You are on page 1of 54

Department of Science and Technology

Laser Physics and Nonlinear Optics


RF mode locking and ber
amplication of diode lasers
M.G. Hekelaar
Graduation thesis
Enschede, June 2005
University of Twente
Department of Science and Technology
Laser Physics and Nonlinear Optics
RF mode locking and ber
amplication of diode lasers
M.G. Hekelaar
Graduation committee:
Prof. Dr. K.-J. Boller
Dr. P. Gro
Dr. ir. H.L. Oerhaus
Enschede, June 2005
Parts of the results in this research will be presented at the Pacic Rim Conference on Lasers
and Electro-Optics 2005 (IQEC/CLEO-PR 2005) in Japan [1]. Also, a large part of the results has
been prepared to be submitted to Optics Express [2].
1
Abstract
Semiconductor lasers are compact and very ecient sources of coherent radiation. By modulating
the driving current of such lasers, picosecond pulses can be generated. One important application
of ultrashort pulses is that the eciency of frequency of conversion stages, like optical parametric
oscillators (OPOs), can be greatly enhanced. By a synchronously pumped optical parametric
oscillator access into the mid-infrared spectral region can be created. For an ecient wavelength
conversion process, these OPOs require very high pump powers of several watts.
In this work, we present the rst mode-locked diode laser around 1.06 m wavelength with
subsequent amplication of the pulses in an ytterbium-doped ber amplier.
To obtain ultrashort pulses, the diode laser has been actively mode-locked by modulation of the
injection current. The antireection coated diode laser has been placed in an external cavity with its
length adjusted to match the modulation frequency. Diculties caused by the residual reectivity
of the antireection coated facet have been resolved by spectral limitation of the emitted light. A
diraction grating provides selective feedback, after which emission of nearly Fourier-limited pulses
results. The pulses have a duration of approximately 35 ps at a repetition rate of 1.4 GHz.
For power amplication of the mode-locked pulses from the diode laser, we use a cladding
pumped double-clad ytterbium-doped ber. After amplication, the average output power is more
than 9 W. With slightly shorter pulse durations than in the unamplied case, this corresponds to
peak powers of more than 100 W.
By grating tuning the diode laser, the ber output is tunable from 1050 to 1085 nm with average
output powers of more than 9 W and corresponding peak powers of more than 100 W. Based upon
our previous experience, these output parameters provide ideal conditions to synchronously pump
mid-IR OPOs. We note that no other experiments so far have shown such high-power pulses in
combination with such large wavelength tunability.
2
Acknowledgement
Ten months ago I started my graduation period at the Laser Physics and Nonlinear Optics group
at the University of Twente. Starting from almost zero, my knowledge of experimental optical work
has grown tremendously during these months. The results presented in this thesis would have been
almost impossible without the help of some members of the group (and, of course, the nature of
the universe).
First of all I would like to thank my supervisor, Petra Gro, for her time, explanation, support
and elitehaver. Secondly, Balaji Adhimoolam, thank you for the lot of joyful time we spent in the
lab. Further I would like to thank Ian Lindsay and Prof. Klaus Boller for their useful suggestions
and advices.
I also wish to thank Remco Nieuwland, Rolf Loch, Arie Irman and Ab Nieuwenhuis for the
many discussions and their experimental help. I am grateful to Martin Fransen for his assistance
in revealing some mysteries of the RF electronics. Besides the graduation students from the group,
Rolf Loch, Marten de Wit, Arjan Verkerk, Mark Luttikhof, Willem Beeker, Edip Can and Johan-
Martijn ten Hove, also Marieke van As, Remco Nieuwland, Bert Borger and Maarten van Zalk
often contributed to very pleasant lunch breaks.
Finally, I would like to thank the graduation committee members for reading my thesis, espe-
cially Herman Oerhaus from the Optical Techniques group for willing to be the external committee
member.
Thijs Hekelaar,
June 2005
3
Contents
1 Introduction 6
2 Theory 8
2.1 Diode lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.1 Lasers in general . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.2 Principle of operation of diode lasers . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Ultrashort pulse generation in diode lasers . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Mode locking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Active mode locking in diode lasers . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Pulse amplication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3 Measurement 17
3.1 Ultrashort pulse measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.1 Intensity autocorrelation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.2 Interferometric autocorrelation . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.3 Rotating-mirror autocorrelator . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Spectral measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.1 Optical spectrum analyzer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.2 Scanning Fabry-Perot interferometer . . . . . . . . . . . . . . . . . . . . . . . 24
4 Experimental results 26
4.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.1.1 Diode laser setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.1.2 RF electronics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Continuous wave diode laser characteristics . . . . . . . . . . . . . . . . . . . . . . . 29
4.3 Measurements on mode-locked diode lasers . . . . . . . . . . . . . . . . . . . . . . . 32
4.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.3.2 Spectral measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3.3 Autocorrelation measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.3.4 Power measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.3.5 Wavelength tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.3.6 Inuence of parameter changes on pulse duration . . . . . . . . . . . . . . . . 39
4.3.7 Summary of experimental results . . . . . . . . . . . . . . . . . . . . . . . . . 40
5 Pulse amplication 41
5.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2 Measurements on amplied mode-locked pulses . . . . . . . . . . . . . . . . . . . . . 43
5.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4
CONTENTS
5.2.2 Spectral measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.2.3 Autocorrelation measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.2.4 Power measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.2.5 Wavelength tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.2.6 Summary of experimental results . . . . . . . . . . . . . . . . . . . . . . . . . 47
6 Summary 49
5
Chapter 1
Introduction
Diode lasers are highly ecient, compact sources of coherent radiation, which are widely used in
science and technology, for example in CD players, for communication technology or for optical
pumping of solid-state lasers.
By modulating the driving current of such diode lasers, pulses with picosecond duration can
be generated, thereby enabling an even larger range of other applications [3]. One important
possibility is that the eciency of frequency of conversion stages (e.g. second harmonic generation,
optical parametric oscillation, etc.) can be greatly enhanced.
Over the last decade, there has been considerable interest in powerful mode-locked light sources
in the wavelength range around 1 m. Such sources are ideal for sensitive environmental monitoring
and for test and measurement purposes. Pulses of picosecond duration, for example, are suitable
for ultrafast spectroscopy within the lifetime of molecular energy levels. Access further into the
mid-infrared spectral region can be readily created by ecient wavelength conversion processes,
such as synchronously pumped optical parametric oscillators (OPOs) [4, 5].
Interest in the 1 m region has been boosted by recent developments in cladding pumped
ytterbium-doped silica ber lasers and ampliers. Oering a broad gain bandwidth and a high pump
conversion eciency, they form an almost ideal gain medium for the generation and amplication of
wavelength-exible and short pulses in the spectral region between 1000 and 1100 nm. Short pulse
generation has been demonstrated using ytterbium-doped ber lasers, followed up by a ytterbium-
based amplication [6, 7, 8]. However, due to the high dispersion in such a system, rather complex
cavities are required. Additionally, the extended ber lengths lead to an inherently low repetition
rate (tens of MHz), compelling large cavity lengths for synchronous pumping of, e.g., OPOs.
Mode-locked semiconductor lasers are a highly attractive alternative to mode-locked bulk and
ber solid state laser oscillators. Such diodes, in addition to the simple cavity design, oer high
(GHz) and variable repetition rates, suitable for driving compact synchronously pumped OPOs.
The electronically controllable repetition rate can be employed for rapid OPO tuning [9], while
the narrow spectral bandwidth of pulses in the picosecond range ensure ecient conversion in long
nonlinear crystals. In addition, the limited spectral bandwidth of these pulses yields a spectral
resolution suitable for spectroscopic detection schemes in ambient pressure-broadened molecular
species.
Surprisingly, despite the potential as seed sources for Yb-doped ber ampliers, there is only
a single experimental demonstration of a mode-locked laser at wavelengths above 1 m [10], in
which an internal absorber section was utilized to produce pulses of around 1 ps duration. Fiber
amplication of a gain-switched Fabry-Perot (FP) diode laser with 20 ps pulse duration has been
demonstrated very recently [11]. However, none of the experiments presented until today exploits
6
CHAPTER 1. INTRODUCTION
the superior wavelength exibility of mode-locked diode lasers in connection with amplication to
high power over a large spectral range.
In this work, we present the rst mode-locked diode laser around 1.06 m wavelength with subse-
quent amplication of the pulses in an ytterbium-doped ber amplier. The pulses are wavelength-
tunable over 45 nm and have a pulse duration of approximately 35 ps at a repetition rate of 1.4 GHz.
The average power of around 15 mW from the mode-locked diode laser is amplied to a maximum
average power of 9.5 W, corresponding to a peak power of more than 100 W.
The basic principles of lasers and diode lasers in particular are described in Chapter 2. Mode
locking, the technique that is employed in this work to generate short pulses, is explained and how
to achieve this in diode lasers is discussed in more detail. Furthermore, a short introduction to the
basic properties of ytterbium-doped ber ampliers is given, as such a ber is used to boost the
power level of the generated pulses.
To characterize the pulses before and after amplication, an autocorrelator is designed. In
Chapter 3, rst the autocorrelation technique is claried, before the specic autocorrelator that
is used in the actual work is explained in detail. After that, the instruments to perform spectral
measurements are listed and details about the scanning Fabry-Perot interferometer are given.
The experimental realization of the mode-locked diode lasers is described in Chapter 4. First the
experimental setup is presented and characteristics of the diode lasers in continuous-wave operation
are determined. During the course of the measurements with the mode-locked diode laser, spectral
and temporal characteristics are presented, together with output power and wavelength tuning
properties.
Chapter 5 deals with the amplication of the mode-locked pulses in an ytterbium-doped ber.
First, changes in the experimental setup are described, after which the results obtained from am-
plication of the mode-locked pulses are presented.
7
Chapter 2
Theory
Overview In this chapter a brief theoretical characterization of diode lasers and ultra-short pulse
generation in diode lasers is given. First, in section 2.1, the fundamental concepts of lasers in gen-
eral are described and of diode lasers in particular, which provides basic understanding of how diode
lasers emit a coherent beam. In section 2.2 several methods of pulse generation in diode lasers are
briey examined. Mode locking is set out theoretically in detail, after which specic details about
mode locking in diode lasers are outlined. Unless indicated otherwise, the majority of this theory is
taken from [12] and [13]. Finally, a short introduction to ytterbium-doped ber ampliers is given
in section 2.3.
2.1 Diode lasers
2.1.1 Lasers in general
The word laser is an acronym and stands for Light Amplication by Stimulated Emission of Radi-
ation. This explains roughly how a laser actually emits light.
In its simplest form a laser is formed by the amplifying medium and a set of mirrors to provide
optical feedback of the light into the amplier for growth of the light intensity. The amplifying
medium gains energy from a pump source, e.g. an electric current or a ash light. The light that
is created in the active medium travels back and forth between the mirrors and develops into a
beam with unique properties by an amplifying light-matter interaction, called stimulated emission.
A part of this light is transmitted through one of the mirrors and can be used for the specic
application.
A laser is quite dierent from a light bulb. Radiation from a light bulb is emitted in all directions,
whereas a laser concentrates the light that would usually be radiated in all directions into a single
high intensity beam in one direction. This is one of the unique properties of stimulated emission.
In addition to the direction, also the frequency, phase and polarization are all equal for all photons
in the beam. Another dierence, as compared to the light bulb, are the photon statistics. In a
light bulb, the output intensity uctuates randomly, because all light is generated by spontaneous
emission. In a laser, a repetitive pattern of output intensity is emitted because of the repetitive
stimulated emission occurring in the resonator once per round trip.
Pumping the gain medium excites, e.g., atoms, molecules or ions to higher energy levels. Elec-
trons decaying to lower levels can emit photons. These photons can be absorbed again by the
medium to excite another atom, but can also leave the medium. When they are not absorbed, they
8
CHAPTER 2. THEORY
can, on their way out, stimulate other exited atoms, molecules or ions to emit phase-coherently.
Without this process there would be no laser. When the rate of emission exceeds the rate of ab-
sorption, i.e. when is the gain in the system is greater than the total loss (absorption, outcoupling,
scattering, etc.), stimulated emission develops into a coherent beam. The dimension of the laser
resonator forms a condition for this emission, namely that a wave must overlap with itself after
one round trip, i.e. constructive interference within the resonator. Fields that satisfy this condition
become standing electromagnetic waves in the cavity and are called eigenmodes or simply modes.
Depending on the gain bandwidth, the number of longitudinal modes may be very large or as small
as only a few or just one, which is sometimes very advantageous. Studying the transverse beam
cross section, the light intensity can be found in dierent distributions. Such patterns are called
transverse electromagnetic modes (TEM). To indicate the TEM-modes, three indices can be used,
TEM
npq
, where n is the number of longitudinal eld nodes and p and q are the number of radial
and angular nodes respectively. Since the longitudinal mode number is generally very large for
optical frequencies, usually only the two latter transverse indices are used to specify a TEM-mode.
Higher order transverse modes are more dicult to focus, so often the TEM
00
mode, also named
fundamental Gaussian beam, is preferred.
2.1.2 Principle of operation of diode lasers
Semiconductor diode lasers distinguish themselves from the various other types of lasers by their
extremely small size and high eciency. With typical dimensions of less than a millimeter in all
directions, these solid state devices form compact, low-cost sources of coherent light. Due to their
high eciency, optical output powers of several tens of milliwatts can be produced by relatively
low electrical input powers.
As lasers in general, diode lasers consist of a gain medium and two mirrors. The active medium
of a diode laser is an advanced form of a p-n-junction, pumped by an electric current. Usually the
mirrors are the facets of the semiconductor chip.
When forward biased, the p-n-junction starts radiating due to the recombination of holes from
the p-region and electrons from the n-region. Additional non-radiative recombination expresses
itself as heat. The narrow depletion region of the junction in which the recombination process
takes place is called the active region. In principle, the spontaneous emitted light is incoherent and
therefore no lasing occurs. Electrons and holes can also absorb the emitted radiation, at which an
electron-hole pair is generated. Stimulated emission develops when the gain exceeds the loss, which
happens above the threshold injection current. A population inversion is achieved as the holes
and electrons in the active region can coexist for a rather long time (several nanoseconds) before
they recombine. If during this time a photon of exactly the right frequency passes, it forces the
hole and the electron to recombine and another photon is emitted with exactly the same properties,
frequency, direction, phase and polarization as the incident photon. Photons resonating between the
two mirroring edges of the junction maintain the stimulated emission and thus keep the laser eld
oscillating. When the diode is lasing, the optical output rises faster with increasing input current
than below the threshold current. This can be explained as follows. Increasing the injection current
causes an increase of injected charge carriers (electrons and holes) into the active region. Below the
threshold, this results in an increased charge carrier density, whereas above threshold the charge
carrier density remains the same. In the latter case, the increased injection current only favors the
optical output power.
With gain coecients generally between 5.000 m
1
and 10.000 m
1
and losses of approximately
2.000 m
1
[12], semiconductor lasers show much higher single pass gains than other types of lasers.
Therefore the size of diode lasers can be extremely small, usually less than a millimeter. In case the
9
CHAPTER 2. THEORY
facets of the diode form the resonator, the free spectral range of the laser is very large (1/
sep
=
c/2nd = 50100 GHz), i.e. the modes show a large spectral separation. Also the gain bandwidth of
these lasers is extremely broad (tens of nm), so generally multi-longitudinal mode operation results.
2.2 Ultrashort pulse generation in diode lasers
Pulses with picosecond duration are of high interest since decades. They enable a large range
of applications, see e.g. [3]. For example, nonlinear frequency conversion (e.g. second harmonic
generation, optical parametric oscillation (OPO), etc.) can be greatly enhanced in eciency [14, 9].
Several methods to generate short pulses in diode lasers are available, among them Q-switching,
gain switching and mode locking. With the Q-switching technique, the laser medium is pumped
while feedback from the resonator is prevented (low Q value of the resonator). A population
inversion occurs and a large amount of energy is stored in the gain medium. As the amount of
energy reaches a maximum level because of losses by spontaneous emission, the Q-switch (to a high
Q value) suddenly allows feedback and the intensity of light in the resonator builds up very quickly
because of stimulated emission. As a results, the energy stored in the medium then depletes very
quickly and a short pulse of light is emitted from the laser. Q-switching leads to pulses in the
nano- or picosecond regime with repetition rates of up to a few MHz. In gain switching, the gain is
modulated (in stead of the losses), bringing the carrier density within the active region from below
to above the lasing threshold. Typical pulse durations in the picosecond regime are obtained from
gain-switched diode lasers. A common feature of Q-switching and gain-switching is that subsequent
pulses lack phase relation, so they are suitable only for a limited number of experiments.
In particular, the synchronous pumping of OPOs require a phase relation between subsequent
pump pulses, thus a dierent approach to pulse generation is required. Mode locking of diode lasers
can provide the required phase relation. This is why mode locking is elaborated below.
2.2.1 Mode locking
The longitudinal modes in a Fabry-Perot laser cavity are separated by
sep
= c/2nd, in which nd
is the optical length of the cavity with the refractive index n and length d. If the gain bandwidth
is broader than this mode spacing, more than one longitudinal mode can, in principle, oscillate.
Assuming a cavity containing N light modes of, e.g., equal amplitude, E
0
, each of which has a
frequency and a phase , the total amplitude can be mathematically expressed as
E(t) = E
0
N

n=1
e
i(
n
t+
n
)
(2.1)
Normally the phases are independent of each other and for large N an intensity is expected,
which uctuates to some, possibly large, extent around an average value of N times the intensity
of one mode.
I(t) = |E(t)|
2
= E
2
0

n=1
e
i(
n
t+
n
)

2
= NE
2
0
(2.2)
The uctuation around this value has a repeating character induced by the modes circulating
through the cavity, producing the same output every round trip. For small N, beat patterns may
occur due to interference of the few modes.
10
CHAPTER 2. THEORY
Now consider a xed linear relationship between the phases of the laser modes. All oscillating
modes then are phase-locked to each other, and the process is termed mode locking. For this to
happen, an intra-cavity loss or gain modulator operating synchronously with the cavity round trip
frequency,
sep
, is required. The modulation generates side bands which give rise to an energy
transport between neighbouring modes. With it, phase information is transferred and after several
round trips, all modes agree on a common phase.
Inserting the linear phase relationship, e.g. as a constant phase oset from mode to mode,
=
n+1

n
, into Eq. (2.1), the combined eld amplitude can be rewritten as
E(t) = E
0
e
i(
N
t+
N
)

1 e
iN(t+)
1 e
it+

(2.3)
and the intensity becomes
I(t) = E
2
0

1 e
iN(t+)
1 e
it+

2
= E
2
0
sin
2
(N (t + ) /2)
sin
2
((t + ) /2)
(2.4)
where =
n+1

n
and =
n+1

n
.
0
Time
I
n
t
e
n
s
i
t
y
50 modes
5 modes
cw
resonator round trip time
Figure 2.1: Locked modes (upper), showing periodic maxima when the phases of all modes
match (lower).
From Eq. (2.1) and Eq. (2.4) and by taking, e.g.,
n
=
0
and = 0, Fig. 2.1 can be constructed.
The upper part shows the real part of the complex eld of ve laser modes as function of time
(green). The lower part shows the intensity as function of time for two dierent numbers of modes,
5 (magenta) and 50 (blue), and the normalized intensity in case the laser is operated in a single
mode with continuous wave (cw) output (red). The ves modes show an equal phase once per
round trip (t
sep
= 1/
sep
), whereas at all other times the modes interfere destructively. Also,
once per round trip, a pulse with large intensity occurs, containing the same amount of energy
as the continuous output per round trip, but concentrated in a small time window. More modes
have a shorter time of coincidence, which results in narrower peaks with larger peak intensities. By
11
CHAPTER 2. THEORY
inspecting Eq. (2.4), one obtains that the maximum intensity is increased by a factor of N over the
average intensity.
I(t)
max
= N
2
E
2
0
(2.5)
The shortest pulse duration at full width half maximum (FWHM) that can be generated is
related to the optical cavity length, nd, and the number of oscillating modes, N, as follows
t
p,min
=
2nd
Nc
=
2
N
=
1

sep
N

1
gain bandwidth
(2.6)
and is ideally as short as the reciprocal of the gain bandwidth. Experimentally generated pulses
that fulll this condition, i.e. pulses with a duration of the inverse gain bandwidth, are referred to
as bandwidth-limited pulses. For example, in case of pulses with a Gaussian temporal shape, the
minimum pulse duration is t
p,min
= 0.441/N. The value 0.441 is known as the time-bandwidth
product and depends on the pulse shape. The minimum attainable time-bandwidth product of a
secant hyperbolic squared pulse shape is 0.315 and 0.11 for a single sided exponential shape [15].
The actual pulse shape can be estimated by tting an experimental autocorrelation trace to a
theoretical one assuming a certain pulse shape.
Mode locking techniques
A number of mode locking techniques is available, which roughly can be divided into three groups.
Active mode locking
In active mode locking, the gain or loss of the laser is modulated. An external high frequency,
usually a radio frequency (RF) signal is applied to modulate either the laser gain or loss
directly or to drive an intra-cavity active device. Modulating the gain or loss must be done
exactly synchronized to the pulse round trip time through the cavity, t
sep
= 2nd/c. This
modulation causes sidebands to develop, with which phase information is transferred. As
this relates the phases of all modes, pulses are formed as described in the current section.
This technique enables ultrashort pulses in the picosecond regime to develop, with high phase
stability between successive pulses.
Passive mode locking
Mode locking can be obtained with a passive medium like a saturable absorber or a Kerr lens
inside the cavity. These intensity-dependent shutters are transmitting light if the intensity
is high, such that they require no external control (thus the form passive) because they are
controlled by the arrival time of the pulse itself. Mode locking is initiated by peaks in the
at rst randomly uctuating intensity inside the laser cavity. Since the higher intensities are
transmitted, the peaks, consisting of modes with equal phase, grow and eventually lead to
a single high-intensity peak traveling through the cavity by ltering out all other (wrongly
phased) mode superpositions. The nonlinear eect of saturable absorbers enables ultrashort
pulses in the femtosecond regime, providing that the gain prole is suciently large.
Hybrid mode locking
Hybrid mode locking is a combination of active and passive mode locking, where both an RF
signal and a passive medium are used to produce ultrashort pulses. It takes advantage of
active mode-locked stability and the saturable absorbers pulse shortening mechanisms.
12
CHAPTER 2. THEORY
2.2.2 Active mode locking in diode lasers
In active mode locking, modulation the loss or gain of a diode laser at a frequency equal to the
intermodal spacing,
sep
, results in optical sidebands to emerge. Each of these modes is being
driven by the modulation sidebands of its neighbours [16], and consequently, the phases of the
modes are locked to each other.
The intermodal spacing of diode lasers (typically 50 to over 100 GHz) is at the upper end of
existing RF technology due to the very small dimensions of the laser medium, which makes direct
modulation technically complicated. To decrease the spacing between the modes, lengthening the
cavity is an easy possibility, so mode locking can be performed at a much lower standard radio
frequency. Lengthening the cavity can be achieved by an external cavity, created by removing one
mirror from the diode laser via an antireection (AR) coating and replacing it by an external one
at some suitable distance.

M DL
M L DL DG L DL
F DL FBG DL
a)
b)
c)
d)
e)
f)
M FP L DL
Figure 2.2: External-cavity congurations of mode-locked diode lasers. DL diode laser, M
mirror, L lens, FP Fabry-Perot etalon, DG diraction grating, F ber, FBG Fiber
Bragg grating
Several commonly used external-cavity congurations are schematically illustrated in Fig. 2.2.
The rst experimentally mode-locked diode lasers used a simple cavity, where a curved mirror
reects the light back into the diode. Such a setup is drawn in (a). It has the disadvantage that
the that mirror curvature has to be chosen according to the applied frequency or vice versa. A
solution to this limitation can be realized using a collimating lens and a plane mirror (b) or a ber
(c). For additional selection of operation wavelength or for bandwidth reduction, implementing a
Fabry-Perot etalon (d), a diraction grating (e) or a ber Bragg grating (f) are commonly used
options.
13
CHAPTER 2. THEORY
Spectral properties
To reduce the required modulation frequency from 50100 GHz to lower values, the diode laser has
to be placed in an external cavity conguration. However, this causes additional complications.
A nite residual reectivity of the front facet, which always remains after antireection coating,
usually results in the formation of a three-mirror cavity. Such a composite cavity consists of the
two diode laser facets and the external reector. Then, also the reectivity of the front diode facet
determines the resultant spectrum of laser emission.
The composite Fabry-Perot resonator eect implies a non-trivial optical output spectrum, which
is a convolution of two Fabry-Perot cavity spectra. An example of such a spectrum of two combined
cavities is given in Fig. 2.3, where the intensity of the modes is plotted as function of the optical
frequency. The external mirror with the diode rear facet usually form a high-Finesse cavity with
Frequency
I
n
t
e
n
s
i
t
y

d

c
Figure 2.3: Typical output spectrum of a mode-locked diode laser with an imperfect antire-
ection coating on the front diode facet. The clusters of external cavity modes are separated
by the intermodal spacing of the diode laser.
larger mirror spacing, i.e. the resulting spectrum consists of rather sharp modes with a small
spacing,
c
. This spectrum is convoluted with the spectrum generated by the two laser diode
facets. As one of the facets is AR coated, the facets from a short, low-Finesse cavity, which
results in a spectrum of broad modes with a large spacing,
d
. The resulting spectrum exhibits a
clustered structure of the modes with the clusters spaced by
d
. Although the modes within such
a cluster are locked, separate clusters are not locked to each other, since they rise from statistically
independent noise sources of spontaneous emission. Generated pulses with such a spectrum have a
temporal substructure and are far from bandwidth limited.
Former studies have shown that a high quality AR coating with a residual facet reectivity
of less than 10
4
[17] is required to prevent mode cluster formation. Otherwise, a weak feedback
in combination with the high single pass gain is sucient to drastically reduce the lockable gain
bandwidth to less than the intermodal spacing of the diode cavity [16]. In that case, the modes of
only one of the clusters in Fig. 2.3 should be locked. If the pump power is high enough for modes of
more than one cluster to reach threshold, only the modes within a cluster are locked. Such emission
consisting of several clusters cannot produce bandwidth-limited pulses. A solution is to restrict the
very broad optical spectrum to a single mode cluster by a Fabry-Perot etalon, diraction grating
or a ber Bragg grating, as shown in Fig. 2.2 (d), (e) and (f).
14
CHAPTER 2. THEORY
Reducing the spectral bandwidth is one option to avoid the eects of the residual reectivity of
the front facet. Another solution is to use an angled-stripe semiconductor laser [18], in which the
gain channel is not exactly perpendicular with respect to the facets. This prevents the formation
of the second resonator between the diode facets. A third solution to the issue is to lock more than
one mode cluster, which can be realized by hybridly mode locking the laser by placing a saturable
absorber in the cavity. A broader spectrum and thus shorter pulses can be achieved.
Pulse quality
The time-bandwidth product, t
p
, is often used to indicate the quality of optical pulses. Band-
width or Fourier-limited pulses are theoretically the shortest pulses that can be obtained from a
given amplitude spectrum. From the mode locking theory follows that a linear phase spectrum
results in the shortest pulse. This minimized pulse duration yields the smallest possible time-
bandwidth product. For this number to give an indication of the pulse quality, the pulse shape
must be known. In the case of a sinusoidal gain modulation, Gaussian pulse shapes have been ob-
served in actively mode-locked solid state lasers. In diode lasers however, the strong carrier photon
interactions [13] and the carrier lifetimes in the order of the modulation period, causes signicant
dierent gain modulation proles, resulting in secant hyperbolic squared [13, 19] of single sided
exponential [15] temporal pulse shapes. In the case of mode-locked diode lasers with an imperfect
AR coating, which show optical spectra like in Fig. 2.3, the random phases of the mode clusters
lead to excessive values for the time-bandwidth product. Reducing the lasing spectrum to a single
mode cluster by inserting a Fabry-Perot etalon in the cavity or by replacing the external mirror by
a diraction grating can result in nearly bandwidth-limited pulses, with t
p
in the range of 0.3
to 0.6 [13, 15].
The time-bandwidth product often is reduced due to a wavelength chirp, which is frequently
observed in experimental work on actively mode-locked diode lasers [20]. Such a wavelength chirp
can be caused by refractive index changes. The carrier density (concentration of electrons and
holes) determines the refractive index of the diode and during the optical pulse emission the carrier
density is aected. The instantaneous frequency over the produced mode-locked pulse typically
decreases (red chirp) [20]. Hence the time-bandwidth product is far from limited. Because the
change of refractive index with carrier density generates a relatively linear chirp [20], the chirp is
found to be linear near the peak of the pulse [21]. Towards the wings of the optical pulse, the
refractive index change due to the electron temperature becomes more dominant, which results in
a nonlinear chirp [20]. According to [21], the amount of chirp and its linearity depends on the
wavelength and the spectral bandwidth.
To counteract the wavelength chirp, dispersion compensating optical components can be placed
into the external cavity like, e.g., a chirped mirror [22]. Linearly chirped pulses can be eectively
compressed using well-known optical pulse compression techniques, like a grating, prism pair or
such chirped mirrors.
2.3 Pulse amplication
High power pulses in the near-infrared (NIR) region, 13 m, are interesting for many applica-
tions, such as spectroscopy, laser and amplier pumping and frequency conversion. Diode-pumped
ytterbium-doped ber ampliers oer a large gain-bandwidth, high eciency and high reliability.
Such optical ampliers are perfectly suitable to reach high powers in the 1-1.1 m region utilizing
low power seed sources [23, 24]. Pumping an OPO with pulses in he wavelength region 1-1.1 m
can give access to the mid-IR region [4, 5, 25], which is interesting for, e.g., spectroscopy purposes.
15
CHAPTER 2. THEORY
Ytterbium-doped (Yb-doped) silica has very broad absorption and emission spectra. Both
absorption and emission as function of wavelength are shown in the right-hand side of Fig. 2.4.
Both spectra show a narrow peak at 975 nm. The left-hand side of Fig. 2.4 shows the energy levels
0
4000
8000
12000
E

(
c
m

1
)
g
f
e
d
c
b
a
2
F
5/2
2
F
7/2
Figure 2.4: Energy levels (left) and absorption and emission spectra [26] (right) of ytterbium-
doped silica.
of Yb-doped silica. The main absorption and emission peak at 975 nm corresponds to an energy
transition between the lowest levels of both manifolds in the energy level structure. Absorption
below this wavelength mainly results from transitions from level a to f and g, whereas transitions
from level e to b, c and d are responsible for the emission at higher wavelengths [27].
When amplication of a particular wavelength is required, a pumping range of 8001064 nm
is available. However, gain is generated only at wavelengths longer than the pump wavelength, so
amplication of pulses in the 1-1.1 m region requires the pump wavelength to be below 1 m.
High output powers can be obtained from so-called double-clad bers. A single mode Yb-doped
core is surrounded by a larger multi-mode undoped inner cladding. The pump light is sent into
the multi-mode inner cladding and due to the overlap these modes have with the doped inner core,
pump light can be absorbed there. A drawback of cladding pumping is that three-level transitions
cause signicant reabsorption of signal in weakly pumped regions and hardly any amplication is
achieved. However, for wavelengths above 1040 nm, four-level transitions dominate and ecient
cladding pumping is possible [28].
The lower pump absorption per unit length, by cladding pumping in stead of core pumping,
can be overcome by a longer ber length or a larger absorption cross-section. Because of its large
absorption cross-section, the sharp absorption line at 975 nm is an ideal pump wavelength for
cladding pumped Yb-doped ber ampliers.
16
Chapter 3
Measurement
Overview In Chapter 2 mode locking as a method of pulse generation was introduced. The
present chapter will discuss how these ultrashort pulses can be measured using the autocorrelation
technique (section 3.1). A general theoretical description is given and certain practical issues are
discussed. Finally, the design of an autocorrelator is worked out and the details on the actual
experimental setup used in this work are given. The theoretical approach of autocorrelation used
in the following sections is mainly based on [29]. In addition to the measurement of the temporal
properties, Chapter 3 also deals with the measurement of spectral properties of diode lasers. In
section 3.2, two ways of doing this are presented, namely the optical spectrum analyzer and the
Fabry-Perot interferometer.
3.1 Ultrashort pulse measurement
Short light pulses of durations down to about one picosecond, can today be measured by optoelec-
tronic methods. However, to visualize the signal from a photodiode with a response time of a few
hundred femtoseconds, an oscilloscope with a response time of the same order is required. Another
possibility is to use a streak camera, a high precision device that not only gives the pulse width,
but also details of the temporal prole.
The main drawback of these methods is that they require very expensive equipment. Because
of that, other techniques are considered to measure ultrashort events. One such method is opti-
cal autocorrelation, a widely used diagnostic technique to determine whether a laser is actually
producing short pulses.
3.1.1 Intensity autocorrelation
In order to measure the temporal prole of an ultrashort event, an even shorter reference event of
known shape is required. Let I
s
(t) be the unknown ultrashort pulse and I
r
(t) be a reference pulse.
With the delay parameter , the intensity cross-correlation is dened as
A
c
() =

I
s
(t)I
r
(t )dt (3.1)
The measured signal, A
c
(), diers from zero only if the pulse to be measured and its reference
pulse overlap temporally. The reference pulse represents a measurement window which is shifted
17
CHAPTER 3. MEASUREMENT
across I
s
(t) in order to scan the temporal prole. With the idealistic assumption of I
r
(t) being a
delta-function, A
c
() is identical to I
s
(t). However, all phase information of the analyzed pulse is
lost in this intensity cross-correlation through conjugate multiplication.
The prerequisite of a reference signal shorter than the optical pulse to be analyzed is often not
available. Even if it is, the shape of this signal must be determined. A very short signal in the form
of the optical pulse itself however, is available at all time. Although the pulse of course is not shorter
than itself, one should consider to use it as the reference signal. In this case I
r
(t) = I
s
(t) = I(t)
and Eq. (3.1) becomes the intensity autocorrelation.
A() =

I(t)I(t )dt (3.2)


As can be understood from Eq. (3.2), the autocorrelation function is always symmetric. De-
spite the fact that the autocorrelation provides little information about the pulse shape and no
information about the phase, the pulse duration can be estimated. For this, most often a certain
pulse shape is assumed and tted to the autocorrelation trace. With the help of the known ratio
between the FWHM of the autocorrelation and that of an assumed pulse shape, the approximate
pulse duration is determined.
3.1.2 Interferometric autocorrelation
In practice, an optical autocorrelator is often realized with a setup based on the Michelson in-
terferometer. Two replicas of the input pulse are generated, e.g. using a 50% beamsplitter, and
then overlapped again with a variable delay, . In the Michelson interferometer, one replica is sent
back to the beamsplitter via a static arm, i.e. E(t), while the other travels a variable path length,
E(t ).
From the theory of the Michelson interferometer, one can nd the eld autocorrelation, dened
as
A() + c.c =

E(t)E

(t )dt + c.c. (3.3)


A rather slow detector at the output of the interferometer can be used to experimentally measure
this eld autocorrelation. In fact, for pulse trains, the detector gives an output signal which is
averaged over many pulses. The signal is integrated, which results in the measured intensity being,
expressed as
G
1
() =

|E(t) + E(t )|
2
dt
=

{I(t) + I(t )} dt + A() + c.c.,


(3.4)
where the term A() +c.c. represents the eld autocorrelation as given in Eq. (3.3). From this rst
order autocorrelation still no reliable information about the pulse duration can be obtained. Noise
sources, chirped pulses or ultrashort pulses may produce exactly the same output, as it reveals only
the coherence time of the input signal, which is short in all these cases.
To acquire a correct pulse duration from the interferometer a nonlinear process in front of the
detector is necessary. An often used, suitable nonlinear process is a second harmonic generating
(SHG) crystal, in which the second harmonic is proportional to E
2
. The second order nonlinear
eects can be accredited to an energy transfer between electromagnetic elds of dierent frequencies
18
CHAPTER 3. MEASUREMENT
within the material [30]. Second harmonic generation is a special case of the more general conversion

, namely that

, so:

= 2

(3.5)
In this case the incoming light at the fundamental frequency,

, is converted into light with twice


that frequency,

. Second order nonlinearities only occur in materials with a noncentrosymmetric


crystal structure such as Beta Barium Borate (BBO).
Employing such a second order process, the detected signal changes to the second order inter-
ferometric autocorrelation.
G
2
() =

|E(t) + E(t )|
2

2
dt (3.6)
Decomposition of this equation after substituting the complex elds by the real amplitude, ,
and phase dependence, , via E(t) = (t) exp
i(t)
, shows that the second order autocorrelation
has four components.
G
2
() =
constant
. .. .

I
2
(t) + I
2
(t )

dt +
intensity autocorrelation
. .. .
4

I(t)I(t )dt
+ Re

I
2
+ I
2
(t )

(t)(t )e
i
e
i((t)(t))
dt + c.c.

+ Re

2
(t)
2
(t )e
2i
e
i((t)(t))
dt + c.c.

. .. .
interferogram of second harmonic
(3.7)
These four components of the interferometric autocorrelation of an arbitrary pulse as a function
0
0
2
4
6
8
Delay
A
m
p
l
i
t
u
d
e

(
a
.
u
.
)
0
4
2
0
2
4
Delay
A
m
p
l
i
t
u
d
e

(
a
.
u
.
)
Figure 3.1: (left) The interferometric autocorrelation (blue) and its intensity autocorrelation
(gray). (right) The four components of the interferometric autocorrelation: the constant back-
ground (blue), the intensity autocorrelation (cyan), the eld autocorrelation (green) and the
second harmonic autocorrelation (magenta).
of the delay are graphically represented on the right-hand side of Fig. 3.1 as function of the delay,
19
CHAPTER 3. MEASUREMENT
in addition to the combined trace on the left. The four components are centered around three
oscillation frequencies, namely: zero, and 2. The rst two components of the expansion are
at zero frequency and together these are referred to as the intensity autocorrelation (cyan) with
background (blue). At = 0, a maximum value of 6


4
dt is reached, whereas at = a
constant background value of 2


4
dt is found. The peak-to-background ratio of this intensity
autocorrelation is thus 3 to 1 (gray line in the left gure). The other two terms represent the
oscillating components, of which the former is the eld autocorrelation (green), oscillating at ,
and the latter the second harmonic autocorrelation (magenta), at 2, with maxima of 8


4
dt and
2


4
dt respectively.
Constructive interference of all components together results in a maximum value of 16


4
dt
at = 0. Taking into account the constant background level of 2


4
dt explains the peak to
background ratio of 8 to 1 for the interferometric autocorrelation (blue). The sum-frequency signal
as a function of relative time delay is proportional to the shape of the pulse. As any autocorrelation,
the interferometric autocorrelation is also symmetrical.
Unlike the intensity autocorrelation the interferometric autocorrelation does contain some phase
information. The ability to quantitatively measure a linear chirp is the main advantage of this,
which is demonstrated below.
Linear chirp
Unchirped Gaussian pulses show correlation traces like in Fig. 3.1. The lower and upper envelopes
split evenly from the constant background level, indicating the coherence time of the pulse is as
long as the pulse itself.
Next, consider a chirped Gaussian pulse, (t) =
0
exp

(1 + ia)(t/
G
)
2

. Depending on the
amount of chirp, dened by the chirp parameter a, the interference pattern is smaller and wings
appear at both sides. These wings exactly correspond to the shape of the intensity autocorrelation
and the width of the interference pattern points out the shorter coherence time. The level at which
the interference pattern starts in relation to the peak level indicates to what extent the pulse is
chirped.
The interferometric autocorrelation of a linearly chirped Gaussian pulse is expressed as
G
2
() = 1 + 2 exp

+ 4 exp

a
2
+ 3
4

cos
a
2

2
cos
+exp

(1 + a
2
)

cos 2
(3.8)
and is shown in Fig. 3.2. Equation 3.8 shows the same constant background and the same intensity
autocorrelation component as the unchirped Gaussian pulse (gray line). A dierence is noticed for
the other two terms, which include the chirp factor a. This term causes the interference pattern
to narrow (magenta line) and the intensity autocorrelation to partially appear. The right-hand
side of Fig. 3.2 shows the envelopes of the autocorrelation for three values of the chirp factor. It
demonstrates that a larger chirp factor causes a larger narrowing the interference pattern.
As was demonstrated in this section, interferometric autocorrelation is capable of revealing
more information about the pulse than the intensity autocorrelation technique. For a practical
application, usually this type is preferred.
20
CHAPTER 3. MEASUREMENT
0
0
2
4
6
8
Delay
A
m
p
l
i
t
u
d
e

(
a
.
u
.
)
0
0
2
4
6
8
Delay
A
m
p
l
i
t
u
d
e

(
a
.
u
.
)
a=0
a=2
a=8
Figure 3.2: Interferometric (magenta) en corresponding intensity autocorrelation (gray) of
a chirped Gaussian pulse (left) and upper and lower envelopes for three values of the chirp
parameter a (right).
Types of autocorrelators
Practical measurement of picosecond or femtosecond pulses requires one variable arm of the Michel-
son interferometer. The most common method to produce an optical path dierence is mounting
a mirror on a linear motion device like an audio loudspeaker or a motorized translation stage and
driving that with a sinusoidal signal. A second method utilizes a material with a signicantly larger
refractive index than air, which is rotated to induce a delay. Another method relies on a rotating
platform containing one or more optical components to produce a varying path length [31].
Pulses longer than a few picoseconds require large scan ranges. For example, for a delay of
100 ps, a path length dierence of 3 cm has to be scanned. The scan range is limited by the type
of autocorrelator. Up to 300 ps of delay can be obtained by a rotating-mirror autocorrelator, while
larger scan ranges can be reached with a motorized linear translator, although the size of such a
device can be considerable.
Because of the large scan range required to measure the expected picosecond pulses from our
diode laser and the exibility to adapt to dierent pulse lengths, we decided to build a rotating-
mirror autocorrelator, which is explained below.
3.1.3 Rotating-mirror autocorrelator
The rotating-mirror autocorrelator consists of a Michelson interferometer where the variable arm
employs two antiparallel mirrors mounted on a rotating platform symmetrically to its rotation axis.
The setup is schematically illustrated in Fig. 3.3.
Splitting the incoming pulses by means of a beam splitter is the rst step in the autocorrelator.
One beam is transmitted to mirror M4 and the other is reected to the rotating platform. The two
mirrors on the platform, M1 and M2, steer the beam to a third xed mirror, M3, from which it is
retroreected back to the beam splitter, again via the two rotating mirrors. The returned pulses
are combined with the ones returning from M4 and focussed into a SHG crystal where the second
harmonic signal is generated. The power of the SHG signal depends on the extent of overlap. After
21
CHAPTER 3. MEASUREMENT

PMT
M
M
BS SHG
crystal
delay
L L
h
d
r
p
4

p
3

p
1

p
2

Figure 3.3: Schematic setup of a rotating-mirror autocorrelator, utilizing a rotating platform


with two mirrors.
the fundamental wavelength is ltered out, the second harmonic signal is sent onto a detector. The
second harmonic is usually very weak, so often the detector is chosen to be a photomultiplier tube.
Due to the rotation of the platform, the path length, L, changes. The path length should be
varied around the point at which the path length in both arms is equal, i.e. the point at which the
maximum overlap in the crystal occurs. Calculation of the path length of the variable arm as a
function of the rotation angle of the platform, , is relatively easy, once four crucial spacial points
are dened. These are four points at which the beam intercepts the beam splitter, the mirrors on
the platform and the xed mirror, p
1
p
4
in Fig. 3.3. Taking the origin to be the axis of rotation,
a mathematical expression of these points in their x and y coordinates can be formulated as
x
p
() =

C
1
htan( + ) + r [sin() cos() tan( + )]
htan( + ) + r

3 4 cos
2
( + )

[sin() cos() tan( + )]


C
2

(3.9)
y
p
() =

h
h
h 4r sin() sin( + )
h 4r sin() sin( + )

(3.10)
in which represents the mirror angle with respect to the line, of length 2r, through the centers
of both mirrors [32]. h is assigned to the incoming beam height, that is the vertical distance in the
view of Fig. 3.3 from the center of the beam splitter to the origin of rotation. The locations of the
beam splitter and of M3 determine the constants C
1
and C
2
, but they do not aect the path length
dierence L. Making use of these coordinates, the path length, L(), can be easily derived.
L() =

[x
p
1
() x
p
0
()]
2
+ [y
1
() y
0
()]
2
+

[x
p
2
() x
p
1
()]
2
+ [y
2
() y
1
()]
2
+

[x
p
3
() x
p
2
()]
2
+ [y
3
() y
2
()]
2
(3.11)
From [32] the optimum mirror angle and and separation can be taken to maximize the path
length dierence, L. It appears that L is maximum for a mirror separation of little more than
22
CHAPTER 3. MEASUREMENT
twice the mirror width. For any mirror angle and separation, the optimum incoming beam height,
h
opt
, can be calculated to maximize L [32]:
h
opt
=
2r sin()

(d/2)
2
+ dr cos() + r
2

d
2
+ r cos()

(3.12)
Usually, the angular velocity of the rotating platform is constant, leading to the drawback of
a non-uniform rate of delay. The smaller the mirror separation, the higher the nonlinearity. In
[31], the nonlinearity NL over the whole scan is estimated to be NL d/8r, leading to a change
of delay rate of 25% for d = 2r. In contrast to pulses in the fs regime, for pulses in the ps
regime this nonlinearity cannot be neglected if a highly accurate autocorrelation trace is required.
Compensation is possible using an interferogram of a cw laser.
Experimental realization
For the characterization of the ultrashort pulses generated in de course of this work, we constructed
home-made rotating mirror autocorrelator. Pulse durations of several tens of picoseconds, i.e. a
spatial length of about 2 cm, are to be measured, so a path length dierence of a least a threefold
of that is required. The mirrors on the rotating platform have a diameter of d = 6.3 cm and are
separated by 2r = 7 cm. The angle at which the mirrors are mounted on the rotating platform is
= 55

. These values allow a path length dierence corresponding to about 200 ps to be traced.
The SHG crystal used is a rather long BBO crystal with dimensions of 5x5x12 mm, cut under an
angle of 22.8

, for type I second harmonic generation of 1064 nm to 532 nm. Instead of a fundamen-
tal lter and a photomultiplier tube, a PerkinElmer channel photomultiplier (CPM) module (type
MD983) is used. Its spectral sensitivity is in the range 185650 nm, so the fundamental wavelength
of around 1060 nm remains undetected, making the fundamental lter abundant. This module is a
newly developed detector, as an enhancement of a conventional photomultiplier tube. It oers an
extremely low dark current and a very high anode sensitivity, both one order of magnitude better
as compared to conventional PMTs. The CPM uses a unique detection principle by converting a
very low light level into electrons via a semitransparent photocathode. On their way to the anode,
the electrons move through a narrow, curved semiconductive channel, hit the inner wall and emit
secondary electrons upon every bounce. This process repeats itself several times along the path
resulting an avalanche eect with a very high gain.
For calibration of the autocorrelator, an event of a known temporal duration is required as
calculation of the path dierence gives only an estimation of the practical scan range. In this work,
the pulses generated within the semiconductor chip itself have been used for that purpose. Their
temporal separation can be veried from spectral measurements. This will be explained in detail
in section 4.3.3.
Evaluation of the recorded traces
From the recorded traces by the autocorrelator, the pulse durations have to be estimated. The
traces are recorded by an oscilloscope (Tektronix TDS2022) and stored in data les. These data
les contain the time axis of the oscilloscope and the measured output of the CPM in Volts. We have
written a MATLAB program to analyze the data les. It generates a plot of the autocorrelation
trace, ts the data by a chosen function (see section 2.2.2) and calculates the pulse duration based
on that t. Once the scan setup and trace calibration parameters are entered, our program can
calculate the delay time and the normalized signal levels directly from the data le. A plot of the
23
CHAPTER 3. MEASUREMENT
autocorrelation trace is generated as well as a plot of a t for a Gaussian or secant hyperbolic
squared function.
From the autocorrelation traces recorded with this autocorrelator, no clear pulse prole can be
determined by comparing it to ts with Gaussian or sech
2
shaped pulse. A Gaussian pulse shape
assumption results in longer pulse durations than with a secant hyperbolic squared assumption.
To avoid overestimation of the shortness of the pulses, all traces in this thesis are tted assuming
a Gaussian pulse shape, even though the actual pulse shape may be closer to the sech
2
function.
The pulse duration for a sech
2
function, calculated using the same tting procedure, is mentioned
several times for comparison.
3.2 Spectral measurements
Apart from an estimation of the phase and temporal pulse properties, information about the op-
tical spectrum is desired for a better characterization of the laser. For coarse, but quantitative
measurements, an optical spectrum analyzer (OSA) is used, whereas ner details are revealed by a
scanning Fabry-Perot interferometer.
3.2.1 Optical spectrum analyzer
Quantitative measurements of the optical spectrum are performed using two optical spectrum
analyzers
1
. The spectral bandwidth of these grating-based devices covers the wavelength range
from 600 nm to 1750 nm. The incorporated detectors have a huge dynamic range, which enables
the measurement of light powers from 20 dBm (100 mW) down to -90 dBm with 0.3 dB accuracy.
The rst of the two (Ando AQ6317) has a wavelength resolution of 0.015 nm with an accuracy of
0.05 nm. The second (Ando AQ6315A) diers from the rst by its resolution of 0.05 nm and its
spectral range, which is enhanced by 250 nm into the UV.
Even though the resolution is not sucient to resolve the single longitudinal modes of a laser
resonator of several centimeters in length, it is sucient for the large intermodal spacing of the
diode laser itself, which in this case is of particular importance. To be able to make fully use of the
high resolution, the light is coupled into the spectrum analyzer using a single mode ber. These
outstanding features allow high resolution and sensitivity measurements of the optical spectrum.
3.2.2 Scanning Fabry-Perot interferometer
In order to resolve the longitudinal modes of the external cavity that we set up, we use a scanning
Fabry-Perot interferometer. Such an instrument usually consists of two partially transmitting mir-
rors, in between which light waves bounce back and forth [33]. This multiple-beam interferometer is
illuminated along the axis, such that incident light can interfere with itself within the Fabry-Perot
cavity. To get a signal, the interferometer cavity has to be lled with light. This can only be the
case if there is positive interference of the waves that oscillate between the mirrors. Standing waves
are formed as a result of the optical distance between the mirrors, nd, being an integer multiple of
half a wavelength. With m being the order of interference, this can be expressed as
nd = m/2, (3.13)
which shows that the separation between the mirrors determines the transmitted wavelength. By
smoothly varying the position of one of the mirrors along the optical axis, this characteristic feature
1
Which OSA was actually used for an experiment was based on the availability at that specic day.
24
CHAPTER 3. MEASUREMENT
can be exploited to continuously tune the interferometer. Usually, the mirror separation is scanned
over several wavelengths by a piezoelectric transducer, leading to a repetitive (because of the
dierent orders of constructive interference) output pattern. The frequency dierence between
adjacent peaks of two successive interference orders of the same incident wavelength is termed the
free spectral range (FSR). For a plane mirror Fabry-Perot interferometer, i.e. one that uses two
at mirrors aligned exactly parallel, the FSR is given by FSR = c/2nd. It denes the frequency
bandwidth over which a measurement is possible without overlapping dierent interference orders,
which is normally in the order of some 100 MHz to several GHz. Meaningful measurements can
only be obtained if the FSR is greater than the spectral bandwidth of the input beam. On the
other hand the highest resolution measurement requires the FSR to be as small as possible.
Another type of Fabry-Perot interferometer uses two concave mirrors with radii of curvature
equal to their separation. This type is called a confocal mirror Fabry-Perot interferometer. The
FSR of these is directly related to the mirror curvature by FSR = c/4nd. In this case the FSR
diers from that of a plane mirror Fabry-Perot interferometer, since the incident light falls back
upon itself only after four traversals in stead of two, which means that the transmitted spectrum
is reproduced after changing the mirror distance a quarter of a wavelength.
A common measure to quantify the performance of the Fabry-Perot interferometer is the ratio
of the distance between two adjacent maxima to the width of the maximum (FWHM), which is
called the nesse of the cavity, F. The nesse is highly inuenced by the reectance, R, of the
mirrors. As the distance between two maxima corresponds to a wavelength dierence of , or to a
frequency dierence of 2, the nesse can be calculated as follows:
F =

R
1 R
(3.14)
Higher nesse values enable higher resolution measurements, but reduce the transmission of the
incident light as a consequence, so choosing an appropriate mirror reection is eminent.
Experimental realization
Two confocal mirrors with a radius of curvature of 25 mm in a confocal setup form the cavity of the
scanning Fabry-Perot interferometer we use. It has a free spectral range of 3 GHz, corresponding to
a wavelength range little more than 0.01 nm at a wavelengths of about 1 m. One of the mirrors is
mounted on a piezoelectric activator, allowing it to be shifted several micrometers. A high-voltage
supply, fed by a signal generator, drives this mirror in a triangular way back and forth, varying the
cavity length to scan the frequency range of 3 GHz. With a mirror reectance of 98%, the nesse
can be calculated to be around 150, leading to a minimum resolvable bandwidth of approximately
20 MHz. In practice however, the minimum resolvable bandwidth reduces to about 100 MHz, but
is still very suitable to distinguish laser modes of about 1 GHz apart.
25
Chapter 4
Experimental results
Overview This chapter describes the experimental setup, the process of mode locking a diode
laser and the measurement results. At rst the experimental setup is explained in detail. A com-
mercially available single stripe diode laser rst is characterized in continuous operation, after which
it is actively mode-locked, which should lead to pulses with picosecond duration. Optical properties
and pulse durations are examined.
4.1 Experimental setup
4.1.1 Diode laser setup
Active mode locking of a diode laser by modulating the input current at RF frequencies requires
a longer cavity than that of the semiconductor chip itself. Therefore the diode laser has to be
positioned in an external cavity conguration. The diverging output beam is collimated using a
lens, CL (Thorlabs C220 TM-B, f = 11 mm AR coated for wavlengths 6001050 nm). Figure 4.1
shows the diode laser placed in an external cavity, where feedback is provided by a diraction
grating in Littrow conguration. Such an external Littrow cavity provides a coarse selection of the
operation wavelength by adjusting the angle of the feedback grating.
For our experiments, presented in the following section, two diode lasers of identical structure
and similar specications have been compared and tested for their suitability for generating mode-
locked pulses. The two GaInAs diode lasers are SAL1060-60 AR coated single stripe diode lasers
from Sacher Lasertechnik, SN281 and SN286. The front facet of these diode lasers is anti-reection
coated, necessary to facilitate mode locking in diode lasers, as was discussed in section 2.2.2. With
residual reectivities specied to be R = 4 10
5
for the SN281 and R = 2 10
6
for the SN286,
both seem to fulll the minimum requirement of R = 1 10
4
[3], demanded because of the large
single pass gain of diode lasers.
The output of the external cavity laser is sent through an optical Faraday isolator (Linos FI1040-
TI), to prevent undesirable optical feedback into the diode laser. Around the Faraday isolator, two
plano-convex cylindrical lenses, ZL
1
(f = 100 mm, AR810) and ZL
2
(f = 50 mm, AR1064) form a
telescope to shape the beam and reduce the beam diameter. This beam shaping is required such
that it can be more easily analyzed or, later, focussed into the ber amplier core. The lens in
front of the Faraday isolator is tilted slightly to avoid reections back into the diode.
Research on the optical properties of the output is performed by three measurement devices, the
rotating mirror autocorrelator, an optical spectrum analyzer and the Fabry-Perot interferometer.
26
CHAPTER 4. EXPERIMENTAL RESULTS

Optical
spectrum
analyzer
Bias-T
DG
M
HR AR
RF
DC
M
BS
BS
ZL
1

M
Scanning
Fabry-Perot
interferometer

Autocorrelator
ZL
2
FI
CL DL
Figure 4.1: Experimental setup for mode locking a diode laser and characterizing its output.
DC DC current source, RF RF current source, DL diode laser, HR Highly reective,
AR Antireection coated, CL Collimating lens, DG diraction grating, M mirror,
ZL Cylindrical lens, FI Faraday isolator, BS Beam splitter.
All these are described in Chapter 3. A noteworthy, unique feature of our setup is that all three
optical instruments can be monitored simultaneously, so realigning can be kept minimal.
4.1.2 RF electronics
As mentioned before, active mode locking a diode laser by modulating the gain requires an RF
modulation signal to be superimposed onto a DC bias current. A device capable of doing that
is a so called bias-T, from which the combined signal has to be coupled into the diode laser. To
ensure an excellent coupling and to avoid reections back into the RF source, the circuit has to be
impedance matched to the RF source at the desired frequency. A home-made circuit has been used
for both purposes. The circuit to supply the combined current to the diode laser contains a bias-T
and a part to take care of the impedance matching the diode laser to the input RF frequency. This
circuit is drawn in Fig. 4.2. The DC bias current is supplied by an ILX Lightwave LDC-37248
laser diode controller. Temperature stabilization of the diode laser by means of a peltier element
integrated in the housing, is also achieved using this controller. The RF signal is provided by a
synthesized sweep generator (Anritsu 69147A, 0.0120 GHz, max. 20 dBm). To avoid possible
27
CHAPTER 4. EXPERIMENTAL RESULTS
reections to reach the generator, a Philips 2272 162 03951 circulator (1.72.3 GHz, -20 dB), acting
as an isolator due to a 50 termination on the third port, is used. The RF signal is amplied by
a Trontech P1020-33 amplier (12 GHz, 33 dB) to reach high modulation powers.

DC
RF GND C
1
C
b

C
a

LD
L
Bias-T Impedance match
Figure 4.2: Bias-T circuit including an impedance matching addition to ensure high coupling.
DC DC current source, RF RF current source, C Capacity, L Inductor, LD Laser
diode.
Impedance is dened as the total opposition to current ow in an alternating current circuit.
It diers from simple resistance in that it takes into account a possible phase oset between the
current and the voltage, i.e. it can change value with frequency. Usually, the impedance is expressed
in complex notation, Z = R+jX, where j is the imaginary unit, R the ohmic resistance and X the
reactance. In case of a sinusoidal signal of angular frequency , inductors have an impedance of
jL and capacitors 1/jC, thus the higher the frequency, the larger the inductive and the smaller
the capacitive impedance.
Maximum power transfer in high-frequency circuits happens when the impedances of dierent
devices are matched. In this case the impedance of the diode laser has to be matched to the
impedance of the RF source. Apart from the maximum power transfer, unwanted reections are
another important reason to match the impedances. High power reections back into the RF source
may cause damage to that apparatus, as well as low power reections may distort the input signal.
Two adjustable capacitors are used for this, C
a
and C
b
in Fig. 4.2. The rst, C
a
, is to compensate
the induction originated from the parallel inductor and capacity, L and C
1
respectively. This
way, the impedance is matched to that of the input signal. The second adjustable capacity, C
b
,
compensates for the induction in the supply channels to the diode. Summarized, the two adjustable
capacitors are required to let the circuit accept the chosen input frequency.
Superimposing the RF signal on top of a DC current is accomplished by means of a bias-T,
integrated in the circuit of Fig. 4.2. Via this bias-T, the DC and RF sources are connected to
the laser diode by separate paths. That for the DC signal is equipped with an inductor, L, which
transmits low frequency signals, but is of high impedance for high frequency signals. The RF
source on the other hand, is connected through a capacitor, C
1
, which acts just the opposite, low
frequencies undergo a high impedance whereas high frequencies can transmit easily. This way, the
signals from each source are prevented from entering the other source, which otherwise can load
each other in stead of the diode and possible leads to the damaging of one of the sources.
To correctly adjust the two capacities C
a
and C
b
, we use setup in Fig. 4.3. The circuit is
frequency swept by a very low RF power of 30 dBm, transmitted through a directional coupler.
Backward reections in the transmission line at all frequencies in the sweep are partially separated,
through a known coupling loss. The source of the frequency sweep is integrated in the RF spec-
28
CHAPTER 4. EXPERIMENTAL RESULTS

Directional coupler
RF source
RF spectrum analyzer

Bias-T DC
DL
reflections RF
0.5 1 1.5 2
90
80
70
60
50
Frequency (GHz)
R
e
f
l
e
c
t
e
d

p
o
w
e
r

(
d
B
m
)
Figure 4.3: (left) Setup for impedance matching the diode laser to an RF frequency. DC
DC current source, DL Diode laser. (right) Reected power with a steep dip at the impedance
match frequency.
trum analyzer (Agilent E4407B), which also measures the reected power. The result of such a
measurement is presented in the right-hand side of the gure. It shows the power reections on
a logarithmic scale over a frequency range of 500 MHz2 GHz. At 1.4 GHz a steep dip of more
than 25 dB below the surrounding powers is observed. At this frequency, minimal reections occur,
indicating that the impedances of the source and the diode are matched. By adjusting the two
capacitors, a steep dip in the RF spectrum can be positioned exactly at the desired frequency any-
where between 1 and 1.5 GHz for this particular diode (SN286). Once the impedances are matched,
a large RF power may be applied at that specic frequency without the danger of damaging the
source.
4.2 Continuous wave diode laser characteristics
For preliminary characterization, the properties of the diode lasers are rst investigated in continu-
ous wave (cw) operation, i.e. no RF power is supplied to the diode. The power-current characteristic
and the output spectrum are of particular importance. From the power-current curve, the thresh-
old value can be deduced, whereas the tunability and mode distance can be determined from the
spectrum.
First of all the laser is activated without any other optical components besides the collimating
lens. The temperature of the diode is stabilized at 19

C. While increasing the injection current,


the optical output power is monitored using a power meter (Newport 1815-C with detector head
883-SL). Figure 4.4 shows the power-current curves for two cases, namely for the free running laser,
i.e. without feedback (purple), and for the case that grating feedback is provided (blue).
In both plots, the power remains low (below 1 mW) until a certain injection current is reached.
Beyond that threshold value, the power increases approximately linear with the current. Also the
slope increases, indicating the output has changed from spontaneous to stimulated emission. With
grating feedback, the threshold value lowers from 43 mA to 33 mA. Looking back on section 2.1.2,
the shift of the threshold current can be explained. Optical feedback allows stimulated emission
to develop at a lower injection current because of the higher optical density due to the feedback.
Secondly, feedback aects the slope in the lasing regime. With feedback, the slope decreases because
29
CHAPTER 4. EXPERIMENTAL RESULTS
0 20 40 60 80 100
0
5
10
15
20
25
30
35
40
Injection current (mA)
O
p
t
i
c
a
l

o
u
t
p
u
t

p
o
w
e
r

(
m
W
)
no feedback
feedback
Figure 4.4: Power-current characteristics of the SAL 1060-60 SN286 diode laser for the case
of the diode laser without feedback (purple) and with grating feedback (blue) at 1040 nm.
of the charge carrier density is still relatively low at the threshold, but stays constant when further
increasing the injection current. This results in a less rapidly growing output power as compared
to the case without external feedback. At 100 mA, a power of 30 mW with and of 34 mW without
feedback is reached, despite the lower threshold injection current in the case of optical feedback.
Furthermore, the optical spectra with and without optical feedback are recorded using the OSA
(Ando AQ6317), at a resolution of 0.2 nm. Both spectra, which are plotted in Fig. 4.5, represent
the output of the laser at 40 mA injection current. The power in a logarithmic scale is plotted
as function of wavelength for the case of no feedback (purple) and with grating feedback (blue).
Either spectrum covers the huge range from 10001100 nm. While the output of the diode laser
without feedback is relatively at, the spectrum of the diode laser with feedback shows a high peak
at a specic wavelength, 1040 nm in the gure. The very broad and relatively at output spectrum
is typical for a diode laser below threshold. Amplied spontaneous emission (ASE) dominates the
spectrum as no lasing occurs as a consequence of the absence of high optical intensities. Above
threshold, the laser lases around a wavelength 1025 nm, but still shows a broad spectrum. The
tendency towards the 1025 nm already appears in the spectrum, paying attention to higher output
levels in that region. In contrast to the case without feedback, an extremely high peak is found
when the laser is in the external cavity conguration. The selective character of the diraction
grating emerges. With over 30 dB of power dierence and a spectral width of less than 0.2 nm, by
far most of the optical power is concentrated within this narrow peak, suppressing the ASE to some
extend. As evident from Fig. 4.4, a higher output power at the same injection current is observed.
The tuning range of the SN286 laser diode in Littrow conguration has been measured with the
OSA. The operation wavelength has been changed via the angle of feedback, while the current is
kept constant at 75 mA. From the very broad ASE spectrum a large tuning range can be predicted.
For the SN286 diode laser, the tuning range proves to be 1010 nm to 1085 nm.
30
CHAPTER 4. EXPERIMENTAL RESULTS
1000 1020 1040 1060 1080 1100
90
80
70
60
50
40
30
20
Wavelength (nm)
P
o
w
e
r

(
d
B
m
)
no feedback
feedback
Figure 4.5: Optical output spectra of the SAL 1060-60 SN286 diode laser running at 40 mA
injection current for the case without feedback (purple) and for the case there is (blue). In the
presence of optical feedback by the diraction grating, a narrow peak in the spectrum is greatly
intensied, at the expense of the amplied spontaneous emission.
Below 1040 nm a ripple in the spectrum can be perceived. Careful examination shows that
the ripples appear with a period of approximately 0.2 nm. Taking into consideration the very
small dimensions of diode lasers, this 0.2 nm may be the separation of individual laser modes. A
corresponding frequency spacing of 57 GHz used in
sep
= c/2nd leads to an optical length of
the cavity of 2.6 mm. The refractive index is in the order of n = 3, which implies a plausible
diode length of less than a millimeter. This supports the assumption that the ripple is the result
of a residual facet reectivity, indicating that the quality of the AR coating is less at wavelengths
shorter than 1040 nm.
The diode laser SN281 has been characterized analogously to the investigations above. In
comparison to the SN286, this laser shows a slightly lasing threshold of 33 mA without feedback
and its tuning range of 10201070 nm is narrower. Both characteristics can be attributed to a
slightly higher residual reectivity of the AR coated facet.
To get an impression of the quality of the AR coating of the front facet of the diode laser,
we explored the number of composite cavity modes spontaneously willing to run. To observe the
modes, the diode is placed in an external cavity with a 70% reective plane mirror. The wavelength
is tuned by slowly increasing the DC current and monitored using a wave meter (HP 86120B). The
cavity length corresponds to an intermodal spacing of 1.4 GHz, and a wavelength jump of around
0.005 nm between two successive modes is expected.
Figure 4.6 shows a portion of the possible wavelengths for both the SN281 and the SN286 diode
lasers. At wavelengths at which a resonating mode was detected, a circle is plotted. Clearly, the
possible modes form clusters, with approximately 0.005 nm separation between two modes and
approximately 0.12 nm between the centers of two mode clusters. This 0.12 nm diers from the
31
CHAPTER 4. EXPERIMENTAL RESULTS
1058.6 1058.7 1058.8 1058.9 1059
Wavelength (nm)
1042.3 1042.4 1042.5 1042.6 1042.7
Wavelength (nm)
Figure 4.6: Possible free running wavelengths in an 107 mm external cavity setup for the
SN281 diode laser (upper) and the SN286 diode laser (lower). The modes caused by the
composite cavity cluster around the original modes of the diode lasers.
0.2 nm mentioned above due to the dispersion of the diode chip. On average, the SN286 diode laser
appears to allow oscillation of a few more modes per cluster than the SN281 diode laser. Broader
modes from the resonator of the two facets can be explained by the less residual reectivity of AR
coating of the front facet of this laser. Presumably a number of additional modes can be excited by
the sideband generation from the modulation of the injection current, but it cannot necessarily be
expected that the gap between two clusters can be completely lled up and that clusters become
locked to each other. In this case no phase information can be transferred between the mode
clusters owing to the absence of modes in the gap. However, based upon the number of modes
willing to run spontaneously, the SN286 diode laser is expected to be able to generate the shortest
pulses of the two.
4.3 Measurements on mode-locked diode lasers
As section 4.2 has shown, additional modes apart from the modes allowed in the diode chip itself
are generated. In this section, the eect of modulating the gain by a large RF modulation is
investigated, after which the characteristics of the output of the diode laser in mode-locked operation
are explored.
4.3.1 Introduction
Gain modulation in diode lasers is often achieved by modulating the injection current. Active mode
locking requires a high modulation depth, so a high RF power is needed. Before driving the diode
laser with large RF modulation powers, the impedances of the RF source and the diode have been
matched at 1.4 GHz. The optical cavity length corresponding to this frequency is 107 mm and
external congurations having this length are set up.
Initial experiments are carried out using the SN281 diode laser in an external cavity setup using
a 70% reective plane mirror. The diode laser is operated at a bias current of 52 mA and 27 dBm of
modulation power is applied to modulate the gain. As the initial experiments did not result in high
quality pulses, a second series of experiments is carried out with the same diode laser employing a
32
CHAPTER 4. EXPERIMENTAL RESULTS
1800 lines/mm diraction grating in a Littrow conguration to limit the lasing bandwidth. A bias
current of 60 mA and a RF power of 30 dBm have been used in this case. Further improvement in
the pulse duration is gained utilizing SN286 diode laser, which, according to the specications, has
a lower residual reectivity, in a Littrow conguration with a 1200 lines/mm diraction grating.
The laser was biassed at 75 mA, and the RF power was set to 30 dBm again. Wavelength tuning
and amplication of the output pulses is also performed using this setup.
For optimal pulse generation, the cavity length has to be matched exactly to the modulation
frequency, so the sidebands generated from one mode will exactly fall into the next possible cavity
mode. Once this condition is fullled, pulses are expected. Average output powers will be the
same for all cavity lengths, whether there are pulses or not. However, utilizing a second order
process, a distinction between pulsating or continuous output can be made. The BBO second
harmonic generating crystal and the channel photomultiplier from the autocorrelator are used to
investigate. The cavity length is optimized by expanding or shortening it, while monitoring the
output of the second harmonic detector. For the plane mirror external cavity, varying the cavity
length over a range of 5 mm around the optimal position, shows a change of output signal from
45 mV to 180 mV at an RF power of 27 dBm. At the highest output level, the optimum cavity
length has been established. Considering that the repetition rate of output pulses is equal to the
modulation frequency of 1.4 GHz and expected pulse durations of several picoseconds, this change is
insucient. However, the pulse shape and duration are unknown, as well as possibly some constant
background. From this small change, the pulse duration can be estimated to be around 200 ps
when no background is taken into account.
4.3.2 Spectral measurements
Due to the modulation of the injection current of the diode laser, optical sidebands are generated.
This causes changes in the optical spectrum, which can be observed by the scanning Farby-Perot
interferometer and the optical spectrum analyzer.
Scanning Farby-Perot interferometer
The development of sideband generation can be easily visualized by the scanning Fabry-Perot
interferometer. Figure 4.7 shows the output of the interferometer at three dierent modulation
strengths of the SN281 diode laser at a bias current of 60 mA with grating feedback.
A frequency span of the FSR of 3 GHz of the interferometer on the horizontal axis is dened
by the two vertical large intensity peaks in left-hand side of the gure. The other two subgures
are equally scaled. Operating the diode laser above threshold current in the external cavity setup
results in a single peak per FSR (left). When modulating the injection current at a frequency of
1.4 GHz, sidebands occur at a distance of 1.4 GHz. From those sidebands additional sidebands
are generated and emerge, again at 1.4 GHz distance. As long as the modulation power is still
low (20 dBm), these sidebands can be distinguished (center), whereas at higher modulation powers
(30 dBm) this becomes impossible, as too many sidebands are generated to be resolved individually
(right). With increasing modulation power, the power per wavelength decreases, which results in
an average Fabry-Perot output level lower than the peak level without current modulation.
Similar results are obtained for all three external cavity congurations described in section 4.3.1.
Optical spectrum analyzer
The previous measurements already demonstrate that the modulation aects the optical spectrum.
For all three series of experiments, the spectral changes are recorded using an optical spectrum
33
CHAPTER 4. EXPERIMENTAL RESULTS
0 1 2 3
0
0.1
0.2
0.3
0.4
Frequency (GHz)
I
n
t
e
n
s
i
t
y

(
a
.
u
.
)
0 1 2 3
0
0.1
0.2
0.3
0.4
Frequency (GHz)
I
n
t
e
n
s
i
t
y

(
a
.
u
.
)
0 1 2 3
0
0.1
0.2
0.3
0.4
Frequency (GHz)
I
n
t
e
n
s
i
t
y

(
a
.
u
.
)
Figure 4.7: Spectra from a Fabry-Perot interferometer with a FSR of 3 GHz for three dierent
stages of modulation at 1.4 GHz. (left) Single mode operation without modulation of the
injection current. (center) At weak modulation, 20 dBm, sidebands and sidebands of those,
etc., are clearly visible exactly spaced by the modulation frequency. (right) Individual sidebands
cannot be distinguished any more at high power modulation of 30 dBm.
analyzer. These spectra are shown in Fig. 4.8 (Ando AQ6317 for (a) and (c) and Ando AQ6315A
for (b)).
In all graphs, the power at all wavelengths is plotted in a logarithmic scale. Each graph shows
the spectra in case the current is modulated (red). In (a) and (b), the spectrum in the unmodulated
case is also plotted (blue). Without modulation of the gain, the diode lasers in all setups are running
on a very narrow bandwidth, as can be seen from the narrow peak in the spectra of (a) and, although
the displayed wavelength range is much smaller, (b). Mode competition results in a single mode to
survive at the expense of all others. Upon modulation, the spectrum broadens. The plane mirror
conguration shows a large portion of the spectrum to become amplied and mode clusters are
clearly present, visualized in graph (a), whereas due to the diraction grating the spectral width
in (b) and (c) remains limited, 0.11 and 0.17 nm respectively.
A detailed view of the spectrum of (a) reveals that between two spectral maxima, the power
drops down into the noise. The clusters are not linked to each other and output pulses with a con-
siderable temporal substructure are generated. External cavity congurations without bandwidth
limiting components are the most likeliest to develop mode clusters (section 2.2.2). Improvement
of the output pulses can be achieved by selecting a single mode cluster. This is done by imple-
menting an external cavity setup in which the plane mirror is replaced by a diraction grating. For
the SN281 diode laser, a diraction grating with 1800 lines/mm is used since only this is selective
enough to pick one single mode cluster. Applying the RF modulation of 30 dBm to the injection
current broadens the bandwidth with about 11 GHz to 26 GHz and pulses are generated. This
broadening stays within one mode cluster because of the high selectivity of the grating. A dirac-
tion grating with 1200 lines/mm is too little selective and more than one mode cluster is picked
out, still resulting in a temporal complexity of the output pulses. Even if all energy is concentrated
in just two mode clusters, the clusters are not willing to lock together. An explanation can be
found in the residual reectivity of the AR coating on the front facet of the diode. R = 4 10
5
is just below of the minimum requirement of R = 10
4
for the AR coating to prevent mode clus-
ter formation in mode-locked diode lasers [3]. Although the feasible spectral width stays limited,
locking the modes within a single cluster is the best option to deal with these AR coatings. The
AR coating of the SN286 diode laser is specied to be 20 times better than the that of the SN281
34
CHAPTER 4. EXPERIMENTAL RESULTS
1045 1050 1055 1060 1065
90
80
70
60
50
40
Wavelength (nm)
P
o
w
e
r

(
d
B
m
)
continuous wave
27 dBm modulation
1028.4 1028.6 1028.8 1029 1029.2
90
80
70
60
50
40
Wavelength (nm)
P
o
w
e
r

(
d
B
m
)
continuous wave
30 dBm modulation
1047 1047.5 1048 1048.5 1049
90
80
70
60
50
40
30
Wavelength (nm)
P
o
w
e
r

(
d
B
m
)
30 dBm modulation
c) a)
b)
Figure 4.8: Spectral characteristics of (a) the SN281 diode laser with a plane mirror external
cavity, both the continuous and mode-locked case; (b) spectral limitation by a 1800 lines/mm
diraction grating of the SN281 diode laser, revealing a 10 GHz spectral broadening between
the continuous and the mode-locked case; (c) the mode-locked SN286 diode laser spectrally
limited by a 1200 lines/mm diraction grating locking three mode clusters.
laser. In this laser multiple mode clusters can be locked, although the number of clusters probably
must stay small for the energy per sideband to be large enough to overcome the still present, but
low, residual reectivity. As the forcedly generated sidebands are able to ll up the gap between
successive mode clusters, phase information between two mode clusters can be exchanged and the
modes from all clusters all become locked to each other. This allows shorter pulses to be generated.
To achieve a larger bandwidth than that of one mode cluster, a less selective diraction grating is
chosen to retroreect the light back into the diode. Operating the diode laser in an external cavity
setup with a 1200 lines/mm diraction grating enables to couple about two to ve mode clusters.
However, the modes locking the former separate clusters are weaker than those in the center of the
former clusters. Small dips in the spectrum occur at the position of the former gaps and output
pulses are less intense with a bit more uctuations between them than expected from the spectral
width. The spectrum of (c) shows three locked clusters, of which two are clearly visible due to the
small dip between them. The third is exposed because of the small node in the falling edge on the
right-hand side of the spectrum.
35
CHAPTER 4. EXPERIMENTAL RESULTS
4.3.3 Autocorrelation measurements
The spectral measurements presented in the previous section have shown that upon modulation
of the injection current, sidebands emerge and the optical spectrum broadens. These observations
imply that the diode laser is mode-locked. Verication if the laser is indeed mode-locked and
generates pulses can be deduced from an autocorrelation trace.
Plane mirror external cavity conguration and autocorrelator calibration
The recorded trace from the output of the laser in the plane mirror external cavity provides the
interesting fact of an intensity-like autocorrelation trace with large peaks superimposed. An exam-
ple of such a trace is given in Fig. 4.9 (a). As explained in [3], these coherence spikes occur when
the optical spectrum is rather complex. The spectrum shows mode clusters (see section 2.2.2) due
to a residual reectivity on front facet of the diode laser. Pulses with a temporal substructure are
formed, heavily chirped in a nonlinear way and with durations far from the Fourier limit. When
such pulses are shifted with respect to each other and overlapped in time, a lack of coherence
prevents the formation of fringes in the autocorrelation trace.
50 0 50
0
2
4
6
8
Delay (ps)
A
m
p
l
i
t
u
d
e

(
a
.
u
.
)
Figure 4.9: Example of an autocorrelation trace recorded by the interferometric rotating-
mirror autocorrelator, showing only an intensity autocorrelation trace containing a number of
coherence spikes.
Coherence spikes are caused by pulses within a longer pulse resonating in the diode laser itself.
They are separated by the round trip time of the cavity of the semiconductor chip itself [3]. This
knowledge enables an easy calibration of the autocorrelator. From the optical spectrum, the optical
cavity length of the diode laser can be calculated. As said before, the laser modes are separated by
572 GHz and an optical cavity length of 2.60.1 mm was determined. Therefore, the round trip
time of a pulse inside the cavity is 181 ps, which is equal to the time separation of the coherence
spikes. This can be exploited to calculate the corresponding delay time for the autocorrelation
trace and all future traces as long as the revolutions per minute is not changed.
The nonlinearity in the rotating-mirror autocorrelator can also be deduced from such a trace.
36
CHAPTER 4. EXPERIMENTAL RESULTS
When more than two coherence spikes are visible, the nonlinearity can be calculated since the
distances between successive spikes dier. As this nonlinear behaviour is below 10% over the scan
range -50 to 50 ps, this is neglected and a pulse duration of approximately 42 ps is found, assuming
a Gaussian pulse shape as a rst order approximation. Instabilities and alignment diculties of the
autocorrelator reduce the accuracy of the measurements. The peak-to-background ratio of 3:1 as
should be found for this intensity autocorrelation, setting aside the coherence spikes, is not reached
and the ratio of 8:1 in the interferometric autocorrelation is often not realized as well.
External cavity congurations with diraction grating feedback
The autocorrelation trace of the pulses from the diode laser without spectral limitations in the
previous section, clearly shows that the coherence within one pulse is insucient to generate an
interferometric autocorrelation. This can be the consequence of the broad spectrum. To restrict
the emission to a narrower bandwidth, a diraction grating is implemented in the next two cong-
urations. In section 4.3.2 is shown that the spectrum then indeed is restricted to a width of less
than 0.2 nm for the SN281 diode laser with a 1800 lines/mm grating or less than 0.5 nm for the
SN286 diode laser in a 1200 lines/mm diraction grating external cavity conguration.
100 50 0 50 100
0
2
4
6
8
Delay (ps)
A
m
p
l
i
t
u
d
e

(
a
.
u
.
)
100 50 0 50 100
0
2
4
6
8
Delay (ps)
A
m
p
l
i
t
u
d
e

(
a
.
u
.
)
Figure 4.10: Autocorrelation traces of the pulsating output of the mode-locked diode lasers.
(a) SN281 with optical feedback from a 1800 lines/mm diraction grating. (b) SN286 diode
laser with a 1200 lines/mm diraction grating.
Figure 4.10 shows two autocorrelation traces acquired with this spectral ltering with SN281
(left) and SN286 (right) employed. Plotted is the signal from the CPM as function of the delay of
the variable arm of the autocorrelator. Fringes in the interferogram indicate a coherence between
the two replicas of the input pulses shifted in time with respect to each other. This suggests
temporally much smoother pulses. In yellow the theoretical Gaussian intensity autocorrelation is
plotted, according to the t parameters obtained by the tting procedure described in section 3.1.3.
The black curve shows the envelope of the theoretical interferometric autocorrelation.
The autocorrelation from the SN281 diode laser (left) leads to an estimated pulse duration of
42 ps, assuming a Gaussian shape. This is no dierent than the estimated pulse width without the
spectral limitation, however, the pulse quality has improved a lot. With a time-bandwidth product,
t
p
, of 1.0, these pulses approach the transform limit of 0.441 for Gaussian pulses. Since the
measurements of the spectral width are at the edge of the capabilities of the spectrum analyzer
37
CHAPTER 4. EXPERIMENTAL RESULTS
and the accuracy of the autocorrelation is relatively low, the exact determination of the TBP is
dicult.
From the autocorrelation trace in (b) a pulse duration of 30 ps is found, assuming a Gaussian
shape, and 26 ps for a sech
2
shape. The time-bandwidth product of 1.3 is a bit larger than that in
the previous case. The spectrum is wider, but a structure revealing that the spectrum consists of
coupled mode clusters is still visible. This spectral shape is one reason for the larger TBP. Another
cause for this can be seen from the autocorrelation trace. Wings on both sides of the coherence
fringes disclose a linear chirp in the pulses. The envelope of the trace is tted according to Eq. (3.8)
using a chirp parameter of a = 2.8. Chirped pulses from semiconductor lasers have been observed
in many experimental works before [20] and its origin has been explained in section 2.2.2.
Comparing the results, it can be concluded that pulses close to the transform limit can be
obtained by the setup with the largest spectral restriction. However, the pulses from the SN286
diode laser are shorter than those of the SN281, which is in agreement with the expectations towards
the less residual reectivity of the former diode.
4.3.4 Power measurements
For a correct estimation of the peak power, two measurements have been performed. The average
power has been measured and a possible continuous-wave (cw) background, on which the pulses are
superimposed, has been determined. The latter has been measured as follows. The second harmonic
(SH) power generated in a 12-mm long BBO crystal is measured in two cases. First, when the pulsed
radiation was focused in the crystal and second, when the diode laser was operated cw, i.e. without
RF power supplied. The SH power generated by pulses with a certain average power is enhanced
with respect to the SH power generated by cw radiation of the same power by a factor of 8.3.
Under the assumption of Gaussian pulses with a pulse duration of approximately 37 ps, this means
that about 80% of the power is in the pulses, while 20% is cw background. In this case 17 mW of
average power (at a central wavelength of 1060 nm of the SN286 laser) leads to a peak power of
approximately 200 mW.
4.3.5 Wavelength tuning
To achieve a tunable mid-infrared source in the 35 m region an optical parametric oscillator can
be used. Near-infrared pump wavelengths can then be transferred directly to the mid-infrared,
i.e. wavelength tuning of the pump leads to wavelength tuning in the mid-IR.
As in the measurements presented before, the SN286 has proven to posses more advantageous
properties with respect to pulse duration and tunability. In the following, wavelength experiments
are performed using this diode in the ultrashort pulsed regime. As the gain prole of the ytterbium-
doped ber amplier starts at around 1040 nm and reaches beyond 1100 nm, determining the tuning
characteristics of the diode laser has been concentrated on for wavelengths larger than 1040 nm.
Over the range of 10401085 nm, we have measured the pulse duration and average output
power. The wavelength is tuned by varying the angle of the diraction grating. The pulse duration
is measured using the autocorrelator and the average power using a power meter (Melles Griot
13PEM001). The optical spectrum has been observed during this measurement using the OSA
(Ando AQ6317).
The results are shown in Fig. 4.11. The pulse duration and average power are plotted at
function of wavelength. It can be seen that over this wavelength range, the pulse duration (squares)
varies between 27 and 44 ps, while the average power varies between 9 and 20 mW (diamonds).
At rst sight, the pulse duration and average power seem to uctuate randomly. However, an
38
CHAPTER 4. EXPERIMENTAL RESULTS
1040 1050 1060 1070 1080 1090
0
10
20
30
40
50
Wavelength (nm)
P
u
l
s
e

d
u
r
a
t
i
o
n

(
p
s
)
pulse duration
average power
0
10
20
30
40
50
A
v
e
r
a
g
e

p
o
w
e
r

(
m
W
)
Figure 4.11: Pulse duration (squares) and average power (diamonds) during wavelength
tuning over 10401085 nm.
inverse relationship appears to relate the two plots. These uctuations can be explained when the
corresponding optical spectrum is taken into account. Spectra of pulses at dierent wavelengths
show dierent mode cluster structures. This is an eect of the residual reectivity of the front
facet of the diode. Interference eects within the semiconductor chip lead to cluster eects, which
can be observed in the measured spectra. In some particular disadvantageous situations that the
central wavelength lies in a gap between two modes, this can lead to a limited spectral width and
correspondingly in an increase in pulse duration. In contrast, in the more advantageous case that
the selected wavelength lies in the center of a mode of the diode laser chip cavity, the available
energy within the diode is more eciently used and transferred to laser modes lling also the gaps
between mode clusters. The spectrum is wider, the phase relation between all modes is preserved
and shorter pulses are generated.
Examining the spectra indeed shows a relation between the structure of the coupled mode
clusters and the pulse duration. Pulses become shorter if more mode clusters are included in the
locking process. The assumption above is also supported by the measurement that longer pulses
show lower average powers. Again an explanation in terms of total energy in the system can be
given. Longer pulses show a relatively narrow bandwidth. No additional mode clusters can be
locked and losses at the edges of the spectrum are large as compared to the case of shorter output
pulses. Spectra in the latter case reveal more mode clusters to be locked and average powers are
larger.
4.3.6 Inuence of parameter changes on pulse duration
The previous section shows that the choice of the central wavelength with respect to the AR coating
leads to dierent pulse durations and average output powers. Apart from that, the eect of dierent
39
CHAPTER 4. EXPERIMENTAL RESULTS
injection currents on the pulse width and average power is measured as well as the eect of small
changes in the cavity length and driving frequency.
Changes in the bias current and modulation power appear to be of little inuence on the pulse
duration. Over the range of 40 to 75 mA of bias current with a constant RF power of 30 dBm, the
pulse width stays in the range of 25 to 35 ps. The lower regions of this range show little longer
pulses than at higher bias currents, but no line can be drawn. In accordance with the longer pulses,
the spectral width for lower bias currents is less. This can be attributed to the lower total energy
in the system. Sidebands are generated by transferring energy from one mode to another. At some
point an equilibrium is reached and losses at the edges of the lasing spectrum overcome the gain and
no more modes are generated. Higher bias currents increase the total energy in the laser and more
energy is transferred from one mode to another as well as the gain for the new modes is higher. This
leads to a further broadening of the bandwidth. The spectral limitation by the diraction grating
is most likely the largest contributor in preventing new modes to rise. The counteracting eect of
the smaller modulation depth at higher bias currents is therefore of less inuence and in this case
can be neglected according to the measured eects on pulse width with changing RF power. The
pulse width decreases with increasing RF until it reaches 25 dBm. Upon further increasing the RF
power to 30 dBm, no signicant change in pulse duration can be observed. A likely explanation
for this observation is that, again, the maximum spectral bandwidth is reached, which is limited
by the cluster formation due to the before mentioned composite-cavity eect.
Variation in the driving frequency without changing the cavity length in contrast, has some
inuence on the pulse width. Changes are negligible as long as the the frequency stays within
40 MHz from the optimum frequency. Beyond 40 MHz, the pulse duration extends and eventually
the pulses degrade and cannot be recognized in the autocorrelation trace any more. As the eect
of detuning the RF frequency is the same as a detuning of the cavity length with respect to a xed
frequency, one should expect analogous observations when changing the cavity length. Indeed, such
eects can be veried. The cavity length may be extended or shortened by about 3 mm without
signicant change in the pulse duration, which exactly corresponds to a frequency detuning of
40 MHz. Upon further detuning the feedback grating from the optimum position, the pulses also
degrade. Considering the pulse width of about 30 ps, which corresponds to a spatial extension of
the beam of 9 mm, a the cavity length mismatch of 3 mm, i.e. a round trip length mismatch of
6 mm, still allows an overlap of two successive pulses of one third of their length. Larger mismatches
result in less overlap inside the gain medium, preventing an ecient pulse formation to take place.
The large dierences in bias current and RF modulation power and the large resonator tolerance
explain the very good reproducibility of the pulse generation by mode locking this diode laser.
4.3.7 Summary of experimental results
In the present chapter, we have examined the behaviour of mode-locked diode lasers. Due to the
large selectivity of the diraction grating, nearly bandwidth-limited pulses of around 40 ps have
been generated using the SN281 diode laser. A time-bandwidth product of 1.0 is obtained. Even
shorter pulses can be generated by the SN286 diode laser. The time-bandwidth product increases
a little due to a chirp in the pulses, but durations of 27 to 44 ps have been generated over a
wavelength range of 10401085 nm, with peak powers of more than 200 mW.
40
Chapter 5
Pulse amplication
Overview Mode-locked pulses have been produced from diode lasers in Chapter 4, tunable in
the wavelength range from 1040 to 1085 nm. Average powers of these pulses are less than 20 mW
with peak powers up to 200 mW. To bring these low-power pulses into the regime of high peak
powers, such that they become of interest for applications like e.g. parametric processes, the pulse
energies are boosted using post-generation active amplication. In this chapter we describe the rst
amplication of mode-locked pulses in an ytterbium-doped ber. First, the experimental setup for
amplication is described in detail in section 5.1. Then, section 5.2 presents the results we obtained
by the ber-amplication of the diode laser pulses.
5.1 Experimental setup
To acquire high power pulses, we amplied the low-power pulses from the mode-locked diode laser
in an ytterbium-doped double-clad ber amplier.
A modication of the experimental setup that has been used before is necessary to characterize
the amplied pulses. The altered setup is shown schematically in Fig. 5.1. Obviously, the main
change is the inclusion of the ber amplier between the output of the diode laser and the inputs
of the three optical characterization instruments, the autocorrelator, the spectrum analyzer and
the Fabry-Perot interferometer. As both the unamplied and the amplied output pulses and their
spectra are to be characterized, rapid switching between the two output beams is realized by a
ip-mirror between the Faraday isolator and the ber amplier. The beam can be steered down
directly to the instruments as well as, after switching the ip-mirror, to the ber. The beam splitter
is then used from both sides and all optical measurements can be used in both cases.
After transmission through the 60 dB optical isolator, the diode laser output is amplied in
a 36-m long ytterbium-doped double-clad large mode area (LMA) ber. The Yb-ber has a core
diameter of 10 m, a numerical aperture (NA) of 0.07, and a doping concentration of 1000 mol
parts in 10
6
. The energy needed to amplify the pulses is provided by a 25 W pump diode at
976 nm, which is close to the strong absorption peak of ytterbium. This allows most energy to
be absorbed by the ber. The output of this diode is pumped into the 400 m inner cladding of
the ber counter-directionally to the seed signal. The inner cladding or pump core has a D-shape
(see the right-hand side of Fig. 5.2) to avoid the excitation of helical modes and to enable an
ecient coupling of the pump light to the core. The seed signal, in this case being the picosecond
pulses from the mode-locked diode laser, is coupled into the core. A dichroic mirror allows the
41
CHAPTER 5. PULSE AMPLIFICATION

Fiber amplfier
Mode locked
diode laser
DM
Pump
diode
FI
Optical
spectrum
analyzer
BS
2

M
Scanning
Fabry-Perot
interferometer

Autocorrelator
M
FM
BS
1

Figure 5.1: Experimental setup for amplication of mode-locked pulses from the diode laser
and characterizing its output. FI Faraday isolator, FM Flip-mirror, DM Dichroic mirror,
M mirror, BS Beam splitter.

signal
seed
pump
DM L Fiber amplifier L
core
pump core
cladding

Figure 5.2: (left) Schematic ber amplier conguration. The seed is injected from the left,
the pump is injected from the right, and the amplied seed is extracted from the right and
separated from the pump via a dichroic mirror. L Lens, DM Dichroic mirror. (right)
Cross section of the double-clad ber.
42
CHAPTER 5. PULSE AMPLIFICATION
pump to enter the ber whereas the output of the ber is transmitted, as illustrated in the left-
hand side of Fig. 5.2. To prevent the high ber output powers to reach the sensitive measurement
instruments, a mirror steers the signal to a high-power power meter. The small fraction of power
that is transmitted through the mirror is used to characterize the amplied pulse properties.
5.2 Measurements on amplied mode-locked pulses
5.2.1 Introduction
All experiments on amplication of the mode-locked pulses are performed using the SN286 diode
laser in an external cavity setup utilizing a 1200 lines/mm diraction grating in Littrow congura-
tion. A bias current of 75 mA and modulation power of 30 dBm at 1.4 GHz is used to mode lock the
laser. The pump power for the ber amplier is set to a maximum of 25 W in most experiments.
To investigate if amplication alters the properties of the pulses, the spectra before and after
amplication have been measured using the OSA (Ando AQ6317) and pulse durations have been
measured with the autocorrelator.
5.2.2 Spectral measurements
Figure 5.3 shows a typical example of the original (red) and the amplied (green) spectra. No
1045.5 1046 1046.5 1047
90
80
70
60
50
40
30
Wavelength (nm)
P
o
w
e
r

(
d
B
m
)
unamplified
amplified
Figure 5.3: Spectra of the unamplied (red) and amplied at a power of 2 W (green) pulses
from the mode-locked diode laser.
signicant eect on the width of the spectrum is observed. The width of the spectrum of about
0.18 nm (49 GHz) is maintained. Therefore, from the spectrum alone no change in pulse duration
is expected. Both spectra show a substructure, which has its origin in the mode clusters. Such be-
haviour is expected from preliminary observations (compare section 4.3.2). However, from Fig. 5.3
43
CHAPTER 5. PULSE AMPLIFICATION
is becomes clear that after amplication this structure is even more distinct. The ber appears to
amplify the higher intensities in the centers of the mode clusters more strongly than the lower inten-
sities in the former gaps between the clusters. A similar eect is also known as mode competition
from homogeneously broadened multi-mode lasers. This can clearly be seen from the gure, where
the rather smooth spectrum of the original pulses is turned into one where the contrast between
the center of the mode clusters and the former gaps is enhanced, visible as the rabbits ears.
To be able to make a statement on the spectral quality of the amplication, the amplied
spectrum of Fig. 5.3 has to be put in relation to the amplier emission across the whole gain
bandwidth. For this, the spectrum is measured using the same OSA in the range of 10201120 nm.
The result at an average power of 2 W is shown in Fig. 5.4 with the power in a logarithmic scale.
One can see a broad background and one high peak at a wavelength of 1046 nm. This peak is the
1020 1040 1060 1080 1100 1120
90
80
70
60
50
40
30
Wavelength (nm)
P
o
w
e
r

(
d
B
m
)
Figure 5.4: Spectrum amplied pulses from the mode-locked diode laser. A large portion of
the spectrum is occupied by ASE, showing a large peak at the central wavelength of the pulses.
high-power amplied radiation from the injected diode laser pulses, that are shown in Fig. 5.3. The
broad background covers from 1040 to 1115 nm and is noise that is added to the signal, called the
amplied spontaneous emission (ASE). Even though the wavelength of the injected pulse is only at
the very edge of the amplier gain spectrum, the ASE is suppressed to 30 dB below the signal level.
In this case this means that about 93% of the total power is concentrated in the peak taking into
consideration the narrow bandwidth of the amplied peak and the broad bandwidth of the ASE.
At maximum output power a 20 dB dierence is found between the peak and the maximum of the
ASE, corresponding to 73% of the total power in the peak. Gain saturation at the wavelengths
around the signal lead to a less rapid growth of the injected pulses than of ASE, eectively lowering
the ratio of signal-to-ASE power distribution.
When the wavelength of the injected pulses is increased, the peak shifts as well and ASE levels
decrease. As the wavelength approaches the central range of the gain of the ber amplier, most
ASE is suppressed and, at maximum output power, dierences of 40 dB are measured between the
44
CHAPTER 5. PULSE AMPLIFICATION
top of the peak and the maximum of the ASE. This implies that more than 99% of the total power
is concentrated in the peak.
5.2.3 Autocorrelation measurements
After investigating the eect of the amplication process on the spectral properties, in this section
the eects on the pulse duration are studied. Autocorrelation traces of the pulses with a central
wavelength of 1058 nm before (left) and after (right) amplication at an average output power of
2 W are recorded and plotted in Fig. 5.5. Again, the yellow curves give the tted Gaussian intensity
100 50 0 50 100
0
2
4
6
8
Delay (ps)
A
m
p
l
i
t
u
d
e

(
a
.
u
.
)
100 50 0 50 100
0
2
4
6
8
Delay (ps)
A
m
p
l
i
t
u
d
e

(
a
.
u
.
)
Figure 5.5: Autocorrelation traces of the unamplied (left) and amplied (right) pulses from
the mode-locked diode laser, tted for a Gaussian pulse shape. The amplied pulse is shorter,
29 ps versus 37 ps for the unamplied, and shows a smaller chirp.
autocorrelations and the black ones the envelopes calculated using Eq. (3.8). After amplication,
the autocorrelation trace becomes narrower, indicating a shorter pulse width. The pulse is shortened
from 37 ps to 29 ps for the traces illustrated here, assuming a Gaussian pulse shape, or from 33 ps
down to 21 ps in case a sech
2
shape is assumed. These shorter pulses can be explained as follows.
The ber is pumped continuously, continuously storing enery by exciting ytterbium ions into higher
energy levels. At the moment the pulse arrives, a high energy storage level is reached and the front
of the pulse gets amplied by a high gain. As the pulse proceeds through the gain material, the
stored energy is depleted, such that the amplication experienced by the later parts of the pulse is
decreased. Due to the non-uniform amplication factor, an asymmetrical output pulse is expected,
a high intensity front with a long weaker tail. When the pulse has passed, gain starts to build up
again until the next pulse arrives.
The second strikingness of the autocorrelation traces is the decreased wings in the right-hand
trace. This indicates a reduction of the chirp. The autocorrelation trace of the unamplied pulse is
tted by Eq. (3.8) in which the chirp factor a = 2.5, whereas for the amplied pulse a = 1.5. As well
as the pulse shortening, the chirp reduction can be attributed to the asymmetrical magnication of
the pulse. When the already chirped pulse enters the amplier, frequencies at the front of the pulse
are amplied more than frequencies at the end of the pulse. This means that the amplication
is non-uniform not only in a temporal view, but that it is also spectrally non-uniform due to the
temporal distribution of the frequency components. This leads to the fact that the less amplied
45
CHAPTER 5. PULSE AMPLIFICATION
frequency components in the tail of the pulse contribute less to the chirp and the chirp factor is
eectively lowered.
Due to the shorting of the pulses, the time-bandwidth product before amplication of 1.5
reduces to 1.1 after amplication. This gives an indication that the pulse quality has improved
during amplication.
5.2.4 Power measurements
After amplication, the peak power of the pulses is determined. As for the unamplied pulses the
continuous-wave (cw) background radiation of the pulses is of interest for an estimation of the peak
power. These experiments have been performed as described in section 4.3.4 with the amplied
pulses at the same wavelength of 1060 nm. With a pulse duration of 32 ps and at an average
output power of 7 W, the enhancement factor is reduced to 4.6, which corresponds to about 50%
of the power in the pulses, while the other half remains as a cw background. With this data, the
estimated peak power is about 80 W in this case. In the case of maximum output power of 9.5 W,
the peak power can be extrapolated to be more than 100 W.
5.2.5 Wavelength tuning
High power ultrashort pulses have been generated as shown in the previous sections. As has been
shown in section 4.3.5, the wavelength of the mode-locked diode laser can be tuned. We studied the
eects of tuning the diode laser over a broad wavelength range on pulse duration and optical output
power as well. As in section 4.3.5, pulse durations have been measured using the autocorrelator
and the average output power using the high power power meter (Newport 1825-C with Thermopile
detector). Also, the optical spectra are recorded using the OSA (Ando AQ6317).
Figure 5.6 shows the pulse duration before (red) and after (green) amplication in the wave-
length range from 1040 to 1085 nm. The autocorrelation traces to determine the pulse width have
been recorded at an average power of approximately 2 W. At this power level, the pulse duration is
shortened over the whole wavelength range with respect to the unamplied ones. An explanation
for the pulse shortening is given in section 5.2.3. At higher average powers, is it likely that the
pulse duration will lengthen again, due to the presence of additional gain when the tail of the pulse
passes through the ber.
To determine the average power and the spectral quality of the output pulses, the maximum
achieved average optical power over the whole wavelength range have been measured and the
optical spectra over the wavelength range 10201120 nm have been recorded. Figure 5.7 shows
the maximum output power(circles) and the signal-to-ASE power ratio (squares) as function of
the wavelength. A gradual increase of the average power with increasing wavelength is shown,
as well as a gradual increase of the signal-to-ASE ratio. From 1040 nm to 1044 nm, the average
power increases gradually. Output powers are limited by self-pulsing of the ber and the maximum
pump power cannot be reached. The pump power is limited at 19 W in these cases. Due to the
wavelength being at the edge of the gain bandwidth of the ber, a lack of gain depletion the the
central region (around 1080 nm) causes gain competition of ASE and the pulses. At wavelengths
longer than 1044 nm, the self-pulsing behaviour reduces and the average powers to be reached at
most are limited by the pump power. This explains the reasonably constant power of 8.59.5 W in
the wavelength range from 1046 to 1085 nm, with the maximum power of 9.5 W between 1070 nm
and 1082 nm.
To determine the signal-to-ASE ratio, the power in the narrow spectral width of the signal and
the power in the ASE have been compared. Considering the broad ASE spectrum present after
46
CHAPTER 5. PULSE AMPLIFICATION
1040 1050 1060 1070 1080 1090
0
10
20
30
40
50
60
Wavelength (nm)
P
u
l
s
e

d
u
r
a
t
i
o
n

(
p
s
)
unamplified
amplified
Figure 5.6: Pulse durations of the unamplied and amplied pulses. Over the whole range,
the pulse duration is shorter after amplication. All traces are taken at approximately 2 W of
average optical power.
amplication of the pulse as in section 5.2.2, the ratio of the power in the signal to the power in
the ASE is determined from the spectra. This signal-to-ASE ratio as function of wavelength is
plotted in Fig. 5.7 as well. It is clear that the ratio increases quickly from the lower wavelengths
(10401044 nm) due to the declining limitation by self-pulsing. From 1052 nm, more than 90% of
the total power is concentrated in the signal wavelengths. The maximum gain of the amplier at
a wavelength of around 1080 nm is conrmed by the further increase of the signal-to-ASE to more
than 99% in the region of 10741085 nm.
5.2.6 Summary of experimental results
With the experiments described in this chapter, we show that a double-clad Yb-doped ber am-
plier is a suitable amplier to boost the power of mode-locked pulses with central wavelengths of
around 1.06 m from a semiconductor laser. The injected pulses with an average power of less than
20 mW are amplied to up to 9.5 W of average power, which at the several tens of picoseconds
pulse durations means more than 100 W of peak power. The central wavelength of the injected
pulses can be readily tuned in the wavelength range 10401085 nm. Amplication in this whole
wavelength range is possible due to the very large gain bandwidth (10401115 nm) of the ber
amplier. No other experiments so far have shown these high-power pulses in combination with
the large wavelength exibility.
47
CHAPTER 5. PULSE AMPLIFICATION
1040 1050 1060 1070 1080 1090
0
5
10
M
a
x
i
m
u
m

a
v
e
r
a
g
e

p
o
w
e
r

(
W
)
maximum average power
signaltoASE ratio
1040 1050 1060 1070 1080 1090
0
5
10
Wavelength (nm)
M
a
x
i
m
u
m

a
v
e
r
a
g
e

p
o
w
e
r

(
W
)
maximum average power
signaltoASE ratio
1040 1050 1060 1070 1080 1090
0
0.5
1
S
i
g
n
a
l

t
o

A
S
E

r
a
t
i
o

(

)
Figure 5.7: The maximum average optical power and the signal-to-ASE power ratio of the
amplied pulses. The limited output powers below 1044 nm are caused by the self-pulsing of
the ber amplier. The signal-to-ASE power ratio has a similar curve.
48
Chapter 6
Summary
In this thesis, we characterize the output performance of GaInAs diode lasers emitting ultrashort
pulses with subsequent amplication in an ytterbium-doped ber to peak powers of more than
100 W. Such sources, particulary when tunable, are of high importance for ecient nonlinear
frequency conversion, e.g., to the mid-infrared using synchronously pumped optical parametric
oscillators (OPOs).
To obtain ultrashort pulses, the diode lasers have been actively mode-locked by modulation of
the injection current. The antireection coated diode lasers have been placed in an external grating
feedback cavity with its length adjusted to match the modulation frequency, and mode-locked pulses
at a repetition rate of 1.4 GHz have been generated.
The pulse duration measured behind the diode laser is 35 ps, as determined by interferometric
autocorrelation, assuming a Gaussian pulse shape. The spectral selection of the emitted light
by properly adjusting the grating, results in the emission of nearly Fourier-limited pulses (time-
bandwidth product of about 1.4). The remaining deviation from the theoretical value of 0.441 can
be accredited to a chirp in the pulse and a residual modulation of the output spectrum, which
indicates a temporal substructure. The average output power of the mode-locked laser is 15 mW,
which corresponds to maximum peak powers of 200 mW at a 1.4 GHz repetition rate.
A second advantage of grating feedback is that the diode output can be tuned over a wide range
(10401085 nm), while output power and pulse duration remain approximately unchanged.
For power amplication of the mode-locked pulses from the diode laser, we use a cladding
pumped double-clad ytterbium-doped ber of 36 m length. After amplication, the average output
power is more than 9 W. With slightly shorter pulse durations than in the unamplied case, this
corresponds to peak powers of more than 100 W.
By grating tuning of the diode laser, the ber output is tunable from 1050 to 1085 nm with
average output powers of more than 9 W. Spectral measurements show that the ASE background
is suppressed to up to 40 dB below the power at the signal wavelength. Based upon our previous
experience, these output parameters provide ideal conditions to synchronously pump mid-IR OPOs.
This means, in particular, an ultra-wide tuning of the OPO output wavelength across the molecular
ngerprint region, by grating tuning our high-power diode-ber source. We note that no other
experiments so far have shown such high-power pulses in combination with such large wavelength
tunability around 1 m.
49
Bibliography
[1] M.G. Hekelaar, B. Adhimoolam, P. Gross, I.D. Lindsay, and K.-J. Boller. Amplication of
short pulses from a mode-locked diode laser in an ytterbium-doped ber. CLEO/IQEC 2005,
2005.
[2] B. Adhimoolam, M.G. Hekelaar, P. Gross, I.D. Lindsay, and K.-J. Boller. 30-ps pulses from a
mode-locked ber-amplied diode laser tunable around 1.06 m. 2005.
[3] P.J. Delfyett, L.T. Florez, N. Stoel, T. Gmitter, N.C. Andreadakis, Y. Silberberg, J.P. Her-
itage, and G.A. Alphonse. High-power ultrafast laser diodes. IEEE Journal of Quantum
Electronics, 28:22032219, 1992.
[4] A. Robertson, M. E. Klein, M. A. Tremont, K.-J. Boller, and R. Wallenstein. 2.5-GHz
repetition-rate singly resonant optical parametric oscillator synchronously pumped by a mode-
locked diode oscillator amplif ier system. Optics Letters, 25(9):657659, 2000.
[5] S. Lecomte, R. Paschotta, S. Pawlik, B. Schmidt, K. Furusawa, A. Malinowski, D.J. Richard-
son, and U. Keller. Optical parametric oscillator with a pulse repetition rate of 39 GHz
and 2.1-W signal average output power in the spectral region near 1.5 m. Optics Letters,
30(3):290292, 2005.
[6] A. Hideur, T. Chartier, M. Brunel, S. Louis, C.

Ozkul, and F. Sanchez. Generation of high
energy femtosecond pulses from a side-pumped Yb-doped double-clad ber laser. Applied
Physics Letters, 79(21):33893391, 2001.
[7] O.G. Okhotnikov, L. Gomes, N. Xiang, T. Jouhti, and A.B. Grudinin. Mode-locked ytterbium
ber laser tunable in the 9801070-nm spectral range. Optics Letters, 28(17):15221524, 2003.
[8] A. Malinowski, A. Piper, J.H.V. Price, K. Furusawa, Y. Jeong, J. Nilsson, and D.J. Richardson.
Ultrashort-pulse Yb3+-ber-based laser and amplier system producing 25-W average power.
Optics Letters, 29(17):20732075, 2004.
[9] M.E. Klein, A. Robertson, M.A. Tremont, R. Wallenstein, and K.-J. Boller. Rapid infrared
wavelength access with a picosecond PPLN OPO pumped by a mode-locked diode laser. Ap-
plied Physics B, 73(1):110, 2001.
[10] M.J. Brennan, A.J. Budz, B.J. Robinson, P. Mascher, and H.K. Haugen. Ultrashort optical
pulse generation with a mode-locked long-wavelength (1075-1085 nm) ingaasgaas semiconduc-
tor laser. IEEE Photonics Technology Letters, 16(8):17981800, 2004.
[11] A. Piper, A.Malinowski, B. C Thomsen, D. J. Richardson, L. M. B. Hickey, and M. N. Zervas.
11.1 W average power, 20 ps pulses at 1 GHz repetition rate from a ber-amplied gain-
switched 1.06 m fabry-perot laser diode. CLEO/IQEC 2005, 2005.
50
BIBLIOGRAPHY
[12] W.T. Silvast. Laser fundamentals, chapter 13. Cambridge University Press, Cambridge, UK,
2nd edition, 2004.
[13] P. Vasilev. Ultrafast diode lasers: fundamentals and applications, chapter 4. Artech House,
Boston London, 1995.
[14] D. Woll, J. Schumacher, A. Robertson, M.A. Tremont, R. Wallenstein, M. Katz, D. Eger, and
A. Englander. 250 mW of coherent blue 460-nm light by single-pass frequency doubling of the
output of a mode-locked high-power diode laser in PPKTP. Optics Letters, 27(12):10551057,
2002.
[15] J.P. van der Ziel. Active mode locking of double heterostructure lasers in an external cavity.
Journal of Applied Physics, 52(7):44354446, 1981.
[16] K.Y. Lau. Short-pulse and high-frequency signal generation in semiconductor lasers. IEEE
Journal of Lightwave Technol., 7(2):400419, 1989.
[17] D. Marcuse. Reection loss of laser mode from tilted end mirror. IEEE Journal of Lightwave
Technology, 7(2):336339, 1989.
[18] G.A. Alphonse, D.B. Gilbert, M.G. Harvey, and M. Ettenberg. High-power superluminescent
diodes. IEEE Journal of Quantum Electronics, 24(12):24542458, 1988.
[19] J.E. Bowers, P.A. Morton, A. Mar, and S.W. Corzine. Actively mode-locked semiconductor
lasers. IEEE Journal of Quantum Electronics, 25(6):14261439, 1989.
[20] M. Schell, M. Tsuchiya, and T. Kamiya. Chirp and stability of modelocked semiconductor
lasers. IEEE Journal of Quantum Electronics, 32(7):11801190, 1996.
[21] A.S. Hou, R.S. Tucker, and E.P. Ippen. Chirp in avtively mode-locked diode lasers. In C.B.
Harris, E.P. Ippen, G.A. Mourou, and A.H. Zewail, editors, Ultrafast phenomena VII, Springer,
pages 1416, Berlin, 1990.
[22] J. Hebling, E.J. Mayer, J. Kuhl, and R. Szipocs. Chirped-mirror dispersion-compensated
femtosecond optical parametric oscillator. Optics Letters, 20(8):919921, 1995.
[23] S.V. Chernikov, J.R. Taylor, N.S. Platonov, V.P. Gapontsev, G. Tastevin P.J. Nacher,
M. Leduc, and M. J. Barlow. 1083 nm ytterbium doped ber amplier for optical pump-
ing of helium. Electronics Letters, 33:787789, 1997.
[24] P. DeNatale G. Modugno M. Inguscio R. Paschotta, D.C. Hanna and P. Laporta. Power
amplier for 1083nm using ytterbium-doped bre. Optics Communications, 136(34):243246,
1997.
[25] I. D. Lindsay, B. Adhimoolam, P. Gro, M. E. Klein, and K.-J. Boller. 110GHz rapid, con-
tinuous tuning from an optical parametric oscillator pumped by a ber-amplied DBR diode
laser. Optics Express, 13(4):12341239, 2005.
[26] J. Swiderski, A. Zajac, M. Skorczakowski, Z. Jankiewicz, and P. Konieczny. Rare-earth-doped
high-power ber lasers generating in near infrared range. Opto-Electronics Review, 12(2):169
173, 2004.
51
BIBLIOGRAPHY
[27] H.M. Pask, R.J. Carman, D.C. Hanna, A.C. Tropper, C.J. Mackechnie, P.R. Barber, and J.M.
Dawes. Ytterbium-doped silica ber lasers: Versatile sources for the 11.2 m region. IEEE
Journal of Selected Topics in Quantum Electronics, 1(1):213, 1995.
[28] R. Paschotta, J. Nilsson, A. C. Tropper, and D. C. Hanna. Ytterbium-doped ber ampliers.
IEEE Journal of Quantum Electronics, 33(7):10491056, 1997.
[29] J. C. Diels and W. Rudolph. Ultrashort laser pulse phenomena, chapter 8. Academic press,
San Diego, 1996.
[30] H.A. Haus. Waves and Fields in Optoelectronics, chapter 13. Prentice-Hall, Englewood Clis,
New Jersey, 1984.
[31] Z.A. Yasa and N.M. Amer. A rapid-scanning autocorrelation scheme for continuous monitoring
of picosecond laser pulses. Optics Communications, 36(5):406408, 1981.
[32] D.M. Rie and A.J. Sabbah. A compact rotating-mirror autocorrelator design for femtosecond
and picosecond laser pulses. Review of Scientic Instruments, 69(9):30993102, 1998.
[33] J. Kondziela. Accurately measure laser spectral characteristics. Technical report, EXFO
Burleigh Products Group, 2003.
52

You might also like