Lasers; II

9.1 Introduction
In the last chapter we showed that for a medium to behave as an
amplifier of light waves, one must create a state of population inversion
between two energy levels of the system. In order to convert this amplifier
to an oscillator i.e., a source of radiation, one must provide optical feedback
to the amplifier. This is brought about by a pair of mirrors between which
is enclosed the amplifying medium. This pair of mirrors forms what is
known as the optical resonator. Fig. 9.1 shows a simple optical resonator
formed by a pair of plane mirrors. Since the sides of the resonator cavity
are open such resonators are also known as open resonators. The resonators
can only support certain specific field configurations and certain specific
oscillation frequencies. These are referred to as the modes of oscillation of
the resonator. The need for an open resonator rather than a closed resonator
(as in microwaves) arises because of the fact that the number of modes
that a closed resonator can support within the frequency band in which
the amplifying medium can provide gain is so large that the output from
such a system would be far from monochromatic. By removing the side
walls of a closed cavity and thus obtaining an open cavity, one increases
the loss of most of the modes to such an extent that they cannot be sustained
in the cavity. Thus one can achieve oscillation at a very few or even a

Fig. 9.1 A simple optical resonator formed by a pair of plane mirrors facing
each other.

M.

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 246 Lasers: II

single frequency. In this chapter we first obtain the modes of a dosed cavity
and show that the number of modes within the oscillating linewidth is very
large. In Sec. 9.3 we discuss the quality factor of a resonator and in Sec. 9.4
we obtain the ultimate linewidth of a laser. Sec. 9.5 discusses various mode
selection techniques and Sees. 9,6 and 9.7 discusses Q-switching and mode
locking in lasers. In Sec, 9.8 we obtain the modes in resonators formed by
spherical mirrors.

9.2 Modes of a rectangular cavity and the open planar resonator

We consider a closed rectangular cavity of dimensions 2a x 2b x d
with perfectly conducting walls (see Fig. 9.2), The field inside the cavity
must satisfy the wave equation (see Eq, (1.16)) given by
d2s
(9.1)

where c represents the velocity of light in the medium filling the cavity.
Since the walls of the cavity are assumed to be perfectly conducting, the
tangential component of the electric field must vanish at the walls. If n
represents the unit normal to the wall surface then we must have
#xi =0 (9.2)
In Appendix A we have obtained the solution of the wave equation by
using the separation of variables technique. The time dependence is of the
form e tot where co is the angular frequency of the wave. By considering
each Cartesian component of Eq. (9.1), one can show that the x,y and z
dependences are linear combinations of sine and cosine functions. Thus we

Fig, 9.2 A closed rectangular cavity bounded by perfectly conducting walls.

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.2 Modes of a rectangular cavity 247

may write
£x = (A 1 sin kxx + Bt cos kxx)(A2 sin kyy + B2 cos kyy)
x (A3 sinfczz+ B3 cos fczz)ebjl (93)
where kx, ky and lcz are the x, y and ^-components of the propagation vector
k and
/c = KX -t- fc -\- kz = cy /c
Now, since Sx is a tangential component on the planes y = 0, y = 2b,
z = 0 and z = 4 it has to vanish on these planes. Thus we have from Eq. (9.3)
B2 = 0, B3 = 0 (9.4)
sin(ky2h) = 0 (9.5)

sin(M) = 0 (9.6)
The last two equations imply
ky = nn/2b, kz = qn/d; n,q = 0,\,2,... (9.7)
where we have intentionally included the value 0, which, in this case, would
lead to the trivial solution of £x vanishing everywhere. In a similar manner
we can show that the x and z dependences of £y will be sin kxx and sin kzz,
respectively with
kx = mn/2a; m = 0,1,2,... (9.8)
and kz again given by Eq. (9.7). Similarly the x and y dependences of £z
would be sin kxx and sin kyy.
Now, due to the above forms of the x dependences of £y and £z, dSJdy
and dSJdz will vanish on the surface x = 0 and x = 2a, Hence on the planes
x = 0 and x = 2a, the condition ¥ - # = 0 leads to dSJdx = 0. Hence the x
dependence of £x most be of the form cos kxx with kx given by Eq. (9,8).
Notice that the case m = 0 now corresponds to a nontrivial solution.
In a similar manner, one can obtain the solutions for £y and £r The
complete solution inside the cavity can be written as
Ex = EOx cos kxx sin kyy sin kzz (9.9)

Ey = EOy sin kxx cos kyy sin kzz (9.10)

Ez ~ EOz sin kxx sin kyy cos fczz (9.11)

where EQx, EOy and JSOz are constants. Since fex, ky and fcz are given by

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 248 Lasers: II

Eqs. (9.7) and (9.8), the allowed frequencies of oscillation in the cavity are
m n q
a) = ck — en — — s - - j - •-••
2 2f -) - (9.12)
Aa Ah d

The field configurations given by Eq. (9.9)—(9.11) are called the modes of
the cavity and correspond to standing wave patterns in the cavity.
If we use Eqs. (9.9)—(9.11) along with the equation V-«f = 0, we will have

where k = kxt + h¥$ + kz% and E o = EOxt + EOy$ + £ Oz i. Since the coef-
ficients EOxi EOy and EOz have to satisfy Eq. (9,13), it follows that for a
given mode, i.e., for given values of m, n and q, only two of the components
of E § can be chosen independently. Thus a given mode can have two
independent states of polarization.

Example: As a specile example, we consider a mode with

Thus kx = 0fky = TI/26,fcz= n/d and using Eqs. (9.9H9.11), we have

Ex = EOxsinkyvsinkzz = ^yjl sin^z

(9.14)
£, =
Using the time dependence of the form eimt and expanding the sine functions into
exponentials, we may write
1
i

(9.15)
Thus the total field inside the cavity has been broken up into four propagating

Fig. 9.3 A mode corresponding to m = 0, n=t,q = l consists of four plane

waves (whose directions of propagation are shown as 1-4) and whose mutual
interference produces a standing wave pattern in the cavity.

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.2 Modes of a rectangular cavity 249

plane waves (see Fig. 9,3); in Eq. (9.15) the irst term on the RHS represents a wave
propagating along the ( + y, + indirection, the second along the ( + y, — z)
direction, the third along the (— y, + z) direction and the fourth along the
(— y, — z) direction, These four plane waves interfere at every point inside the
cavity to produce a standing wave pattern. However, since ky and kz take discrete
values, the plane waves which constitute the mode make discrete angles with the
axes.
As another example if we take a = b — 1 cm, d = 20 cm, and consider the mode
with
m = 0, n=l, ^=106
then
kx = 0, ky = n/2 cm " l , kz = 1 06TC/20 cm " l

implying
k « 10 6 7i/20cm^ l and v = ck/2n = 7.5 x 10 14 Hz
which lies in the optical region. For such a case
0y. = cos^ l (ky/k) % 89.9994°
0Z = co&^1 (kjk) = cos~ l {[1 + (fcy/Jy2] " *} « 0.0006°
and ^ = 0 because of which dy + dz = 90°. It may be noted that the component
waves are propagating almost along the z-axis. In general
cos 2 0X + cos 2 9y + cos 2 Bz = 1
Further, for m # 0, n # 0, q # 0, the cavity mode can be thought of as a standing
wave pattern formed by eight plane waves with components of k given by (± kx,
+ fcv, +kz).
Using the above formulation, we show in Appendix. C that the number
of modes per unit volume of the cavity in the frequency interval v to v + dv
is
p(v)dv = (87iv2/c3)dv (9.16)
Thus, if we take for dv a typical value of the linewidth of an atomic system
such as dv = 3 x 109 Hz at v = 3 x 10 14 Hz, then the number of modes per
unit volume will be
8 x 3.14 x l(3 x 10 14 ) 2 „ inQ 9 „ <rt88 ,
p(v)dv = x t olo3 ^ x 3 x l O « 2 x l0 cm^ 3 (9.17)

Thus for cavities of practical dimensions, the number of oscillating modes

which would lie within the atomic linewidth will be extremely large. Thus
all these oscillating modes will draw energy from the atomic system and
the resulting emission will be far from monochromatic. In order to have
very few oscillating modes within the cavity, the dimensions of the cavity

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 250 Lasers: II

should be chosen to be of the order of the wavelength of the radiation.

This is, of course, impractical at visible or infrared wavelengths.
The problem of the extremely large number of oscillating modes can be
overcome by the use of open cavities which consist of two plane or curved
mirrors facing each other. As we have seen earlier a mode can be considered
to be a standing wave pattern formed between waves propagating within
the cavity with k given by (±kx, ± kp ± kz); here kx, ky and kz represent
respectively the x, y and z-components of the propagation vectors of the
component plane waves. Thus the angles made by the component plane
waves with the x,y and z-directions will respectively be cos" l (ml/la),
CQ$^l(n/,/2b) and cos~ l (qlfd). Since in open resonators, the side walls of
the cavity has been removed, those modes which are propagating almost
along the z-direction (i.e., with a large value of q and small values of m
and n) will have a loss which will be much less then the loss of the modes
which make a large angle with the z-axis (i.e., large values of m and/or n).
Thus on removing the side walls of the cavity, only modes having small
values of m, n ( ~ 0,1,2,...) will have a small loss and thus as the amplifying
medium placed inside the cavity is pumped, only these modes will be able
to oscillate. Modes with large values of m and n will have a very large loss
and thus will not be able to oscillate.
It should be noted here that since the resonator cavity is now open, all
modes are lossy. Thus even the modes that have plane wave components
travelling almost along the z-axis will suffer losses. Since m and n specify
the field patterns along the transverse directions x and y, and q that along
the longitudinal direction z, modes having different values of (m, n) are
referred to as various transverse modes and modes differing in q values as
various longitudinal modes.
The oscillation frequencies of the various modes of the closed cavity are
given by Eq. (9.12), In order to obtain an approximate value for the
oscillation frequencies of an open cavity, we may again use Eq. (9.12) with
the condition n%n<^q. Thus making a binomial expansion in Eq. (9.12),
we obtain

q fm2 n2 g

The difference in frequency between two adjacent modes having same values
ofTOand n and differing in q by unity will be very nearly given by
Avqxc/2d (9.19)
which corresponds to the longitudinal mode spacing. In addition if we

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.2 Modes of a rectangular cavity 251

completely neglect the term containing m and «» we will obtain

vq»(c/2d)q (9.20)
The above equation is similar to the frequencies of oscillation of a stretched
string of length d.

Example: For a typical laser resonator d % 100 cm and assuming free space filling
the cavity, the longitudinal mode spacing comes out to be ~ 150 MHz.

Problem 9.1: Show that the separation between two adjacent transverse modes is
much smaller than Av r

Solution: The frequency separation between two modes differing in m values by

unity will be

Xd

5
where we have used q % 2d/X (see Eq, (9.20)), For typical values / = 6 x 10 cm,
d = 100cm, a = 1 cm; Xd/M2 = 7.5 x 10" 3, Thus for m m 1, Avm « Av r

It is of interest to mention that an open resonator consisting of two

plane mirrors facing each other is, in principle, the same as a Fabry-Perot
interferometer or etalon (see Sec. 2.4), The essential difference in respect of
the geometrical dimensions is that in a Fabry-Perot interferometer the
spacing between the mirrors is very small compared to the transverse
dimension of the mirrors while in an optical resonator, the converse is true.
In addition, in the former radiation is incident from outside while in the
latter, the radiation is generated inside the cavity.
Earlier we showed that the modes in closed cavities are essentially
superpositions of propagating plane waves. Because of diffraction effects,
plane waves cannot represent modes in open cavities. Indeed, if we start
with a plane wave travelling parallel to the axis from, one of the mirrors,
it will undergo diffraction as it reaches the second mirror and since the
mirror is of finite transverse dimension, the energy in the diffracted wave
which lies outside the mirror would be lost. The wave reflected from the
mirror will again undergo losses when it is reflected from the irst mirror.
This loss is termed the diffraction loss. Fox and Li (1961) performed a
numerical calculation of such a planar resonator. The analysis consisted
of assuming a certain field distribution at one of the mirrors of the resonators
and calculating the Fresnel diffracted field at the second mirror. The field

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 252 Lasers: II

reflected by the second mirror is used to calculate back thefielddistribution,

on the first mirror. It was shown that after many traversals the transverse
field distribution settles down to a steady pattern i.e., it does not change
between successive reflections but only the amplitude of the ield decays
exponentially in time due to diffraction losses. Such a field distribution
represents a normal mode of the resonator and by changing the initial field,
distribution on the first mirror, other modes can also be obtained. For
more details readers are referred to Fox and Li (1961); a brief analysis is
also given in Thyagarajan and Ghatak (1981).
The use of curved mirrors instead of plane mirrors in the resonator has
the advantage of having smaller diffraction losses. In fact, if the mirrors
are sufficiently large in the transverse dimensions, the diffraction losses can
almost be neglected. In Sec. 9.8 we will show that the modes of spherical
mirror resonators are Hermite-Gauss functions.

9.3 The quality factor

Since an optical resonator is an open cavity, all modes suffer losses.
These losses arise from thefinitesizes of the mirrors (at the cavity ends) due
to which there is diffraction spillover, the finite reflectivities of the mirrors
and scattering and absorption by the medium filling the resonator cavity.
In Sec. 8.4 we related the various losses to the cavity lifetime tG, One can
also describe the losses in the cavity by what is known as the quality factor.
This is defined by
_ energy stored in the mode ,n „.,
w
n —m , _ (9.22)
energy lost per unit time
where coo is the oscillation frequency of the mode. If W(t) represents the
energy in the mode at time t, then from Eq. (9.22), we obtain
W{t)
n „w
* °^dW/df
or
d W7df = — (a) JO) Wit) (9 23)
whose solution is
'mt'Q (9.24)
Thus if te represents the cavity lifetime i.e., the time in which the energy
in the mode decreases by a factor 1/e, then

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.3 The quality factor 253

We can write for the field associated with the mode

E(t) = EoQimote ~m°tf2Q (9.26)
The frequency spectrum of this wave train can be obtained in a manner
similar to that used in Sec. 8,8 to obtain the spontaneous emission spectrum
and it comes out to be

which again represents a Lorentzian. The FWHM of the spectrum is

Av, = vo/Q (9.28)
Thus the line width of the passive mode depends inversely on the quality
factor. The higher the quality factor (i.e., the longer the cavity lifetime, see
Eq. (9,25)), the smaller will be the FWHM.
In Sec. 8.4 we derived an expression for te in terms of the length of the
resonator and the reflectivity of the mirrors. Thus using Eqs. (8.28), (9.25)
and (9.28), we have

Q=
and
j(2a,J-lnK1K2) (9.30)

Example; Let us consider a typical cavity of a He Ne laser with the following

specifications:
» o ssl, R = 1, i? 2 =0.98, a, s^O (9.31)
For such cavity
Avp«2.4MHz (9.32)
For the same cavity, the frequency separation between adjacent longitudinal
modes is
(9.33)
Thus the width of each mode is smaller than the separation between adjacent modes.
Example; As another example we consider a GaAs semiconductor laser (see Sec.
10.9) with the following values of the various parameters:
d = 500nm, n0 = 3.5, Rt=R2= 0.3, ac « 0 (9.34)
For such a cavity we obtain
Avp%3.3x 1010Hz (9.35)

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 254 Lasers: II

9.4 The ultimate linewidth of the laser

One of the most important properties of a laser is its ability
to produce light of high spectral purity or high temporal coherence. The
finite spectral width of a laser operating continuously in a single mode
is caused by two mechanisms. One is the external factors which tend to
perturb the cavity, for example, temperature fluctuations, vibrations, etc.
randomly alter the oscillation frequency which results in a inite spectral
width. The second more fundamental mechanism which determines the
ultimate spectral linewidth of the laser is that due to the ever present
random, spontaneous emissions in the cavity. Since spontaneous emission
is completely incoherent with respect to the existing energy in the cavity
mode, it leads to a finite linewidth of the laser. In this section, we shall
give a heuristic derivation for this ultimate linewidth of the laser (Gordon,
Zeiger and Townes, 1955 and Maitland and Dunn, 1969).
In order to obtain a value for the ultimate laser linewidth, we assume
that the radiation arising out of spontaneous emission represents a loss as
far as the coherent output energy is concerned. This loss will then lead to
a finite linewidth for the laser. We recall from. Sec. 8.6 that the number of
spontaneous emissions per unit time into a mode of the cavity is given by
KN2 (where K is defined by Eq. (8.98)) and N2 represents the number of
atoms/unit volume in the upper laser level. We are assuming that N%^0.
When the laser oscillates in a steady state then we know from Eq. (8.105)
that N2 * 1/Ktc where we are assuming n » 1 and tc is the passive cavity
lifetime. Thus above threshold, the number of spontaneous emissions per
unit time will be KN2 = i/tc. Thus the energy appearing per unit time in
a mode due to spontaneous emission will be hvo/tG where v0 is the
oscillation frequency of the mode.
Now, the total energy contained in the mode is nhv0 and since the output
power Pmt is given by Pom = nhvo/tcy the energy contained in the mode is
P t
We now use Eq. (9,28) and denote the linewidth of the oscillating laser
caused by spontaneous emission by <5vsp to obtain

Q = p - = 2nv0 p ^ = l7t^f^ (936)

Now if Avp is the passive cavity linewidth then tc = 1/2EA?P and thus from
Eq. (9.36) we obtain

ppfhv0/Pom (9.37)
The above equation gives the ultimate linewidth of an oscillating laser and

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.4 The ultimate line width of the laser 255

is similar to the one given by Schawlow and Towoes (1958). It is interesting

to note that Svsp depends inversely on the output power Pouv This is
physically due to the fact that for a given, mirror reflectivity, an increase
in P ou , corresponds to an increase in the energy in the mode inside the cavity
which in turn implies a greater dominance of stimulated emission over
spontaneous emission. Fig. 9.4 shows a typical measured variation of the
linewidth of a GaAlAs semiconductor laser which shows the linear increase
of linewidth with inverse optical power.
The above derivation is rather heuristic; an analysis based on the random
phase additions due to spontaneous emission is given by Jacobs (1.979)
which gives a result half that predicted by Eq. (9.37).
Example; Let us first consider the He- Ne laser given in Eq. (9.31) and we assume
that it oscillates with an output power of 1 mW at Xo ~ 6328 A. The spontaneous
emission linewidth of the laser will be
0.01 Hz (9.38)
which is extremely small. To emphasize how small these widths are, let us try to
estimate the precision with which the length of the cavity has to be controlled in
order that the oscillation frequency changes by 10~ 2 Hz. We know that the
approximate oscillation frequency of a mode is v = qc/2d. Thus the change in
frequency Av caused by a change in length Ad is
Av = (v/d)Ad (9.39)
Using d = 20 cm, Ao = 6328 A and Sv x 0,01 Hz, we obtain
Arf»4 x 10^ 14 cm (9.40)
which corresponds to a stability of less than even nuclear dimensions!
Fig. 9.4 Experimentally obtained variations of laser linewidth in a single
frequency GaAlAs semiconductor diode laser as a function of inverse output
power. (After Mooradian (1985).)
200 r

0.5 1.0 1.5 2.0 2.5

Inverse power (mW"1)

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 256 Lasers: II

Example; As another example, we consider a GaAs semiconductor laser operating

at Xo = 0.85 /zm with an output power of 1 mW with cavity dimensions as given by
Eq, (9.34), For such a laser
Svspv 1.6MHz (9.41)
which is much larger than that of a He^Ne laser. This is primarily due to the very
large value of Avp in the case of semiconductor laser. Since Avp can be reduced
by increasing the length of the cavity, one can use an external mirror for feedback
and thus reduce the linewidth. For a more detailed discussion of semiconductor
laser linewidth, readers are referred to Mooradian (1985).

9.5 Mode selection

Since optical resonators have dimensions which are large compared
to the optical wavelength, there are, in. general, a large number of modes
which fall within the atomic linewidth and which can oscillate in the laser.
Hence the output may consist of various transverse and longitudinal modes.
In this section we shall describe some techniques which are used to select
a single transverse and a single longitudinal mode to oscillate in the laser.
For more details readers are referred to Smith (1972).

9.5.1 Transverse mode selection

We discussed in Sec. 9.2 that different transverse modes are
characterized by different transverse amplitude distributions. In Sees 9.8
and 9.9 we shall show that for a stable spherical resonator, the transverse
field distributions at the waist of the various modes are Hermite-Gauss
functions given by
Emn(^y) = E0Hmy2x/w0)Hny2y/w0)e <* a + ^o (9 .42)

where m,n represent the transverse mode numbers, Hm(yj2x/wQ) and

Hn{^/2y/w0) represent Hermite polynomials (see Eq. (11.117)) and w0 is
the characteristic mode width which depends on the wavelength of
operation and the resonator dimensions such as the radii of curvatures of
the two mirrors and the distance between them. Fig. 9.5 shows a photograph
of various transverse modes of a laser.
The fundamental transverse mode corresponds to m — 0 n = 0 for which
we get
Ew{xyy) = EQ^2^^< (9.43)
Thus, the fundamental mode of spherical resonators is Gaussian with a
1/e amplitude radius w0 at the waist. (In Sec. 5.4 we discussed in detail
the propagation of a Gaussian beam through free space.) Most applications

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9,5 Mode selection 257

Fig. 9.5 Photograph showing some of the lower order transverse modes of an
optical resonator, TEMmB corresponds to the \m. njth transverse mode and
TEM 0 0 is the fundamental mode with Gaussian distribution (After Kogelnik
and Li (1966). Photograph courtesy Dr. H. Kogelnik.)

:
mi- M&/--'-:--:l-.:-:t-:.': "'•<•

TEM M TEM, TEM 9

TEM, TEM 4 t TEM $0

TEM. TEM,, TEM U

TEM» TEM, TEM,

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 258 lasers: II

of lasers, such as holography, rangefinding, nonlinear optical experiments,

etc., require the laser to oscillate in its fundamental transverse mode which
is a Gaussian field distribution. This is because it is the fundamental mode
which produces the minimum beam divergence, the highest power density
and is also uniphase across its wavefront. In addition it is the fundamental
transverse mode which produces the largest power per unit area in a
focussed diffraction limited spot. In view of the above it is usually necessary
to restrict oscillation of the laser only in its fundamental transverse mode.
Since the fundamental Gaussian mode is the mode with the narrowest
transverse dimensions, an aperture placed inside the resonator will preferen-
tially introduce higher losses for higher order modes. Thus if a circular
aperture is introduced into the laser such that the loss suffered by all higher
order modes is greater than the gain while the loss suffered by the
fundamental mode is still lower than the gain, then one will have a single
transverse mode oscillation.
Fig. 9.6 shows a typical configuration for producing single transverse
mode oscillation (Davis, 1968) in a ruby laser. For the resonator shown in
Fig. 9,6, the diffraction loss for a single pass for the fundamental mode is
~ 20% and for the first excited mode is about 50%. Thus the aperture
introduces much higher losses for the higher order modes as compared to
the fundamental mode.
It is interesting to note that specific higher order transverse modes can
also be selected by choosing complex apertures which introduce high loss
for all modes except the required mode or by profiling the reflectivity of
one of the mirrors to suit the desired mode. Thus a thin wire placed normal
to the axis passing through the resonator axis will select the TEM01 mode.
9,5,2 Longitudinal mode selection
We have seen in Sec. 9.2 that the various longitudinal modes
corresponding to a transverse mode are approximately separated by c/2d,

Fig. 9.6 A typical configuration for producing only fundamental transverse

mode operation of the laser.

J?= co M= 10 m

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9,5 Mode selection 259

where d is the separation between the mirrors of the cavity. Hence if we

consider a 50 cm long He-Ne laser operating at A = 6328 A and having an
oscillating bandwidth of 1500 MHz (which is approximately the FWHM of
the gain profile - see Sec. 8.8) then the longitudinal mode spacing will be

Thus even if the laser is oscillating only in its fundamental transverse mode,
there will still be five longitudinal modes which can oscillate in the laser
(see Fig. 9.7). Thus the output will consist of ive adjacent frequencies and
will have only a short coherence length (see Problem 9.3). Thus in
applications where a long coherence length is necessary (e.g., in holography,
interferometry etc.) or where a well defined frequency is required (e.g., in
spectroscopy) one will require single longitudinal mode operation of the
laser in addition to its single transverse mode oscillation.

Fig. 9.7 (a) The longitudinal mode spacing of a resonator of length d is c/2d.
For an oscillating bandwidth of 1500 MHz and an intermode spacing of
300 MHz, five different longitudinal modes can oscillate, (h) If d <c/Sv and
there is a mode at the line centre then one can have single frequency oscillation
of the laser.
c/2d

I \ C

\,5v

Threshold

Modes with
gain > loss

(a)

— | c/2d

~ A
1 k
\ r Gain

\ Thresholc

Oscillating
mode

(b)

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 260 Lasers: II

Referring to Fig. 9.7 we can have a simple method of obtaining single

longitudinal mode oscillation by reducing the cavity length to a value such
that the intermode spacing is larger than the spectral width over which
oscillation can take place. Thus if the oscillating bandwidth is <5v (see
Fig. 9.7) then for single mode oscillation the cavity length must be such that
c/2d > Sv (9,45)
For a He-Ne laser, <5v» 1500 MHz and for single mode oscillation we
must have
d<c/2Sv*l0cm
We should note here that if one can ensure that a resonant mode exists
at the centre of the gain profile then single mode oscillation can be obtained
even with a cavity length of ~ c/dv (see Fig. 9.7(b)).
One of the major drawbacks with the above method is that since the
volume of the active medium gets very much reduced due to the restriction
on the length of the cavity, the output power is small In addition, in solid
state lasers where the bandwidth is large, the above technique becomes
impractical. Hence other techniques have been developed which can lead
to single frequency oscillations without restricting the length of the cavity
and hence are capable of high powers.
Oscillation of the laser in a given resonant mode can be achieved by
introducing frequency selective elements such as Fabry-Perot etalons (see
Sec. 2.4) into the laser cavity. The element should be so chosen that it
introduces losses in all but the required mode so that their losses are larger
than the gain. Fig. 9.8 shows a tilted Fabry-Perot etalon placed inside the
resonator. The etalon consists of a pair of highly reflecting parallel surfaces
which has a transmission versus frequency as shown in Fig. 9.9. As discussed
in detail in Sec. 2.4 such an etalon has transmission peaks centred at
frequencies given by
c
p — a n y mteger

Fig. 9.8 A laser resonator with a Fabry-Perot etalon placed inside the cavity
for the selection of single longitudinal mode oscillation.

''-0
I'abrv Perot etalon

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9,5 Mode selection 261

where t is the thickness of the etalon and n the refractive index of the
medium between the reflecting surfaces. The width of each peak depends
on the reflectivity of the surfaces; the higher the reflectivity, the sharper
are the resonance peaks1" (see Fig. 9.9). The frequency separation between
two adjacent peak transmittances is
Av = c/2nfcos0 (9,47)
which is also referred to as the free spectral range of the etalon.
If the etalon is so chosen that its free spectral range is greater than the
spectral width of the gain proile then the Fabry-Perot etalon can be tilted
inside the resonator so that one of the longitudinal modes of the resonator
coincides with the peak transmittance of the etalon (see Fig. 9,10) and other
modes are reflected away from the cavity. If in. addition, the finesse of the
etalon is so high as to introduce sufficiently high losses for the modes
adjacent to the mode selected, then one may have a single mode oscillation
(see Fig. 9.10).

Fig, 9.9 Transmittance versus frequency of a Fabry-Perot etalon for two

different values of the reflectivity of the surfaces bounding the etalon. The peaks
of transmission correspond to frequencies satisfying Eq. (9.46).
1.0

i
T
0.5

A __-^, V
f
v_^y
/

/ i v
I

.
1 \R = 0.5

= 0.8
—
y
—--^

The FWHM in transmittivity is approximately given by

1
/

l
V
\

^~~~

Av'i = c/nnt cos OF*

where F is the coefficient of finesse and is given by

and I? is the reflectivity of the two surfaces.

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 262 Lasers: II

Example: Consider an argon ion laser for which the FWHM of the gain profile is
about 8 GHz, Thus for Dear normal incidence (0 « 0), the free spectral range of the
etalon must be greater than about 10 GHz. Thus
cflnt Z\Gl0Mz
Taking fused quartz as the medium of the etalon, we have n 1,462 (at A ~ 0,51 fim)
and thus
t< 1cm
Another very important method used to obtain single frequency oscilla-
tion is to replace one of the mirrors of the resonator with a Fox-Smith
Fig. 9.10 The figure shows how by inserting a Fabry-Perot etaloe inside the
resonator, one can achieve single frequency oscillation. Without the etalon there
are five modes above threshold. With the etalon only one of these modes is
transmitted by the etalon and hence only that mode can oscillate. The other
modes fall below threshold due to increased losses.

Gain

Threshold without
etalon

J V

Threshold
with etalon

Gain

Only mode above

threshold

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.5 Mode selection 263

interferometer as shown in Fig. 9.11. Waves incident on the beam splitter

BS from. Mx will suffer multiple reflections as follows:
Reflection 1: A^ -> BS -»Af 2 -> BS ->M,
Reflection 2: Ml^BS^M1^BS^Mi
^BS^M2^BS-^Ml etc, (9.48)
Thus the structure will behave much like a Fabry-Perot etalon if the beam
splitter BS has a high reflectivity1*. For constructive interference among
waves reflected towards M 1 from the interferometer, the path difference
between two consecutively reflected waves must be ml, i.e.,
(2d2 + 2d3 + 2d2 - 2d2) = mk
or
=m_
v __; m = any integer (9.49)
4" 2 + "3/
Thus frequencies separated by Av = cj2(d2 + d3) will have a low loss. Hence
if Av is greater than the bandwidth of oscillation of the laser, then one can
achieve single mode oscillation. Since the frequencies of the resonator
formed by mirrors Mt and M 2 are
¥ = qc/2{dt + d2); q = any integer (9,50)
for the oscillation of a mode one must have
+d2) (9.51)
which can be adjusted by varying d3 by placing the mirror M 3 on a piezo
electric movement.

Problem 9.2; If one wishes to choose a particular oscillating mode out of the
possible resonator modes which are separated by 300 MHz, what is the approximate

Fig. 9.11 The Fox-Smith interferometer arrangement for selection of a single

longitudinal mode.
4 - »,

/
*/ / t V
A-U
1
f
It is interesting to note that if mirror M 3 is put above BS (in Fig, 9,11), then it would
correspond to a Michelson interferometer arrangement and the transmittivity
would not be sharply peaked.

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 264 Lasers: II

change in d3 required to change oscillation from one mode to another? Assume

A A
Solution; Differentiating Eq. (9.49), we have
me v
Mi=
*~wr^ WTd7)idi (9 52)
-
which gives us
dd3 = 0.025 ftm.

Problem 9.3; Consider a laser which is oscillating simultaneously at two frequ-

encies vt and v2: (v2 - Vj « cf2d> where d is the resonator length). If this laser is used
in an interference experiment, what is the minimum path difference between the
interfering beams for which the interference pattern disappears?

Solution: The interference pattern disappears when the interference maxima

produced by one wavelength fall on the interference minima produced by the other
wavelength, i.e., when the path difference / is such that
I = mc/v, = ( m + |)c/v 2 ; m = 0,1,2,,.. (9.53)
we can write the above equations as
w = fv,/c = lv 2 /t*~|
Thus the interference fringes disappear when
/ = c/2(v 2 -Vj) = d (9.54)
i.e., the source can be considered coherent only for path differences 1 < d, the length
of the laser. On the other hand if the laser was oscillating in just a single mode,
the coherence length would have been much larger and would be determined by
the frequency width of the oscillating mode only.

9.6 g-switching
As discussed in Sec. 9,3, the quality factor (or Q factor) of a cavity
is determined by the losses in the cavity; the smaller the losses, the larger
is the Q-value. Consider a laser cavity in which a shutter is introduced in
front of one of the mirrors as shown in Fig, 9.12. If the laser medium is

Fig. 9.12 A laser resonator with a shutter placed in front of one of the mirrors
to achieve ^-switching.

Amplifying medium

1i1 i
Pump Shutter

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.6 Q-switching 265

continuously pumped, the population inversion in the cavity will go on

increasing and will reach a very high value. This value could be much
larger than the threshold inversion required for the same laser in the absence
of the shutter. If the shutter is now suddenly opened, then the existing
population inversion will correspond to a value much above the threshold
value for oscillation. Thus the gain per round trip will be many times the
loss per round trip and the radiation in the cavity mode will build up very
rapidly, This rapid increase in the intensity will deplete the population
inversion which will go below threshold. This results in the generation of
an intense pulse of light from the cavity. Since the Q of the cavity is being
switched from a small value to a large value, the above technique is referred
to as Q-switching. Fig. 9.13 shows schematically the time variation of the

Fig, 9,13 Schematic representation of how the various quantities namely

(a) loss, (b) Q, (c) population inversion AN and (d) laser output power vary
with time when a laser is Q-switched.

(a)

f
Loss

(*)

= 0

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 266 Lasers: II

cavity loss, cavity Qt population inversion and the output power. As shown
in the figure, an intense pulse is generated with the peak intensity appearing
when the population inversion in the cavity is equal to the threshold value,
We will now write rate equations corresponding to ^-switching and
obtain the most important parameters such as peak power, total energy,
and duration of the pulse. We shall consider only one mode of the laser
resonator and shall examine the specific case of a three level laser system
such as that of ruby. In Sec. 8.6 while writing the rate equations for the
population N2 and the photon number n, we assumed the lower laser level
to be essentially unpopulated. If this is not the case then instead of Eq.
(8.100), we will have
d(N2V)/dt = - KnN2 + KnNl - T2lN2V + RV (9.55)
where the second term on the RHS is the contribution due to absorption
by JYj atoms/unit volume in the lower level. Since the (^-switched pulse is
of a very short duration, we will neglect the effect of the pump and
spontaneous emission during the generation of the g-switched pulse. It
must, at the same time be noted that, for the start of the laser oscillation,
spontaneous emission is essential. Thus we get from Eq. (9.55)
dN'2/dt = - (KnfV)AN' (9.56)
where
(N2-Nl)V, N'2 = N2V (9.57)
and V is the volume of the amplifying medium. Similarly one can also
obtain for the rate of change of population of the lower level
dN'Jdt = (KnfV)A.Nr (9.58)
where N\ = NXV. Subtracting Eq. (9.58) from Eq. (9.57) we get

d(AJV')/df = - (2Kn/V)ANf (9.59)

We can also write the equation for the rate of change of the photon number
n in the cavity mode in analogy to Eq. (8.104) as

dn/dt = Kn(N2 - AT,) - n/tG + KN2

&{Kn/V)AN'-n/tG (9.60)
where we have again neglected the spontaneous emission term KN2- From
Eq. (9.60) we see that the threshold population inversion is
(AJV')t = V/Kic (9.61)

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.6 Q-swhching 267

when the gain represented by the first term on the RHS becomes equal to
the loss represented by the second term (see Eq. (8.105)). Replacing V/K
in Eqs. (9.59) and (9.60) by (AN')tte and writing
T = t/te (9.62)
we obtain
d(AN')/dr = - 2nAN'/(AN% (9.63)
and
dn/dt = n[AN'/{AN\ - 1] (9.64)
Eqs. (9.63) and (9.64) give us the variation of the photon number n and
the population inversion AM' in the cavity as a function of time. As can
be seen the equations are nonlinear and solutions to the above set of
equations can be obtained numerically by starting from an initial
condition

AN'(x = 0) = ANJ and ?i(x = 0) = «.j (9.65)

where the subscript i stands for initial values. Here n-t represents the initial
small number of photons excited in the cavity mode through spontaneous
emission. This spontaneous emission is necessary to trigger laser oscillation.
From Eq. (9.64) we see that since the system is initially pumped to an
inversion AN' > (AN% dn/dt is positive, thus the number of photons
increases with time. The maximum number of photons in the cavity appear
when dn/dt = 0 i.e., when AN' = (AN% At such an instant n is very large
and from Eq. (9.63) we see that AN' will further reduce below (AN'\ and
thus will result in a decrease in n.
Although the time dependent solution of Eqs. (9.63) and (9.64) requires
numerical computation, we can analytically obtain the variation of n with
AN' and from this we can draw some general conclusions regarding the
peak power, the total energy in the pulse and the approximate pulse
duration. Indeed, dividing Eq. (9.64) by Eq. (9.63) we obtain

dn/dAN' = i[(AN') t /AN' - 1]

Integrating we get
n - «j = U AN; In [AN7(AN'y + t(ANr\ - AN']} (9.67)

Peak power Assuming the only loss mechanism to be output coupling and
recalling our discussion in Sec. 8.7, we have for the instantaneous power

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 268 Lasers: II

output
f out = nhv/tc (9.68)
Thus the peak power output will correspond to maximum n which occurs
when AN' = AN'V Thus
P —n hvli
1
max max / c

_ \ — ^ ' n " • / A nr/\ ' vv-"-" /i \'-"-' n> (y.vy)

where we have neglected n{ (the small number of initial spontaneous photons

in the cavity). This shows that the peak power is inversely proportional to
cavity lifetime.

Total energy In order to calculate the total energy in the Q-switched pulse
we come back to Eq. (9.64) and substitute for AN'/(AN'\ from Eq. (9.63)
to get
dn/dx = -1 d(AAO/dT - n (9.70)
Integrating the above equation from x = 0 to oo we get

nr - n, = i[(AAT)i - (AJV')f] - f * ndt

jo
or

ndt = |[(Ai¥') i - (Ai¥')f] - (nf - W|) (9.71)

where the subscript f denotes final values. Since n-x and »f are very small
in comparison to the total integrated number of photons we may neglect
them and obtain

(9.72)

Thus the total energy of the Q-switched pulse is

E= -Poutdr
Jo

= hv \ n dt
Jo
(9,73)

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.6 Q-s witching 269

The above expression could also have been derived through physical
arguments as follows: for every additional photon appearing in the cavity
mode there is an atom making a transition from the upper to the lower
energy level and for every atom making this transition the population
inversion reduces by 2. Thus if the population inversion changes from
(AiV')j to (AiV')f, the number of photons emitted must be |[(AJV')j — (AJY)f ]
and Eq. (9.73) follows immediately.

Pulse duration An approximate estimate for the duration of the Q-switehed

pulse can be obtained by dividing the total energy by the peak power. Thus

JL = (AADi-lAV),.

mm

In the above formulas, we still have the unknown quantity (AiV')f the final
inversion. In order to obtain this, we may use Eq. (9.67) for t-+ oo. Since
the final number of photons in the cavity is small, we have
(AN'), - (AN% = (AN'\ In t(AN%/(AN% ] (9.75)
from which we can obtain (AiV')f for a given set of (ANr){ and (AN'%.
As an example we consider the ^-switching of a ruby laser with the
following characteristics:
length of ruby rod = 10 cm; area of cross section = 1 cm2. ,
resonator length = 10 cm; mirror reflectivities = 1 (9.76)
and 0.7; Cr 3 + population density = 1.58 x 10 19 cm 3;
A0 = 6943A;» 0 =1.76»! sp = 3 x 10" 3 s,^(w 0 )= 1.1 x 10^ 1
The above parameters yield a cavity lifetime of 3.3 x 10" 9 s and the required
threshold population density of 1.25 x 1017cmT3*. Thus
(AAO t =1.25x 10 18
choosing
(AJV')i = 4(AN% = 5 x 10 18
we get
fmM«8.7xl07W
Solving Eq. (9,75) we obtain (AN% « G.G2AJY;, Thus

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 270 lasers: II

and
td % 8 ns

9.6.1 Techniques far Q-switching

As discussed, for g-switching the feedback to the amplifying
medium must be initially inhibited and when the inversion is well past the
threshold inversion, the optical feedback must be restored very rapidly.
ID order to perform this, various devices are available which include
mechanical movements of the mirror or shutters which can be electronically
controlled. The mechanical device may simply rotate one of the mirrors
about an axis perpendicular to the laser axis which would restore the Q
of the resonator once every rotation. Since the rotation speed cannot be
made very large (typical rotation rates are 24000rpm)» the switching of
the Q from a low to a high values does not take place instantaneously and
this leads to multiple pulsing.
In comparison to mechanical rotation, electronically controlled shutters
employing the electrooptic or acoustooptic effect can be extremely rapid.
A schematic arrangement of ^-switching using the electrooptic effect is
shown in Fig. 9.14. As we shall discuss in Chapter 15, the electrooptic effect
is the change in the birefringence of a material on application of an external
electric field. Thus the electrooptic modulator (EOM) shown in Fig. 9.14

Fig. 9.14 A typical arrangement for achieving g-switchlng using an electrooptic

shutter. When a certain voltage is applied across the EOM it acts as a \l plate
and thus the combination of polarizer-electrooptic modulator suppresses the
reflection from M2. When the voltage is removed, EOM does not change the
state of polarization and thus Q becomes very high.

Electrooptic
modulator

Polarizer

Laser rod

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.6 Q-swtching 271

could be a crystal which is such that in the absence of any applied electric field
the crystal does not introduce any phase difference between the two ortho-
gonally polarized components travelling along the laser axis. On the other
hand, if a voltage Vo is applied, then the crystal introduces a phase difference
of \n between the orthogonal components, i.e., it behaves as a \k plate.
If we now consider the polarizer-modulator-mirror system, when there is
no applied voltage, the state of polarization (SOP) of the light incident on
the polarizer after reflection by the mirror is along the pass axis of the
polarizer and thus corresponds to a high Q state. When a voltage Fo is
applied, the linearly polarized light on passage through the EOM becomes
circularly polarized (say right circularly). Reflection from the mirror
converts this to left circularly polarized and passage through the EOM
makes it linearly polarized but now polarized perpendicular to the pass
axis of the polarizer. Thus there is essentially no feedback and this
corresponds to the low Q state. Hence Q-switehing can be accomplished
by first applying a voltage across the crystal and removing it at the instant
of highest inversion in the cavity. Some important electrooptic crystals
used for (2-switching include potassium dihydrogen phosphate (KDP) and
lithium niobate (LiNbO3).
An acoustooptic Q-switch is based on the acoustooptic effect which
is discussed in detail in Chapters 17 and 18. The acoustooptic effect is
simply the diffraction of an incident light wave by an acoustic wave
propagating in a medium. The acoustic wave generates a phase grating
in the medium which is responsible for the diffraction. Thus if an
acoustooptic cell is placed inside the resonator, it can be used to deflect
the light beam out of the cavity thus leading to a low Q state. The Q can
be switched to a high value by pulsing the acoustic wave. Details of electro-
optic and acoustooptic ^-switches can be found in Koechner (1976).
^-switching can also be obtained by using a saturable absorber inside
the laser cavity. In a saturable absorber (which essentially consists of an
organic dye dissolved in an appropriate solvent), the absorption coefficient
of the material reduces with an increase in the incident intensity. This
reduction in the absorption is caused by the saturation of a transition (see
Sec. 8.5.1). In order to understand how a saturable absorber can be made
to Q-switch, consider a laser resonator with the amplifying medium and
the saturable absorber placed inside the cavity as shown in Fig. 9.15. As
the amplifying medium is pumped, the intensity level inside the cavity is
initially low since the saturable absorber does not allow any feedback from
the mirror M 2 . As the pumping increases, the intensity level inside the
cavity increases which, in turn, starts to bleach the absorber. This leads to

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 272 Lasers: II

an increase in feedback which gives rise to an increased intensity and so

on. Thus the energy stored inside the medium is released in the form of a
giant pulse leading to ^-switching. If the relaxation time of the absorber
is short compared to the cavity transit time then as we will discuss in
Sec. 9.7, the saturable absorber would simultaneously mode lock and
<2-switch the laser. For more details of Q-switehing, readers may refer to
Koechner (1976).

9,7 Mode locking in lasers

In this section we shall describe the technique of mode locking by
which it is possible to produce ultrashort optical pulses (with pulse durations
~ 10"~12 s). In order to understand the concept of mode locking, we consider
a laser formed by a pair of mirrors separated by a distance d. If the
bandwidth over which gain exceeds losses in the cavity is Av (see Fig. 9.16)
then, since the intermodal spacing is c/2d, the laser will oscillate simul-
taneously in a large number offrequencies.The number of oscillating modes

Fig. 9,15 A laser resonator with a saturable absorber placed inside the cavity
to achieve passive g-switching.

M, M,

111 Saturable
Pump absorber

Fig. 9.16 The gain profile of an active medium centred at %.

Gain curve

I K
A \
N Threshold

cfld

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.7 Mode locking in lasers 273

will be approximately (assuming that the laser is oscillating only in the

fundamental transverse mode)
Av
N -f 1 » 1 + integer closest to (but less than ) - -prr: (9.77)
\C{ L(l)

For example if we consider a ruby laser with an oscillating bandwidth of

6 x 1010 Hz with a cavity of length about 50 cm, the number of oscillating
modes will be ~ 200. This large number of modes oscillates independently
of each other and their relative phases are, in general, randomly distributed
over the range — n to + n. In order to obtain the output intensity of the
laser when it oscillates in such a condition, we note that the total field E of
the laser will be given by a superposition of the various modes of the laser.
Thus
N/2
E(t)= £ 4,exp(27rivnt + i«k) (9.78)
n = -N/2
where An and 4>n represent the amplitude and phase of the nth mode whose
frequency is given by
vB = v0 + n$v; n = - i J V , - f M + l,...,fiV (9.79)
with dv = c/2d, the intermode spacing, and ?0 represents the frequency of
the mode at the line centre.
In general the various modes represented by different values of n oscillate
with different amplitudes An and also different phases (f>n. The intensity at
the output of the laser will be given by
I = K\E(t)\2 =

where K represents a constant of proportionality and we have used Eq.

(9.79) for vB. Eq. (9.80) can be rewritten as

(9.81)
From the above equation the following observations can be made:
(a) Since the phases (f)n are randomly distributed in the range — n to + n
for the various modes, if the number of modes is sufficiently large, the
second term in Eq. (9.81) will have a very small value. Thus the
intensity at the output would have an average value equal to the first
term which is nothing but the sum of the intensities in various modes.
(Compare with Eq. (4.105).)

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 274 Lasers: II

(b) Although the output intensity has an average value of the sum of the
individual mode intensities it is fluctuating with time (see Fig. 9.17)
due to the second term in Eq. (9.81), It is obvious from Eq. (9,81) that
if t is replaced by t + q/Sv where q is an integer, then the intensity
value repeats itself. Thus the output intensity fluctuation repeats
itself every l/Sv = life seconds which is nothing but the round trip
transit time in the resonator.
(c) It also follows from Eq. (9.81) that, within this periodic repetition in
intensity, the intensity fluctuates. The time interval of this intensity
fluctuation (which is being caused by the beating between the two
extreme modes) will be
k =* [(>'(> + jNSv) - (v0 - iAtf v)] " l « 1/Av (9.82)
i.e., the inverse of the oscillation bandwidth of the laser medium.
When the laser is oscillating below threshold, the various modes are
largely uncorrelated due to the absence of correlation among the various
spontaneously emitting sources. The fluctuations become much less on
passing above threshold but the different modes remain essentially un-
correlated and the output intensity fluctuates with time.
Let us now consider the case in which the modes are locked in phase
such that (j)n = (p0, i.e., they are all in phase at some arbitrary instant of
time I — 0. For such a case we have from Eq. (9.80),
I = K\J^AnQ2ninSvt\2 (9.83)
If we also assume that all modes have the same amplitude
i.e., An = Ao, then we have
l = / 0 |£e 2 w i " a M I 2 (9.84)
where Jo = KAQ is the intensity of each mode. The sum in Eq. (9.84) can
Fig. 9,17 Time variation of the output power from a laser oscillating In a number of
longitudinal modes which have no phase relationship amongst themselves.

t' + 2d/c

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9J Mode locking in lasers 275

easily be evaluated and we obtain

'sin|>(iV+l)<$vfp 2
= I0 (9.85)
sin (nSvt)
The variation of intensity with time as given by Eq. (9.85) is shown in
Fig, 9.18, It is interesting to note that Eq, (9.85) represents a variation in
time similar to that exhibited by a diffraction grating in terms of angle.
Indeed the two situations are very similar; in the case of diffraction, grating,
the principal maxima are caused due to the constructive interference among
waves diffracted by different slits and in the present case it is the interference
in the temporal domain among modes of various frequencies.
From Eq. (9.85) we can conclude the following:
(1) The output of a mode locked laser will be in the form of a series of
pulses and the pulses are separated by a duration
l/dv = 2d/c (9.86)
i.e., the cavity round trip time. Thus the mode locked condition can
also be viewed as a condition in which a pulse of light is bouncing
back and forth inside the cavity and every time it hits the mirror, a
certain fraction is transmitted as the output pulse.
(2) From Eq. (9.85) it follows that the intensity falls off very rapidly
around every peak (for large N) and the time interval between the
zeroes of intensity on either side of the peak is
(9.87)

Fig. 9.18 The intensity variation as a function of time of a mode locked pulse
train as described by Eq. (9.85). Jo represents the intensity of each individual
mode. The pulses are separated by a time interval of 2dfc which is just the
cavity round trip time.

jJSl A A A /W/\A/\J
2d/c

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 276 Lasers: II

Table 9.1, Some typical laser systems and their mode locked pulse widths

Observed pulse
Laser Av(Hz) rD = (Av) ' (s) width (s)

He-Ne 0.6328 ~ 1 . 5 x 10 9 ~ 6.7 x 10 - 1 0 ~ 6 x 10^'10

Argon ion 0.488 ~ 7 x 10" ~ 1 . 5 x 10 10 - 2 . 5 x 10" 1 0
Nd:YAG 1.060 ~ 1 . 2 x 10 10 ~ 8.3 x 10 11 -7.6x 10""
Ruby 0.6943 ~ 6 x 10 10 - 1 . 7 x 10 11
- 1.2 x 10" "
Nd; glass 1.06 ~ 3 x 1012 - 3.3 x 10 - 13 - 3 x 10"13
Dye 0.6 ~10» -10"»

Table adapted from Yariv (1985) and Demtroder (1981).

We may define the pulse duration as approximately (like the

FWHM)
tD - 1/Av (9.88)
i.e., the inverse of the oscillating bandwidth of the laser. Thus the
larger the oscillating bandwidth the smaller will be the pulse width.
Table 9.1 shows the bandwidth Av, tD and observed pulse width for
some typical laser systems. As can be seen one can obtain pulses of
duration 1 ps or shorter; such pulses, also referred to as ultrashort
pulses, find wide applications in the study of ultrafast phenomena in
physics, chemistry and biology (Shapiro, 1977).
(3) The peak intensity of each mode locked pulse is given by
I = (N+l)2I0 (9.89)
which is (N + 1) times the average intensity when the modes are not
locked. Thus for typical solid state lasers which can oscillate
simultaneously in 10 3 -10 4 modes, the peak power enhancement due
to mode locking can be very large.

Fig. 9.19(a) shows the output from a mode locked He-Ne laser operating
at 6328 A. Notice the regular train of pulses separated by the round trip
time of the cavity. Fig. 9.19(fe) shows an expanded view of one of the mode
locked pulse; the pulse width is about 330 ps.

9.7.1 Techniques far mode locking

As we have seen above mode locking essentially requires that the
various longitudinal modes be coupled to each other. In practice this can

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9,7 Mode locking in lasers 277

be achieved either by modulating the loss or optical path length of the

cavity externally (active mode locking) or by placing saturable absorbers
inside the laser cavity (passive mode locking).
In order to understand how a periodic loss modulation inside the
resonator cavity can lead to mode locking, we consider a laser resonator
having a loss modulator inside the cavity with the modulation frequency
equal to the intermode frequency spacing Sv, Consider one of the modes
at a frequency v r Since the loss of the cavity is being modulated at a
frequency <§v, the amplitude of the mode will also be modulated at the same

Fig. 9,19 (a) Output pulse train of a mode locked He-Ne laser, (b) Expanded
view of a single pulse from the pulse train. (After Fox, Schwarz and Smith
(1968).)

Intensity

Time (5 ns/div.)

(a)

Intensity

^r.V' fF

Time (0.1 ns/div.)

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 278 Lasers: II

frequency Sv and thus the resultant field in the mode may be written as
(A 4- B cos 2ndvt) cos 2nvqt = A cos 2nvq t + \B cos [2n(vf + Sv)t]
+ | B cos [2TT(V, - dv)t] (9.90)
Thus the amplitude modulated mode at a frequency vq generates two waves
at frequencies v^ + Sv and vq — Sv, Since Sv is the intermode spacing, these
new frequencies correspond to the two modes lying on either side of v r
The oscillating field at the frequencies vq±Sv = vq+l forces the modes
corresponding to these frequencies to oscillate such that a perfect phase
relationship now exists between the three modes. Since the amplitudes of
these new modes are also modulated at the frequency Sv, they generate
new side bands at vq + 2<5v = vq+2 and vq — 2Sv — vq^2. Thus all modes are
forced to oscillate with a definite phase relationship and this leads to mode
locking.
The above phenomenon of mode locking can also be understood in the
time domain by noticing that the intermode frequency spacing Sv = cfld
corresponds to a time period of 2d/c which is exactly one round trip time
inside the cavity. Hence considering the fluctuating intensity present inside
the cavity (see Fig. 9.17), we observe that since the loss modulation has a
period equal to a round trip time, the portion of the luctuating intensity
incident on the loss modulator at a given value of loss would after every
round trip be incident at the same loss value. Thus the portion incident at
the highest loss instant will suffer the highest loss at every round trip.
Similarly, the portion incident at the instant of lowest loss will suffer the
lowest loss at every round trip. This will result in the build up of narrow
pulses of light which pass through the loss modulator at the instant of
lowest loss. The pulse width must be approximately the inverse of the gain
bandwidth since wider pulses would experience higher losses in the
modulator and narrower pulses (which would have a spectrum broader
than the gain bandwidth) would have lower gain. Thus the above process
leads to mode locking.
The loss modulator inside the cavity could be an EOM or an acoustooptic
modulator (AOM), The electrooptic and acoustooptic effects are discussed
in detail in Chapters 15 and 19. Fig. 9.14 shows a typical arrangement for
introducing loss modulation inside the cavity. The EOM changes the state
of polarization (SOP) of the propagating light beam and the output SOP
depends on the voltage applied across the modulator. Thus the SOP of
the light passing through the modulator, reflected by the mirror and
returning to the polarizer can be changed by the applied voltage and conse-
quently the feedback provided by the polarizer-modulator-mirror system.

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.7 Mode locking in lasers 279

This leads to a loss modulation. Some of the electrooptic materials used

include potassium dihydrogen phosphate (KDP) and lithium niobate
(LiNbO3).
Acoustooptic modulators can also be used for mode locking, Fig, 9,20
shows a typical arrangement in which an AOM is introduced into the
cavity. Standing acoustic waves are generated inside the modulator which
diffracts the light beam through Bragg diffraction effect. If the acoustic
frequency is Q, then the standing wave would oscillate at a frequency 2O
and thus the loss would be modulated at a frequency 20. When this
frequency equals the intermode frequency spacing, one would obtain mode
locking.
The above techniques in which an external signal is used to mode lock
the laser are referred to as active mode locking. One can also obtain mode
locking using a saturable absorber inside the laser cavity. This technique
does not require an external signal to mode lock and is referred to as
passive mode locking. As discussed in Sec. 9.6 in a saturable absorber, the
absorption coefficient decreases with an increase in the incident light
intensity (see Sec. 8,5.1). Thus, the material becomes more and more
transparent as the intensity of the incident light increases. In order to
understand how a saturable absorber can mode lock a laser, consider a
laser cavity with a cell containing the saturable absorber placed in a cell
adjacent to one of the resonator mirrors (see Fig. 9.15). Initially the saturable
absorber does not transmit fully and the intensity inside the resonator has

Fig. 9.20 A typical configuration for mode locking of lasers using the acousto-
optic effect In the AOM standing acoustic waves are formed which through
Bragg diffraction effects diffract the laser beam periodically out of the cavity
leading to a mode locking.
Diffracted beam

Acoustooptic cell

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 280 Lasers: II

Table 9.2 , Some organic dyes used to mode lock laser systems

Eastman Eastman
DDF Cryptocyanine No. 9740 No. 9860

Laser Ruby Ruby Nd Nd

/ s (W/cm 2 :) - 2 x 107 ~ 5 x 10* ~ 4 x 107 - 5.6 x 107
%(ps) -14 -22 -8.3 -9.3
Solvents Methanol Nitrobenzene 1,2-Dichloroethane
Ethanol Acetone Chlorobenzene
Ethanol
Methanol

°DDI stands for 1, l'-diethyl-2,2'-dicarbocyaninc iodide. Table adapted from Koechner

(1976).

a noiselike structure. The intensity peaks arising from this fluctuation bleach
the saturable absorber more than the average intensity values. Thus the
intensity peaks suffer less loss than the other intensity values and are
amplified more rapidly as compared to the average intensity. If the saturable
absorber has a rapid relaxation time (i.e., the excited atoms come back
rapidly to the lower level) so that it can follow the fast oscillations in
intensity in the cavity, one will obtain mode locking. Since the transit time
of the pulse through the resonator is 2d/cy the mode locked pulse train will
appear at a frequency of c/2d. Table 9.2 gives some organic dyes used to
mode lock ruby and Nd laser systems; /, and % represent the saturation
intensity and relaxation time respectively.

9.8 Modes of a confocal resonator system

In Sec. 9.5 we mentioned that the amplitude distribution in the
fundamental transverse mode is Gaussian. In this section we will first

Fig, 9.21 The symmetric confocal cavity consisting of two mirrors of radii of
curvature R and separated by a distance JR.

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.8 Modes of a confocal resonator system 281

consider a confocal resonator cavity consisting of a pair of mirrors of equal

radii of curvature R and separated by a distance R and obtain the modes
of the cavity. Fig. 9.21 shows the symmetric confocal cavity. A mode of
the cavity will be thefielddistribution which is such that after one complete
round trip through the resonator, the resultant field distribution is the
same as the distribution with which we started. Since the cavity shown in
Fig, 9.21 is symmetric about the midplane NNf, the modes can be simply
obtained from the condition that the field distribution across the plane
AA' must be the same as that across the plane BB' which is obtained after
one half round trip. The allowed frequencies of oscillation will be those
for which the phase change after one half round trip is an integral multiple
of n.
In order to obtain the modes, we notice that a given field in the plane
AA" propagating to the right first diffracts over a distance R and then
suffers a reflection at the mirror M2 to complete one half round trip. Hence
in order to obtain thefieldacross BB\ we must know the effect of diffraction
and also the effect of reflection by a mirror.
The effect of diffraction is given by Eq, (5.5) which we recall below:

g(xtyiz)-1-exp(-ikz) ;',/,<>)

(9.91)

where z is the distance between the observation plane and the initial plane
where the field distribution is f{z, y, 0).
In order to obtain the effect of the mirror on the incident field we use
the same procedure as in Sec. 6.3 to obtain the effect of a lens. Fig. 9.22

Fig, 9.22 Diverging spherical waves from a point P coiwerge after reflection
from the mirror to the point Q.

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 282 Lasers: II

shows a point source F placed at a distance u in front of a concave mirror

of radios of curvature R, Spherical waves emerging from P are focussed
by the mirror to the image point Q at a distance from the mirror such that

(9,92)
_
Thus the mirror converts an incident diverging spherical wave of radius
of curvature u to a converging spherical wave of radius of curvature v.
Now as discussed in Sec. 5.4, the phase distribution in the transverse plane
AM of the incident diverging spherical wave is

exp (9.93)

(apart from a constant phase term). Similarly the phase distribution of the
reflected converging spherical wave on the same plane will be

exp (9.94)

Hence if pm represents the factor by which the incident phase distribution

is to be multiplied to obtain the reflected phase distribution, then

exp

or

ex
Pu = P i (9.95)

The above equation may be compared with Eq, (6.14) for the effect of a lens.

Problem 9.4: Use Eq. (9.95) for an incident plane wave on (a) a concave mirror
(represented by a positive R) and (b) a convex mirror (represented by a negative R)
and discuss the result.
We will now use Eqs. (9.91) and (9.95) to obtain the modes of the resonator
depicted in Fig. 9.21. Let f(x,y) represent the ield on the plane AA'. This
field diffracts from mirror Mt to M2 to produce a field given by
I
XR e x p ( - i « * ) | | f(x\y')

. k
x exp (9.96)

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.8 Modes of a confocal resonator system 283

where we have used Eq, (9.91) and the fact that the distance between the
mirrors is R; $tf represents the area of the mirror. Hence the field on the
plane BB' after one half round trip will be

h(x, y) = g(x, y) exp | i | (x2 + y2) (9.97)

If the field distribution f(x, y) is to represent a mode then

Kx>y) = aflxyy) (9.98)

where a is a complex constant. The losses suffered by the field would be
determined by the magnitude of a and the phase shift in one half round
trip by the phase of <r. We now substitute Eqs. (9.97) and (9.98) in Eq. (9.96)
to obtain

c, y) = ~ exp ( _ ikR) dx' dy'f{x', / )

x exp I - i ^ (x/2 + y'2 - 2xx' - 2yy')

(9.99)

We now introduce a function s(x, y):

s(x, y) =/(x, y) exp [" - i A (X2 + ^2) | (9J00)

and also a set of dimensionless variables

(9.101)
Using Eqs. (9.100) and (9.101), Eq. (9.99) becomes

• ^ y ^ ' + w^df'di/' (9.102)

where sff represents the modified limits of integration.

In order to simplify the analysis we assume the mirrors to have a
rectangular cross section with dimensions 2a x 2b. Thus Eq. (9.102) can be
written as
no
s(f, n')^m> + m>)
d? di (9.103)

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 284 Lasers: II

where
(9.104)
We now try a separation of variables technique and write
s(£, rj) — p(i;)q(t]) (9.105)
a = KX
Substituting in Eq. (9.103) and separating the variables we get

*'df (9.106)

m
'di (9.107)

The integrals appearing in Eqs. (9.1.06) and (9.107) are similar to a Fourier
transform and are referred to as finite Fourier transforms. They tend to a
normal Fourier transform for £,Q -* oo and rjQ-*-oo. Slepian and Pollack
(1961) have shown that the solutions of the above integral equations are
prolate spheroidal functions. Here we will consider the case when

(9.108)

We recall from Eq. (5.39) that a2ll0R and b2fA0R are nothing but the
Fresnel numbers corresponding to the size of one mirror when viewed from
the centre of the other mirror. Thus Eq. (9.108) essentially implies that the
Fresnel numbers are large. We also assume that the functions p(£) and q(rj)
tend to zero as £ and r\ tend to infinity; we shall later show the consistency
of the assumption. Under the above assumptions we obtain

fj(f)e^'df (9.109)

m
'4rjf (9.110)

Notice that Eq. (9.109) tells us that we are looking for functions p(£) which
are their own Fourier transforms. It is well known that Hermite-Gauss
functions are a class of functions which are their own Fourier transforms.
Eq. (9.109) can indeed be transformed to a differential equation for

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.8 Modes of a confocat resonator system 285

Hermite-Gauss functions (see Thyagarajan and Ghatak (1981)). Here we

assume the solution to be Gaussian and obtain the beam width which
satisfies Eq. (9.109). Thus we let
nf<?) = /4e~*2/a2 (9 111)
where a is an unknown parameter to be determined such that p(£) satisfies
Eq. (9.109). Substituting in Eq. (9.109) and using Eq. (5.9) we obtain

(9.112)

If the above equation is to be valid for all values of 4 the exponents on

the LHS and RHS must be equal i.e.,
a = 72 (9.113)
Substituting this value of a in Eq. (9.112), we get
i(
K==e- ** / 2 ~* / 4 ) (9.114)
where we have used i = em/2. In an identical fashion, we can obtain
ci(fj) = BQ ~' (9.115)
with
T = e -i(W2-«/4) (9.116)

Hence we obtain

fix, y) = s(x, y) exp I i ^ (x2 + y2)

(9.117)

with K = AB and
o — ICT —e (?, 11 oj

Eq. (9.117) gives the transverse distribution of the Gaussian mode of the
confocal resonator at either mirror. We now note the following points:
(a) From Eq. (9.118) it follows that |<r| = 1, i.e., the amplitude of the field
after one half round trip is the same as at the starting point, i.e., there is
no loss. This is just due to the fact that we have assumed ^ 0 and f/0-> oo.
(I?) From Eq. (9.118), we note that the phase shift after one half round
trip is kR — JTI. Hence only those frequencies are allowed for which this

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 286 Lasers:. II

quantity is an integral multiple of n, i.e.,

kR — jn = qn; # = 0,1,2,...
or

which gives us the allowed frequencies of oscillation. Different value of q

give us the different longitudinal modes of the cavity corresponding to a
Gaussian transverse ield distribution. The intermode frequency spacing is
Avq = c/2R (9.120)
For a typical cavity with I? = 100 cm, Avq — 150 MHz.
(c) Recalling the discussion in Sec. 5,4 we observe from Eq. (9.117) that
the Gaussian mode has a curved phase front of radius of curvature R at
the mirror. This is exactly equal to the radius of curvature of the mirror.
Thus the Gaussian mode is incident normally on the mirror and will retrace
its path after reflection. The spherical wavefront described by Eq. (9.11?)
is converging which is consistent with the fact that Eq. (9.117) represents
the field after reflection.
id) The 1/e width of the Gaussian field at the mirror can be obtained
from Eq. (9.117) and is
w = (2R/k)i (9.121)
(e) We have shown in Sec. 5.4 that a Gaussian beam remains always
Gaussian when it diffracts in space. Since the mode field inside the resonator
cavity is formed by the diffraction of the field given by Eq. (9,117) into the
cavity, it follows that the mode field is Gaussian everywhere both inside
and outside the cavity. Since a Gaussian beam, diffracts symmetrically on
either direction of the waist, it follows from the symmetry of the resonator
shown in Fig. 9.21 that the waist must lie at the centre of the resonator.
Recalling the equation representing the change of the beam width of a
Gaussian beam as it propagates (see Eq. (5.11))
2
w (z) = w|(l + X2z2/n2wt) (9.122)
we can easily obtain the waist size w0 of the Gaussian mode by using the
fact that at z = R/2, w(z) - {IRIkf •• see Eq. (9.121). Thus we get
(9.123)
which is independent of the transverse dimension of the mirror and depends

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.9 General spherical resonator 287

only on i? and X. Fig, 9,23 shows the variation of w(z) and the phase front
of a Gaussian mode in a confocal resonator.
The 1/e beam radius of the Gaussian mode at the mirror can be obtained
from Eq. (9.117) as (2R/kfs. Obviously if the transverse dimension of the
mirrors is very large compared to {2R/kf\ then the mirrors will behave
practically as infinite size mirrors. Thus if the transverse dimension of the
mirror is 2a x 2b, then for the approximation of extending the limits of
integration in Eqs. (9.106) and (9.107) to be valid, we must have
a »{k 0 R/n)*\ b » (A0JR/TT) (9.124)
which is almost the same as Eq. (9.108).
As an example we consider a symmetric confocal resonator JR = 1 m
operating at l 0 = 1 fim. For such a case we have

0.4 mm
2n
and the beam width at the mirrors is ^/2w0 » 0.57 mm.

Problem 9,5: Obtain the angle of divergence of the Gaussian mode discussed above.

9.9 General spherical resonator

In Sec. 9.8 we have shown that the fundamental mode of a symmetric
confocal resonator has a Gaussian field distribution. As we discussed in
Sec. 9.2, in general, one can form a resonator with a pair of spherical mirrors
of radii of curvature Rl and R2 and separated by a distance d (see Fig. 9.24).
If we can find a Gaussian beam which has a phase front with radius of
curvature Rt on mirror Af, and a radius of curvature R2 on mirror M 2 ,

Fig. 9.23 Variation of w(z) and the phase front of a Gaussian mode in a
confocal resonator.

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 288 Lasers: II

and if the beam widths at both the mirrors are small compared to the
mirror sizes, then such a Gaussian beam can resonate in the resonator and
will correspond to the fundamental mode of the resonator (see Fig. 9.24).
Before we consider a general spherical resonator, we first consider a
resonator consisting of a plane mirror and a concave mirror of radius of
curvature JR (see Fig. 9.25). In order to find the Gaussian mode of the
resonator, we use the fact that the Gaussian mode must have radii of
curvature equal to those of the mirrors at the position of the mirrors. Hence
it must have an infinite radius of curvature at the plane mirror and a radius
of curvature JR on the concave mirror. Since the radius of curvature of a
Gaussian beam at its waist is infinite, the Gaussian mode must have its

Fig. 9.24 A Gaussian beam resonating between two spherical mirrors of radii of
curvature Rt and R2 and separated by a distance d; z = 0 is the position of the
waist of the beam where the phase front is plane.

= R,

Fig. 9.25 A resonator consisting of a plane mirror and a concave mirror of

radius of curvature I?.

2= 0

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.9 General spherical resonator 289

waist at the plane mirror. It should also have a radius of curvature R at

a distance d from the plane mirror. Hence using Eq. (5,17) we may write
n2w$/A2d2)
which can be solved to give

This gives us the Gaussian mode of the resonator shown in Fig. 9.25.

Problem 9.6; If d = JR/2 in Fig. 9.25 then it will correspond to a symmetric confocal
resonator. Show from Eq. (9.125) that vv0 is consistent with Eq. (9.123).
From Eq. (9.125) we see that if d > R,. then w0 becomes complex. Such
a resonator cannot support a fundamental Gaussian mode and is unstable.
It also follows from Eq. (9,125) that w0 can be increased by increasing R
while keeping d fixed. Thus if we consider d~50cm, I = l ^ m then
w o »O.4mm for K = l m , w o « 1 . 5 m m for J? = lG0m, The effect of
increasing w0 is to increase the volume of the region occupied by the mode
in the cavity. This volume which is also referred to as the cavity mode
volume essentially determines the volume over which interaction between
the beam and the amplifying medium is taking place. Thus for higher laser
powers, one must have larger cavity mode volumes which can be obtained
by using larger R values. Eq. (9.125) also gives the waist size for symmetric
spherical resonators but d then represents half the distance between the
mirrors.
We now consider a general spherical resonator as shown in Fig. 9.24.
We have to find a Gaussian beam having a phase front of radius of curvature
Rl at mirror Mi and R2 at mirror M 2 . If we let the waist lie at some plane
z = 0 as shown in Fig. 9.24 and let the coordinates of the poles of the
mirrors M% and M 2 be z, and z2, then we have
R{zt) = z, + K2/ZU R(z2) = z2 + K2/Z2 (9,126)
where
K2 = n2wt/X2 (9.127)
We choose a sign convention such that the radius of curvature is positive
if the mirror is concave towards the resonator. Then for the structure shown
in Fig. 9.24, we have R(z2) = R2 and R(z1)=-Rv Thus
-R 1 =Z 1 + K 2 /Z 1 , R2 = Z2 + K2/Z2 (9.128)

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 290 Lasers: II

Since the distance between the mirrors is d, we also have

z2-zl=d (9.129)
We can solve for K2 from Eqs. (9.128) and (9.129) and using Eq. (9.127) obtain

A )M2 (1-
(9.130)
n2 (gl+g2-2g1g2)
where
(9.131)
Knowing w0 and z, and z2 from Eq. (9.128), we obtain

W2(z ) = —— £JL___ (9,132)

/ , )\ _=
W22(Z 2 (9.133)

where w(zt) and w(z2) give the beam radii at the two mirrors. For the above
analysis to be valid, the transverse dimensions of the mirrors must be large
compared to w(zt) and w(z2). From Eqs. (9.132) and (9.133) it follows that

Fig. 9,26 The stability diagram for optical resonators. The shaded regions
correspond to stable resonator configurations.

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9.10 Higher order modes 291

for the existence of a stable Gaussian mode we must have

^l (9.134)
The above condition is referred to as the stability condition for an optical
resonator. Resonators satisfying 0^glg2^'i are termed stable; outside
this range, the resonators become unstable. Fig. 9.26 depicts the stability
diagram which is a plot of gx versus g2. A point in the diagram represents
a resonator. Resonators falling in the shaded region satisfy the condition
(9.134) and hence are stable. Resonators in the unshaded region correspond
to unstable resonators.

9.10 Higher order modes

In Sec. 9.8 we mentioned that the mode of a resonator is such that
after every complete round trip in the resonator it repeats itself. Applying
this condition we have shown that the modal field distribution must satisfy
Eqs. (9.109) and (9.110) under the approximation that the mirrors are of
large transverse extent. We have also shown that a Gaussian field
distribution satisies Eqs. (9.109) and (9.110) and thus they represent a mode
of the resonator. It can be shown that Eq. (9.109) is actually satisfied by
an infinite set of functions known as Hermite-Gauss functions represented
by /jfn(£)e~*Z/2; here Hn(c,) represents the Hermite polynomial of order n,
A few lower order Hermite polynomials are given below:
H 0 (c)=l, HM) = 2l H2(£) = 4Z2-2 (9.135)
The Hermite-Gauss functions satisfy the following equation

HJZ'fr-WeM'd? (9.136)

Comparing with Eq. (9.109) we can write

p(« = H m ( « e ^ 2 ' 2 (9.137)

K = exp < —i

Similarly we can write

(9.139)
IkR . ,47
(9.140)

Substituting for p(0 and q(rj) from Eqs. (9.137) and (9.139) in Eq. (9.105)

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 292 Lasers: II

and using Eq, (9,100) we obtain the field distribution corresponding to the
(m, n)th mode on the mirror as

tf. (*2+y2)'
exp

(9.141)
where C is a constant. The above field distribution corresponds to what
is known as the TEMOTB mode; TEM stands for Transverse Electric and
Magnetic. As discussed earlier, the last factor represents the curvature of
the wavefront at the mirrors. (For a more detailed analysis see, e.g., Ghatak
and Thyagarajan (1978), Sec. 4.15.)
The fundamental mode corresponds to m = 0, n = 0 and using Eqs. (9,135)
in Eq. (9.141), we find that it is indeed Gaussian. The mode TEM, 0 will
correspond to a field pattern given by
x r (x2 + y
/(x,)>) = 2C —exp -^r-i-^lexp
W, (9.142)
L
The corresponding intensity pattern will be given by
/(x, y) = J()x2 exp [ - (x2 + >>2)/w§] (9.143)
where / 0 is a constant. The intensity distribution along the j-direction is
Gaussian and along the x-direction is of the form x2e x !'w°. The intensity
peaks at x = ± w0 and is zero at x = 0. Thus if a laser is oscillating in the
TEM, i 0 mode and we let it fall on a screen we will observe two distinct
blobs of light as shown in Fig. 9.5. Fig. 9.5 also shows the observed pattern
corresponding to other higher order modes. It is evident from Fig. 9.5 that
m corresponds to the number of zeroes in intensity along the x-direction
and n to the number of zeroes in intensity along the ^-direction.
In, Sec. 9.9 we have shown that the fundamental mode of a stable general
spherical resonator is Gaussian. The higher order modes corresponding to
the general resonator are also Hermite-Gaussian.
Substituting for K and T in Eq. (9.105) we get
<7 = exp{-i|>/*-(ro + n + 1)TE/2]} (9.144)
Thus the phase shift in one half round trip is kR — (m + n+ l)n/2 which
for a standing wave pattern must be an integral multiple of n. Thus we get
kR - (m + n + \)nj2 ~ qn
or (9.145)
vmna = (2q + m + n+ l)c/4R

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press

 9,10 Higher order modes 293

The frequency separation between two adjacent longitudinal modes dif-

fering in q by unity will be
Avq = c/2R (9.146)
Eq. (9.145) gives the frequencies of oscillation of a mode characterized by
(m, n, q) in a confocal resonator.

https://doi.org/10.1017/CBO9781139167857.010 Published online by Cambridge University Press