ = . (8)
(Ix + Iy )P
Adding (7) and (8) we obtain cos2 θ + sin2 θ = 1, so that if (7) is satisfied, (8) is
automatically satisfied.
Let us now verify (6). From (3), we find

2
G2xy = (Ix + Iy ) P2 sin2 θ cos2 θ. (9)
Substituting (7) and (8) into (9) yields
G2xy = [Ix − (Ix + Iy )(1 − P)/2] [Iy − (Ix + Iy )(1 − P)/2]

= Ix Iy + 14 (Ix + Iy )2 (1 − P)2 − 12 (Ix + Iy )2 (1 − P)
= Ix Iy + 41 (Ix + Iy )2 (1 − P)[(1 − P) − 2]
= Ix Iy − 14 (Ix + Iy )2 (1 − P2 ). (10)
From the definition of P provided in (12.4-13), we find
4 (1 − |gxy |2 ) Ix Iy
1 − P2 = , so that (10) gives
(Ix + Iy )2
G2xy = Ix Iy − Ix Iy (1 − |gxy |2 ) = Ix Iy |gxy |2 , indicating that (6) is also satisfied.
C H A P T E R
13
PHOTON OPTICS
13.1 THE PHOTON
EXERCISE 13.1-1
Photon in a Gaussian Beam
a) In accordance with (3.1-12), the intensity of a Gaussian beam at z = 0 is I(ρ, 0) ∝

exp (−2ρ2 /W02 ). The probability p of detecting the photon within a circle of radius
W0 is thus given by the ratio
R W0
I(ρ, 0)2πρ dρ
p = R0∞ .
0
I(ρ, 0)2πρ dρ
ρ2 2ρ dρ
Transforming the integration variable to x = 2 , so that dx = , we have
W0 W02
R1
exp (−2x)dx (1 − e−2 )
p = R 0∞ = = 0.86.
0
exp (−2x)dx (1 − 0)
Indeed, recall from the discussion following (3.1-17) that the power contained within
a circle of radius W0 is 86% of the total power in the beam.
b) The average number of photons is p n = 0.86 n.
EXERCISE 13.1-2
Photon-Momentum Recoil
Photon momentum = }k = }ω/c = E/c. The recoil momentum p = Mv , where M is

the mass of the 198 Hg atom and v is its velocity, so that v = E/M c.
Substituting E = 4.88 eV = 4.88 × 1.6 × 10−19 J; M = 198 × 1.66 × 10−27 kg;
and c = 3 × 108 m/s, we obtain v = 7.9 × 10−3 m/s.
p
The RMS thermal velocity of the atom is v thermal = 3k T /M .
At T = 300◦ K, k T = 1.38 × 10−23 × 300, so that v thermal = 194 m/s, which is much
larger than the recoil velocity.
EXERCISE 13.1-3
Single Photon in a Mach–Zehnder Interferometer
Using the interference formula for the Mach–Zehnder
interferometer (2.5-6), the intensity in the detector
branch is
I ∝ I0 [1 + cos (2πd/λ)] ∝ I0 cos2 (πd/λ), where 2I0 is
the total incident intensity. If the wave contains a single
photon, the probability of its detection by the detector sin2(πd /λ) cos2(πd /λ)
is 1 + cos (2πd/λ) ∝ cos2 (πd/λ), as shown in the
figure. The probability of finding the photon in the other 1
output branch of the interferometer is 1−cos (2πd/λ) ∝
sin2 (πd/λ), which is also shown in the figure. The
probability of finding the photon in either of the two 0
branches is the sum cos2 (πd/λ) + sin2 (πd/λ) = 1, as 0 0.5 1 dλ
/
expected.
36
EXERCISE 13.1-4
Single Photon in a Gaussian Wavepacket
a) The Gaussian function f (t) = |a(t)|2 = exp (−t2 /2τ 2 ) has an RMS width, as defined
by Equation (A.2-1), that is σt = τ . Since z = ct, the time uncertainty of the function
a(t − z/c) is σt and the positional uncertainty is σz = cσt .
√
b) The Fourier transform of a(t) is also Gaussian, A(ν) = (1/2 πσν ) exp (−ν 2 /4σν2 ),
where σν = 1/4πσt . The RMS width of |A(ν)|2 is σν . Since the energy E = hν, the
energy uncertainty is σE = hσν = h/4πσt = }/2σt , from which (13.1-20) follows.
Because the momentum p = h/λ = (h/c)ν, the momentum uncertainty is σp =
(h/c)σν = h/4cπσt = }/2cσt . Therefore, σz σp = (cσt )(}/2cσt ) = }/2, from
which (13.1-21) follows.
13.2 PHOTON STREAMS
EXERCISE 13.2-1
Average Energy of a Resonator Mode in Thermal Equilibrium
The average number of photons n for a single mode of thermal light is given by (13.2-
21). The average energy E = hνn so that E = hν/ [exp (hν/k T ) − 1]. A plot of E
versus hν for two values of k T is shown below. In the limit hν/k T 1, i.e., when
the photon energy is much smaller than the unit of thermal energy, exp (hν/k T ) ≈
1 + (hν/kT ) so that E ≈ k T . In this limit, the average energy is what would be obtained
if the light were not quantized.
0.052
k T = 0.052
0.026
kT = 0.026
0 -3
10 10-2 10-1 1
hν (eV)
C H A P T E R
14
LIGHT AND MATTER
14.3 INTERACTIONS OF PHOTONS WITH ATOMS
EXERCISE 14.3-1
Frequency of Spontaneously Emitted Photons
In accordance with (14.3-1) the probability density of spontaneous emission into a

single prescribed mode is psp = (c/V ) σ(ν). The probability density of spontaneous
emission into any of the modes in the band ν to ν + dν is therefore Psp dν =
(c/V ) σ(ν) M(ν) V dν, where M(ν) = 8πν 2 /c3 is the density of modes per unit volume.
Using M(ν) = 8πν 2 /c3 , σ(ν) = Sg(ν), and S = λ2 /8πtsp , we thus obtain
Psp dν = (1/tsp ) g(ν) dν. The probability that the emitted photon has a frequency be-
tween ν and ν + dν is therefore proportional to g(ν)dν. Hence, when many photons are
emitted the distribution of their frequencies is proportional to g(ν).
EXERCISE 14.3-2
Doppler-Broadened Lineshape Function
R∞
a) The average lineshape function is g(ν) = −∞ g(ν − v ν0 /c) p(v ) dv . It is convenient
to transform
R ∞ the integration variable from v to x = (ν0 /c)v , which gives rise to
g(ν) = −∞ g(ν − x) px (x) dx, (1)
where px (x) = (c/ν0 ) p(cv /ν0 ). This result follows because transforming a random
variable v to another random variable av , where a is a constant, modifies the
probability density function to (1/a) p(v /a). Since p(v ) is a Gaussian function of
width σv , px (x) is a Gaussian function of width σx = (ν0 /c)σv . Note that x has units
of frequency. Equation (1) is the convolution of a Lorentzian function g(ν) of width
∆ν with a Gaussian function of width σx .
b) If ∆ν ν0 σv /c, then ∆ν σx , i.e., the Lorentzian function g(ν) in the convolution

integral (1) is much narrower than the Gaussian function. Since g(ν) is a narrow
function of unit area, it can be treated Rfor the purposes of integration as a delta
∞
function δ(ν). Thus (1) gives: g(ν) ≈ −∞ δ(ν − x) px (x) dx = p(ν), i.e., g(ν) is
p
approximately Gaussian with width σD = σx = (ν0 /c)σv = σv /λ = k T /M / λ.
c) At T = 300◦ K, for the Ne transition we substitute the following into (14.3-43)

and (14.3-44): λ ≈ λo = 632.8 × 10−9 m, M ≈ 20 mp = 20(1.67 × 10−27 kg)
so that ∆νD = 2.35 σD = 1.3 × 109 Hz = 1.3 GHz. For the CO2 transition,
λ ≈ λo = 10.6 × 10−6 m, M ≈ 44 mp = 44(1.67 × 10−27 kg), so that ∆νD =
5.3 × 107 Hz = 53 MHz.
d) The maximum value of σ(ν) is h √ i

σ 0 = σ(ν0 ) = (λ2 /8πtsp ) g(ν0 ) = (λ2 /8πtsp ) 1/ 2πσD
h √ i 0.94(λ2 /8π)
= (λ2 /8πtsp ) 2.35/ 2π∆νD ≈ .
tsp ∆νD
(2/π)(λ2 /8π) 0.64(λ2 /8π)
For the Lorentzian lineshape σ 0 = = , which is similar to
tsp ∆ν tsp ∆ν
the Gaussian result. Note, however, that ∆νD is typically much greater than ∆ν so
that σ 0 is much smaller in the Doppler-broadened-Gaussian case.
38
14.4 THERMAL LIGHT
EXERCISE 14.4-1
Frequency of Maximum Blackbody Energy Density
[8π(k T )3 /c3 h2 ]x3

Defining x = hν/k T , (14.4-9) gives %(ν) = .
[ex − 1]
The frequency at which %(ν) is maximum is obtained by equating d%/dx to zero.
This yields 3x2 [ex − 1] − x3 [ex ] = 0, or 3(ex − 1) = xex , from which x = 3(1 − e−x ).
Numerical solution of this nonlinear equation provides x ≈ 2.821. For T = 300◦ K, we
thus have ν = νp = xk T /h = 1.76 × 1013 = 17.6 THz, which is consistent with the plot
presented in Fig. 14.4-4.
C H A P T E R
15
LASER AMPLIFIERS
15.1 THEORY OF LASER AMPLIFICATION
EXERCISE 15.1-1
Attenuation and Gain in a Ruby Laser Amplifier
Parameters: λo = 694.3 × 10−9 m; n = 1.76; λ = λo /n; T = 300◦ K; ∆ν = 330 × 10 9 Hz;
tsp = 3 × 10−3 s; N a = N 1 + N 2 = 1028 m−3 ; h = 6.62 × 10−34 J-s; k = 1.38 × 10−23 J/◦ K.
a) In thermal equilibrium
N 2 /N 1 = exp [−(E 2 − E 1 )/k T ] = exp (−hν/k T ) = exp (−hco /λo k T ) ≈ 10−30 .
Therefore N 2 N 1 so that N 1 ≈ N a , i.e., almost all the atoms are in the lower-
level energy state. The attenuation coefficient at the central frequency = α(ν0 ) =
−γ(ν0 ) = −N(λ2 /8πtsp ) g(ν0 ) = −N(λ2 /8πtsp ) (2/π∆ν), where N = N 2 − N 1 ≈
−N a . Therefore, α(ν0 ) = N a (λ2 /8πtsp ) (2/π∆ν) = 3.98 × 104 m−1 = 398 cm−1 .
b) For γ(ν0 ) = N(λ2 /8πtsp ) (2/π∆ν) = 50 m−1 , the population becomes inverted
for N = N 2 − N 1 = (50)(4)π 2 tsp ∆ν/λ2 = 1.254 × 1025 m−3 = 1.254 × 1019 cm−3 .
c) To attain a gain G = exp [γ(ν0 )d] = 4, we require d = ln(4)/γ(ν0 ) = 2.77 cm.
15.2 AMPLIFIER PUMPING
EXERCISE 15.2-1
Optical Pumping
The populations of the three energy levels (2, 1, and the ground state) are N 2 , N 1 , and
N g , respectively. The total population is N 1 + N 2 + N g = N a . Since level 1 is short lived,
N 1 ≈ 0, so that N 2 + N g ≈ N a and
N g ≈ N a − N 2. (1)
The system is pumped by transitions between the ground state and level 2, so that
R 2 = (N g − N 2 )W . Using (1), we therefore obtain R 2 = (N a − 2N 2 )W . In this case,
it is apparent that the rate R 2 is dependent on N 2 . Now, from (15.2-7) the population
difference N 0 ≈ R 2 tsp = (N a − 2N 2 )W tsp . But N 0 = N 2 − N 1 ≈ N 2 . Thus N 0 =
(N a − 2N 0 )W tsp . Solving for N 0 we have N 0 = N a tsp W /(1 + 2tsp W ). In the limit
where W 1/2tsp , we obtain N 0 ≈ N a tsp W , which is proportional to W . However for
larger W , in the domain where it is not negligible in comparison with 1/2tsp , saturation
sets in and N 0 loses its proportionality to W .
EXERCISE 15.2-2
Saturation Time Constant
If tsp τnr (i.e., nonradiative transitions are slow), and

tsp τ20 (i.e., decay to levels other than level 1 is slow), and
tsp τ1 (i.e., decay from level 1 is fast, i.e., level 1 is short-lived),
then 1/τ2 = 1/τ20 + 1/tsp + 1/τnr ≈ 1/tsp , so that τ2 ≈ tsp ; furthermore 1/τ21 = 1/tsp +
1/τnr ≈ 1/tsp , so that τ21 ≈ tsp . Under these conditions, it follows that the saturation
time constant provided in (15.2-11) can be approximated as τs = τ2 + τ1 (1 − τ2 /τ21 ) ≈
tsp + τ1 (1 − tsp /tsp ) ≈ tsp , thereby demonstrating that τs ≈ tsp .
40
EXERCISE 15.2-3
Pumping Power in Three- and Four-Level Systems
3-level laser amplifier:

For a three-level scheme, in accordance with (15.2-30), when N 0 = 0 we have tsp W = 1
so that W = 1/tsp . Now, when the pumping transition probability is twice as large, as
assumed in the problem for the three-level scheme, namely W = 2/tsp , (15.2-30)
yields N 0 = 13 N a . The pumping power is then P = hν31 R, where R = (N 1 − N 3 )W .
Since N 3 ≈ 0, we obtain R ≈ N 1 W . Because N 2 + N 1 = N a and N 2 − N 1 = N 0 , by
subtraction we obtain 2N 1 = N a − N 0 = N a − 13 N a = 23 N a , so that N 1 = 31 N a . It follows
that R = 13 N a W = 13 N a (2/tsp ) = 23 N a /tsp , which leads to P = 32 hν31 N a /tsp .
4-level laser amplifier:

For a four-level scheme, in accordance with (15.2-20), when N 0 = 0 we have W = 0.
Now, when the pumping transition probability is W = 1/2tsp , as assumed in the
problem for the four-level scheme, (15.2-20) yields N 0 = 13 N a . The pumping power
is then P = hν30 R, where R = (N g − N 3 )W ≈ N g W . But since N g = N a − N 0 =
N a − 13 N a = 23 N a , we have R = 23 N a (1/2tsp ) = 13 N a /tsp , from which P = 31 hν30 N a /tsp .
Comparison:
Under these special conditions, and assuming that the two systems have the same
values of N a and tsp , the ratio of the 4-level to 3-level pumping powers required to
achieve this population difference is ν30 /2ν31 .
15.4 AMPLIFIER NONLINEARITY
EXERCISE 15.4-1
Saturation Photon-Flux Density for Ruby
Parameters: λo = 694.3 × 10−9 m; n = 1.76; τs = 2tsp ; ∆ν = 3.3 × 1011 Hz; co = 3 × 108

m/s; h = 6.63 × 10−34 J-s.
From (15.4-2), we have 1/φs (ν0 ) = (λ2 /8π) (τs /tsp ) g(ν0 ) = (λ2 /8π)(2)(2/π∆ν) =
(λo /n)2 /2π 2 ∆ν, where we have made use of (15.1-8) for g(ν0 ). Inserting the numer-
ical parameter values leads to φs (ν0 ) = 4.186 × 1025 m−2 s−1 . This corresponds to a
saturation intensity Is = hν0 φs (ν0 ) = (co h/λo ) φs (ν0 ) = 1.2×107 W/m2 = 1200 W/cm2 .
EXERCISE 15.4-2
Spectral Broadening of a Saturated Amplifier
Making use of (15.4-2), (15.4-3), (15.4-4), and (15.1-8), we have:
γ(ν) = γ0 (ν)/[1 + φ/φs ], where γ0 (ν) = ag(ν), a = N 0 λ2 /8πtsp ,

and 1/φs = bg(ν), b = (λ2 /8π)(τs /tsp ),
and g(ν) = (∆ν/2π)/[(ν − ν0 )2 + (∆ν/2)2 ].
Therefore,
ag(ν)
γ(ν) =
1 + bφ g(ν)
a (∆ν/2π) a (∆ν/2π)
= = ,
(ν − ν0 )2 + (∆ν/2)2 + bφ ∆ν/2π (ν − ν0 )2 + (∆νs /2)2
where
(∆νs /2)2 = (∆ν/2)2 + bφ ∆ν/2π

= (∆ν/2)2 [1 + bφ (2/π∆ν)]
= (∆ν/2)2 [1 + bφ g(ν0 )]
= (∆ν/2)2 [1 + φ/φs (ν0 )].
Taking thep
square-root of both sides of this equation yields the desired result:
∆νs = ∆ν 1 + φ/φs (ν0 ).
15.5 AMPLIFIER NOISE
EXERCISE 15.5-1
Amplified Spontaneous Emission (ASE)
a) In the unsaturated case γ(ν) ≈ γ0 (ν), whereupon the differential equation (15.5-
3) becomes dφ/dz = γ0 (ν)φ + ξsp (ν). To solve this differential equation, we
use a trial solution of the form φ(z) = A exp [γ0 (ν)z] + B. Substitution yields
γ0 (ν)A exp [γ0 (ν)z] = γ0 (ν)A exp [γ0 (ν)z] + γ0 (ν)B + ξsp (ν), from which it is
clear that B = −ξsp (ν)/γ0 (ν). The initial condition φ(0) = 0 is satisfied if
A + B = 0, or A = −B = ξsp (ν)/γ0 (ν). It follows that the solution is φ(z) =
φsp {exp [γ0 (ν)z] − 1}, where φsp (ν) = ξsp (ν)/γ0 (ν). At z = d, we therefore find
φ(d ) = φsp {exp [γ0 (ν)d ] − 1}.
b) Following (15.1-9) for spontaneous emission with a Lorentzian profile, the unsatu-
rated gain coefficient is
γ0 (ν0 )(∆ν/2)2
γ0 (ν) = .
(ν − ν0 )2 + (∆ν/2)2
The frequency dependence of this gain coefficient, normalized to unity height, is
then
γ0 (ν) (∆ν/2)2
g1 (ν) = = .
γ0 (ν0 ) [(ν − ν0 )2 + (∆ν/2)2 ]
This quantity differs from the Lorentzian lineshape function provided in (15.1-8),
which is normalized to unit area. The function g1 (ν) is plotted in the figure below for
∆ν = ν0 /100.
In the same figure, we present the frequency

dependence of the equivalent function applicable
1
for ASE, also normalized to unit height:
{exp [γ0 (ν)d ] − 1} {exp [ag1 (ν)] − 1}
g2 (ν) = = , g1(ν)
{exp [γ0 (ν0 )d ] − 1} {exp (a) − 1}
g2(ν)
with a = γ0 (ν0 )d = 5.
0
ν0 ν
It is clear that g2 (ν) is narrower than g1 (ν), indicating that the amplification of spon-
taneous emission is accompanied by spectral narrowing.
C H A P T E R
16
LASERS
16.1 THEORY OF LASER OSCILLATION
EXERCISE 16.1-1
Threshold of a Ruby Laser
Parameters: λo = 694.3 × 10−9 m; n = 1.76; λ = λo /n; ∆ν = 330 × 10 9 Hz;

N a = N 1 + N 2 = 1.58 × 1019 cm−3 ; h = 6.62 × 10−34 J-s; k = 1.38 × 10−23 J/◦ K;
co = 3 × 108 m/s; c = co /n; α(ν0 ) = −γ0 (ν0 ) ≈ 0.2 cm−1 ; d = 10 cm; A = 1 cm2 .
a) γ0 (ν) = N 0 σ(ν), where N 0 = N 2 − N 1 .
At thermal equilibrium at T = 300◦ K, N 2 /N 1 = exp (−hν/k T ) = exp (−hco /λo kT ) ≈
10−30 (see Exercise 15.1-1). Therefore N 2 N 1 and N 1 ≈ N a , i.e., almost all of
the atoms are in the lower energy state. In this case N 0 = N 2 − N 1 ≈ −N a . The
gain coefficient, γ0 (ν0 ) = N 0 σ(ν0 ) ≈ −N a σ(ν0 ), is then negative and corresponds
to an absorption coefficient α(ν0 ) ≈ N a σ(ν0 ). Since α(ν0 ) ≈ 0.2 cm−1 , σ(ν0 ) ≈
α(ν0 )/N a = 1.27 × 10−20 cm2 .
b) The resonator has parameters d = 0.1 m, R1 = R2 = 0.8, and αs = 0. Its loss

coefficient is αr = αs + (1/2d) ln(1/R1 R2 ) = 2.231 m−1 = 0.0223 cm−1 . The photon
lifetime is thus τp = (αr c)−1 = 1.49 × 10−9 = 1.49 ns.
c) The threshold population difference is N t = αr /σ(ν0 ) = (0.0223 cm−1 )/(1.27 ×

10−20 cm2 ) = 1.76 × 1018 cm−3 .
16.2 CHARACTERISTICS OF THE LASER OUTPUT
EXERCISE 16.2-1
Number of Modes in a Gas Laser
The gain coefficient is given by γ0 (ν) = γ0 (ν0 ) exp [−(ν − ν0 )2 /2σD2 ] with ∆νD =
a) √
8 ln 2 σD . The allowed oscillation band is obtained from equating the gain coeffi-
cient γ0 (ν) to the loss coefficient αr :
γ0 (ν0 ) exp [−(ν − ν0 )2 /2σD2 ] = αr or (ν − ν0 )2 /2σp

2
D = ln [γ0 (ν0 )/αr ], so that
(ν − ν0 )2 = 2σD2 ln [γ0 (ν0 )/αr ] or (ν − ν0 ) = ±σD 2 ln [γ0 (ν0 )/αr ] .
p
Thus B = 2σD 2 ln [γ0 (ν0 )/αr ] , from which we obtain
B = 2∆νD (8 ln 2)−1/2 [2 ln (γ0 (ν0 )/αr )]1/2 . (1)
b) ∆νD = 1.5 × 109 Hz; γ0 (ν0 ) = 2 × 10−3 cm−1 ; d = 100 cm; R1 = 1; R2 = 0.97; and
αs = 0, so that αr = αs + (1/2d) ln (1/R1 R2 ) = 1.52 × 10−4 cm−1 .
Bandwidth: From (1) above, we have B = 2.89 × 109 Hz = 2.89 GHz.

Modal spacing: νF = c/2d = (co /n)/2d. Using n = 1, d = 100 cm, and
co = 3 × 1010 cm/s, we obtain νF = 1.5 × 108 Hz = 0.15 GHz.
Number of modes: M = B/νF = 19.3, corresponding to a maximum of 19 modes.
43
16.4 PULSED LASERS
EXERCISE 16.4-1
Population-Difference Rate Equation for a Four-Level System
Since τ1 τsp , level 1 is short lived and we may therefore assume that N 1 ≈ 0 so that
N = N 2 − N 1 ≈ N 2 . Substituting N 1 = 0 and N 2 = N into (15.2-8), and assuming that
τ2 ≈ tsp , we obtain
dN/dt = R 2 − N/tsp − NW i . (1)
Under steady-state conditions (dN/dt = 0), with W i = 0, (1) yields R 2 − N/tsp = 0

so that the steady-state population difference in the absence of amplifier radiation is
N 0 = R 2 tsp . Substituting R 2 = N 0 /tsp into (1), we obtain
dN/dt = N 0 /tsp − N/tsp − W i N. (2)
Equation (2) is identical to (16.4-5) except for the factor of two in the W i term. This can
be understood as follows: In the 3-level laser system, a photon emitted from level 2
decreases N 2 by unity and simultaneously increases N 1 by unity, so that the population
difference N = N 2 − N 1 decreases by two.
In the 4-level system, on the other hand, level 1 is short-lived and cannot maintain any
additions to its population. Thus, a photon emitted from level 2 decreases N 2 by unity
but entails no change in N 1 , which is 0 at all times. The result is a decrease of N by
unity and the absence of the factor of two.
EXERCISE 16.4-2
Pulsed Ruby Laser
Given: λo = 694.3 × 10−9 m; n = 1.76; σ(ν0 ) = 1.27 × 10−24 m2 ;
h = 6.62 × 10−34 J-s; co = 3 × 108 m/s; and N i /N t = 6.
Resonator: The resonator has parameters d = 0.1 m; A = 1 cm2 ; R1 = R2 = 0.8;
and αs = 0. Its loss coefficient is αr = αs + (1/2d) ln(1/R1 R2 ) = 2.231 m−1 =
0.0223 cm−1 . The photon lifetime is thus τp = (αr c)−1 = 1.49 × 10−9 = 1.49 ns.
Threshold population difference: N t = αr /σ(ν0 ) = 1.76 × 1024 m−3 .
Peak Pulse Power: Using (16.4-14), together with N t /N i = 16 , we have
np = 12 · 6N t [1 + 16 ln 16 − 16 ] = 3 × 1.76 × 1024 × [1 + 16 ln 16 − 16 ] = 2.82 × 1024 m−3 .
Furthermore, np /N t = 3[1 + 16 ln 16 − 16 ] = 1.604, which is consistent with the
curve labeled N i /N t = 6 in Fig. 16.4-8. From (16.4-15), the peak power is P p =
hν T (c/2d) V np . Substituting c = co /n, ν = co /λo , T = 1 − R1 , and taking the
resonator cross-sectional area to be A = 1 cm2 so that the resonator volume is
V = 10−4 d = 10−5 m3 , we obtain P p = 1.38 × 109 W = 1.38 GW.
Pulse Energy: The energy of the pulse is determined from (16.4-23), which in turn
requires knowledge of the final population difference N f . To determine N f , (16.4-
22) can be rewritten in the form Y exp (−Y ) = X exp (−X) where X = N i /N t
and Y = N f /N t . Given that X = N i /N t = 6, we have X exp (−X) = 6 exp (−6) =
0.015. It follows that Y exp (−Y ) = 0.015, which has the solution Y = 0.015, so
that N f = 0.015N t . Using (16.4-23), we obtain E = 12 hν T (c/2d) V τp (N i − N f ) =
3.83 J.
Duration and Shape of Laser Pulse: The shape of the laser pulse is provided by the
curve labeled N i /N t = 6 in Fig. 16.4-8. From this figure the pulse width at half
maximum value is roughly estimated to be 1.5 τp . An approximate calculation for
the duration of the pulse can be obtained by dividing the energy [3.83 J as obtained
from (16.4-23)] by the peak pulse power [1.38 GW as obtained from (16.4-15)].
This leads to τpulse ≈ E/P p = 2.78 × 10−9 s = 2.78 ns. This calculation, which
yields τpulse ≈ 1.87τp , assumes that the pulse is square and thus provides only a
rough approximation for τpulse .
EXERCISE 16.4-3
Demonstration of Pulsing by Mode Locking
a) When the magnitudes and phases are equal, the intensity can be obtained from
(16.4-31), with the substitutions A = 1 and M = 11:
2 2
sin (M πt/TF ) sin (11πt/TF )
I(t) = |A(t)|2 = |A|2 = .
sin (πt/TF ) sin (πt/TF )
This function is plotted as a function of t/TF in Fig. (a) below. It is a periodic set of
narrow pulses of height M 2 = 121.
b) When the magnitudes are exp (−q 2 /50) and the phases are equal (say 0), the total
complex amplitude, from (16.4-28), is:
P5
A(t) = q=−5 exp (−q 2 /50) exp (jq2πt/TF )
P5
= 1 + q=1 2 exp (−q 2 /50) cos (q2πt/TF ).
The function I(t) = |A(t)|2 is plotted as a function of t/TF in Fig. (b) below. Again,
this is a periodic set of narrow pulses. Note the reduction of the side lobes in
comparison with Fig. (a).
c) Here, the magnitudes are equal and the phases are random so that A(t) =
P5
q=−5 exp (jq2πt/T F + jϕq ), where the ϕq are random variables chosen from a
uniformly distributed probability density function between 0 and 2π. A MATLAB pro-
gram was written to compute I(t) = |A(t)|2 . The random phases ϕq were generated
using the random-number generator in MATLAB. The result, plotted as a function
of t/TF in Fig. (c) below, is a random periodic function whose values typically lie
between 0 and 50.
150
(a)
100
I(t)
50
0
0 0.5 1 1.5 2 2.5 3
t /TF
100
(b)
I(t)
50
0
0 0.5 1 1.5 2 2.5 3
t /TF
40 (c)
I(t)
20
0
0 0.5 1 1.5 2 2.5 3
t /T F
C H A P T E R
17
SEMICONDUCTOR OPTICS
17.1 SEMICONDUCTORS
EXERCISE 17.1-1
Energy–Momentum Relation for a Free Electron
a) The one-dimensional time-independent Schrödinger equation for a particle of mass

m0 in a potential V = 0 (which is appropriate for a free particle) is
−}2 ∂ 2 ψ(x)
= E ψ(x) .
2m0 ∂x2
Substituting a plane-wave trial solution of the form ψ(x) = A exp(−jkx), where A is
a constant, leads to
−}2
(−jk)2 e−jkx = E e−jkx ,
2m0
}2 k 2
so that E = .
2m0
b) The relativistic energy–momentum relation for a free particle of mass m0 is

E 2 = p2 c2 + m20 c4 . (1)
For a free electron of mass m0 , the rest energy m0 c2 has a value 0.511 MeV.
For a nonrelativistic electron, it is thus convenient to carry out√ a Taylor-series
expansion for the energy E , retaining the first term. Recalling that 1 + x ≈ 1 + x/2
for x 1, we have
1/2
p2 c2
q
E= p2 c2 + m20 c4 = m20 c4 1 + 2 4
m0 c
2 2

p c
≈ m0 c2 1 +
2m20 c4
p2
= m0 c2 + .
2m0
Since m0 c2 is the rest energy of the particle, the kinetic energy of a free nonrelativis-
tic electron of mass m0 is E = p2 /2m0 . With p = }k, this becomes E = }2 k 2 /2m0 ,
which varies quadratically with k, in accordance with (17.1-1).
A free photon, on the other hand, is massless so that m0 = 0, whereupon (1)

becomes E = pc. Substituting p = }k, this becomes E = c}k, which varies linearly
with k, in accordance with (17.1-2).
The distinction results in different behavior for the dispersion diagrams of electrons
in semiconductors (Fig. 17.1-5) and photons in photonic crystals (Fig. 7.2-5).
46
EXERCISE 17.1-2
Exponential Approximation of the Fermi Function
For E − E f kT , (17.1-9) becomes the exponential function

f (E) ≈ exp [−(E − E f )/kT ]. (1)
Substituting
R∞ (1) into (17.1-11), and making use of (17.1-7) and (17.1-10), we obtain:
n = E c A(E − E c )1/2 exp [−(E − E f )/k T ] dE, (2)
where A = (2mc )3/2 /2π 2 }3 is a constant. To perform the integral in (2) we use the
transformation u = (E −E c )/k T , with du = dE/kT , whereupon exp [−(E − E f )/k T ] =
exp (−u) exp [−(E c − E f )/k T ], and the integral becomes:
Z ∞
Ec − Ef
n = A(k T )3/2 exp − u1/2 exp (−u) du
kT 0
r
4π(2mc k T )3/2

π Ec − Ef
= exp − ,
h3 4 kT
from which (17.1-12) follows. A similar analysis leads to (17.1-13), and (17.1-14)
follows by multiplication.
If mv = mc , then N c = N v , whereupon (17.1-12) and (17.1-13) provide

n/p = exp [+(E f − E v )/k T − (E c − E f )/k T ]. Thus, if (E c − E f ) < (E f − E v ), the
argument of the exponent is positive and therefore so is n/p, i.e., if E f is closer to the
conduction band than to the valence band, then n > p, and vice versa.
EXERCISE 17.1-3
Determination of the Quasi-Fermi Levels Given the Electron and Hole Concen-
trations
a) At T = 0◦ K, the Fermi function fc (E) = 1 for E < E f c and 0 otherwise. When this
is used together with (17.1-7) and (17.1-10) to evaluate the integral in (17.1-11),
we obtain:
RE
n = E cf c A(E − E c )1/2 dE = 23 A(E f c − E c )3/2 , where A = (2mc )3/2 /2π 2 }3 is a
constant. It follows that E f c − E c = (3n/2A)2/3 from which (17.1-18a) follows.
Equation (17.1-18b) can be similarly obtained.
b) The concentration n is the area under the function %c (E)fc (E). When T > 0◦ K,
fc (E) no longer assumes the values 1 and 0 with a transition at E f c (see middle
panel of figure below). However, if the quasi-Fermi level lies deep within the con-
duction band, for T > 0◦ K the product function %c (E)fc (E) will be a smooth curve
with an area close to that for the T = 0◦ K case, as is evident in the right panel of
the figure below. In that case the expression in (17.1-18a) will be approximately ap-
plicable. A parallel argument for the valence band leads to the approximate validity
of (17.1-18b).
E E E
T>0 T>0 T=0
T=0
Ec
ϱc (E ) fc(E ) ϱc (E ) fc (E )
EXERCISE 17.1-4
Electron–Hole Pair Injection in GaAs
Parameters for GaAs: E g = 1.42 eV; mc = 0.07m0 ; mv = 0.5m0 ; m0 = 9.11 × 10−31 kg,
r = 10−11 cm3 /s; T = 300◦ K.
a) Using the value of n i = 1.8 × 106 cm−3 from Table 17.1-3, together with n0 =
1016 cm−3 , we obtain p0 = n2i /n0 = 3.24 × 10−4 cm−3 . In this case n0 p0 .
b) With injection at a rate R = 1023 cm−3 s−1 , the steady-state concentrations can be
determined from (17.1-22), which provides: R = r(np − n0 p0 ) = r∆n(n0 + p0 +
∆n) ≈ r∆n(n0 + ∆n), so that ∆n2 + n0 ∆n − R/r = 0. Solving this quadratic
equation for ∆n yields: ∆n = 12 −n0 + (n20 + 4R/r)1/2 = 9.5 × 1016 cm−3 . Thus,
∆n is about 9.5 times greater than n0 .
c) Since ∆n = 9.5 × 1016 cm−3 n0 , we use (17.1-24) to obtain τ ≈ 952 ns.
d) The separation between the quasi-Fermi levels at T = 0◦ K may be determined by

subtracting (17.1-18b) from (17.1-18a):
E f c − E f v = E g + (3π 2 )2/3 (}2 /2) n2/3 /mc + p2/3 /mv .
Converting the values for n = n0 +∆n and p = p0 +∆n ≈ ∆n obtained above from
units of cm−3 to m−3 by multiplying by them by 106 , and dividing by the electronic
charge e to convert from J to eV, substitution yields the following:
(3π 2 )2/3 }2 (n × 106 )2/3 (p × 106 )2/3

Efc − Efv = Eg + +
2 m0 e 0.07 0.5
(3π 2 )2/3 }2 (10 × 1021 + 95 × 1021 )2/3 (95 × 1021 )2/3

= Eg + +
2 m0 e 0.07 0.5
2 2/3 2 14 14

(3π ) } 22.3 × 10 20.8 × 10
= Eg + +
2 m0 e 0.07 0.5
−68 14
20.8 × 1014

43.8 × 10 22.3 × 10
= E g + 4.785 · +
5.74 × 10−48 0.07 0.5
= E g + 0.013 eV .
Thus, E f c −E f v is 0.013 eV greater than the bandgap energy E g so that E f c −E f v =

1.433 eV. Using (17.1-18a) and (17.1-18b) separately, we find E f c − E c ≈ 0.011 eV
and E v − E f v ≈ 0.002 eV, so that the quasi-Fermi levels lie within, but very near the
edges of, the conduction and valence bands.
However, neither E f c − E c nor E v − E f v are k T = 0.026 eV at T = 300◦ K, so

that (17.1-18a) and (17.1-18b) should not be used for this carrier concentration at
T = 300◦ K (see Exercise 17.1-3); hence T = 0◦ K was expressly specified for this
part of the problem.
EXERCISE 17.1-5
Energy Levels of a Quantum Well
Inside the well (0 < x < d), V = 0 and the one-dimensional time-independent
Schrödinger equation is (−}2 /2m) d 2 ψ/dx2 = Eψ or d 2 ψ/dx2 + k 2 ψ = 0, where
k 2 = 2mE/}2 . This equation has the general solution ψ(x) = A sin(kx) + B cos(kx).
At the boundaries of the infinite well (x = 0 and x = d), we require ψ(x) = 0. Therefore,
B = 0 and sin(kd) = 0. This is possible if kd = qπ , q = 1, 2, 3, ..., so that k must
have one of the values kq = qπ/d, just as for the standing waves in a Fabry–Perot
resonator [see (11.1-2) and (11.1-3)]. The corresponding energy E = (}2 /2m)k 2
is thus quantized to the values E q = (}2 /2m)(qπ/d)2 . The first three energy levels
(q = 1, 2, 3) are therefore: E 1 = 4.9}2 /md 2 , E 2 = 19.7}2 /md 2 , and E 3 = 44.4}2 /md 2 .
By comparison, a quantum well of finite depth V 0 = 32}2 /md 2 has energies: E 1 =

3.2}2 /md 2 , E 2 = 11.9}2 /md 2 , and E 3 = 25.9}2 /md 2 , as illustrated in Fig. 17.1-26(b).
Finiteness of the well depth is seen to compress the energy-level spacings and to yield
a continuum of energy levels above V 0 .
17.2 INTERACTIONS OF PHOTONS WITH CHARGE CARRIERS
EXERCISE 17.2-1
Requirement for the Photon Emission Rate to Exceed the Absorption Rate
a) In thermal equilibrium E f c = E f v = E f and, in accordance with (17.1-9), f (E) =

1/ {exp [(E − E f )/k T + 1]}. The difference between the emission and absorption
conditions, given by (17.2-10) and (17.2-11), respectively, is fe (ν)−fa (ν) = fc (E 2 )−
fv (E 1 ). Since fc (E) = fv (E) = f (E) in thermal equilibrium, we have fe (ν) − fa (ν) =
f (E 2 )−f (E 1 ). Because f (E) is a monotonically decreasing function of E, we obtain
f (E 2 ) < f (E 1 ) so that fe (ν) − fa (ν) < 0. Thus, fe (ν) < fa (ν), which indicates that
the rate of emission is smaller than the rate of absorption.
b) In quasi-equilibrium, we have fe (ν) − fa (ν) = fc (E 2 ) − fv (E 1 ) =
(1/ {1 + exp [(E 2 − E f c )/kT ]}) − (1/ {1 + exp [(E 1 − E f v )/k T ]}).
This is a positive quantity if exp [(E 2 − E f c )/k T ] < exp [(E 1 − E f v )/kT ], or equiva-
lently if (E 2 −E f c ) < (E 1 −E f v ), or if E 2 −E 1 < E f c −E f v . Since E 2 −E 1 = hν, the
emission rate is greater than the absorption rate if hν < E f c − E f v , or equivalently
if E f c − E f v > hν. This implies that the separation between the two Fermi levels
is greater than the bandgap energy, namely, that E f c and E f v must lie within the
conduction and valence bands, respectively.
EXERCISE 17.2-2
Wavelength of Maximum Interband Absorption
In accordance with (17.2-29), α(ν) is proportional to (hν − E g )1/2 (hν)−2 . This function
has its maximum value, νp , when its derivative with respect to ν is 0. This occurs when
−2(hνp − E g )1/2 + 21 hνp (hνp − E g )−1/2 = 0 or 14 hνp = (hνp − E g ) so that hνp = 34 E g .
To find the maximum value of the wavelength, λp , however, we need to write α(ν) as
α(λo ) instead, and then take the derivative with respect to λo . Since ν = co /λo , we have
1/2 2 1/2 2
α(λo ) ∝ (hco /λo − hco /λg ) (λo /hco ) ∝ (1/λo − 1/λg ) (λo ) . Setting the deriva-
1/2 −1/2
tive of α(λo ) equal to zero yields 2 (1/λp − 1/λg ) λp − 21 (1/λp − 1/λg ) (λ2p /λ2p ) =
3 3
0 so that 4 (1/λp − 1/λg ) λp = 1, which leads to λp = 4 λg or λp (µm) = 4 · 1.24/E g
(eV).
3
For GaAs, E g = 1.42 eV so that λp = 4
· 1.24/1.42 = 0.65 µm, which lies in the red.
In view of the results obtained in Prob. 14.4-5, we know that λp cannot necessarily
be evaluated as co /νp . In this case, however it turns out that evaluating λp in terms of
co /νp also leads to 43 co h/E g , so that both approaches yield the same result.
C H A P T E R
18
LEDS AND LASER DIODES
18.1 LIGHT-EMITTING DIODES
EXERCISE 18.1-1
Quasi-Fermi Levels of a Pumped Semiconductor
a) At T = 0◦ K, the Fermi function fc (E) = 1 for E < E f c , and 0 otherwise. This
expression may be used together with (17.1-7) and (17.1-10) to evaluate the
integral in (17.1-11). Using the substitution x = (E − E c ) to evaluate the integral,
we obtain
R Efc
∆n = Ec A(E − E c )1/2 dE = 32 A(E f c − E c )3/2 ,
where A = (2mc )3/2 /2π 2 }3 is a constant. Thus, E f c − E c = (3/2A)2/3 ∆n2/3 , from

which (18.1-12a) follows. Equation (18.1-12b) is similarly obtained, and (18.1-12c)
follows from simple subtraction, where 1/mr = 1/mc + 1/mv [see (17.2-5)]. The
calculation is the same as that provided in Exercise 17.1-3.
b) From (18.1-5)–(18.1-7), we have fe (ν) = fc (E 2 )[1 − fv (E 1 )], where E 2 = E c +

(mr /mc )(hν − E g ) and E 1 = E 2 − hν. At T = 0◦ K, the Fermi function fc (E 2 ) is
unity as long as E 2 < E f c and is 0 otherwise. Similarly, the Fermi function fv (E 1 ) is
unity for E 1 < E f v and is 0 otherwise. For hν > E g , as hν increases, we see that
E 2 increases and E 1 decreases. But as long as these two values lie below E f c and
above E f v , respectively, fc (E 2 ) = 1 and 1 − fv (E 1 ) = 1, so that fe (ν) = 1. When hν
exceeds the value E f c − E f v , we see that E 2 exceeds E f c and E 1 lies below E f v ,
so that fc (E 2 ) = 0 and 1 − fv (E 1 ) = 0, indicating that fe (ν) = 0. The function fe (ν)
is therefore a rectangular function with value 1 for E g < hν < E f c − E f v , and value
0 otherwise, as shown in Fig. (a) below.
According to (18.1-3), the rate of spontaneous emission r sp is proportional to

%(ν)fe (ν), where %(ν) ∝ (hν − E g )1/2 . Therefore, the dependence of r sp on ν is as
illustrated in Fig. (a) below for T = 0◦ K. The effect of increasing the temperature
(T > 0◦ K) is to smooth the Fermi function so that the functions fe (ν) and r sp (ν)
take the forms shown in Fig. (b) below.
T = 0° K T > 0° K
fe(ν)
ϱ(ν)
rsp(ν)
Eg Efc-Efν hν Eg Efc-Efν hν
(a) (b)
50
EXERCISE 18.1-2
Spectral Intensity of Injection Electroluminescence under Weak Injection
From (18.1-3)–(18.1-5), we have r sp (ν) = τr−1 %(ν)fe (ν), where

%(ν) = [(2mr )3/2 /π}2 ](hν − E g )1/2 and fe (ν) = fc (E 2 )[1 − fv (E 1 )].
When the Fermi distributions are approximated by their tails, we have
fc (E 2 ) ≈ exp[−(E 2 − E f c )/k T ] and 1 − fv (E 1 ) ≈ exp[−(E f v − E 1 )/k T ]
whereupon fe (ν) ≈ exp [(E f c − E f v )/k T ] · exp [−(E 2 − E 1 )/k T ]

= exp [(E f c − E f v )/k T ] · exp (−hν/k T ).
Substituting this approximate expression for fe (ν) into the above expression for r sp (ν)
leads to (18.1-13a) and (18.1-13b).
EXERCISE 18.1-3
Electroluminescence Spectral Linewidth
a) Equation (18.1-13a) may be written in the form r sp (ν) = D(k T )1/2 u1/2 exp(−u),
where u = (hν − E g )/k T . The function u1/2 exp(−u) has its peak value when its
derivative with respect to u vanishes, i.e., when −u1/2 exp(−u) + 21 u−1/2 exp(−u) =
0, from which we obtain u = 21 , i.e., (hν − E g )/k T = 12 or hν = E g + 12 k T .
b) The peak of the function u1/2 e−u occurs at u = 21 , where the function has the value
( 12 )1/2 e−1/2 . The function reaches half its peak value where
u1/2 e−u = 12 × ( 21 )1/2 e−1/2 , i.e., where u1/2 e−u = ( 12 )3/2 e−1/2 . Squaring both
sides of this equation leads to ue−2u = ( 12 )3 e−1 = 0.046. Computation shows
that the roots of this equation are approximately u1 = 0.051 and u2 = 1.84.
The difference between these values, u2 − u1 = 1.79 ≈ 1.8, corresponds to
[(hν2 − E g )/k T − (hν1 − E g )/k T ] ≈ 1.8 so that h(ν2 − ν1 ) ≈ 1.8 k T . The FWHM
spectral width is, therefore, ∆ν ≈ 1.8 k T /h, confirming (18.1-15). Note that ∆ν is
independent of ν.
c) Since ν = c/λ, we have ∆ν = −(c/λ2 )∆λ. The magnitude of the wavelength

spectral width ∆λ that corresponds to the frequency spectral width ∆ν ≈ 1.8 k T /h
is therefore ∆λ ≈ (λ2p /c)∆ν = (λ2p /c)(1.8 k T /h) = 1.8 (λ2p /hc) k T . If we express λ in
µm, and k T in eV, the foregoing equation becomes
∆λ (in µm) × 10−6 ≈ 1.8 · [λ2p (in µm2 ) × 10−12 /hc] · [k T (in eV) · e] or
∆λ (in µm) ≈ [1.8/(106 × hc/e)] · [λ2p (in µm2 )] · [k T (in eV)].
Now, since (106 × hc/e) = 1.24 and 1.8/1.24 ≈ 1.45, we obtain the final result
∆λ (in µm) ≈ 1.45 · [λ2p (in µm2 )] · [k T (in eV)], in agreement with (18.1-16).
In contrast with the frequency spectral width ∆ν, which is independent of ν, the
wavelength spectral width ∆λ increases as λ2p .
d) At T = 300◦ K, we have k T = 0.026 eV. The frequency spectral width is given by

∆ν = 1.8 kT /h = 1.8 · 0.026 · 1.6 × 10−19 /6.6 × 10−34 = 11.3 × 1012 Hz = 11.3 THz.
It is independent of the wavelength λp .
The wavelength spectral width is ∆λ (in µm) ≈ 1.45 · [λ2p (in µm2 )] · [k T (in eV)]. For
λp = 0.8 µm, we have ∆λ ≈ 1.45 · [0.82 ] · [0.026] ≈ 0.024 µm = 24 nm. For λp =
1.6 µm, on the other hand, we have ∆λ ≈ 1.45 · [1.62 ] · [0.026] ≈ 0.096 µm = 96 nm,
confirming that ∆λ increases as λ2p (doubling the wavelength, from 0.8 to 1.6 µm,
results in quadrupling of the wavelength spectral width, from 24 to 96 nm).
EXERCISE 18.1-4
Extraction of Light from a Planar-Surface LED
a) We begin with η3 = 12 (1 − cos θc ) and make use of Snell’s law for the critical angle:
p
sin θc = 1/n and therefore cos θc = 1 − sin2 θc
p
so that η3 = 12 1 − 1 − 1/n2 .
1/2
1 1 1
Since 1 − 2 ≈ 1 − 2 for 1, we have
n 2n n2

1 1
η3 ≈ 21 = .
2n2 4n2
b) θc = sin−1 (1/n) and η3 ≈ 1/4n2 so

θc (GaAs) = sin−1 (1/3.6) = 16.1◦ and η3 (GaAs) = 0.019
θc (GaN) = sin−1 (1/2.5) = 23.6◦ and η3 (GaN) = 0.040
−1 ◦
θc (polymer) = sin (1/1.5) = 41.8 and η3 (polymer) = 0.111.
c) From GaAs (n1 = 3.6) to polymer (n2 = 1.5), the critical angle θc1 is obtained from
n1 sin θc1 = n2 so that θc1 = sin−1 (1.5/3.6) = 24.6◦ . Thus, η3 = 12 [1 − cos(24.6◦ )] =
0.045. As shown in part b) above, light escaping from GaAs into air has η3 (GaAs) =
0.019 so the enhancement in the fraction of extracted light is 0.045/0.019 ≈ 2.4.
d) From n1 = 3.6 to n2 = 1.5, using generalizations of (18.1-21) and (18.1-22) we have:

 s 
2
(n1 − n2 ) 2

1 n2
η2 η3 = 1 − · 1− 1− 
(n1 + n2 ) 2 2 n1
4n1 n2 1 n22 n32

≈ 2
· 2
= .
(n1 + n2 ) 4 n1 n1 (n1 + n2 )2
Similarly, from n2 = 1.5 to n3 = 1, we have:
n33
η20 η30 ≈ .
n2 (n2 + n3 )2
∂(η2 η3 η20 η30 )

The product η2 η3 η20 η30 is maximized for = 0 or
∂n2
n32 n33 n33 ∂ n22

∂
· = = 0.
∂n2 n1 (n1 + n2 ) 2 n2 (n2 + n3 ) 2 n1 ∂n2 (n1 + n2 ) 2 (n2 + n3 ) 2
Thus, (n1 +n2 ) 2 (n2 +n3 ) 2 ·2n2 −n22 [(n1 +n2 ) 2 ·2(n2 +n3 )+(n2 +n3 ) 2 ·2(n1 +n2 )] = 0,
which provides n2 = n1 , indicating that the introduction of an intermediate layer of
arbitrary thickness is not helpful in maximizing the fraction of light emitted from the
LED into air if Fresnel reflection is accommodated.
The use of an intermediate-index material in the form of a quarter-wave film can be

useful in this connection, however, as shown in Exercise 7.1-1.
C H A P T E R
19
PHOTODETECTORS
19.6 NOISE IN PHOTODETECTORS
EXERCISE 19.6-1
Signal-to-Noise Ratio of a Resistance-Limited Receiver
Parameters: η = 1; RL = 50Ω; T = 300◦ K; B = 100 MHz = 108 Hz; e = 1.6 × 10−19 C;

λ = 1.55 µm = 1.55 × 10−6 m; h = 6.63 × 10−34 J · s; k = 1.38 × 10−23 J/◦ K.
Resistor thermal-noise current variance: σi2 = 4k T /RL .

Photoelectron-noise current variance: 2eıB = 2e2 ΦB.
When the two variances are equal, we have 4k T /RL = 2e2 ΦB, so that the
Photon flux Φ = 2k T /e2 RL B = 6.5 × 107 photons/sec, and the
Optical power P = hνΦ = hcΦ/λ = 8.3 × 10−12 W = 8.3 pW.
EXERCISE 19.6-2
Sensitivity of an Analog APD Receiver
2 2 2
From (19.6-39), we have SNR0 = G m0 /(G F m0 + σq2 ), from which we obtain
2 2
m0 − SNR0 F m0 − SNR0 σ 2 /G = 0.
This is a quadratic equation in m0 whose positive solution is

q
1 2 2
2
m0 = 2 F · SNR0 + F SNR0 + 4σq SNR0 /G2
q
1 1 2
= F · SNR0 2
+ 4
+ σq2 /F 2 G SNR0 .
In the limit as σq2 → 0, this reduces to m0 = F · SNR0 as promised.
EXERCISE 19.6-3
Effect of Quantum Efficiency and Background Noise on Receiver Sensitivity
a) State 0: Neither signal nor background photons are present. Hence, the probability
is unity that zero photoelectrons are detected in this OOK system; there is thus
no possibility for error and p 0 = 0.
State 1: An average of n photons is present in a receiver counting time T . This gives
rise to an average of m = ηn photoelectrons, which follow the Poisson distribu-
m
tion p(m) = m exp(−m)/m! . An error (a “miss”) occurs if zero photoelectrons
are observed in the receiver counting time T ; this occurs with probability
p1 = p (0) = exp(−m) = exp(−ηn).
The bit error rate for this system is thus BER = 12 (p1 + p 0 ) = 21 exp(−ηn) =
1
2
exp(−2ηn0 ) since ηn0 = 21 ηn. For a BER = 10−9 , we thus have ηn0 = 10,
corresponding to m0 = 10 photoelectrons per bit and to n0 = 10/η photons per bit.
53
b) State 0: The number of photons n is Poisson distributed with mean nB associated

with the background, and so too is the number of photoelectrons m since the
quantum efficiency η is assumed to be unity.
State 1: The number of photons n is Poisson distributed with mean nB + n, where
n represents the mean number of signal photons.
Decision rule: Select a threshold nth ; If n > nth , decide 1; otherwise decide 0.
Error probabilities:
p0 = probability that n > nth if p(n) is Poisson distributed with mean nB .
p1 = probability that n ≤ nth if p(n) is Poisson distributed with mean nB + n.
BER = 12 p0 + 12 p1
∞ nth
n
X X
= 1
2
nB exp[−nB ]/n! + 1
2
(nB + n)n exp[−(nB + n)]/n!.
n=nth n=0
The expression for the BER is a function of nB , n, and nth . The required plots
can be generated numerically; for given values of nB and n, we can determine the
value of nth that minimizes the BER. The optimal threshold turns out to be nth =
n/ ln(1 + n/nB ) as shown, for example, in B. E. A. Saleh, Photoelectron Statistics,
Springer-Verlag, 1978, p. 315 (in this reference, BER is denoted Pe , nB is denoted
nb , and n is denoted ns ).
C H A P T E R
20
ACOUSTO-OPTICS
20.2 ACOUSTO-OPTIC DEVICES
EXERCISE 20.2-1
Parameters of Acousto-Optic Modulators
Modulator 1
n = 1.46; vs = 6×103 m/s; f = 50 MHz = 5×107 Hz; λo = 633×10−9 m; δθ = 10−3 rad;
λ = λo /n = 433 × 10−9 m; Λ = vs /f = 1.2 × 10−4 m.
Bragg Angle θB = sin−1 (λ/2Λ) = 1.8 mrad.
Bandwidth B = vs /D = vs /(λ/δθ) = 13.9 MHz.
Modulator 2
n = 4.8; vs = 2.2 × 103 m/s; f = 100 MHz = 108 Hz; λo = 10.6 × 10−6 m; D = 10−3 m;
λ = λo /n = 2.2 × 10−6 m; Λ = vs /f = 22 × 10−6 m.
Bragg Angle θB = sin−1 (λ/2Λ) = 50 mrad.
Bandwidth B = vs /D = 2.2 MHz.
EXERCISE 20.2-2
Parameters of an Acousto-Optic Scanner
Parameters: vs = 6 × 103 m/s; n = 1.46; λo = 633 × 10−9 m; fmin = 4 × 107 Hz;

fmax = 6 × 107 Hz; N = 100.
Beam width D: From (20.2-8) we have N = T B = (D/vs )B, where B = fmax − fmin =
2 × 107 Hz. Therefore, D = N vs /B = 3 cm.
Scan angle ∆θ: Since N = ∆θ/δθ and δθ = λ/D, we have ∆θ = N λ/D. This
is the angle within the medium. The corresponding angle outside the medium is
nN λ/D = N λo /D = 2.11 mrad = 0.12◦ .
Slower sound: We have N = (D/vs )B, which is inversely proportional to vs . Thus,

if vs is reduced from 6 to 3.1 km/s, with all other parameters remaining the same, N
increases from 100 to 100 × 6/3.1 = 193.5.
EXERCISE 20.2-3
Resolving Power of an Acousto-Optic Filter
Let θB = sin−1 (λ/2Λ) be the Bragg angle at wavelength λ. Consider the consequences
of fixing the angle θ at the value θB and altering the wavelength λ. The Bragg angle
is then altered and since θ is no longer the Bragg angle, the reflection efficiency
decreases. Considering small angles, it is evident from Fig. 20.1-3 that when θ differs
from θB by λ/2L, where L is the length of the cell, the reflection efficiency diminishes
to zero. This occurs when λ/2Λ − λ/2Λ ≈ λ/2L. Defining ∆λ = λ − λ as the minimum
resolvable wavelength difference, we thus have ∆λ/2Λ ≈ λ/2L, so that ∆λ/λ ≈ Λ/L =
(1/f )(vs /L) = 1/f T , where T is the transit time. It follows that the spectral resolving
power of the acousto-optic filter is given by λ/∆λ = f T .
55
20.3 ACOUSTO-OPTICS OF ANISOTROPIC MEDIA
EXERCISE 20.3-1
Transverse Acoustic Wave in a Cubic Crystal
As indicated in Example 20.3-2, all strain components of the transverse acoustic

wave vanish except s13 = s31 = S0 cos(Ωt − qz). In accordance with Table 21.2-1 for
contracted indices, this component is denoted s5 .
The photoelasticity matrix for the cubic crystal is provided in (20.3-4) so that the
components of the impermeability tensor η are given by
    
η11 p11 p12 p12 0 0 0 0
 η22   p12 p11 p12 0 0 0   0 
 η   p p p 0 0 0  0
    
∆  33  =  12 12 11  .
 η32   0 0 0 p44 0 0   0 
 η31   0 0 0 0 p44 0   s5 
η12 0 0 0 0 0 p44 0
The sole nonzero component is therefore ∆η31 = ∆η13 = p44 s5 .

cubic, η11 = η22 = η33 = 1/n2 .
Moreover, since the crystal isP
The index ellipsoid, given by ij ηij xi xj = 1, i, j = 1, 2, 3, may therefore be written in
the form
(x21 + x22 + x23 )/n2 + 2p44 s5 x1 x3 = 0, or
x22 /n2 + [(x21 + x23 )/n2 + 2p44 s5 x1 x3 ] = 0.

√ √
The transformation u1 = (x1 − x3 )/ 2; u3 = (x1 + x3 )/ 2; u2 = x2 yields the ellipsoid
u21 /n21 + u22 /n22 + u23 /n23 = 1, with
1/n21 = 1/n2 + p44 s5 (1)

n2 = n (2)
1/n23 = 1/n2 − p44 s5 . (3)
For p44 s5 1, Taylor-series expansions of (1) and (3) provide the desired results.
C H A P T E R
21
ELECTRO-OPTICS
21.1 PRINCIPLES OF ELECTRO-OPTICS
EXERCISE 21.1-1
Coupling-Efficiency Spectral Response √
According to (21.1-22), the switching voltage at wavelength λ0 is V0 = 3 C λ0 d/πn3 r.
The coupling efficiency at V = V0 for light at wavelength λ0 is 0. However, at a different
wavelength, say λ0 , the coupling efficiency is given by
√(21.1-23):
T = (π/2)2 sinc2 12 [1 + 3(V /V01 )2 ]1/2 , where V01 = 3 C λ0 d/πn3 r is the appropriate

switching voltage at the wavelength λ0 . Fixing the applied voltage at V = V0 and
substituting (V0 /V01 ) = (λ0 /λ0 ) leads to T = (π/2)2 sinc2 12 [1 + 3(λ0 /λ0 )2 ]1/2 .

The distance between λ0 and λ0 is conveniently framed in terms of the relative deviation
u ≡ (λ0 − λ0 )/λ0 , so that λ0 /λ0= 1/(1 + u). Expressing
the coupling efficiency in terms
of u provides T = (π/2)2 sinc2 21 [1 + 3/(1 + u)2 ]1/2 , which is plotted below. For u = 0
the coupling efficiency is 0, as expected. As |u| increases, representing increasing
wavelength deviation, T increases so that light is coupled by the device. At u = 0.1,
for example, we obtain T = 0.0127, indicating that a 10% relative wavelength deviation
away from λ0 results in a 1.27% coupling efficiency.
0.2
0 λ0 - λ0
-0.5 0 0.5 u =
λ0
21.2 ELECTRO-OPTICS OF ANISOTROPIC MEDIA
EXERCISE 21.2-1
Intensity Modulation Using the Kerr Effect
When an electric field E is applied to an isotropic material exhibiting the Kerr electro-
optic effect, the material becomes uniaxial with the optic axis along the direction of the
electric field, and with refractive indices given by (21.2-23) and (21.2-24), respectively:
no (E) = n − 21 n3 s12 E 2 and ne (E) = n − 21 n3 s11 E 2 . For a longitudinal electro-optic
modulator, the light propagates along the direction of the electric field so the refractive
index is no (E). For a cell of length d with an applied voltage V , we have E = V /d.
Phase
Shift:
2 2
2π 2π π V V
ϕ= no (E)d = nd − n3 s12 d = ϕ0 − π ,
λo λo λo d Vπ
where ϕ0 = (2π/λo )nd and Vπ = (λo d/n3 s12 )1/2 .
Phase Retardation: Since the light is traveling along the optic axis there is no phase
retardation (Vπ = ∞).
57
C H A P T E R
22
NONLINEAR OPTICS
22.1 NONLINEAR OPTICAL MEDIA
EXERCISE 22.1-1
Intensity of Light Required to Elicit Nonlinear Effects
a) The ratio of the second to first terms in (22.1-2) is 2dE/o χ, which is chosen to
be 0.01 and therefore requires E = o χ/200d = o (n2 − 1)/200d. Substituting
o = 8.85 × 10−12 F/m, along with n = 1.5 and d = 6.8 × 10−24 C/V2 for
ADP (NH4 H2 PO4 ), we obtain E = 8.13 × 10 9
p V/m. This corresponds to an
intensity I = E2 /η, where η = ηo /n and ηo = µo /o = 377 Ω. This in turn gives
I = 2.63 × 1017 W/m2 = 2.63 × 1013 W/cm2 , which is very large.
b) The ratio of the third term to the first in (22.1-2), 4χ(3) E2 /o χ, is also taken to be
0.01, which requires E2 = o χ/400χ(3) = o (n2 − 1)/400χ(3) . Substituting o = 8.85 ×
10−12 F/m, along with n = 1.6 and χ(3) = 4.4 × 10−32 Cm/V3 for CS2 , leads to
E = 8.86 × 108 V/m. The corresponding intensity is I = E2 /η = nE2 /ηo = 3.33 ×
1015 W/m2 = 3.33 × 1011 W/cm2 .
22.2 SECOND-ORDER NONLINEAR OPTICS
EXERCISE 22.2-1
Non-Collinear Type-II Second-Harmonic Generation (SHG)
From (22.2-26) we have
no (ω) sin θ1 = n(θ + θ2 , ω) sin θ2 (1)

no (ω) cos θ1 + n(θ + θ2 , ω) cos θ2 = 2n(θ, 2ω). (2) ω
Therefore e
θ1 2ω
θ2 e
n2o (ω) sin2 2
θ1 = n (θ + θ2 , ω) sin θ2 2
(3) ω Axis
o Optic
θ
n2o (ω) cos2 θ1 = [2n(θ, 2ω) − n(θ + θ2 , ω) cos θ2 ] . 2
(4)
Adding (3) and (4), we obtain
n2o (ω) = n2 (θ + θ2 , ω) + 4n2 (θ, 2ω) − 4n(θ, 2ω) n(θ + θ2 , ω) cos θ2
so that −1 −1

cos2 (θ + θ2 ) sin2 (θ + θ2 ) cos2 θ sin2 θ

n2o (ω) = + +4 + 2
n2o1 n2e1 2
no2 n e2
−1/2 −1/2
cos2 θ sin2 θ cos2 (θ + θ2 ) sin2 (θ + θ2 )

− 4 cos θ2 2
+ 2 + . (5)
n o2 n e2 n2o1 n2e1
58
For a KDP crystal, λ1 = 1.06 µm and λ2 = λ1 /2, and we have no1 = 1.494, ne1 = 1.4599,
no2 = 1.5123, ne2 = 1.4707.
The procedure for solving (5) is as follows:

a) Substitute no1 , no2 , ne1 , ne2 into (5).
b) Substitute θ from 0 → 90◦ .
c) Use a Matlab program to solve for the value of θ2 that satisfies (5).
d) Use (1) to compute θ1
A plot of the resultant values of θ1 and θ2

6
versus the angle between the optic axis θ and
the direction of the SH wave is shown in the θ2
θ1, θ2 (deg)
figure. 4 θ1
0
60 70 80 90
θ (deg)
22.3 THIRD-ORDER NONLINEAR OPTICS
EXERCISE 22.3-1
Third-Order Nonlinear Optical Media Exhibit the Kerr Electro-Optic Effect
PNL = 4χ(3) E3 = 4χ(3) [E(0) + 21 E(ω)ejωt + 21 E(ω)∗ e−jωt ]3 .
Carrying out the expansion shows that the term proportional to ejωt has amplitude
1
P (ω), where PNL (ω) = 4χ(3) [3E 2 (0)E(ω) + 21 |E(ω)|2 E(ω)].
2 NL
If |E(ω)| E(0), the second term above is negligible and PNL (ω) ≈ 12χ(3) E 2 (0)E(ω),
which can be cast in the form o ∆χ E(ω) with ∆χ ≈ 12χ(3) E 2 (0)/o .
Since χ = n2 − 1, we have ∆χ = 2n∆n and ∆n = ∆χ/2n. Thus, ∆n ≈ 6χ(3) E 2 (0)/o n,

which is equivalent to a refractive-index change associated with the Kerr electro-optic
effect given by ∆n = − 21 s n2 E 2 (0), provided that s = −12χ(3) /o n4 .
EXERCISE 22.3-2
Optical Kerr Lens
The intensity I ≈ I0 [1 − (x2 + y 2 )/W 2 ] induces a nonlinear refractive index n(I) =

n + n2 I = n + n2 I0 [1 − (x2 + y 2 )/W 2 ] in a thin sheet of material that exhibits the
optical Kerr effect. The result is a medium whose complex amplitude transmittance
is given by exp[−jko d n(I)] = exp[−jko d (n + n2 I0 )] · exp[jko d n2 I0 (x2 + y 2 )/W 2 ] =
h0 exp[jko (x2 + y 2 )/2f ], where h0 = exp[−jko d (n + n2 I0 )] and 1/2f = n2 I0 d/W 2 .
Hence, f = W 2 /2n2 I0 d, revealing that the medium acts as a lens whose focal length
f is inversely proportional to I0 .
EXERCISE 22.3-3
Optical Kerr Effect in the Presence of Three Waves
PNL = 4χ(3) E3
= 12 χ(3) [E(ω1 ) exp(jω1 t) + E ∗ (ω1 ) exp(−jω1 t)
+E(ω2 ) exp(jω2 t) + E ∗ (ω2 ) exp(−jω2 t)
+E(ω3 ) exp(jω3 t) + E ∗ (ω3 ) exp(−jω3 t)]3 .
The term that varies as exp(jω1 t) has an amplitude 21 PNL (ω1 ) where
PNL (ω1 ) = χ(3) [ 3|E(ω1 )|2 E(ω1 ) + 6|E(ω2 )|2 E(ω1 ) + 6|E(ω3 )|2 E(ω1 ) ].
Substituting I1 = |E(ω1 )|2 /2η, I2 = |E(ω2 )|2 /2η, and I3 = |E(ω3 )|2 /2η, we obtain
PNL (ω1 ) = 2ηχ(3) [ 3I1 + 6I2 + 6I3 ] E(ω1 ) = 2o n ∆n E(ω1 ), where
∆n = n2 I, n2 = 3ηχ(3) /o n = 3ηo χ(3) /o n2 , and I = I1 + 2I2 + 2I3 .
The wave travels with a velocity co /(n + ∆n) = co /(n + n2 I) controlled by the intensities
of the three waves.
22.4 SECOND-ORDER NONLINEAR OPTICS: COUPLED WAVES
EXERCISE 22.4-1
SHG as Degenerate Three-Wave Mixing
As in the non-degenerate 3-wave mixing case, we make use of (22.4-1), (22.4-2),

and (22.4-3), but here we have only two waves at frequencies ω1 = ω and ω3 = 2ω.
Substituting E = 21 {E1 exp(jωt) + E1∗ exp(−jωt) + E3 exp(j2ωt) + E3∗ exp(−j2ωt)}
into (22.4-3), we obtain
PNL = 12 {P1 exp(jωt) + P1∗ exp(−jωt) + P3 exp(j2ωt) + P3∗ exp(−j2ωt)}, where
P1 = 2dE3 E1∗ and P3 = dE1 E1 . Substituting this in turn into (22.4-2) then leads to
SNL = 12 {S1 exp(jωt) + S1∗ exp(−jωt) + S3 exp(j2ωt) + S3∗ exp(−j2ωt)}, where
S1 = µo ω 2 P1 = 2µo ω 2 dE3 E1∗ and S3 = µo (2ω)2 P3 = µo ω32 dE1 E1∗ , from which (22.4-16)
follow.
EXERCISE 22.4-2
Photon-Number Conservation: The Manley–Rowe Relations
These results follow directly from (2a), (2b), and (2c) in the solution to Exercise 22.4-3.
EXERCISE 22.4-3
Energy Conservation
Multiply (22.4-20a) by a∗1 :

a∗1 da1 /dz = −jga∗1 a∗2 a3 exp(−j∆kz). (1)
Add (1) to its conjugate and note that a∗1 da1 /dz + a1 da∗1 /dz = (d/dz)|a1 |2 , to obtain:
(d/dz)|a1 |2 = −jga∗1 a∗2 a3 exp(−j∆kz)+ c.c. (2a)
Similarly,
(d/dz)|a2 |2 = −jga∗1 a∗2 a3 exp(−j∆kz)+ c.c. (2b)
(d/dz)|a3 |2 = −jga1 a2 a∗3 exp(j∆kz)+ c.c.
= jga∗1 a∗2 a3 exp(−j∆kz)+ c.c. (2c)
Now multiply (2a) by }ω1 , (2b) by }ω2 , and (2c) by }ω3 and add the three equations:
(d/dz)[ }ω1 |a1 |2 + }ω2 |a2 |2 + }ω3 |a3 |2 ] = −jg}(ω1 + ω2 − ω3 )a∗1 a∗2 a3 exp(−j∆kz)+ c.c.
(3)
Because ω3 = ω1 + ω2 , the right-hand side of (3) vanishes and we obtain:
(d/dz)[ }ω1 |a1 |2 + }ω2 |a2 |2 + }ω3 |a3 |2 ] = (d/dz)(I1 + I2 + I3 ) = 0. (4)
EXERCISE 22.4-4
Coupled-Wave Equations for SHG
Write E1 and E3 as E1 = (2η}ω)1/2 a1 exp (−jk1 z) and E3 = (2η}2ω)1/2 a3 exp (−jk3 z),
respectively, and insert these formulas into (22.4-16a). Use of the slowly varying
envelope approximation in (22.4-19) on the resulting equation then leads to
(2η}ω)1/2 (−j2k1 )(da1 /dz) exp (−jk1 z) =

− 2µo ω 2 d(2η}ω)1/2 (4η}ω)1/2 a3 exp (−jk3 z)a∗1 exp (jk1 z),
whence (da1 /dz) = (µo ω 2 d/jk1 )(4η}ω)1/2 a3 a∗1 exp (−j∆kz) = −jga3 a∗1 exp (−j∆kz),
where ∆k = k3 − 2k1 and g = (µo ω 2 d/k1 )(4η}ω)1/2 , or g2 = (µo c ωd)2 (4η}ω) =
(ηωd)2 (4η}ω) = 4}ω 3 η 3 d2 .
Equation (22.4-27b) can be similarly obtained.
EXERCISE 22.4-5
Infrared Up-Conversion
2
Parameters: d = 1.5 × 10−22 C/V ; n = 2.6; λ1 = 10.6 × 10−6 m; λ2 = 1.06 × 10−6 m;
−8 2 −2
P 2 = 1 W; A = 10 m ; L = 10 m.
Wavelengths: Since ω3 = ω1 +ω2 , we have 1/λ3 = 1/λ1 +1/λ2 or λ3 = λ1 λ2 /(λ1 +λ2 ).

Hence, λ3 = 0.9636 × 10−6 m = 963.6 nm.
Up-conversion efficiency: As provided in (22.4-43), the up-conversion efficiency is

expressed as ηOFC = 2ηo3 ω32 (d2 /n3 )(L2 /A)P 2 . Substituting ηo = 377 Ω; ω3 = 2πco /λ3 =
1.96 × 1015 rad/s; d2 /n3 = 1.3 × 10−45 C2 /V4 ; L2 /A = 104 ; and co = 3 × 108 m/s, leads
to ηOFC ≈ 5.4 × 10−3 = 0.54%.
EXERCISE 22.4-6
Gain of an OPA
Parameters: λ1 = 2.5 µm; λ3 = 1.064 µm; L = 2 cm; G = 3 dB;
For KTP: n = 1.75 and d = 2.3 × 10−23 C/V2 .
a) Since ω2 = ω3 − ω1 , we have 1/λ2 = 1/λ3 − 1/λ1 or λ2 = λ3 λ1 /(λ1 − λ3 ). Thus,
λ2 = 1.85 µm.
r
d2
r
2 ηo 3
b) From (22.4-47), C = 2ω1 ω2 ηo3 3 = 2πco d = 9.0 × 10−5 W−1/2 .
n λ1 λ2 n
√
c) A gain of 3 dB signifies
√ G = cosh2 (γL/2) = 2 so that 2 = cosh(γL/2) and
that √
γL/2 = cosh−1 2 = ln(1 + 2). Thus,
√
γ = (2/L) ln(1 + 2) = 88 m−1 . Also, from (22.4-47) we have
p
γ = 2C P 3 /A, so that
P3 γ 2
= = 2.39 × 1011 W/m2 .
A 2C
d) If the laser power is 2.39 W, then the beam cross-sectional area is A = 10 µm2 .
22.5 THIRD-ORDER NONLINEAR OPTICS: COUPLED WAVES
EXERCISE 22.5-1
THG in the Undepleted-Pump Approximation
We begin with the Helmholtz equations (22.5-4) at the fundamental and third-harmonic
frequencies,
(∇2 + kq2 )Eq = −Sq , q = 1, 3,
where (22.5-23) provides
S1 = µo ω12 χ(3) 3E3 E1∗2 + 3E1 |E1 |2 + 2|E3 |2

S3 = µo ω32 χ(3) E13 + 3E3 |E3 |2 + 2|E1 |2 .

Using the relations Eq = Aq exp(−jkq z), q = 1, 3, and the slowly varying envelope
approximation (22.4-19), (∇2 + kq2 )[Aq exp(−jkq z)] ≈ −j2kq (dAq /dz) exp(−jkq z), the
Helmholtz equations become
dA1
= −j 32 ηω1 χ(3) A3 A∗2 2 2

1 exp(j∆k z) + A1 |A1 | + 2|A3 |
dz
dA3
= −j 12 ηω3 χ(3) A31 exp(−j∆k z) + 3A3 |A3 |2 + 2|A1 |2 ,

dz
where ∆k = 3k1 − k3 .
Under the undepleted pump approximation (|A3 | |A1 |), the amplitude of the funda-
mental wave A1 is assumed to be approximately constant (i.e., it does not vary with z),
so the only equation of interest is
dA3
= −j 12 ηω3 χ(3) A31 exp(−j∆k z) + 3A3 |A3 |2 + 2|A1 |2 .

dz
The first term on the right-hand side of this equation represents third-harmonic
generation while the second term represents the optical Kerr effect.
This equation may simplified since |A3 | |A1 |. The term |A3 |2 in the sum |A3 |2 +2|A1 |2
may therefore be neglected. The term 6A3 |A1 |2 , which is much smaller than A31 , may
also be neglected. In any case, since A1 is constant, the term 6A3 |A1 |2 represents
a constant change of the refractive index due to the optical Kerr effect, and may be
ignored in the context of third-harmonic generation.
Thus, the final outcome is

dA3
≈ −j 21 ηω3 χ(3) A31 exp(−j∆k z).
dz
p
With the substitution Aq = 2η}ωq aq , this result may be written as
da3
q
= −jga31 exp(−j∆k z) with g = } ω13 ω3 η 2 χ(3) .
dz
22.7 DISPERSIVE NONLINEAR MEDIA
EXERCISE 22.7-1
Polarization Density for an Anharmonic-Oscillator Medium
In accordance with Newton’s law, F = ma, we have

eE − (κx + κ2 x2 ) − mζ dx/dt = m d2 x/dt2 .
Dividing by m, reordering terms, and substituting κ/m = ω02 ,

we obtain d2 x/dt2 + ζ dx/dt + ω02 x + (κ2 /m) x2 = (e/m) E. (1)
For a medium containing N atoms per unit volume, the polarization density is P = Nex.
Substituting x = P/Ne into (1), we obtain
d2 P/dt2 + ζ dP/dt + ω02 P + (κ2 /m) P2 /Ne = (Ne2 /m) E. (2)
Defining two parameters, χ0 and b, such that

ω02 0 χ0 = Ne2 /m (3)
and
ω02 0 χ0 b = κ2 /mNe, (4)
respectively, leads to (22.7-8).
Equation (3) provides that χ0 = Ne2 /mω02 0 while (4) gives b = κ2 /(mNeω02 0 χ0 ).
Finally, inserting (3) into (4) yields b = κ2 /e3 N 2 , as promised.
EXERCISE 22.7-2
Miller’s Rule
Consider the superposed waves E = Re {E(ω1 ) exp(jω1 t) + E(ω2 ) exp(jω2 t)}.

The first iteration (ignoring the nonlinear effect) gives a polarization density
P = Re {P (ω1 ) exp(jω1 t) + P (ω2 ) exp(jω2 t)}, where
P1 (ω1 ) = 0 χ(ω1 )E(ω1 ) (1a)

P1 (ω2 ) = 0 χ(ω2 )E(ω2 ) (1b)
In the second iteration, we have a driving force F = E − b P 2 , i.e.,

F = Re {E(ω1 ) exp(jω1 t) + E(ω2 ) exp(jω2 t)}
2
− b [Re {P1 (ω1 ) exp(jω1 t) + P1 (ω2 ) exp(jω2 t)}] .
This force has many components, including a component Re {F (ω3 ) exp(jω3 )} of fre-
quency ω3 = ω1 + ω2 and complex amplitude F (ω3 ) = −(b/2)P1 (ω1 )P1 (ω2 ). This
force creates a polarization density at frequency ω3 with complex amplitude P2 (ω3 ) =
0 χ(ω3 )F (ω3 ) = 0 χ(ω3 )(−b/2)P1 (ω1 )P1 (ω2 ). Substitution from (1) yields
P2 (ω3 ) = 30 (−b/2)χ(ω1 )χ(ω2 )χ(ω3 ), from which (22.7-14) follows.
C H A P T E R
23
ULTRAFAST OPTICS
23.3 PULSE PROPAGATION IN OPTICAL FIBERS
EXERCISE 23.3-1
Dispersion Compensation in Optical Fibers
a) The dispersion length z0 = πτ02 /Dν , where Dν = −(λ2o /co )Dλ , so that z0 =
πτ02 co /λ2o Dλ . For the first fiber segment we have:
2
Dλ = 20 ps/km-nm = 2 × 10−5 s/m ,
−11
τ0 = 10 ps = 10 s, and
λo = 1.55 µm = 1.55 × 10−6 m,
so that z0 = 1.96 km.
At a distance d 1 = 100 km, the chirp parameter and the pulse width are, respec-
tively,
a = d 1 /z
p0 = 51
τ1 = τ0 1 + (d 1 /z0 )2 ≈ 510 ps.
b) The dispersion compensation condition is d 1 Dλ + d 2 Dλ0 = 0 so that d 2 =

−d 1 Dλ /Dλ0 . If the dispersion coefficient of the second fiber segment is Dλ0 = −100
ps/km-nm, we have d 2 = 100 km · (20/100) = 20 km.
EXERCISE 23.3-2
Dispersion Compensation by Use of a Periodic Sequence of Phase Modulators
The effect of GVD on pulse propagation over the distance d between its minimum
width (where it is unchirped) and its maximum width is described by the following
equations (see Table 23.3-1):
√
Pulse width: τ = τ0 1 + a2 (1)
Chirp parameter: a = z/z0 (2)
Dispersion length: zo = πτo2 /Dν . (3)
The quadratic phase modulator does not alter the pulse width, but it changes the chirp
parameter. A change by a factor of −2a is obtained if
−2a = ζτ 2 . (4)
This change guarantees that the pulse is modified periodically, as shown in Fig. 23.3-7.
Substituting (1), (2), and (3) into (4) leads to (23.3-23).
64
C H A P T E R
24
OPTICAL INTERCONNECTS AND
SWITCHES
24.1 OPTICAL INTERCONNECTS
EXERCISE 24.1-1
Interconnection Capacity
Assume that the hologram is divided into L sub-holograms, each of which contains M
spatial harmonic functions (M 1/2 in the x direction and M 1/2 in the y direction). The
incident ray on each sub-hologram is directed into M simultaneous directions, so that
each of L points is connected to M points. If a×a is the area of the hologram, then a2 /L
is the area of the sub-hologram. A width a/L1/2 corresponds to a spatial frequency
uncertainty ∆ν = L1/2 /a (or angular uncertainty λL1/2 /a). The M harmonic functions
on a sub-hologram must be separated from one another by a spatial frequency equal
to the uncertainty (L1/2 /a) in each direction, so that the spatial bandwidth B in one
direction must be at least M 1/2 L1/2 /a. It follows that B ≥ M 1/2 L1/2 /a or (Ba)2 ≥ M L.
If B = 1000 lines/mm and a = 1 mm, then (Ba)2 = 1000. If every point at the input
connected to every point at the output plane, i.e., if L = M , then M 2 ≤ (Ba)2
plane is p
or M ≤ (Ba)2 = 31.6. Thus, at most, each of 31 points at the input are connected to
each of 31 points at the output.
EXERCISE 24.1-2
The Logarithmic Map
The local spatial frequencies are

νx = (1/2π)∂ϕ/∂x = (1/λd)(ln x + 1 − 1 − x) = (1/λd)(ln x − x),
νy = (1/2π)∂ϕ/∂y = (1/λd)(ln y − y).
The angles of deflection are therefore

θx = λνx = (1/d)(ln x − x), θy = λνy = (1/d)(ln y − y).
Rays originating at location (x, y) at the hologram thus reach the location (x0 , y 0 ) in
a plane a distance d away via x0 = x + θx d = ln x, y 0 = y + θy d = ln y, thereby
indicating that the transformation x0 = ln x and y 0 = ln y is implemented.
The
R phase function ϕ(x, y) specified in (24.1-9) is obtained by recognizing that
ln(x) dx = x ln(x) − x.
65
24.4 PHOTONIC LOGIC GATES
EXERCISE 24.4-1
Nonlinear Transmittance Functions that Exhibit Bistability
a) T(x) = 1/[(x -1)2 + a2], a = 0.1.

100 2
T(x) Io
0 0
0 2 0 1
x Ii = Io /T(Io)
b) T(x) = 1/[1 + a2 sin2 (x +θ)], a = 5, θ = π/4 .
1 3
T(x) Io
0 0
0 3 0 40
x Ii = Io /T(Io)
c) T(x) = (1/2)[1 + cos (x +θ)], θ = 3π/4.
1 6
T(x) Io
0 0
0 x 6 0 Ii = Io /T(Io) 30
d) T(x) = sinc2 [(a2 +x2)1/2], a = 2.

0.02 3
T(x) Io
0 0
0 3 0 1000
x Ii = Io /T(Io)
e) T(x) = (x +1)2 /(x +a)2, a = -5.

4 3
T(x) Io
0 0
0 3 0 5
x Ii = Io /T(Io)

(答案) 官方ExerciseSolutionsManual

Uploaded by

Copyright:

Available Formats

(答案) 官方ExerciseSolutionsManual

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(答案) 官方ExerciseSolutionsManual

Uploaded by

Copyright:

Available Formats

Saleh & Teich Fundamentals of Photonics, Third Edition: Exercise Solutions ©2019 page i

FEBRUARY 20, 2019

MALVIN CARL TEICH

JOHN WILEY & SONS, INC.

The pathlength is a function of θ1 and θ2 , which are related by

d 1 tan θ1 + d 2 tan θ2 = d. (2)

The pathlength is minimized when ∂ [n1 d 1 sec θ1 + n2 d 2 sec θ2 ] = 0,

From (2), we have ∂ [d tan θ + d tan θ ] = 0,

1.2 SIMPLE OPTICAL COMPONENTS

A ray originating at P1 = (y1 , z1 ) at angle θ1 meets the mirror at height

Given z1 and z2 , (1) relates y to z and thus defines the surface.

At the second surface, the angle is altered again to θ2 = nθ − 1 − n y , (2)

Substituting (1) into (2) we obtain:

Using (1.2-11), we invoke θ2 = θ1 − (y/f ). (3)

If θ1 = 0, then θ2 = (−y/f ), and z2 ≈ (y/− θ2 ) = f , where f is the focal length. In

a) The rays within the six cones of half angle θc = θc

1.3 GRADED-INDEX OPTICS

For a step-index fiber

1.4 MATRIX OPTICS

• If A = 0, then y2 = B θ1 , i.e., for a given θ1 , we

• If C = 0, then θ2 = D θ1 , i.e., we see that all θ1 θ2

• If D = 0, then θ2 = C y1 , i.e., we see that all rays y1

a) This system is composed of 5 subsystems:

The ray transfer matrix of subsystem 2) is:

The ray transfer matrix of subsystems 2) and 3) is:

The ray transfer matrix of the entire system is:

Grouping together the terms proportional to d 1 , d 2 , and d 1 d 2 , we have

Using the definitions

(1) becomes: (d 1 + d 2 )(1 − h/f ) − d 1 d 2 /f + d/n = 0. (4)

We now rewrite (4) in terms of z1 and z2 by substituting d 1 = z1 − h and d 2 = z2 − h.

b = d/n − h2 /f − 2h(1 − h/f ) = d/n + h2 /f − 2h

f − h in air. This composite system has ray transfer matrix

We now examine A = (1 − d/nf1 ) + (f − h)(−2/f1 + d/nf12 ), and show that it is 0.

The A and D elements of this product are:

Given: λ = 633 × 10−9 m, d = 1 m.

A = (A0 /z) exp[−jk(x2 + y 2 )/2z], (1)

2.4 SIMPLE OPTICAL COMPONENTS

Substituting d(x, y) ≈ αx into (2.4-5) leads to the desired result.

jko (x2 + y 2 ) jko (x2 + y 2 )

U1 (x, y) = exp (−jkz), and t(x, y) = h0 exp [jk (x2 + y 2 ) /2f ].

= h0 exp −jk z − x2 − 2f θx + y 2 /2f .

This is a paraboloidal wave centered about the point (f θ, 0, f ), as illustrated below.

U2 (x, y) ≈ exp −jk x2 + y 2 /2z1 exp jk x2 + y 2 /2f

Substituting n = n0 [1 − α2 (x2 + y 2 ) /2] into (2.4-14), we obtain

t = exp (−jnko d 0 ) = h0 exp [jn0 α2 ko d 0 (x2 + y 2 ) /2], with h0 = exp (−jn0 ko d 0 ).

This is the amplitude transmittance of a lens of focal length f .

ϕ = k (x2 + y 2 ) /2z = π (x2 + y 2 ) /λd.

Therefore, I = 2I0 [1 + cos (2πxθ/λ)], where θ = 2a/d.

The phase difference between two reflections is ϕ = k (Λ2 − Λ1 ).

For ϕ = 2π, we have kΛ sin θ = π so that 2Λ sin θ/λ = 1, or equivalently,

2.6 POLYCHROMATIC AND PULSED LIGHT

b) ϕ = k (z2 − z1 ) = k (2v t) = (2πν/c) 2v t = 2π 2 vc ν t, so that the beat frequency is

3.1 THE GAUSSIAN BEAM

Given: λ = 633 nm = 633 × 10−9 m; P = 10−3 W; W0 = 0.05 × 10−3 m

a) θ0 = λ/ (πW0 ) = 4.03 × 10−3 rad = 4.03 mrad.

c) At beam center, I = I0 = 2P/πW02 = 2.546 × 105 W/m2 = 25.46 W/cm2 .

The condition (2.2-21) is ∂A/∂z  kA.

In summary, the conditions z0  λ, W0  λ, and θ0  1 guarantee that ∂A/∂z  kA

to obtain W 2 /R = (z/z0 ) W02 /z0 = (z/z0 ) (λ/π), from which

The condition (2.2-21) is ∂A/∂z kA.

In summary, the conditions z0 λ, W0 λ, and θ0 1 guarantee that ∂A/∂z kA