6

Laser Pump and Threshold

6.1 Quantum Theory of Absorption and Emission
Quantum mechanics allows the energy of bound atomic states to take only certain
values. Thus the allowed energy values of an atom might look like those shown in
Fig. 6.1. These energies are referred to as the energy levels of an atom. Because
E
3
E
2
E
1
E
0 Ground level
First excited level
Second excited level
Third excited level
:
:
:
:
FIGURE 6.1
Energy levels of an atom.
of this discrete nature of atomic energy, an atom can absorb or emit energy only in
discrete amounts. An atom, initially in level E
i
, can make a transition to a higher
energy level E
j
> E
i
by absorbing a quantum (photon) of energy h
ji
= E
j
E
i
.
Similarly, it can make a transition to a lower energy level E
j
< E
i
by emitting a
photon of energy h
ij
= E
i
E
j
1
. Thus the exchange of energy between the eld
and atom occurs either by absorption or emission of photons. Figure 6.2 shows
absorption and emission processes for atomic transition between levels E
1
and E
2
.
1
Assuming that the transitions are permitted by the conservation laws.
161
162 Laser Physics
(a) Absorption (b) Emission
h
21
E
1
E
2
h
21
E
3
FIGURE 6.2
Absorption and emission of light by an atom.
When we shine light on the atom, when will it absorb a photon? Quantum
Mechanics does not tell us precisely when this will happen, but it tells us that the
probability of absorbing a photon during a short time interval t is proportional to
the ux density of photons
2
with the right amount of energy, incident on the atom.
Thus the probability of transition from level E
1
to E
2
is given by
transition probability = constant
12
t , (6.1)
where
12
is the ux density of photons of energy E
2
E
1
. Photon ux density is
related to eld intensity by
=
I

h
=
1
2

0
nc|E|
2
h
. (6.2)
Equation (6.1) is known as Fermis golden rule. Then if there are initially N
1
atoms per unit volume in level E
1
, the fraction of atoms excited to level E
2
by the
absorption of photons in the time t will be
_
|N
1
|
N
1
_
12
= constant
12
t ,
where N
1
is the change in the number of atoms in level N
1
. The number of atoms
excited in time t is then N
1
= |N
1
|. In the limit t 0, we obtain the rate of
atomic transitions
_
|N
1
|
t
_
12

dN
1
dt
_
12
= constant
12
N
1
. (6.3)
From dimensional considerations it is easily seen that the constant of proportionality
has the dimensions of a cross-section (area). It is usually denoted by the Greek letter
and can be calculated from quantum mechanics. The values of can be measured
2
Flux density number of photons per unit area per unit time (#/m
2
-s). For narrowband elds
it is related to the intensity (power density) by

= I

/h
Laser Pump and Threshold 163
and are often compiled in tables. Let us denote the cross-section for the transition
from E
1
to E
2
by
12
. Then the rate at which the population of level E
1
is changing
due to the transition of atoms from level E
1
to level E
2
[Eq.(6.3)] can be written in
the form
_
dN
1
dt
_
12
=
12

12
N
1
. (6.4)
The minus sign on the right hand side of this equation indicates that the population
N
1
of level E
1
is decreasing due to absorption. The subscript 12 labels terms which
refer to the transition E
1
E
2
. Thus we have
_
dN
1
dt
_
12
= rate at which atoms make transition from E
1
to E
2

12
= cross-section for the transition E
1
to E
2

12
= number ux density of photons with energy h = E
2
E
1
What about the population N
2
of the excited state? We can see that the rate at
which atoms leave the lower level E
1
is equal to the rate at which atoms arrive in
the excited state E
2
_
dN
2
dt
_
12
= +
12

12
N
1
. (6.5)
where the + sign indicates that the excited state population N
2
is increasing due
to transitions from level E
1
to E
2
.
All that has been said about the transition E
1
to E
2
also applies to other ab-
sorptive transitions. For example, the rates for E
1
to E
3
transition are written
as
_
dN
1
dt
_
13
=
13

13
N
1
_
dN
3
dt
_
13
= +
13

13
N
1
Once excited to a certain level, an atom will eventually decay to a lower level
emitting a photon in the process. This is, of course, known as emission. Suppose
N
2
atoms are excited into level E
2
. Then the rate at which atoms decay from this
level to level E
1
will be
_
dN
2
dt
_
21
=
21

21
N
2

rad
21
N
2
(6.6)
Quantum mechanics tells us that the cross-sections for E
1
to E
2
and E
2
to E
1
transitions are, in fact, equal
21
=
12
and, since the emitted and absorbed photons
have the same frequency, the ux densities are equal
21
=
12
.
The rst term
21

21
N
2
on the right hand side of Eq. (6.6) is similar to the one
in Eq.(6.3) with the ones and twos interchanged. This term represents the rate
at which the population of E
2
decreases due to atomic transitions to level E
1
by the
emission of photons. It tells us that like absorption, emission is enhanced by the
164 Laser Physics
E
2
E
1

12

12
N
1

21

21
N
2

21
N
2
FIGURE 6.3
Various rates leading to transfer of populations between levels 1 and 2, of a two-level
system. Transitions to other levels are not shown.
presence of photons of energy E
2
E
1
. This is known as stimulated emission.
Photons emitted in stimulated emission have the same direction, frequency and
polarization as the incident photon ux.
The second term
rad
21
N
2
on the right hand side of Eq. (6.6) corresponds to the
process known as spontaneous emission. It allows an atom to decay to lower
states by emitting a photon even when no photons of energy E
2
E
1
are present.
Unlike stimulated emission, photons in spontaneous emission can be emitted in any
direction. By setting
21
= 0 in Eq. (6.6), we can solve the dierential equation,
and obtain N
2
= N
2
(0)e

rad
21
t
. This means the population of level E
2
decays
exponentially due to spontaneous transitions to level E
1
. The lifetime
rad
21
of the
state due to E
2
to E
1
transitions is dened as the time it takes N
2
to decay to
1/e 0.37 of its initial value N
2
(0). This gives a relation between parameter
rad
21
and radiative lifetime of the state

rad
21
=
1

rad
21
.
Equations like (6.4)-(6.6) hold for any optical transition that involves absorption
or emission of photons. Nonradiative processes, such as collision, can also transfer
populations from one level to another. Corresponding rate equations are of the form
_
dN
dt
_
nr
=
nr
N (6.7)
where
nr
is the fractional decay rate for population transfer due to nonradiative
processes and
nr
= 1
nr
is the corresponding life time. The decay rate (independent
of photon ux density) from level E
2
to E
1
is then the sum
21
=
rad
21
+
nr
21
. In what
follows we will sometime assume that
21

rad
21
and will not distinguish between
the two. All of the population changing rates between levels E
1
and E
2
are shown
in Figure 6.3.
The transitions from level E
2
to E
1
will change the population of level E
1
just as
the transitions from E
1
to E
2
changed the population of E
1
. Since each atom that
Laser Pump and Threshold 165
leaves E
2
by emitting a photon of energy E
2
E
1
ends up in level E
1
, the rate of
change of the population of level E
1
will be given by
_
dN
1
dt
_
21
=
_
dN
2
dt
_
12
=
12

12
N
2
+
21
N
2
,
21
=
12
,
12
=
21
. (6.8)
If we focus our attention on the transitions between two levels E
1
and E
2
due to
absorption or emission of photons, the total rate of change of populations in levels
E
1
and E
2
can be written as
dN
1
dt
=
_
dN
1
dt
_
12
+
_
dN
1
dt
_
21
=
12
(N
2
N
1
)
12
+
21
N
2
, (6.9a)
dN
2
dt
=
_
dN
2
dt
_
21
+
_
dN
2
dt
_
12
=
12
(N
2
N
1
)
12

21
N
2
. (6.9b)
These are the basic equations governing the evolution of population of two levels
connected by absorption and emission of radiation. When transitions to other levels
are included these equations must be modied by adding terms corresponding to
absorption and emission of appropriate frequency photons. We will see examples of
this in later sections. It now remains to relate these changes to the change photon
ux density.
6.2 Photon Equations
Now if the population N
1
(number per unit volume) increases by N
1
in time t,
the same number of photons have been created per unit volume. The change in
photon ux density
12
= I/h is then

12
=
c
n
N
1
=
c
n
[
12

12
(N
2
N
1
) +
21
N
2
] t ,
where we have used Eq. (6.9a) to substitute for N
1
. Taking the limit t 0 then
gives
d
12
dt
=
c
n
[
12

12
N +
21
N
2
] . (6.10)
Let us compare this equation with the energy balance equation for absorption from
an em eld. The average power absorbed by the atoms (the rate at which the eld
does work on the atoms) per unit volume is
dw
dt
= E
P
t
=
1
2
Re
_
E
P

t
_
=
1
2
E (iP

) =
1
2

at
|E|
2
.
Since this power comes at the expense of eld power, the rate at which eld energy
density changes is given by
du
dt
=
1
2

at
|E|
2
166 Laser Physics
Now photon ux density is related to the eld energy density u by
= photon density speed =
energy density
energy of one photon
speed
=
u
h

c
n
=
1
2

0
n
2
|E|
2
h

c
n
=
1
2

0
cn|E|
2
h
Using this relation we can write the power transfer equation into an equation for
photon ux density
d
dt
=
c
hn
du
dt
=

nc

at
c
n
_
1
2

0
nc|E|
2
h
_
=
c
n
. (6.11)
Comparing this equation with Eq.(6.10)
3
, we once again recover our earlier result
=

nc

at
= N
21
= (N
2
N
1
) = (N
1
N
2
) .
Now let us examine if we can achieve inversion in a two-level system. This means
we focus on transitions between two xed levels and ignore transitions to or from
all other levels to these two levels. By adding and subtracting Eqs. (6.9) we nd
equations for the total number of atoms N = N
2
+ N
1
and population inversion
N
21
= N
2
N
1
[N
1
= (N N
21
)/2, N
2
= (N +N
21
)/2]
d(N
1
+ N
2
)
dt
= 0 (6.12a)
dN
21
dt
= 2
12

12
N
21
2
21
N
2
= 2
12

12
N
21

21
(N +N
21
) (6.12b)
The rst of these equations simply expresses the fact that the total number of atoms
N
1
+ N
2
= N remains constant when the transitions between E
1
and E
2
and all
other levels are ignored. For a strictly two-level system, the system can be only in
one of two states available to it.
6.2.1 Steady-state inversion in 2-level system
In the steady-state dN
21
/dt = 0, the population equation (6.12b) leads to
N
21
=

21

21
+ 2
12

12
N =
N
1 + 2I/h
21
. (6.13)
We see that atomic inversion depends on eld strength via I. We note that the
quantity
I
sat
=
h

21
= one quantum of energy per cross-section per life time (6.14)
3
In making this comparison we ignore the spontaneous emission contribution.
Laser Pump and Threshold 167
has dimensions of intensity. This intensity, called the saturation intensity, sets the
scale for the eld intensity to alter the level populations signicantly. In terms of
this intensity, the steady-state inversion can be written as
N
21
=
N
1 + 2(I/I
sat
)
. (6.15)
-1
-0.8
-0.6
-0.4
-0.2
0
0.01 0.1 1 10 100

N
I/I
sat
FIGURE 6.4
Atomic inversion N = N
21
/N for a two-level system as a function of intensity.
The factor of two in the denominator is peculiar to two-level system and will not
appear later when we consider multi-level systems. The reason for this is that in a
two-level system, the transition of an atom from one state to another changes the
population dierence by 2.
Two-level atomic inversion (6.15) is plotted in Fig. (6.4) as a function of eld
intensity I/I
sat
. Field intensities small or large compared to I
sat
do not alter the
inversion signicantly, whereas the intensities comparable to I
sat
cause a rapid vari-
ation in inversion. It is clear from the graph that it is not possible to achieve
inversion in the steady-state for a two-level system. In fact, for very high intensities
(I I
sat
) we approach the limit N = 0 corresponding to equal level populations
N
1
= N
2
.
The steady-state absorption coecient (I) = N
21
is given by
(I) =

12
N
0
1 + 2(I/I
sat
)
=

0
1 + 2(I/I
sat
)
. (6.16)
where
0
= N
0
is the absorption coecient when I = 0. This value of absorp-
tion coecient is called the unsaturated absorption. As eld strength increases
absorption decreases, approaching zero for very high intensities (I I
sat
. In other
168 Laser Physics
words, the medium becomes transparent for very strong elds. This phenomenon
is sometimes referred to as bleaching and is an example of saturation behavior.
0
0.2
0.4
0.6
0.8
1
0.01 0.1 1 10 100

(
I
)
/

o
I/I
sat
FIGURE 6.5
Absorption coecient for a two-level medium as a function of intensity.
6.2.2 Transient response
We have seen steady-state inversion is not possible in a two-level system. Let us
investigate if transient inversion is possible. Using the scaled variables
=
21
t , N = N
21
/N
0
, I/I
sat
= I
we can write the population equation and its solution as
dN
d
= (1 + 2I)N 1 (6.17a)
N = N
0
e
(1+2I)
+
1
1 + 2I
_
e
(1+2I)
1
_
(6.17b)
where N
0
is the initial value of atomic inversion. Figure (6.6) shows the transient
response of a two-level system if initially (at time t = 0) an inversion N = 0.5
Laser Pump and Threshold 169
was created. It can be seen that as time increases atomic inversion decreases and
relaxes toward its steady-state value monotonically. This picture changes somewhat
when dynamical energy exchange between the eld and atom is taken into account.
In that case the approach to the steady-state can be oscillatory but the conclusion
that population inversion in the steady-state remains unchanged.
1
0.5
0
0.5
1
0 0.5 1 1.5 2 2.5 3

I/I
sat
=0
1
2
3
N

=0.5
FIGURE 6.6
Transient response of two level inversion when initially a positive inversion is created.
6.3 Three and Four Level Inversion Schemes
Pump schemes for achieving inversion in a laser can be divided into three or four
level schemes as shown in Figs. (6.7) and (6.8). Ruby is the best known example of
a three level laser system. Other examples include Erbium based lasers. Gas lasers
such as He:Ne and Ar-ion and dye lasers are examples of four level lasers.
The pump mechanism which supplies energy to excite atoms from the lower pump
level (level 1 in Fig. (6.7) and 0 in (6.8)) to the upper pump level (level 3 in both
170 Laser Physics
Figs. (6.7) and (6.8)) varies greatly from one laser system to another. For example,
in Ruby or dye lasers the pump is the optical energy from a ash lamp or a pump
laser, in a He:Ne laser it is the collision of excited He atoms with Ne atoms in
an electrical discharge, in semiconductor lasers it is the electrical current, and in
chemical lasers it is a chemical reaction. Whatever its nature, the pump mechanism
eectively induces transitions at a rate W
p
between the lower and upper pump
levels. In our discussion we will consider an optical pump. Let us consider two
pump schemes separately and compare their them.
6.3.1 Three-level Pump Scheme
Various transition rates for a three-level pump scheme are shown in Figure (6.7).
These rates lead to the following equations for population densities (N
i
is the num-
ber of atoms per unit volume that are in state i)
dN
3
dt
= W
p
(N
3
N
1
)
3
N
3
, (6.18a)
dN
2
dt
=
32
N
3

2
N
2
W(N
2
N
1
) , (6.18b)
dN
1
dt
= W
p
(N
3
N
1
) +
31
N
3
+
2
N
2
+ W(N
2
N
1
) , (6.18c)
N = N
1
+ N
2
+ N
3
. (6.18d)
Here W
p
=
13

p
is the transition rate induced by an optical pump of photon ux
density
p
between the lower and upper pump levels E
1
and E
3
and W = is
the transition rate between lower and upper lasing levels E
1
and E
2
induced by
the presence of laser photon ux density . Each decay rate
ij
, in general, has a
radiative contribution as well as a nonradiative contribution
4
.
In writing these Eqs. (6.18), we have assumed level 1 to be the ground state.
Therefore it does not decay to any other level. Equation (6.18d) assumes that
atomic decays to levels other than those shown in Fig. (6.7) are negligible so that
the sum of population densities equals the total density of active atoms. We have
also assumed that the levels are non-degenerate, and that no radiation eld is present
at the 23 transition frequency. Finally, the rate W
P
is not necessarily induced
by an optical eld. It could be one of many mechanisms including collisions in
an electrical discharge, an electric current, or a chemical reaction, that can excite
atoms from the ground state to the upper pump level.
In the steady-state, the population densities are independent of time (

N
1
= 0 =

N
2
=

N
3
). Then without saturation of the 2 1 transition (W = 0), we obtain
4
Radiative decay is accompanied by the emission of a photon, whereas the nonradiative decay is
not.
Laser Pump and Threshold 171
E
2
E
1
W
p
N
1
W
p
N
3
WN
1
WN
2

31
N
3
E
3

32
N
3

2
N
2
FIGURE 6.7
Three-level pump scheme using an optical pump.
the following equations for the population densities
0 = W
p
(N
3
N
1
)
3
N
3
(6.19a)
0 =
32
N
3

2
N
2
(6.19b)
0 = W
p
(N
3
N
1
) +
2
N
2
+
31
N
3
(6.19c)
Solving these equations for the population densities in the steady-state we get
N
3
=
W
p
W
p
+
3
N
1
=

21rad

3
W
p
/
21rad
W
p
/
3
+ 1
N
1
=

2

32

32

21rad

2
W
p
/
2
W
p
/
3
+ 1
N
1
=
W
p
/
2
W
p
/
3
+ 1
N
1
=

32

21rad

2
=

2

32
(6.19d)
N
2
=

32

2
N
3
=

32

21rad

2
W
p
/
21rad
W
p
/
3
+ 1
N
1
=
W
p
/
21rad
W
p
/
3
+ 1
N
1
(6.19e)
Using these in Eq.(6.18d) we nd
N = N
1
+ N
2
+ N
3
= N
1
_
1 +
(1 +) W
p
/
2
W
p
/
3
+ 1
_
= N
1
_
1 +W
p
/
21rad
+ (1 +)W
p
/
21rad
W
p
/
3
+ 1
_
= N
1
_
1 + (1 + 2)W
p
/
21rad
W
p
/
3
+ 1
_
. (6.19f)
This gives us N
1
in terms of the total number density N
N
1
= N
_
W
p
/
3
+ 1
1 + (1 + 2)W
p
/
21rad
_
(6.19g)
172 Laser Physics
Using Eqs. (6.19e) and (6.19g) for N
2
and N
1
, unsaturated inversion density [N
2

N
1
]
unsat
N
(o)
for three-level pump scheme can be written as
N
(o)
=
_
W
p
/
2
W
p
/
3
+ 1
1
_
N
1
=
_
(1 )W
p
/
2
1
W
p
/
3
+ 1
_
N
1
= N
_
(1 )W
p
/
21rad
1
1 + (1 + 2)W
p
/
21rad
_
, (6.20a)
where
=

2

32
, =

32

21rad

2
. (6.20b)
From Eq. (6.20a) we see that to achieve population inversion, the pump rate W
p
must exceed the value
[W
p
]
min
=

21rad
(1 )
. (6.21)
The conditions favoring inversion are those that lower [W
p
]
min
. These conditions
from Eq. (6.21) are
1 , 0 . (6.22)
Physical meaning of and
is the product of two factors. The rst factor
32
/
3
is the ratio of the spontaneous
decay rate
32
of level 3 into level 2 to its total spontaneous decay rate
3
. The
second factor is the radiative decay rate
21rad
of level 2 into level 1 to the total
decay rate of level 2. Therefore, the fraction of level 3 population that reaches level
2 and then ends up in level 1 by emitting a photon at signal frequency. For optical
pumping, W
p
N
1
is the number of quanta absorbed from the pump per unit time.
This is also the rate at which atoms are pumped into level 3 since the absorption
of each pump quantum results in the transfer of one atom to the upper pump level.
A fraction (
32
/
3
) of these transitions end up in level 2 and of these a fraction

21rad
/
2
decays into level 1 by emitting a photon at signal frequency. Hence the
rate at which laser quanta are emitted is W
p
N
1
. The ratio of laser quanta emission
rate to pump quanta absorption rate is precisely . Thus represents the eciency
(uorescent quantum eciency) with which pump quanta are being converted to
laser quanta. For an ideal system 1.
To see the meaning of let us consider the ideal case. In an ideal system, level
3 decays predominantly to level 2 so that all pumped atoms reach the upper laser
level. In this case
3
=
32
+
31

32
and the parameter is the ratio of the upper
pump level lifetime to the upper laser level lifetime. Now, from the expression for
N
(o)
we see that to achieve inversion must be less than unity and ideally 0.
This means the rate at which atoms arrive in level 2 is fatser than the rate at which
they leave.
In view of the preceding discussion, the conditions in (6.22) for an ideal system
mean that (i) level 3 must decay primarily to level 2 so that most of the atoms
pumped to level 3 end up in level 2 and (ii) level 2 must be longer lived compared
to level 3 so atoms can accumulate there.
Laser Pump and Threshold 173
E
2
E
1
W
p
N
0
W
p
N
3
WN
1
WN
2

31
N
3
E
3

32
N
3

21
N
2
E
0

30
N
3

20
N
2

1
N
1
FIGURE 6.8
Four-level scheme using an optical pump.
6.3.2 Four-level Pump Scheme
Figure (6.8) shows various transition rates for a four-level pump scheme. These
rates lead to the following equations of motion for the population densities
dN
3
dt
= W
p
(N
3
N
0
)
3
N
3
, (6.23a)
dN
2
dt
=
32
N
3

2
N
2
W(N
2
N
1
) , (6.23b)
dN
1
dt
=
31
N
3
+
21
N
2

1
N
1
+ W(N
2
N
1
) , (6.23c)
dN
0
dt
=
30
N
3
+
20
N
2
+
1
N
1
+ W
p
(N
3
N
0
) . (6.23d)
N = N
1
+ N
2
+ N
3
+ N
4
. (6.23e)
Here, once again, the last equation assumes that atomic decays to levels other than
those shown in Fig. (6.8) are negligible so that the sum of population densities
equals the density of active atoms. Level 0 is the ground state and we have assumed
the levels to be non-degenerate. We have also assumed that no radiation eld is
present at 23 and 10 transitions.
Ignoring saturation (W = 0), we nd that in the steady-state, the level population
174 Laser Physics
densities satisfy
0 = W
p
(N
3
N
0
)
3
N
3
(6.24a)
0 =
32
N
3

2
N
2
(6.24b)
0 =
31
N
3
+
21
N
2

1
N
1
(6.24c)
0 =
30
N
3
+
20
N
2
+
1
N
1
+ W
p
(N
3
N
0
) (6.24d)
Solving these equations for the population densities, we nd
N
3
=
W
p
/
3
W
p
/
3
+ 1
N
0
=

2

32

32

21rad

2
W
p
/
21rad
W
p
/
3
+ 1
N
0
=

2

32
W
p
/
21rad
W
p
/
3
+ 1
N
0
(6.24e)
N
2
=

32

2
N
3
=

32

2
W
p
/
3
W
p
/
3
+ 1
N
0
=

32

21rad

3
W
p
/
21rad
W
p
/
3
+ 1
N
0
=
W
p
/
21
W
p
/
3
+ 1
N
0
=

32

21rad

2
(6.24f)
N
1
=
_

21

1
N
2
+

31

1
N
3
_
=
_

21

1
+

2

31

32
_
W
p
/
21rad
W
p
/
3
+ 1
N
0
=
W
p
/
21rad
W
p
/
3
+ 1
N
0
=
_

21

1
+

2

31

32
_
(6.24g)
N
2
N
1
=
(1 )W
p
/
21rad
W
p
/
3
+ 1
N
0
(6.24h)
Using these expressions we can express N
0
in terms of atomic number density N as
N = N
0
+ N
1
+ N
2
+ N
3
= N
0
_
1 +
W
p
/
21rad
W
p
/
3
+ 1
+
W
p
/
21rad
W
p
/
3
+ 1
+

2

32
W
p
/
21rad
W
p
/
3
+ 1
_
N
0
=
_
W
p
/
3
+ 1
1 + (1 + + 2
2
/
32
)W
p
/
21rad
_
N .
Using this in the expressions for the population densities, we nally obtain the
unsaturated inversion density for a four-level system
N
(o)
[N
2
N
1
]
unsat
=
_
(1 )W
p
/
21rad
1 + (1 + + 2
2
/
32
)W
p
/
21rad
_
N , (6.25a)
where
=

32

21rad

2
, (6.25b)
=

21

1
+

2

31

32
. (6.25c)
Laser Pump and Threshold 175
In an ideal system, the upper pump level decay rate
32
to upper laser level will be
large compared with the upper laser level decay rate
2
(
32

2
so that
N
(o)

_
(1 )W
p
/
21rad
1 + (1 +)W
p
/
21
_
N . (6.26)
As in the three-level case, we note that is the eciency with which pump photons
are converted to laser photons. The rst factor in is the fraction of atoms pumped
into level 3 that relax to the upper laser level and the second factor is the fraction
of atoms in level 2 that decay radiatively into level 1.
The parameter is the relaxation ratio. In an ideal system level 3 decays primarily
to the upper laser level 2 (
31

32
) and the upper laser level primarily to the
lower laser level 1 so that
2
/
1
is essentially the ratio of the lower and upper
lasing level lifetimes.
-1
-0.5
0
0.5
1
0 1 2 3 4

N

(
0
)
/
N
W
p
/
2
or
3-level inversion
4-level inversion
W
p
/
21
FIGURE 6.9
Three and four-level inversion as a function of normalized pump W
p
/
21rad
or
W
p
/
21rad
Equation (6.25a) shows that in a four-level system, the inversion on the laser
transition 2 1 is easily achieved for any nonzero W
p
as long as < 1. The
176 Laser Physics
conditions favoring inversion are again
1 , 0 (6.22*)
Under these conditions unsaturated inversion is given by
N
(o)
=
NW
p
/
21
1 + W
p
/
21
(6.27)
These conditions mean to have good inversion (i) the 3 1 relaxation is very fast
and almost entirely to level 2, but the upper level 2 has comparatively long lifetime
so that its population can buildup, and (ii) atoms should relax out of the lower laser
level much faster than the atom relax into level 2 from above.
6.3.3 Three vs four-level systems
We have seen that population inversion can be achieved using either three or four
level pump schemes. Which schemes is to be used depends on the laser system.
A comparison of the three and four-level pump schemes reveals that its is easier to
achieve inversion in a four-level scheme where any amount of pumping can produce
inversion as long as the upper laser level is longer lived than the lower laser level
( < 1). In contrast, in the three-level scheme the pump must exceed the minimum
given by Eq. (6.21) to achieve inversion. This dierence between the two schemes
can be traced to the fact that in the three level scheme the lower level is shared
both by the pump and laser transitions. Since all the atoms are normally in the
lowest level 1, at least half of them must be pumped out from level to to higher
levels in order to achieve population inversion between levels 2 and 1.
6.3.4 Eective two-level pump schemes
Under certain conditions four-level pump scheme can be reduced to an eective two-
level pump scheme. We consider two cases which lead to two qualitatively dierent
saturation behavior.
CASE I:
3
W
p

1
The condition
3
W
p

1
implies a bottleneck at level 2. This means that
pumping is so strong that it pumps an atom from level 0 to level 3 as soon as it
decays to level 0, and once the atom is in level 3, it decays almost instantaneously
to the lower levels. We can then solve for the equilibrium values of N
3
and N
0
N
3
=
W
p
W
p
+
3
N
0
, (6.28)
N
0
=
W
p
+
3
W
p
_

1

32
+
31
N
1
+

20

32
+
31
N
2
_
. (6.29)
Laser Pump and Threshold 177
Substituting these into the equations of motion for N
1
and N
2
we obtain
dN
2
dt
=
32
W
p
W
p
+
3
_
W
p
+
3
W
p
__

1

32
+
31
N
1
+

20

32
+
31
N
2
_
(6.30)

2
N
2
W(N
2
N
1
)
=

1

32

32
+
31
N
1

21
_
1 +

31

32
+
31

20

21
_
N
2
W(N
2
N
1
) ,
=

12
N
1

21
N
2
W(N
2
N
1
) (6.31)
dN
1
dt
=

12
N
1
+

21
N
2
+ W(N
2
N
1
) (6.32)
Note that we still have N
1
+ N
2
= N. From the form of these equations we see
that, eectively, the atoms are being pumped directly from level 1 to level 2 at a
rate

12
=
1

32
/(
32
+
31
).
The pump term appears independent of W
p
but only because of the assumption
W
p

1
. Solving these equation in the steady-state we nd N
2
N
1
N
N =

12

21

12
+

21
+ 2W
=
(

12

21
)/(

12
+

21
)
1 + 2W/(

12
+

21
)
=
N
(o)
1 + 2W/(

12
+

21
)
(6.33)
Note that the factor of 2 in the saturation term which is typical of a two-level
system.
Case II: N
3
, N
2
, N
1
N
0
If the population of the ground level 0 remains essentially unchanged so that
N
3
, N
2
, N
1
N
0
, and level 3 decays almost instantaneously to level 2, we nd
N
3
=
W
p
W
p
+
3
N
0

W
p
N
0

R
p

3
, W
p

3
. (6.34)
Equations of motion for N
1
and N
2
then become
dN
2
dt
=

32
W
p

3
N
0

2
N
2
W(N
2
N
1
)
R
2

2
N
2
W(N
2
N
1
) , (6.35)
dN
1
dt
=

31
W
p

3
N
0
+
21
N
2

1
N
1
+ W(N
2
N
1
) ,
R
1
+
21
N
2

1
N
1
+ W(N
2
N
1
) R
1
=

31

32
R
2
(6.36)
If
31

32
, the pumping term for N
1
can be ignored.
In the steady-state we obtain
_
W +
2
W
(W +
21
) W +
1
__
N
2
N
1
_
=
_
R
2
R
1
_
Solving these equations we obtain
_
N
2
N
1
_
=
1
detM
_
W +
1
W
W +
21
W +
2
__
R
2
R
1
_
178 Laser Physics
where
detM = (W +
2
)(W +
1
) W(W +
21
) =
1

2
+ W(
1
+
20
)
The inversion is then, using R
1
=

31

32
R
2
,
N
2
N
1
=
R
2
(
1

21
) R
1

2
+ W(
1
+
20
)
=
(
21rad
/
2
)(R
2
/
21rad
)(1 )
1 + W/
2eff
=
(1 )(R
2
/
21rad
)
1 + W/
2eff
where
=

21rad

2
(6.37a)
=

21

1
+

2

31

32

1
, for
20

21
(6.37b)

2eff
=

2
1 + (
20
/
1
)

2
, for
20

1
(6.37c)
Note that for I = 0 we recover the unsaturated result for a four level system.
This equation indicates that population inversion is a nonlinear function of eld
intensity and as the signal intensity increases, the inversion decreases. Because
of this intensity dependence of inversion, both absorption and gain are nonlinear
functions of eld intensity.
Once we are able to obtain inversion, we can write down the expression for gain.
6.4 Gain Saturation
Saturation of gain in homogeneous and inhomogeneous media has qualitatively dif-
ferent characteristics. To appreciate this we consider two extreme cases of strong
homogeneous and strong inhomogeneous broadened media.
6.4.1 Gain saturation in homogeneously broadened media
Using the expression for W = ()I/h, we can write the inversion as
N =
N
(o)
1 + (()/h
2eff
)I
=
N
(o)
1 + I/I
sat
()
(6.38)
where the unsaturated inversion N
(o)
and saturation intensity I
sat
are given by
N
(o)
= (1 )
R
2

21rad
=0

R
2

2
, (6.39)
I
sat
() =
h
2eff
()
. (6.40)
Laser Pump and Threshold 179
This expression for the saturation intensity says that the saturation intensity cor-
responds to one photon energy (h) incident per transition cross-section () per
lifetime (1/
2eff
) of the transition. When the signal intensity equals I
sat
inversion
falls to half of its unsaturated value. Saturation intensity is thus a characteristic
intensity that determines the degree of gain saturation with signal intensity. For
a given signal intensity, gain saturation is stronger for small saturation intensities
than it is for higher saturation intensities.
Using the expression for atomic inversion, we can write the gain coecient g() =
()N as
g() =
()N
(o)
1 + I/I
sat
()
=
g
o
()
1 + I/I
sat
()
(6.41)
where
g
o
() = ()N
(o)
=

2
o

21rad
8n
2
S()N
(o)
(6.42)
is the unsaturated gain. Unsaturated gain is also referred to as the small signal
gain.
From Eq.(6.41), we see that the presence of a signal of intensity I reduces the gain
below its unsaturated value g
o
(). This reduction of gain due to signal intensity is
referred to as gain saturation. Reduction of gain below its unsaturated value, when a
signal of intensity I is applied, is not uniform with frequency because the saturation
intensity depends on frequency. The smallest value of saturation intensity occurs
at line center, where its value is
I
sat
(
o
) I
s
=
h
o

2eff
(
o
)
=
2c
2eff

o
8n
2

2
o

12rad

H
2
=
c
2eff
8
3
n
2

3
o

12rad
(6.43)
In terms of I
s
, the saturation intensity at an o line center frequency can be
written as
I
sat
() = I
s
S(
o
)
S()
= I
s
(
o
)
2
+ (
H
/2)
2
(
H
/2)
2
= I
s
_
1 +
_
2(
o
)

H
_
2
_
. (6.44)
This expression indicates that the saturation intensity o line center is larger than
it is at line center and this means a given intensity signal will saturate gain more
strongly at line center than o line center. The largest reduction in gain occurs at
the line center where
g(
o
) =
g
o
(
o
)
1 + I/I
s
=

2
o

21rad
N
(o)
4
2
n
2

H
(1 + I/I
s
)
(6.45)
Overall frequency dependence of saturated gain is determined by both the fre-
quency dependence of atomic line shape function S() and the frequency dependence
of the saturation behavior in Eq.(6.41). With the help of Eqs. (6.41) and (6.43),
180 Laser Physics

g
0
()
I=I
s
g()
I=5I
s
FIGURE 6.10
Saturated gain curve is lower and broader than the unsaturated gain curve.
the explicit frequency dependence of the saturated gain coecient can be written
as
g() =
g
o
(
o
)S()/S(
o
)
1 + IS()/I
s
S(
o
)
=
g
o
(
o
)
1 + (2(
o
)/
H
)
2
+ I/I
s
=
g
o
(
o
)
1 + I/I
s
_
1
1 + (2(
o
)/
b
)
2
_
(6.46)
where
b
is the power broadened linewidth given by

b
=
H
_
1 + I/I
s
, (6.47)
where g
o
(
o
) is the unsaturated gain at the line center.
Saturated gain curve g() for a homogeneously broadened atomic line is plotted
in Fig. (6.10) for several dierent values of signal intensity. The saturated gain
curve g() is still a lorentzian but it is broader than the unsaturated gain curve
g
o
(). The width (FWHM) of the gain curve
b
increases with signal intensity.
This broadening of the gain curve due to the saturation of atomic inversion is called
power broadening.
6.4.2 Gain saturation of an inhomogeneous line
So far, in writing g() we have assumed that all atoms have the same broadening.
Our considerations therefore apply to a homogeneously broadened gain medium.
If atomic resonance frequencies are distributed according to, say, a gaussian, gain
saturation has a dierent behavior. We shall illustrate this by considering two
limiting cases.
Laser Pump and Threshold 181
Assume that atoms are subjected to both homogeneous and inhomogeneous broad-
ening. Let the atomic resonance frequencies be distributed according to
G(
a
) =
1

D
_
4 ln2

exp
_

D
_
2
4 ln2
_
(6.48)
Then the overall gain coecient will be
g() =

_
0
d
a
G(
a
)

2
o

21rad

H
N
(o)
16
2
n
2
1
(
a
)
2
+ (
H
/2)
2
(1 + I/I
s
)
=

2
o

21rad
N
(o)
b
2
8
2
n
2
1

H
_
4 ln2

dy
e
y
2
(x + y)
2
+ b
2
(1 + I/I
s
)
, (6.49a)
where
y =
(
a

o
)2

ln 2

D
, (6.49b)
x =
(
o
)2

ln 2

D
, (6.49c)
b =

H
2

ln 2
2
D
. (6.49d)
If homogeneous broadening is dominant the lorentzian is broad compared to the
gaussian. Then so long as we are not in the far wing limit we can factor out the
lorentzian by evaluating it at the peak (y = 0). The resulting expression for gain is
given by Eq.(6.41).
If Doppler broadening dominates (the gaussian is broader than the power broad-
ened lorentzian,
D

H
_
1 + I/I
s
) and we are not in the far wing limit (x is
small) we can evaluate the gaussian near the peak y = x of the Lorentzian and
factor it out of the integral. Then we are led to the following expression for gain in
the strong inhomogeneous broadening limit
g() =

2
o

21rad
N
(o)
8n
2
_
1 + I/I
s
1

D
_
4 ln2

exp
_

_

o

D
_
2
4 ln2
_
=
g
o
()
_
1 + I/I
s
, (6.50)
where g
o
() is the unsaturated gain given by
g
o
() =

2
o

21rad
N
(o)
8n
2
1

D
_
4 ln2

exp
_

_

o

D
_
2
4 ln2
_
. (6.51)
Comparing this to a homogeneously broadened line, we note that in the inhomo-
geneous case saturation of gain is independent of frequency and it sets in at a
slower rate than for a homogeneous line. In this extreme Doppler broadening limit
power broadening does not show up in the overall gain prole. It is nevertheless
present for a particular velocity group whose frequency is close to the frequency of
the applied eld.
182 Laser Physics
6.5 Saturation Spectroscopy
Another manifestation of the dierence in the saturation behavior of homogeneous
and inhomogeneous lines is in saturation spectroscopy. The main idea here is that a
strong eld (signal or saturating eld) interacting with a collection of atoms modies
the gain (or loss) seen by a second weaker eld (probe eld). The nature of this
modication depends on the nature of broadening and the frequency of the weaker
beam relative to the frequency of the stronger beam. It has important implications
for multimode laser oscillation.
Consider a group of atoms with resonance frequencies in the range [
a
,
a
+d
a
].
These atoms are a fraction G(
a
)d
a
of the total number of atoms per unit volume.
Then the gain coecient g() for the weak eld at frequency in the presence of a
strong (saturating ) eld at frequency
s
can be written
g() =

2

21rad
8n
2
_

0
d
a
G(
a
)S(
a
, )N(
s
), (6.52)
where N(
s
) is the inversion produced essentially by the strong eld at frequency

s
. This is reasonable since the eld at frequency
s
is assumed to be much stronger
than the probe eld at frequency . The inversion is given by
N(
s
) =
N
(o)
1 + I/I
sat
(
s
)
=
N
(o)
[(
a

s
)
2
+ (
H
/2)
2
]
(
a

s
)
2
+ (
H
/2)
2
(1 + I/I
s
)
, (6.53)
where I is the intensity of the strong eld and I
s
is the peak saturation intensity
given by
I
s
= h
o

2eff
_
4
2
n
2

2
o

21rad
_
=
c
2eff
8
3
n
2

3
o

21rad
. (6.43*)
Substituting this in the expression for g() we obtain
g() =

2

21rad
N
(o)
8n
2
b
2

3/2
2

ln 2

H
_

dy
e
y
2
(x + y)
2
+ b
2

_
(x
s
+ y)
2
+ b
2
(x
s
+ y)
2
+ b
2
(1 + I/I
s
)
_
, (6.54)
where the scaled variables are dened by
y =
(
o

a
)

D
2

ln2 (6.55a)
x =
(
o
)

D
2

ln2 (6.55b)
x
s
=
(
o

s
)

D
2

ln2 (6.55c)
b =

H
2
D
2

ln2 (6.55d)
In Eq. (6.54) we have again extended the lower limit to because the integrand
is vanishingly small well before the lower limit
o
/
D
is reached.
Laser Pump and Threshold 183
6.5.1 Spectral Holes in inhomogeneous gain prole
In the extreme Doppler limit b 1 [
H

D
, Eq.(6.55d)] the gaussian is much
broader than either of the two Lorentzians. Moreover, because of power broadening
of the second lorentzian, the rst Lorentzian is the narrowest of the two. It is
centered at y = x so that we can approximately evaluate the integral as
g() =

21rad
N
(o)
8n
2
G()
_
(x
s
x)
2
+ b
2
(x
s
x)
2
+ b
2
(1 + I/I
s
)
_
=

21rad
N
(o)
8n
2
G()
_
(
s
)
2
+ (
H
)
2
(
s
)
2
+ (
H
/2)
2
(1 + I/I
s
)
_
=g
o
()
_
(
s
)
2
+ (
H
)
2
(
s
)
2
+ (
H
/2)
2
(1 + I/I
s
)
_
(6.56)
where g
o
() is the unsaturated gain at frequency
g
o
() =

2

21rad
N
(o)
8n
2
G(). (6.57)
If probe beam frequency is far from the frequency
s
of the saturating beam
[(
s
)
H
], the factor inside square brackets is essentially unity. This means
that small signal gain remains unchanged far from the frequency of the strong
saturating beam. For probe beam frequencies in the neighborhood of
s
this factor
is less than unity so that the gain curve has a dip (hole) centered at =
s
[See
Fig. 6.11]. This phenomena is known as (spectral) hole burning.

s

g()

o
FIGURE 6.11
Saturated gain as a function of the frequency as measured by a weak probe beam
in the presence of a strong saturating beam of frequency
s
in an inhomogeneously
broadened medium.
What is the width of this hole? We note that the depth of the hole is determined
184 Laser Physics
by the intensity I of the saturating beam relative to the peak saturation intensity
I
s
. For strong saturation I I
s
, the width of the whole is, approximately,

hole

H
_
1 + I/I
s
(6.58)
Unlike in an homogeneous gain medium, in an inhomogeneously broadened gain
medium a strong signal aects gain only in a small neighborhood centered at its
frequency
s
.
Spectral hole-burning has important consequences for multimode operation of
lasers with an inhomogeneously broadened gain medium, where several modes can
oscillate simultaneously because each mode saturates gain only in a small neighbor-
hood centered at its frequency. It also helps us understand an unexpected aspect
of single-mode standing wave gas lasers. If we tune the frequency of a single-mode
standing wave cavity across the Doppler broadened gain prole, the power output
as a function of laser frequency shows a narrow dip when the cavity frequency co-
incides with the center of gain prole. This dip is known as Lamb dip. We can
consider the eld inside a standing wave laser as a superposition of two traveling
waves running in opposite directions z. Then the +z traveling-wave will interact
with a group of atoms with velocity v/c = (
o
)/
o
while the z traveling-wave
will interact with a group of atoms with velocity v/c = (
o
)/
o
. When the
cavity frequency is suciently detuned from the line center, two holes are burned in
the gain prole, which are located symmetrically about the line center at
o
,
where
o
is the detuning of mode frequency from the line center. As the
detuning is reduced to zero, the two holes merge and both traveling wave compo-
nents interact with the same set of atoms corresponding to v = 0. Thus whereas
o-resonance the laser sees gain from two dierent groups of atoms, at line center
it sees gain from a single group of atoms. Hence as the detuning is reduced to
zero, the laser power output increases rst but as the holes begin to overlap, it
drops reaching a relative minimum at line center. Thus there is a slight but denite
dip in laser power output at line center. This output power dip in a single-mode
standing wave gas laser is known as Lamb dip. The center of the Lamb dip serves
to mark the exact center of the atomic transition and can be used as a reference
in frequency stabilized lasers. In addition the width of the dip provides a way of
measuring homogeneous linewidth of an atomic transition even in the presence of
strong inhomogeneous broadening.
1. W. E. Lamb, Jr., Theory of an optical maser, Phys. Rev. 134, 1429 (1964).
2. R. A. Mcfarlane, W. R. Bennett, Jr., and W. E. Lamb, Jr., Single mode
tuning dip in the power output of an He-Ne optical maser, Appl. Phys. Let.
2, 189 (1963).
6.5.2 Homogeneously Broadened Gain Prole
For strong homogeneous broadening
H

D
[b 1]. Moreover because of
power broadening the second Lorentzian is broader than the rst. We can then
Laser Pump and Threshold 185
approximately evaluate the integral in Eq.(6.54) by factoring out the gaussian and
the second lorentzian, both evaluated at the peak of the rst lorentzian at y = x.
We then obtain
g()

2
o

21
N
(o)
8n
2
b

4 ln2

H
e
x
2 (x
s
x)
2
+ b
2
(x
s
x)
2
+ b
2
(1 + I/I
s
)
b

dy
1
(x + y)
2
+ b
2
=

2
o

21
N
(o)
8n
2
b

4 ln2

H
e
x
2 (x
s
x)
2
+ b
2
(x
s
x)
2
+ b
2
(1 + I/I
s
)
=

2
o

21
N
(o)
8n
2

4 ln2

D
e
(
o
)
2 (
s

o
)
2
+ (
H
/2)
2
(
s

o
)
2
+ (
H
/2)
2
(1 + I/I
s
)
= g
o
()
_
(
s

o
)
2
+ (
H
/2)
2
(
s

o
)
2
+ (
H
/2)
2
(1 + I/I
s
)
_
(6.59)
The eect of the presence of a strong eld on the gain prole of a weak signal is
contained in the factor inside square brackets. Note that this factor is independent
of probe frequency. This behavior is in contrast to Eq.(6.57) for a strongly Doppler
broadened medium where the factor multiplying g
o
depends on the probe frequency
relative to the saturating eld frequency. Hence for a homogeneously broadened
line, g() has the same frequency dependence as the unsaturated gain g
o
() but it
is modied (reduced) everywhere by the factor inside the square brackets. This is
in contrast to behavior of an inhomogeneously broadened line. Thus the presence
of a strong eld at frequency
s
reduces gain at all frequencies if the line is homo-
geneously broadened. The factor by which the gain is reduced, of course, depends
on the frequency of the saturating eld relative to line center.
It should be kept in mind that in all of our considerations, we had assumed that
the or
s
were not far from line center relative to
H
. Far in the wings, all lines
have predominantly homogeneous character.
6.6 Threshold Condition
1. A. Siegman, Lasers (University Science Boks, Mill Valley, 1986), Chapters
12,13
2. P. W. Milonni and J. H. Eberly (John Wiley and Sons, Inc., Hoboken, NJ,
2010), Chapter 4.
3. A. Yariv, Quantum Electronics (John Wiley and Sons, Inc., Hoboken, NJ,
1989), Chapter 8.
An amplier with feedback is an oscillator where, provided certain conditions are
met, nonzero eld intensity can be sustained without any input elds. Thus, strictly
speaking, a laser should be called a loser but it is inappropriate to call a winner a
loser. So we will continue to use the term laser for a light oscillator.
186 Laser Physics
z =0 z = L
Pump
Gain medium
I
+
I

R
1
=1T
1
R
2
=1T
2
l
m
FIGURE 6.12
A laser is an oscillator formed by a gain medium surrounded by mirrors to provide
optical feedback.
In a laser, amplication or gain is provided by stimulated emission by the gain
medium and optical feedback is arranged by placing mirrors around the gain medium
so as to provide a closed optical path [see Fig. (6.12)]. Light emitted by the gain
medium in directions close to the axis is thus trapped between the mirrors and
bounces back and forth between them making multiple passes through the gain
medium. With each passage through the medium the intensity of light will grow
because of gain due to stimulated emission. The intensity will also decrease due
to absorption and scattering inside the medium and, if the mirrors are partially
transmitting, transmission through the mirrors. In order to reach a steady-state
with nonzero intensity (oscillation) the gain due to stimulated emission must be
sucient to overcome these losses. This sets a lower limit on g() below which laser
oscillation cannot occur. This value of gain is called threshold gain. An expression
for threshold gain can be derives as follows.
We consider only the steady-state, continuous wave (cw) operation. Near thresh-
old the intracavity eld intensity is small so that the gain can be replaced by un-
saturated gain. Then, for a standing wave cavity, the eld inside the cavity can be
described in terms of the intensities of the left and right traveling waves
dI
+
dz
=(g
o
)I
+
, (6.60)
dI

dz
=(g
o
)I

, (6.61)
where I
+
is the intensity of the wave traveling to the right and I

is the intensity of
the wave traveling to the left. Assuming that the gain lls the entire cavity, we can
integrate these equations from z = 0 to z for I
+
and z = L to z for I

and obtain
I
+
(z) = I
+
(0)e
(g
o
)z
(6.62a)
I

(z) = I

(L)e
(g
o
)(Lz)
(6.62b)
Laser Pump and Threshold 187
The right-traveling wave is reected by mirror 1 at z = L to give rise to the left-
traveling wave. Similarly the left-traveling wave is reected by mirror 2 to produce
the right-traveling wave. The intensities of the two waves at z = 0 and z = L are
related by the boundary conditions
I
+
(0) = R
1
I

(0) (6.63a)
I

(L) = R
2
I
+
(L) (6.63b)
Using Eqs. (6.62) and (6.63b) in Eq. (6.63a), we nd that I
+1
(0), the intensity of
the right traveling wave at z = 0 after one pass, is related to its value I
+0
(0) the
same point initially, by
I
+1
(0) = R
1
I

(0) = R
1
e
(g
o
)L
I

(L) = R
1
e
(g
o
)L
R
2
e
g
o
L
I
+0
(0)
= R
1
R
2
e
2(g
o
)L
I
+
(0)
Such a relation can be established for any starting point z in the cavity. If the factor
R
1
R
2
e
2(g
o
)L
< 1, the steady-state intensity is zero because with each pass the
initial intensity is multiplied by a number smaller than 1. In this case no oscillation
can occur and the oscillator is said to be below threshold. If R
1
R
2
e
2(g
o
)L
> 1,
oscillations can build up and the laser is said to be above threshold. The threshold
of laser oscillations is then dened by
R
1
R
2
e
2(g
o
)L
= 1 . (6.64)
This equation means that the round trip gain e
2g
o
L
balances the round trip loss
e
2Lln(R
1
R
2
)
. Thus the minimum value g
o
must have before oscillation can buildup
is given by
2g
th
L = 2L lnR
1
R
2
. (6.65)
If the gain medium doesnt ll the whole space between the mirrors but only a
segment of length
m
, the above formula is modied to
g
th
=
1
2
m
[2L lnR
1
R
2
] (6.66)
For high reectivity mirrors, lnR
1
R
2
= ln(1 T
1
)(1 T
2
) (T
1
+ T
2
), so that
threshold gain is
g
th
=
1
2
m
[2L + T
1
+ T
2
]
L
i
+ L
T
2
m

L
c
2
m
(6.67)
This condition on unsaturated gain implies that inversion must have a minimum
value before laser action can occur. Recalling that
g
o
() =

2
o

21rad
8n
2
S()N
(o)
, (6.42

)
we nd that the minimum inversion to reach threshold is given by
N
th
=
L
c
2
m
8n
2

2
o

21rad
S()
=
L
c
2
m
1
()
. (6.68)
To minimize threshold inversion we look for narrow linewidth, strong radiative decay
on laser transition, long wavelength, low cavity losses and long gain length.
188 Laser Physics
Threshold gain and inversion in He:Ne laser : A 4-level laser system
Consider a two mirror cavity employing a high reector (R
1
= 99.8%) and an
output coupler (R
2
= 98.0%) and the gain medium lling the entire cavity length
(
m
= L = 50 cm). The gain medium is an electrical discharge in a 7:1 mixture of
He:Ne with Neon partial pressure (P
Ne
= 0.3 torr) and temperature T = 300 K.
Under these conditions the gain is predominantly Doppler broadened with 1/S(
o
) =
1.5 10
9
Hz. Radiative decay rate of the upper level for the = 633 nm transition
is
21rad
= 1.4 10
6
s
1
. Taking n 1 and losses L
i
0 and L
T
T
1
+T
2
= 2.2%,
we obtain
N
th
=
0.022
2 50

8 1.5 10
9
(0.633 10
4
)
2
1.4 10
6
cm
3
= 1.6 10
6
cm
3
.
This may seem like a large number of atoms but it is a small fraction of the total
number of Ne atoms
N = 9.66 10
18
0.3
300
cm
3
= 9.66 10
15
cm
3
.
Thus only a fraction
N
th
N
= 1.7 10
7
,
or one out of every ten million atoms needs to be inverted to reach threshold. This
inversion corresponds to a threshold gain coecient of
g
th
=
0.022
2 50
cm
1
= 2.2 10
4
cm
1
.
and threshold gain, in decibels (db), of
G
th
(db) = 4.34 g
th

m
= 4.34 2.2 10
4
50 = 0.048 .
Threshold gain and inversion in Ruby laser : A 3-level laser system
Naturally occurring Ruby is a crystal of Al
2
O
3
(corundum) in which some of the
Al
3+
ions have been replaced by Cr
3+
ions. For laser applications articial crystals
grown from a molten mixture of Al
2
O
3
with a small percentage of Cr
2
O
3
(0.05%
by weight corresponding to 1.6 10
19
cm
3
) are used. The colorless crystal formed
without the addition of Cr
2
O
3
is known as sapphire. The active species are Cr
3+
ions as impurities in a sapphire host crystal (n = 1.77).
At room temperature (300 K), the = 694 nm laser transition is predominantly
homogeneously broadened (phonon broadening) with
H
= 330 GHz. The upper
level radiative lifetime is
rad
= 3 ms. For standing wave cavity with a 5 cm
long Ruby rod, absorption losses of L
i
= 0.03, and mirror reectances of R
1
1,
R
2
= 0.96, we nd
N
th
=
L
c
2
m
8n
2

rad

H
2
=
0.07
2 5
8(1.77)
2
3.3 10
11
(0.694 10
4
)
2
333 2
cm
3
= 1.8 10
17
cm
3
.
Laser Pump and Threshold 189
This is much larger than that for the He:Ne laser and is a signicant fraction of
the total Cr
3+
ion concentration 1.6 10
19
ion/cm
3
. The reason for this dierence
is that the He:Ne laser has much larger stimulated emission cross-section which
results from a larger radiative decay to the lower laser level and narrower linewidth
compared to the Ruby laser.
6.6.1 Laser power output
Once the gain exceeds threshold value, oscillations can be sustained. In this case we
cannot ignore saturation. The equations for a standing wave cavity then become
dI
+
dz
= (g )I
+
(6.69a)
dI

dz
= (g )I

(6.69b)
If the end mirror reectivities are close to unity (high-Q cavity), I
+
(z) and I

(z)
remain nearly constant along the length of the cavity. We can then use the approx-
imation
I
+
(z) I

(z) I
c
(6.70)
in the gain coecient. Since both I
+
and I

will contribute to saturation we can

write the gain coecient as
g =
g
o
1 + (I
+
+ I

)/I
sat
=
g
o
1 + 2I
c
/I
sat
(6.71)
and treat it approximately a constant. Using this in Eqs. (6.69) and integrating
them from z = 0 to z = L we obtain
ln
I
+
(L)
I
+
(0)
= (g )L (6.72a)
ln
I

(0)
I

(L)
= (g )L (6.72b)
These dier from Eqs. (6.62) only in that unsaturated inversion g
o
has been replaced
by saturated inversion g. Adding these and using the boundary conditions (6.63)
we obtain
ln
I
+
(L)I

(0)
I

(L)I
+
(0)
= 2(g
m
L)
lnR
1
R
2
= 2(g
m
L) (6.73)
Applying this to a homogeneously broadened line we obtain
2L lnR
1
R
2
=
2g
o

m
1 + 2I
c
/I
sat
(6.74)
190 Laser Physics
The quantity on the left hand side of this equation is simply the total cavity loss
L
c
= L
i
+ L
T
L
i
= 2L
L
T
= lnR
1
R
2
R
1
, R
2
1
T
1
+ T
2
As written, Eq. (6.74) say that in the steady-state saturated laser gain is clamped
at its threshold value L
c
. The laser will adjust its intensity such that its round trip
gain just compensates for the losses it encounters in one round trip. Solving Eq.
(6.74) for the circulating intensity I
c
we obtain
I
c
=
I
sat
2
_
2g
o

m
L
c
1
_
=
I
sat
2
_
2g
o

m
2g
th

m
1
_
=
I
sat
2
(r 1) , (6.75)
where r is the pump ratio given by
r =
g
o
g
th
=
N
(o)
N
th
. (6.76)
In terms of r the laser threshold is at r = 1. Pump ratio r is the ratio of laser pump
rate to the pump rate needed to reach the threshold. The circulating intensity
depends linearly on r and since laser intensity cannot be a negative quantity, laser
pump ratio r must exceed 1.
Once the circulating intensity is known, the output from the laser is written
simply as the intensity transmitted through the mirrors
I
o
= L
T
I
c
= T
I
sat
2
_
2g
o

m
T + L
i
1
_
(6.77)
The dependence of output intensity on transmission is not linear. We can expect
an increase in the output as transmission increases but as transmission increase
overall cavity losses and therefore the minimum gain needed to reach threshold also
increases. This increased threshold gain lowers the pump ratio and, therefore, the
output. Hence, for a given gain 2g
o

m
there must be an optimum output coupling
for maximizing laser output [See Fig. (6.13)]. This optimum can be found from Eq.
(6.77) by dierentiating I
o
with respect to T
dI
o
dT
= 0 =
I
sat
2
_
2g
o

m
L
i
+ T
1
_
T
I
sat
2
2g
o

m
(L
i
+ T )
2
0 =
I
sat
2(L
i
+ T )
2
_
2g
o

m
(L
i
+ T ) (L
i
+ T )
2
T 2g
o

=
I
sat
2(L
i
+ T )
2
_
2g
o

m
L
i
(L
i
+ T )
2

Solving for optimum output coupling we nd

T
opt
=
_
2g
o

m
L
i
L
i
(6.78)
Laser Pump and Threshold 191
0
1
2
3
4
5
0 2 4 6 8
2
I
o
/
I
s
a
t
T
T
opt
FIGURE 6.13
Laser output intensity as a function of output coupling for xed gain 2g
o

m
= 10%
and internal losses L
i
= 2%.
Rewriting this as
L
i
+ T
opt
=
_
2g
o

m
L
i
and using the result in Eq. (6.77) the optimum output from the laser is given by
[I
o
]
opt
=
I
sat
2
_
_
2g
o

_
L
i
_
2
(6.79)
Exercise 1: How does the pump ratio vary with frequency for (i) a homogeneously
broadened gain medium, (ii) inhomogeneously broadened gain medium?
Homogeneous r =
g
o
(
o
)
g
th
1
1 + [2(
o
)/
H
)
2
]
=
r
o
1 + [2(
o
)/
H
)
2
]
Inhomogeneous r =
g
o
(
o
)
g
th
exp
_

_

o

D
_
2
4 ln2
_
= r
o
exp
_

_

o

D
_
2
4 ln2
_
,
where r
o
= g
o
(
o
)/g
th
is the pump ratio at the line center.
Exercise 2: Find the output intensity for a single-mode unidirectional ring laser.
From
g
o
()
m
1 + I/I
s
= L + T = g
th

m
we obtain
I
o
= T I
s
_
g
o
g
th
1
_
= I
s
(r 1)
192 Laser Physics
Exercise 3: For an inhomogeneous line saturated gain is given by
g() =
g
o
()
_
1 + I/I
s
where g
o
() is unsaturated gain. Find the output intensity from the laser as a
function of laser pump ratio for an inhomogeneously broadened gain medium.
Assuming a running wave cavity, the output as a function of laser pump ratio r is
I
o
= L
T
I
s
2
(r
2
1) (6.80)
Formally, the same expression holds for a standing wave cavity at line center.
6.7 Operating Frequency
Presence of gain (and loss) also modies the operating frequency of the laser. Con-
sider a cavity of length L lled with a medium of refractive index n and a gain
medium of length
m
. Then the operating frequency of the laser
p
(p
th
mode) is
determined by the resonance condition
2
n2
p
c
_
_
L
m
+
m

1 +

at
(
p
)
n
2
_
_
= 2p (6.81)
This condition is equivalent to saying that after one round trip around the cavity,
the accumulated phase change 2kL must equal an integer multiple of 2. The
condition for laser frequency then becomes

p
_
_
1

m
L
+

m
L

1 +

at
(
p
)
n
2
_
_
=
pc
2nL

(o)
p
(6.82)
In general, this leads to a complicated equation for the operating frequency of the
laser, which must be solved numerically or graphically. We can, however, gain some
insight into the change that gain brings about in the operating frequency of the
laser without going through a complicated analysis if we use the fact that for most
lasers

at
is small compared with the host medium refractive index n. Then we can
use binomial expansion for the quantity under the radical in Eq.(6.82). The laser
frequency equation becomes

p
_
1 +

m

at
(
p
)
2n
2
L
_
=
(o)
p
(6.83)
Laser Pump and Threshold 193
This relation can be rewritten by using the relation between the real and imaginary
parts of the atomic susceptibility
5

at
() =
2(
o
)

at

at
() =
_
2(
o
)

at
_
g()nc
2
,
where g() =

at
()
nc
=
2

at
()
nc
and
at
is given by

at
=
_

H
Homogeneous gain broadening

D
_

4 ln 2
, Inhomogeneous gain broadening
(6.84)
Substituting this in the frequency equation (6.83) we obtain

p
_
1
_

at
_
cg(
p
)
m
2n
p
L
_
=
(o)
p
. (6.85)
For most lasers gain per pass g
m
is small, being of the order of a few percent.
In such cases the operating frequency is expected to be close to the bare cavity
frequency
(o)
p
. Then replacing
p
by
(o)
p
in the gain coecient, we nd that the
operating frequency of the laser is given by

(o)
p
1
_

(o)
p

at
_
cg(
p
)
m
2n
(o)
p
L

(o)
p
+
_

(o)
p

at
_
cg(
(o)
p
)
m
2nL
. (6.86)
Subtracting both sides of this equation from line center frequency
o
we can express
the separation
o

p
of the operating frequency of the laser from the line center
in terms of the separation
o

(o)
p
of bare cavity frequency from line center as

p
= (
o

(o)
p
)
_
1
cg(
(o)
p
)
m

at
2nL
_
(6.87)
We note that the factor inside square brackets on the right hand side of this equation
is less than unity for gain (g > 0) and greater than unity for absoption (g < 0).
Then this equation says that if a bare cavity mode
p
experiences gain its operating
frequency (dressed cavity frequency) is closer to the atomic frequency than the
bare cavity mode frequency. This phenomena is known as frequency pulling [Fig.
6.14a].
On the other hand, if the cavity mode experiences absorption (g < 0), the op-
erating frequency (dressed cavity frequency) of a mode experiencing absorption is
farther from the atomic frequency than its bare cavity frequency. This phenomena
is known as frequency pushing [Fig. 6.14b].
5
Such a relation holds for a homogeneously broadened gain line. A similar relation holds for an
inhomogeneous line if
at
=
D
p
/4 ln 2.
194 Laser Physics

o

p
g()

(0)
(a)

p

o

p
g()

(0)
(b)

p
FIGURE 6.14
cavity frequency (a) pulling in the presence of gain (g > 0) and (b) pushing in the
presence of absorption (g < 0).
For a laser operating above threshold (g > g
th
), Eq. (6.85) can be cast in a
dierent form. In the steady-state the saturated gain of the laser is clamped at its
threshold value
2g
m
= 2g
th

m
= L
c
=
_
cL
c
2nL
_
2nL
c
=
1

c
2nL
c
= 2
c

2nL
c
,
where
c
is the FWHM for the cavity transmission prole. Using this in the
frequency equation and assuming that the gain medium lls the whole cavity (
m
=
L), we obtain

p
_
1
(
o

p
)

at
_
=
(o)
p
. (6.88)
This equation can be rewritten as

p
=

(o)
p

at
+
o

at
+
c
. (6.89)
In terms of Q-factors for the atomic transition and the cavity
Q
at
=
o
/
at
, Q
c
=
(o)
p
/
c
, (6.90)
we can write the cavity operating frequency
p
as

p
=

(o)
p
Q
c
+
o
Q
at
Q
c
+ Q
at
. (6.91)
According to this equation, dressed cavity frequency is the weighted average of bare
cavity frequency
(o)
p
and atomic frequency
o
. It lies between
(o)
p
and
o
, being
closer to whichever frequency has larger Q associated with it. Usually Q
at
Q
c
so
that the operating frequency is given approximately by

p

(o)
p
+ (
o

(o)
p
)
Q
at
Q
c
=
(o)
p
+ (
o

(o)
p
)

c

at
(6.92)
in agreement with Eq. (6.86).
Laser Pump and Threshold 195
Estimates of frequency pulling
For a He:Ne laser with threshold gain of g
th
= 0.001 cm
1
,
D
= 1500, = 633
nm, n = 1, the gain medium lling the whole cavity (
m
= L) and recalling

c
=
gc
2n
,
at
=
D
_

4 ln2
we obtain

d
=
gc
2n
D
_
4 ln2

=
10
3
(3.00 10
10
)
2 1.5 10
9

_
4 ln2

= 3.0 10
3
= 0.3%
Thus the corrections are only about 0.3% of the dierence |
o

(o)
p
|.
For the = 3.39 m transition with
D
= 280 MH and g = 0.03 cm
1
, we
obtain

D
=
3 10
2
(3.00 10
10
)
2 2.8 10
8

_
4 ln2

= 0.48 = 48%
In this case the correction is substantial being approximately 48% of the dierence
|
o

(o)
p
|.
Finally, for a high gain laser such as a He:Xe laser operating at = 3.51 m,

D
= 100 MHz, g = 0.5 cm
1
we obtain

D
=
0.5 (3.00 10
10
)
2 10
8

_
4 ln2

= 22
For such large ratios, cavity mode structure is greatly perturbed and the pertur-
bative analysis presented here cannot be applied. We then have to go back to Eq.
(6.85) to determine the mode structure.
Both strongly absorbing and amplifying media change the cavity mode structure
and novel eects can arise. For example, it can happen that several frequencies
corresponding to a single wavelength can exist in the cavity. In such cases analysis
is further complicated by spectral hole burning and mode competition eects.
See, for example: Longitudinal modes in a high gain laser, Lee Casperson and A.
Yariv, Appl. Phys. Lett. 17, 259 (1970)].