Julius Vengelis • Audrius Dubietis

LASER
PHYSICS
Lecture notes
VILNIUS UNIVERSITY PRESS
 Faculty of
Physics

Julius Vengelis • Audrius Dubietis

LASER
PHYSICS
Lecture notes

VILNIUS
UNIVERSITY
PRESS
2023
Reviewed by:
assoc. prof. dr. Domas Paipulas (Vilnius University)

Bibliographic information is available on the Lithuanian Integral Library

Information System (LIBIS) portal ibiblioteka.lt
ISBN 978-609-07-0800-2 (Digital PDF)

© Julius Vengelis, 2023

© Audrius Dubietis, 2023
© Vilnius University, 2023 ­
Foreword
Lasers and laser-related technologies are an integral part of modern science
and are widely used in many areas of everyday life. The number of Nobel
Prizes awarded to relevant discoveries related to lasers or using lasers as a
tool in the past 60 years is extraordinary compared to any other technology
or device. This proves that lasers are unique, incredibly versatile (in terms of
generated pulse durations, wavelength, power and other parameters) devices
pushing frontiers in many fields of science ranging from investigations of ul­
trafast phenomena on a femtosecond time scale to material science, nonlinear
optics, frequency metrology, laser cooling, medicine, etc. Apart from shed­
ding light on fundamental scientific problems, lasers have also found their
place in our everyday life. Laser pointers, bar code readers, information pro­
cessing devices (DVDs, Blu-Ray devices), laser range finders, laser surgery
tools, laser material processing systems (laser engraving, cutting, welding
and other devices) and many more. Therefore it is important to understand
principles of their operation and key technological developments that enable
such broad field of applications.
The guiding theme of this textbook is principles of laser operation with
the intent for the reader to arrive at basic understanding how the laser works,
what are its key components and what are the technological innovations that
allow to achieve unique characteristics of laser radiation (e.g., generate and
amplify the shortest pulses, convert laser wavelength by nonlinear optical
methods). The readers will also be able to understand why laser radiation
is unique compared to other natural or artificial light sources, e.g., the Sun,
light bulbs, LEDs, etc. Chapter 1 provides a brief historic retrospect of laser
development, introduces the key principles of laser operation, outlines the
basic characteristics of laser radiation, their measurement tools and laser
safety. The relevant properties of photons, light waves, laser beams and
laser pulses, as well as coherence properties of laser radiation are discussed
in Chapter 2. Chapter 3 is devoted to optical resonators: resonator algebra,
stability conditions, resonator modes, sources of resonator energy losses and
their importance. Chapter 4 addresses the fundamental light-matter inter­
action processes that govern laser operation, laser gain, amplification band­
width, laser energy level systems and pump sources. Chapter 5 introduces
the regimes of pulsed laser operation: free running, Q-switching and mode-
locking along with methods of their practical implementation Chapter 6 fo­
cuses on the generation and amplification of femtosecond laser pulses, their
measurement techniques. Chapter 7 gives a brief overview of solid-state, gas
and semiconductor lasers, while detailed descriptions of miscellaneous lasers
and their practical applications are intentionally left for students’ seminar
presentation. Finally, Chapter 8 introduces the principles of laser wavelength
conversion via nonlinear optical effects: second harmonic, sum frequency gen­
eration and optical parametric amplification. Simply put, these chapters are
intended for undergraduate students gaining a deeper understanding of laser
physics. The short english-lithuanian glossary at the very end is added to
help native lithuanian students for correct usage of laser physics terms.
CONTENTS 3

Contents
1 Principles of lasers in a nutshell 6
1.1 What is a Laser? . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Historical perspective . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Light Amplification by Stimulated Emission of Radiation . . 13
1.4 Quantifying laser radiation . . . . . . . . . . . . . . . . . . . . 16
1.5 Laser safety classes . . . . . . . . . . . . . . . . . . . . . . . . 25

2 Light beams and pulses 29

2.1 Photons and their properties . . . . . . . . . . . . . . . . . . . 30
2.2 Light waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3 Gaussian beams . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.4 Coherence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.5 Spatial coherence . . . . . . . . . . . . . . . . . . . . . . . . . 45

3 Optical resonators 49
3.1 Light rays and ray (or ABCD) matrices . . . . . . . . . . . . . 49
3.2 Laser resonator stability . . . . . . . . . . . . . . . . . . . . . 53
3.3 Hermite-Gaussian beams and Laguerre Gaussian beams . . . . 57
3.4 Resonance frequencies . . . . . . . . . . . . . . . . . . . . . . 61
3.5 Fabry-Perot resonator . . . . . . . . . . . . . . . . . . . . . . 64
3.6 Losses in resonators . . . . . . . . . . . . . . . . . . . . . . . . 68

4 Light interaction with atomic systems 72

4.1 Einstein’s treatment of spontaneous and induced transitions . 72
4.2 Gain coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.3 Homogeneous and inhomogeneous spectral linewidth broadening 80
4.4 Three-level and four-level lasers . . . . . . . . . . . . . . . . . 83
4.5 Lasing at multiple wavelengths . . . . . . . . . . . . . . . . . 86
4.6 Pump sources . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.7 Laser oscillation . . . . . . . . . . . . . . . . . . . . . . . . . . 91

5 Pulsed laser operation 95

5.1 Free-running regime and relaxation oscillations . . . . . . . . . 96
5.2 Q-switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.3 Methods of Q-switching . . . . . . . . . . . . . . . . . . . . . 105
5.4 Mode-locking . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.5 Methods of mode-locking . . . . . . . . . . . . . . . . . . . . . 114
CONTENTS 4

6 Generation and amplification of ultrashort light pulses 122

6.1 Dispersion control in the resonator . . . . . . . . . . . . . . . 122
6.2 Measurement of ultrashort laser pulses . . . . . . . . . . . . . 127
6.3 Amplification of ultrashort light pulses . . . . . . . . . . . . . 131
6.4 Pulse stretchers and compressors . . . . . . . . . . . . . . . . 135
6.5 Laser amplifier media . . . . . . . . . . . . . . . . . . . . . . . 138

7 Types of lasers 142

7.1 Solid-state lasers . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.2 Gas lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
7.3 Semiconductor lasers . . . . . . . . . . . . . . . . . . . . . . . 149

8 Laser wavelength conversion by nonlinear optical methods 153

8.1 Nonlinear optical susceptibility . . . . . . . . . . . . . . . . . 153
8.2 Second harmonic generation . . . . . . . . . . . . . . . . . . . 154
8.3 Optical parametric amplification . . . . . . . . . . . . . . . . . 157
8.4 Optical parametric amplifiers . . . . . . . . . . . . . . . . . . 160

9 EN - LT Glossary 165
This page is intentionally left blank
1 Principles of lasers in a nutshell 6

1 Principles of lasers in a nutshell

In this chapter readers will briefly get acquainted with key principles of lasers:
what is a laser, how it works, what is so special about light coming out of a
laser and what quantities are used to characterize laser radiation. In addition,
a brief historic overview and a short note about laser safety is also presented.

1.1 What is a Laser?

The first acquaintance that most of us had with lasers was seeing how dif­
ferent is light from the laser and other light sources. Let us compare light
patterns coming from an incandescent light bulb, light emitting diode (LED)
flashlight and a green laser pointer (Fig.1).

Figure 1: Comparison of light patterns coming from three different light sources.
The left image is adapted from Wikimedia under GFDL 1.2 license.

Stark differences between light coming from a laser and the other two sources
are immediately visible allowing to recognize distinctive features of laser ra­
diation:
• Laser light has low divergence (is highly directional): laser
beam is almost aligned to the direction of propagation, with little an­
gular spread from the point of origin. In fact, if the authors had chosen
a He-Ne laser instead of a cheap laser pointer, the laser beam in the
photo would have been even smaller, i.e., with even lower divergence.
For He-Ne laser typical full angle beam divergence can be as low as
1 mrad.
• Laser light is nearly monochromatic: in general, laser emits light
with very narrow spectrum. When it passes through a dispersive prism,
1.1 What is a Laser? 7

we see negligible dispersion since its spectrum consists of essentially a

single color (see Fig.2). Note that some LEDs have certain colors which
can even be defined by a specific wavelength (e.g., 700 nm ” red” LED),
however, such specification states only the central wavelength with
spectral bandwidth being several tens of nanometers, whereas lasers
are able to emit light such that all the emitted power is concentrated
within extremely narrow spectral range.

Figure 2: The concept of monochromaticity can be understood using a dispersive

prism. White light consists of many different wavelengths which can be distin­
guished when light passes through a dispersive prism. Light coming from a LED
has a relatively narrow spectrum, so dispersion effect after passing the prism is
small. Laser radiation has very narrow spectrum and dispersion after passing the
prism is negligible.

• Laser light is coherent: emitted waves are in phase in time and in

space. In practice we can check this by observing interference when
passing a He-Ne laser beam through an interferometer (e.g., Michelson
interferometer) and viewing the interference pattern on a screen (see
Fig.3). High contrast fringes are clearly visible and changing length of
one interferemeter arm causes the fringes to move. Temporal coherence
is defined by the monochromaticity of light, whereas spatial coherence
is defined by low divergence of the light beam.
1.1 What is a Laser? 8

Figure 3: Interference pattern of a collimated He-Ne laser beam after passing

Michelson interferometer.

These properties make laser a unique light source, which offers a wealth of
specific applications not possible with any other light source. To mention a
few: laser pointers make use of low divergence of laser light; laser-induced
fluorescence utilizes the fact that laser radiation is monochromatic, so we can
efficiently excite fluorescence of only the desired molecules; interferometry,
holography and optical sensors rely on coherence of laser light.
Summarizing the above, a following definition of a laser could be given:
Laser is a light source producing low divergence, coherent and monochro­
matic light in the optical region of the electromagnetic spectrum.

Restriction to the optical range of electromagnetic spectrum, which includes

ultraviolet (UV), visible (VIS) and infrared (IR) radiation is very convenient
and necessary, since lasers that emit light in the microwave region due to
historic reasons are called – masers. However, the remaining part of the
definition is not very accurate since a certain degree of monochromaticity,
relatively low divergence and some degree of coherence can be achieved with
other light sources. Mercury vapor lamps emit a relatively narrow band of
spectrum which can be described as nearly monochromatic. Specific optical
systems allow sufficient collimation of light from non-laser sources effectively
reducing beam divergence to a small value. Superluminescent diodes (SLD)
are devices that have some properties of LEDs and some properties of laser
diodes including a certain degree of coherence; such sources of light are used
in low coherence interferometry. Another example is Young’s interference ex­
periment where he used one slit to make a point light source from a burning
candle light and another set of two slits to make two point sources. Light
coming out of the two slits was coherent enough to observe interference when
1.2 Historical perspective 9

the beams were overlapped on the screen. Although none of the presented ex­
amples can match low divergence, monochromaticity and coherence of lasers,
they prove that the initial definition of a laser is rather quantitative, depend­
ing on quantitative description of what we call low divergence, coherence and
monochromaticity.
A truly accurate description of what is a laser comes from the principle
of its operation, i.e. how light with such properties is generated:

Laser is a quantum light generator emitting light through a pro­

cess of optical amplification based on stimulated emission of ra­
diation.

The definition above can be explained as follows. ”Quantum light gener­

ator” means that laser is a device converting one form of energy to another
(the definition of term ”generator”) via certain quantum process – optical
amplification based on stimulated emission of radiation. In the next section of
this chapter we will discuss the basic principles of this quantum process. The
rigor and accuracy of this definition is that it is tied to qualitative definition
of principle of operation rather than a certain quantitative measure. This
is important since different types of lasers can generate light with specific
properties, e.g. femtosecond laser radiation has a relatively broad spectral
bandwidth (several tens of nanometers or even more), whereas a stack of
high power laser diodes can emit light with considerably high divergence,
nonetheless in both cases the final definition of a laser is correct.

1.2 Historical perspective

The fundamental background behind laser generation (optical amplification)
theory was laid in 1916, when A. Einstein theoretically predicted the exist­
ence of a novel emission phenomenon – stimulated (induced) emission of
radiation, which is caused by a presence of external radiation field1,2 . This
discovery had utmost importance since stimulated emission of radiation lies
on the physical basis of laser generation process. The phenomenon of stimu­
lated emission was first experimentally observed by R. Landenburg in 1928.
It was called ”negative absorption” to reflect the fact that light and matter
interaction is stimulated by external radiation field, but occurs in the oppos­
ite way to absorbtion. At that time stimulated emission was thought to have
little practical use since the energy levels of an atom are populated accord­
1
J. Hecht, Short history of laser development, Applied Optics 49, F100–F121 (2010)
2
W. Zinth, A. Laubereau, W. Kaiser, The long journey to the laser and its rapid
development after 1960, European Physical Journal H 36, 153–181 (2011)
1.2 Historical perspective 10

ing to Boltzman law and absorption is the dominating effect. The idea of
optical amplification using stimulated emission in gases was first formulated
by Russian physicist V. Fabricant, but it still was not clear how to realize
this idea in practice. After World War II, W. Lamb and R. C. Retherford
noticed that population inversion can be achieved during Nuclear Magnetic
Resonance (NMR) and later stimulated emission of radio waves was observed.
Population inversion at that time was interpreted as ”negative absolute tem­
perature” as would follow from the Boltzmann distribution. In 1951 C. H.
Townes conceived an idea that electromagnetic radiation could be ampli­
fied via stimulated emission in a resonator, and in 1954 Townes and J.
P. Gordon invented the first device operating on that principle, the maser,
which emitted electromagnetic radiation in the microwave (millimeter wave)
range. The term ”maser” is an acronym for Microwave Amplifier based on
Stimulated Emission of Radiation. In 1964 A. Penzias and R. Wilson using
maser as a sensitive microwave detector, discovered the cosmic microwave
background radiation, which is the relic radiation from the Big Bang. For
this fundamental discovery they received the Nobel Prize in Physics in 1978.
In 1958 C. H. Townes and A. L. Schawlow theoretically predicted the
possibility to amplify electromagnetic radiation in the optical range and put
forward the key principles of laser generation3 . They also foresaw that a
Fabry-Perot etalon could be used as a simple resonator in the optical range.
What was still missing, was to find suitable materials and methods to cre­
ate population inversion, which became a subject of very intense research by
many groups at the time. At present, it is widely recognized that the first
laser was invented in 1960 by T. Maiman, when he reported the laser
action in a ruby crystal (where natural chromium impurities exist) with alu­
minum coated ends serving as a resonator mirrors. To achieve the population
inversion, Maiman used a white light flashlamp as a pump source. The word
”recognized” is used here by purpose, since Maiman described his results very
vaguely and it might be that he only observed the stimulated emission below
the laser generation threshold. These doubts are also based on the fact that
Physical Review Letters rejected Maiman’s manuscript, but he managed to
publish a short message in Nature 4 . It can be claimed with a confidence that
the first laser, where light amplification and directional radiation in the form
of a coherent beam, was reported a few months later by R. J. Collins, D.
F. Nelson, A. L. Schawlow, W. Bond, C. G. B. Garrett and W. Kaiser5 . At
3
A. L. Schawlow and C. H. Townes, Infrared and optical masers, Physical Review112,
1940 (1958).
4
T. H. Maiman, Stimulated optical radiation in ruby, Nature 187, 493-494 (1960).
5
R. J. Collins, et al., Coherence, narrowing, directionality, and relaxation oscillations
in the light emission from ruby, Physical Review Letters 5, 303 (1960).
1.2 Historical perspective 11

the end of 1960 A. Javan, W. R. Bennett and D. R. Herriott demonstrated

the first gas laser, where they used helium and neon gas mixture which was
pumped by an electric discharge6 . Interestingly, the first He-Ne laser emitted
radiation at 1.15 µm wavelength and the well-known He-Ne laser emitting
632.8 nm radiation was demonstrated several years later. Even today He-Ne
lasers are widely used in many laboratories.
Soon after invention of the first laser, a variety of lasers, which used
solid-state materials doped with rare-earth metal ions as active media, were
demonstrated. However, most of these first solid-state lasers operated only
at cryogenic temperatures. In 1961 the first Nd:glass laser was demonstrated,
which later became one of the most important solid-state lasers. Subsequent
observations of lasing in various active media led to a laser boom. In 1962
the first semiconductor laser was invented. The semiconductor laser oper­
ated only at cryogenic temperatures due to contemporary state of the art of
semiconductor technologies. In the same year, the first chemical laser was
invented, which afterwards served as the basis of laser weapons. In 1964 the
first CO2 laser was demonstrated and soon became one of the most important
technological lasers. At the same year, the first fiber laser was demonstrated,
where the active medium was an optical fiber with a rare-earth metal doped
core7 . In 1966 organic dye lasers were invented, serving as the main ul­
trashort pulse laser sources for a long time until the invention of femtosecond
Ti:sapphire laser in 1991. In 1970 efficient excimer lasers were demonstrated.
The first pulsed lasers were able to generate light pulses with durations ran­
ging from a few microseconds to a few nanoseconds. The first picosecond
laser pulses were generated soon after, when passive mode-locking was dis­
covered in 1967, while the first femtosecond pulses were produced in 1974.
Although the first lasers were pumped by flashlamps or electrical discharges,
the idea to use other lasers as pump sources was proposed in 1963, when it
was noticed that radiation from GaAs laser diode is efficiently absorbed by
Nd-doped solid-state media. However, this idea became widely employed in
practice only after semiconductor laser technology was sufficiently developed,
giving rise to the whole generation of the so-called diode-pumped solid-state
(DPSS) lasers.
The first lasers were called ”optical masers” and the term ”laser” became
widespread later. Both masers and lasers are based on the same principle,
but the acronym ”laser” stands for ”Light Amplification by Stimulated Emis­
6
A. Javan, W. R. Bennett, Jr., and D. R. Herriott, Population inversion and continuous
optical maser oscillation in a gas discharge containing a He-Ne mixture, Physical Review
Letters. 6, 106 (1961).
7
C. J. Koester and E. Snitzer, Amplification in a fiber laser, Applied Optics 3,
1182–1186 (1964)
1.2 Historical perspective 12

sion of Radiation”, highlighting the fact that, unlike masers which amplify
microwaves, lasers amplify the electromagnetic radiation in the optical range.
Interestingly, the term ”laser” was first used in 1959 in a conference paper8 by
a graduate student G. Gould, who at that time worked on a doctoral thesis
about the energy levels of excited atoms of thallium. Two years before, he
coined some general ideas about how lasers could be built and where they
could be used in a lab notebook which he notarized. In 1959 he filed a patent
application with the high-tech company he worked in. The U. S. Patent Office
denied his application and granted it to Bell Labs, which resulted in a twenty
eight year-long ”patent war”. Eventually, in 1987 G. Gould won the first pat­
ent lawsuit and was issued patents for optically pumped and gas-discharged
lasers. However, the historical question about assigning the credits for laser
invention is difficult since patents tell little about how the ideas arose and
spread among researchers. C. H. Townes, N. Basov and A. Prochorov were
awarded the Nobel Prize in Physics in 1964 for their input which led to in­
vention of the laser. Another key researcher in the field, A. Schawlow, was
awarded the Nobel Prize in Physics in 1981 for laser applications in spectro­
scopy. The beginning of the laser era in 1960 gave birth to a new branch of
optics – nonlinear optics. Since the invention of the laser, a huge variety of
new laser materials was developed, the power of lasers increased enormously
and the duration of emitted pulses decreased significantly, yielding a coher­
ent light with unprecedented intensity. The very first pulsed lasers generated
light pulses with a duration of a few microseconds, while modern ultrashort
pulse lasers can provide light pulses with a few femtosecond duration – a
billion of times shorter! The pulse energy and power grew tremendously,
and at present lasers are reliable turn-key devices, without which modern
physics, biology, chemistry, medicine and materials science laboratories and
future progress of these sciences is unimaginable. Interestingly, the laser was
not invented for a specific purpose, the skeptics even claimed that invention
of the laser was ”not a solution of a specific problem, but, on the contrary,
a solution that created many problems”. Despite this, lasers gradually and
imperceptibly became an integral part of many scientific, industrial, techno­
logical, military and many other fields, so it can be concluded that laser is
the most important optical device invented in the last six decades.
8
G. Gould, The LASER, Light Amplification by Stimulated Emission of Radiation, in
The Ann Arbor Conference on Optical Pumping (University of Michigan, USA, June 15-18
1959)
1.3 Light Amplification by Stimulated Emission of Radiation 13

1.3 Light Amplification by Stimulated Emission of Ra­

diation
The term ”laser” is an acronym of a phrase ”Light Amplification by Stimulated
Emission of Radiation”. The same is true for ”maser”, only the first word
in the phrase is ”Microwave” instead of ”Light”. We will now discuss key
principles of laser operation and stimulated emission phenomenon.
To understand key principles of laser generation, we must first get ac­
quainted with principal components of a laser:

Figure 4: Principal components of a typical laser.

1. Lasing medium (gain medium): a medium which is excited and

emits laser radiation via stimulated emission process. It is also called
active medium. It can be a solid-state medium (crystal, glass), semi­
conductor, liquid (organic solvents or dyes) or gas medium (He-Ne,
CO2 ). Lasers are usually named after the lasing material, e.g., Nd:YAG
laser means that the lasing material is YAG (yttrium aluminum gar­
net) crystal doped with Nd3+ ions, rhodamine 6G laser means that the
lasing material is a rhodamine 6G dye, He-Ne laser means that the
active material is a mixture of He and Ne gas (Fig.5).

Figure 5: Examples of different lasing materials. Images adapted from Wikimedia

under CC BY 1.0, CC BY 2.5 and CC BY 3.0 licences respectively.
1.3 Light Amplification by Stimulated Emission of Radiation 14

2. Pump source: source of energy for excitation of the active medium

and subsequent laser radiation emission. In other words, energy of the
pump source is converted into laser radiation. Pump sources are classi­
fied into optical (e.g., flashlamp, laser diode), electrical (e.g., electric
current or electric discharge) and chemical (exothermic chemical re­
action).

3. Optical resonator (optical cavity): a part of the laser that provides

positive feedback: returns some part of emitted radiation back to
the active medium. Optical resonator consists of at least two mirrors
between which the lasing medium is placed (Fig.4). These mirrors
reflect laser radiation back to the active medium, where it is amplified.
One of the mirrors is made partially reflective, allowing a fraction of
light to leave the cavity producing the laser output beam. This mirror
is called the output coupler.

Another important aspect behind laser operation principles is under­

standing what happens in the gain medium when it is excited. The simplest
example is a two-level atom energy system depicted in Fig.6.

Figure 6: Possible ways of light and atom interaction.

An atom becomes excited when it absorbs a certain portion of energy. An

electron from the ground state orbital jumps to a higher energy orbital ab­
sorbing amount of energy equal to the difference between the ground and
higher state orbital energies. This energy is called excitation energy and the
light-matter interaction process is called absorption. Another light matter
interaction process takes place when the electron spontaneously returns from
the higher energy orbital to the ground state orbital, emitting a photon with
energy equal to the energy difference between higher and ground state orbit­
als. Such process is called spontaneous emission. The term ”spontaneous”
is used to emphasize the fact that it is not triggered by any external factor
1.3 Light Amplification by Stimulated Emission of Radiation 15

and the phase, direction and polarization of the emitted light wave are ran­
dom. Apart from the two aforementioned processes, yet another interaction
process is also possible. This process is called stimulated emission. In this
case when an electron is in the higher energy orbital (the atom is excited),
another photon can trigger the electron to jump to the ground orbital state
emitting a photon. The stimulated emission process occurs only when the
triggering photon energy is equal to the energy difference between higher
energy state and ground state orbitals. More importantly, the emitted and
corresponding light wave has identical phase, direction, polarization and
wavelength to the initial triggering photon.
Laser operation is based on light amplification by stimulated emission of
radiation. Let us discuss this in more detail.

Figure 7: Principal view of stimulated emission of radiation and lasing.

Consider that the lasing medium is at ground state (Fig.7a). When a pump
source (e.g., an electric flash) is turned on, some energy is transferred to
the lasing medium and some atoms become excited (denoted by red circles
in Fig.7b). The first key moment is spontaneous emission that starts from
the excited atoms and the emitted photons can then trigger stimulated emis­
sion of other excited state atoms (Fig.7c). The second important moment
is that the amount of excited electrons (population of an excited energy
level) must be greater than the amount of electrons in the lower (ground
state) energy level. Such state is called population inversion. The term
”inversion” indicates that this kind of population is impossible to achieve un­
der usual conditions: according to Boltzman distribution, the population of
higher energy levels decreases exponentially. When the emitted light (both
1.4 Quantifying laser radiation 16

due to spontaneous and stimulated emission) reaches resonator mirrors, it

is reflected back to the lasing medium. This increases stimulated emission
(Fig.7d) since after each reflection more and more light in the lasing medium
has the same wavelength, phase, direction and polarization. The stimu­
lated emission starts dominating (Fig.7e) and the number of photons with
the same wavelength, phase, direction and polarization rapidly grows with
each round-trip. After certain amount of round-trips, light amplification
coefficient (power increase per unit length of the lasing medium) exactly
compensates losses in the resonator. This equilibrium state is called laser
generation threshold. When it is exceeded, laser generation starts. Since
one of the resonator mirrors is partially reflective (output coupler), some
light always leaves the resonator in the form of a laser beam and can be
used for intended purposes.
The population inversion is necessary in order to achieve amplification of
light (lasing) due to the following reasons. Atoms are complex systems with
multiple energy levels, so photons that are emitted during spontaneous and
stimulated emission can be absorbed by other atoms in the groundstate –
this is called reabsorption. If population inversion condition is not achieved,
this is exactly what happens. In the opposite case (population inversion),
more photons are emitted than reabsorbed and the net result is amplification
of light by stimulated emission or, in other words, lasing.
It is important to note that here we described laser operation when its
dynamics (how electron population changes) are not interfered with by any
external means. This is the free-running regime of a laser. Practically used
lasers usually have components in the resonator which modulate lasing dy­
namics, e.g., enable generation of ultrashort pulses. This will be discussed in
detail later.

1.4 Quantifying laser radiation

There is a great variety of lasers in terms of their emitted radiation char­
acteristics which also determine potential applications of a particular laser.
We will now discuss how to characterize laser radiation in practice. For more
clarity, it helps to distinguish between the relevant parameters of lasers that
emit light continuously (continuous wave regime – CW) and lasers that
emit light in the form of short bursts that are called laser pulses (pulsed
regime) as shown in Fig.8.
1.4 Quantifying laser radiation 17

Figure 8: Time evolution of light electric field emitted from pulsed (black lines)
and continous wave (CW) lasers (red line). τp is the pulse duration, tr is the pulse
repetition period (inversely proportional to pulse repetition rate).

For lasers that operate in CW regime we can define these main paramet­
ers:

• Wavelength: emission wavelengths of modern lasers covers an ex­

tremely broad spectral range from the vacuum UV to far IR (Fig.9),
e.g., F2 excimer laser emits radiation of 157 nm wavelength, while Al-
GaAs quantum cascade laser emits radiation of up to 250 µm wavelength9 .
The diversity in emission wavelengths in fact defines an extremely broad
range of laser applications. There are two important features to note
regarding laser wavelength. First, the same gain medium can emit
several different wavelengths, e.g., He-Ne laser has these lasing
wavelengths: 543.36 nm, 593.93 nm, 604.61 nm, 611.8 nm, 632.8 nm,
1.15 µm; 3.39 µm. This is related to the fact that atoms/molecules
have many energy levels or their combinations that are suitable for
lasing. The particular lasing wavelength is selected by choosing an
appropriate resonator mirrors which reflect the desired wavelength.
Second, wavelength of several lasers, e.g. Ti:sapphire, Fe:ZnSe,
Fe:ZnS, can be continuously tuned over a certain wavelength
range. However, these are rare cases related to specific features of the
gain medium.

9
C. Walther, M. Fischer, G. Scalari, R. Terazzi, N. Hoyler, and J. Faist,”Quantum
cascade lasers operating from 1.2 to 1.6 THz”, Applied Physics Letters 91, 131122 (2007)
1.4 Quantifying laser radiation 18

Figure 9: Emitted radiation wavelengths of various lasers. Lasers with continu­

ously tunable radiation wavelength are denoted by square bracket marking.

Wavelength of light (including laser radiation) is measured using a

device called optical spectrometer (usually simply called ”spectro­
meter”). This instrument separates different wavelengths by refraction
in a prism or diffraction by a diffraction grating and measures their
intensity. The measured light intensity as a function of wavelength (or
frequency) is called the optical spectrum. Image of a spectrometer
and example of He-Ne laser radiation spectrum are depicted in Fig.10.

Figure 10: Optical spectrometer (left) and example of measured He-Ne laser
spectrum (right).

• Power, that is an amount of energy per unit time. It is important to

note that when characterizing pulsed laser radiation the term ”power”
is usually used to quantify the average power of laser radiation. Vari­
ous lasers have average powers ranging from milliwatts to megawatts,
e.g., solid-state microlasers produce output power of several tens of
milliwatts while chemical lasers typically produce output power of hun­
dreds of kilowatts and the most powerful versions can exceed megawatt
power level. The term, peak power, is also used to characterize pulsed
laser radiation and will be discussed below with other parameters of
1.4 Quantifying laser radiation 19

pulsed laser radiation.

The power of laser radiation is measured using a power meter. The
main types of these devices are: photodiode-based, thermal and pyro­
electric. In photodiode-based power meters the electric current
signal proportional to the power of laser radiation is generated in a
photodiode. Such detectors are fast, sensitive and have linear response
function over a broad range of powers: from fraction of nanowatts to
several milliwatts. On the other hand, they can be easily damaged if
too much power is applied and their response function differs signific­
antly for different wavelengths of light. Thermal power meters are
based on the heating effect of laser radiation: a temperature gradi­
ent between laser radiation absorbing area and a peripheral reference
area (where heat is dissipated) is measured using thermopiles. Thermal
power meters can measure tens of watts of power, with water-cooled
versions being able to measure even greater powers. They are also
durable, accurate and their response is almost constant for different
wavelengths of light. However, such power meters are quite slow (re­
sponse time is several seconds) and require thermal equilibrium with
the environment. An example of a thermal power meter consisting of
a sensor and measuring device is shown in Fig.11.

Figure 11: Thermal power meter. It consists of two connected parts: power sensor
and power meter device which displays the measured power and is used to control
the power meter.

Pyroelectric power meters, as the name suggests, are based on the

pyroelectric effect: temperature gradient induced by the incident laser
radiation in the pyroelectric material generates current signal which
is converted into a voltage signal that is proportional to the power of
1.4 Quantifying laser radiation 20

laser radiation. Such power meters are very fast, broadband (response
is independent of wavelength in a broad spectral range, usually from
200 nm to 20000 nm) and do not require thermal equilibrium with
the environment. However, pyroelectric power meters are fragile, less
accurate, cannot measure CW laser radiation power and are sensitive to
vibrations. Therefore, they are usually used to measure low repetition
rate laser power for each pulse.

• Beam size is usually measured as a diameter of the laser beam at a

certain power/intensity level (Fig.12).

Figure 12: Laser beam size measured at different intensity levels: w1/e2 – radius
at 13.6% of maximum intensity level; wF W HM – radius measured as half of beam
diameter at full width at half maximum intensity level.

The beam diameter can differ in the horizontal (x) and vertical (y)
directions; such beams are called elliptical since the beam spot on a
screen looks like an ellipse instead of a circle. Commonly used intens­
ity level values: FWHM – full width at half maximum, 1/e2 – 1/e2
(13.6%) of maximum intensity level, 1/e – 1/e (36.8%) of maximum
electric field amplitude level. There is also 4σ notation defined as 4
times the standard deviation of the horizontal or vertical marginal dis­
tributions respectively. It is the ISO international standard definition
for beam width. Despite the fact that 4σ notation is the international
ISO standard for beam width definition, other beam width definitions
are also commonly used. Beam size can be measured in several differ­
ent ways but the simplest one is to use a charge-coupled device (CCD)
camera which displays laser beam intensity distribution and estimates
its size (Fig.13). Professional beam width measurement software usu­
ally provides beam width values at various intensity levels. Although
1.4 Quantifying laser radiation 21

CCD camera is a simple and straightforward way to measure beam

profile and size, great care has to be taken not to damage it. Therefore
neutral filters that attenuate laser radiation are usually placed in front
of the CCD sensor.

Figure 13: CCD camera (left) and example of measured laser beam intensity pro­
file in 2D (top right) and 3D (bottom right). Color coding indicates the intensity.

Laser beams usually have intensity distribution described by a Gaus­

sian function and are therefore called Gaussian beams. Properties
of Gaussian beams will be discussed in more detail in Chapter 2. An­
other important aspect to note is that laser beam size tends to change
with propagation. If no additional optics is used, beams coming out
of a laser either diverge (diameter increases) or converge (diameter
decreases). If, using specific optical systems, size of the propagating
beam is maintained constant in the near field, such laser beam is called
collimated.

• Intensity is defined as power per unit area:

P ower
I= (1)
Area

As with the term ”power”, when characterizing pulsed laser radiation

it is important to clearly distinguish whether by using the term ”in­
tensity” we mean average intensity or peak intensity of laser ra­
diation. Unlike ”power”, the term ”intensity” in the context of pulsed
lasers usually means peak intensity unless the text states otherwise.
1.4 Quantifying laser radiation 22

In addition to the above parameters, characterization of pulsed laser op­

eration involves a set of supplementary parameters:

• Pulse duration is usually defined as width of a pulse temporal shape

at FWHM power/intensity level (see Fig.14 and Fig.19). A huge vari­
ety of pulsed lasers cover a broad range of possible pulse durations:
lasers operating in free-running regime generate microsecond dura­
tion pulses, but lasers operating at advanced regimes generate nano­
second (1 ns=10−9 s), picosecond (1 ps=10−12 s) and even femtosecond
(1 fs=10−15 s) pulses. Moreover, using nonlinear optical techniques
and femtosecond laser systems, attosecond (1 as=10−18 s) duration
pulses can be generated with current world record standing at 43 as!10
There is no universal method for laser pulse duration measurement: it
essentially depends on the pulse duration itself. Long pulse (several
nanoseconds or longer) duration can be measured simply by using a
photodiode and observing the generated pulse temporal profile on the
oscilloscope screen (Fig.15 left).

Figure 14: Electric field oscillations in an ultrashort light pulse and its duration
estimation at FWHM level (τp ). Thick blue line above the oscillations represents
what is called the pulse envelope.

10
T. Gaumnitz et al, Streaking of 43-attosecond soft-X-ray pulses generated by a pass­
ively CEP-stable mid-infrared driver, Optics Express 25, 27506-27518 (2017).
1.4 Quantifying laser radiation 23

Figure 15: Signal in an oscilloscope screen corresponding to temporal profiles of

laser pulses detected with a photodiode (left); example of spectrogram of a complex
laser pulse measured with a streak camera (right).

For shorter duration laser pulses more sophisticated devices/techniques

have to be used, ultrafast photodiodes are used for pulse durations
down to 50 ps, they are expensive and require ultrafast oscilloscopes
and special cables. Streak camera is used for pulse durations from
nanoseconds to 500 fs, it is an extremely expensive device, though it is
also able to measure pulse spectrum (Fig.15 right). Correlation tech­
niques are used to measure pulse durations ranging from hundreds of
picoseconds to 30 fs, these are relatively cheap and straightforward
methods but cannot measure complex pulses. Time-frequency tech­
niques comprise a broad range of very complex methods that enable
characterization of pulse duration (to the shortest possible durations),
shape, phase and more. In the context of laser pulse duration, there
are several widely used definitions to classify the temporal width of
laser pulse: long pulses are pulses whose duration is several tens of
ns and longer; short pulses are- pulses with duration from tens of ns
to 100 ps; ultrashort pulses are pulses which duration is less than
100 ps.

• Pulse repetition rate is the number of pulses emitted per second.

It is the inverse of pulse repetition period tr (Fig.8). Different lasers
typically have pulse repetition rates from tens of Hz to 100 MHz, though
in specific cases it can be from below 1 Hz to hundreds of GHz. Pulse
repetition rate is usually measured using a photodiode and estimating
the temporal interval between pulses (looking at the oscilloscope screen
– Fig.15 left).
1.4 Quantifying laser radiation 24

• Pulse energy is the ratio of average power (P ) and pulse repetition

rate (frep ):
P
Ep = (2)
frep

• Peak power is the maximum power of a single pulse. It is defined as

the ratio of pulse energy (Ep ) and pulse duration (τp ):

Ep
Pp = (3)
τp

Peak power of ultrashort laser pulses can vary in the range from MW
to PW, however, we should have in mind that such extreme power is
achieved only for a very short moment of time (duration of the pulse).

• Central (carrier) frequency is defined as a frequency of maximum

spectral intensity (example is shown Fig.16). The shorter is the pulse,
the broader is its spectrum. Ultrashort laser pulses have a very broad
spectrum (unlike CW lasers whose spectrum is extremely narrow), thus
the term central frequency is used. In the wavelength domain we can
also define the central (carrier) wavelength.

Figure 16: Spectrum of the ultrashort laser pulse and its main characteristics:
central frequency (ωc ) and spectral width at FWHM level (∆ωF W HM ).

Laser spectra usually have intensity distribution described by a Gaus­

sian function and are therefore sometimes called Gaussian spectra.
In this case the central frequency matches the spectrum peak. For
1.5 Laser safety classes 25

complex non-Gaussian spectra the central frequency is defined as:

c
νcentroid = , (4)
λcentroid
where νcentroid is center of mass of the spectrum in the frequency do­
main, called spectral centroid:
N −1
n=0 νn In
νcentroid = N −1
, (5)
n=0 In

here νn is central frequency of a discrete bin, in this case – frequency

step of the used spectrometer and In is the weighted coefficient in this
case – measured intensity value of the frequency component.
• Spectral width (bandwidth) is the width of the pulse spectrum. As
already mentioned, ultrashort pulses have a relatively broad spectrum,
so their spectrum width is usually defined at FWHM level (Fig.16).
The estimation in the case of Gaussian spectrum is simple as Gaussian
spectrum is symmetrical (in the frequency domain) and smooth.
Spectral bandwidth is inversely proportional to pulse duration; the
shorter is the pulse, the broader is its spectrum and vice versa. Ul­
trashort pulse spectrum width, when its central wavelength is in the
visible range, in some cases can exceed hundreds of nanometers. The
spectrum of such pulse covers the entire visible range and the laser ap­
pears to be generating white light. Such lasers are called white light
lasers.
• Peak intensity is the peak power per unit area:
Pp
I= (6)
Area
Due to unique properties of laser radiation which enable to achieve
extremely high peak power and focus the beam to a spot size that
is close to wavelength, modern laser systems can achieve significantly
greater peak intensities than any other light source.

1.5 Laser safety classes

Laser radiation is especially harmful to the eyes and skin. It can cause tem­
porary or permanent retina and skin damage, which could severely disrupt
eyesight and cause skin burns. In the case of repetitive exposure to laser
radiation eye or skin illnesses can develop. Laser radiation effect to the eye
or skin cells depends on laser source parameters:
1.5 Laser safety classes 26

• Wavelength of radiation. The effect of ultraviolet, visible and in­

frared radiation to biological systems is very different. Due to large
photon energy, the UV radiation produces photo-ionization, while the
IR radiation has thermal effect. Visible light has the largest penetra­
tion depth in living tissues, since their main component is water, which
is fully transparent in the visible range, therefore visible radiation can
affect deeper tissues.
• Laser operation regime (continuous wave or pulsed), which determ­
ines the exposure time. In the case of pulsed regime, the strength of
effects is determined by the pulse repetition rate and duration. The
peak intensity (the amount of energy per unit area per unit of time) of
nanosecond, picosecond and femtosecond pulses is very different, there­
fore, physical mechanisms of their interaction with biological tissues are
also significantly different.
• The power or pulse energy of a laser source, or in the most general
case, the fluence (power or energy per unit area), which depends on
the parameters of a laser beam and character of the beam propagation
(focused, collimated or scattered beam).
Cell or tissue damage due to laser radiation occurs because of photochem­
ical reactions, thermal effects and photo-ionization. When the laser fluence
increases, these effects manifest themselves in the above order and the result
is different for each of these effects. Low fluence (up to 1 W/cm2 ) usually
causes reversible photochemical reactions which have short-term effects and
do not cause any irreversible changes. Laser radiation with 10-105 W/cm2
fluence produces a strong thermal effect (especially during longer exposure
times) due to which the biological tissue is heated, proteins in the cells break
up, and the cell eventually dies. The fluence greater than 1010 W/cm2 is
achieved only with short pulse lasers. Short laser pulses produce strong ion­
izing effect since the electric field strength can exceed electron bond energy.
In the wake of intense laser radiation microplasma is formed, which can cause
ablation (evaporation) of the tissue’s outer layer, dissociation of intermolecu­
lar bonds (optical damage), while plasma expansion generates acoustic waves,
which affect and disrupt cells and tissues that even were not directly in the
radiation exposure area.
When evaluating laser radiation danger, the laser sources are divided into
4 classes. Laser safety class is marked on the laser housing or on the device
part where the laser is located.
• Class 1: The power density of these lasers is so small that their ra­
diation is not dangerous to skin or to the eyes in the case of any use
1.5 Laser safety classes 27

of the laser. Sometimes two subclasses are distinguished: class 1M

laser system is a class 1 laser using magnifying optics which is incap­
able of causing injury during normal operation unless collecting optics
are used; class 1C laser (established in July 2015) is a class 1 laser
that covers laser systems that are designed for direct contact with the
”objective”. This can be laser systems for hair removal, reduction of
wrinkles, treatment of akne, tattoo removal, etc.

• Class 2: Only continuous wave (CW) lasers emitting visible (400 nm–
700 nm) light whose power does not exceed 1 mW fall into this class.
Direct exposure to such laser radiation is considered not dangerous if
the exposure time does not exceed 0.25 s (the duration of eye blinking
reflex). Sometimes a subclass is distinguished: class 2M laser system is
a class 2 laser using magnifying optics. It includes visible wavelength
lasers incapable of causing injury in 0.25 seconds unless collecting optics
are used.

• Class 3: This class is also divided into 3R and 3B subclasses. 3R class

includes lasers whose power in the visible range (400 nm–700 nm) does
not exceed 5 mW. Direct exposure to such laser radiation is considered
not dangerous, but focused radiation can cause temporary vision dis­
order. 3B class includes CW lasers emitting in the entire optical range
(ultraviolet, visible and infrared), whose power does not exceed 0.5 W
and pulsed lasers emitting in the visible range, whose pulse energy does
not exceed 30 mJ. Direct exposure to such laser radiation is danger­
ous to skin and eyes, therefore, when working with such lasers it is
necessary to wear laser safety goggles and avoid direct contact of laser
radiation with skin.

• Class 4: It is the most dangerous class. All the lasers whose power
or pulse energy is greater than 3B subclass lasers fall into this class.
All ultrashort pulse (picosecond and femtosecond) lasers also fall into
this class. Both direct and scattered (reflected from matted or diffuse
surfaces e.g., sheet of paper) radiation as well as reflections from op­
tical or mechanical element surfaces (lenses, filters, mounts, etc.) is
dangerous to skin and eyes. The most dangerous in that regard are
lasers emitting ultraviolet and infrared radiation since it is not directly
visible to human eye. Maximum safety precautions are necessary when
working with such lasers.
1.5 Laser safety classes 28

Summary of Chapter 1
• Laser operation is based on stimulated emission of radiation. The
term ’laser’ stands for Light Amplification by Stimulated Emis­
sion of Radiation.

• Laser radiation is unique compared to light coming from other

sources: it has low divergence, is monochromatic and coherent.

• Laser radiation is characterized by a set of parameters: central

wavelength, spectral width, pulse duration, beam width, intens­
ity, peak and average power.

• Any laser consists of three principal components: active medium,

optical resonator and pump source.

• Laser invention has a long and rich history and is a result of

many competing scientific groups rather than a single person.

• Laser radiation is harmful to the eyes and skin. According to the

level of danger, the laser sources are divided into 4 safety classes.
2 Light beams and pulses 29

2 Light beams and pulses

Light can be described in two ways: either as electromagnetic waves
or as particles (photons) (Fig.17). From the viewpoint of laser physics,
these concepts are tightly related although they describe different aspects
of light. The wave description enables accurate analysis of light beam and
pulse propagation phenomena: beam diffraction, pulse dispersion, and wave
interference. Equations describing wave optics are derived from Maxwell
equations, which describe propagation of electromagnetic radiation in any
medium.

Figure 17: 1 ns duration 1064 nm wavelength Nd:YAG laser pulse pictured as

waves (upper image) or particles (lower image).

In some cases wave optics can be simplified to geometrical optics, when

light propagation is analyzed using only reflection and refraction laws. This
assumption is very convenient when analyzing light propagation in optical
resonators and determining their stability. Geometrical optics can be applied
when diffraction and interference phenomena can be ignored e.g., when the
size of resonator optical elements is considerably greater than wavelength of
light.
However, light absorption and emission processes cannot be described by
Maxwell equations. Quantum optics is based on the postulate that elec­
tromagnetic field and matter interaction is quantized: atoms and molecules
absorb and emit only quantized energy. Therefore, the operation of laser me­
dium (laser generation and amplification), laser-matter interaction, detection
of radiation are described by means of quantum optics.
2.1 Photons and their properties 30

2.1 Photons and their properties

The smallest quantum of light is a photon. The energy of a photon is defined
as:
E = hν , (7)
where h = 6.62 × 10−34 J/s is the Planck’s constant and ν is the frequency of
the corresponding electromagnetic wave. The wavelength of light is defined
as λ = c/ν, where c is the speed of light in a vacuum. The photon propagates
in a certain direction and the quantity ⃗k = 2π⃗e/λ is called the wavevector.
Here ⃗e is the unity vector which characterizes the propagation direction. The
momentum of a photon is defined as p⃗ = h⃗k/2π with and its modulus equals
to p = h/λ.
Photon as any other quantum particle is a subject to the Heisenberg
uncertainty principle, which states that relevant parameters, i.e. the po­
sition and momentum, can only be known (or determined) with a certain
accuracy. In fact, the uncertainty principle poses certain limitations that are
important in laser physics. Namely, light beams can be ”diffraction-limited”
due to the smallest possible photon position-momentum uncertainty, whereas
light pulses can be ”bandwidth-limited” due smallest possible time-energy
uncertainty. We will now discuss this in detail.

Figure 18: Diffraction of the light beam and relevant parameters. Red circles
mark an idealized view of photons in the beam.

Consider a light beam propagating in the z direction in the coordinate

system, as schematically illustrated in Fig. 18. It can be analyzed as a col­
lection of many photons each with a certain lateral position and propagation
direction which is related to momentum of the photon. The uncertainty
principle states that the photon position (or its localization) in the plane
perpendicular to the propagation direction and the propagation direction
2.1 Photons and their properties 31

in that plane have finite precision defined by the uncertainty relations for
photon position and momentum
h h
∆x∆px ≥ ; ∆y∆py ≥ . (8)
π π
These expressions show that the photon propagation direction in one
plane will be undefined as much as is defined its localization in the x-y
plane (the plane perpendicular to propagation direction). For a collection
of propagating photons (the light beam) with a radius w0 = ∆x (Fig.18) the
propagation direction uncertainty for photons in the beam is

h∆⃗k
∆⃗p ≥ . (9)
2π
For wavevector uncertainty of ∆⃗k = ⃗kθ, where θ is the beam divergence
angle, the uncertainty relation for the beam radius and divergence angle reads
as

λ
w0 θ ≥ . (10)
π
This expression can be interpreted as follows: the smaller is the beam
radius (w0 ), the greater is its divergence angle (θ) and vice versa. For ex­
ample, taking the laser wavelength of 500 nm and the beam radius of 100
µm then, according to expression in Eq.(10), the divergence angle of such a
λ
beam cannot be smaller than πw 0
≈ 1.6 mrad. This is a fundamental lim­
itation following from the aforementioned quantum mechanical uncertainty
principle.
Two boundary cases can be pointed out. Firstly, if we can accurately
define photon propagation direction, then we cannot define where in the x-
y plane the photon is localized. In this case the light is described as an
infinite plane wave which does not diffract. Secondly, if we can accurately
define localization of the photon, then its propagation direction is completely
undefined. In this case the light is described as a spherical wave emitted from
a point source. The real light beams represent an intermediate case between
these two boundary cases. Eq.(10) describes the minimum beam divergence
angle for a given beam radius, such a beam is called diffraction-limited.
Beam with an ideal Gaussian intensity distribution (Fig. 22) are diffraction-
limited beams.
The time-energy uncertainty is expressed as
h
∆E∆τ ≥ (11)
2π
2.1 Photons and their properties 32

and relates energy uncertainty ∆E with characteristic time ∆τ at which

changes occur in the given system, e.g. the photon is absorbed or emitted.
Having in mind that ∆E = h∆ν, this expression is equivalent to photon
time-frequency uncertainty relation

1
∆ν∆τ ≥ . (12)
2π
In the case of emission, ∆τ is directly related to the lifetime of an excited
energy level of an atom or molecule: the shorter is the lifetime of the excited
level, the greater is spread in the energy (frequency uncertainty ∆ν) of the
emitted photons or, in other words, the greater is spectral broadening of
the energy level. For example, energy levels that have long lifetime (the so-
called metastable levels) represent an opposite case and emit radiation with a
narrow spectrum. The above considerations may be applied to light pulses as
well; then parameters in expression ∆ν and ∆τ (Eq.(12)) have meanings of
the spectral bandwidth and pulse duration, respectively. Then the expression
in Eq.(12) could be rewritten in a slightly different form:

∆ν∆τ ≥ C , (13)
where C is a certain constant related to the shape of the pulse envelope (Fig.
14). For example, in the case of a Gaussian-shaped pulse C = 0.441, for
sech2 shaped pulses C = 0.315, for rectangular pulses C = 0.886, etc (Fig.
19).

Figure 19: Examples of different pulse shapes. τ0.5 marks pulse duration at
FWHM level. Note that some pulse shape models (such as rectangular) are purely
theoretical models and in practice only approximations of such shapes are possible
to achieve.

In the case of equality in Eq.(13), pulse duration exactly corresponds

to its spectral bandwidth. Such pulses are called bandwidth-limited or
2.2 Light waves 33

transform-limited. This also means that the shorter is the pulse, the
broader is its frequency spectrum and vice versa. For example, continu­
ous wave (infinite pulse duration) lasers are essentially monochromatic (their
emission spectrum corresponds to their intrinsic spectral bandwidth), while
femtosecond lasers are polychromatic, they generate a broadband frequency
spectrum. For example, in the near-IR range (λ=1 µm) the spectral band­
width of a ∆τ =1 ps pulse (FWHM level) Gaussian pulse is:
0.44
∆υ = = 4.4 · 1011 (Hz) (14)
∆τ
or

∆υλ2
∆λ = = 1.47(nm), (15)
c
whereas the spectral bandwidth of the same central wavelength ∆τ =10 fs
Gaussian pulse is ∆λ=147 nm at FWHM. In the visible range the spectrum
of a latter pulse would essentially be white light. It is important to mention
that artificial narrowing (extension) of the bandwidth-limited pulse spectrum
results in an increase (decrease) of its duration. If the left hand side of
expression in Eq.(13) is significantly greater than the right hand side, the
light pulse is long, but its frequency spectrum is broad. Such pulses are called
phase modulated (or chirped). Manipulation of the pulse phase and
frequency modulation and its control is the basis of temporal compression and
stretching, which is widely exploited in modern laser amplifiers. Compared
with other light sources, lasers are unique in their ability to generate light
pulses and beams with the smallest possible uncertainty: diffraction-limited
beams and bandwidth-limited pulses.

2.2 Light waves

The wave nature of light means that it can be described as electromagnetic
waves: alternating electric and magnetic fields propagating in space and
carrying energy. Electromagnetic wave propagation is fully described by a
set of coupled equations, called Maxwell equations:
⃗
⃗ = ⃗j + ∂D ,
∇×H (16)
∂t
⃗
⃗ = − ∂B ,
∇×E (17)
∂t
∇·D⃗ = ρ, (18)
2.2 Light waves 34

⃗ = 0,
∇·B (19)
→ → →
where H is the magnetic field strength, j is the electric current density, D
→
is the electric field flux density (electric induction), B is the magnetic field
flux density (magnetic induction), ρ is the electric charge density. × is the
curl operator
p indicating
d vector product, while · indicates scalar product and
∂ ∂ ∂
∇ = ∂x , ∂y , ∂z is the nabla operator (expressed in Cartesian coordinate
system).
These equations, named after the physicist and mathematician James
Clerk Maxwell, who published them in early forms in the 1860s, provide a
mathematical model unifying the phenomena of electricity and magnetism
and put forward the foundation of classical electrodynamics. Eq.(16) essen­
tially states the Ampere’s circuital law: electric current and time-varying
electric fields induce magnetic fields curling around them. Eq.(17) represents
the Faraday’s law of electromagnetic induction: time-varying magnetic
fields induce electric fields curling around them. The third equation (Eq.(18))
is essentially the Gauss’ law: electric charges create electric fields. The last
equation (Eq.(19)) can be interpreted as the Gauss’s law for magnetism:
magnetic fields do not diverge, they curl around. In other words, Eq.(19)
indicates that magnetic charges do not exist.
For propagation of electromagnetic waves in free space (vacuum) a more
practical equation can be derived from Maxwell equations. This equation
describes light propagation in free space and is called Helmholtz equation.
In one-dimensional case it is expressed as:

∂ 2E 1 ∂ 2E
− 2 2 =0 (20)
∂z 2 c ∂t
where E is the wave amplitude. Solutions of this equation are certain func­
tions E(z, t) = f (z ± ct), which are called fundamental waves. Helmholtz
equation is also valid for a harmonic wave:

E(z, t) = A0 sin(kz − ωt), (21)

which is described by its amplitude A0 , wavenumber k = 2π/λ, which is
inversely proportional to wavelength λ, and frequency ω which is inversely
proportional to wave period. Quantity (kz − ωt) is called the wave phase.
Superposition of harmonic waves is also a solution of the Helmholtz equa­
tion. However, summation of waves described by trigonometric functions
yields complex mathematical expressions that are inconvenient to use, there­
fore complex wave notation is usually applied which is derived using Euler’s
2.2 Light waves 35

formula:
eiα = cos α + i sin α. (22)
Using it and a derived expression:
1
sin α = [exp (iα) − exp (−iα), (23)
2i
it is convenient to express the harmonic wave in the complex form:
1
E(z, t) = E0 exp [i(kz − ωt)] + c.c., (24)
2
where E0 is the complex amplitude of the wave and c. c. marks the complex
conjugate part of the equation: if we analyze wave propagation in free space,
the complex conjugate part of expression in Eq.(24) can be ignored.
There are several approximations of the light waves that are important
when analyzing propagation of a laser beam. Firstly, for the sake of sim­
plicity, let us assume that light waves are infinite in time, i.e. they have
only a single frequency component (are monochromatic). Plane waves are
waves whose phase at fixed moments of time is stationary in the entire plane
perpendicular to the propagation direction (Fig. 20 a).

(a) (b)

Figure 20: Illustration of plane (a) and spherical waves (b).

It is written as
E(z, t) = E0 exp [i(⃗k⃗r ± ωt)], (25)
where ⃗r = (x, y, z), ⃗k = (kx , ky , kz ) and ⃗k⃗r = const. Such wave repeats itself
in the ⃗k direction after a distance equal to λ. This wave is infinite in space
and its k vectors are unidirectional, so such wave does not diffract. In reality
plane waves do not exist, but the concept of plane waves is often applied as
a convenient approximation when diffraction of light can be neglected.
2.2 Light waves 36

Another wave optics approximation is a spherical wave (Fig. 20 b). In

spherical coordinate system these waves are expressed as
E0
E(r, t) = exp [i(kz ± ωt)]. (26)
r
The amplitude of a spherical wave decreases during propagation as a
function of 1/r and when r becomes very large (approaches infinity), the
spherical wave becomes a plane wave (see Fig. 21). Interestingly, spherical
waves strongly diffract away from the source.

Figure 21: Principal view of spherical wave wavefront evolution during propaga­
tion in free space: far from the source the wavefront can be approximated by a
paraboloidal function and as it propagates further the wavefront becomes very
similar to plane.

Bearing in mind the above, it is clear that plane and spherical waves
represent the two opposite cases considering angular and spatial energy dis­
tributions. A plane wave does not have angular spread and its energy is
evenly distributed through the entire space, whereas a spherical wave comes
from a point source and spreads at every direction with its amplitude rapidly
decreasing at a given direction. The light beams represent an intermediate
case between plane and spherical waves. They possess features of both, plane
and spherical waves. The wave nature of light rules out the possibility for
light to propagate in free space without angular spread, but in the case of
light beams, the energy can be sufficiently well (as much as the uncertainty
principle allows) localized along the propagation direction.
The light waves whose wavefront normal vectors make a small angle with
the propagation axis are called paraxial. Such waves are also solutions of the
Helmholtz equation. One of the most important approximations of paraxial
waves is the Gaussian beam – the light beam whose spatial intensity dis­
tribution is expressed by a Gaussian function (Fig.22).
2.3 Gaussian beams 37

Figure 22: Intensity distribution of a Gaussian beam.

2.3 Gaussian beams

The Gaussian beam is of major importance in laser physics. First of all,
as mentioned before, Gaussian beam has the smallest uncertainty of photon
position and propagation direction, therefore, the energy in a Gaussian beam
is localized very well along the cylinder surrounding the propagation axis.
This implies that the angular spread of an ideal Gaussian beam is minimal
for a given beam radius in the waist, so in other words, the Gaussian beam
has very high quality and is diffraction-limited; it can be focused to a spot
which size is limited only by diffraction. Secondly, the intensity distribution
of a Gaussian beam in the plane perpendicular to the propagation axis is
described by a Gaussian function, which has a maximum located on the
propagation axis. Thirdly, all lasers generate beams with a Gaussian intensity
distribution (the reason for this will be explained later).
We will now discuss the main parameters of the Gaussian beam and how
they evolve during propagation in free space. The light intensity distribution
of an ideal Gaussian beam (Fig.22) is described as

2
w0 2r2
I(r, z) = I0 exp − , (27)
w(z) w2 (z)
where I0 = |E02 |, E0 is the amplitude of electric
l field, w0 is the minimum
beam radius at the waist (see Fig.23), r = x + y 2 . For every propagation
2

distance z the transverse (x-y plane) intensity distribution of the beam is a

Gaussian function with a maximum intensity at z = 0 and monotonically
decreasing when going away from beam symmetry axis. During propagation
(when z increases) the beam intensity changes as:
2.3 Gaussian beams 38

I0
I= 2 , (28)
1 + zz2
R

where zR is the Rayleigh length. Let us discuss the meaning of zR . The

minimum Gaussian beam radius w0 is at the beam waist (z = 0) and the
beam diameter is 2w0 . During propagation (away form the waist)√ the beam
radius continuously increases and at z = zR it becomes w(zR ) = 2w0 . Then
we can define the Rayleigh length as:

πw02
zR = . (29)
λ
√
It defines the distance at which beam radius increases by 2 times. The
minimum Gaussian beam wavefront radius of curvature is at z = ±zR , thus
the Rayleigh length is often called the diffraction
l length. On the other
hand, the beam radius can be expressed as w0 = λzR /π.
The entire power of a Gaussian beam can be determined by integrating
the following expression:
∞
1
P = I(r, z)2πrdr = I0 πw02 . (30)
0 2
The beam radius can be defined at different intensity levels (see Fig.12).
Usually, the beam radius is defined at 1/e2 ≈ 0.135 intensity level, as shown
in Fig.22. Eq.(30) is valid for such beam radius definition. For the electric
field amplitude, this would correspond to 1/e ≈ 0.368 level. If the integration
limit in Eq.(30) is changed from ∞ to w(z), such definition of w(z) corres­
ponds to concentration of 86% of the whole beam power, while 99% of the
beam power fits into 1.5w(z) intensity level. In general, an ideal Gaussian
beam is infinite, but only a negligible part of its power is contained in the
far periphery. In practice, Gaussian beam radius can also be defined at a
half maximum intensity level (HWHM – half width at a half maximum) and
the beam diameter is determined at the same intensity level (FWHM – full
width at a half maximum). Beam radii definitions at FWHM level and at
1/e2 level are related as:
wF2 W HM
exp −2 2
= 0.5. (31)
w1/e 2

Simplifying this expression yields the mathematical relation for different

beam radii definitions:
ln 2
wF W HM = w1/e2 , (32)
2
2.3 Gaussian beams 39

√
which for the diameter of the beam is 2wF W HM = 2 ln 2w1/e2 , thus a factor
of 2 ln 2 is necessary in the Gaussian beam power expression if the beam
radius is defined at FWHM level.
The radius (w(z)) and wavefront curvature (R(z)) of a Gaussian beam
change according to expressions (Fig.23):

z2 1/2
w(z) = w0 1 + , (33)
zR2
and

zR2
R(z) = z 1 + , (34)
z2

Figure 23: Gaussian beam parameters and their evolution during propagation.

Let us determine how radius of a Gaussian beam changes far from the
beam waist. When z ≫ zR in Eq.(33) ”1” can be omitted and the beam
radius increases linearly with z as depicted in Fig.24. This linear dependence
can be expressed as
w0
w(z) ≈ z = θz (35)
zR
where θ is the divergence angle:
w0 λ
θ= = . (36)
zR πw0
2.3 Gaussian beams 40

Figure 24: Definition of divergence angle of a Gaussian beam.

The divergence angle basically reflects the uncertainty principle which

was discussed in the first section of this chapter. In addition, the divergence
angle depends not only on the radius of beam waist (the divergence of beam
with a small waist is large, and vice versa, so for the latter case a plane
wave approximation could be applied) but also on the wavelength. In that
regard, the ultraviolet Gaussian beams have lower divergence than infrared
beams of the same waist size. For radio waves which could be considered as
a boundary case, the concept of beams is meaningless since they have huge
divergence and are essentially spherical waves.
Sometimes the Gaussian beam is characterized by a depth of focus (con­
focal parameter). The main idea is that the beam spreads equally on both
sides with respect to its waist, so the confocal parameter is b = 2zR =
2πw02 /λ. For example, for λ = 633 nm wavelength (He-Ne laser wavelength)
and beam diameter of 2w0 = 2 cm, b = 1 km, whereas for the beam diameter
of 2w0 = 20 µm, b = 1 mm.
Now let us discuss the phase factor of a Gaussian beam which is expressed
as:

kr2
ϕ(r, z) = kz − ξ(z) + . (37)
2R(z)
When considering the phase of the beam we usually use the concept of
wavefront. The wavefront is a surface of equal phase with respect to the
beam propagation direction. The wavefront of a plane wave is flat (the plane
perpendicular to the propagation direction) and the wavefront of a spher­
ical wave is spherical. As follows from expression in Eq.(37), the Gaussian
beam wavefront is neither flat, nor spherical. The phase of the beam on the
propagation axis (r = 0) is expressed as ϕ(0, z) = kz − ξ(z). The first term
describes a plane wave. So at the waist (z = 0) the Gaussian beam is a plane
wave, i.e. its wavefront is flat and radius of curvature R(z) is infinite. The
2.3 Gaussian beams 41

second term describes the phase shift of the on-axis beam part compared to
either plane or the spherical wave. When z = ±zR , the phase shift is ±π/4
and approaches ±π/2 when z → ±∞. This means that during propagation
of the Gaussian beam through infinitely long distance the maximum accumu­
lated axial phase shift is π. This is called the Gouy effect. The third term in
Eq.(37) describes the phase shift for off-axis points compared with the phase
of the on-axis points. In other words, it shows the degree of wavefront bend­
ing. It can be shown that a surface of equal phase is parabolic with a radius
of curvature R. The evolution of the radius of curvature for a Gaussian beam
during propagation is described by Eq.(34), and is graphically illustrated in
Fig.25. It is clear that the minimum radius of curvature of a Gaussian beam
is at z = ±zR : R = ±2zR , i.e., the wavefront here is maximally curved. The
radius of curvature both sides from zR is only increasing with propagation.

Figure 25: Evolution of a Gaussian beam radius of curvature during propagation.

At the beam waist R = ∞, so at this particular location the Gaussian beam

can be approximated as a plane wave as illustrated in Fig.26. The same
approximation can also formally be applied at z = ∞, however, the axial
phase shift due to Gouy effect has to be taken into account. Far from the
beam waist of a Gaussian beam wavefront is almost spherical.
2.4 Coherence 42

Figure 26: Comparison of wavefronts:(a) plane waves, (b) spherical waves and (c)
Gaussian beam.

2.4 Coherence
Coherence is a distinctive feature of laser radiation. In general, the term
coherence is used to describe phase correlation between two monochro­
matic waves. Waves with correlated phases are called coherent and waves
with uncorrelated (random) phases are called incoherent. Two types of
coherence can be distinguished:

• Temporal (longitudinal) coherence is the measure of the correla­

tion of light wave’s phase at different points in time along the direction
of propagation and it describes how monochromatic the light source is
(Fig.27).

• Spatial (transverse) coherence is the measure of the correlation of

a light wave’s phase at different points transverse in the plane perpen­
dicular to the direction of propagation. It describes how uniform the
phase of a wavefront is (Fig.27). In other words, it shows the ability
for two points in space, y1 and y2 , in the extent of a wave to interfere,
when averaged over time.
2.4 Coherence 43

Figure 27: Principal illustration of temporal and spatial coherence of light waves.
tc and lt mark temporal coherence time and spatial coherence length, respectively.
In the spatial coherence part interference intensity pattern is depicted in the cases
of perfectly coherent, partially coherent and incoherent light source.

Let us first discuss temporal coherence in more detail. From a practical

point of view, temporal coherence characterizes how well a wave can interfere
with itself at a different moments of time. When we observe interference
pattern using an interferometer, greater temporal coherence corresponds to
sharper contrast between fringe minima and maxima. If we increase the
length difference between interferometer arms, the interference pattern blurs
and eventually disappears (Fig.28). Let us discuss why.

Figure 28: Example of interference fringes observed from a monochromatic light

source using a Michelson interferometer. d is the optical path difference between
Michelson interferometer arms.
2.4 Coherence 44

An ideal monochromatic light source is temporally coherent, but it is clear

that in the reality strictly monochromatic light sources do not exist. Light
sources that we are called ”monochromatic” in reality radiate a sequence of
harmonic waves with finite lengths Fig.(29), which are interrupted by phase
jumps due to various processes occurring in a system of excited atoms.

Figure 29: A sequence of harmonic waves with random length and phase.

Such a light source is described by an average duration of a harmonic

wave tc (with a constant phase), which is an average of τ1 , τ2 , τ3 , τ4 , .... Con­
sequently, any light source can be described with the duration tc which is
called the coherence time. The main parameter which determines the co­
herence time of a given source is the natural (intrinsic) spectral linewidth,
i.e., ∆ω = 2π/tc or ∆ν = 1/tc . It is obvious that a continuous phase and
infinitely monochromatic wave would have an infinite coherence time, and
vice versa: a radiation with infinitely broad spectral bandwidth would have
an infinitely short coherence time, approaching the delta function. The same
applies to the coherence length which is defined as lc = ctc :

c λ2
lc = ≈ , (38)
∆ν ∆λ
which essentially reflects the boundary case of the uncertainty principle; here
∆λ is the natural spectral linewidth of a given light source. Having this in
mind, we can now explain what is depicted in Fig.28. When optical path
difference between interferometer arms (d) approaches the temporal coher­
ence length (lc ), the contrast of interference fringes drops and when d exceeds
temporal coherence length, the interference pattern disappears. This simply
reflects the fact that there are no ideal monochromatic light sources, there­
fore any real light source has a finite coherence time and temporal coherence
length.
The temporal coherence of several representative light sources is com­
pared in Table 1. For any CW light source (including CW lasers) the tem­
poral coherence in this case describes the monochromaticity of such light
2.5 Spatial coherence 45

Table 1: Comparison of spectral linewidths, coherence times and coherence lengths

of some representative light sources.

Source Spectral linewidth, Hz tc , s lc , mm

The Sun 6 × 10 14
10−15 3 × 10−4
Blue LED 6 × 10 13
3 × 10−14 9 × 10−3
Filter 1.2 × 1012 5.2 × 10−13 0.2
Flashlamp 109
6.2 × 10−10 190
Monochromator 108 6.2 × 10−9 1900
CW laser 105 6.2 × 10−6 106
fs laser 5 × 10 13
3.5 × 10−14 10−2

source. It is clear that CW lasers in regard to their monochromaticity by far

surpass any non-laser light source.
For pulsed lasers, the coherence time is interpreted differently. Such lasers
emit light in bursts or portions, which are called pulses, whose durations
depend on the methods how they are generated. The spectrum of a light
pulse represents a collection of many monochromatic waves with different
frequencies and if all the waves are in phase, the pulse is bandwidth limited
and its duration equals to a coherence time, which is inversely proportional to
its spectral width. However, if the pulse is not bandwidth-limited (chirped),
its duration its always longer than its coherence time.
For example, a femtosecond laser emitting 100 fs bandwidth-limited pulse
has temporal coherence time of tc =100 fs and temporal coherence length lc =
ctc =30 µm. When observing interference of such laser pulse with a Michelson
interfereometer (Fig.28) optical path difference between interferometer arms
cannot exceed 30 µm.

2.5 Spatial coherence

In the previous section we considered the phase correlation at different mo­
ments of time along the propagation direction. Therefore, temporal coher­
ence is also called the longitudinal coherence. Now let us discuss the co­
herence between two points of the same light source (or beam) separated in
space. As already mentioned, the measure of the correlation of a light wave’s
phase at different points transverse to the direction of propagation is called
spatial coherence (Fig.27 right). The spatial (transverse) coherence
length is usually defined as a maximum distance lt between the two points
where the interference can still be observed, as schematically depicted in
Fig.30.
2.5 Spatial coherence 46

Figure 30: Definition of spatial coherence of light.

Then the coherence length is defined as

rλ λ
lt = = , (39)
s θ
where r is the distance from the light source, s is the distance between the
points of the source and θ is the angle at which these points are visible.
According to this definition, an ideal plane waves and point light sources
are fully spatially coherent. Any other incoherent light source can be made
coherent by turning it into a point light source, however, this is possible only
at the cost of its radiation power..
Let us discuss the meaning of spatial coherence in the case of a Gaussian
beam. Comparing expressions in Eq.(36) and Eq.(39), it is clear that they
match with precision of a constant, so the radius of a Gaussian beam waist
corresponds to the length of spatial coherence.
To characterize a real Gaussian beam, another parameter called M 2 is
also introduced. This parameter indicates the difference between the diver­
gence angles of real and ideal beams with the same beam waist. Moreover,
this parameter is often used as a beam ”quality” parameter. The term ”qual­
ity” is more often defined as a measure for how well the beam can be focused
(Fig.31) and it is quantified as either M 2 parameter:

πw0 θ
M2 = , (40)
λ
or beam parameter product (BPP):

BP P = w0 · θ . (41)
When we calculate diameter of a focused beam (df ) we have to take into
account the M 2 parameter:

4λf
df = M 2 , (42)
πd0
2.5 Spatial coherence 47

here f is lens focal length, d0 is the initial (unfocused) beam diameter. M 2 =1

for diffraction-limited Gaussian beam, so the M 2 factor is a quantitative
measure how much the beam is different from a diffraction-limited Gaussian
beam.

Figure 31: Gaussian beam size evolution when it is focused with a lens. The
red contours depict laser beam intensity distribution and d1/e2 denotes laser beam
diameter at 1/e2 of maximum intensity level.

Using the light interference terminology, the Gaussian beam can be viewed
as a result of interfering multiple plane waves propagating at different dir­
ections. Since these waves are phased in a certain way, due to interference
they create a specific spatial modulation which is observed as the light beam.
Thus, laser light sources are also unique with regard to spatial coherence as
laser beams are spatially coherent and carry all the power emitted by the
laser source.
Concluding the coherence issues, the concepts of temporal and spatial
coherence could be merged by defining the volume (three-dimensional) co­
herence:

lv = lc lt2 , (43)
which describes the spatio-temporal coherence and directly indicates the
volume where the energy of light is localized. The concept of volume (spatio­
temporal) coherence applies to pulsed lasers.
2.5 Spatial coherence 48

Summary of Chapter 2
• Light is described either as electromagnetic waves or as particles
(photons). Each concept attributes to different properties of laser
radiation.

• Photons are quantum particles and are subject to uncertainty

principle, which poses fundamental limitations to relevant para­
meters of laser radiation.

• Diffraction is related to the photon position-momentum uncer­

tainty which in turn is the fundamental the reason for laser beam
divergence (spreading in space during propagation). Laser beams
with minimum radius for a given divergence angle are called
diffraction-limited.

• The time-energy uncertainty relates the laser pulse duration with

its spectral bandwidth. The shorter is laser pulse, the broader is
its frequency spectrum, and vice versa. Laser pulses that have
minimum duration for a given bandwidth are called bandwidth-
limited (transform-limited).

• Laser produces a low divergence light beam which has an in­

tensity distribution (in the plane perpendicular to propagation
direction) that is described by Gaussian function.

• Laser emits spatially and temporally coherent light.

3 Optical resonators 49

3 Optical resonators
An optical resonator (optical cavity) is an extremely important component
of any laser. The simplest optical resonator consists of two reflecting mirrors,
which play several relevant roles in laser operation:

• Energy accumulation and feedback: the laser resonator accumu­

lates electromagnetic energy by confining radiation within a certain
volume and provides a feedback function by returning radiation to the
active element, where the process of stimulated emission takes place.

• Formation of transverse modes: the laser resonator forms a dir­

ectional radiation in the shape of a light beam, which has with high
spatial quality and high spatial coherence and which has a particular
intensity distribution in the transverse plane that is called transverse
mode.

• Formation of longitudinal modes: the laser resonator has its in­

trinsic temporal frequencies which satisfy the standing wave condition
(the longitudinal modes) and which determine the spectral quality of
the radiation and ensure high temporal coherence.

In this chapter we will discuss the role of resonators in lasers: light propaga­
tion analysis in optical resonators, resonator stability criteria, transverse and
longitudinal modes and energy loss mechanisms in optical resonators.

3.1 Light rays and ray (or ABCD) matrices

Light propagation in various optical elements and resonators in the paraxial
approximation can be successfully described using an approximation of geo­
metrical optics: light rays. A light ray is defined as a straight line per­
pendicular to the wavefront. Therefore, if we know how rays behave when
propagating through a certain optical medium or element, we also know how
light waves and beams behave in such medium or element. Furthermore,
light ray approximation can be applied not only for light propagation in iso­
tropic optical media but also in optical elements (lenses, mirrors, etc.) and
for light propagation in media with refractive index gradient, absorption,
amplification or scattering.
Propagation of light rays (as well as their refraction or reflection) in op­
tical elements or media can be described using simple 2 × 2 matrices which
are called ray or ABCD matrices. A light ray can be completely described
by two parameters: position x and inclination angle θ with respect to the
3.1 Light rays and ray (or ABCD) matrices 50

Figure 32: The main parameters defining propagation of a light ray.

optical axis, as depicted in Fig.32. These two parameters define the ray vec­
tor (x, θ). An effect of any optical element is equivalent to multiplication of
the ray matrix defining the optical element by ray vector, resulting in new
values of the parameters characterizing the ray:
[ [ [
xout A B xin
= (44)
θout C D θin

where indices in and out mark vector parameters of the initial (incident) and
final (outgoing) rays respectively. We will now discuss the principle how ray
matrices for various optical elements are constructed and how their A, B, C
and D values are determined. Assuming that inclination angle of a ray to the
optical axis is small (which is essentially the condition of paraxial approxim­
ation as was already stated), the ray parameter changes are described by a
two equation system:

∂xout ∂xout
xout = xin + θin
∂xin ∂θin
∂θout ∂θout
θout = xin + θin . (45)
∂xin ∂θin
These equations are equivalent to expression in Eq.(44), so matrix ele­
ments can be expressed as partial derivatives where
∂xout ∂xout ∂θout ∂θout
A= , B= , C= , D= . (46)
∂xin ∂θin ∂xin ∂θin
Matrix elements A and D can be understood as spatial and angular mag­
nification, respectively. With accordance to these expressions, let us write
ABCD matrices for several optical media and elements which are important
for optical resonators. To start with, let us determine the ABCD matrix for
3.1 Light rays and ray (or ABCD) matrices 51

Figure 33: Propagation of a light ray in free space.

free space propagation MS . Light propagation in free space is depicted in

Fig.33.
The system of equations (Eq.(45)) for light ray propagation in free space
can be written as:

xout = xin + z θin

θout = θin . (47)

In matrix form the system of equations transforms into:

[ [ [
xout 1 z xin
= (48)
θout 0 1 θin
or
[
1 z
MS = . (49)
0 1
Now let us determine the ABCD matrix for intersection of two optical
media with different refractive indices n1 ir n2 . Geometrically the situation
is depicted in Fig.34.

Figure 34: Propagation of a light ray through the intersection of two media.
3.1 Light rays and ray (or ABCD) matrices 52

At the intersection of two media, a distance of the light ray from the
optical axis remains unchanged: xout = xin , whereas the change of the in­
clination angle can be calculated from the Snell’s law: θout = nn12 θin . Con­
sequently, the ray matrix for the intersection of two media with different
refractive indices is
[
1 0
MI = . (50)
0 n1 /n2
In analogy, the ray matrix for the intersection of two optical media with
different refractive indices and curved (with R radius of curvature) intersec­
tion surface can be written as:
[
1 0
MCI = n1 /n2 −1 . (51)
R
n1 /n2
In that case, if we assume a lens as two curved surfaces, the ray matrix
for a lens can be derived as a product of two ray matrices of curved surfaces:
[
1 0
ML = MCI1 MCI2 = . (52)
− f1 1
Here we also assumed that the refractive index of the surrounding medium
is n1 = 1 and the refractive index of a lens material is n2 = n. Then f is the
focal distance of the lens and is expressed as:

1 1 1
= (n − 1) − , (53)
f R2 R1
where R1 and R2 are the radii of curvature of the lens’ surfaces. This expres­
sion is also known as the Lens Maker’s formula. Notice that if f > 0 the
lens is convex and if f < 0 it is concave. Based on a similar consideration, we
can determine ABCD matrix for a curved mirror, whose center of curvature
is located on the optical axis:
[
1 0
MCM = , (54)
− R2 1
where R is the mirror’s radius of curvature. Comparing the expressions for
ML and MCM it is clear that a curved mirror focuses rays as a lens with a
focal length of fv = R/2. Therefore a mirror with R > 0 is regarded as
concave and a mirror with R < 0 as convex. In the case of a plane mirror,
the same ray matrix can be used by setting R = ∞, i.e.:
[
1 0
MPM = . (55)
0 1
3.2 Laser resonator stability 53

Concluding the results of this section, we can note that an ABCD matrix
can be written for any optical element no matter how complex it is. This
means that light ray propagation can be modelled in sophisticated optical
systems consisting of various optical elements whose ABCD matrices are
known. Such method is called ray tracing. Moreover, ABCD matrix form­
alism can be directly applied to a Gaussian beam assuming that the optical
element transforms its complex propagation parameter in the following way:
Aqin + B
qout = , (56)
Cqin + D

where qin and qout are incident and passing (through the optical element)
complex propagation parameters of the Gaussian beam, respectively. The
complex propagation parameter is expressed via two real functions: beam
radius w(z) and wavefront curvature R(z), which fully ascribe the Gaussian
beam:
1 1 λ
= −i 2 . (57)
q(z) R(z) πw (z)
From this we can determine how relevant parameters (i.e., w(z) and R(z))
of the Gaussian beam change during beam propagation in various optical
systems.

3.2 Laser resonator stability

Let us discuss light propagation in an optical resonator consisting of two
curved mirrors with radii of curvature R1 and R2 separated by distance l
as shown in Fig.35. The readers may notice that the depicted resonator
does not have any gain medium inside: such resonators are called ”cold” or
empty. We can say in advance that such simplification does not alter light
propagation analysis or any important results that we will arrive at in this
section.
3.2 Laser resonator stability 54

Figure 35: Laser beam in an empty optical resonator consisting of two curved
mirrors.

It is obvious that in order for the laser to work, light should not escape
the resonator: after each reflection the beam must maintain a stable size fit­
ting into the resonator mirror aperture which is finite. Light will be confined
in such optical system only if it fully reproduces its spatial distribution after
each round-trip between the resonator mirrors. This is called beam conver­
gence condition, which implies that only the laser beams with a specific
spatial distributions of electric field (intensity) are able to do so. These are
called spatial (transverse) modes of the resonator. On the other hand,
beam convergence condition poses strict requirements on the resonator para­
meters and geometry: mirrors radii of curvature R1 and R2 and resonator
length l. By applying ABCD law for a Gaussian beam that fully reproduces
itself after each resonator round-trip (qin = qout ):

Aq + B
q= , (58)
Cq + D
where A, B, C, D values are derived from Mresonator = MCM 2 MS MCM 1 MS
and skipping mathematics for the sake of simplicity we can derive laser
resonator stability condition:

l l
0≤ 1− 1− ≤ 1. (59)
R1 R2

Here R1 and R2 are resoantor mirros radii of curvature and l is the length of
a resonator. gi = 1 − Rli are called g parameters of the particular resonator
mirror. This inequality is depicted graphically in Fig.36 which is called res­
onator stability diagram. Blue area in the diagram shows the space of g
parameter values, where resonator is stable.
3.2 Laser resonator stability 55

Figure 36: Resonator stability diagram, where stability zone is shown by blue
area, and several resonator types corresponding to different points in the stability
diagram. Image adapted from Wikimedia under CC BY-SA 3.0 license.

The stability diagram also shows that distinct resonator arrangements

(which have different l and R ratio – Fig.37) correspond to different g1 and
g2 parameter values in the diagram. It is important to understand that when
g parameter values are at the boundary of stability zone, such resonator is
very sensitive to any perturbations due to mechanical vibrations, temperat­
ure instabilities, etc. Therefore, concentric, confocal and plane resonators
are very sensitive to external perturbations. Even very small changes of the
environment make the resonator unstable, i.e., its diffraction losses increase
significantly. Such resonator configurations are also critically sensitive to
mirror alignment with respect to the optical axis. Therefore, in practice,
asymmetrical resonators are usually used, e.g. consisting of plane and con­
cave mirrors (hemispherical resonator). Such resonators are in the middle
of the stability zone in the stability diagram, therefore, they are much less
sensitive to external perturbations. On the other hand, we considered only
”cold” (empty) resonators which do not have any other optical elements in­
side. In a real laser, apart from resonator mirrors there is also an active
element whose operation in some cases can significantly change resonator
stability conditions. For example, active elements of solid state lasers often
have thermal lens effect, which must be taken into account when designing
3.2 Laser resonator stability 56

Figure 37: Types of laser resonators.

resonator geometry.
Therefore, summarizing both facts, we can claim that Gaussian beam is
a spatial mode of a spherical mirror resonator. Generally, any resonator can
be regarded as an approximation of a spherical resonator, so the Gaussian
beam is a spatial mode of any optical resonator, and consequently, all lasers
produce radiation in the form of a Gaussian beam. Furthermore, the
M 2 factor introduced in the previous chapter quantitatively indicates how
much a beam generated in a certain resonator differs from an ideal Gaussian
beam generated in a spherical mirror resonator.
The current discussion may raise a question whether a laser with an un­
stable resonator can operate at all? It is possible indeed, however, lasers
operating with unstable resonators typically have high gain which balances
large diffraction losses. In an unstable resonator light will be ejected after a
certain number of round-trips and output coupling is such lasers is also differ­
ent: light leakage due to diffraction is often taken as the useful laser output
(Fig,38). Unstable resonators also have transverse modes, but their intensity
distributions are very complex and beyond the scope of this textbook. In­
terestingly, unstable resonators are less sensitive to alignment instabilities.
3.3 Hermite-Gaussian beams and Laguerre Gaussian beams 57

Figure 38: Light propagation in an unstable resonator. Output coupling in this

case is at a hard-edge mirror.

3.3 Hermite-Gaussian beams and Laguerre Gaussian

beams
When analyzing resonator stability conditions we have showed that Gaussian
beam is an intrinsic spatial mode of a spherical mirror resonator (and in
general, of any other resonator). The complex amplitude of electric field of
a Gaussian beam can be expressed in a very general form as:
2 2
E(x, y, z) = E0 × exp − xw2+y
w0
w(z) (z)
×
, (60)
[
k(x2 +y 2 )
exp [−i(kz) + iξ(z)] × exp −i 2R(z)

where ξ(z) = tan−1 (z/zR ) and R(z) is the wavefront radius of curvature, E0 is
the amplitude of electric field, w0 is the minimum beam radius (see Fig.23).
The first line in this expression describes the amplitude evolution during
propagation in the z direction, whereas the second line in the expression
describes the phase evolution.
Lasers with very stable resonators may produce beams with more complex
electric field (intensity) distributions, which reproduce themselves after each
round-trip, and satisfy the convergence condition. These beams are called
higher order spatial (transverse) resonator modes.
Two cases of higher order modes can be distinguished. In stable optical
resonators with no cylindrical symmetry higher order transverse modes
can be ascribed as Hermite-Gaussian beams, whose their electric field
amplitude is expressed as
3.3 Hermite-Gaussian beams and Laguerre Gaussian beams 58

�√ � �√ � [ 2
2x 2y w0 x + y2
E(x, y, z) = E0 Hm Hn × exp − 2 × (61)
w(z) w(z) w(z) w (z)
k(x2 + y 2 )
[
exp [−i (kz − (1 + m + n)ξ(z))] × exp −i .
2R(z)

where quantities E0 , w0 , w(z), R(z) and ξ(z) are defined in the same manner
as in the case of a Gaussian beam, while functions Hm and Hn are called
Hermite polynomials, whose indices m and n mark the order of the polyno­
mial. Hermite polynomials are defined as

H0 (u) = 1 (62)
H1 (u) = u (63)
H2 (u) = 2u2 − 1 etc. (64)

In the case of the lowest order Hermite polynomial m, n = 0, we have an

expression for the usual Gaussian beam Eq.(60), so the lowest order Hermite-
Gaussian beam is simply a Gaussian beam. Comparing the expressions in
Eq.(60) and in Eq.(56) we can notice that Hermite-Gaussian beam has all
the relevant requisites of a Gaussian beam, but there are several key differ­
ences. The first multiplier of the product in expression Eq.(56) shows that
the Hermite-Gaussian beam amplitude at z = 0 has interesting variations in
the transverse plane (i.e., the plane perpendicular to the beam propagation
direction) and values of m and n indicate how many times the field amp­
litude is equal to 0 on x and y axes, respectively (see Fig.39). The second
multiplier of the product is identical to that of a Gaussian beam describing
the evolution of field amplitude along the z axis: radius of the beam increases
further away from the waist. The third multiplier describes the evolution of
longitudinal phase: it is clear that the wavefront of Hermite-Gaussian beam
is delayed on the axis with respect to spherical or plane wave. The delay is
greater for higher order of the Hermite polynomials. The last multiplier de­
scribes the transverse phase (wavefront radius of curvature) and is identical
to the term describing Gaussian beam wavefront.
Examples of Hermite-Gaussian beam intensity distributions are illus­
trated in Fig.39. Hermite-Gaussian beams are also the intrinsic spatial modes
of a resonator. The spatial modes of a resonator are often called Transverse
Electromagnetic Modes and abbreviated as TEMnm . The lowest index
TEM mode is TEM00 and represents the Gaussian beam. The number of the
amplitude peaks in higher order spatial resonator modes can be empirically
evaluated as (m + 1)(n + 1).
3.3 Hermite-Gaussian beams and Laguerre Gaussian beams 59

Figure 39: Intensity distributions of various Hermite-Gaussian modes.

It is useful to compare Gaussian and higher order spatial resonator mode

parameters: beam radius and divergence. These are expressed as

√
wmn = 2m + n + 1 w00
√
θmn = 2m + n + 1 θ00 , (65)

where index 00 stands for the Gaussian beam (TEM00 mode). Higher order
spatial resonator modes, compared to a Gaussian beam, have larger beam
radius and larger angular divergence. This tells us that such laser beam
quality is lower than that of a Gaussian beam and that higher order modes
can only be excited in a very stable optical resonator where diffraction losses
are small. On the other hand, the laser output with such a complex intensity
distribution is rarely used in practice, thus usually higher order mode gener­
ation is suppressed: mitigated by specifically inducing losses (e.g., diffraction
losses) to force laser resonator generate the TEM00 mode.
In stable optical resonators with cylindrical symmetry higher order
transverse modes can be ascribed as Laguerre-Gaussian beams and their
3.3 Hermite-Gaussian beams and Laguerre Gaussian beams 60

electric field amplitude is expressed as

l
2ρ2 ρ2
[ [ [
w0 ρ l
E(ρ, φ, z) = E0 Lm × exp − × (66)
w(z) w(z) w2 (z) w2 (z)
kρ2
[
× exp [−ikz − ilφ + i(l + 2m + 1)ξ(z)] × exp −i ,
2R(z)

where ρ, φ and z are the coordinates (in cylindrical coordinate system) and
Llm is the generalized Laguerre polynomial with indices m and l. These
polynomials are the solutions of the Laguerre differential equation. They are
defined via recurrence relation with the first ones being:

Ll0 (x) = 1 (67)

Ll1 (x) = 1 + l − x (68)

Fig.40 depicts the intensity distributions of Laguerre-Gaussian modes in res­

onators with cylindrical symmetry. As in the previous case, the lowest in­
dex TEM00 and represents the Gaussian beam. As in the case of Hermite-
Gaussian beams, the Laguerre-Gaussian beam has lower quality than that of
a Gaussian beam and such higher order Laguerre-Gaussian modes can only
be excited in a very stable optical resonator.

Figure 40: Intensity distributions of Laguerre-Gaussian modes. Indices mark the

TEMml mode index in cylindrical coordinate system.

A specific and particularly interesting case seen in Fig.40 is when TEMml

mode indices are 01*. This is a special case of Laguerre-Gaussian beam when
3.4 Resonance frequencies 61

m=0 and rotational phase factor (exp(+iφl)) is complex conjugate, which is

called an optical vortex:
l
2ρ2 ρ2
[ [ [
w0 ρ l
E(ρ, φ, z) = E0 L0 × exp − × (69)
w(z) w(z) w2 (z) w2 (z)
kρ2
[
× exp [−ikz + ilφ + i(l + 1)ξ(z)] × exp −i
2R(z)

Its phase front has a helical trajectory with an undefined phase and in­
tensity minimum at the center of the beam (Fig.41).

Figure 41: Intensity distribution (left) and wavefront (right) of an optical vortex.
Dashed black line marks phase front rotation pattern.

Mode conversion in optical resonators is one way to obtain such peculiar

laser beams which have several specific and unique applications. However,
in depth analysis of optical vortices is beyond the scope of this textbook.

3.4 Resonance frequencies

Up to now we discussed spatial resonator modes which define the intensity
distribution of radiation in the plane perpendicular to beam propagation
direction. It was also assumed that radiation is monochromatic. However,
laser resonator can sustain many temporal frequencies (Fig.42) which are
defined by the requirement that phase change ∆φ of a light wave after one
round-trip must be multiple of 2π:

∆φ = qπ , (70)

where q is an integer number called longitudinal mode index.

3.4 Resonance frequencies 62

Figure 42: Standing waves in plane parallel resonator and the corresponding
longitudinal mode index q.

In other words, Eq.(70) represents a standing wave condition and fre­

quencies that satisfy this condition are called resonance frequencies or lon­
gitudinal resonator modes.
Let us determine how different transverse and longitudinal modes fit to­
gether, i.e., determine the resonance frequencies of a given spatial resonator
mode. Assume a spherical mirror resonator where mirrors are placed at z1
and z2 . The resonance condition for spatial mode with mn indices can be
written as a difference of longitudinal phases provided that during reflection
from a mirror the phase does not change:

ϕmn (z2 ) − ϕmn (z1 ) = qπ, (71)

where ϕmn (z) is the longitudinal phase of Hermite-Gaussian beam:
z
ϕmn (z) = kz − (m + n + 1) arctan . (72)
zR
Inserting the latter expression into Eq.(71) and defining the resonator
length as l = z2 − z1 , we get

z2 z1
kq l − (m + n + 1) arctan − arctan = qπ. (73)
zR zR
Setting m and n constant, the former expression can be rewritten as
π
kq+1 − kq = , (74)
l
3.4 Resonance frequencies 63

and expressing k = 2πνn0 /c, where n0 is the refractive index of the medium
filling the resonator, and transferring into frequency domain we get

c
νq+1 − νq = . (75)
2n0 l
This frequency difference is called the intermodal distance (Fig.43),
which is inversely proportional to resonator length. Assuming that m and
n are equal to 0 (the case of a Gaussian beam), the intermodal distance is
independent of resonator configuration.

Figure 43: Intermodal distance in the case of a TEM00 mode (Gaussian beam).

Without going into mathematical details, it can be summarized that in

a more complex case, i.e. for higher order transverse modes, the intermodal
distance changes and the amplitudes of distinct longitudinal modes are not
equal and the characteristics depend on the type (or configuration) of reson­
ator. This can be understood from very simple assumptions: due to complex
field amplitude distribution such modes can be expanded as a certain set
of plane waves, where individual plane waves travel a slightly different dis­
tances between the resonator mirrors compared to the TEM00 mode (Gaus­
sian beam), so standing waves in the resonator occur for slightly different
frequencies. Furthermore, it is important to note that a resonator sustains
only a single transverse mode, i.e., electromagnetic field amplitude or intens­
ity distribution in a plane perpendicular to propagation axis, which in turn
is composed of many longitudinal modes (resonance frequencies).
A set of longitudinal modes comprises spectrum of the laser radiation.
However, it is important to note that here we analyzed an ideal ”empty”
resonator which can sustain an infinite number of longitudinal modes (the
index q can be from 0 to infinity and their width is infinitively narrow, as
depicted in Fig.43. In fact, any real laser supports (amplifies) only a finite
number of longitudinal modes, which fit under gain contour (the laser band­
width) of the active medium. For example, a fine structure of He-Ne laser
spectrum which consists of 6 discrete longitudinal modes is shown in Fig.44.
The overal spectral intensity distribution is described by a Gaussian func­
tion, therefore such spectrum is called Gaussian spectrum despite the fact
3.5 Fabry-Perot resonator 64

that it actually consists of only 6 discrete frequencies (longitudinal modes).

Figure 44: Example of multimode He-Ne laser spectral structure with 500 MHz
intermodal distance.

The intensity peaks corresponding to intrinsic resonator frequencies (lon­

gitudinal modes) are rarely detected during laser radiation measurements.
Since the intermodal distance is inversely proportional to the resonator length,
longitudinal modes can be resolved only in the case of short resonators, e.g.,
with lengths of a few mm. In the case of long resonator with a meter length
or more, the intermodal distances are very small, the adjacent modes are
very close, so very often the resolution of the spectral device is insufficient to
resolve them. To do this, interferometers with very high spectral resolution
have to be used.

3.5 Fabry-Perot resonator

So far, the longitudinal modes were considered as monochromatic. In reality,
these modes have a certain spectral bandwidth (linewidth). Let us find out
what is the spectral bandwidth of a longitudinal mode and what determines
its width. For this purpose let us analyze wave propagation in a Fabry-
Perot etalon, which is a prototype of an optical resonator. Here we note
that the first ruby laser had exactly this type of resonator, and currently
Fabry-Perot type resonators are commonly used in solid-state microlasers
and semiconductor lasers..
The Fabry-Perot etalon (resonator) is a piece of transparent material
with parallel surfaces, whose thickness is l and refractive index is n0 . Let
us see how such etalon transmits and reflects light waves. We mark Ai as
3.5 Fabry-Perot resonator 65

the complex amplitude of the incident wave, while B1 , B2 , ... and A1 , A2 , ...
as complex amplitudes of reflected and transmitted waves, respectively, as
shown in Fig.45.

Figure 45: Light wave propagation in Fabry-Perot resonator.

The amplitudes of reflected waves can be written as

B1 = rAi
B2 = tt′ r′ Ai eiδ
B3 = tt′ r′3 Ai e2iδ ... (76)
etc, where r and r′ are external and internal reflection coefficients for wave
amplitudes, t and t′ are respective transmittance coefficients for wave amp­
litudes when the wave propagates from the etalon and vice versa. δ is
the phase shift, which can be expressed from simple geometric assumptions
through the optical path difference (δl) = |AB| + |BC|:

2πn0 (δl) 4πn0 l cos θ

δ= = . (77)
λ λ
The full reflected wave complex amplitude is Ar = B1 + B2 + B3 + ...:

Ar = r + tt′ r′ eiδ 1 + r′2 eiδ + r′4 ei2δ + ... Ai . (78)

In the analogy, we can write transmitted wave complex amplitudes:

A1 = tt′ Ai
A2 = tt′ r′2 Ai eiδ
A3 = tt′ r′4 Ai e2iδ ... (79)
3.5 Fabry-Perot resonator 66

etc, and the full transmitted wave complex amplitude is At = A1 +A2 +A3 +...:

At = tt′ 1 + r′2 eiδ + r′4 ei2δ + ... Ai . (80)

Assuming no energy losses (due to absorption or scattering) in reflecting
surfaces (which essentially are resonator mirrors) and in the resonator itself,
we can claim that r = r′ and at the same time energy conservation law is
also satisfied, which in this case can be expressed as r2 + tt′ = 1. Introducing
reflection and transmittance coefficients for wave intensities R = r2 , T = tt′
and R + T = 1 and summing infinite series, we get

√
1 − eiδ R
Ar = Ai
1 − Reiδ
T
At = Ai . (81)
1 − Reiδ
Let us rewrite the latter expressions as wave intensity ratio assuming that
I = AA∗ . The ratio of reflected and incident wave intensities then is

Ir Ar A∗r 4R sin2 2δ
= = . (82)
Ii Ai A∗i (1 − R)2 + 4R sin2 2δ
In a similar manner, the ratio of transmitted and incident wave intensities
is:

It At A∗t (1 − R)2
= = . (83)
Ii Ai A∗i (1 − R)2 + 4R sin2 2δ
Since there are no losses in the resonator, it is easy to determine that It +
Ir = Ii . From expression in Eq.(83) it follows that a maximum transmittance
of the Fabry-Perot resonator is when δ = 2πm, where m is any integer
number. Then It /Ii = 1 and does not depend on the reflection coefficient
R. On the other hand, we can notice that maximum Fabry-Perot resonator
transmittance coincides with a standing wave condition (Eq.70) and intrinsic
frequencies can be expressed as
c
νm = m , (84)
2n0 l cos θ
and the intermodal distance is

c
∆νm = νm+1 − νm = . (85)
2n0 l cos θ
3.5 Fabry-Perot resonator 67

Assuming normal incidence (cos θ = 1), we get the same expression as

in Eq.(75). Consequently, the maximum transmittance of the Fabry-Perot
etalon (resonator) is for intrinsic longitudinal modes. In other words, these
modes experience no losses inside the resonator and have infinitely long life­
time there. On the other hand, the minimum transmittance depends on R.
Fig.46 illustrates the transmittance of Fabry-Perot resonator as a function of
mirror reflection coefficient.

Figure 46: Fabry-Perot resonator transmittance as a function of mirror reflection

coefficient.

In the case of large reflection coefficient, the transmittance maxima are sharp
and minima are deep. When mirror reflectance decreases, the maxima spread
and the amplitudes of minima increase. In that regard it is convenient to
define the spectral width of a longitudinal mode which depends on the res­
onator length and mirror reflectance coefficient:

c 1−R
δν = √ . (86)
2n0 l cos θ π R
Fabry-Perot etalon can be also used as a spectral device: the intermodal
distance can be varied by adjusting l and the spectral linewidth can be varied
by changing R. In practice, Fabry-Perot etalon is inserted into a resonator of
a CW laser when single longitudinal mode operation and very narrow spec­
tral bandwidth is desired. Fig.47 shows an example of He-Ne laser spectral
structure and a Fabry-Perot etalon modal structure. The infinite set of res­
onator longitudinal modes is limited by the width of the lasing medium gain
curve and the Fabry-Perot etalon transmittance: without the Fabry-Perot
etalon there would be a number of longitudinal modes in the emitted He-Ne
3.6 Losses in resonators 68

laser spectrum, but the Fabry-Perot etalon filters most of them. Leaving just
a single longitudinal mode that falls under the gain curve.

Figure 47: He-Ne laser spectral structure and Fabry-Petor etalon mode structure.
The etalon filters longitudinal modes effectively forcing He-Ne laser to emit only
a single longitudinal mode.

For ultrashort pulse (broad bandwidth) lasers Fabry-Perot etalon is not

recommended since beatings between the intrinsic etalon and laser resonator
modes occur, resulting in modulation of radiation spectrum and deterioration
of the pulse temporal shape. This phenomenon is called the etalon effect.
More generally, the etalon effect can be caused by any other optical element
in the resonator which has parallel surfaces perpendicular to the optical axis
of resonator. For this reason all surfaces of optical elements in the resonator
(except for resonator mirror surfaces) are adjusted or cut at a slight angle
with respect to the optical axis of resonator (e.g., a few degrees).

3.6 Losses in resonators

In order to understand optical resonator operation, it is important to evaluate
sources of energy loss, i.e., find out where the electromagnetic wave energy
dissipates. Resonator losses can be determined using various parameters:
losses per round-trip, resonator quality or photon lifetime in a resonator. In
this section we will provide simple evaluations of these parameters and we
will also show how they are mutually related.
Firstly, let us determine the lifetime of a photon in a resonator. Assume
that initially we have a number of photons circulating in the resonator Np0 .
If the reflection coefficients of resonator mirrors are r1 and r2 , respectively,
after one round-trip the number of photons decreases to r1 r2 Np0 and the lost
3.6 Losses in resonators 69

photon number is (1 − r1 r2 )Np0 . The full round-trip time in a resonator is

2n0 l/c, so the evolution of photon number in time can be expressed as
dNp 1 − r1 r2
=− Np . (87)
dt 2n0 l/c
The solution of this equation is
t
Np (t) = Np0 exp − , (88)
tp
where tp is the photon lifetime in a resonator:

2n0 l/c
tp = . (89)
1 − r 1 r2

The former expression can be interpreted in the following way: the photon
lifetime in a resonator equals to a resonator round-trip time divided by
round-trip losses. On the other hand, photon lifetime can be related with
characteristic resonator mode extinction time:
dE E
=− , (90)
dt tp
where E is the energy of a resonator mode. Resonator quality is usually
defined as a ratio of accumulated energy and dissipated power:

ωE ωE
Q= =− , (91)
P dE/dt
where ω is the frequency of radiation. Comparing both expressions we get
that Q = ωtp . Now let us discuss the causes of optical losses in resonators.
These are produced by several mechanisms:

• Losses due to reflections. These losses are unavoidable since mir­

ror reflections are not perfect, e.g., aluminum mirrors reflect of around
96% incident energy in the visible and near-infrared (an even less in
the ultraviolet). The mirrors with dielectric coatings provide reflection
coefficients as high as 99.9%. On the other hand, lower reflection of
one of the resonator mirrors is set by purpose to enable the radiation
to come out of the resonator. Such semitransparent mirrors are called
output couplers. This is especially important for CW lasers which do
not have any other elements for radiation control in the resonator. A
certain amount losses are introduced by reflections from the surfaces of
3.6 Losses in resonators 70

optical elements inside the resonator. For example, Fresnel reflection

from a single uncoated fused silica lens surface is roughly 4% and if
there are multiple such elements, considerable losses are accumulated
during a single round-trip. Therefore, surfaces of all elements inside the
resonator are coated with anti-reflection dielectric coatings which re­
duce losses related to reflections by several or dozen of times. Moreover,
reduction of reflection losses is also achieved by positioning elements at
a Brewster angle when laser radiation is polarized.

• Losses due to absorption and scattering. Firstly, if the quantum

defect (the difference between pump and laser photon energies) of a
laser is small, i.e., the energy levels from which lasing occurs and the
energy levels to which pump is absorbed are close, a fraction of gener­
ated radiation is absorbed in the active element itself. Secondly, optical
elements in the laser resonator have to be transparent to laser radiation,
but sometimes absorption losses due to impurity-related energy levels
or due to being close to absorption band edge are notable. For ultra­
violet lasers, the durability of resonator optical element maintaining
the same high quality is of paramount importance. At long UV ex­
posure times, high energy UV photons in some transparent materials
(e.g. fused silica, etc.) can create long living color centers, which have
large absorption. Thirdly, there is always scattering of light form dust,
dirt or imperfections of optical elements. In that case light is scattered
at large angles and does not return to the resonator. Losses due to
scattering increase significantly when surface or volume of the optical
element is damaged.

• Diffraction losses play a double role. On the one hand, when diffrac­
tion losses can be decreased or increased by selecting a certain resonator
configuration: mirror radius of curvature and resonator length. Phys­
ical apertures of elements inside the resonator are often smaller than
apertures of the resonator mirrors, so these diffraction losses due to
the former must be taken into account. On the other hand, diffrac­
tion losses help to suppress higher order TEM modes whose generation
is usually not desirable. Finally, diffraction losses, as a rule, are the
largest in long resonators, but it is the diffraction-related spatial filtra­
tion of transverse modes which enables formation of light beams with
very high spatial quality (in simple words, all beam irregularities due to
not ideal optical surfaces, dust, etc are filtered out due to diffraction).
Lasers with short resonators and low diffraction losses (e.g., semicon­
ductor lasers and laser diodes) generate low spatial quality beams, their
3.6 Losses in resonators 71

M 2 is large and such beams are difficult to focus.

In conclusion, resonator losses are important not only when constructing

high quality resonators: without them it is impossible to set the control on
the output parameters of laser radiation. Diffraction losses enable generation
of high quality light beams. By modulating losses in the resonator, pulsed
laser operation and ultrashort pulse generation are performed. This last issue
will be discussed in the coming chapters.

Summary of Chapter 3
• Optical resonator is an indispensable component of any laser,
that is responsible for energy accumulation, feedback and forma­
tion of spatial and spectral characteristics of the laser radiation.

• The resonator stability condition is equivalent to convergence

condition of light rays and beams, which states that a light field
is confined in a resonator after infinite number of round-trips.

• Transverse (spatial) mode of a resonator is a certain distribution

of light electric field that reproduces itself after each round-trip.

• The lowest order transverse mode is Gaussian beam. Higher-

order transverse mode are Hermite-Gaussian and Laguerre-
Gaussian beams, which possess complex intensity distributions.

• Longitudinal mode (resonance frequency) of a resonator is the

light wave which satisfies standing wave condition.

• At a time laser resonator sustains many longitudinal modes, but

only a single transverse mode.

• Any real resonator experiences energy losses due to a variety of

reasons: reflections, diffraction, absorption and scattering. The
resonator quality could be expressed by the photon lifetime in
the resonator.
4 Light interaction with atomic systems 72

4 Light interaction with atomic systems

This chapter addresses the fundamental light-matter interaction processes
that govern laser operation. Various general aspects of laser media, such
as laser gain, amplification bandwidth, laser energy level systems and pump
sources are discussed.

4.1 Einstein’s treatment of spontaneous and induced

transitions
In 1900 M. Planck postulated that light can be absorbed or emitted in the
form of certain energy portions, called quanta. In addition, N. Bohr proved
that electrons in atoms and molecules can occupy energy levels which can
have only quantized energy states. At that time it was believed that light and
matter interaction can by fully ascribed by the two processes: absorption
and emission. In 1916 A. Einstein investigated the fundamental aspects
of electromagnetic radiation interaction with matter and discovered that a
steady state of electromagnetic field and matter is possible only when yet
another previously unknown process is taken into account, which was called
stimulated (induced) emission. Let us discuss these processes in more
detail.
If the atom is in an excited state (the electron occupies a higher energy
level), it spontaneously returns from the higher energy level E2 to the lower
energy level E1 , which for the sake of simplicity is regarded as the ground
level. This transition is accompanied by emission of a photon, whose energy
is equal to the difference between the energy levels (as schematically depicted
in Fig.48):

Eph = hν = E2 − E1 . (92)
4.1 Einstein’s treatment of spontaneous and induced transitions 73

Figure 48: Spontaneous emission of radiation.

This process is spontaneous and is called spontaneous emission. The photon

is emitted in a random direction and the corresponding light wave has random
phase and polarization. The probability of such transition is represented by
Einstein’s coefficient A21 , which determines the probability of transition
from higher to lower energy level per unit time.
Provided that the number of atoms in excited level per unit volume of
material is N2 (this is called population), the number of atoms returning to
the ground energy level is N2 A21 . The equation describing evolution of the
excited level population can be written as
dN2
= −A21 N2 , (93)
dt
and its solution is

N2 (t) = N20 exp(−A21 t), (94)

where N20 is the initial (t = 0) population of the excited level. So the popula­
tion of excited level N2 decreases in time with a characteristic time constant
t21 = tspont = 1/A21 . It is obvious that in a two-level system, which is illus­
trated in Fig.48, the population of ground level N1 increases with the same
time constant. The time tspont is called lifetime of the excited level, which is
inversely proportional to the probability of the corresponding electron trans­
ition. This probability is very different for energy levels of various atoms and
molecules. Typically, probabilities for transitions allowed by the selection
rules are of the order of 106 ÷ 108 s−1 , or, in other words, lifetimes of such
excited energy levels are in the range of 10−6 ÷ 10−8 s. If the electron trans­
itions are forbidden (there are no absolutely forbidden transitions in nature;
the probability of such transition is simply much smaller), the lifetime of
such energy level is much longer and may vary from 10−4 s to even a few
seconds. Such energy levels are called metastable. Their lifetime depends
4.1 Einstein’s treatment of spontaneous and induced transitions 74

on the physical state of the matter (e.g., in rarified gases metastable levels
can exist for tens of seconds and even longer). Note that tspont is a material-
related characteristic. In general, lifetime of an energy level depends on the
frequency of radiated photon: the greater is the frequency (also meaning
greater energy difference between the levels), the shorter is the correspond­
ing lifetime. This is a logical result since electrons in atoms always try to
occupy the state with a lowest energy.
According to the uncertainty principle, the real energy levels are broadened.
This means that these energy levels emit not infinitely narrow (monochro­
matic) light, but a spectrum with finite bandwidth, which is characterized
by a certain linewidth (the so-called natural linewidth) which is described by
a spectral function g(ν):
∞
g(ν)dν = 1. (95)
0

The quantity g(ν)dν indicates the probability that the frequency of emitted
photon is in the frequency interval ν + dν, and the spectral intensity of
radiation is defined as I(ν) = I0 g(ν).
Einstein discovered that besides the spontaneous emission there also ex­
ists another form of emission of radiation, which was called the stimulated
emission. This means that excited atoms can return to the ground (un­
excited) state not only spontaneously but also can be forced to do that.
Consider an excited atom (or an ensemble of atoms) which is in the field of
external radiation that has spectral energy density ρ(ν). If there are frequen­
cies in the external field that match the frequency of transition from excited
to the ground energy level, stimulated (induced) transitions occur, as shown
in Fig.49. An important property of stimulated emission is that emitted
photon has the same direction and energy as the incident photon
of the external radiation field, and the corresponding electromagnetic
wave has the same polarization and phase as the incident one.

Figure 49: Stimulated emission of radiation.

4.1 Einstein’s treatment of spontaneous and induced transitions 75

′
The rate of stimulated transitions can be expressed as B21 (ν)N2 , where
′
B21 (ν) is a specific function describing the probability of such transition,
′
and is expressed as B21 (ν) = B21 g(ν). Here B21 denotes the Eintein’s
coefficient for stimulated emission. In this case, the evolution of excited
energy level population can be expressed by the following equation
∞
dN2
= −B21 N2 g(ν)ρ(ν)dν = −N2 B21 ρ(ν), (96)
dt 0

for simplicity, assume that ρ(ν) is a narrow line and ρ(ν) = ρ(ν0 ). Quantity
B21 ρ(ν) is called rate of stimulated emission.
Now let us consider absorption. Since absorption is a process which occurs
only in the presence of external field, it is a stimulated (induced) process,
which is depicted schematically in Fig.50.

Figure 50: Stimulated absorption of radiation.

During stimulated absorption, a photon, whose energy is equal to the energy

difference between the energy levels of interest, is absorbed and the atom
becomes excited. Then the evolution of the ground energy level population
N1 can be described as:
∞
dN1
= −B12 N1 g(ν)ρ(ν)dν = −N1 B12 ρ(ν), (97)
dt 0

where B12 is the Einstein’s coefficient for (stimulated) absorption and

has a meaning of absorption probability.
Since all processes in nature must obey energy conservation laws, the dis­
cussed light matter interaction processes, which essentially represent energy
transfer via certain mechanism, are related. Let us find the relation between
the Einstein‘s coefficients. To make the task simpler, consider a two-level
system, as illustrated in Fig.51.
Einstein’s theory is based on several important assumptions:
4.1 Einstein’s treatment of spontaneous and induced transitions 76

Figure 51: Equilibrium between stimulated absorption, stimulated emission and

spontaneuos emission in a two-level system.

• Spectral density distribution of the external radiation corresponds to a

black body radiation with temperature T :

8πhν 3 1 8πν 2 1
ρ(ν) = hν = × hν × hν . (98)
c3 e kT − 1 c 3
e kT − 1
Here the first multiplier denotes the number of modes within a fre­
quency interval, the second denotes the photon energy and the third
denotes the modal population factor, i.e., number of photons per mode;

• Population distribution of the atom system obeys the Boltzman distri­

bution:

N2 hν
= exp − ; (99)
N1 kT
• There is a thermodynamic equilibrium between the atom system and
radiation, which means that the population of energy levels does not
change when temperature is stable.

In the case of thermodynamic equilibrium we have

dN2 dN1
= = 0. (100)
dt dt
Let us also assume that radiation energy density of a black body has very
small variations in the frequency domain of interest, which is determined
by the electron transitions between excited and ground energy levels. Then
the resulting population of an excited level is described as an equilibrium of
stimulated absorption, stimulated emission and spontaneous emission:
4.1 Einstein’s treatment of spontaneous and induced transitions 77

dN2
= N1 B12 ρ(ν) − N2 B21 ρ(ν) − N2 A21 = 0. (101)
dt
Inserting expression of ρ(ν) we get:

� � � �
8πhν 3 8πhν 3
N2 B21 3 hν/kT + A21 = N1 B12 3 hν/kT . (102)
c e −1 c e −1

With a reference to Eq.(99), we can rewrite the latter equation as

8πhν 3 A21
= hν/kT
. (103)
3
c ehν/kT −1 B12 e − B21
In this way we get one equation with three unknowns. It is satisfied when:

A21 8πhν 3
B12 = B21 , = . (104)
B21 c3
These are Einstein relations. Let us discuss their meaning. From ex­
pressions in Eq.(104) it is clear that Einstein’s coefficients A21 , B21 and B12
are mutually related: if we know at least one of them, we can find the other
two. Another important thing to note is that coefficients for stimulated
emission and stimulated absorption are equal (in the case of non-degenerate
energy levels). This implies that stimulated emission and stimulated absorp­
tion are the opposite processes, and not spontaneous emission and stimulated
absorption, as might seem at the first sight. If energy level population is
N1 > N2 , then external radiation field is absorbed and attenuated, while if
N2 > N1 , photons that appear as a result of stimulated emission add to the
external radiation field which is thus amplified. A situation when (N2 > N1 )
is called population inversion. It is clear that in any laser the condition
of population inversion must be satisfied.
Now let us discuss the role of spontaneous emission. The rate of stim­
ulated emission depends on the energy density of external radiation field:
W21 = B21 ρ(ν), while the rate of spontaneous emission does not. So if we
ignore transitions related to spontaneous emission when describing the ther­
modynamic equilibrium between two energy levels, we get N2 B21 = N1 B12
which, knowing that B21 = B12 , would yield N2 = N1 ; this is obviously
impossible in the case of thermodynamic equilibrium.
An important aspect is the ratio between the rates of spontaneous and
stimulated emissions, which can be defined as
4.2 Gain coefficient 78

A21 hν
R= ≃ exp −1 . (105)
B21 ρ(ν) kT
Let us compare this ratio in the case of microwave and optical range at
equal ambient temperature, T = 300 K. Microwave frequency is roughly
ν ≈ 1010 Hz, then hν/kT ≈ 1.6 × 10−3 and R ≈ 0.0016. This indicates that
in the microwave range the rate of stimulated emission significantly exceeds
the rate of spontaneous emission. The opposite result is achieved in the
optical range, assuming ν ≈ 1015 Hz. Then we get R ≈ 160. This means
that the rate of the stimulated emission in the optical range is very low
compared to that of the spontaneous emission, therefore, it is much harder
to achieve laser amplification in this spectral range. With this in mind, it is
easy to understand why masers were invented much earlier than lasers. It is
also obvious that laser generation and amplification is much harder to realize
in the blue and ultraviolet spectral range. Of course, emission rate ratio for
metastable energy levels may be substantially different in the optical range
but the general tendency is the same.
Although Einstein’s relations were derived in the case of thermodynamic
equilibrium, they are also valid in diverse conditions as transition probabilit­
ies are specific characteristics for a given material. In reality, a working laser
is hardly in the thermodynamic equilibrium, but despite this, the meaning
of Einstein relations remains the same.

4.2 Gain coefficient

Assume that a monochromatic wave with a frequency ν propagates in a two-
level atom ensemble and the unit volume population of atom ensemble is
as defined earlier: N1 (the ground energy level) and N2 (the excited energy
level) which are measured in atoms per cm3 . For the sake of simplicity, as­
sume that no spontaneous transitions occur in the atom ensemble (or their
rate is very low compared to that of stimulated emission). In fact, spon­
taneous emission can be ignored if we analyze propagation of radiation at
small angles with respect to the optical axis, e.g., in an optical resonator.
Since the spontaneous emission occurs in random directions, its power is dis­
tributed evenly throughout the whole spatial angle and power density at a
given direction (in this case, along the optical axis of the resonator) is very
low. In contrast, the stimulated emission is directional. Expressing spectral
intensity of the light field as Iν = ρν /c, let us find the change of intensity
versus propagation distance z:
4.2 Gain coefficient 79

dIν
= N2 B21 g(ν)ρν − N1 B12 g(ν)ρν hν. (106)
dz
Using Einstein relations Eq.(104) and definition of intensity we get

dIν c2 A21
= N2 − N1 g(ν)Iν . (107)
dz 8πν 2
The solution of this equation is

Iν = Iν (0) exp γ(ν)z , (108)

where Iν (0) is the initial intensity of radiation and γ(ν) is the gain coeffi­
cient, which is expressed as:

c2 A21 λ2
γ(ν) = N2 − N1 g(ν) = ∆N g(ν) . (109)
8πν 2 8πtspont
The amplification of light is achieved when γ(ν) > 0, while attenuation
occurs when γ(ν) < 0. The sign of gain coefficient depends only on the differ­
ence of energy level population ∆N , since all other coefficients are positive.
If the ensemble of atoms is in a thermodynamic equilibrium, population of
the energy levels obeys Boltzmann‘s distribution, then the following is always
true: N2 < N1 and ∆N < 0 when T > 0. Consequently, in a two-level system
we always have only attenuation (absorption) of the external field and in this
case γ(ν) < 0 is simply equivalent to the absorption coefficient γ(ν) ≡ α(ν).
On the other hand, even if the system is far from thermodynamic equilib­
rium and external energy source is present, the population inversion is still
impossible because the rates of stimulated emission and stimulated absorp­
tion are equal. In such a two-level system of atoms, amplification is only
possible when T < 0, which has no physical sense according to the classical
(Boltzmann) definition of the absolute temperature. To this end, in the early
studies, the population inversion was related to an effective negative absolute
temperature.
It is clear that in order to achieve amplification of light (laser generation),
a two-level system is insufficient to create the population inversion (∆N > 0).
Leaving the issues how the population inversion is achieved in practice, we
currently only note that population inversion is achieved not only by setting
the system out of thermodynamic equilibrium (with the use of an external
energy source) but also by involving additional energy levels.
4.3 Homogeneous and inhomogeneous spectral linewidth broadening 80

4.3 Homogeneous and inhomogeneous spectral linewidth

broadening
Although the spectrum of resonator longitudinal modes is infinitely broad,
the expression of gain coefficient in Eq.(109) has the same frequency depend­
ence as g(ν), which means that the laser has its own (intrinsic) frequency
(amplification) bandwidth. Indeed, amplification bandwidth determines the
frequency spectrum of laser radiation. Let us discuss how broad the gain
bandwidth could be and which physical mechanisms define its spectral width.
Due to uncertainty principle, both emission and absorption (in that regard
these processes could be considered identical) spectral linewidths are not
infinitely narrow, they are broadened. The linewidth broadening can be of
two types: homogeneous and inhomogeneous.
In the case of homogeneous broadening, excited atoms (or mo­
lecules) in the ensemble are treated as identical (undistinguish­
able). This means that spectrum of emitted radiation is identical for each
atom and we cannot distinguish which atom exactly emitted one photon or
another photon. The fundamental reason for homogeneous spectral broad­
ening is the time-energy uncertainty. Moreover, spectral lines of emission
are broadened due to elastic collisions between excited atoms with other
atoms or phonons, all of which produce phase jumps of the emitted mono­
chromatic waves. As a result of superposition of monochromatic waves with
different phases, we get a certain temporal modulation which in turn makes
emission spectrum not monochromatic, but broadened. There are several
other reasons due to which homogeneous spectral line broadening occurs,
or in other words, due to which additional spectral line broadening is in­
duced. Firstly, the radiative and non-radiative transitions to other energy
levels, which shorten the lifetime of the excited level and hence broaden the
emission spectrum. Secondly, spectral line can additionally homogeneously
broaden when an ensemble of atoms interacts with external field of radiation.
This effect comes into play when gain saturation in the laser operation occurs.
A special type of homogeneous broadening occurs in certain laser media
when transition metal ions are embedded into the structure of crystal lattice.
Then the spectral line broadens due to crystal lattice oscillations (various
phonon interactions) which reduce the lifetime of the excited level. Such
oscillations occur in all crystalline media, but a considerable broadening of
ion emission spectrum occurs only when the emitting energy levels are weakly
shielded from the net crystal lattice field. This occurs exactly in the case of
transition metal ions, e.g., Ti, Cr. Such lasers are often called vibronic and
the most prominent example is the Ti ion doped sapphire which has a very
wide homogeneously broadened emission band.
4.3 Homogeneous and inhomogeneous spectral linewidth broadening 81

Figure 52: Shape and width of spectral line in the case of homogeneous broadening
of energy levels.

In an ideal case the shape of homogeneously broadened spectral line

is described by the Lorentz function:

∆ν
gH (ν) = , (110)
2π (ν − ν0 )2 + (∆ν/2)2
where ∆ν is the spectral linewidth at half intensity level as shown in Fig.52
and equals to
A21 1
∆ν = = . (111)
2π 2πtspont
Such spectral line broadening is identical for any atom in the ensemble and
since we cannot distinguish between the radiating atoms, it is called homo­
geneous.
Inhomogeneous spectral line broadening is caused by a certain
distribution of features of the individual atoms (and consequently,
their energy levels). In this case we can distinguish between radiating
atoms or their groups, as each atom or a group of atoms emit radiation
with slightly different frequency. There are several physical reasons of in­
homogeneous spectral line broadening, which are different for various laser
materials. Here we will discuss several of them which are the most important
and common.
In the case of solid-state lasers, there are two main reasons for inhomo­
geneous spectral line broadening. The first is attributed to amorphous struc­
ture of laser host material. Various glasses doped with rare-earth metal ions
(Nd, Yb, etc) represent the most illustrative example. Glasses are amorph­
ous materials, which from a macroscopic view are isotropic. However, from
4.3 Homogeneous and inhomogeneous spectral linewidth broadening 82

the microscopic view of structure, they are not strictly isotropic but com­
posed of many small ordered structures which are ”frozen” in the isotropic
matrix of the glass. Orientation and properties of distinct structures (such
as mechanical stress, concentration of impurities and their energy levels) is
slightly different, which in turn slightly modify the energy levels of ions that
are embedded in the glass matrix. Then the overall radiation spectrum is
a sum of all spectra emitted by these structures which consequently makes
the spectrum relatively broad compared to that of homogeneously broadened
line. For example, the spectral line of Nd ions embedded in a YAG crystal
is homogeneously broadened, whereas the spectral line of the same Nd ions
embedded in a glass is inhomogeneously broadened and its width is roughly
50 times broader.
Inhomogeneous spectral line broadening in crystalline materials occurs
due to various defects and impurities in the crystal lattice, so each radiating
ion is surrounded by a slightly different environment. However, in most cases
such inhomogeneous broadening mechanism is relatively weak, as modern
technologies enable production of laser materials with very high optical and
chemical quality.
Several mechanisms of inhomogeneous spectral line broadening are spe­
cifically attributed to gas lasers. The first occurs in gas mixtures due to iso­
tope impurities, which emit radiation with slightly shifted frequencies. The
second mechanism is attributed to spectral line broadening due to Doppler
effect. Random velocity distribution of gas atoms or molecules suggests that
a stationary observer due to Doppler effect sees slightly shifted frequency of
radiation, which depends on the direction of the moving particle with re­
spect to the observer. Distribution of atom or molecule velocities depends
on the mass and temperature and is described by the Maxwell-Boltzmann
distribution function, which can be converted to the distribution function
of radiation frequency. For an inhomogeneously broadened spectral line the
function is
� �
1/2 2
4 ln 2 1 ν − ν0
gI (ν) = exp −4 ln 2 . (112)
π ∆ν ∆ν

In the ideal case, the shape of the inhomogeneously broadened spectral line
is a Gaussian function whose width is defined as
1/2
8kT ln 2
∆ν = ν0 (113)
M c2
and is a function of atomic mass M and temperature T . In the case of
4.4 Three-level and four-level lasers 83

inhomogeneous broadening, the spectral line can be viewed a superposition

of multiple homogeneously broadened lines as shown in Fig.53.

Figure 53: The shape of the spectral line in the case of inhomogeneous broadening
which consists of multiple homogeneously broadened lines shown by dashed curves.

In many real laser systems both, homogeneous and inhomogeneous line

broadenings are not symmetrical with respect to the central frequency and
the emission spectrum of each laser material has a specific shape. The spec­
tral line of any laser material is often broadened both homogeneously and
inhomogeneously, but one of the broadening types is usually dominant. Spec­
tral linewidth and its broadening strongly depends on the temperature, so
lasers generating a very narrow spectrum of radiation operate only at low
temperatures, e.g., cryogenic temperature. In conclusion, notice that in the
case of homogeneous spectral line broadening, the linewidth of emitted spec­
trum is inversely proportional to the lifetime of the excited level, while for
inhomogeneous spectral line broadening this assumption does not apply.

4.4 Three-level and four-level lasers

Now let us discuss possible energy level systems in the laser medium in which
the population inversion can be achieved. In principle, all laser energy level
systems in laser media, no matter how complex they are, can be approxim­
ated as either three-level or four-level systems regardless of the type of
spectral line broadening.
To achieve laser operation, the energy levels must obey certain require­
ments. Let us first consider a four-level laser, whose energy level diagram
is schematically illustrated in Fig.54. In a four-level laser, atoms by absorb­
ing pump energy are excited from the ground level (0) to the level (3), which
may consist of an entire system of energy levels. The lifetime of level (3)
4.4 Three-level and four-level lasers 84

Figure 54: Principal diagram of a four-level laser system.

has to be short, it is also desirable that any radiative transitions from level
(3) do not occur or their rate is very low. In this way, excitations of level
(3) are rapidly transferred by non-radiative transitions to level (2), which
is metastable, i.e. has long lifetime. Levels (2) and (1) are the upper and
lower laser levels, respectively. To achieve population inversion between laser
levels (2) and (1), lifetime of energy level (1) has to be short. i.e. its excit­
ations rapidly relax to the ground state (0) by radiative and non-radiative
transitions. Therefore, in a four-level laser, any excited atom in level (3)
quickly transfers its excitation to level (2) and since level (1) is empty (very
quickly relaxing), the population inversion is achieved by very simple means.
Examples of four-level laser systems include Nd-doped solid-state lasers (e.g.
Nd:YAG, Nd:glass), Ti:sapphire, excimer and most of gas lasers.
A different situation occurs in a three-level laser (Fig.55), where the
lower energy level (1) matches the ground level. In this case, to achieve
population inversion it is necessary to excite at least half of the atoms from
the ground level to level (3). Obviously, this can be obtained only using
sufficiently high pumping rate. The most prominent example of three-level
laser is ruby laser.
4.4 Three-level and four-level lasers 85

Figure 55: Principal diagram of a three-level laser system.

Compared to a three-level laser, four-level laser offers a great advantage

that population inversion is ideally achieved when just one atom is excited to
level (3) that consequently quickly decays to level (2). Therefore four-level
lasers are much more used whenever possible.
More recently, the so called quasi-three-level lasers became an import­
ant laser category. In these lasers the energy level scheme is similar to that
of four-level lasers (Fig.56), with the difference that energy levels (0) and (1)
are very closely spaced (even partially overlapped). The advantage of such
energy level configuration is small quantum defect; that is the energy dif­
ference between absorbed and emitted energy. Small quantum defect means
that only a small amount of pump energy is converted to heat. On the other
hand, since ground (0) and laser emission end (1) energy levels are closely
spaced, pumping level (3) can absorb a fraction of laser radiation. This ef­
fect is called reabsorption. Examples of quasi-three level lasers are Yb-doped
solid-state (Yb:YAG, Yb:KGW) and Yb-doped and Er-doped fiber lasers.
The most distinguished difference between four-level and quasi-three-level
systems is that an unpumped four-level system is transparent to the laser
light, while a quasi-three-level system has significant laser light absorption in
that state: such laser system can absorb part of the radiation that it emits
(reabsorption). Therefore, some level of pump intensity is required to obtain
transparency of such system (when most electrons are excited in the up­
per energy levels, absorption of such medium significantly reduces effectively
making it transparent), and laser gain is achieved only for intensities higher
than that.
4.5 Lasing at multiple wavelengths 86

Figure 56: Principal diagram of a quasi-three-level laser system.

4.5 Lasing at multiple wavelengths

Due to complexity of the real laser medium energy level system, the same las­
ing medium may have several possible three-level or four-level combinations
that enable lasing at multiple wavelengths. Nd ion energy level structure
and possible lasing transitions are depicted in Fig.57. Nd:YAG has multiple
possible laser transitions. The most important fact is that different trans­
itions have different cross-sections (probabilities) with transition emitting at
wavelength of 1064.1 nm being the most probable. In practice, the main las­
ing wavelength (which is usually associated with a particular lasing medium
and offered in commercial lasers) is chosen to be that of the highest cross-
section (probability): this corresponds to lowest laser generation threshold
and highest gain. This is done by using resonator mirrors with high reflectiv­
ity for that wavelength.
The pump wavelength is usually chosen such that its absorption would be
maximum. It is 808 nm is the case of Nd:YAG, but pumping at wavelength of
869 nm is also possible. Interestingly, as seen from Fig.57, in case of 869 nm
pump wavelength Nd:YAG medium operates as a three-level system.
Another example of multiple lasing wavelengths is the He-Ne laser. It
has several possible emission wavelengths at 543.36 nm, 593.93 nm, 604.61
nm, 611.8 nm, 632.8 nm, 1.15 µm, 3.39 µm, some of which are shown in
Fig.58. Interestingly, the first demonstrated He-Ne laser operated at 1.15 µm
wavelength, but later a more efficient transition was discovered at 632.8 nm,
which is the common commercially offered He-Ne laser wavelength.
4.5 Lasing at multiple wavelengths 87

Figure 57: Energy level structure of Nd ion embedded in YAG crystal, pump and
laser transitions.

Figure 58: Simplified view of excited states of He and Ne atoms and possible
transitions.

There are also some unique laser media, e.g., dye lasers, which possess an
extremely broad emission spectrum. This is related to the energy structure
of dyes. They are usually in liquid phase and their electronic states have
multiple vibrational energy sub-levels. Lasing can occur between any (or all)
of these vibrational energy levels, therefore, lasing is broadband (see Fig.59).
This unique feature is important in femtosecond lasers: the shorter is the
4.6 Pump sources 88

pulse duration, the broader laser gain bandwidth is required to generate such
pulse (see Eq.13). Dye lasers very historically the first femtosecond lasers.

Figure 59: Simplified principal view of dye laser energy levels.

The relevant parameters of several widely used laser media are listed in
Table 2. Gain bandwidth for various laser media differs by more than 6
orders of magnitude. Rhodamine 6G dye and Ti:sapphire crystal are two
unique media with very large gain bandwidth.

Table 2: The main parameters of several laser media. λ0 is the laser wavelength, σ
is the laser transition cross-section, tspont is is the lifetime of the upper laser level,
∆ν is the gain bandwidth, I and H stand for inhomogeneous and homogeneous
line broadening, respectively.
Laser medium λ0 ,µm σ,cm2 tspont ,µs ∆ν Broadening
He-Ne 0.6328 5.8 × 10−13 0.03 1.7 GHz I
Ar+ 0.5145 2.5 × 10−13 0.01 3.5 GHz I
CO2 10.6 3 × 10−18 2.9 × 106 60 MHz I
Nd:YAG 1.064 2.8 × 10−19 230 135 GHz H
Nd:glass 1.054 4.1 × 10−20 350 8 THz I
Ti:sapphire 0.8 3.8 × 10−19 3.9 100 THz H
Rhodamine 6G 0.57 2 × 10−16 0.005 45 THz H/I

4.6 Pump sources

External energy sources of lasers are called pump sources. They are very
different and their choice depends on the laser operation mode, properties
of the laser medium and lifetime of laser energy levels. All pump sources,
according to the physical mechanism of their operation, can be classified into
three types:
4.6 Pump sources 89

• Optical pumping. The lasing medium is pumped by electromagnetic

radiation (light). All solid-state lasers are pumped optically. A source
of optical pumping can be:
1) A low pressure flashlamp filled with noble gasses (Fig.60).
(a)

(b)

Figure 60: (a): noble gas-filled flashlamps used for optical laser pumping. Image
adapted from Wikimedia under CC BY-SA 2.0 license. (b): solid-state laser res­
onator pumping geometries using helical or linear shape flashlamps.

Principle of operation of flashlamps is based on electric discharge. An

electric discharge produces an electric current that travels through the
tube and ionizes the gas inside. As ions recombine with electrons,
photons are emitted. A flashlamp emits broadband (white light) ra­
diation, hence multiple energy levels of the active laser medium are
excited at the same time. This is not energetically efficient since, as
4.6 Pump sources 90

a rule, radiative and non-radiative transitions between various energy

levels cause significant losses and unwanted thermal effects. Flashlamps
are used to pump various Nd-doped lasers, as Nd ion has multiplicity
of absorption lines in the UV, visible and near-infrared.
2) A laser diode or another laser (usually gas or solid-state). Laser
diodes are semiconductor lasers with electrical pumping. It is import­
ant to note that the term ”semiconductor laser” also includes optically
pumped semiconductor lasers which are not to be confused with laser
diodes. As compared with flashlamp pumping, diode pumping is much
more efficient, since emission wavelength of a laser diode may be chosen
to exactly match a single absorption line of the laser medium. Fur­
thermore, laser diodes have better quality (focusability) of the beam
compared to flashlamps, which enables advanced pumping setups with
almost 100% pump consumption efficiency (Fig.61).

Figure 61: End- and side-pumping arrangements using laser diodes.

Most laser diodes emit in the near-IR and are of particular importance
for Yb-doped laser pumping, as Yb ion has absorption lines only in
the near-IR. Laser pumping is used only when the absorption lines or
bands of the active laser medium are in the visible spectral range and
excited laser energy levels have short lifetimes. This is the case for the
4.7 Laser oscillation 91

Ti:sapphire laser, which is pumped by the second harmonic of solid-

state lasers operating in CW or Q-switch regime or by CW argon ion
laser.
3) The Sun. Sun may be considered as an exotic continuous-wave
white-light pump source, a kind of natural flashlamp analogue. Inter­
estingly, the first laser pumped by the sunlight (Nd:YAG) was demon­
strated back in 196611 . Currently Sun-pumped iodine, Nd:Cr:YAG and
several semiconductor lasers have been demonstrated.
4) Microwaves can be used to pump certain gas lasers, e.g., the CO2
laser. Along with the Sun, this is also a rather and rarely used pump
source.

• Electrical pumping. This pumping type uses electrical particles to

excite the lasing medium and achieve population inversion. Electrical
pumping includes:
1) Electrical discharge. This type of electrical pumping is used in
gas lasers (e.g., He-Ne, Kr, etc), where high voltage electric discharge
produces and accelerates free electrons, which transfer their energy
to atoms or molecules of the gas via collisions, creating population
inversion.
2) Electric current. This type of electrical pumping is usually used in
laser diodes. An electric current flowing through the p-n junction which
injects electrons/holes into the depletion region. When an electron and
a hole are present in the same region, they may recombine producing
spontaneous emission (photons) which is used as initial light (seed) for
amplification via stimulated emission.
3) Electron-ion beams. Free electron lasers and some excimer lasers
are pumped by electron/ion beams, which excite atoms or molecules in
a similar fashion as electric discharge.

• Chemical pumping. Chemical lasers use this type of pumping when

population inversion is achieved due to energy released during exo­
thermal (emitting heat) chemical reactions.

4.7 Laser oscillation

From what we discussed earlier, we can easily formulate conditions for laser
generation. In this context it is important to distinguish between the terms
11
C. G. Young, A Sun-pumped CW one-watt laser, Applied Optics 5, 993–997 (1966)
4.7 Laser oscillation 92

”amplifier” and ”generator”. If the population inversion (∆N > 0) and

γ(ν) > 0 are achieved, the medium acts as a laser amplifier: radiation that
passes the medium is amplified (middle image of Fig.62). The resonator for
a laser amplifier is not necessary. For the laser generation to occur, the
laser medium needs to be inserted into a resonator (or made to work as a
resonator) and laser oscillation condition needs to be fulfilled – this setup is
called a laser oscillator or simply a laser (Fig.62 top).

Figure 62: Principal setup of laser oscillator (generator), amplifier and MOPA
system.

For laser amplification it is enough to create the population inversion and

then the amplifier is able to amplify the launched signal. However, it is clear
that achieving population inversion ∆N > 0 and positive gain coefficient is
not enough for laser oscillation in a resonator. Taking into account losses in
the resonator, laser oscillation occurs only when amplification per round-trip
exceeds losses per round-trip:

r1 r2 exp 2γ0 (ν)ll ≥ 1, (114)

where ll is the length of the laser medium and we also assume that the
resonator losses due to diffraction, absorption and scattering are very small;
r1 and r2 are coefficients of reflection for resonator mirrors. Then the gain
coefficient necessary to achieve laser oscillation is

1 1
γ0 (ν) ≥ ln =α. (115)
2ll r1 r2
4.7 Laser oscillation 93

Figure 63: Frequency range of laser generation.

Having depicted the former expression in Fig.63, we notice that from

all longitudinal modes supported by the resonator laser radiation consists
only from those longitudinal modes that fall under the frequency interval in
which expression in Eq.(115) is satisfied. In the case of equality, expression
in Eq.(115) defines the laser oscillation threshold and the threshold pump
rate as well.
Now let us discuss the key difference between the laser oscillator and the
laser amplifier. Since the purpose of the laser amplifier is only to amplify a
certain external (injected) signal, the laser amplifier does not require a res­
onator and laser amplification takes place as long as the population inversion
∆N > 0 is achieved. In contrast, there is no such external signal in the laser
oscillator. Therefore the essential question is from where the initial signal
comes from? In a laser oscillator, once the generation threshold is exceeded,
the oscillation starts from spontaneous emission, a small fraction of which
occasionally propagates along the optical axis of a resonator. The resonator
mirrors return the spontaneous emission back to the laser medium, where it
is amplified and the further oscillation process is dominated by the stimu­
lated emission. In this sense, the resonator performs the function of positive
feedback amplifying only the radiation that propagates close to the optical
axis and forming a high quality beam.
A combination of laser oscillator and amplifier (or multiple amplifiers)
is called master oscillator power amplifier – MOPA (Fig.62 bottom).
Such systems produce laser pulses with high energy.
4.7 Laser oscillation 94

Summary of Chapter 4
• The stimulated emission is the fundamental effect that produces
laser radiation.

• A necessary condition for laser operation is population inversion:

a state when population of electrons in a higher energy level
exceeds population of electrons in a lower energy level, which
cannot be achieved in a two-level system.

• All energy level systems in laser media, no matter how complex

they are, can be approximated as either three-level or four-level
systems.

• Laser gain bandwidth is broadened due to various physical mech­

anisms or fundamental limitations. The broadening can be ho­
mogeneous or inhomogeneous.

• Laser generation starts when laser gain exceeds losses per round-
trip in the resonator. This is the laser oscillation condition.

• The purpose of laser pump source is to excite the medium in

order to create population inversion. There are three types of
laser pumping: optical, electrical and chemical.
5 Pulsed laser operation 95

5 Pulsed laser operation

Compared to all other (non-laser) light sources, a unique feature of lasers is
that they emit light not only in the form of a coherent beam but also they can
emit light in certain portions, called light pulses. Pulsed laser operation
impleas that energy accumulated in the laser medium is emitted within a
very short time interval, which is called pulse duration, or pulse width. With
a reference to the fact that laser radiation is localized in space in the form
of a light beam, pulsed laser operation allows achieving high instantaneous
power (termed peak power) and high intensity (instantaneous power per unit
of area).
There are three distinct types of pulsed laser operation (re­
gimes): free-running, Q-switching and mode-locking. These opera­
tion regimes differ in terms of achievable pulse durations and their technical
realization. Free-running is the simplest laser operation regime, which starts
when the laser oscillation threshold is exceeded. If the pump is steady (con­
tinuous), free-running regime is also continuous (with several exceptions) and
if the pump is pulsed, free-running becomes pulsed as well. Achievable pulse
durations in the free-running regime are of the order of several microseconds
(1 µs=10−6 s) and the peak power of such pulses can reach several kW.
Q-switching and mode-locking regimes are always pulsed regardless of the
pumping type. In Q-switching regime the laser generates light pulses with
durations ranging from several to several tens of nanoseconds (1 ns=10−9 s),
with peak powers reaching hundreds of MW. The shortest light pulses with
durations ranging form 100 picoseconds (1 ps=10−12 s) to a few femtoseconds
(1 fs=10−15 s) are generated in mode-locking operation regime. In this re­
gime, the pulse duration depends on the properties of the laser medium
(achievable gain bandwidth) and the methods of mode-locking, and the peak
power of such pulses may reach several hundreds of MW as well. Light pulses
with durations shorter than several picoseconds are called ultrashort. Both
Q-switching and mode-locking regimes are usually employed in solid-state
lasers since these lasers have low laser transition cross-section (see Table
2) and broad gain bandwidth, i.e., can accumulate large population inver­
sion and generate broadband laser radiation. Other lasers, such as CO2 ,
excimer, dye and semiconductor lasers, can also operate in Q-switching or
mode-locking regimes, but these are rare exceptions. The most powerful light
pulses with peak powers reaching PW (1015 W) and higher are achieved by
amplifying ultrashort pulses in laser amplifiers. Their principles and config­
urations will be discussed at the end of this chapter.
5.1 Free-running regime and relaxation oscillations 96

5.1 Free-running regime and relaxation oscillations

Free-running is the simplest laser generation regime which starts simply when
laser generation threshold is exceeded. Even in the case of continuous pump,
for a certain transient time interval periodic oscillations of laser intensity
may occur. These oscillations are called relaxation oscillations. They
are observed in all lasers and their period is significantly longer than the
resonator round-trip time and even than the photon lifetime in the resonator.
The reason of relaxation oscillations is the dynamic interaction between the
light field in the resonator and variation of population of the laser medium.
When the light field is amplified (the rate of stimulated transitions increases),
this inevitably causes a decrease of the upper laser level population, which in
turn leads to a decrease of gain coefficient. As a result, the intensity of the
light field itself starts decreasing. Since an external pump source is involved,
whose pumping rate is steady (assuming continuous pump operation), the
temporal evolution of intensity develops a certain oscillating pattern which
depends on the relation of these parameters.
To characterize relaxation oscillations quantitatively, let us assume an
ideal four-level laser with a homogeneously broadened spectral line which
generates a transverse mode with the lowest index (TEM00 ) and a single
longitudinal mode. For the sake of simplicity, we assume that the lower laser
level is depleted very quickly, so N1 = 0 and ∆N = N2 (this is also part of our
idealization), and the lifetime of the upper laser level is tspont . Assuming that
the rate of stimulated emission is Wi = σ(ν)ϕ and the intensity is defined
as I = hνϕ, let us write the rate equation for population of the upper laser
level, which is analogous to the rate of population difference:
d∆N ∆N
= R − Wi ∆N − . (116)
dt tspont
Since the rate of stimulated transitions Wi is proportional to the intensity
of radiation, the middle term on the right hand side of the expression can be
rewritten as Wi ∆N = ϕB∆N , where ϕ is the number of photons per unit
volume and B is a certain proportionality constant (a modified Einstein coef­
ficient). This term describes the rate at which photons occurr. Consequently,
we can write an equation for the evolution of the photon flux density as:
dϕ ϕ
= ϕB∆N − , (117)
dt tp
where tp is the photon lifetime in the resonator. When the equilibrium is
established, i.e.,
5.1 Free-running regime and relaxation oscillations 97

dϕ d∆N
= = 0, (118)
dt dt
we can find the equilibrium values for population and photon flux densities:

1
∆N0 =
Btp
RBtp − 1/tspont
ϕ0 = . (119)
B
From the above expressions it follows that when Rs = (Btp tspont )−1 , ϕ0 =
0. Here Rs marks the threshold pump rate. When it is exceeded, the laser
starts to generate and photon flux density increases. We define a ratio r =
R/Rs indicating how many times the threshold pump rate is exceeded. Then
the second equation in the system of equations (119) can be written as
r−1
ϕ0 = . (120)
Btspont
Let us examine what happens when we have small deviations of popula­
tion and photon flux densities from their equilibrium values:

∆N (t) = ∆N0 + N ′ (t)

ϕ(t) = ϕ0 + ϕ′ (t), (121)

where N ′ (t) ≪ ∆N0 and ϕ′ (t) ≪ ϕ0 are small quantities. Inserting these
expressions into the initial equations and using relations in Eq.(119), we get
a system of equations for small variations of population and photon flux
densities:

dN ′ ϕ′
= −RBtp N ′ −
dt tp
′
dϕ 1
= RBtp − N ′. (122)
dt tspont

We will search for solutions of these equations in the form ϕ′ (t) = eAt sin(ωm t),
i.e., assuming that a perturbation of the photon flux density has the shape of
a damped sine function with a damping parameter (a characteristic time con­
stant) A and oscillation frequency ωm . Solving equations (without focusing
on mathematical subtleties) we get:
5.1 Free-running regime and relaxation oscillations 98

r
A= ,
tspont
�
r−1
ωm ≈ . (123)
tp tspont

Figure 64: Relaxation oscillations in a laser and evolution of energy level popula­
tion.

An example of relaxation oscillations is provided in Fig.64. The period

and amplitude of relaxation oscillations depend on the pump rate, namely,
how many times the threshold pump value is exceeded. On the other hand,
the period of relaxation oscillations also depends on the photon lifetime in
the resonator, i.e., the quality of the resonator, which in turn is determined
by the losses in the resonator.
Continuous or pulsed pumping produce different dynamics of relaxation
oscillations. In the case of a continuous pump, the relaxation oscillations are
observed only when the laser is turned on and cease when a steady state of
laser operation is established. In the case of pulsed pumping, the relaxation
oscillations repeat with every pump pulse.
Here we derived parameters of relaxation oscillations in the case when
the laser generates a lowest-order transverse and a single longitudinal mode.
If there are more longitudinal modes, the situation does not change qualit­
atively, the relaxation oscillations still have a damped sinusoidal form with
exponentially decreasing intensity. An interesting situation occurs when the
laser is able to generate higher order transverse modes, i.e., the resonator
5.2 Q-switching 99

quality is very high. Then the relaxation oscillations become chaotic and
emerge as a sequence of non-periodic intensity spikes. The reason for this
behavior is explained as follows. The lowest order mode TEM00 is gener­
ated first and strongly depletes population of the laser medium within its
volume. Since the volume of higher order modes is larger, they are easily
excited because the population of laser medium in the peripheral part re­
mains large and exceeds the threshold value. Each of the transverse mode
produces a sequence of its own relaxation oscillations, which overlap produ­
cing non-periodic intensity spikes. Thereafter the TEM00 mode is generated
again and such mode ”hopping” can continue for quite a long time. The
shape of relaxation oscillations essentially indicates the ability of the laser
resonator to support only the lowest index transverse mode and such simple
observations of relaxation oscillations in free-running mode are readily used
for laser adjustment.

5.2 Q-switching
Q-switching regime is used in order to concentrate laser radiation in a form
of a single pulse and so to increase the peak power of laser output. Q-
switched lasers produce nanosecond pulses with a peak power up to several
hundreds of MW. The first Q-switched laser was demonstrated back in 1962.
At present, Q-switched lasers are used for many scientific and technological
applications such as precision materials processing, laser ranging and detec­
tion and pumping ultrashort pulse lasers and amplifiers, to mention a few.
The idea of Q-switching technique is based on the control of losses
in the resonator and consequently the resonator quality. This is
performed as follows: after the pump is turned on, resonator losses are artifi­
cially increased. Therefore, the laser generation is prevented, but the active
laser medium at that time efficiently accumulates population acting as an
energy storage. When the population inversion reaches its maximum value,
the resonator quality is rapidly increased (returned to its initial high value
state). Then the gain coefficient rapidly becomes very high (exceeding the
threshold value by many times) and radiation intensity inside the resonator
rapidly (during just several round-trips) increases, ”emptying” the whole ac­
cumulated population in the upper laser energy level. Such laser pulse is
often called giant (gigantic) pulse.
Before going into technical details, how Q-switching is performed in prac­
tice, we first analyze the basic features of operation of such laser. The prin­
cipal scheme of a Q-switched laser is depicted in Fig.65. Initially, let us
estimate the parameters of such laser radiation and factors from which they
depend on. Since the entire radiation (laser generation) process usually oc­
5.2 Q-switching 100

curs faster than in 20 ns, we can ignore slow processes, such as change of the
pump rate in time and consequent change of the population.

Figure 65: Principal scheme of a Q-switched laser. r1 and r2 are the mirror
reflection coefficients, l is the length of the resonator, L is the length of laser
medium, R denotes the pump source and Q denotes the resonator Q-switching
unit.

We also assume that resonator quality is switched on instantly. The laser

emission, the laser medium and the resonator itself are described by the
following parameters: photon number in a resonator ϕ, mode volume V ,
population difference in the mode volume n = ∆N V and photon lifetime in
passive resonator tp . Since the gain coefficient γ is proportional to population
difference n, the radiation intensity during propagation in a resonator grows
exponentially:

I(z) = I0 exp(γz), (124)

or dI/dz = γI. An observer traveling with the light wave sees the following
intensity evolution in time:
dI dI dz c
= = γI , (125)
dt dz dt n0
where c is the speed of light and n0 is the refractive index of laser medium.
It is clear that temporal growth constant of the intensity is γc/n0 . Since the
intensity increases only during propagation in the laser medium with length
L, the growth of photon number after propagation from one resonator mirror
to the other is only L/l, as the rest of the resonator length is just a free space.
The growth constant of photon number is then ϕγc/n0 (L/l). On the other
hand, the number of photons in the resonator is reduced due to losses, which
are described by the photon lifetime tp , so the constant of photon number
loss is ϕ/tp . When both of these factors are added together, evolution of
photon number in time can be expressed as
5.2 Q-switching 101

dϕ γcL 1
=ϕ − . (126)
dt n0 l tp
Let us introduce a dimensionless time τ = t/tp , which is normalized to the
photon lifetime in a resonator. Then after multiplying both sides of Eq.(126)
by tp , we get

dϕ γ γ
=ϕ −1 =ϕ −1 , (127)
dτ n0 l/cLtp γt
where γt = n0 l/cLtp is the threshold gain coefficient when number of photons
in the resonator is constant: dϕ/dτ = 0. As the gain coefficient is directly
proportional to population inversion, the former expression can be rewritten
in terms of population inversion:

dϕ n
=ϕ −1 , (128)
dτ nt
where nt = ∆Nt V is the threshold population inversion. The term ϕ(n/nt )
indicates the number of photons generated per unit of normalized time due
to stimulated emission. Because each generated photon is a result of electron
transition between the laser energy levels, after emission of a single photon
the difference of population is reduced by a factor of two: population of
the upper energy level is decreased by one and population of the lower en­
ergy level is increased by one. This means that a single optical transition
corresponds to change of population ∆n = −2 and then its evolution is:
dn n
= −2ϕ . (129)
dτ nt
Eq.(128) and Eq.(129) describe the time evolutions of photon number
and population, respectively. Solving these equations yields the radiation
parameters. Dividing Eq.(128) by Eq.(129), we get photon number evolution
which is related to the change of population:
dϕ nt 1
= − . (130)
dn 2n 2
After integration we get:
[
1 n
ϕ − ϕi = nt ln − (n − ni ) , (131)
2 ni
where ϕi and ni are photon number and population, respectively, at the initial
moment of time. Assuming that the initial photon number in the resonator
5.2 Q-switching 102

is infinitely small (ϕi ≈ 0 and this is essentially the case, because at the
very beginning only photons produced by spontaneous emission exist), we
get expression of photon number in the resonator:
[
1 n
ϕ= nt ln − (n − ni ) . (132)
2 ni
The former expression describes the relation between the photon number
and population at any moment of time. Let us take time t ≫ tp , when we
know that there are no photons left in the resonator (ϕ = 0), then we can
evaluate the final population nf :

nf nf − ni
= exp . (133)
ni nt
n −n
The quantity η = i ni f indicates which portion of the accumulated pop­
ulation is converted into radiation and η approaches unity when ni /nt → ∞,
i.e., the greater is the initial accumulated population at the moment
of switching to high quality, the more efficient is its conversion to
radiation. The time evolutions of photon number and population, as de­
scribed by Eq.(128) and Eq.(129), are plotted in Fig.66.

Figure 66: Time evolutions of photon number and population.

It can be assumed that the generation of a gigantic pulse essentially starts

from a single photon (ϕi = 1) which appears due to spontaneous emission
and whose propagation direction coincides with the optical axis of resonator.
The giant pulse develops in a few resonator round-trips. The time delay τd
between switching the resonator to high quality and temporal peak of the
pulse is expressed as
5.2 Q-switching 103

tp
τd = (134)
ni /nt − 1
and is inversely proportional to the ratio of initial and threshold populations.
The greater is this ratio, the faster is pulse evolution when Q-switching is
turned on. The pulse duration is also a function of this parameter and is
expressed as

η ni /nt
τp = t p . (135)
ni /nt − ln(ni /nt ) − 1
Note that duration of the gigantic pulse is directly proportional to photon
lifetime in the resonator: in order to get shorter pulses, some losses
in the resonator should be introduced, e.g., choosing an output coupler
with greater transmission.
Another important parameter of the laser output is the peak power of
the pulse which is defined as P = ϕhν/tp . According to Eq.(132), the peak
power is expressed as:
[
hν n
P = nt ln − (n − ni ) . (136)
2tp ni
The maximum peak power is achieved at the moment of time when n = nt
(it can be derived from the condition ∂P/∂n = 0). Inserting n = nt to
expression in Eq.(136) we get
[
hν nt
Pmax = nt ln − (nt − ni ) . (137)
2tp ni
Assuming that the initial population significantly exceeds the threshold value,
i.e., ni ≫ nt (this is essentially the case), the expression for peak power sim­
plifies to

ni hν
Pmax = . (138)
2 tp
In practical setups Q-switching mode is optimized for either maximum peak
power (minimum pulse duration) or energy, depending on the planned ap­
plications. An important feature of the Q-switched regime: the longer is the
photon lifetime in the resonator tp (the greater resonator quality), the lower
is the peak power.
Up to now we have analyzed the case when the quality of resonator is
switched instantaneously. In practice, this means that Q-switching time is
5.2 Q-switching 104

shorter than the development time of the pulse τd . What happens when res­
onator quality is switched slowly (the resonator losses α are reduced gradu­
ally), is depicted in Fig.67. Then the gain coefficient γ can exceed losses
several times during Q-switching phase (moments t1 and t2 ) and as a result
several pulses are generated, which is often not a desired result.

Figure 67: Dynamics of photon number and gain coefficient in the case of slow
Q-switching. α marks the level of losses in the resonator.

Figure 68: Dynamics of the Q-switching operation in the cases of (a) pulsed pump
and (b) continuous pump.
5.3 Methods of Q-switching 105

Q-switching can be implemented with both continuous and pulsed

pumping, these operation regimes are compared in (Fig.68). In the case
of continuous pumping, the resonator quality can be switched at kHz and
greater repetition rates, which depend on the time required for the population
to grow from nf to ni ≫ nt . The corresponding repetition rate of the gigantic
pulses will be the same. In the case of pulsed pumping, the repetition rate of
gigantic pulses will be determined by the repetition rate of the pump pulse,
which is typically of the order of several or several hundred of Hz.

5.3 Methods of Q-switching

Up to now the Q-switching unit in the Q-switched laser setup (Fig.65) was
depicted as a ”black box”. Now let us discuss the physical principles and
methods of Q-switching in practical setups. The Q-switching methods are
be classified into active (the resonator quality is controlled by an external
signal) and passive (the resonator quality changes by itself due to nonlinear
losses introduced by a certain optical element).
Active Q-switching methods include:
1. Rotating prism or mirror. This is a very simple Q-switching
method, when the resonator quality is adjusted by temporarily aligning or
misaligning one of the resonator mirrors. A rotating right-angle prism instead
of a resonator mirror could be used as well. This Q-switching method is
schematically depicted in Fig.69.

Figure 69: Q-switching employing a rotating prism.

When reflection from the prism or mirror aligns with the optical axis of the
resonator, the quality of the resonator suddenly increases and a giant pulse
is generated at this moment. This is a very simple and cheap Q-switching
method, which performs equally well at any laser wavelength. The first Q-
switched ruby laser operated exactly according to this principle. However,
this method has one key drawback: the speed of prism rotation is limited
(in real conditions it is up to several hundred rotations per second), so Q-
switching is slow and, as a consequence, several light pulses are generated.
5.3 Methods of Q-switching 106

Therefore in practice this method is no longer used and has only a historic
value as a simple idea of Q-switching .

2. Electro-optic switch. The electro-optic switch is the most commonly

used method for Q-switching. This method is based on controlling the losses
by controlling the polarization of light in the resonator. An electro-optic
switch consists of an electro-optical modulator (Pockels cell) and a polariza­
tion analyzer. Operation of an electro-optical modulator is based on electro-
optical effect. A birefringent crystal is oriented in such a way that light
propagates along its optical axis. When no external electric field is applied,
the refractive indices for ordinary (o) and extraordinary (e) waves in the dir­
ection of the optical axis are equal, the crystal shows no birefringence and
has no effect on light propagation. When a constant external electric field is
applied, the refractive index ellipsoid of the birefringent crystal is modified
along the light propagation direction; because of that a certain phase differ­
ence between the ordinary and extraordinary waves is induced. The induced
phase difference depends linearly on the strength of the applied electric field
(the phenomenon is called the Pockels effect):

2πln3o rnm E ωn3o rnm V

∆ϕ = = , (139)
λ c
where E is the strength of external electric field, l is the crystal length,
V = El is the applied voltage voltage, ω is the frequency of light, no is the
refractive index for the ordinary wave and rnm is a certain element of crystal
electro-optic tensor (electro-optic coefficient). In other words, when voltage
is applied, the birefringent crystal, which is oriented along the optical axis,
acts as a voltage-controlled wave-plate. The voltage value when ∆ϕ = π is
called half-wave voltage. When a half-wave voltage is applied, the crystal
acts as a half-wave (λ/2) plate: it rotates the initial linear polarization by
90◦ . When twice less voltage is applied, the phase difference is ∆ϕ = π/2 and
the crystal acts as a quarter-wave (λ/4) plate; as a result linearly polarized
light becomes circularly polarized. The polarization of light then is analyzed
by means of a polarization analyzing element (e.g., Glan prism or a thin-film
polarizer) which is inserted in the resonator.
The simplified layout of a Q-switched laser that uses an electro-optical
modulator is depicted in Fig.70. The laser operation is controlled by applying
the voltage to the Pockels cell. When no voltage is applied, Pockels cell does
not rotate polarization, so the electro-optic switch is open and the quality
of resonator is high. When any voltage is applied, Pockels cell rotates the
polarization of light circulating in the resonator and the analyzer reflects part
of it out of the resonator axis, so introducing losses and reducing the resonator
5.3 Methods of Q-switching 107

Figure 70: The schematic of a Q-switched laser that uses an electro-optic modu­
lator. V is the voltage applied to the electro-optical crystal (Pockels cell), PA is
the polarization analyzing element.

quality. If the half-wave voltage to the Pockels cell is applied, introducing

a phase difference of π, the resonator becomes completely blocked, the laser
generation does not start and the active element can accumulate population
inversion. As the maximum value of population inversion is achieved, the
voltage is switched off, the resonator opens and a giant pulse is emitted.
The size of electro-optic coefficient is determined by physical parameters
of the crystal and its symmetry class. One of the most common crystals
used in these devices is KD2 PO4 (potassium dideuterium phosphate, KD*P)
which has one of the largest electro-optical coefficients (r63 ). A half-wave
voltage for KD*P crystal is ≈ 8 kV for a laser wavelength of 1 µm. In prac­
tice, twice less (4 kV) voltage is applied, since light in the resonator passes
the Pockels cell twice after bouncing the resonator mirror, so the same po­
larization rotation effect as per single pass applying a half-wave voltage is
achieved. Because of that, the Pockels cell is placed close to one of the res­
onator mirrors. Switching on and off 4 kV voltage with a high speed is not
a technically simple task, however, devices and methods enabling control of
high voltage electrical pulses with high repetition rate (kHz) are developed,
and high voltage electrical pulse fronts (which essentially determine the speed
of Q-switching) are steeper than 10 ns. So electro-optical Q-switching is es­
sentially performed by changing the polarization of light: losses are increased
or decreased by analyzing polarization of light. The Q-switching moment is
matched with the time moment when the maximum population inversion is
accumulated, yielding the highest gain coefficient.
There are two types of Pockels effect: longitudinal and trans­
verse. In the former case voltage is applied along the light propagation dir­
ection (as in the case just discussed), while the in the latter case, voltage is
applied transverse to light propagation direction. The geometries of electro-
optical modulators corresponding to these two types of Pockels effect are
illustrated in Fig.71.
5.3 Methods of Q-switching 108

Figure 71: Electro-optical modulator based on (a) longitudinal and (b) transverse
Pockels effect.

The phase difference due to transverse Pockels effect is:

n3 V
[
ωl
∆ϕ = (no − ne ) − o r63 , (140)
c 2 d
where ne is the refractive index for the extraordinary wave, l is the length of
crystal, d is the width of crystal. Unlike the longitudinal Pockels effect, the
polarization rotation due to transverse Pockels effect depends on the crystal
length. It may seem convenient, since the half-wave voltage can be reduced
significantly, however, excitation of strong high frequency acoustic waves can
considerably modulate the transmission of such electro-optic switch in time.

3. Acousto-optic switch. The method of acousto-optic switch relies on

the control of resonator quality by changing propagation direction of light in
a resonator that is equivalent to a misalignment of a resonator. An acousto­
optic switch (modulator) is based on photoelastic effect. The photoelastic
effect is a change of the material density induced by propagating acoustic
wave, which in turn causes modulation of the material refractive index. A
light wave propagating in such medium is diffracted; the situation is equival­
ent to light diffraction from a transparent diffraction grating. In the present
case, the period of the grating corresponds to the wavelength of the acoustic
wave.
The operation principle of an acousto-optic modulator is schematically
depicted in Fig.72. The most commonly used media for acousto-optic mod­
ulation are fused silica (SiO2 ) and tellurium oxide (TeO2 ). These materials
have excellent optical and mechanical properties such as broad transmission
range and sufficiently large photoelastic coefficient. High frequency travel­
ing acoustic waves are generated by applying an alternating voltage at radio
5.3 Methods of Q-switching 109

Figure 72: Principle of operation of an acousto-optic modulator.

frequency to a piezoelectric transducer which is attached at one end of the

crystal. An acoustic wave absorber is attached at the other end of the crys­
tal. Diffraction of a light wave from an acoustic wave is described by Bragg
diffraction condition:
mλ
sin θB = , (141)
2λa
which for small angles and first diffraction order (m = 1) can be rewritten
as θB = λ/2λa , where θB is the angle between the light and acoustic waves
and λa is the wavelength of acoustic wave. In that case the diffraction angle
of a light wave is 2θB .
Q-switching is performed by simply placing an acousto-optic modulator
into the resonator. The control of resonator quality is simple: when altern­
ating voltage at radio frequency is switched on, the light beam changes its
propagation direction due to diffraction and is deflected from the optical axis
of the resonator. Then the resonator becomes misaligned, its quality reduces
to 0 and such laser cannot generate. However, at this time the active element
accumulates population inversion. When the maximum population inversion
is achieved, the radio frequency signal is turned off and the resonator quality
suddenly increases and a gigantic pulse is generated. Acousto-optic modu­
lators can operate at high repetition rates (tens of kHz) and their switching
times are very short, e.g., 5 ns for TeO2 crystal.

To achieve passive Q-switching, losses in the resonator are modulated

using a saturable absorber which is inserted into the resonator. It is a
passive Q-switch since it does not require any source of external control. A
5.3 Methods of Q-switching 110

saturable absorber is essentially a two-level system which absorbs laser radi­

ation very well and has low saturation intensity. The absorption coefficient
of a saturable absorber is a function of intensity:
α0
α= , (142)
1 + I/Is
where α0 is the absorption coefficient for low intensity (when no saturation is
present) and Is is the saturation intensity. When intensity is low, the satur­
able absorber is opaque: it simply absorbs all laser radiation and so induces
significant losses in the resonator. Therefore, the laser cannot generate but
just accumulates the population inversion. When the gain coefficient exceeds
the absorption coefficient, intensity in the resonator rapidly increases, while
the absorber starts to become transparent and losses in the resonator rapidly
decrease. When the absorber saturates, the maximum resonator quality is
achieved. In order to have an efficient and timely switching of the resonator
quality, it is important to match parameters of the pump and saturable ab­
sorber. The lifetime of saturable absorber excited level has to be sufficiently
long, longer than duration of generated pulses, i.e., several tens of ns, so that
after generation has started it would relax slowly and would not introduce
any further unnecessary losses.
Historically, cyanine dyes dissolved in certain solvents (usually methyl or
ethyl alcohols) were widely used as saturable absorbers. Their advantage
is that by adjusting dye molecule concentration, it is easy to establish the
necessary saturation intensity of the absorber. On the other hand, choice of
dyes is very wide, so one can always choose a dye with the desired absorption
and relaxation properties. However, dye solutions have several of drawbacks:
they are toxic, chemically unstable and when irradiated by intense laser light,
they degrade quite fast (in a few months or faster). Another drawback is
strong thermal effects which manifest themselves in the form of a thermal
lens, which dynamically changes resonator stability conditions. Due to this
fact, dye solutions have to be permanently mixed. The next generation
saturable absorbers are based on chromium-doped crystalline media (e.g.,
Cr:YAG) which do not have the aforementioned problems.
Typically, the most efficient in the Q-switching operation are solid-state
lasers, especially those having long lifetimes of their upper laser energy levels,
such Nd;glass, Nd:YAG, Nd:YLF, etc. However, efficient Q-switching was
also demonstrated with gas lasers: CO2 and iodine.
5.4 Mode-locking 111

5.4 Mode-locking
Another laser operation regime is mode-locking. Mode-locking represents
a technique that allows phasing of all frequencies (longitudinal modes)
circulating in the laser resonator. Mode-locking produces the shortest and
the most intense laser pulses with durations varying from several hundred
of ps to < 10 fs (depending on the gain bandwidth of the laser medium
and mode-locking method). The first mode-locked laser was demonstrated
in 196612 . Absolutely all ultrashort pulse lasers employ a certain type of
mode-locking technique. Let us discuss the differences between unphased
and phased (locked) mode laser operation.
Consider a laser medium with inhomogeneously broadened spectral line.
Only longitudinal modes for which the gain is greater than losses are amp­
lified. Assuming sufficiently long resonator, a number of such modes can
be very large and the intermodal distance (the difference of adjacent mode
frequencies) which is defined as
πc
∆ω = ωq+1 − ωq = (143)
n0 l
is much smaller than the laser amplification bandwidth ∆ωL , as schematically
illustrated in Fig.73. Only in this case the discussion on mode-locking is
meaningful.

Figure 73: Intermodal distance and laser amplification bandwidth.

Without any additional means, such laser generates N independent lon­

gitudinal modes whose phases are uncorrelated. The complex amplitude of
the n-th mode can be written as
12
A. J. De Maria, D. A. Stentser, and H. Heynau, Self-mode locking of lasers with
saturable absorbers, Applied Physics Letters 8, 174–176 (1966)
5.4 Mode-locking 112

En (t) = En ei(ωn t+ϕn ) , (144)

where En is amplitude of the mode electric field, ωn is the mode frequency,
ϕn is its phase. The complex amplitude of the total radiation field in the
resonator is a superposition of all N modes with random phases:
N
N −1
E(t) = E0 ei(ωn t+ϕn ) , (145)
n=0

and the intensity is

N
N −1
∗
2
I(t) = |E(t) | = EE = E02 ei(ωn t+ϕn ) e−i(ωn t+ϕn ) = N E02 . (146)
n=0

This result shows that the total radiation intensity is a sum of intensities
of the individual modes: I = N Imode . This is the average value because
a certain number of modes may accidentally have the same phase, but the
average result is exactly as specified. Such (unphased mode) radiation is
generated by free-running and Q-switched lasers.
Now let us consider the situation when all longitudinal resonator modes
are phased, i.e., have the same phase ϕn = ϕ0 . Then the the complex field
amplitude is
N
N −1 N
N −1
i(ωn t+ϕ0 ) iϕ0
E(t) = E0 e = E0 e eiωn t . (147)
n=0 n=0

After performing certain mathematical operations, the intensity can be

expressed as

2
2 ∗ 1 − e−iN ∆ωt 2
2 sin (N ∆ωt/2)
I(t) = |E(t) | = EE = E02 = E0 2 . (148)
1 − e−i∆ωt sin (∆ωt/2)

The latter result suggests that phasing of longitudinal modes induces certain
intensity modulation in time with intensity maxima at
∆ωt
= 0, π, 2π, ..., nπ. (149)
2
It is then easy to determine the temporal separation (period) between the
adjacent intensity maxima:
5.4 Mode-locking 113

2(n + 1)π 2πn 2π 2n0 l

∆t = tn+1 − tn = − = = = tr , (150)
∆ω ∆ω ∆ω c
where tr is the resonator round-trip time.
This means that the light pulses in the resonator are generated and their
repetition period coincides with the resonator round-trip time tr . Let us
estimate the maximum intensity of such a pulse. Let us rewrite expression
in Eq.(148) assuming that ∆ωt/2 = 0:

2 2
2 sin (N ∆ωt/2) N (∆ωt/2)2
I(t)max = lim E0 2 = lim E02 2
= N 2 E02 .
∆ωt/2→0 sin (∆ωt/2) ∆ωt/2→0 (∆ωt/2)
(151)
In the case of mode-locking, intensity of the light pulse is I = N 2 Imode ,
i.e., significantly greater than in the case of modes with random phases.
Therefore, mode-locking produces light pulses with very high intensity. A
comparison of the intensity evolutions I(t) in the cases of unsynchronized
and synchronized modes is shown in Fig.74.

Figure 74: Evolution of light intensity in a resonator in the cases of (a) unsyn­
chronized and (b) synchronized mode (mode-locked) cases. The intensities in (a)
and (b) are not to scale.

Let us now estimate the minimum duration of such pulses:

5.5 Methods of mode-locking 114

Table 3: Relevant parameters of several mode-locked lasers. τi is the minimum

theoretical pulse duration that is inversely proportional to amplification bandwidth
∆νL and τp is the experimentally achieved minimum pulse duration.
Laser medium λ, µm ∆νL , THz τp τi
Nd:YAG 1.064 0.135 5 ps 3.3 ps
Nd:YLF 1.047 0.390 2 ps 1.1 ps
Nd:glass 1.054 8 60 fs 55 fs
Cr:LiSAF 0.85 57 18 fs 8 fs
Ti:sapphire 0.8 100 6 fs 4.4 fs
Rhodamine 6G 0.57 45 27 fs 10 fs

2π tr 2n0 l 1
τp = = = = . (152)
∆ωN N Nc ∆ωL
This result suggests that the more modes are phased, the shorter is the
generated light pulse. Assuming that all the longitudinal modes are phased,
the minimum pulse duration is inversely proportional to the width of laser
medium amplification band. In a real laser, it is possible to synchronize a
major fraction of longitudinal modes, the relevant examples are presented in
Table 3. Note that with several exceptions (e.g., dye or semicoductor lasers)
only solid-state lasers operate in mode-locking regime as only these lasers
provide sufficiently broad laser amplification bandwidth.

5.5 Methods of mode-locking

Up to now we considered mode-locking in the spectral domain, without pay­
ing attention how the mode-locking is performed in a real laser. For that
reason, it is convenient to understand the process of mode-locking in the
time domain. In practice, all the mode-locking methods rely on the con­
trol of resonator losses. The principal layout of a mode-locked laser is very
similar to that of the Q-switched laser, which is schematically is depicted in
Fig.65. However in the case of mode-locking, the modulation depth of the
resonator losses is considerably smaller and the modulation speed is much
faster, matching the resonator round-trip time. Formally, the methods of
mode-locking are classified into active and passive, depending on the way
how the resonator losses are controlled.

Active mode-locking. The term active mode-locking itself means that

mode-locking is performed by active interference into the laser operation us­
ing a certain external signal. Active mode-locking can be performed in several
5.5 Methods of mode-locking 115

ways and their choice depends on specific properties of the laser medium, i.e.,
the gain coefficient and lifetime of the upper laser energy level. According to
the principle of operation, active mode-locking is performed via amplitude
or phase modulation. The amplitude mode-locking relies on periodic modu­
lation of resonator losses, while the phase mode-locking is based on periodic
adjustment of the resonator length. In lasers which have short upper level
lifetime (e.g., dye lasers) active mode-locking is implemented by modulation
of gain coefficient (e.g., by pumping with short pulses from another laser).
Such mode-locking method is called synchronous pumping.
We will now discuss the basic principle of only the most commonly used
active amplitude mode-locking method.

Figure 75: Principle of active amplitude mode-locking.

Active amplitude mode-locking is performed by placing either an electro-

optical or acousto-optical switches into the resonator, which are controlled
by an external signal, whose period exactly matches the resonator round-trip
time (a typical signal frequency is 100 MHz for a 1.5 m long resonator). In
the case of electro-optical modulator, it is impossible to switch its half-wave
voltage at such high repetition rate, so the actual modulation depth of the
introduced losses is small and the desired effect is achieved by introducing
additional steady losses at a certain constant level by applying a bias voltage.
The periodic control signal basically creates a periodic loss modulation and
the light pulse is formed only within the time interval when laser gain exceeds
the resonator losses, as shown in Fig.75. The amplitude modulator is usually
placed very close to one of the resonator mirrors to avoid possible generation
of several pulses at a time and their subsequent competition.
How the introduction of losses makes longitudinal modes synchronized
and how can this process be envisaged in the time domain? If the mod­
ulation period of losses matches the resonator round-trip time, the light
pulse generated at the moment of minimum resonator losses returns back to
5.5 Methods of mode-locking 116

the modulator exactly at the time when resonator losses are reduced again.
Moreover, with each round-trip the pulse shortens due to uneven amplifica­
tion: its peak is amplified more efficiently than its fronts. Mode-locking in
the time domain can be understood in a quite straightforward way: when
the pulse gets shorter, its frequency spectrum broadens and this automat­
ically implies that more resonator longitudinal modes are made to oscillate
in phase. When such process repeats many times, in an ideal case of in­
homogeneous linewidth broadening the minimum pulse duration is defined
as
0.44
tp ≈ . (153)
∆ωL
In the reality, a complete locking of all the longitudinal modes is never
achieved, and the final pulse duration which establishes after many round-
trips is determined by the modulation depth or frequency. In the case of
homogeneous amplification linewidth broadening the minimum pulse dura­
tion can be defined as
0.45
tp ≈ √ , (154)
∆ωL fm
where fm is the modulation frequency of resonator losses. The active mode-
locking critically depends on how accurately the modulation frequency matches
the resonator length (or its round-trip time): even a small mismatch between
these two, e.g., length change due to environmental temperature variations,
mechanical vibration, etc, can cause large instabilities in the mode-locking
process, which eventually can be lost.
Using active mode-locking methods, light pulses with durations ranging
between 1 ns and 50 ps are usually generated. This pulse duration interval is
quite unique, as there are no other methods to generate bandwidth-limited
pulses with such duration. Usually, electro-optical modulators are used in
lasers with pulsed pumping, where the gain coefficient is large, while acousto­
optical modulators are used in lasers with continuous wave pumping where
the gain coefficient is small.

Passive mode-locking. Passive mode-locking is implemented by pla­

cing into the resonator a specific element, which operation is based on intensity-
dependent transmission or reflection. One of such elements is a saturable ab­
sorber, or any other element whose operation is based on a similar principle
and which has fast relaxation time (much faster that the round-trip time of a
resonator). Let us discuss operation of these elements and how they perform
mode-locking in more detail.
5.5 Methods of mode-locking 117

The operation principle of a saturable absorber was explained when we

described Q-switching. The same principle holds also in the case of passive
mode-locking. However, in this case, the saturable absorber has very fast
relaxation time (several ps and shorter, 10−12 s). Fast absorber relaxation
ensures that the light pulse propagating through the absorber gets shorter
with each round-trip: the pulse peak which has the highest intensity, propag­
ates without any losses, while the pulse slopes with lower intensity are par­
tially absorbed, consequently making pulse fronts steeper. This situation is
schematically depicted in Fig.76. Eventually, the established pulse duration
approximately equals to
�
0.79 γ Is
tp ∼
= , (155)
∆νL α Ip
and is determined by saturable absorber parameters, i.e., losses due to ab­
sorption α, gain coefficient γ, saturation intensity Is , pulse peak intensity Ip ,
and gain bandwidth of the laser medium ∆ωL .

Figure 76: Principle of passive mode-locking. tr is the resonator round-trip time.

It is important to understand how passive mode-locking starts, since no

external signal that initiates the process is applied. The laser oscillation
starts from a certain noise background (an example is presented in Fig.74(a))
when many longitudinal modes (falling under the laser medium amplification
contour) having random phases are excited at once. The saturable absorber
absorbs all this ”mode noise” and in the laser medium new ”mode noise”
with random phases is being constantly generated. Sooner or later (this time
interval is of the order of µs) there emerges a fluctuation with sufficiently
high intensity (random combination of modes with the same phase) which
turns the absorber partially transparent and then returns to the laser me­
dium. Here this fluctuation is further amplified and eventually overgrows all
5.5 Methods of mode-locking 118

the remaining noise background. In this way, a single light pulse remains
circulating in the resonator and becomes shorter and shorter with every pass
through the absorber. As a result, its frequency spectrum broadens that is
equivalent to adding more and more longitudinal modes in phase.
We have already mentioned that organic dye saturable absorbers posses
several drawbacks related to their chemical instability (lack of durability).
Furthermore, they are liquid and have to be constantly mixed to avoid un­
desired thermal effects. Solid-state saturable absorbers (such as Cr:YAG)
also have drawbacks related to relatively long relaxation time and limited op­
erating wavelength range, so they are practically used only for Q-switching.
In 1992 qualitatively new type of saturable absorbers based on semiconductor
media was proposed 13 . These saturable absorbers were called SESAM
(Semiconductor Saturable Absorber Mirror). Unlike organic dye ab­
sorbers which are based on saturable transmission, semiconductor absorbers
are based on saturable reflection (Fig.77).

Figure 77: Examples of SESAM mirror reflectivity as a function of laser pulse

fluence.

SESAM consists of a semiconductor layer (e.g., AlAs/AlGaAs) with an intensity-

dependent reflection coefficient and serves as one of the resonator mirrors.
By choosing various combinations of semiconductor materials, such absorber
can operate in a relatively broad spectral range (limited by the near IR range,
however) and its relaxation time is very fast, in the order of 100 fs, which
makes it possible to generate very short light pulses. Although SESAM is
chemically stable, its main drawbacks are relatively small modulation depth
(the intensity-dependent reflection coefficient may vary only within ∼ 5%
13
U. Keller et al, Semiconductor saturable absorber mirrors (SESAMs) for femtosecond
to nanosecond pulse generation in solid state lasers, IEEE Journal of Selected Topics in
Quantum Electronics 2, 435–453 (1996)
5.5 Methods of mode-locking 119

range) and low saturation intensity. Therefore, the use of SESAM is lim­
ited to very low pulse energies; such pulses are generated by pulsed lasers
that employ continuous wave pumping (quasi-cw lasers). With high energy
pulses, the coating of the absorber degrades in time as is eventually optically
damaged.
Another method of passive mode-locking is based on the optical Kerr
effect in transparent materials, which results in the intensity-dependent re­
fractive index, which is expressed as

n = n0 + n2 I , (156)
where n2 is the nonlinear index of refraction. For all solid-state media that
are transparent in the visible and IR range, this quantity is in the order
of 10−16 cm2 /W. The typical intensities present in laser resonators induce a
refractive index change of the order of 10−5 −10−6 , and this is enough that an
intense light beam starts to self-focus when propagating in such a medium.
Self-focusing occurs because beam in the resonator has intensity distribution
I(r) described by the Gaussian function: beam parts with different intensities
acquire different phase shift, which is the largest at the center of the beam:
2πn2 ln
δϕ(r) = I(r), (157)
λ
where ln is the length of nonlinear medium. Due to that intensity-dependent
phase shift, the initially flat phase front of the beam bends as if the beam
passes through a convex lens and the beam starts to self-focus, as shown in
Fig.78.

Figure 78: Bending of the phase front of intense Gaussian beam due to intensity-
dependent phase shift which causes self-focusing.

The lens induced by nonlinear medium is called Kerr lens and its focal
length
5.5 Methods of mode-locking 120

1 4n2 ln
= Ip (158)
fn n0 w2
is a function of the peak intensity Ip . Here w is the radius of a Gaussian
beam.

Figure 79: Principle of Kerr lens operation.

Now let us discuss how the self-focusing effect can be exploited for mode-
locking. The method is called Kerr lens mode-locking (KLM) and its prin­
ciple of operation is schematically depicted in Fig.79. A nonlinear medium
is placed in the laser resonator together with a diaphragm. The diameter of
the diaphragm is chosen such that only a high intensity beam that carries a
very short light pulse passes the diaphragm without any losses. Only a very
short (mode-locked) light pulse has sufficient intensity to induce a nonlinear
lens, while a noise-like radiation with low intensity experiences large losses.
Hence a Kerr lens operates similarly to a saturable absorber with very fast
relaxation time, which is only few fs (10−15 s), and which is determined by
the response time of electron cloud to the electric field of light. Kerr lens
can be induced in any transparent material, since all media possess electronic
nonlinearity. Moreover, this method is suitable to any wavelength and is non-
resonant: the losses are created without absorption. A small problem is that
often the initial fluctuations appearing in the laser resonator have insufficient
intensity to initialize Kerr lens mode-locking. Then the laser generation is
triggered by an external perturbation e.g., by gently hitting the resonator
mirror mount. Of course, KLM can only be applied in the case of quasi-cw
laser and would be difficult to implement in pulsed pumping regime.
Kerr lens mode-locking was first demonstrated in 1991 by generating
femtosecond pulses from Ti:sapphire oscillator14 . Interestingly, in that con­
figuration, sapphire crystal served itself as a nonlinear medium, while the
14
D. E. Spence, P. N. Kean, and W. Sibbett, 60-fsec pulse generation from a self-mode­
locked Ti:sapphire laser, Optics Letters 16, 42–44 (1991)
5.5 Methods of mode-locking 121

focused pump from an argon ion laser served as a diaphragm. Mode-locking

operation was achieved by simply aligning the laser resonator in a certain
way. Such mode-locking method was termed self-mode-locking since no spe­
cific element directly modulating losses was used. Thanks to this discovery,
Ti:sapphire quickly became the most reliable and widely used femtosecond
laser, which outperformed the whole generation of dye lasers, which were
the only available femtosecond lasers at that time. At present, Kerr lens
mode-locking is used in many modern solid-state laser oscillators.
In conclusion, we note that since passive mode-locking starts from random
fluctuations, the time moment when the sufficient fluctuation appears in the
resonator and a pulse is formed is quite undefined. For quasi-cw lasers the
exact start time of mode-locking is not important: when the laser is turned
on and once mode-locking starts, it will continue as long as the laser keeps
running. However, for lasers with pulsed pumping (such as Nd:glass and
Nd:YAG) the start time of mode-locking is important, as mode-locking has
to restart with every pump pulse. For this purpose, along passive mode-
locking, active mode-locking is employed as well, which forms the initial
light pulse (although quite long), which is capable to initiate the process of
passive mode-locking at rather well-defined moment of time. Such mode-
locking method is called hybrid mode-locking.

Summary of Chapter 5
• A unique feature distinguishing lasers from other light sources is
that they can emit light not only in the form of a low divergence
beam but also in the form of very short pulses.
• There are three types of pulsed laser operation regimes: free-
running, Q-switching and mode-locking.
• Free-running is the simplest laser operation regime, which starts
when the laser oscillation threshold is exceeded and produces µs
pulses with kW peak power.
• Q-switching regime is achieved via control of resonator losses and
produces ns pulses with a peak power up to hundreds of MW.
• Mode-locking regime is achieved when longitudinal modes in the
laser resonator are made to oscillate (locked) in phase. Mode-
locking produces ultrashort pulses with durations ranging form
100 picoseconds to a few femtoseconds and peak powers of up to
several hundreds of MW.
6 Generation and amplification of ultrashort light pulses 122

6 Generation and amplification of ultrashort

light pulses
The shortest (the so-called, ultrashort) light pulses are generated only by
employing passive mode-locking. Although passive mode-locking could be
implemented in many solid-state lasers, the shortest light pulses are gen­
erated only in laser media which possess the broadest amplification
(gain) bandwidth. In that regard, there are two unique femtosecond lasers
to note: dye lasers and Ti:sapphire lasers, which have the broadest gain
bandwidth and which served as main femtosecond laser sources at different
decades15 . The minimum pulse duration for both of these lasers has been
reduced by more than two orders of magnitude (from several ps by the end
of 1960s to tens of fs in 1980s) as a result of development of mode-locking
techniques. It is important to mention that the shortest pulse durations are
achieved by additionally compressing the pulses outside the resonator – the
method is generally termed the extracavity compression.
In this chapter we will discuss the main problems related to ultrashort
pulse generation, amplification and techniques how such ultrashort pulse dur­
ation is measured.

6.1 Dispersion control in the resonator

One of the main problems facing the generation of ultrashort light pulses
is their dispersive broadening in a resonator due to very broad frequency
spectrum. There is always at least one optical element in the resonator (i.e.,
the laser medium) in which dispersive effects occur: longitudinal modes with
different frequencies propagate at different phase velocities in the dispersive
medium, which causes uneven phase shifts between adjacent modes, so the
pulse broadens in time and becomes chirped (phase modulated). This prob­
lem is important for all ultrashort pulses as dispersion is a cumulative
effect: the phase shifts accumulate with each resonator round-trip. The
shorter is the pulse, the broader is its frequency spectrum and the
more pronounced is the dispersion effect. For example, it is estimated
that in Ti:sapphire laser resonator a light pulse with a duration of 10 fs can
broaden up to several times during just a single round-trip. Naturally, due
to temporal broadening, the intensity of a pulse decreases, and as a result,
the process of mode-locking could be disturbed or completely lost. To avoid
this, dispersion must be compensated by placing a specific element in the
15
U. Keller, Recent developments in compact ultrafast lasers, Nature 424, 831–838
(2003)
6.1 Dispersion control in the resonator 123

resonator. Let us discuss what these elements are and how they work.
The group velocity dispersion of transparent dielectric media in the vis­
ible and near-IR spectral range is normal: the long-wavelength (”red”) spec­
tral components propagate faster than the short-wavelength (”blue”) spectral
components and therefore the pulse becomes chirped (phase modulated).
A chirped pulse is ”colored”: its leading front is ”red” and its trailing front
is ”blue”, and the frequency across the pulse changes linearly (the phase
changes quadratically). The pulse duration during propagation changes ac­
cording to
�
z2
tp = t0 1+ , (159)
L2d

where t0 is the initial pulse duration, z is the propagation distance and Ld is

the dispersive broadening length, which is expressed as Ld = t20 /|2g0 |, where
∂2k
g0 = ∂ω 2 is the group velocity dispersion coefficient. g0 > 0 denotes the

normal group velocity dispersion, whereas g0 < 0 denotes the anomalous

group velocity dispersion. The dispersive broadening length is proportional
to the pulse duration squared, and its physical meaning could be deduced
from the case when z = Ld : it is√the propagation distance when the pulse
duration increases by a factor of 2. The phase modulation occurring due
to dispersive broadening is expressed as

d2 ϕ λ3 L ∂ 2 n
= , (160)
dω 2 2πc2 ∂λ2
where L is the length of the dispersive medium and n(λ) is the dispersion
law of the refractive index.
Compensation of dispersion requires delaying the spectral com­
ponents of the light pulse that propagate faster with respect to
spectral components that propagate slower. There are several ways to
do this.
It would be very simple if one could introduce into the resonator an
optical element, which has anomalous group velocity dispersion
(g0 < 0), where the short-wavelength (”blue”) spectral components propag­
ate faster than long-wavelength (”red”) spectral components. Unfortunately,
the range of anomalous group velocity dispersion in transparent dielectric
materials lies close to the mid-IR region, e.g., in glass or sapphire, the dis­
persion is anomalous for wavelengths longer than 1.3 µm, while most of the
solid-state lasers emit at shorter wavelengths.
A solution was found by using a pair of prisms, where anomalous group
velocity dispersion occurs due to optical path difference between distinct
6.1 Dispersion control in the resonator 124

spectral components, as illustrated in Fig.80. In this configuration, the op­

tical path of the long-wavelength spectral components is always longer than
that of the short-wavelength spectral components, so propagation through a
prism pair is equivalent to propagation through a medium featuring anom­
alous group velocity dispersion. The prism pair yields the following phase
modulation:
2
d2 ϕ lω0 dα
2
≈− ω0
, (161)
dω c dω
where l is the distance between prisms and the last term of the product
describes the angular dispersion of a prism.

Figure 80: Anomalous group velocity dispersion created in a pair of prisms.

The amount of introduced anomalous dispersion can be easily adjusted

by either varying the distance between prisms or by simply translating one of
the prisms perpendicularly with respect to the beam propagation direction.
In such a way dispersive broadening of the pulse in other resonator elements
can be fully compensated. In fact, all lasers producing light pulses shorter
than 0.5 ps employ dispersion compensating element in the resonator. A
typical layout of Ti:sapphire laser with dispersion compensation is depicted
in Fig.81

Figure 81: Layout of Ti:sapphire laser resonator with dispersion compensation.

6.1 Dispersion control in the resonator 125

Note that after passing the prism pair, the laser beam becomes ”colored”,
as its colors become separated in space due to different refraction angles.
The coloration of the beam is cancelled after back reflection from the mirror
(providing yet another pass), which also doubles the temporal dispersion of
the setup.
If the pulse duration is very short (consequently, its frequency spectrum
is very broad), dispersion cannot be fully compensated using a pair of prisms
because of higher-order dispersion effects (related to higher order derivat­
ives: dϕ3 /dω 3 , etc.) that become significant. The effects related to higher
order dispersion manifest themselves as nonlinear distortions of the pulse
chirp preventing optimal compression of the pulse. In that case, optimal
dispersion compensation (pulse compression) is achieved using the so-called
chirped mirrors16 . Chirped mirror is a mirror with a specific dielectric
coating which consists of e.g. alternating SiO2 and TiO2 layers of variable
thickness, deposed in such a way that light penetration depth (where it is
reflected) depends on its wavelength, as illustrated in Fig.82.

Figure 82: Structure and operation principle of a chirped mirror.

The long-wavelength (”red”) spectral components of the pulse (which

are faster) are reflected from a deeper layer of the coating, while the short-
wavelength (”blue”) spectral components are reflected from a layer close to
the surface, so the group velocity dispersion is compensated. By selecting
an appropriate layer thickness and positioning even a very complex law of
group velocity dispersion can be fully compensated. Chirped mirrors have
only one drawback: when the dielectric layers are deposited their dispersion
and its sign cannot be adjusted.
One of the most flexible dispersion compensation techniques employs the
combination of a prism pair and chirped mirrors: the prism pair is
used to compensate the second order (group velocity) dispersion and the
chirped mirrors are used to compensate higher order dispersion.
16
N. Matuschek, F. X. Kartner, and U. Keller, Theory of double-chirped mirrors, IEEE
Journal of Selected Topics in Quantum Electronics 4, 197–210 (1998)
6.1 Dispersion control in the resonator 126

Kerr lens mode-locking enables generation of the shortest possible pulses

whose duration is comparable to a single optical cycle17 . An optical cycle is
a time interval corresponding to a period of a single oscillation of the electric
field. The duration of an optical cycle is defined as
λ0
τc =
, (162)
c
where λ0 is the central (carrier) wavelength. It is easy to calculate that one
optical cycle at λ0 = 300 nm equals to τc = 1 fs, at λ0 = 600 nm, τc = 2 fs
and so on. For the central wavelength of Ti:sapphire laser, λ0 = 800 nm,
one optical cycle is τc ≃ 2.7 fs. Spectral bandwidth which corresponds to a
bandwidth-limited single optical cycle pulse can also be estimated:
c
∆λ = . (163)
ν0 ± ∆ν/2
In the visible range (e.g., λ0 = 550 nm) the spectrum of a single optical cycle
pulse will extend through the entire range and could be regarded as white
light. In that regard we can view such laser sources as white light lasers.
In some cases, laser pulses can be compressed outside the resonator by ap­
plying the above discussed dispersion compensation techniques. The shortest
optical pulses to date were generated by employing the extracavity com­
pression technique. It is based on increasing the spectral bandwidth via
self-phase modulation (SPM) in a nonlinear medium and subsequent removal
of the frequency modulation by using an appropriate dispersive delay line.
This constitutes a simple and robust method for obtaining few optical cycle
pulses whose spectral widths extend well beyond the gain bandwidth suppor­
ted by the ultrashort pulse lasers. Pulse compression based on SPM-induced
spectral broadening in a bulk solid-state medium offers the advantages of a
wide variety of suitable nonlinear materials, technical simplicity, low cost,
and easy implementation to virtually any existing ultrashort pulse laser sys­
tem. For instance, this pulse compression technique attracts an increased
practical interest, in particular due to its applicability to novel Yb-based
lasers, which produce either relatively long (few hundreds of fs) femtosecond
or sub-picosecond pulses with very high average power.
17
G. Steimeyer, D. H. Suter, L. Gallmann, N. Matouschek, and U. Keller, Frontiers in
ultrashort pulse generation: Pushing the limits in linear and nonlinear optics, Science 286,
1507–1512 (1999)
6.2 Measurement of ultrashort laser pulses 127

6.2 Measurement of ultrashort laser pulses

Measurement of the laser pulse duration is a simple task if the pulse duration
is long (several ns or longer). The duration of free-running and Q-switched
laser pulses is easy to measure using a sufficiently fast photodetector and an
oscilloscope. However, this method is unsuitable in the case of ultrashort
laser pulses because of insufficient temporal resolution of electronic detection
components, i.e., the detector response time is significantly longer than the
pulse duration. Therefore, ultrashort pulse duration measurement techniques
are essentially different and are based on the measurement of correlation
functions. One of the most common used correlation techniques is meas­
urement of the second order (intensity) correlation function:
∞
IAC (τ ) = I(t)I(t − τ )dt, (164)
−∞

where I(t) is a function describing temporal envelope (profile) of the pulse,

I(t − τ ) is the same temporal envelope (its replica) shifted in time by τ .
By changing the temporal interval τ (i.e., the delay between the original
pulse and its replica), the entire correlation function, which is called the
autocorrelation function, is measured. It is obvious that IAC (τ ) = IAC (−τ ),
i.e., the autocorrelation function has a maximum at τ = 0 and is always
symmetrical regardless of the shape of the measured pulse.
In practice, the autocorrelation function of the light pulse is measured
using an optical setup which is in principle identical to Michelson interfero­
meter setup, as depicted in Fig.83. The incident light beam is divided into
two identical beams (carrying identical light pulses) by means of a beams­
plitter. One arm of the interferometer is kept fixed, while the other arm is
translated (by means of mechanical translation stage) and in this way tem­
poral delay between the pulse and its replica is changed. Thereafter both
beams are focused by a lens into a nonlinear crystal where second harmonic
is generated. Since the intensity of the second harmonic radiation is propor­
tional to the product of intensities of the pulse and its replica:

I2ω (t, τ ) ∝ I(t)I(t − τ ) = IAC (t, τ ), (165)

by changing the delay τ between them and measuring the intensity of the
second harmonic, we essentially record the autocorrelation function of the
pulse. Moreover, the photodetector performs integration in time, since its
response time is significantly longer than the duration of the second har­
monic pulse, so the time variable t is eliminated and the measured function
is described by the expression in Eq.(164).
6.2 Measurement of ultrashort laser pulses 128

Figure 83: Setup for measurement of the autocorrelation function. PD is the beam
splitter, L is the lens, NK is the nonlinear crystal for second harmonic generation,
D is the diaphragm and FD is the photodetector. The inset depicts measured
autocorrelation function of an ultrashort pulse.

An example of the autocorrelation function is shown in the inset of Fig.83.

In order to increase the accuracy of measurement, tens to hundreds autocor­
relation function values are measured at each delay, which thereafter are
averaged. To further increase the measurement accuracy, each point in the
autocorrelation function is normalized to the intensity of the incident pulse,
so eliminating pulse-to-pulse intensity fluctuations from the laser source. The
step of time delay variation is chosen to provide sufficient time resolution
for the measurement. Modern computer-controlled mechanical translation
stages ensure high accuracy, so the delay between the pulse and its replica
can be varied with an accuracy of 1 µm, i.e., 3 fs. Once the autocorrela­
tion function is measured, it is important to know how the measured dur­
ation of the autocorrelation function τAC is related to the duration of the
pulse τp . This relation depends on the shape of the measured pulse. The
autocorrelation
√ function of a Gaussian pulse also has a Gaussian shape, so
τAC /τp = 2 = 1.414. For optical pulses whose shapes are different from
Gaussian, the ratio is different as well, e.g., for pulses with squared hyper­
bolic secant (sech2 ) shape this ratio is τAC /τp = 1.543.
The above discussed method is well suited for measurement of autocor­
relation functions of pulses at high repetition rate. However, if the pulse
6.2 Measurement of ultrashort laser pulses 129

Figure 84: Setup for measurement of the autocorrelation function in a single-shot.

PD is the beam splitter, NK is the nonlinear crystal for second harmonic generation
and CCD is the CCD camera. The bottom picture depicts how the overlap zone
of beams carrying the ultrashort pulses is transformed into the autocorrelation
function.

repetition rate is low, measurement of the autocorrelation function (data ac­

cumulation and averaging) can take a relatively long time. Therefore, in that
case a slightly different method is used which is based on the measurement of
autocorrelation function in a single laser shot (i.e., single pulse). The layout
of a single-shot autocorrelator is depicted in Fig.84. The light beams in a
nonlinear crystal are crossed at a certain angle Φ and their longitudinal di­
mensions correspond to the temporal intensity distribution of the pulse I(t).
The beam and pulse intersection zone in the nonlinear crystal has the shape
of a rhombus and only in this zone the second harmonic is generated. The
second harmonic signal has a certain spatial intensity distribution which is
recorded by the CCD camera. From the intersection geometry it is easy to
estimate that spatial and temporal coordinates are related as follows
n 0 x0 Φ
τ= sin , (166)
c 2
where n0 is the refractive index of the nonlinear crystal. The spatial intens­
ity distribution of the second harmonic corresponds to the autocorrelation
function
6.2 Measurement of ultrashort laser pulses 130

∞
IAC (x0 ) = I(x)I(x − x0 )dx. (167)
−∞

The temporal delay in such setup is calibrated very simply: when one
of the pulses is delayed, the image of autocorrelation function on the x axis
shifts accordingly. Since the temporal autocorrelation function is converted
into spatial intensity distribution, the measurement accuracy also depends
on the beam diameter. In this case it is chosen such that it would be much
greater than longitudinal pulse parameters (its duration).
The intensity autocorrelation function has one significant advantage. This
method background free: if the pulses do not overlap in time, the measured
signal is zero. On the other hand, since the autocorrelation function is always
symmetrical even when the intensity profile of the measured pulse is not, or
the pulse shape is complex (i.e. containing several intensity peaks, sub-
pulses, etc.), this kind of information is not retrieved. This shortcoming is
circumvented by the measurement of cross-correlation function. The cross-
correlation function is produced in a similar manner as the autocorrelation
function, just instead of the pulse replica, a pulse with a well-known known
shape (usually much shorter as well), which is called the probe pulse, is used:
∞
ICC (τ ) = I(t)Iz (t − τ )dt, (168)
−∞

where Iz (t) is the intensity of probe pulse. It is obvious that the cross-
correlation function is not symmetrical and the cross-correlation technique
is often applied for the measurement of complex pulses.
The temporal information can also be extracted by measuring higher-
order, e.g., third order correlation function:
∞
I (3) (τ ) = I(t)2 I(t − τ )dt. (169)
−∞

Third-order correlation function is obtained by making use of third-order

nonlinear processes, which take place in media with cubic nonlinearity. These
media are isotropic materials, e.g., fused silica. It is easy to notice that
in the case of third-order autocorrelation function I (3) (τ ) =
̸ I (3) (−τ ), the
information on the direction of the time axis is maintained. It is important
that the wavelength of the third-order autocorrelation signal is the same
as that of the incident radiation. Third-order autocorrelation functions are
measured when wavelength of radiation is short (e.g., in the UV spectral
region) and where usual second-order correlation methods cannot be applied
for some reason.
6.3 Amplification of ultrashort light pulses 131

Very often, besides the duration and shape of the ultrashort pulse, the
knowledge of its spectral and temporal phase is required, especially concern­
ing optimization of pulse compression setups. Measurement methods that
enable phase retrieval are based on the measurement of instantaneous fre­
quency of the autocorrelation function. The measurement setup is basically
identical to that shown in Fig.83, only in the present case instead of a pho­
todetector a spectrometer is used. These methods are called FROG; the
abbreviation stands for frequency-resolved optical gating. The obtained res­
ults (the so-called frogograms) are processed using mathematical algorithms
which are able to retrieve the actual shape and instantaneous phase of the
pulse.

6.3 Amplification of ultrashort light pulses

Amplification of ultrashort light pulses is one of the most important tasks in
laser physics. Mode-locking technique enables the generation of very short
light pulses, but their energy is low (varies from several nJ to several µJ) and
the peak power does not exceed tens of MW. To increase the peak power up
to GW, TW or even PW level, laser amplifiers are used.
All modern laser systems which produce ultrashort pulses with high peak
power consist of: 1) a laser oscillator which operates in the mode-locking
regime and generates very short light pulses with low energy; 2) a laser amp­
lifier, whose purpose is to increase the pulse energy (and the peak power)
as much as possible without distorting its temporal and spatial characterist­
ics. Here we also note that there is a principal difference between the pulse
repetition rate of the oscillator and the amplifier. In a quasi-cw oscillator,
the pulse circulates in a resonator and repeats itself after each round-trip,
so at the output of the oscillator a train of pulses with a period inversely
proportional to resonator round-trip time (a repetition rate of tens on MHz)
is obtained. The amplifier usually amplifies only one pulse from the sequence
(e.g., every 100th or every 1000th), so the pulse repetition rate at the output
of the amplifier is determined by the repetition rate of amplifier pump source,
thermal effects in the amplifier medium and the ability to electronically con­
trol operation of the amplifier.
The purpose of the amplifier is to increase energy of the ul­
trashort pulse without introducing any distortion to the pulse and
beam. The main factors limiting efficient amplification of ultrashort pulses
are attributed to self-action phenomena, which may occur in the optical
elements of the amplifier or in the amplifying medium itself. Self-action phe­
nomena arise from the optical Kerr effect, i.e. the intensity-dependent re­
fractive index of transparent materials: self-phase modulation, which causes
6.3 Amplification of ultrashort light pulses 132

spectral broadening of the pulse and distortion of its phase characteristics

(bandwidth-limited pulse becomes chirped) and self-focusing of the beam,
which causes uncontrolled decrease of the beam diameter and increase of the
intensity, which eventually leads to damage of amplifier components.

Figure 85: The layout of a two-stage single-pass laser amplifier. T1 and T2 are
telescopes which increase the size of the beam.

The simplest amplifier is a single-pass (sometimes called linear) ampli­

fier which is schematically depicted in Fig.85. The gain per single pass of
such amplifier is not large, and varies from several to several tens of times,
depending on the amplifier active medium. Since the intensity is inversely
proportional to the beam area, the beam dimensions must be enlarged be­
fore each amplification stage as to keep the intensity below the threshold for
self-action effects. It is obvious that in order to effectively increase pulse en­
ergy, the size of the amplifier medium in each successive stage increases very
quickly. This causes new problems: how to achieve homogeneous population
inversion within the entire volume of the amplifier medium, at the same time
avoiding thermal effects due to its limited thermal conductivity. As a result,
the repetition rate of the amplifier drops significantly: such amplifier systems
can only operate at very low (from several to several tens of Hz) repetition
rates.
The physical limits of such amplifier setups were reached around 1985,
and further increase of the amplified pulse energy became hardly possible.
These fundamental difficulties in short pulse amplification were circumvented
by the invention of Chirped Pulse Amplification (CPA) technique in
1985 by D. Strickland and G. Mourou in 198518 . The principle of CPA is
depicted in Fig.86. The idea of the CPA concept is to boost the energy of an
ultrashort pulse, while avoiding very high fluence in the laser amplifier. In
contrast to the above discussed approach, this is done in an elegant way: the
ultrashort, broadband pulse is first stretched (chirped) in time without the
18
G. Mourou and D. Strickland, Compression of amplified chirped pulses, Optics Com­
munications 56, 219–221 (1985)
6.3 Amplification of ultrashort light pulses 133

loss of its spectral content by the use of dispersive elements. The stretching
factor may vary from hundreds to thousands of times, in this way significantly
reducing the pulse intensity. The long chirped pulse is then amplified in a
laser amplifier by a factor of 104 − 106 and then recompressed to the original
duration by dispersive elements, which introduce an opposite dispersion as
the stretched does. The CPA technique solved the long-standing problem of
safe and efficient amplification of ultrashort optical pulses without the onset
of optical damage of the amplifier material and other optical components,
enabling a tremendous leap in the peak power and intensity of laser pulses,
boosting an exciting progress in ultrafast laser technology. At present, all
modern ultrashort pulse laser systems are based on this principle19 , while the
inventors of this technique were awarded the Nobel Prize in Physics in 2018.

Figure 86: The principal layout of chirped pulse amplification.

After the advent of the CPA technique, construction of the amplifiers

themselves was also improved since stretching of the pulse in time and the
reduction of its intensity eliminated the need to enlarge beam dimensions
each amplification stage. Currently, two configurations of laser amplifiers
are in use: multi-pass and regenerative.

Multi-pass amplifier. A multi-pass amplifier is essentially a ”folded”

single-pass amplifier, its layout is schematically depicted in Fig. 87. The
multi-pass amplification is very efficient when the gain coefficient of the laser
medium is large and the beam area is relatively small. This setup perfectly
fits Ti:sapphire, which provides large amplification per single-pass and which
makes laser pump of the active medium easy to implement in a small volume.
19
S. Backus, C. G. Durfee III, M. M. Murnane, and H. C. Kapteyn, High power ultrafast
lasers, Review of Scientific Instruments 69, 1207–1223 (1998)
6.3 Amplification of ultrashort light pulses 134

The number of passes through the active medium is adjusted with the mirrors
and the main advantage of such amplifier is simplicity of the setup which does
not require any complex control elements. However, the mechanical stability
of such amplifier system is low: even a small misalignment of mirrors signi­
ficantly alters direction the light beam and small beam distortions occurring
after each pass only increase.

Figure 87: Layout of the multi-pass amplifier.

Regenerative amplifier.
The setup of a regenerative amplifier (Fig. 88) is very similar to that
of a Q-switched laser. Here the electro-optic switch (the Pockels cell and
polarization analyzer) performs functions of coupling the seed pulse (the
stretched oscillator pulse) into the resonator and dumping the pulse after
amplification.
The principle of operation of the regenerative amplifier is as follows. Let
us assume that initially the resonator is closed (λ/4 voltage is applied to the
Pockels cell). Then the amplifier medium accumulates population inversion.
When the seed pulse arrives, it is reflected from the polarizer and passes
the Pockels cell which turns its polarization into circular. After reflection
from the mirror and second pass through the Pockels cell, pulse polarization
becomes perpendicular with respect to the initial polarization, therefore, it
passes the polarizer and is coupled into the resonator.
At this moment the voltage applied to the Pockels cell is turned off and
the resonator quality suddenly increases, making the pulse ”trapped” and
amplified in the resonator. The pulse makes several tens to several hundreds
of round-trips in the resonator to reach the gain saturation. When the gain
saturates, the λ/4 voltage is again applied to the Pockels cell which rotates
the polarization by 90◦ (note a double pass) and the amplified pulse is coupled
out (dumped). It is obvious that the regenerative amplifier operates as a Q-
switched laser when there is no seed pulse, but the resonator is opened.
6.4 Pulse stretchers and compressors 135

The main advantage of the regenerative amplifier is its resonator which

ensures very high spatial quality of the amplified beam due to diffraction
losses. Any spatial distortions occurring during amplification or present in
the seed signal are filtered out. Precise timing between the pulse seeding and
dumping moments and the pump pulse is adjusted by the electronic control
system.

Figure 88: Layout of the regenerative amplifier. PE is the Pockels cell, PA is the
polarization analyzer.

6.4 Pulse stretchers and compressors

Pulse stretchers and compressors are very important parts of the CPA-based
laser system. Their purpose is to stretch and compress the light pulse in
time without changing its spectral bandwidth. As discussed in the previous
chapter, these manipulations could be performed using dispersive elements
(prism pair and chirped mirrors). CPA requires very large pulse stretching
and compression factors (300÷10000), so for this purpose a diffraction grating
pair is used which yields very high group velocity dispersion.
Let us briefly discuss the performance of pulse stretchers and compressors.
Let us first discuss operation of a pulse compressor. A pair of two paral­
lel identical diffraction gratings, like a pair of dispersive prisms, introduces
anomalous group velocity dispersion through the difference of optical paths.
Due to fundamental diffraction law, angles for the short-wavelength radiation
are smaller, so the ”blue” spectral components travel a shorter distance than
the ”red” spectral components, as shown in Fig. 89, so the grating pair com­
pressor introduces large anomalous group velocity dispersion.
The difference in optical paths between distinct frequency components
can be estimated according to the diffraction grating equation:
λ
sin γ + sin θ =, (170)
d
where γ and θ are the incidence and diffraction angles, respectively and d
is the period of the diffraction grating. The optical path of any frequency
6.4 Pulse stretchers and compressors 136

Figure 89: Principal setup of a diffraction grating pulse compressor. G1, G2 are
the diffraction gratings, V is the mirror.

component can be estimated as

Lg
P = 1 + cos(γ − θ) , (171)
cos θ
where Lg is the distance between the gratings. The optical path difference
between the marginal spectral components of the pulse determines the tem­
poral delay between them ∆t = ∆P/c and so the compression factor. In
analogy with a prism pair compressor, the light beam is reflected back by a
plane mirror (see Fig. 89) to pass the grating compressor for a second time
to eliminate the spatial chirp (coloration of the beam). As a result, the time
delay between the marginal spectral components (the compression factor) is
doubled (2∆t). Here we note that the compression factor of grating com­
pressor depends on the spectral bandwidth of the pulse, and not on its initial
duration.
The same considerations apply to the pulse stretcher. However, the
pulse stretcher must introduce the group velocity dispersion with an opposite
sign as compared to pulse compressor, implying that the configuration of the
pulse stretcher is different from that of the pulse compressor. Let us discuss
how the pulse stretcher works. First of all, a pulse stretcher must provide
large normal group velocity dispersion. This is determined also by the fact
that the group velocity dispersion of the laser medium and other optical
elements in the amplifier is also normal, so it adds to the present dispersion.
The simplest way to stretch the pulse in this way is to make a pulse propagate
in a long dispersive medium, e.g. optical fiber. Indeed this approach is used,
however it is feasible only in the case of the pulses with very low peak power as
to avoid accumulation of nonlinear effects during propagation in the optical
fiber.
6.4 Pulse stretchers and compressors 137

A considerably more flexible approach is to use a pair of diffraction

gratings with a telescope in between, as shown in Fig.90.

Figure 90: Layout of a grating pulse stretcher. G1, G2 are the diffraction gratings,
L1, L2 are the telescope lenses, V is the mirror.

If both gratings are placed exactly at the focal planes of the lenses (s = f ),
group velocity dispersion of such system is zero, as the optical paths of all
spectral components are equal, and pulse duration does not change when
passing through such arrangement. Moving the gratings closer to the lenses
(as depicted in Fig.90), the group velocity dispersion of the system becomes
normal: the optical path of ”red” spectral components becomes shorter than
that of the ”blue” spectral components. The closer to the lenses are the
gratings, the larger is the introduced normal group velocity dispersion. The
effective length of such pulse stretcher is expressed as:

Leff = −2(f − s), (172)

where s is the distance between the grating and the closest lens. A reflect­
ing mirror plays the same role as in the compressor: the second pass through
the pulse stretcher eliminates spatial distribution of the spectral components
(spatial chirp) and doubles the dispersion. If the distance between the grat­
ing and lens is made longer than the focal length of the lens (s > f ), the setup
introduces anomalous group velocity dispersion, and the stretcher works as
a simple grating compressor. The lengths of pulse stretcher and compressor
are adjusted so that the latter also compensates the amount of normal group
velocity dispersion accumulated by the propagation in the amplifier optics.
The optical path lengths of grating stretchers and compressors in ul­
trashort pulse laser systems are of the order of a meter and typical grating
periods are d = 1200 ÷ 1800 mm−1 . The diffraction gratings are aligned
in such a way that incidence and reflection angles for the central frequency
(wavelength) of the pulse are equal: γ = θ(λ0 ). This angle is termed the
6.5 Laser amplifier media 138

autocollimation (or Littrow) angle. For this angle a specific profile of

diffraction grating blaze is used to achieve the maximum diffraction efficiency
(a fraction of the incident energy diffracted to the first diffraction maximum).
In the near-IR spectral range the diffraction efficiency of aluminum-coated
diffraction grating is up to 90%, for gold-coated grating it is up to 95% and
gratings with special dielectric coatings have up to 98% diffraction efficiency.
The diffraction efficiency of the compressor is very important, for example if
the grating reflects 90% of the incident energy to the first diffraction max­
imum, then after four reflections in the compressor only 66% of amplified
pulse energy remains (0.94 ≈ 0.66), while the increase of reflection up to
98% yields the throughput increase up to 92%. For the pulse stretcher this
parameter is of much less importance, since this kind of energy losses can be
easily compensated during amplification.

6.5 Laser amplifier media

There are strict requirements for laser media which can be used to amplify
ultrashort pulses:

• One of the most important requirements is high saturation fluence of

the laser amplifier medium, which is defined as

hν
Jsat = . (173)
σ

The saturation fluence for dye and most of gas lasers is ∼ 2 mJ/cm2
which means that in order to get, for example, 5 mJ energy ampli­
fied pulse, the diameter of the active medium should be around 2 cm.
Therefore, due to low saturation fluence dye and gas lasers are very
inefficient amplifiers. For solid-state lasers the saturation fluencies are
much higher, e.g. 0.9 J/cm2 for Ti:sapphire, 7 J/cm2 for Nd:glass and
32 J/cm2 for Yb:glass. In a 1 cm diameter Ti:sapphire amplifier it is
possible to obtain the amplified pulse with an energy of 1 J.

• Another important requirement for laser amplifier medium is broad

amplification bandwidth, which ensures that all spectral components
of the chirped pulse are amplified; otherwise, the amplified pulse cannot
be compressed to its initial duration.

• Long lifetime of the upper laser energy level, which determines the
ability to accumulate large population inversion.
6.5 Laser amplifier media 139

Summarizing these requirements we can define a certain figure of merit

for the amplifier medium:

tspont
M= . (174)
tp Jsat
The evolution of achieved laser powers from the invention of the laser is
illustrated in Fig.91.

Figure 91: The maximum focused laser radiation intensity versus year.

The early rise of the focused intensity was achieved thanks to the invention
of laser pulse generation techniques, Q-switching and mode-locking and sub­
sequent amplification of mode-locked pulses. Thereafter, the progress has
slowed down due to physical limits of short pulse amplification, which were
imposed by the optical damage of the laser amplifier material. A major
breakthrough came in 1985, when the chirped pulse amplification technique
was invented, which currently lies on the basis of every modern ultrashort
pulse laser system. Apart from tremendous impact on laser physics, invention
of CPA facilitated rapid development of experimental sciences and opening
new areas of physics, technology and multidisciplinary research. In partic­
ular, CPA-based lasers provided experimental access to strong-field physics
6.5 Laser amplifier media 140

and relativistic nonlinear optics, eventually drawing the guidelines for prob­
ing vacuum nonlinearity and reaching the fundamental laser intensity limits.
At present, the most powerful laser systems employ Nd:glass and/or Ti:sapp­
hire as amplifier media. Fabrication technology of Nd:glass is well developed,
there are many efficient pump sources for this amplifying medium and the
wavelength of Ti:sapphire oscillator can be matched with the Nd:glass emis­
sion spectrum. However, Nd:glass is an amorphous medium, its thermal
conductivity is low which significantly reduces repetition rate of the ampli­
fied pulses. Moreover, amplification bandwidth of Nd:glass can sustain only
pulses with sub-picosecond duration. The highest peak power of 1.5 PW
was achieved in a hybrid Ti:sapphire and Nd:glass laser system, which pro­
duced 0.44 ps pulses with an energy of 660 J, yielding the focused intensity
of > 7 × 1020 W/cm2 20 . More recently, 2 PW femtosecond laser system
based on sole Ti:sapphire laser medium was reported to produce the com­
pressed pulses with a duration of 26 fs and an energy of 72.6 J21 . Currently
the highest reported focused beam peak intensity is 1.1 × 1023 W/cm2 22 .
Let us determine what the maximum peak power can be achieved by amp­
lifying ultrashort light pulses. Assuming that tp is duration of the amplified
pulse, which is inversely proportional to its spectral width (tp = 0.441/∆ν):
Jsat hν
Pth = = ∆ν. (175)
tp 0.441σ
Assuming that we can focus such a beam into a spot size of the order of a
wavelength, we get the maximum intensity:

Pth hν 3 ∆ν
Ith = = . (176)
λ2 0.441σ c2
It is easy to estimate that for known laser materials Ith ≈ 1024 W/cm2 .

20
M. D. Perry et al, Petawatt laser pulses, Optics Letters 24, 160–162 (1999)
21
Y. Chu et al, High-contrast 2.0 Petawatt Ti:sapphire laser system, Optics Express 21,
29231–29239 (2013)
22
J. Woo Yoon et al, Realization of laser intensity over 1023 W/cm2 , Optica 8, 630–635
(2021)
6.5 Laser amplifier media 141

Summary of Chapter 6
• Dispersive broadening in laser resonator prevents the generation
of ultrashort light pulses carrying very broad spectrum. Disper­
sion compensation is performed by introducing a delay of the
fastest (long-wavelength) spectral components with respect to
the slowest (short-wavelength) ones by using prism pairs and/or
chirped mirrors

• Since mode-locked lasers generate ultrashort pulses with very low

energy (from several nJ to several µJ), the energy and peak power
scaling of these pulses is performed by amplification in single-
pass, multi-pass and regenerative laser amplifiers.

• Chirped pulse amplification (CPA) is the technique that allows

boosting the energy of an ultrashort pulse, while avoiding optical
damage of the laser amplifier. In doing this, the ultrashort os­
cillator pulse is stretched in time, amplified in a laser amplifier
(either multi-pass or regenerative) and then re-compressed to the
original duration.

• CPA technique constitutes the standard conceptual basis of mod­

ern ultrafast solid-state lasers and laser systems.
7 Types of lasers 142

7 Types of lasers
In this chapter we will briefly discuss the basic features, materials and prin­
ciples of operation of solid-state, gas and semiconductor lasers.. The key
difference between these three types of lasers is that the laser radiation is
generated by very different means. In solid-state lasers, the laser radiation is
produced by rare-earth or transition metal ions embedded in a solid-state host
material. In gas lasers the laser radiation is produced by neutral atoms, ions
or molecules, while in semiconductor lasers the laser radiation is generated
at the p-n junction by recombination of electron-hole pairs, i.e. is produced
by the semiconductor crystal itself. These key differences determine the con­
struction principles, operation regimes, parameters of output radiation and
the range practical applications of these laser types.

7.1 Solid-state lasers

Solid-state lasers are the main laser sources for the generation and ampli­
fication of ultrashort light pulses and therefore play a central role in the
development of modern ultrashort pulse science and technology. Solid-state
lasers can efficiently perform in all regimes of laser operation: free-running
mode, Q-switching and mode-locking.
Fabrication of solid-state laser materials essentially relies on three criteria,
which determine the basic features of such laser operation, construction and
characteristics of the generated pulse:

• the activator ions, which have a specific configuration of energy levels;

• the laser host material having specific mechanical, optical and thermal
properties;

• the optical pump source with its spectral and temporal properties and
pump geometry.

Solid-state laser materials are crystals or glasses (laser hosts) doped with
small amounts of activator ions which produce laser radiation. In general,
solid-state laser materials have to have sufficient narrow emission lines (ex­
cept those used for ultrashort pulse generation), strong absorption bands or
lines and high quantum efficiency of the laser transition. The host material,
where these activator ions are embedded, must be transparent to both pump
and laser radiation, and the absorption of an ion must be in the spectral
7.1 Solid-state lasers 143

region where optical pumping with flashlamps, semiconductor laser diodes

or other lasers is possible23 .

Activator ions.
Ions of a certain group of chemical elements termed rare-earth metals,
lanthanide and actinide, serve as activator ions in solid-state lasers. Rare-
earth metal ions (see the two special rows positioned at the very bottom
of the periodic table shown in Fig. 92) possess narrow emission linewidths
(this essentially ensures that the lifetime of the upper laser level is long)
and electronic transitions occur between the inner layers which are not fully
occupied. Since the electronic transitions occur between the inner layers
which are shielded by the outer layers, emission spectrum of such ion is
almost identical to that of a free ion and is almost independent of the laser
host the ion is embedded in.

Figure 92: Periodic table of elements.

The configuration of electron layers in rare-earth metals is similar to Xe, in

which layers with quantum numbers n = 1, 2, 3 are fully occupied, sublayers
s, p, d of layer n = 4 are also fully occupied and 4f sublayer which can contain
up to 14 electrons is empty, whereas sublayers 5s and 5p are fully occupied
again:
23
M. Eichhorn, Quasi-three-level solid-state lasers in the near and mid infrared based
on trivalent rare earth ions, Applied Physics B 93, 269–316 (2008)
7.1 Solid-state lasers 144

Xe : 1s2 2s2 2p6 3s2 3p6 3d10 4s2 4p6 4d10 5s2 5p6
In the case of rare-earth metals, electrons further fill the empty orbital of
sublayer 4f :

Ce : ...4s2 4p6 4d10 4f 2 5s2 5p6 6s2

Nd : ...4s2 4p6 4d10 4f 4 5s2 5p6 6s2
(177)

As a rule, rare-earth metal ions are trivalent; the neutral atoms have lost
three electrons from 6s, 5d (if present) and 4f sublayers:

Ce3+ : ...4s2 4p6 4d10 4f 1 5s2 5p6

Nd3+ : ...4s2 4p6 4d10 4f 3 5s2 5p6
(178)

After excitation of such ions, electronic transitions between 4f sublayer

levels occur, which are fully shielded by 5s and 5p sublayers.
The lanthanide ions emit radiation in the near-IR spectral range, e.g.,
the wavelengths of strongest emission lines are 1.06 µm in Nd3+ , 1.03 µm
in Yb3+ , 1.53 µm in Er3+ , 2.0 µm in Ho3+ , etc. Actinides also possess a
shielded inner electron sublayer (which is shielded by 6s and 6p sublayers),
however all these elements and their ions are radioactive and therefore are
not used.
Another metal ion group consists of trivalent and bivalent transition metal
ions: Cr3+ , Ti3+ , Ni2+ , etc. Laser levels of transition metal ions are not
shielded, as the electronic transitions occur between outer 3d sublayer levels.
Therefore embedding these ions into different materials, significantly alters
their emission wavelength. Due to strong interaction between the electronic
energy levels and the electric field of the crystal, the energy levels of transition
metal ions may become significantly broadened. Such lasers have a broad
emission band (although the linewidth broadening is homogeneous) and often
are called vibronic.

Laser hosts.
Laser hosts are the materials into which the activator ions are embedded.
Laser hosts have to comply with strict requirements which can be summarized
as follows:
7.1 Solid-state lasers 145

• good optical quality, which includes broad transmission range (trans­

parency to both laser and pump radiation), homogeneity (smoothness
of refractive index) and purity;
• good mechanical, thermal and chemical properties: large thermal con­
ductivity, rigidity, chemical stability and resistance to color center form­
ation;
• activator ions in laser host must maintain their energy level structure;
• crystal lattice structure of laser host must not be altered during dop­
ing and should preserve its optical, chemical, mechanical and thermal
properties;
• fabrication technique of laser host material has to be as simple and
cheap as possible.
Laser hosts, according to their structure are divided into two large groups:
glasses and crystals. Glasses are essentially isotropic and amorphous materi­
als produced by cooling of liquids and alloys. Among huge variety of glasses,
the main types of glasses that serve as laser hosts are silicate (SiO2 ), phos­
phate (P2 O5 ) and fluoride (usually a mixture of various fluorides: ZrF4 ,
BaF2 , LaF3 , AlF3 and NaF, the so-called ZBLAN). Glasses can be molded
with high optical quality and in large dimensions; the optical fibers are also
made from these materials. Glasses are easily processed and could be doped
to a high degree. Since glasses are amorphous materials, their thermal con­
ductivity is low, so imposing limits on the repetition rate of glass-based lasers
and amplifiers. Nd3+ , Yb3+ and Er3+ ions serve as the main activators used
for glass host materials. In general, the marking of laser material includes
the names of a doping ion and laser host, e.g. Nd:glass, which reads as glass
host material doped with Nd3+ ions. The charge of an ion is usually omitted.
Compared to glasses, crystalline hosts possess very good optical proper­
ties and thermal conductivity in particular. However, production technology
of crystalline hosts is more complex, since during the crystal growth, the
original ion in the crystal lattice is replaced with the activator ion. For this
exchange to be successful and no residual defect to appear in the lattice, the
charge of the activator ion has to be identical and its physical size has to be
very close to the replaced element.
The main and most widely used crystalline hosts are oxides and fluorides,
which have very broad transparency range. Basically, these are dielectric
crystals with a large energy bandgap.
Sapphire (Al2 O3 ) is an excellent optical material possessing high crystal­
line quality, and high optical damage threshold, its production technology
7.1 Solid-state lasers 146

is relatively simple and well elaborated. While doping the sapphire crys­
tal, Al3+ ions are replaced with trivalent activator ions. However, there is
a slight problem: the size of Al3+ ion is very small, only ≈ 0.6 Å, while all
rare-earth metal ions are almost twice larger. Therefore, A3+ ions in the
sapphire crystal could be replaced only by transition metal ions: Cr3+ and
Ti3+ , which have similar size (≈ 0.7 Å). Consequently, two doped sapphire
materials are in use: Ti:sapphire and Cr:sapphire (ruby), the latter exists
naturally in nature.
Apart sapphire, other gem class crystals could be doped with trans­
ition metal ions: Cr:BeAl2 O4 (alexandrite), Cr:Be3 Al2 (SiO3 )6 (emerald), etc.
Some fluoride crystals, called colquirites, are also doped with transition metal
ions (usually Cr3+ ): LiCaAlF6 (LiCAF), LiSrAlF6 (LiSAF) and LiSrGaF6
(LiSGAF). Compared to sapphire, their thermal properties are significantly
poorer, but they have a very broad amplification bandwidth (they are so
called vibronic materials) that affords either generation of femtosecond pulses
or production of wavelength-tunable laser radiation.
Garnets comprise the most widely used oxide group. Out of these, the
most popular and widespread is yttrium aluminium garnet, Y3 Al5 O12 (YAG).
YAG exhibits outstanding mechanical, thermal, and optical properties, which
makes it a versatile optical material that is widely used in optoelectronics,
laser physics, and nonlinear optics. YAG is one of the most important laser
host materials, which could be doped with various rare earth metal ions, since
radius of yttrium ions (≈ 1.1 Å) is practically identical to rare-earth metal
ions radii. Neodymium doped YAG (Nd:YAG) is one of the most popular
laser materials, which has low generation threshold, large gain and could be
pumped with flashlamps and laser diodes. Currently, several other artificial
garnets were produced: Gd3 Ga5 O12 (GGG), Ga3 Sc2 Al3 O12 (GSGG), etc.
Garnets could also be doped with Yb3+ ions. For example, in the meantime
a lot of research is carried out to develop Yb:YAG based laser systems that are
capable of operating at very high repetition rates and producing ultrashort
light pulses with sub-picosecond duration. Among the other commonly used
laser hosts it is worth to mention yttrium-based crystals: yttrium vanadate
(YVO4 ) and yttrium lithium fluoride YLiF4 (YLF) which all are doped with
Nd3+ ions.
The search for new efficient laser materials continues nowadays, for ex­
ample, a recently invented but already commercially used laser material is yt­
terbium ion doped potassium gadolinium tungstate KGd(WO4 )2 (Yb:KGW).
Another laser research line is directed toward the development of ultrashort
pulse lasers operating in the mid-IR spectral range. To this end, novel laser
sources based on Ho-doped gain media (Ho:YLF and Ho:YAG), operating
around 2 µm and delivering multimilijoule, multigigawatt pulses with dur­
7.2 Gas lasers 147

ation of a few picoseconds were developed more recently. A particularly

promising and interesting research line concerns the development of trans­
ition metal doped chalcogenide lasers (Cr or Fe doped ZnS and ZnSe), which
produce femtosecond pulses in the wavelenght range from 2.3 µm to 3 µm
and are often termed as ”Ti:sapphire of mid-infrared”.
In conclusion, a huge variety of solid-state laser materials was invented
and demonstrated in operation. However, only a limited cnumber of these
have a commercial perspective, which is defined by ability to provide a specific
wavelength of laser emission, pulse duration and energy, repetition rate, etc.,
which are essential for practical applications.

7.2 Gas lasers

Unlike in solid-state lasers, linewidth broadening in gaseous media is insigni­
ficant (practically the only important broadening mechanism is the Doppler
effect), so gas lasers, as a rule, emit narrowband radiation. Most of gas
lasers (except CO2 , metal vapor and excimer lasers) operate in only continu­
ous wave (CW) regime; their ability to generate or amplify light pulses is
poor due to narrow gain bandwidth and low saturation fluence. However,
due to their diversity, gas lasers cover a very wide spectral range, from the
vacuum ultraviolet to the mid-infrared. So the main applications of gas lasers
are oriented to medicine and industry.
Gases are transparent over a very broad spectral range (almost through­
out the entire optical range), and electronic energy levels in gas atoms and
molecules are usually distributed quite widely. Therefore gas lasers cannot be
pumped optically and due to large quantum defect gas lasers are less efficient
than solid-state lasers. Almost all of gas lasers are pumped by an electric
discharge which is created along the gas tube by applying high voltage. Gas
atoms or molecules are excited by electrons, which are accelerated by the
electric field and transfer their kinetic energy during collisions. In many
cases low pressure gases are used as laser media, since it is much easier to
create a homogeneous discharge at low pressure. Mixtures of gases are used
to significantly increase the efficiency of electrical pumping: additive gas is
selected in such a way that its energy levels are very similar to that of radi­
ating gas. Some of the gas lasers are pumped chemically, where population
inversion is achieved due to excess energy during chemical reactions; these
lasers are called chemical lasers, or temporarily creating molecules which ex­
ist only in excited state, the so called excited dimers; these lasers are called
excimer lasers.
In general, it is more difficult to achieve the population inversion in gases
than in solid-state media. Although general requirements for laser energy
7.2 Gas lasers 148

levels are the same for all laser media, gas media have certain specifics. It
can be briefly described as factors reducing the lifetime of the upper laser
level:

• collisions between gas atoms (molecules or ions) and electrons: the

excitation energy can be converted into electron kinetic energy;

• collisions between excited and not excited atoms. In this way the ex­
citations of the upper laser level are lost because they are become re­
distributed between the other energy levels;

• collisions of excited atoms with the gas tube walls;

The above factors suggest that in order to achieve population inversion

in gases, the pump rate has to be sufficiently high. With a reference to the
gases that are used, gas lasers can be classified into 5 types:

• Neutral atom gas lasers. The laser media are noble gases and their
mixtures, also metal vapors. All noble gas lasers emit in the infrared,
except the He-Ne laser which emits in the visible spectral range, at
the wavelength of 632.8 nm. Metal vapor lasers emit visible light and
operate only in a pulsed free-running regime. Continuous population
inversion cannot be achieved in these media, because their lower laser
energy level is depleted only through collisions with gas tube walls.
Metal vapor lasers operate only at high (> 1000◦ C) temperature, since
the metal has to be evaporated. The best-known example of metal
vapor laser is copper vapor laser (CVL).

• Ion lasers. The energy levels of an ionized atom are shifted with
respect to those of a neutral atom due to an unbalanced charge. As
a result, the gaps between energy levels increase, and all ion (noble
gas, metal vapor) lasers emit at shorter wavelengths, in the visible and
ultraviolet spectral range. The most important ion gas lasers are argon
ion Ar+ laser and helium-cadmium He-Cd laser.

• Molecular gas lasers.

In molecular gas lasers the laser radiation is generated during the trans­
itions between vibrational levels. The energy levels in molecular gases
split into vibrational and rotational, due to relative motions of the
constituent atoms or their vibrations, respectively. The gaps between
vibrational energy levels are much smaller than those between elec­
tronic energy levels, therefore the emitted wavelength is in the mid-IR
7.3 Semiconductor lasers 149

spectral range. Out of variety of molecular gas lasers, the most import­
ant laser is CO2 laser emitting at wavelengths of around 10 µm. CO2
laser is one of the most powerful continuous wave laser (the average
power can exceed kW and more) and one of the most important lasers
for industrial applications, such as laser cutting and welding.

• Excimer lasers. An excimer (termed after an excited dimer) is a

diatomic molecule that exists only in an excited state. When excimer
molecule emits a photon, the temporary bond between atoms breaks
and the molecule disintegrates. The excimer laser is an ideal four-
level laser, in which the lower laser level is always empty. Excimers
are composed of noble gas and halogen atoms: ArCl, ArF, KrF, etc.
To form an excimer molecule, a gas mixture has to be pumped with
an electron beam or electric discharge. All excimer lasers emit light
in the ultraviolet spectral range and operate only in the pulsed free-
running regime. Due to broad gain bandwidth, the excimer lasers can
be used as efficient amplifiers for ultrashort UV pulses, although they
can not produce ultrashort pulses themselves. As the UV beams can be
focused very sharply, the main applications of excimer lasers are related
to biological research, medicine and lithography. However, halogens
are very volatile, reactive and poisonous, which does not make them
attractive materials.

• Chemical lasers. The chemical lasers are lasers in which population

inversion is created during chemical reactions. All chemical lasers are
oriented to military applications, e.g., long range (outside the atmo­
sphere) shooting or destruction of enemy satellites or rockets. Chem­
ical lasers emit in the IR, in wavelength range of 1.3 − 3 µm, where the
Earth’s atmosphere is transparent. The power of chemical lasers can
exceed MW in continuous wave regime and the beam diameter is up
to 1.5 m, however, not much is known about these lasers, since most of
the data is classified. Currently, several types of chemical lasers were
developed: HF laser, AGIL (All Gas-phase Iodine Laser), COIL (Chem­
ical Oxygen Iodine Laser) and deuterium fluoride (DF) laser MIRACL
(Mid-Infrared Advanced Chemical Laser).

7.3 Semiconductor lasers

In a semiconductor laser, the radiation is produced by stimulated electron-
hole recombination that is based on transitions between the conduction and
the valence band of the semiconductor crystal (interband transitions). Since
7.3 Semiconductor lasers 150

the first demonstration in 1962, semiconductor lasers were recognized as a

potential breakthrough in telecommunication technology. However, the first
semiconductor lasers were based on simple p-n junction and operated only at
cryogenic temperature. The development of double-heterostructure allowed
to reduce the threshold current and achieve operation at room temperature.
Further progress in semiconductor lasers was facilitated by the development
of material processing technologies, such as metal-organic chemical vapor
deposition (MOCVD) and molecular beam epitaxy (MBE), which allowed
fabrication of semiconductor crystal structure with atomic layer accuracy.
At present, semiconductor lasers provide the output power up to several
watts and their emission wavelengths cover very broad spectral range, from
the UV to the mid-IR with. Besides numerous applications in various fields
of modern technology, semiconductor lasers are used for pumping of fiber
and solid-state lasers.
The most important class of semiconductor diode lasers is based on III­
V semiconductors. An example for this is GaAs and Ga1−x Alx As, where x
indicates the molar fraction fraction of Ga in GaAs replaced by Al. The
emission wavelength of such lasers depends on doping fraction and is in the
750–880 nm range. A generic double-heterojunction GaAs laser represents a
thin (typically 0.1 − 0.2 µm) GaAs region sandwiched between two regions
of p- and n-doped Ga1−x Alx As, as illustrated in Fig 93.

Figure 93: A typical double-heterostructure formed by GaAs and Ga1−x Alx As

regions.

Since the bandgap energy of Ga1−x Alx As depends on the Al mole frac­
tion, this double-heterostructure is fabricated so as the potential well for
electrons of height ∆Ec coincides spatially with a well for holes of height
∆Ev . Applying the forward bias of eVa ∼ Eg , large densities of electrons are
7.3 Semiconductor lasers 151

injected from the n side and holes from the p side into the well, providing
the population inversion condition, which is expressed as

EF c − EF v > ℏω,
where EF c and EF v are the quasi-Fermi levels for the conduction and valence
bands, respectively. The GaAs layer where stimulated emission takes place
is called the active region, as shown in Fig. 94(a).

Figure 94: (a) Schematic energy diagram under positive (forward) bias in
GaAlAs/GaAs/GaAlAs double-heterojunction laser diode. (b) The refractive in­
dex profile that confines the optical field.

For the maximum gain, the light has to be confined within an active re­
gion. This is achieved by a waveguiding effect in a GaAlAs/GaAs/GaAlAs
sandwich, since reduction of the energy gap of a semiconductor by dop­
ing, causes increase of its refractive index. The refractive index distribution
and the output profile of a typical double-heterojunction laser is shown in
Fig. 94(b). In this case the gain scales as d−1 , where d is the height of the act­
ive region. The concept of double-heterojunction GaAlAs laser is transferred
to other semoconductor lasers, e.g. InAlP.
Conventional heterostructures confine the gain into a small zone (the act­
ive region) and prevent electron and hole diffusion. Further miniaturization
of the active region leads to quantized electron motion that is characterized
7.3 Semiconductor lasers 152

by discrete energy levels, giving rise to quantum film, quantum wire and
quantum dot lasers. These lasers are not just simply smaller, but also show
qualitatively novel radiation properties. Structures with reduced dimension­
ality offer lower threshold currents, larger gain and lower temperature sens­
itivity.
The most unwanted property of a laser diode is the mode hop that results
in the change of laser wavelength and consequently the gain with temperat­
ure. To overcomes this, the simplest approach is based on an external cavity
concept, which makes use of a diffraction grating at the Littrow configura­
tion, which serves as a cavity mirror. Various modifications of semiconductor
lasers were developed with regard of control the laser wavelength and coher­
ence properties by integrating additional components during laser crystal
manufacturing. In such a way, the so-called distributed feedback (DFB),
distributed Bragg reflector (DBR) lasers are designed.

Summary of Chapter 7
• In solid-state lasers the laser radiation is produced by either rare-
earth or transition metal ions embedded in a solid-state host
material (transparent crystals and glasses). Solid-state lasers ef­
ficiently operate in all regimes: free-running, Q-switching and
mode-locking.

• In gas lasers the laser radiation is produced by neutral atoms,

ions or molecules. Gas lasers are efficient only in free-running
regime (either CW or pulsed).

• In semiconductor lasers the laser radiation is produced by the

crystal itself via stimulated electron-hole recombination in p-n
junction.

• Electrically pumped semiconductor lasers are called laser diodes.

They are the most common lasers.
8 Laser wavelength conversion by nonlinear optical methods 153

8 Laser wavelength conversion by nonlinear

optical methods
All solid-state lasers emit in the near infrared spectral range and the pos­
sibilities to tune their emission wavelengths are very limited or absent at
all. These limitations come from laser energy level broadening, which in
most cases is very small (Ti:sapphire and other vibronic lasers are the excep­
tions). Nonlinear optics offers simple methods how to efficiently change the
wavelength of laser radiation. These methods can be formally divided into
frequency up-conversion and down-conversion, which increase or de­
crease the frequency of laser radiation, respectively. Frequency up-conversion
is based on the second harmonic and sum-frequency generation, while fre­
quency down-conversion is based on the optical parametric generation and
amplification. In the latter case the laser radiation at a fixed wavelength
is converted into a broadly tunable radiation preserving original duration
of laser pulses. All these phenomena take place in transparent birefringent
crystals and rely on the second-order nonlinear optical susceptibility of these
materials. In this chapter we will discuss the basic principles and practical
approaches how laser frequency conversion can be performed.

8.1 Nonlinear optical susceptibility

Nonlinear optical phenomena are nonlinear in the sense that polarization re­
sponse of the material depends on the electric field strength of a light wave
in a nonlinear fashion. To understand the origin of the optical nonlinearity,
it is useful to recall how material dipole moment per unit volume (polariz­
ation) depends on the strength of electric field. In linear optics, assuming
that medium response is instant (non-inertial), this dependence is expressed
simply as:

P (t) = ϵ0 χ(1) E(t) , (179)

where ϵ0 is the vacuum dielectric permittivity constant and χ(1) is the linear
optical susceptibility. In general, quantities P (t) and E(t) are vectors and
rapidly oscillate (at the optical frequency) in time. The electric field of a
light wave propagating in a transparent dielectric medium shifts the charges
inducing a dipole moment µ(t), which oscillates at a frequency of the light
wave. The induced polarization is then P (t) = N < µ(t) >, where N is
the number of dipoles and <> denotes the averaging of all dipole moments.
Note that all dipole moments which may naturally exist in the medium are
not accounted for, since they do not emit electromagnetic waves.
8.2 Second harmonic generation 154

The nonlinear optical phenomena occur when the electric field of light is
strong enough to induce the nonlinear polarization, which is however much
smaller that the linear counterpart. This is called perturbative regime of
light-matter interaction or perturbative approximation. In that case, the
polarization can be expressed in power series of the electric field strength:

P (t) = ϵ0 χ(1) E(t) + ϵ0 χ(2) E 2 (t) + ϵ0 χ(3) E 3 (t) + ..., (180)

where χ(2) and χ(3) are second and third order nonlinear optical suscept­
ibilities, respectively. It is also assumed that the material is transparent,
dispersionless and its nonlinear response is instantaneous. The former ex­
pression can be rewritten by separating linear and nonlinear polarizations:

P (t) = P (1) (t) + P (2) (t) + P (3) (t) + ... = PL + PNL . (181)
Here P (2) (t) = ϵ0 χ(2) E 2 (t) and P (3) (t) = ϵ0 χ(3) E 3 (t) are called second order
(quadratic) and third order (cubic) nonlinear polarizations, respectively. In
what follows, we will further restrict to second order nonlinear optical
phenomena which occur in a certain class of materials: noncentrosym­
metric dielectric crystals, i.e. crystals that exhibit birefringence.

8.2 Second harmonic generation

The simplest nonlinear optical phenomenon is the second harmonic gen­
eration: a nonlinear process when frequency of light is doubled. It can be
explained in general details as follows. Consider the incident monochromatic
light wave whose electric field is expressed as

E(t) = Ee−iωt + c.c., (182)

where ω is the frequency of the light wave. The electric field induces the
nonlinear polarization in the medium

P (2) (t) = ϵ0 χ(2) E 2 (t) = 2ϵ0 χ(2) EE ∗ + (ϵ0 χ(2) E 2 e−2iωt + c.c.). (183)

The induced quadratic polarization has two components. The first polariza­
tion component (the first term on the right hand side) has zero frequency and
describes the nonlinear phenomenon called optical rectification, i.e. creates
a static electric field in the crystal which exists as long as the optical field is
present. The second polarization component (the second term on the right
hand side) denotes the polarization which oscillates at the doubled frequency
of the incident light wave and is responsible for second harmonic generation.
8.2 Second harmonic generation 155

It is convenient to ascribe the nonlinear optical phenomena using virtual

energy level diagrams. Second harmonic generation can be viewed as
simultaneous absorption of two photons at frequency ω into a virtual energy
level and subsequent emission of a single photon at frequency 2ω, as depicted
in Fig. 95.

Figure 95: Energy level diagram depicting second harmonic generation. Eg is the
energy bandgap of dielectric medium. Dashed lines denote virtual energy levels.

Note that virtual energy levels are not real, and they are associated with a
small perturbation of bound electron quantum mechanical state by a light
wave. Therefore, lifetime of virtual energy level can be formally estimated
from the uncertainty principle: ℏ/δE, where δE is the energy difference
between virtual and the closest real energy level. So the lifetime of virtual
energy level is always much shorter than the lifetime of any real energy level,
suggesting that processes involving virtual energy levels are very fast.
Figure 95 also represents the energy conservation law: ℏω1 + ℏω1 = ℏω2 ,
where 2ω1 = ω2 . However, for efficient second harmonic generation, the mo­
mentum conservation law also has to be satisfied, i.e., ℏk1 +ℏk1 = ℏk2 , where
k1,2 = n1,2 ω1,2 /c denote fundamental (incident) and second harmonic wave
vector lengths, respectively. Assuming that all interacting waves propagate
at the same direction, we get that the momentum conservation is equivalent
to the condition

n(ω1 ) = n(2ω1 ). (184)

The above expression represents the phase matching condition as it
imposes the phase velocities (c/n) of the interacting waves to be equal. In
other words, the dipoles excited by the wave at fundamental frequency emit
waves at doubled frequency (second harmonic) which are always in phase
8.2 Second harmonic generation 156

with an exciting wave. This in turn guarantees that emitted second harmonic
waves interfere constructively during propagation and the intensity of second
harmonic radiation increases quadratically versus the propagation distance.
The fulfillment of the phase matching condition makes use of crystal bi­
refringence and calls for setting appropriate polarizations of the interacting
waves. For instance, in a negative uniaxial crystal (ne < no , where ne is the
principal value of the extraordinary refractive index and no is the ordinary
refractive index), the phase matching condition could be satisfied for the fun­
damental wave of an ordinary polarization, and the second harmonic wave
of an extraordinary polarization. This is achieved only for certain propaga­
tion direction of the interacting waves with respect to the optical axis of the
crystal, and which is called the “phase matching angle”:

ne (2ω, θ) = no (ω). (185)

The angle θ is computed from the refractive index ellipsoid equation.

Figure 96: Phase-matching angle for the second harmonic generation in BBO and
KDP crystals.

The phase matching angle for second harmonic generation changes by

varying the fundamental wavelength, as illustrated in Fig.96. In practice, the
nonlinear crystal is cut at the phase-matching angle for a given fundamental
laser wavelength, i.e., so the phase-matching condition is satisfied for normal
incidence.
The above considerations apply also to sum-frequency generation.
In the case of sum-frequency generation, there are two incident waves with
8.3 Optical parametric amplification 157

frequencies ω1 and ω2 , and the process is described as ω1 + ω2 = ω3 , where ω3

is the sum frequency. The incident waves with different frequencies can be
produced by either two different lasers or the same laser. For example, the
most common case of sum-frequency generation is generation of higher order
(third, fourth, fifth) harmonics. Third harmonic generation is equivalent to
sum-frequency generation between the fundamental and second harmonics:
ω + 2ω = 3ω. Fourth harmonic can be generated via ω + 3ω = 4ω, and so
on.

8.3 Optical parametric amplification

Although harmonic generation is a very simple and efficient method to change
laser radiation wavelength and to extend the wavelength range accessible
by a particular laser, many applications in physics, biology, medicine and
chemistry demand continuous wavelength tuning over a broad frequency
(wavelength) range. This could be achieved by making use of optical para­
metric amplification. The physical mechanism of the optical parametric
amplification is essentially different from the stimulated emission, which lies
on the basis of laser operation. At the photon level, the parametric amplific­
ation process does not require the population inversion between real energy
levels or bands of the medium. Only virtual energy levels, which occur as
a result of a weak perturbation of the electron states via nonlinear polar­
ization induced by an intense pump wave, and which are located within a
transparency band of the medium, are involved, highlighting instantaneous
and non-resonant nature of the parametric amplification process. The prin­
ciple of optical parametric amplification is schematically depicted in Fig. 97.
The energy from an intense pump wave at frequency ωp is transferred to a
much weaker incident signal wave at frequency ωs , which is thus amplified
alongside the occurrence of a third wave at frequency ωi , which is called the
idler wave. A simplified physical picture of the process could be understood
as follows. As a pump photon at frequency ωp is absorbed by a virtual en­
ergy level, the presence of a photon at the signal frequency stimulates the
emission of a photon pair: the signal photon with identical characteristics to
the incident one and the idler photon at difference frequency ωi = ωp − ωs ,
as schematically shown in Fig. 97. Since the roles of the signal and idler
photons are interchangeable, the idler photon stimulates the emission of an­
other identical idler photon and the signal photon at difference frequency
ωs = ωp − ωi . The process repeats a multiplicity of times consuming the
energy of the pump. In the macroscopic scale, the amplification of the sig­
nal wave reinforces the generation of the idler wave, and vice versa, thus
the amplitudes of the interacting waves (pump, signal and idler) become
8.3 Optical parametric amplification 158

coupled, and the intensities of the signal and idler waves grow exponentially
with propagation distance, at the cost of the pump wave, whose energy is
continuously depleted.

Figure 97: The principle of optical parametric amplification.

Optical parametric amplification is based on the same physical principles

as the second harmonic or sum-frequency generation and requires momentum
conservation (phase matching condition) to be fulfilled. In a collinear case
(assuming that pump, signal and idler beams propagate in the same direc­
tion), for a negative uniaxial crystal (no > ne ) the phase-matching condition
can be written as:

ωp ne (ωp , θpm ) = ωs no (ωs ) + ωi no (ωi ), (186)

which implies that the pump wave has an extraordinary polarization, whereas
the signal and idler waves have ordinary polarizations. The angle θpm between
the propagation direction and optical axis of the crystal is called the phase-
matching angle.
When propagation direction of the pump beam with respect to crystal op­
tical axis is changed (i.e., the crystal is rotated), a different pair of signal and
idler frequencies satisfies collinear phase-matching condition. Hence, continu­
ous signal and idler wave frequency tunability can be achieved in a very broad
spectral range as shown in Fig. 98. Note that when the signal wavelength
decreases, the idler wavelength increases and vice versa, as defined by the en­
ergy conservation law. In the ideal case, the signal wavelength can be tuned
from ωp to ωp /2, whereas the idler wavelength can be tuned from 0 to ωp /2
respectively. However, in real crystals wavelength tuning range is limited by
crystal absorption and/or phase-matching condition.
Optical parametric amplification is a process when energy of the pump
wave is transmitted to signal and idler waves, therefore, their intensity in­
creases exponentially from propagation distance in the crystal z:
8.3 Optical parametric amplification 159

Figure 98: Phase-matching curve for BBO crystal. The pump wavelength is
400 nm, which corresponds to the second harmonic of Ti:sapphire laser.

1
Is (z) = Is (0) exp(2Γz) (187)
4
ωi
Ii (z) = Is (0) exp(2Γz) (188)
4ωs
here

ωi ωs d2eff Ip2
Γ2 = , (189)
ns ni np ϵ0 c3
and quantity Γ is called the amplification increment, which depends on
the effective nonlinearity of the crystal deff , proportional to certain matrix
elements of the second order nonlinear susceptibility χ(2) , and pump wave
intensity Ip . Here we notice that only radiation intensity of a laser source
is high enough for amplification, while using non-laser sources makes optical
parametric amplification practically impossible.
In this sense the nonlinear crystal performs as a certain catalyzer and
gain coefficient is very high (conversion from pump energy to signal and idler
waves can be more than 80%). The gain coefficient of optical parametric
amplification is defined as:

Is (z) 1
G= = exp(2Γz) , (190)
Is (0) 4
here Is (0) and Is (z) are the signal wave intensities at crystal input and output
respectively. At GW/cm2 level pump intensity, which is easy to achieve with
8.4 Optical parametric amplifiers 160

ultrashort pulse lasers, the gain coefficient can be 103 − 106 for a crystal of
a few mm length. So optical parametric amplification is a unique process
allowing fixed wavelength radiation to be converted into tunable wavelength
radiation with high efficiency and without loss of coherence properties.

8.4 Optical parametric amplifiers

There are two types of optical parametric amplifiers, whose design depends
on the pump laser parameters: pulse duration, power and intensity.

Optical parametric oscillators

The first type of optical parametric amplifiers are made of nonlinear crys­
tal which is placed in a resonator as shown in Fig.99. Such optical parametric
amplifiers are called standing wave amplifiers or optical parametric oscillators
(OPO). Here the pump beam is coupled through one of the resonator mir­
rors and the mirrors are chosen to reflect either signal or idler waves (singly
resonant OPO, Fig.99(a)) or both (doubly resonant OPO, Fig.99(b)).
For OPO pumping either low power lasers which operate in CW or Q-
switched modes or femtosecond laser oscillators generating ultrashort but low
energy pulses at high repetition rate are usually used. Since the peak power
of such pump laser radiation is low, single-pass amplification is relatively low
as well (from several to a dozen of times). The resonator performs the same
roles of energy accumulation, feedback and formation of high spatial quality
beam (due to diffraction losses) as in a conventional laser. During the first
round-trip parametric superfluorescence is generated in the nonlinear crystal;
a part of which propagating on the resonator axis is amplified during the
following round-trips through the nonlinear crystal. As in conventional laser,
the oscillation condition is defined in the same way: parametric amplification
after one resonator round-trip must exceed losses due to light scattering,
absorption, diffraction and mirror transmission.
In the case of the simplest singly resonant OPO a small fraction (usually
few percent) of resonating signal is coupled out. In addition, the entire idler
wave is coupled out which after each pass is generated again. Notice that
all resonator stability criteria discussed at the beginning of this textbook
also apply to OPO resonator. Wavelength tuning of OPO is performed in
a simple way – by rotating the nonlinear crystal, i.e., by changing pump
beam direction with respect to crystal optical axis. In such case at each
time collinear phase-matching is achieved for a different signal and idler fre­
quency pair. Wavelength tuning range of OPO is usually limited by spectral
characteristics of resonator mirrors because fabricating mirrors with uniform
8.4 Optical parametric amplifiers 161

Figure 99: The main configurations of optical parametric oscillators: (a) singly
resonant, (b) doubly resonant.

coefficient of reflection over a broad spectral range is very difficult or even

impossible. Therefore, in such device the entire wavelength tuning range
allowed by phase-matching conditions is not achievable.

Traveling wave optical parametric amplifiers

The second type of parametric amplifiers operate according to traveling

wave principle, i.e., parametric amplification is performed at a straight setup
using several nonlinear crystals. The main difference of traveling wave optical
parametric amplifiers (simply termed OPA) from OPO is the absence of a
resonator, a number of passes through the crystal is small but the gain factor
is large. High power ultrashort pulses, which are generated by amplifying the
laser oscillator pulses, are used to pump OPA. The peak power and intensity
of the pump pulses are very high, so a single pass through the nonlinear
crystal provides huge (several hundreds to thousands of times) amplification.
The simplest OPA consists of several nonlinear crystals. In the first crys­
tal parametric superfluorescence is generated which is then amplified using
one or more amplification stages. The only drawbacks of such amplification
system is the requirement to form spatial and spectral characteristics of para­
metric superfluorescence and the fact that any distortions occurring in the
beam are amplified at each successive stage. However, the main advantage
of such device is large energy of output radiation and broad frequency tuning
range which is only limited by phase-matching conditions.
The principal setup of traveling wave optical parametric amplifier is de­
picted in Fig.100. The pump laser beam is divided (by means of beam
splitters BS) into several independent channels used to pump parametric
8.4 Optical parametric amplifiers 162

Figure 100: The principal setup of traveling wave optical parametric amplifier:
SHG - second harmonic generator (optional), BS - beamsplitters, DM - dichroic
mirrors, PSFG - parametric superfluorescence generator, OPA - optical parametric
amplification stages.

superfluorescence generator (PSFG), first and second amplifiers OPA1 and

OPA2. Spatial and temporal characteristics of the PSFG signal are formed
using an aperture (A) and diffraction grating (DG). The signal and pump
beams are directed to OPA1 and OPA2 using dichroic mirrors (DM) that
have high transmittance for the signal and high reflectance for the pump.
The pump energies for each amplification stage are chosen to achieve gain
saturation. Since amplification of ultrashort light pulses is considered, precise
temporal matching of pump and signal pulses is required, which is provided
by mechanical delay lines. Wavelength tuning is performed in the same way
as in OPO only in this case all three nonlinear crystals have to be rotated
synchronously.
One of the most important and commonly used nonlinear crystals is beta
barium borate (BaB2 O4 , BBO). This crystal has large nonlinearity and phase
matching can be achieved in the whole transparency range (from 190 nm to
3.5 µm). Moreover, growth technology for this crystal is well developed and
relatively cheap.
Fig.101 depicts wavelength tuning ranges of BBO crystal-based OPA
when pumping with commonly used ultrashort pulse lasers and their har­
monics. More than that, frequency tuning range of OPA can be additionally
extended by employing second harmonic generation, sum-frequency and dif­
ference frequency generation. Consequently, wavelength tuning range of OPA
may cover a considerable part of the optical spectrum: from the ultraviolet
to the mid-infrared. Because of that, optical parametric amplifiers are often
called multi-color lasers.
Finally, replacement of the laser amplifier by an optical parametric amp­
8.4 Optical parametric amplifiers 163

Figure 101: Tuning ranges of BBO crystal based optical parametric amplifiers
pumped with commonly used ultrashort pulse lasers and their harmonics.

lifier24 in CPA system has led to novel high power amplification technique,
termed Optical Parametric Chirped Pulse Amplification (OPCPA).
OPCPA favorably combines the advantages of CPA with the advantages of
optical parametric amplification, which offers very high gain, low thermal ef­
fects, great wavelength flexibility and intrinsically broad amplification band­
width, extending well beyond that afforded by existing solid-state laser amp­
lifiers
At present, compact optical parametric chirped pulse amplifiers comprise
a unique class of ultrafast light sources, which currently amplify octave-
spanning spectra and produce few optical cycle pulses with multi-gigawatt
to multi-terawatt peak powers and multi-watt average powers, with carrier
wavelengths spanning a considerable range of the optical spectrum.

24
A. Dubietis, G. Jonušauskas, A. Piskarskas, Powerful femtosecond pulse generation by
chirped and stretched pulse parametric amplification in BBO crystal, Optics Communic­
ations 88, 437–440 (1992)
8.4 Optical parametric amplifiers 164

Summary of Chapter 8
• Nonlinear optics offers simple and efficient way to change the
wavelength of laser radiation.

• Frequency conversion via nonlinear optical phenomena involves

only virtual energy levels, so none of the interacting waves is
really absorbed.

• Efficient second harmonic generation, as well as any other non­

linear optical phenomenon require energy and momentum con­
servation laws to be satisfied. The latter is called phase matching
and makes use of crystal birefringence.

• Optical parametric amplification is a nonlinear process when en­

ergy from intense pump (higher frequency) wave is transferred to
the seed (lower frequency) wave and a third wave with difference
frequency is generated.

• OPOs and OPAs allow generation of broadly tunable radiation

from a pump laser source operating at the fixed wavelength.
9 EN - LT Glossary 165

9 EN - LT Glossary

No. English term Lithuanian term (and additional re­

marks)

1. Acoustooptic modulator Akustooptinis moduliatorius

2. Active Q-switch Aktyvus kokybės moduliatorius
3. Amplification Stiprinimas
4. Angle of incidence Kritimo kampas (spindulio ar panašiai)
5. Anti-reflective coatings Skaidrinančios dangos
6. Arc lamp Elektros lanko lempa
7. Attenuation Silpimas (silpninimas)
8. Autocorrelation Autokoreliacija
9. Bandwidth Spektro plotis
10. Bandwidth-limited pulse Spektriškai ribotas impulsas
11. Beam divergence Pluošto skėstis (matuojama radianais/laipsni­
ais)
12. Beam parameter product Pluošto parametru, sandauga
(BPP)
13. Beam waist Pluošto sasmauka
, (mažiausio diametro vieta)
14. Birefringence Dvejopalaužiškumas
15. Broadening Plitimas (išplitimas)
16. Cavity Rezonatorius (vartojamas turint omenyje
tuščia, rezonatoriu, )
17. Chemical laser Cheminis lazeris
18. Chirped pulse Čirpuotas (faziškai moduliuotas) impulsas
19. Chirped pulse amplification Čirpuotu, impulsu, stiprinimas
(CPA)
20. Coherence Koherentiškumas
21. Continuous wave (CW) laser Nuolatinės veikos lazeris
22. Convergence Konvergavimas
23. Correlation Koreliacija
24. Coupled amplitude wave equa­ Susietu, amplitudžiu, bangu, lygtys
tions
25. Cross-correlation Kryžminė koreliacija
26. Cross-section Skerspjūvis (atomo fizikoje tai reiškia
tikimybe)
,
27. Curved mirror Kreivo paviršiaus veidrodis
28. Delay Vėlinimas [s]
9 EN - LT Glossary 166

No. English term Lithuanian term (and additional re­

marks)
29. Dye laser Dažu, lazeris
30. Dielectric permittivity Dielektrinė skvarba
31. Diffraction-limited beam Difrakciškai ribotas pluoštas
32. Dynamical equilibrium Dinaminė pusiausvyra
33. Dipole moment Dipolinis momentas
34. Dispersive prism Dispersinė prizmė
35. Doubly resonant oscillator Dvibangis osciliatorius
36. Electric discharge Elektros išlydis
37. Electrooptic modulator Elektrooptinis moduliatorius
38. Energy fluence Energijos tankis [J/m2 ]
39. Excimer laser Eksimerinis lazeris
40. Excited Sužadintas
41. Experimental setup Eksperimento schema
42. Feedback Gri̧žtamasis ryšys
43. Fiber laser Šviesolaidinis lazeris
44. Figure of merit Naudingumo skaičius
45. Fission Skilimas (pvz. branduoliu, )
46. Flashlamp Blykstinė lempa (blykstė)
47. Fourier transform Furjė transformacija
48. Free space resonator Laisvos erdvės rezonatorius
49. Free-electron laser Laisvu, ju, elektronu, lazeris
50. Free-running mode laser Laisvos veikos lazeris
51. Full width at half maximum Pusės maksimalaus lygio plotis
(FWHM)
52. Fusion Sintezė/susijungimas (pvz. branduoliu, )
53. Gain bandwidth Stiprinimo juosto plotis
54. Gain coefficient Stiprinimo koeficientas
55. Gas laser Duju, (dujinis) lazeris
56. Gaussian beam Gauso pluoštas
57. Glass laser Stiklo lazeris
58. Group delay dispersion (GDD) Grupinio vėlinimo dispersija [s2 ]
59. Group velocity Grupinis greitis [m/s]
60. Group velocity dispersion Grupiniu, greičiu, dispersija [s2 /m]
(GVD)
61. Harmonic generation Harmoniku, generacija
62. Heat induced birefringence Šilumos indukuotas dvejopalaužiškumas
63. Higher order mode Aukštesnės eilės moda
9 EN - LT Glossary 167

No. English term Lithuanian term (and additional re­

marks)
64. Highly reflective coatings Didelio atspindžio dangos
65. Yb doped medium Yb jonais legiruota terpė
66. Idler wave Šalutinės banga
67. Impurities Priemaišos
68. Intensity Intensyvumas [W/m2 ]
69. Intensity autocorrelation Intensyvumo autokoreliacija
70. Kerr effect Kero efektas
71. Kerr lens mode-locking Kero lešio
, modu, sinchronizmas
72. Lase Generuoti lazerio spinduliuote,
73. Laser Lazeris
74. Laser amplifier Lazerinis stiprintuvas (vartojama pabrėžiant,
kad tai lazerinio stiprinimo terpė be rezonat­
oriaus)
75. Laser beam Lazerio pluoštas
76. Laser diode Lazerinis diodas
1) Lazerio sukelta pažaida (reiškinys)
77. Laser induced damage
2) Lazerio sukeltas pažeidimas (efektas)
78. Laser oscillator Lazerio osciliatorius (lazeris – vartojama
pabrėžiant, kad tai lazerinio stiprinimo terpė
su rezonatoriumi)
79. Laser pulse Lazerio impulsas
80. Lasing Lazerio generacija
81. Lifetime Gyvavimo trukmė
82. LED (Light-Emitting Diode) Šviestukas
83. Longitudinal mode Išilginė moda
84. Losses Nuostoliai
85. Maser Mazeris
86. Mechanical stress Mechanins i, tempimas
87. Metal vapor laser Metalo garu, lazeris
88. Microlaser Mikrolazeris
89. Mode-locked laser Sinchronizuotu, modu, veikos lazeris
90. Multimode beam Daugiamodis pluoštas
91. Multi-pass Daugelio praėjimu,
92. Multi-shot autocorrelator Daugiašūvis autokoreliatorius
93. Noncentrosymmetric media Terpės be simetrijos centro
94. Nonlinear coupling coefficient Netiesinės saveikos
, koeficientas
95. Nonlinear medium Netiesinė terpė
9 EN - LT Glossary 168

No. English term Lithuanian term (and additional re­

marks)
96. Nonlinear response Netiesinis atsakas
97. Nonlinearity Netiesiškumas
98. Nonradiative transitions Nespinduliniai šuoliai
99. Optical density Optinis tankis
100. Optical fiber Šviesolaidis
101. Optical parametric amplifica­ Parametrinis šviesos stiprinimas
tion (OPA)
102. Optical parametric chirped Optinis parametrinis čirpuotu, impulsu, sti­
pulse amplification (OPCPA) prinimas
103. Optical parametric generation Parametrinė šviesos generacija
(OPG)
104. Optical parametric oscillator Parametrinis šviesos osciliatorius (generat­
(OPO) orius, turintis rezonatoriu, )
105. Optical rectification Optinis lyginimas (netiesinis efektas)
106. Optical susceptibility Optinis jautris
107. Optical tweezer Šviesos pincetas
108. Optical vortex Šviesos sukūrys
109. Output coupler Išvadinis veidrodis
110. Output radiation Išvadinė (išeinanti iš rezonatoriaus) spinduli­
uotė
111. Parametric fluorescence Parametrinė fluorescencija
112. Parametric gain Parametrinio stiprinimo koeficientas
113. Parametric superfluorescence Parametrinė superfluorescencija
114. Paraxial approximation Paraksialinis artinys
115. Passive Q-switch Pasyvus kokybės moduliatorius
116. Peak power Smailinė galia [W]
117. Periodically poled crystal Periodiškai poliuotas (orientuotas) kristalas
118. Perturbation Perturbacija (trikdis)
119. Phase matching condition Fazinio sinchronizmo salyga ,
120. Phase mismatch Fazinis nederinimas
121. Phase velocity Fazinis greitis [m/s]
122. Photon Fotonas
123. Photon flux density Fotonu, srauto tankis
124. Photon lifetime Fotonu, gyvavimo trukmė
125. Plane wave Plokščia banga
126. Pockels cell Pokelso elementas
9 EN - LT Glossary 169

No. English term Lithuanian term (and additional re­

marks)
127. Population Užpilda (elektronu, tankis energijos lyg­
menyje)
128. Population inversion Užpildos apgraž, a,
129. Power Galia [W]
130. Propagation Sklidimas
131. Pulse compressor Impulsu, spaustuvas
132. Pulse duration Impulso trukmė [s]
133. Pulse repetition period Impulsu, pasikartojimo periodas [s]
134. Pulse repetition rate Impulsu, pasikartojimo dažnis [Hz]
135. Pulse stretcher Impulsu, plėstuvas
136. Pulsed beam Impulsinis pluoštas (retai vartojamas termi­
nas)
137. Pulsed laser Impulsinės veikos lazeris
138. Pump Kaupinimas
139. Pump-enhanced singly reson­ Vienbangis osciliatorius su kaupinimo
ant oscillator gražinimu
,
140. Pump rate Kaupinimo sparta [Hz]
141. Q-switched laser Moduliuotos kokybės veikos lazeris
142. Quantum Kvantas (mažiausias i, manomas dydis)
143. Quantum defect Kvantinis defektas
144. Quasi-phase matching Kvazifazinis sinchronizmas
145. Radiation Spinduliuotė (vienas iš galimu, vertimu, )
146. Radiative transitions Spinduliniai šuoliai
147. Rayleigh range Reilėjaus atstumas
148. Regenerative amplifier Regeneracinis stiprintuvas
149. Relaxation oscillation Relaksaciniai svyravimai
150. Resonator Rezonatorius
151. Rod Strypas (taip kartais vadina dideliu, matmenu,
lazerinius kristalus)
152. Round-trip time Lėkio (pilno apėjimo) trukmė (pvz. lazerio
rezonatoriuje)
153. Running wave Bėganti banga
154. Saturable absorber I̧sisotinantis sugėriklis
155. Seed radiation Užkrato spinduliuotė
156. Semiconductor laser Puslaidininkinis lazeris
9 EN - LT Glossary 170

No. English term Lithuanian term (and additional re­

marks)
157. Semiconductor saturable ab­ Puslaidininkinis i̧sisotinančios sugerties
sorber mirror (SESAM) veidrodis (dažniausiai tiesiog vartojamas
akronimas SESAM)
158. Signal wave Signalinė banga
159. Single-pass Vieno praėjimo
160. Single-shot autocorrelator Vienašūvis autokoreliatorius
161. Singly resonant oscillator Vienbangis osciliatorius
162. Solid-state laser Kietojo kūno lazeris
163. Spatial Erdvinis
164. Spectral responsivity Spektrinis atsakas
165. Spherical wave Sferinė banga
166. Spontaneous emission Savaiminė spinduliuotė
167. Standing wave Stovinti banga
168. Stimulated emission Priverstinė spinduliuotė
169. Stimulated emission depletion Priverstinio spinduliavimo nuskurdinimo mik­
(STED) microscopy roskopija
170. Streak camera Fotoelektronu, kamera
171. Temporal Laikinis
172. Thermal lens Šiluminis lešis
,
173. Thermal stress Šiluminis itempimas
,
174. Third order dispersion (TOD) Trečios eilės dispersija [s3 ]
175. Threshold Slenkstinis (slenkstis)
176. Time-bandwidth product Impulso trukmės ir spektro pločio sandauga
177. Timing jitter Laikinis tirtėjimas (nukrypimas nuo peri­
odiškumo)
178. Transform-limited pulse Spektriškai ribotas impulsas
179. Transverse (spatial) mode Skersinė (erdvinė) moda
180. Transverse mode suppression Skersiniu, modu, slopinimas
181. Triply resonant oscillator Tribangis osciliatorius
182. Uncertainty principle Neapibrėžtumo principas
183. Wave vector Bangos vektorius
184. Wavefront Bangos frontas
185. Waveguide Bangolaidis
186. Wavelength Bangos ilgis
187. Wavelength tuning Bangos ilgio derinimas
This page is intentionally left blank
Julius Vengelis, Audrius Dubietis
Laser Physics: Lecture notes. Vilnius University, Vilnius 2023.

Lasers and laser-related technologies are an integral part of modern science and
are widely used in many areas of everyday life. This textbook provides the basic
principles of laser operation, describes the underlying physical effects a nd gives
an overview of relevant scientific and technological innovations that make possible
generation and amplification o f u ltrashort l ight p ulses. T he m aterial presented
here is intended for Vilnius University Faculty of Physics undergraduate students
of Physics and Light engineering study programmes, but is also suitable to anyone
with a background in natural sciences, interested in this subject.
knygynas.vu.lt
journals.vu.lt