ISSN 1054660X, Laser Physics, 2012, Vol. 22, No. 1, pp. 153.

Pleiades Publishing, Ltd., 2012.

Original Text Astro, Ltd., 2012.

REVIEWS

Advances in Intense Femtosecond Laser Filamentation in Air1

S. L. China, T.J. Wanga, C. Marceaua, J. Wub, J. S. Liuc, O. Kosarevad, *, N. Panovd,
Y. P. Chena, J.F. Daiglea, S. Yuana, A. Azarma, W. W. Liue, T. Seidemanf,
H. P. Zengb, M. Richardsong, R. Lic, and Z. Z. Xuc
a

Center for Optics, Photonics and Laser (COPL), Laval University, Quebec City, Qc G1 V 0A6, Canada
State Key Laboratory of Precision Spectroscopy, East China Normal University, Shanghai 200062, China
c Shanghai Institute of Optics and Fine Mechanics (SIOM), Chinese Academy of Sciences, Jiading, Shanghai, China
d
International Laser Center, Dept. of Physics, Moscow State University, Moscow, Russia
e Institute of Modern Optics, Nankai University, Tianjin, China
f
Department of Chemistry, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 602083113, USA
g
Townes Laser Institute, CREOL College of Optics and Photonics, University of Central Florida, Orlando, FL, USA
b

*email: [email protected]
Received April 12, 2011; in final form, April 26, 2011; published online September 2, 2011

AbstractThis is a review of some recent development in femtosecond filamentation science with emphasis
on our collective work. Previously reviewed work in the field will not be discussed. We thus start with a very
brief description of the fundamental physics of single filamentation of powerful femtosecond laser pulses in
air. Intensity clamping is emphasized. One consequence is that the peak intensity inside one or more fila
ments would not increase significantly even if one focuses the pulse at very high peak power even up to the
petawatt level. Another is that the clamped intensity is independent of pressure. One interesting outcome of
the high intensity inside a filament is filament fusion which comes from the nonlinear change of index of
refraction inside the filament leading to cross beam focusing. Because of the high intensity inside the fila
ment, one can envisage nonlinear phenomena taking place inside a filament such as a new type of Raman red
shift and the generation of very broad band supercontinuum into the infrared through fourwavemixing. This
is what we call by filamentation nonlinear optics. It includes also terahertz generation from inside the fila
ment. The latter is discussed separately because of its special importance to those working in the field of safety
and security in recent years. When the filamenting pulse is linearly polarized, the isotropic nature of air
becomes birefringent both electronically (instantaneous) and through molecular wave packet rotation and
revival (delayed). Such birefringence is discussed in detailed. Because, in principle, a filament can be pro
jected to a long distance in air, applications to pollution measurement as well as other atmospheric science
could be earned out. We call this filamentation atmospheric science. Thus, the following subjects are dis
cussed briefly, namely, lightning control, rain making, remote measurement of electric field, microwave guid
ance and remote sensing of pollutants. A discussion on the higher order Kerr effect on the physics of filamen
tation is also given. This is a new hot subject of current debate. This review ends on giving our view of the pros
pect of progress of this field of filamentation in the future. We believe it hinges upon the development of the
laser technology based upon the physical understanding of filamentation and on the reduction in price of the
laser system.
DOI: 10.1134/S1054660X11190054
1
CONTENT

Section
1. Introduction
2. The physics of filamentation
3. Limitation to reaching very high intensity inside a filament zone
4. Nonlinear interaction of intense femtosecond filamentsin air
5. Filamentation nonlinear optics: Raman spectral shift, fourwavemixing and supercontinuum
6. Terahertz pulse generation in filaments
7. Birefringence inside a filament in air.
8. Molecular alignment in a filament and revival
9. Filamentation atmospheric science
10. Nonionizing filament and higher order Kerr contribution
11. Challenges ahead
1 The article is published in the original.

page
2
3
7
10
20
22
29
30
40
44
47

CHIN et al.

List of abbreviations/special terms

4WM: fourwavemixing
CEP: carrier envelope phase
CPA: chirped pulse amplification
CPM: crossphasemodulation
EOS: electrooptical sampling
FIB: femtosecond laser induced breakdown spec
troscopy
FIR: far infrared
fs: femtosecond (1015 s)
FW: fundamental wave
GVD: group velocity dispersion
GW: gigawatt(109 W)
Ic: clamped intensity
ICCD: intensified CCD detector
LIBS: laser induced breakdown spectroscopy
LIDAR: light detection and ranging
MW: megawatt (106 W)
NIR: near infrared
ns: nanosecond (109 s)
Pc: critical power for selffocusing
ps: picoseconds (1012 s)
PW: petawatt (1015 W)
SC: supercontinuum
SH: second harmonic
SHW: second harmonic wave
SPM: selfphasemodulation
Teramobile: terawatt laser in a mobile unit, a
FrenchGerman consortium.
TH: third harmonic
THG: third harmonic generation
THz: terahertz (1012 Hz)
TW: terawatt (1012 W)
UV: ultraviolet
XPM: cross phase modulation
XKerr effect: cross Kerr effect
1. INTRODUCTION
Filamentation of an intense laser pulse is a phe
nomenon arising from selffocusing. The latter was
discovered in the midsixties [1, 2] following the
invention of the laser. Since then, it was studied exten
sively for a decade or so using nanosecond laser pulses
[3]. In those years, nanosecond pulses were considered
to be very short, but compared to modernday femto
second (1015 s) and attosecond (1018 s) time scale,
nanosecond is now considered to be very long. It is
precisely because a nanosecond laser pulse is so long
compared to the interaction time scale with matters
that optical breakdown occurs and masks the clear and
easy observation of filamentation for about twenty

years. In 1995, in the Center for Ultrafast Optical Sci

ence (CUOS) of University of Michigan under the
leadership of Gerard Mourou, an accident took
place in the laboratory. In the beam path of the femto
second Tisapphire laser pulses, damaged spots on
mirrors led to the first manifestation of filamentation
by an intense femtosecond laser pulse [4]. Since then,
there have been a lot of activities in this newfound
area of frontier optical research. The fundamentals of
filamentation and some of its applications are by now
wellknown. Several review papers and books [511]
have already been published. It is not the intention of
the current paper to review everything in the field.
Instead, we shall focus on some recent experimental
progress with emphasis on our collective work and dis
cuss some challenges in the field. Thus, many interest
ing subjects which have been reviewed are not dis
cussed. One example is xwave [12, 13] where an
excellent review has been given by Couairon et al. [14].
We shall thus only give a very brief description in
Section 2 of the fundamental physics of single fila
mentation, intensity clamping, selfpulse compres
sion, multiple filamentation, filaments competition
and the independence of the clamped intensity on air
(gas) pressure. One important consequence of inten
sity clamping is immediately discussed in detail in
Section 3. It pertains to the question of significantly
increasing the intensity inside a bundle of filaments in
air using, for example, the highest power available
(petawatt (PW) level) lasers. A natural physical idea is
to fuse together these filaments in the focal region of
an externally focusing element. This idea is tested in
experiments and simulations. Common sense physics
would predict that through mutual interference of the
many fusing filaments, the intensity would be greatly
increased. Intensity clamping simply sets a limit. So
far, experiments in air show that the increase is less
than 50% of the clamped intensity of a single filament
using up to 70 TW pulses.
Although filament fusion cannot give rise to very
high intensity, it still has a lot of new physical conse
quences. So, following Section 3, we present recent
advances in filament fusion and interaction in Section 4.
Filament fusion and interaction have been studied
extensively in the past few years. We shall thus discuss
this problem in more detail. This section includes new
advances in harmonic generation which partially
belong to the next section (5) on the advances in fila
mentation nonlinear optics. Section 5 includes a new
type of Raman red shift inside the filament and the
generation of very broad band supercontinuum into
the infrared through fourwavemixing. One branch of
filamentation nonlinear optics, namely, THz genera
tion from inside the filament, is discussed separately in
Section 6 because of its special importance to those
working in the field of safety and security in recent
years.
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

In the high intensity zone of the filament of a lin

early polarized pump pulse, the isotropic air medium
becomes birefringent. Section 7 discusses the elec
tronic birefringence effect in air induced by the pump
and sampled by a linearly polarized probe pulse
through cross phase modulation. The probe pulse
becomes elliptically polarized and rotates as it propa
gates through the filament.
Since air is a molecular medium, the high intensity
inside the filament will induce yet another type of bire
fringence, namely, delayed alignment of molecular
wave packets of nitrogen and oxygen molecules which
later dephase and revive rotationally over a long time
(more than 100 ps), long after the pump pulse is gone.
The physics of this molecular phenomenon is ana
lyzed by a probe pulse and is discussed in Section 8.
Although intense laser induced molecular alignment,
dephasing and rotational revival were known for a
number of years, we believe that the general readers
interested in filamentation might not be very familiar
with the phenomenon. We thus take an extra effort to
describe in more detail the physics of laser induced
molecular alignment before presenting the phenome
non of filament induced alignment and revival.
Some applications in air are discussed in Section 9
which we call filamentation atmospheric science.
They are lightning control, rain making and remote
measurement of electric field, microwave guidance
and remote sensing. A renewed debate of nonionizing
filamentation has arisen in the past year. The higher
order terms of the nonlinear Kerr index were mea
sured and were found to be very large and effectively
negative. Simulation suggested that even without the
plasma, these higher order Kerr terms would be suffi
cient to balance the initial Kerr selffocusing. Plasma
generation is then considered to be a byproduct.
However, this point of view is not unanimous. There
are questions and arguments about this finding. We try
to confront this problem in Section 10.
This review ends in Section 11 in which we give our
view on the prospect of progress of this field of fila
mentation in the future. We believe it hinges upon the
development of the laser technology which should give
a much better beam quality to minimize multiple fila
ments' competition. Commercialization of some of
the practical applications depends very much on the
reduction in price of the laser system.
At this point, we would like to mention that a full bib
liography on filamentation has been nicely documented
by Stelios Tzortzakis of the Institute of Electronic Struc
ture and Laser (IESL), Foundation for Research and
TechnologyHellas (FORTH), Heraklion, Greece in
the following freely accessible website: http://
www.filamentation.org/Publications.aspx/.
Those
who need to have more references should consult this
website.
LASER PHYSICS

Vol. 22

No. 1

2012

2. THE PHYSICS OF FILAMENTATION

In this section, a brief qualitative description of the
physical mechanism of generating filaments in air will
be given in Subsections 2.1 to 2.5, from single filamen
tation to multiple filaments and their competition for
energy from the reservoir. For details, the readers are
referred to references [511]. This is followed by a
description of selfpulse compression (Subsection 2.6)
and the independence of intensity clamping on pres
sure (Subsection 2.7). This latter phenomenon might
have an important consequence of opening up a way to
increase the energy content inside a filament at high
altitude.
2.1. Single Filamentation
The first question one would ask would be what
could have been unknown or still hidden in selffocus
ing in 1995 when the first sign of filamentation was
observed? The answer is nothing except for the tool,
the femtosecond (fs) laser pulses. In particular, self
focusing in air needs to have a laser peak power higher
than the high critical power so that the pulse will col
lapse. In air, the critical power for selffocusing is of
the order of a few gigawatts (GW). In the olden days,
using nanosecond laser pulses, the threshold for laser
induced airbreakdown was at a much lower power, a
few megawatts (MW) followed by focusing by a rela
tively short focal length lens (about 10 cm). Using
nanosecond pulses, laser induced breakdown is due to
the threestepprocesses all occurring in the laser
field; namely, multiphoton ionization of impurities in
the gas, inverse Bremsstrahlung and electron cas
cade/avalanche ionization [15]. The latter two are col
lisional processes and they take time (many ps to ns
time scale) to accomplish the work. Using nanosecond
pulses, even if the peak power is higher than the critical
power for selffocusing, the breakdown threshold is
reached at the rising part of the pulse before GW level
is reached. Thus, a plasma in the focal volume is cre
ated that further absorbs and reflects the laser radia
tion. Selffocusing is thus prohibited. With the use of
the fs laser pulse (10 to a few hundred fs), the pulse
duration is even shorter than the time it takes for a sin
gle inverse Bremsstrahlung process to occur (accelera
tion of a free electron in the laser field while colliding
with another heavy particle (ion or atom or molecule).
The mean free time of an electron in one atmospheric
air is of the order of 1 picosecond (ps). Thus, even at a
peak power higher than the critical power for self
focusing, breakdown would not occur. This leaves the
playground open to other processes.
From now on, we shall assume that a Tisapphire
laser pulse is used because it is currently the most pop
ular high power fs laser used in almost all the intense fs
laser laboratories in the world. Most of these systems
are based upon the CPA (chirped pulse amplification)
technique [16]. Assume that the peak power of the

CHIN et al.

laser is many times higher than the critical power for

selffocusing, Pc [3]
2

3.77
P c =  .
8n 0 n 2

(1)

Its duration is of the order of a few tens of fs to a few

hundred fs at a central wavelength of 800 nm. The laser
pulse will undergo the socalled slicebyslice self
focusing process. Each slice has an intensity distribu
tion, say Gaussian; i.e., high intensity at the center and
low at the edge. The higher index of refraction at the
center of the slice due to the Kerr nonlinearity is (n0 +
n2I), where n0 is the linear index of refraction, n2, the
nonlinear Kerr coefficient, and I, the intensity. The
nonlinear index of refraction, n2I, is positive in most
normal transparent medium. The index of refraction
at the edge of the slice is the linear one, n0. This inten
sity distribution would lead to selffocusing because
the central part will propagate slower than the edge.
The thickness of each slice could be conceptually con
sidered to be at least c, where c is the speed of light
and , the period of the light wave. This is because we
consider the distribution of intensity across each slice
while intensity is defined as the Poynting vector aver
aged over at least one cycle of oscillation of the field, .
When the power of a slice is higher than Pc, the self
focusing will overcome natural linear diffraction and
collapse resulting in tunnel/multiphoton ionization of
air molecules. The position at which the collapse
occurs is given by [3]
2

0.367ka
z f = 

,
1/2
2
P 1/2

 0.852 0.0219

P crit

(2)

where k and a0 are the wave number and the radius of

the beam profile at 1/e2 level of intensity, respectively,
and P is the peak power. Successive slices at the front
part of the pulse will selffocus at different position
giving rise to the perception of a filament. Each slice,
after the collapse and ionization, will be defocused by
the selfcreated plasma into the background reservoir
[5, 10, 17]. The background reservoir plays an impor
tant role in supporting the continuous generation of
the selffoci slice after slice. Conceptually, a filament
is the perception of a series of selffoci. Each selffocus
comes from the collapse of one slice of the reservoir of
the pulse. A filament does not correspond to a light
bullet piercing through the medium.
Interestingly, Eq. (2) which was derived for a long
pulse or a continuous wave laser beam works also for
femtosecond laser pulses [5]. The peak power is inde
pendent of the chirp in the pulse; i.e., it could be that
of a chirped pulse or a transform limited pulse.
Eq. (2.1) can be coupled to the following Talenov
equation when a focusing/defocusing element with
focal length f is used

1/z f + 1/f = 1/z 'f .

(3)

Here, z f' is the new self focusing distance. Based on

this principle, the selffocusing distance of intense
laser pulses in the atmosphere can be experimentally
controlled with a simple doublelens setup [18].
In practice, the short laser pulse will suffer group
velocity dispersion (GVD) because of the large band
width of the fs laser pulse. That is to say, as the pulse
propagates in an optical medium it will become longer
so that the peak power becomes lower. This will have
an impact on the collapse of the pulse at the selffocus.
Slightly higher power than Pc is necessary to overcome
GVD as well as diffraction so as to have a collapse.
2.2. Intensity Clamping
Equations (2) and (3) give rise to the confidence
that z 'f , (and zf) could be varied at will to any long dis
tance. Lets see how it could be done. We look at zf first
(Eq. (2)). Given a certain beam radius a0, one can
change the peak power P to change the selffocusing
position zf. This could be easily done by changing the
chirp of the pulse, both positively and negatively [5].
The lengthening of the pulse by the chirp would
decrease the peak power P hence lengthening the self
focusing position zf. One can also change the self
focusing position by increasing the beam radius a0
according to Eq. (2). These manipulations would
result in moving the filament to a longer distance. The
addition of a focusing element (Eq. (3)) will further
control the selffocusing distance. Thus, in principle,
any distance up to infinity is possible [10].
Moreover, common sense would say that if one
could keep increasing the laser pulses energy, the
intensity in the selffocus would be increased accord
ingly. Since in principle there is no limit to increasing
the laser pulse energy, one would expect that a very
high intensity could be delivered onto a remote target
generating even a relativistic interaction. Such opti
mism could lead to disastrous consequences if one
were to immediately jump this conclusion and invest
resources, financial and otherwise, into an ambitious
research program. That ever increasing intensity does
not happen is because of intensity clamping.
Equations (2) and (3) predict only the position of
the mathematical collapse of the pulse. Such a col
lapse leading to a singularity does not happen in real
ity. This is because, as the selffocal size becomes
smaller and smaller, the intensity becomes higher and
higher. Soon, multiphoton/tunnel ionization takes
place (with femtosecond laser pulses, there is no colli
sional ionization [15]) resulting in a defocusing
plasma. It is exactly the plasma and not the higher
order Kerr nonlinearity that stops selffocusing col
lapse in femtosecond laser filamentation in air [19].
The local intensity is thus clamped at this balancing
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

point of selffocusing and selfdefocusing by the self

generated plasma. The clamped intensity of a free
propagating beam (no external focusing) is of the
order of 5 1013 W/cm2 [2022]. This phenomenon of
intensity clamping is very profound and will impact
many new phenomena and applications.
Recently, a theoretical work has studied the subcy
cle spatiotemporal dynamics of the laser transforma
tion during filamentation by retaining the full electric
field of the laser pulse. The result predicts a subfs
intensity spike, having intensity exceeding the clamp
ing intensity by a factor of 3, can last for a short prop
agation distance in the filamentation process [23].
2.3. SelfSpatial Filtering
Normally, in many experiments, the laser pulse
contains a lot of spatial modes. The fundamental
mode having the smallest diameter would selffocus
and reach the ionization stage first according to the
dependence of zf on the diameter 2a0 in Eq. (2); i.e., zf
is shortest. Higher order modes having larger diame
ters would have a longer zf during their individual self
focusing process. Before they reach the point of high
intensity ionization, they will have to pass through the
plasma zone created by the fundamental mode and be
diffracted out into the reservoir. Thus only the funda
mental spatial mode would keep on selffocusing into
a series of selffoci which constitute the perfectly
round filament core. We call this phenomenon self
modecleaning or selfspatial filtering [10, 24, and ref
erences therein]. The selffiltered laser pulse by fila
mentation can be applied to improve the outcome of
nonlinear optical process, such as the fourwave mix
ing [25] and second harmonic generation [26].
2.4. Multiple Filamentation
So what would happen to the intensity inside the
filament if we keep increasing the energy or peak
power in the pulse? Lets say we start at a power level
that gives rise to one filament. We now increase the
power/energy of the pulse. Intensity clamping would
stop any further increase in intensity. The extra energy
coming into the filament zone would be spilled over
resulting in a filament with a larger diameter [27].
However, further increase of the energy would lead to
multiple filamentation in practice because all practical
high power fs Tisapphire laser pulses do not have a
homogeneous beam pattern. Hot zones are plentiful
across the beam cross section. Each hot zone will self
focus into a filament if the intensity and power are high
enough. Each of the multiple filaments will have a core
intensity clamped down to that of a single filament.
Thus, no increase of intensity is possible with multiple
filamentation.
LASER PHYSICS

Vol. 22

No. 1

2012

2.5. Multiple Filaments Competition

With multiple filamentation comes an unexpected
phenomenon called filament competition [10]. This
arises from the basic mechanism of a single filament
formation; i.e., slicebyslice selffocusing. Each self
focal spot comes from the surrounding energy reser
voir. For each slice that selffocuses toward the central
axis, it is as if energy is sucked from the reservoir
towards the center and then, due to plasma genera
tion, diverges out into the reservoir. Thus, if there are
two or more well separated hot zones in the pulse each
having a tendency to selffocus, they will each suck
energy from the same reservoir thus competing for
energy. Depending on the relative distances among the
hot zones, sometimes one or two very strong (larger in
diameter) probably fused filaments occur. However,
most of the time, in a currently available TW level laser
pulse, there are many hot zones. If we let a large diam
eter TW level beam to propagate freely without using a
lens, the competition among the various filamenting
hot zones results in no winner; i.e., very few filaments
occur [2832]. But if the beam diameter of the same
TW level beam is reduced by a few times, filament
cooperation takes place. That is to say, along the prop
agation direction, consistently strong filaments are
generated [2932].
Physically, these are due to the interference
between neighboring filaments and the reservoir.
Assume a free propagating fs pulse with a plane wave
front and with two hot zones each having a Gaussian
intensity distribution across the hot front. Lets say
each hot zone has a local peak power higher than the
critical power for selffocusing. Thus each hot zone
will start to selffocus by itself. After the generation of
a plasma, the focusing wave front will be defocused by
the plasma. The two defocusing (diverging) spherical
waves would interfere with the now distorted wave
front of the reservoir. If the distance between the two
hot zones is large, the interference would not result in
a strong local field so that further selffocusing of the
interfering zone is not possible. But if the two hot
zones are close to each other, the interfering field
would be large; i.e., another hot zone is created. This
new hot zone will further selffocus into a new child
filament. If there are many hot zones in the initial
wave front of the pulse each with a different peak
power, they will again undergo interference. If they are
all far apart from one another, the interference fringes
would be too weak to generate further filamentation.
But if they are close to one another, strong local inter
ference fields will give rise to many more child fila
ments downstream in the propagation [5, 32].
According to this multiple filaments competition
scenario, when every individual filament is displaced
from each other sufficiently, thus avoiding strong
interference, each of them will propagate as if it were a
single stationary filament. By using adaptive optics,
this kind of twodimensional organization of multiple

CHIN et al.

M1

(a)

f = 1.5 m A1

(b)
Intensity, arb. units

CEP controlled CPA Ti:s

laser system @ 1 kHz

0.8

~5 fs

Gas chamber 1
f = 1.5 m
M2

CM1
Gas chamber 2 M3

A2
f = 1.5 m

f = 1.5 m
CM4

CM3

BS
A3

0.4
0.2

M4
M5

0.6

CM2

Autocorrelator

0
0.05

0.05
Delay, ms

Fig. 1. (a) Experimental setup: Ti:s, Ti:sapphire; A1A3, softaperture; CM1CM4, chirped mirrors; M1M5, silvercoated high
reflection mirrors; BS, beam splitter. (b) Autocorrelation signal of the final compressed 5 fs pulses.

filaments has been experimentally demonstrated [33,

34]. The key is indeed to generate such an initial laser
pattern with multiple hot spots well separated.
2.6. SelfCompression of Femtosecond Laser Pulses
The effect of selfcompression of femtosecond
laser pulse during filamentation has received extensive
attention owing to the farranging application of few
cycle pulses in attosecond physics [35, 36]. In order to
producing intense fewcycle infrared laser pulses, sev
eral compression schemes have already been pro
posed. One wellestablished technique for pulse short
ening makes use of spectral broadening of laser pulses
by cascaded filamentation in gases as well as selfphase
modulation in a rare gasfilled hollow fiber followed by
chirp compensation with suitable dispersive elements
such as chirp mirrors [37, 38]. Chen et al. have dem
onstrated the generation of twocycle optical pulses
with duration of 5 fs and energy of 0.7 mJ at 1 kHz by
compressing a 38 fs laser pulse from a carrierenvelope
phase (CEP) controlled Ti:sapphire laser system
through a cascade filamentation compression scheme
[39].
The physics of the selfcompression arises from the
slicebyslice selffocusing mechanism of filamenta
tion. The central and front part of the pulse undergo
slicebyslice self focusing leaving behind a series of
plasma dots in such a way that the front part
becomes narrower and narrower. The back part of the
pulse would interact with the plasma zone (filament)
dispersing the energy of this part of the pulse into the
back ground reservoir. Thus, the pulse becomes a
structure sharp at the front part and flat at the back
part. This means selfcompression. Later in the prop
agation, pulse splitting occurs followed by a higher
peak at the back part of the pulse while the front part
runs out of steam to selffocus and generate a plasma.

A detailed analysis is given in [40]. We describe a typi

cal experiment in what follows.
The experimental setup is illustrated in Fig. 1. The
laser source is a conventional chirped pulse amplifica
tion (CPA) Ti:sapphire laser system operating at 1 kHz
repetition rate, with the oscillator CEP stabilized. This
system can provide 4.5 mJ intense pulses with duration
of 38 fs (full width at half maximum (FWHM)), but
only 2 mJ is exploited in the experiment. The laser
pulses are attenuated to 1.4 mJ after passing through a
circular soft aperture, A1, the diameter of which is
6 mm, and the aperture edge is irregularly toothed
with a modulation depth of 1 mm and anominal mod
ulation period of 0.5 mm. Then the beam is loosely
focused by a thin f = 1.5 m lens into the first argon
filled chamber at a pressure of 900 mbar, where a
15cmlong single filament is generated. The broad
ened spectrum is dispersion compensated by chirped
mirrors 200 fs2, delivering stable 9.7 fs pulses and
retaining 93% of the input energy. By using another
soft aperture, A2 (of the same size as A1), to remove
the peripheral lowintensity beam energy, the remain
ing 1.2 mJ, 9.7 fs pulses are refocused by a silver
coated f = 1.5 m concave mirror into the second
argonfilled chamber at a pressure of 434 mbar, where
a 20cmlong filament is formed. The output beam
after the second chamber comprises an intense white
light core and a lowintensity energy reservoir during
filamentation. After several bounces on a pair of
chirped mirrors 160 fs2, the recompressed pulse of
the inner core turns out to be 5 fs (FWHM), which is
shown in Fig. 1b. The diameter (at 1/e2 of the peak
intensity) of this intense central core is about 5.5 mm,
which carries an energy of more than 0.7 mJ and con
tains 67% of the total beam energy. This corresponds
to an effective energy efficiency of 50% for the whole
compression system. Moreover, a simple and effective
method is developed in this work to suppress multiple
filament formation and stabilize a single filament by
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

inserting a soft aperture with an appropriate diameter

into the driving laser beam prior to focusing, resulting
in an excellent compressed beam quality [39]. The
accessible wavelengths of this new type of compression
technique have been extended to deep ultraviolet [41]
and infrared [42].
2.7. Pressure Independence of Clamped Intensity
What would be the clamped intensity in air (or any
gas) when the pressure is changed? An intuitive reac
tion to this question is naturally that it should depend
on the pressure, be it linear or nonlinear. It turns out
that it is independent of pressure simply because of the
very short laser pulse duration which could not induce
any collisional ionization during electron/plasma gen
eration in the selffocusing collapse of the pulse. The
oretical analysis and experimental observation has jus
tified this unexpected result [43]. A quick observation
is the following. When the intensity is clamped, the
Kerr nonlinear index of refraction for selffocusing
(n2I) is balanced by the index of refraction of the self
generated electron/plasma (Ne(I)/2Ncrit). Here,
Ne(I) is the electron/plasma density generated at the
laser intensity I and Ncrit is the critical density. In
mathematical form, the balance of the change of index
of refraction n is given by (the sign implies that so
far other higher order Kerr terms are neglected in the
description of filamentation; we shall come back to
this question at the end of this review)
nKerr(neutral) n(plasma)
(4)
which becomes
n 2 I c N e ( I c )/2N crit ,

(5)

I c N e ( I c )/ ( 2N crit n 2 ) .

(6)

where
Equation (6) can be solved to obtain the clamped
intensity Ic. It is a function of Ncrit, n2 and Ne. The crit
ical density is given by
2

N crit = 0 m /e ,

(7)

with e, m, and being the elementary charge, the

electron mass and the central frequency of the laser
field, respectively. This term is independent of the
pressure. n2, a material property, is related to the non
linear susceptibility of the gas medium and is linearly
proportional to the pressure. The electron/plasma
density Ne depends on the physical process of its gen
eration. In the case of fs laser pulses, only multipho
ton/tunnel ionization of the gas is relevant. There is no
collisional ionization such as inverse Bremsstrahlung
or cascade ionization [15] because the pulse duration
is too short, shorter than the mean free time of an elec
tron in one atmospheric pressure of air (~1 ps). That
means that it depends linearly on the pressure. From
Eq. (6), the linear dependences on pressure of the two
LASER PHYSICS

Vol. 22

No. 1

2012

terms, one in the nominator (Ne) and one in the

denominator (n2), cancel each other out leaving
behind an equation for the clamped intensity indepen
dent of pressure. (For the detail of the analysis and
experimental verification, see [43].)
Thus, if one sends a very powerful fs laser pulse ver
tically through the atmosphere in which the pressure
decreases with the altitude and assuming that one can
control at will the filament position, the intensity in
the filament will be clamped to the same value inside
the filament at any height. However, the diameter of
the filament will increase at higher altitude. This is
because the critical power for selffocusing (Eq. (1)) is
inversely proportional to n2. The latter is proportional
to pressure. Hence, the critical power is inversely pro
portional to pressure or directly proportional to the
altitude; i.e., the higher the altitude is, the lower the
pressure is and the higher the critical power is. Thus,
one needs to deliver more power in the fs laser pulse to
high altitudes than at sea level in order to have fila
mentation. Note that the clamped intensity is
I Pc/A,

(8)

where Pc is the critical power for selffocusing and A,

the cross sectional area of the filament core. At high
altitude, according to Eq. (8), Ic being unchanged
(clamped) while Pc has to be higher, the area A of the
filament should be larger. Roughly speaking, since Pc
is directly proportional to the altitude, A should also be
proportional to the altitude (Eq. (8)) or inversely pro
portional to the pressure. A consequence of this is that
the filament diameter is larger at higher altitude. Thus,
since the air pressure or air density is inversely propor
tional to the altitude, the total number of air molecules
inside a filament is roughly independent of the altitude
but is proportional to the filament length L. For a fila
ment with a constant length L, the nitrogen fluores
cence signal emitted from inside the filament is con
stant whatever the altitude is so long as there is fila
mentation.
Another consequence of this phenomenon (see
Eq. (8)) is that at higher altitude, there will be more
energy inside the larger diameter filament. This could
be important from the point of view of remote sensing
because the interaction volume would be larger; hence
there would be more return signal with respect to the
signal from a lower altitude assuming a constant pol
lutant density.
3. LIMITATION TO REACHING VERY HIGH
INTENSITY INSIDE A FILAMENT ZONE
One can use an external focusing element to focus
a fs laser pulse and obtain filamentation. In the case of
single filamentation, the intensity inside the filament
could be increased but not significantly. However, the
increase in the plasma density inside the filament
could be rather high because of the high order nonlin

CHIN et al.

Laser
Dump
Filaments in air

Spectro
meter

ICCD

Fig. 2. Experimental set up to measure the fluorescence from the filament zone in air.

ear generation process (multiphoton and/or tunneling

ionization) that depends nonlinearly on the intensity.
The reason for the insignificant increase in intensity is
intensity clamping even if external focusing helps in
reducing the focal spot size [27]. Could we then reach
a very high intensity by focusing many filaments
together into the geometrical focus using an external
lens or focusing element?

we were to focus a large diameter petawatt (PW,

1015 W)) laser pulse in air, hundreds or more filaments
would be created. (Note that the critical power for self
focusing of the Tisapphire laser pulse in air is between
5 and 10 GW, depending on the pulse length [44].)
Thus interference would result in an enormous inten
sity. This, unfortunately, would not happen because of
intensity clamping as the following shows.

Let us analyze this problem. If two filaments could

be fused into one simultaneously in space and time,
the overlap of the two filament cores would in principle
gives rise to interference zones [5]. One would expect
that in the constructive interference zone, the inten
sity would be four times higher than a single filament
would give. This could be done by external focusing
one single pulse with two hot zones or by crossing two
pulses at a very small angle. In the case of multiple fil
aments, one could focus the pulse such that all the fil
ament cores would overlap near the geometrical focus.
The interference zone would give rise to intensity as
high as 2nI, where n is the number of overlapping fila
ments and I, the clamped intensity of one filament. If

A series of experiments have been performed [45,

46] by focusing 50 fs Tisapphire laser pulses in air and
argon at power levels ranging from 0.1 to 70 TW (Pc in
air of 50 fs pulses is about 10 GW [44]). These pulses
did not have a good beam quality and hence multiple
filaments naturally occurred. Figure 2 shows an exper
imental set up.
Tisapphire laser pulses of nearly 1.5 TW (47 fs,
70 mJ, 10 Hz, diameter ~2.5 cm) were focused by a
short 20 cm focal length lens in air. The spectral broad
ening of atomic oxygen O I (777.4 nm) triplets from
I, 1013 W/cm2
10

Ne, 1016 cm3

10

100 fs (chirped)

8
100 fs (chirped)
Focal length is 20 cm

Ne ~ 5.3 1016 cm3

f = 20 cm
I ~ 6.3 1013 W/cm2
Ie ~ 4.5 1013 W/cm2

4
2

10
0

10

20

30

40

50

60

70
E, mJ

Fig. 3. Measured electron density in the focal region of a

100 fs, negatively chirped pulse in air by varying the pulse
energy from 570 mJ. The focal length is 20 cm.

20

30

40

50

60

70
E, mJ

Fig. 4. Measured intensity (dots) in the focal region of a

100 fs, negatively chirped pulse in air by varying the pulse
energy from 570 mJ. The focal length is 20 cm. The
intensity stays the same and is slightly higher than the sin
gle free propagating filaments clamped intensity (dashed
line).
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

inside the filament zone was measured from the side.

This broadening is proportional to the plasma density
and is calibrated [47]. Figure 3 gives the plasma den
sity as a function of the laser pulse energy. It is constant
over the whole range of energy used.
Since the intensity dependence of the ionization
probability of air molecules (nitrogen and oxygen) can
be calculated using the Smatrix theory [45, 46], one
can estimate the intensity at which the measured
plasma density is produced assuming that the ioniza
tion process is purely unimolecular (i.e., no colli
sional ionization is involved). Figure 4 (black dots)
shows the corresponding calculated intensity inside
the filament as a function of laser pulse energy.
This result shows that the intensity inside the fila
ment does not change even if the energy per pulse is
changed by more than 10 times. The highest energy
used was 70 mJ using 100 fs negatively chirped pulses.
The peak intensity was 0.7 TW while the critical power
for selffocusing at this pulse length is around 5 GW
[44]. Many filaments would have been created near the
geometrical focus. At least some overlap is expected.
Yet, the measured intensity is unchanged and is ~40%
higher than the clamped intensity of a single filament
at free propagation (no external focusing) condition
(dash line, Fig. 4).
Another experiment was done by tightly focusing
60 fs, 4 J (67 TW) Tisapphire laser pulses in air by Ji
et al. [46]. The beam diameter was ~15 cm and the
focal length of the lens was 1 m. In this experiment, the
total nitrogen fluorescence was measured from the
side similar to the set up in Fig. 2. Two pulse energies,
4 J and 50 mJ (ratio is 80), of the same duration (60 fs)
were used. The ratio of the two peak fluorescence was
measured to be about 420/41 = 10.24. Considering
that the nitrogen fluorescence [50, 51] is the result of
tunnel/multiphoton ionization [52] and depends on
the plasma density inside the ionization zone, one
could write down an over simplified empirical relation
between the (measured) fluorescence strength, If, and
the laser intensity, Ilaser [46] which is that the fluores
cence strength is proportional to the ionization rate of
nitrogen

If ionization rate of nitrogen I laser .

(9)

The value of (8) comes from the experimentally

measured slope of the ion yield vs intensity curve dur
ing ionization of nitrogen molecules by a fs Tisap
phire laser pulse [4, 53].
This analysis gives an orderofmagnitude type of
argument but it is enough to give a physical feeling.
The ratio of the intensities inside the filaments is given
by
1/8
I 4 J I f ( 4 J ) 1/

 < 
= 420
 1.34 ,
41
I 50 mJ I f ( 50 mJ )

LASER PHYSICS

Vol. 22

No. 1

2012

(10)

8.5 mJ/100 fs
/800 nm

0.85 mm dia. 1.9 mJ, P/Pc = 1.7

2 mm dia. (FWHM)
f = 30 cm
0.85 mm dia. 1.9 mJ, P/Pc = 1.7
0.85 mm 2

Fig. 5. A theoretical scheme to focus two fs laser pulses

towards the geometrical focus of a lens. A screen with two
apertures transmits two symmetrically spaced beams from
a Gaussian fs laser beam. They pass through a lens and
cross at the focus. The dimensions are given in the figure.

where I4 J, I50 mJ are laser intensities with pulse energy

4 J and 50 mJ, respectively. Since more filaments are
produced when a laser energy of 4 J is used, it is rea
sonable that the fluorescence arising from the 4 J pulse
is higher than that from the 50 mJ pulse. However,
what is interesting is that from Eq. (10), the fluores
cence intensity was increased by only about 34%
inside the tight focal region while the energy (power)
was increased by 80. Let us make another rough esti
mation with the previous experiment using 0.7 TW
pulses. The f/number of the lens used to focus the
50 mJ pulses is 20/2.5 = 8 while that of the present lens
is 100/15 7. Let us say they are roughly equal. Then,
as in the case of the previous 70 mJ/100 fs pulse, the
intensity in the filament of the current 50 mJ pulse
would be roughly equal to that of the previous
70 mJ/100 fs pulse; i.e., I50mJ in Eq. (10) would be
~40% higher than the single filaments clamped inten
sity Ic
I4 J 1.34 I50mJ (from Eq. (10))
1.34 1.40Ic (from Fig. 4)

(11)

1.88Ic.
Thus, the intensity inside the tightly focused filament
zone of the 67 TW pulse is roughly less than twice that
of a single filament in air generated normally by a pulse
of ~1 mJ/50 fs = 20 GW. That is, while the initial
power ratio of the two pulses is 67/0.02 = 3350, the
ratio of the intensity in the filament zones is less than
2. Even though the estimation is rough, it conveys the
principal idea that multiple filaments could not be
forced into multiple interference to produce very high
intensity.
However, using the argument given above, it is
expected that the intensity at the focal zones in both
sets of experiments be significantly increased because
there would be many filaments fusing together near
the focus. This seemingly counter intuitive observation
can be explained by the physical process during inten

CHIN et al.
1.0
0.5
0

(a)

0.5

1.0
80
60
40
20
0
1.5 1017

(b)

Peak intensity,
TW/cm2

Peak electron
density, cm3

y, mm

10

1.0 10

One aperture
Two apertures

(c)

One aperture
Two apertures

17

0.5 1016
0
10

20
30
Propagation distance, cm

40

Fig. 6. The simulation results of the interaction scheme in Fig. 5: (a) fusion of two beams; (b) peak intensity distribution along
the propagation axis where 30 cm is the geometrical focus; (c) peak electron density distribution.

sity clamping. Kosareva et al. [45] have calculated the

fusion of two filaments by passing a fs laser beam with
a Gaussian cross section through two adjacent circular
apertures before being focused by a lens as shown in
Fig. 5.
Figure 6a shows the crossing/fusion of the two fila
ments generated from the two parts of the beam sym
metrically going through the same lens. The fusion
position is before the geometrical focus at 30 cm. The
two filaments cross and merge into one as they propa
gate. However, the calculated peak intensity in the
crossing region is only increased by 30% as compared
to the single filament case coming from one aperture
(Fig. 6b). The peak electron density (Fig. 6c) increases
by roughly two times. This is again due to the balanc
ing act between the plasma generated by the two pulses
and their simultaneous selffocusing act similar to the
balance in the single filament case.
Two recent experiments trying to observed the
enhancement of intensity inside multiple filaments
crossing/fusion and the conclusion again is negative
[54, 55].
4. NONLINEAR INTERACTION OF INTENSE
FEMTOSECOND FILAMENTS IN AIR
4.1. Introduction
As two or more intense light filaments spatiotem
porally approach one another, there would be strong
interaction. This has attracted much attention for
many potentially important applications and for its
distinctive features of particlelike repulsion, attrac
tion, fusion, collision, and spiraling [5660].
It was recently demonstrated that the strong cou
pling between two noncollinear filaments could result

in the energy exchange between them [61, 62], which

could be further controlled by tuning the frequency
chirp of the incident pulses. Such kind of the energy
exchange is partly based on the dynamic plasma grat
ing [63] created by the intense noncollinear filaments.
Furthermore, it was demonstrated that the thirdhar
monic generation from the filaments could be dramat
ically enhanced based on the noncollinear filament
interaction [64, 65]. The filament interaction also was
demonstrated to be very useful for the nonlinear fre
quency conversion of ultrashort laser pulses [25]
including the efficient generation of ultraviolet pulses
[66, 67]. By optimizing the relative phases of the fila
ments, the interaction among multiple filaments was
demonstrated to be controllable [68].
In air, the filament interaction can be induced by
the instantaneous Kerr effect, the long lifetime plasma
effect, and the recently revealed fieldfree molecular
alignments and revivals. (The physics of molecular
alignment by a laser field and the revival of a molecular
wave packet will be discussed in detail in Section 8.)
The instantaneous Kerr effect is raised by the pure
electronic response which exists within the laser pulse
duration, as shown in Fig. 7 (black solid curve). This
kind of filament interaction occurs only when the fila
ment pulses are temporally and spatially overlapped. It
has been the most popular kind of filament interaction
that was studied in the recent past. Laser ionization
induced plasma which counterbalances the Kerr effect
also participates in the filament interaction. The
plasma influences the filament interaction as it is gen
erated by the laser pulse and slowly decays with a typ
ical lifetime of the order of several hundred picosec
onds to a few nanoseconds, as shown in Fig. 7 (grey
dashed curve). Filament interaction can also be
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

Strength, arb. units

(a)
CCD

Kerr effect
Plasma effect
Molecular alignment

0.2

11

f=1m
TFP

0
0.2
Time, ps

Objective

PumpFW
ppol.

f=1m
ProbeFW
spol.
(b)
~
~

1
0

50

100

HWP f = 1 m

150
Time, ps

Fig. 7. Different temporal ranges of the Kerr, plasma, and

molecular alignment effects for filament interaction in air.
The word strength along the vertical coordinate has dif
ferent meanings for different processes. It will be clear later
in this section and in Section 8. The inset shows that there
is a time delay of ~80 fs for the molecular alignment effect
with respect to the instantaneous Kerr effect, and the
plasma effect participates in the filament interaction since
its generation during the pulse duration. The zero time
corresponds to the location of the pulse peak.

induced by the noninstantaneous rotational Raman

effects related to molecular alignment of diatomic
molecules in air, as shown by the grey dashdotted
curve in Fig. 7. The molecules are impulsively aligned
by the ultrashort laser pulses through rotational
Raman excitation [69], which later dephase and revive
periodically after the extinction of the pump pulse.
(See Section 8 for detail.) Depending on the relative
orientation of the linear molecules with respect to the
field polarization, the refractive index experienced by
a second properly matched ultrashort filamenting
probe pulse is changed by following the molecular
alignment and revivals. This index of refraction modi
fication hence dominates the succeeding filament
interaction and is indeed fieldfree achievable. The
strength of the molecular alignment and revivals
decreases gradually due to the timedependent decay
of the excited coherent rotational states of the mole
cules and shows a typical lifetime of a hundred pico
seconds for air at room temperature as indicated in
Fig. 7.

CCD

DM
Objective

PumpFW
ppol.
ProbeSH
spol.

f=1m

BBO

Fig. 8. The experimental setup for interaction between fil

aments launched in the same direction in air. The wave
lengths of the pump and probe pulses are (a) both at
800 nm, and (b) at 800 and 400 nm, respectively. TFP: thin
film polarizer; HWP: half wave plate at 800 nm; DM:
dichromatic mirror: HR at 400 nm and AR at 800 nm; FW:
fundamental wave; SH: second harmonic wave.

approached, the refractive index experienced by each

filament increases due to the Kerr effect. The local
refractive index changes owing to the Kerr effect are
XPM

given by [70] nKerr_pump(r) = n2Ipump + n 2

nKerr probe(r) = n2Iprobe +

Iprobe and

XPM
n 2 Ipump, respectively, in the
XPM

pump and probe filaments, where the n 2 is the Kerr

nonlinear coefficient induced by the crossphase
modulation (XPM) of the pump (or probe) pulse. The
XPM

n2

of the parallel field component is 3 times larger

XPM

4.2. Filament Interaction due to Kerr Effect

than the n 2 of the orthogonal field component [10,

70, 71]. (See also Subsection 7.1.) Therefore, the
pump and probe pulses experience increased refractive
indices when they are temporally overlapped due to
the instantaneous XPM (XKerr) effect.

Proportional to the field intensity, the strength of

the electronic Kerr effect is described by nKerr ~ n2I,
where n2 is the Kerr nonlinear coefficient related to
the third order susceptibility (3) and I is the laser field
intensity. When two light filaments, named pump and
probe in the following, are spatiotemporally

4.2.1. Fusion of parallel filaments. The XKerr

effect based interaction can occur between filaments
of different polarizations and wavelengths. As a dem
onstration, the interaction of two filaments launched
in the same direction in air was studied as shown in
Fig. 8a.

LASER PHYSICS

Vol. 22

No. 1

2012

12

CHIN et al.
(a)
Probe/FW

(b)

(a)

Pump/FW

(d)

Plasma
array

Probe/SH
Initial
ortho.

Pump/FW

(b)

(c)

FW
Pr
ob

Pump/FW

Fusion
ortho.

100 nm

Initial
ortho.

FW

4f imaging system

CCD

Fusion
ortho.
(e)

(f)

(c)

Probe/SH

Initial
paral.

Fusion
paral.

Fig. 9. The fusion between two parallellaunched femto

second filaments of the same wavelength and orthogonal
polarizations [(a) and (b)], different wavelengths and
orthogonal polarizations [(c) and (d)], and different wave
lengths and parallel polarizations [(e) and (f)] around the
zero time delay due to the XKerr effect. The filaments
propagate along the z direction as indicated by the arrow.
The bottom sheets are the projections of the filaments. The
initially launched pump and probe filaments merged into
one as the fusion occurred.

The filament interaction was characterized by

imaging the ionizationinduced fluorescence of the
nitrogen molecules (see Section 9) in the filament
wake to a monochrome digital charge coupled device
(CCD) with a microscope objective (10). The fila
ment separation is defined by the distance between the
fluorescence peaks of the filaments. A more detailed
description of the experimental setup can be found in
[70, 72, 73]. As shown in Fig. 9a, two parallel pump
and probe filaments (~5 cm in length) at 800 nm with
an initial separation of 106 m were generated by
focusing the 1.6 mJ pump and 0.8 mJ probe pulses
with two separated fusedsilica lenses (f = 1 m) before
a thinfilm polarizer (see Fig. 8a).
The polarizations of the pump and probe filaments
were orthogonal to each other. As shown in Figs. 9a
and 9b, filament attraction and even fusion were
observed as a result of the XKerr effect around the
zero time delay [70, 72]. Furthermore, filament inter
action due to the XKerr effect could be observed for
filaments at different wavelengths, such as for the fun
damental wave (FW) at 800 nm and its second har
monic (SH) at 400 nm. As shown in Fig. 8b, a SH

~ 50 ps

~ 5 ps

Fig. 10. (a) The onedimensional plasma array created by

noncollinear interaction of two filaments. (b) The probe
system for the plasma array imaging. (c) The measured 1D
plasma array at different time delays after its generation.
The bottom sheets in (c) are the projections of plasma
arrays.

pulse was generated by using a 200mthick BBO

crystal, and the polarization of the FW pulse could be
rotated with a halfwave plate. The energies of the FW
and SH pulses were measured to be ~1.00 and 0.25 mJ
after the combining dichromatic mirror, which led to
two 100mseparated parallel filaments of ~6.5 and
14.0 cm, respectively. As shown in Figs. 9c and 9d,
attraction and fusion of the FW and SH filaments [73]
could be ensured by the XKerr effect as they were
temporally overlapped. Since the Kerr nonlinear coef
XPM
ficient n 2
induced by the XPM was larger for the
parallel polarizations than that of the orthogonal
polarizations, as shown in Figs. 9e and 9f, the filament
attraction and fusion were much stronger for the par
allel polarized filaments with more florescence inten
sity as compared to the orthogonally polarized fila
ments. As a result, for the same pulse duration and
intensity, the time ranges (the relative time delay
between the pump and probe pulses) for filament
attraction and fusion were much larger for parallel
polarized filaments than that of orthogonally polar
ized ones as demonstrated in detail in [73].
For parallel polarized filaments of the same wave
length, interference is involved which affects the fila
ment interaction. The Kerr effect induced refractive
index in the interaction region is increased and
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

decreased for the constructive and destructive interfer

ence, respectively. Therefore, by controlling the rela
tive phase of the filamenting pulses, the Kerr effect can
steer the filaments to be attracted or fused for inphase
filaments and repulsed for outofphase filaments,
respectively. This interesting filament interaction was
numerically predicted [74] and was very recently
experimentally confirmed [75].
4.2.2. Two crossing filaments. The filament interac
tion becomes more interesting for crossed filaments
when they are temporally synchronized. As shown in
Fig. 10a, two femtosecond laser pulses (50 fs, 1 mJ,
800 nm) were focused in air with two separated lenses
(f = 1 m), and crossed at an angle of ~7.
The constructive interference between the fila
menting pulses increased the field intensity in the
interaction zone, where the local refractive indices
were increased due to the Kerr effect. Correspond
ingly, more energy was attracted into the interference
peaks which further increased the field intensities. On
the other hand, the increase of the field intensities was
limited by the correspondingly increased local plasma
densities that defocused the interference peaks. Con
sequently, selfchanneled parallel subfilaments were
formed along the interference peaks. The inset of
Fig. 10a shows the onedimensional plasma
waveguide array could be created by means of interfer
enceassisted coalescence of the noncollinear fila
ments, which featured wavelengthscale periodic
modulation of the plasma density and could guide the
input pulses equivalently as a photonic crystal plasma
waveguide [63]. As shown in Fig. 10b, the plasma array
with periodically modulated refractive index could
also be probed by a timedelayed pulse crossed at a
right angle. The imaging system in 4f configuration
composed of two lenses enabled us to have the real
spatial distribution of the plasma array. The measured
plasma arrays at various time delays after its creation
are shown in Fig. 4c, indicating quite a long lifetime of
the plasma array. Such a plasma array could be
adjusted by changing the intensity ratios, crossing
angles, and field polarizations of the generating pulses
[63]. A maximal modulation depth of the plasma array
was achieved for filaments with parallel polarization
with an intensity ratio of 1:1, whose transverse period
was determined by the noncollinear crossing angle of
the incident beams.
Beyond the onedimensional plasma arrays, two
dimensional plasma arrays could be created by the
noncollinear interaction between three filaments. The
noncollinear filament interaction induced photonic
crystal plasma waveguides could be used to guide high
intensity pulses over long lengths in gases benefiting
from the periodically modulated plasma density.
LASER PHYSICS

Vol. 22

No. 1

2012

(a)
Probe/FW

13
(b)

Pump/FW
z

Initial

Repulsion
~ 300 fs

Fig. 11. The repulsion of the parallel probe filament by

plasma generated by the advancing pump for two parallel
launched femtosecond filaments. The probe pulse was
~300 fs delayed with respect to the pump pulse in (b) as
compared to that in (a). The filaments propagated along
the z direction as indicated by the arrow.

4.3. Filament Interaction due to Plasma Effect

The multiphoton ionization induced plasma also
participates in the filament interaction due to the neg
ative contribution of the refractive index change, pro
portional to nplas(r) = 0.5(r)/c, where is the
electron density and c (~1.7 1021 cm3 at 800 nm) is
the critical plasma density. As shown in Fig. 7, the
plasma is generated within the pulse duration and
decays gradually after the extinction of the laser pulse
with a typical lifetime of the order of a few hundred
picoseconds depending on its initial density. As it is
well established for single filaments, the multiphoton
ionization induced plasma acts as a defocusing lens
against the Kerr selffocusing to support long distance
propagation of light filament (see Section 2). For two
parallel launched filaments propagating closely in
space, the plasma of sufficient density in one filament
wake that acts as a negative optical potential also influ
ences the propagation of the other filament. For fem
tosecond filaments of orthogonal polarization at 800
nm launched in the same direction in air [see Fig. 8a
for the experimental setup, the pump and probe pulse
of 1.6 and 0.8 mJ were focused with two separated
fusedsilica lenses (f = 1 m) before the thinfilm polar
izer], filament repulsion by the plasma effect was
clearly observed [70, 72] at a time delay of ~300 fs as
shown in Fig. 11b, resulting in a larger filament sepa
ration as compared to Fig. 11a. Such filament repul
sion could also be observed for a wide range of time
delay until significant decay of the plasma density.
4.4. Filament Interaction due to Molecular Alignment
For a typical light filamentation in air, most of the
molecules survive in the filament wake due to weak
molecular ionization. These neutral molecules are
impulsively aligned by the intense ultrashort filament
pulse through rotational Raman excitation. In gen
eral, as shown in Fig. 12a, the molecular alignment is
characterized by the statistical term cos2 [69],
where is the crossing angle between the laser field

14

CHIN et al.
cos2

(a)

0.36

1/4 TO

Pump pulse

cos2 1/3

TO

0.34
0.32
0.30

TN

1/2 TN
0

Air

12

Time, ps
(c)

Paral. align
positive lens

4.0
4.5
Time, ps

m
r, m

Perp. align
negative lens

0.2
0
0.2

cos2 1/3

(b)
0.02

cos2 1/3

Falling edge
blueshift

0.02 Rising edge

redshift
4.0
4.5
Time, ps

Fig. 12. (a) The simulated molecular alignment metric cos2 of the N2 and O2 molecules in air as a function of time delay. The
blackdashed curve stands for the envelope of the pump pulse. (b) The spatiotemporal profile of the refractive index change expe
rienced by a linearly polarized ultrashort laser pulse (marked with arrows) at the half revival time of N2 as its time delay is tuned
to match the molecular alignment revival. (c) The timedependent phase modulation induced by the rapid evolution of the
molecular alignment revival.

polarization and molecular axis (see Section 8 for the

detailed physics of alignment and revival).
4.4.1. Principle of molecular alignment effect. The
impulsive kick of the ultrashort laser pulse leaves the
molecules aligned into an excited rotational wave
packet parallel to the lasers linear polarization. Peri
odic dephasing and revivals of the impulsive molecu
lar alignment follow due to the quantum beatings of
the excited rotational states in the wave packet. For
molecular N2 and O2 in air, the revival periods, deter
mined by the molecular rotational constants, are
T N2 = 8.3 ps and T O2 = 11.6 ps, respectively, as shown
in Fig. 12a. The inphase (positive peak value of
cos2 ) and antiphase (negative peak value of
cos2 ) beatings of the molecular wave packets cor
respond to the parallel and perpendicular alignment of
the molecules, respectively.
The great interest of the molecular alignment in fil
ament interaction originates from the orientation
dependent refractive index of the molecules, which is
controllable and modulates the spatiotemporal phase
of the properly matched filamenting probe pulse [76,
77]. The additional refractive index change induced by
the molecular alignment is nmol(r, t) =
2(0/n0)[cos2 (r, t) 1/3], where = ||
, 0, and n0 denote the molecular polarizability dif
ference between the components parallel and perpen

dicular to the molecular axis, initial molecular density,

and linear refractive index of randomly orientated
molecules, respectively. For molecular alignment par
allel or perpendicular to the field polarization with
cos2 greater or less than 1/3, the refractive index is
increased or decreased, respectively, as compared to
the randomly orientated molecules. Since the degree
of molecular alignment is proportional to the field
intensity before its saturation [78] and evolves rapidly
in the time domain, it will introduce an additional spa
tiotemporal phase modulation to the properly
matched ultrashort probe laser pulse [76]. This modu
lation becomes easily observable around the revival
periods as alignmentinduced change of the refractive
index follows the ultrafast evolution of the molecular
rotational states in the typical time scale of a few hun
dreds fs.
Excited by a femtosecond pulse of Gaussian
shaped transverse profile, Fig. 12b shows the numeri
cal simulation of the alignmentinduced refractive
index change nmol(r, t) for roomtemperature air
around the halfrevival time of molecular N2. The
refractive index change at the beam center is larger
than that at its periphery due to the larger molecular
alignment degree at the beam center. The parallel and
perpendicular molecular alignments with increased
and decreased refractive index change then account
for additional effects of crossfocusing (positive lens)
LASER PHYSICS

Vol. 22

No. 1

2012

Refractive index
change, arb. units

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

1.0

Paral. alignment
pos. potential
attraction

(a)

0
(b)

Pump

Probe

0.5

0.5

Perp. alignment
neg. potential
repulsion

Probe
1.0

Pump
0

15

70

0
r, m

70

70

0
r, m

70

Fig. 13. Schematic illustrations of (a) the positive refractive index potential and (b) the negative refractive index potential created
by the pump pulse induced parallel and perpendicular molecular alignment revivals, respectively. The properly matched probe
pulse propagates nearby the positive and negative refractive index potentials is attracted and repulsed by the pump pulse, respec
tively.

(a)
Pump/FW
Probe/FW
z
Attraction
paral. align.

Initial
ortho.
polarization

Repulsion
perp. align.
(b)

Pump/FW

Probe/SH
Attraction/fusion
paral. align.

Initial
paral.
polarization

Repulsion
perp. align.
(c)

Pump/FW

Probe/SH

Initial
perp.
polarization

Attraction
paral. align.

Repulsion
perp. align.

Fig. 14. The controllable attraction, fusion, and repulsion between two parallel launched femtosecond filaments by the fieldfree
molecular alignment revivals. The pump and probe filaments are set to have (a) perpendicular polarizations of the same wave
length at 800 nm, (b) parallel polarizations of different wavelengths at 800 and 400 nm, and (c) perpendicular polarizations of
different wavelengths at 800 and 400 nm, respectively. The filaments propagate along the z direction as indicated by the arrow.
The bottom sheets are the projections of the filaments.

and crossdefocusing (negative lens), respectively, as

shown in Fig. 12b. Such a crossdefocusing effect has
been used to characterize the fieldfree molecular
alignment [79] by measuring the molecular alignment
induced diffraction of a probe pulse.
Similar to the selfphase modulation induced by
the intensitydependent Kerr effect, the timedepen
dent revival of the molecular alignment introduces an
additional nonlinear phase shift mol to a properly
matched ultrashort probe pulse, corresponding to a
spectral phase modulation given by mol(t) =
mol/t ~ nmol(t)/t as shown by the black
dashed curve in Fig. 12c. Hence, the rising and falling
LASER PHYSICS

Vol. 22

No. 1

2012

edges of the molecular alignment revival account for

red and blue shifts of the properly matched ultrashort
laser pulse, respectively. This controllable spectral
modulation of a time delayed femtosecond laser pulse
propagating in prealigned molecules was demon
strated in [80], where additional red and blueshifted
frequency components were generated when its time
delay was tuned to match the rising and falling edges of
the molecular alignment revival. More details of spec
tral shifts will be given in Section 8.
4.4.2. Filament attraction and repulsion by molecu
lar alignment. It is clear that based on the orientation
dependent refractive index change of the molecules,

16

CHIN et al.

controllable optical potentials with modulated refrac

tive index can be created by the molecular alignment.
For a general spatially Gaussianshaped excitation
pulse with higher field intensity at the beam center
than its periphery, the refractive index optical potential
gradually decreases with the distance away from the
beam center. For an ultrashort laser pulse propagating
nearby the optical potential, as shown in Fig. 13, it is
attracted or repulsed by the positive or negative refrac
tive index potentials [70, 72], respectively.
Since pumpexcited molecular alignment and
revival could induce a change of index of refraction in
a probe pulse, filament attraction and repulsion would
occur in a way similar to the effect of XPM demon
strated in Fig. 9. This is shown in Fig. 14. An advanc
ing ultrafast pulse aligns molecules and a second time
delayed probe filamenting pulse experiences the
refractive index change in the molecular rotational
revivals, typically a few (or tens) picoseconds delay
from the aligning pulse. For two orthogonally polar
ized pump and probe femtosecond filaments at
800 nm launched in the same direction (see Fig. 8a for
the experimental setup), the parallel launched probe
filament (initially separated by ~106 m) could be
attracted by the pump for parallel molecular align
ment revival, resulting in a decreased filament separa
tion of ~94 m as shown in Fig. 14a around the half
revival of molecular N2 in air.
Furthermore, using pump and probe of different
wavelengths (800 and 400 nm) with either parallel or
perpendicular polarization, fusion or repulsion of fila
ments was observed at the full revival time of N2 mol
ecule in air, as shown in Figs. 14b and 14c. More
detailed discussions can be found in [70, 72, 73].
4.4.3. Some consequences of molecular alignment.
Recently, a cross(de)focusing assisted polarization
spectroscopic technique [81] was demonstrated to
measure the transient alignment of the molecules,
where both the spatial and polarization modulation of
the probe pulse by the molecular alignment are con
sidered. By detecting only a small part at the center of
the probe pulse, the crossdefocusing assisted polar
ization spectroscopy measurements show us whether
the molecules are transiently parallel or perpendicu
larly aligned. Meanwhile, the orientationdependent
ionization of the molecules during the pump pulse
excitation could be extracted from the measured
molecular alignment signal [81], which could there
fore be used to reveal the involved ionization during
the impulsive molecular alignment and might be
extended to characterize the weakly ionized filaments
in molecular gases.
For the low dispersion of the gaseous molecules,
the molecular alignment was recently proposed to
tune the central wavelength of fewcycle laser pulses
[82]. The fewcycle laser pulse was accurately put in an
advancing femtosecond filament channel in a molec
ular gas, such as air, where the molecules were impul

sively aligned. The central wavelength of the time

delayed fewcycle pulse could be effectively tuned with
desired red or blue shifts, while keeping the temporal
duration almost unchanged. Traditionally, wavelength
tunable fewcycle laser pulses can be obtained by
means of noncollinear optical parametric amplifica
tion, where a speciallycut, thin nonlinear crystal is
required and the spectral bandwidth is critically lim
ited by the phase matching condition. Nevertheless,
the timedependent molecular alignment revivals
break these limits for fewcycle pulses of tunable cen
tral wavelengths. By using the XPM of the molecular
alignment revivals, similar spectral modulations of the
timedelayed femtosecond laser pulses were also
observed [73, 83, 84].
Because of its advantages of controllable dynamics
and noninterference influence on the succeeding
processes, fieldfree molecular alignment has been
extensively studied for molecular orbital reconstruc
tion [85], ultrashort laser pulse generation [8689],
high harmonic generation [89, 90], and so forth. It was
recently demonstrated that the propagation dynamics
of a timedelayed ultrashort probe pulse could be sig
nificantly influenced by properly matching it to the
molecular alignment wakes in an advancing femtosec
ond filament wake [83], where periodical trapping and
diffraction of the probe filamenting pulse were
observed for the parallel and perpendicular revivals,
respectively. Similar trapping and destruction of
intense femtosecond filament were reported [84]. The
molecular alignment in the femtosecond filament
wake was also demonstrated as a polarization separa
tor [91] for a timedelayed ultrashort laser pulse.
The propagation of ultrashort laser pulses in pre
aligned gaseous molecules is dominated by several
effects in the addition to spatiotemporal XPM: the
spatial confinement or diffraction by the cross
(de)focusing of the parallel or perpendicular molecu
lar alignment which functions as a positive or negative
lens; the timedependent phase accumulations that
induces red or blue shifts for the rising or falling edge
of the molecular alignment revivals; the enhanced or
weakened selfphase modulation, selfsteepening, and
other high order nonlinear processes that contribute to
the spectral modulation. Assisted with the molecular
alignment, it was proposed [87] that a femtosecond
light filament could be dramatically enhanced with an
increased filament length and a significantly broad
ened spectrum, leading to the generation of a self
compressed fewcycle pulse. For the cross(de)focus
ing effect of the molecular alignment, the collapse dis
tance of a given incident laser pulse was increased and
decreased for the perpendicularly and parallel aligned
molecules, respectively. The effective nonlinearity of
the molecules is equivalently increased for the cross
focusing effect, which correspondingly decreases the
threshold for filamentation. For instance, in air,
Xwave light bullets can be generated along the intense
filament with input ultrashort laser pulses of several mJ
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

FW

17

Incident SH

FW
SH

Thin plasma grating

created by filament interaction

Diffracted SH

Fig. 15. The energy transfer of a timedelayed (delay time = 1 ps) secondharmonic pulse from the pump arm to the probe arm
by the pumpprobe filament interaction induced onedimensional plasma array.

(b)
Fraunhofer
diffraction

Onaxis
Noncollinear
phasematching

Spectral intensity, arb. units

(a)
1.0

With inter.
z = 100 cm

Without inter.
z = 102.5 cm

0.5
0
1.0 z = 105 cm

z = 107.5 cm

0.5
0
200 240 280 320 200 240 280 320
Wavelength, nm

Fig. 16. (a) The measured beam profile of the generated thirdharmonic pulse from the filament. The bottom sheet is the projec
tion of the beam profile. (b) The measured spectra of the generated thirdharmonic pulse from the probe filament at various prop
agation distances. z = 0 is located at the focusing lens and along the propagation direction of the FW pulse. The grey solid and
light grey dashed curves account for the cases when the noncollinear interaction with the pump filament is turned on and off,
respectively.

in energy [92], which is much higher than those in the

quadratic nonlinear media [93, 94]. However, by using
molecular alignment, it was shown [88] that Xwaves
could be generated in air for modest input pulse power
and temporal duration. Much more details on the few
cycle shocked Xwave with controllable rising or fall
ing edges can be found in [88].
4.5. Applications of Filament Interaction
4.5.1. Spatial transfer of probe energy. As men
tioned above, plasma arrays or waveguides could be
generated by noncollinear interaction of femtosecond
filaments ensured with the interference Kerr effect
(see Fig. 10). As demonstrated in [63], selfguiding of
the interacting pulses by the plasma waveguides could
result during their generations where the input femto
second pulses were partly guided by the periodically
modulated plasma waveguide and emitted from the
bisector. For the small crossing angle (few degrees) of
the noncollinear filaments, a thin plasma grating is
generated as shown in Fig. 15, which could be used to
diffract a timedelayed ultrashort pulse.
LASER PHYSICS

Vol. 22

No. 1

2012

The diffraction angle of the SH pulse (seeded along

the pump FW pulse) was governed by m = md/
(under the approximation of sinm m) (m = 0, 1,
2), where m denotes the mthorder diffraction
angle, d is the wavelength of the SH light wave, and
is the spatial period of the generated thin plasma grat
ing. As shown in Fig. 15, by propagating a time
delayed SH pulse (delay time is 1 ps) along one of the
FW filament, the first and second orders of the dif
fracted SH pulses locate along the bisector and propa
gation direction of another FW filament were
observed, respectively. A maximum energy transfer
efficiency of ~2.3% was observed due to the second
order diffraction of the created plasma grating by two
noncollinear FW filaments at a crossing angle of ~3.
The energy transfer efficiency decreased gradually as
the time delay was increased due to the expansion and
decay of the created plasma array. Therefore, efficient
energy transfer from one filament to another could be
ensured by the filament interaction assisted plasma
array [95], which functioned as a plasma grating to dif
fract the timedelayed ultrashort laser pulses.

18

CHIN et al.
100

Spectral intensity, arb. units

101
102

CO2

0.4
100
101

(a)

Input
Random, 2.0 atm
0.5 mJ, 2.0 atm
0.5 mJ, 1.8 atm

0.6

0.8

Random, 2.0 atm

0.57 mJ, 2.0 atm
0.57 mJ, 1.8 atm

(b)

O2

102
0.4
100
101

0.6

0.8

(c)

Random, 2.0 atm

0.57 mJ, 2.0 atm
0.57 mJ, 1.8 atm

N2

102
0.4

0.6
0.8
Wavelength, m

Fig. 17. The measured supercontinuum spectra at the out

put of the probe filament as its time delay was tuned to
match the parallel molecular alignment revivals of molec
ular (a) CO2, (b) O2, and (c) N2. The energy in each figure
indicates the 35fs pump pulses energy.

4.5.2. Efficient third harmonic generation. The self

guided filament was demonstrated to be a suitable
channel for nonlinear frequency conversion, such as
thirdharmonic generation (THG) whose phase is
nonlinearly locked to the FW filament pulse over a dis
tance much longer than the characteristic coherent
length [66, 96]. However, the energy conversion effi
ciency is still limited. As compared with a single fem
tosecond filament in air, it was recently demonstrated
that the THG could be significantly enhanced by non
collinear filament interaction [97], which was criti
cally dependent on the crossing angle, intensity ratio,
and relative polarization between the incident pulses.
For example, two noncollinearly crossed FW pulses
(0.85 mJ for each pulse, 50 fs) are focused by using two
separate lenses of the same focal length of f = 100 cm.
We measured the TH generation from one of the FW
pulses by changing the relative time delay, polariza
tion, and energy of the other incident FW pulse. Two
fused silica prisms were used to separate the generated
TH pulse from the FW. The beam profile of the gener
ated TH was taken by CCD. As shown in Fig. 16a, the
generated thirdharmonic (TH) pulse from the intense
selfguided filament shows a welldefined onaxis core
surrounded by multiple offaxis rings.

The onaxis TH core could be understood to have

originated from the FW filament with nonlinearly
modulated refractive index. The offaxis multirings
came from the offaxis noncollinear phase matching
between the onaxis FW and offaxis TH pulses as well
as the Fraunhofer diffraction of the generated onaxis
TH pulse [98]. For a single filament, the energy and
multiringed beam profile of the generated TH pulse
decreased rapidly and eventually disappeared with the
termination of the FW filament. However, in the pres
ence of filament interaction, the generated TH pulse
showed almost constant energy and a multiringed
beam profile over a much longer distance. Figure 16b
shows the measured spectra of the generated TH pulse
from the probe FW filament when its noncollinear
interaction with the pump FW filament was on and
off, respectively. The maximum enhancement factor
of the THG was achieved when the intensities of the
two pulses were approximately equal for the maximum
modulation depth of the field interference induced
grating. By adjusting the relative polarization of the
FW pulses from parallel to orthogonal, the THG
decreased rapidly as the field interference induced
grating was switched off. For an optimal crossing angle
of ~13, a maximal THG enhancement factor of ~174
in energy was achieved for parallel polarized input FW
pulses with an intensity ratio of one. Here, the
enhancement of the generated TH pulse was measured
by focusing in a photomultiplier tube (PMT).The
observed THG enhancement is understood to be a
consequence of the elongated filamentation due to the
noncollinear filament interaction [97].
4.5.3. Molecular alignment enhanced supercontin
uum generation. Usually, broadband supercontinuum
(SC) spectrum can be generated by intense filament in
air [98] but needs very high power pulses with multiple
filaments. Interestingly, it was recently demonstrated
[99, 100] that the SC generation by a timedelayed
femtosecond laser pulse with modest energy could be
readily controlled/enhanced by collinear interaction
with another advancing femtosecond filament through
the impulsive molecular alignment. At the parallel
revival of the molecular alignment created by an
advancing pump filament, the SC generation by the
timedelayed femtosecond probe pulse could be
turned on gradually with extended blueshifted fre
quency components as the degree of molecular align
ment increased. The parallel molecular alignment
induced crossfocusing effect and the consequently
promoted selfphase modulation as well as selfsteep
ening effect played important roles for the observed
control of the SC generation. For a 35fs probe pulse
with an input energy of 1.5 mJ focused in a gas cell
with a lens of f = 60 cm, Fig. 17 shows the dependence
of the output spectra of the collinearly propagating
probe pulse on the pump pulse energy and gas pressure
when its time delay was tuned to properly match the
parallel halfrevivals of the molecular CO2, O2, and N2
created by an advancing pump filament, respectively.
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

Among them, the molecular CO2 is the most suitable

molecule for broadband SC pulse generation using
molecular alignment, since its polarizability is
larger than those of the O2 and N2 molecules.
Meanwhile, the beam breakup of the probe pulse
due to the multifilamentation was suppressed as the
spectral energy was redistributed to the SC frequency
components. Eventually, a linearly polarized ultra
broadband SC pulse with the pulse energy of 1.2 mJ
was generated in a single filament core. The impulsive
molecular alignment in the wake of an advancing
pump filament provides us an additional tool to effec
tively control the SC generation from another time
delayed intense femtosecond filament in molecular
gases.
As the input energy of the timedelayed femtosec
ond filament further increased, highorder nonlinear
processes such as multifilamentation and multipho
ton ionization were much stronger to compete with
the SC generation [101]. This makes the highenergy
SC pulse generation different from the lowenergy
case. As very recently demonstrated [102], a stable SC
pulse of 1.8 mJ in a single core with the spectrum rang
ing from 400 to 900 nm can be generated when the
time delay of the 35fs probe pulse with the pulse
energy of 2.7 mJ (focused with a lens of f = 2 m) was
tuned to match the perpendicular molecular align
ment revival. It was quite different from the case when
the lowenergy probe pulse was used for the SC gener
ation by the parallel molecular alignment revivals [99,
100]. The crossdefocusing effect loosened the focus
ing condition of the highenergy probe laser pulse,
which increased the filament length. The elongation
of the filament length was indeed observed [87]. The
ionization loss was also decreased due to the reduced
ionization probability for the perpendicularly aligned
molecules [103]. The SC generation was therefore
enhanced, and a linearly polarized ultrabroadband
SC pulse with the pulse energy of 1.8 mJ was eventu
ally generated.
Since for most filamentbased applications, a long
filament length is desired, the above observation of fil
ament elongation is important. An example is shown
in Fig. 18. By propagating a probe filamenting pulse
with an input pulse duration of 35 fs and energy of
2.26 mJ in the perpendicularly aligned molecules cre
ated with an advancing pump filament, the probe fila
ment length were controllably increased from 27 to
51 cm as the input pump energy was increased [104].
Meanwhile, as compared with the case of randomly
orientated molecules, the onset position of the probe
filament moved backward due to the crossdefocusing
effect by the perpendicular molecular alignment
revival. Consistent with the highenergy SC genera
tion [102], the length of highenergy femtosecond fil
ament was elongated at the perpendicular molecular
alignment revival with the crossdefocusing effect.
However, for a lowenergy probe filament, as pre
LASER PHYSICS

Vol. 22

No. 1

2012

19

Fluorescence intensity, arb. units

Random
Pump: 1.2 mJ
Pump: 1.4 mJ

100

101
27 cm
51 cm

20

40
60
80
Propagation distance, cm

Fig. 18. The dependence of the probe filament length on

the input energy of the pump filament, which accounts for
the degree of the molecular alignment. The filament length
is characterized by measuring the fluorescence of the pho
ton ionized N2. In order to eliminate the influence of the
input pulse and supercontinuum emission, a dielectric
mirror (HR 800 nm @ 0) and a bandpass filter (UG11,
transmission 200400 nm) were placed in front of the
detector.

dicted in [87], filament length increase was observed at

the parallel revival of the molecular alignment with the
crossfocusing effect which enhanced the filament
dynamics.
4.5.4. Broadband energetic UV pulse generation.
Apart from the filaments in the near infrared range,
molecular alignment assisted filament interaction can
also be extended to the ultraviolet region. Highenergy
ultrashort ultraviolet pulses are significant for preci
sion spectroscopy measurements. The spectral band
width of the generated ultraviolet pulses is usually lim
ited by the phase matching condition of the frequency
conversion processes in nonlinear crystals, and the
energy conversion efficiency was influenced by the
group velocity mismatch between the ultrashort laser
pulses. These limits can be broken by using the molec
ular alignment assisted filament interaction, leading to
broadband highenergy ultraviolet pulse generation.
As shown in Fig. 19, based on group velocity compen
sated cascaded secondorder nonlinear frequency
mixing processes in BBO crystals with input FW pulse
at 800 nm (13.8 mJ/70 fs), a high energy TH pulse of
~1.1 mJ per pulse at 267 nm was generated with a
spectral bandwidth of ~2 nm.
The spectrum of the UV pulse was then ~6.5 times
broadened by interacting with the molecular align
ment revivals in the intense filament wake of a FW
pulse (7.0 mJ/70 fs). The FW pulse used for molecular
alignment was actually the residual part of the FW
pulse after the TH generation, which separated from
the TH pulse by using dichroic mirror. The filamenta
tion of the UV pulse was enhanced by the parallel half

1.1 mJ
FWHM = 2 nm

0.5
0

264
276
288
Wavelength, nm

1.0
0.5
0
264
276
288
Wavelength, nm

267 nm

Phasecompensated
cascaded frequency mixing

CO2
Molecular
alignment

800 nm
7 mJ/70 fs

800 nm
13.8 mJ/70 fs

1.0

Spectral intensity

CHIN et al.
Spectral intensity

20

1.1 mJ
FWHM = 13 nm
Beam profile

Fig. 19. Schematic illustration of the experimental setup for highenergy broadband third harmonic generation by cascaded fre
quency mixing and subsequent molecular alignment based filament interaction. The pulse energies of the 70fs FW pulses used
for TH generation (left) and molecular alignment (bottom) were 13.8 and 7.0 mJ, respectively.

Movable table
800 nm/50 fs
L1 L2

Wedge
Integrating sphere
Filament zone
Spectrometer

Fig. 20. Measuring the spectra in the forward direction of a filament. L1: pianoconvex lens (f = 50 cm); L2: biconcave lens (f =
25 cm); beam diameter: 2.75 mm (after L2). The beam is collimated after L2 at low energy. Distance in Fig. 21 is measured from
lens L2.

revival of the prealigned CO2 molecules, which

broadened the output spectrum of the UV pulse. As a
result, energetic broadband UV laser pulse with pulse
energy of ~1.1 mJ and spectral bandwidth of ~13 nm
(full width at half maximum) could be generated as
recently demonstrated in [105].
4.6. Summary of Filament Interaction
Filament interaction, either by the instantaneous
Kerr effect, long lifetime plasma effect, or impulsive
molecular alignment, definitely enriches the laser fila
ment physics; it will bring much more promising
applications in ultrafast optics in the near future.

Since then, efficient generation of UV through the

same 4MW scheme as [25] was reported in gases [110,
111]. The UV pulse energy was much higher than that
from the hollow core fiber method while the alignment
was considerably simpler since no fiber was used.
Recently, broad band infrared supercontinuum gener
ation through 4WM was reported [112, 113]. Terahertz
(THz) pulses are generated and will be discussed in a
separate section below. Also, a Raman type of spectral
red shift of the pump wavelength (800 nm) was
observed in air [114]. A brief description of these new
developments (apart from THz) is given below.
5.1. Raman Type Spectral Shift

5. FILAMENTATION NONLINEAR OPTICS:

RAMAN SPECTRAL SHIFT, FOURWAVE
MIXING AND SUPERCONTINUUM
Filamentation nonlinear optics was proposed to be
a new way to do high intensity nonlinear optics in gases
[106] in relation to efficient third harmonic generation
[107, 108] and fourwavemixing (4WM) [25, 109].

The spectra from inside the filament of a fs Tisap

phire laser pulse in air was measured in the forward
direction as shown in Fig. 20. The collimated laser
pulse underwent free propagation and filamentation.
The wedge was set at a near grazing angle on a movable
table and partially reflected the pulse towards the inte
grating sphere followed by a spectrometer.
LASER PHYSICS

Vol. 22

No. 1

2012

Intensity, arb. units

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

460 cm
370 cm
330 cm
280 cm
270 cm

10

21

280 cm
370 cm
460 cm

103

102

260 cm

700

750

800
850
900
Wavelength, nm

950

700

800
900
Wavelength, nm

1000

Fig. 21. Spectral red and blue shifts of a filamenting fs laser pulse as a function of the position inside the filament. The distances
were measured from lens L2 in Fig. 20.

LASER PHYSICS

Vol. 22

No. 1

2012

Halfangle divergence, mrad

quency down shift along the propagation of the pulse

inside the filament zone. In the case of a fiber, all
pulses are confined inside the fiber core. In the present
case of filamentation, we suspect that the Raman
pulse, once created inside the filament, is phase
Frequency, THz
102

20

103
Xwave
4 WM

40

1E 0
1E3
1E6

Pump

30

TH

1E9

20
1E12

10

0
1.00030
1.00028
1.00026
15

Index of
refraction

It was discovered [114] that the spectrum under

went strong continuous selffrequency red shift in air
as the filament length increased. The filament started
at 260 cm and ended at around 500 cm from L2.
Before the filament is formed, there is no spectral
shift. Figure 21 (left) shows the spectral shifts with a
linear scale in the vertical axis. There is a new spec
trum moving out from the peak spectral position
towards the red side as the position of the detector
(wedge + integrating sphere + spectrometer) moves
through the filament in the propagation direction.
Figure 21 (right) shows three spectra at three different
positions in the filament using a log vertical scale. Both
the Stokes (red) shift and antiStokes (blue) shift are
evident.
When using argon gas, such shifts disappeared indi
cating that the shifts pertained to molecular gas. This
suggests that the Raman effect would be the reason.
Also, the measured continuous self frequency down
shift during filamentation in air has a strong depen
dence on the length of the filament. It was explained in
[114] that this phenomenon is similar to the observa
tion of continuous Raman selffrequency down shifts
in a fiber [115118].
The first Stokes shift is due to rotational effect from
N2 (12 cm1) and O2 (8.4 cm1) corresponds approxi
mately to a wavelength shift of 1.5 and 1.1 nm, respec
tively. These Stokes shifts are much smaller than the
initial spectral bandwidth of the laser pulse, and thus,
we cannot resolve them. During filamentation, the
cascaded Stokes shift occurred and the generated IR
pulse is continuously shifted towards longer wave
length. Because the spectral bandwidth of the newly
generated IR pulse is much wider than the first Stokes
shift due to rotational effects from N2 and O2, the low
frequency spectral components of the pulse, acting as
a seed, can be amplified by Raman gain while the
highfrequency spectral components of the same pulse
work as a pump. This results in continuous self fre

4
2
0.8 0.4
Wavelength, m

0.2

Fig. 22. (a) Angularly resolved spectrum generated during

the filamentation in air by laser pulses having an energy of
60 mJ and a pulse duration of 42 fs. The horizontal axis is
a logarithmic scale for the frequency. The false colors asso
ciated with the spectrum are related to the relative spectral
intensity. The gray curve is the angular position of the CE
according to the Xwave dispersion equation, and the
white curve is the angular position of the infrared CE
according to the offaxis 4WM process. The white arrows
point out the central frequency of the pump pulse and its
TH. (b) Linear refractive index of air calculated using a
Sellmeiertype model for a pressure of 1 atm, a tempera
ture of 15C, and a relative humidity of 40%. Adapted with
permission from [112] ( 2008 OSA), image courtesy of
Thberge, Defence Research and Development Canada
Valcartier.

22

CHIN et al.
0.01
(a)
200

Delay, fs

100
0
100
200
300
400

0.1

Fluence, arb. units

0.08 (b)
0.04
0
0.8 (c)
0.4
0
0.2 (d)
0.1
0
0.10 (e)
0.05
0
5 6 7 8 9 10 11 12
5 6 7 8 9 10 11 12
Wavelength, m
Wavelength, m

Fig. 23. (a) FIR spectrum generated during the cofilamentation in air of a 20mJ NIR laser pulse and a 1mJ SH laser pulses. The
vertical axis is the relative delay between the pump pulses. A positive delay corresponds to the SH pulse in front of the NIR pulse
and viceversa for the negative delay. The zero delay was arbitrarily set to the optimum FIR energy. The false colors associated
with the spectrum are related to the relative spectral fluence. The white dashed lines indicate the delay of the measured spectrum
for (b) 200, (c) 0, (d) +200, and (e) +365 fs. Adapted with permission from [113] ( 2010 APS), image courtesy of Thberge,
Defence Research and Development Canada Valcartier.

locked to the pump pulse. Thus, the Raman pulse stays

inside the filament (i.e., propagates together with the
pump pulse inside the filament core) while undergoing
cascaded Stokes shift. Once outside the filament core,
linear diffraction would make the pulse intensity too
weak to undergo further cascaded Stokes shift.
5.2. UltraBroad Band Supercontinuum Generation
Focusing the 800 nm pump pulse at 1.4 TW
(60 mJ/42 fs) in air with a 4 m focal length concave
mirror generating multiple filaments, ultrabroad
band conical supercontinuum from ~ 200 nm to
~14 m in the infrared was observed by Thberge et al.
[112]. The band width was so broad that five different
types of spectrometers had to be used in different
wavelength ranges. Longer wavelengths in the infrared
have larger divergence at the Stokes side of the 800 nm
pump while shorter wavelengths in the UV at the anti
Stokes side also have larger divergence, though not as
large as that of the infrared. Figure 22 shows the mea
sured angular resolved spectrum.
There are two principal physical processes respon
sible for the spectrum according to the authors.
Towards the antiStokes side of the pump wavelength,
normal conical emission or the socalled Xwave
model [119] could explain the observation. In the
Stokes side of the pump wavelength of up to 14 m, a
4WM model involving 2 pump photons and one blue
shifted photon at the antiStokes side of the pump
explains well the large divergence.

5.3. TwoColor Filament Generation

of Broadband Infrared Pulses
A special scheme [113] was designed in which an
initially ~800 nm/38 fs laser pulse (diameter ~2.5 cm
FWHM) was spectrally split into two pulses centered
at around 795 and 813 nm by using a dichroic mirror.
The latter had a smooth cutoff reflectivity at 805 nm.
Thus, this mirror reflected a pulse centered at 795 nm
while the transmitted a pulse centered at 813 nm. The
transmitted 813 nm beam passed through a KDP dou
bling crystal. The resulting second harmonic (SH)
pulse was focused collinearly with the reflected 795
nm near infrared (NIR) beam by a silver coated con
cave mirror (focal length: 1.5 m). The 795 nm pulse
generated a filament in air. Before the filament, the
pulse duration of the 20 mJ, 795 nm pulse was 50 fs at
FWHM (transform limited), while the 1 mJ, SH pulse
was positively chirped to 130 fs (transformlimited
pulse duration was 70 fs). By changing the delay
between the two pulses, pulses whose spectrum
spanned from 5 m in the infrared to beyond 12 m in
the far infrared (FIR) were generated. The spectral
content varied with the delay time. Figure 23 shows
the results.
The physical mechanism of generating these infra
red pulses was analyzed using both the 4WM model
and the photocurrent model. It was found that the
4WM model was the dominant physical process
responsible for the above observation [113].
6. TERAHERTZ PULSE GENERATION
IN FILAMENTS
This is another example of filamentation nonlinear
optics. Very broadband THz pulses can be generated
from filaments in air or other gases. Because of its
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

potential practical use in remote sensing of materials

such as explosives, it is a popular subject of scientific
and technical investigation.

1.0

LASER PHYSICS

Vol. 22

No. 1

2012

THz waveform, arb. units

0.5
0

0.5

0.5
= 15
= 105

1.0

Y, = 105

6.1. General
Generation of intense terahertz (THz) pulses with
large bandwidth from inside the filament in air is an
active area of current research on THz science. Such
broadband and rather powerful THz pulses would pro
vide a new prospective tool for remote THz nonlinear
optics and spectroscopy. THz pulses have the potential
to combine safetouse highresolution imaging and
identification through spectroscopy; however, con
ventionally generated THz pulses from emitters such
as semiconductor antennae or nonlinear crystals can
not propagate over a long distance in the atmosphere
because of beam diffraction and strong attenuation
due to water vapor. THz generation from femtosecond
laser remote filamentation in gases, especially in air, is
a promising alternative.
Singlecycle THz generation from a photo
induced plasma was first demonstrated by Hamster
et al. in 1993 by focusing laser pulses with a power of
1 TW and a duration of 100 fs into He gas [120]. Since
the emitter target is a gas, which is endlessly replenish
able for each laser shot, there is no damage threshold
limit at high intensities. The THz emission mecha
nism in their experiments was based on the space
charge separation by the ponderomotive force gener
ated by the optical beam [121], leading to a conical
THz emission. Since then, other techniques have been
used to generate stronger THz emission. Lffler et al.
[122, 123] applied a DC bias to the plasma region that
resulted in an orderofmagnitude increase in the THz
field. At the same time, Cook and Hochstrasser[124]
demonstrated a fourwave rectification method by
focusing the fundamental wave (FW) and secondhar
monic wave (SHW) of an amplified Ti:sapphire laser
into air. More than 3 orders stronger THz signal has
been recorded compared with only fundamental wave
pump.
Since then, intense THz generations by the two
color pump have been investigated extensively [125
141]. THz electric field as high as 400 kV/cm [126]
and a superbroadband THz spectrum up to 75 THz
( = 4 m) with J pulse energy [135] have been
reported. Dai et al. [129] reported that THz emission
from a twocolor plasma filament could be amplified
by another twocolor filament in nitrogen gas through
fourwavemixing parametric processes during the
plasma formation. Later, Chen et al. [130] demon
strated that the intensity of the terahertz wave
increased as the number of twocolor air plasmas
increased. It was explained as the coherent superposi
tion of the THz waves generated by each individual air
plasma. The terahertz wave amplitudes from two
color filament can also be enhanced by more than
eight times with an optimal aperturelimited pump

23

0
1
Time, ps

0.5
1.0
Measured THz polarization
Fit

1.0

0.5
0
X, = 15

0.5

1.0

Fig. 24. THz polarization from singlecolor filament. Solid

squares and solid curve are experimental data and simula
tion, respectively. The arrow indicated the rotating direc
tion of the THz electric field vector. The inset shows typi
cal experimental THz electric field waveforms obtained at
two orthogonal directions (solid triangles: = 15 and
solid circles: = 105).

laser beam [136]. Blanchard et al. reported more than

2 times increase in THz generation efficiency from
twocolor plasma by replacing the focusing lens with a
gold offaxis parabolic mirror [140]. More recently, by
using a converging lens followed by an axicon resulting
in a tailored twocolor laser filament in air, Manceau
et al. reported significant THz pulse shortening from
the uniform plasma strings [141]. For the physical
mechanism of the generation of the THz pulses in a
gas filament, two theoretical models have been
reported. The first model frequently used to explain
the THz emission is fourwave mixing [124127]
based on the third order nonlinearity. The second is
the microscopic polarization model [134, 135], which
has been suggested to interpret the underlying dynam
ics, attributing the THz emission to the freeelectron
drifting current driven by the combined field of the FW
and its SHW. We call this photocurrent model.
Recently, Karpowicz and Zhang [137] interpreted the
terahertz radiation in two steps based on the photo
current quantum mechanical model. First, terahertz
photons are emitted due to the acceleration of the
electron wave packet which is given a net dipole
moment through the asymmetry introduced by the
twocolor field. Next, when these wave packets collide
with neighboring atoms, they emit bremsstrahlung in
the THz range. The theory can explain the single
cycle (broadband) electric field and the following
echo in the THz waveform. The polarization prop
erties of THz emission have also been extensively
investigated. The FWM model describes them reason
ably well [127, 131133]. Typical focal lengths used
above for intense THz emission from twocolor
scheme are from 5 cm to 25 cm.

24

CHIN et al.
(a)

0.8

(b)

0.2

THz without Eext.

0.2

0.25 kV

0.2 0

0.2

0.2

(e)

0.2
Eext.

1.5

0.4

0.2

Fit: e = 0.25, = 15

0.8
0.8
0.8 (f)

1.5
1 kV

0.8

0.2
0.2 0
X

0.2

0
1 kV
Fit: e = 0.05
= 88

0.4
0.8
0.8

1.5

0
0.25 kV
Fit: e = 0.05
= 88

5 kV

3.0
3
3.0 (g)

0.4

(d)

3.0

0.4
Y

0.2

(c)

0.2

0
X

0.8

5 kV
Fit: e = 0.05
= 88

1.5
3.0
3

0
X

Fig. 25. Polarization trajectories of THz emission from a filament with a DC field of (a) 0, (b) 0.25, (c) 1, and (d) 5 kV/cm. THz
polarizations in (e, f, g) were obtained at 0.25, 1, and 5 kV/cm by subtracting the waveforms without DC field, respectively.

The other important reason why THz generation

from femtosecond laser induced plasma filaments in
air recently attracts much attention [142159] is that
THz emission can be generated close to a remote tar
get. This is because the onset of the filament to remote
distances can be controlled, in principle, by the initial
laser parameters: beam diameter, divergence and pulse
duration, etc. Remote THz generation minimizes the
importance of propagation issues as mentioned above.
DAmico et al. reported the first remote THz genera
tion from the filaments induced by the FW of a Ti:sap
phire based laser system at a distance of 30 m from the
laser in 2007 [142]. A radially polarized forward coni
cal emission was observed whose origin they attribute
to a Cherenkovtype mechanism from the laserpulse
induced plasma wake. Their conclusion was based on
the measurement obtained with a heterodyne detector
(sensitive only to 0.1 THz with a bandwidth of 4 GHz).
Later, by using an electrooptic sampling (EOS)
technique (sensitive to frequencies below 4 THz), it
was demonstrated that the THz pulses in the forward
direction of a filament are elliptically polarized [143,
144] as shown in Fig. 24. The physical mechanism is
that filament induced birefringence in gases provides a
phase delay between the two orthogonal components
of the THz field, leading to an elliptically polarized
THz emission.
6.2. External DC Electric Field Effect on THz
from SingleColor Filaments

aments, leading to 3 orders enhancement of THz

amplitude with 10 kV/cm external electric field [145].
In the case of transverse external electric field, the
polarization of this enhanced terahertz signal was
found to be collinear with the external DC field. The
physical origin is that the external electric field sepa
rates the electrons and ions in the plasma filament,
which results in a transverse current responsible for
terahertz emission with polarization parallel to the
direction of the applied external electric field [146].
Typical THz polarization trajectories are depicted in
Fig. 25. Without external electric field, an elliptical
THz polarization trajectory similar to the results in
[143] was observed (Fig. 25a). In the presence of the
external electric field, the THz emission changed its
polarization trajectory with the increase of the applied
electric field, Figs. 25b25d. When the THz electric
field obtained without external electric field was sub
tracted from the corresponding ones obtained with
external electric field, the recomposed THz polariza
tion trajectories were surprisingly uniform: linearly
polarized along the orientation of the external electric
field, as shown in Figs. 25e25g. The total THz emis
sion from a DCbiased singlecolor laserinduced fil
ament could be decomposed into two independent
processes: THz generation by the neutrals in the strong
laser field inside the filament which is elliptically
polarized [143] and THz induced by the externally
applied electric field through acceleration of the free
electrons from inside the filament which corresponds
to a linearly polarized THz source [146].

External electric fields were applied transversely

[145, 146] and longitudinally [147] to singlecolor fil

By taking the Fourier transform of the THz fields,

it is found that the central frequency of this linearly
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

800 nm separated by less than 3 ns, forming two over

lapping filaments in air (so called bifilamentation) can
generate 1 order of magnitude more intense THz
emission than transientCherenkov THz emission
[142] from the plasma filaments. The origin is attrib
uted to the emission from a bimodal transmission line
created by a pair of neighboring plasma filaments.

THz pulse energy, nJ

Two Si
Two Si and teflon
102

101

Twocolor (FW and SHW) femtosecond laser fila

mentation in air was also investigated to generate high
energy THz emission under loose focusing condition.
Under a fixed high energy pump of 24 mJ from a Ti
sapphire laser, Wang et al. [149] observed more than
4 times enhancement of THz pulse energy by chirping
the 42 fs transform limited pump pulse either nega
tively or positively to around 150 fs as shown in Fig. 26.
The reason why there is an increase of THz pulse
energy when the pulse is chirped is because, at the
same energy, the transform limited pulse has a higher
peak power. Thus, initial hot spots in the beam profile
will selffocus earlier during the propagation. In the
focusing geometry, multiple filaments will occur closer
to the focusing optic (farther from the geometrical
focus) in a zone of larger diameter. Filament competi
tion [28] would occur in this case because the fila
ments are far apart from each other and would
reduce the effectiveness of generating strong fila
ments. With longer chirped pulses, the peak power
being lower, multiple filaments occur nearer the geo
metrical focus where the diameter of the beam is
smaller. This smaller diameter favors multiple fila
ments cooperation (stronger constructive interfer
ence) resulting in stronger childfilaments [150], and
hence more THz emission. When the chirp becomes
even larger, the pulse duration becomes so long that
the peak power decreases to the extent that the number

100
210

100

42

110
240
Pulse duration, fs

Fig. 26. Pump pulse duration dependence of THz emission

with different filters in air. The vertical dashed line denotes
the Fourier transformlimited pump pulse duration. Dif
ferent filters indicate the different THz spectra measured.

polarized THz field is independent of the external

voltage. This means that it is the weak plasma inside
the filament that is set into motion at its own charac
teristic plasma frequency. Because of intensity clamp
ing, the plasma density is constant and hence the
plasma frequency is constant. Apparently, this latter
observation does not support the claim of THz ampli
fication in an external electric field [145]. Only the
characteristic plasma frequency is amplified.
6.3. Enhancement of THz from Filaments
Liu et al. [148] demonstrated that by sending a
sequence of two femtosecond IR laser pulses at
(a)

Xe

330

1.6

30

270

90

THz signal, V

1.2

210

0.8

330

30
60

300
270

90

210

180

150

330
270

30

90 300

150

330

60

20
40

300

60
80

Xe
30

300

120 270
180

280
60

210

180 150

30

120 240

240

0.4

300

340
320

120

240

(b)

60

300

330

25

260

90
270

90

120

240

240
210

180 150

100

60

120
210

120

240
140

220
200

180 150

180

160
75 mbar
880 mbar

200

400

600
800
Pressure, mbar

1000

1200

Fig. 27. THz power as a function of xenon gas pressure (a) and the linear polarization state of THz pulse at low pressure to the
eircular polarized THz at 880 mbar (b). The insets in (a) show the THz polarization states at different pressure levels. Adapted
with permission from [151] ( 2010 OSA), image courtesy of J.M. Manceau, Foundation for Research and TechnologyHellas
(FORTH).
LASER PHYSICS

Vol. 22

No. 1

2012

CHIN et al.

E800
E400

15

(b)
EDC
E800
E400

0
6 4 2 0 2 4 6
Horizontal DC field, kV/cm
(d)
Vertical polarization
Horizontal polarization

(e)

(g)

0 kV/cm
1 kV/cm
3 kV/cm
5 kV/cm

Vertical polarization
Horizontal polarization

15
+

5 kV/cm
3 kV/cm
1 kV/cm

THz electric field,

arb. units

EDC

(c)
THz pktopk amplitude, arb. units

(a)

6 4 2 0 2 4 6
Vertical DC field, kV/cm

2 3 4 5 6
Time, ps
(f)
5 kV/cm
3 kV/cm
1 kV/cm

0 kV/cm
1 kV/cm
3 kV/cm
5 kV/cm

2 3 4
Time, ps

5 6

Spectrum amplitude,
arb. units

26

5 kV/cm
4 kV/cm
3 kV/cm
2 kV/cm
1 kV/cm
0 kV/cm

0.5 1.0 1.5 2.0 2.5 3.0

Frequency, THz
(h)
5 kV/cm
4 kV/cm
3 kV/cm
2 kV/cm
1 kV/cm
0 kV/cm

0.5 1.0 1.5 2.0 2.5 3.0

Frequency, THz

Fig. 28. THz peaktopeak amplitude (c, d), waveform (e, (f) and frequency spectra (g, h) from a twocolor laserinduced fila
ment in air with horizontal (a) and vertical (b) external DC field. The THz waveforms and frequency spectra are for the compo
nents polarized along to the direction of external DC field.

of strong filaments will reduce, and thus, decrease the

THz signal.

pressure of 75 mbar was changed to circular polariza

tion at high pressure of 880 mbar (Fig. 27b).

6.4. Coherent Polarization Control of THz Pulses

from TwoColor Filaments

6.5. External DC Electric Field Effect on THz

from TwoColor Filaments

By changing the optical phase between the FW and

its SHW, the polarization of THz emission from two
color filaments in air can be coherently controlled
when at least one of the optical pump pulses is ellipti
cally polarized [138]. This phenomenon was simulta
neously reported by Wen and Lindenberg [139] and
Dai et al. [138] either through the positioning of the
frequency doubling crystal along the laser propagation
axis [139] or using an inline phase controller [138]. In
particular, when both the FW and the SHW are circu
larly polarized, the THz polarization angle can be
arbitrarily rotated by changing the phase between the
FW and the SHW, with the THz amplitude kept con
stant [138]. More recently, Manceau et al. [151] dem
onstrated the THz polarization control via the accu
rate control of the surrounding gas pressure of two
color filaments. The change of gas pressure will induce
a wavelength dependent change of refractive index,
resulting in the relative phase change of the two laser
fields. As expected, the polarization of the THz pulse
is rotated as the pressure of the gas medium is changed
in nitrogen gas. Beyond the rotation of the THz pulses
polarization, elliptically polarized THz pulses at dif
ferent elliptical states and up to fully circular polarized
THz pulses have been observed in xenon gas as shown
in Fig. 27. Figure 27a shows the THz radiation as
function of xenon gas pressure. The polarization rota
tion of THz pulse was observed by increasing the gas
pressure. The linear polarization of THz pulse at low

More recently, the generation of THz emission

from a DCbiased twocolor femtosecond laser
induced filament in air was systematically investigated
[152]. THz pulses were characterized using the EOS
technique with the external DC field parallel
(Fig. 28a) and perpendicular (Fig. 28b) to the FW
polarization direction, respectively. THz peakto
peak amplitudes as a function of the external DC field
are shown in Figs. 28c and 28d. The external DC field
could amplify only one THz component whose polar
ization is parallel to the direction of the external DC
field. The higher the external DC field is, the stronger
the THz emission is. The THz electric fields parallel to
the DC field are depicted in Figs. 28e and 28f. The
THz waveform can be modulated by controlling the
external DC field. The corresponding Fourier spectra
of THz emission are shown in Figs. 28g and 28h. Sim
ilar to the case of DC field biased singlecolor fila
ment, the peak frequencies stay at around 0.54 THz
when increasing the external DC field for both cases.
The peak frequency without DC field is around
0.9 THz. These two different peak frequencies indi
cate two different generation processes, one induced
by external DC field alone and the other only bichro
matic excitation. Polarization plots similar to Fig. 25
were obtained which reconfirms the above claim
[152]. Therefore, the total THz emission could be
interpreted as a sum of two contributions. One is the
linearly polarized THz component induced by the
external DC field with polarization parallel to the
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

0
2.7
3.0 3.3
Delay time, ps

THz amplitude

2.4

1/4 TO2

0.1

0.5

0.92

0.88
1/2 TN2

0.84
3.6

1.02

1.05 (d)

1.0

1.00

0.5

0.95

(e)

TN2+ 3/4 TO2

0.85

1.0

7.6 8.0 8.4 8.8 9.2 9.6

Delay time, ps

0.2

0.95

0.5

0.94

1.0

0.93

TO2 + 3/2 TN2

0.99

0.2
5.2 5.6 6.0 6.4 6.8
Delay time, ps
1.0
0.5
0

0.96

0.5

0.90

0.4

0.96

4.0 4.4 4.8

Delay time, ps

Air R(t)
THz amplitude

0.94

0.96

1/2 TO2 + 3/4 TN2

0.97 (c)

1.0

Air R(t)

0.95

(b)

Air R(t)

0.1

0.96

1.00

Air R(t)
THz amplitude

0.2

(a)

Air R(t)
THz amplitude

THz amplitude

0.97

27

0.5
0.93

1.0
11

12
13
Delay time, ps

Fig. 29. Peaktopeak: amplitude of THz electric field as a function of delay time around different air molecule revivals: (a) quar
ter revival of O2; (b) half revival of N2; (c) half revival of O2 and threequarter revival of N2; (d) full revival of N2 and threequarter
revival of O2; (e) full revival of O2 and one and half revival of N2. Solid line in (ae) is the calculated response function R(t) of
molecular Raman alignment in air (80% N2 and 20% O2).

direction of the DC field; it corresponds to the plasma

frequency of the filament. The other is an emission
from the twocolor laserinduced filamentation due to
the neutrals; i.e., 4WM.
However, depending on the external DC field
polarity, the THz component parallel to the DC field
was slightly asymmetric in Fig. 28c and a minimum
was observed around +1 kV/cm. This small decrease
of the THz amplitude, as compared to the case with
out DC field, could be attributed to partial destructive
interference between the two THz sources. On the
other hand, if the photocurrent [134] did partly con
tribute to the THz generation from the twocolor
pump pulses, the external DC field with positive
polarity could cancel the freeelectron drift at
+1 kV/cm, and thus slightly decreases the total THz
amplitude by leaving alone the contribution from the
fourwave mixing. Since the THz emission from the
electric current induced by an external DC field is
mainly independent of the THz emission generated by
the bichromatic excitation, it could be reasonably
concluded that THz emission from the twocolor fila
mentation is mostly due to the fourwave mixing pro
cess.
6.6. Molecular Alignment Effect on THz Emission
from Filaments in Air
During the filamentation of femtosecond laser
pulses in air, most of the molecules survive in the
plasma filament due to weak molecular ionization
LASER PHYSICS

Vol. 22

No. 1

2012

[27]. However, the laser pulses with clamped intensity

of ~5 1013 W/cm2 in the filaments are strong enough
to align the neutral molecules through rotational
Raman excitation, resulting in fieldfree periodic
revivals. The typical revival periods of N2 and O2 mol
ecules in air are 8.4 and 11.6 ps, respectively. Recently,
a control of THz emission from bifilamentation of two
FW pulses based on molecular lensing effect was char
acterized by a heterodyne THz detector, which is sen
sitive at 0.1 THz with a bandwidth of 4 GHz [153].
More recently, Wu et al. [154] and Wang et al. [155]
reported the molecular alignment effect on THz emis
sion from twocolor filaments in air. Air molecules are
prealigned by FW pulses [154] or a pair of twocolor
laser pulses [155]. In the latter case, the first pair of
twocolor pulses (800 nm/400 nm) prepared the fila
ment. THz emission is generated by the second pair of
twocolor laser pulses. By tuning the delay time
between the two pairs of twocolor pulses, significant
modulations of THz emission are observed around the
air molecule revival time. Figure 29 show the peakto
peak amplitudes of THz electric field as a function of
delay time (dotted curves) and the calculated response
function of rotational Raman process (solid curves).
The delay times are set around the quarter revival of O2
(Fig. 29a), the half revival of N2 (Fig. 29b), the half
revival of O2 and the threequarter revival of N2
(Fig. 29c), the full revival of N2 and threequarter
revival of O2 (Fig. 29d), the full revival of O2 and one
and a half revival of N2 (Fig. 29e).

28

CHIN et al.

KDP

Ultrathin
wedges
BBO
DWP
THz
Reflective type
focusing telescope

Fig. 30. Schematic setup for remote high energy THz generation. The bottom indicates the time sequence of the two color beams.

When the rotational Raman response function,

R(t), is >0, the average molecular axis is parallel to the
polarization of the exciting laser field and the refrac
tive index change of the probe laser pulse and its SH is
positive. When R(t) is <0, the average molecular axis is
perpendicular to the polarization of the exciting laser
field and the refractive index change of the probe laser
pulse and its SH is negative. (See Section 8 for more
discussion of such index change.) The refractive index
change due to molecular alignment results in focusing
or defocusing of the probe beam and hence, in the
changes of laser intensity and plasma density. For
example, positive refractive index change corresponds
to the increase of intensity and plasma density. Thus
when R(t) > 0, the intensity of the FW and the SHW
should slightly increase and the THz efficiency should
increase and vice versa. Meanwhile, the positive
refractive index change will increase the nonlinearity,
(3), and the negative refractive index change will
decrease the nonlinearity. Generally, enhanced THz
emission was observed when R(t) is >0 and vice versa
as depicted in Fig. 29. The observed phenomenon on
the modulation of THz emission is attributed to
molecular alignment induced refractive index change,
resulting in the changes of the nonlinearity ((3)) in
neutrals and laser intensities inside the filaments.
6.7. Remote High Energy THz Generation
For remote high energy THz generation, so far,
twocolor filamentation is strongly desired and is also
challenging. Phase compensation between the two
color beams is necessary and needs to be precisely
controlled. One can simply set up a delay stage and a
pair of ultrathin wedges in one arm of the twoarm
scheme (FW and SHW in each arm, respectively).
However, fluctuation and mechanical instability is
serious in the twoarm scheme especially for long dis
tance experiments according to our experience.
Recently, an inline phase compensation setup was
reported by Zhang et al. [138, 156] to improve the sta
bility. The setup in [138] together with a reflectivetype
telescope (Fig. 30) was used to remotely generate high

energy THz pulses in air [157] using high energy Ti

sapphire laser pulses.
In order to decrease the nonlinear effects inside the
optics of the inline setup and also considering the
large pump beam size, optics with large aperture had
to be used. For frequency doubling, a 0.5 mm thick
KDP crystal with aperture >30 mm was chosen. When
the FW and SHW with orthogonal polarizations pass
through a birefringent crystalXcut BBO, the FW
will be delayed compared to the SHW (socalled neg
ative delay). All the following optics and air will pro
vide positive delays, which will result in the temporal
overlap in the predetermined filament zone. A pair of
ultrathin wedges can precisely control the relative
phase between the two color beams. A dualwave
length wave plate (DWP) can rotate one of the polar
izations by 90 and the other by 180. The latter means
that the polarization is unaffected. A reflectivetype
telescope can project the twocolor filament at a long
distance.
So far, a record of THz generation at a distance of
16 m is detected. Pulse energies more than 250 nJ in
the frequency range below 5.5 THz are recorded using
the current detection system with a pyroelectric
energy meter by some of the authors of this review
[157]. In the experiments, less than 30 mJ pump was
used, which was limited by the selffocusing effect
after the intense pump beam passes through the centi
meterthick BBO crystal.
As an improvement, Xcut BBO wedges combin
ing the current BBO crystal and ultrathin wedges
would be helpful. The current setup has a capability to
generate THz radiation at few tens of meters away by
carefully managing the delay between two color
beams. In order to generate stronger THz emission at
further distances, good pump beam quality, larger
aperture optics and more sensitive detection will be
more helpful. When an intense singlecycle THz pulse
is applied on a plasma filament in air, THz radiation
enhanced emission of fluorescence (REEF) [160]
from the nitrogen fluorescence from the filament was
observed. The THzenhanced fluorescence intensity
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

carries THz waveform information, which is ideal for

standoff detection of THz pulse. For THz standoff
detection, one can refer to some recent work [156,
161], which is out of the scope of this review.
7. BIREFRINGENCE INSIDE A FILAMENT
IN AIR
7.1. Birefringence
When a linearly polarized fs laser pulse undergoes
filamentation in air, the high intensity inside the fila
ment will turn the normally isotropic medium into
anisotropic during the short time when the filamenting
laser pulse exists. If simultaneously, a second linearly
polarized weak probe pulse propagates with the fila
ment, it will experience this anisotropy through cross
phasemodulation (XPM). That is to say, in the eyes
of the probe, the normally isotropic air medium
becomes birefringent. The probe pulse would be
guided by the filament, become elliptically polarized
in general and would rotate inside the filament core
[71, 10]. We explain this phenomenon as follows.
The anisotropy arises because inside the filament,
the pumps linear polarization induces a stronger
polarization parallel to the field of the pump than that
perpendicular to the pump field. The probe would
experience this anisotropy through XPM. Thus, the
probe would experience two different extra indices of
refraction, n|| and n, corresponding to the polar
ization components of the probe parallel and perpen
dicular to the pump polarization, respectively. From,
for example, [10, 71],
3 0
(3)
filament 2
,
n || = Re
( xxxx ) E x
4n 0

(12)

3 0
(3)
filament 2
.
n = Re
( yyxx ) E x
4n 0

(13)

Here, x denotes the direction of the pump polarization

filament
and E x
is the electric field of the pump inside the
(3)

filament; y is the direction perpendicular to x, xxxx ,

(3)

yyxx are the corresponding third order susceptibility

tensors.
(3)

(3)

probe

ijkl = ijkl ( i

probe

; j

filament

; k

filament

; l

),

(14)

ijkl (= x, y) being directions of the electric fields of the

(3)
corresponding frequencies . Since Re( xxxx ) =
(3)

3Re( yyxx ) in an isotropic medium [10, 78]

n|| = 3n.

(15)

Thus, the parallel component of the probe propagates

slower than the perpendicular component resulting in
a phase difference between them. This phase differ
ence is given by (n|| n)2z/, where is the probe
LASER PHYSICS

Vol. 22

No. 1

2012

29

wavelength, z is the distance measured from the begin

ning of the filament whose length is l. It increases as
the probe propagates through the filaments length l. It
is at the origin of changing the probe polarization from
linear to elliptical with various degree of rotation in
general.
Because both the extra indices n|| and n are pos
itive (Eqs. (12) and (13)), they give rise to the guiding
of the probe by the filament in a way similar to the
guiding by a fiber; however, it is more complicated
because of the inequality of n|| and n and because
of the existence of plasma inside the filament. Conse
quently, the plasma part of the probe beam is guided
and part of it is diffracted by the plasma [162]. This
phenomenon was reviewed in [10].
7.2. Polarization Separator
Another result is what we call by polarization sepa
rator [91, 162] reviewed in [10]. It is related to initial
molecular alignment (see next section) in air inside a
filament after a delay of about 100 fs. By this time, the
pump pulse is gone. The probe pulse would see only
the aligned molecules. The parallel polarization com
ponent of the probe, being parallel also to the aligned
molecular wave packet, would see an increase of index
of refraction and would be guided inside the filament
core even in the presence of the (weak) plasma. The
perpendicular component would see a reduction in the
index of refraction as well as the plasma. It would be
diffracted out of the filament core into a ring pattern.
The result is the separation of the initial full beam pat
tern into a central part with parallel polarization and
an outer ring with perpendicular polarization.
7.3. Polarization Rotation
The rotation of the ellipse comes from the phase
difference between the two orthogonal components.
When the phase difference is the effect is similar to a
halfwave plate. Because the x and ycomponents of
the probe are generally not equal, the polarization is
an ellipse rotated by twice the angle between the
probes initial linear polarization axis and the pumps
linear polarization axis. The latter could be considered
as the optic axis of the filament.
An experiment was done by setting the angle
between a probe (400 nm/~50 fs/2.2 J) and the pump
(800 nm/50 fs/0.85 mJ) polarizations at two arbitrary
angles of 15 and 75 with respect to the pump polar
ization (defined as = 0) in argon at one atmo
spheric pressure [163]. The reason for using argon was
to look at only the electronic effect and avoid the
molecular complications. The probe was superim
posed on the pump; it passed through (hence sam
pling) only the filament core of the pump by col
linearly focusing the two pulses into the gas. Using an
analyzer to measure the polarization of the probe at
the end of the filament, the result is shown in Fig. 31.

30

CHIN et al.

120

Initial, = 75
Rotate 15.8 2
e2 = 0.055
60
Initial, = 15

0
e = 0.085
Rotate 16.4 2

180

Experiment 75
Experiment 15
Simulations 75

300

Simulations 15
Fig. 31. Polarization analysis of the transmitted probe at the end of filament for a pair probe polarization angle probe at 75 (light
grey circles, solid curve) and 15 (black circles, solid curve) with respect to the pump polarization defined as = 0. Dash straight
lines and solid straight lines correspond to probe polarization after the analyzer when the pump is turned off (initial) and on (fin
ial), respectively. XPM induced polarization rotations are marked by the blue arrows and characters.

The lineartoelliptical transformation of a 400 nm

fsprobe pulse in the birefringent filament in argon of
an 800 nm linearly polarized fspump pulse was also
studied numerically [164]. The rotation of the probe
elliptical polarization was the largest in the high
intensity filament core. With propagation the rotated
radiation diffracted outward by the pumpproduced
plasma. Figure 32 shows the numerical results of the
rotation of the polarization of the probe.
An analyzer was used to measure the probes polar
ization change. The transmission of the analyzer
crossing the probes polarization was a maximum at
the pumpprobe angle of 45 and gave equal values for
each pair of angles symmetrically situated at both sides
of the maximum. This prediction was confirmed
experimentally as shown in Fig. 33.
8. MOLECULAR ALIGNMENT
IN A FILAMENT AND REVIVAL
8.1. Preface
Recent experiments provide enticing evidence for
the potential of alignment to introduce an intriguing
range of novel applications of laser filamentation.
Thus, for instance, it was experimentally found that
alignment can increase the length of the plasma chan
nel generated in the wake of the filament [104], modify
the filamentgenerated supercontinuum [99], give rise
to shock Xwaves in the filament and tune their prop
erties [88], control the wavelength of a fewcycle pulse
[82], and, interestingly, modulate interfilament inter
actions [73]. Equally inviting are the experiments of

[91], where it is shown that nonadiabatic alignment

can serve as a polarization separator, the results of
[162], where the ability of a filament in an aligned
medium to affect ultrafast birefringence was illus
trated, the work of [165], where the timemodulations
of the refractive index caused by alignment was mea
sured and computed, and the recent work of [166],
where the interplay between laser alignment and pop
ulation trapping in a filament was explored.
In order to understand, and hence also to control,
these and other phenomena, it is important to first
understand the concepts and mechanism responsible
for nonadiabatic alignment. This is the goal of the
present section. We begin in Subsection 8.2 with a dis
cussion of the qualitative physics underlying intense
laser alignment, noting the roles of the aligning field
parameters and the material system properties. In
Subsection 8.3 we briefly review the theoretical and
experimental methods currently being used to quan
tify alignment, and in Subsection 8.4 we discuss nona
diabatic alignment and population trapping in the
context of laser filamentation. We conclude in Subsec
tion 8.5 with a discussion of opportunities and avenues
for future research on filamentation from aligned
media.
8.2. Basic Concepts
Essentially, the alignment of an ensemble of molec
ular rotational states by a light pulse is a phenomenon
of the excitation of a wave packet. A rotational wave
packet is analogous in some respects to an electronic
(Rydberg) wavepacket or a vibrational wavepacket.
LASER PHYSICS

Vol. 22

No. 1

2012

0.2
0
0.2

J/cm2

r, mm

0.2
0
0.2

r, mm

(a)
Fluence of 800 nm pump pulse
0.5 J/cm2

Fluence of 400 nm pump pulse

31

mJ/cm2

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

2.6

5.2

1.3

2.6

1 mJ/cm

Pump intensity,
TW/cm2

120

Pump
Probe

80

200

40

100

0
0

y, mm

0.4

300

50

z, cm
(b)
z = 55 cm

z = 40 cm

0
150

100

Probe intensity,
GW/cm2

Peak intensity of 800 and 400 nm pulses

z = 85 cm

z = 112 cm

90

90

0.4
0.4

0.4

Core

90

180

Periphery

Pump

0
x, mm
90

180

270

270

270

270

Fig. 32. (a) Simulated fluences and peak intensity of the pump and the probe pulses copropagating in 1 bar argon. The initial
pumpprobe angle is = 45, the pump direction is 0. Dashed vertical lines on the intensity plot indicate the positions z = 40,
55, 85, 112 cm, where the polarization is analyzed in (b) (each of the four columns corresponds to one zposition). (b) The upper
row shows the polarization ellipses at several radial positions across the beam plotted over the transverse fluence distribution. The
polar plots show the probe fluence (solid curve) in the beam center (radius 100 m, the middle row) or the periphery ring (320
370 m, the lower row). The dashed line is the probe polarization direction at z = 0.

One gains localization in the coordinate space (here

the angle) by populating a broad superposition in the
conjugate quantum number space (here the angular
momentum quantum number). The superposition is
localized (here it is aligned) but it will dephase in the
LASER PHYSICS

Vol. 22

No. 1

2012

course of time under field free conditions (even if

coherence is maintained) because the energy level
spacings are not equal.
In this section we provide a qualitative picture of
coherent laser alignment. We start by discussing the

Experiment
Simulations

(a)

1.0

30
20

0.5
10
0

0
0

20
40
60
80
100
Pump probe angle , deg

Probe energy after the crossed

analyzer, arb. units (experiment)

CHIN et al.
Probe energy after the crossed
analyzer, % of initial (simulations)

32

(b)

y' Crossed analyzer

direction

Probe, 400 nm

x'

Pump, 800 nm

Fig. 33. (a) The experimentally obtained (stars) and simulated (dashed curve) energy of the overall probe beam transmitted by the
crossed analyzer at the end of the filament for the pairs of pumpprobe angles = (; 90 ). (b) The initial pump and probe
directions.

Jo+3

(a)

(b)

Jo+2
Jo+1
Jo+1
Jo1
Jo3

Jo2 Jo1
Jo4 Jo3

Jo2
Jo4

Jo+4
Jo+3

Energy

Energy

Jo+3

Jo+4
Jo+3

Jo+3

Jo+2

Jo+2

Jo+2

Jo+1

Jo+1

Jo+1

Jo
Jo1
Jo+2
Jo4

Jo

Jo
Jo2
Jo4

Jo1

Jo+4

Jo+4

Jo+4

Jo+3

Jo+3

Jo+3

Jo+2

Jo+2

Jo+2

Jo+1

Jo+1

Jo+1

Jo

Jo

Jo

Jo2 Jo1
Jo4 Jo3

Jo1
Jo2
Jo3

Jo2
Jo4

Fig. 34. (a) Rotational excitation in nearresonance laser fields. In the weak field limit, electric dipole selection roles restrict the
degree of rotational excitation to |J| = 0, 1. In a moderately intense field, by contrast, Rabitype sequential transitions between
the two electronic manifolds are allowed, each of which is accompanied by the exchange of another unit of angular momentum
between the molecule and the field, leading to the population of a rotationally broad wavepacket. (b) Rotational excitation in
nonresonant laser fields. At faroff resonance frequencies, well below electronic transition frequencies, rotational excitation takes
place via Raman type, |J| = 0,2, transitions, resulting again in a population of a broad rotational wavepacket.

mechanism underlying laser alignment within a quan

tum mechanical framework (Subsection 8.2.1) and
noting the role played by the field and system parame
ters (Subsection 8.2.2). To that end we focus on the
simplest case scenario of a linear, isolated rigid rotor
subject to a linearly polarized field. In Subsection 8.2.3
we discuss briefly the possibility of orienting molecules
with intense pulses, and in the last subsection we con
sider the extension of alignment from isolated mole
cules to dissipative media.
8.2.1. Rotational excitation and coherent align
ment. Consider a rigid, linear molecule subject to a
linearly polarized laser field whose frequency is tuned

near resonance with a vibronic transition. In the weak

field limit, if the system has been initially prepared in
a rotational level J0, electric dipole selection rules
allow transition to J0 and J0 1 in the excited state.
The interference between these levels gives rise to a
mildly aligned excited state population, depending on
the type of the dipole transition, parallel or perpendic
ular to the lasers linear polarization. At nonperturba
tive intensities, the system undergoes Rabitype oscil
lations between the two resonant rotational manifolds,
exchanging one unit of angular momentum with the
field on each transition. Consequently, a rotationally
broad wavepacket is produced in both states. The asso
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

ciated alignment is considerably better than the weak

field distribution, as it arises from the interference of
many levels [167, 168] (Fig. 1a). The degree of rota
tional excitation is determined by either the pulse
duration or the balance between the Stark shift that
depends on the laser intensity and the detuning from
resonance. In the limit of a short, intense pulse,
2
pulse < (BeR)1, the number of levels that can be
excited is roughly the number of cycles the system has
1
undergone between the two states, Jmax ~ pulse/ R ,
where pulse is the pulse duration, Be is the rotational
1

constant, R is the Rabi coupling and R is the cor

responding period. In the case of a long, or lower
2
intensity pulse, pulse > (BeR)1, the degree of rota
tional excitation Jmax is determined by the accumu
lated detuning from resonance, i.e., by the require
ment that the Rabi coupling be sufficient to access
rotational levels that are further detuned from reso
nance as the excitation proceeds, Jmax ~ R /2B e . A
quantitative discussion is given in [169].
A rather similar coherent rotational excitation pro
cess takes place at faroff resonant frequencies, well
below electronic transition frequencies. In this case a
rotationallybroad superposition state is produced via
sequential Ramantype (|J| = 0,2) transitions,
(Fig. 34) the system remaining in the ground vibronic
state [169].
The qualitative criteria determining the degree of
rotational excitation remain as discussed in connec
tion with the nearresonance case. In the case of near
resonance excitation, however, the Rabi coupling is
proportional to the laser electric field, whereas in the
case of nonresonant excitation it is proportional to the
intensity, i.e., to the square of the field, due to the two
photon nature of the cycles. Consequently, much
higher intensities are required to achieve a given
degree of alignment, as expected for a nonresonant
process. On the other hand, much higher intensities
can be exerted at nonresonant frequencies, as undes
ired competing processes scale similarly with detuning
from resonance. It is interesting to note that the
dynamics of rotation excitation and the wavepacket
properties are essentially identical in the near and
nonresonance excitation schemes. In fact it is possi
ble to formally transform the equations of motion cor
responding to one case into those corresponding to the
other [169]. In reality the near and the nonresonant
mechanisms may act simultaneously, with the former
typically corresponding to a vibrational resonance
[170]. The degree of spatial localization and the time
evolution of the alignment are essentially independent
of the frequency regime; they are largely controlled by
the temporal characteristics of the laser pulse, as illus
trated in the following subsection.
LASER PHYSICS

Vol. 22

No. 1

2012

33

8.2.2. Time evolution. The simplest case scenario is

that of alignment in a continuous wave (CW) field,
proposed in [171, 172], where dynamical consider
ations play no role. In practice, intense field experi
ments are generally carried out in pulsed mode, but,
provided that the pulse is long with respect to the rota
tional period, rot = /Be, each eigenstate of the field
free Hamiltonian is guaranteed to evolve adiabatically
into the corresponding state of the complete Hamilto
nian during the turnon, returning to the original (iso
tropic) fieldfree eigenstate upon turnoff [169]. For
mally, the problem is thus reduced to alignment in a
CW field of the peak intensity. The latter problem is
formally equivalent to alignment of nonpolar mole
cules in a strong DC field, since the oscillations of the
laser field at the light frequency can be eliminated at
both near and nonresonance frequencies [167, 169].
Thus, in the long pulse limit, pulse rot, laser align
ment reduces to an intensively studied problem that is
readily understood in classical terms, namely the
problem of fieldinduced pendular states [173, 174].
The Stark shift gives rise to a potential well for the
angular motion, which can readily exceed the thermal
energy and convert the free rotation of molecules into
small amplitude librations about the polarization vec
tor. In this limit, the sole requirement for alignment is
that the Rabi coupling be large compared to the rota
tional energy of the molecules at a given rotational
temperature. Intense laser alignment in the CW limit
(termed adiabatic alignment) grew out of research on
alignment and orientation in DC fields, and is very
similar to DC alignment conceptually and numeri
cally. It shares the advantage of DC field alignment of
offering an analytical solution in the linear, rigid rotor
case [171]. It shares, however, also the main drawback
of DC field alignment, namely, the alignment is lost
once the laser pulse has been turned off. For applica
tions one desires fieldfree aligned molecules.
Shortpulseinduced alignment (termed nonadia
batic, or dynamical alignment) was introduced [167]
at the same time as the adiabatic counterpart and is
similar in application but qualitatively different in
concept. This field of research grew out of research on
wavepacket dynamics and shares several of the well
studied features of vibrational and electronic wave
packets while exhibiting several unique properties
(vide infra). A short laser pulse, pulse < rot, leaves the
system in a coherent superposition of rotational levels
that is aligned upon turnoff, dephases at a rate pro
portional to the square of the wavepacket width in
Jspace, and subsequently revives and dephases peri
odically in time. As long as coherence is maintained,
the alignment is reconstructed at predetermined times
and survives for a controllable period. As with other
discrete state wavepackets of stable motions, all
observables obtained from the wavepacket are periodic
in the full revival time, given, in the case of rotational
wavepackets, by the rotational period rot. Interest

34

CHIN et al.

ingly, under rather general conditions, the alignment is

significantly enhanced after turnoff of the laser pulse
[168]. The origin of the phenomenon of enhanced
fieldfree alignment is unraveled in [168] by means of
an analytical model of the revival structure of rota
tional wavepackets.
The first experiment to realize the prediction of
fieldfree, post pulse alignment was reported in 2001
[175]. During the past decade, a large number of
quantitative experiments that probe and utilize nona
diabatic alignment have been published (see also Sec
tion 4 and 7), of which the ones most relevant to laser
filamentation are discussed in the following sections.
In the ultrashort pulse limit, where the interaction is
an impulse as compared to rotations, the dynamics
following immediately the laser pulse is again readily
understood in classical terms [168]. Here the pulse
imparts a kick to the molecule, which rapidly trans
fers a large amount of angular momentum to the sys
tem and gives rise to alignment only after the turnoff.
The molecule is rotationally frozen during the interac
tion but the sudden kick is encoded in the wave
packet rotational composition and gives rise to align
ment upon turnoff. The subsequent long time
dynamics, by contrast, exhibit strongly quantum
mechanical dephasing and rephrasing dynamics that
does not have a classical analog. At very low rotational
temperatures, the revival structure is rich and dense,
differing qualitatively from the generic revival struc
tures of bound state wavepackets [176]. As the rota
tional temperature increases, however, the structure
simplifies drastically. At each multiple of the rotational
period the initial alignment is precisely reconstructed.
At half the period the molecules align perpendicular to
the field. Depending on the spin statistics of the mol
ecule studied, one observes features (termed fractional
revivals) also at other fractions of the rotational
period.
A third mode of alignment is introduced in [177] in
the context of simultaneous alignment and focusing of
molecular beams, potentially an approach to nanos
cale surface processing [178]. A route to that end is
provided by the combination of long turnon with
rapid turnoff of the laser pulse. During the slow turn
on, the isotropic free rotor eigenstate evolves into an
eigenstate of the complete Hamiltonianan aligned
manyJ superpositionas in the longpulse limit.
Upon rapid turnoff the wavepacket components are
phased together to make the full Hamiltonian eigen
state but now evolve subject to the fieldfree Hamilto
nian. The alignment dephases at a rate determined by
the rotational level content of the wavepacket (hence
the intensity, the pulse duration and the rotational
constant), and subsequently revives. At integer multi
ples of the rotational period, t = t0 + nrot, t0 being the
pulse turnoff, the fully interacting eigenstate of the
complete Hamiltonian attained at the pulse peak is
precisely reconstructed and the alignment character
izing the adiabatic limit is transiently available under

fieldfree conditions. Experimental realization of

fieldfree alignment via the combination of slow turn
on with rapid turnoff was reported in [179].
8.2.3. Molecular orientation. A purely alternating
current (AC) electromagnetic field polarized along the
(say) spacefixed zaxis preserves the symmetry of the
fieldfree Hamiltonian with respect to z
z (by con
trast to a DC electric field that defines a direction, as
well as an axis, in space); it aligns, but cannot orient
molecules. The motivation for augmenting laser align
ment to produce orientation comes from the field of
stereochemistry, where orientation techniques have
proven to make a valuable tool for elucidating reaction
mechanisms [180, 181]. Several different methods
have been proposed to that end [182, 183], two of
which have been also tested in the laboratory [184,
185].
Within the method of [182], an intense laser field is
combined with a relatively weak DC electric field, with
the former serving to produce a broad superposition of
Jlevels via the nonlinear polarizability interaction
and the latter providing a means of breaking the z
z symmetry. This approach is applied in [184] to orient
HXeI, and in [186, 187] to orient OCS molecules. An
alternative approach, advanced by several groups,
takes advantage of the natural asymmetry of halfcycle
pulses to orient molecules. Substantial experience has
been gained in recent years in optimizing the ensuing
orientation and manipulating its temporal character
istics by different control schemes [188192]. A third
method, which likewise received significant attention
[185, 193195], exploits the possibility of breaking the
z symmetry through coherent interfer
spatial z
ence, e.g., by two color phaselocked laser excitation
or by combining the fundamental frequency with its
second harmonic. (The second harmonic is often
taken to be resonant with a vibrational transition but a
nonresonant variant of the same procedure has also
been suggested [183].)
8.2.4. Alignment in dissipative media. The foregoing
discussion has been restricted to the limit of isolated
molecules, corresponding to a molecular beam exper
iment, where collisions do not take place on the time
scale of relevance and coherence is fully maintained.
This limit applies to the vast majority of alignment
studies reported to date. In a laser filament, however,
the pressure is typically significant and hence the mol
ecules suffer collisions and, in the course of time,
gradually lose their relative phase relations. Since the
time scales of interest are typically short, and since the
medium is only mildly dissipative, the isolated mole
cule description is often qualitatively correct, depend
ing on the experimental parameters. Rigorously, how
ever, dissipation need be considered (see Subsection
8.3).
Alignment subject to dissipative media, where col
lisions give rise to decoherence and population relax
ation on relevant timescales, is also fundamentally
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

interesting. First, one can show [196] that the unique

coherence properties of rotationallybroad wavepack
ets provide a sensitive probe of the dissipative proper
ties of the medium. In particular, it is found that the
experimental observables of alignment disentangle
decoherence from population relaxation effects, pro
viding independent measures of the relaxation and
decoherence dynamics that go beyond rate measure
ments. Second, we expect laser alignment to become a
versatile tool in chemistry, once the effects of dissipa
tive media on alignment are properly understood.
Reference [196] explores the evolution of nonadia
batic alignment in dissipative media within a quantum
mechanical density matrix approach, illustrating both
the sensitivity of rotationally broad wavepackets to the
dissipative properties of the medium and the possibil
ity of inhibiting rotational relaxation, so as to prolong
the alignment lifetime, by choice of the field parame
ters. A classical study of alignment in a liquid is pro
vided in [197]. The application of intense laser align
ment in solutions to control charge transfer reactions
is illustrated in [198]. Experimentally, nonadiabatic,
intense pulse alignment in the dense gas environment
has been probed in several studies. A particularly
quantitative study is provided in [199], which com
pares the alignment measured in rotationallycold
molecular beam environments with that obtained in
the dense gas medium.
Our discussion so far has been limited to linear sys
tems, which are the ones of relevance to laser filamen
tation studies at present. We note in Subsection 8.5
that the case of nonlinear, in particular asymmetric,
rotors offers new and interesting physics; see [200].
8.3. Theoretical and Experimental Methods
In this section we discuss several of the commonly
used approaches to observe nonadiabatic laser align
ment, experimentally (Subsection 8.3.1) and numeri
cally (Subsection 8.3.2).
8.3.1. Experimental approaches. The most com
mon way to probe nonadiabatic (and adiabatic)
molecular alignment is through femtosecond pump
probe Coulomb explosion fragment ion imaging.
Nonadiabatic alignment was observed in I2, [175] and
nonadiabatic orientation was achieved in CO using a 2
color pump laser field and a single color pump field
[201]. In a typical experimental scheme to observe
nonadiabatic alignment, a first femtosecond laser
pulse (pump) prepares a molecular rotational wave
packet. The periodically aligned molecules are then
multiply ionized by a second, more intense, time
delayed laser pulse (probe). The doubly or highly
charged molecular ions then dissociate rapidly
through Coulomb explosion and the final momentum
of each ionic specie is revealed through a position sen
sitive timeofflight ion mass spectrometer. For simple
molecules, the molecular configuration of the precur
LASER PHYSICS

Vol. 22

No. 1

2012

35

sor state (bond lengths and 3D spatial orientation)

prior to the dissociation can often be reconstructed
from classical mechanics (Coulombs law), recovering
the alignment sense and quality as well as the fascinat
ing femtosecond dynamics of the different ions. A
noteworthy example is the ultrafast hydrogen migra
tion studied in [202, 203]. In several laboratories,
coincidence detection of ions, imposing the total
momentum conservation for the fragments, has been
used to make sure that all fragments come from the
same molecule and to identify unambiguously com
peting dissociation pathways [204206]. For detailed
discussions of the stateoftheart of Cold Target
Recoil Ion Momentum Spectroscopy (COLTRIMS)
detectors, see [207, 208].
If the medium density is sufficiently high, the
refractive index modulation n caused by the molecu
lar alignment can be probed by a second short laser
pulse in an alloptical measurement. This approach is
particularly relevant in the case of filamentation,
where the medium pressure is high and hence ion
imaging becomes impossible. Thus, instead of mea
suring ions under ultrahigh vacuum conditions, the
measurements are performed directly on the light,
employing well established optical methods (interfer
ometry, polarization optics, etc.) and photonic detec
tors such as photomultiplier tubes, CCD cameras and
spectrometers. This approach can be used as long as
the molecular alignment induces a macroscopic phase
modulation that can be measured. (See Section 4
and Subsection 8.4 for some examples.) The condition
is that should not be vanishingly small compared to
rad. Typically, pressures (p) of a few torrs to a few
atmospheres are used for practical interaction lengths
L ranging from micrometers to meters. ( nL and
n p). The actual optical phase modulation mea
surements of a weak probe pulse induced by molecular
alignment can be implemented in several ways. For
instance, spectral interferometry [209, 210], polariza
tion gating [162, 211], spatial [212] and spectral mod
ulation [86, 83, 165, 213] were used to observe align
ment.
A novel approach involving electron diffraction was
demonstrated in [214]. In this work, a lasersynchro
nized 4ps electron pulse was used to probe selectively
aligned C2F4I2 molecules undergoing dissociation. By
imaging the electron diffraction pattern for several
delays between the UV pump and the electron probe
pulse, the authors could observe the alignment and its
picosecond timescale decay through the azimutal
dependence of the electron diffraction pattern.
Monitoring the laser induced high harmonic emis
sion spectra is also an indirect way to observe adiabatic
[215] and nonadiabatic [216] molecular alignment,
and one that is compatible with high pressure. High
harmonic generation from aligned molecules is a
young but intensively studied subject, and its informa
tion content remains the topic of scientific debate

36

CHIN et al.

[217219]. It has been illustrated, however, that har

monic signals convey unique information regarding
rotational wavepacket dynamics [220].
8.3.2. Numerical approaches. In the isolated mole
cule limit, nonadiabatic alignment is most commonly
described by nonperturbative solution of the time
dependent Schroedinger equation subject to the com
plete Hamiltonian H = H0 + Hint. Here, H0 is the field
free Hamiltonian and Hint is the interaction term.
Within the (almost uniformly applied) rigid rotor
approximation, H0 reduces to the rotational Hamilto
nian, and in the (most commonly considered) faroff
resonance limit Hint is the induced dipole Hamilto
nian. In the simplest case scenario of a linear, rigid,
isolated molecule subject to a linearly polarized,
2
faroffresonance field, H = B J and H =
0
e
int
1/42(t)cos2 , where Be is the rotational constant,
2

J is the angular momentum squared operator, (t) is

the pulse envelope, the polarizability anisotropy is
defined in terms of the components of the molecular
polarizability tensor as = || , and is the polar
Euler angle (the angle between the molecular axis and
the field polarization direction in the present context).
The general forms of H0 and Hint are derived in [200].
Given the complete Hamiltonian H, the time
dependent Schroedinger equation is typically solved
by expanding the wavepacket in the complete set of
eigenstates of H0,
| J, M ( t ) =

J
J, M ( t ) |JM,

(16)

where ,|JM = YJM(,) are the spherical harmon

ics, and propagating a set of coupled differential equa
j
tions for the expansion coefficients C Ji M ( t ) . In
Eq. (16) J is the total angular momentum quantum
number, Ji is its initial value, and M is the magnetic
quantum number. The azimuthal Euler angle, , is the
angle of rotation about the field vector in the present
context. Given the timedependent wavefunction (t)
as a function of time, all observables of interest can be
computed in a nonperturbative, fully quantum
mechanical fashion. A common measure of the degree
and timeevolution of the alignment is the expectation
value of cos2 , computed as (t)|cos2 |(t). Up to a
constant, cos2 is proportional to the second
moment of the rotational distribution, hence its com
mon use as a measure of the alignment. We remark,
however, that this measure does not convey all the
information that is available in the rotational wave
packet and is sensitive only to second order interfer
ences.
In the context of laser filamentation, where the
pressure is not negligible and collisions take place on
timescales of relevance, a density matrix approach is

appropriate. Accordingly, we expand the density oper

ator in terms of the rigid rotor eigenstates as ( t ) =
JMJ'M' ( t ) |JMJ'M'|. The time evolution of

J, M, J', M'

the elements of the density matrix is given by the quan

tum Liouville equation, and within a multilevel
Bloch model for the dissipative operator, they obey the
following equation of motion:

d JMJ'M' ( t )

 = i ( J J' ) JMJ'M' ( t )
dt

R ( t )

[V

JMJ 1 M 1 J 1 M 1 J'M' ( t ) MM1

J1

 V J1 M1 J'M' JMJ1 M1 ( t ) M'M1 ]

(17)

1 ( K JMJ1 M1 JMJ'M' ( t ) + K J'M'J1 M1 JMJ'M' ( t )

J1 M1 2

JJ' MM' 2K J1 M1 JM J1 M1 J1 M1 ( t ) )
( pd )

+ ( 1 JJ' MM' ) JMJ'M' JMJ'M' ( t ),

where R(t) = 1/42(t), VJMJ'M' = JM|cos2 |J'M',
the dependence on the initial rotational quantum
number is omitted to avoid cluttering the notation,
and we note that optical transitions conserve M. The
population transition rates between states |JM and
|J1M1 are given by K JMJ1 M1 with K JMJ1 = JM1 = M = 0 and
( pd )

JMJ'M' is the rate of pure dephasing of the |JMJ'M'|

coherence.
In terms of the reduced density operator, ( t ) , the
expectation value of cos2 is given as cos2 =
Tr{cos2 ( t ) }. We remark that cos2 is sensitive only
to rotational coherences and rotational relaxation.
The diagonality of the operator cos2 with respect to
all quantum numbers except for the angular momen
tum quantum number J follows from its dependence
on a single coordinate, the polar Euler angle that is
conjugate to J (in a rough sense). It thus disentangles
the effects of Jchanging collisions from those of
purely Mchanging collisions, and, in case a rovi
bronic wavepacket is excited, also from the effects of
vibrational and electronic relaxation and decoher
ence. Numerical examples along with experimental
data are given in the next section.
8.4. Nonadiabatic Alignment in the Context
of Laser Filamentation
In a filamentation process in gases, there are two
major observables as a consequence of the molecular
alignment on the laser pulse. First, through nonadia
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

NL ( r, t ) = n ( r, t ) 0 L/c 0 ,

(18)

where r denotes the radial cylindrical coordinate, t is

time, n is the nonlinear refractive index, c0 is the
speed of light in vacuum, 0 is the probe central angu
LASER PHYSICS

Vol. 22

No. 1

2012

404
Air||

408
Mean wavelength, nm

batic Ramantype excitation of the different molecu

lar rotational levels (see Subsection 8.2), the driving
laser spectrum slowly shifts toward lower frequencies
(i.e., to the Stokes side) of the spectrum through prop
agation [114]. This is a consequence of the energy
transfer from the light field to the molecules. Second,
the macroscopic refractive index modulation caused
by the alignment of molecules with anisotropic polar
izability tensors is felt by the tail of the driving laser
pulse and/or by a delayed probe pulse. This transient
refractive index change has several consequences for
the probe pulse propagation. For instance, it can mod
ulate its spectrum [83, 165], spatially split its polariza
tion state through polarization selective guiding/defo
cusing processes (see Section 7) [91, 162], and
strongly affect its divergence properties after the fila
ment zone [83]. A detailed analysis based on numeri
cal propagation simulations was carried out by Cale
gari et al. [212]. Essentially, all the complicated prop
agation effects can be explained by the refractive index
modulation of the medium, which is a function of both
space (namely, the radial cylindrical coordinate r from
the propagation axis) and time. (See also Section 4.)
Also, one could observe the effect of nonadiabatic field
free molecular alignment in the other properties of
molecules such as their fluorescence. We could dem
onstrate the possibility of using the filament behavior
as a tool for probing phenomena in lasermolecule
interaction that is difficult to study with other means.
The next two case studies deal with the effect of field
free alignment and revival from inside the filament on
the anisotropic changes in the refractive index of a
probe pulse and in the nitrogen fluorescence signal.
8.4.1. Case study A: probing filament induced
anisotropy in air by measuring refractive index change.
In a recent article, Marceau et al. [165] demonstrated
a simple way to characterize the degree of alignment
through the spectral modulation of the probe pulse
and we summarize the main results here. The tech
nique consists of measuring the spectral modulation of
a weak (~J) copropagating linearly polarized probe
pulse as a function of its delay with respect to the driv
ing, 1.1 mJ, 50 fs, 800 nm linearly polarized pump
pulse. In order to distinguish the probe from the pump
in the spectrometer, the second harmonic of the
Tisapphire laser was used.
The centroid of the probe spectrum as a function of
the delay is shown in Fig. 35 for air and argon for both
parallel (||) and perpendicular () pump and probe
polarization.
The theory can be briefly explained as follows. The
nonlinear phase NL accumulated by the probe pulse is
given by [78]:

37

Air
400

Argon||
Argon

404

396
0.4

0.4

400

396
0

2
3
Delay, ps

Fig. 35. Centroid wavelength of the probe spectrum as a

function of the delay in air and in argon for parallel and
perpendicular polarizations of the probe field relative to
that of the pump. The initial alignment peaks near delay
0.1 ps. The quarter and half revival of N2 occur near 2.1
and 4.2 ps, respectively, while the quarter revival of O2
occurs near 2.9 ps. In argon (atomic gas), spectral modu
lation only occurs near delay 0 and is attributed to the elec
tronic Kerr effect. Inset: Zoom in around delay 0.

lar frequency and L is the effective interaction length

(i.e., the filament length here). The probe angular fre
quency modulation associated with this timedepen
dent refractive index modulation is given by:
d n ( t ) L/c .
( t ) = d NL ( t ) = 
0
0
dt
dt

(19)

Converting angular frequencies () into wavelengths

(), in the limit of small spectral modulation ( 0)
and neglecting dispersion (dn/d 0), the spectral
modulation is given by:
L d
( t ) 0  n ( t ) .
c 0 dt

(20)

Here, 0 stands for the probe central wavelength

(400 nm) and is the measured shift (i.e., spectral
modulation) in the centroid of the probe pulse spec
trum. The refractive index modulation can be com
puted from this spectral modulation measurement
through a simple integral:
t

c0
n ( t ) 
 ( t' ) dt' .
L 0

(21)

The integration is from negative infinity up to the par

ticular delay time. The birefringence of the gas
medium (air) in the wake of the filament is an obvious
consequence of the polarization dependence of the
refractive index (see also Section 7). The latter refrac
tive index is obtained from Eq. (21) in which (t) for

38

CHIN et al.
cos2()

Exp.||

10

0.40

Air

Exp.
Sim.||

Sim.

n(106)

0
0.35

(a)

10

N2
0.30

O2
1

Argon

0
1

4
5
Time, ps

0.4

Fig. 36. Calculated alignment factor of N2 and O2 due to a

50 fs, 800 nm, 5 1013 W/cm2 Gaussian laser pulse. The
initial gas temperature is 295 K.

various polarization is shown in Fig. 35. In fact, for

delays >0.1 ps in the case of air, the curves for parallel
and perpendicular relative polarizations of the pump
and probe are exactly out of phase and the amplitude
of the parallel curve is about twice larger than that of
the perpendicular curve. This observation agrees well
with the theoretical prediction that nmol, ||(r, t) =
2nmol, (r, t) (see below) for a homonuclear diatomic
molecule such as N2 and O2 [78].
A theoretical calculation of the averaged alignment
are shown in Fig. 36 and the refractive index are shown
in Fig. 37.
The alignment is computed using the density
matrix approach discussed in Subsection 8.3 and the
associated refractive index modulation is determined
as [78],
2
n mol, X ( r, t ) = 2N
 ( cos ( r, t ) 1/3 ) , (22)
n0

N ( cos ( r, t ) 1/3 )
n mol, Y ( r, t ) = 
, (23)
n0
2

n mol, X ( r, t ) = 2n mol, Y ( r, t ) ,

(b)

(24)

where N is the medium density, n0 is the weak field

refractive index, is the polarizability anisotropy
and cos2 (r, t) is the most commonly applied mea
sure of alignment, see Subsection 8.3. The refractive
index change caused by the electronic Kerr effect is
then given as [78],
n Kerr, X ( r, t ) = 2n 2 I pump ( r, t ) ,

(25)

2
n Kerr, Y ( r, t ) =  n 2 I pump ( r, t ) ,
3

(26)

0.2

0
0.2
Time/delay, ps

0.4

Fig. 37. Transient refractive index modulation (n)

induced by a filamenting laser pulse and felt by a 400 nm
probe pulse in (a) air (b) argon. Simulation (Sim.) and
experimental results (Exp.) are shown for both polariza
tion components. Pump pulse conditions for the simula
tion: 800 nm, 5 1013 W/cm2, 50 fs Gaussian pulse, 3 cm
long filament, plasma density of 1016 cm3. The horizon
tal axis represents the delay between the pulses for the
experimental curves and is the time axis for the simula
tions. There is a small difference between the simulation
and the experiment since experimental curves were not
deconvoluted for the finite probe pulse duration.

n Kerr, X ( r, t ) = 3n Kerr, Y ( r, t ) ,

(27)

where n2 for the case of air is determined empirically in

[44] to be 1 1019 cm2/W at 800 nm and 40 fs. Finally,
the plasmas isotropic contribution to the refractive
index is given (in CGS units) by [5]
2

2e N e ( r, t )
n plasma ( r, t ) = 
,
2
me 0

(28)

where e is the elementary charge, Ne is the plasma den

sity and me is the electron mass. Good agreement is
found between the computed and the measured
refractive indices. More interestingly, the calculation
illustrates that the measured refractive index responds
directly to the rotational wavepacket dynamics.
8.4.2. Case study B: probing filament induced align
ment and revival in air by measuring nitrogen fluores
cence. In this section, the filament induced alignment
and revival can be observed through the behavior of
nitrogen fluorescence in a pumpprobe experiment;
and this in turn reveals an important phenomenon in
lasermolecule interaction, namely, population trap
ping [166]. We studied the fluorescence of two photo
emission bands from inside the filament in air, namely,
+
those of N 2 (the first negative band systems
+

(B2 u

X2 g ) (v' = 0
LASER PHYSICS

v = 0) at 391 nm) and

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

Fluorescence signal, arb. units

0.576
Fluorescence signal, arb. units

(a)
1.0
391 nm 100 cm
337 nm 30 cm

0.8

0.6

0.512
~
~

337 nm 30 cm

(b)
0.455
0.420

10

10

20

30
40
50
Delay time, ps
+

Fig. 38. Fluorescence signal of first negative system of N 2

+

(B2 u X2 g transition) at 391 nm and second positive

system of N2 (C3uB3g) at 337 nm versus delay time
with corresponding plan convex lens whose focal lengths
are indicated.

neutral N2 (the second positive band systems

(C3u
B3g) (v' = 0
v = 0) at 337 nm). We
measured the fluorescence strength versus the delay
between the strong fundamental pump and the weak
second harmonic probe pulses of the Ti:sapphire laser.
The delay time dependence of the fluorescence signal
of the 391 nm band using a f = 100 cm lens and the
337 nm band using a f = 30 cm lens are presented in
Fig. 38. There is a strong enhancement at zero time
delay. When the probe pulses are behind the pump
pulses (positive delay time), periodic weaker enhance
ment (revival) is observed in the 391 nm band. For the
337 nm band, apart from the decay behavior, an oscil
latory structure of the fluorescence signal at the revival
positions is observed. Figures 39a and 39b show at a
higher resolution the enhancement and the oscillation
of the fluorescence signal of ionic and neutral nitrogen
molecules at 391 and 337 nm, respectively. The delay
time in Fig. 39 is in the vicinity of the full revival
period of nitrogen molecules. Figure 39c depicts the
rotational parameter of nitrogen obtained from theo
retical calculation. The photoinduced alignment
dynamics of nitrogen molecules is calculated using a
density matrix approach which is explained in Subsec
tions 8.18.3. We shall explain the fluorescence sig
nals observed in Fig. 38 through field free alignment
and population trapping.
The origin of the fluorescence signals from the 337
and 391 nm bands have been explained in references
[52, 221], respectively. Briefly, for the 391 nm band,
the strong laser field, through inner valence electron
LASER PHYSICS

Vol. 22

No. 1

2012

cos2

0.378
20

391 nm 100 cm

0.544

~
~

0.4

39

Alignment parameter

(c)

0.336
0.294
8.0

8.4
8.8
Delay time, ps

9.2

Fig. 39. (a, b) Observed fluorescence signal of ionize and

neutral nitrogen molecules and (c) calculated alignment
parameter versus delay time in the vicinity of full revival
period of nitrogen. One of the parameters influencing flu
orescence signal is field free alignment.

excitation, populates the excited B2 u ionic band

which decays through fluorescence to the ground ionic
+
X2 g band. (In this work, we measured only the stron
gest fluorescence signal from the 00 band around 391
nm.) The 337 nm bands is the transition of the neutral
nitrogens C3u band to B3g band, the strongest tran
sition amongst all vibrational transitions from these
bands is the 00 vibrational transition, around
337 nm. The upper band is dominantly populated by
+
dissociative recombination through the process N 2 +
+
+
N2
N 4 followed by N 4 + e
N *2 (C3u) + N2
in atmospheric pressure [52]. That is to say, the 337 nm
fluorescence depends very much on the nitrogen ion
density.

Besides nonadiabatic field free alignment, there is

population trapping (or interference stabilization)
[222, 223] induced by the pump pulse. Lowest order
perturbation theory (LOPT) predicts that the increase
in ion yield versus the laser peak intensity (I) will fol
low an In power law (n being the number of the pho
tons participating in ionization). Stabilization is man
ifested by a decrease in ion yield with increasing inten
sity of the laser field. During the interaction, the pump
pulse gives rise to population trapping in highly excited
(Rydberg) states of nitrogen molecules converging to

40

CHIN et al.

both the ground and excited state ions. Such neutral

excited molecules can be aligned by the linearly polar
ized pump pulse at 800 nm and revive periodically.
The arrival of the probe pulse, whose linear polar
ization is parallel to that of the pump pulse will ionize
molecules that are trapped in the highly excited states
through one or two photon absorption. Ionization of
nitrogen molecules aligned in the polarization direc
tion of the pump is more probable as compared to
those molecules perpendicularly aligned [224]. At
near zero time delay, enhancement of the fluorescence
signals at both wavelengths (391 and 337 nm) occurs.
This enhancement includes the effect of two color ion
ization and alignment enhancement of ionization. At
longer delay, ionization is enhanced at the periodic
revival times. This results in periodic fluorescence
enhancement of the first negative band at 391nm as it
is shown in Fig. 38. Figure 39a shows the same
enhancement in higher temporal resolution around
the revival period of nitrogen molecules rotational
wave packet. The consequence of this process is that
+
the number of N 2 ions is increased, which influences
the signal at 337 nm.
The fluorescence signal from the 337nm band in
Fig. 39b, shows a clear oscillatory structure. This fluo
rescence is the result of collision between neutral and
ionic nitrogen molecules as explained above. Another
process that influences the signal detection is the
anisotropy of the fluorescence with respect to the
alignment. The detection direction is perpendicular to
the laser propagation axis and parallel to the pump and
probe polarization axis. Hence, the detected fluores
cence is minimized when the molecules are aligned in
the observation direction and maximized when the
alignment direction is perpendicular to the observa
tion axis. The result is the oscillation of the fluores
cence signal observed at 337 nm. The period corre
sponds to the periodic revival of the aligned nitrogen
molecular wave packet.
In summary, the effect of molecular alignment on
the fluorescence signal has been confirmed by theoret
ical calculations [166]. Population trapping of nitro
gen molecules in highly excited states seems to be at
the origin of the enhancement. These results show that
trapping is an important phenomenon accompanying
ionization during ultrafast strong laser field interac
tion with atoms and molecules.
8.5. Opportunities and Directions
Our goal in this chapter has been to provide a per
sonal view of the role of nonadiabatic alignment in
laser filamentation. To that end we first briefly
reviewed the essential physic underlying nonadiabatic
alignment, the underlying theory and several of the
most pertinent observation methods. We next focused
on a number of the recently observed consequences of

alignment in femtosecond laser filamentation experi

ments.
The combination of nonadiabatic alignment with
filamentation observables is a very young field of
research and several opportunities remain to be real
ized. On the one hand, alignment can serve as a con
trol tool of filamentation dynamics and enrich the
range of applications of laser filaments. On the other,
the spectroscopic techniques of filamentation physics
offer an easy and efficient way of studying alignment,
orientation and wavepacket revivals under high pres
sure. Although attention so far has been exclusively
focused on the constituents of airrigid diatomic
homonuclear molecules, we expect the fundamental
interest in oriented polar molecules, in threedimen
sionally aligned nonlinear molecules, and in nonrigid
systems to further enrich the field in future years.
9. FILAMENTATION ATMOSPHERIC SCIENCE
This section describes some attempts to fulfill some
ultimate dreams; i.e., using filamentation to control
lightning and rainfalls, remotely measure pollutants
and high electric fields and deliver (guiding) high
power microwaves onto remote targets. Lightning
control experiments have been reviewed previously in
some detail [5]. So was remote sensing of pollutants
molecules in air using the fluorescence and LIDAR
technique [10, 225].
9.1. Lightning Control
Since the beginning of filamentation research, it
was already realized that one could use the filament to
guide lightning, in principle [226]. A few serious
experiments were carried out in high voltage laborato
ries resulting in the possibility to guide high voltage
discharge over a distance of up to a few meters in air
[227240]. The discharge could even go through a
shower imitating rain fall [240]. An experiment was
done in the field. The FrenchGerman Teramobile
group moved their terawatt level laser system inside a
transportable container (hence, Teramobile) to the
Langmuir Laboratory of the New Mexico Tech, on
South Baldy peak (3200 m altitude) [241]. Positive
results were obtained. According to the authors, we
quote [242], the conducting filaments generated by
the laser pointed toward the thundercloud have
behaved like a metallic tip directed towards a loaded
electrode; they have initiated corona discharges at
their tip. Our result provides observable evidence that
allows us to optimize the laser parameters in future
field campaigns.
The laser group in CRIEPI (Central Research
Institute of Electric Power Industries), Japan, in an
attempt to understand why lightning occurs at a very
low electric field of the order of 1 kV/cm, has recently
studied laserfilamentassisted corona bursts near a
highvoltage electrode at both positive and negative
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

polarities [233, 238, 239]. Filaments were created

below a high voltage electrode by focusing
800 nm/60 fs/84 mJ Tisapphire laser pulses by a long
focal length concave mirror (f = 10 m). Using imaging
technique, they measured the UV fluorescence from
the second positive band of nitrogen molecule
(N2 : C3u
N2 : B3g) in and around the filament
zone near the high voltage electrode. The position of
the UV source moved away from the electrode after the
laser filament was formed. The result was correlated
with the still images of the corona bursts. The latter
image showed that streamer shot out away from the
electrode up to the position where the electric field was
below 5 kV/cm. They analyzed and concluded that the
runaway (high energy) electrons' avalanche would
have ignited the corona bursts.
In spite of the strong interest among many scien
tists, the development of filament induced lightning,
so far, has not gone very fast. There are at least two fac
tors that might have slowed down a smooth develop
ment of filament lightning control. One is the laser
technology on which we shall comment at the end of
this review. The other is the detailed physics of fila
ment induced discharge which is still far from being
understood. That is to say, lightning control needs to
have more fundamental physical studies.

41

Laser filamentation in air

Pollutant cloud
Filament zone

Spectrometer

Fs laser system

Fig. 40. Experimental scheme to remotely measuring pol

lutants in air using filamentation induced fluorescence
spectroscopy.

Fluorescence, arb. units

1% CH4 in one atm. air
5 mJ, 100 shots

50000
40000
30000
20000

CH(8 >X)
CH(C >X)

9.2. Remote Detection of Polluting

and/or Dangerous Molecules in Air

10000
310

We shall only discuss a technique that has been pro

posed and the feasibility demonstrated recently in the
laboratory to remotely measure pollutant and/or dan
gerous molecules. It is the fluorescence coupled to a
LIDAR detection system. This is because the physical
process involved in this technique reveals some new
physics, namely, superexcited states excited by
intense fs laser. Another technique will be very briefly
discussed. It is based upon stimulated Raman scatter
ing by the pollutant molecules in the path of the fila
ment.
9.2.1. Fluorescence + LIDAR technique. It was
proposed by one of us (SLC) that molecules inside a
strong fs laser field would emit fluorescence that was
characteristic of the molecules; i.e., each molecule
would emit a finger print fluorescence [10]. Systematic
experiments were carried out in our laboratory and
indeed, it was discovered that different molecules,
after interaction with the strong laser field inside the
filament, would exhibit distinctive finger print fluores
cence [10, 225, 243]. However, the fluorescence
comes mainly from the fragments of the molecules
which are dissociated by the strong laser. An exception
is nitrogen molecule which fluoresces from its neutral
and ionic states [51, 52]. The claim is that using only
one intense (fs) laser, one could remotely detect most
if not all pollutants in air [225].
LASER PHYSICS

Vol. 22

No. 1

2012

CH(A >X)

330

350

370

390 410 430

Wavelength, nm

Fig. 41. LIDAR spectrum from inside a filament in 1%

CH4 in air at one atmospheric pressure. The three CH
bands are mixed with the normal nitrogen fluorescence.

Fluorescence, arb. units

CH
7000
6000

Contaminated air
Air

N2

5000

N2
N2

4000
3000
2000
424

426

428

430
432
434
Wavelength, nm

Fig. 42. An enlarged section of the spectrum in Fig. 41

showing that one can resolve the fluorescence of the
CH(A
X) band from the background nitrogen fluo
rescence.

42

CHIN et al.

So far, we were able to remotely detect fluorescence

in the near i.r. through the visible to the near UV spec
tral ranges where sensitive detectors such as photo
multiplier and, in particular, ICCD (intensified CCD)
detector exist. The experimental technique is based
upon the observation of what we call by clean fluores
cence [10, 225]. That is to say, even though plasma
exists inside a filament, there is no optical breakdown
of air. Only multiphoton/tunnel ionization takes
place. The plasma density is very low. In air, it is two to
four orders of magnitude lower than the density of air
depending on the external focusing condition [27].
Thus, it is the unimolecular interaction process that
leads to the fragmentation and fluorescence. We shall
comment on the physics of interaction later. But now,
let us dwell upon the experimental observation.
Figure 40 is a schematic drawing of what one would
like to do in atmospheric sensing of pollutants. Fila
menting fs laser pulses are sent into the atmosphere.
Pollutant molecules in the strong field inside the fila
ment zone will be dissociated and the fragments will
fluoresce. The back scattered fluorescence is captured
by a spectrometer coupled to an ICCD detector. This
latter technique is the LIDAR (light detection and
ranging) technique.
In the laboratory, the same scheme was used to test
this idea except that the laser was shot horizontally
through a shorter distance of up to 100m into an inter
action chamber where gas mixtures were prepared.
Figures 41 and 42 show the spectrum obtained from a
mixture of 1% CH4 in air at one atmospheric pressure
[244]. Figure 41 gives the spectrum spanning from the
near UV to the blue showing the normal nitrogen flu
orescence bands [51, 52] emitted from air filaments as
well as the three CH bands from the dissociation of the
CH4 molecules.
Figure 42 is an enlarged section of the spectrum
showing that the CH(A
X) band could be resolved
from the background nitrogen fluorescence when
compared to the spectrum of pure air. This means that
one can identify CH4 in air. Using the LIDAR equa
tion [245], one can estimate the distance at which one
could still measure the signal down to three times the
rootmeansquare noise level (3 level). Under the
experimental conditions in our laboratory (up to
~1 TW peak power), the distance could be as far as a
few hundred meters to one or two km. This distance
could be easily extended by increasing the peak power
and other parameters of the laser pulse, in principle.
More will be discussed on the current challenge of the
laser technology later.
We have tested the feasibility of remotely (up to
100 m) detecting many gaseous, aerosols, biological
and solid targets inside the laboratory environment.
(In the case of solids, the targets are placed in the path
of the filament zone (Fig. 40). They are: CH4, C2H2
(gases), ethanol (vapor), powders of egg white and
yeast, grain dusts of barley, corn and wheat, water

aerosols containing multiple solutes (NaCl, PbCl2,

CuCl2 and FeCl2), smoke from mosquito repellant and
metallic targets (lead, copper and aluminum). The
technique for the detection of solid targets is called
femtosecond laser induced breakdown spectroscopy
(FIBS) and is the same as the popular LIBS (laser
induced breakdown spectroscopy) except that the lat
ter usually uses nanosecond laser pulses to interact
with the target. All targets show finger print fluores
cence and are scalable to long distance detection
[225].
9.2.2. Physics of fluorescence from pollutant tar
gets. The interaction physics and chemistry of pollut
ants inside the filament (relatively high intensity)
yielding fluorescence have not been sufficiently stud
ied yet. They cover a vast range of subjects. Much of
these are seemingly well known physics; however, they
sometime would yield surprisingly new physics and
chemistry when fs time scale is involved. In this sub
section, we shall focus on the interaction of the high
field inside the filament with molecules in the gas
phase while giving some comments on the interaction
with aerosols.
In the case of dry aerosol such as the [246] smoke
coming from burning mosquito coil (repellant), the
interaction physics is not known. In the case of water
droplet aerosol containing chemical salts, experimen
tal observation indicates that the water droplets were
exploded and later end up as neutral atomic fluoresc
ing species. Fluorescence from atomic Na, Fe, Cu,
and Al as well as the hydrogen atomic fluorescence
were all observed when the filament interacts with
water aerosols containing salts of these atoms [247].
Since hydrogen atomic lines were also observed, the
water droplets might have been exploded through a yet
unknown process yielding excited hydrogen atoms.
The interaction of the high field inside the filament
with gas phase molecules reveals some hitherto unsus
pected intense field physics and chemistry. The case of
nitrogen molecules is discussed in the section on pop
ulation trapping, molecular alignment and revival. For
oxygen molecule and polyatomic molecules, excita
tion of superexcited states was observed [248251]
and reviewed [10]. So, we shall be brief.
Superexcited states were known to chemists who
use the XUV Synchrotron radiation to excite mole
cules into some highly excited states whose total
energy is above the first ionization energy. According
to Hatano [252, 253], most observed SESs are
assigned to Rydberg states that are vibrationally
(or/and rotationally), doubly, or inner core exited and
converge to each of ion states. Such single photon
(low field) excitation could be extended to multipho
ton (high field) excitation when the laser field is
strong. Indeed, this is how we explain our observation
of superexcited states [10, 248251] using a pump
probe technique.
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

After multiphoton excitation, the superexcited

states would decay quickly into various channels one
of which is the neutral dissociation of the parent mol
ecule into neutral excited fluorescing fragments. The
fluorescence coming from the neutral fragments of the
molecules such as that shown in Fig. 42 is the result of
the fast decay from the superexcited states into the
fragmentation channel. In the case of oxygen mole
cules, the superexcited states were assigned to the
various Rydberg states converging to various excited
ionic states. There are many fluorescence lines coming
from the dissociated neutral oxygen atoms. By moni
toring the fluorescence using a fs pumpprobe scheme,
the life time (fast decay time) of the superexcited
states could be measured. The physical process is such
that before the superexcited states decay into the frag
mentation and fluorescence channels, the fs probe
pulse would deexcite the superexcited state cut
ting the route towards the fragmentation and fluores
cence channel. This would result in the decrease in the
fluorescence signal. The very short decay time could
thus be measured by changing the delay between the
pump and probe pulses. For the superexcited state in
CH4 decaying into the various CH fluorescence, the
life time is about 200 fs [248, 250]. For oxygen mole
cules, it is also of the order of 200 fs [251].
9.2.3. Filamentation Raman detection of pollutants.
Another technique used by the group of Levis to
remotely detect pollutants is based upon stimulated
Raman spectroscopy inside the filament. Both gas and
liquid phase targets [254257] as well as solid targets
[258, 255] were studied. By superposing a supercon
tinuum (SC) pulse with another ultrafast laser pulse,
stimulated Raman transition was measured. For
example, in the case of acetone and methanol (liquid),
the ultrafast pulse is a 400 nm/fs pump pulse [254]. In
the case of gas phase target such as N2, O2, CO2, H2,
toluene, ammonia, and chloroform, the SC pulse
came from a filamenting 1.9 mJ/40 fs Tisapphire laser
pulse focused by a 2 m lens in air resulting in the SC
pulse with selfpulse compression. It superposed with
the ~0.5 mJ/40 fs Tisapphire laser pulse resulting in
the stimulated Raman signatures [255257]. Solid
targes were studied using fs LIBS spectroscopy on
graphite samples, for example [258, 250].
9.3. Filament Induced Water Condensation in Air
The dream of rain making through filamentation
should start from the first step of filament induced
condensation. In the first review published by one of
us (SLC) and coworkers in 2005 [5], we mentioned
this possibility. We quote the following from [5].
An experiment by the Teramobile group [259]
shows that filamentation inside a cloud chamber can
induce nucleation of the supersaturated water vapor.
The physics is similar to the standard nuclear physics
experiment in which ion tracks are left behind by
LASER PHYSICS

Vol. 22

No. 1

2012

43

Gamma rays, for example, and water molecules would

nucleate around such ions resulting in water droplets
formation. Ionization inside the filaments would have
similar consequence. We could now stretch our imag
ination and ask if this observation could be applied to
artificial rain making by generating filaments inside a
rain cloud containing supersaturated water vapor.
Since then, not much was published until this year
when the same Teramobile group repeated and further
advanced this experiment [242, 260]. They found that
even subsaturated water vapor could be condensed by
the Tisapphire fs laser filament. They attribute this to
the generation of acid such as HNO3 through the cre
ation of NO, NO2 inside the filament. They also shot
the multiTW pulses into a humid sky in Berlin and
could measure the increase of scattering of a ns probe
laser pulse by the filament generated condensation
(water droplets). The door is now opened for more
research because efficient rain making would certainly
help in relieving draught, cooling down cities during
hot and humid summers, etc. The challenge is to fully
understand the physics and chemistry of both natural
rain fall and filament induced rain fall.
9.4. Guiding Microwaves
If one could create a sheath of filaments, it would
be similar to an instantaneous and flying conduct
ing sheath. Microwave could thus be guided through
the sheath overcoming its natural large divergence over
a long distance while maintaining a high energy den
sity. Since filaments in principle could be both long
and projected to long distances, a filament sheath
could guide intense microwave to a distant target for
any appropriate application. At least one experimental
paper was published showing positive results [261].
A deformable mirror was used to spatially shape
10 Hz/27 fs/1.5 J linearly polarized Tisapphire laser
pulses (diameter; 95 mm) into a ring in air. The ring
shaped pulse generated more than 1000 closely spaced
filaments resulting in a plasma hollow cylindrical zone
with a diameter of about 45 mm. This plasma zone
would be able to guide through internal reflection
electromagnetic waves whose frequency is lower than
the plasma frequency which in the case of filament
plasma is limited to about 300 GHz [262]. In this
work, 10 GHz microwave was guided by the filament
wave guide over a distance of 16 cm. The detected
guided signal was increased by a factor of greater than
6 compared to the free propagation of the microwave
radiations.
9.5. Filament Assisted Remote Measurement
of High Static Electric Field
There is a practical need to measure high voltage
remotely. For example, in high voltage (~1000 kV)
electric power transmission lines and similar facilities,
it is necessary to measure the high voltage for diagnos

44

CHIN et al.

(a)
Filter transmission: 1.25%

2 mm

(b)
Filter transmission: 0.15%

Fig. 43. Beam patterns measured after the ionization segment using bandpass filters at (a) 850 and (b) 800 nm at 16 m from the
compressor. The former shows the redshifted postfilament core and the latter the energy reservoir which suffered from diffrac
tion of the plasma filament.

tics. During a thunder storm, it is desirable to know

the high electric field at the tip of lightning rods for
better prevention of lightning strike on buildings. It is
also desirable to measure the electric field distribution
in air under a thunder cloud. Thus, in parallel to their
study of filament assisted corona bursts, the CRIEPI
group studied remote measurement of high electric
field through measuring the UV fluorescence from
nitrogen molecules at 337.1 nm; the latter corresponds
to the strongest emission of the second positive band of
N2 : B3g) [238,
nitrogen molecules (N2 : C3u
239].
Tisapphire laser pulses (800 nm, 5060 fs, 4 mJ)
were focused by a 10 m focal length concave mirror
generating a long plasma filament at various positions
below a high voltage electrode in open air. Two sets of
high voltage spherical electrodes (radii: 60 and
250 mm) were used so as to observe interaction in dif
ferent electric fields. The 337.1 nm fluorescence from
inside the filament was measured as a function of elec
tric field. The fluorescence strength increased as the
electric field increased. The explanation is the increase
of electron avalanche ionization in the high electric
field resulting in more electronion recombination,
hence, more fluorescence from excited nitrogen mol
ecules. This measurement paved the way towards
remote measurement of electric field for practical pur
poses such as those mentioned above.
10. NONIONIZING FILAMENT
AND HIGHER ORDER KERR CONTRIBUTION
The question of nonionizing filaments has been a
subject of debate in the mid2000 [5, 263]. Up until
recently, it seems to have been settled; i.e., a nonion
izing filament is essentially the prolongation of the
propagation after the end of the plasma filament [264].
That is to say, after the end of a filament when plasma
generation has stopped, the intensity in the filament

core of the pulse is still high but not so high as to ionize

the medium (air). Thus, selffocusing through the
Kerr term n2I is still significant but diffraction is now
stronger. This would result in a slowly diverging beam
core in which the intensity is still respectable over a
rather long distance without ionization and is consid
ered to be the nonionizing filaments observed in ref.
[263].
Recently, the group of one of us (SLC) has per
formed a systematic experiment to check this point.
Indeed, we found that the divergence of the filament
core after the ionization zone is more than one order
of magnitude smaller than the linear divergence of the
same pulse over a long distance [265]. The experiment
was done using short laser pulses (800 nm/45 fs/5 mJ)
emitted at a 10 Hz repetition rate from a typical
Ti:sapphire amplifier. The initial beam diameter after
the pulse compressor was elliptical in shape with major
and minor axes at FWHM of 4.2 and 3.9 mm, respec
tively. The laser pulses formed a single filament 1015
m after the pulse compressor and their postfilamenta
tion propagation was rigorously investigated up to 50
m. In order to avoid any damage to the equipment, a
weak reflection from the original pulse incident on
fused silica a wedge was sent to a white diffuser. A
CCD camera protected with a bandpass filter at 800
nm and appropriate ND filters was used to capture the
images.
Figure 43 presents beam patterns measured after
the ionization zone (16 m from the compressor) using
bandpass filters with 10 nm transmission window cen
tered at 850 (a) and 800 nm (b). Figure 43a shows the
perfectly circular Raman redshifted postfilament
core [24, 114]. The measured beautiful beam pattern is
the result of spatial filtering during filamentation
caused by the selffocusing of the beams fundamental
mode into the filament core [24, 25, 266]. The Raman
frequency red shift (see Subsection 5.1) phenomenon,
occurs only in the high intensity core at the front part
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

45

FWHM beam diameter, mm

Simulations
5
Experiment
Simulation, no Kerr nonlinearity after z = 24 m

5 mm

10

20

30
40
50
Propagation distance, m

Fig. 44. Full width at half maximum beam diameter of 800 nm radiation measured along the propagation direction in the exper
iment (circles) and the simulations (solid line). In the experiment no data points were placed between 1225 m because a clear
axial hot spot could not be observed due to strong diffraction. The dashed curve is obtained if the nonlinear contribution to the
Kerr nonlinearity nk = 0 (see Eq. (2)) at z 24 m. The inserted beam patterns correspond to characteristic events of the pulses
propagation.

of the pulse; hence, it does not encounter any plasma

and is perfectly circular. However, the pattern in
Fig. 43b is the result of the diffraction of the back part
of the pulse including the background reservoir [17] by
the cylindrical plasma filament left behind by the front
part of the pulse. Even though the core profile around
850 nm (Fig. 43a) is of excellent quality, its energy
content is rather weak. For comparison, we need to use
a higher filter transmission (1.25%) to obtain the pic
ture in Fig. 43a while in Fig. 43b, the filter transmis
sion was lower (0.15%).
It was demonstrated numerically that in the far
field, diffraction on a circular opaque obstacle [267]
(plasma), develops into an intense axial spike sur
rounded by weak Bessellike rings. Because of its high
energy content, the formation of this axial spike, also
known as Aragos spot, is present everywhere along the
propagation axis in the exact center of the dark shadow
[267]. It is at the source of intense ionizationfree light
channels. However, as described in the following para
graphs, diffraction only cannot explain the low diver
gence observed in the experiment.
Figure 44 shows the evolution of the beam diameter
vs propagation distance measured using the beam pat
terns at 800 nm through a filter of 10 nm bandwidth.
The beam patterns inserted present characteristic
events during the propagation of the laser pulses mea
sured at 800 nm. The laser pulses first underwent fila
mentation giving rise to the minimum zone between
LASER PHYSICS

Vol. 22

No. 1

2012

12 and 25 m where plasma was generated [268]. In this

zone, diffraction of the back part of the pulse by the
plasma generated by the peak of the pulse (as discussed
above in Fig. 43) prohibited us to measure the core
diameter correctly. The axial hot spot could not be
observed due to strong diffraction effects such that no
data points were placed in this range. However, beyond
25 m, the Bessellike pattern evolved into a central
spike with extremely good spatial mode and a low
divergence of 0.03 mrad. 51 m after the pulse compres
sor, the channels intensity was estimated to be 0.5
TW/cm2 which could not ionize air, but could induce
sufficiently high selffocusing to reduce linear diffrac
tion over a long distance. The solid line in Fig. 44
shows the evolution of the simulated FWHM diameter
of the radiation filtered at 800 nm. The filament starts
3 m earlier in the simulations as compared to the
experiment (filled circles in Fig. 44) due to the purely
Gaussian and not elliptical initial intensity distribu
tion used for the calculation. In addition, the larger
postfilament diameter observed in the experiment
was caused by the weak reflected pulse which suffered
diffraction during its propagation between the wedge
and the CCD. At the same time the divergence angle
after the end of the filament is in quantitative agree
ment with the experimental one. The physical reason
for the slow divergence is the (3) Kerr nonlinearity of
the neutral air molecules, which prevents the postfil
ament beam to diverge. In fact, the dashed curve in
Fig. 44 corresponds to the simulated evolution of the

46

CHIN et al.

Plasma
2

Blue: 700 nm

Higher
order Kerr
700 nm

y, mm

1
0

1
2
750 nm

Fig. 45. Beam patterns in the middle of the filament in air during the filamentation of a Tisapphire laser at 800 nm/45 fs/5 mJ.
Bottom: experimentally obtained pattern through a filter at 750 nm (bandwidth: 10 nm). Top row: simulated beam patterns; left,
first order Kerr term plus plasma generation; all Kerr terms without plasma generation.

pulse if the Kerr nonlinearity were switched off for

distances larger than 24 m. Consequently, selffocus
ing must have played a dominant role to maintain a
small diameter over long distances.
Another experiment was performed using low peak
power pulses (low energy long ~300 ps uncompressed
pulses) to measure the divergence in the linear regime
under the same propagation conditions. It was found
that the divergence started to decrease from ~1 MW
peak power onwards. That is to say, nonlinear regime
starts at around 1 MW under our experimental condi
tion. This shows that nonionizing filaments are
indeed the slowly diverging postionization core.
According to this observation, one could also con
clude that nonionizing filament could be defined as
any beam with divergence smaller than linear diver
gence due to weak selffocusing. It could be generated
without prior ionization. In our case, so long as the
peak power is higher than 1 MW, the beam could be
considered as a nonionizing filament.
A new development occurred in the past year.
Loriot et al. [269] measured quantitatively the higher
order terms of the Kerr nonlinear index of refraction
of air in a series expansion with alternate positive and
negative signs. Bjot et al. [270] calculated the forma
tion of filament in air by including these higher order
Kerr terms. The simulated result of the latter showed
some very interesting results. They found that even
without including the ionization term in the wave

equation, the higher order terms, which together are

negative, are strong enough to stop the selffocusing
due to the first Kerr term n2I. Thus, according to the
authors of [244], any filamentation would not involve
ionization. The plasma generation in all experiments
so far is simply considered by them as a side effect
[270].
To confront this new development, a comparative
analysis was performed recently by one of us (Olga
Kosareva of Moscow State University and her group)
in the simulation of conical emission using the con
ventional model and the higher order model. The
results of the simulation were compared with experi
ments from the group of one of us (SLC) in Laval Uni
versity [19]. It was found that in the experiment, con
ical emission is evident in the antiStokes side of the
supercontinuum. In the simulation, only the conven
tional model reveals conical emission while the new
higher order model does not. Figure 45 shows such a
comparison.
It is evident that the simulated pattern using the
usual first Kerr term and plasma generation looks sim
ilar to the experimental result; i.e., with ring structure
(diffraction by the plasma). The pattern with all the
Kerr terms without plasma does not show such diffrac
tion rings.
We shall not draw a conclusion on this matter at
this stage of development yet. Comparison with other
experiments is being carried out by other groups. It
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

probably will take another year or so before a definitive

comment could be made on this interesting subject.
11. CHALLENGES AHEAD
The field of intense fs laser filamentation in air has
aroused a lot of attention in recent years probably
because of the prospect of projecting the pulse to long
distances at high intensity, a characteristic that no
other laser possesses. This, in principle, would open up
a lot of possibilities to do remote sensing, remote
interaction, remote high laser field applications,
remote nonlinear optics, etc. We have talked about
lightning control, rain making, THz generation,
remote sensing of pollutants, remote measurement of
high electric field, etc. However, it is precisely the
word remote that is challenging us. How far away
could we really project the pulse and obtain real fila
mentation with high energy at high (clamped) inten
sity?
Most of the past and current research work is still
being carried out in the laboratory. The possibility to
carry out the socalled remote experiment is mostly
extrapolation of laboratory results (see for example,
ref. [225] in the case of remote detection of pollut
ants). Limited success was achieved in the field with
the very few mobile TW laser systems such as the
European Teramobile and the Canadian T&T system,
the latter being mainly reserved for defense related
work. There are other TW level mobile laser systems
but the number, the whereabout and the results are
unknown because they are classified. Yet, there is a
need to carry out more systematic field tests after lab
oratory demonstration of the feasibility of a certain
application.
A practical difficulty is related to the cost of the
currently available fs high power laser system; it is far
too expensive. When applying the fruit of the research
to application in the commercial market, it would
become too expensive for any investor to seriously
consider putting money into developing a product
using multiTW lasers. The cost of the laser system
would most probably go down in the future when more
commercializable applications are demonstrated.
To the best of our knowledge, classified or not, cur
rent high power fs laser system could not properly
deliver very high energy, high (clamped) intensity
plasma filaments to distances of a few km or more
away. In the open literature, the longest is a few hun
dred meters [263]. Since any practical application
would need to deliver much energy and high
(clamped) intensity at targets situated at longer dis
tances, we should question why it is so difficult to
achieve this goal since, in principle, there should have
no difficulty.
The difficulty is twofold. One is multiple filament
competition and the other is beam quality. The two of
them are closely related. Bad beam quality with irreg
LASER PHYSICS

Vol. 22

No. 1

2012

47

ular intensity distribution across the wave front would

lead to multiple filaments. For long distance propaga
tion, it is necessary to increase the beam diameter. If it
is under free propagation condition, multiple filament
competition is unavoidable reducing the efficient con
version of the laser energy into useful filament energy.
One could use a focusing lens to focus the large diam
eter beam to a predetermined distance so that only
near the geometrical focusing zone will there be fila
ment formed [150]. Under this condition, only fila
ment cooperation occurs in the external focal region
where the diameter is small. Multiple filaments under
this condition of small diameter will be cooperative as
discussed in Section 2. This technique is successful up
to 100 m or so. For longer distances, the geometrical
focal diameter becomes larger and multiple filaments
competition occurs. Using deformable mirror would
correct the wave front aberration but not the intensity
distribution. Hence, the success is limited [271]. The
key correction scheme, in our opinion, is to improve
the beam quality. Until this happens, real long dis
tance filamentation (1 km or more) would have to be
patient. All current TW to multiTW level lasers
(mostly Ti:sapphire laser) use multipass bowtie
configuration along with a nonuniform pumping.
Hot spots/zones in the pulse front are unavoidable.
These are the sources of multiple filaments.
On the theory side, there is a need to have a fully
quantitative 3D model to accurately predict any
experimental outcome. So far, this has not yet happen.
ACKNOWLEDGMENTS
The authors from Laval University acknowledge
the financial support of the Natural Sciences and
Engineering Research Council of Canada (NSERC),
Defence Research Development CanadaValcartier,
Canada Research Chairs, le Fonds Qubecois pour la
Recherche en Nature et Technologie (FQRNT), the
Canada Foundation for Innovation (CFI) and the
Canadian Institute for Photonics Innovation (CIPI).
They thank the contributions from Yousef Kamali,
Jens Bernhardt, Sima Hosseini, Patrick Simard,
Abdulsamad Talebpour and Huailiang Xu as well as
the technical support from Mario Martin. H.Z. and
J.W. acknowledge the financial support by the
National Natural Science Fund, National Key Project
for Basic Research, and Projects from Shanghai Sci
ence and Technology Commission. Contributions
from H. Z.s group members and students (H. Cai,
X. Yang, J. Ding, and H. Pan) and from Profs. Z. Xu
and L. Ding are highly appreciated. H.Z. and J.W.
thank Dr. A. Couairon for valuable discussions.
T. Seideman would like to acknowledge the US
Department of Energy (grant no. DEFG02
04ER15612). O.G. Kosareva and N.A. Panov
acknowledge the support from Russian Foundation
for Basic Research, grant no. 090201200a, RFASI

48

CHIN et al.

no. 02.740.11.0223, the President of the RF grant

MK2213.2010.2.
REFERENCES
1. R. Y. Chiao, E. Garmire, and C. H. Townes, Phys. Rev.
Lett. 13, 479 (1964).
2. M. Hercher, J. Opt. Soc. Am. 54, 563 (1964).
3. J. H. Marburger, Prog. Quantum Electron. 4, 35
(1975).
4. A. Braun, G. Korn, X. Liu, D. Du, J. Squier, and
G. Mourou, Opt. Lett. 20, 73 (1995).
5. S. L. Chin, S. A. Hosseini, W. Liu, Q. Luo, F. The
berge, N. Akzbek, A. Becker, V. P. Kandidov,
O. G. Kosareva, and H. Schroeder, Can. J. Phys. 83,
863 (2005).
6. A. Couairon and A. Mysyrowicz, Phys. Rep. 44, 47
(2007).
7. L. Berg, S. Skupin, R. Nuter, J. Kasparian, and
J.P. Wolf, Rep. Prog. Phys. 70, 1633 (2007).
8. J. Kasparian and J.P. Wolf, Opt. Express 16, 466
(2008).
9. V. P. Kandidov, S. A. Shlenov, and O. G. Kosareva,
Quantum Electron. 39, 205 (2009).
10. S. L. Chin, Femtosecond Laser Filamentation, Monog
raphy, Springer series on Atomic, Optical, and Plasma
Physics, Vol. 55 (Springer Sci. Business Media, New
York, 2010).
11. SelfFocusing: Past and Present, Ed. by R. W. Boyd,
S. G. Lukishova, and Y. R. Shen (Springer, New York,
2009).
12. P. Di Trapani, G. Valiulis, A. Piskarskas, O. Jedrk
iewicz, J. Trull, C. Conti, and S. Trillo, Phys. Rev. Lett.
91, 093904 (2003).
13. M. Kolesik, E. M. Wright, and J. V. Moloney, Phys.
Rev. Lett. 92, 253901 (2004).

14. A. Couairon, E. Gai z auskas, D. Faccio, A. Dubietis,

and P. Di Trapani, Phys. Rev. E 73, 016608 (2006).
15. S. L. Chin, From Multiphoton to Tunnel Ioniza
tion, in Advances in Multiphoton Processes and Spec
troscopy, Ed. by S. H. Lin, A. A. Villaeys, and
Y. Fujimura, Vol. 16 (World Sci., Singapore, 2004), pp.
273306.
16. D. Strickland and G. Mourou, Opt. Commun. 56, 219
(1985).
17. M. Mlejnek, E. M. Wright, and J. V. Moloney, Opt.
Lett. 23, 382 (1998).
18. G. Fibich, Y. Sivan, Y. Ehrlich, E. Louzon, M. Fraen
kel, S. Eisenmann, Y. Katzir, and A. Zigler, Opt.
Express 14, 4946 (2006).
19. O. Kosareva, J.F. Daigle, N. Panov, T.J. Wang,
S. Hosseini, S. Yuan, G. Roy, V. Makarov, and
S. L. Chin, Opt. Lett. 36, 1035 (2011).
20. J. Kasparian, R. Sauerbrey, and S. L. Chin, Appl.
Phys. B 71, 877 (2000).
21. A. Becker, N. Akzbek, K. Vijayalakshmi, E. Oral,
C. M. Bowden, and S. L. Chin, Appl. Phys. B 73, 287
(2001).
22. W. Liu, S. Petit, A. Becker, N. Akozbek, C. M. Bow
den, and S. L. Chin, Opt. Commun. 202, 189 (2002).

23. M. B. Gaarde and A. Couairon, Phys. Rev. Lett. 103,

043901 (2009).
24. W. Liu and S. L. Chin, Phys. Rev. A 76, 013826 (2007).
25. F. Thberge, N. Akzbek, W. Liu, A. Becker, and
S. L. Chin, Phys. Rev. Lett. 97, 023904 (2006).
26. D. Shwa, S. Eisenmann, G. Marcus, and A. Zigler,
Opt. Express 17, 6451 (2009).
27. F. Thberge, W. Liu, P. Tr. Simard, A. Becker, and
S. L. Chin, Phys. Rev. E 74, 036406 (2006).
28. M. Mlejnek, M. Kolesik, J. V. Moloney, and
E. M. Wright, Phys. Rev. Lett. 83, 2938 (1999).
29. Q. Luo, S. A. Hosseini, W. Liu, J.F. Gravel,
O. G. Kosareva, N. A. Panov, N. Akzbek, V. P. Kan
didov, G. Roy, and S. L. Chin, Appl. Phys. B 80, 35
(2005).
30. O. G. Kosareva, N. A. Panov, N. Akozbek, V. P. Kan
didov, Q. Luo, S. A. Hosseini, W. Liu, J.F. Gravel,
G. Roy, and S. L. Chin, Appl. Phys. B 82, 111 (2006).
31. S. A. Hosseini, Q. Luo, B. Ferland, W. Liu,
N. Akzbek, G. Roy, and S. L. Chin, Appl. Phys. B 77,
697 (2003).
32. S. A. Hosseini, Q. Luo, B. Ferland, W. Liu, S. L. Chin,
O. G. Kosareva, N. A. Panov, N. Akzbek, and
V. P. Kandidov, Phys. Rev. A 70, 033802 (2004).
33. C. P. Hauri, J. Gautier, A. Trisorio, E. Papalazarou,
and P. Zeitoun, Appl. Phys. B 90, 391 (2008).
34. P. Rohwetter, M. Queisser, K. Stelmaszczyk, M. Fech
ner, and L. Woste, Phys. Rev. A 77, 013812 (2008)
35. M. Hentschel, R. Kienberger, Ch. Spielmann,
G. A. Reider, N. Milosevic, T. Brabec, P. Corkum,
U. Heinzmann, M. Drescher, and F. Krausz, Nature
414, 509 (2001).
36. G. Sansone, E. Benedetti, F. Galegari, C. Vozzi,
L. Avaldi, R. Flammini, L. Poletto, P. Villoresi,
C. Altucci, R. Velotta, S. Stagira, S. De Silvestri, and
M. Nisoli, Science 314, 443 (2008).
37. C. P. Hauri, W. Kornelis, F. W. Helbing, A. Heinrich,
A. Couairon, A. Mysyrowicz, J. Biegert, and U. Keller,
Appl. Phys. B 79, 673 (2004).
38. M. Nisoli, S. D. Silvestri, and O. Svelto, Appl. Phys.
Lett. 68, 2793 (1996).
39. X. Chen, X. Li, J. Liu, P. Wei, X. Ge, R. Li, and Z. Xu,
Opt. Lett. 32, 2402 (2007).
40. D. Uryupina, M. Kurilova, A. Mazhorova, N. Panov,
R. Volkov, S. Gorgutsa, O. Kosareva, A. Savelev, and
S. L. Chin, J. Opt. Soc. Am. B 27, 667 (2010).
41. M. Ghotbi, P. Trabs, and M. Beutler, Opt. Lett. 36, 463
(2011).
42. O. D. Mcke, S. Aliauskas, A. J. Verhoef, A. Pugzlys,
A. Baltuka,V. Smilgevi@cus, J. Pocius, L. Gininas,
R. Danielius, and N. Forget, Opt. Lett. 34, 2498
(2009).
43. J. Bernhardt, W. Liu, S. L. Chin, and R. Sauerbrey,
Appl. Phys. B 91, 45 (2008).
44. W. Liu and S. L. Chin, Opt. Express 13, 5750 (2005).
45. O. G. Kosareva, W. Liu, N. A. Panov, J. Bernhardt,
Z. Ji, M. Sharifi, R. Li, Z. Xu, J. Liu, Z. Wang, J. Ju,
X. Lu, Y. Jiang, Y. Leng, X. Liang, V. P. Kandidov, and
S. L. Chin, Laser Phys. 19, 1776 (2009)
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

46. Z. G. Ji, J. S. Liu, Z. X. Wang, J. Ju, X. M. Lu,
Y. H. Jiang, Y. X. Leng, X. Y. Liang, W. Liu, S. L. Chin,
R. X. Li, and Z. Z. Xu, Laser Phys. 20, 886 (2010).
47. J. Bernhardt, W. Liu, F. Thberge, H. L. Xu,
J. F. Daigle, M. Chteauneuf, J. Dubois, and
S. L. Chin, Opt. Commun. 281, 1268 (2008).
48. J. MuthBoehm, A. Becker, and F. H. M. Faisal, Phys.
Rev. Lett. 85, 2280 (2000).
49. A. JaronBecker, A. Becker, and F. H. M. Faisal, Phys.
Rev. A 69, 023410 (2004).
50. A. Talebpour, S. Petit, and S. L. Chin, Opt. Commun.
171, 285 (1999).
51. A. Talebpour, M. AbdelFattah, and S. L. Chin, Opt.
Commun. 183, 479 (2000).
52. H. L. Xu, A. Azarm, J. Bernhardt, Y. Kamali, and
S. L. Chin, Chem. Phys. 360, 171 (2009).
53. A. Talebpour, J. Yang, and S. L. Chin, Opt. Commun.
163, 29 (1999).
54. S. Q. Xu, Y. B. Zheng, Y. Liu, and W. Liu, Laser Phys.
20, 1968 (2010).
55. X. Lu, X. L. Liu, T. T. Xi, X. Liu, and J. Zhang, in Pro
ceedings of the 3rd International Symposium on Fila
mentation (COFIL 2010), Iraklion, Greece, May 31
June 5, 2010, p. 73.
56. D. J. Mitchell, A. W. Snyder, and L. Poladian, Phys.
Rev. Lett. 77, 271 (1996).
57. V. Tikhonenko, J. Christou, and B. LutherDavies,
Phys. Rev. Lett. 76, 2698 (1996).
58. M.F. Shih, M. Segev, and G. Salamo, Phys. Rev. Lett.
78, 2551 (1997).
59. W. Krolikowski and S. Holmstrom, Opt. Lett. 22, 369
(1997).
60. G. I. Stegeman and M. Segev, Science 286, 1518
(1999).
61. A. C. Bernstein, M. McCormick, G. M. Dyer,
J. C. Sanders, and T. Ditmire, Phys. Rev. Lett. 102,
123902 (2009).
62. Y. Liu, M. Durand, S. Chen, A. Houard, B. Prade,
B. Forestier, and A. Mysyrowicz, Phys. Rev. Lett. 105,
055003 (2010).
63. X. Yang J. Wu, Y. Peng, S. Yuan, and H. Zeng, in Pro
ceedings of the CLEO (2008), paper CTuE6; X. Yang,
J. Wu, Y. Peng, Y. Tong, P. Lu, L. Ding, Z. Xu, and
H. Zeng, Opt. Lett. 34, 3806 (2009).
64. X. Yang, J. Wu, Y. Peng, Y. Tong, S. Yuan, L. Ding,
Z. Xu, and H. Zeng, Appl. Phys. Lett. 95, 111103
(2009).
65. T. Xi, X. Lu, and J. Zhang, Opt. Common. 282, 3140
(2009).
66. N. Akzbek, A. Iwasaki, A. Becker, M. Scalora,
S. L. Chin, and C. M. Bowden, Phys. Rev. Lett. 89,
143901 (2002).
67. R. A. Ganeev, H. Singhal, P. A. Naik, J. A. Chakera,
M. Kumar, and P. D. Gupta, Phys. Rev. A 82, 043812
(2010).
68. P. Rohwetter, M. Queisser, K. Stelmaszczyk, M. Fech
ner, and L. Woste, Phys. Rev. A 77, 013812 (2008).
69. H. Stapelfeldt and T. Seideman, Rev. Mod. Phys. 75,
543 (2003).
LASER PHYSICS

Vol. 22

No. 1

2012

49

70. H. Cai, J. Wu, P. Lu, X. Bai, L. Ding, and H. Zeng,

Phys. Rev. A 80, 051802(R) (2009).
71. P. Bjot, Y. Petit, L. Bonacina, J. Kasparian, M. Mo
retand, and J.P. Wolf, Opt. Express 16, 7564 (2008).
72. J. Wu, H. Cai, P. Lu, X. Bai, L. Ding, and H. Zeng,
Appl. Phys. Lett. 95, 221502 (2009).
73. J. Wu, Y. Tong, X. Yang, H. Cai, P. Lu, H. Pan, and
H. Zeng, Opt. Lett. 34, 3211 (2009).
74. T. Xi, X. Lu, and J. Zhang, Phys. Rev. Lett. 96, 025003
(2006).
75. B. Shim, S. E. Schrauth, C. J. Hensley, L. T. Vuong,
P. Hui, A. A. Ishaaya, and A. L. Gaeta, Phys. Rev. A
81, 061803(R) (2010).
76. J. Wu, H. Cai, Y. Peng, Y. Tong, A. Couarion, and
H. Zeng, Laser Phys. 19, 1759 (2009).
77. H. Cai, J. Wu, P. Lu, Y. Tong, X. Yang, X. Bai, H. Pan,
H. Li, and H. Zeng, Sci. China: Phys. Mech. Astron.
53, 1036 (2010).
78. R. Boyd, Nonlinear Optics, 2nd ed. (Academic, New
York, 2003).
79. V. Renard, O. Faucher, and B. Lavorel, Opt. Lett. 30,
70 (2005).
80. H. Cai, J. Wu, A. Couairon, and H. Zeng, Opt. Lett.
34, 827 (2009).
81. J. Wu, H. Cai, Y. Tong, and H. Zeng, Opt. Express 17,
16300 (2009).
82. J. Wu, H. Cai, A. Couarion, and H. Zeng, Phys. Rev.
A 79, 063812 (2009).
83. F. Calegari, C. Vozzi, S. Gasilov, E. Benedetti, G. San
sone, M. Nisoli, S. De Silvestri, and S. Stagira, Phys.
Rev. Lett. 100, 123006 (2008).
84. S. Varma, Y.H. Chen, and H. M. Milchberg, Phys.
Rev. Lett. 101, 205001 (2008).
85. J. Itatani, J. Levesque, D. Zeidler, H. Niikura,
H. Ppin, J. Kieffer, P. B. Corkum, and D. Villeneuve,
Nature 432, 867 (2004).
86. R. Bartels, T. Weinacht, N. Wagner, M. Baertschy,
C. Greene, M. Murnane, and H. Kapteyn, Phys. Rev.
Lett. 88, 013903 (2002).
87. J. Wu, H. Cai, H. Zeng, and A. Couairon, Opt. Lett.
33, 2593 (2008).
88. J. Wu, H. Cai, A. Couarion, and H. Zeng, Phys. Rev.
A 80, 013828 (2009).
89. T. Kanai, S. Minemoto, and H. Sakai, Nature 435, 470
(2005).
90. J. Wu, H. Qi, and H. Zeng, Phys. Rev. A 77, 053412
(2008).
91. Y. Chen, C. Marceau, F. Thberge, M. Chateauneuf,
J. Dubois, and S. L. Chin, Opt. Lett. 33, 2731 (2008).
92. D. Faccio, A. Averchi, A. Lotti, P. D. Trapani,
A. Couairon, D. Papazoglou, and S. Tzortzakis, Opt.
Express 16, 1565 (2008).
93. H. Zeng, J. Wu, H. Xu, K. Wu, and E. Wu, Phys. Rev.
Lett. 92, 143903 (2004); H. Zeng, K. Wu, H. Xu, and
J. Wu, Appl. Phys. Lett. 87, 061102 (2005); H. Zeng,
J. Wu, H. Xu, and K. Wu, Phys. Rev. Lett. 96, 083902
(2006).
94. H. Xu and H. Zeng, Opt. Lett. 32, 1944 (2007); H. Xu
and H. Zeng, Opt. Lett. 32, 820 (2007); W. Lei, J. Wu,
H. Cai, and H. Zeng, Opt. Lett. 34, 166 (2009).

50

CHIN et al.

95. X. Yang, J. Wu, Y. Tong, L. Ding, Z. Xu, and H. Zeng,

Appl. Phys. Lett. 97, 071108 (2010).
96. F. Thberge, N. Akzbek, W. Liu, J. Filion, and
S. L. Chin, Opt. Commun. 276, 298 (2007).
97. X. Yang, J. Wu, Y. Peng, Y. Tong, S. Yuan, L. Ding,
Z. Xu, and H. Zeng, Appl. Phys. Lett. 95, 111103
(2009).
98. J. Yu, D. Mondelain, G. Ange, R. Volk, S. Nieder
meier, J. P. Wolf, J. Kasparian, and R. Sauerbrey, Opt.
Lett. 26, 533 (2001).
99. H. Cai, J. Wu, Y. Peng, and H. Zeng, Opt. Express 17,
5822 (2009).
100. J. Wu, H. Cai, Y. Peng, and H. Zeng, Phys. Rev. A 79,
041404(R) (2009).
101. F. A. Ilkov, L. Sh. Ilkova, and S. L. Chin, Opt. Lett. 18,
681 (1993).
102. H. Cai, J. Wu, X. Bai, H. Pan, and H. Zeng, Opt. Lett.
35, 49 (2010).
103. J. Wu, H. Zeng, and C. Guo, Phys. Rev. A 75, 043402
(2007).
104. H. Cai, J. Wu, H. Li, X. Bai, and H. Zeng, Opt.
Express 17, 21060 (2009).
105. Y. Wang, X. Dai, J. Wu, L. Ding, and H. Zeng, Appl.
Phys. Lett. 96, 031105 (2010).
106. S. L. Chin, F. Thberge, and W. Liu, Appl. Phys. B 86,
477 (2007).
107. F. Thberge, Q. Luo, W. Liu, S. A. Hosseini,
M. Sharifi, and S. L. Chin, Appl. Phys. Lett. 87,
081108 (2005).
108. L. Berge, S. Skupin, G. Mejean, J. Kasparian, J. Yu,
S. Frey, E. Salmon, and J. P. Wolf, Phys. Rev. E 71,
016602 (2005).
109. F. Thberge and S. L. Chin, Patent US No. 7729044
B2 (2010).
110. T. Fuji, T. Horio, and T. Suzuki, Opt. Lett. 32, 2481
(2007).
111. T. Fuji, T. Suzuki, E. E. Serebryannikov, and A. Zhel
tikov, Phys. Rev. A 80, 063822 (2009).
112. F. Thberge, M. Chteauneuf, V. Ross, P. Mathieu,
and J. Dubois, Opt. Lett. 33, 2515 (2008).
113. F. Thberge, M. Chteauneuf, G. Roy, P. Mathieu,
and J. Dubois, Phys. Rev. A 81, 033821 (2010).
114. Y. Chen, F. Thberge, C. Marceau, H. Xu, N. Akoz
bek, O. Kosareva, and S. L. Chin, Appl. Phys. B 91,
219 (2008).
115. F. M. Mitschke and L. F. Mollenauer, Opt. Lett. 11,
659 (1986).
116. J. P. Gordon, Opt. Lett. 11, 662 (1986).
117. X. Liu, C. Xu, W. H. Knox, J. K. Chandalia, B. J. Egg
leton, S. G. Kosinski, and R. S. Windeler, Opt. Lett.
26, 358 (2001).
118. G. P. Agrawal, Nonlinear Fiber Optics (Academic, San
Diego, CA, 2001).
119. D. Faccio, A. Averchi, A. Couairon, M. Kolesik,
J. V. Moloney, A. Dubietis, G. Tamosauskas, P. Pole
sana, A. Piskarskas, and P. Di Trapani, Opt. Express
15, 13077 (2007).
120. H. Hamster, A. Sullivan, S. Gordon, W. White, and
R. W. Falcone, Phys. Rev. Lett. 71, 2725 (1993).

121. H. Hamster, A. Sullivan, S. Gordon, and R. W. Fal

cone, Phys. Rev. E 49, 671 (1994).
122. T. Lffler, F. Jacob, and H. G. Roskos, Appl. Phys.
Lett. 77, 453 (2000).
123. T. Lffler and H. G. Roskos, J. Appl. Phys. 91, 2611
(2002).
124. D. J. Cook and R. M. Hochstrasser, Opt. Lett. 25,
1210 (2000).
125. M. Kress, T. Lffler, S. Eden, M. Thomson, and
H. G. Roskos, Opt. Lett. 29, 1120 (2004).
126. T. Bartel, P. Gaal, K. Reimann, M. Woerner, and
T. Elsaesser, Opt. Lett. 30, 2805 (2005).
127. X. Xie, J.M. Dai, and X.C. Zhang, Phys. Rev. Lett.
96, 075005 (2006).
128. M. D. Thomson, M. Kre, T. Lffler, and H. G. Ros
kos, Laser Photon. Rev. 1, 349 (2007).
129. J. Dai, X. Xie, and X.C. Zhang, Appl. Phys. Lett. 91,
211102 (2007).
130. M.K. Chen, J. H. Kirn, C.E. Yang, S. S. Yin, R. Hui,
and P. Ruffin, Appl. Phys. Lett. 93, 231102 (2008).
131. A. Houard, Y. Liu, B. Prade, and A. Mysyrowicz, Opt.
Lett. 33, 1195 (2008).
132. Y. Zhang, Y. Chen, S. Xu, H. Lian, M. Wang, W. Liu,
S. L. Chin, and G. Mu, Opt. Lett. 34, 2841 (2009).
133. D. Dietze, J. Darmo, S. Roither, A. Pugzlys,
J. N. Heyman, and K. Unterrainer, J. Opt. Soc. Am. B
26, 2016 (2009).
134. K. Y. Kim, J. H. Glownia, A. J. Taylor, and G. Rod
riguez, Opt. Express 15, 4577 (2007).
135. K. Y. Kim, A. J. Taylor, J. H. Glownia, and G. Rod
riguez, Nature Photon. 2, 605 (2008).
136. X.Y. Peng, C. Li, M. Chen, T. Toncian, R. Jung,
O. Willi, Y.T. Li, W.M. Wang, S.J. Wang, F. Liu,
A. Pukhov, Z.M. Sheng, and J. Zhang, Appl. Phys.
Lett. 94, 101502 (2009).
137. N. Kaipowicz and X.C. Zhang, Phys. Rev. Lett. 102,
093001 (2009).
138. J. Dai, N. Karpowicz, and X.C. Zhang, Phys. Rev.
Lett. 103, 023001 (2009).
139. H. Wen and A. M. Lindenberg, Phys. Rev. Lett. 103,
023902 (2009).
140. F. Blanchard, G. Sharma, X. Ropagnol, L. Razzari,
R. Morandotti, and T. Ozaki, Opt. Express 17, 6044
(2009).
141. J.M. Manceau, A. Averchi, F. Bonaretti, D. Faccio,
P. Di Trapani, A. Couairon, and S. Tzortzakis, Opt.
Lett. 34, 2165 (2009).
142. C. DAmico, A. Houard, M. Franco, B. Prade,
A. Mysyrowicz, A. Couairon, and V. T. Tikhonchuk,
Phys. Rev. Lett. 98, 235002 (2007).
143. Y. Chen, C. Marceau, W. Liu, Z.D Sun, Y. Zhang,
F. Thberge, M. Chteauneuf, J. Dubois, and
S. L. Chin, Appl. Phys. Lett. 93, 231116 (2008).
144. Y. Zhang, Y. Chen, C. Marceau, W. Liu, Z.D. Sun,
S. Xu, F. Thberge, M. Chteauneuf, J. Dubois, and
S. L. Chin, Opt. Express 16, 15483 (2008).
145. A. Houard, Y. Liu, B. Prade, V. T. Tikhonchuk, and
A. Mysyrowicz, Phys. Rev. Lett. 100, 255006 (2008).
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

146. Y. Chen, T.J. Wang, C. Marceau, F. Thberge,
M. Chteauneuf, J. Dubois, O. Kosareva, and
S. L. Chin, Appl. Phys. Lett. 95, 101101 (2009).
147. Y. Liu, A. Houard, B. Prade, A. Mysyrowicz, A. Diaw,
and V. T. Tikhonchuk, Appl. Phys. Lett. 93, 051108
(2008).
148. Y. Liu, A. Houard, B. Prade, S. Akturk, A. Mysyrow
icz, and V. T. Tikhonchuk, Phys. Rev. Lett. 99, 135002
(2007).
149. T.J. Wang, Y. Chen, C. Marceau, F. Thberge,
M. Chteauneuf, J. Dubois, and S. L. Chin, Appl.
Phys. Lett. 95, 131108 (2009).
150. W. Liu, F. Thberge, J.F. Daigle, P. T. Simard,
S. M. Sarifi, Y. Kamali, H. L. Xu, and S. L. Chin,
Appl. Phys. B 85, 55 (2006).
151. J.M. Manceau, M. Massaouti, and S. Tzortzakis,
Opt. Exp. 18, 18894 (2010).
152. T.J. Wang, C. Marceau, Y. Chen, S. Yuan, F. Th
berge, M. Chteauneuf, J. Dubois, and S. L. Chin,
Appl. Phys. Lett. 96, 211113 (2010).
153. M. Durand, Y. Liu, A. Houard, and A. Mysyrowicz,
Opt. Lett. 35, 1710 (2010).
154. J. Wu, Y. Tong, M. Li, H. Pan, and H. Zeng, Phys. Rev.
A 82, 053416 (2010).
155. T.J. Wang, S. Yuan, Z.D. Sun, C. Marceau, Y. Chen,
F. Thberge, M. Chteauneuf, J. Dubois, and
S. L. Chin, Laser Phys. Lett. 8, 295 (2011).
156. J. M. Dai, X. F. Lu, J. Liu, I. C. Ho, N. Karpowicz,
and X.C. Zhang, THz Sci. Technol. 2, 131 (2009).
157. T.J. Wang, S. Yuan, Y. Chen, J.F. Daigle, C. Mar
ceau, F. Thberge, M. Chteauneuf, J. Dubois, and
S. L. Chin, Appl. Phys. Lett. 97, 111108 (2010).
158. T.J. Wang, J.F. Daigle, Y. Chen, C. Marceau, F. Th
berge, M. Chteauneuf, J. Dubois, and S. L. Chin,
Laser Phys. Lett. 7, 517 (2010).
159. T.J. Wang, C. Marceau, S. Yuan, Y. Chen, Q. Wang,
F. Thberge, M. Chteauneuf, J. Dubois, and
S. L. Chin, Laser Phys. Lett. 8, 57 (2011).
160. J. Liu and X.C. Zhang, Phys. Rev. Lett. 103, 235002
(2009); IEEE J. Sel. Top. Quantum Electron. 17, 229
(2011).
161. J. Liu, J. Dai, S. L. Chin, and X.C. Zhang, Nature
Photon. 4, 627 (2010).
162. C. Marceau, Y. Chen, F. Thberge, M. Chteauneuf,
J. Dubois, and S. L. Chin, Opt. Lett. 34, 1417 (2009).
163. S. Yuan, T.J. Wang, O. Kosareva, N. Panov, V. A. Ma
karov, H. Zeng, and S. L. Chin, in preparation (2011).
164. O. Kosareva, N. Panov, V, Makarov, I. Perezhogin,
C. Marceau, Y. Chen, S. Yuan, T.J. Wang, H. Zeng,
A. Savelev, and S. L. Chin, Opt. Lett. 35, 2904 (2010).
165. C. Marceau, S. Ramakrishna, S. Gnier, T.J. Wang,
Y. Chen, F. Thberge, M. Chteauneuf, J. Dubois,
T. Seideman, and S. L. Chin, Opt. Commun. 283,
2732 (2010).
166. A. Azarm, S. Ramakrishna, A. Talebpour, S. Hosseini,
Y. Teranishi, H. L. Xu, Y. Kamali, J. Bernhardt,
S. H. Lin, T. Seideman, and S. L. Chin, J. Phys. B 43,
235602 (2010).
167. T. Seideman, J. Chem. Phys. 103, 7887 (1995).
168. T. Seideman, Phys. Rev. Lett. 83, 4971 (1999).
LASER PHYSICS

Vol. 22

No. 1

2012

51

169. T. Seideman, J. Chem. Phys. 115, 5965 (2001).

170. C. M. Dion, A. Keller, O. Atabek, and A. D. Ban
drauk, Phys. Rev. A 59, 1382 (1999).
171. B. Friedrich and D. Herschbach, Phys. Rev. Lett. 74,
4623 (1995).
172. B. A. Zon and B. G. Katsnelson, Sov. Phys. JETP 42,
595 (1976).
173. H. Loesch and J. Moller, Faraday Discuss. 113, 241
(1999).
174. P. A. Block, E. J. Bohac, and R. E. Miller, Phys. Rev.
Lett. 68, 1303 (1992).
175. F. RoscaPruna and M. J. J. Vrakking, Phys. Rev. Lett.
87, 153902 (2001).
176. C. Leichtle, I. S. Averbukh, and W. P. Schleich, Phys.
Rev. A 54, 5299 (1996).
177. Z.C. Yan and T. Seideman, J. Chem. Phys. 111, 4113
(1999).
178. T. Seideman, Phys. Rev. A 56, R17 (1997).
179. J. G. Underwood, M. Spanner, M. Y. Ivanov, J. Mot
tershead, B. J. Sussman, and A. Stolow, Phys. Rev.
Lett. 90, 223001 (2003).
180. T. P. Rakitzis, A. J. van den Brom, and M. H. M. Jan
ssen, Science 303, 1852 (2004).
181. V. Aquilanti, M. Bartolomei, F. Pirani, D. Cappelletti,
Vecchiocattivi, Y. Shimizu, and T. Kasai, Phys.
Chem. Chem. Phys. 7, 291 (2005).
182. B. Friedrich and D. Herschbach, J. Phys. Chem. A
103, 10280 (1999).
183. T. Kanai and H. Sakai, J. Chem. Phys. 115, 5492
(2001).
184. R. Baumfalk, N. H. Nahler, and U. Buck, J. Chem.
Phys. 114, 4755 (2001).
185. H. Ohmura and T. Nakanaga, J. Chem. Phys. 120,
5176 (2004).
186. H. Sakai, S. Minemoto, H. Nanjo, H. Tanji, and
T. Suzuki, Phys. Rev. Lett. 90, 083001 (2003).
187. H. Sakai, S. Minemoto, H. Nanjo, H. Tanji, and
T. Suzuki, Eur. Phys. J. D 26, 33 (2003).
188. A. MatosAbiague and J. Berakdar, Phys. Rev. A 68,
063411 (2003).
189. A. MatosAbiague and J. Berakdar, Chem. Phys. Lett.
382, 475 (2003).
190. D. Sugny, A. Keller, O. Atabek, D. Daems, C. M. Dion,
S. Gurin, and H. R. Jauslin, Phys. Rev. A 69, 033402
(2004).
191. D. Sugny, A. Keller, O. Atabek, D. Daems, S. Gurin,
and H. R. Jauslin, Phys. Rev. A 69, 043407 (2004).
192. M. Spanner, E. A. Shapiro, and M. Ivanov, Phys. Rev.
Lett. 92, 093001 (2004).
193. C. M. Dion, A. D. Bandrauk, O. Atabek, A. Keller,
H. Umeda, and Y. Fujimura, Chem. Phys. Lett. 302,
215 (1999).
194. S. Gurin, L. P. Yatsenko, H. R. Jauslin, O. Faucher,
and B. Lavorel, Phys. Rev. Lett. 88, 233601 (2002).
195. O. Atabek, C. M. Dion, and A. B. H. Yedder, J. Phys.
B: At. Mol. Opt. Phys. 36, 4667 (2003).
196. S. Ramakrishna and T. Seideman, Phys. Rev. Lett. 95,
113001 (2005).

52

CHIN et al.

197. J. Ohkubo, T. Kato, H. Kono, and Y. Fujimura,

J. Chem. Phys. 120, 9123 (2004).
198. S. Ramakrishna and T. Seideman, Phys. Rev. Lett. 99,
103001 (2007).
199. V. Renard, M. Renard, S. Gurin, Y. T. Pashayan,
B. Lavorel, O. Faucher, and H. R. Jauslin, Phys. Rev.
Lett. 90, 153601 (2003).
200. T. Seideman and E. Hamilton, Ad. At. Mol. Opt. Phys.
52, 289 (2006).
201. S. De, I. Znakovskaya, D. Ray, F. Anis, N. G. Johnson,
I. A. Bocharova, M. Magrakvelidze, B. D. Esry,
C. L. Cocke, I. V. Litvinyuk, and M. F. Kling, Phys.
Rev. Lett. 103, 153002 (2009).
202. A. Hishikawa, H. Hasegawa, and K. Yamanouchi,
Phys. Scripta T 110, 108 (2004).
203. T. Okino, Y. Furukawa, P. Liu, T. Ichikawa, R. Itakura,
K. Hoshina, K. Yamanouchi, and H. Nakano, Chem.
Phys. Lett. 423, 220 (2006).
204. H. Hasegawa, A. Hishikawa, and K. Yamanouchi,
Chem. Phys. Lett. 349, 57 (2001).
205. F. Lgar, K. F. Lee, I. V. Litvinyuk, P. W. Dooley,
A. D. Bandrauk, D. M. Villeneuve, and P. B. Corkum,
Phys. Rev. A 72, 052717 (2005).
206. J. Gagnon, K. F. Lee, D. M. Rayner, P. B. Corkum,
and V. R. Bhardwaj, J. Phys B.: At. Mol. Opt. Phys. 41,
215104 (2008).
207. R. Drner, V. Mergel, O. Jagutzki, L. Spielberger,
J. Ullrich, R. Moshammer, and H. SchmidtBcking,
Phys. Rep. 330, 95 (2000).
208. E. Gagnon, A. S. Sandhu, A. Paul, K. Hagen, A. Cza
sch, T. Jahnke, P. Ranitovic, C. L. Cocke, B. Walker,
M. M. Murnane, and H. C. Kapteyn, Rev. Sci.
Instrum. 79, 063102 (2008).
209. Y.H. Chen, S. Varma, I. Alexeev, and H. M. Milch
berg, Opt. Express 15, 7458 (2007).
210. Y.H. Chen, S. Varma, and H. M. Milchberg, J. Opt.
Soc. Am. B 25, B122 (2008).
211. V. Renard, M. Renard, A. Rouze, S. Gurir,
H. R. Jauslin, B. Lavorel, and O. Faucher, Phys. Rev.
A 70, 033420 (2004).
212. F. Calegari, C. Vozzi, and S. Stagira, Phys. Rev. A 79,
023827 (2009).
213. J.F. Ripoche, G. Grillon, B. Prade, M. France,
E. Nibbering, R. Lange, and A. Mysyrowicz, Opt.
Commun. 135, 310 (1997).
214. P. Reckenthaeler, M. Centurion, W. F, S. A. Trushin,
F. Krausz, and E. E. Fill, Phys. Rev. Lett. 102, 213001
(2009).
215. R. Velotta, N. Hay, M. B. Mason, M. Castillejo, and
J. P. Marangos, Phys. Rev. Lett. 87, 183901 (2001).
216. Y. Mairesse, D. Zeidler, N. Dudovich, M. Spanner,
J. Levesque, D. M. Villeneuve, and P. B. Corkum,
Phys. Rev. Lett. 100, 143903 (2008).
217. Y. Mairesse, J. Levesque, N. Dudovich, P. B. Corkum,
and D. M. Villeneuve, J. Mod. Opt. 55, 2591 (2008).
218. B. K. McFarland, J. P. Farrell, P. H. Bucksbaum, and
M. Ghr, Science 322, 1232 (2008).
219. N. Kajumba, R. Torres, J. G. Underwood, J. S. Rob
inson, S. Baker, J. W. G. Tisch, R. de Nalda,
W. A. Bryan, R. Velotta, C. Altucci, I. Procino,

I. C. E. Turcu, and J. P. Marangos, New J. Phys. 10,

025008 (2008).
220. S. Ramakrishna and T. Seideman, Phys. Rev. Lett. 99,
113901 (2007).
221. A. Talebpour, M. AbdelFattah, A. D. Bandrauk, and
S. L. Chin, Laser Phys. 11, 68 (2001).
222. M. V. Fedorov and A. M. Movsesian, J. Phys. B: At.
Mol. Opt Phys. 21, L155 (1988).
223. M. V. Fedorov, in Progress Inultrafast Intense Laser
Science, Ed. by K. Yamanouchi, P. Agostini, L. Chin,
and G. Farante, Vol. 1 (Springer, Berlin, 2006), pp. 1
18.
224. I. V. Litvinyuk, K. F. Lee, P. W. Dooley, D. M. Rayner,
D. M. Villeuneuve, and P. B. Corkum, Phys. Rev. Lett.
90, 233003 (2003).
225. S. L. Chin, H. L. Xu, Q. Luo, F. Thberge, W. Liu,
J. F. Daigle, Y. Kamali, P. T. Simard, J. Bernhardt,
S. A. Hosseini, M. Sharifi, G. Mjean, A. Azarm,
C. Marceau, O. Kosareva, V. P. Kandidov,
N. Akzbek, A. Becker, G. Roy, P. Mathieu,
J. R. Simard, M. Chteauneuf, and J. Dubois, Appl.
Phys. B 95, 1 (2009).
226. S. L. Chin and K. Miyazaki, Jpn. J. Appl. Phys. 38,
2011 (1999).
227. X. M. Zhao, J.C. Diel, C. Y. Wang, and J. M. Eli
zondo, IEEE J. Quantum. Electron. 31, 599 (1995).
228. S. Tzortzakis, B. Prade, M. Franco, A. Mysyrowicz,
S. Hller, and P. Mora, Phys. Rev. E 64, 57401 (2001).
229. B. L. Fontaine, F. Vidal, D. Comtois, C.Y. Chien,
A. Desparois, T. W. Johnston, J.C. Keffer, H. P. Mer
cure, H. Ppin, and F. A. M. Rizk, IEEE Trans.
Plasma Sci. 27, 688 (1999).
230. F. Vidal, D. Comtois, C. Y. Chien, A. Desparois,
B. La Fontaine, T. W. Johnston, J. C. Kieffer, H. Mer
cure, H. Ppin, and F. A. Rizk, IEEE Trans. Plasma
Sci. 28, 418 (2000).
231. H. Ppin, D. Comtois, F. Vidal, C. Y. Chien, A. Desp
arois, T. W. Johnston, J. C. Kieffer, B. La Fontaine,
F. Martin, F. A. M. Rizk, C. Potvin, P. Couture,
H. P. Mercure, A. BondiouClergerie, P. Lalande, and
I. Gallimberti, Phys. Plasma 8, 2532 (2001).
232. D. Comtois, H. Ppin, F. Vidal, F. A. M. Rizk,
C. Y. Chien, T. Y. Johnston, J.C. Kieffer, B. La Fon
taine, F. Martin, C. Potvin, P. Couture, H. P. Mercure,
A. BondiouClergerie, P. Lalande, and Ivo Gallim
berti, IEEE Trans. Plasma Sci. 31, 377 (2003); IEEE
Trans. Plasma Sci. 31, 387 (2003).
233. T. Fujii, M. Miki, N. Goto, A. Zhidkov, T. Fukuchi,
Y. Oishi, and K. Nemoto, Phys. Plasma 15, 013107
(2008).
234. T. Fujii, N. Goto, M. Miki, T. Nayuki, T. Shindo, and
K. Nemoto, Trans. Inst. Electr. Eng. Jpn. A 125, 765
(2005).
235. N. Goto, M. Miki, T. Fujii, T. Nayuki, T. Sekiya,
T. Shindo, and K. Nemoto, Trans. Inst. Electr. Eng.
Jpn. A 125, 1059 (2005).
236. M. Rodriguez, R. Sauerbrey, H. Wille, L. Wste,
T. Fuji, Y.B. Andr, A. Mysyrowicz, L. Klingbeil,
K. Rethmeier, W. Kalkner, J. Kasparian, E. Salmon,
J. Yu, and J.P. Wolf, Opt. Lett. 27, 772 (2002).
LASER PHYSICS

Vol. 22

No. 1

2012

ADVANCES IN INTENSE FEMTOSECOND LASER FILAMENTATION IN AIR

237. D. F. Gordon, A. Ting, R. F. Hubbard, E. Briscoe,
C. Manka, S. P. Slinker, A. P. Baronavski, H. D. La
douceur, P. W. Grounds, and P. G. Girardi, Phys.
Plasma 10, 4530 (2003).
238. K. Sugiyama, T. Fujii, M. Miki, M. Yamaguchi,
A. Zhidkov, E. Hotta, and K. Nemoto, Opt. Lett. 34,
2964 (2009).
239. K. Sugiyama, T. Fujii, M. Miki, A. Zhidkov, M. Yama
guchi, E. Hotta, and K. Nemoto, Phys. Plasma 17,
043108 (2010).
240. R. Ackermann, A. K. Stelmaszczyk, P. Rohwetter,
G. Mejean, E. Salmon, J. Yu, J. Kasparian, G. M
chain, V. Bergmann, S. Schaper, B. Weise, T. Kumm,
K. Rethmeier, and W. Kalkner, L. Woeste, and
J. P. Wolf, Appl. Phys. Lett. 85, 5781 (2004).

53

255. G. Heck, E. J. Judge, J. Odhner, M. Plewicki, and

R. J. Levis, in Remote Sensing via Femtosecond Fila
ment Based Technologies, Proceedings of the 43th
Annual GomacTech Conference, Orlando, FL, USA,
1619 March 2009, session 22Laser sensing.
256. J. H. Odhner, D. A. Romanov, and R. J. Levis, SPIE
Photon. West Conf. 758, 75820M (2010).
257. J. H. Odhner, D. A. Romanov, and R. J. Levis, Phys.
Rev. Lett. 103, 075005 (2009).
258. E. J. Judge, G. Heck, E. Cerkez Bader, and R. J. Levis,
Anal. Chem. 81, 2658 (2009).
259. J. Kasparian, M. Rodriguez, G. Mjean, J. Yu, E. Sal
mon, H. Wille, R. Bourayou, S. Frey, Y.B. Andr,
A. Mysyrowicz, R. Sauerbrey, J.P. Wolf, and
L. Woeste, Science 301, 61 (2003).

241. J. Kasparian, R. Ackermann, Y.B. Andr, G. M

chain, G. Mjean, B. Prade, P. Rohwetter, E. Salmon,
K. Stelmaszczyk, J. Yu, A. Mysyrowicz, R. Sauerbrey,
L. Woeste, and J.P. Wolf, Opt. Express 16, 5757
(2008).

260. P. Rohwetter, J. Kasparian, K. Stelmaszczyk, Z. Hao,

S. Henin, N. Lascoux, W. M. Nakaema, Y. Petit,
M. Queir, R. Salam, E. Salmon, L. Woestel, and
J.P. Wolf, Nature Photon. 4, 451 (2010).

242. J. Kasparian, L. Woeste, and J. P. Wolf, OPN Opti.

Photon. News 21, 22 (July/Aug., 2010).

261. M. Chteauneuf, S. Payeur, J. Dubois, and J.C. Kief

fer, Appl. Phys. Lett. 92, 091104 (2008).

243. S. L. Chin, US Patent No. 7184143 B2 (2007).

262. A. E. Dormidonov, V. V. Valuev, V. L. Dmitriev,

S. A. Shlenov, and V. P. Kandidov, Proc. SPIE 6733,
67332S (2007).

244. Y. Kamali, Q. Sun, J.F. Daigle, A. Azarm, J. Bern

hardt, and S. L. Chin, Opt. Commun. 282, 950 (2009).
245. R. M. Measures, Laser Remote Sensing, Fundamentals
and Applications (Krieger, Malabar, Florida, 1992).

263. G. Mechain, A. Couairon, Y.B. Andre, C. DAmico,

M. Franco, B. Prade, S. Tzortzakis, A. Mysyrowicz,
and R. Sauerbrey, Appl. Phys. B 79, 379 (2004).

246. J.F. Daigle, Y. Kamali, G. Roy, and S. L. Chin, Appl.

Phys. B 93, 759 (2008).

264. S. L. Chin, Y. Chen, O. Kosareva, V. P. Kandidov, and

F. Thberge, Laser Phys. 18, 962 (2008).

247. J.F. Daigle, G. Mjean, W. Liu, F. Thberge, H. L. Xu,

Y. Kamali, J. Bernhardt, A. Azarm, Q. Sun, P. Mathieu,
G. Roy, J.R. Simard, and S. L. Chin, Appl. Phys. B
87, 749 (2007).

265. J.F. Daigle, O. Kosareva, N. Panov, T.J. Wang,

S. Hosseini,S. Yuan, G. Roy, and S. L. Chin, Opt.
Commun. (2011, in press).

248. F. Kong, Q. Luo, H. L. Xu, M. Sharifi, D. Song, and

S. L. Chin, J. Chem. Phys. 125, 133320 (2006).

266. B. Prade, M. Franco, A. Mysyrowicz, A. Couairon,

H. Buersing, B. Eberle, M. Krenz, D. Seiffer, and
O. Vasseur, Opt. Lett. 31, 2601 (2006).

249. F. Kong, and S. L. Chin, in Progress in Ultrafast Intense

Laser Science III, Ed. by K. Yamanouchi, S. L. Chin,
P. Agostini, and G. Ferrante (Springer, Berlin, 2008),
vol. 3, Ch. 6, p. 113.
250. A. Azarm, H. L. Xu, Y. Kamali, J. Bernhardt, D. Song,
A. Xia, Y. Teranishi, S. H. Lin, F. Kong, and S. L. Chin,
J. Phys. B 41, 225601 (2008).
251. D. Song, A. Azarm, Y. Kamali, K. Liu, A. Xia, Y. Ter
anishi, S. Lin, F. Kong, and S. L. Chin, J. Phys. Chem.
A 114, 3087 (2010).
252. Y. Hatano, Radiat. Environ. Biophys. 38, 239 (1999).
253. Y. Hatano, Phys. Rep. 313, 109 (1999).
254. M. Plewicki and R. J. Levis, J. Opt. Soc. Am. B 25,
1714 (2008).
LASER PHYSICS

Vol. 22

No. 1

2012

267. A. E. Siegman, Lasers (Univ. Sci. Books, Mill Valley,

CA, 1986).
268. Y. Chen, F. Thberge, O. Kosareva, N. Panov,
V. P. Kandidov, and S. L. Chin, Opt. Lett. 32, 3477
(2007).
269. V. Loriot, E. Hertz, O. Faucher, and B. Lavorel, Opt.
Express 17, 13429 (2009); Opt. Express 18, 3011
(2010).
270. P. Bjot, J. Kasparian, S. Henin, V. Loriot, T. Vieillard,
E. Hertz, O. Faucher, B. Lavorel, and J.P. Wolf, Phys.
Rev. Lett. 104, 103903 (2010).
271. J.F. Daigle, Y. Kamali, J. Bernhardt, W. Liu, C. Mar
ceau, A. Azarm, and S. L. Chin, Opt. Commun. 281,
3327 (2008).