CHAPTER SEVEN

Control of Laser Oscillators

 7

Control of Laser Oscillators

7.1 Introduction
During laser oscillation one or more distinct mode frequencies can be
emitted, which in the general case behave as independent monochro-
matic oscillations. When more than one mode oscillates this multimode
operation usually corresponds to simultaneous oscillation on more than
one longitudinal mode of a single transverse mode of the cavity. It is also
possible for more than one transverse mode to oscillate simultaneously,
but this is not of great importance or interest to us here. In this chapter
we will examine how multimode operation can be suppressed in favor of
single mode operation. We shall also see how many simultaneously oscil-
lating modes can become locked together in phase to produce a form of
laser oscillation in which the laser output becomes a train of very short
pulses. This mode-locked operation is of considerable fundamental and
practical interest and we shall examine the various ways in which the
phenomenon can be induced.

7.2 Multimode Operation

We have seen that in an inhomogeneously broadened laser simultaneous
oscillation is possible on more than one longitudinal mode. These oscilla-
tions can be coupled to each other by applying an external perturbation,
as we shall see shortly. However, in the absence of such perturbation,
150 Control of Laser Oscillators

each mode oscillates independently of the others. We can represent the

total electric field of the laser beam as
3N
E(t) = En (t) cos[ωn t + φn (t)], (7.1)
n=1
where ωn is the frequency of the nth of N modes and φn (t) is its phase.
The frequences ωn , ωn+1 are separated by approximately πc/ , but suc-
cessive adjacent modes are not exactly evenly spaced because of mode-
pulling. The phase term fluctuates with time in a random way because
the coherence time of the beam is finite. These fluctuations can be
attributed to three main causes:
(i) spontaneous emission, of which there is always some even into the
narrow solid angle occupied by the laser beam,
(ii) fluctuation in the index of refraction of the amplifying medium,
(iii) vibrations of the mechanical structure of the laser resonator.
The amplitude En (t) changes with time if the frequency ωn drifts relative
to the center of the gain profile. Such drift can be made very small
by stabilizing the frequency of the laser. This is most simply done by
stabilizing the optical length L of the resonator. The optical length can
be defined in the most general way as
8
L= n(x)dx, (7.2)
0
where n(x) is the refractive index of the medium (or media) between the
resonator mirrors placed at the geometric spatial locations 0 and .
If the laser intensity is monitored with a fast optical detector, it will
be observed to fluctuate with time because of the beating together of the
various longitudinal modes. We saw in Chapter 5 that these beats allow
the number of oscillating modes to be determined. Although these beats
might generally be expected to give only high frequency fluctuation this
is not so. The beat signals from different pairs of modes also beat with
each other to give fluctuations down to quite low frequencies. The laser
intensity also fluctuates because of variations in the mode amplitudes
En (t) caused by fluctuations in the gain of the amplifying medium.
Fluctuations in amplitude and laser frequency can be controlled in
several ways. The simplest approach to ideal behavior is to cause the
laser to oscillate in only a single longitudinal and transverse mode. The
transverse mode most desirable in this context is the so-called funda-
mental mode in which the distribution of intensity across the laser beam
 Single Longitudinal Mode Operation 151

Fig. 7.1.

is Gaussian† . Oscillation in the fundamental transverse mode can be

controlled by appropriate selection of the radii of curvature of the laser
mirrors and the size of apertures in the laser cavity — for example, the
apertures provided by the lateral size of the amplifying medium itself.

7.3 Single Longitudinal Mode Operation

Single longitudinal mode operation in a homogeneously broadened laser
should be automatic. Surprisingly, sometimes this is not so. The
strongest mode should suppress all its neighbors, but because of the
standing wave character of the in-cavity field this may not occur if the
maxima of field amplitudes of two different modes do not overlap in
space. This phenomenon is called spatial hole-burning. Fig. (7.1) shows
schematically how it happens. Mode 1 draws its gain from, for example,
particles located near a field maximum such as A. Mode 2 draws its gain
from particles located near field maximum B. The two modes will not
compete strongly unless particles can move rapidly from A to B. This
phenomena is called spatial cross-relaxation. Clearly, such effects are
more important in a gas than a solid-state laser.
To force a laser into single longitudinal mode operation several meth-
ods have been used:
(i) By making the laser sufficiently short that only one cavity res-
onance lies under the gain profile and above the loss line. This
generally requires c/2L > ∆νD .
(ii) With an intracavity etalon of thickness t << L, as shown in Fig.

† To be discussed in detail in Chapter 16.

152 Control of Laser Oscillators

Fig. 7.2.

(7.2). The etalon is usually tilted slightly to prevent complica-

tions caused by reflections between the etalon faces and laser mir-
rors. The transmission maxima of such an etalon are relatively
wide in frequency compared with the transmission maxima of the
main cavity, but they are separated by a much greater frequency,
∆νetalon >> ∆νcavity ,
c0
∆νetalon = , (7.3)
2nd
where n is the refractive index of the etalon.
For example, for a 10 mm thick fused silica etalon ∆νetalon 10
GHz. If this etalon is used inside the cavity of a laser with a gain
linewidth of 1 GHz (say), then it is likely that only one of the
main cavity resonances will be within a transmission maximum
of the in-cavity etalon. This situation is shown schematically in
Fig. (7.3).
(iii) By placing a very thin absorbing film inside the cavity. The posi-
tion of the film, which could be a thin layer of metal evaporated
onto a glass substrate, is adjusted to be at a node of the intracavity
standing wave of a strong longitudinal mode. Other modes suffer
loss at the film and can be suppressed.
(iv) By making the laser homogeneously broadened. This is sometimes
possible in the case of gas lasers by making them operate at high
gas pressure.
(v) With a Fox—Smith interferometer as shown in Fig. (7.4), which acts
as a frequency-dependent[7.1] cavity mirror. Laser oscillation can
use either mirror A or mirror B, the combination of which behaves
like a Michelson interferometer. A longitudinal mode that divides
into two parts at the beamsplitter will suffer high loss from the
 Single Longitudinal Mode Operation 153

Fig. 7.3.

Fig. 7.4.

cavity unless both parts reflect in phase towards the amplifying

medium.
(vi) If one of the main cavity mirrors is an etalon, both faces of which
are reflective, then a frequency-dependent cavity mirror results and
can discriminate between longitudinal modes.
(vii) In homogeneously broadened lasers, multimode oscillation fre-
quently results because the standing waves corresponding to dif-
ferent modes deplete the population inversion at different spatial
locations. This is the spatial hole-burning already mentioned. Sin-
gle longitudinal mode operation can be obtained if the periodicity
of the intracavity field can be smoothed out. This requires the wave
inside the cavity to look like a travelling rather than a standing
wave, which can be accomplished by building the laser in the form
of a ring. In Fig. (7.5) waves can travel around the three-mirror
cavity in both directions, if one of these waves can be suppressed,
then a travelling wave exists inside the laser medium and no spatial
154 Control of Laser Oscillators

Fig. 7.5.

hole-burning results. To explain how this might be done, it is actu-

ally simplest to describe a clever scheme that uses a conventional
cavity with two mirrors but with elements inside the cavity that
change the polarization state of the wave. The laser configuration
shown in Fig. (7.6) has a Brewster window in the laser cavity and
two quarter-wave plates. These optical components, which will be
described in detail in Chapter 18, change linearly polarized wave
light passing through them into circularly polarized light. At point
A inside the cavity the wave that oscillates with minimum loss will
be linearly polarized in the direction shown. With respect to x and
y axes oriented at 45◦ to the plane of the diagram we can represent
a wave travelling to the right as
(Ex )A = E0 cos(ωt − kz),
(Ey )A = E0 cos(ωt − kz). (7.4)
After passage through quarter-wave plate L1 , which is oriented appro-
priately, the electric fields at point B are
(Ex )B = E0 cos(ωt − kz + π/2),
(Ey )B = E0 cos(ωt − kz). (7.5)
This is left hand circularly polarized light (LHCP). The second quarter-
wave plate L2 is oriented to produce a π/2 phase shift of the y-directed
electric vector, so at point C the right travelling wave can be written as
(Ex )C = E0 cos(ωt − kz + π/2),
(Ey )C = E0 cos(ωt − kz + π/2), (7.6)
which is once again linearly polarized. This wave reflects from mirror
M2 and returns through L2 . At point D this left travelling wave can be
 Mode-Locking 155

Fig. 7.6.

written as
(Ex )D = E0 cos(ωt + kz + π/2),
(Ey )D = E0 cos(ωt + kz + π), (7.7)
which is right hand circularly polarized light (RHCP). After a second
pass through L1 this wave becomes linearly polarized in its original
direction once more.
Inside the amplifying medium the total electric field resulting from
the right and left travelling waves is
Ex = E0 cos(ωt − kz + π/2) + E0 cos(ωt + kz + π/2)
= −2E0 sin ωt cos kz, (7.8)
Ey = E0 cos(ωt − kz) + E0 cos ωt + kz + π
= 2E0 sin ωt sin kz. (7.9)
The intracavity intensity is proportional to
Ex2 + Ey2 = 4E02 sin2 ωt, (7.10)
which is independent of z. No standing wave distribution of energy
density exists in the cavity and single longitudinal mode operation can
be obtained.

7.4 Mode-Locking
So far in this chapter we have discussed ways in which to cause a laser
to generate one or more CW oscillating frequencies corresponding to
cavity modes. However, it is also possible to cause the laser to generate
a train of regularly spaced, generally very short, pulses. A laser that
operates in this way is said to be mode-locked. This kind of behavior
often occurs spontaneously and is then referred to as self-mode-locking.
156 Control of Laser Oscillators

A fundamental mode-locked pulse train consists of a series of pulses sep-

arated by the cavity round-trip time 2 /c. That this kind of behavior
could occur should not seem surprising. Early in the development of the
laser, Fleck[7.2] showed that as a laser oscillation built up from sponta-
neous emission there was a natural tendency for the laser output to show
strong fluctuations on a time scale that corresponded to the round-trip
time in the cavity. A single spontaneous photon emitted into a high Q
mode of the cavity will bounce back and forth in the resonator, and be
amplified leading to a pulse train of spacing 2 /c. Whether such a group
of photons will succeed in competing with other growing oscillations will
determine whether the laser ends up operating CW or mode-locked.†
There are two ways of looking at mode-locking. The more common,
but less physically satisfying, is to treat the independent oscillating cav-
ity modes whose electric fields are given in Eq. (6.61) as modes that have
had their phases locked together, so that each mode has φn (t) = φ0 . By
a shift of our choice of time origin we can arbitrarily set φ0 = 0. The
total electric field of the laser can then be represented as
^N „
3
iωn t
E(t) = R En (t)e . (7.11)
n=1
To simplify the analysis we take En (t) = 1 and
} ]
(N + 1)
ωn = ω0 + n − ∆ωc , (7.12)
2
where ∆ωc = πc/ .
Substituting Eq. (7.12) into Eq. (7.11) and summing the resultant
geometric series gives
sin(N ∆ωc t/2)
E(t) = cos ω0 t . (7.13)
sin(∆ωc t/2)
This is a pure oscillation at frequency ω0 modulated with the envelope
function
sin(N ∆ωc t/2)
f (t) = . (7.14)
sin(∆ωc t/2)
The average power corresponding to this envelope is
sin2 (N ∆ωc t/2)
P (t) ∝ . (7.15)
sin2 (∆ωc t/2)

† In this context the term CW is used to include lasers in which the laser
excitation is pulsed, but in which the output radiation consists of one or more
waves at longitudinal mode frequencies that last for the period of excitation.
 Mode-Locking 157

Fig. 7.7.

Eq. (7.15) represents a periodic train of pulses that have the following
properties:
(a) the pulse spacing is ∆T = 2π/∆ωc = 2 /c;
(b) the peak power in the train is N times the average power, and the
peak field is N times the average for a single mode;
(c) the pulses within the train become narrower as N increases, and
for large N approach a value τ = ∆T /N .
These characteristics are evident from Fig. (7.7), which is a plot of
Eq. (7.15). For an inhomogeneously broadened laser with linewidth
∆νD , the number of independent oscillating modes will be
2π∆νD
N∼ , (7.16)
∆ωc
so the mode-locked pulse width will approach a value τ 1/∆νD .
For a homogeneously broadened laser, which ideally oscillates in a
single longitudinal mode, the frequency explanation of the generation of
short “mode-locked” pulses seems less than satisfactory. This brings us
to a more internally self-consistent model of mode-locking as a mode of
operation that the laser finds energy advantageous, and which does not
require the a priori assumption of independent oscillating cavity modes
that become locked. This argument is illuminated by considering one of
the most common ways in which mode-locking is accomplished — passive
locking with a saturable dye cell in the laser cavity as shown in Fig. (7.8).
The dye cell is frequently placed in close proximity to one of the cavity
mirrors.
The oscillation threshold for an individual cavity mode in this ar-
rangement will be high because of the intracavity loss presented by the
absorbing dye. However, if the laser oscillates in a “bouncing-pulse”
158 Control of Laser Oscillators

Fig. 7.8.

mode in which the intracavity intensity is concentrated, then this pulse

can “bleach” the dye.†
In this situation a bouncing pulse is more energy efficient in over-
coming the intracavity loss presented by the dye. The actual shape of
the pulse will be determined by a self-consistency condition. The pulse
train f (t) will have a Fourier transform (F T ) F (w) that describes the
frequency content of the bouncing pulse. Each of the frequency compo-
nents of this effective pulse train will be amplified by the gain medium
according to a saturated gain profile gs (w). It is clear that the narrow-
est width that the bouncing pulse can have is ∼ 1/∆ν, where ∆ν is the
width of the gain profile. In common with any amplifier, frequencies
outside the gain bandwidth are not amplified. Schematically, we could
write a self-consistency equation
F T [f (t)] = F (ω)gs (ω)αd (ω), (7.17)
where αd (ω) is the effective absorption spectrum of the bleached dye.
This explanation does not impose an inhomogeneously broadened de-
scription of the gain medium. Any broadband gain profile can, in prin-
ciple, amplify a short pulse. Of course, once we accept the bouncing-
pulse model of mode-locking we can examine F (ω) and will find that
the principal components of the Fourier transform will correspond to
longitudinal mode frequencies.‡

† A dilute dye becomes less absorbing at high light intensities, as stimulated

absorption and emission of dye molecules become balanced as the ground
state of the dye is depleted by excitation.
‡ In a sense this is an example of what comes first — the chicken or the egg?
 Methods of Mode-Locking 159

7.5 Methods of Mode-Locking

In general, mode-locked behavior is caused to occur by modulating the
gain or loss of the laser cavity in a periodic way, usually at a frequency
fm = c/2L. In amplitude modulation (AM) mode-locking the mag-
nitude of cavity loss (or gain) is modulated; in phase (or frequency)
modulation (FM) mode-locking only the complex part of the gain is
modulated. In active mode-locking the modulation is introduced by ex-
ternal means, for example, with an intracavity modulator. In passive
or self-mode-locking the modulation is created by the bouncing mode-
locked pulse itself.

7.5.1 Active Mode-Locking

In the simplest form of AM mode-locking the intracavity loss is switched
periodically with an intracavity shutter as shown schematically in Fig.
(7.9). The precise time variation of the periodic intracavity loss is not
very important, except that its period should equal the round-trip time,
∆T , for an optical pulse bouncing back and forth in the cavity. It is
energy advantageous for a bouncing pulse to develop, rather than one or
more CW longitudinal modes, as the pulse will adjust its arrival time at
the intracavity modulator to correspond to the time of maximum trans-
mission. An alternative way of viewing what happens is to realize that
the modulator generates new frequency components in a longitudinal
mode of frequency νm passing through it. These new frequencies are
called side-bands† and have values
νk = νm + kfm k = 0, ±1, ±2 . . . . (7.18)
Each of these side-bands also corresponds to another longitudinal cavity
mode with frequency ν(m+k ).
The interaction between the fundamental longitudinal modes and side-
bands causes the phases of the longitudinal modes to lock together and
generate a mode-locked pulse train. In an inhomogeneously broadened
laser the existence of several independent longitudinal modes whose
phases can become locked is clear. In a homogeneously broadened laser
in which only one such mode should oscillate, the side-bands of the
dominant mode are not suppressed by gain-saturation: they compete
for gain and a self-consistent bouncing pulse results. Almost all the
lasers that are used to generate very short mode-locked pulses are (ide-

† For a more detailed discussion see Chapter 19.

160 Control of Laser Oscillators

Fig. 7.9.

ally) homogeneously broadened: for example Nd:YAG, Nd:glass, dye

and semiconductor lasers.
A common method of intracavity AM mode-locking is to use an
acousto-optic modulator.† . This was the method used in the first ex-
perimental observation of mode-locking using a helium—neon laser[7.3] .
The acousto-optic modulator can be used to deflect light from the cavity.
The undeflected beam is thereby amplitude modulated.
In mode-locking applications the acousto-optic modulator is used in
a standing sound wave configuration; the acousto-optic modulator ma-
terial is designed to resonate for sound waves at frequency fs .‡ For an
acousto-optic medium used in this way the variation of refractive index
is
n(z, t) = n0 + ∆n cos 2πfs t cos ks z. (7.19)
This standing wave acts like a phase diffraction grating that deflects
light. The light travelling orthogonally to the sound wave is deflected at
frequency 2fs , because the maximum index variation in Eq. (7.19) occurs
twice per cycle of the standing sound wave. A typical experimental
arrangement for mode-locking with an intracavity acoustic cell is shown
in Fig. (7.10).
Mode-locking can also be induced through periodic variation of the
gain of the amplifying medium. In a semiconductor laser (see Chapter
13) this can be simply done by modulating the drive current to the laser
at frequency fm . Alternatively, one periodically modulated laser can be

† A full discussion of these devices is given in Chapter 19.

‡ This is in contrast to acousto-optic frequency shifter applications where one
end of the crystal is terminated in a matched acoustic load or absorber so
that only a unidirectional travelling sound wave results.
 Methods of Mode-Locking 161

Fig. 7.10.

used to mode-lock another. This approach is called synchronous pump-

ing. For example, an argon laser that is itself mode-locked, but does
not generate very short mode-locked pulses, can be used to periodically
pump a dye laser (an optically excited laser that uses a dilute organic
dye solution as its gain medium — see Chapter 12). The dye laser gener-
ates much shorter mode-locked pulses than were injected from the pump
laser. The success of this scheme requires that the cavities of pump and
sychronously mode-locked lasers be matched so that
1
fm (pump) = , (7.20)
T
where T is the cavity round-trip time in the pumped laser. Synchronous
pumping allows the use of a convenient mode-locked pump, such as a
Nd:YAG or argon ion laser, which do not themselves intrinsically gener-
ate the shortest mode-locked pulses, to pump dyes or doped glasses and
crystals that support a large gain-bandwidth.
In the ideal case the length of the mode-locked pulses generated by any
laser, whether it be homogeneously or inhomogeneously broadened, can
approach a value τ ∼ 1/∆ν, where ∆ν is the width of the gain profile.
Consequently gain media with large values of ∆ν are desirable for the
shortest mode-locked pulse generation, these include dye lasers,[7.4],[7.5]
Nd:glass lasers[7.6],[7.7] and titanium—sapphire lasers[7.8]−[7.11] . Typical
pulse lengths generated by such lasers are in the picosecond range, al-
though special configurations achieve subpicosecond values. Fig. (7.11)
shows a good example of an advanced mode-locked laser system that
can generate sub-picosecond pulses. In this scheme a CW argon ion
laser pump as acousto-optically mode-locked Nd:phosphate glass oscil-
lator. The mode-locked pulses are then further amplified by causing
162 Control of Laser Oscillators

Fig. 7.11.

them to make multiple passes inside a Nd:phophate glass regenerative

amplifier[7.7] .†
It is usual in a mode-locked system using a linear resonator to place
the intracavity modulator or absorbing dye close to one end mirror to
ensure that the laser pulse passes the modulator only once in every
round trip. If this is not done, then more than one bouncing pulse can
develop within the cavity and the output will not be a regular pulse train
of spacing 2 /c. In laser cavities configured in the form of a ring, the
placement of the intracavity modulator or absorber is not so important.
In this case the possibility exists for two counterpropagating pulses to
develop in the ring. In ring and linear resonators where two bouncing
pulses develop, it is possible to cause the pulses to overlap or collide
as they pass through an intracavity saturable absorber (dye). Because
of the nonlinear variation of absorption of the dye with intensity the
superimposed counterpropagating short pulses actually become shorter
in order to maximize bleaching of the dye. In this scheme, called colliding
pulse mode-locking (CPM), the typical placement of the absorber is at
the center of a linear cavity or one quarter perimeter away from the gain
medium in a ring cavity as shown in Fig. (7.12). CPM lasers can directly
generate mode-locked pulses of sub-picosecond generation[7.5],[7.12] .
The shortest pulses generated to date in ring cavity CPM are 27 fs
long[7.13] . There is also an approach called additive pulse mode-locking

† Synchronous mode-locking can also be achieved if the optical lengths of the

pump laser cavity and the pumped laser cavity have a ratio that is of the form
Lpump /
Lpumped = p/p , where p and p are small integers.
 Pulse Compression 163

Fig. 7.12.

(APM). The basic idea involves two coherent pulses with a relative phase
shift interfering so that their wings tend to cancel out[7.14],[7.15] .

7.6 Pulse Compression

Any short pulse can be represented by its Fourier transform
F (w) = F T {f (t)}. (7.21)
The Fourier transform describes the spectral composition of the pulse
f (t) in both amplitude and phase. For example,
F (w) = A(w)eiφ(w) . (7.22)
If we pass our original short pulse through a system in which the optical
delay varies with frequency as ∆φ(ω), then the emerging pulse will have
a Fourier transform
F (ω) = F (ω)ei∆φ(ω) . (7.23)
If the lower frequency components suffer a greater phase shift than the
higher frequency components, this is called a negative group velocity
delay. In this case the output pulse, g(t) will be
g(t) = F T {F (ω)}. (7.24)
If the original pulse is written as
f (t) = a(t)e−i[ω0 t+φ(t)] , (7.25)
where a(t) is a smooth, slowly varying function of time (the pulse enve-
lope), then the instantaneous frequency of the pulse is
dφ
ω(t) = ω0 + . (7.26)
dt
If φ(t) is nonlinear, the pulse is said to be chirped. If
φ(t) = bE 2 , (7.27)
164 Control of Laser Oscillators

Fig. 7.13.

Fig. 7.14.

the frequency chirping is linear, and with b > 0, is positive. The leading
edge of the pulse has a lower instantaneous frequency than the trailing
edge — it is red-shifted relative to the blue-shifted trailing edge. In a
medium with a negative group velocity delay characteristic, the lower
frequencies in the leading edge of the pulse travel more slowly than the
higher frequencies in the trailing edge. The back of the pulse catches up
with the front and the pulse is compressed, as shown schematically in
Fig. (7.13).
A simple pulse compressor can be constructed with two diffraction
gratings as shown in Fig. (7.14)[7.16] . The first order diffraction angle θ
satisfies
λ
sin θ = − sin θin . (7.28)
d
The longer of two wavelengths diffracts at a larger angle and takes a
longer path between the two gratings — thereby acquiring a larger phase
shift and satisfying the condition for negative group delay.
 Pulse Compression 165

An optical fiber can also be used to produce pulse compression. Be-

cause of the Kerr effect the refractive index of the fiber varies with
intensity as
n = n0 + γI. (7.29)
Accordingly, the peak of a pulse is slowed relative to its tail. There
are additional effects on the pulse because of the dispersion in the fiber
n(λ). If n(λ) lies in a region of negative group velocity dispersion (which
for silica fibers lies beyond 1.3 µm) then the index nonlinearity and
dispersion work together to make the pulse shorter.† To date pulses
have been compressed to only 6 fs in length[7.17] .

† An optical soliton (see Chapter 17) develops for similar reasons.

 References

[7.1] P.W. Smith, “Mode selection in lasers,” Proc. IEEE, 60, 422—440, 1972.
See also A.G. Fox and P.W. Smith, Phys. Rev. Lett., 18, 826—828, 1967.
[7.2] J.A. Fleck, “Mode-locked pulse generation in passively switched lasers,”
Appl. Phys. Lett., 12, 178—181, 1968.
[7.3] L.E. Hargrove, R.L. Fork and M.A. Pollack, “Locking of He-Ne laser modes
induced by synchronous intracavity modulation,” Appl. Phys. Lett., 5, 4—5,
1964.
[7.4] C.V. Shank, and E.P. Ippen, “Subpicosecond kilowatt pulses from a mode-
locked CW dye laser,” Appl. Phys. Lett. 24, 373—375, 1974.
[7.5] R.L. Fork, B.I. Greene, and C.V. Shank, “Generation of optical pulses
shorter than 0.1 psec by colliding pulse mode-locking,” Appl. Phys. Lett., 38,
671—672, 1981.
[7.6] D.J. Bradley and W. Sibbett, “Streak-camera studies of picosecond pulses
from a mode-locked Nd:glass laser,” Opt. Commun., 9, 17—20, 1973.
[7.7] L. Yan, P-T. Ho, C.H. Lee, and G.L. Burdge, “Generation of ultrashort
pulses from a neodymiun glass laser system,” IEEE J. Quantum Electron., QE-
26, 2431—2440, 1989.
[7.8] J.D. Kafka, M.L. Watts, D.J. Roach, M.S. Keirstead, H.W. Schaaf, and
T. Baer, in Ultrafast Phenomena VII, C.B. Harris, E.R. Ippen, G.A. Mourou,
and
A.H. Zewail, Eds., Springer-Verlag, New York, 1990, pp. 66—68.
[7.9] N. Sarnkara, Y. Ishida, and H. Nakano, “Generation of 50-fsec pulses
from a pulse-compressed, CW, passively mode-locked Ti:sapphire laser,”
Opt. Lett., 16, 153—155, 1991.
[7.10] D.E. Spence, P.N. Kean, and W. Sibbett, “60-fsec pulse generation from
a self-mode-locked Ti:sapphire laser,” Opt. Lett., 16, 42—44, (1991).
[7.11] J. Squier, F. Salin, and G. Mourou, “100-fs pulse generation and ampli-
fication in Ti:Al2 O3 ,” Opt. Lett. 16, 324—326, 1991.
 References 167

[7.12] P.M.W. French and J.R. Taylor, “Generation of sub-100 fsec pulses
tunable near 497nm from a colliding-pulse mode-locked ring dye laser,”
Opt. Lett., 13, 470—472, 1988.
[7.13] J.A. Valdmanis, R.L. Fork, and J.P. Gordon, “Generation of optical
pulses as short as 28 femtoseconds directly from a laser balancing self-phase
modulation, group-velocity dispersion, saturable absorption, and saturable
gain,” Opt. Lett., 10, 131—133, 1985.
[7.14] E.P. Ippen, H.A. Haus, and C.Y. Liu, “Additive pulse mode-locking,”
J. Opt. Soc. Am., B6, 1736—1745, 1989.
[7.15] K.J. Blow and D. Wood, “Mode-locked lasers with nonlinear external cav-
ities,”
J. Opt. Soc. Amer., B5, 629—632, 1988.
[7.16] E.B. Treacy, “Optical pulse compression with diffraction gratings,” IEEE
J. Quantum Electron., QE-5, 454—458, 1969.
[7.17] R.L. Fork, C.H. Brito Cruz, P.C. Becker, and C.V. Shank, “Compression
of optical pulses to six femtoseconds by using cubic phase compensation,”
Opt. Lett., 12, 483—485, 1987.