482 Chapter 22 Theory of Multilayer Films

The reflectance R, which determines the reflected irradiance, is defined by

R = ƒrƒ2 (41)

To calculate R, first notice that the reflection coefficient r is complex and that
it has the general form
A + iB
r =
C + iD
so that
A + iB A - iB A2 + B2
ƒ r ƒ 2 = rr* = a ba b = 2
C + iD C - iD C + D2

By inspection then, we may write

normal incidence
(42)
n211n0 - ns22cos2 d + 1n0ns - n2122sin2 d
R =
n211n0 + ns22cos2 d + 1n0ns + n2122sin2 d

Example 1
A 400-Å-thick film of ZrO2 1n = 2.102 is deposited on glass 1n = 1.502. De-
termine the normal reflectance for sodium light of wavelength l0 = 589.3 nm.
Solution
The phase difference is given by

2p 2p
d = 1n1t2 = 12.121402 = 0.8956 rad
l0 589.3

so that cos d = 0.6250 and sin d = 0.7806. Then, substituting into Eq. (42),

2.1211 - 1.52210.625022 + [11211.52 - 2.12]210.780622

R = = 0.174
2.1211 + 1.52210.625022 + [11211.52 + 2.12]210.780622

That is, the irradiance of the reflected beam is 17.4% of the irradiance of
the incident beam.

A plot of reflectance versus the optical path difference ¢ 1 = n1t associat-

ed with one traversal of the film is shown in Figure 2, where the abscissa is cal-
ibrated in ratios of ¢ 1>l, with l = l0>n1 being the wavelength in the film. Each
curve corresponds to a different film index, but the glass substrate index has
been chosen ns = 1.52 in all cases. The magnitude of the film index n1 evidently
determines whether the reflectance is enhanced (for n1 7 ns) or reduced (for
n1 6 ns) from that for uncoated glass. The curves show that quarter-wave
thicknesses, or odd multiples thereof, lead either to optimum enhancement
(high-reflectance coating) or to maximum reduction (antireflection coating).
These minima or maxima points in R can be made to occur at various wave-
lengths by changing ¢ 1 through selection of the film thickness. Notice that for
¢ 1 = l>2 or any even multiple of a quarter-wavelength, the reflectance is just
that from the uncoated glass. An antireflecting single coat, with n1 6 ns , never
reflects more than the uncoated glass at any wavelength. The periodic variation
in R with ¢ 1 , which is proportional to the film thickness, provides a practical
way of monitoring film thickness in the course of a film deposition.
 Theory of Multilayer Films 483

20

18

n1 ! 2
16

14
n1 ! 1.8

12
R (%)

10

8
n1 ! 1.6
6

4
n1 ! 1.4
2 Figure 2 Reflectance from a single film
n1 ! 1.2 layer of index of refraction n1 versus nor-
malized path difference. The dashed line
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 represents the uncoated glass substrate of
Path difference/wavelength index ns = 1.52.

The important case of quarter-wave film thickness,

l l0
t = =
4 4n1

makes the phase difference, Eq. (15), d = 2pn1t>l0 = p>2, so that cos d = 0
and sin d = 1. In this case, Eq. (42) reduces to

n0ns - n21 2
normal incidence quarter-wave thickness R = a b
n0ns + n21
(43)

From Eq. (43), it follows that a perfectly antireflecting film can be fabricated
with a coating of l>4 thickness and refractive index n1 = 1n0ns . If the sub-
strate is glass, with ns = 1.52, the ideal index for a nonreflecting coating is
n1 = 1.23, assuming an ambient with n0 = 1. A compromise choice among
available coating materials is a film of MgF2 , with n1 = 1.38. For this film,
Eq. (43) predicts a reflectance of 1.3% in the visible region, where the
uncoated glass (set n1 = n0) would reflect about 4.3%. This difference repre-
sents a significant saving of light energy in an optical system where multiple
surfaces occur. For example, after only six such interfaces, or three optical
components in series, 93% of the incident light survives in the case of MgF2
coatings, compared with 77% in the case of uncoated glass.

3 TWO-LAYER ANTIREFLECTING FILMS

Durable coating materials with arbitrary refractive indices are, of course, not
immediately available. Practically speaking then, single films with zero re-
flectances cannot be fabricated. By using a double layer of quarter-wave-thick-
ness films, however, it is possible to achieve essentially zero reflectance at one
wavelength with available coating materials. At normal incidence, the transfer
 484 Chapter 22 Theory of Multilayer Films

matrix of a single film of quarter-wave thickness is

i
0
g1
M1 = J K
ig1 0

The transfer matrix M for two such layers is found, according to Eq. (26), by
forming the product

g2
i i - 0
0 0 g1
M = M1M2 = g1 g2 = D T
J KJ K g1
ig1 0 ig2 0 0 -
g2

Matrix components are m11 = - g2>g1 , m22 = - g1>g2 , and m12 = m21 = 0.
Using these values in Eq. (36), the result is

g22g0 - gsg21
r = (44)
g22g0 + gsg21

Incorporating the refractive indices through the use of Eqs. (12) to (14) and
then squaring to get the reflectance,

n0n22 - nsn21 2
normal incidence quarter-wave thickness R = a b
n0n22 + nsn21
(45)

Zero reflectance is predicted by Eq. (45) when n0n22 = nsn21, or

n2 ns
= (46)
n1 A 0
n

For a glass substrate 1ns = 1.522 and incidence from air 1n0 = 12, the
ideal ratio for the two films is n2>n1 = 1.23. The requirement is met quite
well using zirconium dioxide 1n2 = 2.12 and cerium trifluoride 1n1 = 1.652,
both good coating materials. The ratio of refractive indices for CeF3 and
ZrO2 of 1.27 produces a reflectance of only 0.1% according to Eq. (45). The
arrangement is shown in Figure 3 and is plotted as curve (a) in Figure 4.
Air n0 ! 1
Achieving zero reflectance at some wavelength may not satisfy the very
l
4
CeF3 n1 ! 1.65 Low index common need to reduce reflectance over a broad region of the visible spec-
l ZrO2 n2 ! 2.1 High index trum. Curve (a) is rather steep on both sides of its minimum at 550 nm. Broad-
4
er regions of low reflectance result for l>4 –l>4 coatings when the substrate
Glass ns ! 1.52 index is larger than that of the adjacent film layer, that is, ns 7 n2 . In such
cases, the index is “stepped down” consistently from substrate to ambient. In-
dices high enough to satisfy this condition are possible in infrared applications
Figure 3 Antireflecting double layer, where large values of ns are available, as in the case of germanium with ns = 4.
using l>4–l>4 thickness films. A list of useful refractive indices is given in Table 1. Broader regions of low
reflectance also become possible in the visible region of the spectrum, once the
restriction of using equal l>4 coatings is relaxed. For example, curves (b) and
(c) of Figure 4 show two such solutions to the problem, where the inner coat-
ing has a thickness of l>2, as illustrated in Figure 5. At the wavelength of
550 nm, for which the l>4 and l>2 thicknesses are determined, the l>2 layer
 Theory of Multilayer Films 485

10

8 (a)

6
R (%)

(c)
5
Uncoated glass
4

3
(b)
2
Figure 4 Reflectance from a double-layer
film versus wavelength. In all cases n0 = 1
1 and ns = 1.52. Thicknesses are determined
at l = 550 nm. (a) l>4– l>4; n1 = 1.65,
n2 = 2.1. (b) l>4 – l>2: n1 = 1.38, n2 = 1.6.
350 450 550 650 750 850 (c) l>4 – l>2: n1 = 1.38, n2 = 1.85.
Wavelength (nm)

TABLE 1 REFRACTIVE INDICES FOR SEVERAL COATING MATERIALS

n0 ! 1
Material Visible 1 ! 550 nm2 Near infrared 1 ! 2 mm2 l
4
n1 ! 1.38
Cryolite 1.30–1.33 — l n2 ! 1.6 (b); 1.85 (c)
MgF2 1.38 1.35 2
SiO2 1.46 1.44
SiO 1.55–2.0 1.5–1.85
Al2 O3 1.60 1.55 ns ! 1.52
CeF3 1.65 1.59
ThO2 1.8 1.75
Nd2 O3 2.0 1.95
Figure 5 Antireflecting double layer using
ZrO2 2.1 2.0
l>4 – l>2 thickness films. Reflectance
CeO2 2.35 2.2
curves are shown in Figure 4.
ZnS 2.35 2.2
TiO2 2.4 —
Si — 3.3
Ge — 4.0

has no effect on the reflectance and the double layer behaves like a single l>4
layer with R = 1.3%. At nearby wavelengths, however, the l>2 layer helps to
keep R below values attained by a single l>4 layer alone. For n = 1.85 [curve
(c)], two minima near R = 0 appear. Although reflectance at 550 nm is about
1.3%, greater than for the l>4 –l>4 coating of curve (a), it remains at values
less than this over the broad range of wavelengths from about 420 to 800 nm.
For n2 = 1.6 [curve (b)], the spectral response of the double layer, while more
reflective, is flatter over the visible spectrum. Still other practical solutions for
double-layer antireflecting films become possible if the thicknesses of the lay-
ers are allowed to have values other than multiples of l>4.
The curves of Figure 4 have been calculated using the theory presented
in this chapter. The overall transfer-matrix elements are first determined
by forming the product of the transfer matrices of the individual layers.
In these elements, the phase difference d is expressed as a function of l, and
 486 Chapter 22 Theory of Multilayer Films

l n0 ! 1 the film thickness is determined by the l>4 or l>2 requirement at a single

4 wavelength. These matrix elements are then used in Eq. (36) for the reflec-
n1 ! 1.38 tion coefficient. When squared, the reflectance as a function of wavelength is
l n2 ! 2.02 determined. Although the calculations can be tedious, they are easily done
4
n3 ! 1.8 using a programmable calculator or computer.
l
4
ns ! 1.52 4 THREE-LAYER ANTIREFLECTING
FILMS

(a) The procedure just outlined can be used to calculate the spectral reflectance
of three-layer films as well. The use of three or more layer coatings makes
n0 ! 1 possible a broader, low-reflectance region in which the response is flatter. If
l
each of the three layers is of l>4 thickness, one can show that a zero reflectance
n1 ! 1.38
4 occurs when the refractive indices satisfy
n2 ! 2.2 l
2 n1n3
= 2n0ns (47)
n3 ! 1.7 l
4
n2

One such practical solution is shown in Figure 6a and plotted as curve (a) in
ns ! 1.52
Figure 7. Some improvement results when the middle layer is of l>2 thick-
ness, as in Figure 6b and curve (b) of Figure 7.
(b)
Figure 6 Antireflecting triple layers. (a) 5 HIGH-REFLECTANCE LAYERS
Quarter-quarter-quarter wavelength layers.
(b) Quarter-half-quarter wavelength layers. If the order of the layers in a l>4 – l>4 double-layer film optimized for antire-
Reflectance curves are shown in Figure 7.
flection is reversed, so that the order is air–high index–low index–substrate,
all three reflected beams are in phase on emerging from the structure, and the
reflectance is enhanced rather than reduced. A series of such double layers
increases the reflectance further, and the structure is called a high-reflectance
stack, or dielectric mirror.

10

6
R (%)

5
Uncoated glass
4

3
(a)
2
(b)
Figure 7 Reflectance from triple-layer films
versus wavelength. In all cases n0 = 1 and 1
ns = 1.52. Thicknesses are determined at
l = 550 nm. (a) l>4 – l>4: n1 = 1.38,
n2 = 2.02, n3 = 1.8. (b) l>4 –l>2 –l>4: 350 450 550 650 750 850
n1 = 1.38, n2 = 2.2, n3 = 1.7. Wavelength (nm)
 Theory of Multilayer Films 487

We derive now an expression for the reflectance of this type of struc-

ture, shown schematically in Figure 8, where High and Low signify high- n0
and low-refractive indices, respectively. The transfer matrix for one double High
l
layer of l>4-thick coatings at normal incidence is the product of the individ- Low 4
ual film matrices, just as in the case of the double-layer antireflecting films: High
l
Low
4
MHL = MHML

or High
-gL Low
i i 0
0 0 gH (48)
MHL = gH gL = D T ns
J KJ K - gH
igH 0 igL 0 0
gL
Figure 8 High-reflectance stack of double
layers with alternating high- and low-
For N similar double layers in series,
refractive indices. Reflectance curves are
shown in Figure 9.
M = 1MH1ML121MH2ML22 Á 1MHNMLN2 = 1MHML2N = 1MHL2N

(49)

Substituting the double-layer matrix, Eq. (48),

N
-gL - gL N
0 a b 0
gH gH
M = D T = D T
- gH - gH N
0 0 a b
gL gL

For normal incidence,

gL nL gH nH
= and =
gH nH gL nL

so that

- nL N
a b 0
nH
M = D T (50)
-nH N
0 a b
nL

The matrix elements of the transfer matrix representing N high-low double

layers of l>4-thick coatings in series are thus

- nL N -nH N
m11 = a b , m22 = a b , m12 = m21 = 0
nH nL
(51)

Using these matrix elements in the expression for the reflection coefficient,
Eq. (36), we arrive at

n01- nL>nH2N - ns1- nH>nL2N

r = (52)
n01- nL>nH2N + ns1- nH>nL2N
488 Chapter 22 Theory of Multilayer Films

When the numerator and denominator of Eq. (52) are next multiplied by the
factor 1 - nL>nH2N>ns and the result is squared to give reflectance, we have

1n0>ns21nL>nH22N - 1 2
Rmax = c d (53)
1n0>ns21nL>nH22N + 1

Example 2
A high-reflectance stack like that of Figure 8 incorporates six double layers
of SiO2 1n = 1.462 and ZnS 1n = 2.352 films on a glass 1n = 1.482 sub-
strate. What is the reflectance for light of 550 nm at normal incidence?
Solution
Substituting directly into Eq. (53), we get
11>1.48211.46>2.35212 - 1 2
R = c d
11>1.48211.46>2.35212 + 1

or R = 99.1%.

Equation (53) predicts 100% reflectance when either N approaches

infinity or when 1nL>nH2 approaches zero. Some data indicating these
tendencies are given in Table 2. One sees that the reflectance quickly
approaches 100% for several double layers. Since the smallest ratio of nL>nH
yields best reflectances, high-reflectance stacks may be fabricated from alternat-
ing layers of MgF2 1nL = 1.382 and ZnS 1nH = 2.352 or TiO2 1nH = 2.402.

The reflectance given in Eq. (53) represents the maximum reflectance at

the wavelength l0 , for which the layers have optical thicknesses of l0>4. For
other wavelengths the transfer matrix must be used in its general form, contain-
ing the wavelength-dependent phase differences. Spectral reflectance curves
for N = 2 and N = 6 double-layer stacks have been calculated and plotted
in Figure 9. Curve (c) shows the improvement in the maximum reflectance
that results for N = 2 stacks when an extra high-index layer is inserted
between the substrate and the last low-index layer. The width of the high-
reflectance region in these curves is nearly independent of the number of
double layers used but increases when the ratio nL>nH decreases. This ratio is
0.587 in Figure 9, representing alternating MgF2 and ZnS layers on glass.
Outside the central stopband—the region of highest reflectance—the

TABLE 2 REFLECTANCE OF A HIGH-LOW QUARTER-WAVE STACK

Reflectance for N ! 3 high-low layers Reflectance versus N when nL>nH ! 0.587 for
versus nL>nH alternating double layers of MgF2 and ZnS

nL>nH R (%) N R (%)

1.0 4.26 1 39.71

0.91 21.01 2 73.08
0.83 40.82 3 89.77
0.77 57.77 4 96.35
0.71 70.44 5 98.72
0.67 79.35 6 99.56
0.625 85.48 7 99.85
0.59 89.67 8 99.95
0.56 92.55
0.53 94.56
0.50 95.97
 Theory of Multilayer Films 489

100
(b)

90
(c)
80

70 (a)

60
R (%)

50

40

30

20 Figure 9 Spectral reflectance of a high-

low index stack for (a) N = 2 and
(b) N = 6 double layers. Curve (c) repre-
10 sents an N = 2 stack with an additional
Uncoated glass
high-index layer adjacent to the substrate.
Layers are l>4 thick at l = 550 nm. In all
350 450 550 650 750 850 cases, nH = 2.35, nL = 1.38, ns = 1.52, and
Wavelength (nm) n0 = 1.00.

reflectance oscillates between a series of maxima and minima. The center of

the stopband can be shifted by depositing layers whose thickness is l>4 at an-
other l. Except for light energy lost by absorption and scattering during pas-
sage through the dielectric layers, the percent transmission of the structure is
given by T1%2 = 100 - R1%2. Thus such structures can be designed as
band-pass filters with high spectral transmittance in the wide region of low
spectral reflectance. Narrow band-pass filters that behave like Fabry-Perot
etalons can be fabricated by separating two dielectric-mirror, multilayer struc-
tures with a spacer of, say, MgF2 film. Narrow wavelength regions that satisfy
constructive interference can be produced far enough apart in wavelength so
that all but one such region is easily filtered out by a conventional absorption
color filter. The result is a filter with a pass-band width of perhaps 15 Å and
40% transmittance.

PROBLEMS
B
1 Show that when the incident E-field is parallel to the plane 4 A single layer of SiO2 1n = 1.462 is deposited to a thick-
of incidence, g1 has the form given in Eq. (37). ness of 137 nm on a glass substrate 1n = 1.522. Determine
the normal reflectance for light of wavelength (a) 800 nm;
2 A transparent film is deposited on glass of refractive index (b) 600 nm; (c) 400 nm. Verify the reasonableness of your
1.50. results by comparison with Figure 2.
a. Determine values of film thickness and (hypothetical)
5 A 596-Å-thick layer of ZnS 1n = 2.352 is deposited on glass
refractive index that will produce a nonreflecting film
1n = 1.522. Calculate the normal reflectance of 560 nm
for normally incident light of 500 nm.
light.
b. What reflectance does the structure have for incident
light of 550 nm? 6 Determine the theoretical refractive index and thickness of
a single film layer deposited on germanium 1n = 4.02 such
3 Show from Eq. (42) that the normal reflectance of a single
that normal reflectance is zero at a wavelength of 2 mm.
half-wave thick layer deposited on a substrate is the same
What actual material could be used?
as the reflectance from the uncoated substrate.
7 A double layer of quarter-wave layers of Al2O3 1n = 1.602
1n0 - ns22
R = and cryolite 1n = 1.302 are deposited in turn on a glass
1n0 + ns22 substrate 1n = 1.522.
 490 Chapter 22 Theory of Multilayer Films

a. Determine the thickness of the layers and the normal 12 Using the materials given in Table 1, design a three-layer
reflectance for light of 550 nm. multifilm of quarter-wave thicknesses on a substrate of ger-
b. What is the reflectance if the layers are reversed? manium that will give nearly zero reflectance for normal
incidence of 2 mm radiation.
8 Quarter-wave thin films of ZnS 1n = 2.22 and MgF2 1n =
1.352 are deposited in turn on a substrate of silicon 1n = 13 Determine the maximum reflectance in the center of the
3.32 to produce minimum reflectance at 2 mm. visible spectrum for a high-reflectance stack of high-low
index double layers formed using nL = 1.38 and nH = 2.6
a. Determine the actual thickness of the layers. on a substrate of index 1.52. The layers are of equal optical
b. By what percentage difference does the ratio of the thickness, corresponding to a quarter-wavelength for light
film indices differ from the ideal? of average wavelength 550 nm. The high-index material is
c. What is the normal reflectance produced? encountered first by the incident light, as in Figure 8. As-
sume normal incidence and stacks of (a) 2; (b) 4; (c) 8 dou-
9 By working with the appropriate transfer matrix, show that
ble layers.
a quarter-wave/half-wave double layer, as in Figure 5,
produces the same reflectance as the quarter-wave layer 14 A high-reflectance stack of alternating high-low index lay-
alone. ers is produced to operate at 2 mm in the near infrared. A
stack of four double layers is made of layers of germanium
10 Write a computer program that will calculate and/or plot
1n = 4.02 and MgF2 1n = 1.352, each of 0.5-mm optical
reflectance values for a double layer under normal inci-
thickness. Assume a substrate index of 1.50 and normal in-
dence. Let input parameters include thickness and indices
cidence. What reflectance is produced at 2 mm?
of the layers and the index of the substrate. Check results
against Figure 4. 15 What theoretical ratio of high-to-low refractive indices is
needed to give at least 90% reflectance in a high-re-
11 Prove the condition given by Eq. (47) for zero reflectance
flectance stack of two double layers of quarter-wave layers
of three-layer, quarter-quarter-quarter-wave films when
at normal incidence? Assume a substrate of index 1.52.
used with normal incidence. Do this by determining the
composite transfer matrix for the three quarter layers and 16 Show that Rmax in Eq. (53) approaches 1 when either N
using the matrix elements in the calculation of the reflection approaches infinity or when the ratio nL>nH approaches
coefficient in Eq. (36). zero.
 u
u Incident light

Circularly polarized u
light

23 Fresnel Equations

INTRODUCTION

The basic laws of reflection and refraction in geometrical optics were derived
earlier on the basis of either Huygens’ or Fermat’s principles. In this chapter
we regard light as an electromagnetic wave and show that the laws of reflec-
tion and refraction can also be deduced from this point of view. More impor-
tantly, this approach also leads to the Fresnel equations, which describe the
fraction of incident energy transmitted or reflected at a plane surface. These
quantities will be seen to depend not only on the change in refractive index
and the angle of incidence at the surface but also on the polarization of the
incident light. Finally, the important differences between internal and external
reflection are clarified.

1 THE FRESNEL EQUATIONS

Consider Figure 1, which shows a ray of light incident at point P on a plane

interface—the xy boundary plane—and the resulting reflected and refracted
rays. The plane of incidence is the xz-plane. Let us assume the incident light
consists of plane harmonic waves, expressed by
B
#
E = E0ei1k r - vt2
B B B

(1)
B
where the origin of coordinates is taken to be point O. The wave vector E of
the incident wave is chosen in the + y-direction, so that the wave is linearly

491
 492 Chapter 23 Fresnel Equations

n2

Interface Et
kt

O Plane of
r Bt incidence
ut
Br Er
P

ur
kr E
n1 u
k x
B
z Normal
Figure 1 Defining diagram for incident,
reflected, and transmitted rays at an xy-
plane interface when the electric field is
perpendicular to the plane of incidence, the
TE mode.

B
polarized. The direction of the correspondingBmagnetic
B
field vector B is then
determined to B
ensure that the direction of E * B is the
B
direction of wave
propagation k. This mode of polarization,B
in which the E -field is perpendicu-
lar to the plane of incidence and the B-field lies inB the plane of incidence, is
called the transverse electric (TE) mode. If instead B is transverse to the plane
of incidence, a case to be considered later, the mode is a transverse magnetic
(TM) mode. An arbitrary polarization direction represents some linear com-
bination of these two special cases. The reflected and transmitted waves in
Figure 1 can be expressed, respectively, in forms like that of the incident
wave of Eq. (1):
B
#
Er = E0rei1kr r - vrt2
B B B

(2)
B
#
Et = E0tei1kt r - vtt2
B B B

(3)

In the boundary plane xy, where all three waves exit simultaneously,
there must be a fixed relationship between the three wave amplitudes (and
thus their irradiances) that has yet to be determined. Since such a relation-
ship cannot depend on the arbitrary choice of a boundary point Br nor a time
t, it follows that the phases of the three waves, which depend on Br and t, must
themselves be equal:

1k # Br - vt2 = 1kr # Br - vrt2 = 1kt # Br - vtt2

B B B
(4)

In particular, at the boundary point Br = 0 of Figure 1,

- vt = - vrt = - vtt
or
v = vr = vt (5)

so that all frequencies are equal. On the other hand, at t = 0 within the
 Fresnel Equations 493

boundary plane, Eq. (4) yields:

k # Br = kr # Br = kt # Br
B B B
(6)

Several conclusions can be drawn from the relations of Eq. (6). First notice
that by subtracting any two members, these relations are equivalent to

1k - kr2 # Br = 1k - kt2 # Br = 1kr - kt2 # Br = 0

B B B B B B
(7)
B B
Equation (7) requires that the vectors kr and kt lie in the plane determined
B
by the vectors k and Br . Thus all three propagation vectors are coplanar in
the xz-plane, and we conclude that the reflected and refracted waves lie in
the plane of incidence. Next, consider the first two members of Eq. (6), which
govern the relationship between the incident and reflected waves. In terms of
the angles designated in Figure 1, they are equivalent to

kr sin u = krr sin ur

Since both waves travel in the same medium, their wavelengths are identical
and so k = kr . Therefore, we have the

law of reflection: u = ur (8)

Finally, the last two members of Eq. (6) are equivalent to

krr sin ur = ktr sin ut (9)

Writing kr = v>yr = nrv>c and kt = ntv>c, Eq. (9) becomes Snell’s

law of refraction: nr sin ur = nt sin ut (10)

We continue now to specify further the situation at the boundary with

the help of boundary conditions arising out of Maxwell’s equations and treated
in texts on electricity and magnetism. We employ them here without proof.
These boundary conditions require that the components of both the electric and
magnetic fields parallel to the boundary plane be continuous as the boundary is
crossed.

Boundary Conditions for TE Waves

As mentioned earlier, TE waves have electric fields that are perpendicular to
the plane of incidence and therefore are parallel to the boundary plane sepa-
rating the two media. In terms of the choices made for the direction of the
electric fields in Figure 1, the vector amplitudes of the complex fields of
Eqs. (1)–(3) can be written as

B B B
E0 = EyN E0r = EryN E0t = EtyN (11)

Here, E, Er and Et are the complex field amplitudes associated, respectively,

with the incident, reflected, and transmitted waves. The requirement that the
component of the electric field parallel to the boundary plane be continuous
at the boundary then gives

E + Er = Et (12)
 494 Chapter 23 Fresnel Equations

The magnetic fields associated with the electric fields of Figure 1 have the
form,
B
#
B = 1B cos uxN - B sin uzN 2ei1k r - vt2
B B

B
#
Br = 1 - Br cos urxN - Br sin urzN 2ei1kr r - vt2
B B

(13)
B
#
Bt = 1Bt cos utxN - Bt sin utzN 2ei1kt r - vt2
B B

Continuity of the parallel components of the magnetic field requires that the
field amplitudes be related by

B cos u - Br cos u = Bt cos ut (14)

where
B
we have
B
made use of Eq. (8). Equations (12) and (14) are correct for
the E and B vectors as chosen
B
in Figure 1. If a different choice is made,B for
example, by reversing the E vector of the incident wave (and also B to
keep the direction of wave propagation the same), Eqs. (12) and (14) appear
with a change of signs. However, the physical import of these equations is
the same when they are interpreted in terms of their original figures.

Boundary Conditions for TM Waves

Before pursuing the significance of Eqs. (12) and (14) for the TE mode, we
parallel their development for the TM mode pictured in Figure 2. The electric
and magnetic fields in this figure can be written as
B
#
E = 1E cos uxN - E sin uzN 2ei1k r - vt2
B B

B
#
Er = 1Er cos urxN + Er sin urzN 2ei1kr r - vt2
B B

(15)
B
#
Et = 1Et cos utxN - Et sin utzN 2ei1kt r - vt2
B B

n2

Interface kt

O Bt Et Plane of
r ut incidence

Br P
kr
ur
n1 Er u k
E x
z Normal
Figure 2 Defining diagram for incident,
reflected, and transmitted rays at an xy- B
plane interface when the magnetic field is
perpendicular to the plane of incidence, the
TM mode.
 Fresnel Equations 495

and
B
#
B = - ByN ei1k r - vt2
B B

B
#
Br = BryN ei1k r - vt2
B B

(16)
B
#
Bt = - Bt yN ei1kt r - vt2
B B

Requiring continuity of the components of the electric and magnetic fields

that are parallel to the boundary gives, in this case,

-B + Br = - Bt (17)

E cos u + Er cos u = Et cos ut (18)

Reflection and Transmission Coefficients

The magnetic field amplitudes of Eqs. (14) and (17) can be expressed
in terms of the corresponding electric field amplitudes through the generic
relation
c
E = yB = a bB (19)
n

Writing the index of refraction for incident and refracting media as n1 and n2 ,
respectively, Eqs. (12), (14), (17), and (18) can be recast as follows:

E + Er = Et (20)
TE: e
n1 E cos u - n1 Er cos u = n2 Et cos ut (21)

-n1 E + n1 Er = - n2 Et (22)
TM: e
E cos u + Er cos u = Et cos ut (23)

Next, eliminating Et from each pair of equations and solving for the reflection
coefficient r = Er>E,

Er cos u - n cos ut
rTE = = (24)
E cos u + n cos ut

Er -n cos u + cos ut
rTM = = (25)
E n cos u + cos ut

where we have introduced a relative refractive index n K n2>n1 . Note that we

use subscripts to distinguish between the TE and TM cases. Finally, since n
and ut are related to u through Snell’s law, sin u = n sin ut , ut may be elimi-
nated using

n cos ut = n21 - sin2 ut = 2n2 - sin2 u (26)

The results are then

Er cos u - 2n2 - sin2 u

rTE = = (27)
E cos u + 2n2 - sin2 u
496 Chapter 23 Fresnel Equations

Er -n2 cos u + 2n2 - sin2 u

rTM = = (28)
E n2 cos u + 2n2 - sin2 u

Returning to Eqs. (20) through (23), if Er is eliminated instead of Et , similar

steps lead to the following equations describing the transmission coefficient
t = Et>E:
Et 2 cos u
tTE = = (29)
E cos u + 2n2 - sin2 u

Et 2n cos u
tTM = = (30)
E n cos u + 2n2 - sin2 u
2

Eqs. (29) and (30) can also be found more quickly by using Eqs. (20) and (22)
written in the form

tTE = 1 + rTE

ntTM = 1 - rTM

into which the results expressed by Eqs. (27) and (28) can be conveniently
substituted. Equations (27) through (30) are the Fresnel equations, giving reflec-
tion
B
and transmission coefficients, the
B
ratio of both reflected and transmitted
E-field amplitudes to the incident E-field amplitude. Note that, for normal
incidence, the reflection and transmission coefficients for the TE case are
identical to those for the TM case. This is sensible since for normal incidence
there is no distinction between the two cases.1 In practice, measured reflection
and transmission coefficients also depend on scattering losses from a nonplanar
surface.

Example 1
Calculate the reflection and transmission coefficients for both TE and TM
modes of light incident from air at 30° onto glass of index 1.60.
Solution
Using Eqs. (27) and (28),

cos130°2 - 21.62 - sin2130°2

rTE = = - 0.2740
cos130°2 + 21.62 - sin2130°2

-1.62 cos130°2 + 21.62 - sin2130°2

rTM = = - 0.1866
1.62 cos130°2 + 21.62 - sin2130°2

Using the relations below Eq. (30),

tTE = 1 + rTE = 1 - 0.2740 = 0.7260

1 - rTM 1 + 0.1866
tTM = = = 0.7416
n 1.60

1
Some texts use a different convention, in which the positive direction of the reflected electric field
for the TM case is opposite to that shown in Figure 2, leading to an expression for the reflection coef-
ficient for the TM case that differs from ours by a factor of - 1. Of course, both conventions lead to the
same physical result since the extra factor of -1 simply reverses the direction of the reflected electric field.
 Fresnel Equations 497

2 EXTERNAL AND INTERNAL

REFLECTIONS

When interpreting these equations, it is useful to distinguish between two

physically different situations:

n2
external reflection: n1 6 n2 or n = 7 1
n1
n2
internal reflection: n1 7 n2 or n = 6 1
n1

Figure 3 is a plot of Eqs. (27) through (30) for the case of external
reflection with n = 1.50. Notice that at both normal and grazing incidence—
angles of 0° and 90°, respectively—TE and TM modes have reflection coeffi-
cients of the same magnitude and transmission coefficients of the same
magnitude. Negative BvaluesB of r for both the TE and TM modes indicate a
phase change of the E- or B-field vectors on reflection and will be discussed
presently. The fraction of power P in the incident wave that is reflected or
transmitted, called the reflectance and the transmittance, respectively, de-
pends on the ratio of the squares of the amplitudes.

Pr Er 2
reflectance = R = = r2 = a b (31)
Pi E

Pt cos ut 2
transmittance = T = = na bt (32)
Pi cos u
These expressions are justified later in this chapter.
In Figure 4, reflectance is plotted as a function of the angle of inci-
dence u. The curve for the case of external reflection, TM mode, indicates that
no wave energy is reflected when the angle of incidence is near 60°. The angle
up at which RTM = 0 is known as Brewster’s angle or the polarizing angle and
takes the value,
up = tan-11n2 = tan-11n2>n12

1.0

0.8
tTM
0.6 tTE

0.4

0.2 rTM

!0.2 up " 56.31#

!0.4 rTE

!0.6

!0.8 External reflection (air-to-glass)

!1.0 Figure 3 Reflection and transmission co-

10 20 30 40 50 60 70 80 90 efficients for the case of external reflection,
Angle of incidence u with n = n2>n1 = 1.50.
 498 Chapter 23 Fresnel Equations

100

90

80

70

60

R (%)
50 External reflection
(air-to-glass)

40

Internal reflection
30
(glass-to-air)

20
TE
10 TE TM
TM

10 20 30 40 50 60 70 80 90
Figure 4 Reflectance for both external
u$p " 33.69# uc " 41.81# up " 56.31#
and internal reflection when n1 = 1 and
n2 = 1.50. Angle of incidence u

This condition is also evident in the vanishing of rTM in Figure 3 and the
vanishing of the numerator of Eq. (28). See problem 1. For the case n = 1.50
used in Figures 3 and 4, up = 56.31°. RTE does not go to zero under this con-
dition, so reflected light contains only the TE mode and is linearly polarized,
with RTE = 15%. At normal incidence 1u = 0°2, for both TE and TM modes,
Eqs. (24) and (25) simplify to give

1 - n 2
R = r2 = a b (33)
1 + n

Equation (33) gives a reflectance of 4% from an air/glass interface with

n = 1.5. Keep in mind, however, that n is a function of wavelength. As the
angle of incidence increases to grazing incidence 1u = 90°2, both RTE and
RTM become unity, although RTM remains quite small until Brewster’s angle
has been exceeded.
The reflection coefficient for the case of internal reflection is shown in
Figure 5 with n = 1>1.50, as when light encounters a glass/air interface
from the glass side. Evidence of phase changes and of a polarizing, or Brew-
ster’s, angle may also be seen here.
œ
For the case of internal reflection we give
Brewster’s angle the symbol up. Examination of Figures 4 and 5 shows that,
for the case of internal reflection, both RTE = r2TE and RTM = r2TM reach
values of unity before the angle of incidence u reaches 90°. This is the
phenomenon of total internal reflection, which occurs at the critical angle
uc = sin-11n2 = sin-11n2>n12. For the example of glass 1n = 1>1.52 used in
Figure 5, upœ = 33.7° and uc = 41.8°. When sin uc 7 n, the radical
2n2 - sin2 u is negative and both rTE and rTM are complex. Their magnitudes,
however, are easily shown to be unity in this range, giving total reflection
for u 7 uc .
 Fresnel Equations 499

0.8

0.6

0.4
TE
0.2
TM
r 0

!0.2 u$p " 33.7# uc " 41.8#

!0.4

!0.6

!0.8 Internal reflection

(glass-to-air)
!1 Figure 5 Reflection coefficient for the case
0 10 20 30 40 50 60 70 80 90 of internal reflection with n = n1>n2 =
Angle of incidence u 1>1.50.

3 PHASE CHANGES ON REFLECTION

The negative values of the reflection coefficient in Figures 3 and 5 indicate that
Er = - ƒ r ƒ E in certain situations. Evidently, the electric field vector may reverse
direction on reflection. Equivalently, in such cases there is a p-phase shift of E
on reflection, as the following mathematical argument demonstrates:
B
# B
#
Er = - ƒ r ƒ E = eip ƒ r ƒ E0ei1k r - vt2 = ƒ r ƒ E0ei1k r - vt + p2
B B

Thus in the case of external reflection, Figure 3, a p-phase shift of E occurs

at any angle of incidence for the TE mode and for u 6 up for the TM mode.
When reflection is internal, Figure 5, we conclude that a p-phase shift occurs
for the TM mode for upœ 6 u 6 uc . However, the situation in the region u 7 uc ,
where r is complex, requires further investigation. When u 7 uc = sin-11n2, the
radical in Eqs. (27) and (28) becomes imaginary, and the equations may be
written in the form

cos u - i 2sin2 u - n2
rTE = (34)
cos u + i 2sin2 u - n2

-n2 cos u + i 2sin2 u - n2

rTM = (35)
n2 cos u + i 2sin2 u - n2

The reflection coefficients can be written in polar form as r = ƒ r ƒ eif and we

shall refer to f as the phase shift on reflection. In Eq. (34), the reflection
coefficient takes the form rTE = 1a - ib2>1a + ib2. Since the real and imagi-
nary parts of the numerator and denominator are the same, except for a sign,
the magnitudes of the numerator and denominator are equal, and rTE has
unit amplitude. The phase of rTE may be investigated by expressing Eq. (34)
in complex polar form, as

e -ia
rTE = = e -i12a2
eia

where tan a = 2sin2 u - n2>cos u. So, for the TE case, the phase shift on
reflection is fTE = - 2a. A similar analysis (see problem 6) can be used to
 500 Chapter 23 Fresnel Equations

show that rTM , like rTE , has unit magnitude when the angle of incidence ex-
ceeds the critical angle and enables one to find the phase shift on total internal
reflection fTM for the TM case. The phase shifts on total internal reflection
for the two cases have the form,

fTE 2sin2 u - n2
tan a b = - (36)
2 cos u

fTM - p 2sin2 u - n2
tan a b = - (37)
2 n2 cos u

Clearly, the phase shift on reflection, for total internal reflection, may take on
values other than 0 and p, depending on the angle of incidence. The phase
shift f, as determined from Eqs. (36) and (37), is plotted in Figure 6.

3p/4

p/2
Phase shift on reflection f

TM
p/4
uc # 41.8$

u"p # 33.7$
!p/4
TE
!p/2

!3p/4 Internal reflection

(glass-to-air)

Figure 6 Phase shift f on reflection of the !p

electric field for internally reflected rays, 10 20 30 40 50 60 70 80 90
with n = n1>n2 = 1>1.5. Angle of incidence u

It happens that the relative phase shift fTE - fTM is about - 3p>4 at an
angle of incidence near 53°. Two consecutive internal reflections thus pro-
duce a relative phase shift of 21 - 3p>42 = - 3p>2 (equivalently, +p>2) be-
B
tween the perpendicular components of the E-field. Recall that circularly
polarized light consists of equal amplitude components with phases that differ
by ; p>2. Thus linearly polarized incident light with equal TM and TE compo-
nents, after two internal reflections at 53°, will be transformed into circularly
polarized light. This technique is utilized in the Fresnel rhomb (Figure 7).
Summarizing these results for the case of internal reflection,

0, u 6 upœ
p, upœ 6 u 6 uc (38)
fTM = d 2 2
2sin u - n
- 2 arctana b + p, u 7 uc
n2 cos u

0, u 6 uc
2 2 (39)
fTE = c - 2 arctana 2sin u - n b , u 7 uc
cos u

Phase shifts for both TM and TE modes and for both internal and external re-
flection are summarized in Figure 8.
 Fresnel Equations 501

u
u Incident light

Figure 7 The Fresnel rhomb. With the

incident light polarized at 45° to the plane of
incidence, two internal reflections produce
Circularly polarized u equal-amplitude TE and TM amplitudes
light with a relative phase of p>2, or circularly
polarized light. For n = 1.50, the angle
should be u = 53°. The device is effective
over a wide range of wavelengths.
Phase shift on reflection f

% Phase shift on reflection f 0 0

uc
!% !%
0
0 30 60 90 30 60 90
Angle of incidence u Angle of incidence u
(a) TE mode, external reflection (air to glass) (b) TE mode, internal reflection (glass to air)
Phase shift on reflection f

% %
Phase shift on reflection f

% %

up up" uc
0 0 0 0 Figure 8 Phase changes on reflection f
between incident and reflected rays versus
30 60 90 30 60 90
angle of incidence. Discontinuities occur at
Angle of incidence u Angle of incidence u uc = 41.8°, up = 56.3°, and upœ = 33.7° for
(c) TM mode, external reflection (air-to-glass) (d) TM mode, internal reflection (glass-to-air) refractive indices of n1 = 1 and n2 = 1.50.

Example 2
What is the phase shift of the TM and TE rays reflected both externally and
internally for the situation discussed in Example 1?

Solution
For this interface,
1
uc = sin-1 a b = 38.7°
1.6
up = tan-111.62 = 58.0°
1
upœ = tan-1 a b = 32.0°
1.6

Since the angle of incidence of 30° is less than either upœ or uc , Eqs. (38) and
(39) or Figure 8 require that for internal reflection, fTM = 0 and fTE = 0,
while Figure 8 shows that for external reflection, fTM = p and fTE = p.
502 Chapter 23 Fresnel Equations

A general conclusion can be drawn from the phase changes for the TE
and TM modes under internal and external reflection: Near normal inci-
dence, for both TE and TM modes, the phase shift for an internally reflected
beam differs from that of an externally reflected beam by p. For a thin film in
air, we are interested in the relative phase shift between rays reflected from
the first surface (external) and the second surface (internal). Inspection of
Figure 8 shows that a relative phase shift of p occurs in the TE mode for
internal angles of incidence less than uc and in the TM mode for internal
angles of incidence less than upœ . The corresponding ranges of external angles
of incidence at the first surface are 0° to 90° (TE mode) and 0° to up (TM
mode). Thus in the TE mode a relative phase shift of p occurs for all external
angles of incidence, but in the TM mode this is true only for external angles
less than up .

4 CONSERVATION OF ENERGY

To conserve energy, at a given boundary, it must be true that the power inci-
dent on the boundary be equal to the sum of the power reflected at that
boundary and the power transmitted through the boundary. That is,

Pi = Pr + Pt (40)

If we represent the reflectance R as the ratio of reflected to incident power

and the transmittance T as the ratio of transmitted to incident power,

Pr Pt
R = and T = (41)
Pi Pi

then Eq. (40) takes the form

1 = R + T (42)

The irradiance I is the power density 1W>m22, so that we may write, in place
of Eq. (40),

IiA i = IrA r + ItA t (43)

The cross-sectional areas of the three beams (see Figure 9) that appear in Eq. (43)
are all related to the area A intercepted by the beams in the boundary plane
through the cosines of the angles of incidence, reflection, and refraction. We
may then write

Ii1A cos u2 = Ir1A cos ur2 + It1A cos ut2

Of course, u = ur by the law of reflection. Also using the relation between ir-
radiance and electric field amplitude,

ey
I = a b E 20
2
and the facts that yi = yr , and ei = er , since they correspond to the same
medium, we arrive at the equation

ytet cos ut 2
E20 = E20r + a ba bE0t (44)
yiei cos u
 Fresnel Equations 503

Ai Ar

u ur

ut

At

Figure 9 Comparison of cross sections of

incident, reflected, and transmitted beams.

The quantity 1ytet>yiei2 is just a complicated way of expressing the relative re-
fractive index n, which we can show as follows:

ytet yt y2i mi yi
= 2
= = n (45)
yiei yi yt mt yt

In arriving at this result we have used

mi = mt = m0

for nonmagnetic materials and the relation

1
y2 =
me
for the velocity of a plane electromagnetic wave. Incorporating Eq. (45) in
Eq. (44),
cos ut 2
E20i = E20r + na b E0t (46)
cos u

Dividing the equation by the left member, it becomes

cos ut 2
1 = r2 + na bt (47)
cos u

where the reflection and transmission coefficients r and t have been intro-
duced. Now the quantity r2 is just the reflectance R:

Pr Ir E0r 2
R = = = a b = r2
Pi Ii E0i

Comparing Eq. (47) with Eq. (42), it follows that the transmittance T is expressed
by the relation
cos ut 2
T = na bt (48)
cos u
504 Chapter 23 Fresnel Equations

Notice that T is not simply t2 since it must take into account a different
speed and direction in a new medium. The change in speed modifies the rate
of energy propagation and thus the power of the beam; the change in direc-
tion modifies the cross section and thus the power density of the beam. How-
ever, for normal incidence, Eq. (48) reduces to T = nt2 and Eq. (47) becomes

normal incidence: 1 = r 2 + nt2 (49)

Note that throughout this section we have assumed that the reflection and
transmission coefficients are real, as they will be for all external reflections
and all internal reflections with angles of incidence less than the critical angle.

Example 3
Calculate the reflectance R and transmittance T for both TE and TM modes
of light incident at 30° on glass of index 1.60.
Solution
The reflection and transmission coefficients for this situation are given in
the solution to Example 1. Using these, the reflectance and transmittance
are found to be

RTE = r2TE = 1- 0.274022 = 0.075 and TTE = 1 - RTE = 0.925

RTM = r2TM = 1- 0.186622 = 0.035 and TTM = 1 - RTM = 0.965

5 EVANESCENT WAVES

In discussing the propagation of a light wave by total internal reflection

(TIR) through an optical fiber, we mentioned the phenomenon of cross talk,
the coupling of wave energy into another medium when it is brought close
enough to the reflecting wave. This loss of energy is described as frustrated
total internal reflection. The theory presented in this chapter allows us to
describe this phenomenon quantitatively.
The transmitted wave at a refraction can be represented as
B
#
Et = E0tei1kt r - vt2
B

where, according to the coordinates chosen in Figure 1,

kt # Br = kt 1 -sin ut xN - cos ut zN 2 # 1x xN + z zN 2
B

kt # Br = kt 1- x sin ut - z cos ut2

B

We can express cos ut as

sin2 u
cos ut K 21 - sin2 ut = 1 -
B n2

where we have used Snell’s law, n sin ut = sin u in writing the last equality. At
the critical angle, sin u = n and cos ut = cos190°2 = 0. For angles such that
sin u 7 n, when TIR occurs, cos ut becomes purely imaginary and we can write

sin2 u
cos ut = i - 1
B n2
 Fresnel Equations 505

Thus the exponential factor

sin2 u sin2 u
kt # Br = - ktx
B sin u sin u
- iktz - 1 = - ktx + ikt ƒ z ƒ - 1
n B n2 n B n2

In writing the last equality we have noted that, for the situation depicted in
Figure 1, the transmitted wave exists in the region for which z 6 0, and so
in this region z = - ƒ z ƒ . With the definition of the real, positive number,

sin2 u
a K kt - 1
B n2

the transmitted wave may be expressed as

B
#
Et = E0tei1kt r - vt2 = E0te -ivte -ixkt sin u>ne -aƒ z ƒ
B

The last factor on the right-hand side of this relation describes an exponential
decrease in the amplitude of the wave as it enters the medium of lesser re-
fractive index along the negative z-direction. When the wave penetrates into
the medium of lesser refractive index by an amount

1 l
ƒzƒ = = (50)
a sin2 u
2p - 1
B n2

the amplitude is decreased by a factor of 1/e. The energy of this evanescent

wave returns to its original medium unless a second medium is introduced
into its region of penetration. Although detrimental in the case of cross talk
in closely bound fibers lacking sufficient thickness of protective cladding, the
frustration of the total internal reflection is put to good use in devices such as
variable output couplers, made of two right-angle prisms whose separation
along their diagonal faces can be carefully adjusted to vary the amount of
evanescent wave coupled from one prism into the other. Another application
involves a prism face brought near to the surface of an optical waveguide so
that the evanescent wave emerging from the prism can be coupled into the
waveguide at a given angle (mode) of propagation.

Example 4
Calculate the penetration depth of an evanescent wave undergoing TIR at
a glass- 1n = 1.502 to-air interface, such that the amplitude is attenuated to
1/e of its original value. Assume light of wavelength 500 nm is incident on the
interface at an angle of 60°.
Solution
Since uc = sin-111>1.52 = 41.8°, TIR occurs at 60°. The penetration depth is
given by Eq. (50):

0.500 mm
ƒzƒ = = 0.096 mm
sin2 60
2p - 1
B 11>1.522
506 Chapter 23 Fresnel Equations

6 COMPLEX REFRACTIVE INDEX

We wish now to show that when the reflecting surface is metallic, the Fresnel
equations we have derived continue to be valid, with one important modifi-
cation: The index of refraction becomes a complex number, including an
imaginary part that is a measure of the absorption of the wave.
When the reflecting surface is that of a homogeneous dielectric—the
case we have been discussing in this chapter—the conductivity s of the mate-
rial is zero. The conductivity is the proportionality constant in Ohm’s law,

B B
j = sE

where j is the current density 1A>m22 produced by the field E. In such cases,
B B

B B
both the E- and B-fields satisfy a differential wave equation of the form

1 0 2E
§2E = a b (51)
c2 0t2

We have written harmonic waves satisfying Eq. (51) in the form

B
#
E = E0ei1k r - vt2
B

(52)

Now if the material is metallic or has an appreciable conductivity, the funda-

mental Maxwell equations of electricity and magnetism lead to a modifica-
tion Bof Eqs. (51) and (52). The differential wave equation to be satisfied by
the E-field is then

1 0 2E s 0E
§2E = a 2
b 2 + a 2b (53)
c 0t e0c 0t

Note that, compared with Eq. (51), the new wave equation, given as Eq. (53),
includes an additional term involving the conductivity and the first time deriva-
tive of E. As a result, when a harmonic wave in the form
B
of Eq. (52) is substitut-
ed into Eq. (53), we find that the propagation vector k must have the complex
magnitude

' v s 1>2
k = c1 + ia bd (54)
c e0v

Since the refractive index n is related to k by n = 1c>v2k, the refractive index

is now the complex number

1>2
' s
n = c1 + ia bd (55)
e0v

or we write, in general,
'
n = nR + inI (56)

' '
where Re 1n2 = nR and Im 1n2 = nI . Combing Eqs. (55) and (56) and equat-
ing their real and imaginary parts, the optical constants nR and nI can be
 Fresnel Equations 507

found in terms of the conductivity by the equations

n2R - n2I = 1
(57)
s
2nRnI =
e0v

Furthermore, if the complex character of k in the form

' v ' v
k = a b n = a b [nR + inI] (58)
c c

is introduced into the harmonic wave, Eq. (52), the result is

E = E0e -1vnIs>c2eiv1nRs>c - t2 (59)

where s is the directed distance along the propagation direction. We conclude

from Eq. (59) that the wave propagates in the material at a wave speed c>nR
and is absorbed such that the amplitude decreases at a rate governed by the
'
exponential factor e -1vnIs>c2. Thus, Re 1n2 = nR must behave as the ordi-
'
nary refractive index, and Im 1n2 = nI , called the extinction coefficient,
determines the rate of absorption of the wave in the conductive medium. This
absorption, due to the energy contributed to the production of conduction
current j in the material, is usually described by the decrease in power density I
with distance, given by

I = I0e -as (60)

By comparison with the power density as determined from Eq. (59), where
I r ƒ E ƒ 2,

I = I0e -2vnIs>c (61)

Thus, the absorption coefficient a is related to the extinction coefficient nI by

2vnI 4pnI
a = = (62)
c l

7 REFLECTION FROM METALS

'
Replacing n by n in the Fresnel equations, Eqs. (27) and (28), we have for metals,
'
Er cos u - 2n 2 - sin2 u
TE: = ' (63)
E cos u + 2n 2 - sin2 u

' '
Er - n 2 cos u + 2n 2 - sin2 u
TM: = ' ' (64)
E n 2 cos u + 2n 2 - sin2 u
'
Introducing n as nR + inI into Eqs. (63) and (64) gives

Er cos u - 21n2R - n2I - sin2 u2 + i12nRnI2

TE: = (65)
E cos u + 21n2R - n2I - sin2 u2 + i12nRnI2
 508 Chapter 23 Fresnel Equations

Er - [n2R - n2I + i12nRnI2] cos u +21n2R - n2I - sin2 u2 + i12nRnI2

TM: =
E [n2R - n2I + i12nRnI2] cos u + 21n2R - n2I - sin2 u2 + i12nRnI2
(66)

In calculating the reflectance R = ƒ Er>E ƒ 2, the complex quantity Er>E can

first be reduced to a ratio of complex numbers in the form 1a + ib2>1c + id2,
so that
R = 1a2 + b22>1c2 + d22
In the process, we must take the square root of a complex number, which is
done by first putting it into polar form. For example, if z = A + iB, then, in
polar form,
-1
1B>A2]
z = 1A2 + B221>2ei[tan
and the square root becomes
-1
1B>A2]
z1>2 = 1A2 + B221>4ei[11>22 tan (67)

The complex expression in Eq. (67) can then be returned to the general com-
plex form C + iD using Euler’s equation. These mathematical steps are easily
performed with a programmable calculator or a computer. In Figure 10, the
results of such calculations are shown for two metal surfaces, solid sodium and
single-crystal gallium. High reflectance in the visible spectrum is characteristic
of metallic surfaces, as shown by the curves for solid sodium at a wavelength of
589.3 nm. Strong discrimination between the TE and TM modes in the inci-
dent radiation is exhibited by the curves for single-crystal gallium surfaces.

TE
100
TM
90

80 TE

70

60
R (%)

TM
50
Single-crystal gallium:
40
nR # 3.7, nI # 5.4
30 Solid sodium:
nR # 0.04, nI # 2.4
20

10
Figure 10 Reflectance from metal sur-
faces by using Fresnel’s equations. The val-
10 20 30 40 50 60 70 80 90
ues of nR and nI are given for sodium light
of l = 589.3 nm. Angle of incidence u

PROBLEMS
1 Show that the vanishing of the reflection coefficient in 3 Determine the critical angle and polarizing angles for (a)
the TM mode, Eq. (28), occurs at Brewster’s angle, external and (b) internal reflections from dense flint glass
up = tan-11n2. of index n = 1.84.
2 The critical angle for a certain oil is found to be 33°33¿. 4 For what refractive index are the critical angle and (ex-
What are its Brewster’s angles for both external and inter- ternal) Brewster angle equal when the first medium is
nal reflections? air?
 Fresnel Equations 509

5 Show that the Fresnel equations, Eqs. (27) to (30), may also 15 Unpolarized light is reflected from a plane surface of fused
be expressed by silica glass of index 1.458.
a. Determine the critical and polarizing angles.
sin 1u - ut2 2 cos u sin ut b. Determine the reflectance and transmittance for the
TE: r = - t = TE mode at normal incidence and at 45°.
sin 1u + ut2 sin 1u + ut2
c. Repeat (b) for the TM mode.
tan 1u - ut2 2 cos u sin ut d. Calculate the phase difference between TM and TE
TM: r = - t = modes for internally reflected rays at angles of incidence
tan 1u + ut2 sin 1u + ut2 cos 1u - ut2
of 0°, 20°, 40°, 50°, 70°, and 90°.

6 Show that Eq. (37) follows from Eq. (35). 16 A Fresnel rhomb is constructed of transparent material of
index 1.65.
7 Using Eqs. (27) through (30) and a computer program
a. What should be the apex angle u, as in Figure 7?
or a computer algebra system, reproduce Figures 3 and 5.
b. What is the phase difference between the TE and TM
Also, change the value of n to produce graphs for the case
modes after both reflections, when the angle is 5% below
of external and internal reflection from diamond
and above the correct value?
1n = 2.422.
17 Determine the reflectance for metallic reflection of sodi-
8 Use a computer to calculate and plot the reflectance curves
um light (589.3 nm) from steel, for which nR = 2.485 and
of Figure 4. Also plot the corresponding transmittance.
nI = 1.381. Calculate reflectance for (a) TE and (b) TM
9 Use a computer to calculate and plot the phase shifts on re- modes at angles of incidence of 0°, 30°, 50°, 70°, and 90°.
flection as a function of angle of incidence for u 7 uc . Take
18 Determine the reflectance from tin at angles of incidence
n = 1>1.5 to reproduce Figure 6 and then make similar
of 0°, 30°, and 60°. Do this for the (a) TE and (b) TM modes
plots for n = 1>1.3 and n = 1>2.42.
of polarization. Real and imaginary parts of the complex
10 A film of magnesium fluoride is deposited onto a glass sub- refractive index are 1.5 and 5.3, respectively, for light of
strate with optical thickness equal to one-fourth the wave- 589.3 nm.
length of the light to be reflected from it. Refractive indices for
19 a. What is the absorption coefficient for tin, with an imagi-
the film and substrate are 1.38 and 1.52, respectively. Assume
nary part of the refractive index equal to 5.3 for 589.3-nm
that the film is nonabsorbing. For monochromatic light inci-
light?
dent normally on the film, determine (a) reflectance from the
b. At what depth is 99% of normally incident sodium light
air–film surface; (b) reflectance from the film–glass surface;
absorbed in tin?
(c) reflectance from an air–glass surface without the film;
(d) net reflectance from the combination. 20 a. From the power conservation requirement, as ex-
pressed by Eq. (47), show that for an external reflec-
11 Calculate the reflectance of water 1n = 1.332 for both
tion the transmission coefficient t must be less than 1,
(a) TE and (b) TM polarizations when the angles of inci-
but for an internal reflection t¿ may be greater than 1.
dence are 0°, 10°, 45°, and 90°.
b. Show further, using the Fresnel Eqs. (29) and (30), that
12 Light is incident upon an air–diamond interface. If the as the angle of incidence approaches the critical angle,
index of diamond is 2.42, calculate the Brewster and critical t¿ must approach a value of 2 in the TE mode and 2/n in
angles for both (a) external and (b) internal reflections. In the TM mode.
each case distinguish between polarization modes. c. Plot the transmission coefficient t¿ for an interface be-
13 Calculate the percent reflectance and transmittance for tween glass 1n = 1.52 and air.
both (a) TE and (b) TM modes of light incident at 50° on a 21 A narrow beam of light 1l = 546 nm2 is rotated through
glass surface of index 1.60. 90° by TIR from the hypotenuse face of a 45°–90°–45°
14 Derive Eqs. (29) and (30) for the transmission coefficients prism made of glass with n = 1.60.
both by (a) eliminating Er from Eqs. (20) to (23) and by a. What is the penetration depth at which the amplitude
(b) using the corresponding equations for the reflection of the evanescent wave is reduced to 1/e of its value at
coefficients, together with the relationships between reflec- the surface?
tion and transmission coefficients implied by Eqs. (20) b. What is the ratio of irradiance of the evanescent wave
and (22). at 1 mm beyond the surface to that at the surface?
 First-order diffraction
u1

Incident
light Zero order Modulated
beam beam

Piezoelectric
crystal
AM signal

24 Nonlinear Optics and the

Modulation of Light

INTRODUCTION

Much of the optics you are familiar with, including the processes of transmis-
sion, reflection, refraction, superposition, and birefringence, fall in the category
of what is called linear optics. When we speak of linear optics, we assume that
an optical disturbance propagating through an optical medium can be de-
scribed by a linear wave equation. As a consequence of this assumption, two
harmonic waves in the medium obey the principle of superposition, traveling
without distortion due to the medium itself or as a result of the mutual inter-
ference of the waves, regardless of the intensity of the light. Only the wave-
length and velocity of a light beam in a transparent material are required to
describe its behavior.
When the light irradiance becomes great enough, linear optics is not ade-
quate to describe the situation. With the advent of the more intense and coher-
ent light made available by the laser, we find that the optical properties of the
medium, such as its refractive index, become a function of the electric field of
the light. When two or more light waves interfere within the medium, the princi-
ple of superposition no longer holds. The light waves interact with one another
and with the medium. These nonlinear phenomena require an extension of the
linear theory that allows for a nonlinear response of optical materials to the
electromagnetic radiation.
In this chapter we define more precisely the area of nonlinear optics, de-
scribe and categorize some nonlinear phenomena, and discuss some of their
practical applications.

510
 Nonlinear Optics and the Modulation of Light 511

1 THE NONLINEAR MEDIUM

Nonlinear phenomena are due ultimately to the inability of theBdipoles in the

optical medium to respond in a linear fashion to the alternating E-field associ-
ated with a light beam. Atomic nuclei are too massive B
and inner-core elec-
trons too tightly bound to respond to the alternating E-field at the frequency
of light 1 ' 1014 –1015 Hz2. Thus the outer electrons of the atoms in a material
are primarily responsible for the polarization of the optical medium by the
beam’s E-field.1 When the oscillations of these electrons
B

B
in response to the
field are small, the Bpolarization is proportional to the E-field. However, as
the strength of the E-field increases, strict proportionality begins to fail, just
as the harmonic oscillations of a simple spring become increasingly anhar-
monic as the amplitude of the oscillations increases. Another means of excit-
ing nonlinear behavior without using high beam irradiances is to choose the
exciting optical frequency near a resonant frequency of the oscillating
dipoles, a technique widely utilized in nonlinear spectroscopy and known as
resonance enhancement.2 B
The polarization of a linear medium by an electric field E is usually writ-
ten in the form
B B
P = e0xE (1)
where x is the susceptibility and e0 is the vacuum permittivity. When depar-
tures from linearity are small, it is possible to represent the modification of
the susceptibility in a nonlinear medium by a power series in the form

x = x1 + x2E + x3E 2 + Á (2)

When substituted into Eq. (1), the polarization strength takes the form

P = e01x1E + x2E 2 + x3E 3 + Á 2 (3)

or
P = P1 + 1P2 + P3 + Á 2
()* (''')'''*
linear small nonlinear terms

where the subscripts on x match the powers of E and reflect the decreasing
magnitude of the higher-order terms. The linear and nonlinear susceptibility
coefficients characterize the optical properties of the medium, and this rela-
tion between P and E completely characterizes the response of the optical
medium to the field. We note that in linear media, where only the x1 contri-
bution is important, it is common to define the permittivity of the medium to
be e = e0x1 so that P1 = eE. The speed of light in a linear nonmagnetic ma-
terial can be written as y = 1> 1em0 and the index of refraction of such a ma-
terial is n = c>y = 1em0> 1e0m0 = 1e>e0.
The first term P1 in Eq. (3) represents linear optics in which the polariza-
B B
tion of the medium is simply proportional to the E-field. Unless the E-field
amplitude is very large, the x-coefficients of the higher-power E-terms are too
small to allow these terms to influence the polarization appreciably. Only with
the availability of intense, coherent light have these higher-order terms become
important. The high coherence of laser light allows the beam to be focused onto

1
We speak here of the electric polarization, rather than wave polarization.
2
P. N. Butcher, and D. Cotter, The Elements of Nonlinear Optics (New York: Cambridge Uni-
versity Press, 1990), Ch. 6.
512 Chapter 24 Nonlinear Optics and the Modulation of Light

small spots with wavelength dimensions, producing electric field strengths ex-
ceeding 1010 V>m—on the order of the strengths of the fields binding electrons
to nuclei in the optical medium. Peter Franken and his associates are credited
with the first nonlinear coherent optics experiment,3 conducted at the Univer-
sity of Michigan in 1961. The team focused the coherent 694.3-nm output from
a pulsed ruby laser onto a quartz crystal and detected second harmonic genera-
tion, the presence in the output of a weak ultraviolet coherent radiation com-
ponent at 347.15 nm, twice the frequency or half the wavelength of the exciting
light. This nonlinear phenomenon is discussed in the next section. Materials
used in electro-optic applications typically have second-order nonlinear sus-
ceptibility constants x2 in the range of 10-10 m>V to 10-13 m>V, and third-order
nonlinear susceptibilities in the range of 10-17 m2>V2 to 10-22 m2>V2. In the fol-
lowing example we estimate the field strengths and irradiances required to
make the nonlinear contributions to the polarization sizeable compared to the
linear contribution.

Example 1
a. Estimate the electric field amplitude and irradiance that would cause
the second-order nonlinear contribution to the polarization to be 1%
of the linear contribution in KDP 1KH2PO42. KDP has an index of re-
fraction of about 1.5 and a second-order nonlinear susceptibility x2 of
about 10-12 m>V.
b. Estimate the electric field amplitude and irradiance that would cause
the third-order nonlinear contribution to the polarization to be 1% of
the linear contribution in a certain mix of b-carotene in ethanol. This
mix has an index of refraction of about 1.3 and a third-order nonlinear
susceptibility x3 of about 10-20 m2>V2.

Solution
a. The linear susceptibility for KDP can be found from the relation given
earlier,
c e e0x1
n = = = = 1x1
y A e0 A e0

Therefore, x1 = n2 = 1.52 = 2.25. Then for the x2 term of the polar-

ization to be 1% of the linear term, the electric field amplitude E0 must
satisfy the relation
x2E 20 = 10.012x1E0
or
10.012x1 10.01212.252
E0 = = = 2.25 * 1010 V>m
x2 10-12 m>V

This electric field strength E0 corresponds to an irradiance of

1 1
I = e0cnE 20 = 18.85 * 10-12213 * 108211.52
2 2
12.25 * 101022 W>m2 L 1018 W>m2 = 1014 W>cm2

3
P. A. Franken, A. E. Hill, C. W. Peters, and G. Weireich. “Generation of Optical Harmonics,”
Phys. Rev. Letters, 7, 1961: 118.
 Nonlinear Optics and the Modulation of Light 513

This irradiance would result, for example, from a pulsed laser field of
power 100 MW focused to a spot of radius about 5 mm. While achiev-
able, this is a very large irradiance.
b. Following the routine used in part (a), the linear susceptibility of
b-carotene in ethanol is x1 = n2 = 1.32 = 1.69. Then for the x3 term
of the polarization to be 1% of the linear term,

x3E 30 = 10.012x1E0

or,

x1 1.69
E0 = 10.012 = 10.012 -20 V>m = 1.3 * 109 V>m
A x3 A 10
The corresponding irradiance is about
1 1
I = e0cnE 20 = 18.85 * 10-12213 * 108211.32
2 2
11.3 * 10922 W>m2 L 3 * 1015 W>m2 = 3 * 1011 W>cm2

2 SECOND HARMONIC GENERATION

AND FREQUENCY MIXING

In this section we will discuss the manner in which a nonlinear crystal can be
used to convert energy in an electromagnetic wave of a given frequency—or
from several electromagnetic waves of given frequencies—to energy in an
electromagnetic wave at a different frequency. Such processes provide a
means of producing intense and coherent electromagnetic radiation at fre-
quencies at which there are no efficient laser transitions.
Second Harmonic Generation
Second harmonic generation results from the contribution of the second-
order term in Eq. (3):
P2 = e0x2E 2 (4)

in which the second-order polarization term P2 of the optical medium is pro-

portional to the square of the electric field. Figure 1 shows the polarization as
a function of the electric field for the linear case and the deviation from lin-
earity due to this second-order term.
It can be shown that the second-order term makes no contribution to
P
polarization in an isotropic optical material or one having a center of symme-
try. A crystal having a center of symmetry is characterized by an inversion
center, such that if the radial coordinate r is changed to -r, the crystal’s "P0
atomic arrangement remains unchanged and so the crystal responds in the !E0
same way to a physical influence. In such a crystal, reversing the applied field "E0
E
should not—except for a change in sign—change any physical property, such
as its polarization. Thus we should have both !P0

P2 = e0x21 +E22 and -P2 = e0x21-E22

Figure 1 Linear and typical nonlinear
Because the E-field is squared, P2 = - P2 , which can only be true if P2 = 0. response of polarization P to an applied
The quartz crystal used by Franken, and many other crystals as well, do not electric field E. For equal positive and nega-
possess inversion symmetry. They can, therefore, manifest second harmonic tive fields, the response of the optical medi-
um is not symmetrical in the case of the
generation in addition to other second-order phenomena to be described
nonlinear (curved line) response. In this
presently. case, the negative field E0 produces a greater
The appearance of a second harmonic in the polarization is expected polarization than a positive field of the same
from the following mathematical argument. If the applied electric field, or magnitude.
514 Chapter 24 Nonlinear Optics and the Modulation of Light

one of its Fourier components, is of the form

E = E0 cos vt
substitution into Eq. (4) gives

P2 = e0x2E20 cos2 vt = e0x2E20 [1211+ cos 2vt2]

where we have substituted the double-angle identity for cos2 vt. Then

P2 = 12 e0x2E 20 + 12 e0x2E 20 cos 2vt (5)

Evidently, the second-order polarization P2 consists of a term of twice the fre-

quency of the applied optical field as well as a constant or DC component
that represents optical rectification. Optical rectification results in a time-
independent polarization in the medium that manifests itself as a DC voltage,
in a direction transverse to the field propagation, across the nonlinear crystal.
The component of P2 oscillating at 2v corresponds to dipole oscillations in
the medium at this same frequency. These dipole oscillations generate elec-
tromagnetic radiation of angular frequency 2v, which is present in the resul-
tant field, together with the stronger first-order field at the fundamental
frequency v. Second harmonic generation is commonly used, for example, to
produce light of wavelength 532 nm by passing the 1064-nm light produced
by a Nd:YAG laser through a nonlinear crystal.

Phase Matching
The same nonlinear interaction that allows conversion of energy in the funda-
mental electromagnetic wave of frequency v into energy in the second har-
monic electromagnetic wave of frequency 2v also allows for energy conversion
in the other direction. That is, the nonlinear interaction can convert energy
from the second harmonic wave into the fundamental wave. The direction of
energy flow depends critically on the phase of the fundamental field, which
drives the nonlinear polarization, relative to the phase of the second harmon-
ic field, which is absorbed and emitted by the dipole oscillations comprising
the nonlinear polarization. Because of dispersion, the light continuously gen-
erated at frequency 2v travels at a different speed in the optical material than
the light at frequency v. Thus, the two waves are periodically in and out of
step as they traverse the crystal. One can show4 that the irradiance in the sec-
ond harmonic field is proportional to the irradiance factor
¢kL
sinc2 a b
2
where L is the distance into the crystal and k is the wave propagation con-
stant, equal to nv>c. Here, ¢k = k2v - 2kv . When ¢k = 0 we say the fields
are phase matched in the crystal and the irradiance factor above is a maxi-
mum. Because dispersion is present in materials, ¢k is not typically zero and
the irradiance factor describes the consequent reduction in the irradiance of
the generated second harmonic field due to the phase difference that devel-
ops between the two fields as they propagate through the nonlinear crystal.
The coherence length LC is defined by the relation ¢kLC = p. Thus,
LC = p> ¢k
When L = LC , the sinc-squared intensity factor is reduced to about 0.4 of its
maximum value and so represents an estimate of the useful length of a crystal

4
Ammon Yariv, Optical Electronics, 3d ed. (New York: Holt, Rinehart and Winston, 1985), Ch. 8.
 Nonlinear Optics and the Modulation of Light 515

designed for efficient second harmonic generation using light with a phase
mismatch ¢k. The coherence length can also be expressed as
p p p
LC = = =
¢k 2v 4p
¢n ¢n
c l0
l0
LC = (6)
4 ¢n

where we have used

nv n2v12v2 2nv1v2 2v
k = and ¢k = - = ¢n
c c c c
In Eq. (6), l0 is the vacuum wavelength of the fundamental and ¢n the
amount the index of refraction of the fundamental differs from that of the
second harmonic.

Example 2
Consider a situation in which an incident fundamental field of wavelength
l0 = 0.8 mm is incident on a KDP crystal with refractive indices 1.4802 for
the second harmonic and 1.5019 for the fundamental. What is the maximum
crystal thickness useful in generating second harmonic light?
Solution
Substitution into Eq. (6) gives
0.8 mm
LC = = 9.2 mm
14211.5019 - 1.48022

This calculation shows that the maximum crystal thickness useful in gener-
ating second harmonic light is typically quite small, in this case around 10
times the wavelength of the fundamental. Crystals with thicknesses equal to
their coherence length are impractically small.

Birefringent Phase Matching

A technique commonly used to circumvent the small coherence length of
nonlinear crystals makes use of their birefringence. The refractive index (and
so the velocity) of the extraordinary (E) ray varies with direction through the
crystal. If a direction through the crystal is chosen such that n2v for the E-ray
equals nv for the ordinary (O) ray, the fundamental and second harmonic
waves remain in step and the crystal can be a centimeter or so thick. This
technique is called index matching or birefringent phase matching and is clar-
ified by Figure 2a, which shows how the ellipsoids representing the velocity
versus crystal direction for the E- and O-rays intersect along the direction of
matching.
Quasi-Phase Matching
An alternative to birefringent phase matching is so-called quasi-phase matching,
QPM. In this technique a phase mismatch ¢k between the fundamental and
second harmonic fields can be compensated for by using a nonlinear crystal with
a periodic structure such that the sign of x2 changes every coherence length LC
of the crystal. Such a periodic structure could be constructed by slicing a crystal
into many thin slabs, each having a width equal to LC and then placing the slabs
back together in a manner such that each slab is rotated 180° relative to its
neighbors. Since materials with nonzero second-order susceptibilities x2 lack in-
version symmetry, the resulting crystal will have a second-order susceptibility
that changes sign each coherence length of the crystal, as shown in Figure 2b.
 516 Chapter 24 Nonlinear Optics and the Modulation of Light

Optic axis
Direction of
matching

" ! " ! " ! " Sign of x2

Figure 2 Phase matching techniques. (a) Propagation
Velocity ellipsoids for orthogonally polar- direction
ized light beams in a birefringent medium. E-ray velocity
The O-ray ellipsoid is spherical and inter- Lc
surface
sects the E-ray ellipsoid along a direction
O-ray velocity
(shown relative to the optic axis) for which
surface
both rays have the same velocity. (b) QPM
crystal. (a) (b)

The coherence length of the crystal is the length over which the direction of en-
ergy flow is from the fundamental field to the second harmonic field. Thus, just
as the direction of energy flow in the crystal is about to switch so as to cause at-
tenuation of the second harmonic field, the sign of x2 changes. This change in
sign restores the proper phase relation between the dipoles and the second har-
monic field and so ensures continued amplification of the second harmonic
field. The small coherence length associated with the phase mismatch in many
crystals (see Example 2) makes the construction of a QPM crystal by the
method just described somewhat impractical. Instead, external fields can be
used to perform a periodic poling of a nonlinear ferroelectric material or a non-
linear polymer. This permanent periodic poling produces the alternating sign of
x2 needed for QPM. A common structure of this type is commonly referred to
as periodically poled lithium niobate, or PPLN.
Frequency Mixing
When two or more incident beams with different frequencies are allowed to
interfere within a nonlinear dielectric material, frequency mixing can occur.
For example, consider two interfering incident waves of frequencies v1 and
v2 represented in the form
E = E01 cos v1t + E02 cos v2t
or, in the equivalent exponential form,
E = 12 E011eiv1t + e -iv1t2 + 12 E021eiv2t + e -iv2t2

The second-order nonlinear polarization P2 = e0x2E 2 is proportional to

the square of this incident field and so evidently produces harmonic fields at
2v1 , 2v2 , v1 - v2 , and v1 + v2 .
A special case of frequency mixing is the process known as parametric am-
plification. Instead of 2v1 : v3 , as in second harmonic generation, it is possible
to have sum frequency generation such that v1 + v2 : v3 or difference fre-
quency generation such that v1 - v2 : v3 . In common parlance, the two wave
components input into a nonlinear crystal are called the pump and signal waves.
The generated difference wave is known as the idler wave. Suppose, then, that a
small signal wave at vs = v1 and a powerful pump wave at vp = v2 interact
within a nonlinear medium. A difference or idler frequency v3 = vi = vp - vs
can be produced. This idler frequency can in turn beat with the pump frequency
to enhance the signal frequency, vs = vp - vi . In this process, therefore, both
idler and signal waves can be amplified, drawing power from the pump wave.
When signal and idler frequencies correspond to resonant frequencies in the
nonlinear crystal acting as a tuned Fabry-Perot cavity, the parametric oscillator
is a tunable source of coherent radiation. Tuning the cavity is accomplished by
varying the refractive index of the cavity through control of temperature or an
applied DC field.
 Nonlinear Optics and the Modulation of Light 517

TABLE 1 LINEAR AND NONLINEAR PROCESSES

Linear first order: Nonlinear second order: Nonlinear third order:

P1 = e0x1E P2 = e0x2E2 P3 = e0x3E3

Classical optics: Materials lacking inversion Materials with inversion

symmetry: symmetry:
Superposition Second harmonic generation Third harmonic generation
Reflection Three-wave mixing Four-wave mixing
Refraction Optical rectification Kerr effect
Birefringence Parametric amplification Raman scattering
Absorption Pockels effect Brillouin scattering
Optical phase conjugation

Second harmonic generation is not the only nonlinear phenomenon

that results from the quadratic dependence of the polarization on the electric
field. Table 1 lists others, as well as several that depend on the third-order
contribution to the polarization of the medium, P3 = e0x3E 3. For example,
notice that for third-order nonlinear processes, third harmonic generation can
occur. We describe several of the nonlinear phenomena listed in Table 1 in
the remaining sections of this chapter.

3 ELECTRO-OPTIC EFFECTS

Nonlinear electro-optics effects result from the application of a DC (or low-

frequency) electric field to a medium. In this section we discuss two such ef-
fects, the Pockels effect and the Kerr effect, and show that these effects can be
used in light modulators. By light modulation, we mean the modification of
the amplitude (AM), frequency (FM), phase, polarization, or direction of a
light wave. One purpose of modulation is to render the wave capable of car-
rying information. Of course, light choppers and shutters can accomplish some
modulation mechanically. We are interested here in describing the modulation
that is accomplished by varying the refractive index of a material through the
use of an applied electric field. In subsequent sections of this chapter, we exam-
ine so-called magneto-optic and acousto-optic devices.
The basic equation describing nonlinear behavior was given as Eq. (3), a
relation between the polarization of the medium and the applied electric field.
In dealing with crystalline media, which represent most of the useful electro-
optic materials, it is customary to express nonlinearity of the refractive index n
by an equation5 analogous to Eq. (2) for the susceptibility:

1 1
= 2 + rE + RE 2 (7)
n2 n0

where r and R are the linear and quadratic electro-optic coefficients,6 respec-
tively, and we assume that there is no other effect present (like crystal strain)
that can modify n. The refractive index in the absence of an applied field is n0 .
In general, the refractive index depends on theB propagation direction and wave
polarization relative to the crystal axes. Since E is a vector field, the coefficients
r and R are tensors that reflect the crystal symmetry. Depending on the degree

5
Ivan P. Kaminov, An Introduction to Electrooptic Devices (New York: Academic Press, 1974),
Ch. 3.
6
It is important to distinguish the “order” of an electro-optic process as determined from
Eq. (3) for the polarization and as determined from Eq. (7) for 1>n2 or ¢n. For example, the Pockels
effect is a second-order effect (involving x2) by the first criterion and a first-order effect (involving r)
by the second. Both criteria are used.
 518 Chapter 24 Nonlinear Optics and the Modulation of Light

of symmetry, many tensor components may vanish or become equal to others,

reducing the total number of independent elements7 required to represent a
particular crystalline material.

The Pockels Effect

The Pockels effect results from the linear term in Eq. (7), where E is an applied
DC field. This effect can be considered a special case of two-wave mixing, where
one of the waves is the incident optical wave and the other a field of zero fre-
quency. The optical electric field can be small, since the DC field is itself large
enough to produce nonlinear behavior. In general, the DC field redistributes
electrons in such a way that birefringence is induced in an otherwise isotropic
material, or new optic axes appear in naturally birefringent crystals. Since the
Pockels effect is a second-order effect relative to the polarization [Eq. (3)], it is
not found in isotropic materials having inversion symmetry. All crystalline
materials exhibiting a Pockels effect8 are also piezoelectric; that is, they show
induced birefringence due to mechanical strain. Since the effect was discovered
in 1893, long before the discovery of the laser, it was well known even before in-
tense optical fields became available.
In one configuration of the Pockels cell, the natural optic axis of the crys-
tal is aligned parallel to the applied field. Fast and slow axes are induced in a
plane normal to the applied field, as shown in Figure 3. If the Pockels cell crys-
tal is rotated until theBFA and SA are at 45° to the x- and y-axes, a vertically
polarized light wave E0 incident on the crystal along the field direction has
equal amplitude components on FA and SA. These components experience
different refractive indices and different speeds through the crystal. The

E0

y Incident
field

45#
45#
FA

x
SA

Figure 3 Pockels cell schematic. The retar-

dation action due to an applied voltage V is
suggested by an on-axis separation of the
polarized components transmitted by the
z
fast and slow axes of the crystal.

7
The linear electro-optic tensor pij is defined by the relation ¢11>n22i = © jpijEj , with
i = 1, 2, 3, Á , 6 and j = x, y, z. For example, in crystals of triclinic symmetry, all the 18 possible ten-
sor elements are required; in zinc blende (GaAs), only one is required; in centrosymmetric crystals, all
elements are zero. See for example, Ammon Yariv, Optical Electronics, 3d ed. (New York: Holt, Rine-
hart and Winston, 1985), Ch. 9.
8
Single crystals can be divided into 32 symmetry classes; of these, 20 show the Pockels effect.
 Nonlinear Optics and the Modulation of Light 519

crystal therefore behaves as a phase retarder, and the component waves

emerge with a phase difference. One component advances in phase by ¢w
and the other lags in phase by ¢w while traversing the crystal of length L, so
that their relative phase on emerging is given by £ = 2 ¢w. Now
¢w = 12p>l02L¢n, where l0 is the vacuum wavelength and L¢n represents
the optical-path difference induced in each component by the applied field. We
find ¢n from Eq. (7), which, for small changes, can be approximated by
d11>n22 = rE. The E 2 term is considered negligible in the Pockels effects. Thus,
1 dn
da 2
b = - 2 3 ! rE
n n
r
ƒ ¢n ƒ ! n30E
2
Substituting into the phase equations, we find

2p 2p 3 2p 3
£ = 21¢w2 = 2 L¢n = rn0EL = rn V
l0 l0 l0 0

where V = EL is the voltage applied across the length L of the cell. Notice
that the phase difference £ is independent of the crystal length. For example,
if the Pockels cell is to behave as a half-wave plate, we need to make £ = p.
The half-wave voltage required is then
l0
VHW = (8)
2rn30

Example 3
Suppose the Pockels cell is made from a KD * P crystal of 1-cm thickness
and the optical wave has a wavelength of 633 nm. What half-wave voltage
is required?
Solution
From Table 2, we find r = 24.1 * 10-12 m>V and a refractive index of 1.51.
Then,
633 * 10-9
VHW = = 3800 V
2124.1 * 10-12211.5123

Thus an applied voltage of 3.8 kV transforms the crystal into a half-wave plate.

TABLE 2 LINEAR ELECTRO-OPTIC COEFFICIENTS FOR REPRESENTATIVE MATERIALS

Linear electro-optic
Material (wavelength if not 633 nm) coefficienta r (pm/V) Refractive index n0

KH2PO4 (KDP) 11 1.51

KD2PO4 1KD * P2 24.1 1.51
1NH42H2PO4 (ADP) l = 0.546 nm 8.56 1.48
LiNbO3 (lithium niobate) 30.9 2.29
LiTaO3 (lithium tantalate) 30.5 2.18
GaAs (gallium arsenide) l = 10.6 mm 1.51 3.3
ZnS (zinc sulfide) l = 0.6 mm 2.1 2.36
Quartz 1.4 1.54
a
Depending on crystalline symmetry, materials have more than one electro-optic coefficient. Only one
has been listed here for use in a Pockels cell. These and others may be found in standard references.9

9
For example, see Amnon Yariv, Optical Electronics, 3d ed. (New York: Holt, Rinehart and Win-
ston, 1985), Ch. 8.
 520 Chapter 24 Nonlinear Optics and the Modulation of Light

TA x
Unpolarized
light

Pockels Linear
crystal polarizer

Figure 4 The effect of applying a half- Analyzer

wave voltage to a Pockels cell when the
system includes a crossed polarizer and
analyzer. The vertically polarized beam in-
cident on the Pockels cell is transmitted as a VHW
horizontally polarized beam. For other val- TA
ues of applied voltage, the beam incident
on the analyzer is elliptically polarized and Linear
is only partially transmitted. polarizer

Recall that for a half-wave plate the linear polarization of the emergent
light is rotated by 90° relative to the linear polarization of the incident light.
If a linear polarizer with TA along the x-direction intercepts the emergent
light, as in Figure 4, the transmittance of the system is zero when V = 0 and
maximum when V = VHW . Variations in V therefore modify the polarization
state of the emergent light, rendering it elliptical, in general, with an x-
component that can be transmitted by the analyzer. In effect, the polarizer-
analyzer pair transforms phase modulation into amplitude modulation. Thus
we see that the transmittance of the system can be modulated by variations in
the applied voltage. Variations of a signal voltage superimposed on V are
transformed into variations in light intensity in such a device, known as a
Pockels electro-optic modulator.
The Pockels cell can be used also with the field oriented orthogonally to
the beam direction, an arrangement that simplifies placement of the elec-
trodes. In the geometry we have been describing, the electrodes are usually
endrings that allow the light beam to pass through and still provide a reason-
ably uniform field in the crystal.
The transmittance of the beam can be expressed by the relation10

p V
I = Imax sin2 a b (9)
2 VHW

and is plotted in Figure 5. To take advantage of the more linear region of the
transmittance, a quarter-wave plate is often inserted between the initial
polarizer and the Pockels crystal. This has the effect of producing 50% trans-
mittance when V = 0 so that the operating point is located at P in the figure,
rather than at the origin. Variations in modulating voltage, if not too large,
then occur with a system response that is linear.
Two other closely related applications of the Pockels cell are illustrated
in Figures 6 and 7. In Figure 6, the cell is used as a Q-switch that allows

10
J. Wilson, and J. F. B. Hawkes, Optoelectronics: An Introduction (London: Prentice-Hall Inter-
national, 1983).
 Nonlinear Optics and the Modulation of Light 521

100
Transmittance
(%)

P
50
Transmitted intensity

VQW VHW Applied voltage V

p/2 p Phase difference $

Figure 5 Transmittance curve for the

Pockels cell modulator. Without a quarter-
wave plate, the transmittance is zero when
Modulating the applied voltage is zero. Using a quarter-
voltage wave plate between polarizer and modula-
tor, the transmittance is 50% at operating
point P when the applied voltage is zero.
Under these conditions, the modulator re-
sponds more linearly to an input signal.

sudden dumping of the energy stored in a laser. Without an applied voltage, the
Pockels crystal transmits the horizontally polarized beam from a laser cavity
without changing its state of polarization. A Glan-laser prism, tuned to pass
vertically polarized light, rejects the horizontally polarized beam. With half-
wave voltage applied, the horizontally polarized beam from the laser cavity is
rotated by 90° and is accordingly passed by the Glan-laser prism, allowing the
beam to be backreflected from a high-reflectance mirror. The configuration
now permits rapid traversal of the beam back and forth through the laser cavity,
and stimulated emission occurs, producing an energetic pulse of laser radiation.
In Figure 7, the Pockels cell is used to initiate cavity dumping. When half-
wave voltage is applied, the polarization state of the laser radiation is rotated
by 90°, so that it can be extracted with the help of the polarizing prism, as
shown.
The Kerr Effect
When the optical medium is isotropic, as in the case of liquids and glasses,
the Pockels effect is absent and the polarization is modified by the third-
order [in the expansion of the polarization of the medium, see Eq. (3)]
electro-optic effect, better known as the Kerr effect. This effect, like all
third-order effects, occurs whether or not a material possesses inversion
symmetry. Actually, the Kerr effect was the first electro-optic effect to be
discovered (1875). Kerr cells usually contain nitrobenzene or carbon disul-
fide in the space between two electrodes across which a voltage is applied,
as indicated in Figure 8. The applied electric field induces birefringence
with an optic axis parallel to the applied field. Light traversing the cell thus
encounters two refractive indices, ne and no , for polarizations parallel and
perpendicular to the optic axis, and phase retardation results. In this case,
 522 Chapter 24 Nonlinear Optics and the Modulation of Light

Horizontally
polarized beam
rejected

V%0
Horizontally
polarized
beam
HR mirror

Glan-laser
prism
Laser rod
Pockels
crystal

(a)

VHW
Horizontally
polarized
beam
Figure 6 Light-controlling action of a
Pockels cell, used as a Q-switch. The config-
uration in (a) produces low transmission at
zero-cell voltage and in (b) high transmis-
Vertically
sion at half-wave voltage. In (b), the inci-
polarized
dent and reflected beams are separated for beam
clarity. Repeated reentries of the beam into
the laser cavity initiates stimulated emis-
sion that produces the laser pulse. (b)

Output Other pulse-

Pockels beam shaping components
cell

Figure 7 Light-controlling action of a

Pockels cell, used as a cavity dumper. When
half-wave voltage is applied to the Pockels
cell, it rotates the linear polarization of the Laser rod
laser beam so that it can be dumped by the Polarizing
polarizing prism. HR mirror prism HR mirror

plied
y E- ap

Output

Analyzer

Input L
d
x
Figure 8 Kerr cell. The applied voltage Polarizer
creates a field that is perpendicular to the
beam direction. As with the Pockels cell, a
modulating voltage produces phase modu-
lation that is converted to amplitude modu- Modulating
lation by the polarizer-analyzer pair. voltage
 Nonlinear Optics and the Modulation of Light 523

Eq. (7) becomes

1 1 R 3 2
= 2 + RE 2 or ƒ ¢n ƒ = nE (10)
n2
n0 2 0

Experimentally, the difference between ne and no is found to obey a relation

of the form
¢n = KE 2l (11)
where K is the Kerr constant.
Equating Eqs. (10) and (11), we find the relationship between K and R
to be
Rn30
K =
2l
As explained for the Pockels cell, the relative phase retardation for the ordi-
nary and extraordinary components is
2p
£ = L ¢n
l
Introducing the Kerr constant through Eq. (11),

2pKV2L
£ =
d2
where we have set V = Ed and d is the interelectrode distance. To function
as a half-wave plate, £ = p, and we find that the required voltage VHW is
given by
d
VHW = (12)
22KL

Example 4
Consider a nitrobenzene Kerr cell for which K = 2.4 * 10-12 m>V2 (see
Table 3) at room temperature and l = 589 nm. If d = 1 cm and L = 3 cm,
what half-wave voltage is needed?
Solution
Substituting into Eq. (12),

0.01
VHW = = 26.4 kV
212212.4 * 10-12210.032

Thus the Kerr cell behaves as a half-wave plate at a voltage of around

26 kV, considerably higher than for a typical Pockels cell.

TABLE 3 KERR CONSTANT FOR SELECTED MATERIALS

Material (l = 589 nm, room temperature) K (pm> V2)

Nitrogen (STP) 4 * 10-6

Glass (typical) 0.001
Carbon disulfide 1CS22 0.036
Water 1H2O2 0.052
Nitrotoluene 1C5H7NO22 1.4
Nitrobenzene 1C6H5NO22 2.4
524 Chapter 24 Nonlinear Optics and the Modulation of Light

Kerr cells can be used as modulators in the manner described for Pock-
els cells. Because of the higher voltages required, and because of the toxic
and explosive nature of nitrobenzene, Pockels cells are usually preferred.
Nevertheless, Kerr cells find application as high-speed shutters and as a sub-
stitute for mechanical light choppers. They are capable of response to fre-
quencies in the range of 1010 Hz and they can often be found operating as
Q-switches in pulsed lasers.

4 THE FARADAY EFFECT

In contrast to the electro-optic effects discussed to this point, the Faraday effect
is a first-order (i.e., ¢n r B) magneto-optic interaction.11 When a transpar-
ent material is placed in a magnetic field and linearly polarized light is passed
through it along the direction of the magnetic field, the emerging light is
found to remain linearly polarized, but with a net rotation b of the plane of
polarization that is proportional both to the thickness d of the sample and the
strength of the magnetic field B, according to the empirical relation,
b = VBd (13)
Here V is the Verdet constant for the material, usually expressed in minutes of
angle per Gauss-cm (G-cm). The Verdet constant is both temperature and
wavelength dependent.
An interesting aspect of the Faraday rotation is that the sense of rota-
tion relative to the magnetic field direction is, for a given material, indepen-
dent of the propagation direction of the light. Thus, repeated forward and
backward traversals of the material by a light beam has a cumulative effect
on the angle of rotation b. This behavior contrasts with that exhibited in the
closely related phenomenon of optical activity.
The optical rotation of the polarized light can be understood as circular
birefringence, the existence of different indices of refraction for left-circularly
and right-circularly polarized light components. Recall that linearly polarized
light is equivalent to a combination of right- and left-circularly polarized com-
ponents. Each component is affected differently by the applied magnetic field
and traverses the sample with a different speed, since the refractive index is
different for the two components. The end result consists of left- and right-
circular components that are out of phase and whose superposition, upon
emerging from the Faraday rotator, is linearly polarized light with its plane of
polarization rotated relative to its original orientation.
A classical derivation12 of the angle of rotation b predicts a relation of
the form
e l dn
b = a bBd (14)
2m c dl

with e and m the electronic charge and mass, c the speed of light, l the wave-
length, and dn>dl the rotatory dispersion. By comparison with Eq. (13), the
theory predicts a dependence of the empirical Verdet constant V given by
e dn
V = l (15)
2mc dl

11
The magnetic field analogs of the Kerr effect (where ¢n r E 2) are the Voigt effect (in gases)
and the Cotton-Mouton effect (in liquids), in which a constant magnetic field is applied normal to the
light-beam direction. Both are very small effects and will not be discussed further here.
12
Frank L. Pedrotti, and Peter Bandettini. “Faraday Rotation in the Undergraduate Advanced
Laboratory,” Am. J. Phys., 58, June 1990: 542.
 Nonlinear Optics and the Modulation of Light 525

TABLE 4 VERDET CONSTANT FOR SELECTED MATERIALS

Material V (min/G-cm) l = 589 nm

H2O 0.0131
Crown glass 0.0161
Flint glass 0.0317
CS2 0.0423
CCl4 0.0160
NaCl 0.0359
KCl 0.02858
Quartz 0.0166
ZnS 0.225

When the constants are evaluated and V is expressed in the standard units of
min/G-cm, Eq. (15) becomes
dn
V = 1.0083l
dl
We note that the Verdet constant is proportional to both the wavelength of
the light and the induced rotatory dispersion in the medium. Measured val-
ues for V at 589 nm are given in Table 4.
The Faraday effect can be used for light modulation, although it is diffi-
cult, practically speaking, to modulate a magnetic field at very high frequen-
cies. Figure 9 shows schematically the Faraday rotator, a crystal or liquid cell
whose axis of symmetry is aligned with a magnetic field. The figure shows a
field B established by current windings and indicates a rotation of the polariza-
tion in the same sense as the current producing the field. This particular geom-
etry defines a positive Verdet constant. Figure 10 illustrates the principal

b
B
I E
E

L
Plane of polarization rotates
inside material
Incident Exit
polarization polarization
(horizontal) (vertical)
B

B
Head-on b % 90# Head-on Figure 9 Faraday effect producing rota-
view view tion of the plane of polarization.

Analyzer
at 45#
Faraday 45#
rotator
Polarizer Transmitted
vertical light
Incident
light 45# rotation of
Retropulse polarization
Figure 10 Faraday rotator used between a
polarizer-analyzer pair to produce optical
isolation of the optical system providing the
Retropulse incident light.
 526 Chapter 24 Nonlinear Optics and the Modulation of Light

application of the Faraday rotator as an optical isolator. The isolator consists

of a Faraday rotator situated between a polarizer-analyzer pair. As shown, the
incident vertically polarized light is rotated 45° counterclockwise by the Fara-
day rotator and in this orientation is fully transmitted by the analyzer. Optical
elements (not shown) farther down the line are responsible for undesirable
back reflections (retropulses) of this radiation along the optical axis. In tra-
versing the Faraday rotator a second time, the polarization vector of the re-
flected light is rotated an additional 45° in the same rotational sense, so that it
emerges horizontally polarized and encounters the polarizer at an angle of
90° with the polarization direction of the original beam. In this state, it is re-
jected by the polarizer, preventing it from continuing back into the optical
system, where, in high-power laser systems, it can damage optical compo-
nents. Thus the optical isolator effectively isolates the optical system from
stray retropulses.

Example 5
Let us calculate the required length of SF58 flint glass, having a Verdet con-
stant of 0.112 min/G-cm for 543.5-nm light, if it is to produce the 45° rotation
of the polarization vector required in an optical isolator when the magnetic
field has a value of 9 kG.
Solution
Using Eq. (13), we have
b 45° * 60 min>°
d = = = 2.68 cm
VB 10.112 min>G-cm2 * 9000 G

5 THE ACOUSTO-OPTIC EFFECT

Photoelasticity is the change in refractive index of a crystal due to mechanical

stress. This phenomenon makes possible the AO or acousto-optic effect—the
interaction of optical and acoustic waves—in which a longitudinal acoustic wave
launched by means of a piezoelectric transducer produces a periodic mechanical
stress in the crystal. The acoustic wave (see Figure 11) consists of a series of com-
pressions and rarefactions (longitudinal vibrations) in atomic density and so a
periodic—although small—variation of the refractive index about its normal

Train of
compressions and
rarefractions

Incident ks Diffracted
High
Low light beam k Bragg k# light beam
High angle
Low u u
High
Low !D " 2u
High
Low Path
difference
High
Low ls
Undeflected
Figure 11 Variations in the refractive High
Low ls light beam
index of a medium due to the passage of a
harmonic acoustic wave (left) and the scat- k#
h ks
tering of an incident optical beam by the in- Index of RF
duced “planes” (right). The inset shows the Moving sound waves k
refraction signal
relationship of wave vectors required by at speed y
momentum conservation when the acoustic Piezoelectric
wave has the propagation direction indicated. transducer
 Nonlinear Optics and the Modulation of Light 527

value. Light incident on this structure is scattered to a greater extent from re-
gions of higher refractive index. The scattering is called Brillouin scattering and
is another third-order effect that does not require a medium possessing inver-
sion symmetry.
If the crystal is thin enough, variations in refractive index along its
length lead to corresponding variations in the speed of light so that the crys-
tal behaves as a transmission phase grating. Some of the incident light beam
is diffracted into various orders, according to the diffraction grating equa-
tion, ml = d sin um , where the grating constant d in this case should be
taken to be the acoustic wavelength lS , and um is the angle of diffraction in
mth order. This is the so-called Raman-Nath regime.13 If the crystal is thicker,
regions of higher refractive index represent planes normal to the direction of
the acoustic wave, as suggested in Figure 11. In this case, the light wave is dif-
fracted in a way similar to that of X-rays from crystalline planes in Bragg
scattering (the Bragg regime). The light wave traverses the crystal with a
speed that is around five orders of magnitude greater than the speed of the
acoustic wave. This means that the induced grating is essentially stationary
relative to the light wave.
A diffracted light beam appears in any direction in which portions of
the wavefront reflected (1) from different parts of a given plane and (2)
from successive planes obey the usual condition for constructive interfer-
ence; that is, the path difference must be an integral number of wavelengths.
The first requirement is satisfied by an angle of scatter that is equal to the
angle of incidence. This condition is shown satisfied in Figure 11. The second
requirement is the Bragg condition. The path difference between the incident
and diffracted waves is made up of the two segments indicated in the figure.
By geometry, each has a magnitude of lS sin u. This leads to the equation

£D
ml = 2lS sin u = 2lS sin (16)
2

a relation worked out in this context by Brillouin in 1921 and identical to

the Bragg equation for X-ray diffraction.14 In Eq. (16), the angles and opti-
cal wavelength are those measured within the medium. (However, see
problem 15.)
The Bragg condition for diffraction maxima can be found by an alterna-
tive argument that makes useB of the particle nature of waves. The incident
light beam with wave vectorBk can be considered a flux of photons, each of
energy Uv and momentum Uk, where v is the angular frequency and U is the
Planck constantB
divided by 2p. In the same way, the acoustic wave with
wave vector kS can be considered a flux of quantized particles called
B
phonons, each of energy UvS and momentum UkS . The acousto-optic interaction
then consists of the interaction or collision of these particles
B
in which
B
both en-
ergy and momentum B
are conserved. The two wave vectors k and k S , as well as
the wave vector k ¿ of the diffracted light beam, are also shown in Figure 11.
The acoustic waves are shown propagating downward in the figure. Conser-
vation
B
of momentum
B B
in a collision between a photon and a phonon requires
that k ¿ = k - kS , as indicated in the inset vector triangle. On the other hand,

13
Robert Guenther, Modern Optics (New York: John Wiley and Sons, 1990), Ch. 14.
14
An important difference between X-ray diffraction and light diffraction in the Bragg regime
is that light scattering occurs in a continuous manner from a thick, sinusoidal grating, rather than from
discrete planes. As a consequence, only the first-order diffraction, m = 1, occurs in the acousto-optic
effect.
 528 Chapter 24 Nonlinear Optics and the Modulation of Light

if the acoustic Bwave Bis reversed

B
in direction, the corresponding vector triangle
is satisfied by k ¿ = k + kS . In general,
B B B
k ¿ = k ; kS (17)
With the plus sign, we interpret this to mean that an incident photon com-
bines with a phonon to produce the diffracted photon; with the negative sign,
the incident photon is considered to yield an additional phonon to the
k
acoustic wave, as well as a photon to the diffracted beam. Thus Eq. (17) de-
k sin u scribes both the emission and the absorption of a phonon by the crystal lattice.
We now show that Eq. (17) is equivalent to the Bragg diffraction condi-
u
kS
tion. Since light frequencies are of the order of 1014 Hz, while acoustic fre-
u quencies are generally less than 1010 Hz, both v and v¿ are much greater than
vS , that is,
k sin u
k B B
v¿ ! v and accordingly k¿ ! k

The vector triangle (Figure 12) for the wave vectors then shows that,
Figure 12 Wave vector triangle in the ap- B B
B B
proximation ƒ k ¿ ƒ = ƒ k ƒ = k. The angle u is
ƒ k ƒ S = 2 ƒ k ƒ sin u, or, in terms of wavelength,
the Bragg angle. The geometrical relation-
ship of sides is equivalent to the Bragg dif- l = 2lS sin u (18)
fraction condition.
which is Bragg’s equation, with m = 1.
Conservation of energy of the interacting particles requires that
Uv¿ = Uv ; UvS , or

v¿ = v ; vS (19)

This result shows that the diffracted photon differs in frequency—however

little—from the incident photon by the amount vS , greater or less, depending
on the direction of the acoustic wave. This turns out to be another way of
arriving at the Doppler effect for light. When the incident light encounters an
approaching wave, the scattered frequency is greater, and when it encounters
a receding wave (as in Figure 11), the scattered frequency is less.

Example 6
To get an idea of the magnitude of the diffraction angle in first order, let us
consider a typical case in which the incident light has a wavelength of 550 nm
and the acoustic wave has a frequency nS of 200 MHz and a speed yS of 3000
m/s.

Solution
Then,
yS 3 * 103
lS = = = 1.5 * 10-5 m
nS 2 * 108
Using Eq. (16),

l 550 * 10-9
sin u = = = 0.0183
2lS 12211.5 * 10-52
so that

u = 1.05°
The acousto-optic (AO) effect can be applied to the modulation of a
light beam by controlling its amplitude (AM), its frequency (FM), or its di-
rection. Figure 13 illustrates one means of achieving AM modulation using
 Nonlinear Optics and the Modulation of Light 529

First-order diffraction
u1

Incident
light Zero order Modulated
beam beam

Figure 13 Modulation of a light beam by

an acousto-optic grating in the Raman-
Nath regime. The modulated signal driving
Piezoelectric the piezoelectric crystal is transferred to
crystal
the output beam in zero order. Only the
AM signal zero- and first-order diffracted beams are
shown.

90# prism
Output
Laser cavity beam

HR mirror
Prism
(1) Figure 14 Cavity dumping of a laser using
HR mirror AO beam an acousto-optic beam deflector. Turning on
(2) deflector the acoustic wave deflects the beam outside
the laser cavity and initiates cavity dumping.
HR mirror

a thin AO material in the Raman-Nath regime. The fraction of light re-

moved from the zero-order diffracted beam depends on the magnitude of
the induced stress and so on the amplitude of the modulating RF signal. The slit
allows only the modulated zero-order beam to be transmitted. Other applica-
tions make use of the beam deflection capabilities of the AO effect. Referring
back to Figure 11, it should be evident that both a change in frequency and
a change in direction of the acoustic wave cause a change in the direction of
the diffracted beam. If the frequency-sensitive position of the output beam is
detected by a photodetector array, the AO device can be used as a spectrum
analyzer. Again, because the frequency of the diffracted beam is shifted by an
amount equal to the acoustic frequency, the beam can be frequency modulat-
ed. In this case, the design aims to minimize the angular spread of the dif-
fracted light; when used as a spectrum analyzer, the design aims instead to
maximize it.
Another application of the device as a beam deflector to initiate laser-
cavity dumping is illustrated in Figure 14. When no acoustic wave is applied,
the beam (1) bounces back and forth in the laser cavity, building up energy
to a maximum value. Turning on the acoustic wave causes a deflection of
the beam (2) out of the cavity, thereby dumping the energy stored in the
cavity.

6 OPTICAL PHASE CONJUGATION

Optical phase conjugation (OPC) represents another third-order, nonlinear

phenomenon with some fascinating applications. It was first observed in
1971–1972 by researchers in the Soviet Union. It is so named because the
nonlinear interaction produces a light beam that is the spatial complex conju-
gate of one of the waves incident on the nonlinear medium. A process called
 530 Chapter 24 Nonlinear Optics and the Modulation of Light

degenerate four-wave mixing, illustrated in Figure 15, can be used to produce

A1
Signal
the phase conjugate beam. In this process, two strong, counterpropagating
Pump beam
A3 pump beams of complex amplitude A 1 and A 2 interact with a third, weaker
Nonlinear
medium
signal or probe beam of complex amplitude A 3 in a nonlinear medium
A4
A4 ! A1 A2 A3
with inversion symmetry. In such a medium, the lowest-order contribution to
PC output the nonlinear polarization of Eq. (3) will be proportional to the cube of the
A2
Pump beam total electric field in the medium. If the two pump beams and the signal beam
have the same frequency (that is, are degenerate), a fourth beam of this fre-
Figure 15 Conventional geometry for
quency will be produced. This fourth beam, pictured with complex amplitude
phase conjugation by four-wave mixing.
Pump beams A 1 and A 2 are antiparallel and A 4 in Figure 15, will be the spatial complex conjugate of the signal beam, A 3 .
are much stronger than the signal beam A 3 . A detailed analysis of this interaction15 shows that the phase conjugate
The phase conjugate output of the signal output beam has an amplitude that is proportional to the product of the am-
beam is the beam shown as A 4 .
plitudes of the pump beams and the complex conjugate of the amplitude of
the signal beam:
A 4 r A 1A 2A *3 (20)
Now, since the two pump beams are oppositely directed, the complex ampli-
tudes on the right-hand side of Eq. (20) can be written as
B
# B
# B
#
A 1 = ƒ A 1 ƒ eik1 r , A 2 = ƒ A 2 ƒ e -ik1 r , A 3 = ƒ A 3 ƒ eik3 r
B B B

so that
B
#
A 4 = ƒ A 1 ƒ ƒ A 2 ƒ ƒ A 3 ƒ e -ik3 r
B

B B
Thus, k4 = - k3 and A 4 is proportional to the complex conjugate of A 3 . Note
B
that the k-vectors representing the two pairs of counterpropagating waves
comprising the degenerate four-wave mixing process satisfy the phase match-
ing condition,
B B B B B B B B
k1 + k2 + k3 + k4 = k1 - k1 + k3 - k3 = 0

We have seen that the phase conjugate wave 1A 42 exactly reverses the
direction and overall phase factor of the signal wave 1A 32. Thus the phase
conjugate wave precisely retraces the path of the original beam and, at each
position, reproduces the exact shape of the original wavefront. Thus optical
phase conjugation can be viewed as a unique type of reflection, and we refer
to the nonlinear medium that creates the phase conjugate wave as a phase
conjugate mirror (PCM). To appreciate the uniqueness of the process, consider
Figure 16, which shows reflections from ordinary mirrors. In Figure 16a, a plane
wave is reflected from an ideal (infinite) plane mirror. The reflected wave is
also a plane wave. To express mathematically the reversal in direction of the
incident wave, the sign of the kz term is changed from minus to plus. Notice
then that, except for the sign of the vt term, the reflected wave is the complex
conjugate of the original wave and has the properties of a phase conjugate
wave as described above: It retraces the path of the incident beam and is its
phase-reversed replica. Figure 16b shows the same process for an incident
spherical wave. All that is needed to produce the phase-reversed replica on
reflection is a concave spherical mirror whose curvature exactly matches that
of the wavefront at incidence. For an arbitrary incident wavefront, as shown
in Figure 16c, the amplitude °1r2 is complex and includes the amplitude and
phase factors that describe its deviation from a plane wave. We imagine such
a wave as a plane wave that has been shaped by passing through a distorting
medium, by diffraction, or by modulation. Its phase-reversed replica is ex-
pressed by taking the complex conjugate of both °1r2 and ikz, in other

15
Amnon Yariv, Optical Electronics, 3d ed. (New York: Holt, Rinehart and Winston, 1985), Ch. 16.
 Nonlinear Optics and the Modulation of Light 531

E0 i(vt " kr)

E1(z, t) ! E0ei(vt " kz) E1(z, t) ! e
r

M
M

E0 i(vt # kr)
E2(z, t) ! E0ei(vt # kz) E2(z, t) ! e
r
(a) Plane wave (b) Spherical wave

E1(r, t) ! Re[$(r)ei(vt " kz)]

M
Figure 16 Three examples of a phase-
reversed replica of an incident wavefront pro-
duced by an ordinary mirror. In each case, the
rays corresponding to the wavefront are
everywhere normal to the mirror surface on
E2(r, t) ! Re[$*(r)ei(vt # kz)] reflection. A phase-conjugate mirror handles
all cases and also responds to instantaneous
(c) "Nearly plane" wave changes in the incident wavefront.

words, the spatial part of the wave equation. To produce this wave with
ordinary mirrors, we would need to construct a mirror surface that matched
exactly the wavefront of the incident wave at the instant of reflection. The
uniqueness of the PCM is that the phase conjugate replica is produced
regardless of the shape of the incident wavefront, as long as the PCM has an
aperture large enough to receive the entire wavefront. Unlike an ordinary
mirror, the PCM is able to respond immediately to varying spatial and tem-
poral features of an incident wave such that a phase conjugate wave is con-
tinually produced.
Applications of OPC include aberration correction and pointing and
tracking. These properties follow from the basic nature of the PC wave, as de-
scribed above. First, consider a transparent, distorting medium like frosted
glass, which is placed in the path of a light wave on its way to a PCM. The
wave is modified by the frosted glass in a nonuniform manner, but its phase
conjugate—after reflection by the PCM—exactly retraces its path and reverses
its modifications, so that the return passage through the distorting medium
undoes or “heals” the original distortion. This means that a beam distorted by
passing through an optical system with severe aberrations can be recovered
by “reflection” from a PCM that sends it back through the same system. It
also means that, if the light beam originates from a point source, divergence
and diffraction effects are reversed when the beam is returned through the
system, so that the PC beam converges to the original point. Furthermore, if
the source point moves, the returning beam adjusts so that it continues to
point to the source, the pointing and tracking property.

7 OPTICAL NONLINEARITIES IN FIBERS

We conclude this chapter with a survey of some aspects of the nonlinear in-
teraction of light waves propagating in optical fibers. We begin by dis-
cussing some nonlinear effects that inhibit the successful transmission, at
high rates, of information through fibers and then describe a nonlinear fiber
amplifier.
532 Chapter 24 Nonlinear Optics and the Modulation of Light

Stimulated Raman Scattering

When light of frequency v is scattered from a molecule, the scattered light
consists of a strong component at frequency v and components of lesser
strength at frequencies above or below v. The scattered light at v is elastic or
Rayleigh scattering. An inelastic process leading to scattered light at frequen-
cies different from the incident light is known as spontaneous Raman scatter-
ing. The inelastically scattered light of frequency less than v is known as the
Stokes field, and that component with frequency greater than v is known as
the anti-Stokes field. In the Raman process the energy lost or gained by the
electromagnetic field is accounted for by a change in energy of the molecule.
The photons scattered in the spontaneous Raman process can act as seed
photons, which in turn stimulate additional Raman scattering processes. The
predominant frequencies in the Stokes and anti-Stokes fields are characteris-
tic of the molecule and so their detection can be used to determine the com-
position of, for example, the gases in a particular combustion process.
Classically, Raman scattering can be described as arising from the third-order
nonlinear polarization of Eq. (3). The nonlinearity of silica molecules in opti-
cal fibers leads to the generation of, primarily, a Stokes field shifted from the
fundamental frequency by an amount in the range 1012 – 1013 Hz. The attenu-
ation, due to stimulated Raman scattering, of the power carried by an optical
fiber is typically not significant unless the power carried by the fiber
approaches 1 W. Most fiber-optical communication systems use signals of
powers significantly less than this and so do not suffer significantly from
Raman-induced attenuation. Unfortunately however, the Stokes frequency
shift is comparable to the frequency spacing between channels in a wavelength-
division-multiplexed (WDM) fiber. The Stokes field can couple the ideally
separate channels in the WDM fiber, leading to the degradation of the signal
carried by each channel. This effect begins to be important at signal powers of
only a few mW.

Stimulated Brillouin Scattering

In Section 5 we discussed ordinary Brillouin scattering in which light is scat-
tered by an acoustic wave. Stimulated Brillouin scattering is a nonlinear inter-
action in which light from an incident field is scattered by an acoustic wave,
which in turn was generated in the medium by the incident electromagnetic
field. The scattered light has a frequency that differs from that of the incident
light by an amount equal to the frequency of the acoustic waves in the medi-
um. In optical fibers this frequency shift is typically too small to couple the
different channels in a WDM fiber. However, stimulated Brillouin scattering
results in a wave that is scattered in the backward direction and so both
attenuates and mixes with the forward-traveling signal wave. In contrast to
the stimulated Raman process, stimulated Brillouin scattering can lead to sig-
nificant attenuation and distortion of signal power in a given fiber channel
when the signal power is just a few mW.

Self-Phase Modulation and Cross-Phase Modulation

In order to carry information in a fiber, the irradiance of the signal wave must
vary in time at a given point in the fiber. Due to fiber nonlinearities, this irra-
diance variation induces a time-varying index of refraction at this given point.
This time-varying index of refraction in turn gives rise to a time-dependent
phase shift of the signal. This process is known as self-phase modulation and
leads to what is known as frequency chirping. Generally, this phenomenon al-
ters the shape of the information-carrying pulses in a communications sys-
tem. This effect becomes increasingly important as the pulse width is reduced
and the propagation length is increased, and so may be important in long-
haul, high-bit-rate systems. Similarly, irradiance variations in one frequency
channel of a WDM fiber affect the index of refraction seen by the light signals
in the other frequency channels, leading to cross-phase modulation.
 Nonlinear Optics and the Modulation of Light 533

Raman Amplification in Fibers

The stimulated Raman effect can be used to amplify the signal in an optical
fiber, thus counteracting attenuation and increasing the length of fiber that
can carry a usable signal. This can be accomplished by propagating a pump
laser field through the fiber. The pump field will induce a Stokes field at a fre-
quency less than that of the pump field. In fact, the Stokes field is generated
in a range of frequencies known as the Raman bandwidth of the fiber. If the
signal field input into the fiber has a frequency that is near the peak of the
Raman bandwidth, it will strongly stimulate the Raman process and transfer
energy from the pump field into the signal field. Raman amplification in
fibers is attractive in part because it utilizes the entire unaltered fiber in the
amplification process. Competing amplification technologies require either
that the signal exit the fiber and be amplified before being inserted into a sec-
ond fiber or that portions of the fiber be doped with an amplifying medium
like erbium. The gain in the Raman process is sufficient to produce a Raman
laser in which the Raman gain compensates for the attenuation in a fiber
loop, and a self-sustained oscillation at the overlap between a fiber ring mode
frequency and the peak of the Raman gain bandwidth occurs.
Nonlinear effects in optical fibers can lead also to optical pulse com-
pression and to the propagation of pulses of unchanging shape, solitons,
through an optical fiber. Nonlinear optics is a burgeoning field driven by the
availability of high-power/short-pulse electromagnetic fields. In this chapter
we have discussed but a few of the many important effects and applications
associated with the nonlinear interaction of light and matter.

PROBLEMS
1 Write out the third-order terms of the polarization for a the light beam is from a He-Ne laser at 632.8 nm. The
single beam described by a plane wave with amplitude E0 length of the crystal is 1 cm.
and frequency v. What frequencies appear in the polariza-
8 Using Eq. (9), show that the transmittance of a Pockels cell
tion wave?
can also be written as I = Imax sin21£>22.
2 Write out the third-order terms of the polarization for two-
a. At what values of V and £ (greater than zero) is the
beam interaction, where the beams are plane waves having
transmittance zero?
amplitudes E01 and E02 and frequencies v1 and v2 , respec-
b. If the Pockels cell is preceded by an ordinary half-wave
tively. What frequencies are radiated by the polarization
plate, what is the irradiance when V = 0 and when
wave?
V = VHW?
3 Write out the second-order terms of the polarization for
three-beam interaction, where the beams are plane waves 9 In what kinds of media are both longitudinal Pockels and
having amplitudes E01 , E02 , and E03 and frequencies v1 , v2 , Kerr effects present? To get some idea of their relative
and v3 , respectively. What frequencies are radiated by the strengths, compare them by calculating the ratio of retarda-
polarization wave? tions produced by an appropriately applied 10 kV. Derive
an expression for this ratio. Then do a numerical calculation
4 Arguing from Eq. (7), show that the linear electro-optic ef- by assuming a hypothetical medium with “typical” values
fect is found only in crystals lacking inversion symmetry. of r = 10 pm>V, K = 1 pm>V2, L = 2 cm, d = 1 cm, and
5 a. Determine the coherence length for second harmonic n0 = 2. Take l = 550 nm.
generation in KDP when subjected to pulsed ruby laser 10 Calculate the length of a Kerr cell using carbon disulfide
light at l0 = 694 nm. Appropriate refractive indices required to produce half-wave retardation for an applied
are n1694 nm2 = 1.505 and n1347 nm2 = 1.534. voltage of 30 kV. The electrodes of the cell have a separa-
b. The measured coherence length of barium titanate at tion of 1.5 cm. Is this cell practical?
l0 = 1.06 mm is 5.8 mm. Calculate the expected change
in refractive index at l = 0.53 mm. 11 Show that Eq. (19) is equivalent to the Doppler effect for
light. Use the fact that the Doppler frequency shift ¢n for
6 Determine the half-wave voltage for a longitudinal Pockels
light reflected from a moving object is twice that of light
cell made of ADP (ammonium dihydrogen phosphate) at
emanating from a moving object, or ¢n = 2nup>y, where n
l = 546 nm. What is its length?
is the light frequency, y its velocity in the medium, and up
7 A longitudinal Pockels cell is made from lithium niobate. is the component of the object velocity parallel to the
Determine the change in refractive index and the phase light wave’s propagation direction. Use the geometry of
difference produced by an applied voltage of 426 V when Figure 11 and the Bragg condition.
 534 Chapter 24 Nonlinear Optics and the Modulation of Light

12 The speed of sound in glass is 3 km/s. For a sound wave hav- acoustic crystal is sapphire, with n = 1.76 and a sound
ing a width of 1 cm, calculate the advance of the sound speed of 11 km/s.
wave while it is traversed by a light wave. Take n = 1.50 for
17 Design an optical isolator, as in Figure 10, that uses ZnS as
the glass. What is the significance of this result?
the active medium. Let the magnetic field be produced by
13 a. Show that a small change in angle ¢u around the direc- winding a solenoid directly onto the ZnS crystal at a turn
tion of the diffracted beam in Figure 11 can be expressed density of 60 turns/cm. Assume l = 589 nm.
approximately by ¢u = ¢kS>k.
18 A sample of SF57 glass with polished, parallel sides and
b. Show that this result can be expressed as
2.73 cm in length is placed between the tapered poles of an
¢u = 1l>yS2 ¢nS electromagnet. A small, central hole is drilled through the
pole pieces to allow passage of a linearly polarized He-Ne
where l is the wavelength in the medium.
laser beam through the sample and parallel to the magnet-
c. The factor by which ¢u exceeds the beam divergence is
ic field direction. The magnetic field is set at 5.098 kG.
a practically useful number N called “number of resolv-
able spots.” This serves as a figure of merit, giving the a. When red He-Ne laser light (632.8 nm) is used, the mea-
number of resolvable positions that can be addressed sured rotation is 900 min. Determine the Verdet constant
by the beam deflector. If the beam divergence is ex- for the glass.
pressed by the diffraction angle uD = l>D, with D the b. When green He-Ne laser light (543.5 nm) is used, the
beam diameter, show that measured rotation is 1330 min. Determine the Verdet
¢u constant for the glass.
N = = t¢nS 19 A 5-cm-long liquid cell is situated in a magnetic field of 4 kG.
uD
The cell is filled with carbon disulfide and linearly polarized
where t is the time for the sound to cross the optical
sodium light is transmitted through the cell, along the B-field
beam diameter.
direction. Determine both the net rotation of the light and
d. As a numerical example, consider modulation of the
the circular dispersion of CS2 at this wavelength.
sound frequency in the range 80–120 MHz in fused
quartz, where yS = 5.95 * 105 cm>s. If the beam diam- 20 Sketch the shape of a nonsymmetrical pulse before and after
eter is 1 cm, determine the number of resolvable spots. reflection from an ordinary mirror and before and after re-
flection from a PCM. In the latter case, assume that the PCM
14 What acoustic frequency is required of a plane acoustic
is “turned on” by initiating the pump beams at the instant the
wave, launched in an acousto-optic crystal, so that a He-Ne
entire pulse has moved inside the PC medium. Show how
laser beam is deflected by 1°? The speed of sound in the
this effect might be used to correct dispersion broadening in
crystal is 2500 m/s and its refractive index at 632.8 nm is 1.6.
an optical fiber. (If necessary, consult Vladimir V. Shkunov,
15 In Bragg’s equation (18), the wavelength of the light and and Boris Ya. Zel’dovich, “Optical Phase Conjugation,” Sci-
the angle are those measured within the medium. Show entific American, Dec. 1985: 54.)
that, if the medium is isotropic and its sides are parallel to
21 Sketch an arrangement using a PCM to project a sharp,
the direction of a plane acoustic wave, the equation also
high-intensity image of a mask onto the photo-resist layer
holds for the wavelength and angle of diffraction measured
on a semiconducting chip without using lenses. This pro-
outside the medium.
vides a means of doing photolithography without placing a
16 Determine the difference in deflection angle for a He-Ne mask in direct contact with the chip. (If necessary, consult
laser beam that is Bragg-scattered by an acoustic plane Vladimir V. Shkunov, and Boris Ya. Zel’dovich, “Optical
wave when the frequencies are 50 MHz and 80 MHz. The Phase Conjugation,” Scientific American, Dec. 1985: 54.)
 nR 1.0

Extinction coefficient nI
1.0

Refractive index nR
0.1

0.1

0.01

nI

1015 1016 1017

Frequency (rad/s)

25 Optical Properties of Materials

INTRODUCTION

Electromagnetic waves that encounter materials create a complex of interac-

tions with the charged particles of the medium. Forces are exerted on the
charges by the electric field of the waves and, because of the motions of the
charges, also by the magnetic field of the waves. In responding to these oscil-
lating fields, the charges themselves oscillate and act as radiators of sec-
ondary electromagnetic waves. Thus, in determining the net field at some
point, the fields of both the source waves and the waves emitted by the
charged oscillators must be taken into account. In the case of ordinary fields,
smaller than those now attainable with high-energy lasers, the net fields are
assumed to be a linear superposition of the constituent fields. The complicat-
ed effects of all the microscopic contributions to the resultant field by the
charges in the material can, for certain purposes, be simply described by macro-
scopic material parameters, the optical constants of the material. In this chapter,
we show in particular how the refractive index and the absorption coefficient for
isotropic conducting (metals) and nonconducting (insulators or dielectrics)
materials can be understood. In order to do this we use Maxwell’s equations
and the mathematical techniques of vector calculus.

1 POLARIZATION OF A DIELECTRIC
MEDIUM

We take as our model a simple dielectric, that is, a nonconducting material

whose properties are isotropic. By nonconducting we mean that the medium,
unlike a metal, contains no free charges. Positive charges are associated with

535
 536 Chapter 25 Optical Properties of Materials

the constituent nuclei and negative charges with the electrons bound to such
nuclei. By isotropic we mean that the relevant physical properties we consid-
er are independent of direction in the medium, so that we may treat the phys-
ical constants as scalar quantities. Application of an electric field to such a
medium causes charge displacement, in which the negative charge distribu-
tion bound to the nuclei shifts in a direction opposite to the electric field. The
shift may occur in a polar molecule, like H2O, because the molecule has a
permanent electric dipole, that is, the effective centers of its positive and neg-
ative charge distributions do not coincide. In this case, application of the field
produces some reorientation of the molecules so that, on the average, the
positive end of the dipole is in the direction of the field. The tendency toward
alignment is counteracted by the thermal motions of the molecules. The shift
in charge distribution may also occur in nonpolar molecules, such as O2 , in
which positive and negative charge distributions normally have the same ef-
fective center. Application of the field results in a slight shift of the electron
cloud relative to its nucleus, producing an induced dipole. In either case, the
E dipole moment p B
due to each atom or molecule is given by the product of the
magnitude of the displaced charge q and the vector Br that locates the effec-
!q r "q
tive negative charge center relative to the effective positive charge center in
p the dipole, or
(a)
p
B
= - qrB (1)
"mgv
"KS r as indicated in Figure 1a. The direction of the dipole moment is from the
negative toward the positive charge. The magnitude of the dipole moment for
e
v a given material depends on how easily charge B
is displaced under the influ-
r ence of a given electric field. The polarization P of the medium is then said to
F # "eE
be the collective dipole moment per unit volume, the sum of dipole moments
E given by
(b)
B
P = - NerB (2)
Figure 1 The elementary electric dipole.
(a) Alignment with the field. (b) Forces act-
ing on a dipole when the electric field has
where N is the number of elementary dipoles per unit volume and e is the
the direction indicated. magnitude of the electronic charge.
Electrons behave as though the forces binding them to the nuclei are
elastic forces given by Hooke’s law, where the restoring force is proportional
to the displacement and oppositely directed. The more massive nuclei can be
considered stationary since they are unable to respond to the rapid changes
in the field representing an electromagnetic wave in the optical region of the
spectrum. A simple model in which electrons are held by springlike forces to
a fixed nucleus is therefore applicable. In an alternating electric field, howev-
er, forced oscillations of electrons remove a certain amount of energy from
the incident radiation, the energy that the electrons radiate in turn and the
energy of interaction with neighboring atoms that shows up as thermal ener-
gy. The model of the oscillating electron is therefore that of a damped, har-
monic oscillator, with a frictional force proportional to the velocity. Newton’s
second law, applied to the electron in the model of Figure 1b, then leads to
the equation of motion,

drB B d2Br
- KSBr - mg - eE = m 2 (3)
dt dt

In Eq. (3), KS is the force constant of the effective spring, m is the electronic
mass, and g is a frictional
B
constant with dimensions of reciprocal time. Notice
that the force 1evB * B2 on the electron due to the magnetic field of
B
the radiation is omitted; it is, in fact, negligible compared with the force 1e E2
due to the electric field.
 Optical Properties of Materials 537
B
When the applied E-field is static, there is no oscillation of the dipoles,
so that both velocity and acceleration of the electron vanishes. In this special
case, Eq. (3) reduces to
B
- KSBr = eE

or, eliminating Br with the help of Eq. (2), the static polarization is given by

Ne2E
B
B
P = (4)
KS
B
Suppose now that E is a harmonic field with a time dependence given
by E = E0e -ivt and that the oscillations respond with a similar dependence,
B B

r = Br 0e -ivt. Note that we are representing the electric field E and the dipole
B B

displacement vector r by complex vectors. The real field and displacement

B

vectors are formed by taking the real part of these complex vectors. All of the
equations used in this chapter are linear in these quantities and so hold for
both the real fields and displacements and their complex representations. In-
serting the derivatives of the complex displacement vector drB>dt = - ivrB and
d2Br >dt2 = - v2Br into Eq. (3) leads to

B
-eE
B
r = 2
(5)
-mv - imvg + KS

which, whenB substituted into Eq. (2), yields a time-dependent complex

polarization P given by

B Ne2 B
P = a 2
bE (6)
-mv - imvg + KS

Note that Eq. (6) agrees with Eq. (4) in the case of a static 1v = 02
B
E-field. In general, the polarization
B
is a function of the radiation frequency v
and, because the coefficient of E in Eq. (6) is complex,
B
the polarization may
possess a frequency-dependent
B
phase relative to E . If we were to associate
the symbol E with the externally applied field, our analysis, thus far, would B
apply only to the case of a single dipole oscillator. The electric field E in
Eq. (6) should represent Bthe actual field at the dipole, in the interior B
of the
medium. This local field Eloc is a superposition of the applied field Eapp and
the field that results from all the other dipoles aligned in a polarized medium.
Therefore, in order to treat a macroscopic assembly of dipoles we must make
use of a result given in standard texts on electricity and magnetism. That is, the
contribution to the total electric field at the position
B
of a given dipole, due to
all the other dipoles in the medium, is given by P>3e0 , where e0 is the permit-
tivity of free space. Thus,

B
B P B
Eloc = + Eapp (7)
3e0
B
Choosing to retain the symbol E for the applied field leads to

B
B P B
Eloc = + E
3e0
538 Chapter 25 Optical Properties of Materials

Substituting this expression for the local electric field into the right-hand side
of Eq. (6) gives

Ne2
B
B B P
P = a b aE + b (8)
- mv2 - imvg + KS 3e0

B
We can solve for P explicitly by, for the moment, setting the prefactor in
Eq. (8) equal to F. We conclude

B F B
P = a bE (9)
1 - F>3e0
B
The multiplier of E is

F Ne2>m
=
1 - F>3e0 1KS>m - Ne2>3me02 - v2 - ivg

Let us define v20 as the quantity in parentheses, that is,

KS Ne2
v20 K - (10)
m 3me0

Then, Eq. (9) becomes

B Ne2>m B
P = E (11)
v20 - v2 - ivg

Forming the magnitudes of both sides of this relation shows that the magnitude
of the polarization is related to the magnitude of the applied field by the relation

B Ne2>m B
ƒPƒ = ƒEƒ
21v20 2 2
- v2 + vg 2 2

Clearly, ƒ P ƒ can increase dramatically as v : v0 , so that v0 represents a

B

resonance frequency for the dipoles of the medium. Equation (11) has the
same form as the equation of motion of a driven harmonic oscillator with
damping. As the driving frequency approaches the resonance frequency v0 of
the oscillator, the amplitude of the vibrations becomes very large and sub-
sides again as the frequency increases beyond v0 . For a dielectric medium, the
increase of dipole moments at resonance results in a large maximum polariza-
tion. Equation (11) also indicates that there is a frequency-dependent
phase shift between the applied field and the polarization. Far from reso-
nance, vg V v20 - v2, and so the damping term in the denominator of the
expressionB
on the
B
right-hand side of Eq. (11) can be ignored. Then for
v % v0 , P and E have the same sign and the dipoles are oscillating in phase
B B
with the field. Beyond resonance, however, when v $ v0 , P and E have op-
posite signs, indicating a phase difference of p. Free electrons respond in
this manner. When v ! v0 , near resonance, the vibrations are large. The
damping term in the denominator in this case is not negligible, and the divi-
sion byB- i, equivalent
B
to multiplication by i, indicates a p>2 phase shift be-
tween E and P.
 Optical Properties of Materials 539
B B
The dependence of P on E, as given in Eq. (11), can now be used to
discover the conditions under which plane waves are able to propagate in a
dielectric. The fundamental wave equation for electromagnetic waves in the
dielectric is a consequence of the Maxwell equations.

2 PROPAGATION OF LIGHT WAVES

IN A DIELECTRIC

The four Maxwell equations may be written in the general form

§#E =
B r
(12)
e0

B
B 0B
§ * E = - (13)
0t

§#B = 0
B
(14)

B B
0E J
c2 § * B =
B
+ (15)
0t e0

In these equations r is the charge density, which in general includes both the
free charge density rf and the bound charge density rb , so that r = rb + rf .
In a dielectric, however, rf = 0. It is standard practice in a course in electric-
ity and magnetism to show that the bound-charge density is related to the po-
larization by
rb = - § # P
B
(16)
B
The quantity J similarly represents the current
B B
density
B
and can arise from
both free and bound charge, as indicated by J = Jb + Jf . In a dielectric where
B
rf = 0, Jf = 0 also. Furthermore, it can be shown that
B
B 0P
Jb = (17)
0t

With these constraints, the four Maxwell equations for a dielectric can be
written

-§ # P
B

§#E =
B
(18)
e0

B
B 0B
§ * E = - (19)
0t

§#B = 0
B
(20)

B B
2 0E
B 1 0P
c § * B = + (21)
0t e0 0t
540 Chapter 25 Optical Properties of Materials

Now we take the curl of both sides of Eq. (19), giving

B
B 0B 0 B
§ * 1§ * E2 = § * a - b = - 1§ * B2 (22)
0t 0t

where we have interchanged the order of differentiation with respect to

space and time in the last step. The left member of Eq. (22) can be reex-
pressed by the identity

§ * 1§ * E2 K §1§ # E2 - §2E
B B B
(23)

In a homogeneous dielectric, the effect of polarization is to produce a net sur-

face charge density, while leaving the internal charge density rb = 0 unchanged.
The internal charge density is zero because, in any internal closed surface,
every bit of charge that moves into the enclosed volume in response to a po-
larizing field is balanced by an equal bit of charge that moves out. The surface
charge density appears because such balancing is not possible there. Thus by
Eqs. (16) and (18) we conclude that § # E = 0 and substitute the remainder of
B

Eq. (23) into Eq. (22), giving

0
§2E =
B B
1§ * B2 (24)
0t

For the right member we may make use of Maxwell’s equation (21) and
write
0 2E 1 0 2P
B B
2 2B
c §E = 2 + (25)
0t e0 0t2

B
The last term is expressible in terms of E using Eq. (11), so we have

Ne2 0 2E
B

c2 §2E = c 1 +
B
d (26)
me01v20 - v2 - ivg2 0t2

For a harmonic wave expressed as E = E0ei1kz - vt2, in which case §2E = - k2E
B B B B

and 0 2E>0t2 = - v2E, Eq. (26) solved for k2 becomes

B B

v2 Ne2 1
k2 = 2
c 1 + 2 2
d (27)
c me0 1v0 - v - ivg2

We conclude that the analysis of plane waves propagating in a homogeneous

dielectric requires in general that the propagation constant k be a complex
number. Consequently, we write

k = kR + ikI (28)

Inserting this form into the expression for a harmonic wave, we have

E = E0ei1kRz + ikIz - vt2 = E0e -kIzei1kRz - vt2

B B B
(29)

The exponential factor in kI represents a depth-dependent absorption of an

otherwise harmonic wave, and kI measures the amplitude attenuation of the
wave. By taking the square of the magnitude of both sides of Eq. (29), the
 Optical Properties of Materials 541

result describes instead the energy flux density, giving

I = I0e -az

where a = 2kI is the absorption coefficient of the medium. If the propagation

constant is complex, so must be the refractive index, since

2p 2pn v
k = = = a bn (30)
l y c

If we identify the real and imaginary parts of the complex refractive index by

n = nR + inI (31)

where nR is the usual refractive index and nI is called the extinction coeffi-
cient, it follows from Eqs. (28) and (30) that

v
kR + ikI = a b1nR + inI2
c

yielding the relations

v
kR = a b nR (32)
c

and

v
kI = a b nI (33)
c

Writing n2 as

ck 2
n2 = 1nR + inI22 = a b
v

and relating this equation to Eq. (27) gives

Ne2 1
n2 = 1nR + inI22 = 1 + a b 2 (34)
me0 v0 - v2 - ivg

Expressions for the real and imaginary parts of the refractive index can be
found by equating real and imaginary parts in Eq. (34). The left member is

1nR + inI22 = 1n2R - n2I2 + i12nRnI2 (35)

The right member can also be written as the sum of a real and imaginary part.
The complex term is first rewritten by multiplying numerator and denominator
by the complex conjugate of the denominator. The result, after simplification, is

Ne2 v20 - v2 vg
1nR + inI22 = 1 + a 2 2 2 2 2
+ i 2 b
me0 1v0 - v 2 + v g 1v0 - v222 + v2g2
(36)
 542 Chapter 25 Optical Properties of Materials

Now by comparing the right members of Eqs. (35) and (36),

Ne2 v20 - v2
n2R - n2I = 1 + c 2 d (37)
me0 1v0 - v222 + g2v2

and

Ne2 gv
2nInR = c 2 d (38)
me0 1v0 - v222 + g2v2

The equations can be solved simultaneously for nI and nR . The appear-

ance of the mass m in the denominator of these equations shows that elec-
tronic oscillations are more important than ionic oscillations in determining
the index of refraction. Ionic polarization may be significant in the region of res-
onance, however, where the large-bracketed terms in Eqs. (37) and (38) balance
the small prefactors containing the mass.
Figure 2 shows both nR and nI calculated from Eqs. (37) and (38) as a
function of driving frequency v. The absorption described by the extinction
coefficient nI is seen to peak at the resonant frequency v0 . The real refractive
index experiences a sharp rise and then a fall as v increases toward and then
passes through resonance, after which it increases again, approaching the
value nR = 1 at high frequencies. The narrow region where nR decreases with
frequency is contrary to the usual dispersion of transparent media and is
called the region of anomalous dispersion.
A resonance frequency such as v0 for the dielectric means that, for inci-
dent photons of frequency v0 , there is a high probability of absorption. Ab-
sorption of such a photon corresponds to a transition between states differing
in energy by Uv0 = hn0 in the energy-band structure of the material. As v is
varied, there will be a series of resonance frequencies characteristic of the
material. If such a resonance occurs in the visible range of frequencies, for ex-
ample, the material absorbs a portion of the spectrum and appears colored,
while transmitting the remainder. Transparent materials like glass have reso-
nance frequencies in the infrared and ultraviolet regions but not in the visi-
ble. In terms of our simplified model of a dielectric, we interpret the existence

nR 1.0
Extinction coefficient nI

1.0
Refractive index nR

0.1

0.1

0.01

nI

Figure 2 Angular frequency dependence of

the refractive index nR and the extinction co- 1015 1016 1017
efficient nI for a dielectric. Assumed values v0
are v0 = 1 * 1016 s-1, g = 1014 s-1, and
N = 1 * 1028 m-3. Frequency (rad/s)
 Optical Properties of Materials 543

of a number of resonance frequencies to mean that electrons experience dif-

ferent degrees of freedom in response to the applied field. To take this into
account formally, Eq. (34) is usually generalized to include a number of terms
summed over the resonant frequencies vj , given by

a
Ne2 fj
n2 = 1 + (39)
me0 j vj - v2 - igjv
2

where fj , called the oscillator strength for the resonance vj , represents the
fraction of dipoles having this resonant frequency. The determination of the
oscillator strength for a given resonant transition requires the application of
quantum theory.

The Dispersion Equation

The variation in refractive index with frequency, described by Eq. (39), is
what we mean by dispersion. We wish to show that the Cauchy dispersion
equation can be deduced from Eq. (39) under certain simplifying assump-
tions. We shall assume a single resonant frequency v0 in the ultraviolet, such
that frequencies in the visible obey the inequality, v % v0 We shall also as-
sume that, in this regime gv V 1v20 - v22. In this case, the index of refrac-
tion is real and Eq. (34) takes the form

Ne2 1
n2 = 1 + a 2 b
me0 v0 - v2

Notice that, for v % v0 , as for a gas, the refractive index is nearly constant.
As v increases toward v0 , the refractive index increases slightly, as shown in
Figure 2. The slowly increasing index with frequency (decreasing with
wavelength) is characteristic of normal dispersion.
To derive the Cauchy dispersion equation, let us first expand the fre-
quency factor in a binomial series:

1 1 v2 -1 1 v2 v4
= 2 a 1 - 2 b = 2 a1 + 2 + 4 + Á b
v20 - v2
v0 v0 v0 v0 v0

so that

Ne2 v2 v4
n2 = 1 + a 1 + + + Áb
me0v20 v20 v40

Writing v = 2pc>l and gathering constants A¿, B¿, and C¿ appropriately, we

can express

B¿ C¿
n2 = A¿ + 2
+ 4 + Á
l l

We may take the square root of each side and, since each higher-order term
in the expression is less than A¿, use the binomial expansion again on the
right member. After re-collecting constants, we get
B C
n = A + 2
+ 4 + Á
l l
This is the Cauchy relation introduced earlier to describe normal dispersion.
544 Chapter 25 Optical Properties of Materials

3 CONDUCTION CURRENT IN A METAL

In metals, the existence of “free” electrons, not bound to particular nuclei,

modifies the treatment outlined above for dielectrics. Although there are also
bound electrons, the response of the free electrons dominates the electrical
and optical properties of the medium. So, in Eq. (3), we set KS = 0, and the
equation of motion becomes

dvB B
m + mgvB = - eE (40)
dt

The equation B
may be conveniently expressed in terms of the conduction cur-
rent density J, defined by
B
J = - NevB (41)
B
where
B
J has (SI) units of amperes per square meter. Writing Eq. (40) in terms
of J rather than vB,

Ne2 B
B
dJ B
+ gJ = a bE (42)
dt m

In the case where the applied field is the harmonic wave EB = EB0e -ivt, we ex-
B B

pect the current density to vary at the same rate and write J = J0e -ivt. Equa-
tion (42) then takes the form

B Ne2 B
1- iv + g2J = a bE (43)
m

In the static, or DC, case specified by v = 0,

B Ne2 B
J = a bE (44)
mg

The static conductivity s, defined by Ohm’s law,

B B
J = sE (45)

then takes the theoretical form

Ne2
s = (46)
mg

Since conductivities are usually measured, we rewrite Eq. (43) in terms of s,

giving

B s B
J = a bE (47)
1 - iv>g

4 PROPAGATION OF LIGHT WAVES

IN A METAL

An electromagnetic wave propagating in the conducting medium satisfies

Maxwell’s equations (12) through (15). Although free charge exists in
 Optical Properties of Materials 545

the metal, the internal free-charge volume density rf is zero. The free charge
is so mobile that it quickly redistributes in response to an applied field, pre-
venting the buildup of local charge densities. The appropriate Maxwell equa-
tions are then

§#E = 0
B
(48)

B
B 0B
§ * E = - (49)
0t

§#B = 0
B
(50)

B B
2 0EB J
c § * B = + (51)
0t e0

As before, § * 1§ * E2 = - §2E because § # E = 0 in the identity of

B B B

Eq. (23). So, taking the curl of Eq. (49), we have

1 0 2E
B B B
2B 0B 0 B 1 0J
- § E = § * a- b = - 1§ * B2 = - 2 2 - a b
0t 0t c 0t e0c2 0t

B
where we have used Eq. (51) in the last step. Representing J with the help
of Eq. (47), we conclude

1 0 2E
B B
2B 1 s 0E
§E = 2a 2b + a
2 1 - iv>g
b (52)
c 0t e0c 0t

For plane, harmonic waves given by E = E0ei1kz - vt2, the appropriate space
B B

and time derivatives required by Eq. (52) can be calculated to give

v2 svm0
k2 = 2
+ ia b (53)
c 1 - iv>g

where we have also made use of the fact that e0m0 = 1>c2 with m0 the perme-
ability of vacuum. Again, we find that the propagation constant must be a
complex number to properly describe the propagation of the wave in a metal.

5 SKIN DEPTH

Before proceeding with the general case described by Eq. (53), we pause
to consider the special case in which the frequency v of the incident radiation
is small enough to allow as a good approximation to Eq. (53)

k2 = ivsm0

Expressing i as eip>2 and taking the square root of each side,

sm0v 1>2
k = 11 + i2a b (54)
2
546 Chapter 25 Optical Properties of Materials

Writing k in the complex form k = kR + ikI , as before, we can identify the

real and imaginary coefficients by

sm0v 1>2
kR = kI = a b (55)
2

and the real and imaginary refractive indices by

c c2sm0 1>2 s 1>2

nR = kR = a b = a b (56)
v 2v 2ve0

and

c s 1>2
nI = kI = a b (57)
v 2ve0

The complex character of k, when introduced into the plane, harmonic wave
equation, leads as in Eq. (29) to

E = E0e -kIzei1kRz - vt2

B B

The real exponential factor e -kIz describes absorption. When the radiation
has penetrated a depth of z = 1>kI , therefore, the amplitude has decreased
to 1/e of its surface value. This particular distance is called the skin depth, d,
where

1 2
d K = (58)
kI A 0v
sm

and is evidently smaller for better conductors with larger s. For 3-cm mi-
crowaves, for example, the skin depth in copper, with conductivity of
5.8 * 107> Æ-m, is only about 6.6 * 10-5 cm.

6 PLASMA FREQUENCY

Returning to the general case of Eq. (53) and introducing there the complex
refractive index, we write

c 2
isc2m0
n2 = a kb = 1 +
v v11 - iv>g2

After multiplying the second term on the right-hand side of this equation by
ig>ig, this becomes

m0sc2g
n2 = 1 - (59)
v2 + ivg
 Optical Properties of Materials 547

The numerator in the second term must have the same dimensions as v2 and
is identified as the square of a plasma frequency given by

Ne2 Ne2
v2p = m0c2gs = m0c2g a b = (60)
mg me0

where we have made use of both Eq. (46) and the relation e0m0 = 1>c2.
The plasma frequency is a resonant frequency for the free oscillations of the
electrons about their equilibrium positions. Inserting it into Eq. (59),

v2p
n2 = 1 - (61)
v2 + ivg

The plasma frequency turns out to be a critical frequency whose value deter-
mines whether the refractive index is real or imaginary. This can be seen by
neglecting the g-term, valid for high-enough frequency 1v $ g2, in which
case Eq. (61) is simply

v2p
n2 = 1 - (62)
v2

Equation (62) now shows that for v 6 vp , the refractive index of the metal
is complex and radiation is attenuated, whereas for v 7 vp , the index is real
and the metal is transparent to the radiation.
Returning to Eq. (61), we find, as before, two equations from which the
real and imaginary parts of the refractive index can be calculated. Equating
real and imaginary parts in

v2p
n2 = 1nR + inI22 = 1n2R - n2I2 + i12nRnI2 = 1 - a b (63)
v2 + ivg

we find

v2p
n2R - n2I = 1 - a b (64)
v2 + g2

2
g vp
2nRnI = a 2 b (65)
v v + g2

These equations, solved simultaneously, permit calculation of curves such as

those in Figure 3. The curves cross at v = 1v2p - g221>2, as is evident from
Eq. (64). Since typically vp $ g, the crossover occurs at v ! vp , dividing
the transparent and the opaque (and highly reflecting) regions. The plasma
frequency for metals falls in the visible to near-ultraviolet regions, so that
they are opaque to visible and transparent to ultraviolet radiation at suffi-
ciently high frequency.
Intermediate to the good insulator and good conductor we have treated
separately are materials, like semiconductors, for which neither of these extreme
cases suffices to explain the properties. Such materials manifest appreciable
contributions to their optical properties from both free and bound charges and
accordingly must be treated by allowing for both types of behavior.
 548 Chapter 25 Optical Properties of Materials

nI

nR
1.0 1.0

Extinction coefficient nI
Refractive index nR
0.10 0.10

0.01 0.01
Figure 3 Angular frequency depen-
dence of the refractive index nR and the
extinction coefficient nI for copper. Val-
ues assumed are vp = 1.63 * 1016 s-1 and
g = 4.1 * 1013 s-1. The crossover point
1014 1015 1016 1017 1018
of the curves coincides with the plasma
frequency. Frequency (rad/s)

PROBLEMS
1 In general, the “electrical constant” K, the dielectric con- 6 Calculate the skin depth in copper for radiation of (a)
stant, is related to the refractive index by 60 Hz and (b) 3 m. First ensure that the approximations of
problem 5 are satisfied. (Handbook data for copper:
K = n2 s = 5.76 * 107>Æ-m.)
a. Show that if KR and KI are the real and imaginary parts 7 Compare the skin depth of (a) aluminum, with conductivity
of the dielectric constant, then of 3.54 * 107>Æ-m and (b) seawater, with conductivity of
4.3>Æ-m, for radio waves of 60 kHz.
KR + 1K2R + K2I21>2 1>2
nR = c d 8 Calculate the skin depth of a solid silver waveguide compo-
2
nent for 10-cm microwaves. Silver has a conductivity of
and 3 * 107> Æ-m. Explain why a more economical silver-plated
brass component will work as well.
- KR + 1K2R + K2I21>2 1>2
nI = c d 9 The energy density of red light of wavelength 660 nm is re-
2
duced to one-quarter of its original value by passage
b. Calculate nR and nI for a dielectric, in terms of KI , at through 342 cm of seawater.
frequencies high enough such that KI = KR . a. What is the absorption coefficient of seawater for red
2 Show that in a nearly transparent medium, the absorption light of this wavelength?
coefficient is related to the conductivity and refractive b. At what depth is red light reduced to 1% of its original
index by energy density?
1377 Æ2s 10 Calculate and/or plot the real and imaginary parts of the re-
a = fractive index for a metal, given the frictional parameter g
nR
and the plasma frequency. Check your results against
3 Calculate and/or plot real and imaginary parts of the re- Figure 3.
fractive index for a dielectric given the frictional parameter
11 Determine the theoretical content of the constants A, B,
g, the resonant frequency v0 , and the dipole density N.
and C used to express the Cauchy dispersion equation.
Check your calculations against Figure 2.
12 In writing Eq. (3) we neglected to include a contribution
4 Assume that aluminum has one free electron per atom and
due to the magnetic force on the electron. Under what con-
a static conductivity given by 3.54 * 107>Æ-m. Determine
dition is the magnitude of the magnetic force exerted by a
(a) the frictional constant g, (b) the plasma frequency vp , (c)
harmonic electromagnetic wave on an electron much less
the real and imaginary parts of the refractive index at 550 nm.
than the magnitude of the electric force exerted by the
5 Show that Eq. (58) for the skin depth at low frequency is an same harmonic electromagnetic wave acting on the
adequate approximation when v % g and v % s>e0 . electron?
 E3

k32

E2

spIp sI
k31 k21
hnp hn!

E1

k30 k10 k20

E0

26 Laser Operation

INTRODUCTION

In this chapter we give a quantitative treatment of laser operation. We begin

by developing the rate equations governing the population densities in a
medium interacting with an electromagnetic field. These rate equations are
developed following an approach taken by Albert Einstein in 1916. The pop-
ulation-density rate equation, together with equations representing the effect
of a cavity on an electromagnetic field, are used to develop a relation that
predicts the output irradiance from a laser given the characteristics of the
pump, gain medium, and cavity that comprise the laser system. We then dis-
cuss the gain bandwidth of laser gain media describing both homogeneous
and inhomogeneous broadening. The use of Q-switching and mode-locking
to produce pulsed output fields is then considered. Finally, we give a qualita-
tive description of the operation and characteristics of diode lasers.

1 RATE EQUATIONS

Rate equations relate the population densities (number of atoms or

molecules per unit volume in a given energy state) to the properties of an
electromagnetic field incident on the atoms or molecules. Recall that the
different energy states of interest in an atom correspond to different
configurations of the charge cloud associated with one of the outermost

549
 550 Chapter 26 Laser Operation

E2 electrons in the atom. In molecules, the energy states are also distinguished
r(n)
hn0 by the vibrational and rotational state of the molecule. For the moment we
E1 will concentrate on but two of the many energy states in an atomic system.
We denote the energies of these levels E1 and E2 with the nominal differ-
Figure 1 Electromagnetic field spectral ence in energy E2 - E1 = hn0 and take, again for the moment, the lower
energy density r1n2 incident on a two-level energy state to be the ground state of the system, shown in Figure 1.
atom. Level 1 is the ground state and the
We denote the population densities of these states N1 and N2 . The
nominal energy difference between the lev-
els is E2 - E1 = hn0 . interaction of an electromagnetic field with a pair of atomic or molecular
energy states can be described by three processes: stimulated emission,
stimulated absorption, and spontaneous emission. The rates associated with
these processes are valid for the case of interaction with a nearly mono-
chromatic laser field of frequency n¿. In the more general case of the inter-
action with an electromagnetic field with spectral energy density r1n2, these
relations must be generalized as
q
RSt.Em. = B21N2 ! 0
g1n2r1n2 dn (1)

RSt.Abs. = B12N1 ! 0
g1n2r1n2 dn (2)

and
q
RSp.Em. = A21N2 ! 0
g1n2 dn = A21N2 (3)

In the last relation we have used the fact that the lineshape function g1n2 is
normalized so that its integral over all frequencies is unity. Note that the rate
of spontaneous emission is independent of the spectral energy density of the
electromagnetic field. The spectral energy density r1n2 has a dimension of
energy per volume per frequency interval. The integration of the spectral en-
ergy density over all frequencies gives the time-averaged energy density 8u9
of the electromagnetic field,
g(n) r(n)

r(n) q
r(n0)
0.035 0.5 ! r1n2 dn = 8u9
0
g(n)

0
n0 n
0 Evaluation of the integrals in Eqs. (1) and (2) requires knowledge of the form
of the spectral energy density r1n2 and lineshape function g1n2. These inte-
(a) grals simplify in two general cases that are of particular interest. These cases
are described in the following subsections.

Broadband Electromagnetic Energy Density

0.035 r(n0)g(n) Consider the case in which the spectral energy density function r1n2 is much
broader than the lineshape function g1n2. The behavior of these functions is
illustrated in Figure 2. Note that, for such a situation, the product function
0
n0 n r1n2g1n2 is significant only over the range of frequencies near n0 for which
(b) g1n2 is significant. Over this range, the energy density function is nearly con-
stant and can be taken to be r1n02. As a consequence the energy density func-
Figure 2 Broadband spectral energy den-
sity. (a) A lineshape function g1n2 (left axis)
tion can be “pulled out” of the integral so that, to a good approximation,
and a broadband spectral energy density
r1n2 (right axis) are shown. (b) The product q q
of the two functions plotted in (a) is
r1n2g1n2 L r1n02g1n2.
! g1n2r1n2 dn = r1n 2 ! g1n2 dn = r1n 2
0
0
0
0
 Laser Operation 551

Thus, for a broadband electromagnetic field, the stimulated rates are

RSt.Em. = B21r1n02N2

RSt.Abs. = B12r1n02N1

Here, n0 is the center frequency of the 2 : 1 atomic transition. The rate of

spontaneous emission is, of course, still given by Eq. (3).
The rate equations describing the interaction of the two-state atomic
system with a broadband electromagnetic field then have the form

dN2
= - A 21N2 - B21r1n02N2 + B12r1n02N1 (4)
dt

and

dN1
= + A 21N2 + B21r1n02N2 - B12r1n02N1 (5)
dt

Note that
dN1 dN2
= -
dt dt

since we are accounting only for processes that couple the energy levels 2 and 1.
Thus a reduction in the population density of state 2 must be accompanied by
an increase in the population density of state 1.
The coefficients A 21 , B21 , and B12 are characteristic of the two energy
states. Their form can be determined using a fully quantum-mechanical treat-
ment1 of the interaction of the atom with an electromagnetic field, but such a
treatment is beyond the scope of this chapter. However, in 1916 Einstein was
able to develop relationships between these so-called A and B coefficients,
without relying on a fully quantum treatment, by considering the situation in
which the atoms in the system come to thermal equilibrium with an electro-
magnetic field. We summarize his approach in what follows.

Relationships Between A and B Coefficients

The spectral energy density in an electromagnetic field in thermal equilibri-
um with its surroundings at temperature T takes the form

8phn3 1
r1n2 = 3 hn>kBT
(6)
c e - 1

[See problem 1.] In thermal equilibrium at temperature T the population

densities N1 and N2 of the atoms satisfy a Boltzmann relation,
N2
= e -1E2 - E12>kBT = e -hn0>kBT (7)
N1

These relations can be compared to the requirements imposed by the rate

equations, Eqs. (4) and (5). In thermal equilibrium the rates of change of the

1
See, for example, M. O. Scully and M. S. Zubairy, Quantum Optics (Cambridge, UK: Cambridge
University Press, 1997).
552 Chapter 26 Laser Operation

population densities should be zero. Solving either Eq. (4) or (5) in steady
state (that is, setting the left-hand side of either of these equations to 0)
results in the relation

A 21
r1n02 =
B121N1>N22 - B21

In thermal equilibrium, Eqs. (6) and (7) can be used in this relation to give
8phn30 1 A 21
3 hn0>kBT
= hn0>kBT
c e - 1 B12e - B21

Rearranging to isolate multipliers of the term ehn0>kBT,

A 21 8phn30 B12 hn0>kBT A 21 8phn30

a - be - a - b = 0
B21 c3 B21 B21 c3

In order for this relation to be true at all temperatures T, the term that multi-
plies ehn0>kBT and the remaining term in parentheses must each be identically
zero. Then, it follows that

A 21 8phn30
= (8)
B21 c3

and

B12 = B21 (9)

Thus we have derived the relations between the Einstein A and B coefficients
by considering a thermal equilibrium situation. However, the A and B coeffi-
cients are properties of the atom alone2 and so Eqs. (8) and (9) are true
whether or not the atom is in thermal equilibrium.

Rate Equations for Monochromatic Light

When the frequency width of the spectral energy density r1n2 is much nar-
rower than the frequency width of the lineshape function g1n2, as is typically
the case for laser light, the stimulated emission and absorption rates appear-
ing in Eqs. (1) and (2) can be approximated as,

q
RSt.Em. = B21N2g1n¿2 ! r1n2 dn = B
0
21N2g1n¿28u9 = B21N2g1n¿21I>c2

(10)

and
q
RSt.Abs. = B12N1g1n¿2 ! r1n2 dn = B
0
12N1g1n¿28u9 = B12N1g1n¿21I>c2

(11)

2
Actually, these coefficients depend also on the structure of the enclosure surrounding the
atom, but this dependence typically becomes important only for enclosures of dimension not too much
larger than the wavelength of light.
 Laser Operation 553

Here n¿ is the center frequency of the spectral energy density of the electromag-
netic field. For nearly monochromatic fields we typically say simply that n¿ is the
frequency of the field. Finally we have noted [see problem 2] that for nearly
monochromatic fields, 8u9 = I>c, where I is the irradiance of the electromag-
netic field. Using Eqs. (10) and (11) allows the rate equations governing the
population densities in the states 2 and 1 to be written as

dN2
= - A 21N2 - B21g1n¿21I>c2N2 + B12g1n¿21I>c2N1 (12)
dt

and
dN1
= + A 21N2 + B21g1n¿21I>c2N2 - B12g1n¿21I>c2N1 (13)
dt

Following a particular convention, we choose to define the stimulated emis-

sion cross section s as

s = B21g1n¿2hn¿>c

For an atomic system in which the degeneracy of the upper and lower states
each is 1, s is also the stimulated absorption cross section3 since, in that case,
B12 = B21 . With this definition, the rate equations can be written as

dN2 sI
= - A 21N2 - 1N - N12 (14)
dt hn¿ 2

and
dN1 sI
= + A 21N2 + 1N - N12 (15)
dt hn¿ 2

Now s has a dimension of area which accounts for its designation as a cross
section. Further, the factor sI>hn¿ has a dimension of inverse time and so is a
rate. In fact, it is the stimulated emission rate and can be compared to the
spontaneous emission rate A 21 . As we see in following sections, the ratio of
the stimulated emission rate and the spontaneous emission rate governs the
behavior of the interaction of light with an atomic system.

2 ABSORPTION

Light may either be attenuated or amplified as it propagates through a

medium. The population densities of the different energy states in the medium
determine the amount of attenuation or amplification that a given electro-
magnetic field will undergo. In this section we derive steady-state population
densities for a two-level atomic system, in which the lower energy level is the
ground state of the system. We let laser light of irradiance I and frequency n¿
be incident on the atomic medium. This system is described by Eqs. (14) and
(15). At room temperature nearly all the atoms in a medium will be in the
atomic ground state. When laser light is incident on such a medium
a significant population density can accumulate in an excited state only if the

3
More generally, sabs = 1g2>g12s, where g2 and g1 are the degeneracies of the upper and lower
levels, respectively.
 554 Chapter 26 Laser Operation

transition to the excited state is resonant with the electromagnetic field so that
E2 - E1 L hn¿. Let us assume that only such an excited state 2 accumulates a
significant population density as a result of the interaction with the electro-
magnetic field. In this case we can set

N1 + N2 = NT (16)

where NT is the total population density of the atoms in the medium. Equa-
tion (14), considered in steady state 1dN2>dt = 02, and Eq. (16) can be solved
jointly to give the steady-state values of the population densities N1 and N2 .
The result (see problem 3) is

sI>1hn¿A 212
N2 = NT a b (17)
1 + 2sI>1hn¿A 212

1 + sI>1hn¿A 212
N1 = NT a b (18)
1 + 2sI>1hn¿A 212

We shall see that the population inversion Ninv K N2 - N1 is of physical in-

terest. For the case at hand, the population inversion is always negative since

1
Ninv = N2 - N1 = - NT a b (19)
1 + 2sI>1hn¿A 212

Note that the key parameter governing the distribution of population

between the two levels is the ratio of the stimulated emission rate sI>hn¿ and
the spontaneous emission rate A 21 . Plots of the population densities of the
two levels and the population inversion as functions of light irradiance are
given in Figure 3. Note that the inversion is always negative and decreases in
magnitude as the irradiance of the light incident on the medium increases. In
the limit of very large irradiances, the inversion tends to zero. This behavior
can be understood by noting that two processes, stimulated emission and
spontaneous emission, drive an atom from level 2 to level 1. Only one process,
stimulated absorption, drives the atom from level 1 to level 2. The rates of
stimulated emission and absorption are equal, and so the “downward” (2 to 1)
rate always exceeds the “upward” (1 to 2) rate; thus a steady-state population
inversion cannot occur in a two-level atomic system. In the limit of very large
irradiances, the spontaneous emission process becomes negligible, leading to
a near equalization of population densities.

N1/NT
Scaled population density

0.5
N2 /NT

Ninv /NT
"0.5
Figure 3 Scaled population densities of the
excited (2) and ground (1) states of a two-
level atomic system as a function of the scaled
irradiance I>1A 21hn¿>s2. The scaled popula- "1
tion inversion density Ninv>NT = 1N2 - N12> 0 2 4 6 8 10
NT is also shown. Scaled irradiance
 Laser Operation 555

Absorption Coefficient and Beer’s Law Medium

Thus far we have concentrated on the change in population density induced I0 I(z) IL
by an incident electromagnetic field. In doing so we have implicitly assumed
that the irradiance of the field is constant temporally and spatially. Con-
z#0 z#L
sider the situation depicted in Figure 4. Light of irradiance I0 is incident on
an atomic medium from the left as shown. As the light propagates in the Figure 4 Light propagating through a
z-direction across the medium, there can be an exchange of energy between medium of length L. The irradiance I
the light and the atoms in the medium. The absorption or loss coefficient a changes due to interaction with the atoms
in the medium.
characterizes the spatial rate of change of the irradiance, that is,

dI
= - aI (20)
dz

Let us now form the left-hand side of Eq. (20). Let ¢n>¢t be the incremental
rate of change of the number n of photons (traveling in the + z-direction)
through a small volume ¢V of cross-sectional area ¢A and length ¢z due to
interaction with a gain medium. The incremental change in irradiance ¢I
across this small volume then can be written as

hn¿ ¢n
¢I =
¢A ¢t

Dividing each side of this relation by ¢z gives

¢I hn¿ ¢n
= (21)
¢z ¢V ¢t

Now, the net rate of photon production or loss in the small volume ¢V is the
result of spontaneous emission, stimulated emission, and stimulated absorp-
tion. For sizable incident irradiances, the rate of spontaneous emission into
the direction of the beam is much less than the stimulated emission and stim-
ulated absorption rates, and so spontaneous emission makes a negligible con-
tribution to the net rate of photon production. Then, since each stimulated
emission event creates one photon and each stimulated absorption event re-
moves one photon from the field, the rate of change of the number of pho-
tons in the volume ¢V (see the last term in Eq. 14) is

¢n sI
= 1N2 - N12 ¢V
¢t hn¿

Using this in Eq. (21) leads to

¢I
= sI1N2 - N12
¢z

Letting the increments pass into differentials gives the seminal relation

dI
= s1N2 - N12I (22)
dz

Thus we have discovered the form of the loss coefficient a appearing in

Eq. (20):
a = - s1N2 - N12 (23)

Note that the loss coefficient has a dimension of inverse length.

556 Chapter 26 Laser Operation

As is evident from Eq. (18) and Figure 3, for sufficiently small input
irradiance (or sufficiently small cross section s) the population density
remains concentrated in the ground state, N1 L NT . In this case,

a L sNT K a0 (24)

is a constant, independent of the irradiance of the input field. Here we have

introduced a0 as the small-signal 1I : 02 loss coefficient. For this case,
Eq. (22) takes the form
dI
= - a0I
dz

and can be directly integrated as

dI
= - a0 dz
I
IL L
dI
!I0 I
= - a0 !0
dz

ln1IL>I02 = - a0L

IL = I0e -a0L (25)

This result is known as Beer’s law. The simple exponential-decay nature of this
expression follows only for the case of a weakly absorbing system for which
nearly all of the population remains in the ground state. The treatment of a
system in which a significant population is transferred to an excited state is
mathematically more complex. In such a situation, which is the subject of
problem 4, the loss coefficient decreases (saturates) with increased incident
irradiance.

Example 1
The cross section s, for a transition from the ground state to an excited state
that is resonant with an electromagnetic field of wavelength 808 nm, for a
neodymium (Nd) atom doped into a YAG (yttrium aluminum garnet) crystal4
is about 3 * 10-20 cm2. Assume that the dopant density (number of atoms
per cm3) of Nd in the YAG crystal is 1020 atoms>cm3 and that the YAG crys-
tal itself is transparent to 808-nm light. Assume that a diode laser with an
emission at a wavelength of 808 nm is to be used to pump an Nd:YAG laser
rod. (Pump energy absorbed by the crystal, as we shall see in the next sec-
tion, can be converted to Nd:YAG laser output.)

a. Estimate the small-signal absorption coefficient for the Nd:YAG crystal.

b. Estimate the depth to which the diode laser beam would penetrate sig-
nificantly into the Nd:YAG crystal.

Solution
a. The small-signal absorption coefficient is

a0 = sNT = 13 * 10-202110202 cm-1 = 3 cm-1

4
Modeling absorption in the the Nd:YAG crystal by a two-level system with no degeneracies is
a rather drastic oversimplification of the real situation. Example 1 should be regarded as an “order-of-
magnitude, back of the envelope” estimation.
 Laser Operation 557

b. Let us take the depth of “significant penetration” to be the depth L at

which the irradiance of the diode laser has decreased to 1>e L 0.368 of
its initial value. Then assuming that it is appropriate to use the small-
signal result given in Eq. (25),

IL = I0e -1 = I0e -a0L

L = 1>a0 = 0.33 cm

Thus Nd atoms deeper than about 0.33 cm would not absorb much of
the pump energy.

3 GAIN MEDIA

We have seen that a simple two-level system with the lower energy state being
the ground state always acts as an absorber. Laser systems require a gain medi-
um that provides energy to a field that propagates through it. As alluded to in
Example 1, gain media must be pumped. That is, the pump energy stored in the
gain medium is converted to irradiance by interaction of the electromagnetic
field with the gain medium. A common atomic- level structure that serves as a
E3
gain medium is illustrated in Figure 5. In this figure four important levels are k32
shown,5 with the level of energy E0 being the ground state. The light field to be E2
amplified is taken to have irradiance I and frequency n¿ and should be nearly spIp sI
k31 k21
resonant with the 2 to 1 transition. Pump energy is provided to the system via hn!
hnp
the 0 to 3 transition. For concreteness, let us assume that the pump is a laser of E1
frequency np = (E3 - E0)/h and irradiance Ip . The rate equations governing k30 k10 k20
population densities of the four levels shown in Figure 5, due to the indicated
E0
processes, take the form
Figure 5 Level structure of a four-level gain
medium. The thicker, double-headed arrows
dN3 spIp indicate stimulated processes, and the lighter,
= - k3N3 - 1N3 - N02 (26) single-headed arrows indicate decay process-
dt hnp
es. Gain is provided to a field of irradiance I
and frequency n¿ L 1E2 - E12>h as a result
dN2 sI of an optical pump of irradiance Ip and fre-
= k32N3 - k2N2 - 1N - N12 (27)
dt hn¿ 2 quency np L 1E3 - E02>h.

dN1 sI
= k31N3 + k21N2 - k10N1 + 1N - N12 (28)
dt hn¿ 2

dN0 spIp
= k30N3 + k20N2 + k10N0 + 1N3 - N02 (29)
dt hnp

Here we have replaced the spontaneous emission A coefficients with

decay rates k in order to account for any decay process that causes a down-
ward transition in the atom, including spontaneous emission and, for exam-
ple, inelastic collisions with other atoms, molecules, or particles. These latter
processes are examples of what is called nonradiative decay. Also, sp is the
stimulated emission cross section for the 3-to-0 transition and we have again
ignored, for simplicity, level degeneracies. Finally, we will assume that the
four-level system is closed, in the sense that all atoms in the sample exist in
one or the other of these four levels so that

NT = N0 + N1 + N3 + N4 (30)

5
See problem 7 for another common arrangement involving three levels.
558 Chapter 26 Laser Operation

If this is to be true, the right-hand sides of the Eqs. (26) though (29) must sum to
zero. This is so provided that k3 = k32 + k31 + k30 and k2 = k21 + k20 . The
lifetime t of an energy level is defined to be the inverse of the total decay rate
from the level so that, in the present case,

1 1 1
t3 = , t2 = , and t1 =
k3 k2 k10

The lifetime of a level is the time for the population density of a given level to
decay to 1/e of its initial value, when the decay process is the only process that
occurs.
The analysis leading to Eq. (22) holds as well for the four-level gain
medium being presently considered. Thus gain will occur provided that a
population inversion 1N2 7 N12 exists in steady state. Under this condition,
stimulated emission will exceed stimulated absorption and a net production
of photons will occur. Any three of the four-level rate equations, together
with Eq. (30), can be solved simultaneously to give the steady-state popula-
tion densities of the four levels and so to find the steady-state population
inversion Ninv = N2 - N1 . This full analysis is left as an exercise (problem 5). It
is instructive to make a series of simplifying assumptions that sometimes
apply in real atomic gain media and which dramatically reduce the complexi-
ty of the system of equations. We do so in the following subsections.
Undepleted Pump Approximation It is often true that the laser pump
does not significantly empty the ground state of the four-level system so that
N0 W N3 and N0 L NT . In this case, Eq. (26) can be solved in steady state
to give
1 spIp
N3 L N (31)
k3 hnp T

It is important to note that, in the undepleted pump approximation, N3 is ap-

proximately constant and although N3 V N0 , the rate k3N3 is not negligible.
Using Eq. (31) in Eqs. (27) and (28) leads to, in steady state,

dN2 sI
= 0 = Rp2 - k2N2 - 1N - N12 (32)
dt hn¿ 2

dN1 sI
= 0 = Rp1 + k21N2 - k10N1 + 1N - N12 (33)
dt hn¿ 2

where we have defined the effective pump rate densities

k32 spIp
Rp2 = a N b
k3 hnp T

k31 spIp
Rp1 = a N b
k3 hnp T

Equations (32) and (33) can be solved for the population inversion, giving

11 - k21>k102Rp2 - 1k2>k102Rp1
Ninv = N2 - N1 = (34)
k2 + 11 + k20>k1021sI>hn¿2
 Laser Operation 559

When describing gain media, Eq. (22) is often recast by defining a gain
coefficient g as

g = s1N2 - N12 = sNinv (35)

so that

dI
= gI (36)
dz

It is important to note that the gain coefficient depends on the irradiance I,

since the population inversion is, in general, dependent on the irradiance.
Evidently, maximizing the population inversion maximizes the gain coeffi-
cient. Examination of Eqs. (34) and (35) indicates that population inversion and
so the gain coefficient are made larger under the following conditions:

1. The gain coefficient becomes larger as k10 is increased relative to the

decay rates associated with level 2. This is sensible since stimulated
absorption attenuates the irradiance, and a large decay rate from level 1
indicates that the atoms do no not dwell a long time in level 1 and so the
chances of stimulated absorption events occurring are reduced.
2. The gain coefficient becomes larger by reducing Rp1 . This is again sensi-
ble since the effective pump rate Rp1 feeds the population of level 1 and
so increases the likelihood of stimulated absorption. The effective pump
rate density Rp1 is less in an atom for which k31 is small.
3. The gain coefficient is increased by reducing the decay rate from level 2.
By doing so the dwell time of the atoms in the upper lasing level 2 in-
creases and so the likelihood of the occurrence of stimulated emission,
which leads to gain, is increased. Energy levels with such long lifetimes
are often referred to as metastable states.
4. The gain coefficient is increased by increasing the effective pump rate
density Rp2 that feeds the upper lasing level and so increases the likeli-
hood of occurrence of stimulated emission events. This rate can be in-
creased by increasing the irradiance Ip of the pump laser.

Ideal Four-Level Gain Medium

In the ideal case, Rp1 = 0 and k10 : q . Then, the gain coefficient takes the
simple form

sRp2>k2 g0
g = K (37)
1 + 1sI>hn¿2>k2 1 + I>IS

Here we have introduced two important parameters that describe gain media,
the small-signal gain coefficient g0 and the saturation irradiance IS . For the
ideal four-level gain medium, these parameters take the form
g0
g0 = sRp2>k2 = sRp2t2 (38) g

and g0 / 2

IS = hn¿k2>s = hn¿>st2 (39)

0
0 1 2 3 4 5
Note that the small-signal gain coefficient is the approximate gain coefficient
I/IS
when the irradiance I in the medium is much less than the saturation irradiance
IS . As illustrated in Figure 6, when I = IS the gain coefficient is reduced by a Figure 6 Gain coefficient g as a function of
factor of 2 from its small-signal value. Further, when I W IS the gain coeffi- scaled irradiance I>IS . Note that g = g0>2
cient becomes inversely proportional to the irradiance, and the spatial rate of when I = IS .
 560 Chapter 26 Laser Operation

change of the irradiance gI given in Eq. (36) becomes constant. Although the
middle member of Eq. (37) is valid only for an ideal four-level gain medium,
the general form of the gain coefficient given as the rightmost member of
that equation holds for a wide range of homogeneously broadened gain
media (see Section 5), but with definitions of g0 and IS different from those
given in Eqs. (38) and (39).
That the gain coefficient is less for fields of larger irradiance is said to be
due to gain saturation. Gain saturation occurs because large irradiances cause
more stimulated emission, which depletes the steady-state population density
of the upper lasing level, thus reducing the likelihood of a stimulated emis-
sion event.
Integrated Gain
Using Eq. (37) in Eq. (36) results in a differential equation that can be
integrated to provide a relation between the irradiance I0 input into a gain
medium and the irradiance IL output from that gain medium. Proceeding,

dI g0
= gI = I (40)
dz 1 + I>IS

Separating variables and integrating both sides of the resulting relation gives
IL L
1 1
!I0
a
I
+
IS
b dI = g0 !0
dz

Integration gives

IL 1
ln a b + 1IL - I02 = g0L (41)
I0 IS

The important result given in Eq. (41) is a transcendental relation that can
be numerically solved for the output irradiance IL given the input irradiance
and the characteristics of the gain medium, g0 , L and IS . Important and
interesting features of this relationship are illustrated in a series of problems
at the end of this chapter.
Figure 7 indicates the effects that varying the small-signal gain coeffi-
cient g0 and the saturation irradiance IS have on the output irradiance IL
from a gain cell. Note that for all three curves, the irradiance input to the cell
is I0 = 1 W>cm2. The lower curves have the same small-signal gain coefficient

20

g0 ! 2/cm
15
IS ! 10 W/cm2
IL (W/cm2)

g0 ! 1/cm
10 IS ! 20 W/cm2

Figure 7 Irradiance output IL from a gain

5
cell as a function of the length L of the gain
g0 ! 1/cm
cell. The plots correspond to the different
small-signal gain coefficients g0 and satura- IS ! 10 W/cm2
tion irradiances IS indicated. In each case, 0
the irradiance input into the gain cell is 0 1 2 3
1 W>cm2. Gain length L (cm)
 Laser Operation 561

but different saturation irradiances. These two lower curves show that, for
gain cell lengths L short enough that the irradiance in the cell remains well
below the saturation value, the output irradiance from the cell is nearly inde-
pendent of the saturation irradiance. However, for gain cells that are long
enough to allow the irradiance to grow to an appreciable fraction of the satu-
ration irradiance, the output irradiance is less for cells with smaller saturation
irradiances. The gain cell represented by the upper curve has twice the small-
signal gain coefficient of the cells represented by the lower two curves, and as
a consequence the output irradiance shown in the upper curve is significantly
larger, even for short gain lengths, than those shown in the lower curves.
In closing this section, we point out that both the saturation irradiance
IS and the small-signal gain coefficient g0 are functions of the electromagnetic
field frequency n¿, since they each depend on the cross section s, which in
turn is proportional to the lineshape function g1n¿2. In fact,

g0 r g1n¿2
and
IS r 1>g1n¿2

The import of these relations for the operation of a gain cell is explored in
problem 13.
M1

4 STEADY-STATE LASER OUTPUT

Equation (41) allows for the prediction of the steady-state or continuous

wave (CW) output irradiance of a laser given the natures of the pump, gain g0, IS
I0 IL Iout # T3 IL
medium, and cavity that constitute the laser. We will develop first an expres- Gain
sion for the CW output irradiance for a ring laser and then write down the M2 M3
L
corresponding result for the more common case of a laser system that uses a
two-mirror linear cavity. Figure 8 Ring laser. The field is con-
strained to oscillate in the counterclockwise
direction by means not shown. The gain cell
Ring Laser
has length L and is characterized by small
We choose to illustrate the development of an expression for the steady-state signal gain coefficient g0 and saturation ir-
irradiance output from a laser cavity by considering a ring laser consisting of radiance IS . The output mirror has trans-
three mirrors as shown in Figure 8. Take the reflectances of the three mirrors mittance T3 .
M1 , M2 , and M3 to be R1 , R2 , and R3 , respectively. Further, let mirror M3 be
the output mirror with a transmittance T3 . The gain medium is characterized
by its saturation irradiance IS , length L, and small-signal gain coefficient g0 .
We assume that lasing action occurs only in the counterclockwise direction
around the ring.
In steady state, the losses due to imperfect reflection and transmission
at the mirrors must be, in each round-trip, offset by the increase in irradiance
that results from interaction with the gain medium. As indicated in Figure 8, we
will take the steady-state irradiance at the input end of the gain cell to be I0
and that at the output end to be IL . Tracking the irradiance from the output
end around the ring to the input end leads to the relation

I0 = SIL (42)

Here, S is the fraction of the irradiance that survives the trip around the cavity
from the output to the input of the gain cell. For the simple cavity considered
here, S = R1R2R3 . More generally, the survival factor S should include factors
describing each process that reduces the irradiance as the beam traverses the
ring. Equation (42) can be used in Eq. (41) to give the desired expression for
the steady-state irradiance at the output end of the gain cell. After a bit of
562 Chapter 26 Laser Operation

rearrangement, we find

g0L - ln11>S2
IL = IS a b
1 - S

The irradiance of the laser beam exiting the ring cavity is, then,

g0L - ln11>S2
Iout = T3IL = T3IS a b (43)
1 - S

Let us examine the features of this expression. Larger irradiances occur for
gain media with larger saturation irradiances. However, the saturation irradi-
ance is a property of the gain media and cannot be manipulated. With a given
gain medium one can increase the output irradiance by increasing the pump
density Rp2 , since this increases the small signal gain coefficient g0 . A bit of
analysis (see problem 17) indicates that decreasing the cavity losses not
associated with the laser output also increases the laser output irradiance.
Note that Eq. (43) is only meaningful when the small-signal gain
coefficient exceeds a certain threshold value,

1
gth = ln 11>S2 1Ring cavity2 (44)
L

since it predicts a negative irradiance when g0 6 gth . Of course, there are

no negative irradiances. Rather when this condition holds, the cavity loss
per round-trip exceeds the gain per round-trip and no irradiance builds up
in the cavity. Note that gth characterizes the cavity losses. Using Eq. (44) in
Eq. (43) yields the indicative relation

1g0 - gth2L
Iout = T3IS a b (45)
1 - S

Let us obtain the expression for the threshold small-signal gain coeffi-
cient gth in another fashion. The field in a laser cavity builds from sponta-
neous emission events occurring into a laser cavity mode. These spontaneous
emission events are then amplified by the action of stimulated emission with-
in the gain medium. However, the gain provided must exceed the losses en-
countered per round-trip due to imperfect mirrors, transmission from the
output mirror, and other cavity loss mechanisms. The irradiance due to the
spontaneous emission is typically much less than the saturation irradiance of
the gain medium. Thus, during the initiation of the laser field the gain coeffi-
cient g has its small-signal value g0 . In this case, the irradiance I0 input into
the gain cell and the irradiance IL output from the gain cell are simply relat-
ed (see problem 8) by

IL>I0 = eg0L

In order for the irradiance to grow during each round-trip, this single-pass
small-signal gain factor eg0L must more than offset the fractional loss in a
round-trip. That is,
Seg0L 7 1 (46)

This condition implies that for the laser field to grow,

ln11>S2
g0 7 = gth
L
 Laser Operation 563

If g0 7 gth , the field grows with each round-trip through the cavity, but as it
does so the gain coefficient g decreases due to saturation. When the gain co-
efficient g is reduced so that the gain per pass through the gain cell just off-
sets the cavity loss per round-trip, steady state is reached. In summary, the
small-signal gain coefficient g0 must exceed the threshold gain coefficient gth
for the laser field to grow in the cavity. This field will then continue to grow
until the gain coefficient g saturates to the threshold value gth .

Example 2
A ring laser cavity system like the one shown in Figure 8 uses mirrors with
reflectances R1 = R2 = 0.99 and R3 = 0.95. The transmittance of the out-
put mirror is T3 = 0.04. The gain medium is a 10-cm-long Nd:YAG crystal
with a saturation irradiance of 2300 W>cm2. The crystal is optically pumped
at a rate that leads to a small-signal gain coefficient of g0 = 0.05>cm.
a. Find the threshold gain coefficient for this cavity.
b. Find the irradiance of the laser field that exits this cavity.
c. Assuming that the laser beam has cross sectional area A = 0.1 cm2,
find the output power of the laser.
d. Assuming that the overall efficiency of this laser system is 3%, find the
pump power required to operate this laser system.
Solution
a. Using Eq. (44),

1 1 1 1 1
gth = ln11>S2 = ln a b = ln a b = 0.0071>cm
L L R1R2R3 10 cm 0.99 # 0.95
2

b. According to Eq. (45),

1g0 - gth2L 10.05 - 0.0071210

Iout = T3IS a b = 10.04212300 W>cm22
1 - S 1 - 0.99210.952

Iout = 570 W>cm2

c. The output power is, then, approximately

Pout L IoutA = 1570 W>cm2210.1 cm22 = 57 W

d. The required pump power would be

Pout 57 W
Ppump = = = 1900 W
efficiency 0.03

Two-Mirror Linear Cavity M1 M2

Gain
The analysis leading to the output irradiance for the more common two-mirror Iout
linear laser is somewhat more complicated than that given for the ring laser,
but nevertheless can be carried out in a similar fashion.6 The result for a two- L
mirror cavity like the one shown in Figure 9 (see problem 19) is Figure 9 Linear laser cavity.

T2IS g012L2 - ln A R11R2 B

Iout = (47)
2 A 1 - 2R1R2 B A 1 + 2R2>R1 B

6
See, for example, Joseph T. Verdeyen, Laser Electronics, 3d ed. (Prentice-Hall: Englewood
Cliffs, New Jersey, 1995), Ch. 9.
564 Chapter 26 Laser Operation

Here, T2 is the transmittance of the laser cavity output mirror, R1 and R2 are
the reflectances of the cavity mirrors, and the gain cell is characterized by
length L, small-signal gain coefficient g0 , and saturation irradiance IS .
Equation (47) implies that for a two-mirror linear cavity, the threshold gain
coefficient is
1 1
gth = ln a b
2L R1R2

More generally, an analysis like that leading to Eq. (46) leads to

1 1
gth = ln a b 1Linear cavity2 (48)
2L S

Example 3
Estimate the minimum length of a linear cavity with mirror reflectances
R1 = 0.99 and R2 = 0.98 that can be used in a He-Ne laser system given a
small-signal gain coefficient of 0.001/cm.
Solution
The small-signal gain coefficient must exceed the threshold gain coefficient
for lasing action to occur. Thus, using Eq. (48),

1
gth = ln11>S2 6 g0
2L
so
1 1 1
L 7 ln11>S2 = ln a b = 15 cm
2g0 210.001>cm2 10.99210.982

Since the gain cell must fit into the laser cavity, the laser cavity must be longer
than 15 cm. One rarely finds a He-Ne laser shorter than 15 cm.

5 HOMOGENEOUS BROADENING

You should be familiar with the notion of the lineshape factor g1n¿2 that gov-
erns the strength of the interaction of an atomic system with an incident field
of frequency n¿. In this section we discuss the underlying broadening mecha-
nisms that determine the width ¢n of the lineshape function and distinguish
between homogeneous and inhomogeneous broadening mechanisms. Briefly,
homogeneous broadening mechanisms are those physical influences that
broaden the linewidth of the frequency response of each atom in the medium
in the same manner, whereas inhomogeneous broadening mechanisms affect
different groups of atoms in different ways—typically making the central fre-
quency of the lineshape function different for different atoms.
Two important homogeneous broadening mechanisms are called lifetime
broadening and pressure broadening. These effects can be shown to lead to a
Lorentzian lineshape function,

¢nH
g1n2 = (49)
2p[1n - n022 + 1¢nH>222]

This function peaks at n = n0 and has a full-width at half-maximum of ¢nH . The

full-width at half-maximum, sometimes called the linewidth of the transition or
 Laser Operation 565

the gain bandwidth, can be shown to be well approximated by

1 1 1
¢nH = a + + 2rcol b (50)
2p t2 t1

Here t2 and t1 are the lifetime of the upper and lower levels, respectively,
participating in the transition. The terms involving these lifetimes arise from
the process called lifetime broadening. The remaining term on the right-hand
side of Eq. (50) is due to pressure broadening, which contributes a term pro-
portional to the rate of collisions, rcol , of the gain atoms with each other and
with other species in the gain medium. This term is important primarily in
gas media. We now describe briefly each of these homogeneous-broadening
mechanisms.

Lifetime Broadening
Classically, an atom driven by a nearly resonant electromagnetic field oscil-
lates at the frequency of the driving field. This oscillation appears in a quan-
tum-mechanical treatment of the atom as an alternating transition between
the upper and lower laser states. Spontaneous decay from either of these lev-
els removes that atom from the coherent interaction with the field. The net
effect, in an assembly of oscillating atoms, is to introduce an additional time
dependence related to the random deletion of the contribution of one of the
oscillators from the net charge oscillation in the medium. As a result, the
charge oscillation in the gain medium cannot be represented by a function
oscillating with a single frequency but rather has a broadened Fourier spec-
trum. The lifetime broadening contribution to the lineshape function g1n2 is
related to this broadened spectrum. The finite lifetime of the atomic levels is
due both to the fundamental process of spontaneous emission and to transi-
tions induced by inelastic collisions with other atoms. In these inelastic inter-
actions, the collisions induce a change in internal state of one or both of the
atoms participating in the collisions.

Pressure Broadening
Pressure broadening is a result of elastic collisions between atoms in the gain
medium. Although these elastic collisions induce no change in the internal
state of the colliding atoms, they act to interrupt the regular oscillation of the
charge in the atom. These interruptions introduce additional Fourier compo-
nents into the spectrum of the oscillations of the gain medium and hence con-
tribute to the broadening of the atomic response. The pressure-broadening
contribution to the gain bandwidth is, of course, pressure and temperature
dependent. For many gas gain media, pressure broadening makes the domi-
nant contribution to the homogeneous bandwidth.

Example 4
Consider a homogeneously broadened transition in a carbon dioxide 1CO22
laser.
a. Find the lifetime-broadening contribution to the homogeneous linewidth
for this transition if the lifetime of the upper lasing level is 10 ms and that
of the lower lasing level is 0.1 ms.
b. The linewidth of this transition is measured to be 1 GHz. Estimate the
pressure-broadening contribution to the linewidth of this transition.
c. Gain bandwidths are often expressed as wavelength spreads ¢l rather
than frequency spreads ¢n. Express the linewidth of this transition as
a wavelength spread given that the light resonant with this transition
has a wavelength of 10.6 mm.
 566 Chapter 26 Laser Operation

Solution
g0(n)
a. According to Eq. (50), the lifetime broadening contribution ¢nH,t to
the linewidth is
gth

1 1 1 1 1 1
¢nH,t = a + b = a + b Hz
2p t1 t2 2p 10 * 10-6 0.1 * 10-6
nfsr n0 = 1.61 * 106 Hz
(a)
b. The contribution due to pressure broadening ¢nH,p is evidently

¢nH,p = ¢nH - ¢nH,t = 1 GHz - 0.00161 GHz L 1 GHz

g(n)
gth Under the operating conditions leading to a linewidth of 1 GHz, this
transition is predominantly pressure broadened.
c. The wavelength spread associated with the linewidth is found by not-
ing that
nfsr n0
dn = d1c>l2 = - 1c>l22 dl
(b)
For small spreads, then, the magnitude of the wavelength spread is

¢l = l2 ¢n>c = 110.6 * 10-6 m221109 Hz2>13 * 108 m>s2

= 3.75 * 10-10 m = 0.375 nm
gth
g(n) Gain Saturation in Homogeneously Broadened Media
It is important to note that the expression for the gain coefficient given in
Eq. (37) is valid only for homogeneously broadened media since the deriva-
nfsr n0
tion of this relation implicitly relies on the assumption that all atoms partic-
ipating in the interaction with the light are of the same type. Further,
(c) recall that the low-loss “cavity modes” of a linear laser cavity are separated by
Figure 10 Gain saturation in a homoge- c/2d, where d is the cavity length. (For a ring cavity, the mode separation is c/P,
neously broadened laser. Cavity mode fre- where P is the cavity perimeter.) This mode separation is sometimes called
quencies are separated by nfsr = 0.15 GHz. the free spectral range, nfsr , of the cavity. In general, several different cavi-
For all three plots the gain bandwidth is 1
ty modes might have frequencies within the bandwidth of the gain medium
GHz. The small-signal gain coefficient at
linecenter g01n02 is twice the threshold gain and so see significant gain. Let us be a bit more specific. Consider a 1-meter-
coefficient gth . (a) At laser turn on the gain long linear cavity with a mode spacing nfsr = 1.5 * 108 Hz. Let this cavity
coefficient has its small-signal value g01n2 contain a gain medium of bandwidth 1 GHz pumped so that at the frequen-
and six cavity modes are above threshold. cy of the transition line center n0 the small-signal gain coefficient g01n02 is
(b) As the irradiance in the cavity modes
twice the threshold gain coefficient gth . Recall that the small-signal gain co-
above threshold grows, the gain coefficient
is reduced due to gain saturation. At the in- efficient is proportional to the lineshape function so that the width of the
stant shown, four cavity modes are above small-signal gain coefficient is ¢nH = 1 GHz. The gain coefficient and cavi-
threshold. (c) Gain saturation has reduced ty modes for this system are shown in Figure 10.
the gain coefficient so that only the mode When the laser is turned on, the irradiance is very small and so the
nearest linecenter survives.
gain coefficient g1n2 takes on its small-signal value g01n2. This situation is
illustrated in Figure 10a. Notice that in this case six cavity modes are at fre-
quencies that have gain larger than the threshold value. The irradiance in each
of these six modes, therefore, begins to grow. As the irradiance grows, the gain
saturates and the gain coefficient is reduced. As the gain coefficient is
reduced, the number of cavity modes above threshold decreases. Figure 10b
shows the gain coefficient at a time in the buildup to steady state when only
four cavity modes remain above threshold. These four modes continue to
grow while all others die out. As the irradiance in these four modes grows,
gain saturation further reduces the gain coefficient, causing still more
modes to drop below threshold. Steady state is reached when only the cavi-
ty mode with the frequency closest to the line center frequency n0 experiences
 Laser Operation 567

a gain coefficient equal to the threshold value. This situation is shown in

Figure 10c.
Note that this scenario indicates that only a single cavity mode can be
present in the CW output of a laser containing a homogeneously broadened
gain medium. In fact, it is possible that more cavity modes could be present in
the output, but only if these modes utilize different spatial portions of the
gain medium. In contrast, as we shall see in the next section, many different
cavity modes can lase in a system that uses an inhomogeneously broadened
gain medium.

6 INHOMOGENEOUS BROADENING

The homogeneous gain bandwidth results from the broadening of the fre-
quency response of each atom in the medium. In addition, there can be envi-
ronmental factors that cause the center frequencies of each of the atoms in an
assembly of atoms to differ from one another. Then, the frequency response
of the gain medium reflects both the homogeneous bandwidth of each atom
and the distribution of the center frequencies of the atoms. In gas lasers the
most important inhomogeneous broadening mechanism is Doppler broaden-
ing, due to the Doppler effect. Recall that the frequency measured by a
detector depends on the velocity of the source of the wave relative to the
velocity of the detector. Since the atoms in a gas gain medium have a distrib-
ution of velocities (the Maxwell Boltzmann distribution), the radiation from
different atoms, each of which emits light at frequency, say, n0 in its own rest
frame, will be perceived in the laboratory as radiation with a spread of fre-
quencies related to the spread of velocities of the atoms in the gain medium.
Without proof we state that the lineshape function, resulting from this
process, is well described by a Gaussian function,

4 ln122 1>2
2
g1n2 = a b e -4 ln122[1n - n02>¢nD] (51)
p¢n2D

Here, n0 is the center frequency of the emission from the atoms in their rest
frame and ¢nD is the full width at half-maximum of the Doppler-broadened
lineshape function. For an assembly of gain atoms each of mass M at temper-
ature T,
1>2
8kBT
¢nD = a ln122b n0 (52)
Mc2

Figure 11 shows the lineshape function g1n2 for a Doppler-broadened gain

medium consisting of atoms each of which has a homogeneous linewidth
¢nH . In general, both homogeneous and inhomogeneous broadening occur

g (n)
$nD
Figure 11 A Doppler-broadened line-
shape function g1n2 (solid curve) is shown
as a function of n. The lineshape functions
$nH of ten of the continuum of “groups” of
atoms of homogeneous linewidth ¢nH and
differing center frequencies that contribute
to the overall lineshape function are shown
n0 n
by the dotted curves.
 568 Chapter 26 Laser Operation

within the same gain system. The forms of the lineshape functions given in the
preceding paragraphs are appropriate for those relatively common cases in
which either homogeneous broadening or Doppler broadening is dominant.
In the intermediate case, the lineshape function does not have a simple closed
form. The bandwidth of many gas gain media, like He-Ne and Ar+, for exam-
ple, are primarily due to Doppler broadening.

Example 5
Estimate the linewidth of an Ar+ gain medium. Consider a transition wave-
length of l0 = 488 nm and take the temperature of the gas under the oper-
ating conditions to be 3000 K.
Solution
The atomic mass of an argon atom is M = 6.64 * 10-26 kg. Then, from
Eq. (52),

8 * 1.38 * 10-23 * 3000 1>2

3 * 108
¢nD = a ln122b Hz
-26
6.64 * 10 13 * 10 28 2
488 * 10-9

= 3.8 * 109 Hz = 3.8 GHz

There are many other important mechanisms leading to inhomogeneous

broadening. For example, in solid-state lasers in which the active atom is doped
into a transparent host, inhomogeneities in the host lead to broadening. In
many real gain media the gain broadening is due to a complicated mix of many
physical processes and is difficult to model accurately.
Gain Saturation in Inhomogeneously Broadened Media
Let us now turn to a description of gain saturation in an inhomogeneously
broadened medium. Consider Figure 12, which shows the behavior of the
gain coefficient during the buildup to steady state in such a medium. As in
the corresponding figure for a medium that is homogeneously broadened
(Figure 10), let us consider a case in which the small-signal gain coefficient at
line center g01n02 is twice the threshold gain coefficient gth . Once again we
consider a case in which several (for the case shown, three) cavity modes have
frequencies at which the small-signal gain coefficient exceeds the threshold
gain coefficient, as shown in Figure 12a. Irradiance grows at each of these fre-
quencies. The gain coefficient in an inhomogeneously broadened medium
has contributions from groups of atoms with different center frequencies
and relatively narrow homogeneous bandwidths. Consequently, the grow-
ing irradiance only reduces the population inversion, and so the gain coef-
ficient, in atoms within, roughly, one homogeneous gain bandwidth of the
corresponding cavity-mode frequency. Thus, as shown in Figure 12b, the

g0(n) g(n)

Figure 12 Gain saturation in an inhomo-

geneously broadened laser. (a) At laser
turn on, the gain coefficient has its small- gth gth
signal value and three cavity modes are
above threshold, and so the irradiance
grows at these frequencies. (b) In steady
state, the irradiance in each of the three
cavity modes has grown large enough to re- nfsr n0 n nfsr n0 n
duce the gain coefficient g1n2 at these fre-
quencies to the threshold value gth . (a) (b)
 Laser Operation 569

laser output can consist of frequencies corresponding to several different cavity

modes in a laser using an inhomogeneously broadened gain medium. The dips
in the gain curve have a frequency width roughly equal to the homogeneous
gain bandwidth ¢nH of the atoms whose center frequency corresponds to the
cavity-mode frequency. If the cavity-mode separation is less than the homo-
geneous gain bandwidth of the atoms, the cavity modes will “compete” for
the same group of atoms and not all cavity modes initially above threshold
will lase.
The gain coefficient, near a cavity mode above threshold, in an inhomo-
geneously broadened gain medium saturates according to a law that is differ-
ent from that given in Eq. (37), which is appropriate for homogeneously
broadened media. For an inhomogeneous medium with a Gaussian lineshape
function, as is appropriate for Doppler-broadened media, the gain saturation
takes the form
g0
g = (53)
21 + I>IS

In Eq. (53), the saturation irradiance IS is that for the group of atoms with a
center frequency at the cavity-mode frequency. Note that the gain coefficient
for an inhomogeneously broadened gain medium saturates less, for a given
increase in irradiance, than does the corresponding gain coefficient given in
Eq. (37) for a homogeneously broadened medium. For example, according
to Eq. (53), when I = IS , g = g0> 22 L 0.71g0 rather than 0.5g0 as in a
homogeneously broadened medium. This slower saturation results because,
although irradiance at a frequency equal to that of a given cavity mode inter-
acts primarily with the group of atoms whose center frequency is at the cavity-
mode frequency, the irradiance also can use other groups of atoms with
“nearby” center frequencies. As the irradiance at a given cavity-mode fre-
quency grows, it “reaches” groups of atoms with center frequencies further
and further from the cavity-mode frequency. This ability to extract energy
from new groups of atoms as the irradiance grows reduces, somewhat, the
rate of saturation of the gain coefficient. Equation (53) can be used to relate
the output irradiance to the input irradiance of an inhomogeneously broad-
ened gain cell using the same procedure that led to Eq. (41). This result can
then be used together with the cavity survival factor to develop an expression
for the output irradiance from a laser system using an inhomogeneously
broadened medium. This procedure is cumbersome and the resulting expres-
sion inelegant, and so we leave this as an unlisted problem for the ambitious
student.

7 TIME-DEPENDENT PHENOMENA

In the previous sections of this chapter, we have developed a means for pre-
dicting the steady-state output irradiance from a laser. In this section, we
write down the equations that govern the time-dependent exchange of ener-
gy between the energy stored in a gain medium and that stored in the intra-
cavity field. For simplicity we will treat the ideal, homogeneously broadened,
four-level system discussed earlier. Recall that in this case the population of
the lower lasing level N1 is negligible since we take t1 L 0. Equation (32),
describing the atomic population density N2 , then becomes

dN2 sI
= Rp2 - k2N2 - N
dt hn¿ 2
We wish to develop a rate equation governing the photon number density Np
in the cavity. Let us adopt again the ring cavity model used earlier. In that
570 Chapter 26 Laser Operation

case the photon number density in the cavity can be related to the field irra-
diance I in the cavity by the relation
I = hn¿cNp

Using this in the rate equation for N2 gives,

dN2
= Rp2 - k2N2 - scNpN2 (54)
dt

The rate equation for the photon number density can be formed by adding
the contribution originating from cavity losses to the contribution originating
from interaction with the gain medium. For the case at hand, the gain contri-
bution is found by noting that each stimulated emission event decreases the
population of level 2 by 1 and increases the number of photons in the cavity
by 1. Recalling that N2 and Np are densities and that the atomic population is
spread through the volume Vg of the gain medium, whereas the photons are
distributed throughout the field in the cavity occupying an effective volume
Vc , we write

dNp
= -≠Np + 1Vg>Vc2scNpN2 (55)
dt

Here, ≠ is the cavity loss rate and the Vg>Vc is approximately the ratio of the
length of the gain cell to the perimeter P of the ring cavity. Using the method
that is valid for a linear cavity, one can show (see problem 23) that the cavity
loss rate ≠ for the ring cavity is

c
≠ = 11 - S2
P

Note that Eq. (55) predicts a zero-growth rate if there are no photons in
the cavity. This results because we have ignored spontaneous emission in the
derivation of this rate equation. Whereas, as noted earlier, spontaneous emis-
sion is typically unimportant in steady-state laser operation, it is essential in
initiating the growth of the laser field. To treat the initial growth of the
laser field, Eq. (55) can be rectified by replacing the factor Np in the last term
on the right with the factor 1NpVc + 12>Vc . This follows from a fully quan-
tum-mechanical treatment of the interaction between the atomic system and
the electromagnetic field and indicates that the ratio of stimulated emission
to spontaneous emission into a given cavity mode is the same as the ratio of
the number of photons in the cavity to 1. In practice, one often follows the
evolution of a cavity photon number density by using Eqs. (54) and (55),
starting from a nonzero “seed” photon number density. The former approach
was used in producing the curves, in Figure 13, showing the population inver-
sion Ninv = N2 and photon number density Np as functions of time after laser
turn on. Note that, for the parameters listed in the figure caption, the inver-
sion population density grows to a value which is roughly a factor of 2 larger
than its steady-state, threshold value. As a result the initial gain exceeds the
cavity losses and the photon number density begins to grow rapidly. When
the photon number density grows to near its steady-state value, however, the
rate of stimulated emission, which reduces the population inversion, exceeds
the pump rate that feeds the inversion, and so the population invers-
ion decreases rapidly. This decrease in population density, of course, is
accompanied by a rapid increase in photon number density. However, this
large photon number density initiates sufficient stimulated emission events
 Laser Operation 571

2.5 % 1016

2.0 % 1016
Number density (cm"3)

1.5 % 1016

Ninv
1.0 % 1016 Figure 13 Laser turn-on and approach to
20Np steady state. The solid curve represents the
population inversion Ninv as a function of
time. The dotted curve shows the photon
0.5 % 1016
number density Np in the cavity (multiplied
by 20 so that both curves have similar vertical
scales) as a function of time. The parameters
0 used to produce this curve are s = 10-18
0 0.2 0.4 0.6 0.8 1 1.2 1.4 cm2, ≠ = 108 s-1, k2 = 107 s-1, g0>gth = 2,
Time (ms) and Vg>Vc = 0.3.

to drive the population inversion below its threshold value, at which time
cavity loss exceeds gain and the photon number density begins to decrease.
The decrease in photon number density allows the population inversion to
once again grow and there follows a back-and-forth trading of energy be-
tween the atomic population and the electromagnetic field as the system set-
tles into steady state. The oscillations in population inversion in the approach
to steady state are known as relaxation oscillations. The strength and dura-
tion of the relaxation oscillations are sensitive functions of the parameters
governing the laser system. (See problems 24 and 25.)
Note that, as shown in Figure 13, in the approach to steady state a laser
system can produce pulses of intracavity photon number densities Np that far
exceed the steady-state photon number density. Consequently, in the
approach to steady state, the laser system produces pulses that have peak irra-
diances far larger than the steady-state irradiance. We describe other means of
producing pulsed laser output in the next section.

8 PULSED OPERATION

In this section we turn to a discussion of the means of producing laser output

pulses. The ability to control the temporal delivery of laser energy is impor-
tant in a wide variety of applications, including materials processing, charac-
terization of fast processes, and laser fusion technology. Perhaps the simplest
means of producing pulsed energy output from a laser would be to pulse the
pump—that is, turn the source of laser energy on and off. Such a pulsed-pump
or gain-switched system can produce usable pulses as suggested by Figure 13. In
general, however, there is a complex exchange of energy between the field
and atomic population in such a system that makes the pulse characteristics
difficult to control. Two important methods used to control the characteristics
of laser light pulses are Q-switching and mode-locking. These are discussed in
the following subsections.
Q-switching
You should be familiar with the quality factor Q as a measure of the loss
rate of the cavity. Cavities with higher loss rates have lower quality factors.
Q-switching refers to the periodic change of the loss rate of a cavity in
572 Chapter 26 Laser Operation

order to pulse the output of a laser. Q-switching proceeds in the following

time sequence:
1. A laser gain medium is pumped while the cavity has high loss (low Q).
The low Q of the cavity prevents growth of the intracavity irradiance and
so spontaneous emission and incoherent decay processes are the only
drains on the population of the upper lasing level. As a result, a large
population inversion grows in the gain medium. But since the cavity loss
is high, the small-signal gain coefficient is less than the threshold value
needed for lasing. That is, g0 = s1N2 - N12 6 glow th
Q
.
2. The Q of the cavity is rapidly switched to a high value so that the loss
rate of the cavity is reduced to a low value. The small-signal gain coeffi-
cient, built while the cavity had a high loss rate, now exceeds the high-Q
threshold value, g0 7 ghigh
th
Q
. As a result, a large irradiance builds rapidly
within the cavity, leading to output pulses of high peak irradiance.
3. The large irradiance in the cavity induces a high stimulated-emission
rate, which depletes the population inversion, driving the laser system
below threshold.
4. Before the population inversion can grow again to sustain a small-signal
gain coefficient larger than the threshold value, the cavity Q is switched
back to the low value. This prevents the reinitiation of lasing action and
allows the population inversion to grow, once again to a large value,
storing a large amount of energy in the gain medium.
5. Steps 2–4 are repeated.
Note that this description of the production of a Q-switched pulse has
much in common with the description of the first pulse shown in Figure 13.
There are, however, several important differences between the underlying
systems and their behaviors. In a Q-switched system, the population inversion
grows to a value larger than that in the equivalent gain-switched system be-
cause the low-Q, high-loss cavity prevents the growth of the cavity field until
the population inversion grows to the maximum value allowed by the pump-
rate/atomic-decay-rate dynamics. Further, when the cavity in a Q-switched
system is made to have low loss (high Q), the system is far above the lasing
threshold and so the irradiance grows more rapidly and to a larger value than
in the gain-switched system. Finally, in a Q-switched system, the cavity is
switched to the high-loss state after the pulse has driven the population in-
version below threshold and before the inversion can begin to grow again.
That is, switching to the low-Q state prevents additional relaxation oscilla-
tions characteristic of the approach to steady state. A Q-switched laser sys-
tem, then, produces a series of irradiance pulses similar in character to the
first pulse in a gain-switched system. The Q-switched pulses are, in general, of
higher peak irradiance and narrower pulse width than the initial pulse in a
similar gain-switched system. A Q-switched laser system is an optical ana-
logue to a capacitor. The high loss rate in the low-Q cavity allows for large en-
ergy storage in the gain medium, which can then be released rapidly as light
energy when the cavity is switched to the low-loss, high-Q state.
Q-switching can be accomplished by mechanical means, such as rotating a
cavity mirror into and out of alignment or passing a miniature fan blade
through the cavity. More commonly, Q-switching is accomplished through elec-
tro-optic or acousto-optic means. A particularly elegant means of Q-switching
involves the insertion of a saturable absorber into the cavity. A saturable ab-
sorber is a system that is absorbing for low light irradiances but transmitting for
high light irradiances. In the high-loss state, the population inversion grows,
which in turn leads to an increase in spontaneous emission. If the spontaneous
emission grows to a value large enough to saturate the absorber, the cavity will
suddenly become low loss. Thus, the saturable-absorber Q-switch requires no
 Laser Operation 573

external intervention and the cavity is naturally Q-switched when the

irradiance due to spontaneous emission reaches a value large enough to satu-
rate the absorber.
To a first approximation, Q-switching a laser system does not change
the overall efficiency of the system. As a result, the average power output Pav
from a Q-switched system is roughly the same as the CW power of the same
system. Q-switching simply redistributes the energy so that it comes out in
large bursts separated by periods of nearly zero energy output. Q-switched
systems typically have pulse widths on the order of a cavity lifetime tp = 1>≠
since the pulse must leak out of the cavity. This leads to pulse widths on the
order of 1 ms or less. The pulse repetition time (time between pulses) can be no
less than the time needed for an inversion to build, which is governed by the
lifetime t2 of the upper lasing level. The pulse repetition time in a Q-switched
system is typically on the order of 1 ms. As a result, the peak power Ppeak in a
Q-switched pulse can be a factor of 1000 or more larger than the CW output
power from the same laser system.

Example 6
A certain Nd:YAG laser is reported to have a CW power output of
PCW = 10 W. If the system is to be Q-switched with a pulse repetition rate of
1 kHz and a pulse width of ¢tp = 0.25 ms, (a) estimate the peak power in a
Q-switched pulse and (b) estimate the energy in each pulse.
Solution
To estimate the peak power, we shall model the pulses as rectangles of
width equal to the pulse width and height equal to the peak power Pp , as
shown in Figure 14. Since the pulse repetition rate is 1 kHz, the time be-
tween pulses is T = 1 ms. The average power in the Q-switched system is
the power averaged over the time T. The energy in each cycle is the energy
contained in a single pulse:

Ep = Pp ¢tp

a. The average power is thus Pav = Ep>T, so that

Pp ¢tp
Pav =
T

Finally, setting Pav = PCW and solving for the peak power gives

T 0.001
Pp = P = 1 10 W2 = 4 * 104 W
¢tp CW 0.25 * 10-6

b. Using the expression given earlier,

Ep = Pp ¢tp = 14 * 104 W210.25 * 10-6 s2 = 0.01 J = 10 mJ

P T

Figure 14 Construction showing a rectan-

Pp gular pulse approximation to Q-switched
pulses used in Example 6. Typically, the pulse
width ¢tp is a much smaller fraction of the
t pulse repetition time T than is shown in this
$tp
diagram.
 574 Chapter 26 Laser Operation

Gate Mode-Locking
Gain The physical arrangement of a mode-locked system, shown in Figure 15,
is similar to that for a Q-switched system. Both systems require the inser-
Mirror Pulse Mirror tion into the laser cavity of an element that performs loss modulation.
Figure 15 A mode-locked laser system. However, as noted, in a Q-switched system the cavity is left in the high-loss
The gate opens for a brief time during each state for a time long enough to allow the population inversion to build to a
round-trip to let the pulse pass. large value. This process requires many cavity round-trips. In contrast, in a
mode-locked system the loss modulation occurs once each round-trip in
the following fashion. In a linear cavity, the loss modulation “gate” is placed
near one of the mirrors and is switched to the low-loss state for a brief pe-
riod once each round-trip. Here, “a brief period” means a fraction of a
round-trip time. As a result, only pulses timed to pass through the gate
when it is open can survive in the cavity. All other field shapes see a high
loss and so do not lase. In a mode-locked system, the loss modulation oc-
curs too rapidly for the population inversion in the gain medium to re-
spond and so the population inversion, more or less, retains the same value
that it would have in a CW system. The mode-locking “gate” can be any of
the devices, described earlier, that could be used as Q-switches, provided
that the gate can be opened and closed over a time period shorter than a
cavity round-trip time.
Inhomogeneously broadened laser systems in which many different cavi-
ty modes lase are commonly mode-locked. The total field in the cavity is a su-
perposition of these fields with frequencies corresponding to the cavity-mode
frequencies that are separated by the cavity free spectral range c/2d, where d
is the cavity length. In order that these modes add to a pulse, they must be
locked to a common phase at a particular place that moves back and forth
through the cavity at the speed of light. The opening and closing of the gate,
just described, provides such a phase-locking mechanism. The initiation of
such a pulse can be viewed as the fortuitous spontaneous emission into the
different cavity modes in such a way that the fields in these modes happen to
constructively interfere at the gate when the gate is open. Once formed, this
pulse passes through the gate, bounces off the mirror, and passes through the
gate again before the gate closes. The pulse is then amplified by the gain
medium and returns as a larger pulse just as the gate opens again. In this
way the pulse builds from phase-matched spontaneous emission events. Field
components not phase-locked to form constructive interference, at the posi-
tion of the gate when the gate is opened, are blocked by the closed gate. The
mode-locked pulse is the low-loss mode of a cavity that is loss modulated once
each round-trip.
Phase-locking the fields in the cavity allows for the formation of a nar-
row pulse centered on the positions of common phase. Between positions of
common phase, since the fields have different frequencies, the fields tend to
destructively interfere. At a given place in the cavity, the condition for a
common phase for all of the fields recurs at times equal to the inverse of the
constant frequency separation of the modes. (See problem 26.) As noted
earlier, the frequency difference between modes is nfsr = c>2d so that the
recurrence time for the pulses is the round-trip cavity time 2d/c.
Let us compare the peak power in a mode-locked, multimode laser to
the average power in the same multimode laser system with CW output. Let
the number of nearly equal-amplitude modes each of power P0 in the system
be N. In CW operation the phases of these modes are independent from each
other and wander randomly since the fields are not perfectly coherent. Thus
the total average power, for the CW laser, is simply the sum of the powers in
the individual fields. That is,

PCW = NP0 (56)

 Laser Operation 575

In the mode-locked pulse, at the positions of equal phase, the fields constructively
interfere so that the power Pp in the total field at these positions is given by

Pp = N 2P0 (57)

Now the average power in the mode-locked laser is roughly the same as the
average power in the same CW system. Thus the relation given in part (a) of
Example 6, derived there for a Q-switched system, also applies for the
mode-locked case. That is, with Pav = PCW ,

Pp ¢tp
PCW =
T

Using Eqs. (56) and (57), together with the relation for the pulse repetition
time T = 2d>c, leads to an approximate expression for the temporal width
¢tp of a mode-locked pulse,

PCWT NP012d>c2 1 1
¢tp = = 2
= =
Pp N P0 N1c>2d2 Nnfsr

The number of modes that are above threshold depends on the bandwidth
¢n of the gain medium and on the ratio of the small-signal gain coefficient to
the threshold gain coefficient. In an inhomogeneously broadened system
pumped so that g0 = 2gth , all modes within ¢n>2 of linecenter will be above
threshold (see, for example, Figure 12a). In that case, Nnfsr = ¢n and the
expression for the pulse width becomes

1
¢tp L (58)
¢n

This important relation is an example of the more general bandwidth theo-

rem, which states that the width of a pulse is inversely proportional to the
range of frequencies in the fields that constitute the pulse. So gain media
with larger bandwidths can be used to form narrower mode-locked pulses.
The relationships given in Eqs. (57) and (58) are explored in problems 27
through 31.

9 SOME IMPORTANT
LASER SYSTEMS

Detailed descriptions of the characteristics of different laser systems are

beyond the scope of this chapter. In this section, we describe briefly the prop-
erties of the different types of atomic and molecular gain media and list a few
examples of each. In Section 10 we discuss, in a bit more detail, the operation
of diode lasers.
Gas Atomic Lasers
By an atomic laser we mean one in which the upper and lower lasing levels are
different electronic states of one of the atoms in the gain medium. Helium-neon
(He-Ne) and argon-ion 1Ar+2 lasers are two important gas atomic lasers. In the
He-Ne laser neon atoms are the lasing species and the helium atoms aid in the
pumping process. The lasing levels in the Ar+ laser are two different electronic
576 Chapter 26 Laser Operation

energy states of a singly ionized argon atom. Gas atomic lasers are typically
Doppler broadened and have low efficiencies (less than 1%). The two types
mentioned here are known for their coherence. Gas atomic lasers typically emit
laser light with frequencies in the visible to near-infrared range. These lasers are
usually pumped by an electric discharge.
Gas Molecular Lasers
Molecules have energy levels corresponding to different rotational and vibra-
tional states of the molecule, in addition to the states associated with different
electronic configurations. The upper and lower laser levels of gas-molecular
gain media are typically different vibrational-rotational states associated
with the ground electronic state. These vibrational-rotational states typically
differ in energies such that the emitted light is in the mid-infrared range. The
most important gas-molecular laser is the carbon dioxide 1CO22 laser that
emits radiation of wavelength 10.6 mm. The CO2 laser is one of the most
versatile and efficient laser systems, capable of CW power outputs ranging
from mW to MW with efficiencies ranging from 10% to 40%. CO2 lasers are
typically pumped by an electric discharge. High-power CO2 lasers are some-
times pumped by electron beams or gas dynamic processes.
Excimer Lasers
The gain medium in an excimer laser is a rare-gas halogen mixture. Rare-gas
atoms have filled outer shells and so tend not to bond to other atoms. How-
ever, an excited rare-gas atom (like Kr*) has a single electron in its outer
shell and so can form an ionic bond with a halogen atom (like F), which is one
electron short of a filled outer shell. A typical reaction might be
Kr… + F2 : 1Kr+ F-2… + F
The production of the excited rare-gas species is accomplished by electric-
discharge or e-beam pumping. The molecule 1Kr+F-2… is the upper state of
the lasing transition. If the Kr is not excited, it repels the F atom and so the
lower lasing level is unstable with an extremely short lifetime 1" 10-13 s2.
Thus this system approximates an ideal four-level laser system. Excimer
lasers are typically pulsed systems distinguished by their relatively high aver-
age power 1 ' 50 W2 and, importantly, lasing wavelengths in the UV.
Liquid-Dye Lasers
The gain media in liquid-dye lasers are solutions of certain long-chain or-
ganic molecules in alcohol or other solvents. Dye lasers have very large
small-signal gain coefficients (4/cm or more) and very large gain bandwidths
(50–100 nm). Dye lasers emit radiation in the visible region of the spectrum,
and a few different dyes are sufficient to provide coherent radiation tunable
across the entire visible spectrum. The broad gain bandwidth of a given dye
allows for the production of extremely short (ps-width) mode-locked pulses.
Solid-State Lasers
Solid-state lasers are a class of lasers in which the lasing atomic species is
doped into a transparent host material. The most common solid-state laser
is the Nd:YAG laser in which neodymium (Nd) atoms replace about 1% of
the yttrium (Y) atoms in the host crystal, yttrium aluminum garnet (YAG).
Solid-state lasers typically have outputs with wavelengths in the near infra-
red. The Nd:YAG system provides laser output at 1.064 mm. An Nd:YAG
system can provide a high power, good beam quality CW output, and can
be either mode-locked or Q-switched. The output of an Nd:YAG laser is
commonly frequency doubled to produce coherent radiation at 532 nm.
Solid-state lasers are typically flash-lamp pulsed or optically pumped by
 Laser Operation 577

another laser. Efficient, portable Nd:YAG lasers, pumped by an array of

diode lasers, are available.

10 DIODE LASERS

Semiconductor or diode lasers have gain media that differ significantly from
the atomic and molecular gain media discussed in the preceding sections of
this chapter. A diode laser is essentially a p-n junction whose cleaved edges
act as reflecting surfaces that supply the cavity feedback. From a packaging
perspective, the main distinguishing feature of the diode laser is its size. The
photograph in Figure 16 allows one to appreciate the difference in scale
between diode lasers and conventional table-top laser systems. In this
section we provide a qualitative description of the operating principles of
diode lasers and begin with a very brief review of semiconductors and p-n
junctions.
Each of the electrons in a solid insulator is tightly bound to a given
atom in the lattice that forms the solid. These tightly bound electrons are
Figure 16 Two views of a packaged laser
said to occupy states in the valence band of electronic energy states of the diode. The packaged diode is about 6 mm in
material. In a conductor, on the other hand, one or more electrons per atom diameter. The laser diode chip itself is locat-
in the solid are not tightly bound to a given atom but rather are relatively ed in the small rectangular depression visi-
free to roam about the solid as a whole. These electrons are said to be in the ble in the mount shown in the top image.
conduction band of electron energy states in the solid. In an intrinsic (pure) Courtesy of John Sohl. (Photo by Sheri Tr-
bovich.)
semiconductor, each electron in the solid is bound to a given atom (that is,
it is in the valence band). However, the outermost electrons in a semicon-
ductor are not as tightly bound as those in an insulator. The strength of the
binding of the outermost valence electrons is described by the band-gap
energy, which is the energy required to free an electron from its host nucle-
us. That is, the band-gap energy is the energy required to promote an elec-
tron from the valence band to the conduction band. In an insulator, the
band-gap energy is typically about 4 eV. In a semiconductor, the band-gap
energy is between 1 and 2 eV. (In a conductor, the conduction and valence
bands overlap.)
In an intrinsic semiconductor, the atoms bond so as to “fill” the outer
shells of the atoms. For example, in gallium arsenide (GaAs), gallium has
three outer-shell electrons and arsenic has five outer-shell electrons. These
atoms bond in the solid lattice in such a way that each atom sees a full outer
shell (eight electrons) with some of these electrons shared with neighboring
atoms. It is this sharing of the electrons that forms the crystal bond. In a p-type
semiconductor, the material is doped with a small amount of an impurity that
has one less electron in its outer shell than does the atom that it replaces in the
lattice. For example, zinc has two outer-shell electrons. In GaAs doped with
zinc, the zinc replaces gallium (three outer-shell electrons) in the lattice. As a
result, for every zinc atom added, there are empty spaces in the lattice re- Electrical
served for an electron. These empty spaces are called holes and act as positive Metal lead
contact
charge carriers. In an n-type semiconductor, the material is doped with an im-
Cleaved
purity whose outer shell has one electron more than the outer shell of the end e
atom that it replaces in the lattice. For example, if selenium (six outer-shell p-typ 1 mm
p e
electrons) is doped into GaAs, it replaces As (five outer-shell electrons) in n-ty
the lattice. This results in one electron per selenium atom that does not have
a slot in the charge clouds surrounding the nuclei of the atoms in the lattice. Heat sink
Junction
These “extra” electrons act as negative charge carriers. Of course, both n-type and
Output
and p-type materials have zero net charge since the dopant atoms are neutral. electrical
contact
Laser diodes use forward-biased p-n junctions, in which the voltage source
drives the holes in the p-type material and the conduction electrons in the Figure 17 Simple p-n junction, laser diode
n-type material toward the junction (Figure 17). pumped by an injection current.
 578 Chapter 26 Laser Operation

E Figure 18 shows the band structure in the junction region. Note that
pc the bands are curved and contain states of differing energy E since the carri-
ers in the bands can have different momenta pc . A population inversion is
shown near the center of the bands where there is an excess of electrons in
the conduction band. In the junction region the electrons can “fall” into the
available holes, releasing energy in the form of a photon as they do so. This
spontaneous emission process is responsible for the output of light-emitting
(a) Spontaneous emission
diodes and is illustrated in Figure 18a. The cleaved edges of these p-n
E
junctions, of submillimeter dimension, act as mirrors, providing feedback for
the laser system. The spontaneously emitted photons that are reflected back
pc
into the thin junction region can cause stimulated absorption (the creation of
a conduction electron/valence hole pair and the loss of a photon, shown in
Figure 18b) or stimulated emission (the creation of a twin photon and the
demotion of an electron from the conduction band into a hole in the valence
band, shown in Figure 18c).
(b) Stimulated absorption
Diode Laser Operating Characteristics
E Important favorable characteristics of diode lasers are that they are relatively
pc
inexpensive, are small and efficient, and can be engineered to have a variety of
wavelengths of interest in many applications. In addition, the output irradiance
from laser diodes can be easily modulated by varying the injection current that
pumps the diode. Thus, it is relatively easy to encode information into laser-
diode light fields. Efficient diodes have a more complicated layered structure
than does the simple p-n junction device illustrated in Figure 17. A more
(c) Stimulated emission typical stripe-heterojunction device is shown in Figure 19. Arrays of diode
Figure 18 The interaction of an electro- lasers allow for relatively high average power devices. Diode lasers do have
magnetic field with a semiconductor gain several unfavorable characteristics. The small asymmetric output aperture
medium. The upper energy band is the con- leads to highly divergent nonsymmetrical output beams, as indicated in
duction band and the lower energy band is
Figure 17. As a result, the output of a diode laser is typically coupled directly
the valence band. Filled circles indicate
electrons and empty circles indicate holes. into a fiber or collimated by a short focal length lens. Further, it is difficult to
The thick, straight arrows indicate transi- limit a diode laser to stable single-mode output, and so diode devices gener-
tions of an electron from one band to anoth- ally have shorter coherent lengths than, for example, do Nd:YAG, He-Ne, or
er, and the lighter, curved arrows represent Ar+ laser systems. Still, the advantages of the diode laser design make them
photons. The energy E of the charge carriers
the preferred choice in an ever-increasing array of devices, including laser
varies across the band since the carriers in a
given band vary in momentum, pc . pointers, optical pumps, and optical data storage and read-out.
Semiconductor materials can be engineered to have a variety of band-
gap energies resulting in a variety of operating wavelengths. Some of these
are listed in Table 1 below.

100 mm
Metal electrode
p& GaAs
10 mm p GaAlAs
1 mm
GaAlAs active region
n GaAlAs
n GaAs substrate

Emitting
Figure 19 Stripe-heterojunction semicon- area
m
ductor laser. Note the dimensions of the 0m
20
cavity region.

TABLE 1 LASER DIODE WAVELENGTHS

Material GaN AlGaInP GaAlAs InGaAsP Sb mixtures

Wavelength (nm) 400–480 630–690 750–900 1200–2000 2000–4000

 Laser Operation 579

PROBLEMS

1 The spectral energy density r1n2 in an electromagnetic a. Write down the rate equations for the population densi-
field in thermal equilibrium at temperature T is given by ties of the levels.
Eq. (6). Recall that r1n2 is the energy per unit volume per b. Find and plot the steady-state small-signal population
unit frequency interval in the field. inversion N2 - N1 as a function of the pump irradiance.
a. Show that r1l2, which we define to be the energy per (Recall that “small signal” is code for “set I = 0.”)
volume per unit wavelength interval in the field, is c. Find the pump irradiance Ip required to sustain a steady-
state population inversion.
8ph 1 d. Find the pump irradiance Ip required to sustain a small-
r1l2 = 5
a hc>lkBT
b signal gain coefficient of 0.01/cm.
l e - 1
e. Find the pump irradiance Ip required to sustain a small-
b. Confirm that the relation found in (a) is in agreement signal gain coefficient of 1/cm.
with the spectral exitance associated with a blackbody f. Compare N0 to N1 , N2 , and N3 for the pump irradiances
given as of parts (d) and (e). Is it reasonable to set N0 L NT for
either of these irradiances?
2phc2 1 g. Use the ideal four-level gain medium relation given as
Ml = 5
a hc>lkBT
b
l e - 1 Eq. (38) together with the definition of the effective
Note that Ml is the the exitance per wavelength interval pump rate density given following Eq. (33) to estimate
emitted by a blackbody source. (Hint: The spectral ener- the pump irradiance required to sustain a small-signal
gy density in (a) includes the energy in field components gain coefficient of 0.01/cm and 1/cm. Compare these re-
moving in all directions while the spectral exitance ac- sults to those obtained in parts (d) and (e).
counts for the power moving normally away from the
6 Show that Eq. (34) follows from Eqs. (32) and (33).
blackbody surface.)
7 Consider an amplifying medium composed of homoge-
2 Consider a monochromatic electromagnetic field traveling neously broadened three-level atoms. Amplification is to
with speed c in a given direction. Use a conservation of en- occur on the 2-to-1 transition. The medium is pumped by a
ergy argument to show that the time-averaged energy den- laser, of irradiance Ip , that is resonant with the 3-to-1 transi-
sity 8u9 associated with this field is related to the irradiance tion. Level 3 decays spontaneously only to level 2 and level
I of the field by 8u9 = I>c. 2 decays spontaneously to level 1, which is the ground state
of the system. The total number density of gain atoms is
3 Show that Eqs. (17), (18), and (19) follow, in steady state,
NT = N1 + N2 + N3 . The various parameters are
from Eq. (16) and Eq. (14).
4 One can define a saturation irradiance IS,abs for an absorp- k32 = 108>s, k21 = 103>s
sp = s31 = 3 # 10-19 cm2,
tive medium as the irradiance for which the loss coefficient
s21 = 10-18 cm2
a is reduced by a factor of 2 from its small-signal value.
a. Show that for the two-level absorptive medium consid- l31 = 400 nm, l21 = 600 nm 1in free space2
ered in Section 2, NT = 1.5 # 1026>m3
hn¿ a. Sketch a level diagram like Figure 5 appropriate for this
IS,abs =
2st2 case and indicate the various stimulated and decay
processes with arrows on the level diagram.
where t2 = 1>A 21 . b. Write down the rate equations for the population densi-
b. Compare this relation to the saturation irradiance for ties of the levels. Include the presence of a field of irra-
the ideal four-level gain medium given in Eq. (39) and diance I resonant with the 2-to-1 transition, the pump
account, with a conceptual argument, for the factor of interaction, and the decay processes.
two difference between the two saturation irradiances. c. Find and plot the steady-state small-signal population
inversion N2 - N1 as a function of pump irradiance.
5 Consider an amplifying medium composed of homoge- (Recall that “small signal” is code for “set I = 0.”)
neously broadened four-level atoms like the one depicted in d. Find the pump irradiance Ip required to sustain a steady-
Figure 5. Amplification is to occur on the 2-to-1 transition. state population inversion. Compare this result to the
The medium is pumped by a laser of intensity Ip , which is answer obtained for part (c) of problem 5.
resonant with the 3-to-0 transition. The spontaneous decay e. Find the pump irradiance Ip required to sustain a small-
processes are as indicated on the diagram. The total number signal gain coefficient of 0.01/cm. Compare this result to
density of gain atoms is NT = N0 + N1 + N2 + N3 . The the answer obtained for part (d) of problem 5.
various parameters are f. Find the pump irradiance Ip required to sustain a small-
k32 = 108>s, k21 = 1000>s, k10 = 108>s, k30 = k31 = k20 L 0 signal gain coefficient of 1/cm. Compare this result to
the answer obtained for part (e) of problem 5.
sp = 3 # 10-19 cm2, s = 10-18cm2 g. Summarize the important differences in the behavior
l30 = 400 nm, l21 = 600 nm 1in free space2 of the three-level gain medium considered in this prob-
lem and the four-level gain medium considered in
NT = 1.5 # 1026>m3 problem 5.
 580 Chapter 26 Laser Operation

8 Show that if the irradiance throughout a gain cell described c. Explain why, even when every pump event leads to an
by Eq. (41) is much less than the saturation irradiance IS , output photon, the efficiency of the laser system is less
the output irradiance IL is related to the input irradiance I0 than 100%.
by the simple relation
16 In this problem and the following two problems, consider a
IL = I0e g0L ring cavity like the one depicted in Figure 8. Let the cavity
mirrors M1 and M2 have reflectances R1 = R2 and let mir-
That is, show that, in the small-signal regime, the irradiance ror M3 have reflectance R3 = 1 - T3 - A 3 , where A 3 char-
exhibits exponential growth. acterizes the output mirror absorption. Let the gain
medium be homogeneously broadened and have length
9 Show that if the irradiance throughout a gain cell described L = 10 cm and a saturation irradiance (at the lasing fre-
by Eq. (41) is much greater than the saturation irradiance quency) of IS = 2000 W>cm2.
IS , the output irradiance IL is related to the input irradiance
I0 by the simple relation a. Find the threshold gain coefficient if R1 = R2 = 1,
R3 = 0.95, and T3 = 0.05.
IL = I0 + ISg0L b. If the small-signal gain coefficient is twice the threshold
value, find the irradiance of the laser output field.
That is, show that for a very large input irradiance, the irra-
17 Consider again the ring laser described in problem 16 but
diance exhibits linear growth. [It may be somewhat simpler
now take the small-signal gain coefficient to be 0.01/cm,
to implement the relation I W IS in Eq. (40) and then in-
R3 = 0.95, and T3 = 0.05. Plot the laser output irradiance
tegrate, than to use Eq. (41) directly.]
as a function of the variable reflectance R = R1 = R2 of the
10 Consider the limit described in problem 9. other two cavity mirrors.
a. Show that in this limit and for an ideal four-level gain 18 Consider again the ring laser described in problem 16. Let
medium, R1 = R2 = 0.99 and A 3 = 0.01. Let the small-signal gain
coefficient be 0.01/cm.
IL - I0 = hn¿R2pL
a. Plot the laser output irradiance as a function of the vari-
b. Argue that the relation in part (a) implies that for the able transmittance T3 of the output mirror.
large input-irradiance case of problem 9 every pump b. Using the plot produced in part (a), determine the value
event leads to one photon added to the electromagnetic of T3 that maximizes the laser output irradiance. Explain
field being amplified. why, for this system in which there are unavoidable losses,
c. For the small input-irradiance of problem 8, even for an the output irradiance is reduced from its maximum
ideal four-level gain medium, it is not true that every value if T3 is either too large or too small.
pump event leads to one photon added to the electro-
magnetic field being amplified. Conceptually, account 19 Derive Eq. (47) by a procedure similar to that leading to
for the missing pump events. Eq. (43). The linear cavity case is complicated by the fact
that the field encounters the gain medium twice in each
11 A homogeneously broadened gain medium has a length of round-trip with the losses encountered at the mirrors inter-
L = 2 cm, a small-signal gain coefficient at the transition spersed between passes through the gain medium. It may be
linecenter of g01n02 = 1>cm, and a saturation irradiance at useful to research and then summarize the solution to this
the transition linecenter of IS1n02 = 100 W>cm2. Assume problem.
that light of frequency n¿ = n0 is input into the cell. Find the
20 Show that Eq. (47) for a linear cavity reduces to
irradiance IL exiting the gain cell when the irradiance I0
input to the cell is (a) 1 W>cm2, (b) 10 W>cm2, (c)
100 W>cm2, (d) 1000 W>cm2, and (e) 10,000 W>cm2. T2IS g012L2 - ln11>S2
Iout = a b
2 1 - S
12 For each case of problem 11, find the irradiance added by
passage through the gain cell IL - I0 and describe how this for a cavity with R1 = 1. Here, S is the survival fraction in
added irradiance changes with increasing input irradiance. the linear cavity without gain 1S = R22. Compare this
13 Repeat problems 11 and 12 for the case in which the field result with the similar result given in Eq. (43) for the ring
input into the cell has a frequency n¿ = n0 + ¢n>2, where cavity and account for the differences between the two
¢n is the homogeneous linewidth of the gain medium. results.
14 Reproduce the curves shown in Figure 7 but extend the 21 Consider the CO2 transition described in Example 4. In addi-
maximum length of the gain cell on the plot to 10 cm. tion to the information given in the example note that the
spontaneous emission rate for the transition is A 21 = 0.34>s.
15 Consider an ideal four-level gain medium in a ring cavity
like the one of Figure 8 but with R1 = R2 = 1 and R3 = a. What is the stimulated emission cross section s for this
1 - T3 . transition?
b. What must be the population inversion in the gain medi-
a. Show that, for this case, um to produce a small-signal gain coefficient (at linecen-
ter, n¿ = n0) of 0.03/cm?
Iout = Isat 1g0 - gth2L
c. Treating this system as an ideal four-level system, esti-
b. Show that, for this case and for g0 W gth , essentially mate the saturation irradiance for this transition.
every pump event leads to an output photon.
 Laser Operation 581

22 Find the Doppler-broadened gain bandwidth of the 633-nm 30 Use the relation in problem 29 to verify Eqs. (57) and (58).
He-Ne transition. Assume that the operating temperature is
31 Estimate the peak power and pulse repetition rate in a
400 K and recall that neon is the lasing species.
mode-locked Nd:YAG laser pulse of pulse width 70 ps if the
23 Show that the loss rate ≠ for a ring cavity with round-trip Nd:YAG laser cavity is 1.5 m long, and the CW output
survival factor S and perimeter P is power of the Nd:YAG laser system is 10 W.
c 32 Estimate the diffraction-limited far-field divergence angles
≠ = 11 - S2 of a beam output from the heterojunction laser diode illus-
P
trated in Figure 19.
24 Reproduce the curves shown in Figure 13 using the parame- 33 What is the band-gap energy of an AlGaAs semiconductor
ters given in the figure caption. Note that the effective used in a laser diode device that emits light of wavelength
pump rate can be found from the listed condition g0 = 2gth . 800 nm?
25 Produce curves like those shown in Figure 13 for the parame- 34 What must the reflectance of the cleaved ends of the laser
ters given in the figure caption except let (a) k = 10-8 s-1, (b) diode illustrated in Figure 19 be if the small-signal gain
k = 10-6 s-1, (c) g0>gth = 1.1, and (d) g0>gth = 4. In each coefficient of the medium is 40/cm?
case describe how changing the indicated parameter
changes the curves. 35 a. Show that solving Eqs. (54) and (55) for the steady-state
photon number density Np and population inversion N2
26 Consider the electromagnetic fields given by gives,
E1 = E0 cos12pn1t - k1z2
sc1Vg>Vc2Rp2 - k2≠
E2 = E0 cos[2p1n1 + dn2t - k2z] Np =
sc≠
a. Show that at z = 0, E1 + E2 = 2E0 at times given by
t = n>dn, where n is an integer. ≠ Vc
N2 =
b. Discuss the relevance of the result shown in (a) for a sc Vg
mode-locked laser.
b. Use the result for Np in (a) to form the following expres-
27 The gain bandwidth (in nm) and the transition wavelength sion for the steady-state output irradiance from a ring
for three different laser systems are given below. Estimate laser like the one discussed in connection with Figure 8:
the pulse width attainable with these laser systems if they
are mode-locked. k2hn¿ 1sRp2L>k22 - 11 - S2
Iout = T3
Ar+ l = 488 nm ¢l = 0.004 nm s 1 - S
He-Ne l = 633 nm ¢l = 0.002 nm
c. Show that the relation from part (b) agrees with Eq. (43)
Rhodamine 6G dye l = 590 nm ¢l = 80 nm
only if the survival fraction is close to 1. (Hint:
28 In order to investigate the bandwidth theorem, plot the given ln11 - x2 L x, for small x.)
function F as a function of time t for (a) N = 5, (b) N = 10, d. Which relation, the one from part (b) or the one given in
and (c) N = 50. In each case estimate the pulse width from Eq. (43), is correct when S is not close to 1? Explain.
the plot and compare the pulse width to the range of fre-
quencies in the superposition.

F = a cos[2p1110 + 0.2j2t>s]
N

j=0

29 Show that the sum E of the electric fields associated with N

mode-locked cavity modes of equal amplitude and with fre-
quencies nj = n0 + jnfsr can be written as

a
1N - 12>2
E = E0 cos12pnjt + w02
j = -1N - 12>2

sin1Npnfsrt2
= E0 cos12pn0t + w02
sin1pnfsrt2