Professional Documents
Culture Documents
(Scottish Graduate Series) Adam S. Wyatt, Ian a. Walmsley (Auth.), Robert Thomson, Christopher Leburn, Derryck Reid (Eds.) - Ultrafast Nonlinear Optics-Springer International Publishing (2013)
(Scottish Graduate Series) Adam S. Wyatt, Ian a. Walmsley (Auth.), Robert Thomson, Christopher Leburn, Derryck Reid (Eds.) - Ultrafast Nonlinear Optics-Springer International Publishing (2013)
(Scottish Graduate Series) Adam S. Wyatt, Ian a. Walmsley (Auth.), Robert Thomson, Christopher Leburn, Derryck Reid (Eds.) - Ultrafast Nonlinear Optics-Springer International Publishing (2013)
Robert Thomson
Christopher Leburn
Derryck Reid Editors
Ultrafast
Nonlinear
Optics
Ultrafast Nonlinear Optics
Scottish Graduate Series
The Scottish Graduate Series is a long-standing series of graduate level texts
proceeding from the Scottish Universities Summer Schools in Physics (SUSSP).
SUSSP was established in 1960 to contribute to the dissemination of advanced
knowledge in physics, and the formation of contacts among scientists from different
countries through the setting up of a series of annual summer schools of the highest
international standard. Each school is organized by its own committee which is
responsible for inviting lecturers of international standing to contribute an in-depth
lecture series on one aspect of the area being studied.
123
Editors
Dr. Robert Thomson Dr. Christopher Leburn
Heriot-Watt University Heriot-Watt University
Edinburgh, UK Edinburgh, UK
v
vi Previous SUSSP Schools
Guest Lecturers
Thomas M. Baer Stanford University
Chris Dorman Coherent Scotland
Allister Ferguson University of Strathclyde
Wilson Sibbett University of St Andrews
Executive Committee
Derryck T. Reid Heriot-Watt University Director and Co-Editor
David Townsend Heriot-Watt University Secretary
Henry T. Bookey Heriot-Watt University Treasurer
Robert R. Thomson Heriot-Watt University Editor
vii
viii Lecturers
ix
x Preface
The executive committee hope that this book will act as part of a lasting legacy of
an extremely interesting and fulfilling school, where participants not only expanded
their knowledge, but also formed lasting friendships and networks.
The chapters contained in the book are based on the lectures given by the lecturers
at SUSSP66. The chapters are aimed at graduate-student level and are intended to
provide the student with an accessible, self-contained and comprehensive gateway
into each subject. Chapters 1, 2, 3, and 4 are concerned with the generation
and measurement of ultrashort pulses. Chapters 5, 6, and 7 are concerned with
fundamental applications of ultrashort pulses in metrology and quantum control.
Chapters 8 and 9 are concerned with ultrafast nonlinear optics in optical fibres.
Chapters 10, 11, 12, and 13 are concerned with the applications of ultrashort pulses
in areas such as particle acceleration, microscopy and micromachining. The editors
sincerely thank the authors for their excellent and timely contributions. Matthew
Edmonds is acknowledged and thanked by the editors for his help in proofreading a
number of chapters.
xiii
Contents
xv
xvi Contents
Participants .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 357
Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 367
Chapter 1
Measuring Ultrashort Optical Pulses
1.1 Introduction
Modern laser and laser-driven sources can generate light pulses of unprecedented
brevity, with durations in the range of picoseconds to attoseconds. Such durations
are significantly shorter than any photodetector response time. Further, there is
a need for more information about the pulse than the temporal intensity profile
obtained from a simple photodetector. Sophisticated applications, such as coherent
control of atomic and molecular dynamics demand a detailed knowledge of the
electric field of the pulse, and not merely its duration [1].
The need for metrology has increased along with the development of new sources
and their application in a wide range of new fields. Of course, the need to determine
the pulse duration remains a primary application, both because this parameter is an
important specification of the laser output needed for other applications, and because
it acts as a diagnostic of the system operation.
Modern mode-locked lasers, for example, generate pulses with spectral band-
widths exceeding one octave and with durations below 10 fs, well beyond anything
that can be characterized by means of fast photodetectors. The operation of such
lasers relies on a complex combination of linear pulse propagation, influenced
by the chromatic dispersion of the laser material, the mirrors and the intra-cavity
dispersion compensating devices, together with nonlinear effects, such as self-phase
modulation of the pulse in the laser material or by saturation of an intracavity
absorption, such as in a semiconductor saturable absorber mirror (SESAM), as
well as, in some cases, space-time coupling. The optimization of a mode-locked
laser is made practicable by means of a diagnostic providing the electric field as a
An electromagnetic pulse may be specified by its electric field alone, at least below
intensities that give rise to fields that will accelerate electrons to relativistic energies.
Thus a useful notation is that of the analytic signal, whose amplitude and phase we
seek to determine via measurement. The (real) electric field of the pulse is given in
terms of the analytic signal by Eq. (1.1).
where ".t/ is an analytic function of time (and space, although we suppress other
arguments here for clarity). The signal ".t/ is taken to have compact support in the
domain .T; T /, and we shall refer to it henceforth as the “field of the ultrashort
pulse”.
The spectrum of the pulse is then defined by the Fourier transform (Eq. (1.2))
ZT
"Q .!/ D dt ".t/ e i !t; (1.2)
T
so that EQ .!/ D "Q .!/ C "Q .!/. Note that "Q .!/ contains only positive frequency
R1
components, since ".t/ D dt "Q.!/e i !t . This is therefore a reasonable descrip-
0
tion for the fields of pulses propagating in charge-free regions of space, for which
RT
the pulse area, ‚.T / D dt ".t/ D ".!
Q D 0/ must be zero.
T
A single pulse is said to be completely characterized if the function ".t/ is
known on the domain .T; T /. In practice one usually adopts the approximation
that the pulse is also characterized by the function "Q .!/ on the domain .; /,
where >> 1 = with the rms pulse duration. The sampling theorem prevents a
function from having compact support in both domains, but it is usually a reasonable
approximation to truncate the spectral function at large frequencies, where the
spectral energy falls below the noise level of the detector.
The analytic signal is complex and therefore can be expressed uniquely in terms
of an amplitude and phase
One can think of all elements, except the beamsplitters, as two-port devices: a pulse
enters at one port and exits at another. There may be ancillary ports for control
signals, such as the timing signal for the shutter opening, for example, but these are
essentially linear systems, in that the output pulse field scales linearly with the input
pulse field. Thus the input/output relations for these devices are all of the kind
ZT
"o .t/ D dt 0 H t; t 0 "i t 0 ; (1.5)
T
Where ".t/ is the analytic signal (with subscripts i and o representing the input and
output fields respectively), and H .t; t 0 / is the (linear, causal) response function of
the device. We will specify the functional forms of the common linear filters given
above in subsequent paragraphs.
The beamsplitter is a four-port device, having two input and two output ports. The
input-output relations for this device are well known, and the main utility in pulse
measurement applications is either in providing a means to generated a replica of a
pulse (one input and two outputs) or to combine the unknown pulse with a reference
pulse (two inputs and two outputs), or as elements of a interferometer in which
phase to amplitude conversion takes place.
We take it that all detectors available have a response that is slow compared to the
pulse itself, though they need not be integrating. For pulses with temporal structure
of duration less than 100 fs or so, this is usually the case. The measured signal from
an integrating detector is related to the incident field, for our purposes, via
ZTR
ˇ ˇ2
S.f / D dt 0 ˇ"f t 0 ˇ ; (1.6)
TR
In general, the signal will be averaged over a train of pulses. If each pulse in the
train is not identical then the root quantity characterizing the ensemble of pulses is
˝ ˛
C t; t 00 D " .t/" t 00 : (1.8)
where the brackets indicate either a time average over the pulse train, or an
ensemble average over repeated experiments. Eqs. (1.7) and (1.8) are identical
only if each pulse in the train is identical with all others. Note that C .t; t 00 / is
not the same as the correlation function that is derived from the pulse spectral
intensity jQ" .!/j2 . In that case, the Fourier transform yields the reduced correlation
6 A.S. Wyatt and I.A. Walmsley
R1
C 0 ./ D dt C .t; t C /. This obviously contains no more information than the
1
spectrum itself, in contrast to C .t; t 00 /.
A key issue for all methods of pulse characterization is that it is assumed that
Eq. (1.7) is the correct form of the correlation function. All inversion algorithms
assume this. Of course, for single shot measurements, based on just one pulse from
an ensemble, the assumption is necessarily valid. Further, single-shot methods can
be used to determine whether the ensemble consists of identical pulses by repeated
measurements on single realizations of the ensemble on individual pulses drawn
from the train.
It is frequently productive to work with a variation of the correlation function
that uses a two dimensional space of time and frequency – the chronocyclic phase
space [9]. The intuitive concept of chirp (that is, time-dependent frequency in the
pulse) can be most easily seen within this space. A particularly useful function in
this regard is the chronocyclic Wigner function, defined as;
ZT
t0 t0 0
W .!; t/ D dt 0 C t C ; t e i !t (1.9)
2 2
T
Note also that the Wigner function is sufficient to characterize both individual
pulses and partially coherent pulse ensembles. However, the function is not in
general positive definite, and cannot therefore be considered a probability distri-
bution of the pulse field. Indeed negative Wigner functions characterize many of
the complicated pulse shapes that are in current use in, say, quantum control. For
example, a pair of phase-locked Gaussian pulses has a significant region of phase
space where its Wigner function is negative. The restrictions on the pulse duration
and bandwidth required by Fourier’s theorem are inherent in the Wigner function,
and there is a minimum area of the chronocyclic phase space that it may occupy.
Example Wigner functions for a number of simple pulse shapes are shown in
Fig. 1.1. The concept of a time-dependent frequency or chirp is clearly visible in
Figs. 1.1b, d, and the coherence between two separate pulses, that is a well-defined
relative phase, is shown in Fig. 1.1c.
This representation sheds some light on the general form of measurements, since
Eq. (1.5) may be written in terms of the Wigner representation of the pulse field and
that of the measurement apparatus as an overlap integral
1 Measuring Ultrashort Optical Pulses 7
Fig. 1.1 Wigner functions of (a) a Fourier-transform limited Gaussian pulse, (b) a pulse with
Gaussian spectrum and quadratic spectral phase, (c) a pair of identical Fourier-transform-limited
Gaussian pulses, and (d) a pulse with Gaussian spectrum and third-order spectral phase. In each
case, the temporal and spectral marginals are plotted
Z1 Z1
S .; T / D dt d!W .!; t/ WS .!; tI ; T / : (1.12)
1 1
is similar to that of the pulse shown in Fig. 1.1a. The parameters describing this
function are W and T noted in the previous paragraph. The window function moves
around the phase space as these parameters are adjusted: parallel to the time axis
as T changes, and parallel to the frequency axis as W changes. At each location a
“sample” of the field Wigner function is taken. The overlap integral in this case is
a convolution, and the inversion algorithm is a deconvolution in two dimensions.
Because these dimensions are complementary variables, the inversion is unique.
The second approach makes use of a window function that approximates a line
in the phase space. The orientation of this line is adjusted by one parameter and
the offset with respect to the origin by another. For example, the window function
may be oriented parallel to the time axis, intersecting the frequency axis at ˝.
The overlap integral determining the signal is therefore a projection of a “slice”
of the pulse Wigner function W .!; t/ onto the frequency axis. As the orientation is
changed about the intersection frequency, a different section of W .!; t / is projected
onto this frequency, thereby building up a set of spectra parameterised by the
window function orientation, so fSi g S .I fi g/. This is called phase-space
tomography, by analogy to the process used in medical diagnostics to assemble a
3-dimensional representation of an organ from a set of 2-dimensional projections.
The third approach makes use of an apparatus that shifts the pulse Wigner
function in time or frequency (or some combination of the two) and then mixes
it with the original, unshifted, version. As can be imagined from Fig. 1.1c, this
gives rise to fringes that reveal the relative phase between different components
of the pulse. The spectral or temporal phase can be read off from the position of
these fringes, and this, together with the direct measurement of the pulse spectral
or temporal intensity (the marginals of W .!; t/) gives the pulse field directly. This
approach has the simplifying feature that the shifted pulse replica need not be moved
around the phase space, since the interference property of Wigner representations
enables the important phase information to be mapped into the amplitude domain.
These operations need to be implemented in laboratory apparatus. We may use
the linear filter model to help delineate the necessary and sufficient conditions for
such an apparatus. Linear filters are those for which the output field scales linearly
with the input field. These filters modulate the pulse being measured and possess a
characteristic response time of the order of the pulse itself, though not significantly
shorter than it. They change the pulse in a way that is prescribed by an external
signal, say the voltage applied to an electroabsorption modulator. Linear filters
may be separated into two classes: those with time-stationary response functions
and those with time-nonstationary responses. For the former class, which includes
the spectrometer and dispersive delay line, the shape of the output pulse does not
depend on which time the input pulse arrives. For the latter class, which includes
the modulator and the shutter, the output pulse shape clearly depends on the timing
of the input pulse with respect to the shutter opening or the modulator drive signal.
Time stationary filters are characterized by response functions of the form
H .t; t 0 / D S .t t 0 /, and non-stationary filters by H .t; t 0 / D N.t/ı.t t 0 /.
Equivalently in the frequency domain, stationary filters take the general form
1 Measuring Ultrashort Optical Pulses 9
HQQ .!; ! 0 / D SQ .!/ ı .! ! 0 /, and nonstationary the form HQQ .!; ! 0 / D NQ .!! 0 /,
where the tilde represents a Fourier transform.
Representative response functions for the various common elements that facili-
tate analysis of all pulse measurement apparatuses, are:
2
Shutter W N A .t/ D e .t / =g2 ; (1.13a)
2
Modulator W N P .t/ D e i'.t / ; (1.13b)
phase. Roughly speaking, if one wishes to reconstruct the field at N time points,
then at least 2N independent data points are required. While interferometric
techniques are capable of reconstructing the field by recording only the necessary
2N points, spectrography and tomography require the measurement of N2 points.
The acquisition of excess data is necessary to obtain a reliable estimate of the pulse
shape. Of course, an overcomplete data set is available from direct measurement of
the entire correlation function as well.
ZT
1
S2 ./ D dt j".t/" .t C /j2 : (1.14)
T
T
However, the autocorrelation provides very little information about the temporal
phase structure of the pulse, so from it alone, there is no way to determine
whether all the frequencies of the pulse arrive at the same time or not. If they
do, then the pulse is said to be “transform-limited”, and has the shortest possible
duration consistent with a given spectrum. The pulse duration obtained from the
1 Measuring Ultrashort Optical Pulses 11
1.3.2 Spectrograms
Where ! is the optical frequency and the relative delay between the gate and the
test pulse. It is important that the resolution of the spectral filter is very high in order
to ensure that the measured trace is effectively the spectrogram of the test pulse.
A sonogram can be measured by reverting the order of the temporal and spectral
gate [13, 14].
The spectrogram is the double convolution of the Wigner function of the pulse
with the Wigner function of the gate with a change of sign on the frequency
variable [12]:
“
S.!; / D WE .t; ! 00 /WR .t ; ! ! 00 /dtd! 00 (1.17)
12 A.S. Wyatt and I.A. Walmsley
Thus, the pulse field may be estimated from a spectrogram by means of phase
retrieval that implements a deconvolution [15]. In fact, this is the only option if the
gate is unknown. The spectrogram of Eq. (1.16) is the modulus square of the short-
time Fourier transform of the pulse. The trick in phase retrieval is to estimate the
phase of the transform, since then, a Fourier transform would directly lead to the
recovery of the pulse under test and additionally the gating function.
Phase retrieval is usually ambiguous in one dimension, but is usually unique
in two dimensions. The excess data available in the spectrogram enables iterative
reconstruction of N complex numbers specifying the field from the N2 data points,
and this can also lead to the simultaneous reconstruction of the gate [16, 17]. Fur-
thermore, in the case of the nonlinear spectrogram, there is often a known functional
relation between the pulse and the gate, since the gate is often implemented as a
nonlinear interaction with replicas of the pulse under test. Also, other information
might be available, such as the spectrum of the pulse or the transfer function of the
gate. The recovery can be performed by means of several algorithms. A very robust
approach is based on the Principal Component Generalised Projections Algorithm
[18]. The protocol works as follows: from sampled representations of the field En
and gate Rn , one calculates the complex “square root” spectrogram. The modulus
of the calculated “square-root” spectrogram is replaced by the measured signal,
while the reconstructed phase is kept. A new set of representations (EnC1 , RnC1 )
is calculated by decomposing this constructed function into its singular values.
The pair of vectors corresponding to the largest singular value is taken as the set
(EnC1 , RnC1 ). The convergence of the algorithm can be monitored by examining
the difference between the measured spectrogram and the calculated one using, for
example, the rms difference. Also, the consistency of the decomposition into an
outer product can be quantified by considering the eigenvalues of the decomposition
since, for a perfect decomposition, there is only one non-zero singular value.
Moreover, the precision of the estimate of the field can be obtained from the
distribution of the eigenvalues [19].
1.3.3 Interferograms
interfering it with a laterally shifted (or sheared) replica. The resulting intensity
interferogram can be measured with a square-law detector, and the phase simply
extracted. The spectral analogue, in which two spectrally sheared pulses are
interfered also allows direct reconstruction of the electric field in the spectral domain
using the measured spectral phase and a pulse spectrum [21].
We will focus here on techniques that use the two-frequency correlation
Q
function E.!/ EQ .! /, the phase '.!/'.! / of which can be concatenated
or integrated to get the spectral phase of the initial pulse (note that the spectral
intensity can be measured directly with an optical spectrum analyser). The spectral
shear is set by the sampling theorem, and it is typically a few percent of the total
bandwidth of the pulse under test. Too large a shear would lead to undersampling
of the pulse spectrum, while too small a shear could lead to increased sensitivity
to noise, and thus reduced precision and in some circumstances reduced accuracy
of the reconstruction. The spectral intensity can be obtained either from a separate
measurement using the spectrometer, or can be extracted from the correlation
function directly.
The quantity E.!/Q EQ .! / can be obtained by measuring the interference
of the pulse under test with its sheared replica with an optical spectrum analyser
(Fig. 1.3) The frequency shear can be implemented for example using a linear
temporal phase modulation exp.i t/. The spectral intensity of the two interfering
ˇ ˇ2 ˇ ˇ
pulses is ˇE.!/
Q ˇ C ˇE.!Q /ˇ2 C E.!/Q EQ .! / C EQ .!/E.! Q /. The
interferometric component of interest can be extracted from several measurements
of the spectral density for various relative phases between the two interfering
pulses. However, if a delay is introduced between the non-shifted and the shifted
replica, this leads to spectral fringes with small spacing, by virtue of the phase
'.!/ '.! / C !. In this case, the interferometric component can be directly
extracted using Fourier processing of a single interferogram [22].
1.3.4 Tomograms
It can be shown that an arbitrary rotation of the Wigner function can be imple-
mented using a combination of a quadratic temporal and spectral phase modulation.
Therefore, the chronocyclic tomography approach to the characterization of the
electric field consists in rotating the Wigner function by various angles using a
combination of quadratic temporal and spectral phase modulation and measuring
the spectrum of the field after manipulation, as shown in Fig. 1.4. From the set
of spectra, the Wigner function can be reconstructed using algorithms used in
tomography, such as the back-projection algorithm. This leads to the reconstruction
of the Wigner function of the source. Complete phase space rotations are hard,
however, so it is often preferred to use a simplified approach.
The simplified version of chronocyclic tomography uses a particular relation
between the frequency marginal of the rotated Wigner function and the electric field
[23, 24]. The fractional power spectrum of the pulse is obtained from the rotated
Wigner function:
Z
I˛ .!/ D W Œt cos.˛/ C ! sin.˛/; ! cos.˛/ t sin.˛/ dt (1.18)
and therefore to
@I˛ @ @'
D I ; (1.20)
@˛ @! @!
This is the form most amenable to experiment, since the bandwidth required to
generate a small shear using a phase modulator is modest.
1 Measuring Ultrashort Optical Pulses 15
Thus the spectral phase is reconstructed on a grid with spectral resolution . This
procedure is exact, but of course may undersample the spectral phase if the shear is
too large. An alternative method of reconstructing the spectral phase is via trapezium
integration as follows
XN
!n !n1
'TP !N D '.!0 / .!n /: (1.23)
2 nD1
This method is exact for reconstructing the global GDD and third order dispersion
(TOD), which are normally of primary interest, and exact for the group delay locally.
Thus higher order phase can be reconstructed with sufficient accuracy if the shear
is suitably small such that the spectral phase can be approximated linearly over a
spectral range equal to the shear.
When designing a new SPIDER or any other device, or using an existing setup, it
is necessary to consider the gross properties of the optical pulses that one wants to
be able to measure, and the range of values that these properties take on. Any given
device is designed for a particular range of parameters, such that the reconstructed
pulse will be accurate and within a specified precision [25]. When utilizing the
device outside this range, it may still be possible to obtain some information on the
pulse, but in general the accuracy of this information will diminish the further away
from this ideal range the test pulse becomes. Such considerations are a necessary
design prerequisite for every measurement device. Here, we concentrate on the first-
order design for a SPIDER instrument.
The primary parameters of importance in the design include the spectral range,
the amount of chirp, and the pulse complexity. Clearly the optics must be suitable
for the pulse spectrum and that of the frequency shifted pulses and the detector must
also be sensitive to the full spectral range spanned by both the fundamental and
frequency shifted pulses. However, the pulse bandwidth also determines the amount
of dispersion that must be applied to the ancilla to ensure upconversion with a quasi-
monochromatic frequency. Additionally, the ancilla dispersion, test pulse bandwidth
and test pulse dispersion are constrained for the same reason. The accuracy of the
reconstruction can also depend on the pulse complexity; unfortunately this can
be difficult or even impossible to quantify in general. However, there are some
important features that can be considered.
The most important parameter in any SPIDER device is setting the shear to the
correct value. According to the sampling theorem, for a given pulse with compact
support T, the spectral resolution and therefore the shear is constrained to:
2
D (1.24)
T:
This upper bound may not be attainable. Since the spectral phase is reconstructed
by concatenation of the SPIDER phase, the precision of the reconstruction is
proportional to the square root of the number of sampling points, which is itself
1 Measuring Ultrashort Optical Pulses 17
inversely proportional to the shear. Therefore it is best to set the shear close to the
sampling limit to optimize the precision.
In the case of a spectral phase that varies smoothly, it is possible to use a
shear larger than determined by the sampling limit, although it is still necessary
to reconstruct on a sampling grid with better resolution than the shear. This can be
achieved by using an integration scheme generalising Eq. (1.23). However, if the
GDD of the test pulse becomes too large, then the SPIDER phase will be aliased
(i.e. it will vary by more than ˙ between pixels). This increases the maximum
acceptable shear to:
; (1.25)
' 00 TP ı!max
where ı!max is the maximum spectral resolution of the spectrometer within the
spectral region covered by SPIDER interferogram.
Another important aspect of a standard SPIDER (i.e. one that uses sum/difference
frequency generation to create the spectrally sheared pulses) is the chirp rate of the
ancilla. This is important since to be able to create spectrally sheared replicas of the
test pulse, one needs to ensure that the instantaneous frequency of the ancilla does
not vary significantly over the duration of the test pulse. However, increasing the
chirp rate also reduces the efficiency of the nonlinear process, and thus reducing the
sensitivity of the SPIDER apparatus. Therefore there is a balance between efficiency
and precision.
Calculating the optimal chirp rate would require exact knowledge of the test
pulse. However, it is possible to place some bounds on the chirp by considering
that the frequency sweep during the nonlinear interaction must be much smaller
than the spectral shear, i.e. >> ı!CP Š T =' 00 CP . Combining this relationship
with Eq. (1.24) allows one to determine the maximum temporal support of the test
pulse that can be accurately reconstructed, given the ancilla chirp rate:
T 2' 00 CP (1.26)
Note that in practice, this does not place a strict limit on the temporal window/chirp
rate because the frequency sweep is applied across the whole temporal window and
so the error in the spectral phase at a given frequency will still remain small. In
addition, the error will depend on the exact pulse shape. As an example, in [26],
a glass block is used to stretch the ancilla, introducing a GDD of 2 104 fs2 /rad;
this will enable accurate reconstruction of pulses up to at least 300 fs in duration.
Simulation has shown that a temporal window larger than a picosecond is required
before the reconstruction becomes unacceptably inaccurate.
18 A.S. Wyatt and I.A. Walmsley
It is also necessary to ensure that the stretcher used to chirp the ancilla applies
enough GDD to significantly stretch the pulse, and thus requires that the test pulse
has a minimum bandwidth. In the highly chirped limit, the pulse duration is given by
tchirp Š ' 00 !. This is valid when the chirped pulse duration is much larger than
the Fourier transform limited (FTL) pulse duration, i.e. tchirp >> tFTL ˛= !,
where ’ is a scaling factor that depends on the shape of the spectral intensity. Thus:
r
˛
! >>
' 00 CP
p
tFTL << ˛' 00 CP (1.27)
For a Gaussian spectral intensity and defining the pulse duration/bandwidth over
the full width at half maximum (FWHM), ’ D 4ln2. Eq. (1.27) is a stringent limit,
therefore the much greater (less) than symbols should be read as at least (most) 10
(.1) times the right hand side. To illustrate, a spectral bandwidth of at least 0.1 rad/fs
( œ 30 nm @ œ0 D 800 nm) or FTL pulse duration of at most 30 fs is required for
the device in [26].
It should be noted that Eqs. (1.26) and (1.27) place sufficient bounds on the pulse
duration that can be measured for a given ancilla chirp (or equivalently the ancilla
chirp required to measure a particular pulse). This can be understood more clearly if
one asks the following questions: (1) what is the minimal spectral bandwidth of the
pulses and (2) what is the maximum dispersion (and hence maximum time window)
that one can measure with any given device.
Another design parameter is that of the temporal separation between the test pulse
replicas. This parameter is constrained by the resolution of the spectrometer, such
that it must be necessary to resolve the fringes; and it is also constrained by the need
to be able to isolate the sidebands from the DC term during Fourier filtering. The
sidebands are located at pseudo times tQ Š ˙, and the width of the DC term and
the sidebands are given by at most 2T, the limits are given by:
2T < < (1.28)
ı!max
The final design parameter is the crystal thickness. If the crystal is too thin, then
the signal may be too weak to measure. However, if the crystal is too thick, then
phasematching can restrict the bandwidth of the upconverted pulses to less than
the fundamental. It should be noted that the effect of the phasematching function
1 Measuring Ultrashort Optical Pulses 19
of the crystal is to reduce the signal in the wings of the interferogram, and thus
reduce the precision at those locations and not the precision overall. By considering
longitudinal phase matching, the length of the crystal is given by
2
L< ; (1.29)
k .!0 ; !/
where:
is the momentum mismatch between the ancilla, test pulse and signal pulse. In [26],
a 30 m crystal was used, which can support upconversion of pulses spanning 550–
1200 nm when upconverting with 750 nm. The wavelength range can be extended
by angle tuning (dithering) the crystal.
One final note should be made about SPIDER’s ability to measure complex pulses.
Provided the pulse being measured is within the bounds specified by the parameters
noted above, one can in principle measure a pulse of any shape and therefore any
complexity. This is true except for one caveat: the interferometric term extracted
from the SPIDER interferogram must have sufficient intensity at every sampling
point of the reconstruction grid to enable accurate determination of the SPIDER
phase at those points. If the intensity falls below the noise floor, then the SPIDER
phase at that point becomes undetermined, and therefore the relative phase of the
spectral regions either side of that point is also undetermined. This can have quite
significant consequences for the temporal reconstruction. Such a problem can be
alleviated by making the spectral shear large enough to span any spectral holes.
This of course will reduce the sampling of the spectral phase within the separated
spectral regions. In the case that this needs to be sampled with a fine resolution,
one may take multiple measurements with different sized shears, such that the small
shear measurement can be used to accurately reconstruct the individual spectral
regions, and the larger shear measurement to obtain the relative phase between the
separated regions, as demonstrated in [27, 28].
1.5 Conclusions
proven commercially viable, so that they have passed from being laboratory tools to
being workaday devices.
The concepts of spectrography, tomography and interferometry serve as the
major classifications for pulse shape characterization, and each may be implemented
using appropriate linear optical elements. The use of linear temporal modulators as
time-nonstationary filters in these classes of measurement has been detailed and
experimental implementations for characterization presented. As research in both
signal analysis and ultrafast optics is in continual development, it can be expected
that further exciting discoveries will be made in the near future.
The field of ultrashort pulse measurement at present is moving a number of
different directions. Now that the basic principles are well established, and the most
simple and reliable apparatuses have been demonstrated, the application to more
complex fields is underway. This includes determining the space-time structure of a
pulse, rather than simply its temporal structure, as well as identifying the fields of
pulses that are not close to the transform limit, or which may have highly structured
spectra. In another direction, the techniques developed for the optical domain are
finding application in the characterization of pulses in quite different spectral and
temporal domains, including that of attosecond duration XUV pulses generated via
high harmonic radiation from atoms.
Underlying all of this, however, is fundamental notion, well exemplified in
optics, that the development of new tools underpins scientific discovery, so that
as new methods arise, and old ones evolve, we can be sure there will emerge new
phenomena. Conversely, new science underpins the development of new technology,
so we can also be sure that there will be new techniques and applications arising
from new discovery. The symbiosis between science and technology is clear: the
generation, amplification and measurement of short electromagnetic pulses has
opened new vistas in physics, chemistry and materials science, as well as in
applications such as biomedicine and telecommunications. It should be anticipated
that accurate and precise measurement methods will continue to play a key role in
discovery and innovation.
References
1. For a recent review, see I.A. Walmsley, C. Dorrer, Characterization of ultrashort electromag-
netic pulses. Adv. Opt. Photon. 1, 308–437 (2009)
2. L. Gallmann, D.H. Sutter, N. Matuschek, G. Steinmeyer, U. Keller, C. Iaconis, I.A. Walmsley,
Characterization of sub-6-fs optical pulses with spectral phase interferometry for direct
electric-field reconstruction. Opt. Lett. 24, 1314–1316 (1999)
3. D. Strickland, G. Mourou, Compression of amplified chirped optical pulses. Opt. Commun.
56, 219–221 (1985)
4. S. Backus, C.G. Durfee, M.M. Murnane, H.C. Kapteyn, High power ultrafast lasers. Rev. Sci.
Instrum. 69, 1207–1223 (1998)
5. B. Kohler, V.V. Yakovlev, K.R. Wilson, J. Squier, K.W. Delong, R. Trebino, Phase and intensity
characterization of femtosecond pulses from a chirped-pulse amplifier by frequency-resolved
optical gating. Opt. Lett. 20, 483–485 (1995)
1 Measuring Ultrashort Optical Pulses 21
2.1 Introduction
Many modern studies of light-matter interaction require optical pulses with short
durations as well as broad frequency tuneability. Both properties are of utmost
importance for time-resolved optical spectroscopy and high-field physics. In fem-
tosecond pump-probe spectroscopy, the system under study (atom, molecule or
solid) is resonantly excited by a “pump” pulse and its subsequent evolution is
monitored by measuring the pump-induced transmission change of a delayed
“probe” pulse [1]. To be able to probe elementary photophysical and photochemical
relaxation processes, it is necessary to push the temporal resolution towards the
limit set by the oscillation period of the carrier wave. The need to excite a system
on resonance and probe optical transitions occurring at different photon energies
requires broad frequency tuneability of both pump and probe pulses [2]. On the
other hand, by focusing high-intensity ultrashort pulses in a noble gas jet, it
is possible to produce coherent bursts of XUV radiation by the so-called High
Harmonic Generation (HHG) process [3] (See Chap. 3 by Marangos et al. and
Chap. 7 by Mathias et al.). Also for the HHG process, wavelength tuneability of
the driver pulses is important, since it influences both the frequency extension and
the phase-matching conditions of the XUV pulses. In view of the above mentioned
applications, the generation of tuneable few-optical-cycle light pulses is one of the
major challenges of ultrafast optics [4].
Nowadays, several reliable sources of femtosecond pulses exist, based for
example on Ti:sapphire or Ytterbium technologies; however they emit only at fixed
frequencies. Despite this limitation, the very high peak power of these sources
enables exploiting the second order nonlinear optical effect known as Optical
Parametric Amplification (OPA) [5, 6] to extend their tuning range. The principle of
OPA is quite simple (Fig. 2.1a): in a suitable nonlinear crystal, energy is transferred
from a high frequency and high intensity beam (the pump beam, at frequency ! 3 ) to
a lower frequency, lower intensity beam (the signal beam, at frequency ! 1 ) which
is thus amplified; in addition a third beam (the idler beam, at frequency ! 2 ) is
generated. The OPA process can be given a simple corpuscular interpretation (see
Fig. 2.1b): a photon at frequency ! 3 is absorbed by a virtual level of the material
and a photon at frequency ! 1 stimulates the emission of two photons at frequencies
! 1 and ! 2 . In this interaction energy conservation:
In this section we briefly derive the equation for the propagation of an ultrashort
pulse in a dispersive nonlinear medium, that will be used in the remainder of the
chapter [6–8]. We start from Maxwell’s propagation equation for the electric field E
of a scalar plane wave:
@2 E 1 @2 E @2 P
D 0 (2.2)
@z2 c02 @t 2 @t 2
where c0 is the speed of light in vacuum and P is the polarisation describing the
medium response. We write the electric field as:
1
E.z; t / D RefA .z; t/ expŒi .!0 t k0 z/ g D fA .z; t / expŒi .!0 t k0 z/ C c:c:g
2
(2.3)
where A(z,t) is the complex field envelope. In the following we will for simplicity
drop the notation Re f: : :g. The polarisation can be decomposed in the sum of a
linear (PL ) and a non-linear (PNL ) component:
@2 E 1 @2 E @2 PL
D 0 (2.6)
@z2 c02 @t 2 @t 2
Equation (2.6) is more easily solved in the frequency domain, by introducing the
Fourier transforms:
and becomes:
@2 EQ !2
2
C 2 EQ D 0!
2
PQL (2.8)
@z c0
26 G. Cerullo and D. Brida
@2 AQ @AQ 2 Q !2 Q
2i k0 k0 A C AD 0!
2
pQL (2.9)
@z2 @z c02
We now make the Slowly Varying Envelope Approximation (SVEA), which consists
@2 AQ @AQ
in assuming that 2
k0 ; this corresponds to neglecting variations of
@z @z
the envelope over propagation lengths of the order of the wavelength. With this
assumption we obtain:
@AQ !2
2i k0 k02 AQ C 2 AQ D 0!
2
pQL (2.10)
@z c0
and obtain:
@AQ
2i k0 D k 2 .!/ k02 AQ (2.12)
@z
!
with k .!/ D n .!/. In a dispersive medium, the refractive index is a function
c0
of frequency, and the wavenumber thus becomes a nonlinear function of !. We can
further approximate:
1
k .!/ D k0 C k 00 .! !0 / C k 000 .! !0 /2 C : : : (2.14)
2
dk 1
where k 00 D D , vg0 being the group velocity of the carrier frequency
d! !0 vg0
2
d k
and k 000 D the Group Velocity Dispersion (GVD). Usually an expansion
d ! 2 !0
up to the second order is sufficient, with the third order required only for very broad
pulse bandwidths. Equation (2.12) then becomes:
@AQ .! !0 / 1 1
i Š .! !0 / AQ C k 000 .! !0 /2 AQ (2.15)
@z vg0 2
2 Ultra-Broadband Optical Parametric Amplifiers 27
@A .z; t / 1 @A i 00 @2 A
C k0 2 D0 (2.16)
@z vg0 @t 2 @t
@2 PNL
2
Š !02 pNL exp i !0 t kp z (2.18)
@t
and finally obtain the equation:
@A 1 @A i 00 @2 A 0 !0 c
C k 0 2 D i pNL exp Œi kz (2.19)
@z vg0 @t 2 @t 2n0
.2/
"0
P1NL .z; t / D A2 A3 exp fi Œ!1 t .k3 k2 / z C c:c:g (2.22a)
2
.2/
"0
P2NL .z; t/ D A1 A3 exp fi Œ!2 t .k3 k1 / z C c:c:g (2.22b)
2
.2/
"0
P3NL .z; t / D A1 A2 exp fi Œ!3 t .k1 C k2 / z C c:c:g (2.22c)
2
Obviously there are other terms on PNL at different frequencies, such as for example
2! 1 , 2! 2 , ! 1 –! 2 : : : Here we consider only the terms at ! 1 , ! 2 and ! 3 because
we assume that only the interaction between these three fields is efficient, due to the
phase-matching condition. We can then derive the following three equations for the
fields at ! 1 , ! 2 and ! 3 :
@A1 1 @A1 i @2 A1
C k 001 2 D i 1 A2 A3 exp Œi kz (2.23a)
@z vg1 @t 2 @t
@A2 1 @A2 i @2 A2
C k 002 2 D i 2 A1 A3 exp Œi kz (2.23b)
@z vg2 @t 2 @t
@A3 1 @A3 i @2 A3
C k 003 2 D i 3 A1 A2 exp Œi kz (2.23c)
@z vg3 @t 2 @t
!i .2/
where we have defined the nonlinear coupling coefficients: i D and the
4cni
“wave-vector mismatch” as: k D k3 k1 k2 . The three equations are coupled
nonlinear partial differential equations, are in general not amenable to an analytic
solution and must be treated numerically. A first simplification, that captures the
main physics of second order parametric interaction, consists in neglecting the GVD
terms, i.e. putting k 00i D 0. This is justified by considering that the three interacting
pulses are propagating with very different group velocities vgi . Unless one considers
ultrashort pulses, the effects of this group velocity mismatch (GVM) are much more
relevant than those of GVD between the different frequency components of a single
pulse. The coupled equations can be further simplified by moving to a frame of
z
reference translating with the group velocity of the pump pulse D t , and
vg3
become:
@A1 @A1
C ı13 D i 1 A2 A3 exp Œi kz (2.24a)
@z @
2 Ultra-Broadband Optical Parametric Amplifiers 29
@A2 @A2
C ı23 D i 2 A1 A3 exp Œi kz (2.24b)
@z @
@A3
D i 3 A1 A2 exp Œi kz (2.24c)
@z
d A1
D i 1 A3 A2 exp Œi kz (2.25a)
dz
d A2
D i 2 A3 A1 exp Œi kz (2.25b)
dz
d 2 A1 d A1
D i k C 2 A1 (2.26)
d z2 dz
2deff 2 !1 !2
where 2 D I3 and I3 D 12 n3 c"0 jA3 j2 is the pump beam intensity. The
c03 "0 n1 n2 n3
nonlinear coefficient deff is proportional to the element of the (2) tensor that keeps
trace of the polarisation of the beams and their propagation directions within the
nonlinear crystal lattice.
In the hypothesis of an initial signal field amplitude A1 .0/ D A10 (the “seed
beam”) and no initial idler (A2 .0/ D 0), the solutions of Eqs. (2.25) are:
( 2 )
I1 .z/ D I10 1C sinh .gz/ (2.27a)
g
2
!2
I2 .z/ D I10 sinh .gz/ (2.27b)
!1 g
q
k 2
where g D 2 2
. For the case of large gain (gz >> 1) Eqs. (2.27) further
simplify to:
I10 2
I1 .z/ D exp .2gz/ (2.28a)
4 g
30 G. Cerullo and D. Brida
!2
I2 .z/ D I1 .z/ (2.28b)
!1
giving an exponential growth of both signal and idler intensities with crystal length,
characteristic of an optical amplifier. It should also be noted, that, in the large gain
limit, signal and idler intensities are related by energy conservation, since for each
annihilated pump photon a signal and an idler photon are simultaneously generated.
The parametric gain for the signal beam, in the large gain limit, can be written as:
I1 .z/ 1 2
GD D exp .2gz/ (2.29)
Is0 4 g
Let us now discuss the factors influencing this gain:
1. G / exp(g) exponentially depends on the parameter g, which is maximum when
k D 0 (phase-matching condition). G rapidly decreases for non-zero values of
k, suggesting that phase-matching is a key condition to be fulfilled in order to
get significant amplification from the non-linear material.
2. G / exp(deff ) depends exponentially on the second order nonlinear optical
coefficient of the crystal deff ; one should therefore select the crystal with the
largest nonlinear response. There are however other considerations leading to the
choice of the crystal, such as phase matching range, dispersive properties, and
optical damage threshold.
3. G scales as the exponential of the square root of the pump intensity. This indicates
the suitability of ultrashort pulses for OPAs, due to their high peak powers. One
should try to use the highest possible pump intensity before the onset of other
nonlinear optical phenomena such as self-focusing, self-phase modulation and
beam breakup. In order to be able to use high pump intensities, it is however
important to have a spatially clean beam profile, without hot spots.
4. G scales as the exponential of the crystal length, as in an optical amplifier. With
ultra-short light pulses, however, the optimum crystal length has to be chosen
considering the durations and group velocities of the interacting pulses.
5. G scales as the exponential of the square root of the product of signal and idler
frequencies. This seems to indicate an advantage to use higher pump frequencies.
However we will see that with ultrashort pulses this advantage is often offset by
the larger difference in group velocities of the interacting pulses.
The gain of an OPA is similar to that of a classical optical amplifier based on
population inversion, but with some important differences;
1. an OPA does not have any energy storage capability, i.e. the gain is present only
during the pump pulse;
2. the gain centre frequency is not fixed, but can be continuously adjusted by
varying the phase-matching condition;
3. the gain bandwidth is not limited by the linewidth of the transition, as in an
amplifier based on population inversion, but rather by the possibility of satisfying
the phase-matching condition over a broad range of frequencies.
2 Ultra-Broadband Optical Parametric Amplifiers 31
So far we have considered the theory of OPAs with monochromatic waves. When
dealing with ultrashort pulses, one should in principle perform a full numerical
solution of Eqs. (2.23). However, considerable physical insight can already be
derived from the simpler Eqs. (2.24), showing that the efficiency of the process
is governed by the GVM between the interacting pulses. In particular, GVM
between the pump and the signal/idler pulses limits the interaction length over
which parametric amplification takes place, while GVM between the signal and
the idler beams limits the phase matching bandwidth [9]. Given a pump pulse with
duration £, one can define a pulse splitting length Lsp as the propagation length
after which the signal (or the idler) pulse temporally separates from the pump pulse
in the absence of gain, which can be expressed as:
Lspj3 D ˇ ˇ j D 1; 2 (2.30)
ˇı j 3 ˇ
Note that the pulse splitting length becomes shorter for decreasing pulse duration
and for increasing values of GVM, depending on the crystal type, pump wavelength
and type of phase matching. There is a qualitatively significant difference between
the cases in which •13 and •23 have the same or different signs. When •13 •23 > 0,
both the signal and the idler pulses walk away from the pump in the same direction
so that the gain rapidly decreases for propagation distances longer than the pulse
splitting length and eventually saturates. On the other hand, when •13 •23 < 0, signal
and idler pulses move in opposite direction with respect to the pump; in this way
the signal and idler pulses tend to stay localized under the pump pulse and the gain
grows exponentially even for crystal lengths well in excess of the pulse splitting
length. This situation is illustrated in Fig. 2.2, showing simulations of the evolution
of the temporal intensity profiles inside the gain medium, obtained by numerical
solution of Eqs. (2.24).
Figure 2.2a shows a simulation for the case •13 •23 > 0; one can see that,
for propagation lengths greater than the pulse splitting length, the parametric
amplification process stops as the signal and idler pulses linearly walk away from
the pump. Figure 2.2b shows a simulation for the case •13 •23 < 0. Here we see that
the signal growth stays exponential for propagation distances well exceeding the
pulse splitting length and that the signal pulse tends to stay localized under the
pump. To understand this effect we can consider the situation in which the signal
pulse has moved slightly to the left and the idler pulse to the right of the pump
pulse. During the parametric amplification process, the signal pulse generates idler
photons, which move to the right, i.e. towards the peak of the pump; in turn, the
idler pulse generates signal photons which move to the left, again towards the peak
of the pump. This concentration of photons under the peak of the pump explains the
exponential gain growth overcoming GVM effects.
32 G. Cerullo and D. Brida
Fig. 2.2 Simulated signal intensity profiles of OPAs in a frame of reference moving with the pump
pulse: (a) for the case •13 > 0, •23 > 0: signal and idler are slower than the pump and the effective
crystal length is equal to the pulse splitting length between pump and signal, after which both
pulses walk away from the pump; (b) for the case •13 > 0, •23 < 0: signal and idler walk in different
directions with respect to the pump, as a result they stay trapped under the pump, increasing the
interaction length
Let us now discuss which conditions determine the gain bandwidth of an OPA.
Ideally one would like to have a broadband amplifier, i.e. an amplifier which, for a
fixed pump frequency !N 3 , provides a more or less constant gain over an as broad
as possible range of signal frequencies. To this end, one needs to keep the phase
mismatch k as small as possible over a large bandwidth. Practically, however,
the phase-matching condition can be satisfied only for a given set of frequencies
(!N 1 ; !N 2 ; !N 3 ), such that:
k D k .!N 3 / k .!N 1 / k .!N 2 / D 0 (2.31)
@k1 @k2
kD !C ! D ı12 ! (2.33)
@!1 @!2
1 1
where ı12 D is the GVM between signal and idler pulses. The full
vg2 vg1
width at half maximum parametric gain bandwidth for a crystal of length L can
then be calculated from Eq. (2.29), within the large gain and low pump-depletion
approximations, as:
2 Ultra-Broadband Optical Parametric Amplifiers 33
4 log .2/1=2 1=2 1
!D (2.34)
L jı12 j
Equation (2.34) shows that the gain bandwidth is inversely proportional to the GVM
between signal and idler and has only a square root dependence on small-signal gain
and crystal length. For the case when vg1 D vg2 (group velocity matched OPA), Eq.
(2.34) loses validity and (2.33) must be expanded to the second order in !, giving:
4 log .2/1=4 1=4 1
!D ˇ 00 ˇ (2.35)
L ˇk1 C k2 00 ˇ1=2
In this case the gain bandwidth is inversely proportional to the square root of the sum
of the GVDs of signal and idler pulses. The conditions for achieving group-velocity
matched OPAs will be discussed in Sect. 2.6.
Fig. 2.3 Scheme of an ultrafast optical parametric amplifier. BS beam splitter, OPA optical
parametric amplification stage
Fig. 2.4 Scheme of a near-IR OPA. BS beam splitter, DM dichroic mirror, LPF long-pass filter,
VA variable attenuator
consisting of a 3-mm-thick “-barium borate (BBO) crystal cut for type II phase
matching (™ D 26ı , ® D 0ı ). Wavelength tuning is achieved by tilting the crystal,
thus changing the phase-matching condition. Typical signal energies after the pre-
amplifier are up to 6 J. The power amplifier stage consists of an identical BBO
crystal pumped by 450 J; in this case amplified pulse energies up to 200 J are
generated. The signal beam is tuneable from 1.1 to 1.6 m and the idler up to
2.8 m, limited by absorption in the BBO crystal; pulsewidth ranges from 30 to
50 fs according to wavelength.
2 Ultra-Broadband Optical Parametric Amplifiers 35
Equation (2.34) makes it clear that, in order to obtain broad phase matching
bandwidths, one must achieve, for a given signal frequency !N 1 , •12 D 0, i.e. group
velocity matching between signal and idler pulses. It will then become possible
to amplify a broad bandwidth centred around !N 1 . We will show in the following
that, by using suitable non-linear crystals, pump frequencies and phase-matching
configurations !N 1 can be varied over a very broad range, spanning from the visible
to the mid-infrared, enabling the generation of widely tuneable few-optical-cycle
light pulses from OPAs.
In an OPA using a collinear interaction geometry, the propagation direction
inside the nonlinear crystal is selected to satisfy the phase-matching condition
(k D 0) for a given signal wavelength. In this configuration the signal and
idler group velocities are in general not matched. Group velocity matching can
be obtained in a type I (or type 0 in a periodically poled crystal) degenerate
configuration, in which signal and idler have the same frequency (! 1 D ! 2 D ! 3 /2)
and the same polarisation.
If the signal wavelength is tuned away from degeneracy, then the •12 D 0
condition is generally not fulfilled in a collinear configuration, leading to narrow
phase-matching bandwidths. An additional degree of freedom can be introduced
using a non-collinear geometry [11], in which pump and signal wave-vectors form
an angle ’ (independent of signal wavelength) and the idler is emitted at an angle
with respect to the signal. In this case the phase matching condition is a vector
equation, k3 D k1 C k2 that, when projected on directions parallel and perpendicular
to the signal wave-vector, becomes:
Note that the angle is not fixed, but it depends on the idler frequency according to
Eq. (2.36b). If the signal frequency increases by !, the idler frequency decreases
by ! and the wave-vector mismatches along the two directions parallel and
perpendicular to the signal wavevector can be approximated, to the first order, as:
@k1 @k2 @
kpar Š !C cos ! k2 sin ! (2.37a)
@!1 @!2 @!2
@k2 @
kperp Š sin ! C k2 cos ! (2.37b)
@!2 @!2
To achieve broadband phase matching, both kpar and kperp must vanish. Upon
multiplying (2.37a) by cos and (2.37b) by sin and adding the results, we get
@k2 @k1
cos D0 (2.38)
@!2 @!1
36 G. Cerullo and D. Brida
Fig. 2.5 Scheme of a BBO based NOPA pumped at 400 nm. BS beam splitter, SPF shortpass filter
crystal, cut for type I phase matching (™ D 32ı ), using a single-pass configuration to
increase the gain bandwidth. The chosen crystal length is close to the pulse-splitting
length Lsp13 for signal and pump in the visible wavelength range. To minimize the
effects of self-focusing, the BBO crystal is beyond the focus of the pump beam. The
WLC seed is imaged by a spherical mirror in the BBO crystal, with a 100- m spot
size matching that of the pump beam. A thin short-pass filter removes the strong
residual 800-nm component from the WLC, preventing its parasitic amplification.
When the BBO crystal is illuminated by the pump pulse and aligned perpendic-
ularly to the pump beam, it emits a strong off-axis parametric superfluorescence in
the visible in the form of a cone with apex angle of 6.2ı (corresponding to an
angle of 3.82ı inside the crystal); this is the direction for which the group velocities
of signal and idler are matched and therefore the gain bandwidth is maximized.
The visible cone gives a visual aid to the identification of the optimum condition
for broadband generation, which is found when the pump-signal angle matches the
cone apex angle. In this condition, for optimum pump–seed delay, an ultrabroad
gain bandwidth that extends over most of the visible (500–750 nm) is observed.
The amplified pulses from a single stage have energy of approximately 2 J; much
higher energies, up to 300 J [20], can be extracted by a second amplification
stage. After the gain stage the amplified pulses are collimated by a spherical mirror
and sent to the compressor.
Several compressor schemes have been implemented for the visible NOPA.
Simple prism pairs can correct the second but not third-order dispersion and thus
can only compress the pulses down to 10–15 fs; sub-10-fs pulses can be achieved
by using either prism-grating [17] or prism-chirped mirrors combinations, as well as
adaptive compressors based on deformable mirrors [20–22]. It is also possible to use
exclusively chirped mirrors [19, 23], greatly simplifying the system design, allowing
for compactness, insensitivity to misalignment and high day-to-day reproducibility,
which are of great importance in practical applications.
where A(t) is the pulse envelope with its maximum at t D 0, ! c is the carrier
frequency and is the CEP. If D 0, a maximum of the electric field corresponds
to the peak of the pulse envelope (cosine pulse), while if D /2 the electric
field at the peak of the pulse envelope is zero (sine pulse). Control of the CEP
38 G. Cerullo and D. Brida
Fig. 2.6 Frequency and phase summation rules for (a) sum-frequency generation (SFG), (b)
difference-frequency generation (DFG), equivalent to the generation of the idler wave in an OPA,
and (c) a four-wave mixing process (FWM)
becomes important for few-optical-cycle light pulses and for extreme nonlinear
optics experiments, which are sensitive to the electric field rather than the intensity
of the pulse.
Typical femtosecond laser sources generate a pulse train in which the CEP is not
stabilized, so that varies randomly from shot to shot. CEP can be stabilized using
either active or passive methods. Active control starts by measuring the pulse-to-
pulse CEP slip in a mode-locked oscillator and stabilizing it by active electronic
feedback [24]; by picking pulses at a fraction of the laser repetition rate, it is possible
to obtain a train of pulses with a constant absolute phase, which can be further
amplified. A completely different approach is passive CEP stabilization, which
exploits nonlinear optical effects to generate a pulse train with stable CEP [25].
To understand passive CEP stabilization, let us review the effect of different non-
linear processes on the CEPs of the interacting pulses. For the second-order process
of Sum-Frequency Generation (SFG), in which two pulses at frequencies ! 1 and ! 2 ,
with CEPs 1 and 2 , generate the sum frequency ! 3 D ! 1 C ! 2 , the CEP of the
resulting pulse is 3 D 1 C 2 /2 (Fig. 2.6a). For the particular case of second
harmonic generation (SHG) starting from a pulse with frequency ! 0 and CEP , one
obtains a pulse with frequency ! SH D 2! 0 and CEP SH D 2 /2. In the second-
order process of Difference-Frequency Generation (DFG) between two pulses at
frequencies ! 1 and ! 3 , with CEPs 1 and 3 , the resulting difference-frequency
pulse has frequency ! 1 D ! 3 ! 2 and CEP 1 D 3 2 /2 (Fig. 2.6b). In
the third order non-linear process of Four-Wave Mixing (FWM), four waves at
frequencies ! 1 , ! 2 , ! 3 and ! 4 , with ! 4 D ! 1 ! 2 C ! 3 , are coupled in a (3)
nonlinear medium. In this case, the CEPs of the waves are linked by the relationship:
4 D 1 2 C 3 /2 (Fig. 2.6c). Self-Phase Modulation (SPM), i.e. spectral
broadening due to the instantaneous intensity-dependent variations of index of
refraction in a Kerr medium, can be regarded as a special case of FWM, in
which a new frequency ! SPM is generated starting from three frequencies ! 1 ,
! 2 and ! 3 , all belonging to the pulse spectrum, according to the FWM process:
2 Ultra-Broadband Optical Parametric Amplifiers 39
In a broadband OPA both the pump and the seed are femtosecond pulses, typically
generated by a Ti:sapphire system. It is generally difficult and expensive to scale
the energy of femtosecond pump lasers; on the other hand, it is easier to generate
energetic picosecond pulses (5–50 ps duration) exploiting well-established gain
media such as Nd: or Yb:doped crystals. With a long pump pulse, in order to achieve
efficient energy extraction, it is necessary to temporally overlap the seed pulse
with the pump by first stretching the seed pulse and then, after the amplification
40 G. Cerullo and D. Brida
Fig. 2.7 Schematics of three OPA configurations. For clarity, the seed and the pump beams are
shown to intersect non-collinearly in the crystal. However, the conclusions are also valid for the
collinear beam arrangement
3. in an OPCPA, amplification occurs only during the pump pulse, so that amplified
spontaneous emission and the consequent pre-pulse pedestal are greatly reduced.
4. since the OPA/OPCPA is an instantaneous process of energy exchange between
the interacting beams, no fraction of the pump photon energy is deposited in the
medium; thus, if parasitic pump absorption can be neglected, thermal loading
effects are absent, greatly reducing spatial aberrations of the beams. This allows
much higher repetition rates to be achieved while avoiding the heat removal
problems that limit the frequency scaling of CPA systems.
In addition to these clear advantages, the OPA/OPCPA concept also has some
drawbacks:
1. an OPA/OPCPA requires pump pulses with short duration (hundreds of femtosec-
onds to tens of picoseconds), which are technologically challenging to generate,
especially at high energy and/or high repetition rate; on the other hand the CPA,
thanks to energy storage in the excited state, can use much longer nanosecond
pump pulses (or even continuous wave pumping for materials with long excited
state lifetime like Yb-doped crystals);
2. an OPA/OPCPA is very sensitive to the spatial and temporal quality of the pump,
again putting a severe technological constraint on the pump laser.
2.9 Conclusions
This chapter has provided a brief review of the principles of ultrashort pulse OPAs,
focusing in particular on the generation of few-optical-cycle pulses. After discussing
the propagation of an ultrashort pulse in a nonlinear medium and reviewing the
theory of parametric amplification, we considered the typical architecture of an
ultrafast OPA. This chapter went on to discuss non-collinear OPAs, which are
widely used for the generation of few-optical-cycle pulses in the visible range and
capable of producing pulses with energies of 2 J, which can be further amplified to
300 J. These schemes can be extended for the synthesis of pulses with controlled
carrier-envelope phase. Indeed, the last section of this chapter introduces the concept
of optical parametric chirped pulse amplification, which is considered the most
promising approach for energy scaling of few-optical-cycle pulses.
References
1. A.H. Zewail, Femtochemistry: atomic-scale dynamics of the chemical bond. J. Phys. Chem. A
104, 5660–5694 (2000)
2. C. Manzoni, D. Polli, G. Cerullo, Two-color pump-probe system broadly tunable over the
visible and the near infrared with sub-30 fs temporal resolution. Rev. Sci. Instrum. 77,
023103–023112 (2006)
3. F. Krausz, M. Ivanov, Attosecond physics. Rev. Mod. Phys. 81, 163–234 (2009)
42 G. Cerullo and D. Brida
4. F.X. Kärtner, Few-Cycle Laser Pulse Generation and Its Applications (Springer, Berlin, 2004)
5. J.A. Giordamaine, R.C. Miller, Tunable coherent parametric oscillation in LiNbO3 at optical
frequencies. Phys. Rev. Lett. 14, 973–976 (1965)
6. R.A. Baumgartner, R.L. Byer, Optical parametric amplification. IEEE J. Quantum Electron.
15, 432–444 (1979)
7. Y.R. Shen, The Principles of Nonlinear Optics (Wiley, New York, 1984)
8. R. Boyd, Nonlinear Optics (Academic, New York, 1992)
9. G. Cerullo, S. De Silvestri, Ultrafast optical parametric amplifiers. Rev. Sci. Instrum. 74, 1–18
(2003)
10. D. Brida, G. Cirmi, C. Manzoni, S. Bonora, P. Villoresi, S. De Silvestri, G. Cerullo, Sub-two-
cycle light pulses at 1.6 m from an optical parametric amplifier. Opt. Lett. 33, 741–743 (2008)
11. G.M. Gale, M. Cavallari, T.J. Driscoll, F. Hache, Sub-20-fs tunable pulses in the visible from
an 82-MHz optical parametric oscillator. Opt. Lett. 20, 1562–1564 (1995)
12. E. Riedle, M. Beutter, S. Lochbrunner, J. Pielm, S. Schenkl, S. Spörlein, W. Zinth, Generation
of 10 to 50 fs pulses tunable through all of the visible and the NIR. Appl. Phys. B 71, 457–465
(2000)
13. T. Wilhelm, J. Piel, E. Riedle, Sub-20-fs pulses tunable across the visible from a blue-pumped
single-pass noncollinear parametric converter. Opt. Lett. 22, 1494–1497 (1997)
14. G. Cerullo, M. Nisoli, S. De Silvestri, Generation of 11 fs pulses tunable across the visible by
optical parametric amplification. Appl. Phys. Lett. 71, 3616–3618 (1997)
15. A. Shirakawa, T. Kobayashi, Noncollinearly phase-matched femtosecond optical parametric
amplification with a 2000 cm1 bandwidth. Appl. Phys. Lett. 72, 147–149 (1998)
16. G. Cerullo, M. Nisoli, S. Stagira, S. De Silvestri, Sub-8-fs pulses from an ultrabroadband
optical parametric amplifier in the visible. Opt. Lett. 23, 1283–1285 (1998)
17. A. Shirakawa, I. Sakane, T. Kobayashi, Pulse-front-matched optical parametric amplification
for sub-10-fs pulse generation tunable in the visible and near infrared. Opt. Lett. 23, 1292–1294
(1998)
18. A. Shirakawa, I. Sakane, M. Takasaka, T. Kobayashi, Sub-5-fs visible pulse generation
by pulse-front-matched noncollinear optical parametric amplification. Appl. Phys. Lett. 74,
2668–2670 (1999)
19. M. Zavelani-Rossi, G. Cerullo, S. De Silvestri, L. Gallmann, N. Matuschek, G. Steinmeyer, U.
Keller, G. Angelow, V. Scheuer, T. Tschudi, Pulse compression over a 170-THz bandwidth in
the visible by use of only chirped mirrors. Opt. Lett. 26, 1155–1157 (2001)
20. P. Tzankov, J. Zheng, M. Mero, D. Polli, C. Manzoni, G. Cerullo, 300 J noncollinear optical
parametric amplifier in the visible at 1 kHz repetition rate. Opt. Lett. 31, 3629–3631 (2006)
21. M.R. Armstrong, P. Plachta, E.A. Ponomarev, R.J.D. Miller, Versatile 7-fs optical parametric
pulse generation and compression by use of adaptive optics. Opt. Lett. 26, 1152–1154 (2001)
22. A. Baltuška, T. Fuji, T. Kobayashi, Visible pulse compression to 4 fs by optical parametric
amplification and programmable dispersion control. Opt. Lett. 27, 306–308 (2002)
23. P. Baum, M. Breuer, E. Riedle, G. Steinmeyer, Brewster-angled chirped mirrors for broadband
pulse compression without dispersion oscillations. Opt. Lett. 31, 2220–2222 (2006)
24. D.J. Jones, S.A. Diddams, J.K. Ranka, A. Stentz, R.S. Windeler, J.L. Hall, S.T. Cundiff,
Carrier-envelope phase control of femtosecond mode-locked lasers and direct optical frequency
synthesis. Science 288, 635–639 (2000)
25. A. Baltuška, T. Fuji, T. Kobayashi, Controlling the carrier-envelope phase of ultrashort light
pulses with optical parametric amplifiers. Phys. Rev. Lett. 88, 133901–133905 (2002)
26. A. Baltuška, T. Fuji, T. Kobayashi, Self-referencing of the carrier-envelope slip in a 6-fs visible
parametric amplifier. Opt. Lett. 27, 1241–1243 (2002)
27. S. Backus, C.G. Durfee, M.M. Murnane, H.C. Kapteyn, High power ultrafast lasers. Rev. Sci.
Instrum. 69, 1207–1223 (1998)
28. A. Dubietis, G. Jonušauskas, A. Piskarskas, Powerful femtosecond pulse generation by chirped
and stretched pulse parametric amplification in BBO crystal. Opt. Commun. 88, 437–440
(1992)
2 Ultra-Broadband Optical Parametric Amplifiers 43
29. A. Dubietis, R. Butkus, A.P. Piskarskas, Trends in chirped pulse optical parametric
amplification. IEEE J. Sel. Top. Quantum Electron. 12, 163–172 (2006)
30. D. Herrmann, L. Veisz, R. Tautz, F. Tavella, K. Schmid, V. Pervak, F. Krausz, Generation of
sub-three-cycle, 16 TW light pulses by using noncollinear optical parametric chirped-pulse
amplification. Opt. Lett. 34, 2459–2461 (2009)
Chapter 3
Attosecond Generation and High Field Physics
area of great activity over the last two decades (note the award of the 1999 Nobel
prize in Chemistry to Ahmed Zewail for his development of that subject) covering
investigation from diatomic molecules to large biomolecules. The principal tool
for femtosecond chemistry is the mode-locked laser that typically provides pulses
of duration as short as 20 fs which through various non-linear optical processes
can routinely span the range of wavelengths from 200 nm to 5 m. These pulses
comprise multiple cycles of the carrier frequency of the optical domain laser field.
Typically, a pump-probe methodology is adopted where a pump interaction is used
to “start” the event and then it is probed after a short delay t. This allows us to
capture the state of the system at a time t after the event starts. To build up the full
dynamics it is necessary to repeat the measurement over the desired range of t.
One reason for needing measurements with just a few femtosecond resolution
is that some chemical events are very fast (such as proton transfer) and if we
wish to capture them with sufficient resolution for a full understanding, pulses
significantly shorter than the event are needed. This certainly justifies the progress
in generating optical domain pulses to the few femtosecond (few optical cycle)
limit. There is, however, a more profound reason for seeking a sub-femtosecond
temporal resolution and that is to target electron dynamics within matter. The recent
development of new tools enables these measurements and a far more sophisticated
understanding of the fast dynamics in matter after electronic excitation can be
obtained. Electrons move very quickly within matter, often on timescales measured
in 10’s of attoseconds. Moreover, in almost all matter the electrons are bound in
very close proximity to other electrons and so their dynamics are highly correlated
through both the Coulomb interaction and quantum mechanical exchange effects.
Although this may not be apparent in our everyday perception of matter, it is of cru-
cial importance if we are to understand or control the microscopic states of matter,
goals which are at the core of many emerging twenty-first century technologies [1].
Through attosecond science we are developing the tools required to measure these
ultra-fast correlated electron processes (for a detailed review see [2]).
For instance in the last 5 years advanced quantum chemistry calculations by the
groups of Cederbaum in Heidelberg [3] and Levine in USA/Israel [4] have revealed
that sudden photo-excitation in polyatomic molecules leads to large amplitude
charge migration (rapid motion of the hole through the molecular structure) driven
by the electron correlations (See Fig. 3.1). This occurs on timescales of just a few
femtoseconds down to less than 100 attoseconds, and crucially before any nuclear
motion has begun. The universality of this charge migration phenomenon is just
beginning to be recognised and the coupling to subsequent physical and chemical
changes due to the nuclear motion that it drives is an open question. There is,
therefore, a target to understand such processes in gas and condensed phase systems
and this motivates further advances in attosecond sources. Other areas that are the
target of attosecond science include: ultrafast (few fs) nanoplasmonics, ultrafast
x-ray spectroscopy and chemical processes at surfaces (see Chap. 7 by Mathias
et al.). A summary of current approaches to making ultrafast structural dynamics
methods can be found in Altucci et al. [6].
3 Attosecond Generation and High Field Physics 47
Fig. 3.1 Charge migration as an example for electron dynamics in molecular systems. The figure
shows a numerical simulation of the 3D-charge density evolution after initial ionisation of the
molecule 2-phenylethyl-N,N-dimethylamine (PENNA) at three different time steps. The hole
density is shaded in light grey, whereas the electron density is depicted in dark grey. The left
panel shows the evolution for the ground-state geometry of the molecule, whereas the right panel
shows the process for the C–C bond between the benzene ring and the N-atom elongated by 20 pm
[5] (Reproduced with permission from Ref. [5])
A high order non-linearity permits access to timescales that are only a fraction of
the optical cycle time of the laser field. We shall see later there are other potential
methods for creating sub-femtosecond pulses: broadband Raman generation, pulse
shaping of free electron laser emission and relativistic high order harmonic genera-
tion. Nevertheless, all of these also rely in one way or the other on the interaction of
free or bound electrons with few cycle phase stabilized high intensity optical pulses.
Therefore our next task is to gain an understanding of the salient physics of the
interaction of a strong field with matter.
Consider the motion of a free classical electron interacting with a laser electric field
of peak amplitude E0 and angular frequency !. Assuming a linearly polarised field
in the x-direction of the cosinusoidal form (i.e. ignoring the spatial and suppressing
the polarisation properties):
The typical wavelength of optical light (2 107 to 5 106 m) is much larger than
the size of atoms, or even the excursions electrons make in a moderately intense
field, which justifies the suppression of the spatial dependence. We confine our
discussion to lower intensities (<<1017 Wcm2 ) so no relativistic effects arise. The
field will of course exert an oscillatory force on the electron (charge e and mass m)
and thus a time-dependent acceleration of magnitude:
d2 x e
2
D E0 cos.!t/ (3.2)
dt m
The equation of motion can be solved to yield the time dependent oscillatory
velocity and displacement of the electron, taking into account any initial conditions
for velocity or position. The amplitude of the electron oscillation is given by:
eE0
x0 D (3.3)
m! 2
We can see that the displacement depends linearly upon the electric field and
inversely to the frequency squared (i.e. quadratically with the wavelength). For
example, in a field of wavelength 1 m at an intensity of 1014 Wcm2 the oscillation
amplitude is 1 nm (109 m), which is 10 atomic radii. This illustrates that in a
strong field the excursion of the field driven electron can cause it to be far removed
from the influence of the atomic core (which is the basis of the strong field limit
discussed below).
3 Attosecond Generation and High Field Physics 49
The cycle averaged kinetic energy of the electron is called the ponderomotive
energy Up and is given by:
e 2 E02
Up D (3.4)
4m! 2
In general we are most interested in what happens when the strong field interacts not
with an initially free electron but with an electron bound in an atom or molecule. The
interaction of a strong laser field with matter is in general a very complex problem
(see Cohen Tanoudji [7]). As a simplification we will assume that the field is non-
resonant (far detuned from any single photon transition frequency in the system),
low frequency (photon energy much less than the ionisation potential) and that the
field is very strong so that (a) it can be regarded as a classical field and (b) the
binding potential can start to be considered as weak with respect to the interaction
with the interaction with the laser field.
It is important to note that within the multiphoton picture a strong field can give
rise to very high order processes that include absorption of multiple photons into
excited states. Typically multiphoton ionisation starts to appear at intensities in the
range from 1011 to 1013 Wcm2 (depending on the atomic ionisation potential and
possible existence of multiphoton resonances). Transitions between laser dressed
states of the continuum can become probable in a strong field giving rise to high
orders of photoelectron absorption (called above threshold ionisation, ATI). These
excited states can radiate back to the ground state if parity is satisfied (odd number of
photons absorbed) giving rise to the phenomenon termed high harmonic generation
(HHG).
Unfortunately the usual quantum theoretical approaches are either very difficult
to implement accurately (such as high order perturbation theory) or don’t provide
much physical insight into the process (numerical solution of the time-dependent
Schrödinger equation) at these high field strengths. Fortunately the strong field in
the problem gives rise to a great conceptual simplification that has an accompanying
powerful theoretical and simple physical insight for the calculation of HHG: the
Strong Field Limit first developed by Corkum [8] and Schafer et al. [9]. The so
called “simple man’s” or “3 step” model for HHG and other strong field processes
50 J.P. Marangos and M. Oppermann
Fig. 3.2 Illustration of the three step model, with the driving laser electric field E(t) displayed in
the lower part of the figure. Directly above, the two ground state electrons in the atom’s Coulomb
potential are displayed, where Ip is illustrated as the energy required to transfer one of them to the
continuum. Their time-evolution in the laser field is described by the three step model, where in
the last step the electron recombines with the parent ion, releasing a photon of energy -h!
is as follows (see Fig. 3.2). In this limit we can view the effect of the strong field as
a strong periodic distortion of the binding potential due to the interaction with the
combined Coulomb and laser fields.
Ze 2 1
V .x/ D C .e/E.t/x (3.5)
4"0 r
Where r is the radial distance from the nucleus and x the coordinate in the laser field
direction, Z the ion charge (Z D 1 for the singly ionised atom).
The distortion can be so great that the resulting potential barrier is instan-
taneously suppressed below the bound state energy – leading to the classical
freeing of the electron. Yet it is more likely that the potential barrier is lowered
but still higher than the bound state energy. In this case there is a probability
for quantum mechanical tunnelling to occur. If tunnelling has a high probability
of occurring within an optical cycle, then tunnel ionisation will dominate over
multiphoton ionisation. The ionisation rate then becomes strongly time dependent,
with significant magnitude only around the peaks of the laser field jE.t/j [10]. We
now introduce the Keldysh parameter [11]. The case where < 1 corresponds to the
strong field limit, where tunnelling ionisation starts to dominate:
s
Ip
D (3.6)
2Up
The Keldysh parameter decreases as the square root of the laser intensity and as
the inverse of the laser wavelength. It is therefore much easier to reach the tunnelling
regime with near IR lasers than visible-UV, the fact that titanium sapphire lasers at
3 Attosecond Generation and High Field Physics 51
The ionisation rate also depends upon the “shape” of the atomic or molecular
wavefunction in the region of the barrier, this can be critically important in
determining the ionisation dynamics of a specific state. The formula for tunnelling
derived for atoms by Ammosov, Delone and Krainov (ADK) theory [12–15] is a
useful guide that captures these features:
.2l C 1/ .l C jmj/Š 3E.t/ 1=2 2F0 2n jmj1 2F0
w.t/ D Cn2 Ip exp
2jmj jmjŠ .l jmj/Š F0 E.t/ 3E.t/
(3.8)
The above ADK tunnelling rate w(t) is given for an initial state characterized by its
effective principal quantum number n* D Z(2 F0 )1/2 , its principal quantum number
n, orbital angular momentum l and magnetic quantum number m. Furthermore Z is
the ionic charge state, Ip the ionisation potential in atomic units (au), F0 D (2Ip )3/2
the intra-atomic field strength (au) and E(t) the linearly polarised time dependent
electric field amplitude (au). Progress is being made to extending durable theoretical
treatments to the ionisation of molecular systems [16, 17].
If the electron is released after the field peak, the classical trajectory it takes can
lead to it returning to the ion from which it was ionised within an optical cycle.
Of course the electron is not a simple classical particle, instead it is most correctly
described as being in a wavepacket of many continuum states consistent with the
initial localisation at the moment after tunnelling. The initial wavepacket is related
to the boundstate wavefunction but with a transverse Gaussian filter applied from
the tunnelling process [18]. The wavepacket state inevitably undergoes quantum
diffusion (wavepacket spreading) due to the non-zero momentum components that
comprise it. For the fullest description, the continuum electron states must be
regarded as in a laser dressed state continuum state (Volkov state). An important
property is that the continuum wave-packet position expectation value hri is
oscillatory as we can anticipate for the classical free electron case just discussed.
The tunnelling process will result in an amplitude for ionisation and an amplitude
for the electron to remain bound, so the wavefunction of the electron should be
viewed as a superposition of these two components. The consequence of this is that
there is the possibility for the continuum part of the electron wavefunction to return
52 J.P. Marangos and M. Oppermann
and re-interact with ion at a slightly later time (determined by the optical cycle)
and this has big role in control of light emission and other strong field processes.
The recolliding continuum wavepacket interferes with the remaining bound state,
resulting in high frequency oscillations in the electron’s dipole moment, radiation
from which leads to high-order harmonic generation (HHG) (corresponding to the
recombination process).
The nature of the HHG process in a strong laser field, and the importance of
this process to attosecond measurement, is encapsulated in the “three-step” picture
described in the previous section. In the first step the laser field ionises the atom
or molecule through field ionisation (predominately quantum tunnelling through
the field suppressed potential barrier) which occurs with high probability when the
electric field amplitude is near a peak. The exponential dependence of the tunnelling
rate ensures that the ionisation is confined in time to ˙200 attosecond (for an
800 nm laser field) around the cycle peak. In the next step, the liberated electron
wavepacket moves in response to the laser field: first being accelerated away
from the parent ion, and then returning at some later time on the order of half
an optical cycle period (typically 0.5–1.5 fs for a laser field at 800 nm) as the
laser field reverses direction. During this process the electron wavepacket gains
energy from the laser field. The third step is a recombination process in which
the energy gained is emitted as a high energy photon with the electron returning
to the ground state of the parent system. The spectrum extends up to a cut-off
energy set by the maximum possible electron return energy which from classical
considerations is 3.17Up C Ip . The recombination is inherently confined in time to a
period of a few hundred attoseconds, which will be discussed later. The discovery,
elucidation and application of this process to attosecond measurement are very
important developments that have paved the way to attosecond science.
The theoretical description of this process is captured in essence by the so called
Strong Field Approximation (SFA) theory of Lewenstein et al. [19]. A development
from SFA casts the strong field process in terms of all the possible quantum paths
that an electron may follow. Only a few of these turn out to make a significant
contribution to the HHG process, those that satisfy the correct conditions for the
momentum and position at the time of birth (zero initial momentum, initially
located at the exit of the quantum tunnel) and return (returning to atomic centre).
Mathematically these “saddle point” conditions greatly reduce the computation of
all the quantum path integrals. Moreover the concept of quantum paths essentially
allows us to make direct contact with the classical trajectories and in fact the
classical motion of electrons in the combined Coulomb and laser field thus gives
a powerful insight into the process.
The electron interacts with the ionic core if it approaches close enough on its
return to the parent ion. A number of processes can occur during this re-encounter
3 Attosecond Generation and High Field Physics 53
(see Fig. 3.3). These processes include inelastic excitation of the ion during
the collision that may result in excitation or further ionisation. Elastic collision
will also occur and has recently proved to contain useful information on the
molecular structure encoded through the diffraction effect from multiple centres in
the molecule [18], a process termed laser induced electron diffraction. The electron
may also recombine back into the state it left resulting in the molecule returning to
the initial state of energy and, if the resulting harmonic photon is forward scattered
in the direction of the laser field, the momentum is also the same as before. In
this case the molecule is equivalent to all others in the sample. Provided there is a
degree of macroscopic phase matching preserved the emitted harmonics can be a
strong signal, albeit with an overall energy conversion efficiency typically <<106 .
This is the reason for the coherent addition of the electric field signals that is at the
heart of all optical parametric processes and results in a N2 scaling of the intensity of
emitted radiation (N number of atoms/molecules in the interaction region), and the
fact that the emission is confined into a narrow low divergence beam in the forward
scattering direction. We will describe this process in more detail below, particularly
how it can lead to the generation of isolated attosecond duration soft X-ray pulses
that can be used in subsequent measurements.
The quantum orbit model uses the saddle point method to select a finite number of
orbits which can be summed up to approximate the entire integral. This approach
allows individual or groups of trajectories to be studied in isolation. Thus we can
continue to use our purely classical picture of the electron trajectories to gain deep
54 J.P. Marangos and M. Oppermann
insight into the process. To account for quantum contributions to the trajectories we
must also account for the ionisation process and the action S(t,t0 ) accumulated by
the electron during the trajectory which determines the phase:
0ˇ ˇ2 1
ˇ* ˇ
ˇ ˇ
*
0
Z t B ˇp.t; t / C A./ˇ C
B C
S.t; t 0 / D d B C Ip C (3.9)
t0 @ 2 A
The correct summing of these trajectories must include the phase which clearly
depends upon the time spent in the continuum for the trajectory D t t0 . The birth
time t0 also determines the momentum gained by the electron upon its return.
After tunnelling, the electron position x(t) and velocity v(t) are determined by the
field according to the equations [8]:
eE0
x.t/ D Œ cos.!t/ C v0x t C x0x (3.10)
m! 2
eE0
vx .t/ D sin.!t/ C v0x (3.11)
m!
Here, the x-axis corresponds to the polarisation direction of the field and x0x and v0x
are the initial conditions at the time of tunnelling, which are set to 0 in the simplest
approximation.
The solution to these equations shows that half the electrons (those born after the
peak in each half cycle) will return to the origin (x D 0) at least once. Electrons born
before a cycle peak will never return and instead will undergo displaced oscillatory
motion in the field that drifts away from the atom. Further, these equations show that
it is the electrons born at 0.047 of an optical cycle after the peak of the field (i.e. at a
phase of 170 ) that return at the maximum energy of 3.17Up. We shall refer to these
as the maximum return energy electrons. These electrons return after D 0.6 T of
an optical cycle (that is 1.6 fs for an 800 nm field).
Electrons born either immediately before or after the maximum return energy
electrons also return to the atom, but with reduced energy and a different return
time . Those born after the maximum return energy electrons return sooner than
the maximum return energy electrons and so are referred to as having “short trajec-
tories”. Short trajectory electrons have a definite dependence of increasing return
energy with increasing return time, they show a “positively chirped” behaviour (see
Fig. 3.4). Electrons born before the maximum return energy electrons will return
after longer times and have a return energy that decreases with return time (they are
negatively chirped) and this group are referred to as “long trajectories”.
To quantify the relationship between the return energy Er and return time for
the short trajectories, Lein [20] has made a good fit to the exact classical result for
driving field of angular frequency ¨ and ponderomotive energy Up using the form:
3 Attosecond Generation and High Field Physics 55
Fig. 3.4 Time dependent properties of the trajectories describing the free electron motion in a
strong laser field. The top panel shows the excursion distance x(t) as a function of time within an
optical laser cycle for electrons born at different times after the peak of the field. The trajectory with
maximum return energy is highlighted. The bottom panel shows the return energy as a function of
the electron’s return time and thereby illustrates which photon energies and hence frequencies are
emitted at which point in time. This contains information about the chirp of the harmonic emission
1:207 0:492
Er Er
! D 0:786 f C 3:304 f (3.12)
Up Up
Where the function f (x) is defined as f(x) D arccos(1 x/1.5866)/. How close
these results are to the real situation where quantum effects and the full molecular
potential are also present has been the subject of recent research that has so far
56 J.P. Marangos and M. Oppermann
concluded that these approximations are fairly reliable for plateau harmonics up to
the cut-off (up to a shift of less than 100 attoseconds), although the results may be
less reliable for harmonics close to threshold.
In the discussion so far we have not considered phase matching of the harmonic
emission from the macroscopic sample. In general this is a complicated subject
as several contributions to the wavevector mismatch k must be considered that
originate from geometrical (Guoy) phase as the fields propagate through the focus,
material dispersion, intensity dependent dipole phase (related to the action integral)
and dependent upon the local intensity I(r,t), dispersion of free electrons (also time-
space dependent). Accurate computation of phase matching must take into account
the real temporal and spatial distribution of the laser field and typically one-
dimensional calculations are not sufficient for anything other than indicative results.
Nevertheless, by reasonable choices of gas density, gas path length (typically kept
small relative to the laser Rayleigh range), laser focusing and position of the sample
a robust trajectory selection can be achieved e.g. so only short trajectory harmonics
are observed and the emitted spectral amplitude can be a faithful reflection of the
single molecule emission.
At first glance it might appear that the existence of these ranges of return times
and return energies will simply “smear out” the HHG emission so that it is not useful
for attosecond pulse production. In fact this is not so if isolation of short (or long)
trajectories is achieved via phase matching in the sample. Details can be found in
the literature [21], but in essence the action leads to different accumulated phases in
short and long trajectories that can be balanced against the geometrical phase. Short
trajectories are strongly favoured if the gas sample is confined to a location around
a Rayleigh range downstream of the laser focus. Further the mapping of return
energy to time have been experimentally confirmed [22] by the use of the RABBITT
technique that characterises the relative phases between neighbouring harmonics
and so allows the reconstruction of the time history of emission of different
harmonic orders. This work confirms that the harmonics emitted from each half-
cycle of the field will have a chirp on the attosecond timescales. Various methods to
correct for this are being actively considered, including carefully prepared optical
fields to compress the recollision times of the highest energy returning electrons
and use of soft-X-ray optical filters with the appropriate dispersion to correct for the
chirp in the field.
We now turn to the experimental state-of-the-art for HHG based attosecond sources.
From the description of the temporal dependence of HHG emission it is clear that
each half-cycle of an intense optical field will give rise to ionisation events that
lead to electron trajectories that can return and result in emission of HHG radiation.
This radiation will be emitted in a burst with a duration spanning several hundred
attoseconds. For a multi-cycle laser pulse where the optical field strength varies little
3 Attosecond Generation and High Field Physics 57
over several cycles there will therefore be two attosecond bursts per cycle. These
are for electrons colliding from opposites sides each cycle and the interference
between these events leads to the odd order harmonic emission within this picture.
The resulting attosecond pulse train (APT), with oppositely phased fields emitted
every half-cycle, will be radiated in this multi-cycle limit.
It is clear that an attosecond pulse train will accompany all HHG generation
although it was only in the last decade or so that the sub-cycle temporal structure
was fully characterized and explained. Accessing this temporal structure is not
trivial and requires considerable precautions to deliver the distinct pulses in the
train to an experiment and utilise them. Work in this direction followed the first
characterization of an APT by the RABBITT technique and was pursued by
groups in France and in particular in Sweden (Lund) [23]. The approach being
pursued in Lund is to reconstruct the photoelectron wavepacket generated by the
attosecond pulses in the train, which is done by utilising the IR field (synchronised
to the APT but with a controllable relative delay) to shift the momentum of the
electron wavepacket by an amount that depends upon the IR field phase at the
time of photoionisation by the attosecond XUV pulse (see below in the description
of streaking) [24]. The momentum distribution of the generated electron wave
packets is recorded using a velocity map imaging detector. Consecutive electron
wavepackets will thus overlap and interfere at the detector and from the interference
the spectral phase of the wavepacket can be reconstructed. It is hoped that the
photoelectron wavepacket can then be related back with sufficient fidelity to the
original electronic ground state dynamics.
The challenge of producing isolated attosecond pulses has been met by two
successful approaches so far. The first, and most direct, of these is to generate
HHG from very short pulses comprising only a few optical cycles. In such a short
pulse, the variation in field strength from one half cycle to the next is substantial and
only a single cycle generates the highest energy harmonics which will be emitted
only once in the laser pulse. These harmonic orders, near the cut-off, are isolated
by spectral filtering of lower photon energies which forms an isolated pulse. Our
discussion below will concentrate upon this approach. The other approach, suitable
for somewhat longer pulses, applies an idea first proposed some years ago [25]
where the laser polarisation is controlled so that it varies from circular through
linear and back to circular through the pulse, with pure linear polarisation only
achieved for a period less than a single cycle. This technique relies upon the fact that
the efficiency of high-harmonic generation decreases rapidly with increasing laser
field ellipticity due to a suppression of electron returns in the recollision model.
Due to this strong polarisation dependence of the HHG efficiency this “polarisation
gating” method leads to isolated attosecond pulses. By using two counter-rotating
circularly polarised pulses, with an appropriate short mutual time delay there is a
small time window (or gate) when the combined field becomes linear. It can be
contrived for this gate to be smaller than a half-optical cycle, then there will only be
one burst of XUV emission during the gate resulting in the production of an isolated
attosecond XUV pulse. Pulses as short as 130 attosecond have been reported by this
method [26].
58 J.P. Marangos and M. Oppermann
The approach that has resulted in the shortest isolated pulses so far reported is,
however, the direct approach requiring a laser pulse comprising just a few optical
cycles. This is described by Goulielmakis et al. [27]. It requires several key technical
features to be fully controlled; (a) the pulse must be very short with a FWHM of
fewer than 2 optical cycles, (b) the CEP must be tightly locked so that within this few
cycle pulse only a single dominant recollision accounts for the cut-off harmonics,
(c) the HHG spectrum must be efficiently spectrally filtered so that the cut-off can
be isolated. In general (a) requires the use of either focussing the laser through a
gas filled hollow core fibre to ensure self-phase modulation (SPM) causes sufficient
broadening to support a short pulse [28] or a similar broadening mechanism arising
through self-guided filamentation in a gas filled cell. Typically a factor of 4 increase
in spectral width can be achieved with around 50 % transmission. The next step is to
compensate all orders of phase to produce the very short, high power, pulse which
is done through using high order phase compensating stacks of chirped mirrors
and additional thin glass matched wedges for fine adjustment. The technology of
CEP stabilisation of chirp pulsed amplified titanium sapphire lasers [29] is now
well established and such lasers are commercially available, although it is often
challenging to achieve the degree and duration of phase stability required for
the purposes of attosecond pulse production. A range of transmission filters and
reflective dielectric mirror coatings are now available for selection of a number
of XUV spectral ranges, for example, Al (bandpass 20–75 eV), Zr (bandpass
70–120 eV), MoSi (reflective 85–95 eV). Figure 3.5 presents the results of a
numerical simulation under the conditions of the experiment by Goulielmakis et al.
[27, 30], leading to the required XUV emission for the creation of an isolated
attosecond pulse.
Future developments are promising to extend the photon energy range to both
deeper into the X-ray [31] and to lower VUV photon energies [32] which, providing
the appropriate VUV – soft X-ray optics are available, will allow the techniques
of attosecond measurement to be extended to a wider range of systems. Increased
photon flux is also being achieved, for instance by limiting the HHG generation to
a single optical cycle so reducing the effect of ionisation saturation.
Fig. 3.5 Numerical simulation illustrating the creation of isolated attosecond pulses from a sub-
1.5 cycle, linearly polarised, 720 nm driving laser pulse in Neon atoms. The top panel shows the
maximum energy electron trajectories generated near the peaks of the laser field EL (t). The number
next to those trajectories denotes their ionisation probability. The bottom panel shows the resulting
XUV emission spectra from the set of trajectories emitted close to the laser field peaks, with the
line formatting corresponding to the trajectories displayed in the top panel. The spectrum resulting
from the solid line trajectory in the top panel is not visible due to its low ionisation probability and
hence intensity. The dash-dotted line in the bottom panel shows the bandpass filter, which filters
out the harmonic emission for the isolated XUV pulse (Reproduced and adapted with permission
from Ref. [30])
field which is locked to the XUV pulse in phase, but with the possibility to vary
the phase of the arrival of the XUV pulse. The measured photoelectron energies
correspond to a mapping of the XUV pulse spectrum. The “streaking effect” of the
IR field on the photoelectrons shifts the momentum by an amount that depends upon
its electric field amplitude and the phase at which ionisation occurred and therefore
provides the spectral phase information of the XUV pulse.
The momentum of the electron observed in the laser polarisation direction at
some time t following ionisation by the XUV pulse at time t0 is given by the
expression [33]:
eE0
p.t; t0 / D O
Œcos.!t0 /Oy cos.!t/Oy C ı u (3.13)
!
Where the momentum imparted by the undressed XUV pulse is given by:
q
ıD .Exuv Ip /=2Up (3.14)
60 J.P. Marangos and M. Oppermann
where e is the electron charge, E0 and ! the IR field amplitude and carrier frequency,
respectively, Exuv and Ip being the XUV photon energy and atomic ionisation
potential, respectively, Up D e 2 E0 2 =4m! 2 the quiver energy of the released
electron of mass m in the IR pulse, yO and uO are the IR field polarisation and initial
direction of the outgoing electron, respectively. The final electron momentum at the
end of the pulse, i.e. that measured at the detector, is given by:
eE0
pf .t0 / D O
Œcos.!t0 /Oy C ı u (3.15)
!
Note the final momentum depends on a term E0 ı which for modest IR field
strength and moderate initial photo-electron energy (a few 10’s of eV) leads to a
significant momentum shift that is then modulated by the oscillating term depending
on the field. The momentum that is observed therefore encodes the time t0 of
ionisation by the XUV.
Streaking measurements can be analysed by composing a spectrogram from a
series of streaked electron spectra recorded for different delays (i.e. different t0 )
between the laser and the attosecond pulse. Such a data set contains both temporal
and spectral information about the physical process under study. Measurement
of the “streaked” photoelectron spectrum thus provides information on when the
XUV pulse arrives within the IR pulse and (from the spread in final measured
photoelectron energies) a mapping of the XUV pulse. In the limit of a single XUV
pulse shorter than the optical cycle time a delay scan of the XUV pulse through the
(attenuated) IR few-cycle pulse provides sufficient information to retrieve both the
IR field history and the XUV pulse duration (from algorithms such as FROG-CRAB
[34]). (See Chap. 1 by Wyatt et al. for a discussion of ultrashort pulse measuring
techniques).
A related application of HHG has recently been exploited to carry out a new class
of attosecond measurements. The approach is to characterise very carefully the
harmonic radiation emitted from a molecule. It has been found that the spectrum,
phase and polarisation of this radiation can encode information of molecular
structure and internal dynamics. The recombination step of HHG depends upon
the matrix element between the continuum state ((k) with returning electron
momentum k) and the bound state for a particular molecular ion intermediate state:
D E
Aj D '0 jkj.k/'jC (3.16)
X
D.!/ D Aj recombination Œt.!/Aj i propagation Œt.!/; t 0 .!/Ai ionization Œt 0 .!/ (3.17)
ij
where, reading from right to left, are the amplitudes for ionisation at t’, propagation
between t’ and t, and recombination at t; the summation is over different ‘quantum
trajectories’ that result in the same return energy. Crucially, different phase-
matching properties of the radiation associated with different trajectories allow
one to effectively select a single term from the sum in the measurements. This
simplification makes the analysis of the connection between the harmonic spectrum
and the underlying physics practical and feasible.
The possibility to extract molecular orbitals from this approach has been widely
discussed and the reader is recommended to read several original research papers
and reviews on this topic (for example Ref. [35]), although there is not space here
to review this further in depth.
An intriguing side of HHG spectroscopy is the ability to probe the dynamics that
the parent molecular ion undergoes while the “probe” electron is away. Note that
these dynamics are only launched upon ionisation and curtailed by recombination.
For time resolution it is the inherent chirp of the harmonic emission that is the
crucial aspect. Clearly, D(t) encodes attosecond dynamics taking place between
ionisation and recombination and the presence of the strong laser field allows one
to dramatically simplify the formal expression, making such recovery possible.
High harmonic spectroscopy can probe such dynamics with time-resolution
better than 100 asec. First steps in using HHG for attosecond probing have been
made in [36]. In a molecule, ionisation simultaneously launches an electronic
wavepacket into the continuum and a nuclear wavepacket onto the groundstate
electronic potential energy surface of the molecular ion. As the nuclei move between
ionisation and return, recombination to the initial state requires an overlap between
the evolving nuclear wavepacket (R, t(¨), t0 (¨)] in the ion and the original
nuclear wavefunction 0 (R) in the neutral (R labels the nuclear coordinates, e.g.
the internuclear distance for a diatomic). With the corresponding modification of
the recombination amplitude, the harmonic emission becomes [20]:
Z
D.!/ D 0 .R/ R; t.!/; t 0 .!/ D0 .R; !/dR (3.18)
Fig. 3.6 Illustration of the impact of nuclear dynamics on the observed high harmonic emission as
described in the text. Three electron trajectories are displayed at different ionisation and recollision
times. The recollision time and hence time delay t is encoded in the emitted harmonic frequency,
whereas the corresponding intensity depends on the internuclear stretching during that time. Here,
a lower intensity is suggested by a thinner linewidth of the harmonic emission (Reproduced and
adapted with permission from Ref. [36])
to a known relationship (see Eq. (3.12)) The emission of each harmonic order
therefore probes the molecular ion at a different time delay. The strength of the
harmonic signal at different frequencies therefore has encoded within it the position
of the molecular nuclei at a range of ionisation-recollision time delays. In this
way a remarkable temporal resolution is achieved, since the emission of successive
harmonic orders is typically separated in time by less than 100 as.
The strength of the harmonic signal at different frequencies also depends on
ionisation and propagation steps contributing to D0 (R, t). In the first PACER
measurement, the effect of the nuclear motion was isolated from any effects of the
electronic parts of Eq. (3.18) by comparing the harmonic emission in two isotopes
of the same species. This procedure circumvents the need to consider details of the
electronic contributions such as the recombination and ionisation amplitudes for a
given R – these are similar for the two isotopes. The same applies to the impact of the
macroscopic propagation of the emitted light. The PACER technique is described
in detail in [36, 37]. In essence it is simply necessary to compare high harmonic
generation in protonated and deuterated samples of a molecule at the same laser
intensity and gas density. As an example, the harmonic signals in both H2 and D2
as well as CH4 and CD4 were recorded, and any difference in the spectra could be
attributed largely to the different nuclear dynamics in the two species. This principle
is illustrated in Fig. 3.6 for H2 and D2 .
More recently the concept of PACER has begun to be applied to studying
the electronic dynamics that are initiated when the HHG ionisation event simul-
taneously excites a superposition of two ionic electronic states [38, 39]. In the
3 Attosecond Generation and High Field Physics 63
case of CO2 , interference between these channels has been obseved in addition
to the structural interference. It is anticipated that this method can be extended
to investigate a wide range of molecular hole dynamics. To achieve this it is
necessary to decouple the structural dependence of the HHG signal from the
temporal dependence. There are several possible strategies to do this, the isotope
comparison of the original PACER measurement being an example. The HHG
spectrum at a given photon energy is of course emitted at different times if the laser
wavelength and/or intensity is changed [38, 39] opening the way to make these
separations without a reference molecule or atom.
HHG remains very promising and is currently the only demonstrated path for
making attosecond measurements. There are, nevertheless, several potential short-
comings in HHG based routes coming principally from the limited photon energy
range of the sources (confined to the VUV (>20 eV) to very soft X-ray range
<100 eV) and so presently not capable of accessing inner shell processes. Also the
low pulse energy of the sources demonstrated so far (nJ) is still not sufficient for
many potential applications. We now turn to other routes for generating attosecond
pulses that may overcome these limits within the next few years.
close to the vacuum speed of light c. This mechanism has been termed the relativistic
oscillating mirror and along with the coherent wakefield emission mechanism that
is observed at somewhat lower laser intensities gives rise to a very broad spectrum
of efficient harmonics (for details see [47]). Such surface oscillations are driven
when a high intensity laser is reflected from either a solid or liquid surface. Here,
the electric field direction is into the surface. The maximum upshift that can be
observed scales with laser intensity and allows the up-conversion of NIR lasers to
100 eV for 5 1019 Wcm2 increasing to multiple keV for intensities exceeding
1020 Wcm2 . At high intensities the RHHG conversion efficiency is independent of
laser intensity (i.e. it becomes a linear process) with a conversion efficiency into the
nth harmonic scaling as n2.6 [48].
RHHG intrinsically produces attosecond pulses (or trains) with one pulse emitted
per laser cycle and it is currently believed to be the brightest known source of
attosecond pulses [49] (see Table 3.1). Scaling to higher drive energies and higher
photon fluxes is currently an area of active research but there is already a clear
advantage over gas target HHG in terms of efficiency and photon flux. Currently
the laser sources used to drive RHHG have been relatively low repetition rate
(10 Hz or less) and this makes the application of the attosecond pulses so far
produced challenging. Future sources based upon RHHG will require high intensity
(>>1019 Wcm2 ) laser source to drive attosecond pulse train sources, with a need
for kHz and higher repetition rates.
Moreover to drive an isolated attosecond source at full efficiency requires a high
intensity (>1019 Wcm2 ) few cycle (sub 10 fs) laser source or a demonstration of
polarisation gating (as in gas HHG) is needed. A high pulse contrast (>1011) is
essential as the surface must not be disrupted prior to exposure to the high intensity
laser. This is required either from the laser itself or as so far shown by use of plasma
mirrors. Work on optical parametric amplification (OPCPA) lasers may be able to
remove the need for plasma mirrors [50] by delivering high contrast, few-cycle drive
pulses of the required intensity. To match the future high repetition rate lasers high
repetition rate target sources must be developed e.g. by use of liquid jets or tape
targets (for details see [51] and references therein).
In the last few years the first true X-ray lasers have become available for scientific
applications. These free-electron sources are based upon high charge, low emittance,
bunches of electrons accelerated to relativistic energies which emit copious amounts
of coherent radiation if they pass through a sufficiently long magnetic undulator
structure. An undulator consists of a periodic lattice of permanent magnets that
creates a strong alternating transverse magnetic field. This forces the electron
bunch to oscillate and emit electromagnetic radiation at a fundamental radiation
wavelength which co-propagates with the electrons. The fundamental wavelength
66
Table 3.1 Comparison of ultra-short XUV pulse sources and related techniques. Material collated for NLS project (www.newlightsource.org)
Accelerator based
HHG based sources Free Electron Lasers
HHG RHHG Slicing Current Energy Chirp Single Spike Mode-Locking
Pulse Length <100 as <100 as 250 as <200 as 3 fs–300 as 400 as–23 as pulse train
Photon energy 100 eV–1 keV 1–10 keV 12 keV 12 keV 300 eV–12 keV 100 eV–8.6 keV
Photons per 106 –104 1011 (1 keV) 109 1010 1010 –108 1010 –108
pulse
Peak Power 200 W (1 keV) 1 GW 10 GW 100 GW 250 MW–5 GW 1–5 GW
Contrast Good Good Poor Good Excellent Good
Rep-rate Laser <10 kHz Laser 10 Hz Laser seed Laser seed LINAC Laser seed
Synchronisation Yes Yes Yes Yes No Yes
(with laser)
J.P. Marangos and M. Oppermann
3 Attosecond Generation and High Field Physics 67
is given by 1 D u 1 C aN u2 =2ƒ2 ; where u is the period of the undulator, ƒ
is the energy of the electrons in units of their rest-mass and aN u2 is the undulator
parameter, proportional to the root means square (RMS) magnetic field strength
between its poles. The continued oscillation of the electrons in the presence of this
radiation results in an exponential instability that forces the electrons to bunch at the
fundamental radiation wavelength and to emit coherently at that wavelength – the
so called Self-Amplified Spontaneous Emission process (SASE).
At the time of writing, three such facilities are operational: XUV machines
at DESY, Hamburg (FLASH) and SPring-8 Japan and the world’s first X-ray
FEL the LCLS (Linac Coherent Light Source) at SLAC, Stanford, USA. They
are based upon SASE [52] that generates radiation pulses of limited temporal
coherence consisting of a series of phase uncorrelated spikes. Pulse durations as
short as a few femtoseconds are present in a single emission “spike” but typically
contain many such spikes within a pulse of 30–200 fs duration. The radiation
is thus multimode, and because SASE is triggered by spontaneous processes, is
inherently stochastic with a significant temporal jitter (typically larger than the pulse
duration). Nevertheless the lasing bandwidth in a FEL is sufficient to support sub
100 attosecond pulses and so ways to control the emission are now being sought to
permit the intrinsic attosecond capability to be harnessed.
One of the simplest methods to reduce the poor temporal coherence of the
SASE output is to seed the FEL amplifier with a resonant seed source of power
significantly greater than that of the intrinsic spontaneous noise from which SASE
starts. Successful seeding of the Japanese VUV FEL by a HHG source was
demonstrated at 160 nm [53] and it is anticipated that HHG seeding up to 100 eV
photon energies will be possible (10 nm). Where seed sources have insufficient
power to dominate initial noise towards shorter wavelengths, currently .10 nm
for HHG sources, it is possible to use an Harmonic Optical Klystron [54] (HOK) or
High Gain Harmonic Generation [55] (HGHG) configuration of the high-gain FEL
to reach shorter wavelengths while retaining good temporal coherence.
Several techniques have been identified that may shorten the pulse durations of
<20 fs possible from seeding yet further into the attosecond regime. A conceptually
simple method has been suggested [56] that ‘spoils’ all but a short region of
the electron bunch. Only this region subsequently lases in the FEL which could
potentially operate under either SASE or seeded mode to generate a short radiation
pulse of duration <1 fs.
A wide range of proposals have been made to pre-modulate the electron bunch
energy with a carrier envelope phase-locked high power laser before the bunch
enters the radiator undulator. The resonant FEL wavelength is correlated to this
energy modulation, and it may be possible to selectively filter [57] or amplify [58] a
restricted wavelength band to generate short pulses with widths of a fraction of the
modulation period. Another approach, called E-SASE [59], uses laser modulation
of the electron bunch energy at an intermediate acceleration stage. Regions of
enhanced current are created that subsequently generate short pulses in a final
radiator undulator. The above techniques might generate single attosecond pulses
68 J.P. Marangos and M. Oppermann
(there may be smaller satellite pulses about the main pulse) of typical duration
100–300 as at a target wavelength of 1 1 Å.
Two different methods from those discussed above are viable candidates for short
pulse X-ray generation. The first employs a relatively low charge, short electron
bunch of length of a few cooperation lengths [60]. Such electron pulses emit only
one or two spikes in the SASE mode (see above). This mode has been successfully
employed at LCLS [61] where the short pulse has been utilised, e.g. for studies
of double core hole formation as the pulse is shorter than the characteristic Auger
decay times in most atoms [62]. The second method has not been demonstrated
experimentally yet. It uses techniques familiar from mode-locking in conventional
cavity lasers to generate an attosecond pulse train [63]. A cavity-like longitudinal
mode structure can be synthesised in a FEL amplifier by applying a series of
spatiotemporal shifts between the radiation and copropagating electron bunch.
These shifts are achieved by repeatedly delaying the electron bunch using magnetic
chicanes between undulator modules. By introducing an interaction modulation
at the synthesized mode spacing the modes may become phase locked to give a
temporal train of short, equally spaced, high-power pulses. At 12 nm a train of
400 as pulses, evenly spaced at 2.5 fs, of peak power 1 GW are predicted for
typical FEL parameters for this wavelength range. For operation at 1.5 Å, again
using typical parameters, a pulse train consists of 23 as pulses separated by 150 as
with peak powers up to 6 GW are predicted.
The methods for attosecond measurement that arise from attosecond generation in
strong laser fields have been summarised in this chapter. The possible alternatives
to gas phase HHG have been discussed, the majority of these also relying on strong
field interactions e.g. RHHG and most of the viable FEL based methods. Despite
the potential of the alternatives, gas phase HHG remains the dominant technique
for attosecond measurement since it is the only route that has demonstrated actual
attosecond measurement applications. This is partly because HHG is now a rather
well understood process both theoretically and technically. Moreover, although
state-of-the-art few cycle lasers are required, these are based upon commercially
available and relatively small scale laser systems that offer access to “attosecond
science” to any motivated university laboratory (i.e. do not require the resources
of large national laboratories and research institutes). Moreover a critical fact of
many attosecond experiments where the measurement process dictates low count
rates per laser shot (e.g. the need to avoid space charge effects in photoelectron
studies or to make correlated or coincidence measurements) is access to high
repetition rates (kHz or greater). RHHG and FEL based methods are therefore
still requiring technical breakthroughs in lasers technology and superconducting
accelerator technology before these sources can be as versatile and high rep-rate
despite their potential advantages.
3 Attosecond Generation and High Field Physics 69
References
1. G.R. Fleming, M.A. Ratner, Grand challenges in basic energy sciences. Phys. Today 61, 28–33
(2008)
2. F. Krausz, M. Ivanov, Attosecond physics. Rev. Mod. Phys. 81, 163–234 (2009)
3. L.S. Cederbaum, J. Zobeley, Ultrafast charge migration by electron correlation. Chem. Phys.
Lett. 307, 205–210 (1999)
4. F. Remacle, R.D. Levine, An electronic time scale in chemistry. Proc. Natl. Acad. Sci. USA
103, 6793–6798 (2006)
5. S. Lünnemann, A.I. Kuleff, L.S. Cederbaum, Ultrafast charge migration in 2-phenylethyl-N,N-
dimethylamine. Chem. Phys. Lett. 450, 232–235 (2008)
6. C. Altucci, R. Velotta, J.P. Marangos, Ultra-fast dynamic imaging: an overview of current
techniques, their capabilities and future prospects. J. Mod. Opt. 57, 916–952 (2010)
7. C.C. Tannoudji, G. Grynberg, J. Dupont-Roe, Atom-Photon Interactions (Wiley, New York,
1992)
8. P.B. Corkum, Plasma perspective on strong field multiphoton ionization. Phys. Rev. Lett. 71,
1994–1997 (1993)
9. K.J. Schafer, B. Yang, L.F. DiMauro, K.C. Kulander, Above threshold ionization beyond the
high harmonic cutoff. Phys. Rev. Lett. 70, 1599–1602 (1993)
10. G.L. Yudin, M.Y. Ivanov, Physics of correlated double ionization of atoms in intense laser
fields: quasistatic tunneling limit. Phys. Rev. A 63, 033404 (2001)
11. L.V. Keldysh, Ionization in the field of a strong electromagnetic wave. JETP 20, 1307–1314
(1965)
12. M.V. Ammosov, N.B. Delone, V.P. Kranov, Tunnel ionization of complex atoms and atomic
ions in an alternating electromagnetic field. JETP 91, 2008–2013 (1986)
13. A.M. Perelomov, V.S. Popov, Ionization of atoms in an alternating electrical field. 9. JETP 25,
336 (1967)
14. A.M. Perelomov, V.S. Popov, M.V. Terent’ev, Ionization of atoms in an alternating electrical
field. II. JETP 24, 207 (1967)
15. A.M. Perelomov, V.S. Popov, M.V. Terent’ev, Ionization of atoms in an alternating electrical
field. JETP 23, 924 (1966)
16. X.M. Tong, Z.X. Zhao, C.D. Lin, Theory of molecular tunneling ionization. Phys. Rev. A 66,
033402 (2002)
17. R. Murray, M.Y. Ivanov, Tunnel ionization of molecules and orbital imaging. Phys. Rev. Lett.
106, 173001 (2011)
18. M. Meckel, D. Comtois, D. Zeidler, A. Staudte, D. Pavičić, H.C. Bandulet, H. Pépin, J.C.
Kieffer, R. Dörner, D.M. Villeneuve, P.B. Corkum, Laser-induced electron tunneling and
diffraction. Science 320, 1478–1482 (2008)
19. M. Lewenstein, P. Balcou, M.Y. Ivanov, A. L’Huillier, P.B. Corkum, Theory of high-harmonic
generation by low-frequency laser fields. Phys. Rev. A 49, 2117–2132 (1994)
20. M. Lein, Attosecond probing of vibrational dynamics with high-harmonic generation. Phys.
Rev. Lett. 94, 053004 (2005)
70 J.P. Marangos and M. Oppermann
41. A.V. Sokolov, D.R. Walker, D.D. Yavuz, G.Y. Yin, S.E. Harris, Raman generation by phased
and antiphased molecular states. Phys. Rev. Lett. 85, 562–565 (2000)
42. J.Q. Liang, M. Katsuragawa, F.L. Kien, K. Hakuta, Sideband generation using strongly driven
Raman coherence in solid hydrogen. Phys. Rev. Lett. 85, 2474 (2000)
43. S. Baker, M.P. Anscombe, A.M. Abdulla, E. Sali, J.W.G. Tisch, P. Kinsler, G.H.C. New, J.P.
Marangos, Ultrashort-pulse modulation in adiabatically prepared Raman media. Opt. Lett. 30,
180–182 (2005)
44. P.J. Bustard, B.J. Sussman, I.A. Walmsley, Amplification of impulsively excited molecular
rotational coherence. Phys. Rev. Lett. 104, 193902 (2010)
45. N. Zhavoronkov, G. Korn, Generation of single intense short optical pulses by ultrafast
molecular phase modulation. Phys. Rev. Lett. 88, 203901 (2002)
46. S. Baker, J.W.G. Tisch, I.A. Walmsley, J.P. Marangos, Femtosecond to attosecond light pulses
from a molecular modulator. Nat. Phot. 55(11), 664–671 (2011)
47. G.D. Tsakiris, K. Eidmann, J. Meyer-ter-Vehn, F. Krausz, Route to intense single attosecond
pulses. New J. Phys. 8, 19 (2006)
48. B. Dromey, M. Zepf, A. Gopal, K. Lancaster, M.S. Wei, K. Krushelnick, M. Tatarakis, N.
Vakakis, S. Moustaizis, R. Kodama, M. Tampo, C. Stoeckl, R. Clarke, H. Habara, D. Neely, S.
Karsch, P. Norreys, High harmonic generation in the relativistic limit. Nat. Phys. 2, 456–459
(2006)
49. B. Dromey, S. Kar, C. Bellei, D.C. Carroll, R.J. Clarke, J.S. Green, S. Kneip, K. Markey,
S.R. Nagel, P.T. Simpson, L. Willingale, P. McKenna, D. Neely, Z. Najmudin, K. Krushelnick,
P.A. Norreys, M. Zepf, Bright multi-keV harmonic generation from relativistically oscillating
plasma surfaces. Phys. Rev. Lett. 99, 085001 (2007)
50. B. Dromey, S. Kar, M. Zepf, P. Foster, The plasma mirror—a subpicosecond optical switch for
ultrahigh power lasers. Rev. Sci. Instrum. 75, 645–649 (2004)
51. R. Hörlein, Y. Nomura, J. Osterhoff, Z. Major, S. Karsch, F. Krausz, G.D. Tsakiris, High
harmonics from solid surfaces as a source of ultra-bright XUV radiation for experiments.
Plasma Phys. Control. Fusion 50, 124002 (2008)
52. R. Bonifacio, L. De Salvo, P. Pierini, N. Piovella, C. Pellegrini, Spectrum, temporal structure,
and fluctuations in a high-gain free-electron laser starting from noise. Phys. Rev. Lett. 73,
70–73 (1994)
53. G. Lambert, T. Hara, D. Garzella, T. Tanikawa, M. Labat, B. Carre, H. Kitamura, T. Shintake,
M. Bougeard, S. Inoue, Y. Tanaka, P. Salieres, H. Merdji, O. Chubar, O. Gobert, K. Tahara,
M.E. Couprie, Injection of harmonics generated in gas in a free-electron laser providing intense
and coherent extreme-ultraviolet light. Nat. Phys. 4, 296–300 (2008)
54. R. Bonifacio, R. Corsini, P. Pierini, Theory of the high-gain optical klystron. Phys. Rev. A 45,
4091 (1992)
55. A. Doyuran, L. DiMauro, W. Graves, R. Heese, E.D. Johnson, S. Krinsky, H. Loos, J.B.
Murphy, G. Rakowsky, J. Rose, T. Shaftan, B. Sheehy, Y. Shen, J. Skaritka, X. Wang, Z.
Wu, L.H. Yu, Experimental study of a high-gain harmonic-generation free-electron laser in
the ultraviolet. Phys. Rev. Spec. Top Accel. Beam 7, 050701 (2004)
56. P. Emma, M. Borland, Z. Huang, Attosecond x-ray pulses in the LCLS using the slotted foil
method, in 26th International Free Electron Laser Conference and 11th FEL User Workshop
2004, Trieste, Italy, 29 Aug–3 Sep 2004, pp. 333–338
57. E.L. Saldin, E.A. Schneidmiller, M.V. Yurkov, Terawatt-scale sub-10-fs laser technology – key
to generation of GW-level attosecond pulses in X-ray free electron laser. Opt. Commun. 237,
153–164 (2004)
58. E.L. Saldin, E.A. Schneidmiller, M.V. Yurkov, A new technique to generate 100 GW-level
attosecond X-ray pulses from the X-ray SASE FELs. Opt. Commun. 239, 161–172 (2004)
59. A.A. Zholents, Method of an enhanced self-amplified spontaneous emission for x-ray free
electron lasers. Phys. Rev. Spec. Top. Acel. Beam 8, 040701 (2005)
60. R. Bonifacio, B.W.J. McNeil, P. Pierini, Superradiance in the high-gain free-electron laser.
Phys. Rev. A 40, 4467–4475 (1989)
72 J.P. Marangos and M. Oppermann
61. Y. Ding, A. Brachmann, F.J. Decker, D. Dowell, P. Emma, J. Frisch, S. Gilevich, G. Hays,
P. Hering, Z. Huang, R. Iverson, H. Loos, A. Miahnahri, H.D. Nuhn, D. Ratner, J. Turner, J.
Welch, W. White, J. Wu, Measurements and simulations of ultralow emittance and ultrashort
electron beams in the linac coherent light source. Phys. Rev. Lett. 102, 254801 (2009)
62. J.P. Cryan, J.M. Glownia, J. Andreasson, A. Belkacem, N. Berrah, C.I. Blaga, C. Bostedt, J.
Bozek, C. Buth, L.F. DiMauro, L. Fang, O. Gessner, M. Guehr, J. Hajdu, M.P. Hertlein, M.
Hoener, O. Kornilov, J.P. Marangos, A.M. March, B.K. McFarland, H. Merdji, V.S. Petrović,
C. Raman, D. Ray, D. Reis, F. Tarantelli, M. Trigo, J.L. White, W. White, L. Young, P.H.
Bucksbaum, R.N. Coffee, Auger electron angular distribution of double core-hole states in the
molecular reference frame. Phys. Rev. Lett. 105, 083004 (2010)
63. N.R. Thompson, B.W.J. McNeil, Mode locking in a free-electron laser amplifier. Phys. Rev.
Lett. 100, 203901 (2008)
64. T. Pfeifer, A. Jullien, M.J. Abel, P.M. Nagel, L. Gallmann, D.M. Neumark, S.R. Leone,
Generating coherent broadband continuum soft-x-ray radiation byattosecond ionization gating.
Opt. Express 15, 17120–17128 (2007)
Chapter 4
Advances in Solid-State Ultrafast Laser
Oscillators
4.1 Introduction
Nearly all modern ultrashort-pulse lasers rely on the process known as passive
modelocking to generate and sustain femtosecond pulses. In an inhomogeneously-
broadened laser, different frequencies in the emission bandwidth experience gain
simultaneously when they correspond to the frequencies of the longitudinal modes
of the resonator. The frequency separation between adjacent longitudinal resonator
modes can be expressed as:
! D 2 c= l (4.1)
where l is the optical roundtrip length of the resonator and c is the speed of light.
When these modes are coupled by a fixed phase relationship – the modelocked
condition – a periodic sequence of ultrashort optical pulses is generated. If the nth
longitudinal mode has an amplitude En , then the total optical field can be denoted as:
X
E.t/ D En exp i Œ.!0 C n !/t C n ' (4.2)
n
Figure 4.1 illustrates how the coherent superposition of such resonator modes
generates a wavepacket whose extent in the time-domain is inversely propor-
tional to its frequency-domain bandwidth. Practical femtosecond lasers require
frequency bandwidths in excess of 1 THz, typically corresponding to the coherent
superposition of around 100,000 resonator modes. This requirement places a
demanding constraint on which laser gain materials can be used for femtosecond-
pulse generation, ruling out gas lasers and focusing attention on media with
broadband fluorescence spectra, such as dyes, vibronic crystals and doped amor-
phous materials.
From the discussion above, it is clear that practical modelocking requires both
! and ' to be constants; ! is a constant when the resonator length is fixed,
however ' must be stabilized by using a modelocking mechanism – any effect
which introduces a periodic modulation of either the amplitude or phase of the field
in the resonator. It can be shown (see Box 4.1) that an amplitude or phase modulation
of a monochromatic field of frequency !0 leads to the creation of coherent sidebands
at frequencies of !0 ˙ , where is the modulation frequency. If we consider
a single longitudinal resonator mode oscillating with frequency !0 , then when
exactly matches the frequency separation between adjacent resonator modes,
!, the generated sidebands experience low-loss propagation in the laser cavity
and can themselves be amplified in the laser gain medium. Furthermore, because
these sidebands are generated coherently, they have a constant phase relationship
to the original mode and so can combine coherently to form a modelocked pulse.
This process repeats over many cavity roundtrips, with the periodic modulation
4 Advances in Solid-State Ultrafast Laser Oscillators 75
Fig. 4.1 (a) Time-domain and (b) frequency-domain representations of a modelocked pulse
synthesised by the coherent superposition of (top to bottom) 3, 5, 7 and 9 resonator modes
creating sidebands on each of the resonant cavity modes, broadening the frequency
bandwidth of the oscillating field and shortening the pulse duration, until a steady-
state condition is reached which results from the finite bandwidth of the gain
medium, or the dispersion characteristics of the cavity.
The requirement for the amplitude or phase modulation frequency, , to exactly
match the frequency separation between adjacent resonator modes, !, means that
the laser field must be modulated with a period equal to roundtrip cavity time, l/c.
In a passively modelocked laser, the periodic modulation of the pulse amplitude
is obtained by using a saturable absorber, an optical element whose loss decreases
as the incident light intensity increases. Incorporating a saturable absorber inside a
laser resonator allows a resonant pulse to modulate its own amplitude – a process
known as self-amplitude modulation (SAM) – and this effect automatically satisfies
the modelocking requirement that D !.
Many different saturable absorption effects and devices have been successfully
used to obtain modelocking at pulse durations below 1 ps, and include: organic
dyes [1]; semiconductor saturable absorber mirrors (SESAMs) [2]; Kerr-effect self-
focusing [3]; Kerr-effect nonlinear polarisation rotation [4]; Kerr-effect self-phase
modulation [5–9]; and carbon nanotubes [10].
The dynamics of the modelocking process depend critically on the speed of the
SAM effect and its recovery time after the loss of the absorber has been saturated
by an intense pulse. Two characteristic timescales are particularly relevant, namely
76 C.G. Leburn and D.T. Reid
the pulse roundtrip time of the laser cavity (typically around 10 ns) and the pulse
duration (typically a few hundred fs or less). Slow saturable absorbers (for example,
organic dyes) exhibit a typical recovery time of many ns which is comparable
with the pulse roundtrip time of the laser. In contrast, fast saturable absorbers (for
example, effects which exploit a Kerr nonlinearity) have a recovery time which is
either instantaneous or comparable to the pulse duration. Many practical saturable
absorbers, such as SESAMs and carbon nanotubes, possess recovery times of
the order of several ps, several times the pulse duration, but much shorter than the
pulse roundtrip time. These saturable absorbers can still be used for femtosecond
pulse generation when the dispersion of the laser resonator is properly controlled.
Box 4.1
The field of a laser cavity mode n can be represented as:
En exp.i !n t/
By using the identity cos t D Œexp.i t/ C exp.i t/=2, this field may
be rewritten as:
m m
E D En exp.i !n t/ C En exp.i.!n C /t/ C En exp.i.!n /t/
2 2
Phase modulation may be analysed in a similar way, and the phase-
modulated field may be written as:
In this notation, A(t) is expressed in units normalized so that jA.t/j2 represents the
instantaneous optical power. It is often useful to express the pulse in the optical
frequency domain, corresponding to a Fourier transform of the complex temporal
amplitude:
Z1
1
A.!/ D p A.t/ exp.i !t/dt (4.4)
2
1
The quantity jA.!/j2 is the power spectrum of the pulse, which is readily measured
in the laboratory by an optical spectrum analyser.
The duration-bandwidth product describes the relationship between the width of
a pulse in time and its spectral bandwidth. Table 4.1 lists the minimum duration-
bandwidth products resulting from pulses with particular temporal profiles. The
duration, p , and the spectral bandwidth, p , are measured at the full-width at
half-maximum (FWHM) points. It can be seen from Table 4.1 that the duration-
bandwidth product increases as the energy content in the wings of the pulse
decreases. This effect arises from the FWHM-definition of the pulse duration.
Chirping of the pulse due to dispersion increases the duration-bandwidth product
to above the minimum values listed in Table 4.1.
With the exception of propagation in a vacuum, a pulse always experiences
dispersive effects as it travels through optical materials and systems. These effects
can be treated generally as an additional frequency-dependent phase, '(!), written
onto the pulse:
Z1
0 1
A.t/ D p A.!/0 exp.i !t/d! (4.6)
2
1
78 C.G. Leburn and D.T. Reid
Table 4.1 Common field envelope profiles of ultrashort pulses and their minimum duration-
bandwidth products
Pulse A.t / vp p
Hyperbolic secant
sech.1:763t = p / 0.315
Gaussian
exp.2 ln 2t 2 = p2 / 0.441
Parabolic s
2t 2 0.728
1
p2
Rectangle
p 0.886
1; jt j
2
p
0; jt j >
2
The local variation of the spectral phase due to an optical medium can be represented
by the Taylor expansion:
ˇ ˇ
@' ˇˇ 1 2 ˇ
2@ 'ˇ
'.!/ D '.!0 / C .! !0 / ˇ C .! !0 /
@! !0 2 @! 2 ˇ!0
3 ˇ
ˇ n ˇ
ˇ
1 3@ 'ˇ 1 n@ 'ˇ
C .! !0 / C C .! !0 / (4.7)
6 @! 3 ˇ!0 nŠ @! n ˇ!0
The first term in Eq. (4.7) is the change in carrier phase after passing through the
medium and has no effect on the shape of the pulse. In the second term we can
observe a linear phase ramp in frequency which corresponds, through the Fourier
4 Advances in Solid-State Ultrafast Laser Oscillators 79
Fig. 4.2 (a) Unchirped pulse; (b) chirped pulse showing the variation of instantaneous frequency
shift theorem, to a delay of the pulse in time. A quantity known as the group delay
has units of fs and corresponds to the time spent by the pulse inside the medium is
given by:
ˇ
@' ˇˇ
g D (4.8)
@! ˇ!0
For a medium of physical length z, the group delay can be related directly to the
group velocity, vg , of the pulse in the medium by using the definition:
ˇ
@k ˇˇ
vg D 1= (4.9)
@! ˇ!0
The quadratic and higher-order terms describe the variation of the group delay with
frequency (or wavelength), and have a direct influence on the shape of the pulse
in time. The .! !0 /2 term is responsible for adding linear chirp which refers to
a time-varying uniform increase or decrease in the instantaneous optical frequency
across the pulse. Figure 4.2 compares chirped and unchirped pulses, showing the
variation in instantaneous frequency of a chirped pulse. The pulse shown has
positive linear chirp, meaning the instantaneous frequency increases linearly with
time across the pulse.
Formally, group delay dispersion (GDD) has units of fs2 and is described by the
quantity:
ˇ
@g @2 ' ˇˇ
GDD D D (4.11)
@! @! 2 ˇ!0
80 C.G. Leburn and D.T. Reid
Table 4.2 Dispersion formulae for light propagating a distance z in a dielectric medium
Expressed in phase/ Expressed in index
Quantity wavevector and frequency and wavelength
! c
Phase velocity vp D
k n
1
@k dn
Group velocity vg D 1= c n
@! d
@' z dn
First-order dispersion (group delay) g D n
@! c d
@g @2 ' 3 z d 2 n
Second-order dispersion (GDD) D
@! @! 2 2 c 2 d 2
2
@3 ' 4 z d n d 3n
Third-order dispersion (TOD) 3 C
@! 3 4 2 c 3 d 2 d 3
Strategies for generating the shortest pulses from ultrafast lasers rely on control-
ling the net intracavity dispersion so that the quadratic and higher-order spectral
phase terms associated with one cavity roundtrip are negligible across the entire
spectral bandwidth of the pulse. Contributions to the cavity dispersion may originate
from a number of sources:
Material dispersion is associated with the wavelength-dependent refractive index
of a dielectric medium. In solid-state lasers this is limited mainly to the gain
medium, but in fibre lasers may be distributed throughout the entire cavity.
For propagation in bulk dielectric media, the material dispersion is commonly
represented in terms of the wavelength-dependent refractive index, n(). It is
therefore often convenient to express the quantities in Eqs. (4.9), (4.10), (4.11),
and (4.12) in a form that can allow them to be evaluated directly from n(), and
Table 4.2 lists the formulae used for this conversion.
Geometrical dispersion arises from the wavelength-dependent path lengths that
can be observed in optical systems exhibiting angular dispersion, such as prism
or diffraction-grating sequences. The GDD and TOD of such systems may be
highly controllable by careful choice of the materials and geometrical layout of the
elements.
Table 4.3 lists the second and third-order dispersion formula for double-passed
prism [11–13] and grating pairs [14, 15], the two most common optical systems used
to obtain geometrical dispersion. The formulae listed can be combined with those
4 Advances in Solid-State Ultrafast Laser Oscillators 81
Table 4.3 Expressions for the second and third-order dispersion of common optical systems
Double-pass prism-pair, apex separation l, d 2' 3 d 2 P
angular deviation within ray of “ 2
D
d! 2c 2 d 2
2
d 3' 4 d P d 3P
D 3 C
d !3 4 2 c 3 d 2 d 3
" 2#
d 2P d 2n dn
D 4l sin ˇ C.2n1=n /3
d 2 d 2 d
2
dn
8l cos ˇ
d
"
d 3P ı 3 dn 3 d n d 2n
D l cos ˇ 24 n 48n 24
d 3 d d d 2
" 3
dn ı 6 ı ı ı
C l sin ˇ 12 n C 12 n4 C 8 n3 16 n2 C 32n
d
ı dn d 2 n d 3n
C 24n 12 n3 C 4
d d 2 d 3
" 2 #3=2
Double-pass grating-pair, separation l, angle
d 2' 3
of incidence ” , line spacing d. Dl 2 2 1 sin
d !2 c d d
d 3' 4
D 6l 1 C sin sin2
d !3 c3d 2 d
" 2 #5=2
1 sin
d
Fig. 4.3 Simulated reflectivity and group-delay dispersion of a Bragg mirror comprising 10 equal
pairs of quarter-wave layers with alternating refractive indices of 1.35 and 2.35
of alternating n D 1.35 and n D 2.35 index quarter-wave layers. Chirping the Bragg
period [17] allows the mirror dispersion to be engineered to achieve a specific
(typically negative) GDD over a certain wavelength range. Unavoidable modulation
in the GDD characteristics of chirped mirror coatings can be addressed by designing
matched pairs of coatings, in which the GDD modulations of one mirror are
cancelled by those of the other mirror [18]. Gires-Tournois interferometer (GTI)
mirrors [19] offer an alternative means of introducing substantial GDD via a Bragg
mirror coating containing an embedded etalon-like structure, and typically offer
several times the GDD of a chirped-mirror coating but over a narrower reflectivity
bandwidth.
Waveguide dispersion is unique to lasers incorporating a waveguide such as an
optical fibre, and may be a significant contributor to net cavity dispersion. Normally
waveguide dispersion is only practical to exploit in fibre lasers [20] or fibre optical
parametric oscillators [21].
@A X ˇk @k A
D Œg l q.t/A C i kC1 C i jAj2 A (4.13)
@z kŠ @t k
k1
4 Advances in Solid-State Ultrafast Laser Oscillators 83
Fig. 4.4 Schematic of a ring resonator showing consecutive operation of dispersion, self-phase-
modulation, gain/loss and saturable absorption operators. A linear resonator may be modelled in a
similar way
where TR is cavity roundtrip time and EL is the saturation energy of the laser gain
medium, which can be related to the saturation fluence of the gain medium and the
laser mode area [23]. The dependence of the gain on the propagation coordinate z
may be important in gain media in which the population inversion varies along their
length, e.g. fibre amplifiers. A parabolic approximation to the gain dispersion can
be treated explicitly by the addition of another term to the ENLSE [24], or it can be
handled implicitly by including g(!) in the numerical model [23].
84 C.G. Leburn and D.T. Reid
The saturable absorber action can be treated in a similar way to the gain, but here
the time-dependence of the absorber must be included. For slow saturable absorbers
the dynamic loss coefficient can be described by [25]:
2 3
Zt ˇ ˇ
6 1 ˇ 0 2 0 ˇ7
q.t/ D q0 exp 4 ˇA.t / dt ˇ5 (4.15)
EA
TR =2
where EA is the saturation energy of the absorber, and the integral term describes
the total pulse energy incident on the absorber until a time t. This expression is
valid when the duration of the pulse is much shorter than the recovery time of the
absorber, which itself is assumed to be shorter than the cavity roundtrip time, TR .
Fast saturable absorbers operate in a regime where the duration of the pulse is
much longer than the recovery time of the absorber. Consequently, such absorbers
respond to the peak power of the pulse, not the integrated energy, and are described
by [24]:
q0
q.t/ D (4.16)
1 C jA.t/j2 =PA
where PA is the saturation power of the absorber.
Both analytical and numerical solutions to the ENLSE have been studied in a
variety of contexts, and for further details the reader is referred to papers dealing
with specific cases [26–29].
Fig. 4.5 Passively mode-locked pulse shaping mechanisms due to gain and loss dynamics of:
(a) a slow saturable absorber, (b) a fast saturable absorber and, (c) solitonic modelocking with a
slow saturable absorber
Figure 4.5 illustrates the three main pulse-shaping mechanisms responsible for
the generation of ultra-short pulses using saturable absorbers. In the case where
the recovery time of the saturable absorber is not much faster than the pulse
86 C.G. Leburn and D.T. Reid
duration, and the upper-state lifetime of the gain medium is short, then the
pulse shaping mechanism cannot shape the trailing edge of the pulse, as the
absorber remains strongly bleached for a duration much longer than the pulse
duration. If the bleaching of the saturable absorber means there is net gain, then
the trailing edge of the pulse is amplified along with any noise, and the pulse
soon becomes unstable. However, if the saturation fluence of the gain medium is
sufficiently low and the upper-state lifetime sufficiently fast, the gain saturates very
quickly after the absorption is bleached. This has become known as dynamic gain
saturation [37]. It allows for a short net-gain window to be formed (as illustrated in
Fig. 4.5a). Due to the long upper state lifetime in gain media such as glass-based or
crystalline solid-state media, dynamic gain saturation cannot be realised. Therefore,
this mechanism is limited to short-lifetime gain media such as semiconductors
or dyes.
The second pulse shaping mechanism (Fig. 4.5b) is known as fast saturable
absorber (FSA) modelocking and requires a saturable absorber with a fast recovery
time so that a suitable short net gain window can be created. For solid-state laser
systems, the most successful FSA-based mode-locking mechanism has been Kerr-
lens modelocking (KLM), whereby the laser crystal itself is encouraged to act like
a fast saturable absorber. In the case of a fast saturable absorber, any spikes in
the random temporal output will suffer low attenuation in the saturable absorber,
whereas low intensity noise will be attenuated to a far greater extent. As a result, it
is possible for a single noise spike to be promoted by a succession of many round
trips into a short, intense pulse. When the intensity of the pulse is large enough to
sufficiently bleach the absorber, then a short window of net gain appears for the
intense central maximum of the pulse and the low intensity wings experience net
loss, and this acts as the pulse shaping mechanism. This makes the assumption that
there is no appreciable variation in the gain during the passage of the pulse, which
requires the upper-state lifetime of the gain medium to be significantly longer than
the absorber recovery time, such as in glass-based solid state lasers where the upper
state lifetime is typically hundreds of microseconds [38]. Of course, there is a limit
to how short the pulse can be, limited by the recovery time of the absorber and the
finite extent of the gain bandwidth.
The third important pulse shaping mechanism in mode-locked lasers is soliton
mode-locking [28, 38, 39]. In this regime, a saturable absorber is responsible for
starting and stabilising mode-locking, however the pulse duration is controlled by
a cancellation of non-linear phase changes and anomalous dispersion contributions
within the cavity. In media that exhibit the nonlinear optical Kerr effect, which is
an intensity dependent modification to the refractive index of the material, high-
intensity pulses produce time dependent phase changes in the material as they
propagate, known as self-phase modulation (SPM). The modulation of the phase
produces a linear frequency chirp of the pulse carrier frequency, which increases
with time, and is therefore identical in effect to normal dispersion. This acts as
a pulse broadening mechanism, and large amounts of SPM can lead to severe
instabilities. However, the effect can be compensated by a suitable amount of
anomalous dispersion, allowing the pulse to propagate without broadening from
4 Advances in Solid-State Ultrafast Laser Oscillators 87
either mechanism. In this case the pulse is a stable solution to a wave equation with
non-linear terms arising from the SPM and anomalous dispersion, which is known
as a soliton solution.
In solitonic modelocking an additional loss mechanism (typically a saturable
absorber) is required to initiate the formation of solitonic pulses from background
noise spikes, and to stabilise the solitonic pulses once they are formed. The short net-
gain windows that are achieved through dynamic gain saturation and FSA are not
necessary for solitonic modelocking because the soliton formation is the dominant
pulse shaping mechanism. This de-couples the saturable absorber from the cavity
and relaxes the constraints of the saturable absorber, thus allowing for a much longer
net-gain window than its dynamic gain saturation and FSA counterparts (Fig. 4.5c).
This greatly relaxes the limitations of the saturable absorber used to initiate the
modelocking. The pulse duration of the soliton solution is comparable for fast and
slow absorbers [28]. The relaxation time of the saturable absorber can be up to
10 times longer than the pulse duration of the laser. This makes semiconductor
saturable absorbers ideal candidates as mode-locking devices.
Fig. 4.6 A sketch of the saturable reflectivity of a typical SESAM under cw illumination, showing
the important parameters of modulation depth, saturation intensity and non-saturable loss
of the Bragg stack (which will be <100%), scattering losses within the material
and excited state absorption. The design and growth characteristics of the device
strongly affect this parameter. For low gain laser systems it is imperative to have
intracavity losses that are as low as possible.
The saturation fluence is defined as the pulse fluence (energy density per unit
area) required to saturate the absorption of a device. The value for the saturation
fluence should be kept relatively small so that the incident fluence needed to saturate
the device does not damage the structure.
The shortest pulse durations that can be generated from a SESAM modelocked
laser system are limited by two factors. One is the bandwidth of the gain material
and the other is the spectral bandwidth of the Bragg mirror of the device. The
bandwidth of SESAM devices is determined by the index contrast of the layers
and the thickness of the layers in the Bragg stack.
In laser systems with long upper-state lifetimes, which are passively mode-
locked with a slow saturable absorber, there is a tendency for instabilities, usually
in the form of Q-switched modelocking [25, 40] (QML). In QML the mode-locked
pulse train output of the laser is modulated within a broader, Q-switched envelope,
typically with kHz frequencies, as opposed to the hundreds of MHz repetition
frequencies of the pulse train. For applications which require constant pulse energies
at high repetition rates, QML is a serious drawback. However, QML has found some
applications in laser machining, as the modulated pulse train has higher peak power
than the continuous wave mode-locking regime.
Honninger et al. [40] showed that the criterion for stability against QML is
given by:
ˇ
dR.EP / ˇˇ TR TR EP
EP < rD C (4.17)
dEP ˇE P L L ESAT;L
4 Advances in Solid-State Ultrafast Laser Oscillators 89
where EP is the energy of a modelocked pulse in the cavity, R(EP ) is the nonlinear
reflectivity, TR is the cavity round trip time, L is the upper-state lifetime of the
laser medium. ESAT,L is the saturation energy of the gain, which is defined as the
product of saturation fluence (FSAT,L D h /(m L )) and the effective laser mode area
in the gain medium. r is the pump parameter, defined as r D 1 C P/PSAT,L where
PSAT,L is the saturation power, given by ESAT,L / L A and describes the number of times
over threshold at which the laser is operating. Eq. (4.17) shows that modelocking
can be achieved when EP Esat , when the laser is operated far beyond threshold,
r 1, or the cavity round trip time is long. These constraints are very significant
for the choice of saturable absorber, especially in the case of SESAMs, for use with
a particular laser system. For example, in a passively mode-locked system using
an intracavity SESAM, a high repetition-rate laser with low intracavity power will
require a SESAM with low saturation fluence in order to minimise the LHS of Eq.
(4.17). However, in the case of lasers with very high intracavity pulse energies,
the saturation fluence needs to be large enough to avoid over-saturation of the
SESAM, which may lead to multiple pulsing, where the mode-locking dynamics
result in two lower power, longer duration pulses [41]. The SESAM also needs
to be resistant to intensity related effects, such as non-linear optical effects like
two-photon absorption. A simple way of correctly matching the intracavity pulse
energy, EP , to the saturation fluence of the SESAM, Fsat , is by controlling the spot-
size of the intracavity beam on the absorber. However, there is obviously a limit to
the parameter range using the spot-size alone. The absorption recovery time of the
absorptive transition is also important, requiring consideration of the mode-locking
mechanism for the particular gain medium as to whether the saturable absorber is
required to be fast or slow.
Several researchers have been able to develop a recipe for the design and production
of tailor-made SESAMs for insertion into specific systems. It should be noted that
semiconductor saturable absorber mirrors (SESAMs) and saturable Bragg reflectors
(SBRs) both have an intensity-dependent reflectivity. Although SBRs and SESAMs
could be classed as separate devices, they are in fact very similar in structure and
in the way they operate. Differences between the two types of devices are usually
attributable to the way they are grown and processed.
The first semiconductor saturable absorber devices used to modelock solid-
state lasers date back to 1992 [2]. Since then, they have become very popular
intracavity devices that can be exploited in lasers to generate picosecond and
femtosecond pulses. Epitaxial growth techniques such as molecular beam epitaxy
(MBE) and metal-organic chemical vapour deposition (MOCVD) can be used to
produce devices with very well defined parameters. Being able to grow devices
that have specific and pre-determined properties makes them very attractive as
modelocking elements. These devices consist typically of alternate layers of high
90 C.G. Leburn and D.T. Reid
Fig. 4.7 Schematic diagram of a typical SESAM structure. The device consists of a substrate onto
which a Bragg stack is grown. AlAs and GaAs have a sufficient refractive index contrast and almost
identical lattice constant to produce very high quality DBR structures. The last few layers contain a
quantum well or quantum dot absorber. Material type and placement of the absorber layer is crucial
and strongly determines the performance and losses of the device
and low index semiconductor materials, which act as a Bragg reflector with a thin
saturable absorber layer near the top of the device (Fig. 4.7). As already discussed,
the modulation that is necessary for modelocking is brought about via bleaching of
the absorber layer when sufficiently intense intracavity pulses are incident on the
device.
The absorber layer consists of either quantum-well or quantum-dot material,
depending on the type of laser that is being modelocked and its wavelength of
operation. The first semiconductor device to be used as a modelocking element was
a GaAs/GaAlAs device that had 75 GaAs quantum wells embedded in the absorber
layer. This device was used in a Ti:sapphire laser and assisted in the formation
and stabilisation of picosecond pulses [42]. This semiconductor approach was then
transferred to a Nd:YLF laser with similar results [2]. This new technique was
termed resonant passive modelocking but was the basis of SESAM modelocking
as we know it today.
Quantum-dot systems have atom-like density of states, which allows for the
growth of devices that have tuneable optical properties, depending on the size and
density of the quantum dots within the absorber layer. Although QD technology
is less well established than its QW counterpart, it has been recognised that QDs
have several advantages over QWs for use in SESAMs operating in the near-
IR. They have ultrafast absorption recovery times [43] and large inhomogeneous
line-widths relative to QWs. This makes them ideal candidates for laser systems
where large absorption bandwidths are required. Typically quantum-dot devices
also exhibit lower saturation fluences when compared with quantum-well devices
that have a similar modulation depth. In 1999, the first quantum-dot devices were
used to modelock semiconductor edge emitter devices [44]. Following on from this
result, Garnach et al. demonstrated the first QD-SESAM in 2001 [45]. Rafailov
et al. then went onto demonstrate the fast recovery dynamics of QD-devices [46]
and pulse durations of 86 fs have been demonstrated at 1.3 m with QD-SESAMs
[47]. A recent review of ultrashort-pulse laser modelocked by QD-based saturable
absorbers can be found in [48].
4 Advances in Solid-State Ultrafast Laser Oscillators 91
This last section will briefly look at a small selection of SESAM modelocked solid-
state laser sources. This is by no means an exhaustive list of the most recent research
in the field of ultrafast laser science. In fact we have restricted this section in order to
highlight advances in: high power systems, high pulse repetition rate laser systems,
2 m laser sources, and SESAMs that are now bringing added functionality to the
systems they are incorporated into.
Fig. 4.8 Reflectivity and GVD characteristics for the resonant-like GaInNAs QW structure
confirmed that the resonant nature of the device led to a modulation depth of 7%
at 1,330 nm. This would typically make the device unsuitable for operation in
a Cr4C :forsterite laser system, but the low insertion losses of the device and its
high reflectivity characteristics at shorter wavelengths (around 1,270 nm) allowed
continuous-wave output powers of 1.1 W to be generated by the laser when the
device was inserted into the laser cavity.
Figure 4.8 illustrates the group velocity dispersion characteristics of the SESAM.
This is a typical plot for a resonant-like device compared against the anti-resonant
case, illustrated in Fig. 4.3, where the GVD is flat across a large proportion of the
bandwidth of the device. From Fig. 4.8 it can be seen that GVD values between
0 and 1,100 fs2 per bounce can be generated by the device when operating at
wavelengths between 1,260 and 1,310 nm respectively. The positive dispersion
generated from the Cr4C :forsterite crystal used in this experimental assessment was
estimated to be 1,050 fs2 per round trip [74].
Given the relatively large modulation depth of the SESAM, the laser was
operated away from its higher gain region (1,270 nm) to ensure the initiation of
modelocking. This was achieved by inserting a fused silica prism into the cavity
to force the laser to operate at a wavelength were modelocking was self-starting.
When modelocked with the 1% output coupler in place, output powers of 270 mW
could be extracted from the system. The laser supported an optical spectrum
of œFWHM D 12 nm and the autocorrelation data implied near-transform-limited
pulses of 150 fs. An RF spectrum analyser was used to confirm the pulse repetition
frequency of 652 MHz. This piece of work illustrates how SESAM devices are now
becoming multi-functional components within solid-state laser systems. Not only
can they start and maintain the formation of modelocked pulses, they can also relax
the need for GTIs or chirped mirrors.
4 Advances in Solid-State Ultrafast Laser Oscillators 95
Researchers have gone a step further to look at ways of disturbing the modelocking
dynamics of the SESAM to force the laser between cw and cw-modelocked
regimes. Zolotovskaya et al. [75] developed a passively mode-locked Cr:forsterite
laser incorporating a dc-voltage-controlled p-n junction quantum-dot semiconductor
saturable absorber mirror. With this system output pulse durations could be varied
from 17.4 ps to 6.4 ps by applying a dc voltage between 0 and 4.5 V across
the SESAM. In this work the QD SESAM structure comprised of a 300-nm Si-
doped GaAs buffer layer deposited on an n-type GaAs (100) substrate, an n-doped
distributed Bragg reflector (DBR), and a long GaAs microcavity, with a design
wavelength of 1,250 nm incorporating a p-doped GaAs cap layer. Metal contacts
were deposited on the p- and n-side of the device. The DBR stack consisted of 33
pairs of Si-doped GaAs–AlGaAs layers, with a stop band centred at a wavelength of
1,260 nm. The QD saturable absorber region was formed in seven groups consisting
of three layers of InAs–InGaAs QDs, with 32-nm-thick GaAs spacer layers in
between. The groups of QDs were positioned at the antinodes of the electric field
standing wave using GaAs spacers. Stable modelocked operation of the Cr:forsterite
laser was attained across the whole bias range with pronounced pulse shortening
observed with increasing applied reverse voltage. These workers were able to infer
that the reduction in the modelocked pulse duration was attributed to a decrease in
the saturable absorber recovery time under reverse biasing.
Several researchers have also looked at ways of disturbing the modelocking
dynamics of the SESAM to force the laser between cw and cw-modelocked regimes
[76–78] by optically disturbing the SESAM. Currently, femtosecond laser systems
operate in a fixed pulsed regime. There are several applications, particularly in
the area of biophotonics, that would profit from a system that provided more
flexible characteristics. These include experiments involving optical transfection
of trapped cells [79] and optical tweezing [80] where the laser system can be
simply and effectively switched between the continuous wave (cw) state and the
cw mode-locked state to enable different functionalities at specific times. Actively
modelocked lasers have demonstrated this type of regime change, but the electronic
modulators required for this complicate the system and add cost to the overall
setup. Mechanical methods have been implemented to switch between cw and
femtosecond operation of passively modelocked systems however switching times
have been relatively long (ms to s) [81, 82]. Savitski et al. [76] demonstrated a
picosecond Nd:YVO4 system that could switch between the cw and cw mode-
locked state, with a switching time 50 s, however picosecond pulses are not
suitable for most biological applications [83]. Savitski et al. have also demonstrated
this switching technique using a SBR device within a femtosecond Ti:sapphire laser
[77], and performed a detailed study of the changes in the nonlinear optical response
of the saturable absorber in the presence of a continuous-wave optical pump source.
They also went on to develop a model of the heating dynamics within the device
under optical pumping and show that fast optically induced heating could be used
96 C.G. Leburn and D.T. Reid
Fig. 4.9 Schematic diagram of the set-up used to switch the Cr:forsterite laser between the cw and
cw-modelocked regime
to externally control the behaviour of the SBR. This external optical pumping of
the SBR allowed for rapid (75 s) switching between CW and femtosecond pulse
operation of a passively mode-locked Ti:sapphire laser.
Work has also been carried out at the biologically relevant wavelength region of
1,300 nm [84] with a Cr4C :forsterite laser system which contained a SESAM that
was optically pumped from outside the cavity, in order to switch the laser from a
state where the laser produced sub-300 fs pulses to a state where the laser operated
in the cw regime. Figure 4.9 illustrates the set up that was used for this experiment.
A standard 4-mirror asymmetric, astigmatically compensated Z-fold cavity was
constructed and a Yb:fibre laser was used as the pump source. This pump beam
passed through a telescope and focusing lens into the Cr4C :forsterite crystal (pump
spot radius, wp 25 m). The 11.6-mm-long Brewster-cut Cr4C :forsterite rod had a
small-signal pump absorption coefficient of 1.3 cm1 , and was mounted in a water-
cooled copper jacket that was maintained at a temperature of 12 ı C. The two curved
folding mirrors in the cavity had radii of curvature of 100 mm and 75 mm for the
long and short arms respectively, high-reflectivity (HR) coatings for wavelengths in
the 1,250–1,370 nm range and high transmission at the pump laser wavelength. A
wedged 0.5% output coupler was located at the end of the long arm of the cavity
and either a HR mirror or SBR structure terminated the short arm of the cavity (spot
size 110 m radius). Two silica prisms were incorporated into the long arm of the
laser resonator to compensate for the positive GVD.
The GaInNAs anti-resonant quantum well SESAM reflectivity characteristics
exhibited a >110 nm-stop-band Bragg reflectivity centred at 1,320 nm and a
photoluminescence peak at 1,340 nm, as detailed in [57]. To control the mode
of operation of the laser, a commercially available external diode laser (EDL)
4 Advances in Solid-State Ultrafast Laser Oscillators 97
Fig. 4.10 Typical temporal response of rapid state-switching of the Cr4C :forsterite laser when
(a) the EDL is activated and (b) deactivated
There is now a push to extend the wavelengths of operation for SESAM devices.
As already been discussed, pushing towards longer wavelengths requires thicker
98 C.G. Leburn and D.T. Reid
layers within the SESAM structure, which leads to increased strain and losses.
To remove the strain within longer wavelength devices researchers have investi-
gated several different schemes (including carbon nanotubes and InGaAsSb-based
saturable absorbers, which produced 9.7 ps and 3.3 ps pulses near 1,950 nm
and 2,060 nm, respectively), but the most promising technique has involved ion-
bombardment of the SESAM devices.
As an example, Lagatsky et al. [85] have used a GaSb/AlAsSb SESAM structure
for the initiation and stabilization of passive modelocking in a Tm,Ho:KYW laser
at 2 m. The distributed Bragg reflector had a high reflectivity at 1,930–2,150 nm
and two InGaAsSb quantum wells as the absorber. In their original work Lagatsky
et al. demonstrated 3.3 ps pulses centred on 2,057 nm from a Tm,Ho:KYW
laser using this as-grown SESAM [86]. They later discovered that no significant
variations in pulse durations or spectral widths were observed for different values of
intracavity pulse energy or net cavity group-velocity dispersion, and they attributed
this restrictive mode-locking regime to the relatively long recovery time (100 ps)
and low self-amplitude modulation in the as-grown SESAM. To decrease the carrier
recombination time [87, 88], the SESAM samples were irradiated with 4 MeV AsC
ions at doses ranging from 5 1010 cm2 to 5 1011 cm2 . As the implantation dose
increased, a damage-induced reduction of the overall reflectivity was observed.
With this ion-bombarded SESAM, stable modelocking was realised at 2,055 nm
with a maximum average output of 130 mW and a pulse repetition frequency of
118 MHz. The pulse durations varied from 2.64 ps at the modelocking threshold to
570 fs at the maximum output power of 130 mW.
These workers are concentrating on the development of diode- pumped Tm,Ho-
codoped femtosecond lasers using SESAMs having better optimized macroparame-
ters for efficient femtosecond modelocking in the 2 m spectral region and illustrate
how the detailed design of SESAM devices are still yielding significant advances in
ultrashort pulse solid-state laser sources.
4.5 Conclusions
The examples of modern ultrafast oscillators discussed here illustrate how higher
repetition rate multi-GHz laser systems have a requirement for SESAMs with faster
recovery times and lower saturation fluences. The longer upper-state lifetime of
contemporary gain media like Yb:KYW also makes it harder to satisfy the stability
criterion for SESAM modelocking (Eq. 4.17). Consequently, the role of both QW
and QD SESAMs will become ever more important as the need for achieving
broader bandwidth and faster recovery times gets stronger. The push towards mid-IR
ultrafast operation through new gain materials will only be possible with the advent
of new SESAM material systems too. Application-led development will also play a
role in determining the path of ultrafast science. This will include, for example, areas
like low-repetition-rate, high-energy mid-IR lasers for high harmonic generation,
and high-repetition-rate, high-peak-power, short-duration pulses for GHz frequency
combs.
4 Advances in Solid-State Ultrafast Laser Oscillators 99
References
1. E.P. Ippen, C.V. Shank, A. Dienes, Passive modelocking of the CW dye laser. Appl. Phys. Lett.
21, 348–350 (1972)
2. U. Keller, D.A.B. Miller, G.D. Boyd, T.H. Chiu, J.F. Ferguson, M.T. Asom, Solid-state low-loss
intracavity saturable absorber for Nd:YLF lasers: an antiresonant semiconductor Fabry-Perot
saturable absorber. Opt. Lett. 17, 505 (1992)
3. D.E. Spence, P.N. Kean, W. Sibbett, 60-femtosecond pulse generation from a self-mode-locked
Ti-sapphire laser. Opt. Lett. 16, 42–44 (1991)
4. K. Tamura, L.E. Nelson, H.A. Haus, E.P. Ippen, Soliton versus nonsoliton operation of fiber
ring lasers. App. Phys. Lett. 64, 149–151 (1994)
5. K.J. Blow, D. Wood, Mode-locked lasers with nonlinear external cavities. J. Opt. Soc. Am. B
5, 629 (1988)
6. P.N. Kean, X. Zhu, D.W. Crust, R.S. Grant, N. Langford, W. Sibbett, Enhanced modelocking
of color-center lasers. Opt. Lett. 14, 39–41 (1989)
7. X. Zhu, P.N. Kean, W. Sibbett, Coupled-cavity modelocking of a KCl:Tl laser using an erbium-
doped optical fiber. Opt. Lett. 14, 1192–1194 (1989)
8. J. Mark, L.Y. Liu, K.L. Hall, H.A. Haus, E.P. Ippen, Femtosecond pulse generation in a laser
with a nonlinear external resonator. Opt. Lett. 14, 48–50 (1989)
9. F. Ouellette, M. Piché, Pulse shaping and passive mode-locking with a nonlinear Michelson
interferometer. Opt. Comm. 60, 99–103 (1986)
10. S.Y. Set, H. Yaguchi, Y. Tanaka, M. Jablonski, Ultrafast fiber pulsed lasers incorporating carbon
nanotubes. IEEE J. Sel. Top. Quantum. Electron. 10, 137–146 (2004)
11. R.L. Fork, C.H. Cruz, P.C. Becker, C.V. Shank, Compression of optical pulses to six
femtoseconds by using cubic phase compensation. Opt. Lett. 12, 483–485 (1987)
12. B.E. Lemoff, C.P.J. Barty, Cubic-phase-free dispersion compensation in solid-state ultrashort-
pulse lasers. Opt. Lett. 18, 57–59 (1993)
13. R.E. Sherriff, Analytic expressions for group-delay dispersion and cubic dispersion on arbitrary
prism sequences. J. Opt. Soc. Am. B 15, 1224–1230 (1998)
14. E.B. Treacy, Compression of picosecond light pulses. Phys. Lett. 28a, 34–35 (1968)
15. E.B. Treacy, Pulse compression with diffraction gratings. IEEE J. Quantum. Electron. 5,
454–458 (1969)
16. M. Yamashita, M. Ishikawa, K. Torizuka, T. Sato, Femtosecond-pulse laser chirp compensated
by cavity-mirror dispersion. Opt. Lett. 11, 504 (1986)
17. R. Szipocs, K. Ferencz, C. Spielmann, F. Krausz, Chirped multilayer coatings for broadband
dispersion control in femtosecond lasers. Opt. Lett. 19, 201 (1994)
18. U. Morgner, F.X. Kärtner, T.R. Schibli, P. Wagenblast, J.G. Fujimoto, E.P. Ippen, V. Scheuer,
G. Angelow, T. Tschudi, M. Lederer, A. Boiko, B. Luther-Davies, Ultrabroadband double-
chirped mirror pairs covering one octave of bandwidth, in Conference on Lasers and Electro-
Optics (CLEO), 39, OSA Trends in Optics and Photonics Series (Optical Society of America,
Washington, DC, 2000), paper CThE 5
19. F. Gires, P. Tournois, Interferometre utilisable pour la compression d’impulsions lumineuses
modulees en frequence. C. R. Acad. Sci. Paris 258, 6112 (1964)
20. A. Isomäki, O.G. Okhotnikov, Femtosecond soliton mode-locked laser based on ytterbium-
doped photonic bandgap fiber. Opt. Express 14, 9238–9243 (2006)
21. J.E. Sharping, M.A. Foster, A.L. Gaeta, J. Lasri, O. Lyngnes, K. Vogel, Octave-spanning, high-
power microstructure-fiber-based optical parametric oscillators. Opt. Express 15, 1474–1479
(2007)
22. For details on the split-step Fourier method see, for example, G.P. Agrawal, Nonlinear Fiber
Optics, 2nd edn. (Academic, San Diego, 1995)
23. F.Ö. Ilday, J.R. Buckley, W.G. Clark, F.W. Wise, Self-similar evolution of parabolic pulses in
a laser. Phys. Rev. Lett. 92, 213902 (2004)
24. H.A. Haus, A theory of fast saturable absorber modelocking. J. Appl. Phys. 46, 3049 (1975)
100 C.G. Leburn and D.T. Reid
25. F.X. Kärtner, L.R. Brovelli, D. Kopf, M. Kamp, I.G. Calasso, U. Keller, Control of solid-state
laser dynamics by semiconductor devices. Opt. Eng. 34, 2024 (1995)
26. H.A. Haus, J.G. Fujimoto, E.P. Ippen, Analytic theory of additive pulse and Kerr lens mode-
locking. IEEE J. Quantum Electron. 28, 2086–2096 (1992)
27. H.A. Haus, K. Tamura, L.E. Nelson, E.P. Ippen, Stretched-pulse additive pulse mode-locking
in fiber ring lasers: theory and experiment. IEEE J. Quantum Electron. 31, 591 (1995)
28. F.X. Kärtner, U. Keller, Stabilization of soliton-like pulses with a slow saturable absorber. Opt.
Lett. 20, 16 (1995)
29. W.H. Renninger, A. Chong, F.W. Wise, Dissipative solitons in normal-dispersion fiber lasers.
Phys. Rev. A 77, 023814 (2008)
30. M. DiDomenico, J.E. Geusic, H.M. Marcos, R.G. Smith, Generation of ultrashort optical pulses
by modelocking the YAIG: Nd laser. Appl. Phys. Lett. 8, 180–183 (1966)
31. H.W. Mocker, R.J. Collins, Mode competition and self-modelocking effects in a Q-switched
ruby laser. Appl. Phys. Lett. 7, 270–273 (1965)
32. A.J. DeMaria, D.A. Stetser, H. Heynau, Self mode-locking of lasers with saturable absorbers.
Appl. Phys. Lett. 8, 174–176 (1966)
33. H. Statz, C.L. Tang, Phase locking of modes in lasers. J. Appl. Phys. 36, 3923–3927 (1965)
34. K. Wundke, S. Potting, J. Auxier, A. Schulzgen, N. Peyghambarian, N.F. Borrelli, PbS
quantum-dot-doped glasses for ultrashort-pulse generation. Appl. Phys. Lett. 76, 10 (2000)
35. S. Yamashita, Y. Inoue, S. Maruyama, Y. Murakami, H. Yaguchi, M. Jablonski, S.Y. Set,
Saturable absorbers incorporating carbon nanotubes directly synthesized onto substrates and
fibers and their application to mode-locked fiber lasers. Opt. Lett. 29, 1581 (2004)
36. R. Paschotta, U. Keller, Passive modelocking with slow saturable absorbers. Appl. Phys. B 73,
653 (2001)
37. G.H.C. New, Pulse evolution in mode-locked quasi-continuous lasers. IEEE J. Quantum
Electron. 10, 115 (1974)
38. O. Svelto, Principles of Lasers, 4th edn. (Plenum Press, New York, 1998)
39. F.X. Kärtner, I.D. Jung, U. Keller, Soliton mode-locking with saturable absorbers. J. Sel. Top.
in Quantum Electron. 2, 540 (1996)
40. C. Honninger, R. Paschotta, F. Morier-Genoud, M. Moser, U. Keller, Q-switching stability
limits of continuous-wave passive modelocking. J. Opt. Soc. Am. B 16, 46 (1999)
41. U. Keller, K.J. Weingarten, F.X. Kärtner, D. Kopf, B. Braun, I.D. Jung, R. Fluck, C. Honninger,
N. Matuschek, J. Aus der Au, Semiconductor saturable absorber mirrors (SESAM’s) for
femtosecond to nanosecond pulse generation in solid-state lasers. IEEE J. Sel. Top. Quantum
Electron. 2, 435 (1996)
42. U. Keller, W.H. Knox, H. Roskos, Coupled-cavity resonant passive mode-locked Ti:sapphire
laser. Opt. Lett. 23, 1377–1379 (1990)
43. P. Borri, S. Schneider, W. Langbein, U. Woggon, A.E. Zhukov, V.M. Ustinov, N.N. Ledentsov,
Z.I. Alferov, D. Ouyang, D. Bimberg, Ultrafast carrier dynamics and dephasing in InAs
quantum-dot amplifiers emitting near 1.3- m-wavelength at room temperature. Appl. Phys.
Lett. 79, 2633 (2001)
44. O. Qasaimeh, W.-D. Zhou, J. Phillips, S. Krishna, P. Bhattacharya, M. Dutta, Bistability and
self-pulsation in quantum-dot lasers with intracavity quantum-dot saturable absorbers. Appl.
Phys. Lett. 74, 1654 (1999)
45. A. Garnache, S. Hoogland, A.C. Tropper, J.M. Gerard, V. Thiery-Mieg, J.S. Roberts,
Pico-second passively modelocked surface-emitting laser with self-assembled semiconductor
quantum dot absorber, in Proceedings of CLEO/Europe-EQEC, Post–deadline Paper (2001)
46. E.U. Rafailov, S.J. White, A.A. Lagatsky, A. Miller, W. Sibbett, D.A. Livshits, A.E. Zhukov,
V.M. Ustinov, Fast quantum-dot saturable absorber for passive mode-locking of solid-state
lasers. IEEE Photonics Technol. Lett. 16, 2439 (2004)
47. M.P. Lumb, E. Clarke, E. Harbord, P. Spencer, R. Murray, F. Masia, P. Borri, W. Langbein,
C.G. Leburn, C. Jappy, N.K. Metzger, C.T.A. Brown, W. Sibbett, Ultrafast absorption recovery
dynamics of 1300 nm quantum dot saturable absorber mirrors. Appl. Phys. Lett. 95, 041101
(2009)
4 Advances in Solid-State Ultrafast Laser Oscillators 101
48. A.A. Lagatsky, C.G. Leburn, C.T.A. Brown, W. Sibbett, S.A. Zolotovskaya, E.U. Rafailov,
Ultrashort-pulse lasers passively modelocked by quantum-dot-based saturable absorbers. Prog.
Quantum Elecron. 34, 1–45 (2010)
49. S. Tsuda, W.H. Knox, S.T. Cundiff, W.Y. Jan, J.E. Cunningham, Mode-locking ultrafast solid-
state lasers with saturable Bragg reflectors. IEEE J. Sel. Top. Quantum Electron. 2, 454–464
(1996)
50. B. Craig, A. Krueger, Saturable Bragg reflectors simplify modelocking. Laser Focus World 36,
227–228 (2000)
51. P.T. Guerreiro, S. Ten, E. Slobodchikov, Y.M. Kim, J.C. Woo, N. Peyghambarian, Self-starting
mode-locked Cr:forsterite laser with semiconductor saturable Bragg reflector. Opt. Comm. 136,
27–30 (1997)
52. H.D. Sun, G.J. Valentine, R. Macaluso, S. Calvez, D. Burns, M.D. Dawson, T. Jouhti, M. Pessa,
Low-loss 1.3- m GaInNAs saturable Bragg reflector for high- power picosecond neodymium
lasers. Opt. Lett. 27, 2124–2126 (2002)
53. O.G. Okhotnikov, T. Jouhti, J. Konttinen, S. Karirinne, M. Pessa, 1.5- m monolithic GaInNAs
semiconductor saturable-absorber modelocking of an erbium fiber laser. Opt. Lett. 28, 364–366
(2003)
54. V. Liverini, S. Schon, R. Grange, M. Haiml, S.C. Zeller, U. Keller, Low- loss GaInNAs
saturable absorber modelocking a 1.3 m solid-state laser. Appl. Phys. Lett. 84, 4002–4004
(2004)
55. Z. Zhang, K. Torizuka, T. Itatani, K. Kobayashi, T. Sugaya, T. Nakagawa, Self-starting mode-
locked femtosecond forsterite laser with a semiconductor saturable-absorber mirror. Opt. Lett.
22, 1006–1008 (1997)
56. Z. Zhang, K. Torizuka, T. Itatani, K. Kobayashi, T. Sugaya, T. Nakagawa, H. Takahashi, Broad-
band semiconductor saturable-absorber mirror for a self- starting mode-locked Cr:forsterite
laser. Opt. Lett. 23, 1465–1467 (1998)
57. A. McWilliam, A.A. Lagatsky, C.G. Leburn, P. Fischer, C.T.A. Brown, G.J. Valentine, A.J.
Kemp, S. Calvez, D. Burns, M.D. Dawson, M. Pessa, W. Sibbett, Low-loss GaInNAs saturable
Bragg reflector for mode-locking of a femtosecond Cr:forsterite-laser. IEEE Photon. Technol.
Lett. 17, 2292–2294 (2005)
58. T. Südmeyer, S.V. Marchese, S. Hashimoto, C.R.E. Baer, G. Gingras, B. Witzel, U. Keller,
Femtosecond laser oscillators for high-field science. Nat. Photon. 2, 599 (2008)
59. S.V. Marchese, C.R.E. Baer, A.G. Engqvist, S. Hashimoto, D.J.H.C. Maas, M. Golling, T.
Südmeyer, U. Keller, Femtosecond thin disk laser oscillator with pulse energy beyond the 10-
microjoule level. Opt. Express 16, 6397 (2008)
60. J. Neuhaus, D. Bauer, J. Zhang, A. Killi, J. Kleinbauer, M. Kumkar, S. Weiler, M. Guina, D.H.
Sutter, T. Dekorsy, Subpicosecond thin-disk laser oscillator with pulse energies of up to 25.9
microjoules by use of an active multipass geometry. Opt. Express 16, 20530 (2008)
61. C.R.E. Baer, C. Kränkel, C.J. Saraceno, O.H. Heckl, M. Golling, R. Peters, K. Petermann, T.
Südmeyer, G. Huber, U. Keller, Femtosecond thin-disk laser with 141 W of average power.
Opt. Lett. 35, 2302–2304 (2010)
62. L. Krainer, R. Paschotta, S. Lecomte, M. Moser, K.J. Weingarten, U. Keller, Compact
Nd:YVO4 lasers with pulse repetition rates up to 160 GHz. IEEE J. Quantum Electron. 38,
1331 (2002)
63. A.E.H. Oehler, T. Südmeyer, K.J. Weingarten, U. Keller, 100 GHz passively mode-locked
Er:Yb:glass laser at 1.5 m with 1.6-ps pulses. Opt. Express 16, 21930 (2008)
64. A. Oehler, M. Stumpf, S. Pekarek, T. Südmeyer, K. Weingarten, U. Keller, Picosecond diode-
pumped 1.5 m Er,Yb:glass lasers operating at 10–100 GHz repetition rate. Appl. Phys. B 99,
53 (2010)
65. A. Bartels, Passively mode-locked 10GHz femtosecond Ti:sapphire laser. Opt. Lett. 33,
1905–1907 (2008)
66. C.G. Leburn, A.A. Lagatsky, C.T.A. Brown, W. Sibbett, Femtosecond Cr4C :YAG laser with
4 GHz pulse repetition rate. Electron. Lett. 40, 805–806 (2004)
102 C.G. Leburn and D.T. Reid
67. A.J. Kemp, B. Stormont, B. Agate, C.T.A. Brown, U. Keller, W. Sibbet, Gigahertz
repetition-rate from directly diode-pumped femtosecond Cr:LiSAF laser. Electron. Lett. 37,
1457–1458 (2001)
68. D. Li, U. Demirbas, J.R. Birge, G.S. Petrich, L.A. Kolodziejski, A. Sennaroglu, F.X. Kärtner,
J.G. Fujimoto, Diode-pumped passively mode-locked GHz femtosecond Cr:LiSAF laser with
kW peak power. Opt. Lett. 35, 1446–1448 (2010)
69. P. Wasylczyck, P. Wnuk, C. Radzewicz, Passively mode-locked, diode-pumped Yb:KYW
femtosecond oscillator with a 1 GHz repetition rate. Opt. Express 17, 5630–5635 (2009)
70. S. Pekarek, C. Fiebig, M.C. Stumpf, A.E.H. Oehler, K. Paschke, G. Erbert, T. Südmeyer, U.
Keller, Diode-pumped gigahertz femtosecond Yb:KGW laser with a peak power of 3.9kW.
Opt. Express 18, 16320–16326 (2010)
71. T.-Z. Liu, F.X. Kärtner, J.G. Fujimoto, C.-K. Sun, Multiplying the repetition-rate of passively
modelocked femtosecond lasers by an intra-cavity flat surface with low reflectivity. Opt. Lett.
30, 439–441 (2005)
72. M.P. Lumb, P.N. Stavrinou, E.M. Clarke, R. Murray, C.G. Leburn, C. Jappy, N.K. Metzger,
C.T.A. Brown, W. Sibbett, Dispersionless saturable absorber mirrors with large modulation
depths and low saturation fluences. Appl. Phys. B 97, 53–60 (2009)
73. C.G. Leburn, W. Lu, S. Vetter, M.D. Dawson, C.T.A. Brown, J.S. Harris, S. Calvez,
W. Sibbett, 650MHz-prf-femtosecond Cr4C :forsterite laser with dispersion-compensating
GaInNAs SESAM, in Conference on Lasers and Electro-optics, Baltimore, Paper JWA76
(2011)
74. V. Yanovsky, Y. Pang, F. Wise, B.I. Minkov, Generation of 25-fs pulses from a self-mode-
locked Cr:forsterite laser with optimized group-delay dispersion. Opt. Lett. 18, 1541–1543
(1993)
75. S.A. Zolotovskaya, K.G. Wilcox, A. Abdolvand, D.A. Livshits, E.U. Rafailov, Electronically
controlled pulse duration passively mode-locked Cr:forsterite laser. IEEE Photon. Technol.
Lett. 21, 1124–1126 (2009)
76. V.G. Savitski, D. Burns, S. Calvez, Optically-pumped saturable absorber for fast switching
between continuous-wave and passively mode-locked regimes of a Nd:YVO4 laser. Opt.
Express 17, 5373–5378 (2009)
77. V.G. Savitski, A.J. Kemp, S. Calvez, D. Burns, Optically pumped saturable Bragg reflectors:
nonlinear spectroscopy and application in ultrafastlasers. IEEE J. Quantum Electron. 46,
1650–1655 (2010)
78. C.G. Leburn, C.T.A. Brown, W. Sibbett, Optically-pumped SESAM for fast switching between
continuous wave and passively modelocked regimes of a femtosecond pulse Cr4C :forsterite
Laser, in Ultrafast Phenomena XVII, Proceedings of the 17th International Conference,
Snowmass, CO, 2010, pp. 715–717
79. D. Stevenson, B. Agate, X. Tsampoula, P. Fischer, C.T.A. Brown, W. Sibbett, A. Riches, F.
Gunn-Moore, K. Dholakia, Femtosecond optical transfection of cells: viability and efficiency.
Opt. Express 14, 7125 (2006)
80. B. Agate, C.T.A. Brown, W. Sibbett, K. Dholakia, Femtosecond optical tweezers for in-situ
control of two-photon fluorescence. Opt. Express 12, 3011 (2004)
81. J. Ando, G. Bautista, N. Smith, K. Fujita, V.R. Daria, Optical trapping and surgery of living
yeast cells using a single laser. Rev. Sci. Instrum. 79, 103705-103705-5 (2008)
82. A. McWilliam, A.A. Lagatsky, C.T.A. Brown, W. Sibbett, A.E. Zhukov, V.M. Ustinov, A.P.
Vasil’ev, E.U. Rafailov, Quantum-dot-based saturable absorber for femtosecond mode-locked
operation of a solid-state laser. Opt. Lett. 31, 1444–1446 (2006)
83. D.J. Stevenson, F.J. Gunn-Moore, P. Campbell, K. Dholakia, Single cell optical transfection. J.
Royal Soc. 7, 863–871 (2010)
84. J.H. Lee, S.Y. Chen, C.H. Yu, S.W. Chu, L.F. Wang, C.K. Sun, B.L. Chiang, Noninvasive
in vitro and in vivo assessment of epidermal hyperkeratosis and dermal fibrosis in atopic
dermatitis. J. Biomed. Opt. 14, 15 (2009)
4 Advances in Solid-State Ultrafast Laser Oscillators 103
85. A.A. Lagatsky, F. Fusari, S. Calvez, S.V. Kurilchik, V.E. Kisel, N.V. Kuleshov, M.D. Dawson,
C.T.A. Brown, W. Sibbett, Femtosecond pulse operation of a Tm, Ho-codoped crystalline laser
near 2 m. Opt. Lett. 35, 172–174 (2010)
86. A.A. Lagatsky, F. Fusari, S. Calvez, G.A. Gupta, V.E. Kisel, N.V. Kuleshov, C.T.A. Brown,
M.D. Dawson, W. Sibbett, Passive modelocking of a Tm, Ho:KY(WO4)2 laser around 2 m.
Opt. Lett. 34, 2587 (2009)
87. E.L. Delpon, J.L. Oudar, N. Bouche, R. Raj, A. Shen, N. Stelmakh, J.M. Lourtioz, Ultrafast
excitonic saturable absorption in ion-implanted InGaAs/InAlAs multiple quantum wells. Appl.
Phys. Lett. 72, 759 (1998)
88. J.W. Tomm, V. Strelchuk, A. Gerhardt, U. Zeimer, M. Zorn, H. Kissel, M. Weyers, J. Jimenez,
Properties of AsC -implanted and annealed GaAs and InGaAs quantum wells: structural and
band-structure modifications. J. Appl. Phys. 95, 1122 (2004)
Chapter 5
Ultrafast Quantum Control in Atoms
and Molecules
Philip H. Bucksbaum
5.1 Introduction
The interaction of short wavelength and strong field lasers with atoms and small
molecules can be described in the language of a control problem: That is, the
electromagnetic field produced by the laser exerts control over the quantum
evolution of the system under study. By exerting this control in a predetermined
way, the system will evolve to some desired final state, such as dissociative products
or photoelectrons in a particular direction. This chapter uses this point of view to
introduce the subject of ultrafast short wavelength strong field interactions. This
subject is quite broad and deep, and this chapter can only discuss a few interesting
aspects of it. Readers who wish to learn more should consult some of the review
papers and textbooks on this subject [1].
To place this in its simplest context, consider the Schrödinger equation for
a molecule in an open environment, and in the presence of a laser field. The
Hamiltonian can be considered as the sum of three parts:
E t/ D i P .x;
.Hmol C Henv C Hlaser / .x; E t/ (5.1)
(Note that this is written using atomic units, in which „ D me D jej D 1). The
first term Hmol is the single isolated molecule Hamiltonian in all its glory, including
all kinetic energy terms, coulomb interactions among electrons and ions, spin-orbit
interactions, and hyperfine interactions. This Hamiltonian has a ground state as
well as bound and continuum excited states with eigenenergies En which satisfy
Schrödinger’s time-independent equation Hmol n .x; E t/ D En n .x; E t/: The second
and third terms describe “everything else,” and here we must adopt a particular
point of view in order to separate this into two parts, due to the laser, and due to the
“environment.” For the purposes of this discussion the laser term Hlaser will be taken
to include both the laser energy itself as well as its interaction with the molecule.
Control is exerted on the quantum system by means of this term.
In order to utilize Hlaser to control the system, we therefore require that it possess
two criteria:
• Hlaser > Henv , i.e. the forces on the system due to the laser must exceed the forces
of the environment, or other forces within the molecule that lead to dephasing or
quantum diffusion of the wave packet away from the optimal path.
• laser < „= E, i.e. the time of the interactions must be less than the natural
dynamical timescale for the system, given by the energy splitting. This timescale
depends on the context. For example, in many molecular problems the relevant
timescale is set by the relative motion of the atoms in the molecule, which
is typically on the order of hundreds of femtoseconds. In other problems, the
relative timescale may be due to the motion of molecular electrons, which can
have dynamics on the few femtosecond or sub-femtosecond scale.
Thus control requires both ultrafast pulses and strong fields, to beat the tendency
towards decoherence and to overcome natural dephasing. Some general protocol is
required to generate a laser pulse corresponding to Hlaser that is optimal for a given
problem.
There are two different types of proposals for quantum control, that are known
generally by the names of their inventors, the “Tannor-Rice” scheme and the
“Brumer-Shapiro” scheme. Tannor, Rice, and the team of Brumer and Shapiro have
each written textbooks or monographs including descriptions of control from their
points of view, and these are excellent introductions [2–4]. Here we will merely
summarize the essential features of these schemes.
Many readers of this chapter may already be completely familiar with the Born-
Oppenheimer approximation in molecules, but since this chapter is intended for a
broader group of laser scientists, we include a “one-page course” on the essential
physics of this powerful idea. A very readable and more complete treatment is given
in a review paper by Worth and Cederbaum [5]. Consider a molecule with N atoms
and many electrons. The quantum Hamiltonian of this molecule can be separated
5 Ultrafast Quantum Control in Atoms and Molecules 107
into a kinetic energy part Tn for the atoms, whose relative positions are described by
a 3N-6 dimensional vector R, (3N-5 for a diatomic molecule); plus all of the terms
describing the kinetic energies of the electrons at their different positions r and the
electrostatic potential energies between all of the pairs of charged particles. It can
be written in a schematic way like this:
The electronic energy eigenvalues Vi .R/ are the potential energy surfaces of
the Born-Oppenheimer approximation. Although they are often depicted as one-
dimensional potential wells or two-dimensional “elastic sheets,” they are really high
dimensional manifolds spanning all of the different relative positions of all of the
atoms in the molecule.
The i .R/ form a real orthonormal basis on which the full solution to the
molecular wave function may be expanded:
X
j .R; r/ D i .R/i .r/ (5.4)
i
Combining this expression for with the molecular Hamiltonian, we can express
the full Schrödinger’s equation H D i „ P . Then we employ the familiar algebraic
procedure to find the coefficients .R/, i.e. multiply on the left by each of the i
and integrate. This yields a set of coupled equations that look almost like a series
of Schrödinger’s equations for the atoms moving on the potential energy surfaces
Vi .Ri /, with wave functions that are the expansion functions i:
X
ŒTn C Vi i C ƒij j D i „ Pi (5.5)
j
5.2.3 Tannor-Rice
David Tannor and Stuart Rice have introduced an intuitive concept for control
based on the motion of coherent wave packet states in laser-excited molecules. This
is often referred to as “pump-dump” control [8]. The essence of this scheme is
demonstrated by considering two Born-Oppenheimer potential energy surfaces in
a molecule. We consider a model molecule consisting of three atoms, A, B, and
C, which have a ground state in which the three are bound together in the order
ABC. There are 3N-6 D 3 dimensions in the Born Oppenheimer potential energy
surfaces, which could, for example, be designated as the AB bond length (R1 ); the
BC bond length (R2 ), and the ABC bend angle (R3 ). The coherent control problem
is to photoexcite the molecule in such a way to control whether A or C dissociates
from the other two atoms. The Born-Oppenheimer surface for the ground state in
the R1 -R2 plane might look like Fig. 5.2.
Note that there are two low places in the energy landscape, where A may
dissociate from BC (along R1 ), or C may dissociate from AB (along R2 ). The initial
state of the molecule is presumed to be the ground vibrational state, which is well-
localized in the region of the potential minimum at (R1 , R2 ) D (2,2). The control
task is to guide the system toward one dissociation channel using coherent light.
110 P.H. Bucksbaum
Fig. 5.4 The path of the pump-dump control scheme is shown above. The molecule is initially at
location 1. It is pumped to 2, and then evolves on the excited surface to 3, where it is dumped to
4 and dissociates. The output channel is selected by controlling the time between the pump and
dump pulses
5.2.4 Brumer-Shapiro
In the alternative Brumer-Shapiro route to quantum control, the final state specificity
comes from the interference between multiple pathways in multiphoton transitions
from the initial state to the final state. This method emphasizes the role of
phase coherence in the control mechanism. Brumer-Shapiro control is multimode
interference between stationary states, whereas Tannor-Rice control is wave packet
evolution of nonstationary states. At first glance it might appear that the control
mechanisms are totally different, since Tannor and Rice emphasize ultrafast pump
and dump pulses to create wave packets that consist of many different eigenstates,
whereas Brumer and Shapiro seem to deal with narrowband resonant processes
with good phase control and excitation in two competing pathways that interfere.
In fact, in many respects these are just two different views, time and spectral, of
the same process. A simple example of how the Brumer and Shapiro idea can
be connected to ultrafast excitation is the two-photon process depicted below in
Fig. 5.5.
We assume that the bandwidth of the pulses is sufficiently broad that A ! B and
A ! C can both occur, and the control problem is whether we can use multipath
interference to select only one of these alternatives. The answer is that we can
control the outcome without changing the power spectrum of the laser pulse, and we
do this by controlling the spectral phase of the light. (Spectral phase is an important
control feature of ultrafast light pulses, and is covered in some detail in other
chapters of this volume, such as Chap. 1 by Wyatt and Professor Walmsley, so I will
assume the reader has some familiarity with it.) The use of phase is seen clearly
by considering the two-photon process within the formalism of time-dependent
perturbation theory, where a non-resonant two-photon excitation is driven by an
operator proportional to E 2 .t/. The Fermi’s Golden Rule transition probability
between the ground state and any excited state is then proportional to the spectral
power of E 2 .t/, i.e. by the square of its Fourier transform. Let’s take a simple
pictorial example: Consider a short pulse E.t/, with a real Fourier transform and
a power spectrum as shown in Fig. 5.6.
112 P.H. Bucksbaum
Fig. 5.6 The four panels show the electric field of a transform-limited ultrafast pulse; the spectrum
of the pulse; the squared field; and the spectrum of the squared field
Fig. 5.7 These panels show the same progression as in Fig. 5.6., but for a phase-shaped pulse. The
spectrum of the squared field displays phase-dependent features that affect two-photon absorption
control. For that we seek some search protocol or algorithm that can be used to
model general systems, to examine their controllability and to find the best control
parameters.
One such protocol is Optimal Control Theory. We begin with a general statement of
the problem: Given i .t/ and H D H0 C H 0 Œ".t/, we want to adjust ".t/ to create
a desired target state f .T /. Furthermore, we want to find the “best” path, i.e. the
path that minimizes some cost, such as the total energy consumed. How do we do it?
This is a deep subject, and there are many deep thinkers who have influenced it.
An excellent introduction is the textbook by Tannor [9]. The best we can do here is to
provide some basic information. We will start with some simplifying assumptions:
Two potential energy surfaces in the Born-Oppenheimer approximation; and pure
states (i.e. no decoherence or thermal averaging) so that we can use the Schrödinger
equation without modifications.
The “pump-dump” picture and the “two-path interference” picture both suggest
that a perturbative approach taken to second order (for two interactions with the
laser field) should suffice to describe the process of control. However, following
Tannor, we will take a more general approach that allows for strong fields and any
number of photons.
Our objective can be quantified as the overlap integral, or projection, of the
system .T / on the target state f , i.e.
ˇ˝ ˇ ˛ˇ2
J D lim ˇ f ˇ .T / ˇ D lim h .T /j Pf j .T /i (5.8)
T !1 T !1
ˇ ˛˝ ˇ
where Pf D ˇf f ˇ is the projection operator for channel f, e.g. a particular final
state channel. J is called the “objective functional” for the optimal control problem.
The system state function .t/ is evolving according to some H.t/ that depends
explicitly on the history of the applied laser field ".t/ at all times t < T:
There may also be some constraints. For example, perhaps we need to do this
RT
job with finite total laser pulse energy dtj".t/j2 . To implement this constraint, we
0
modify the objective functional by decreasing its value in proportion to the pulse
energy, as a penalty:
ZT
J lim h .T /j Pf j .T /i dtj".t/j2 (5.10)
T !1
0
The wave function must obey Schrödinger’s equation, but this may also be
enforced as a constraint, following Kosloff [3], so that we have:
J lim h .T /j Pf j .T /i
T!1
2 T 3
Z ZT
4 @ H.".t// 5
C 2 Re dt h .t/j C j .t/i dtj".t/j2 (5.11)
@t i„
0 0
The “dual” wave function .t/ plays the role of the Lagrange multiplier. Clearly
the second term vanishes for any .t/ which satisfies Schrödinger’s equation,
independent of .t/. Notice by the way, that the opposite is also true: If .t/
obeys Schrödinger’s equation, then the second term vanishes. A convenient physical
interpretation for .t/ is the projection of .T / on the target state, i.e.
.T / D Pf .T / (5.12)
We can use transitions between two states of the system, the ground state (“g”) and
an excited state (“e”), just as in the Tannor-Rice example. The Hamiltonian takes on
a simple 2 2 form:
Hg " .t/ g g
H ; with D ; D (5.13)
".t/ He e e
The job now is to find the extremum value for J by varying its parameters .t/,
.T /, ".t/, and " .t/. The first two variational equations simply direct us to require
that .t/ satisfies Schrödinger’s equation and .T / is the projection of .T / on the
target state, as we supposed from inspection of the form of J. The field variationals
are more interesting. They lead to the principal equation of optimal control:
i ˝ ˇ ˝ ˇ
".t/ D g .t/
ˇ j e .t/i g .t/
ˇ j e .t/i (5.14)
„
This equation states a relationship between the field, the wave function, and the
“dual” function at every time during the quantum evolution of the state. The target
state appears implicitly because of its connection to the dual function .T /, and this
provides a clue about how to solve this equation iteratively:
1. Start with the initial state .t D 0/ and a simple guess for the field ".t/.
2. Propagate the wave function using H Œ".t/ from t D 0 to t D T.
3. Invoke .T / D Pf .T /.
4. Now propagate .t/ backwards in time from T to 0 using H Œ" .t/.
5. Now armed with both .t/ and .t/, we may calculate a new ".t/. This will
generally be different from the first field that was chosen. But this new field can
now be employed in steps 1–5, repeating the iterations until convergence.
5 Ultrafast Quantum Control in Atoms and Molecules 115
Does this work? Often, but not always. The issue of “controllability” is beyond the
scope of this introduction, except to state that not all simple quantum systems are
obviously controllable. A classic example of a “problem” potential is the harmonic
oscillator. Since all energy levels are equally spaced, any transition between two
is potentially a transition among all, which makes control more challenging, if not
impossible.
The ABC dissociation problem that we discussed earlier as an example was
considered by Kosloff et al. for the specific case of mass ratios of 1:1:2 (such as
HHD), as a test of the optimal control algorithm discussed above. The results are
quite instructive regarding the connection to the Tannor-Rice ansatz. They do indeed
find that the best laser field transfers population from g to e in a “pump-dump”
sequence, but the optimal field sometimes requires several pump-dump cycles.
The conclusions of this section are quite powerful. We find that it is possible
to define a protocol for quantum control in many general situations! That is, given
an initial molecule and a more than one possible target product, we can use the
coherence properties of a laser field to selectively direct a transition. The next
question is, so what?! We take up this issue next.
This section has a grand title that it probably doesn’t deserve, because there is no
general prescription for deriving physical insight from the results of optimal control
theory. Nonetheless, that is a primary motivation for the field. We desire to know the
optimal field, but immediately after that to also know why that field works. What is
the underlying physical mechanism? To illustrate the problem, we consider a recent
theoretical study by Artamonov on the isomerization of ozone [1]. The normally
bent O3 molecule has a metastable triangular form. Optimal control theory was used
to find the optimal field for the transition, as shown in Fig. 5.8. The optimal field,
taken by itself, is rather uninformative.
Perhaps one can glean that the field starts out at a lower frequency, then has a
period of near darkness, before resuming a higher frequency. The spectrum shows
that these two features are quite separate:
The correlation between time and frequency is an important feature of any
control pulse, but that correlation is sometimes difficult to pick out of the temporal
representation ".t/, and the information is totally lacking from the power spectrum.
There are representations that contain both time and frequency information in an
easy to read form. One example from audio waves is a musical score. Sheet music
is a form of audio spectrogram, where each note on the page displays spectral
information through its vertical placement on the musical staff, and temporal
information through its shape and its placement order along the horizontal. This
suggests that a two-dimensional representation could also be employed for light,
and there are several different examples in common use.
116 P.H. Bucksbaum
Fig. 5.8 Left: Optimal field for the ozone isomerization transition [1]. Right: Spectrum of the field
shown on left of the figure (Reproduced and adapted with permission from Ref. [1])
Fig. 5.9 Left: Potential energy vs. opening angle for ozone. Right: WFT spectrogram for the
optimal photoisomerisation pulse (Reproduced and adapted with permission from Ref. [1])
of transitions upwards in the potential well at greater amplitudes, until the molecule
can pass across to the metastable well. Then a positive chirped higher frequency
pulse induces de-excitation to the final state. This pulse has a transfer efficiency of
an impressive 94 %. This simple explanation may not be correct, though, because if
the optimal pulse is replaced by a simple combination of two chirped pulses with the
same shape, but without the low amplitude parts that appear to be a noisy pedestal,
then the transfer efficiency goes down to less than 10 %. Evidently, even the small
bumps and wiggles are very important to the success of the experiment, and the
simple view of chirped pulse ladder climbing is not the whole story, nor even the
biggest part of the story! [1].
One missing element in the simple one-dimensional wave packet interpretation of
this calculation is the three-dimensional structure of the potential energy surfaces for
ozone. The other two dimensions are associated with bond stretching. Although they
are hidden from view in the one-dimensional picture, their presence survives in the
energy level splitting, and the optimal field is affected by this as it tries to steer the
molecule away from unnecessary vibrations excited by sub-optimal control fields.
Two-dimensional representation of control fields: This example demonstrates
the value of good visualization of the control field using two-dimensional spectro-
grams. Optical spectrograms can be constructed in different ways, and the analogy
to music only describes them partially. We would like a spectrogram that retains all
of the same information as the temporal representation, E.t/, which is a complete
description of the field. The WFT does not do this, since it convolves the field with
a gate function of finite width and shape. The gate can smear out sudden field
changes, or attenuate rapid fluctuations. One transformation that retains complete
information is the Wigner distribution, which is the inverse Fourier transform of the
autocorrelation of the complex spectrum (Fig. 5.10):
Z
0
S.; t/ D E. C 0 /E . 0 /e 4i t d 0 (5.15)
Like the field itself, the Wigner distribution is a real function. It resembles a
spectral-temporal probability distribution, but differs in two important ways from
this simple interpretation. First, the Wigner distribution can take on both positive and
negative values, so it can’t simply represent a probability. Second, it has features that
appear in locations in space and time where there is no field at all! These are regions
between optical field pulses, and they represent the relative coherence properties
between sub-pulses that are not temporally connected. Still, like the FROG traces
discussed in Chap. 1 of this book, the Wigner distribution does provide a fairly
simple connection to the real pulse after a little practice. The Wigner distribution is
“overcomplete.” This simply means that that if I divide an electric field waveform
into N pieces, for example in order to plot it in Matlab, then the Wigner distribution
I calculate from this has N N D N 2 pixels, even though only N pieces of
information went into its construction. A different
p transformation
p was developed
that contains the same features but with only N N D N pixels, called the
von Neumann distribution [10].
118 P.H. Bucksbaum
Fig. 5.10 Displays of a two-pulse optical field. Top: The spectrum and the intensity vs. time.
Left: Intensity and power spectrum. Left: Wigner time-frequency distribution, showing coherence
features between the two separated spectral peaks. Right: A Husimi distribution eliminates the
spectral interference features but still shows the spectral separation of the two peaks
Finally, for those who wonder if all two-dimensional optical distributions are
named after Princeton faculty, here is a counterexample: A very useful spectrogram
called a Husimi distribution is a two-dimensional convolution of the Wigner
distribution with a minimum uncertainty Gaussian. This procedure removes rapid
oscillating coherence features from the Wigner distribution, leaving something
that much more closely resembles the temporal-spectral probability. Although this
destroys coherence information contained in the complete description of the field, it
shows clearly how the optical energy is distributed in both time and frequency.
Pulse shaping for ultrafast optical fields can be performed in various ways. There
is an excellent pedagogical review by one of the pioneers in the field, Andrew
Weiner [11]. Here I summarize some of the basic ideas and methods.
5 Ultrafast Quantum Control in Atoms and Molecules 119
At its root, pulse shaping is wave filtering. A pulse shaper receives an input wave
form with a complex Fourier transform Ein .!/, transforms it by combining it with
some mask or mixing wave M.!/, and outputs a final waveform Eout .!/, as in the
example in Sect. 5.2 of this chapter. If the mixer is a frequency modulator then we
describe this using the language of nonlinear optics:
0 for ! ¤ !M
Eout .!/ D Ein .!/M.!/ (5.17)
Ein for ! D !M
This kind of mask appears purely subtractive, meaning that it cannot add new
frequencies to Eout .!/, but only filter frequencies that are already present in the
input field. This is an example of a linear filter, whereas the wave mixing modulator
is a nonlinear filter. Most programmable pulse shapers used in quantum control
experiments use a linear frequency filter to shape pulses. Filters used include
grating-mask combinations employing patterned stationary masks, liquid crystal
programmable masks, acousto-optic Bragg cells, and programmable dispersive
filters, such as the “Dazzler.”
The basic configuration of a linear frequency filter, as shown in Fig. 5.11, is
similar to a spectrometer combined with its mirror image:
120 P.H. Bucksbaum
This is sometimes called a “4f ” configuration, because the distance between the
mirrors is chosen to be twice their focal length f. The distance from the gratings to
the mirrors is f, so that the first mirror collimates the expanding fan of wavelengths at
the same time as it focuses each separate wavelength. Such an arrangement does not
change the size of the laser beam from the input to the output. Also, to lowest order
the optical path for each wavelength is the same, so that a short pulse is not dispersed
at the output. The mask is at the image plane of the first spectrometer, which is the
point where the different wavelengths are most cleanly separated. The mask can
then transmit some wavelengths and block others, or apply pre-programmed phase
shifts to different regions of the spectrum. The second half of the device performs
an inverse Fourier transform on the amplitude- and phase-altered spectrum that
emerges from the mask. If the mask is programmable, then this device functions
as an arbitrary waveform generator for light.
There are a few important constraints on the waveform that can be produced by a
Fourier filter pulse shaper. First, if the mask is a static and a linear element as I have
described it, then this device can only attenuate and shift frequencies that are already
present in the input beam. Second, global phase cannot be controlled by this device.
This is sometimes called the “carrier envelope” phase. For any optical field pulse
longer than a few tens of femtoseconds, this phase has no significance, because it
only represents a trivial overall time shift of less than one optical cycle. But for very
short pulses that are only a few cycles long, the carrier phase can determine the
details of the interactions of the light with matter.
Fourier filter pulse shapers can be constructed using several different kinds of
spectral selection, some of which have been commercialized. The simple 4f system
shown in Fig. 5.11 requires a mask-type filter, which can be implemented with
programmable liquid crystals, acousto-optic Bragg-type reflectors, or static masks.
All of these methods are subtractive, and in addition they have various efficiencies.
I have already described how nonlinear processes can be controlled by adjusting the
phase alone, and indeed phase shaping can accomplish many tasks, including the
production of series of short pulses with variable delays. For this purpose, phase-
only shapers are sufficient, such as deformable mirrors that adjust the phase of
different spectral components by introducing sub-micron delays. The “Dazzler” is
a commercial acousto-optic pulse shaper that relies on the spectral selectivity of a
programmable periodic Bragg reflector, and therefore does not require dispersing
gratings [12, 13].
Another property that influences the selection of an appropriate pulse shaper is
the refresh rate. Acousto-optic gratings are transient but can only be refreshed at the
speed of sound, and therefore inappropriate for continuous wave or high repetition
rate applications. On the other hand, they are ideal for applications where a different
pulse shape must be selected on each pulse, up to repetition rates of several kilohertz.
Liquid crystals and deformable mirrors refresh much more slowly, but they make
good static programmable masks, which can therefore be used for continuous wave
applications.
5 Ultrafast Quantum Control in Atoms and Molecules 121
2. Perform the experiment with each member of the generation, and record the
relative amount of the desired target state following each experiment. This
successful fraction is called the “fitness” of the corresponding pulse, since it
indicates its relative suitability.
3. Rank the genomes according to their fitness.
4. Now form the next generation in the learning feedback loop, by combining the
best traits of the previous generation. This step is one where all the improvement
in fitness occurs, and I will have more to say about it below.
5. Go back to step 2, and repeat steps 2–5 until the fitness is optimized.
The fact is that such a fitness directed search works extremely well. Much of the
success, however, comes from clever innovations in step 4 above, i.e. the selection
of a new generation base on the fitness ranking of the previous generation. The
selection method is called an “operator.”
The learning feedback used to select individual pulses with high fitness can also be
used during the algorithm to select the most efficient operators. This technique is
called “adaptive operators.” In a typical implementation, several different operators
are used to create offspring. The particular operator for each individual is chosen
at random, but recorded by associating a genealogy with each individual. After
three generations, the genealogy of the most fitted individuals is a record of the
most efficient operators. The ensuing generations use this information to weight the
probability for using these more effective operators. The genealogy is examined in
every subsequent generation, and the probability for selecting operators is adjusted
to ensure that the most effective operators are used a higher fraction of the time
(Fig. 5.13).
One of the great strengths of learning feedback control is that it achieves a solution
without any prior knowledge of the Hamiltonian. Its efficiency does not depend on
this, nor to any great extent on the size of the control space. It is hardly more effort
to search a large space than a small one, since the search is fitness-directed, and
therefore tends away from unproductive regions or degrees of freedom. The number
of possible solutions in the control space may be huge, but the number of solutions
included in the search is not huge [16].
124 P.H. Bucksbaum
Fig. 5.13 The fitness for multiple operators plotted as a function of generation for a single
run of the learning algorithm while maximizing the fitness for a particular learning feedback
algorithm. This was a search to optimize the peak intensity of a pulse by adjusting its spectral
phase (Reproduced and adapted with permission from Ref. [15])
The price paid for this luxuriously large search space is that the solution may be
difficult to intuit, and it may be nearly impossible to discern the underlying physics
by examining the features of the optimal pulse. The fitness function maximum
may only depend on a few “essential” features of the control field, even though
these are not obvious in the spectral basis used for most control experiments. Other
unimportant features are just ornaments, so long as they do no harm.
This problem has been addressed by adapting a control analysis protocol bor-
rowed from the analysis of large statistical data sets, known as Principal Component
Analysis (PCA). The version adapted to control problems has been named Principal
Control Analysis [17]. PCA is based on the notion that all solutions with high fitness
should share common traits which can be found by examining details of the search
process itself.
The basic idea is that the control Hamiltonian for any particular pulse shape
H ŒE.xi / depends on a vector of control parameter settings xi that produce that
pulse shape. It should be possible to rewrite the Hamiltonian using a new basis
where the most important linear combinations of parameters are basis vectors. In
the old bases, the set of most important control directions forms a lower-dimensional
subspace, like a surface embedded in a cube (see Fig. 5.14). Each essential control
direction is a linear combination of the much larger basis of the control experiment.
Since the trial pulse shapes in the experiment were chosen specifically because
5 Ultrafast Quantum Control in Atoms and Molecules 125
Fig. 5.14 Geometric model of the control Hamiltonian. The space of all pulse shapes is depicted as
a cube on the left. The optimal pulses lie on a lower dimensional space within the cube, here shown
as a surface. All pulse shapes on this surface are equally fit. The same Hamiltonian control space
is shown on the right, but now transformed to the essential control basis. Now the unimportant
degrees of freedom are no longer present, so that the optimal pulses all lie at a point. PCA is a
method to find these essential degrees of freedom (From Ref. [17])
they had the highest fitness, we can get a good sense of these control directions
by examining the control values in the learning search [18].
Principal Control Analysis investigates the covariance of the values of pairs of
genes in the set of pulse shapes used in the search. In other words, how much do
pairs of values vary together in a fitness directed search. The variance of ith gene in
the control parameter genome xi is:
!1=2
X
hxi i D xi2 n (5.18)
n
We want to find a new basis in which the variance of each control direction is
independent of the others. Clearly this is just the basis in which cij is diagonal:
2 3 2 3
c11 : : : c1m 1 0 0
6 :: : : : 7 6 7
4 : : :: 5 ! 4 0 2 0 5 : (5.20)
:
cm1 cmm 0 0 ::
126 P.H. Bucksbaum
The largest eigenvalues correspond to directions in the control space with the largest
variance. Since the search is directed by fitness, these are the directions of greatest
fitness improvement, and therefore correspond to the most important features of
the pulse for control. The eigenvectors of these eigenvalues then form the essential
control basis. Often there are only two or three large eigenvalues, indicating that
control can be established with only a few independent controls.
haf i haihf i
RD
a f
P
N P
N
˛k k f ˛k k hf i
kD1 kD1
D (5.21)
a f
The feature with the highest fitness correlation has a particularly simple form in
the PCA eigenvector basis: Its components are given by:
This is powerful, because it allows us to compute the pulse shape with the highest
correlation to fitness even though this specific pulse shape may never have been
produced in the search. Furthermore, since we know that only a few directions in
the PCA basis have high eigenvalues and therefore high sensitivity to the search
objective, therefore only the first few ˛ k are important.
A basis of most correlated features may be calculated iteratively by Schwartz
decomposition, i.e. subtracting the projection of the data onto the most correlated
feature from the data and repeating the MCF analysis. This basis can be used to
produce a simplified “essential” cut in the fitness landscape.
that has a monotonic relationship to the target functional. However, the connection
between the raw signal and the fitness can be exploited to learn more about the
dynamics initiated by the control fields, and recent work in quantum control has
begun to focus on this.
This chapter has introduced some of the basic concepts in the subject of quantum
control of small molecules. The original impetus for quantum control was to learn
how to use the coherent properties of lasers to accomplish arbitrary and selective
bond rearrangements in molecules, despite the fact that the molecular Hamiltonian is
far too complicated to solve directly, and the number of coupled degrees of freedom
means that any solution will be nearly impossible to implement anyway.
The theory problem has been approached through optimal control methods,
which are essentially computational feedback protocols to iteratively solve complex
problems. On the experimental side, the learning feedback method overcomes
this complexity by using the molecule itself as an analogue computer, where the
experimental apparatus supplies the input and output registers, and the experimenter
is merely the programmer [14]. The ansatz here is that the molecule will solve the
problem for us, if we supply the right experimental protocol, and if we are persistent.
The current state of the field is far from the original goal of bond-selective
manipulation, but quantum control has had an important impact in several areas,
for example in ultrafast and strong-field physics, and ultracold atom physics. This
brief introduction has not gone into these applications, but now we mention a few
of them to provide an outlook for future work.
The theory of quantum control is extremely active. Fundamental issues such as
the dependence of controllability on the nature of the state space, or the topology
of the control landscape, are quite current [19]. Control in open systems, where
there is dissipation, is important to many applications. A particularly active area of
theoretical interest in this regard is quantum information.
Many of the concepts of quantum coherent control are closely related or
even derived from the far more established field of nuclear magnetic resonance.
Very sophisticated pulse shapes and feedback protocols are employed in magnetic
resonance experiments to obtain structure information on the atomic
0
scale [20].
The new field of attosecond science relies on strong (V/Å) laser field, which
can field ionise and subsequently re-scatter a molecular electron, producing high
harmonics (HHG). Control of these quantum processes could been used for
detecting atomic-scale structure and transient processes in small molecules [21].
Finally, quantum control concepts are also beginning to have an impact in the new
x-ray free electron lasers. Control schemes are needed there to prepare molecules
for interrogation by intense x-rays [22].
128 P.H. Bucksbaum
References
1. M. Artamonov, T.-S. Ho, H. Rabitz, Quantum optimal control of ozone isomerization. Chem.
Phys. 305, 213–222 (2004)
2. P. Brumer, M. Shapiro, Laser control of molecular processes. Annu. Rev. Phys. Chem. 43,
257–282 (1992)
3. R. Kosloff, S.A. Rice, P. Gaspard, S. Tersigni, D.J. Tannor, Wavepacket dancing: achieving
chemical selectivity by shaping light pulses. Chem. Phys. 139, 201–220 (1989)
4. S.A. Rice, M. Zhao, Optical Control of Molecular Dynamics (Wiley, New York, 2000)
5. G.A. Worth, L.S. Cederbaum, Beyond Born-Oppenheimer: molecular dynamics through a
conical intersection. Annu. Rev. Phys. Chem. 55, 127–158 (2004)
6. E.P. Wigner, J. von Neumann, Über das verhalten von eigenwerten bei adiabatischen prozessen.
Phys. Z. 30, 467–470 (1929)
7. E. Teller, The crossing of potential surfaces. J. Phys. Chem. 41, 109–116 (1937)
8. D.J. Tannor, S.A. Rice, Control of selectivity of chemical reaction via control of wave packet
evolution. J. Chem. Phys. 83, 5013–5018 (1985)
9. D.J. Tannor, Introduction to Quantum Mechanics: A Time-Dependent Perspective (University
Science, Sausalito, 2007)
10. S. Fechner, F. Dimler, T. Brixner, G. Gerber, D.J. Tannor, The von Neumann picture: a new
representation for ultrashort laser pulses. Opt. Express 15, 15387–15401 (2007)
11. A.M. Weiner, Femtosecond pulse shaping using spatial light modulators. Rev. Sci. Instrum. 71,
1929–1960 (2000)
12. P. Tournois, Acousto-optic programmable dispersive filter for adaptive compensation of group
delay time dispersion in laser systems. Opt. Commun. 140, 245–249 (1997)
13. F. Verluise, V. Laude, J.-P. Huignard, P. Tournois, A. Migus, Arbitrary dispersion control of
ultrashort optical pulses with acoustic waves. J. Opt. Soc. Am. B 17, 138–145 (2000)
14. R.S. Judson, H. Rabitz, Teaching lasers to control molecules. Phys. Rev. Lett. 68, 1500–1503
(1992)
15. B.J. Pearson, J.L. White, T.C. Weinacht, P.H. Bucksbaum, Coherent control using adaptive
learning algorithms. Phys. Rev. A 63, 063412 (2001)
16. J.H. Holland, Genetic algorithms. Sci. Am. 267, 66–72 (1992)
17. J.L. White, B.J. Pearson, P.H. Bucksbaum, Extracting quantum dynamics from genetic learning
algorithms through principal control analysis. J. Phys. B At. Mol. Opt. 37, L399–L405 (2004)
18. C.J. Bardeen, V.V. Yakovlev, K.R. Wilson, S.D. Carpenter, P.M. Weber, W.S. Warren, Feedback
quantum control of molecular electronic population transfer. Chem. Phys. Lett. 280, 151–158
(1997)
19. C. Rangan, A.M. Bloch, C. Monroe, P.H. Bucksbaum, Control of trapped-ion quantum states
with optical pulses. Phys. Rev. Lett. 92, 113004 (2004)
20. T.E. Skinner, T.O. Reiss, B. Luy, N. Khaneja, S.J. Glaser, Application of optimal control theory
to the design of broadband excitation pulses for high-resolution NMR. J. Magn. Reson. 163,
8–15 (2003)
21. B.K. McFarland, J.P. Farrell, P.H. Bucksbaum, M. Gühr, High harmonic generation from
multiple orbitals in N2 . Science 322, 1232–1235 (2008)
22. J.M. Glownia, J. Cryan, J. Andreasson, A. Belkacem, N. Berrah, C.I. Blaga, C. Bostedt,
J. Bozek, L.F. DiMauro, L. Fang, J. Frisch, O. Gessner, M. Gühr, J. Hajdu, M.P. Hertlein,
M. Hoener, G. Huang, O. Kornilov, J.P. Marangos, A.M. March, B.K. McFarland, H. Merdji,
V.S. Petrovic, C. Raman, D. Ray, D.A. Reis, M. Trigo, J.L. White, W. White, R. Wilcox,
L. Young, R.N. Coffee, P.H. Bucksbaum, Time-resolved pump-probe experiments at the LCLS.
Opt. Express 18, 17620–17630 (2010)
Chapter 6
Femtosecond Optical Frequency Combs
6.1 Introduction
Frequency can be measured with by far the highest precision of all physical
quantities. In the radio frequency (rf) domain (! < 2 100 GHz), frequency
counters have existed for a long time. Almost all of the most precise measurements
in physics have been performed with a counter that uses an atomic clock as a time
base. To extend this accurate technique to higher frequencies, so called harmonic
frequency chains have been constructed since the late 1960s [1, 2]. In such a chain
nonlinear elements produce frequency multiples (harmonics) of a given oscillator to
which a subsequent oscillator is phase stabilized. Repeating this procedure several
times makes it possible to multiply a reference radio frequency to much higher
frequencies. Electronic phase locked loops [3] can be used to stabilize the relative
phase of any kind of oscillator, even lasers, provided their intrinsic stability is
sufficient so that there is no need for very rapid frequency corrections. It was not
until 1996 that visible laser light was first referenced phase coherently to a cesium
atomic clock using this method [4].
The disadvantage of these harmonic frequency chains was not only that these
set-ups could fill several large laser laboratories at once, but that they could only
measure a single optical frequency. A new solution to that problem emerged at the
end of the 1990s through the use of a mode locked laser that is capable of producing
pulses of light in the femtosecond domain. Such a laser emits a very broad spectrum,
comparable in width to the optical carrier frequency. The laser stores this light pulse
inside its resonator for an almost unlimited duration emitting a copy of the stored
pulse after each round trip time T that typically lasts between 1 and 100 ns.
In the frequency domain, a train of short pulses from a mode locked laser is the
result of a phase coherent superposition of many continuous wave (cw) longitudinal
cavity modes. These modes form a series of frequency spikes that is referred to as
a frequency comb. The individual modes can be selected by phase locking other
cw lasers to them. The modes are remarkably uniform, i.e. the frequency difference
between adjacent modes is constant across the comb. This strictly regular arrange-
ment is the most important and defining feature of a frequency comb and was first
demonstrated in [5]. The optical frequencies ! n of the comb may be expressed as:
Fig. 6.1 Three consecutive pulses of the pulse train emitted by a mode locked laser and the
corresponding spectrum. There is a pulse-to-pulse phase shift of ' between pulses because
the pulse envelope propagates at the group velocity and the optical carrier wave at ! c moves with
the phase velocity. In the frequency domain this corresponds to an offset frequency ! CE D '/T
that shifts the comb as a whole from the strict harmonics of ! r
X
E.t/ D A.t/e i !c t C c:c: D Aq e i.!c Cq!r /t C c:c:; (6.2)
q
where Aq are Fourier components of A(t). This equation represents a comb of laser
frequencies with spacing ! r that is shifted as a whole from the harmonics of ! r , just
as in Eq. 6.1. The only difference being the numbering of modes which clearly is a
matter of convention.
In a real laser the pulses may be chirped, i.e. their carrier frequency varies across
the pulse. Such a chirp from a laser is periodic with the repetition rate otherwise the
carrier wave would increase/decrease indefinitely. For this reason the chirp can be
accounted for by allowing a complex envelope function A(t) leaving all of the above
arguments valid.
Another obvious difference of the above model to a real laser is that neither
the carrier wave nor the pulse train will be free of noise. To take this into account
the Fourier series in Eq. 6.2 has to be replaced by a convolution of the carrier
wave spectrum with the spectrum of the pulse envelope [11]. Provided the carrier
wave possesses a suitably narrow line width and the noise in the repetition rate is
low enough (to maintain the comb structure), individual modes can be separated.
This condition is easily satisfied, even with a free running mode locked laser. The
noise of the frequency comb will be taken out by phase locked loops that stabilize
both frequencies in Eq. 6.1 relative to a precisely known radio frequency reference.
132 T. Udem et al.
This time dependent phase shift leads to a frequency modulation that is pro-
portional to the time derivative of the self-induced phase shift ˚(t). Self-phase
modulation modifies the envelope function according to:
Because ˚(t) has the same periodicity as A(t), the comb structure of the spectrum,
as previously derived, is not affected.
In an optical fibre this process is more efficient than in bulk material, even though
the nonlinear coefficient in fused silica is comparatively small. This is because the
fibre core carries a high intensity over an extended length. However, dispersion
causes the pulse to broaden temporally so that the large peak intensity required for
this process can diminish significantly after only a few cm of propagation length.
Optical fibres that partially compensate that effect can be much more effective
in terms of spectral broadening. In the near infrared one can use the zero group
velocity dispersion of silica around 1.3 m, but the zero dispersion wavelength
can also be shifted in fibres by properly designing the refractive index profile
of the waveguide. In addition, doping can be used to increase the non-linearity.
This approach is generally used with fibre based mode locked lasers that operate
in this spectral region. For wavelengths around 800 nm, where titanium sapphire
mode locked lasers operate, photonic crystal fibres (or micro structured fibres)
may be used. Such a fibre uses an array of air holes that surround the fibre core
and run the length of a silica fibre to change its effective dispersion [12]. Some
lasers can reach an octave spanning spectrum even without using external self-
phase modulation. This became possible with intra-cavity self-phase modulation
[13] and/or by using extremely broad band intra-cavity mirrors [14]. These titanium
sapphire lasers are an advantage because they avoid problems associated with
6 Femtosecond Optical Frequency Combs 133
mechanical stability when coupling into the tiny core of a photonic crystal fibre.
(For a more comprehensive discussion of nonlinear effects in optical fibres, see the
Chaps. 8 and 9 by Dudley et al. and Wadsworth respectively).
6.3 Self-Referencing
To stabilize the frequency comb it is first of all necessary to generate the two radio
frequencies ! r and ! CE that make up the frequency comb in order to stabilize them
relative to some radio frequency reference. Whereas the pulse repetition frequency
may be detected anywhere in the laser beam, the generation of the carrier envelope
frequency is more involved. In fact, generating ! CE became possible only in 1999
by beating different harmonics (3.5th and 4th) of different components of the
comb [15]. This first demonstration involved considerable effort to generate the
required harmonics, but the task simplified dramatically with the availability of
octave spanning frequency combs [9, 16, 17].
The so called f-to-2f self-referencing technique uses a group of modes
from the low frequency end of the comb around nL ! r C ! CE that is frequency
doubled in a nonlinear crystal and super imposed with another group of
modes around nH at the high frequency end of the comb. This generates
signals at the difference frequency of two optical waves on a photo detector:
2(nL ! r C ! CE )nH ! r C ! CE D (2nL nH )! r C ! CE .
These beat frequencies are created for all possible combinations of nL and nH
that lie within the bandwidth of the photo detector. One can restrict the number
of signals to precisely one by using an rf low pass filter. If this filter has a cut off
frequency of ! r /2, only the beat note at ! CE , belonging to all combinations with
2nL nH D 0, remains. The octave spanning comb ensures that these combinations
exist, i.e. that the modes nL and nH with nH D 2nL are active modes of the comb
with non-vanishing power. A typical set-up for f-to-2f self-referencing sketched
in Fig. 6.2. Knowing both, ! CE and ! r , means that frequencies of all modes are
known using Eq. 6.1 and may be used for optical frequency measurements. This is
accomplished by creating another beat note ! b between an unknown laser at ! l and
a nearby mode of the comb. The frequency of that laser can then be determined
by ! l D n! r C ! CE C ! b . The mode number n may be determined by a coarse
measurement of ! l , say with a wavelength meter or by repeating the measurement
with slight variations in ! r [18].
By adjusting the resonator length, through a piezo-electric transducer, the
repetition rate can be adjusted. For a typical laser the intrinsic stability of the
repetition rate is good enough such that the attainable bandwidth of the transducer
is by far sufficient to keep it in phase with a reasonable rf reference. The other comb
parameter (! CE ) is determined by both the phase and group velocity of the stored
pulse. It was shown that for a soliton like laser such as the Ti:sapphire Kerr-lens
mode locked laser, that the group and phase velocity depend on the peak intensity
of the stored pulse [19]. Therefore, one can adjust the pulse-to-pulse slippage of
134 T. Udem et al.
Fig. 6.2 The carrier envelope offset frequency ! CE that displaces the modes of an octave spanning
frequency comb from being exact harmonics of the repetition rate ¨r is measured by frequency
doubling some modes at the “red” side of the comb and beat them with modes at the “blue” side.
The top figure sketches one possible implementation of this f-to-2f self-referencing method while
the bottom part shows what happens in the frequency domain
the carrier envelope phase, and thereby ! CE , by controlling the average laser power.
For titanium sapphire and fibre mode locked lasers this can be done by controlling
the power of the pump laser. The two controls are not completely independent.
However, any modification of the frequency that maintains the comb properties
can be characterized by a fix-point corresponding to the unaffected mode [20]
(which possibly lies outside the active frequency comb). It turns out that the fix-
point with respect to pump laser modulation is usually located close to the centre of
the frequency comb [20, 21]. For this reason using the pump power and the cavity
6 Femtosecond Optical Frequency Combs 135
length always changes both degrees of freedom of the frequency comb. Keeping the
servo bandwidths for ! r as low as possible and for ! CE as high as possible usually
de-couples the two servo systems. Having the two phase servo systems operational
ensures that, by virtue of Eq. 6.1 and the large integer n, the radio frequencies ! r
and ! CE are phase coherently up-converted to the optical region.
To investigate the integrity of the frequency comb and to check for possible sys-
tematic deviations several tests have been performed. Two independent frequency
combs have been found to agree at a level of 5.1 1016 [9] and 1.3 1017 [10].
Even though Doppler shifts due to mirror motions caused by thermal expansion set
in at this level, no systematic shifts have been found yet.
6.4 Applications
By virtue of Eq. 6.1 a frequency comb can be used to phase coherently connect the
radio frequency domain with the optical domain. This is enabled by the large integer
n ( 105 : : : 106 ) that the pulse repetition frequency is multiplied with. As sketched
in Fig. 6.3, three interesting frequency conversions are possible with the frequency
comb. Locking the radio frequencies ! r and ! CE to a Cs clock references the
optical modes to the SI second in order to determine optical transition frequencies of
interest with the utmost precision. Using the repetition rate as a reference to stabilize
the carrier envelope frequency allows the division of an optical frequency down to
the radio frequency domain or to directly measure optical frequency ratios [22].
An optical clock consists, like any other clock, of an oscillator that defines the ticks
in time and a counter that keeps track of these cycles. In a cesium clock for example
the ground state hyperfine splitting of the 133 Cs atom is used. An electronic counter
advances the second hand every time the counter has completed 9,192,631,770
oscillations. This number was chosen when the SI second was redefined in 1967
to match with the previous definition. If we look at the history of time keeping,
and compare clocks as different as sun dials, pendulum clocks and quartz clocks,
it is obvious that they become more accurate as the oscillation frequency increases.
This is simply because a higher oscillation frequency slices time into smaller pieces.
Since the end of the 1950s the cesium 9.192 : : : GHz oscillator has been used. With
the frequency comb one has a simple tool to count up 5 orders of magnitude faster.
In fact, it can be shown that any clock limited by quantum noise will possess an
instability that decreases with the inverse of the averaging time and is proportional
to the observed line width divided by the transition frequency [23]. For practical
reasons one cannot simply pick a transition with an arbitrary small line width,
because this will at the same time increase the required observation time. For this
reason optical oscillators have been envisioned to be used in clocks for a long time.
For a well-chosen transition the systematic uncertainties may be as low as 1018 .
After tremendous advances in laser spectroscopy in the 1980s and 1990s that
ultimately resulted in trapped atom and ion standards (see for example [24, 25]),
a reliable optical clockwork mechanism was the missing piece for completion of
the all-optical atomic clock. The best of these new clocks are now surpassing the
accuracy of even the best Cs fountain clocks. To demonstrate such a performance a
comparison with another optical clock is required. Recently T. Rosenband and co-
workers at NIST have succeeded with such a measurement that shows the stability
of two optical clocks based on single trapped Hg and Al ions. The frequency ratio
measured as sketched at the right hand side of Fig. 6.3 has been found to be
reproducible within 5.2 1017 [26].
An overview of the evolution of radio frequency (cesium) clocks versus optical
standards over the last decades is presented in Fig. 6.4. Whereas the cesium atomic
clocks have steadily improved by an order of magnitude per decade, the optical
frequency standards have significantly gained in pace since the introduction of
frequency combs. Because existing optical clocks are still quite complicated devices
and different atoms and ions are still being investigated it will be some time before
a particular transition is chosen to re-define the SI second.
The possibility to readily count optical frequencies has opened up new experimental
possibilities. High precision measurements on hydrogen have allowed for improved
6 Femtosecond Optical Frequency Combs 137
Fig. 6.4 Evolution of radio frequency (rf) and optical frequency standards. The rf standards gain
quite steadily an order of magnitude per decade (straight line). The optical measurements are on a
different slope since they can be conducted with frequency combs. Now the best optical clocks are
outperforming even the best cesium atomic clocks [26]
these observations with laboratory measurements, even though the look back time
is typically about 9 orders of magnitude smaller.
In the first round of these experiments the hydrogen 1S-2S [29], and transitions in
single trapped Hg [31] and Yb ions [32] were used by comparing them with the best
cesium atomic clocks. Since the 133 Cs ground state hyper-fine splitting is determined
by the nuclear structure, the strong interaction mainly scales the reference frequency.
It is therefore impossible to distinguish variations of the fine structure constant
from the strong coupling constant using a single atomic system. If such a variation
actually takes place, grand unified theories predict a synchronization of the various
coupling constants [33] such that it may cancel out from the observation of a
single system. Fortunately H, YbC and HgC possess different dependences on the
fine structure constant so that the contributions of the coupling constants can be
separated [34].
An even better way to detect a drifting fine structure constant is to observe optical
frequency ratios as discussed in Sect. 6.3. The above mentioned comparison of the
HgC and AlC optical clocks provides this unique possibility. After only 1 year
of observation this frequency ratio puts the smallest limit of 5.2 1017 year1
on the drift of the fine structure constant [26]. Remarkably, the precision of these
laboratory measurements makes it possible to reach the same sensitivity within
a few years of monitoring, that astronomical and geological observations require
billions of years of look back time. In the laboratory, systematics can be investigated
or challenged, and if in doubt, experiments may be repeated given the relative short
time intervals. So far none of the laboratory tests have been able to detect a temporal
variation of the fine structure constant. Geological and astronomical observations
may be affected by systematic uncertainties that lead to partially contradicting
results (see Refs. [35–37] for results on quasar absorption and [38, 39] geological
data). It should however be pointed out that laboratory tests operate on a different
epoch than astronomical and geological observations so that the data may not be
directly comparable.
A very important application for improved atomic clocks in basic research could
be to test or refine general relativity which still poses one of the major problems
in physics as it refuses proper quantization. It is possible that small, yet undetected,
deviations from the classical theory of gravity give valuable hints on how to proceed
with quantization.
Laser-like pulsed emission of very short wavelength can be produced with a method
generally referred to as high harmonic generation (HHG) [40]. For this purpose one
focuses infrared femtosecond pulses in a gas target employing its extreme nonlinear
response when subject to intensities exceeding 5 1012 Wcm2 . In most HHG
experiments a nozzle emitting a jet of noble gas atoms is placed near the focus of an
6 Femtosecond Optical Frequency Combs 139
Fig. 6.5 Very short pulses driving a very non-linear process such as high harmonic generation
close to the short wavelength cut-off. If the intensity and/or the non-linearity is well chosen such
that it sets in at a certain threshold, a “cos” pulse will produce a spectrum that is periodic with the
infrared carrier frequency because the threshold is reached twice per pulse (left). In contrast the
“sine” pulse (right) produces a single isolated harmonic pulse that is much shorter than an optical
cycle of the infrared driving pulse [73]
used to increase the sensitivity for trace gas analysis by coupling the probing comb
into the sample holding cavity and applying the multi-heterodyne method with the
transmitted light [68].
Non-resonant intra-cavity ring down spectroscopy has been used for a long time
to enhance sensitivity. With the introduction of the frequency comb this method has
been extended to operate simultaneously on many optical frequencies at once [69].
Adjusting the mode spacing of the enhancing cavity to be an integer ratio
(say 100:101) of the combs mode spacing, allows pre-selection of every 100th
mode from the comb. This means that a simple low resolution grating is capable
of resolving the transmitted modes. By scanning the cavity length this cavity
successively resonates with sub groups of modes very much like a Vernier calliper
in the frequency domain [70]. This method combines a simple set-up with only one
frequency comb, the possibility to (sequentially) resolve individual modes of the
comb and the high sensitivity of cavity enhanced sensing.
The application of frequency combs has also enabled advances in another field as
it allows the possibility for stabilizing the carrier envelope phase [71]. According
to Fig. 6.1 a pulse train with a vanishing ! CE has a fixed phase of the carrier with
respect to the envelope. This means that all the pulses have the same electric field.
With the technique of self-referencing this can be readily accomplished. However,
even though the electric field transients of the pulses are identical in this case, it is
unknown what value the carrier envelope phase actually assumes upon stopping its
pulse to pulse slippage.
To detect the value of the carrier envelope phase, very short pulses that drive
processes which depend on the electric field in high order are used. Figure 6.5 shows
two extreme values of the carrier envelope phase that correspond to a “cos” and a
6 Femtosecond Optical Frequency Combs 143
“sine” pulse. Close observation of the field transients reveals that the peak electric
field slightly depends on the carrier envelope phase. In addition the duration of the
optical carrier cycle changes slightly due to the steep pulse envelope when measured
say between two field maxima. These effects are largely enhanced if the pulses are
short and can be detected using HHG’s or above threshold ionisation [72].
Two important aspects are realized from this: For “cos” driving pulse, the high
harmonic pulse is much shorter than one cycle of the generating field. So by
stabilizing the carrier envelope phase to the proper value (“cos” drive pulses), single
isolated attoseconds could be produced for the first time [74]. Secondly the high
harmonic spectrum reveals whether the carrier envelope phase has been fixed at the
proper value or not. A “sine” drive pulse creates two high harmonic pulses per laser
pulse, that have almost, but not quite the same, time separation given by the carrier
frequency. For this reason the spectrum shows well separated harmonic peaks even
at the shortest wavelength that are generated. In contrast to that, a “cos” produces a
shifted harmonic spectrum that loses its periodicity in this region. This is because it
belongs to a single isolated attosecond pulse [73].
Not only that the detection and stabilization of the carrier envelope phase allowed
the production of attosecond pulses, it also allowed to completely recover the
electric field transients of ultra-short pulses. The work in that direction relied
on measuring the pulses autocorrelation which, together with the determination
of the carrier envelope phase via spectral analysis of high harmonics, allows the
calculation of the field transient [74]. A more direct measurement uses cross
correlation between attosecond pulses generated the way described above with the
driving pulses. In this way the field transient of the latter can be sampled with a
temporal resolution significantly shorter than one optical cycle [75].
6.5 Conclusions
Optical frequency combs provide a phase coherent link between the radio frequency
domain and the frequency domain of lasers. This possibility enabled precise
determination of atomic and molecular transition frequencies. In addition, optical
atomic clocks are realized by using a frequency comb as a clock work mechanism.
Optical frequency combs have also found numerous other applications as described
in this contribution.
References
1. L.O. Hocker, A. Javan, D.R. Rao, L. Frenkel, Absolute frequency measurement and spec-
troscopy of gas laser transitions in the far infrared. Appl. Phys. Lett. 10, 147–149 (1967)
2. K.M. Evenson, J.S. Wells, F.R. Petersen, B.L. Danielson, G.W. Day, Accurate frequencies of
molecular transitions used in laser stabilization: the 3.39- m transition in CH4 and the 9.33-
and 10.18- m transitions in CO2 . Appl. Phys. Lett. 22, 192–195 (1973)
144 T. Udem et al.
26. T. Rosenband, D.B. Hume, P.O. Schmidt, C.W. Chou, A. Brusch, L. Lorini, W.H. Oskay, R.E.
Drullinger, T.M. Fortier, J.E. Stalnaker, S.A. Diddams, W.C. Swann, N.R. Newbury, W.M.
Itano, D.J. Wineland, J.C. Bergquist, Frequency ratio of AlC and HgC single-ion optical
clocks; metrology at the 17th decimal place. Science 319, 1808–1812 (2008)
27. B. de Beauvoir, C. Schwob, O. Acef, L. Jozefowski, L. Hilico, F. Nez, L. Julien, A. Clairon,
F. Biraben, Metrology of the hydrogen and deuterium atoms: determination of the Rydberg
constant and Lamb shifts. Eur. Phys. Lett. D 12, 61–93 (2000)
28. G. Santarelli, P. Laurent, P. Lemonde, A. Clairon, A.G. Mann, S. Chang, A.N. Luiten, C.
Salomon, Quantum projection noise in an atomic fountain: a high stability cesium frequency
standard. Phys. Rev. Lett. 82, 4619–4623 (1999)
29. M. Fischer, N. Kolachevsky, M. Zimmermann, R. Holzwarth, T. Udem, T.W. Hänsch, M.
Abgrall, J. Grünert, I. Maksimovic, S. Bize, H. Marion, F. Pereira Dos Santos, P. Lemonde,
G. Santarelli, P. Laurent, A. Clairon, C. Salomon, M. Haas, U.D. Jentschura, C.H. Keitel, New
limits on the drift of fundamental constants from laboratory measurements. Phys. Rev. Lett.
92, 230802 (2004)
30. P.A.M. Dirac, The cosmological constants. Nature 139, 323 (1937)
31. S. Bize, S.A. Diddams, U. Tanaka, C.E. Tanner, W.H. Oskay, R.E. Drullinger, T.E. Parker,
T.P. Heavner, S.R. Jefferts, L. Hollberg, W.M. Itano, J.C. Bergquist, Testing the stability of
fundamental constants with the 199 HgC single-ion optical clock. Phys. Rev. Lett. 90, 150802
(2003)
32. E. Peik, B. Lipphardt, H. Schnatz, T. Schneider, C. Tamm, S.G. Karshenboim, Limit on the
present temporal variation of the fine structure constant. Phys. Rev. Lett. 93, 170801 (2004)
33. X. Calmet, H. Fritzsch, Symmetry breaking and time variation of gauge couplings. Phys. Lett.
B 540, 173–178 (2002)
34. M. Zimmermann, M. Fischer, N. Kolachevsky, R. Holzwarth, T. Udem, T.W. Hänsch, M.
Abgrall, J. Grünert, I. Maksimovic, S. Bize, H. Marion, F. Pereira Dos Santos, P. Lemonde,
G. Santarelli, P. Laurent, A. Clairon, C. Salomon, High-resolution laser spectroscopy and time
variation of fundamental constants. Laser Phys. 15, 997–1009 (2005)
35. J.K. Webb, M.T. Murphy, V.V. Flambaum, V.A. Dzuba, J.D. Barrow, C.W. Churchill, J.X.
Prochaska, A.M. Wolfe, Further evidence for cosmological evolution of the fine structure
constant. Phys. Rev. Lett. 87, 091301 (2001)
36. R. Srianand, H. Chand, P. Petitjean, B. Aracil, Limits on the time variation of the electromag-
netic fine-structure constant in the low energy limit from absorption lines in the spectra of
distant quasars. Phys. Rev. Lett. 92, 121302 (2004)
37. M.T. Murphy, J.K. Webb, V.V. Flambaum, Revision of VLT/UVES constraints on a varying
fine-structure constant. arXiv.org, e-Print Arch., Astrophys. 1, 0612407 (2006)
38. Y. Fujiia, A. Iwamotoc, T. Fukahoric, T. Ohnukic, M. Nakagawac, H. Hidakad, Y. Ourae, P.
Möllerf, The nuclear interaction at Oklo 2 billion years ago. Nucl. Phys. B 573, 377–401 (2000)
39. S.K. Lamoreaux, J.R. Torgerson, Neutron moderation in the Oklo natural reactor and the time
variation of ˛. Phys. Rev. D 69, 121701(R) (2004)
40. J.G. Eden, High-order harmonic generation and other intense optical field-matter interactions:
review of recent experimental and theoretical advances. Prog. Quantum Electron 28, 197–246
(2004)
41. J. Seres, E. Seres, A.J. Verhoef, G. Tempea, C. Streli, P. Wobrauschek, V. Yakovlev, A. Scrinzi,
C. Spielmann, F. Krausz, Laser technology: source of coherent kiloelectronvolt X-rays. Nature
433, 596 (2005)
42. P. Fendel, S.D. Bergeson, T. Udem, T.W. Hänsch, Two-photon frequency comb spectroscopy
of the 6s-8s transition in cesium. Opt. Lett. 32, 701–703 (2007)
43. M.C. Stowe, M.J. Thorpe, A. Pe’er, J. Ye, J.E. Stalnaker, V. Gerginov, S.A. Diddams, Direct
frequency comb spectroscopy. Adv. At. Mol. Opt. Phys. 59, 1–60 (2008)
44. A.L. Wolf, S.A. van den Berg, W. Ubachs, K.S.E. Eikema, Direct frequency comb spectroscopy
of trapped ions. Phys. Rev. Lett. 102, 223901 (2009)
45. S. Reinhardt, E. Peters, T.W. Hänsch, T. Udem, Two-photon direct frequency comb spec-
troscopy with chirped pulses. Phys. Rev. A 81, 033427 (2010)
146 T. Udem et al.
46. Y.F. Baklanov, V.P. Chebotayev, Narrow resonances of two-photon absorption of super-narrow
pulses in a gas. Appl. Phys. Lett. 12, 97–99 (1977)
47. M. Bellini, T.W. Hänsch, Measurement of the temporal coherence of ultrashort harmonic
pulses: towards coherent spectroscopy in the extreme ultraviolet. Appl. Phys. B: Lasers Opt.
65, 677–680 (1997)
48. D.C. Yost, T.R. Schibli, J. Ye, J.L. Tate, J. Hostetter, M.B. Gaarde, K.J. Schafer, Vacuum-
ultraviolet frequency combs from below-threshold harmonics. Nat. Phys. 5, 815–820 (2009)
49. C. Gohle, T. Udem, M. Herrmann, J. Rauschenberger, R. Holzwarth, H.A. Schuessler, F.
Krausz, T.W. Hänsch, A frequency comb in the extreme ultraviolet. Nature 436, 234–237
(2005)
50. R.J. Jones, K.D. Moll, M.J. Thorpe, J. Ye, Phase-coherent frequency combs in the vacuum
ultraviolet via high-harmonic generation inside a femtosecond enhancement cavity. Phys. Rev.
Lett. 94, 193201 (2005)
51. A. Ozawa, J. Rauschenberger, C. Gohle, M. Herrmann, D.R. Walker, V. Pervak, A. Fernandez,
R. Graf, A. Apolonski, R. Holzwarth, F. Krausz, T.W. Hänsch, T. Udem, High harmonic
frequency combs for high resolution spectroscopy. Phys. Rev. Lett. 100, 253901 (2008)
52. D.C. Yost, T.R. Schibli, J. Ye, Efficient output coupling of intracavity high-harmonic genera-
tion. Opt. Lett. 33, 1099–1101 (2008)
53. M. Herrmann, M. Haas, U.D. Jentschura, F. Kottmann, D. Leibfried, G. Saathoff, C. Gohle,
A. Ozawa, V. Batteiger, S. Knünz, N. Kolachevsky, H.A. Schüssler, T.W. Hänsch, T. Udem,
Feasibility of coherent xuv spectroscopy on the 1s-2s transition in singly ionized. Phys. Rev.
A 79, 052505 (2009)
54. M.T. Murphy, T. Udem, R. Holzwarth, A. Sizmann, L. Pasquini, C. Araujo-Hauck, H. Dekker,
S. D’odorico, M. Fischer, T.W. Hänsch, High-precision wavelength calibration of astronomical
spectrographs with laser frequency combs. Mon. Not. R. Astron. Soc. 380, 839–847 (2007)
55. C.H. Li, A.J. Benedick, P. Fendel, A.G. Glenday, F.X. Kärtner, D.F. Phillips, D. Sasselov,
A. Szentgyorgyi, R.L. Walsworth, A laser frequency comb that enables radial velocity
measurements with a precision of 1 cm.s1 . Nature 452, 610–612 (2008)
56. T. Steinmetz, T. Wilken, C. Araujo-Hauck, R. Holzwarth, T.W. Hänsch, L. Pasquini, A.
Manescau, S. D’odorico, M.T. Murphy, T. Kentischer, Laser frequency combs for astronomical
observations. Science 321, 1335–1337 (2008)
57. M. Mayor, X. Bonfils, T. Forveille, X. Delfosse, S. Udry, J.-L. Bertaux, H. Beust, F. Bouchy, C.
Lovis, F. Pepe, C. Perrier, D. Queloz, N.C. Santos, The HARPS search for southern extra-solar
planets XV9. An Earth-mass planet in the GJ 581 planetary system. Astron. Astrophys. 507,
487–494 (2009)
58. C. Lovis, M. Mayor, F. Pepe, Y. Alibert, W. Benz, F. Bouchy, A.C.M. Correia, J. Laskar, C.
Mordasini, D. Queloz, N.C. Santos, S. Udry, J.-L. Bertaux, J.-P. Sivan, An extrasolar planetary
system with three Neptune-mass planets. Nature 441, 305–309 (2006)
59. D.A. Braje, M.S. Kirchner, S. Osterman, T. Fortier, S.A. Diddams, Astronomical spectrograph
calibration with broad-spectrum frequency combs. Eur. Phys. J. D 48, 5766 (2008)
60. M.S. Kirchner, D.A. Braje, T.M. Fortier, A.M. Weiner, L. Hollberg, S.A. Diddams, Generation
of 20 GHz, sub-40 fs pulses at 960 nm via repetition-rate multiplication. Opt. Lett. 34, 872–874
(2009)
61. G. Chang, C.-H. Li, D.F. Phillips, R.L. Walsworth, F.X. Kärtner, Toward a broadband astro-
comb: effects of nonlinear spectral broadening in optical fibers. Opt. Express 18, 12736–12747
(2010)
62. A. Manescau, G.L. Curto, L. Pasquini, J.I.G. Hernández, R. Rebolo, T. Steinmetz, R. Probst,
R. Holzwarth, T. Wilken, T.W. Hänsch, T. Udem, Approaching cm/sec calibration of high
resolution astronomical spectrograph. in Proceedings of the Conference on Lasers and Electro-
Optics paper JS91.5 THU (Baltimore, USA, 2011)
63. A. Bartels, D. Heinecke, S.A. Diddams, 10-GHz self-referenced optical frequency comb.
Science 326, 681 (2009)
6 Femtosecond Optical Frequency Combs 147
64. T. Wilken, C. Lovis, A. Manescau, T. Steinmetz, L. Pasquini, G.L. Curto, T.W. Hänsch, R.
Holzwarth, T. Udem, High-precision calibration of spectrographs. Mon. Not. R. Astron. Soc.
405, L16–L20 (2010)
65. V. Gerginov, C.E. Tanner, S.A. Diddams, A. Bartels, L. Hollberg, High-resolution spectroscopy
with a femtosecond laser frequency comb. Opt. Lett. 30, 1734–1736 (2009)
66. S. Schiller, Spectrometry with frequency combs. Opt. Lett. 27, 766–768 (2009)
67. I. Coddington, W.C. Swann, N.R. Newbury, Coherent multiheterodyne spectroscopy using
stabilized optical frequency combs. Phys. Rev. Lett. 100, 013902 (2008)
68. B. Bernhardt, A. Ozawa, P. Jacquet, M. Jacquey, Y. Kobayashi, T. Udem, R. Holzwarth,
G. Guelachvili, T.W. Hänsch, N. Picqué, Cavity-enhanced dual-comb spectroscopy. Nat.
Photonics 4, 55–57 (2010)
69. M.J. Thorpe, K.D. Moll, R.J. Jones, B. Safdi, J. Ye, Broadband cavity ringdown spectroscopy
for sensitive and rapid molecular detection. Science 311, 1595–1599 (2006)
70. C. Gohle, B. Stein, A. Schliesser, T. Udem, T.W. Hänsch, Frequency comb Vernier spec-
troscopy for broadband, high-resolution, high-sensitivity absorption and dispersion spectra.
Phys. Rev. Lett. 99, 263902 (2007)
71. D.J. Jones, S.A. Diddams, J.K. Ranka, A. Stentz, R.S. Windeler, J.L. Hall, S.T. Cundiff,
Carrier-envelope phase control of femtosecond mode-locked lasers and direct optical frequency
synthesis. Science 288, 635–639 (2000)
72. G.G. Paulus, F. Lindner, H. Walther, A. Baltuška, E. Goulielmakis, M. Lezius, F. Krausz,
Measurement of the phase of few-cycle laser pulses. Phys. Rev. Lett. 85, 253004 (2000)
73. R. Kienberger, E. Goulielmakis, M. Uiberacker, A. Baltuška, V. Yakovlev, F. Bammer, A.
Scrinzi, T. Westerwalbesloh, U. Kleineberg, U. Heinzmann, M. Drescher, F. Krausz, Atomic
transient recorder. Nature 427, 817–821 (2004)
74. A. Baltuška, T. Udem, M. Uiberacker, M. Hentschel, E. Goulielmakis, C. Gohle, R. Holzwarth,
V.S. Yakovlev, A. Scrinzi, T.W. Hänsch, F. Krausz, Attosecond control of electronic processes
by intense light fields. Nature 421, 611–615 (2003)
75. E. Goulielmakis, M. Uiberacker, R. Kienberger, A. Baltuška, V. Yakovlev, A. Scrinzi, T.
Westerwalbesloh, U. Kleineberg, U. Heinzmann, M. Drescher, F. Krausz, Direct measurement
of light waves. Science 305, 1267–1269 (2004)
Chapter 7
Ultrafast Material Science Probed Using
Coherent X-ray Pulses from High-Harmonic
Generation
7.1 Introduction
X-rays represent one of the most powerful tools for understanding molecules and
materials at the nanoscale, uncovering important information related to magnetism,
photochemistry, materials, biology, nanoscience and many other areas of science
and technology.
To date, research in x-ray science has predominantly been carried out using
synchrotron light sources, with great success due to the power of x-rays as probes
of matter. Their short wavelengths provide access to elemental absorption edges, to
site-specific chemical and magnetic states of matter, as well as ultrahigh spatial
resolution imaging capabilities. In recent years, advances in laser science have
made it possible to generate x-ray pulses at the femto- and attosecond frontiers
using either free-electron laser facilities (“4th generation” synchrotron) or high-
harmonic up-conversion of femtosecond lasers (also see Chap. 3 by Marangos
and Malte Oppermann). The availability of such ultrashort x-ray pulses paves the
way for a completely new generation of experiments that can capture the coupled
dynamics of elementary excitations in materials. Ultrashort x-ray pulses can access
the fundamental interactions between charge, lattice, orbital, and spin dynamics in
real time, which eventually determine the intrinsic physical limits at which, for
example, a phase-transition in a correlated-electron material occurs, the magnetic
state of a material can be switched, or a chemical reaction on a surface evolves.
In biology, these new ultrafast x-ray light sources promise to capture the most basic
processes, such as charge transfer, photosynthesis or cell death, on their fundamental
time- and length-scales, using the “diffraction before destruction” concept.
The promise of ultrafast x-ray science has fuelled the impressive free-electron
laser facilities that are coming online worldwide, which deliver high energy bursts
of femtosecond x-rays to capture an image in a single shot. Recent breakthroughs
in coherent x-ray generation using high-harmonic up-conversion provide a com-
plementary small-scale light source that is ideal for the study of ultrafast material
science [1–3]. The key enabling advance for generating bright compact ultrafast
x-ray sources was the extension of phase-matching techniques from the extreme
ultraviolet region (XUV) to the soft x-ray region to photon energies >1 keV.
The goal of this chapter is to give a short introduction to high-harmonic
generation (HHG) (Sect. 7.2), followed by a summary of recent breakthroughs in
using HHG for the study of ultrafast material science (Sect. 7.3). The experiments
discussed in this book chapter were all carried out using photon energies in the
40–100 eV photon energy range, where high flux HHG beams are readily available
at high (> kHz) repetition rates. In the near future, however, equally high photon flux
table-top high-harmonic beamlines will be available at photon energies up to 1 keV,
which will have great potential impact on material science research, as discussed in
Sect. 7.4 “Conclusions and Outlook”.
Fig. 7.1 Illustration of the three-step model for high-harmonic generation. (Top) An electron
tunnels from the atom in the presence of a strong laser field that suppresses the Coulomb potential
binding the electron to the atom. The free electron is then accelerated by the laser field and can
recombine with the parent ion, emitting a high-energy photon. (Bottom) Quantum representation
of the recolliding electron wavepacket in HHG (Reproduced and adapted with permission from
Ref. [1])
the electron tunnel ionises, the energy it gains can range between 0 and 3.17 Up –
the latter representing the maximum kinetic energy that the electron can gain if it
ionises just after the peak of the laser field. As a result, when the electron recombines
with its parent ion, the energy of the highest cutoff harmonic that can be emitted is
given by:
Fig. 7.2 Phase-matched HHG spectrum generated in 20 Torr of Ar using a waveguide geometry
(Reproduced and adapted with permission from Ref. [14])
Ionisation of the medium occurs as a sequence of bursts near the peaks of the
field, and occurs twice per cycle. This leads to harmonic emission that occurs as a
series of short attosecond bursts in time, corresponding to a series of odd harmonics
of the fundamental laser (¨, 3¨, 5¨ etc.) in frequency space, up to the cutoff photon
energy „!cut off . Furthermore, harmonics are emitted only during a fraction of the
laser pulse when the atom is ionising and very rapid changes in the time-dependent
dipole moment are present. As a result, the total HHG pulse duration is considerably
shorter than the driving laser pulse [10, 11]. For sufficiently short driving pulse
duration, the highest harmonics can be confined primarily to a single attosecond
burst [11, 12]. Moreover, when phase matching is taken into account, that specifies
the time interval during which emission from a large number of atoms is coherently
combined, bright harmonic bursts as short as 200 attoseconds can be generated in
the XUV region of the spectrum, even using 15 fs lasers at a wavelength of 0.8 m
[13]. Even shorter HHG pulses, below 10 attoseconds, can be generated in the soft
x-ray region of the spectrum using mid-infrared driving lasers with pulse durations
of tens of femtoseconds (Fig. 7.2) [1].
The development of sub-30 fs solid-state lasers based on Ti:sapphire was
instrumental to the first practical applications of HHG [15–17]. The efficiency of
HHG increases as the driving laser pulse duration is reduced, because the required
laser pulse energy is less in order to reach ionising intensities of 1014 Wcm2 . In
addition, as explained below, phase-matching conditions are dramatically improved
by using 20 fs laser pulses, resulting in efficient generation of harmonics with
efficiency 105 per harmonic per pulse in the XUV region of the spectrum.
For phase matched frequency upconversion, the driving laser and the generated
harmonics must travel with the same phase velocity, to ensure constructive addition
7 Ultrafast Material Science Probed Using Coherent X-ray Pulses. . . 153
of the generated HHG field from a macroscopic number of atoms over an extended
propagation distance. In the XUV, the real part of the refractive index is close
to unity, and thus high harmonics travel at phase velocities close to c (the speed
of light in vacuum). In the case of the laser field, the refractive index of neutral
atoms is greater than unity, corresponding to a phase velocity less than c for the
laser. However, the gas medium is being ionised by the laser as it propagates.
Since the index of refraction of a plasma is less than unity, the phase velocity
of the driving laser is dynamically changing, from less than c on the leading
edge of the pulse to greater than c once significant ionisation of the medium
occurs. In general, there will be a finite time interval during which phase matching
of HHG can occur, provided that the required HHG photon energy (i.e. laser
intensity) is reached while the medium is still only weakly ionised (<5 % for
Argon gas).
A very successful approach to phase matching of HHG that scales to keV photon
energies and that results in perfectly spatially coherent beams, uses a gas-filled
waveguide as the generation medium [1, 2, 18–21]. Laser beams well above the
ionising intensity can be reliably guided in a gas-filled hollow waveguide without
damage to the structure. In this case, phase matching occurs when the phase
velocities of the laser and HHG are equal, or equivalently, the difference in their
k-vectors of propagation is zero:
In Eq. (7.2), q is the harmonic order, u11 is the lowest-order waveguide mode
factor, L is the centre wavelength of the driving laser, a is the inner radius of the
hollow waveguide, P is the pressure, is the ionisation fraction, re is the classical
electron radius, Natm is the number density of atoms at 1 atm, ın is the difference
between the indices of refraction at the fundamental and harmonic wavelengths,
and kquantum , often negligible in a waveguide geometry, accounts for any Gouy or
other phase shifts in the focus. The gas-filled hollow waveguide ensures that a well-
defined spatial mode builds up over an extended propagation distance and allows for
careful control to engineer the phase-matched conditions including the wavelength,
gas pressure, gas species and waveguide size. To achieve phase matching, the gas
pressure is tuned so that the neutral and plasma index of the gas balances the
waveguide dispersion. The waveguide geometry thus makes it possible to obtain
optimal conversion efficiency in a wide range of conditions in a compact geometry
(i.e., 1 m beamline) using pulse energies readily obtainable using kHz and higher
repetition-rate lasers. The waveguide geometry has been used in all experiments
following in Sect. 7.3 of this book chapter.
Expanding the useful range of photon energies that can be generated using HHG
depends on continued progress in understanding how to phase match the process.
154 S. Mathias et al.
Dramatic progress has been achieved in this area in the past 2 years, with the
demonstration of mid-infrared phase matching in the keV region [1, 2, 22]. Other
approaches include new techniques for quasi phase matching using periodic struc-
tures [23–25] or counter-propagating light pulses [26] that periodically modulate
the emission and can thus facilitate coherent signal build-up at higher energies
than is possible through simple phase matching. The most successful approach
to date is to generate HHG using longer wavelengths which allows for full phase
matching of the HHG process at very high multi-atmosphere pressures, enabling
conversion efficiencies into the water window and keV region of the spectrum
predicted to be comparable to those currently obtained in the XUV [1, 2]. These
future developments will continue to expand the capabilities of HHG based ultrafast
material science research.
Fig. 7.3 Photoemission and conservation of the k-vector component parallel to the surface
energy h and is promoted into the final state Ef . The final kinetic energy Ekin of the
photoelectron is measured with the electron energy analyser and is given by:
where W is the work function of the material, and Eb the binding energy of the initial
state with respect to the Fermi-level EF (Fig. 7.3a). Because of the 2D translational
symmetry parallel to the surface, the wave-vector component parallel to the surface
is conserved (see Fig. 7.3b):
kEjjex D kEjji n C G
E jj (7.4)
Where kEjjex and kEjji n are the parallel wave-vector components of the electron outside
and inside the crystal, respectively, and G E jj a reciprocal lattice vector of the crystal.
156 S. Mathias et al.
The kE? component normal to the surface is not conserved during transmission
through the surface. The wave-vector component parallel to the surface outside the
crystal, which is determined from known experimental parameters,
r r
2me p 2me
kEjjex D h jEb j W sin.‚/ D Ekin sin.‚/ (7.5)
„2 „2
therefore directly yields the internal wave-vector component kEjji n for a kinetic energy
Ekin and emission angle of the photoemitted electron. Modern electron analysers
measure the photoelectron energy and emission angle in parallel, and thus directly
map the electronic structure E(kjj ). Figure 7.3c depicts the dependence of the
photoelectron emission angle on the wave-vector component kEjjex parallel to the
surface for three different photon energies and a typical workfunction of W D 5 eV.
For the lowest photon energy, h D 6 eV (and Ekin D h W D 1 eV), which can
conveniently be reached by frequency quadrupling the output of a Ti:sapphire
ultrafast laser system in two beta-barium-borate crystals, the maximum kEjjex wave
vector that can be measured (for a theoretical emission angle of 90ı ) is about
0.5 Å1 . In order to reach typical Brillouin-zone boundaries at about 1.5 Å1 , which
is crucial for the study of ultrafast correlated-electron physics in materials, higher
photon energies and therefore XUV and soft x-ray pulses from HHG or free-electron
laser (FEL) sources are needed.
The first experiment that implemented and used the capabilities of time-resolved
ARPES with HHG XUV light for the study of correlated-electron materials was
recently carried out by Rohwer et al. [40]. Here, the charge-density wave state
in the potential excitonic insulator 1 T-TiSe2 was disrupted on a timescale as
short as 20 fs. The photo-induced phase transition was monitored using ARPES,
enabling real-time stroboscopic imaging of the electronic band dispersions E(kjj ) at
large momentum, in the vicinity of the first Brillouin zone. The transition metal
dichalcogenide 1 T-TiSe2 undergoes a (2 2 2) charge-density wave (CDW)
phase transition below a temperature of 200 K. The transition is accompanied
by a periodic lattice distortion, where the real-space unit cell doubles its size in
all three directions. In momentum space, the dimensions of the Brillouin zone
are correspondingly halved, and this effect is most prominently resolved using
conventional APRES, by monitoring the consequent backfolding of the Se 4p bands
at the Brillouin zone boundary (M -point). Figure 7.4 (from Ref. [40]), top left,
shows the ARPES intensity map of 1 T-TiSe2 recorded with HHG XUV pulses in
the charge-density wave phase. The photoemission intensity in the vicinity of the
M -point is due to the backfolded Se 4p bands (“Folded Se 4p”) in the CDW phase,
and it is this signature that is used to track the ultrafast photo-induced vaporization of
the charge-density long-range order in the material. Figure 7.4b–e shows snapshots
of the electronic structure as a function of delay between the 32 fs infrared pump-
pulses, and the sub 10 fs XUV probe pulses. When pump and probe pulses overlap
(b), the Ti 3d band is filled with hot electrons. These transiently generated free
7 Ultrafast Material Science Probed Using Coherent X-ray Pulses. . . 157
Fig. 7.4 Tracking the photoinduced transition by femtosecond time-resolved ARPES [40]. (a)
ARPES intensity map of 1T-TiSe2 recorded with high- harmonic XUV pulses in the CDW phase
(temperature T D 125 K). (b–e) Time-resolved ARPES snapshots at increasing pump (infrared)–
probe (XUV) temporal delays (pump-pulse fluence, 5 mJ/cm2 ). (f) Photoemission transients of the
folded Se 4p band for different infrared excitation fluences close to time zero. The infrared–XUV
cross-correlation signal, which has been determined from the laser assisted photoemission [40]
(LAPE) signal, is added. Inset, transients up to temporal delays of 3.5 ps (Reproduced and adapted
with permission from Ref. [40])
158 S. Mathias et al.
material. The physical origin of the magneto-optical effect lies in the influence of
exchange interaction and spin-orbit interaction on the transition probability between
initial and final states for the electrons that are excited by the incoming light. In
the so-called longitudinal and polar geometries (in-plane magnetization parallel to
plane of incidence, and out-of-plane magnetization parallel to plane of incidence,
respectively), linearly polarised light that is reflected from the magnetic material
shows a rotation of the polarisation plane (Kerr angle) and it becomes elliptically
polarised. In the transversal geometry (in-plane magnetization perpendicular to
plane of incidence), as used by La-o-vorakiat et al. the M-edge magneto-optical
effect results predominantly from spin-orbit splitting and the unfilled d-band spin
polarisation [48]. In this geometry, the reflected light shows a modulation of the
intensity depending on the magnetization direction of the material. Therefore, no
polarisation analysis and no circular polarised light is needed, which considerably
relaxes experimental requirements.
Taking the two reflection coefficients for the two magnetization directions for
p-polarised modes from [49], the normalized difference for the T-MOKE signal
is [50]:
" #
2"0 "2 sin 2i
A D Re 2 (7.6)
"1 cos i "0 "1 C "20 sin2
2
"1 and "2 are the diagonal and off-diagonal elements of the dielectric tensor,
respectively, and i the angle of incidence with respect to the surface normal.
Therefore, the T-MOKE is directly related to the off-diagonal permittivity tensor.
Figure 7.5 shows the measured T-MOKE asymmetry signal of a Permalloy
sample (using synchrotron light), which contains Fe and Ni atoms, in grey scale
as a function of energy and angle of incidence [51]. The asymmetry signal as given
by Eq. (7.6) shows highest asymmetry values for angles of incidence around 45
degrees. The Fe magnetic asymmetry signal is visible below energies of 60 eV,
while the Ni magnetic asymmetry signal appears above 60 eV, as expected since
the elemental M absorption edges are around 54 eV for Fe and 67 eV for Ni. In the
experiment by La-o-vorakiat et al. which was carried out using coherent XUV pulses
from HHG, the magnetism of a grating structure made of Permalloy (Ni80 Fe20 )
was investigated (where the grating structure was used to spectrally resolve the
harmonics) [47].
The recorded HHG spectrum with this setup is shown in Fig. 7.6, illustrating the
discrete harmonics diffracted from the Permalloy grating. The strong absorption at
72 eV is due to aluminium filters that are used to block the fundamental laser light
that is co-propagating with the harmonics. From this spectrum, it is clear that the
magnetic asymmetry can be mapped over the entire M-edge region of Ni ( 67 eV)
and Fe ( 54 eV).
After the sample was excited by an intense infrared pump pulse, the very short
(<10 fs) duration HHG light was used to probe the ultrafast demagnetization
processes – with element specificity for the first time. Initially, the ultrafast intense
160 S. Mathias et al.
Fig. 7.5 Variations of T-MOKE signal from a Permalloy film consisting of Fe and Ni as a function
of photon energy E and angle of incidence ™ [51] (Data courtesy P. Grychtol, R. Adam and C.M.
Schneider, Forschungszentrum Jülich, Germany)
Fig. 7.6 Experimental data showing the HHG spectrum (bottom, right axis) and measured
magnetic asymmetry (top, left axis) for different incidence angles. No asymmetry is measured
with s-polarised light. (Left to right) The vertical lines denote Fe, Ni and Al M-shell absorption
edge (Reproduced and adapted with permission from Ref. [47])
7 Ultrafast Material Science Probed Using Coherent X-ray Pulses. . . 161
pump pulse coherently interacts with the electron and spin systems during the time
the laser pulse is incident on the material [52]. Thermalisation of the electron and
spin systems then takes place through incoherent scattering processes, resulting in a
subsequent reduction/re-orientation in the magnetization [43, 53–55]. The change
in asymmetry as a function of time after the IR pump pulse can then be used
to follow the demagnetization dynamics (see Figs. 7.6 and 7.7). As expected, the
Permalloy sample demagnetizes on a femtosecond timescale, with an exponential
decay time-constant of below 400 fs for both Fe and Ni in Permalloy. The time
resolution of this first proof-of-principle experiment was about 55 fs, limited by
the experimental geometry. In the future, by using a more-collinear pump probe
geometry and a shorter HHG pulse, unprecedented time-resolution for ultrafast
magnetic studies will be available, which may, e.g., enable the direct investigation
of ultrafast coherent magnetization dynamics [52].
The experiment by La-o-vorakiat et al. clearly demonstrates that T-MOKE with
XUV light from HHG provides an ultrafast, element-specific probe of the magnetic
state. Furthermore, the use of coherent HHG for probing magnetic materials
promises to combine ultrafast, element-specific magnetism with nanometre spatial
resolution (see Sect. 7.3.5), making it possible to answer important fundamental
and technologically relevant questions in ultrafast magnetism. Furthermore, the new
experimental capability presented by La-o-vorakiat et al. is useful in the context of
correlated-electron materials, where ferromagnetic and antiferromagnetic order are
parts of the rich phase diagrams of these materials.
162 S. Mathias et al.
Fig. 7.8 Experimental observation of laser-assisted photoemission from a Pt(111) surface using
HHG light. The XUVCIR spectrum (solid line) shows a strong distortion around the Fermi
edge compared to the XUV-only spectrum (dotted line). This change is due to the appearance of
sidebands in the spectrum. The dashed line shows the fit of the XUVCIR spectrum by convolving
a LAPE response function with the XUV-only spectrum. The inset shows the LAPE response
function associated with the fit. The energy separation between the sidebands is equal to the IR
photon energy, as expected for surface LAPE [64, 67] (Reproduced and adapted with permission
from Ref. [64])
characteristic for the probed-atom species. By comparing the time evolution of the
LAPE response from the Pt substrate with the LAAD signal from the adsorbed
Xe, Miaja-Avila and co-workers measured the lifetime of the 4d core hole initially
created by the XUV pulse (see Fig. 7.10). This measurement was possible despite
the many complexities that result from using the metallic substrate such as above-
threshold photoemission, hot electrons, and space-charge effects. The extracted
lifetime of 7 ˙ 1 fs for the Xe/Pt 4d core hole was consistent with high-resolution
linewidth measurements of Xe in the gas phase [70], clearly validating the success
of this experimental approach. Although carried out in a simple model system of a
noble-gas atom physisorbed on a surface, this experiment paves the way to more
complex charge-transfer processes in surface-adsorbate systems, where complex
electron relaxation processes, which depend on the chemical environment, are
expected.
Understanding heat flow away from a nanoscale hotspot is important for our
fundamental understanding of phonon transport as well as for many areas of
nanotechnology. Applications include thermal management in integrated circuits
or in heat-assisted magnetic recording techniques for future hard-disk devices,
thermoelectrics or thermally-enabled cancer therapies. An optimal method for
164 S. Mathias et al.
Fig. 7.9 Ultrafast XUV and IR pulses are focused onto a Xe/ Pt(111) surface, where the kinetic
energy of the photoemitted electrons is measured by a time-of-flight detector. An electron from
the Xe N4 (4d) shell is ejected by the XUV, followed by filling of the core hole by an O2,3 (5p)
shell electron and ejection of a secondary (Auger) electron from the O2,3 shell. In the presence of
the IR beam, sidebands appear in the photoelectron spectrum of both Xe and Pt. By comparing the
sideband amplitude vs. time delay for the Pt and the Xe Auger photoelectron peaks, the lifetime of
the Xe 4d core hole can be extracted [63] (Reproduced and adapted with permission from Ref [63])
Fig. 7.10 (a) Photoemission spectra from Xe/Pt(111) as a function of the time delay between the
IR and the XUV pulse. The time evolution of the redistribution of electrons due to the IR dressing
at the Pt d-band and at the Xe Auger photoelectron peaks can be observed. (b) The sideband height
for the Pt d-band (grey curve) corresponds to a cross correlation between the XUV and the IR
pulses. The Auger sideband height (black curve) is shifted and broadened with respect to the Pt
d-band curve. A value of 7.1 ˙ 1.1 fs for the Xe 4d core-hole lifetime was extracted using these
data [63] (Reproduced and adapted with permission from Ref. [63])
as a function of time delay between the infrared laser pulse and the XUV probe
(solid black line). Upon cooling of the nanostructure as heat flows into the substrate,
the diffracted signal decays on picosecond-to-nanosecond timescales. Additionally,
166 S. Mathias et al.
Fig. 7.11 Normalized dynamic XUV diffraction signal for Ni lines on sapphire. The signal (solid
line) consists of a sharp rise owing to impulsive laser heating, a thermal decay owing to interfacial
thermal transport and an oscillation owing to surface acoustic wave propagation. The dashed and
dotted lines show the fit using the Fourier law for thermal transport in the substrate and either large-
scale interfacial resistance (dotted line) or a modified effective interfacial resistance (rEffective , units
of 109 K m2 W1 ), which accounts for ballistic effects from a nanoscale heat source (dashed
line) (Reproduced and adapted with permission from Ref. [71])
Fig. 7.12 Measured effective thermal resistivity for nickel nanostructures of width L deposited on
fused-silica and sapphire substrates. The dotted horizontal lines show the large-scale resistivity
rTBR for data from the fused-silica and sapphire substrates. The dashed curves show model
predictions for the ballistic resistivity correction rBC ƒ/L, with ƒFS D 2 nm and ƒSa D 120 nm,
respectively (Reproduced and adapted with permission from Ref. [71])
Figure 7.12 summarizes the systematic measurement and analysis of different sized
gratings on different substrates. Siemens et al. could experimentally show and
reproduce the expected dependence of rBC on /L.
The experiment by Siemens and co-workers is the first direct and systematic mea-
surement of quasi-ballistic thermal transport effects from a nanoscale heated region,
an effect of very high importance for material science research and applications,
e.g., in nanoscale thermal transport in circuits, for heat-assisted magnetic recording,
and for other structures in nanotechnology. The key to this type of measurement
was the use of coherent XUV light from HHG, which enabled for the first time heat
transport measurements with the necessary temporal and spatial resolution.
Coherent lensless imaging with x-rays is currently one of the most promising
techniques for ultrahigh resolution imaging with broad applications in biology,
168 S. Mathias et al.
Fig. 7.13 Setup of the tabletop soft x-ray diffraction microscope. A coherent XUV beam is
produced by high-harmonic generation or by spatially filtering the emission from a capillary
discharge laser in an Ar-filled hollow waveguide. The beam is gently focused by a multilayer
mirror pair onto the sample, and the diffraction pattern is collected on an x-ray CCD. A moveable
beam block allows for brighter portions of the diffraction pattern to be blocked to avoid saturating
the CCD when acquiring the highest angle diffracted light. ROC, radius of curvature (Reproduced
and adapted with permission from Ref. [78])
The presented series of ultrafast x-ray condensed matter physics experiments with
HHG demonstrates the potential of these table-top phase-matched high-flux light
sources in material science research. In future, the combination and extension of
170 S. Mathias et al.
Fig. 7.14 Lensless imaging by using coherent high-harmonic beams at 29 nm. (a) Reflection
image of the sample using an SEM. (Scale bar, 1 m.) (b) Coherent soft x-ray diffraction pattern (in
transmission) after curvature correction (maximum momentum transfer of 0.134 nm at the edge of
the CCD). (c) Reconstructed image. (d) Lineout of the image along the legs (Inset), demonstrating
a resolution of 94 nm (Reproduced and adapted with permission from Ref. [78])
References
1. T. Popmintchev, M.-C. Chen, P. Arpin, M.M. Murnane, H.C. Kapteyn, The attosecond
nonlinear optics of bright coherent X-ray generation. Nat. Photonics 4, 822–832 (2010)
2. T. Popmintchev, M.-C. Chen, A. Bahabad, M. Gerrity, P. Sidorenko, O. Cohen, I.P. Christov,
M.M. Murnane, H.C. Kapteyn, Phase matching of high harmonic generation in the soft and
hard X-ray regions of the spectrum. Proc. Natl. Acad. Sci. USA 106, 10516–10521 (2009)
3. M.-C. Chen, P. Arpin, T. Popmintchev, M. Gerrity, B. Zhang, M. Seaberg, D. Popmintchev,
M.M. Murnane, H.C. Kapteyn, Bright, coherent, ultrafast soft X-ray harmonics spanning the
water window from a tabletop light source. Phys. Rev. Lett. 105, 173901 (2010)
4. Z.H. Chang, A. Rundquist, H.W. Wang, M.M. Murnane, H.C. Kapteyn, Generation of coherent
soft X rays at 2.7 nm using high harmonics. Phys. Rev. Lett. 79, 2967–2970 (1997)
5. C. Spielmann, N.H. Burnett, S. Sartania, R. Koppitsch, M. Schnurer, C. Kan, M. Lenzner,
P. Wobrauschek, F. Krausz, Generation of coherent X-rays in the water window using 5-
femtosecond laser pulses. Science 278, 661–664 (1997)
6. E. Seres, J. Seres, F. Krausz, C. Spielmann, Generation of coherent soft-x-ray radiation
extending far beyond the titanium L edge. Phys. Rev. Lett. 92, 163002 (2004)
7. J.L. Krause, K.J. Schafer, K.C. Kulander, High-order harmonic-generation from atoms and
ions in the high-intensity regime. Phys. Rev. Lett. 68, 3535–3538 (1992)
172 S. Mathias et al.
8. K.C. Kulander, K.J. Schafer, J.L. Krause, Dynamics of short-pulse excitation, ionization and
harmonic conversion, in Super Intense Laser–Atom Physics, ed. by B. Piraux, A. L’Huillier,
K. Rza˛Ùzewski. vol 316 of NATO ASI Series (Plenum, New York, 1993), pp. 95–110
9. P.B. Corkum, Plasma perspective on strong-field multiphoton ionization. Phys. Rev. Lett. 71,
1994–1997 (1993)
10. I.P. Christov, J. Zhou, J. Peatross, A. Rundquist, M.M. Murnane, H.C. Kapteyn, Nonadiabatic
effects in high-harmonic generation with ultrashort pulses. Phys. Rev. Lett. 77, 1743–1746
(1996)
11. I.P. Christov, M.M. Murnane, H.C. Kapteyn, High-harmonic generation of attosecond pulses
in the “single-cycle” regime. Phys. Rev. Lett. 78, 1251–1254 (1997)
12. A. Baltuska, T. Udem, M. Uiberacker, M. Hentschel, E. Goulielmakis, H.R. Gohle Ch, V.S.
Yakovlev, A. Scrinzi, T.W. Hansch, F. Krausz, Attosecond control of electronic processes by
intense light fields. Nature 421, 611–615 (2003)
13. I. Thomann, A. Bahabad, X. Liu, R. Trebino, M.M. Murnane, H.C. Kapteyn, Characterizing
isolated attosecond pulses from hollow-core waveguides using multi-cycle driving pulses. Opt.
Express 17, 4611–4633 (2009)
14. S. Mathias, M. Bauer, M. Aeschlimann, L. Miaja-Avila, H.C. Kapteyn, M.M. Murnane, Time-
resolved photoelectron spectroscopy at surfaces using femtosecond XUV pulses, in Dynamics
at Solid State Surfaces and Interfaces. Vol. 1: Current Developments, ed. by U. Bovensiepen,
H. Petek, M. Wolf (Wiley-VCH, Weinheim, 2010). p. 501
15. J.P. Zhou, C.P. Huang, C.Y. Shi, M.M. Murnane, H.C. Kapteyn, Generation of 21-fs millijoule-
energy pulses by use of Ti-sapphire. Opt. Lett. 19, 126–128 (1994)
16. S. Backus, C.G. Durfee, M.M. Murnane, H.C. Kapteyn, High power ultrafast lasers. Rev. Sci.
Instrum. 69, 1207–1223 (1998)
17. S. Backus, R. Bartels, S. Thompson, R. Dollinger, H.C. Kapteyn, M.M. Murnane, High-
efficiency, single-stage 7-kHz high-average-power ultrafast laser system. Opt. Lett. 26,
465–467 (2001)
18. A. Rundquist, C.G. Durfee, Z.H. Chang, C. Herne, S. Backus, M.M. Murnane, H.C. Kapteyn,
Phase-matched generation of coherent soft X-rays. Science 280, 1412–1415 (1998)
19. C.G. Durfee, A.R. Rundquist, S. Backus, C. Herne, M.M. Murnane, H.C. Kapteyn, Phase
matching of high-order harmonics in hollow waveguides. Phys. Rev. Lett. 83, 2187–2190
(1999)
20. E. Constant, D. Garzella, P. Breger, E. Mevel, C. Dorrer, C. Le Blanc, F. Salin, P. Agostini,
Optimizing high harmonic generation in absorbing gases: model and experiment. Phys. Rev.
Lett. 82, 1668–1671 (1999)
21. R. Bartels, A. Paul, H. Green, H. Kapteyn, M. Murnane, S. Backus, I. Christov, Y. Liu,
D. Attwood, C. Jacobsen, Generation of spatially coherent light at extreme ultraviolet
wavelengths. Science 297, 376–378 (2002)
22. X. Zhang, A. Lytle, T. Popmintchev, A. Paul, N. Wagner, M. Murnane, H. Kapteyn, I. Christov,
Phase matching, quasi-phase matching, and pulse compression in a single waveguide for
enhanced high-harmonic generation. Opt. Lett. 30, 1971–1973 (2005)
23. E.A. Gibson, A. Paul, N. Wagner, R. Tobey, D. Gaudiosi, S. Backus, I.P. Christov, A. Aquila,
E.M. Gullikson, D.T. Attwood, M.M. Murnane, H.C. Kapteyn, Coherent soft x-ray generation
in the water window with quasi-phase matching. Science 302, 95–98 (2003)
24. J. Seres, V.S. Yakovlev, E. Seres, C. Streli, P. Wobrauschek, C. Spielmann, F. Krausz, Coherent
superposition of laser-driven soft-X-ray harmonics from successive sources. Nat. Phys. 3,
878–883 (2007)
25. A.L. Lytle, X.S. Zhang, R.L. Sandberg, O. Cohen, H.C. Kapteyn, M.M. Murnane, Quasi-phase
matching and characterization of high-order harmonic generation in hollow waveguides using
counterpropagating light. Opt. Express 16, 6544–6566 (2008)
26. X.H. Zhang, A.L. Lytle, T. Popmintchev, X.B. Zhou, H.C. Kapteyn, M.M. Murnane, O.
Cohen, Quasi-phase-matching and quantum-path control of high-harmonic generation using
counterpropagating light. Nat. Phys. 3, 270–275 (2007)
7 Ultrafast Material Science Probed Using Coherent X-ray Pulses. . . 173
27. S. Wall, D. Prabhakaran, A.T. Boothroyd, A. Cavalleri, Ultrafast coupling between light,
coherent lattice vibrations, and the magnetic structure of semicovalent LaMnO3 . Phys. Rev.
Lett. 103, 097402 (2009)
28. M. Rini, R. Tobey, N. Dean, J. Itatani, Y. Tomioka, Y. Tokura, R.W. Schoenlein, A. Cavalleri,
Control of the electronic phase of a manganite by mode-selective vibrational excitation. Nature
449, 72–74 (2007)
29. M. Rini, Y. Zhu, S. Wall, R.I. Tobey, H. Ehrke, T. Garl, J.W. Freeland, Y. Tomioka, Y. Tokura,
A. Cavalleri, R.W. Schoenlein, Transient electronic structure of the photoinduced phase of
Pr0.7 Ca0.3 MnO3 probed with soft x-ray pulses. Phys. Rev. B 80, 155113 (2009)
30. L. Perfetti, P.A. Loukakos, M. Lisowski, U. Bovensiepen, H. Berger, S. Biermann, P.S.
Cornaglia, A. Georges, M. Wolf, Time evolution of the electronic structure of 1T-TaS2 through
the insulator-metal transition. Phys. Rev. Lett. 97, 067402 (2006)
31. L. Perfetti, P.A. Loukakos, M. Lisowski, U. Bovensiepen, H. Eisaki, M. Wolf, Ultrafast
electron relaxation in superconducting Bi2 Sr2 CaCu2 O8C• by time-resolved photoelectron
spectroscopy. Phys. Rev. Lett. 99, 197001 (2007)
32. L. Perfetti, P.A. Loukakos, M. Lisowski, U. Bovensiepen, M. Wolf, H. Berger, S. Biermann,
A. Georges, Femtosecond dynamics of electronic states in the Mott insulator 1T-TaS2 by time
resolved photoelectron spectroscopy. New J. Phys. 10, 053019 (2008)
33. F. Schmitt, P.S. Kirchmann, U. Bovensiepen, R.G. Moore, L. Rettig, M. Krenz, J.-H. Chu, N.
Ru, L. Perfetti, D.H. Lu, M. Wolf, I.R. Fisher, Z.-X. Shen, Transient electronic structure and
melting of a charge density wave in TbTe3 . Science 321, 1649–1652 (2008)
34. R.I. Tobey, D. Prabhakaran, A.T. Boothroyd, A. Cavalleri, Ultrafast electronic phase transition
in La1/2 Sr3/2 MnO4 by coherent vibrational excitation: evidence for nonthermal melting of
orbital order. Phys. Rev. Lett. 101, 197404 (2008)
35. K. Miyano, T. Tanaka, Y. Tomioka, Y. Tokura, Photoinduced insulator-to-metal transition in a
perovskite manganite. Phys. Rev. Lett. 78, 4257–4260 (1997)
36. M. Matsubara, Y. Okimoto, T. Ogasawara, Y. Tomioka, H. Okamoto, Y. Tokura, Ultrafast
photoinduced insulator-ferromagnet transition in the perovskite manganite Gd0.55 Sr0.45 MnO3 .
Phys. Rev. Lett. 99, 207401 (2007)
37. J. Gudde, M. Rohleder, T. Meier, S.W. Koch, U. Hofer, Time-resolved investigation of
coherently controlled electric currents at a metal surface. Science 318, 1287–1291 (2007)
38. S. Mathias, L. Miaja-Avila, M.M. Murnane, H. Kapteyn, M. Aeschlimann, M. Bauer, Angle-
resolved photoemission spectroscopy with a femtosecond high harmonic light source using a
two-dimensional imaging electron analyzer. Rev. Sci. Instrum. 78, 083105 (2007)
39. S. Mathias, A. Ruffing, F. Deicke, M. Wiesenmayer, M. Aeschlimann, M. Bauer, Band structure
dependence of hot-electron lifetimes in a Pb/Cu(111) quantum-well system. Phys. Rev. B 81,
155429 (2010)
40. T. Rohwer, S. Hellmann, M. Wiesenmayer, C. Sohrt, A. Stange, B. Slomski, A. Carr, Y. Liu, L.
Miaja-Avila, M. Kallaene, S. Mathias, L. Kipp, K. Rossnagel, M. Bauer, Collapse of long-range
charge order tracked by time-resolved photoemission at high momenta. Nature 471, 490–493
(2011)
41. G.L. Dakovski, Y. Li, T. Durakiewicz, G. Rodriguez, Tunable ultrafast extreme ultraviolet
source for time- and angle-resolved photoemission spectroscopy. Rev. Sci. Instrum. 81, 073108
(2010)
42. S. Mathias, M. Wiesenmayer, F. Deicke, A. Ruffing, L. Miaja-Avila, M.M. Murnane, H.C.
Kapteyn, M. Bauer, M. Aeschlimann, Time and angle resolved photoemission spectroscopy
using femtosecond visible and high-harmonic light. J. Phys. Conf. Ser. 148, 012042 (2009)
43. E. Beaurepaire, J. Merle, A. Daunois, J. Bigot, Ultrafast spin dynamics in ferromagnetic nickel.
Phys. Rev. Lett. 22, 4250 (1996)
44. R.W. Schoenlein, Generation of femtosecond pulses of synchrotron radiation. Science 287,
2237–2240 (2000)
45. C. Boeglin, E. Beaurepaire, V. Halte, V. Lopez-Flores, C. Stamm, N. Pontius, H.A. Duerr, J.-Y.
Bigot, Distinguishing the ultrafast dynamics of spin and orbital moments in solids. Nature 465,
458–461 (2010)
174 S. Mathias et al.
8.1 Introduction
Nonlinear fibre optics is a very mature field of research, and nonlinear propagation
effects have been studied under a wide range of conditions. Very good general
textbook treatments of the whole field as well as specialised monographs treating
particular aspects of the field exist and it is not the purpose of this chapter to fully
reproduce material that is covered more fully in other places. Rather the aim here is
to present a focussed treatment of the subject, concisely covering the most essential
aspects of the field in a self-contained way. A particular feature of the chapter will
be a brief review of the physics of supercontinuum (SC) generation in optical fibre
under different conditions. The term SC generation is generally used to describe
the case where (relatively) narrow-band incident pulses undergo extreme nonlinear
spectral broadening to yield a broadband spectrally continuous output.
SC generation in bulk media and fibres has been studied since the earliest work
in nonlinear optics and nonlinear fibre optics. The broadband spectral output of
SC generation has found numerous applications in many diverse fields including
spectroscopy and the design of tuneable ultrafast femtosecond laser sources. The
dramatic observation of SC generation in photonic crystal fibre in 1999–2000 led to
a revolution in the field of optical frequency metrology, a story that has been well-
told elsewhere. In a telecommunications context, the spectral slicing of broadband
SC spectra has also been proposed as a simple way to create multi-wavelength
optical sources for dense wavelength division multiplexing applications.
There is clearly wide interest in understanding the complex process of SC
generation, but it seems at first sight very difficult because of the complex way in
which the different linear and nonlinear propagation effects can combine. But in fact
the essential features of the spectral broadening can be broken down into a number
of different contributing mechanisms and it is possible to present a reasonably
clear physical interpretation of the major spectral and temporal features observed.
A key element, however, relates to understanding the fundamental principles of
nonlinear fibre optics, and thus we must begin our description of SC generation with
a summary of the nonlinear and dispersive interactions that can occur when a light
pulse is injected in a dielectric waveguide. Our approach follows that in Ref. [1].
Note that this chapter represents a synthetic summary of a number of our previous
publications in this area, prepared with the specific aim of covering the content of
the lectures given by one of us (JMD) at the 66th Scottish Universities Summer
School in Physics, which was held at Heriot-Watt University in August 2010.
The aim here is to summarise the theory that underlies the description of a
guided wave propagating in an optical fibre. We suppose firstly, that the transverse
component of the guided wave field can be decoupled from the time-varying
component along the propagation direction. Then if the electric field is written as:
(where x is a unit polarisation vector), the complex envelope E(r, t) can be expressed
as:
E.r; t/ D F.x; y/ A.z; t/ exp.i“0 z/; (8.2)
where A(z, t) is the envelope describing the evolution of the field, F(x, y) represents
the transverse field profile, and “0 D “(¨0 ) is the wavevector. Note that we are of
course neglecting (vectorial) polarisation effects here. On the other hand, recent
work has begun to develop a comprehensive treatment of this important problem
but this is outside the scope of this chapter. In any case, polarisation effects can be
neglected over quite a large range of experimental parameters. Also, the bandwidths
of ultrashort light pulses are generally sufficiently large that chromatic dispersion in
the material must be described with minimal approximation. The effect of dispersion
can be conveniently examined using a Taylor series expansion:
n D n0 C n2 I .t/; (8.4)
where n0 is the usual linear index, the intensity I(t) is expressed in W.m2 and n2 is
the material nonlinear refractive index.
There is some debate over what value of n2 should be used for optical fibres,
but for fused silica good agreement with many experiments is obtained using
n2 D 3.2 1020 m2 .W1 [1]. There is a need for further clarification here, however,
and it is to be hoped that as experimental techniques improve, some of the ambiguity
associated with the correct value of nonlinear parameter to use in fibres can be lifted.
The combined effects of dispersion and nonlinearity on a pulse with bandwidth
¨ << ¨0 are described in the absence of loss by the nonlinear Schrödinger
equation (NLSE):
@A ˇ2 @2 A
i C jAj2 A D 0: (8.5)
@z 2 @T 2
which has the solution in the frequency domain associated with the development of
a quadratic spectral phase across the field spectrum.
Q Q i
A.z; / D A.0; / exp 2
ˇ2 .2v/ z ; (8.7)
2
Here is the frequency, the reciprocal Fourier variable of T. Note that the solution
(8.7) in fact describes a temporal equivalent of linear diffraction in space, and there
are many important equivalents between temporal propagation in fibres and spatial
diffractive effects (e.g. the imposition of a quadratic spectral phase by a modulator
is often referred to as a “time lens”.)
When the distance of propagation is such that z LNL and z << LD , nonlinear
effects dominate and Eq. (8.5) becomes:
@A
i C AjAj2 D 0; (8.8)
@z
with the solution now in the time domain associated with a nonlinear phase
modulation:
A.z; T / D A.0; T / exp i jAj2 z : (8.9)
It is important to fully understand the implications of the solutions (8.7) and (8.9).
A pulse propagating in the presence of only dispersion is associated with a spectral
intensity that remains invariant during propagation, whereas a pulse propagating in
the presence of only nonlinearity is associated with a temporal intensity profile that
remains invariant during propagation. But in general, both effects occur together,
and numerical techniques must be used to solve (8.5). A very straightforward
technique to achieve this is the split-step Fourier technique, where propagation
is considered in terms of multiple small increments that allow the separation of
dispersive and nonlinear effects. In this case, the analytic solutions we derived above
in (8.7) and (8.9) can be applied iteratively to propagate forward from some initial
condition to follow the evolution of an initial field.
8 Ultrafast Nonlinear Fibre Optics and Supercontinuum Generation 181
Fig. 8.1 Numerical solution of the NLSE showing top: (a) fundamental soliton evolution and
bottom: (b) N D 2 soliton evolution.
Figure 8.1 presents the results of this technique used to study soliton solutions
of the NLSE. Stable fundamental soliton solutions exist when “2 < 0, and under
conditions such that the linear and dispersive effects are exactly balanced. This
is often termed the p
focussing regime of the NLSE. The form of the fundamental
soliton is: A(0,T) D P0 sech(T/T0 ) with the pulse parameters matched to the fibre
parameters through:
LD jˇ2 j
D N 2 D 1 thus P0 D (8.10)
LNL T02
@A X k ˇk @k A
i C i
@z kŠ @T k
k2
0 1
Z1
i @ @
C 1 C A.z; T / R.T 0 /jA.z; T T 0 /j2 d T 0 A D 0 (8.11)
!0 @T
1
higher-order dispersion coefficients, this results in the GNLSE being valid over
very broad bandwidths and even down to pulses of one or two cycles [7]. But
here it is important to realise that just because the GNLSE describes the evolution
of an envelope, it does not mean that the envelope has to be “slowly-varying”. In
fact when one revisits the derivation of the GNLSE from Maxwell’s equations, one
realises that the transformation from the usual second-order differential propagation
equations of electromagnetism to a first order partial differential equation such as the
GNLSE is actually associated with a factorisation step that is equivalent to assuming
that the propagation is purely unidirectional. This is an important point that deserves
careful reflection [7].
Also of importance to fully understand is the form of the nonlinear operator:
12 C 22
hr .T / D exp.T =2 / sin.T =1 / (8.13)
1 22
with £1 D 12.2 fs, £2 D 32 fs [1]. Note that there are some subtle aspects of
normalisation in the way in which the Raman effect is included in the GNLSE; for
example the function hr (T) is normalised to have unity integral, yet the dimensional
Raman gain at the pump wavelength is included through the terms ” and fr , and
its wavelength dependence is included through the time-derivative term. (Recall in
this context that the Fourier transform of a time-derivative term is equivalent to
multiplication by a frequency operator, and thus we can see how a ¨/¨0 scaling can
multiply the strength of the nonlinear response term in the frequency domain form
of the GNLSE.)
In this section, we consider the general features of SC generation with the aim of
discussing the general physical mechanisms involved and illustrating the different
features that can be observed in both the time and frequency domains. The cases
considered are SC generated by femtosecond pump pulses, SC generated by pump
pulses longer than several picoseconds, and SC generated in a particular class of
fibre with a convex dispersion profile where the flat spectral characteristics are
especially useful for telecommunications applications.
184 J.M. Dudley et al.
and highlights in particular the point of soliton fission, the generation of dispersive
wave (DW) radiation, and the continual spectral and temporal shifting of an ejected
soliton undergoing the Raman self-frequency shift (RS).
The initial phase of nonlinear evolution contributes very significantly to the
overall spectral bandwidth. It is often the case that it is the spectral extension
during this phase that results in the injected higher-order soliton “experiencing” the
perturbation of dispersion across its bandwidth. This perturbation ultimately leads
to its breakup. Cutting the fibre before the breakup point has been applied usefully
for applications in pulse compression, and we can apply insights obtained in this
previous work to aid in the practical study of SC generation. In particular, motivated
by studies aiming to optimise compression [9], we introduce a characteristic length
scale to describe the fission point, based on the distance at which the spectral width
of the initial soliton reaches it maximum value. This has been termed the fission
length and can be written approximately as Lfiss LD /N D (LD LNL )1/2 . Simulations,
such as those shown in Fig. 8.2, can be readily extended over a wider parameter
range, allowing the study of femtosecond propagation where the pump pulse is
in the normal GVD regime. In this case, results show that the initial spectral
broadening arises primarily from self-phase modulation (SPM) and can lead to a
spectrum that is significantly less structured than that associated with anomalous
GVD pumping. However, power levels generally required to achieve the same
bandwidth are lower when soliton dynamics are exploited. When pumping in the
normal GVD regime, but approaching closer to the zero dispersion wavelength, the
initial spectral broadening due to SPM transfers spectral content into the vicinity
of the zero dispersion wavelength and across into the anomalous dispersion regime.
186 J.M. Dudley et al.
Once this occurs, soliton dynamics as described above can also begin to contribute
significantly to the spectral and temporal features of the generated SC.
At this point when discussing SC characteristics in different wavelength regimes,
we should also make an obvious but necessary remark. Although we often plot
spectral characteristics using axes scales in nanometres, meaningful physical com-
parisons of spectral widths generated across different wavelength ranges should be
made in frequency units of GHz or THz because “the number of nanometres/THz”
varies significantly with wavelength. For small bandwidths, some useful conversion
factors to remember (using j œ/ j D œ2 /c) are 1 THz/nm at 550 nm; 0.5 THz/nm
at 780 nm; 0.3 THz/nm at 1,000 nm; 125 GHz/nm at 1,550 nm. A 100 nm SC around
1,550 nm using a fibre laser for example is associated with a photon energy spread
equivalent to a 25 nm SC from a Ti:sapphire laser at 780 nm.
Shortly after the initial experiments by Ranka et al. in 2000 [4] reporting SC
generation using femtosecond pulses, several groups reported broadband SC in
PCF using nanosecond and picosecond sources (See Ref. [6] for a full historical
review). Subsequent research has also extended this “long pulse” regime to include
continuous wave sources. For the long pulse regime, soliton effects do not pay
an important role in the initial dynamics, but it is rather the spontaneous growth
of new frequency components outside the pump bandwidth that dominates the
initial spectral broadening. The mechanisms involved in this regime are based on
spontaneous four wave mixing (FWM) or, as it is referred to in the time domain,
modulation instability (MI). To emphasise the fact that these terms are different
ways of discussing the same physics, we refer to this as FWM/MI in what follows.
The relative importance of soliton effects and spontaneous FWM/MI is in fact
quite easy to quantify. If we consider anomalous dispersion regime pumping, as
the soliton order becomes very large ( N >> 10), higher-order soliton evolution (as
described above) become progressively less important because the characteristic
length scale over which soliton fission occurs increases with the pump pulse
duration, so the effect of spontaneous MI/FWM sideband growth dominates. In
such cases, initial MI/FWM dynamics leads to the temporal break-up of the pump
pulse into a large number of sub-pulses. Interestingly, recent work has provided new
insight into this pulse breakup by showing that they can be interpreted in terms of
a particular class of analytic solution to the NLSE known as Akhmediev breathers
[10]. The subsequent evolution of these sub-pulses then leads to additional spectral
broadening and SC formation through a variety of mechanisms, notably the same
soliton dynamical effects such as dispersive wave generation and the Raman soliton
self-frequency shift as seen above. However, the transition regime between breather
structures and solitons remains somewhat lacking of a clear physical picture, but
some insights are being obtained and links with instability processes of turbulence
and convection in other areas of physics are providing new insights. Still, a simple
8 Ultrafast Nonlinear Fibre Optics and Supercontinuum Generation 187
Fig. 8.3 Left: Single-shot simulation results showing spectral evolution of a 500 W peak power
20 ps pulse at 800 nm over 2 m. Right: Similar single-shot results for pulse durations as shown
out/averages much of the fine spectral and temporal structure seen in the figure
[11, 12]. Understanding the mechanisms of these instabilities and developing new
techniques for their measurement is forming the basis of the emerging field of
“optical rogue wave” research – the description here refers to a possible analogy
between fibre instabilities and oceanic instabilities that lead to the formation of giant
high amplitude waves of great destructive power [13]. We consider this below.
Much of the studies of nonlinear propagation in optical fibres in the 1980s were
performed in the context of optical telecommunications, focussing extensively on
applications such as soliton transmission, Raman amplification and wavelength
conversion [1]. Possible applications of SC generation to telecommunications have
also been considered, motivated by the need to develop broadband sources suitable
for spectral slicing in wavelength-division multiplexing. The idea here is that it
is generally easier to increase data rates by using more channels of different
wavelength carrying data at a low bit rate rather than one single channel at a greatly
enhanced bit rate. The spectral slicing approach is based on using one primary
source to generate a broad spectrum that is subsequently filtered into the desired
number of separate wavelength channels. This is an attractive alternative to using an
independent source for each channel.
For such multi-wavelength applications, a crucial practical requirement is spec-
tral uniformity across as wide a wavelength range as possible, centred at around the
telecommunications wavelength range of 1,550 nm. Generating SC spectra across
this range has been pioneered by Japanese groups [14–16]. The optical fibres used
for SC generation at around 1,550 nm generally possess dispersion characteristics
that are very different from the fibres used for SC generation with Ti:sapphire laser
sources. As a result the propagation dynamics and the output spectral characteristics
are also qualitatively very different from the results seen in Figs. 8.2 and 8.3. The
mechanisms of SC generation in this regime have been considered in a number of
publications, notably [16] for standard fibre and, more recently [17] for PCF.
A case of particular interest is when the fibre possesses a convex dispersion
profile such that two zero dispersion wavelengths bracket the pump wavelength.
Although this is a class of fibre design that has seen some recent interest for photonic
crystal fibre, it has existed within the highly nonlinear fibre technology for some
time. One consequence of this class of fibre is that propagating pulses can see a
slope of dispersion that changes sign on either side of the point where the dispersion
reaches its extremum. Propagation in this regime leads to interesting new effects in
the SC spectrum.
When a convex profile is combined with a dispersion at the pump wavelength
that varies longitudinally from anomalous at the fibre input to normal at the
fibre output, spectra are obtained that are simultaneously uniform, symmetric, and
8 Ultrafast Nonlinear Fibre Optics and Supercontinuum Generation 189
Fig. 8.4 A “telecommunications supercontinuum” showing (a) longitudinal evolution and (b) the
corresponding output spectrum after 1 km of propagation. Note the significant qualitative
difference in the spectral shape when compared to supercontinuum generation with soliton or
MI/FWM dynamics
stable. Of course this requires appropriate matching of fibre and source parameters.
Figure 8.4 above shows results from simulations that illustrate the features observed.
Here we show (a) the longitudinal evolution in greyscale as well as (b) the output
spectrum after 1 km of propagation in a convex profile dispersion decreasing fibre
(see Ref. [17]). In this case, the uniform spectrum at the output develops from
a combination of initial compression followed by dispersive wave generation as
the expanding pulse spectrum penetrates into regimes of normal dispersion on
both the short wavelength and long wavelength side of the central wavelength.
This symmetric dispersive wave generation appears to be the primary mechanism
underlying the smooth spectra observed. Note that the spectral asymmetry seen in
Fig. 8.4b arises from Raman gain on the long wavelength edge.
The “stability” of a SC refers to how the properties of the output spectrum and
temporal intensity vary with time. Stability issues are most readily understood when
considering SC generation using pulsed sources [6]. In this case, we consider the
variation in the SC characteristics from shot-to-shot, as each pulse in a mode-locked
laser pulse train excites an independent spectrum. Because there will always be
some level of technical and/or quantum noise, each input pulse will be slightly
different; thus in regimes where the spectral broadening is highly sensitive to noise,
this can lead to significant shot-to-shot variation.
190 J.M. Dudley et al.
8.6 Conclusions
The objective of this short chapter was not to present a complete overview of
every possible scenario of SC generation in optical fibres, but rather to discuss
a selection of specific results that are of particular interest in understanding the
underlying physics. SC generation is a fascinating optical phenomenon that provides
8 Ultrafast Nonlinear Fibre Optics and Supercontinuum Generation 191
a unique testbed to examine numerous areas of nonlinear fibre optics. If there is one
absolutely critical thing to stress in closing, however, it is that one should not be
overly simplistic in trying to explain the dynamics of SC broadening.
Appendix
This MATLAB code using the split-step Fourier technique is a very simple
implementation to show the correspondence between the description in the text and
Fig. 8.1 concerning the basic soliton solutions of the NLSE.
% - - - - - - - - - - - - - - - - - - - - WARNING- - - - - - - - - - - - - - - - - - - - - - - - -
% This code is written in simple dimensional form with no
% effort to be particularly robust numerically.
% Higher order effects are neglected with this implementation
clear,close all
lambda0 D 850e-9; % input wavelength [m]
c D 299792458; % speed of light [m/s]
F0 D c/lambda0;
% Time and Frequency Arrays
Npts D 256;
Tmax D 0.5e-12;
dT D 2*Tmax/(Npts-1);
TT D [Npts/2:(Npts/2)-1]*dT; % [ps]
FF D [Npts/2:(Npts/2)-1] ./(2*Tmax); % [THz]
WW D 2*pi*FF;
% Fibre parameters
beta2 D 1.275e-026;
gamma D 0.10; %nonlinear length
% Soliton order and other parameters
N D 1;
T0 D 30e-15; %nonlinear length
P0 D Nˆ2*abs(beta2)/(T0ˆ2*gamma); %nonlinear length
% Length scales
L D D T0ˆ2/abs(beta2); %dispersion length
L NL D 1/(gamma*P0); %nonlinear length
L sol D pi/2*L D;
% Input pulse
A0 D sqrt(P0)*sech(TT/T0);
% - - - - - - - - - - - - - - - - - - - - - Propagation - - - - - - - - - - - - - - - - - - - - - - -
192 J.M. Dudley et al.
zl D zlabel('Intensity (W)','fontsize',16)
set(gcf,'colormap',[0 0 0])
set(gca,'Xgrid','off','Ygrid','off','Zgrid','off','fontsize',16, : : :
'Linewidth',1)
view(135,50)
axis([0.5 0.5 0 2 0 max(max(Ip))])
References
1. G.P. Agrawal, Nonlinear Fiber Optics, 4th edn. (Academic, San Diego, 2007)
2. P.S.J. Russell, Photonic-crystal fibers. J. Lightwave Technol. 24, 4729–4749 (2006)
3. J.M. Dudley, J.R. Taylor, Ten years of nonlinear optics in photonic crystal fibre. Nat. Photonics
3, 85–90 (2009)
4. J.K. Ranka, R.S. Windeler, A.J. Stentz, Visible continuum generation in air-silica microstruc-
ture optical fibers with anomalous dispersion at 800 nm. Opt. Lett. 25, 25–27 (2000)
5. J.M. Dudley, L. Provino, N. Grossard, H. Maillotte, R.S. Windeler, B.J. Eggleton, S. Coen,
Supercontinuum generation in air–silica microstructured fibers with nanosecond and femtosec-
ond pulse pumping. J. Opt. Soc. Am. B 19, 765–771 (2002)
6. J.M. Dudley, G. Genty, S. Coen, Supercontinuum generation in photonic crystal fiber. Rev.
Mod. Phys. 78, 1135–1184 (2006)
7. G. Genty, P. Kinsler, B. Kibler, J.M. Dudley, Nonlinear envelope equation modeling of sub-
cycle dynamics and harmonic generation in nonlinear waveguides. Opt. Express 15, 5382–5387
(2007)
8. E.M. Dianov, A.Y. Karasik, P.V. Mamyshev, A.M. Prokhorov, V.N. Serkin, M.F. Stel’makh,
A.A. Fomichev, Stimulated-Raman conversion of multisoliton pulses in quartz optical fibers.
JETP Lett. 41, 294–297 (1985)
9. C.M. Chen, P.L. Kelley, Nonlinear pulse compression in optical fibers: scaling laws and
numerical analysis. J. Opt. Soc. Am. B 19, 1961–1967 (2002)
10. J.M. Dudley, G. Genty, F. Dias, B. Kibler, N. Akhmediev, Modulation instability, Akhmediev
Breathers and continuous wave supercontinuum generation. Opt. Express 17, 21497–21508
(2009)
11. J.M. Dudley, S. Coen, Coherence properties of supercontinuum spectra generated in photonic
crystal and tapered optical fibers. Opt. Lett. 27, 1180–1182 (2002)
12. K.L. Corwin, N.R. Newbury, J.M. Dudley, S. Coen, S.A. Diddams, K. Weber, R.S. Windeler,
Fundamental noise limitations to supercontinuum generation in microstructure fiber. Phys. Rev.
Lett. 90, 113904 (2003)
13. D.R. Solli, C. Ropers, P. Koonath, B. Jalali, Optical rogue waves. Nature 450, 1054–1057
(2007)
14. T. Morioka, K. Mori, M. Saruwatari, More than 100-wavelength-channel picosecond optical
pulse generation from single laser source using supercontinuum in optical fibres. Electron.
Lett. 29, 862–864 (1993)
15. K.R. Tamura, H. Kuhota, M. Nakazawa, Fundamentals of stable continuum generation at high
repetition rates. IEEE J. Quantum Electron 36, 773–779 (2000)
16. K. Mori, H. Takara, S. Kawanishi, Analysis and design of supercontinuum pulse generation in
a single-mode optical fiber. J. Opt. Soc. Am. B 18, 1780–1792 (2001)
17. G. Genty, S. Coen, J.M. Dudley, Fiber supercontinuum sources (Invited). J. Opt. Soc. Am.
B 24, 1771–1785 (2007)
Chapter 9
Nonlinear Wavelength Conversion and Pulse
Propagation in Optical Fibres
William J. Wadsworth
9.1 Introduction
The subject of nonlinear optics in optical fibres has a long history, and has developed
in parallel with, and as a spur to, developments in fibre design and laser sources.
Chapter 8 in this book covers in some detail the physics of nonlinear pulse
propagation in optical fibres in terms of the nonlinear Schrödinger equation (Eq. 8.5)
and its application to understanding the mechanisms of supercontinuum generation.
This chapter approaches many of the same issues, but from the perspective of
the engineering and capabilities of real fibres. One external reference that it is
impossible to ignore in this field is Agrawal’s book ‘Nonlinear Fiber Optics’ [1]
which, in any of its four editions to date, will guide the reader through the theory
of pulse propagation and nonlinear wavelength conversion in optical fibres with a
wide range of practical examples from the academic literature. Many phenomena
can only be touched on here, and Ref. [1] will help to lead the reader to a deeper
understanding of the nonlinear optics.
9.2 Nonlinearity
In optics, the term nonlinearity refers to the relationship between the electric
displacement, D, and the electric field, E, which may be expressed either as
or
D D "0 E C P : (9.2)
assuming for the moment that vectors P and E are in the same direction and so
may be treated as scalars. One must, however, bear in mind that in principle the
coefficients (n) are tensors with elements linking each possible component of E
with each component of P. If the electric field oscillates as ei!t , then terms in
E2 contain components at twice the frequency, e2i!t . If the electric field contains
components at frequencies ! 1 and ! 2 , then terms in E2 contain components at
! 1 C ! 2 and ! 1 ! 2 , with similar effects for the terms in E3 . An oscillating induced
polarisation will re-radiate electromagnetic waves, which will now contain new
frequencies or wavelengths of light. This is the basis for nonlinear wavelength
conversion. In this chapter we shall only consider optical fibres, which are made
from glasses. As glasses are centrosymmetric materials, symmetry forbids a non-
zero (2) so the lowest order nonlinearity is (3) . Furthermore, of the 81 elements of
the (3) tensor only five are non-zero
.3/
xxxx D .3/
yyyy D .3/
xxyy C .3/
xyyx C .3/
xyxy ; (9.4)
for electric fields polarised in the (x,y) plane. The cross terms mean that an induced
polarisation, P, and hence an output field, can be generated perpendicular to the
polarisation direction of the input field. In order to achieve a macroscopic conversion
of energy into the new frequencies one obviously needs large nonlinear coefficients
(n)
, but one also needs to ensure that the fields at new frequencies radiated
from different points in the medium add up in phase at the output. This requires
phasematching, which will be discussed later in Sect. 9.6.1.
Another useful way of expressing the optical nonlinearity is in terms of an
intensity dependent refractive index, as in Eq. (8.4). Putting Eq. (9.3) into (9.2)
and re-formulating the result into the form of (9.1) gives (again assuming a scalar
relationship between the fields)
D D "0 E C "0 .1/ E C .2/ E 2 C .3/ E 3 C : : :
D "0 1 C .1/ C .2/ E C .3/ E 2 C : : : E (9.5)
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 197
(2)
Using the fact that is zero in glasses and that the refractive index is defined
by n2 D "r we obtain:
q
nD 1C .1/ C .3/ E 2 C :::
q
1
1C .1/ C p .3/
E2
2 1C .1/
1
D n0 C .3/
E2
2n0 (9.6)
To convert to the intensity units used in Eq. (8.4) one can relate the optical intensity,
I, to the electric field, E, by [2, 3]:
c"0 n 2
I D E : (9.8)
2
Several nonlinear processes may be explained from the induced polarisation
picture or the nonlinear refractive index picture. The (3) component of the induced
polarisation gives an output wave at a frequency
!4 D !1 C !2 ˙ !3 ; (9.9)
The processes of FWM, MI, XPM and SPM all arise from the electronic
nonlinearity. The inelastic processes of Raman scattering or Brillouin scattering are
not considered in detail here.
a b
nclad r
ncore
dn
n nclad
ncore
r
nclad 0
Fig. 9.1 (a) simple (though incorrect) ray picture of guidance in an optical fibre, (b) terms defining
a step-index fibre geometry
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 199
and Maxwell’s equations may be manipulated into the vector wave equation
rt2 E C k 2 n2 ˇ 2 E D rt C iˇOz E:rt ln n2 : (9.11)
Here, E and n are functions of x and y only (the transverse mode field profile and
transverse index profile of the fibre or waveguide), ˇ is the propagation constant, k
the free-space wavevector (! /c), and rt is the transverse gradient operator
@ @
rt D xO C yO C 0Oz: (9.12)
@x @y
The x-component of the left hand side of Eq. (9.11) contains only the
x-component of the field, however the x-component of the right-hand side of Eq.
(9.11) contains terms which contain both the x and y components of the electric field
E. The equation is generally intractable and solutions require numerical methods.
However one can often make approximations for which this reduces back to a
tractable scalar equation. If n(x,y) does not vary very much (weak guiding), then
the term on the RHS of Eq. (9.11) may be neglected:
rt2 C k 2 n2 ˇ 2 D 0; (9.13)
with the electric field vector, E(x,y), related to the scalar field, (x,y), by a spatially
invariant direction vector, eO W E
.x; y/ D eO .x; y/: Note that the polarisation
described by eO is not necessarily linear, however the polarisation is constant
across the (x,y) plane. It can be useful to note that this scalar wave equation is
mathematically identical to the Schrödinger equation for the wavefunction, ‰, of a
particle, mass m, in a 2D potential, V,
2m 2m
rt2 ‰.x; y/ V .x; y/ E ‰.x; y/ D 0: (9.14)
„2 „2
Note that here E is the energy of the state, not the electric field. We can identify
the potential, V, (note the sign change) with the fibre index profile, n2 , and the
energy eigenvalue, E, with the fibre mode propagation constant, ˇ. This analogy is
useful because many students study solutions of the Schrödinger equation, so have
some intuition for the form of these solutions which can then be transferred to give
a basic understanding of the modes of optical fibres.
200 W.J. Wadsworth
a ∞ ∞
p
b c ncore
4 dn nmode
3
dV dV
V 2 V
n
1 nclad
x x r
0 a 0 a -
Fig. 9.2 Schematic representations of: (a) solutions to the Schrödinger equation in a 1-D infinite
potential well, (b) the same for a finite potential well, (c) solutions to the scalar wave for a step-
index optical fibre
The first problem which we often solve for the Schrödinger equation is a particle
in an infinite square potential well. There are sin(p x/a) solutions as shown in
Fig. 9.2a with defined spatial profiles and energies as in Table 9.1. These eigen
solutions are wavefunctions which do not change in time (except for a phase),
analogous to the modes of a fibre which do not change with propagation distance
along the fibre (except for a phase). The solution for a finite potential well is similar
to the solution for an infinite potential well, but with the addition of evanescent tails
of the wavefunction extending into the classically forbidden region outside the well
(Fig. 9.2b). There are then a finite number of bound states with energies below the
potential of the boundaries of the well.
A step-index optical fibre bears a close resemblance to a finite potential well as
shown in Fig. 9.2c. The potential well becomes a positive index step (because of
the sign change between Eqs. 9.13 and 9.14), the energy eigenvalue becomes the
modal propagation constant, the tails of the particle wavefunction in the classically
forbidden regions outside the well become the evanescent tails of the optical field in
the low-index cladding medium. Also, in Eqs. 9.10, 9.11 and 9.13 the propagation
constant ˇ appears where one would normally expect the wavevector in a medium,
kn D n ! /c. We can therefore define a modal index
ˇ
nmode D ; (9.15)
k
where k is the free-space wavevector.
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 201
By analogy with the bound energy eigenvalues in a finite potential well, there will
be a finite number of guided modes in a step index fibre, with modal indices lying
between the core and cladding indices for the fibre, as shown in Fig. 9.2c. There
will also be a continuum of unbound states, the cladding modes, with modal indices
below the cladding index. From further inspection of Eqs. 9.13 and 9.14, with the
definition of modal index in Eq. (9.15), it is clear that the free-space wavevector, k,
is analogous to the particle mass, m. It is not usual to consider particles of different
mass in a box, but it is very common in optics to consider light of widely differing
wavelengths, and hence of differing wavevectors. The overall effect on the modal
index of changing wavelength may be easily seen from the basic ideas of optical
diffraction. Light of a long wavelength will be more difficult to confine in the small
spatial dimension of an optical fibre core. The evanescent tails in the cladding will
be stronger, and the modal effective index will become closer to the index of the
cladding. Conversely a shorter wavelength will have smaller evanescent tails and a
modal index closer to the index of the core. This will be considered further in the
discussion of dispersion in Sect. 9.4.2.
Looking at the order of the modal indices of the fundamental mode and higher
order modes shown schematically in Fig. 9.2c, there appears to be a contradiction
with the ray picture of Fig. 9.1a. Intuitively, one would expect that the fundamental
mode would be the one with the straightest path in the ray picture, which would
also have the shortest path and travel faster down the fibre. However in the modal
indices given in Fig. 9.2c the fundamental mode has the highest index, which implies
a slower velocity. The resolution of this apparent contradiction comes from the
difference between phase and group velocity. The modal index is a phase index,
implying that the fundamental mode has a slow phase velocity. However pulses or
information travel at the group velocity, d! / dˇ, and the fundamental mode does
indeed have the expected low group index and high group velocity.
The number of guided modes is an important consideration in fibre optics, and
it is often convenient or required for a fibre to support only one guided mode – a
single-mode fibre (SMF). We define a fibre parameter
q
2
V D n2core n2clad ; (9.16)
which normalises all the important parameters of the fibre given in Fig. 9.2c into
a single dimensionless number which describes the size of the ‘box’ available for
the modes. It is bigger for a larger index step or a larger core diameter, and bigger
for a shorter wavelength: in other words, when the box is larger compared to the
wavelength. For V < 2.405 there is only one guided mode, for 2.405 < V < 3.832
the second mode is guided as well. For highly multimode fibres with V >> 1 the
number of spatial modes, N, varies as N V2 /4, with each spatial mode having
twofold polarisation degeneracy.
202 W.J. Wadsworth
All of the nonlinear processes discussed in Sect. 9.2 require control of chromatic
dispersion – the relative refractive indices for different wavelengths of light. For
SPM and XPM we have seen in Chap. 8 that the sign of dispersion, normal or
anomalous, has a fundamental impact in the formation of optical solitons (Eq. 8.10).
For FWM/MI phasematching is required, which also is controlled by the dispersion
of the medium. An optical fibre will always have a small core and long length,
so that the intensity length product is extremely large, and fibres with particularly
small cores are often termed ‘highly nonlinear fibres’. However, one can only take
advantage of the nonlinearity induced at this high intensity if the dispersion is
correct for the intended nonlinear process. The dispersion must be the first design
factor, this may result in several possible designs with different fibre nonlinearities.
Only if the correct dispersion is obtained can the high nonlinearity of a small core
fibre be useful. This is the case for supercontinuum generation in photonic crystal
fibres pumped at 800 nm, and is one of the reasons that the phenomenon was so
pronounced at such low energies when first demonstrated by Ranka et al. in 1999
[4, 5]. The exceptions to this rule are Brillouin and Raman scattering which do not
require specific dispersion or phasematching.
where vg is the group velocity and D has units of ps.nm1 .km1 . The two variables
D and ˇ 2 are related simply by a coefficient, but unfortunately this coefficient
contains a minus sign
2
ˇ2 D D : (9.18)
2c
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 203
D (ps / nm km)
anomalous
β2 (ps2 / km)
normal
0 0
λ ω
normal
anomalous
λ
ZDW
Fig. 9.3 Sketch of the dispersion of a fibre expressed in terms of D() or ˇ 2 (!), showing the
regions of normal and anomalous dispersion. The zero dispersion wavelength, ZDW , is marked
a b
ps / nm km)
Mode effective index nmode
ncore
0
(µm
· D (µ
(
nclad
0 l/ 8 0 l/ 8
Fig. 9.4 Dispersion of a waveguide mode: (a) Variation of modal index with normalised wave-
length for fixed core and cladding indices. (b) The resultant normalised dispersion. The horizontal
scale applies for ncore D 1.45, nclad D 1.0, and in that case the vertical scale in (b) is 800 to
400 m.ps.nm1 .km1
Optical fibres are made out of a particular material, usually silica glass, which
will have a material dispersion of its own, however this will be modified by the
dispersion arising from the optical waveguiding. The variation of the modal index
with wavelength was considered in Sect. 9.3: at long wavelengths the modal index
will become close to the index of the cladding and at short wavelengths the modal
index will become close to the index of the core. This effect and the consequent
waveguide group velocity dispersion are illustrated in Fig. 9.4.
204 W.J. Wadsworth
20 Step index
0
1.1 1.2 1.3 1.4 1.6 1.7
Wavelength (μm) m)
Dis
-10 per
sio
n fl
atte
ned
-20
Fig. 9.5 Dispersion of various fibres in the telecommunications wavelength range together with
example radial refractive index profiles
Fig. 9.6 Scanning electron microscope images of silica – air fibre structures. In all images,
black areas are air, grey/white areas are silica glass. (a) Free silica micro/nanowire (a tapered
optical fibre). The structure is suspended from the ends by standard fibre and input and output
tapered transition regions. Useful diameter range <0.5–5 m, fabricated length 1–25 cm. (b)
Silica micro/nanowire suspended from the sides by 100 nm wide silica webs in a high air filling
fraction PCF structure (overall fibre and close-up of the core region). Useful core diameter range
0.8–4 m, fabricated length 100s–1,000s m. (c) Low air filling fraction PCF for further dispersion
control. Useful core diameter range 1–20 m, fabricated length 100s–1,000s m. (d) Hollow-core
PCF. Core diameter 7 m at 800 nm, fabricated length 100s–1,000s m
Taking the extreme case of a silica microwire in air first, the material and waveguide
dispersion are plotted in Fig. 9.7a. It is clear that the dispersion of the waveguide
can now dominate the material dispersion in the visible and near infrared, opening
up the possibility of fibres with anomalous dispersion at wavelengths shorter than
1,270 nm (ZDW for silica glass). Separating the fibre dispersion into a waveguide
part (assuming a constant material index) and the bulk material dispersion gives
insight into the factors affecting the dispersion, but will not give an accurate
dispersion curve. A full calculation of the guided modes of a silica nanowire,
including the material dispersion, yields the dispersion curves shown in Fig. 9.7b.
The dispersion curves for 1.5, 2.0 and 2.5 m diameter strands look qualitatively
similar to the dispersion of silica (Fig. 9.7a solid line) or conventional step-
index fibres (Fig. 9.5) i.e. there is large normal dispersion at short wavelengths,
a zero dispersion (ZDW ) in the red or near infrared and anomalous dispersion at
longer wavelengths. There will be little difference in the fundamental linear and
nonlinear physics observed. However the change of ZDW from 1,270 nm to the
600–800 nm region is technologically important for nonlinear optics. To achieve
206 W.J. Wadsworth
a b
800
d=0
532 nm 780 nm 1064 nm
.5 μm
400
Dispersion D(ps/nm km)
0.7 μm
0
-
-200
0.5 μm
d=1
- 200
-
-400
μm
- 400
400 600 800 1000
1000 2000 3000 Wavelength (nm)
Wavelength (nm)
Fig. 9.7 (a) Material dispersion of silica (solid line) and waveguide dispersion for different
diameters of silica microwires in air; 0.5 m (dotted), 1.0 m (dashed), 2.0 m (dash-dot); (b)
Total dispersion of silica microwires in air with diameters from 0.5 to 2.5 m. The common laser
wavelengths of 532, 780 and 1,064 nm are marked
high nonlinearity one needs a high power laser, and this usually means a high
peak power pulsed laser. The Ti:sapphire laser, which is the established laboratory
workhorse, typically operates at 780 nm. Silica microwires in the form of small
core PCFs were the starting point for visible continuum generation in fibres at low
power because of the match of ZDW to the operating wavelength of modelocked
Ti:sapphire lasers [4, 5, 10, 15]. Frequency doubled pulsed Nd3C and Yb3C lasers
at around 532 nm are also very common, and a ZDW in this region is achievable
with sub-micron silica nanowires (Fig. 9.7b). Nonlinearity and supercontinuum
generation has been observed in these structures [12, 16], although the small size
brings further problems.
Taking next the case of a PCF with a relatively low air filling fraction, gives further
options for selecting appropriate waveguide dispersion. Here one may not only
select the core diameter, but also the core – cladding index step. This would be
useful even with a naive assumption that the index of the silica – air cladding
is simply given by the averaged index, taking into account the area of glass and
air in the cladding. However the reality is somewhat more complex, and luckily
also more effective in dispersion design. A PCF cladding of the type shown in
Fig. 9.6c has a scale that is large or comparable to the wavelength of light. This
means that the modes of the cladding (which define the effective cladding index
[17, 18]) will actually be well confined to the silica regions in the cladding, giving
a high effective index at short wavelengths. At longer wavelengths, or at smaller
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 207
Fig. 9.8 (a) Wavelength dependence of the effective index of a silica – air PCF cladding with
d/ƒ D 0.4, for different values of pitch: ƒ D 10 m (dot), ƒ D 5 m (dash), ƒ D 2 m (dash-
dot), compared to uniform bulk silica (solid), inset: representation of the structure (b) Measured
dispersion of three PCFs with pitch ƒ D 3 m and different air hole diameters close to d/ƒ D 0.4,
compared to uniform bulk silica (solid), inset: Scanning electron microscope image of core region
of one of the fibres (Part (b) reproduced and adapted with permission from Ref. [19])
cladding scale, the modes will spread out into the air (just as the fundamental mode
spreads from the core of a fibre at long wavelengths) and the effective index will be
low. The calculated effective index for an infinite cladding material for a particular
structure is shown in Fig. 9.8a for a range of wavelengths at different structural
scales. Thus a PCF cladding not only has an effective index which is controllable
over a wide range – from 1 to 1.45 – but the cladding index is also highly wavelength
dependent, adding another dimension to the waveguide dispersion. An example of
the application of this type of fibre for nonlinear optics is shown in Fig. 9.8b, where
a set of fibres has been fabricated with ZDW close to the wavelength of common
pulsed Nd3C and Yb3C lasers [19]. Again this is of technological importance
because the lowest cost high power pulsed lasers available are in the 1,030–1,080 nm
spectral range. These fibres serendipitously have one further useful property; they
are endlessly single mode (ESM), that is, they support only one guided mode at all
wavelengths [17, 18]. The variation of the cladding effective index exactly cancels
the term 1/ in the fibre parameter, V, (Eq. 9.16) at short wavelengths so that when
d/ƒ < 0.4 the effective V never rises above 2.4 (using an appropriate scaling factor
to define V for a PCF [17]). ESM fibre is reassuring for supercontinuum systems
discussed in Sect. 9.7, which may generate light over the range from less than
350 nm to beyond 2,400 nm.
Calculation of the propagation constants and dispersion of modes in PCFs is
generally a complex numerical task. However some commercial finite element codes
(such as Comsol) offer techniques for PCF modelling, and a mode solver using
the multipole method is currently available from the University of Sydney [20].
An empirical model provided by Saitoh and Koshiba [21] is an invaluable simple
tool for initial estimates of fibre design. One must always bear in mind the stated
validity limits for this model and also that as it separates out the contribution of
208 W.J. Wadsworth
Hollow-core photonic crystal fibres (HC-PCFs) (Fig. 9.6d) exploit the photonic
bandgaps that can be achieved in periodic dielectric structures – light of a certain
range of frequencies and propagation constant can be excluded from such a structure
[13, 14]. If we imagine an elongated doughnut of a photonic crystal then if light
at a frequency within the bandgap is launched into the central hole it cannot
escape through the photonic crystal and therefore must be guided along the hole.
In the usual implementations of photonic crystals, light propagates in the plane
of periodicity, whereas in a PCF the light travels perpendicular to the plane of
periodicity. For a HC-PCF one also requires that light is in a guided mode in
the hollow core region, which means that at any particular frequency the bandgap
only needs to cover the well-defined propagation constant for that mode. This is a
considerably simpler criterion to fulfil than the full 2-D bandgap required for in-
plane photonic crystal waveguides [22] and means that HC-PCFs can be fabricated
from silica-air structures and do not need high index materials like silicon.
Figure 9.9 shows the relatively narrow guiding band of a HC-PCF, which
may extend to 20 % of the central frequency. Attenuation and dispersion are
fundamentally linked by the Kramers – Kronig relations [23]. With a narrow guiding
band and sharp transitions to extremely high loss, the dispersion of HC-PCF is
determined primarily by the Kramers – Kronig relations and the position of the
guiding band. The dispersion imposed by the bandgap guidance is zero at the centre
of the bandgap and diverges to large normal dispersion at the short wavelength
edge and to large anomalous dispersion at the long wavelength edge. There is also
a contribution from waveguide dispersion, which is anomalous across the whole
guiding band, so the overall ZDW is shifted closer to the short wavelength edge, and
the dispersion is anomalous over much of the guiding band, as shown in Fig. 9.9.
This anomalous dispersion allows propagation of solitons at any wavelength where
HC-PCF can be fabricated, which has been demonstrated at 532, 800, 1,064 and
1,550 nm [24–27].
be true; the solitons occur for a particular pulse shape with defined energy (related
to the pulse duration) in a nonlinear Schrödinger equation that does not include the
Raman effect, fibre attenuation or a slope of the GVD curve (third-order dispersion).
Surely a real pulse of non-ideal pulse shape, with too much or too little energy and
launched into a real fibre with attenuation and higher order dispersive and nonlinear
terms will never be able to excite these magic unchanging pulses. The fact that
solitons are actually readily observable [27, 29] may be seen more easily from the
simple phenomenological description of a soliton, as follows. In a nonlinear medium
the leading edge of a pulse experiences a red-shift from self phase modulation.
With anomalous dispersion red-shifted frequencies have a slower group velocity
and so will tend to move back into the body of the pulse. Similarly the trailing edge
experiences a blue shift from self phase modulation, and the blue-shifted frequencies
travel forward into the pulse under anomalous dispersion. When the nonlinear SPM
and the linear dispersion are balanced then the pulse will propagate unchanged along
the fibre as a soliton. In this picture it is apparent that a soliton is not an impossible
stationary solution (like a pencil balanced on its point) but can be a self-building and
self-healing solution. In fact, solitons are extremely robust; solitons can form from
most input pulse shapes at energies far from the exact soliton energies, and solitons
will react to gradual changes as they propagate (such as reduction in energy through
fibre attenuation or changes in fibre dispersion) by adjusting temporal duration and
spectral bandwidth to suit the new conditions (Sect. 9.5.2).
Table 9.2 Soliton parameters for femtosecond pulses propagating in hollow-core PCF
Wavelength (nm) Pulse duration (fs) Soliton energy (nJ) Fibre length (m) Ref.
1,550 300 500 3 [27]
1,064 500 170 5 [26]
800 300 65 5 [25]
532 100 25 1 [24]
laboratory or from the laser to the work piece in a laser machining rig. However,
the energy transmitted as a soliton cannot be selected arbitrarily but must satisfy the
relation
3 DAeff
Esoliton D 1:76 2 ; (9.19)
2 cn2 0
where Esoliton is the soliton energy, is the wavelength of the light, 0 is the FWHM
pulse duration, D is the group velocity dispersion, Aeff is the modal effective area
in the fibre, c is the speed of light and n2 is the nonlinear refractive index. This
is the same condition as Eq. (8.10) and the explanation in Sect. 8.2, but with the
nonlinearity, , written out explicitly and the pulse duration expressed as FWHM
rather than 1/e2 width. If one uses typical values for fibre and pulse parameters
of 0 D 100 fs, D D 20 ps.nm1 .km1 , Aeff D 80 m2 and n2 D 2 1020 m2 W1
then Esoliton D 0.07 nJ at a wavelength of 800 nm or 0.5 nJ at 1,550 nm. The first
solitons demonstrated were 7 ps in duration at energies of just 8.4 pJ [29]. These
are extremely low energies. In order for soliton propagation to be useful the pulse
energy delivered by the fibre must be sufficient for the application: energies of
<1 nJ are in the range of telecommunications pulses (although solitons have not
proved greatly successful in this application because of the interactions of adjacent
solitons), and they may be enough for seeding amplifiers. For most of the other
applications covered in this book, energies of at least 10s or 100s nJ are required,
and often much more.
The energy in Eq. (9.19) is governed by the dispersion and the nonlinearity of the
fibre. In particular, if the nonlinearity of the fibre, n2 , can be reduced, then the soliton
energy is increased. Hollow-core PCF gives an excellent way of greatly reducing the
nonlinearity by replacing the guiding medium with air, or even a vacuum. There is
still some overlap of the guided mode with the silica glass in the cladding which
contributes to the nonlinearity, but the overall effect of changes in D, Aeff and n2
is an increase in the soliton energy by a factor of 100–1,000 more than the figures
calculated above for solid core fibres. This was first demonstrated at 1,550 nm with
a soliton energy of 500 nJ for 300 fs pulses propagated over 3 m of fibre [27]. Such
an energy is sufficient for micromachining applications (see Chaps. 12 and 13).
Subsequently, soliton propagation has been demonstrated at shorter wavelengths
as shown in Table 9.2. The pulse durations are output pulse durations, rounded to
the nearest 100 fs and the soliton energies are the minimum energy given in the
references for propagation of a soliton. The soliton energy scales with wavelength,
roughly as 3 , even though D and Aeff in Eq. (9.19) also change with wavelength.
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 211
The pulse durations for pulse delivery listed in Table 9.2 are the output pulse
duration, and are longer than the input pulse duration in all but the 532 nm case.
This is partly as a result of the formation of a soliton from an input pulse which may
not be transform limited or may not have the required sech2 pulse shape, partly as
a result of third-order dispersion, and partly as a result of attenuation in the fibre.
Eq. (9.19) is formulated to calculate the input energy required to launch a soliton,
however, once a soliton exists inside a fibre the free parameter is not the energy but
the pulse duration, so it is appropriate to reformulate Eq. (9.19) as
3 DAeff
0 D 1:76 : (9.20)
2 2 cn2 Esoliton
From this it is clear that a soliton experiencing attenuation in a fibre will adjust
to have a longer pulse duration. However, if the dispersion or effective area is
reduced along the length of the fibre, then the soliton will adjust to have a shorter
pulse duration. The length scales for any changes in fibre parameters must be long
compared with the characteristic soliton length (LD D LNL in Sect. 8.2 and [1]) in
order for these transformations to be adiabatic. Adiabatic pulse compression has
been observed in dispersion-decreasing conventional fibres [30, 31], and also for
high energy solitons in HC-PCFs [28, 32–34]. As the dispersion of HC-PCF is
defined by the position of the bandgap (Fig. 9.9), which in turn is defined by the
scale of the cladding structure, it is relatively simple to produce a HC-PCF with
decreasing dispersion by drawing a section of fibre with increasing outer diameter.
This shifts the bandgap and the ZDW to longer wavelengths, which reduces the
dispersion at any given wavelength. As an example, adiabatic soliton compression
has been achieved in a non-uniform HC-PCF at 800 nm [32]. The output pulse
duration was half the input pulse duration, whereas, with a uniform fibre, the
output pulse duration was nearly three times longer than the input. The longer pulse
duration in the uniform fibre is not simply due to chirp, the spectral bandwidth is
simultaneously reduced to maintain soliton pulses close to the transform limit. The
minimum pulse duration achievable in this way is limited by third order dispersion.
The small Raman gain shift in air can also impose a limit in air-filled fibres [35].
Adiabatic soliton compression has also been proposed by varying the nonlinearity
of a fibre using a pressure gradient of the gas in the core [33].
Higher order solitons, which may be launched at higher energies, experience
spectral broadening and pulse compression in the first stage of propagation [1]. With
appropriate pulse energy and fibre length this may be employed for high energy
pulse compression in HC-PCF [24, 36].
Linear compression of chirped pulses is also possible in HC-PCF. This is of
particular use for amplified modelocked fibre lasers, where SPM in the amplifier
can yield an output pulse with a broadened spectrum and linear normal chirp. The
chirp can be compensated for by a grating pair to give short, high energy pulses.
212 W.J. Wadsworth
Fig. 9.10 (a) Normalized experimental output pulse width and (b) spectrum as a function of the
propagation distance in HC-PCF for 5 ps SPM-broadened pulses at 320 nJ laser pulse energy
(Reproduced and adapted with permission from Ref. [26])
The same chirp compensation can also be achieved by the anomalous dispersion
of a HC-PCF, but there is also the benefit that the HC-PCF does have some
small nonlinearity. For an appropriate pulse energy, the pulse formed by linear
compression can undergo spectral re-shaping from SPM and the Raman effect to
form a transform limited soliton. This has been demonstrated for compression of
5 ps pulses to 300 fs [26] where the two stages of compression are clearly visible
from a cut-back measurement (Fig. 9.10). For the first 4.5 m of fibre the pulse
duration reduces because of the linear dispersion. In this stage the spectrum remains
almost unchanged as the peak power is insufficient to initiate nonlinear effects in the
HC-PCF. When the pulse is compressed to 500 fs after 4.5 m (approximately the
transform limit of the structured 12 nm input spectrum), the peak intensity is then
high enough for nonlinearity to re-shape the spectrum and after a total propagation
of 6 m a soliton is formed with a nearly transform limited smooth spectral bandwidth
< 3 nm, containing nearly 70 % of the total output energy [26].
The process of FWM/MI introduced in Sect. 9.2 allows for conversion of a laser at a
particular wavelength to another arbitrary wavelength, governed by phasematching
in a fibre. This is similar to bulk optical parametric generators and oscillators but
with two important differences: firstly the phasematching is provided by a guided
fibre mode, so there is no need for angle tuning and no possibility of spatial walk-
off, secondly the nonlinearity used is (3) rather than (2) and so the generated signal
and idler wavelengths are equally spaced around the pump rather than both being
longer than the pump wavelength. Conversion of, for example, 1,064–700 nm can
then be made in one step, rather than requiring frequency doubling to 532 nm and
subsequent (2) parametric conversion to 700 nm.
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 213
The energy conservation and phasematching conditions for the frequency, !, and
propagation constant, ˇ, for partially degenerate FWM in an optical fibre are
2!p D !s C !i (9.21)
and
where subscripts p, s and i refer to the pump, signal and idler waves respectively,
is the nonlinear parameter defined below Eq. (8.5) and P is the power in the
pump wave. FWM does not rely on a time varying intensity (unlike SPM), so these
relations are valid for CW or pulsed lasers, with P being the peak power for pulsed
lasers. Using Eq. (9.21) one can see that the signal and idler waves are shifted up
and down by equal frequency offset, , from the pump. One may then make a
Taylor expansion of the propagation constant ˇ(!) as in Eq. (8.3) so that Eq. (9.22)
becomes
with further terms if the Taylor expansion is carried beyond 4th order [37]. As
the pump power P is always positive and is positive in silica glass, this has
solutions for negative ˇ 2 (anomalous dispersion), or for sufficiently large negative
ˇ 4 even when ˇ 2 is positive (normal dispersion, close to ZDW ). An example of the
phasematching curve for a particular fibre is shown in Fig. 9.11a.
In the case of anomalous dispersion at the pump wavelength (right hand side
of Fig. 9.11a), the phasematched sidebands are relatively close to the pump and
the FWM gain extends over a relatively broad bandwidth (and is strongly power
dependent). Wavelength conversion has been extensively investigated for telecoms
applications in this regime, with an emphasis on dispersion flattened fibres with
small and constant ˇ 2 < 0 (ˇ 3 0, ˇ 4 0) to maximise the bandwidth [39].
Dispersion control in PCF offers the potential to improve the bandwidth greatly
with fibres with extremely small and flat anomalous dispersion. There are very many
proposed designs in the literature, but very few have been made [37, 40] owing to
the very tight dimensional tolerances. These arise because in order to achieve a
low fibre dispersion one needs to exactly cancel the material dispersion with the
waveguide dispersion. A very small percentage error in the waveguide dispersion
214 W.J. Wadsworth
Fig. 9.11 (a) Calculated phasematching curve for FWM in a particular birefringent PCF when
pumped close to its zero dispersion wavelength on each of the two principal polarisation axes,
ZDW D 655 and 662 nm. (b) Observed output spectra for pumping at 647 nm with different
polarisations (Reproduced and adapted with permission from Ref. [38])
leads to a large deviation from zero in the resultant fibre dispersion. An example
of a dispersion flattened PCF used for signal conversion in the telecommunications
band is shown by Anderson et al. [40] where signal and idler gain bandwidths of
30 nm were demonstrated (compared to 15 nm in [39]). Fibre optical parametric
oscillators in small-core PCF pumped at 750 nm have also been demonstrated based
on this parametric gain [41].
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 215
In the case of normal dispersion at the pump wavelength (left hand side of
Fig. 9.11a, the phasematched sidebands are far from the pump and the FWM gain
extends over only a narrow bandwidth. This only occurs for relatively large ˇ 4 ,
which is a curvature in the spectral variation of the group velocity dispersion.
It is clear from Fig. 9.7 that the large waveguide dispersion required to bring
the zero dispersion wavelength down to the visible and near-infrared in a PCF
usually imparts a large curvature to the dispersion, so this region of phasematching
is particularly common in PCFs at short wavelengths, and less pronounced in
conventional fibres at telecommunications wavelengths. Figure 9.11b shows an
example of this type of phasematching used to produce signal and idler sidebands
close to 500 and 1,000 nm when pumped with 60 ps pulses at 647 nm [38].
Using PCFs with ZDW close to 1,064 nm (Fig. 9.8b) replicates these results in
both normal and anomalous regimes [19, 42]. At this pump wavelength there are
many pump laser options, from nanosecond and sub-nanosecond Q-switched lasers
(in particular compact microchip lasers [19, 43, 44]) to fs or picosecond modelocked
bulk or fibre lasers [45, 46]. The FWM process is efficient even without a feedback
cavity. A single pass of fibre has been shown to give more than 30 % conversion
to signal wavelengths around 700 nm for a variety of laser configurations using
0.5–4 m fibre lengths [19, 43–46]. The range of signal wavelengths observed from
a 1,064 nm pump spans 650–950 nm with corresponding idler wavelengths from
1,200 nm to nearly 3 m, Fig. 9.12. With high power, high repetition rate pump
sources, 2.16 W has been generated at 20 MHz at 740 nm [45, 46] and 450 mW
at 1 MHz at 2,539 nm [43, 44]. The steepness of the phasematching curve on
the left hand side of Fig. 9.11a indicates that for a relatively small pump tuning
range there will be a wide signal and idler tuning range, the potential of which is
discussed extensively in [43–46]. However this also points to a bandwidth constraint
on the pump laser: if the pump bandwidth is too broad then different wavelength
components of the pump will be phasematched for generation of different FWM
wavelengths and the process is not efficient and eventually becomes swamped by
non-phasematched Raman scattering.
In Sects. 9.2 and 9.6.1 the FWM process has been treated in a classical manner.
However it is clear that in a quantised electromagnetic field the FWM process, Eq.
(9.21), corresponds to the annihilation of two pump photons and the creation of a
signal and idler photon with the same total energy, and Eq. (9.22) corresponds to
conservation of photon momentum. This means that the signal and idler field must
contain the same number of photons: FWM cannot generate a signal photon without
an idler photon (or vice versa). If one then imagines turning the pump power down
sufficiently so that there is on average less than one FWM event per pulse, then the
few pulses where anything happens contain exactly one signal photon and one idler
photon. This is the basis of a heralded single photon source: the two photons are
216 W.J. Wadsworth
Fig. 9.12 (a) Output spectra for FWM (658 and 667 nm) in two different PCFs pumped with
30 ps pulses at 1,064 nm [45], (b) Output spectra and signal and idler efficiency for FWM in
PCF pumped with 200 ps pulses at 1,064 nm [43] (Parts (a) and (b) reproduced and adapted with
permission from Refs. [45] and [43] respectively)
separated (they are at different wavelengths) and one (e.g. the signal) is detected.
Detection of a single signal photon from particular pulse determines (heralds) that
there must be an idler photon from that pulse in the experiment. This idler photon
(together with others from similar sources) can then be used as the basic resource of
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 217
Most SCG sources employ a fs or ps modelocked laser and a fibre with a zero
dispersion wavelength, ZDW , slightly shorter than the pump wavelength giving
small anomalous dispersion at the pump wavelength. For pumping at 1,550 nm,
conventional step-index fibres are appropriate (Fig. 9.5), in particular those with
highly Ge-doped cores giving a small core diameter and relatively high nonlinearity
[57]. For pumping at 1,064 nm, PCFs provide appropriate dispersion (Fig. 9.8b) and
generate broad, flat supercontinua that extend into the visible [19, 42]. Pumping at
800 nm requires small core, high index contrast structures (Fig. 9.7b) either as PCFs
[4, 5, 9] or tapered optical fibres [10]. Figure 9.7b shows that this same approach
applies in the visible, down to pump wavelengths at 532 nm [12].
All of these sources will provide a broad spectrum at the output of the fibre,
in a single spatial mode. This is an extremely bright source in terms of mW per
nanometre spectral bandwidth per unit solid angle per unit area of the source.
Indeed, the spectral coverage is similar to the output of an incandescent lightbulb,
but because it is a directed beam, an SCG source is more than 10,000 times brighter
even for just 100 mW of total power. Such sources have obvious applications; in
photonic testing (measuring the attenuation of a single mode optical fibre over a
large spectral range requires a source with a large spectral range and sufficient power
in a single mode); spectroscopy; confocal microscopy and a range of nonlinear
spectroscopic and microscopic techniques. However not all SCG sources are good
for all applications. The details of the different processes involved in SCG require
SCG sources to be application specific for demanding applications.
The biggest issue in SCG is the fundamental pulse-to-pulse fluctuations inherent
in SCG when it arises from soliton breakup or from MI based spectral broadening.
In this context it is useful to consider FWM in the time domain as MI to emphasise
the instability. Both soliton breakup and MI are seeded from noise (Sect. 8.5) which
is necessarily random and different from pulse to pulse. This can lead to pulse-to-
pulse fluctuations in intensity of >10 % over most of a supercontinuum spectrum
(e.g. when pumped with ps to ns pulses at 1,064 nm), with fluctuations rising to
>50 % towards the edges of the continuum spectrum.
Perhaps surprisingly, this fluctuation actually imparts some useful properties
to an SCG source for some applications. As the fluctuations are caused by truly
random noise rather than systematic effects, they will average out over time with
p
the usual 1/ N dependence for N pulses (except for the very edges of the SCG
spectrum with extreme statistics [58]). This, together with the picture of an MI
based SCG containing many individual solitons at different wavelengths, suggests
that with each pulse giving a randomly different spectral structure, the average
spectrum will be smooth. This is indeed the case; long pump pulses (e.g. >1 ps)
which can be expected to produce SCG through MI, usually do give very smooth
and flat averaged spectra. With modelocked pump sources operating at 10 MHz,
the fluctuations average out rapidly, even 50 % pulse-to-pulse variation averages
to 0.5 % in 1 ms (10,000 pulses). These sources are then excellent for almost any
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 219
single point spectroscopic measurement even at high sample rates, but start to show
problems if used for 2-D high speed imaging. A 2-D image even at a relatively
low resolution contains hundreds of thousands of pixels. With a video frame rate
there are now only a few laser pulses available for each pixel and the fundamental
SCG fluctuation will begin to show up. This is potentially an issue for confocal and
other point-by-point microscopes and also for optical coherence tomography (OCT)
which aims towards one pulse per voxel data rates. Other technologies are even more
demanding, for example optical frequency metrology requires the pulse-to-pulse
phase of the laser to be carried through into the continuum ([59, 60], Chap. 6).
Of all the processes involved in SCG, only SPM will allow deterministic, and
therefore stable, transformation of the input pulse into a continuum (see also
Sect. 8.5). SPM requires a large dI/dt so will be particularly prevalent for short pump
pulses. This was recognised experimentally in the context of frequency metrology
(early experiments with 100 fs pulsed lasers simply did not work, and shorter pulses
provided a solution) and then explained through numerical modelling [61]. In these
studies, fibres with anomalous dispersion were considered, so MI will compete with
SPM. However, in the normal dispersion regime the widely separated FWM/MI
bands are a much weaker competitor to SPM. Furthermore, it is possible to fabricate
fibres where the dispersion is always normal, there is no anomalous dispersion, so
there is no zero dispersion wavelength and hence no phasematching for FWM/MI
(Eq. 9.25). In this case only SPM broadening can occur [62]. The spectral bandwidth
may be easily estimated (Section 4.1.2 of Ref. [1]) and is usually relatively narrow.
By designing a fibre with a sufficiently low value of the normal dispersion, the pump
pulse will retain its peak intensity for a long fibre length and the SPM broadening
may extend to several hundred nanometres. It is, of course, imperative that there is
no anomalous dispersion or ZDW over the generated bandwidth and that the low
normal dispersion is maintained [62, 63].
Section 8.4.1 provides a picture of SCG as a process where high order solitons break
up into multiple fundamental solitons, each of which shifts to longer wavelength
under the soliton self frequency shift (SSFS). It can also be shown that short
wavelength radiation, in the normal dispersion regime, becomes trapped by these
solitons and is shifted to even shorter wavelengths [64]. The actual short wavelength
which is trapped is determined by group velocity matching to the long wavelength
soliton. The wavelength coverage of an SCG spectrum is then defined by how far the
SSFS can take a soliton into the infrared, and how the group velocity of this soliton
220 W.J. Wadsworth
is matched to the group velocity at short wavelengths [64, 65]. At short wavelengths
the index of the guided mode is close to the core index, and also the dispersion of
silica becomes increasingly strong and is dominant over the waveguide dispersion,
so that in practice it is not possible to make large changes to the group velocity of a
fibre mode for wavelengths below 500 nm. Extension of SCG to both long and short
wavelength then relies solely on the wavelength and the group velocity of the long
wavelength SSFS shifting solitons [56, 65, 66].
Extension of the SSFS to longer wavelength is determined by attenuation and
dispersion in the fibre. Equation (9.20) shows that if a soliton loses energy, then
the soliton duration increases. The peak power suffers doubly because less energy
is then spread over a longer time. Eventually there is insufficient peak power for
the SSFS to continue. As the SSFS is based on the Raman effect, photon number
is conserved and energy is lost to vibrations of the silica matrix as the wavelength
increases. This, however, is not generally seen to be the main practical limit for
long wavelength generation. Usually the SSFS is limited by absorption in the fibre,
with rising dispersion at long wavelengths sometimes also playing a role. In silica
glass the intrinsic absorption rises to 10 dB/m at 2.7 m and 100 dB/m at
3.5 m, and the extrinsic absorption from OH impurities (usually below the ppm
level) adds absorption bands at 1.38, 2.2 and 2.76 m with strength increasing with
wavelength [67]. The long wavelength of a supercontinuum is then usually limited
to 2.4–2.6 m, but SCG in fibres with high OH content can be limited by the
2.2 or even 1.38 m OH absorption bands. Very high peak power pump pulses and
short fibre lengths mean that even the very high absorption of silicate glasses can be
overcome to produce spectra extending to beyond 3 m [68]. Other glasses, such as
fluorides, tellurites or chalcogenides, with lower intrinsic infrared absorption, have
also been used [56, 69–71] allowing extension to beyond 4 m and even as far as
4.8 m in particularly short fibres [70].
Stone and Knight [65] designed fibres to match the group index of 2.6 m
solitons to the shortest wavelengths possible and achieved short SCG wavelengths
in the 400 nm region from a 1,064 nm pump. A further constraint is that there
must be some short wavelength radiation generated initially, which is group velocity
matched to wavelengths near the pump, in order for it to be able to be swept down
to shorter wavelength by the SSFS solitons. This usually means that the pump
wavelength must be close to the ZDW of the fibre. The ideal PCF for group velocity
matching to very short wavelength would be a silica microwire type PCF (Fig. 9.6b)
with a core diameter of around 2 m. This has a ZDW close to 715 nm and is
therefore not suitable for pumping at 1,064 nm. However, it is simple to generate
715 nm by FWM or SCG from pulses at 1,064 nm, so multistage approaches have
been demonstrated to achieve visible wavelengths, either using a FWM stage to
generate 740 nm followed by SCG using the 740 nm pump [72], or using a series
of concatenated PCFs with different core diameters [73]. The limit of concatenated
fibres is, of course, a continuously varying fibre, which was demonstrated to yield
375 nm from a pump at 1,064 nm [66] and, more recently, wavelengths well below
350 nm have been generated [74].
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 221
Another technique for achieving high power at short wavelength from an infrared
source is to use dual wavelength pumping with both wavelengths far from the ZDW
that lies between them [75]. The short wavelength may be generated as the second
harmonic of the pump in a nonlinear crystal [75], or by FWM in an all-fibre device
[76]. These techniques also allow extension to below 400 nm.
9.8 Conclusion
References
1. G.P. Agrawal, Nonlinear Fiber Optics, 4th edn. (Academic, Burlington, 2007)
2. R. Paschotta, Encyclopedia of Laser Physics and Technology (Wiley-VCH, Berlin, 2008)
3. R. Paschotta, Encyclopedia of Laser Physics and Technology (2010), http://www.rp-photonics.
com/optical intensity.html. Accessed 3 Sept 2010
4. J.K. Ranka, R.S. Windeler, A.J. Stentz, Efficient visible continuum generation in air-silica
microstructure optical fibers with anomalous dispersion at 800 nm, in Conference on Lasers
and Electro-Optics, CLEO ’99, Baltimore, May 1999, postdeadline paper CPD8 (1999)
5. J.K. Ranka, R.S. Windeler, A.J. Stentz, Visible continuum generation in air silica microstruc-
ture optical fibers with anomalous dispersion at 800 nm. Opt. Lett. 25, 25–27 (2000)
6. J.M. Senior, Optical Fiber Communications: Principles and Practice, 3rd edn. (Prentice Hall,
Harlow, 2008)
7. B.J. Ainslie, C.R. Day, A review of single-mode fibers with modified dispersion characteristics.
IEEE J. Lightwave Technol. LT-4, 967–979 (1986)
8. D. Mogilevtsev, T.A. Birks, P.S.J. Russell, Group-velocity dispersion in photonic crystal fibres.
Opt. Lett. 23, 1662–1664 (1998)
9. J.C. Knight, J. Arriaga, T.A. Birks, A. Ortigosa-Blanch, W.J. Wadsworth, P.S.J. Russell,
Anomalous dispersion in photonic crystal fiber. IEEE Photonics Technol. Lett. 12, 807–809
(2000)
10. T.A. Birks, W.J. Wadsworth, P.S.J. Russell, Supercontinuum generation in tapered fibers. Opt.
Lett. 25, 1415–1417 (2000)
11. L. Tong, R.G. Gattas, J.A. Ashcom, S. He, J. Lou, M. Shen, I. Maxwell, E. Mazur,
Subwavelength-diameter silica wires for low-loss optical wave guiding. Nature 426, 816–819
(2003)
12. S.G. Leon-Saval, T.A. Birks, W.J. Wadsworth, P.S.J. Russell, M.W. Mason, Supercontinuum
generation in submicron fibre waveguides. Opt. Express 12, 2864–2869 (2004)
222 W.J. Wadsworth
13. J.C. Knight, Photonic crystal fibers. Nature 424, 847–851 (2003)
14. P.S.J. Russell, Photonic crystal fibers. Science 299, 358–362 (2003)
15. W.J. Wadsworth, A. Ortigosa-Blanch, J.C. Knight, T.A. Birks, T.-P.M. Man, P.S.J. Russell,
Supercontinuum generation in photonic crystal fibers and optical fiber tapers: a novel light
source. J. Opt. Soc. Am. B 19, 2148–2155 (2002)
16. M. Delgado-Pinar, P.J. Mosley, J.C. Knight, T.A. Birks, W.J. Wadsworth, Visible supercontin-
uum generation in the femtosecond regime in submicron structures. in Nonlinear Photonics,
OSA Technical Digest (CD) (Optical Society of America, 2010), paper NWD3, Karlsruhe,
Germany
17. T.A. Birks, D. Mogilevtsev, J.C. Knight, P.St.J. Russell, J. Broeng, P.J. Roberts, J.A. West,
D.C. Allan, J.C. Fajardo, The analogy between photonic crystal fibres and step index fibres.
in Optical Fibre Conference, OSA Technical Digest (Optical Society of America, 1999), paper
FG4-1, San Diego, USA, pp. 114–116
18. T.A. Birks, J.C. Knight, P.S.J. Russell, Endlessly single-mode photonic crystal fibre. Opt. Lett.
22, 961–963 (1997)
19. W.J. Wadsworth, N. Joly, J.C. Knight, T.A. Birks, F. Biancalana, P.S.J. Russell, Supercontin-
uum and four-wave mixing with Q-switched pulses in endlessly single-mode photonic crystal
fibres. Opt. Express 12, 299–309 (2004)
20. CUDOS MOF utilities. University of Sydney (2011). http://sydney.edu.au/science/physics/
cudos/research/mofsoftware.shtml. Accessed 28 July 2011
21. K. Saitoh, M. Koshiba, Empirical relations for simple design of photonic crystal fibers. Opt.
Express 13, 267–274 (2005)
22. E. Yablonovitch, Inhibited spontaneous emission in solid-state physics and electronics. Phys.
Rev. Lett. 58, 2059–2062 (1987)
23. R.d.L. Kronig, On the theory of dispersion of X-rays. J. Opt. Soc. Am. 12, 547–556 (1926)
24. W. Huang, M.G. Welch, P.J. Mosley, B.J. Mangan, W.J. Wadsworth, J.C. Knight, Ultrashort
pulse delivery in hollow-core photonic bandgap fiber at 540 nm. in Conference on Lasers and
Electro-Optics, OSA Technical Digest (CD) (Optical Society of America, 2010), paper CWC3,
San Jose, USA
25. F. Luan, J.C. Knight, P.S.J. Russell, S. Campbell, D. Xiao, D.T. Reid, B.J. Mangan, D.P.
Williams, P.J. Roberts, Femtosecond soliton pulse delivery at 800 nm wavelength in hollow-
core photonic bandgap fibers. Opt. Express 12, 835–840 (2004)
26. F. Gèrôme, P. Dupriez, J. Clowes, J.C. Knight, W.J. Wadsworth, High power tunable fem-
tosecond soliton source using hollow-core photonic bandgap fiber, and its use for frequency
doubling. Opt. Express 16, 2381–2386 (2008)
27. D.G. Ouzounov, F.R. Ahmad, D. Müller, N. Venkataraman, M.T. Gallagher, M.G. Thomas,
J. Silcox, K.W. Koch, A.L. Gaeta, Generation of megawatt optical solitons in hollow-core
photonic band-gap fibers. Science 301, 1702–1704 (2003)
28. M.G. Welch, Compressing and Propagating Solitons in Hollow Core Photonic Crystal Fibre.
PhD thesis (University of Bath, Bath, 2010). http://opus.bath.ac.uk/25235/
29. L.F. Mollenauer, R.H. Stolen, J.P. Gordon, Experimental observation of picosecond pulse
narrowing and solitons in optical fibers. Phys. Rev. Lett. 45, 1095–1098 (1980)
30. K. Tajima, Compensation of soliton broadening in nonlinear optical fibers with loss. Opt. Lett.
12, 54–56 (1987)
31. H. Kuehl, Solitons on an axially nonuniform optical fiber. J. Opt. Soc. Am. B 5, 709–713
(1988)
32. F. Gérôme, K. Cook, A.K. George, W.J. Wadsworth, J.C. Knight, Delivery of sub-100 fs pulses
through 8 m of hollow-core fiber using soliton compression. Opt. Express 15, 7126–7131
(2007)
33. J. Lægsgaard, P.J. Roberts, Theory of adiabatic pressure-gradient soliton compression in
hollow-core photonic bandgap fibers. Opt. Lett. 34, 3710–3712 (2009)
34. M.G. Welch, K. Cook, R. Amezcua-Correa, F. Gérôme, W.J. Wadsworth, A.V. Gorbach, D.V.
Skryabin, J.C. Knight, Solitons in hollow core photonic crystal fiber: engineering nonlinearity
and compressing pulses. IEEE J. Lightwave Technol. 27, 1644–1652 (2009)
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 223
35. A.V. Gorbach, D.V. Skryabin, Soliton self-frequency shift, non-solitonic radiation and self-
induced transparency in air-core fibers. Opt. Express 16, 4858–4865 (2008)
36. D.G. Ouzounov, C.J. Hensley, A.L. Gaeta, N. Venkateraman, M.T. Gallagher, K.W. Koch,
Soliton pulse compression in photonic band-gap fibers. Opt. Express 13, 6153–6159 (2005)
37. W.H. Reeves, D.V. Skryabin, F. Biancalana, J.C. Knight, P.S.J. Russell, F.G. Omenetto, A.
Efimov, A.J. Taylor, Transformation and control of ultra-short pulses in dispersion-engineered
photonic crystal fibres. Nature 424, 511–515 (2003). doi:10.1038/nature01798
38. J.D. Harvey, R. Leonhardt, S. Coen, G.K.L. Wong, J.C. Knight, W.J. Wadsworth, P.S.J. Russell,
Scalar modulation instability in the normal dispersion regime by use of a photonic crystal fiber.
Opt. Lett. 28, 2225–2227 (2003)
39. J. Hansryd, P.A. Andrekson, Broad-band continuous-wave-pumped fiber optical parametric
amplifier with 49-dB gain and wavelength-conversion efficiency. IEEE Photonics Technol.
Lett. 13, 194–196 (2001)
40. P.A. Andersen, T. Tokle, Y. Geng, C. Peucheret, P. Jeppesen, Wavelength conversion of a
40-Gb/s RZ-DPSK signal using four-wave mixing in a dispersion-flattened highly nonlinear
photonic crystal fiber. IEEE Photonics Technol. Lett. 17, 1908–1910 (2005)
41. J.E. Sharping, M. Fiorentino, P. Kumar, R.S. Windeler, Optical parametric oscillator based on
four-wave mixing in microstructure fiber. Opt. Lett. 27, 1675–1677 (2002)
42. M. Seefeldt, A. Heuer, R. Menzel, Compact white-light source with an average output power
of 2.4 W and 900 nm spectral bandwidth. Opt. Commun. 216, 199–202 (2003)
43. D. Nodop, C. Jauregui, D. Schimpf, J. Limpert, A. Tünnermann, Efficient high-power
generation of visible and mid-infrared light by degenerate four-wave-mixing in a large-mode-
area photonic-crystal fiber. Opt. Lett. 34, 3499–3501 (2009)
44. D. Nodop, C. Jauregui, D. Schimpf, J. Limpert, A. Tünnermann, Efficient high power
generation of pulsed red light via four-wave-mixing in a large-mode-area, endlessly single-
mode photonic-crystal fiber. in CLEO/Europe and EQEC 2009 Conference Digest (Optical
Society of America, 2009), paper CJ5 5 (Munich, Germany, 2009) http://www.opticsinfobase.
org/abstract.cfm?URI=CLEO E-2009-CJ5 5
45. L. Lavoute, J.C. Knight, P. Dupriez, W.J. Wadsworth, High power red and near-IR generation
using four wave mixing in all integrated fibre laser systems. Opt. Express 18, 16193–16205
(2010)
46. L. Lavoute, W.J. Wadsworth, J.C. Knight, Efficient four wave mixing from a picosecond fibre
laser in photonic crystal fibre. in CLEO/Europe and EQEC 2009 Conference Digest (Optical
Society of America, 2009), paper CJ5 4 (Munich, Germany, 2009)
47. P.G. Kwiat, K. Mattle, H. Weinfurter, A. Zeilinger, A.V. Sergienko, Y.H. Shih, New high-
intensity source of polarization-entangled photon pairs. Phys. Rev. Lett. 75, 4337–4341 (1995)
48. X. Li, P.L. Voss, J.E. Sharping, P. Kumar, Optical-fiber source of polarization-entangled
photons in the 1550 nm telecom band. Phys. Rev. Lett. 94, 053601 (2005)
49. J.E. Sharping, J. Chen, X.Y. Li, P. Kumar, R.S. Windeler, Quantum-correlated twin photons
from microstructure fiber. Opt. Express 12, 3086–3094 (2004)
50. J. Fan, A. Migdall, A broadband high spectral brightness fiber-based two-photon source. Opt.
Express 15, 2915–2920 (2007)
51. J. Fulconis, O. Alibart, J.L. O’Brien, W.J. Wadsworth, J.G. Rarity, Nonclassical interference
and entanglement generation using a photonic crystal fiber pair photon source. Phys. Rev. Lett.
99, 120501 (2007)
52. C. Söller, B. Brecht, P.J. Mosley, L.Y. Zang, A. Podlipensky, N.Y. Joly, P.S.J. Russell, C.
Silberhorn, Bridging visible and telecom wavelengths with a single-mode broadband photon
pair source. Phys. Rev. A 81, 031801(R) (2010)
53. J.G. Rarity, J. Fulconis, J. Duligall, W.J. Wadsworth, P.S.J. Russell, Photonic crystal fiber
source of correlated photon pairs. Opt. Express 13, 534–544 (2005)
54. A.R. McMillan, J. Fulconis, M. Halder, C. Xiong, J.G. Rarity, W.J. Wadsworth, Narrowband
high-fidelity all-fibre source of heralded single photons at 1570 nm. Opt. Express 17, 6156–
6165 (2009)
224 W.J. Wadsworth
55. A. Kudlinski, A. Mussot, Visible cw-pumped supercontinuum. Opt. Lett. 33, 2407–2409
(2008)
56. C. Xia, Z. Xu, M.N. Islam, F.L. Terry, M.J. Freeman, A. Zakel, J. Mauricio, 10.5 W time-
averaged power mid-IR supercontinuum generation extending beyond 4 m with direct pulse
pattern modulation. IEEE. J. Sel. Top. Quantum 15, 422–434 (2009)
57. J.W. Nicholson, A.D. Yablon, P.S. Westbrook, K.S. Feder, M.F. Yan, High power, single mode,
all-fiber source of femtosecond pulses at 1550 nm and its use in supercontinuum generation.
Opt. Express 12, 3025–3034 (2004)
58. D.R. Solli, C. Ropers, P. Koonath, B. Jalali, Optical rogue waves. Nature 450, 1054–1057
(2007)
59. D.A. Jones, S.A. Diddams, J.K. Ranka, A. Stentz, R.S. Windeler, J.L. Hall, S.T. Cundiff,
Carrier-envelope phase control of femtosecond mode-locked lasers and direct optical frequency
synthesis. Science 288, 635–639 (2000)
60. R. Holzwarth, J. Reichert, T. Udem, T.W. Hänsch, J.C. Knight, W.J. Wadsworth, P.S.J. Russell,
An optical frequency synthesiser for precision spectroscopy. Phys. Rev. Lett. 85, 2264–2267
(2000)
61. J.M. Dudley, S. Coen, Coherence properties of supercontinuum spectra generated in photonic
crystal and tapered optical fibers. Opt. Lett. 27, 1180–1182 (2002)
62. G. Humbert, W.J. Wadsworth, S.G. Leon-Saval, J.C. Knight, T.A. Birks, P.S.J. Russell, M.J.
Lederer, D. Kopf, K. Wiesauer, E.I. Breuer, D. Stifter, Supercontinuum generation system for
optical coherence tomography based on tapered photonic crystal fibre. Opt. Express 14, 1596–
1603 (2006)
63. L.E. Hooper, P.J. Mosley, A.C. Muir, W.J. Wadsworth, J.C. Knight, Coherent supercontinuum
generation in photonic crystal fiber with all-normal group velocity dispersion. Opt. Express 19,
4902–4907 (2011)
64. A.V. Gorbach, D.V. Skryabin, Light trapping in gravity-like potentials and expansion of
supercontinuum spectra in photonic-crystal fibres. Nat. Photonics 1, 653–657 (2007)
65. J.M. Stone, J.C. Knight, Visibly “white” light generation in uniform photonic crystal fiber
using a microchip laser. Opt. Express 16, 2670–2675 (2008)
66. A. Kudlinski, A.K. George, J.C. Knight, J.C. Travers, A.B. Rulkov, S.V. Popov, J.R. Taylor,
Zero-dispersion wavelength decreasing photonic crystal fibers for ultraviolet-extended super-
continuum generation. Opt. Express 14, 5715–5722 (2006)
67. O. Humbach, H. Fabian, U. Grzesik, U. Haken, W. Heitmann, Analysis of OH absorption bands
in synthetic silica. J. Non-Cryst. Solids 203, 19–26 (1996)
68. F.G. Omenetto, N.A. Wolchover, M.R. Wehner, M. Ross, A. Efimov, A.J. Taylor, V.V.R.K.
Kumar, A.K. George, J.C. Knight, N.Y. Joly, P.S.J. Russell, Spectrally smooth supercontinuum
from 350 nm to 3 m in sub-centimeter lengths of soft-glass photonic crystal fibers. Opt.
Express 14, 4928–4934 (2006)
69. M.R. Lamont, B. Luther-Davies, D.-Y. Choi, S. Madden, B.J. Eggleton, Supercontinuum
generation in dispersion engineered highly nonlinear (” D 10 /W/m) As2 S3 chalcogenide
planar waveguide. Opt. Express 16, 14938–14944 (2008)
70. P. Domachuk, N.A. Wolchover, M. Cronin-Golomb, A. Wang, A.K. George, C.M.B. Cordeiro,
J.C. Knight, F.G. Omenetto, Over 4000 nm bandwidth of mid-IR supercontinuum generation
in sub-centimeter segments of highly nonlinear tellurite PCFs. Opt. Express 16, 7161–7168
(2008)
71. M. El-Amraoui, J. Fatome, J.C. Jules, B. Kibler, G. Gadret, C. Fortier, F. Smektala, I.
Skripatchev, C.F. Polacchini, Y. Messaddeq, J. Troles, L. Brilland, M. Szpulak, G. Renversez,
Strong infrared spectral broadening in low-loss As-S chalcogenide suspended core microstruc-
tured optical fibers. Opt. Express 18, 4547–4556 (2010)
72. C. Xiong, A. Witkowska, S.G. Leon-Saval, T.A. Birks, W.J. Wadsworth, Enhanced visible
continuum generation from a microchip 1064 nm laser. Opt. Express 14, 6188–6193 (2006)
9 Nonlinear Wavelength Conversion and Pulse Propagation in Optical Fibres 225
73. J.C. Travers, S.V. Popov, J.R. Taylor, Extended blue supercontinuum generation in cascaded
holey fibers. Opt. Lett. 30, 3132–3134 (2005)
74. J.C. Travers, A.B. Rulkov, S.V. Popov, J.R. Taylor, A. Kudlinski, A.K. George, J.C. Knight,
Multi-watt supercontinuum generation from 0.3 to 2.4 m in PCF tapers. in Conference on
Lasers and Electro-Optics/Quantum Electronics and Laser Science Conference CLEO/QELS,
OSA Technical Digest Series (Optical Society of America, 2007), paper JTuB2 (Baltimore,
USA, 2007)
75. P.A. Champert, V. Couderc, P. Leproux, S. Février, V. Tombelaine, L. Labonté, P. Roy, C.
Froehly, P. Nérin, White-light supercontinuum generation in normally dispersive optical fiber
using original multiwavelength pumping system. Opt. Express 12, 4366–4371 (2004)
76. C. Xiong, Z. Chen, W.J. Wadsworth, Dual-wavelength-pumped supercontinuum generation in
an all-fiber device. IEEE J. Lightwave Technol. 27, 1638–1643 (2009)
Chapter 10
Applications of Ultra-Intense, Short
Laser Pulses
10.1 Introduction
Outstanding progress has been made in high power laser technology over the last
10 years, with laser powers reaching petawatt levels. At present there are 15 petawatt
lasers built or being built around the world and plans are afoot for new even higher
power lasers reaching values of exawatt or even zetawatt levels. Petawatt lasers
generate electric fields of 1012 V m1 with a large fraction of the total pulse energy
being converted to relativistic electrons with energies in excess of 1 GeV. In turn
these electrons result in the generation of beams of protons, heavy ions, neutrons
and high-energy photons. These laser driven particle beams have encouraged many
to think of carrying out experiments normally associated with conventional nuclear
accelerators and reactors. It must be acknowledged right from the outset that to
date laser driven particle beams have yet to compete favourably with conventional
nuclear accelerator generated beams in any way but in principle have the potential
of providing compact accelerators with much reduced shielding.
In this chapter there will be three sections: high intensity laser production
of electrons, protons, ions and photons, applications of ultra-high intensity laser
physics and the future of the field particularly the Extreme Light Infrastructure
(ELI), and the Photo Nuclear Pillar in Romania.
‘Is the future of conventional RF accelerators limited?’ This is the first question
one might ask, motivated by the observation that the size of such accelerators
is becoming prohibitive and the huge electric fields necessary to provide even
higher particle energies is likely to cause breakdown in the accelerator structures.
In his retiral address, Enrico Fermi suggested that a PeV accelerator would have
a circumference that would encircle the earth. Scientists are presently seeking
solutions to these problems to provide more compact accelerators. One of the
possible solutions is to use high power laser acceleration. After the invention
of the laser in 1960, the laser powers have increased dramatically and indeed
recent advances in laser technology have led to the development of compact multi-
terawatt and petawatt pulsed laser systems in many laboratories worldwide. When
focused to an area of a few tens of square microns or less, this laser radiation
can reach intensities close to 1021 W cm2 . Figure 10.1a illustrates the increase
in achievable laser intensity since 1960. Arguably the greatest advance was made
in the mid-1980s with the introduction of ‘chirped pulse amplification’ (CPA) [1].
In CPA a laser pulse of the order of femtoseconds or picoseconds is temporally
stretched by 3–4 orders of magnitude using dispersive gratings, thus preventing
damage to the laser amplifying medium resulting from non-linear processes at
high intensities. After amplification, these laser pulses are recompressed to deliver
103 –106 PW cm2 on target. Proposed techniques, including optical parametric
chirped pulse amplification, (OPCPA) (See Chap. 2 by Cerullo and Brida), [2, 3]
promise to continue to extend the boundaries of laser science in the foreseeable
future and powers of exawatts (1018 W) and zetawatts (1021 W) are being discussed.
High intensity laser radiation may now be applied in many traditional areas
of science usually reserved to nuclear accelerators and reactors. As the laser
intensity and associated electric field is increased, so does the electron quiver
energy, the energy a free electron has in the laser field – this is illustrated in
Fig. 10.1b. When laser radiation is focused onto solid and gaseous targets at
intensities >1018 W cm2 , electrons quiver with energies greater than their rest
mass (0.511 MeV) creating relativistic plasmas [4].
It is very useful to compare laser accelerators (Fig. 10.1a) with conventional
accelerators driven by alternating radio frequency electric fields (Fig. 10.1b) in the
so called Livingstone plot [6]. It should be noted that the electron beam energy of
conventional accelerators has increased by a factor of 10 every 6 years over the
last 70 years while the peak power of the laser has increased on average by an
order of magnitude every 3 years. Moreover there is a widely held view that the
Livingstone plot for conventional accelerators will flatten off above 1 TeV whereas
there is no apparent limitation for laser accelerators. It is perhaps a little naı̈ve to
compare the maximum energy of laser driven electron beams with conventional
accelerators since the existing properties of laser accelerators are still a long way
from being of interest to particle physicists. However, the present completion date
10 Applications of Ultra-Intense, Short Laser Pulses 229
Fig. 10.1 (a) Focussed laser intensity as a function of time [5]. The electron quiver energies are
indicated and as of 2006 have reached energies of 1 GeV. (b) So called Livingstone plot showing
how conventional accelerator energies have increased over the last 70 years. There is a widely held
view that the Livingstone plot will flatten off above 1 TeV – a fate not shared by laser accelerators
230 K.W.D. Ledingham
of the international linear collider is somewhere between 2020 and 2030 and there
is the potential for laser accelerators to play a part in the final design.
It is clear that the high power laser has emerged as a new tool for investigating
advanced physics under extreme conditions. Laser-plasma based accelerators have
the potential to deliver accelerating gradients more than 1,000 times higher than
conventional accelerator technology and on a table-top scale. When laser and
conventional accelerators are compared, one of the main differences between the
two technologies is the one of scale. This large increase in accelerating gradient for
laser technology is the key to reducing the size and associated cost over conventional
accelerators and in addition the shielding requirements are much reduced. Through
acceleration of a range of energetic particles and photons, intense laser-matter inter-
actions may induce nuclear and particle processes. This will be covered in Sect. 10.3
of this chapter. It should be emphasised that many of the applications that have been
demonstrated on very large lasers as a “proof of principle”, will soon be carried
out on the new generation of compact scale high intensity, high repetition rate lasers
being developed. Both types of lasers are necessary to advance this field of research.
Fig. 10.2 (a) A laser is injected into a plasma in a 3 cm capillary tube to generate energetic
electron beams which are analysed by a magnet. (b) A graph of an electron beam reaching energies
of 1 GeV (Reproduced and adapted with permission from Ref. [15])
234 K.W.D. Ledingham
Fig. 10.3 (a) A laser with intensity 3 1019 W cm2 is directed onto a proton rich dot deposited
on a titanium substrate. (b) The peaked proton distribution (black dots) and a particle in cell (PIC)
simulation (black curve). The inset shows the peaked proton spectrum expected at 1021 W cm2
(Reproduced and adapted with permission from Ref. [20])
pointed out that the resulting proton spectrum has a strong correlation with the
spatial distribution of the protons on the target surface, as shown in Fig. 10.3a.
A terawatt laser beam is focused on the front side of a metal target generating
10 Applications of Ultra-Intense, Short Laser Pulses 235
electrons which set up a Debye sheath on the back surface – as described earlier.
Applying a small hydrogen rich dot on the back surface enhances the proton yield in
the central part of the accelerating field which is nearly homogeneous. This produces
a quasi monoenergetic bunch of protons. Figure 10.3b indicates the mono-energetic
bunch and also the PIC simulation for the laser and target conditions used in the
experiment. The inset in Fig. 10.3b estimates the proton peak energy when the laser
intensity is increased to 1021 W cm2 .
A completely different approach was adopted by Toncian et al. [21]. As shown in
Fig. 10.4, two CPA lasers were used; CPA1 produced a broad spectrum of protons
from the target which were directed through a hollow metal cylinder irradiated by a
second CPA2 laser beam.
The second laser beam causes transient focusing electric fields to be set up
in the hollow cylinder which enabled a range of proton energies to be selected from
the broad spectrum generated by the first laser beam, depending on the timing of
the two lasers and on the flight time of the ions through the cylinder. This technique
addresses two of the current drawbacks of laser accelerated proton beams: namely
their broad energy spectrum and the source divergence. A similar outcome could
more easily be achieved using small quadrupole or solenoid magnets, an approach
that is being developed by several groups including the author of this chapter.
The contributions to cancer cure from surgery, radiation therapy and chemotherapy
are about 49, 40 and 11 % respectively and the role of X-ray radiotherapy is assured
for many years in the future. However it has been known for some time that some
236 K.W.D. Ledingham
Fig. 10.5 This graph shows the relative radiation doses as a function of depth of tissue given by
conventional radiation treatments as well as the Bragg peak for protons
patients (as many as 10 %) would be better treated with proton and ion therapy
rather than X-rays. Hadrontherapy is a form of radiotherapy that uses protons,
carbon and other heavy ions to irradiate cancerous tumours. The use of protons
and carbon ions in radiotherapy has several advantages to the more widely used
X-ray radiotherapy. Firstly, the proton beam scattering from the atomic electrons is
greatly reduced, and thus, there is less irradiation of healthy tissues in the vicinity
of the tumour. Secondly the well-known Bragg peak for ions, where the energy loss
dE/dx increases dramatically near the end of its path, can be used to good effect to
coincide with the location of the tumour causing the deposition of the largest fraction
of energy inside the tumour. The Bragg peak results in very little radiation being
deposited after the tumour (spatially) and the oxygen enhancement ratio (OER), as
well as the radiation toxicity, are also reduced compared to X-ray radiotherapy. This
is precisely what is required when passing through healthy tissue. This is shown in
Fig. 10.5 below.
In recent years a number of papers discussing the potential for laser hadron
therapy have been published e.g. [23–27]. These authors point out that proton
beams with the required parameters can already be obtained using conventional
accelerators of charged particles (synchrotrons, cyclotrons and linacs). There are
a number of conventional accelerators being built for hadrontherapy in Europe,
America and Japan and two in particular at Heidelberg and in Italy will treat
several thousand patients per annum. Alternatively the potential use of a laser-
based accelerator is attractive for three reasons: it is compact, requires less radiation
10 Applications of Ultra-Intense, Short Laser Pulses 237
In addition, the building for a conventional cyclotron system would cost about
$40 million with no large costs for housing a laser accelerator facility.
Although we are still a long way from generating laser protons for hadron
therapy, a number of groups are experimenting with low energy laser proton sources
up to a few MeV for biological studies e.g. the demonstration of DNA double-strand
breaking in cancer cells was recently demonstrated [30].
238 K.W.D. Ledingham
Laser driven ion, neutron and photon beams can produce radioactive isotopes similar
to reactors and cyclotrons, but the arguments for pursuing this line are similar to
those described in Sect. 10.3.1 – namely a laser accelerator is potentially more
compact, requires less radiation shielding and perhaps less expensive to construct.
Positron Emission Tomography (PET) is a powerful medical diagnostic/imaging
technique requiring the production of short-lived (2 min ! 2 h) positron emit-
ting isotopes. The PET process involves the patient receiving an injection of a
pharmaceutical, labelled with a short-lived “C emitting source which collects in
‘active’ areas of the body such as tumours. The principal tracers used in the PET
technique are 11 C, 13 N, 15 O and 18 F. Many chemical compounds can be labelled
with positron emitting isotopes and their bio-distribution can be determined by
PET imaging as a function of time. The most commonly used radiopharmaceutical
is 2-fluoro-2-deoxyglucose, 2-[18 F]FDG. Over the last few years, the value of
PET-FDG in the management of cancer patients has been widely demonstrated.
Figure 10.6 highlights the success rate of PET in diagnosing lung cancer compared
with conventional X-ray computed tomography (CT) scanning.
One of the main factors limiting the wider use of FDG-PET imaging is the
requirement for expensive infrastructure at the heart of which lies the cyclotron and
the associated extensive radiation shielding. As was described earlier, a simplified
approach to isotope production would be to develop a miniaturised, on site resource
with capabilities similar to that of a cyclotron. Recent results show when an intense
laser beam (I >1019 W cm2 ) interacts with solid targets, beams of MeV protons
capable of producing PET isotopes are generated. A number of recent reports have
been published using the high power laser VULCAN both as a 100 terawatt laser and
a petawatt laser [31, 32]. In addition some preliminary work has also been carried
out at the 10 Hz table–top “Salle Jaune” laser at the Laboratory of Applied Optics
LOA [33], which has been calculating the strength of PET isotopes which could be
10 Applications of Ultra-Intense, Short Laser Pulses 239
Fig. 10.8 The black dots are data obtained from a plastic target and the open dots from an
aluminium target. The curve represents a two dimensional PIC simulation (Reproduced and
adapted with permission from Ref. [33])
lasers at roughly the same laser intensity. For a pulse of about 300 J, the 11 C activity
per shot on VULCAN was a few 106 Bq. At LOA, for a pulse of typically 1 J, the
activity was about 10 Bq. The ratio of the activities is about 105 , with the ratio of
the energies being 300. Thus at roughly the same laser intensity, the activity scales
approximately with the square of the pulse energy.
One of the major problems of the nuclear power industry today is the management
and disposal of high-level radioactive waste with a long half-life. Vitrified high-level
waste can be stored for about 50 years before ultimate geological disposal. However,
in many countries around the world much research effort is being expended in
the possibility of partitioning and transmuting radioactive waste, which can reduce
its toxicity by a factor of 100 e.g. [34]. Recently a number of roadmaps for
nuclear waste transmutation using GeV energy light particle beams have been
produced [35].
Although the most frequently discussed methods involve transmutation by bom-
bardment with neutrons from a reactor, there have been other suggested approaches
to this problem, for example, laser-driven high-brightness gamma generation for
photo-transmutation. It is possible to generate monochromatic gamma rays by back
scattering laser light from conventional GeV linacs or indeed laser driven GeV
electron beams (to be described later). The energy of the photons should coincide
10 Applications of Ultra-Intense, Short Laser Pulses 241
with the maximum of the giant dipole resonance cross section for (”,n) reactions.
Initial studies have been encouraging and the experiments using the nuclear waste
isotopes 129 I and 137 Cs will be carried out in the future [36].
Transmutation studies have also been carried out at two high power laser centres,
the large single pulse laser centre at the Rutherford Laboratory and compact high
repetition rate laser centre at the University of Jena. Firstly, the experiment carried
out on VULCAN at the Rutherford Appleton Laboratory will be discussed [37]. This
involved using the petawatt laser to drive the photo-transmutation of long-lived 129 I
with a half-life of 15.7 million years to 128 I with a half-life of 25 min. A laser-
generated gamma ray is absorbed by the 129 I nucleus, which releases a neutron to
undergo transmutation to 128 I. This giant dipole resonance (”,n) reaction is observed
in 129 I and 127 I, and the absolute activities of the reaction products 128 I and 126 I were
used for the first time to determine the integrated cross-section for 129 I(”,n)128 I from
the known 127 I(”,n)126 I cross-section. The laser produced relativistic electrons then
interacted with a 4 mm thick Au target to produce high energy gamma rays. These
photons irradiated iodine samples of 85 % 129 I and 15 % 127 I as shown in Fig. 10.9.
The iodine samples were analysed using germanium detectors before and after
laser irradiation of the gold target. The background spectrum in Fig. 10.10 was
measured for 3 h and the principal lines observed resulting from the decay include
125
Sb. The background activity from the samples was of the order of 2.4 108 Bq.
Figure 10.10 shows clear peaks at 443.3 and 527.1 keV, characteristic of the decay
of 128 I. The decay rate of 128 I was determined by measuring the integrated area
under the peaks over successive time intervals. The half-lives of the 443.3 and
527.1 keV peaks were 25.8 ˙ 1.0 and 25.5 ˙ 1.5 min respectively, which agree well
with the literature value of 25.0 min. In addition, the measured ratio of the area
under the 443.3 and 527.1 keV peaks averaged over all of the measured spectra is
10.9, which is in good agreement with the literature value of 10.4. From the data,
the total activity of 128 I produced at the time of the laser shot was deduced to be
1,323 Bq. This corresponds to the production of 2.9 106 nuclei of 128 I for the
360 J laser shot.
242 K.W.D. Ledingham
600
128
500 I
443.3keV
400
Counts
300
200 125
Sb 128
125
428keV Sb I
100 463keV 527.1keV
128
I
0
Background
-100
400 420 440 460 480 500 520 540 560 580
Energy(keV)
Fig. 10.10 Gamma emission spectra from one of the iodine samples measured before (back-
ground) and after laser irradiation of the gold target. Characteristic emission lines of 128 I at 443.3
and 527.1 keV are clearly observed
A transmutation experiment was also carried out at the compact high repetition
laser JETI at the University of Jena using the same iodine sources [38]. The Jena
multi TW laser system generates pulses with an energy of 1 J, <80 fs pulse width,
at a repetition rate of 10 Hz with a centre wavelength of 800 nm. The laser pulses
facilitate the intensities of up to 1020 W cm2 within a focal area of 5 m2 . This
is directed onto a helium gas jet to produce an electron beam and then directed at a
high Z target to produce a bremsstrahlung beam as shown in Fig. 10.11. The nuclear
reaction rate in 129 I was increased by two orders of magnitude by striking a helium
gas jet in front of the thick high Z bremsstrahlung target, compared against striking
the high Z target directly with the laser beam [38].
10 Applications of Ultra-Intense, Short Laser Pulses 243
In this section we shall develop laser acceleration a stage further and investigate
what happens when conventional RF accelerators or laser driven electron accel-
erators are coupled with high power lasers. This is likely to be a very important
development for ELI- Nuclear Physics.
The interaction of optical photon and electron beams is a process called Thomson
scattering when the photons have energies much smaller than the rest energy of the
electron (mo c2 ), or Compton scattering if the photon’s energy is mo c2 .
Figure 10.12 illustrates how backscattered photons are produced. To get the
maximum energy of scattered X-rays, the electron and photon beams are directed
towards each other head on. The photons scatter off the electrons at 180ı with energy
(! max ) given by the following simple equation if the laser beam is only moderately
intense – in other words linear Thomson scattering:
ı
!max D !laser 4 2 1 C 2 2 (10.3)
where is (Ee /mc2 C 1) with Ee being the electron energy, ! laser is the laser photon
energy, is the usually small angle between the electron beam direction and the
angle of observation. The photons are thus backscattered and are Doppler shifted
into the X-ray or gamma-ray part of the spectrum.
At the FZD facility in Dresden there exists all the instrumentation for carrying
out this type of experimentation. The linac Elbe at 40 MeV, is 40/0.511 C 1 D 79.
Elaser is 800 nm D 1.6 eV and hence the back scattered photons have energies
38 keV. The arrangement of the electron and photon beams interacting in the
newly constructed high vacuum target chamber in the Elbe beam vault is shown
in Fig. 10.13.
If the laser beam is sufficiently intense then a more complicated nonlinear process
takes place described by the equation ! sc D n! laser 2 2 (1cos')/(1 C ao 2 /2 C 2 2 ),
where n is the number of laser photon absorbed by the electron. The term ao is a
dimensionless quantity indicating the strength of the laser intensity. n can take
244 K.W.D. Ledingham
Fig. 10.12 A typical arrangement for detecting back scattered X-rays. The electron and photon
beams interact (180ı ) as shown and the X-rays continue through a pinhole in a focusing
parabola and are detected by a germanium detector (Reproduced and adapted with permission
from Ref. [39])
Fig. 10.13 Photograph showing how the Elbe electron beam couples to the photons from the
DRACO laser in the backscattering chamber
any integral value to produce a series of harmonics, although the ao term in the
denominator reduces the maximum X-ray energy (red-shift) for n D 1. This is
shown in Fig 10.14.
The most important application for backscattered monochromatic x-rays in the
range of 20–50 keV is in phase contrast imaging particularly for medical purposes.
10 Applications of Ultra-Intense, Short Laser Pulses 245
The Thomson back scattered photons scale as the square of the electron beam
energy and so if we have GeV electrons with a 2 value of about 4 106 , the
energy of the back scattered photons is about 20 MeV. Thus we are getting into the
MeV monochromatic photon regime where applications to nuclear physics become
important. These GeV electrons can be produced by a laser wakefield accelerator.
ELI at present is the only funded European and International Centre for high-
level research on ultra-high intensity laser, laser-matter interaction and secondary
sources with unparalleled possibilities. Its pulse peak power and ultra-short pulses
will go beyond the current state-of-the-art by several orders of magnitude. Because
of its unique properties, this multidisciplinary facility will provide outstanding
new opportunities to study the fundamental processes unfolded during light-matter
interaction.
The ELI project, which is being co-ordinated by Gerard Mourou, is a collabora-
tion between 13 European countries and will consist of four pillars:
1. High Energy Beam Science devoted to the development and usage of dedicated
beam lines with ultra-short pulses of high energy radiation and particles reaching
almost the speed of light (100 GeV). This part of ELI will be realized in Prague
(Czech Republic).
2. Attosecond Laser Science designed to conduct temporal investigation of elec-
tron dynamics in atoms, molecules, plasmas and solids on the attosecond scale
(10 18 s). Szeged (Hungary) will host the short pulse pillar of ELI.
3. Laser-based Nuclear Physics: The third site in Magurele (near Bucharest/
Romania) will focus on laser-based nuclear physics. While atomic processes
are well suited to the visible or near visible laser radiation, as a third pillar
246 K.W.D. Ledingham
ELI-NP will generate radiation and particle beams with much higher energies
and brilliances suited to studies of nuclear and fundamental processes.
4. Ultra High Field Science will explore relativistic laser-matter interaction in an
energy range where totally new phenomena, like radiation dominated interac-
tions, become dominant. The decision on the location of the technologically most
challenging pillar will be taken in 2012 after validation of the technology.
The ELI-NP pillar will be the only laser facility in the world that will be solely
dedicated to the study of laser driven nuclear processes both at a fundamental level
but also with applications in mind. For ELI-NP two new central instruments are
planned:
1. A very high intensity laser, where three 10 PW Apollon lasers are coher-
ently added to yield a high intensity of 1,024 W/cm2 or electrical fields of
2 1,015 V/m. This is close to four orders more powerful than any existing laser.
2. A very intense, brilliant ” beam (E” 19 MeV), which is obtained by incoherent
Compton back scattering of direct laser light or laser light from a super cavity off
a very brilliant, intense, classical electron beam of GeV energy or laser driven
electron beam.
The ” beam will have unique properties in world-wide comparison and opens
new possibilities for high resolution spectroscopy at higher nuclear excitation
energies. They will lead to a better understanding of nuclear structure at higher
excitation energies with many doorway states, their damping widths, and chaotic
behaviour, but also new fluctuating properties in the time and energy domain. The
detailed investigation of the pygmy dipole resonance above and below the particle
threshold is very essential for nucleosynthesis in astrophysics. In ion acceleration
the high power laser permits the production of 1015 times higher density ion beams
than is achievable with classical RF accelerators. This will permit for the first time,
a possible fission-fusion reaction mechanism that can then be used to produce very
neutron-rich heavy nuclei. These nuclei will permit the investigation of the N D 126
waiting point of the r-process in nucleosynthesis. With this type of new laser acceler-
ation mechanism, very significant contributions to one of the fundamental problems
of astrophysics, the production of the heavy elements beyond iron in the universe,
can be addressed. According to a recent report by the National Research Council
of the National Academy of Science (USA), the origin of the heaviest elements
remains one of the 11 greatest unanswered questions of modern physics [41].
The ” beam also opens many new possibilities for applications. The ” beam
itself can be used to map the isotope distributions of radioactive waste remotely
via Nuclear Resonance Fluorescence (NRF) of radioactive waste [42]. At lower
energies around 100 keV the high resolution of the beam is very important for
protein structural analysis. In addition we want to produce low energy, brilliant,
10 Applications of Ultra-Intense, Short Laser Pulses 247
intense neutron beams and low energy, brilliant, intense positron beams, which open
new fields in material science and life sciences. The possibility of studying the same
target with these very different brilliant beams will be unique and advance science
much faster.
The high power laser allows for intensities of up to 1024 W cm2 . Here very
interesting synergies are achievable with the ” beam and the brilliant high energy
electron beam, which will allow the study of new fundamental processes in high
field QED. When the ” beam is injected into the focus of the high intensity laser,
which in this special case consists of a standing electric-field of two focused lasers,
the most recent non-perturbative QED calculations predict that one can observe
already, at 1024 W cm2 , the catalytic pair creation from the vacuum [43]. If
confirmed, this would constitute a very basic non-perturbative textbook QED result.
Also, the radiation damping theory could be tested very accurately with the very
brilliant electron beam injected into the laser focus. Here reflected high-energy ”
quanta and cascades of e C e pairs could be studied as a function of the ” factor
of the electron beam and the laser field strength. While the predictions for radiation
damping will probably be correct in the perturbative regime, the different theoretical
approximations can be tested very sensitively for nonlinear radiation damping. On
the other hand, the central questions of high-field QED for the ultra-relativistic laser-
plasma interaction with I > 1024 W cm2 will be the subject covered by the fourth
pillar of ELI.
10.5 Conclusions
References
25. K.W.D. Ledingham, W. Galster, R. Sauerbrey, Laser-driven proton oncology — a unique new
cancer therapy? Br. J. Radiol. 80, 855–858 (2007)
26. A.L. Boyer, M. Goitein, A.J. Lomax, E.S. Pedroni, Radiation in the treatment of cancer. Phys.
Today 55(34) (2002)
27. K.W.D. Ledingham, W. Galster, Laser-driven particle and photon beams and some applications.
New J. Phys. 11, 045005 (2010)
28. U. Linz, J. Alonso, What will it take for laser driven proton accelerators to be applied to tumor
therapy? Phys. Rev. Spec. Top.-Accel. Beams 10, 094801 (2007)
29. C.-M. Ma, Private communication (2010)
30. A. Yogo, K. Sato, M. Nishikino, M. Mori, T. Teshima, H. Numasaki, M. Murakami, Y. Demizu,
S. Akagi, S. Nagayama, K. Ogura, A. Sagisaka, S. Orimo, M. Nishiuchi, A.S. Pirozhkov, M.
Ikegami, M. Tampo, H. Sakaki, M. Suzuki, I. Daito, Y. Oishi, H. Sugiyama, H. Kiriyama, H.
Okada, S. Kanazawa, S. Kondo, T. Shimomura, Y. Nakai, M. Tanoue, H. Sasao, D. Wakai,
P.R. Bolton, H. Daido, Application of laser-accelerated protons to the demonstration of DNA
double-strand breaks in human cancer cells. Appl. Phys. Lett. 94, 181502 (2009)
31. I. Spencer, K.W.D. Ledingham, R.P. Singhal, T. McCanny, P. McKenna, E.L. Clark, K.
Krushelnick, M. Zepf, F.N. Beg, M. Tatarakis, A.E. Dangor, P.A. Norreys, R.J. Clarke, R.M.
Allott, I.N. Ross, Laser generation of proton beams for the production of short-lived positron
emitting radioisotopes. Nucl. Instrum. Meth. Phys. Res. B 183, 449–458 (2001)
32. K.W.D. Ledingham, P. McKenna, T. McCanny, S. Shimizu, J.M. Yang, L. Robson, J. Zweit,
J.M. Gillies, J. Bailey, G.N. Chimon, R.J. Clarke, D. Neely, P.A. Norreys, J.L. Collier, R.P.
Singhal, M.S. Wei, S.P.D. Mangles, P. Nilson, K. Krushelnick, M. Zepf, High power laser
production of short-lived isotopes for positron emission tomography. J. Phys. D: Appl. Phys.
37, 2341–2345 (2004)
33. S. Fritzler, V. Malka, G. Grillon, J.P. Rousseau, F. Burgy, E. Lefebvre, E. d’Humières, P.
McKenna, K.W.D. Ledingham, Proton beams generated with high-intensity lasers: applications
to medical isotope production. Appl. Phys. Lett. 83, 3039 (2003)
34. J. Magill, V. Berthou, D. Haas, J. Galy, R. Schenkel, H.-W. Wiese, G. Heusener, J. Tommasi,
G. Youinou, Impact limits of partitioning and transmutation scenarios on the radiotoxicity of
actinides in radioactive waste. Nucl. Energy 42, 263–277 (2003)
35. C.A. Rubbia, European roadmap for developing accelerator driven systems for nuclear waste
incineration, ENEA Report (ISBN 88-8286-008-6) (2001)
36. D. Li, K. Imasaki, S. Miyamoto, S. Amano, T. Mochizuki, Experiment on photonuclear
reaction induced by laser compton scattering gamma-ray. J. Nucl. Sci. Technol. 42, 259–261
(2005)
37. K.W.D. Ledingham, J. Magill, P. McKenna, J. Yang, J. Galy, R. Schenkel, J. Rebizant, T.
McCanny, S. Shimizu, L. Robson, R.P. Singhal, M.S. Wei, S.P.D. Mangles, P. Nilson, K.
Krushelnick, R.J. Clarke, P.A. Norreys, Laser-driven photo-transmutation of I-129 – a long-
lived nuclear waste product. J. Phys. D: Appl. Phys. 36, L79–L82 (2003)
38. B. Liesfeld, K.-U. Amthor, F. Ewald, H. Schwoerer, J. Magill, J. Galy, G. Lander, R. Sauerbrey,
Nuclear reactions triggered by laser-accelerated relativistic electron jets. Appl. Phys. B: Lasers
Opt. 79, 1047–1052 (2004)
39. S.G. Anderson, C.P.J. Barty, S.M. Betts, W.J. Brown, J.K. Crane, R.R. Cross, D.N. Fittinghoff,
D.J. Gibson, F.V. Hartemann, J. Kuba, G.P. LeSage, J.B. Rosenzweig, D.R. Slaughter, P.T.
Springer, A.M. Tremaine, Short-pulse, high-brightness X-ray production with the PLEIADES
Thomson-scattering source. Appl. Phys. B: Lasers Opt. 78, 891–894 (2004)
40. D. Seipt, Private communication (2009)
41. E. Haseltine, http://discovermagazine.com/2002/feb/cover
42. R. Hajima, N. Kikuzawa, N. Nishimori, T. Hayakawa, T. Shizuma, K. Kawase, M. Kando, E.
Minehara, H. Toyokawa, H. Ohgaki, Detection of radioactive isotopes by using laser Compton
scattered gamma-ray beams. Nucl. Instrum. Meth. A 608, S57–S61 (2009)
43. G.V. Dunne, H. Geis, R. Schutzhold, Catalysis of Schwinger vacuum pair production. Phys.
Rev. D 80, 111301 (2009)
Chapter 11
Utilising Ultrafast Lasers for Multiphoton
Biomedical Imaging
11.1 Introduction
E.E. Hoover () • E.V. Chandler • J.J. Field • D.N. Vitek • M.D. Young • J.A. Squier ()
Center for Microintegrated Optics for Advanced Bioimaging and Control, Department of Physics,
Colorado School of Mines, Golden, CO 80401, USA
e-mail: ehoover@mines.edu; jsquier@mines.edu
The basic multiphoton imaging system is quite straight forward and essentially
consists of four components: (1) the laser source, (2) the scanning system, (3) the
microscope platform, and (4) the detection system. Starting with the first com-
ponent, the laser, the most common ultrafast source for this application is the
Ti:sapphire laser, owing to its broad wavelength range (700–1,100 nm) and high
average power capability (>1 W). The tuning range is particularly desirable from
the point-of-view of being able to selectively excite different fluorophores when
performing two-photon excitation fluorescence (TPEF) imaging. In addition, the
average power capability can be usefully exploited in several ways. Given the
average repetition rates of these lasers, which is typically on the order of tens of
megahertz (MHz), the per pulse energy translates into tens of nanojoules (nJ). This
energy makes it possible to image deeply (up to about a millimetre) into scattering
specimens, or quickly (30 frames per second and higher) – the energy can be
distributed over multiple focal points decreasing the scan time needed to complete
a field-of-view.
In terms of scanning (the second component), multiple strategies are used –
the exact choice being dependent upon the system being studied. The beam can
be rastered across the specimen (beam scanning); the specimen can be scanned
about the excitation beam (specimen scanning); or, in the case of non-scattering
specimens, a wide-field approach can be employed – which eliminates or requires
minimal beam scanning. Sometimes several of these approaches are combined
together to create the most optimal exposure conditions possible for exploring and
studying the biological system under consideration. Basic beam scanning for single
and multiple beam geometries is discussed in Sect. 11.3, while a wide-field approach
using spatio-temporal focusing is detailed in Sect. 11.6.1.
The scanning system couples the ultrafast laser source to the microscope
platform – the third component. The platform can be a commercially available
microscope or a home-built unit. It provides delivery of the excitation light to the
specimen and simultaneously collects and routes the signal photons to the detection
system. Often a pathway for separately viewing the specimen with a white light
source either through an eyepiece, a camera, or both is incorporated. In the next
section, we will explore the characteristics of the objective – the most important
component in the microscope platform.
The fourth and final component of the multiphoton imaging system is the
detector, which is used to capture precious signal photons with the highest possible
11 Utilising Ultrafast Lasers for Multiphoton Biomedical Imaging 253
Fig. 11.1 A 1.2 NA, 40X water immersion objective with correction collar (left) for use with a
cover slip, and a 0.80 NA, 40X water immersion objective that requires no cover slip (right)
A good starting point in the design of a multiphoton imaging system is the objective
lens, and as such it is important to be able to read and understand the characteristics
of the optic.
Figure 11.1 depicts two different objectives for use under different imaging
conditions. First, note the band about the barrel of both objectives. The colour
denotes the rough magnification, the exact magnification being imprinted directly on
the side of the objective. Different coloured bands denote different magnifications
and serve as a simple way of quickly identifying the degree of magnification. The
objective on the right has an additional white band denoting that this objective can
be used in water, and in this case, without a cover slip. The marking located on the
side of the barrel is used to indicate that no cover slip is necessary. The ‘0’ indicates
that the objective requires a zero thickness cover slip, while the infinity sign denotes
that this is an infinity-corrected objective. In other words, the lens will produce the
254 E.E. Hoover et al.
least abberated image when the image is at infinity. A second lens, known as a tube
lens, must therefore be used in conjunction with an infinity-corrected lens to create
a real image.
The objective on the left has a similar marking on the side of the barrel:
1 D 0.14 0.18. Again, the infinity sign denotes it is an infinity-corrected objec-
tive, but in this case, a cover slip must be used in conjunction with the objective.
For this objective the cover slip size can range from 0.14 to 0.18 m in thickness.
The grooved correction collar must be adjusted accordingly to match the desired
thickness. Finally, the numerical aperture (NA) is always recorded on the side of the
objective. For the objective shown on the left, the NA is 1.2, and for the objective
on the right the NA is 0.80. The NA describes how tightly the excitation beam
will be focused and is key to determining the resolution that can be achieved with
the objective – the higher the NA, the better the resolution. Specifically, the NA is
defined as:
NA D n sin .˛/
where n is the index of refraction of the medium in which the beam is focused, and
˛ is the half angle of the cone of light (semi-aperture angle) produced at focus.
As the contrast mechanisms in multiphoton microscopy all scale with the excitation
intensity to the power of two or greater, it is important to ensure a short pulse is
created at the focus of the objective to obtain efficient high-resolution imaging.
Optimally, at focus the pulse should be diffraction-limited in space and transform-
limited in time in order to produce the highest intensity. Complications often occur
with refractive optics which prevent either one or both of these conditions from
being satisfied. For example, multi-element objectives (and the scan optics) are
sufficiently long, from an optical path length point-of-view, that there is significant
dispersion. This dispersion causes the pulse to be stretched in time (“chirped”) and
become no longer transform-limited. This dispersion lowers the excitation intensity
and hence also the efficiency at which a particular nonlinearity is driven. There
are multiple strategies for addressing dispersion, all of which involve simply “pre-
chirping” the pulse. When pre-chirping, the pulse is pre-stretched with the opposite
sign of chirp compared to what is produced by the imaging system. The result of
this is that at focus, the system dispersion is effectively cancelled and a transform-
limited pulse is produced. Strategies for effective dispersion compensation and
measuring the pulse-width at the focal plane of the objective are well documented
[4, 6] and will not be covered any further within the context of this chapter.
They are however an important consideration for the end-user of any multiphoton
imaging system, and as such should not be neglected in the final system design and
implementation.
11 Utilising Ultrafast Lasers for Multiphoton Biomedical Imaging 255
The key principle behind the construction of a scan system for a laser scanning
microscope is the concept of the telecentric plane. A telecentric plane occurs when
the back of the objective is image relayed in the beam path (any location where the
marginal rays and the chief ray are parallel for all scan parameters). It is relatively
easy to design such a telecentric scan system by working backwards (from the
objective toward the laser) with an off-axis beam. Anywhere where the chief ray of
this off-axis beam crosses the origin is where the telecentric planes exist. All of the
beam-modifying elements should be located at these positions in order to produce a
good scan system. The majority of scan systems utilise galvanometric scan mirrors,
due to their high optical throughput, but a variety of other technologies exist and
follow the same principle.
Figure 11.2 shows the two most common telecentric scanning systems, though
many other valid configurations exist. While the majority of scanning systems put
the true telecentric plane between the scan mirrors, such a system does not provide
for the best performance. In a scan system where the telecentric plane is located
between the two mirrors (Fig. 11.2a) a situation results where the centre of the
beam does not perfectly hit the centre of the mirror closest to the objective. This
problem is exacerbated when a large tilt is applied to the mirror further from the
objective. An ideal scan system places a telecentric plane at each mirror [7], relaying
the plane from one mirror to the other, but comes at the cost of two additional lenses
(Fig. 11.2b). This sacrifice is important for multifocal systems and systems with
axial beam scanning, since these techniques have beams that are not perfectly on-
axis going into the first scan mirror.
In a multifocal scan system, the multiple beams need to be overlapped on the back
of the objective, which results in laterally separated beams at the focus. To produce
such a system, it is necessary to image relay the back of the objective through the
scan mirror system, thereby placing a telecentric plane at each scan mirror. With this
configuration (Fig. 11.3) it is possible to produce overlapped beams at the back of
the objective by overlapping the beams on the first scan mirror. While it is possible
to build a multifocal scan system using a single telecentric plane placed between the
two scan mirrors, such a system will not scan exactly the same size region for each
beam and is consequently less desirable.
It is important to note that not all multifocal systems are combined with an
external scan system. A rotating disk microlens array [8, 9] is a good example of
such an alternative system. In such a system the microlenses are placed so that
256 E.E. Hoover et al.
rotation of the disk is designed to reach the different regions of the focal plane,
requiring fewer optics to produce the scan system. However, this type of scan
system does not produce a traditional square uniformly-illuminated scan region and
is therefore non-ideal for many imaging applications [9].
11 Utilising Ultrafast Lasers for Multiphoton Biomedical Imaging 257
Fig. 11.4 A remote focusing system for adjusting the axial focal plane position
Moving the axial position of the focal plane is another important issue to consider
when constructing a scan system. Traditional systems accomplish this task by either
moving the sample or the objective, which can be disruptive to the sample and
does not permit rapid scanning. These challenges can be overcome by utilising the
remote focusing technique [10], which relays the telecentric plane at the back of the
objective to a secondary objective focused onto a piezoelectric mirror (Fig. 11.4).
This technique permits the rapid movement of the axial position of the focal plane
by adjusting the divergence of the beam at the back of the objective through moving
the piezo mirror back and forth.
Since the beam divergence can vary drastically depending on the piezo position,
it is important when constructing such a scan system to take into account the
potentially large size of the beam at all of the optics between the two objectives.
Most optical systems are designed to operate with collimated beams over the
majority of the scan system, where a remote focusing system only has a collimated
beam when the piezo mirror is exactly at the focus of the matched objective. This
situation places significant restrictions on the distance between the different optical
elements – a system with too much space between elements will cause the beam to
be clipped for some piezo positions.
Since the 1960s, with the advent of computers, the task of lens design has greatly
shifted. A cursory examination of lens design books prior to the 1960s [11, 12] will
show the laborious process involved in calculating the aberrations introduced by
an optical system. Now, computers can quickly and accurately calculate the Seidel
coefficients and even other higher order terms. Despite the great advantages which
258 E.E. Hoover et al.
computers afford the design process, the design of a custom optical element or
system cannot simply be left to a computer. In the computer’s attempt to optimise
a set of variables on the target parameters, the design may settle into local minima
(of the merit function), which are not ideal and often non-physical.
It is still the role of the lens designer or optical engineer to find an appropriate
and promising initial optical configuration. In this section we will walk through the
design of a simple non-imaging optical component and demonstrate some of the
key general principles of optical design and also the advantage of custom optics for
specific applications [13]. For our design we wish to create a simple optical system
that will collect endogenously produced light from a sample and relay it to the active
area of a point detector, such as a PMT.
Optics literature, such as text books [14–17] and scholarly articles [18, 19], may
provide a good starting point that meets some of our design requirements [13].
One example of such a system is called the substage condenser, which is used to
relay light from a source in such a way that it uniformly illuminates the sample and
matches the cone of light defined by the objective’s NA. The simplest condenser
11 Utilising Ultrafast Lasers for Multiphoton Biomedical Imaging 259
systems involve no lenses and instead utilise only a concave mirror to illuminate
the sample plane. Such a system can fill objectives of small numerical aperture
(about 0.25).
Substage condensers may contain any number of lenses, but one of the simplest
demonstrated designs is the Abbe two-lens condenser (Fig. 11.5), which can
accommodate an objective with a numerical aperture up to 1.2 [15]. By reversing
the design of the Abbe condenser, so that it collects light generated at the specimen
rather than focusing light into the specimen, we have a good starting place for our
initial design. In some ways we may consider this configuration to be a two-lens
infinity-corrected objective. In order to focus this collimated light onto the detection
surface of a point detector, we require an additional singlet to function as a tube lens.
of our first optic. A solution to this would be using a meniscus lens. However, a
meniscus lens may not have a sufficiently short focal length to relay the light to our
next optic. Normally this is remedied by a series of meniscus lenses which slowly
bend the rays into a desired configuration, but we wish to avoid using more than
three lenses total. In this case we are going to use a bi-convex aspheric lens made
from B270 glass with a 12 mm focal length and an 18 mm diameter. This by no
means represents an exhaustive search of available lenses; however, such a setup
will be shown to be suitable for our design. Examining basic designs for objectives
also provides us with a better understanding and familiarity for what designs do and
do not work. Many of these considerations may be identified in Fig. 11.6. In this
figure an initial optic with a small radius of curvature is followed by a meniscus
lens in order to gradually bend the light onto a pair of achromats, providing good
performance across a broad range of colours [15].
If a lens is found to be ultimately unsuitable, then we return to this step and
try another option, such as favouring something with a smaller diameter but a
shorter focal length. It is important to understand which parameters are going to
be most important to meeting the target specifications and which can be adjusted
or sacrificed. Often the best way to gain a familiarity with what works is to try a
variety of lenses and options. Fortunately, computer aided design programs often
contain lens catalogues from typical providers and make it easy and straightforward
to substitute different lenses into a simulated design.
As an initial attempt at the design, we can use the paraxial ray tracing equations
for thin lenses first and then try thick lenses. It is worth mentioning that there
are obvious and significant problems with this approach. Since we are designing
a relatively small optic in terms of total length relative to optic diameter, the thin
lens approximation is poor. Due to our requirement for a very high NA, the paraxial
11 Utilising Ultrafast Lasers for Multiphoton Biomedical Imaging 261
Fig. 11.7 Paraxial designs for (a) A thin lens system and (b) A thick lens system
approximations for small angles (sin() ) are also quite poor [20]. Despite these
bad assumptions we can still proceed, since we simply wish to create a rough initial
configuration that can be entered into a computer design program. Once we have a
rough configuration, we can adjust by eye and then allow the software to create a
more optimised configuration.
After plugging the numbers, the thin lens calculations give us a very simple
design (Fig. 11.7a). Unfortunately, it is obvious that the thin lens approximation
does not give us a clear picture of how the system should be constructed. However,
the thick lens paraxial approximation gives us a much more reasonable starting
point, since we can start to see some physically realisable distances between
elements (Fig. 11.7b).
Starting from the estimates produced from the thick lens paraxial design, we can
now enter the design into a computer aided program. Once these parameters are
entered, we can ray trace the system utilising real lenses (Fig. 11.8).
11.4.2.4 Optimisation
Now that our initial design is entered into the software we can optimise further
by giving the software some variables to adjust. The lens spacings and total track
length are our only variables that we can attempt to optimise. Optimising on
these parameters, we can find a good solution (Fig. 11.9a) by requesting that the
software minimise the final spot size (Fig. 11.9b) and collimate the rays between
the condenser and tube lenses. Considering all of the approximations made in our
first order paraxial design, it is quite remarkable that this design implemented with
real lenses (Fig. 11.8) comes so close to our optimised design (Fig. 11.9a).
Before we finish we can attempt to push the limits of this design to see how
much of the light we can really capture. We can explore this limit by increasing
the numerical aperture until the point just before the software starts to lose rays
(Fig. 11.10a). It is also informative to take a look at water immersion of the objective
262 E.E. Hoover et al.
(Fig. 11.10b), which demonstrates an even better collimation than the air-immersion
objective.
Finally, as a last check, we can test our design to ensure that it can handle the
requisite wavelength range and field of view. Figure 11.11a shows that the rays still
hit our target for all desired wavelengths, even at the edge of our field of view,
where Fig. 11.11b demonstrates that we retain all the rays on the detector even at
this extremum.
In review, Table 11.1 shows how we met or exceeded all of our target specifications.
This design is not really a condenser or a traditional objective; rather it may be
considered a measurement objective, since it is just designed to collect light and
relay it to a PMT. Whatever the case may be, it is a good example of pragmatic
optics. Pragmatic optics allow us to sacrifice many of the qualities of expensive
optical components by realising what is important to our specific design and only
focusing on those attributes. Such an attitude can potentially translate into a less
expensive optic that offers a substantial gain in performance.
11 Utilising Ultrafast Lasers for Multiphoton Biomedical Imaging 263
Fig. 11.9 (a) Optimised computer design at 0.95 NA (b) Optimised spot size
264 E.E. Hoover et al.
Fig. 11.10 Computer aided design at 0.97 NA (a) Air immersion. (b) Water immersion
11 Utilising Ultrafast Lasers for Multiphoton Biomedical Imaging 265
Fig. 11.11 Final design with full spectrum and at extreme point on field of view. (a) Worst case
layout. (b) Spot diagram for worst case layout
266 E.E. Hoover et al.
samples. As such, a large body of research has been accumulated in the past decade
aimed at reducing rates of photobleaching under TPLSM. While some studies aim
to improve the photostability of certain fluorophores, a more global approach is to
manipulate the excitation source in the hope that any improvement in photobleach-
ing rates will apply to many fluorescent tags. Photostability in TPLSM has been
studied by manipulation of the femtosecond pulse trains in terms of their energy
[24–26], repetition rate [25, 27], and shape [26, 28, 29]. In order to understand what
parameters of a femtosecond pulse train are relevant to photobleaching rates and
fluorescence yield, it is useful to examine the mathematical form of TPEF.
R1
˝ 2 ˛ dt f I02 .t/
I0 .t/ 1 gp
g D 2 D ; (11.2)
hI0 .t/i 2
R1 f
dt f I0 .t/
1
where f is the repetition rate of the laser oscillator, I0 (t) is the temporal intensity
envelope, is the full-width at half-maximum (FWHM) of the temporal intensity
envelope, and gp is the shape factor, a unitless quantity that is only dependent upon
the shape of the laser pulse. Equation (11.2) clearly shows that the temporal intensity
envelope is of critical importance to the TPEF signal through the so-called gain
factor, g, also referred to as the TPEF efficiency [32].
To visualise the effect that g has on the strength of TPA in a fluorophore, let
us assume a Gaussian-intensity pulse in the temporal domain. It is well known
that spectral dispersion acquired from linear optics (e.g. lenses, dichroics, and
objectives) causes distortions away from transform-limited (TL) pulse shapes in
the focal plane and can stretch the duration of the excitation pulse significantly. By
applying various amounts of spectral phase, one can calculate g for different orders
of spectral dispersion. Figure 11.13 shows how g varies as a function of bandwidth
as well as second-order dispersion, also called group-delay dispersion (GDD), for a
Gaussian pulse centred at 800 nm. As one can clearly see, the shorter the pulse is in
time, the better the TPA efficiency is and therefore the more fluorescence photons
one can measure.
It is interesting to compare g and gp as spectral dispersion is added to the
pulse. In this case, we will assume a constant bandwidth of 25 nm and a repetition
rate of 21.7 MHz, typical for an extended-cavity Ti:sapphire (Ti:Al2 O3 ) oscillator.
Figure 11.14 shows , g, and gp for a Gaussian pulse as GDD and third-order
dispersion (TOD) are varied. It is interesting to note that for both cases g is maximal
when the pulse is shortest, i.e. when GDD and TOD are zero, while gp is maximal
when GDD D 0 but not when TOD D 0. Since gp is only a measure of how good
the pulse shape is, ignoring temporal duration, one can conclude that a Gaussian
pulse with approximately 20,000 fs3 of TOD is better at exciting TPA than a TL
Gaussian pulse, but that the concomitant increase in temporal duration offsets this
improvement in shape.
The best shape factor is clearly unity, which is possible only when I0 2 (t) D I0 (t).
An example of such a pulse is the square pulse (Eq. 11.3):
11 Utilising Ultrafast Lasers for Multiphoton Biomedical Imaging 269
Fig. 11.13 Second order temporal coherence, or TPA efficiency, for a Gaussian pulse centred on
800 nm with varying amounts of GDD and spectral bandwidth. Note that the shortest pulse (largest
bandwidth and TL) produces the largest value for g, and that spectral dispersion of the excitation
pulse causes a significant reduction in TPEF efficiency
8
<0 t < =2
Isq .t/ D 1 =2 t =2 ; (11.3)
:
0 t > =2
implying that the greater the slope of the pulse, the better the shape factor will
be. Clearly, by minimising the duration of the square pulse to be infinitesimal, we
qualitatively arrive at a Dirac-delta function, ı(t), for the optimal pulse shape, a
pulse that is not realisable in practice due to its infinite peak intensity. Clearly then,
the closest one can get to realising a Dirac-delta pulse in practice is to produce a
TL pulse, which optimises the efficiency of TPEF. Put simply, a TL pulse is optimal
for TPA efficiency because the peak intensity of such a pulse is higher than that of
a stretched pulse of the same bandwidth. The ramifications of this scenario are that
the probability for the fluorophore to absorb two photons simultaneously is higher
for the TL pulse, boosting the TPEF signal.
A more qualitative picture can be used to understand this principle. Intensity is
typically measured in units of W/cm2 . However, using the expression for the energy
of a photon, E D „!, where „ is Planck’s constant divided by 2, and ! is the
angular frequency of the photon, one finds the result in Eq. (11.4) [3]:
Fig. 11.14 Variation in g & gp for a Gaussian pulse as a function of TOD and GDD. The effect of
GDD is shown in (a) and (b), while TOD is shown in (c) and (d). Note that the noise in these
simulations is due to quantisation noise in the numerical determination of the temporal pulse
duration
consider two pulses that have the same spectrum, but differ in that one is TL and the
other is stretched, we know from Parseval’s theorem that both pulses must have
the same energy, and thus the same number of photons per pulse. What differs
between the pulses is the peak flux of these excitation photons. Because the TL
pulse has all fluorescence photons arriving in the shortest possible duration, the
photon flux is greater, and thus there is a higher probability that the fluorophore will
simultaneously absorb two photons as compared to the stretched pulse. This leads
to an increased fluorescence yield in a non-bleaching sample as the pulse duration
is decreased. Mathematically, we see that this qualitative argument is verified by
the inverse relationship between the average fluorescence and the pulse duration
(cf. Eqs. 11.1 and 11.2).
11 Utilising Ultrafast Lasers for Multiphoton Biomedical Imaging 271
overlap in space. This grating is imaged onto the sample such that all of the
frequency components converge at a focal plane and, by Fermat’s principle, the
original pulse duration is re-established. Using SSTF, Oron et al. monitored two
photon excitation fluorescence over an area of 100 m2 with 4.5 m FWHM
depth resolution [35]. In this experiment the lateral image resolution was 0.25 m
giving approximately 100,000 effective pixels per image. Thus, an equivalent image
obtained with a multifocal or scanning beam system would need to cover roughly
100,000 positions.
More complex focusing arrangements in SSTF may be achieved by shaping the
phase and amplitude of the pulses. For example, the position of the temporal focus
may be offset by adding second order phase or group velocity dispersion (GVD)
to the spatially chirped pulses [37–39]. This dispersion can be achieved by several
means – including a prism pair, a dispersive optical fibre, or a liquid crystal array
pulse shaper. Within a certain range, the pulses remain transform-limited in time,
and the temporal focal plane is linearly dependent on the GVD. Such an SSTF
system with an adjustable axial focal position has been demonstrated with a range
of over 200 m, or 17 times greater than the beam’s Rayleigh length of 11.5 m
[37]. A potential application of this capability is remote axial scanning in a fibre
endoscope. While GVD controls the axial focal position, it has also been shown that
the lateral position of the SSTF focus is dependent upon the pulse’s linear phase
[38]. In addition, the shape of the illumination pattern on the sample may be adjusted
by amplitude shaping at the grating [40]. For example, illuminating the grating
with a line-shaped beam produces a line focus on the sample and illumination with
an array of lines produces an array of line-shaped foci on the sample. Amplitude
patterning combined with phase shaping has also been employed in order to generate
multiple foci with independent axial and lateral focal positioning [38].
Photon counting has a long history in microscopy, evolving into a variety of different
methods with various advantages and disadvantages. Since photon counting requires
extremely fast electronics, early photon counting used analogue techniques, such as
the time to amplitude converter (TAC). A TAC is used in time-correlated single pho-
ton counting (TCSPC) to measure the time between the incident pulse and the arrival
of the emitted photon. Later, the gated photon counting technique was introduced,
permitting the digital differentiation of photons arriving at different times after the
incident sample pulse. This technique was then extended to produce the photon
counting histogram (PCH), which performs virtually the same measurement as
TCSPC by utilising digital techniques rather than a time to amplitude converter [47].
Due to the complexity involved in photon counting, an innovative technique for
“fixing” averaging the signal from the PMT was developed. By utilising a discrim-
inator and a pulse shaper on a traditional averaging system, the resulting pulses
are conditioned to have virtually identical width and amplitude [45]. Performing
averaging on this modified system is therefore not as detrimental. Since the number
of photons can be accurately extracted from the averaged value on such a system,
the appropriate fluorophore concentration can be inferred from the averaged value.
However, this technique does not permit for any of the photon timing information
to be extracted, limiting the technique for many applications.
Combining these techniques with synchronised scan mirrors (or other beam
deflecting techniques) opened the door to performing laser scanning microscopy
with a point detector. Early systems designed to perform these measurements
involved a great deal of custom electronics [43], but these systems were eventually
commercialised and turned into more manageable, but expensive, turn-key systems.
However, with the advent of field programmable gate array (FPGA) technology and
complex programmable logic devices (CPLD) these systems are becoming inexpen-
sive enough that laboratory researchers can build on their own. These programmable
logic device technologies also permit novel customisation of the photon counting
system, allowing for unique techniques, such as spectral photon counting [44] and
demultiplexing photon counts from a single detector corresponding to multiple input
beams [48].
Fig. 11.16 Geometric difference between imaging detection and point detection. (a) Imaging
detection. (b) Point detection. CCD charge-coupled device camera, D dichroic, SM scan mirror,
SL scan lens, TL tube lens, OBJ objective, PMT photomultiplier tube
photon counts, rather than collecting all of the precise measurements of the time
between the incident pulse and the emitted photons and post processing that data.
Another technique unique to photon counting is the ability to electronically
demultiplex the information from optically multiplexed beams. This differential
multiphoton microscopy technique works by utilising beams with a time separation
greater than the fluorescence lifetime of typical dyes (several nanoseconds) and
electronically distinguishing the beam corresponding to the emitted photons. This
technique has unique utility in that it permits the nearly simultaneous acquisition
of photons from different beams, allowing for measurements from multiple layers,
multiple polarisations, multiple lateral locations, or multiple beam shapes.
Photon counting is not very complicated, at least in principle. The general principle
of a photon counting system is to detect the edges of a PMT signal corresponding
to single photons and increment a counter every time such an edge is detected.
In practice, photon counting data signals are extremely rapid (up to a couple
hundred megahertz for multiphoton microscopes); and therefore require a great
deal of care to handle properly. A commonly overlooked issue with signals at these
frequencies is that every cable between the photon counting components operates as
an antenna. Due to this problem, it is necessary to use well shielded cable (at least
double shielded) to keep environmental signals from causing extraneous noise to be
interpreted as signal photons. One might suppose that this shielding is only critical
before the pulse amplifier; however, experience shows that due to the combination
of high frequencies and high amplitudes, it is necessary to shield all of the PMT and
photodiode signals prior to the photon counting electronics.
The general architecture of a photon counting system is shown in Fig. 11.17.
This figure shows both the components necessary to construct a basic photon
counting system and the extra components required for an advanced demultiplexing
system. Both systems require a good PMT designed for photon counting; however,
a Geiger-mode PMT is not ideal for such systems, as these PMTs have a long
dead time between pulses. A much better performing system consists of a non-
Geiger-mode photon counting PMT chained with a pulse amplifier and a high-speed
comparator (HSC). Such a PMT does not output enough current for the pulses to be
discriminated directly, hence the need for a pulse amplifier, but provides much better
pulse-pair resolution than a Geiger-mode PMT. In addition to the pulse amplifier,
a HSC is necessary to threshold the pulses and provide a clean logic signal to
the photon counting electronics. Together with these signal electronics, it is also
necessary to synchronise the photon counting system with the scan electronics
(usually scan mirrors), so that a two-dimensional image can be formed from the
intensity readings. It is not strictly necessary to integrate this step with the photon
counting system (this task can be performed by a computer), but properly integrating
this step onto the photon counter means there will be no dead-time between pixels.
278 E.E. Hoover et al.
Fig. 11.17 Photon counting system architecture. PMT photomultiplier tube, PD photodiode,
PA pulse amplifier, HSC high-speed comparator (discriminator), SMC scan mirror controller,
LN line clock signal, PXL pixel clock signal, PhC photon counter. Star (*): Only required for
demultiplexing photon counting systems
11.6.3.1 Spectroscopy
Fig. 11.18 Simplified energy level diagrams. (a) Indicates the Stokes shift, the difference in
energy between excitation and emission wavelengths for linear excitation. (b) Relaxation from
different vibrational levels within excited states contributes to broadening of the output spectrum
spectroscopy – the spectrum of an atomic gas lamp enables mapping specific energy
level transitions to given spectral lines. The same is true for any fluorophore – an
emission wavelength directly maps to a transition from an excited state to the ground
state or another, lower-energy excited state. Unfortunately, complications abound in
real systems, stemming from both the method of excitation and the physical system
of the sample.
Practical considerations for spectroscopy include the following points. In linear
excitation, there is a slight difference between the emission and excitation energies,
as shown in Fig. 11.18a, known as the Stokes shift. The proximity of the source
and emission wavelengths often limits observation of the emission spectrum at
shorter wavelengths. Multiphoton spectroscopy circumvents the issue by pushing
the excitation wavelength far from the emission wavelength, enabling capture of
the entire emission spectrum. In either case, filtering out the source wavelength is
critical, though certainly simpler in the multiphoton excitation method. Dichroics,
absorptive glass, or edge filters are the most common methods of separating source
light from excitation light.
The physical process of optical excitation and relaxation further complicate spec-
troscopy. Each energy level possesses sub-levels generated by different vibrational
and rotational modes, as in Fig. 11.18b. The small contributions to the transition
energy from these sub-levels broaden the emission wavelength well beyond the
near-infinitesimal width from atomic gas lamps. Broadening limits the efficacy of
spectroscopy in two ways. Nearby transitions can overlap, making it difficult or
impossible to discern which transition is responsible for the fluorescence. More
relevant to biological systems, broadening also increases the difficulty of detecting
small spectral shifts (shifts on the order of several nanometres). A fluorophore
will undergo a small spectral shift, if the nearby chemical environment (electron
density) changes. As an example, the fluorophore may be attached to a larger
molecule that has undergone a conformational shift, and the new electron densities
force a slight alteration in the excited state energy. An understanding of the
fluorescence shift often provides insight into the physical process responsible for the
280 E.E. Hoover et al.
11.6.3.2 Spectrometers
Fig. 11.19 A simplified grating spectrometer. The entrance slit is imaged to the detector by the
curved mirror. The grating and focusing mirror together determine the spectral resolution. Optional
grating turret is shown for grating substitution. In the case of a scanning spectrometer, the grating
rotates causing a different colour to be passed through the spectrometer’s exit slit to a non-imaging
detector
proportional to the wavelength, unlike with prisms. However, care must be taken
in selecting the proper grating. The blaze angle tilts the rulings with respect to
the surface normal, forcing more energy into a particular order (typically the
first order). Blazed gratings are most efficient when angled to satisfy the Littrow
condition, where both the incident and diffracted light are normal to the grooves.
Unfortunately, gratings exhibit strongly wavelength-dependent and polarisation-
dependent efficiencies.
If a wide wavelength range is to be observed, grating substitution may be
necessary to maintain sufficient throughput. This is often achieved with multiple
gratings on a rotating turret in commercial systems. Groove density determines
how much dispersion is applied; for example, a grating with 1,200 lines/mm will
introduce twice as much dispersion as a 600 lines/mm grating. Additional dispersion
improves spectral resolution for a set entrance slit width but also limits the spectral
range on an imaging detector. For a set number of pixels on an imaging detector,
twice the dispersion halves the wavelength range but doubles the resolution. Less
dispersive gratings improve spectral range but suffer from mode overlap, where
shorter wavelengths of higher orders can overlap with longer wavelengths of lesser
orders. As an example, 300 nm light from the second-order of a grating will overlap
with the 600 nm light from the first-order [50].
282 E.E. Hoover et al.
Further, one might consider dialling down the entrance slit width to maximise
spectral resolution. However, the smaller the entrance slit width, the fewer signal
photons reach the detector. This is a significant concern for two-photon induced
processes, where the cross-section is already quite small and photon production is
limited. Increasing the exposure time is the obvious solution, but it is not feasible
in mobile biological systems. High-gain and high quantum efficiency detectors,
such as PMTs for grating scanning spectrometers and electron-multiplying CCDs
(EMCCDs), help mitigate this problem.
Most spatial scanning techniques – whether the beam is rastered across the
sample, or the sample itself is translated – use a grating-based spectrometer to
obtain spectral images. A minor variant of the grating-based spectrometer involves
employing a multiple slit mask and deconvolving the resulting image to obtain
the spectrum [58]. The multi-slit approach improves throughput over single-slit
techniques. The spectral image itself must be assembled in post-processing from
individual spectra obtained from each spatial location.
11.7 Conclusion
Acknowledgments This work was funded by the National Institute of Biomedical Imaging and
Bioengineering, Grant EB-003832.
References
1. W. Denk, K. Svoboda, Photon upmanship: why multiphoton imaging is more than a gimmick.
Neuron 18, 351–357 (1997)
2. J. Mertz, Nonlinear microscopy: new techniques and applications. Curr. Opin. Neurobiol. 14,
610–616 (2004)
3. W.R. Zipfel, R.M. Williams, W.W. Webb, Nonlinear magic: multiphoton microscopy in the
biosciences. Nat. Biotechnol. 21, 1369–1377 (2003)
4. K.E. Sheetz, J. Squier, Ultrafast optics: imaging and manipulating biological systems. J. Appl.
Phys. 105, 051101 (2009)
284 E.E. Hoover et al.
5. P.S. Tsai, P. Blinder, B.J. Migliori, J. Neev, Y. Jin, J.A. Squier, D. Kleinfeld, Plasma-mediated
ablation: an optical tool for submicrometer surgery on neuronal and vascular systems. Curr.
Opin. Biotechnol. 20, 90–99 (2009)
6. R. Carriles, D.N. Schafer, K.E. Sheetz, J.J. Field, R. Cisek, V. Barzda, A.W. Sylvester, J.A.
Squier, Invited review article: imaging techniques for harmonic and multiphoton absorption
fluorescence microscopy. Rev. Sci. Instrum. 80, 081101 (2009)
7. E.H.K. Stelzer, Chapter 9: The intermediate optical system of laser-scanning confocal mi-
croscopes, in Handbook of Biological Confocal Microscopy, ed. by J.B. Pawley, 3rd edn.
(Springer, New York, 2006), pp. 207–220
8. J. Bewersdorf, R. Pick, S.W. Hell, Multifocal multiphoton microscopy. Opt. Lett. 23, 655–657
(1998)
9. M. Straub, S.W. Hell, Multifocal multiphoton microscopy: a fast and efficient tool for 3-D
fluorescence imaging. Bioimaging 6, 177–185 (1998)
10. E.J. Botcherby, R. Juskaitis, M.J. Booth, T. Wilson, Aberration-free optical refocusing in high
numerical aperture microscopy. Opt. Lett. 32, 2007–2009 (2007)
11. A.E. Conrady, Applied Optics and Optical Design (Dover Publications, New York, 1992)
12. J. Strong, Concepts of Classical Optics (Dover Publications, New York, 2004)
13. G. Smith, Practical Computer-Aided Lens Design, 1st English edn. (Willmann-Bell, Rich-
mond, 1998)
14. R. Kingslake, Optical System Design (Academic, New York, 1983)
15. L.C. Martin, Technical Optics, vol. 2, 2nd edn. (Sir Isaac Pitman & Sons, LTD, London, 1961)
16. J. Chaves, M. Collares-Pereira, Ideal concentrators with gaps. Appl. Opt. 41, 1267–1276
(2002)
17. R. Winston, Nonimaging Optics (Elsevier Academic Press, Burlington, 2005)
18. K.D. Sharma, Design of slide projector condenser: a new approach. Appl. Opt. 22, 3925 (1983)
19. L.V. Foster, Aspheric enlarging condenser. J. Opt. Soc. Am. 13, 631–634 (1926)
20. J. Geary, Introduction to Lens Design: With Practical ZEMAX Examples (Willmann-Bell,
Richmond, 2002)
21. M. Müller, Introduction to Confocal Fluorescence Microscopy, 2nd edn. (SPIE Press, Wash-
ington, DC, 2006)
22. D. Soumpasis, Theoretical analysis of fluorescence photobleaching recovery experiments.
Biophys. J. 41, 95–97 (1983)
23. E. Sánchez, L. Novotny, G. Holtom, X. Xie, Room-temperature fluorescence imaging and
spectroscopy of single molecules by two-photon excitation. J. Phys. Chem. A 101, 7019–7023
(1997)
24. G. Patterson, D. Piston, Photobleaching in two-photon excitation microscopy. Biophys. J. 78,
2159–2162 (2000)
25. N. Ji, J.C. Magee, E. Betzig, High-speed, low-photodamage nonlinear imaging using passive
pulse splitters. Nat. Methods 5, 197–202 (2008)
26. P. Xi, Y. Andegeko, L. Weisel, V. Lozovoy, M. Dantus, Greater signal, increased depth, and less
photobleaching in two-photon microscopy with 10fs pulses. Opt. Commun. 281, 1841–1849
(2008)
27. G. Donnert, C. Eggeling, S.W. Hell, Major signal increase in fluorescence microscopy through
dark-state relaxation. Nat. Methods 4, 81–86 (2007)
28. H. Kawano, Y. Nabekawa, A. Suda, Y. Oishi, H. Mizuno, A. Miyawaki, K. Midorikawa,
Attenuation of photobleaching in two-photon excitation fluorescence from green fluorescent
protein with shaped excitation pulses. Biochem. Biophys. Res. Commun. 311, 592–596 (2003)
29. J.J. Field, R. Carriles, K.E. Sheetz, E.V. Chandler, E.E. Hoover, S.E. Tillo, T.E. Hughes,
A.W. Sylvester, D. Kleinfeld, J.A. Squier, Optimizing the fluorescent yield in two-photon laser
scanning microscopy with dispersion compensation. Opt. Express 18, 13661–13672 (2010)
30. C. Xu, W.R. Zipfel, Handbook of Biomedical Nonlinear Optical Microscopy (Oxford Univer-
sity Press, New York, 2008)
31. C. Xu, W.W. Webb, Measurement of two-photon excitation cross sections of molecular
fluorophores with data from 690 to 1050 nm. J. Opt. Soc. Am. B. 13, 481–491 (1996)
11 Utilising Ultrafast Lasers for Multiphoton Biomedical Imaging 285
32. C. Bardeen, V. Yakovlev, J. Squier, K.R. Wilson, S. Carpenter, P. Weber, Effect of pulse shape
on the efficiency of multiphoton processes: implications for biological microscopy. J. Biomed.
Opt. 4, 362–367 (1999)
33. C. Eggeling, A. Volkmer, C.A.M. Seidel, Molecular photobleaching kinetics of rhodamine 6g
by one- and two-photon induced confocal fluorescence microscopy. Chem. Phys. Chem. 6,
791–804 (2005)
34. F. Helmchen, W. Denk, Deep tissue two-photon microscopy. Nat. Methods 2, 932–940 (2005)
35. D. Oron, E. Tal, Y. Silberberg, Scanningless depth-resolved microscopy. Opt. Express 13,
1468–1476 (2005)
36. G. Zhu, J.v. Howe, M. Durst, W. Zipfel, C. Xu, Simultaneous spatial and temporal focusing
of femtosecond pulses, in Conference on Lasers and Electro-Optics/Quantum Electronics and
Laser Science and Photonic Applications Systems Technologies, pp. CWP1. Optical Society of
America (2005)
37. M.E. Durst, G. Zhu, C. Xu, Simultaneous spatial and temporal focusing for axial scanning.
Opt. Express 14, 12243–12254 (2006)
38. M.A. Coughlan, M. Plewicki, R.J. Levis, Parametric spatio-temporal control of focusing laser
pulses. Opt. Express 17, 15808–15820 (2009)
39. H. Suchowski, D. Oron, Y. Silberberg, Generation of a dark nonlinear focus by spatio-temporal
coherent control. Opt. Commun. 264, 482–487 (2006)
40. E. Papagiakoumou, V.D. Sars, V. Emiliani, D. Oron, Temporal focusing with spatially
modulated excitation. Opt. Express 17, 5391–5401 (2009)
41. D.N. Vitek, D.E. Adams, A. Johnson, P.S. Tsai, S. Backus, C.G. Durfee, D. Kleinfeld, J.A.
Squier, Temporally focused femtosecond laser pulses for low numerical aperture microma-
chining through optically transparent materials. Opt. Express 18, 18086–18094 (2010)
42. F. He, H. Xu, Y. Cheng, J. Ni, H. Xiong, Z. Xu, K. Sugioka, K. Midorikawa, Fabrication of
microfluidic channels with a circular cross section using spatiotemporally focused femtosecond
laser pulses. Opt. Lett. 35, 1106–1108 (2010)
43. W.G. Wier, C.W. Balke, J.A. Michael, J.R.H. Mauban, A custom confocal and two-photon
digital laser scanning microscope. Am. J. Physiol. Heart Circ. Physiol. 278, H2150–2156
(2000)
44. C. Buehler, K.H. Kim, U. Greuter, N. Schlumpf, P.T.C. So, Single-photon counting multicolor
multiphoton fluorescence microscope. J. Fluoresc. 15, 41–51 (2005)
45. J.B. Pawley, Chapter 2: Fundamental limits in confocal microscopy, in Handbook of Biological
Confocal Microscopy, ed. by J.B. Pawley, 3rd edn. (Springer, New York, 2006), pp. 20–42
46. J.D. Driscoll, A.Y. Shih, S. Iyengar, J.J. Field, G.A. White, J.A. Squire, G. Cauwenberghs, D.
Kleinfeld, Photon counting, censor corrections, and lifetime imaging for improved detection in
two-photon microscopy. J. Neurophysiol. 105, 3106–3113 (2010)
47. W. Becker, Chapter 2: Overview of photon counting techniques, in Advanced Time-Correlated
Single Photon Counting Techniques (Springer, Berlin/Heidelberg, 2005), pp. 11–25
48. W. Amir, R. Carriles, E.E. Hoover, T.A. Planchon, C.G. Durfee, J.A. Squier, Simultaneous
imaging of multiple focal planes using a two-photon scanning microscope. Opt. Lett. 32,
1731–1733 (2007)
49. B.R. Masters, P.T.C. So, E. Gratton, Multiphoton excitation fluorescence microscopy and
spectroscopy of in vivo human skin. Biophys. J. 72, 2405–2412 (1997)
50. E. Hecht, Optics, 4th edn. (Addison Wesley, New York, 2002)
51. B.R. Masters, P.T.C. So, Handbook of Biomedical Nonlinear Optical Microscopy (Oxford
University Press, New York, 2008)
52. R.J. Bell, Introductory Fourier Transform Spectroscopy (Academic, London, 1972)
53. J.E. Chamberlain, The Principles of Interferometric Spectroscopy (Wiley, New York, 1979)
54. Y. Garini, M. Macville, S. Manoir, R.A. Buckwald, M. Lavi, N. Katzir, D. Wine, I. Bar-Am, E.
Schrock, D. Cabib, T. Ried, Spectral karyotyping. Bioimaging 4, 65–72 (1996)
55. H.R. Morris, C.C. Hoyt, P.J. Treado, Imaging spectrometers for fluorescence and raman
microscopy: acousto-optic and liquid crystal tunable filters. Appl. Spectrosc. 48, 857–866
(1994)
286 E.E. Hoover et al.
56. E.S. Wachman, W. Niu, D.L. Farkas, Imaging acousto-optic tunable filter with 0.35-micrometer
spatial resolution. Appl. Opt. 35, 5220–5226 (1996)
57. E.S. Wachman, W. Niu, D.L. Farkas, Aotf microscope for imaging with increased speed and
spectral versatility. Biophys. J. 73, 1215–1222 (1997)
58. M.J.E. Golay, Multi-slit spectroscopy. J. Opt. Soc. Am. 39, 437–444 (1949)
Chapter 12
Femtosecond Laser Micromachining
12.1 Introduction
L. Cerami • E. Mazur
Division of Engineering and Applied Sciences, Harvard University, Cambridge,
MA 02138, USA
S. Nolte ()
Institute of Applied Physics, Friedrich-Schiller-University Jena, Max-Wien-Platz 1,
07743 Jena, Germany
e-mail: stefan.nolte@uni-jena.de
C.B. Schaffer
Department of Biomedical Engineering, Cornell University, Ithaca, NY 14853, USA
Fig. 12.1 Schematic diagram of typical micromachining setup. A femtosecond duration laser
pulse is tightly focused by a microscope objective or lens onto the surface of an absorbing or
transparent material for ablation or into the inside of a transparent sample for bulk micromachining.
The sample is mounted on a three-axis translation stage and its position is scanned relative to the
laser focus to produce the desired microstructure
Laser parameters and material properties have a significant effect on the inter-
action and therefore on the results of the micromachining process. Pulse energy,
repetition rate, wavelength, and pulse duration are several laser properties that
play a role in the absorption of laser pulses by solid materials. Specific properties
of materials, such as bandgap and thermal properties determine whether micro-
machining with femtosecond pulses is possible and how the resulting structures
look and behave. The processes by which energy from the laser is absorbed
by a material and the mechanisms by which the material responds are the key
to understanding the micromachining process. Section 12.2 will illuminate these
processes and give the reader an understanding of the capabilities of femtosecond
laser micromachining. In Sect. 12.3, several applications of micromachining in both
absorptive and transparent materials will be discussed.
The interaction of femtosecond laser pulses with materials differs from the inter-
action of longer pulses or continuous wave light in two ways. First, the energy
deposition occurs on a timescale that is short compared to any relaxation process.
The laser energy is absorbed by the electrons, leaving the ions cold, and only after
the laser pulse has left the focal region does thermalisation take place. Second, the
intensity of a femtosecond pulse, even with very moderate energy, is high enough to
drive highly nonlinear absorption processes in materials that do not normally absorb
at the laser wavelength.
Overall, linear absorption of femtosecond pulses does not differ from the linear
absorption of any other light field. In non-metallic materials, the highest energy
level occupied by an electron (in the valence band) is separated by an energy gap
(the bandgap) from the lowest energy level that is unoccupied (in the conduction
band). If the photon energy exceeds the bandgap energy, light can be absorbed by
the material, promoting electrons from the valence band to the conduction band
(Fig. 12.2, left panel) [32]. In metals, the conduction band is partially occupied,
providing unoccupied energy levels at only slightly higher energy than occupied
ones. In this case, photons can be absorbed through free-carrier absorption, a three-
body interaction where an electron gains energy by absorbing a photon, and gains
momentum through an interaction with a phonon (a lattice vibration) to move to
a higher-lying level in the conduction band [33]. In addition, there remains the
possibility for interband transitions between the conduction band and higher-lying
bands, as in insulating materials. If sufficient laser energy is deposited into the
surface of a material through linear absorption, material ablation can occur.
12 Femtosecond Laser Micromachining 291
Fig. 12.2 Schematic diagram of absorption mechanisms that play a role in femtosecond laser
micromachining. Linear absorption (left panel) dominates for materials with a bandgap that is
less than the photon energy. In transparent materials optical processes such as multiphoton (centre
panel) and avalanche (right panel) ionisation are required to deposit energy, where the photon
energy is less than the material bandgap
In samples that are transparent to the laser wavelength, a single photon of light does
not have enough energy to excite an electron from the valence to the conduction
band, and absorption of laser energy can occur only through nonlinear processes.
If enough laser energy is deposited into the material by this nonlinear absorption,
ablation or permanent structural changes can occur. Because the absorption is
nonlinear, i.e. the absorption probability is a nonlinear function of the laser intensity,
the absorption can be confined to the bulk of a transparent sample by tightly focusing
the laser beam inside the sample, producing a much higher laser intensity in the focal
volume than at the surface.
There are two classes of excitation mechanisms that play a role in absorption in
transparent samples: photoionisation and avalanche ionisation [34]. In photoionisa-
tion, electrons are directly excited from the valence to the conduction band by the
laser field. Depending on the laser frequency and intensity, there are two different
regimes of photoionisation: the multiphoton ionisation regime and the tunnelling
ionisation regime, shown to be two limits of the same process by Keldysh [35], yet
retaining very distinct conceptual pictures.
In tunnelling ionisation, the electric field of the laser suppresses the potential
that binds a valence electron to its parent atom, allowing the electron to tunnel out
and become a free electron. This type of nonlinear ionisation dominates for strong
laser fields and low laser frequency. At higher laser frequencies (but still below
that required for linear absorption) nonlinear photoionisation occurs due to the
simultaneous absorption of multiple photons by an electron (Fig. 12.2, centre panel).
The sum of the energy of all the photons absorbed must exceed the bandgap energy,
i.e. n„! Eg , where Eg is equal to the bandgap energy, ! to the laser frequency,
292 L. Cerami et al.
and „ to the reduced Planck constant, and n is the minimum number of required
photons. Multiphoton absorption requires n photons to interact simultaneously with
a single electron, so the probability for absorption will scale as the density of the
photons raised to the power n. Because photon density is proportional to laser
intensity, I, the probability for multiphoton absorption scales as In .
Avalanche ionisation involves free carrier absorption by an electron already in the
conduction band of the material followed by impact ionisation. An electron in the
conduction band sequentially absorbs several laser photons until its energy exceeds
the conduction band minimum by more than the bandgap energy, then collisionally
ionises another electron, leaving two electrons at the conduction band minimum
(Fig. 12.2, right panel) [34]. As long as the laser field is present this process repeats,
and the electron density in the conduction band grows exponentially.
Note that avalanche ionisation requires some “seed” electrons in the conduction
band of the material. For femtosecond duration laser pulses, photoionisation during
the leading edge of the laser pulse provides these seed electrons for avalanche
ionisation during the rest of the pulse [36]. The electron density grows through
avalanche ionisation until the plasma frequency of the electrons in the conduction
band approaches the frequency of the incident laser radiation (the “critical” plasma
density). This high density plasma strongly absorbs laser energy by free-carrier
absorption [36]. For ultrashort pulse durations, the absorption described above
occurs on a timescale that is short compared to the timescale for energy transfer
to the lattice, decoupling the absorption and lattice heating processes. At the end of
the laser pulse, we are essentially left with “hot” electrons and a “cold” lattice.
As discussed before, the laser radiation interacts primarily with the electronic states
of the valence and the conduction bands. The time to deposit the energy within
these states is determined by the laser pulse duration. For very short laser pulses the
resulting energy distribution of the electrons at the end of the laser pulse may be
non-thermal. Within a fraction of a picosecond this non-thermal energy distribution
will relax to a Fermi-Dirac distribution, which can be properly characterized by a
temperature. Later on, the deposited energy is redistributed over the various energy
states of the system, i.e. energy is transferred from the electrons to the lattice. In
order to understand the difference in the interaction of ultrashort and long laser
pulses with matter, it is instructive to consider the characteristic timescales involved.
This will be done in the following for the interaction with metals. However, similar
derivations can be obtained for other materials as well.
Let us first consider laser material interactions for a long laser pulse. In this case
“long” means that the pulse duration is much longer than any time-scale concerning
12 Femtosecond Laser Micromachining 293
the energy relaxation within the system. This is true for pulses of nanosecond
duration and longer. (Note that in some cases nanosecond pulses are considered to
be short; it is thus obvious that these attributes can only be interpreted in the proper
context!) Since all relaxation processes occur much faster than the pulse duration,
the whole system will be in equilibrium during the interaction process.
Let us consider energy conservation in this system. Since any change in the
internal energy U within a volume V must be either due to the absorption of laser
light within this volume (S V) or due to energy diffusion out of the volume (Ej being
the energy flux), the change in energy can be described by
@U
D r.Ej V / C S V (12.1)
@t
This change in internal energy can be connected to a change in temper-
ature T using the specific heat capacity Ci of the material per unit volume
(@U/@t D Ci V @T/@t). The heat source S can be calculated taking the energy loss
of the laser radiation (S D (1 R) @I/@z) into account (R denotes the surface
reflectivity and I the laser intensity). Using Beer’s law: I.t; z/ D I.t/ exp.˛z/,
this can be simply derived to
Ej D k rT: (12.3)
Here, k denotes the thermal conductivity. We can now rewrite Eq. (12.1) as:
@T
Ci D r.k rT / C S: (12.4)
@t
This is the so-called parabolic heat equation. The temperature dependence of the
thermophysical parameters (Ci , k) and of the optical parameters (A, ˛) makes Eq.
(12.4) non-linear, so that its analytical solution can be obtained only in a few cases
(see Ref. [37]). In general, it has to be solved numerically.
We will restrict ourselves here to only a brief discussion of the heat transport,
in order to obtain an intuitive feeling for the typical temporal and spatial scales
involved. Assuming constant thermophysical parameters (Ci , k) one can derive from
Eq. (12.4) that the resulting heat wave after laser absorption will propagate for
about [37]:
p
lt h D t: (12.5)
294 L. Cerami et al.
lt h is called the heat diffusion length, and D D k/Ci is the thermal diffusivity. For
most metals the thermal diffusivity is typically in the range of 0.1–1 cm2 s1 [37],
which means that the heat wave will travel about 0.1–1 m within the pulse duration
for a 10 ns pulse. Note that this is much longer than the optical penetration depth
’1 , which is on the order of 10 nm for metals. Thus, it is appropriate to speak of
the laser as a surface source of heat. Since typical laser spot sizes are much larger
than the heat diffusion length, heat diffusion will be mainly a one-dimensional heat
flow directed normal to the surface in this case.
It is easy to see from Eq. (12.5) that the heat diffusion length will become equal or
even smaller than the optical penetration depth for pulse durations in the picosecond
or femtosecond range. In addition, there is an even more dramatic change in the
interaction for these short pulse durations. Now, the laser pulse duration is shorter
than the energy relaxation within the system, i.e. the time it takes for the electrons
to transfer their energy to the lattice. As a consequence, heating with ultrashort laser
pulses results in strong non-equilibrium conditions. Thus, electrons and lattice have
to be described by their own temperatures separately. Such a description, called the
two-temperature model (TTM), has been proposed by Anisimov et al. [38],
@Te @Q.z/
Ce D .Te Ti / C S; (12.6)
@t @z
@Ti
Ci D .Te Ti /: (12.7)
@t
Note that a similar two-temperature model had already been used 20 years earlier
to describe the energy relaxation between the electrons and the lattice in order to
explain deviations from Ohm’s law at high currents [39].
In Eqs. (12.6 and 12.7) Te and Ti , denote the electron and lattice temperatures,
z is the direction perpendicular to the target surface, and Ce and Ci are the heat
capacities (per unit volume) of the electron and lattice subsystems, respectively.
The parameter characterizes the electron-lattice coupling, i.e. the energy exchange
between the two systems. Expressions for the coupling constant ” can be found in
[39–41]. Measurements revealed values in the range of D 1016 –1017 Wm3 K1
[42–44]. Q.z/ D ke .Te /@Te =@z is the heat flux, and S D I.t/A˛ exp.˛z/ is the
laser heating source term. A and ˛ are the surface absorptivity and the material
absorption coefficient, I(t) is the laser intensity, and ke is the electron thermal
conductivity. In the above equations, thermal conductivity in the lattice subsystem
(phonon component) is neglected because this heat transport is typically negligible
compared to the electronic contribution for metals.
The strong non-equilibrium conditions associated with ultrashort pulse laser
interaction with solids makes a detailed theoretical analysis difficult. In principle,
12 Femtosecond Laser Micromachining 295
the fast and significantly varying properties, including heat capacity, thermal
conductivity, relaxation time, reflectivity, and absorption coefficient have to be taken
into account by using quantum treatments [41]. Instead, we will discuss here a
simplified analytical solution, which is more instructive for the reader.
By dividing Eqs. (12.6 and 12.7) by the corresponding heat capacities Ce and
Ci , respectively, it can be easily seen that these equations have two characteristic
timescales e D Ce / and i D Ci / , where e is the electron cooling time, and i is
the lattice heating time. Due to the differences in the heat capacities, it is always true
that e i . Typical values are of the order of 1 ps for e and 0.01–1 ns for i . These
timescales define three different regimes of interaction: the laser pulse duration is
shorter than e , longer than i or in between the two values. Below, we will discuss
the differences.
Long (Nanosecond) Pulses If the laser pulse duration L is much longer than the
lattice heating time, thermalisation between the electron subsystem and the lattice
takes place during the laser pulse. In this case the electrons and the lattice can
be characterized by a common temperature T D Te D Ti and Eqs. (12.6 and 12.7)
reduce to the parabolic heat diffusion equation (Eq. 12.4) as discussed above.
Picosecond Pulses For picosecond laser pulses the condition e L i is
fulfilled. For times t e , Eq. (12.6) for the electron temperature becomes qua-
sistationary, i.e., the energy loss due to heat transport and energy exchange with
the lattice is compensated by the energy deposited due to the laser radiation.
Accordingly, Eq. (12.6) can be simplified, because @Te /@t D 0. For times t i it
is also possible to simplify the equation for the lattice temperature. Due to the
quasistationary character of the electron temperature, the lattice temperature will
exponentially approach the electron temperature and can thus be expressed as [29]:
t t
Ti Te 1 exp Te : (12.8)
i i
Here, an initial lattice temperature has been neglected. It can be seen from this
expression that in the picosecond regime, the lattice temperature remains much
lower than the electron temperature during the pulse duration. Equilibrium is
obtained only after the pulse is over.
However, a detailed analysis of this interaction regime depends on the actual ma-
terial properties. Since these parameters will strongly change during the interaction
process, such an analysis is beyond the scope of this chapter.
Femtosecond Pulses For pulse durations much shorter than the electron cooling
time ( L e ), i.e. femtosecond pulses or pulses of a few picosecond duration,
the coupling between the electrons and the lattice can be neglected in a first
approximation for times t e . In this case, the differential Eq. (12.6) can be pretty
easily solved, since it is identical with the parabolic heat Eq. (12.4), except that the
electron heat capacity has a specific temperature dependence.
296 L. Cerami et al.
2 ne kB 2
Ce D Ce0 Te D Te ; (12.9)
2 EF
@Te 2
Ce0 D 2I.t/A˛ exp.˛z/: (12.10)
@t
Assuming a constant intensity I(t) D I0 , Eq. (12.10) can be directly integrated,
yielding
s
2Ia ˛
Te .t/ D t exp.˛z/: (12.11)
Ce0
t
Ti Te .L / : (12.13)
i
The attainable lattice temperature is determined by the average cooling time of the
electrons ea Ce0 Te .L / =2 [29] and is given by:
12 Femtosecond Laser Micromachining 297
Ce0 Fa ˛
Ti Te 2 .L / D exp.˛z/: (12.14)
2Ci Ci
Fa
Ti exp.z=lt h /: (12.15)
Ci lt h
12.2.3.1 Ablation
The non-equilibrium physics associated with ultrashort pulse laser ablation makes a
detailed theoretical analysis difficult. Thermodynamically metastable states such as
superheated liquids are likely to play a role under these highly transient conditions.
Recent experiments revealed that ultrashort laser pulse ablation is a process with
emission of particles up to several hundred nanoseconds after the laser pulse in at
least two steps [50]. In the first step a thin surface layer on the order of the optical
penetration depth is ablated by electron emission, sublimation and transition to the
plasma state. These ablation processes proceed on a nanosecond timescale or faster.
The remaining heat diffuses into the sample and leads to the emission of particles
and droplets after several 10 ns due to a thermal boiling process. Note that the
material ejected in the initial ablation step can carry away much of the deposited
laser energy, especially when working very close to the ablation threshold, thereby
minimizing the amount of thermal energy that later diffuses into the sample. Only
for femtosecond pulses is this true, and it leads to greatly reduced collateral damage
to material surrounding the ablation zone. A full treatment of the ablation process is
thus complex. While numerical approaches including plasma hydrodynamics yield
good agreement with experimental results (see, Ref. [51]), we will present a very
simplified analytical solution below. First we will discuss the scenario of metal
ablation.
In this simplified model, we assume that the energy of the lattice Ci Ti (per
unit volume) has to exceed a certain threshold value in order to obtain significant
ablation. For the following discussion, it will be sufficient to estimate this threshold
value as the heat of evaporation ˝, where is the density and ˝ is the specific heat
of evaporation per unit mass [29]. The condition for significant ablation, Ci Ti ˝,
can be written based on Eq. (12.14) as:
Fa exp.˛z/: (12.16)
˛
If, however, the heat diffusion length for the electrons before they transfer
their energy to the lattice, lth D (D e )1/2 ˛ 1 , is much longer than the optical
penetration depth, Eq. (12.16) has to be changed to
according to Eq. (12.15). The corresponding thresholds for the absorbed laser
fluences are Ft h =˛ and Ft h lt h , respectively. Note that these thresholds
are independent of the pulse duration. This is in contrast to longer pulses ( L > i ),
where the heat diffusion length and therefore the corresponding threshold value
scale with the square root of the pulse duration [52].
12 Femtosecond Laser Micromachining 299
The ablation depths can be easily derived from Eqs. (12.16) to (12.17), giving:
This logarithmic dependence of the ablation depth on the laser pulse fluence has
been demonstrated in several experiments (see, Ref. [30]). However, it is interesting
to note, that as the laser fluence increases, the ablation behaviour will change from
being limited by the absorption depth (Eq. 12.18) to being limited by the diffusion
length (Eq. 12.19). This is due to faster diffusion for the highly energetic electrons
that are produced only in the high fluence regime [47, 53].
For semiconductors and transparent materials the physics of ultrashort pulse
laser ablation is basically the same as for metals as described above. The main
difference is that in metals the electron density can be considered constant during
the interaction with an ultrashort laser pulse, while the electron density changes
in transparent materials and semiconductors due to interband transitions, as well
as nonlinear field and avalanche ionisation (see Sect. 12.2.1.2). In transparent
materials, free electrons are first generated by nonlinear field ionisation, which serve
as seed electrons for the avalanche process. The ablation process thus does not rely
on the presence of free electrons due to defects. As a consequence, the ablation
threshold for femtosecond pulses is much better defined than for long laser pulses,
where initial electrons are required resulting from defects [54].
Concerning the ablation threshold, transparent materials and semiconductors do
not exhibit a constant value as a function of pulse duration (in the short pulse regime)
like metals (see above). This is a direct consequence of the strongly nonlinear
absorption process. Since the absorption increases with increasing intensity, i.e.,
decreasing pulse duration, the threshold fluence also decreases with pulse dura-
tion [52].
The high potential for ultrashort pulse laser micromachining arises from the
fact that heat diffusion can be minimized and the interaction process is mainly
determined by the optical penetration depth. In the regime of low laser fluences,
high precision processing is thus possible with negligible heat affected zones and
practically no burr formation. Moreover, the laser fluence threshold is much lower
than for longer pulses due to the reduced thermal losses and, for transparent
materials, higher intensities to drive nonlinear processes. Therefore, lower pulse
energies can be used for processing, resulting in significantly reduced mechanical
damage around the processed area.
In contrast, the absorption of laser radiation at very high fluences results in the
creation of a dense, high-temperature plasma. In this regime, the material is removed
in plasma, liquid and vapour phases, and the heat penetrates deep into the target. The
plasma expansion and the vaporisation process create a strong recoil pressure that
expels molten material. This results in the formation of droplets and debris on the
300 L. Cerami et al.
target surface. Expelled liquid material, which resolidifies at the edges of the hole,
results in the formation of burr. Moreover, the high ablation pressure causes strong
mechanical deformations and even cracks in the surrounding material [55].
While using ultrashort laser pulses in this processing regime provides no
advantage regarding the achievable machining precision, it should be noted that
relatively high ablation rates can be achieved [56]. This is due to the fact that the so-
called plasma-shielding (the laser pulse is absorbed and reflected by the expanding
plasma and does not reach the target; see, Ref. [57]) which limits the achievable
ablation rate for long laser pulses, does not exist for ultrashort pulses due to the
temporal separation between energy deposition and ablation process (the plasma
expansion occurs mainly after the laser pulse).
When sufficient energy is deposited into the bulk of a transparent material through
nonlinear absorption of tightly-focused femtosecond laser pulses, structural changes
can be produced inside the sample without affecting the surface, allowing three-
dimensional patterns to be fabricated by translating the laser focus throughout the
sample. In contrast to the surface ablation discussed above, no energy deposited by
the laser can be carried away from the sample through material ablation, resulting
in somewhat different regimes of material damage. To date, three qualitatively
different kinds of structural changes have been induced in the bulk of transparent
materials with femtosecond laser pulses: an isotropic refractive index change
[12, 58]; a birefringent refractive index change [59, 60]; and a void [25, 26, 61].
The structure produced depends on laser parameters (pulse duration, wavelength,
energy, repetition rate), material parameters (bandgap, thermal properties), and the
numerical aperture (NA) of the focusing objective. In the case of tightly-focused
(NA > 0.65), 100-fs, 800-nm pulses in fused silica, each of the three morphologies
can be produced by varying only the incident laser energy, with isotropic refractive
index changes produced from the threshold energy of about 40 nJ to about 150 nJ,
birefringent refractive index changes produced from about 150 nJ to about 500 nJ,
and voids formed above about 500 nJ. Figure 12.3 illustrates these three types of
modification.
The isotropic refractive index changes are likely due to localized melting and
rapid resolidification of the glass (Fig. 12.3, left panel) [62–64]. In fused silica, the
density increases if the glass is quenched from a high temperature [65, 66], partially
explaining the higher refractive index observed in femtosecond laser irradiated fused
silica [8, 67, 68]. Glasses with a more typical thermal behaviour show a lower
refractive index and density in the irradiated region [64, 69]. From this evidence,
it seems likely that the energy deposited in the focal volume by near-threshold
femtosecond laser irradiation leads to heating and melting of the glass in the focal
volume. Subsequently this glass rapidly cools, leading to density and refractive
index changes.
12 Femtosecond Laser Micromachining 301
Fig. 12.3 Schematic diagram of the morphology of femtosecond laser-induced structural modifi-
cations in bulk transparent materials. Three different energy regimes, low, intermediate, and high,
are represented from left to right, with examples of the damage morphology produced in glass in
the lower panels. On the left is an optical image of a refractive index change. Centre is an SEM
image of a nanograting. Right is an SEM image of a void (The left, centre and right hand images
are reproduced and adapted with permission from Refs. [61], [70] and [25] respectively)
As discussed earlier, laser oscillators and amplified systems have very different
repetition rates. As a result, oscillator-only micromachining is fundamentally
different from micromachining with amplified laser systems. Because the average
output power is limited by the pump source, there is a trade-off between the
pulse energy and pulse repetition rate of femtosecond lasers. The time interval
between the pulses emitted by a femtosecond laser oscillator is on the order of tens
of nanoseconds, which is significantly shorter than the 1- s heat diffusion time
characteristic of the micrometre-sized focal volume of the high numerical aperture
objectives used for bulk micromachining. Consequently there is not enough time
between successive pulses for the energy deposited by one laser pulse to diffuse
out of the focal volume before the next pulse arrives. Over time, the energy from
successive pulses accumulates in and around the focal volume, producing damage.
The train of oscillator pulses constitutes a “point source of heat” located at the focal
volume within the bulk of the material. The longer the material is exposed to the
train of pulses from an oscillator, the higher the temperature at the focus and the
larger the region that is heated. If the temperature exceeds the material’s melting
point, structural changes can occur. The pulses from an amplified femtosecond laser
system, on the other hand, typically are separated by milliseconds, which far exceeds
the time required for heat to diffuse out of the focal volume. The focal volume
thus returns to room temperature before the next pulse arrives. Consequently, the
structural change caused by an amplified laser is confined to the focal volume,
regardless of the number of pulses that strike the sample. This means that the
shape and volume of the modified area depends on the volume of the focus. For
low numerical aperture (NA) objectives, the confocal parameter is much larger than
the radius of the focal spot, leading to an elliptical focal volume. Only with very
high NA objectives or special beam shaping techniques [16, 77, 78] does the focal
volume become more symmetric. On the other hand, when micromachining with a
femtosecond oscillator, heat causes melting outside of the focal volume, leading to
a more spherically shaped modified region. The shape of the focal volume becomes
a factor when choosing a waveguide writing geometry, which will be discussed in
Sect. 12.3.
size that ensures intensities above the damage threshold. For micromachining in
glass, microscope objectives are typically used for this purpose to achieve a very
tight focus.
When dealing with a focused femtosecond laser beam in a transparent material,
one must consider two issues that can arise: self-focusing and aberrations. Self-
focusing is a process by which the spatial variation in intensity across a laser beam
leads to a spatial variation in the refractive index (through the nonlinear Kerr index
n D n0 C n2 I ), which acts as a lens and causes the beam to focus itself [79]. This
process leads to a more intense laser beam and can cause catastrophic collateral
damage when micromachining precise structures. Fortunately, there are ways to
minimize self-focusing when micromachining with femtosecond laser pulses. Self-
focusing depends on the peak power of the laser, whereas microstructuring in glass
is intensity-dependent. By keeping the peak power below the threshold for self-
focusing, while keeping the intensity above the threshold for damage, one can ensure
precise micromachining. This is done by using high NA focusing optics.
Aberrations arise as distortions to the laser beam at or near the focus and
are a result of the focusing process. Spherical and chromatic aberrations are two
types that should be considered for micromachining, though chromatic aberrations
may be neglected if the bandwidth of the laser is narrow enough. As mentioned
above, microscope objectives are typically used for micromachining in transparent
materials. These complex lens systems are often corrected for aberrations when
used with a glass cover slip of a specific thickness. When using different materials
(different refractive index) or different thicknesses of material, the aberrations
may become noticeable and degrade the focus. The change in the focus affects
the threshold energy above which modification occurs. Higher numerical aperture
objectives exhibit more significant spherical aberration than lower NA objectives.
For surface micromachining, focusing should be aberration free to achieve the
finest feature size, by using, for example, a microscope objective that is designed
to form a sharp focus without a glass coverslip. Only when using very high pulse
energies does self-focusing in the air in front of the sample become a problem. Such
high-energy machining is usually associated with the use of projection masks to
ablate a large, patterned area at once, rather than with the point-scanning systems
discussed here. The self-focusing can be avoided by placing the sample in a vacuum
chamber.
Precise micromachining using long laser pulses is often limited due to thermally
or mechanically induced damage (melting, stress, formation of microcracks) of the
surrounding area. These problems are especially severe for brittle materials, like
semiconductors, or for materials with a high thermal conductivity, like metals. The
collateral damage can be significantly larger than the desired structures in this case.
Moreover, the process efficiency is reduced due to the extensive thermal losses.
In contrast, thermal or mechanical damage can be minimized or almost avoided
as discussed in Sect. 12.2.3.1 when ultrashort laser pulses are used for processing.
Since the heating of surrounding material through thermal diffusion is minimal in
this case, the processing efficiency is much higher compared with the long pulse
case. Moreover, the processing area can be precisely defined by the laser beam
geometry, i.e., by the diffraction limit of the optical system. Thus, structure sizes
within the nanometre range can be obtained.
One application example for nanometre scale structuring is the repair of photolitho-
graphic masks. Photolithographic masks typically consist of a fused silica substrate
coated by a 100-nm thick chrome layer. The chrome layer is patterned using either
electron beam lithography or ultraviolet light. This pattern is then imaged onto the
semiconductor wafer to expose the photoresist by an appropriate lens system which
reduces the pattern size by a factor of 4 [7].
Unfortunately, many masks exhibit defects, which in turn result in failures in the
produced integrated circuits. Due to the high costs of the masks, a reliable defect
removal method is indispensable. In most cases the defects consist of excess chrome
in undesired places, which has to be removed with minimal collateral damage.
12 Femtosecond Laser Micromachining 305
Fig. 12.4 Atomic force microscope images of a chromium layer on top of a fused silica substrate
structured with (a) 270-nJ, 10-ns and (b) 225-nJ, 150-fs pulses using 780-nm laser radiation [6].
Note the different height scaling in the images
Such a defect repair can be achieved using focused ion beams. However, their use
is limited due to damage of the substrate [5], which results in reduced transmission
or phase shifts of the transmitted light.
These drawbacks can be overcome if the defects are removed using laser
radiation. However, long-pulse laser processing suffers from the large thermal
damage created. To illustrate this, Fig. 12.4 shows a comparison of the ablation of a
100-nm thick chrome film on top of a fused silica substrate by 10-ns (Fig. 12.4) and
150-fs pulses (Fig. 12.4) [6]. In both cases the 780-nm radiation has been focused
onto the sample and scanned across it along two different paths. The pulse energies
were chosen to be close to the ablation threshold. In the case of nanosecond pulses
306 L. Cerami et al.
Fig. 12.5 Microscope images (transmission mode at 248 nm) of a photomask before (left) and
after (right) repair. At two sites the chrome has been fully ablated and the transmission was
recovered (Reproduced and adapted with permission from Ref. [7])
the metal film is melted and then resolidifies at the edge of the trench creating a
large burr. It is thus obvious that such an approach is impractical for the defined
chrome removal on a nanometre scale.
In contrast, processing with femtosecond pulses results in clean ablation with
sharp edges, and no metal splatter or debris is evident. Since the fused silica sub-
strate has an ablation threshold significantly higher than the chrome layer selective
ablation is possible without damaging the substrate. Thus, a high transmission at
the repaired site can be recovered. Typical transmission of repaired sites in excess
of 95 % of similar areas requiring no repair have been achieved on photolithographic
masks [5].
As feature sizes of the photolithographic masks decrease, the resolution demands
on the repair tool increase. In principle, the minimum achievable structure size
is determined by the diffraction limit of the optical system and is of the order of
the laser wavelength. Thus, using short laser wavelengths improves the resolution
[80]. A further increase in resolution can be obtained by taking advantage of the
well-defined ablation threshold associated with ultrashort laser pulses. By choosing
the peak of the Gaussian intensity distribution only slightly above the threshold
value, sub-diffraction sized holes can be produced [3]. Combining both approaches,
researchers at IBM were able to achieve a resolution of 75–80 nm by tightly focusing
ultrashort pulses at a wavelength of 266 nm (frequency tripled Ti:sapphire laser)
onto the photomasks [7]. This corresponds to a resolution of /3, i.e., substantially
below the diffraction limit.
An example of the application of this technique is given in Fig. 12.5, where
two regions of chrome in the left image are removed. The transmission image on
the right impressively demonstrates the accuracy of the process as well as the high
transparency obtained after repair [7].
12 Femtosecond Laser Micromachining 307
High quality materials processing without visible thermal and mechanical damage
can be achieved using ultrashort laser pulses at low fluences close to the ablation
threshold as detailed in Sect. 12.2.3.1. However, several applications like the drilling
of deep holes, as required for fuel injection nozzles, spinnerets and extrusion dies,
rely on the use of much higher fluences significantly above the ablation threshold
[4]. Apart from the fact that a high processing speed is necessary in order to
drill the hole in a reasonable time, the high fluences are mainly necessary due to
the following two reasons. On one hand, material expulsion occurs deep within a
typically small hole. Therefore, the ejected particles need to have a high kinetic
energy in order to be able to escape from the drilled hole. On the other hand, laser
drilled holes typically exhibit a conical shape, which is due to the Gaussian intensity
profile of the beam and due to enhanced thermal losses at the walls of the hole. As
a consequence, the laser radiation is interacting at grazing incidence with the walls
increasing the interaction area and therefore, decreasing the effective fluence.
Despite the high fluences applied in hole drilling, high quality processing can be
achieved. This is due to the fact that the drilling with ultrashort pulses is basically
a two-step process [4]. In the first phase of drilling high laser intensities are used
to rapidly drill a through-hole. Once this is accomplished the high intensity part
of the laser pulse directly propagates through the hole without any interaction with
the hole walls. Only the wings of the Gaussian intensity distribution which exhibit
a much lower fluence interact with the hole walls. This results in a very gentle
material removal with all the advantages of low-fluence ultrashort pulse ablation.
Thus, one can consider this low-fluence finishing as an “integrated” post-processing
step, which results in high hole quality without the need for any post-processing.
Apart from the advantages of using ultrashort pulses for drilling as discussed
above, simple percussion drilling could not sufficiently fulfil the requirements
for high precision, high quality, and high processing efficiency at the same time.
Thus, new processing techniques like helical drilling were developed. By applying
these drilling techniques the precision and quality of holes produced could be
considerably improved [81].
Moreover, such advanced processing techniques allow even for the fabrication
of special hole shapes. An example is presented in Fig. 12.6, which demonstrates
the fabrication of an inkjet nozzle in a stainless steel plate [82]. Here, picosecond
pulses are applied in a laser milling process in order to obtain a certain hole taper.
Therefore, a uniform thin layer of material is ablated by moving the laser spot on the
sample surface in a spiral way. The outer diameter of the spiral defines the diameter
of the layer removed. Subsequent layers are ablated while the outer diameter of the
spirals is reduced from layer to layer. This process is repeated until the desired depth
is reached. By controlling the depth of each layer removed and the outer diameter
reduction between successive spirals, the amount of taper of the milled hole can be
precisely adjusted. In order to obtain a high processing speed, up to 400 holes are
drilled simultaneously by this method after splitting the beam using a diffractive
optical beam splitter [82].
308 L. Cerami et al.
Fig. 12.6 Laser milling process to drill holes with precise taper control (Images courtesy of X.
Liu, Panasonic)
Femtosecond lasers have the ability to induce bulk refractive index changes inside
of transparent materials, as discussed in Sect. 12.2.3.2 (and also Chap. 13 by
Thomson and Cerullo). Where the intensity is high enough to induce damage, the
region will be modified, while the rest of the material remains unaffected. The
change in the refractive index must, in general, be positive in the irradiated region
for the fabrication of optical waveguides to be possible. Whether the refractive
index increases or decreases in the laser irradiated volume depends on the material
parameters as well as laser parameters. Note that even in glasses where irradiation
leads to a refractive index decrease in the irradiated volume, there is often a
refractive index increase just outside this volume, likely due to compression and
densification of the material that surrounds the, now less dense, irradiated volume
[69, 88, 89]. By translating the laser beam with respect to the sample, a cylinder
shaped region of index modification, and therefore a waveguide, can be formed.
The experimental parameters used in the micromachining process determine the re-
fractive index profile of the microstructures. Therefore, it is important to understand
their relationships when fabricating waveguide-based devices.
A major advantage of fabricating waveguides using femtosecond microma-
chining is the ability to create three-dimensional structures inside of transparent
materials. Standard waveguide fabrication methods are limited to planar or layer-
by-layer processing. By varying the focus of the laser beam, it can be precisely
controlled where the microstructuring occurs inside of a sample. This micromachin-
ing process has the flexibility to create structures that extend through the depth of the
sample (by varying the focal depth) or that fall in a single plane (by only translating
the sample transverse to the laser beam). Figure 12.7 shows two different geometries
that are used for fabricating straight waveguides inside of transparent materials:
transverse and longitudinal. In the transverse geometry, the sample is translated in
a perpendicular direction with respect to the laser beam. The longitudinal geometry
involves translating the sample in a direction parallel to the incident laser beam,
such that the depth of the laser focus in the material is constantly changing as the
sample is translated. As a result, a 10-mm long waveguide would require a focusing
element with at least a 10-mm focal length.
In choosing which geometry to use, one must consider the laser system,
the working distance of the microscope objective, and the shape of the focal
volume. With oscillator-only micromachining, cumulative heating leads to roughly
symmetric structures [27] (Fig. 12.8). Translating the sample either longitudinally
or transversely would result in the same waveguide shape. The limiting factor
for oscillator-only micromachining is the working distance of the microscope
objective. In order to reach high enough intensities to induce refractive index
changes, typically very high NA objectives are used, which consequently have very
short working distances. When using an amplified laser system, the asymmetry
310 L. Cerami et al.
Fig. 12.7 Geometries used for micromachining waveguides in the bulk of transparent samples.
The self-channelling and longitudinal geometries are sometimes used with amplified laser systems;
oscillator-only micromachining is done transversely
Fig. 12.8 (a) Optical microscope image of waveguides written inside bulk glass using a 25-MHz
train of 5-nJ energy, sub-100-fs duration laser pulses focused with a 1.4-NA microscope objective
while the sample was translated at 20 mm.s1 transverse to the beam propagation direction. The
inset shows an optical image of the end face of a waveguide. (b) Near-field mode of a waveguide
like that shown in (a) for 633-nm light (Reproduced and adapted with permission from Ref. [27])
of the focal volume (when using a low NA microscope objective) limits which
geometry can be used. If the sample is translated transversely to the objective, an
elliptical waveguide cross-section would result. This is sometimes undesirable and
so the longitudinal writing method is better suited for amplified systems. However,
adjustments can be made to use the transverse writing geometry with amplified
systems. For example, it has been demonstrated that circular cross-sections can
be obtained using an elliptical beam (by way of beam shaping or a simple slit)
[16, 77, 78] or by a multiscan approach, where the waveguide is composed of
several overlapping modification lines [90–93]. For a more in depth discussion of
waveguide shaping techniques, see Chap. 13 by Thomson and Cerullo.
12 Femtosecond Laser Micromachining 311
Although one does not usually think of biological tissues as being transparent,
in the infrared wavelength range linear absorption becomes weak enough that
nonlinear absorption can dominate for tightly-focused femtosecond laser pulses.
Femtosecond lasers can thus provide a laser scalpel with sub-micrometre precision
that can disrupt specifically targeted structures several hundred micrometres beneath
a tissue surface without affecting surrounding tissue. This tool is gaining traction
in biomedical research, with important studies in cell biology [111, 112], in
invertebrate neuroscience [113, 114], in studies of drosophila development [115],
and in reproducing neurovascular diseases that affect humans in rodent animal
models for the evaluation of therapeutics [116].
Polymers are a popular class of materials that we find in many places in our daily
life. Poly(methyl methacrylate) or PMMA is a hard plastic polymeric material,
very similar to glass that is used in many applications where more durability
and lower cost is necessary. Polymers provide distinct advantages over glass due
to their flexible processing techniques and adjustable compositions. For example,
base polymers can be doped with molecules with a high nonlinear susceptibility
and used to make optical waveguide devices that allow efficient second and third
order nonlinear processes. Polymers have already been implemented in passive
optical waveguide devices mainly due to their low cost. The extension of waveguide
micromachining from glass into polymers shows promise, although the laser-
material interactions may prove to be very different.
Recently, research on micromachining optical waveguides in polymers has
achieved some success. Zoubir et al. showed that tubular waveguides could be
formed inside of poly(methyl methacrylate) (PMMA) using a 25 MHz repetition
rate femtosecond laser oscillator [28]. The structures have a circular cross-section
with diameter of 25 m but were shown to only guide 633 nm light in the outer
region of the waveguide. Symmetric single mode waveguides were fabricated in
12 Femtosecond Laser Micromachining 313
PMMA using a 1-kHz repetition rate femtosecond laser system [101]. Here, a slit
was placed before the focusing lens for beam shaping in order to produce a circularly
symmetric focal spot.
Other recent work demonstrated microchannel fabrication using an 80 MHz
repetition rate laser oscillator with pulse energies of 0.9 nJ [117]. In this two-
step process, a cylindrical modified region is first formed by focusing the laser
into a PMMA sample using a water immersion lens. Next the structured sample
is annealed at a temperature above the glass transition temperature of PMMA to
cause expansion of the modified region and the creation of a microchannel inside
the bulk.
12.3.3.2 Photopolymerisation
12.3.3.3 Photodeposition
Apart from the precise and practically damage free ablation discussed in Sect. 12.3.1
material deposition using ultrashort pulses is an upcoming application. In pulsed
laser deposition (PLD), where the ablated material is sputtered onto a target, the
314 L. Cerami et al.
Fig. 12.9 Scanning electron microscope images of (a) a “micro-bull” and (b) a woodpile structure
produced using two-photon photopolymerisation (Image in (a) courtesy of S. Kawata, Osaka Univ.
and reproduced and adapted with permission from Ref. [120])
12.4 Conclusions
repetition rate has a significant effect on the modification due to heat accumulation
effects. Pulse duration affects the relaxation processes. Laser wavelength is impor-
tant in determining how the energy is absorbed. Pulse energy limits what type of
structures can be produced. The energy must be above some damage threshold value
to modify a material. In addition, when working above the threshold, the amount of
energy can affect the size and shape of the resulting structures.
Waveguide fabrication is probably the most widely studied application of mi-
cromachining in transparent materials. Over the past 10 years, work has been done
in a variety of glasses, crystals, and polymers, and many waveguide-based devices
have been made, including couplers, interferometers, gratings and sensors. Future
work will rely on using new materials to develop unique waveguide-based photonic
devices. A similar use of femtosecond lasers is the fabrication of microfluidic
channels in glass or transparent polymers. Femtosecond lasers also provide a way
of removing materials from the surface of metals, semiconductors, and transparent
materials. Laser drilling of sub-micron holes and removal of material to repair
photolithographic masks are two applications of micromachining in absorbing
materials.
The interaction of femtosecond laser pulses and solid materials has been widely
studied over the last four decades. In addition to the scientific understanding we
have gained, applications of femtosecond laser micromachining have had an impact
in photonics, electronics and biology. The field of micromachining continues to
provide us with interesting science and engineering challenges, as new materials
are developed and new applications are envisaged.
References
10. A.M. Streltsov, N.F. Borrelli, Fabrication and analysis of a directional coupler written in glass
by nanojoule femtosecond laser pulses. Opt. Lett. 26, 42–43 (2001)
11. K. Minoshima, A.M. Kowalevicz, E.P. Ippen, J.G. Fujimoto, Fabrication of coupled mode
photonic devices in glass by nonlinear femtosecond laser materials processing. Opt. Express
10, 645–652 (2002)
12. C. Florea, K.A. Winick, Fabrication and characterization of photonic devices directly written
in glass using femtosecond laser pulses. J. Lightwave Technol. 21, 246–253 (2003)
13. S. Taccheo, G. Della Valle, R. Osellame, G. Cerullo, N. Chiodo, P. Laporta, O. Svelto, A. Killi,
U. Morgner, M. Lederer, D. Kopf, Er:Yb – doped waveguide laser fabricated by femtosecond
laser pulses. Opt. Lett. 29, 2626–2628 (2004)
14. M. Kamata, M. Obara, R.R. Gattass, L.R. Cerami, E. Mazur, Optical vibration sensor
fabricated by femtosecond laser micromachining. Appl. Phys. Lett. 87, 051106-051101-
051103 (2005)
15. A. Marcinkevicius, S. Juodkazis, M. Watanabe, M. Miwa, S. Matsuo, H. Misawa, J.
Nishii, Femtosecond laser-assisted three-dimensional microfabrication in silica. Opt. Lett.
26, 277–279 (2001)
16. Y. Cheng, K. Sugioka, K. Midorikawa, M. Masuda, K. Toyoda, M. Kawachi, K. Shihoyama,
Control of the cross-sectional shape of a hollow microchannel embedded in photostructurable
glass by use of a femtosecond laser. Opt. Lett. 28, 55–57 (2003)
17. M. Masuda, K. Sugioka, Y. Cheng, N. Aoki, M. Kawachi, K. Shihoyama, K. Toyoda, H.
Helvajian, K. Midorikawa, 3-D microstructuring inside photosensitive glass by femtosecond
laser excitation. Appl. Phys. A – Mater 76, 857–860 (2003)
18. Y. Bellouard, A. Said, M. Dugan, P. Bado, Fabrication of high-aspect ratio, micro-fluidic
channels and tunnels using femtosecond laser pulses and chemical etching. Opt. Express 12,
2120–2129 (2004)
19. Y. Iga, T. Ishizuka, W. Watanabe, K. Itoh, Y. Li, J. Nishii, Characterization of micro-
channels fabricated by in-water ablation of femtosecond laser pulses. Jpn. J. Appl. Phys.
1(43), 4207–4211 (2004)
20. K. Ke, E.F. Hasselbrink, A.J. Hunt, Rapid prototyped three-dimensional nanofluidic channel
networks in glass substrates. Anal. Chem. 77, 5083–5088 (2005)
21. C. Hnatovsky, R.S. Taylor, E. Simova, V.R. Bhardwaj, D.M. Rayner, P.B. Corkum,
Polarization-selective etching in femtosecond laser-assisted microfluidic channel fabrication
in fused silica. Opt. Lett. 30, 1867–1869 (2005)
22. T.N. Kim, K. Campbell, A. Groisman, D. Kleinfeld, C.B. Schaffer, Femtosecond laser-drilled
capillary integrated into a microfluidic device. Appl. Phys. Lett. 86, 201106 (2005)
23. P. Maker, R. Terhune, C. Savage, Optical third harmonic generation, in Proceedings of the
Third International Conference on Quantum Electronics, Paris, 1963, ed. by P. Grivet, N.
Bloembergen (Columbia Unversity Press), p. 15559
24. M. Hercher, Laser-induced damage in transparent media. J. Opt. Soc. Am. 54, 563 (1964)
25. E.N. Glezer, M. Milosavljevic, L. Huang, R.J. Finlay, T.H. Her, J.P. Callan, E. Mazur, Three-
dimensional optical storage inside transparent materials. Opt. Lett. 21, 2023–2025 (1996)
26. E.N. Glezer, E. Mazur, Ultrafast-laser driven micro-explosions in transparent materials. Appl.
Phys. Lett. 71, 882–884 (1997)
27. C.B. Schaffer, A. Brodeur, J.F. Garcia, E. Mazur, Micromachining bulk glass by use of
femtosecond laser pulses with nanojoule energy. Opt. Lett. 26, 93–95 (2001)
28. A. Zoubir, C. Lopez, M. Richardson, K. Richardson, Femtosecond laser fabrication of tubular
waveguides in poly(methyl methacrylate). Opt. Lett. 29, 1840–1842 (2004)
29. B.N. Chichkov, C. Momma, S. Nolte, F. von Alvensleben, A. Tunnermann, Femtosecond,
picosecond and nanosecond laser ablation of solids. Appl. Phys. A – Mater 63, 109–115
(1996)
30. S. Preuss, A. Demchuk, M. Stuke, Subpicosecond UV laser-ablation of metals. Appl. Phys.
A – Mater 61, 33–37 (1995)
31. S. Preuss, E. Matthias, M. Stuke, Subpicosecond UV-laser ablation of Ni-films – strong flu-
ence reduction and thickness-independent removal. Appl. Phys. A – Mater 59, 79–82 (1994)
12 Femtosecond Laser Micromachining 317
32. N.W. Ashcroft, N.D. Mermin, Solid State Physics (Holt Rinehart and Winston,
New York, 1976)
33. P.Y Yu, M. Cardona, in Fundamentals of Semiconductors: Physics and Materials Properties,
3rd, rev. and enlarged edn (Springer, Berlin/New York, 2001)
34. C.B. Schaffer, A. Brodeur, E. Mazur, Laser-induced breakdown and damage in bulk transpar-
ent materials using tightly-focused femtosecond laser pulses. Meas. Sci. Technol. 12, 1784
(2001)
35. L.V. Keldysh, Ionization in the field of a strong electromagnetic wave. Sov. Phys. JETP –
USSR 20, 1307 (1965)
36. B.C. Stuart, M.D. Feit, S. Herman, A.M. Rubenchik, B.W. Shore, M.D. Perry, Nanosecond-
to-femtosecond laser-induced breakdown in dielectrics. Phys. Rev. B 53, 1749–1761 (1996)
37. A. Prokhorov, V. Konov, I. Ursu, I. Mihailescu, Laser Heating of Metals (Adam Hilger,
Bristol, 1990)
38. S.I. Anisimov, B.L. Kapeliovich, T.L. Perelman, Electron-emission from surface of metals
induced by ultrashort laser pulses. Zh. Eksp. Teor. Fiz. C66, 776–781 (1974)
39. M.I. Kaganov, I.M. Lifshitz, L.V. Tanatarov, Relaxation between electrons and the crystalline
lattice. Sov. Phys. JETP – USSR 4, 173–178 (1957)
40. P.B. Allen, Theory of thermal relaxation of electrons in metals. Phys. Rev. Lett. 59,
1460–1463 (1987)
41. L. Jiang, H.L. Tsai, Improved two-temperature model and its application in ultrashort laser
heating of metal films. J. Heat Trans. – T ASME 127, 1167–1173 (2005)
42. P.B. Corkum, F. Brunel, N.K. Sherman, T. Srinivasanrao, Thermal response of metals to
ultrashort-pulse laser excitation. Phys. Rev. Lett. 61, 2886–2889 (1988)
43. S.D. Brorson, A. Kazeroonian, J.S. Moodera, D.W. Face, T.K. Cheng, E.P. Ippen,
M.S. Dresselhaus, G. Dresselhaus, Femtosecond room-temperature measurement of the
electron-phonon coupling constant-lambda in metallic superconductors. Phys. Rev. Lett. 64,
2172–2175 (1990)
44. X.Y. Wang, D.M. Riffe, Y.S. Lee, M.C. Downer, Time-resolved electron-temperature mea-
surement in a highly excited gold target using femtosecond thermionic emission. Phys. Rev.
B 50, 8016–8019 (1994)
45. M.B. Agranat, A.A. Benditskii, G.M. Gandel’man, P.S. Kondratenko, B.I. Makshantsev, G.I.
Rukman, B.M. Stepanov, Inertialess metal glow produced by picosecond pulses. Sov. Phys. –
JETP 52, 27–31 (1980)
46. M. Agranat, S.I. Anisimov, S. Ashitkov, B. Makshantsev, I. Ovchinnikova, Thermal radiation
emitted by metals due to disturbance of an equilibrium between electrons and the lattice. Sov.
Phys. Solid State 29, 1875–1880 (1987)
47. S. Nolte, C. Momma, H. Jacobs, A. Tunnermann, B.N. Chichkov, B. Wellegehausen, H.
Welling, Ablation of metals by ultrashort laser pulses. J. Opt. Soc. Am. B 14, 2716–2722
(1997)
48. A.P. Kanavin, I.V. Smetanin, V.A. Isakov, Y.V. Afanasiev, B.N. Chichkov, B. Wellegehausen,
S. Nolte, C. Momma, A. Tunnermann, Heat transport in metals irradiated by ultrashort laser
pulses. Phys. Rev. B 57, 14698–14703 (1998)
49. A.A. Abrikosov, Fundamentals of the Theory of Metals (Elsevier, Amsterdam, 1988)
50. J. Konig, S. Nolte, A. Tunnermann, Plasma evolution during metal ablation with ultrashort
laser pulses. Opt. Express 13, 10597–10607 (2005)
51. A.M. Komashko, M.D. Feit, A.M. Rubenchik, M.D. Perry, P.S. Banks, Simulation of material
removal efficiency with ultrashort laser pulses. Appl. Phys. A – Mater 69, S95–S98 (1999)
52. B.C. Stuart, M.D. Feit, S. Herman, A.M. Rubenchik, B.W. Shore, M.D. Perry, Optical ablation
by high-power short-pulse lasers. J. Opt. Soc. Am. B 13, 459–468 (1996)
53. K. Furusawa, K. Takahashi, H. Kumagai, K. Midorikawa, M. Obara, Ablation characteristics
of Au, Ag, and Cu metals using a femtosecond Ti:sapphire laser. Appl. Phys. A – Mater 69,
S359–S366 (1999)
54. D. Du, X. Liu, G. Korn, J. Squier, G. Mourou, Laser-induced breakdown by impact ionization
in SiO2 with pulse widths from 7 ns to 150 fs. Appl. Phys. Lett. 64, 3071–3073 (1994)
318 L. Cerami et al.
55. A. Luft, U. Franz, A. Emsermann, J. Kaspar, A study of thermal and mechanical effects on
materials induced by pulsed laser drilling. Appl. Phys. A – Mater 63, 93–101 (1996)
56. C. Momma, B.N. Chichkov, S. Nolte, F. von Alvensleben, A. Tunnermann, H. Welling,
B. Wellegehausen, Short-pulse laser ablation of solid targets. Opt. Commun. 129, 134–142
(1996)
57. D. Baeuerle, Laser Processing and Chemistry (Springer, Berlin/Heidelberg, 1996)
58. K. Miura, J.R. Qiu, H. Inouye, T. Mitsuyu, K. Hirao, Photowritten optical waveguides in
various glasses with ultrashort pulse laser. Appl. Phys. Lett. 71, 3329–3331 (1997)
59. L. Sudrie, M. Franco, B. Prade, A. Mysyrewicz, Writing of permanent birefringent microlay-
ers in bulk fused silica with femtosecond laser pulses. Opt. Commun. 171, 279–284 (1999)
60. L. Sudrie, M. Franco, B. Prade, A. Mysyrowicz, Study of damage in fused silica induced by
ultra-short IR laser pulses. Opt. Commun. 191, 333–339 (2001)
61. C.B. Schaffer, A.O. Jamison, E. Mazur, Morphology of femtosecond laser-induced structural
changes in bulk transparent materials. Appl. Phys. Lett. 84, 1441–1443 (2004)
62. J.W. Chan, T.R. Huser, S. Risbud, D.M. Krol, Structural changes in fused silica after exposure
to focused femtosecond laser pulses. Opt. Lett. 26, 1726–1728 (2001)
63. J.W. Chan, T.R. Huser, S.H. Risbud, J.S. Hayden, D.M. Krol, Waveguide fabrication in
phosphate glasses using femtosecond laser pulses. Appl. Phys. Lett. 82, 2371–2373 (2003)
64. J.W. Chan, T.R. Huser, S.H. Risbud, D.M. Krol, Modification of the fused silica glass network
associated with waveguide fabrication using femtosecond laser pulses. Appl. Phys. A – Mater
A76, 367–372 (2003)
65. R. Bruckner, Properties and structure of vitreous silica. I. J. Non-Cryst. Solids 5, 123–175
(1970)
66. R. Bruckner, Properties and structure of vitreous silica. II. J. Non-Cryst. Solids 5, 177–216
(1971)
67. D. Homoelle, S. Wielandy, A.L. Gaeta, N.F. Borrelli, C. Smith, Infrared photosensitivity in
silica glasses exposed to femtosecond laser pulses. Opt. Lett. 24, 1311–1313 (1999)
68. M. Will, S. Nolte, B.N. Chichkov, A. Tunnermann, Optical properties of waveguides
fabricated in fused silica by femtosecond laser pulses. Appl. Optics 41, 4360–4364 (2002)
69. S. Nolte, J. Burghoff, M. Will, A. Tuennermann, Femtosecond writing of high-quality
waveguides inside phosphate glass and crystalline media using a bifocal approach. Proc. SPIE
5340, 164 (2004)
70. Y. Shimotsuma, P.G. Kazansky, Q. Jiarong, K. Hirao, Self-organized nanogratings in glass
irradiated by ultrashort light pulses. Phys. Rev. Lett. 91, 247405 (2003)
71. E. Bricchi, B.G. Klappauf, P.G. Kazansky, Form birefringence and negative index change
created by femtosecond direct writing in transparent materials. Opt. Lett. 29, 119–121 (2004)
72. R. Taylor, C. Hnatovsky, E. Simova, Applications of femtosecond laser induced self-
organized planar nanocracks inside fused silica glass. Laser Photonics Rev. 2, 26–46 (2008)
73. J.R. Qiu, K. Miura, K. Hirao, Three-dimensional optical memory using glasses as a recording
medium through a multi-photon absorption process. Jpn. J. Appl. Phys. 1(37), 2263–2266
(1998)
74. T. Gorelik, M. Will, S. Nolte, A. Tuennermann, U. Glatzel, Transmission electron microscopy
studies of femtosecond laser induced modifications in quartz. Appl. Phys. A – Mater A76,
309–311 (2003)
75. V.R. Bhardwaj, E. Simova, P.B. Corkum, D.M. Rayner, C. Hnatovsky, R.S. Taylor, B.
Schreder, M. Kluge, J. Zimmer, Femtosecond laser-induced refractive index modification in
multicomponent glasses. J. Appl. Phys. 97, 083102 (2005)
76. A.M. Streltsov, N.F. Borrelli, Study of femtosecond-laser-written waveguides in glasses. J.
Opt. Soc. Am. B 19, 2496–2504 (2002)
77. M. Ams, G.D. Marshall, D.J. Spence, M.J. Withford, Slit beam shaping method for femtosec-
ond laser direct-write fabrication of symmetric waveguides in bulk glasses. Opt. Express 13,
5676–5681 (2005)
78. R. Osellame, S. Taccheo, M. Marangoni, R. Ramponi, P. Laporta, D. Polli, S. De Silvestri, G.
Cerullo, Femtosecond writing of active optical waveguides with astigmatically shaped beams.
J. Opt. Soc. Am. B 20, 1559–1567 (2003)
12 Femtosecond Laser Micromachining 319
79. Y.R. Shen, The Principles of Nonlinear Optics (Wiley, New York, 1984)
80. P. Simon, J. Ihlemann, Machining of submicron structures on metals and semiconductors by
ultrashort UV-laser pulses. Appl. Phys. A – Mater 63, 505–508 (1996)
81. D. Breitling, C. Fohl, F. Dausinger, T. Kononenko, V. Konov, Drilling of metals, in
Femtosecond Technology for Technical and Medical Applications. Topics in Applied Physics,
vol 96. Springer, pp. 131–154 (2004)
82. X. Liu, Industrial applications of ultrahigh precision short-pulse laser processing. Proc. SPIE
5713, 372–386 (2005)
83. H. Booth, Laser processing in industrial solar module manufacturing. J. Laser Mi-
cro/Nanoeng. 5, 183–191 (2010)
84. A. Schoonderbeek, V. Schutz, O. Haupt, U. Stute, Laser processing of thin films for
photovoltaic applications. J. Laser Micro/Nanoeng. 5, 248–255 (2010)
85. P. Gecys, G. Raciukaitis, M. Ehrhardt, K. Zimmer, M. Gedvilas, ps-laser scribing of CIGS
films at different wavelengths. Appl. Phys. A – Mater 101, 373–378 (2010)
86. T. Abeln, U. Klink, Laserstrukturieren zur Verbesserung der tribologischen Eigenschaften von
Oberflächen, in Konferenz Stuttgarter Lasertage (2001), pp. 61–64
87. M. Weikert, F. Dausinger, Surface structuring, in Femtosecond Technology for Technical and
Medical Applications. Topics in Applied Physics, vol 96. Springer, pp. 117–129 (2004)
88. J. Burghoff, C. Grebing, S. Nolte, A. Tunnermann, Efficient frequency doubling in femtosec-
ond laser-written waveguides in lithium niobate. Appl. Phys. Lett. 89, 081108 (2006)
89. J. Burghoff, S. Nolte, A. Tuennermann, Origins of waveguiding in femtosecond laser-
structured LiNbO3 . Appl. Phys. A – Mater 89, 127–132 (2007)
90. J.R. Liu, Z.Y. Zhang, C. Flueraru, X.P. Liu, S.D. Chang, C.P. Grover, Waveguide shaping and
writing in fused silica using a femtosecond laser. IEEE J. Sel. Top. Quant. 10, 169–173 (2004)
91. Y. Nasu, M. Kohtoku, Y. Hibino, Low-loss waveguides written with a femtosecond laser for
flexible interconnection in a planar light-wave circuit. Opt. Lett. 30, 723–725 (2005)
92. R.R. Thomson, H.T. Bookey, N. Psaila, S. Campbell, D.T. Reid, S.X. Shen, A. Jha, A.K. Kar,
Internal gain from an erbium-doped oxyfluoride-silicate glass waveguide fabricated using
femtosecond waveguide inscription. IEEE Photonic. Tech. L. 18, 1515–1517 (2006)
93. R. Osellame, M. Lobino, N. Chiodo, M. Marangoni, G. Cerullo, R. Ramponi, H.T. Bookey,
R.R. Thomson, N.D. Psaila, A.K. Kar, Femtosecond laser writing of waveguides in periodi-
cally poled lithium niobate preserving the nonlinear coefficient. Appl. Phys. Lett. 90, 241107
(2007)
94. T. Calmano, J. Siebenmorgen, O. Hellmig, K. Petermann, G. Huber, Nd:YAG waveguide
laser with 1.3 W output power, fabricated by direct femtosecond laser writing. Appl. Phys.
B – Lasers O 100, 131–135 (2010)
95. D. Ehrt, T. Kittel, M. Will, S. Nolte, A. Tunnermann, Femtosecond-laser-writing in various
glasses. J. Non-Cryst. Solids 345, 332–337 (2004)
96. M. Hughes, W. Yang, D. Hewak, Fabrication and characterization of femtosecond laser
written waveguides in chalcogenide glass. Appl. Phys. Lett. 90, 131113 (2007)
97. R. Osellame, N. Chiodo, G. Della Valle, S. Taccheo, R. Ramponi, G. Cerullo, A. Killi, U.
Morgner, M. Lederer, D. Kopf, Optical waveguide writing with a diode-pumped femtosecond
oscillator. Opt. Lett. 29, 1900–1902 (2004)
98. R. Osellame, S. Taccheo, G. Cerullo, M. Marangoni, D. Polli, R. Ramponi, P. Laporta, S. De
Silvestri, Optical gain in Er-Yb doped waveguides fabricated by femtosecond laser pulses.
Electron. Lett. 38, 964–965 (2002)
99. A. Baum, P.J. Scully, M. Basanta, C.L.P. Thomas, P.R. Fielden, N.J. Goddard, W. Perrie,
P.R. Chalker, Photochemistry of refractive index structures in poly(methyl methacrylate) by
femtosecond laser irradiation. Opt. Lett. 32, 190–192 (2007)
100. E. Mazur, C.R. Mendonca, L.R. Cerami, T. Shih, R.W. Tilghman, T. Baldacchini, Femtosec-
ond laser waveguide micromachining of PMMA films with azoaromatic chromophores. Opt.
Express 16, 200–206 (2008)
101. S. Sowa, W. Watanabe, T. Tamaki, J. Nishii, K. Itoh, Symmetric waveguides in poly(methyl
methacrylate) fabricated by femtosecond laser pulses. Opt. Express 14, 291–297 (2006)
320 L. Cerami et al.
121. T. Baldacchini, C.N. LaFratta, R.A. Farrer, M.C. Teich, B.E.A. Saleh, M.J. Naughton, J.T.
Fourkas, Acrylic-based resin with favorable properties for three-dimensional two-photon
polymerization. J. Appl. Phys. 95, 6072–6076 (2004)
122. P.R. Willmott, J.R. Huber, Pulsed laser vaporization and deposition. Rev. Mod. Phys. 72,
315–328 (2000)
123. M. Womack, M. Vendan, P. Molian, Femtosecond pulsed laser ablation and deposition of thin
films of polytetrafluoroethylene. Appl. Surf. Sci. 221, 99–109 (2004)
124. S. Eliezer, N. Eliaz, E. Grossman, D. Fisher, I. Gouzman, Z. Henis, S. Pecker, Y. Horovitz,
M. Fraenkel, S. Maman, Y. Lereah, Synthesis of nanoparticles with femtosecond laser pulses.
Phys. Rev. B 69, 144119 (2004)
125. S. Amoruso, G. Ausanio, R. Bruzzese, M. Vitiello, X. Wang, Femtosecond laser pulse
irradiation of solid targets as a general route to nanoparticle formation in a vacuum. Phys.
Rev. B 71, 033406 (2005)
126. R. Haight, P. Longo, A. Wagner, Metal deposition with femtosecond light pulses at atmo-
spheric pressure. J. Vac. Sci. Technol. A 21, 649–652 (2003)
Chapter 13
Ultrafast Laser Inscription of Photonic Devices
in Bulk Dielectrics
13.1 Introduction
Over the last decade a powerful new photonic device fabrication technology
has emerged. In this chapter, we shall refer to this technology as ultrafast laser
inscription (ULI), but it is also widely known by a variety of other names, including
femtosecond-laser direct-writing. ULI uses ultrashort laser pulse to directly write,
or inscribe, structures inside transparent materials, and it uniquely provides the
possibility of fabricating micro-optics [1], micro-fluidics [2], micro-mechanics [3]
and optical waveguides [4] all on a single substrate. This can be achieved in a single
ULI process, without removing the substrate from the inscription rig. As a result, all
the components on the substrate may, in principle, be aligned to within the positional
accuracy of the ULI rig, which can be in the region of a few nm. The aim of this
chapter is to introduce the reader to the technology of ULI, the techniques used for
ULI and some of the applications of ULI that are emerging.
It is logical to start this chapter by briefly describing the basic ULI experimental
setup and process. As shown in Fig. 13.1, a typical ULI rig consists of an ultrafast
laser system. For ULI, pulses with temporal duration ranging between 50 fs and
1.0 ps have been successfully used and some laser systems are supplied with a
Fig. 13.1 Schematic diagram of a typical ULI rig. S shutter, HWP half wave plate, PBS
polarisation beam splitter, BD beam dump, BS beam splitter, D detector, AOM acousto-optic
modulator, QWP quarter wave plate, L lens, SUB substrate material, TS translation stages
compressor which allows the pulse duration to be varied. A shutter is used to control
the on/off state of the laser on the substrate, a half wave plate and polarising beam
splitter are used to control the laser power incident on the substrate, a beam splitter
and detector are used to monitor this power, an acousto-optic modulator can be used
to rapidly modulate the laser power during the inscription and a quarter wave plate
and half wave plate pair are used to control the polarisation state of the laser light
incident on the substrate. Once the power and polarisation properties of the laser
beam have been defined, the laser light is coupled into a lens and the pulses are
focused inside the substrate material which is mounted on a set of high precision
x–y–z translation stages. A computer is used to control the trajectory and speed of
the substrate translation, and synchronise the sample movement with other functions
described above.
To perform ULI, the substrate material and laser wavelength are chosen such that
the photon energy (Ep D h, where h D Planck’s constant and v D the oscillation
frequency of the light) is insufficient to excite an electron across the material band-
gap. Under these conditions, the laser radiation would normally travel through
the material without absorption. However, due to the high peak powers facilitated
by ultrashort pulse durations, the electric fields (E-fields) created in the focus
can be sufficient to induce nonlinear light-matter interaction phenomena which
initiate the generation of a free electron plasma [5]. If the irradiation conditions are
appropriately controlled, this plasma can be highly confined to the region directly
in and around the laser focal volume. After the pulse has left the focal volume,
the plasma transfers energy to the lattice, inducing a localised, and in many cases
permanent, modification to the structure of the irradiated material.
13 Ultrafast Laser Inscription of Photonic Devices in Bulk Dielectrics 325
The ULI induced modification can manifest itself in a variety of ways, examples
of which include changes to the refractive index [4, 6] or chemical etch-rate
[7, 8] of the modified material. Since the light-matter interaction is nonlinear,
only occurring in the focal region where the E-field intensity is sufficient, lines of
modified material can be inscribed inside the bulk of the material by translating it in
three-dimensions through the laser focus. The manifestations of the ULI induced
structural modification can thus be used to directly inscribe three-dimensional
components such as optical waveguides, microfluidics and lenses.
In any dielectric material at absolute zero, there is an energy bandgap (Eg ) between
the highest filled electron energy level (the valence band) and the next available level
(the conduction band). If a single photon of light is to excite an electron across the
band gap, its energy must be greater than or equal to this gap. As mentioned in the
previous section, ULI is performed by choosing the laser wavelength and substrate
material such that this condition is not met. However, due to the high peak powers of
ultrashort pulses, the E-field in the focal region can be sufficient to excite electrons
across the energy bandgap via nonlinear excitation processes such as tunnelling
and multi-photon ionisation [5, 9, 10]. During tunnelling ionisation (Fig. 13.2a),
the E-field of the laser suppresses the coulomb potential that binds the electron to
the nucleus. The electron is then able to tunnel through the lowered barrier into the
conduction band with a probability which increases with increasing E-field strength.
Electrons may also be excited through multi-photon ionisation (Fig. 13.2b), during
which n photons interact simultaneously to promote an electron across the band gap.
Fig. 13.2 Nonlinear photoionisation processes responsible for transferring energy from the
ultrashort laser pulse to the material. (a) Tunnelling ionisation (b) Multi-photon ionisation (c)
Avalanche ionisation (Reproduced and adapted with permission from Ref. [11])
326 R.R. Thomson and G. Cerullo
Since all the photons must essentially be coincident on the electron, the probability
of multi-photon ionisation taking place is strongly dependent on the E-field strength.
Electrons may also be excited through a combination of multi-photon and tunnelling
ionisation.
Once an electron is excited to the bottom of the conduction band, it has a
continuum of states available above it, and it repeatedly absorbs single photons of
light through phonon mediated (to conserve momentum) absorption. The excited
electron absorbs photons until it has sufficient excess energy to impact ionise a
second electron residing at the top of the valence band. This ionisation process,
which is known as avalanche ionisation (Fig. 13.2c), repeats continually while the
laser field is present and results in an exponential increase in the density of the free
electron plasma while the laser field is present.
The free electron plasma created by the pulse exhibits a resonant plasma
frequency given by Eq. (13.1), which increases with increasing free electron density.
The plasma frequency therefore increases until it matches the laser frequency.
At this point the laser resonantly drives the plasma oscillation and a large fraction
of the rest of the pulse is absorbed.
s
N e2
!p D (13.1)
"0 m
Once the pulse has left the focal volume, the free electron plasma begins to
transfer energy to the cold lattice. This transfer begins on the picosecond timescale
through electron-ion collisions until the electronic and atomic systems reach thermal
equilibrium at about 105 K, and on the nanosecond timescale due to recombination
of the electrons with the ions [12]. Since these processes only begin on the
picosecond timescale, the heating of the lattice during the pulse is negligible for
sub-picosecond pulses. Ultrashort pulse irradiation therefore induces a highly non-
equilibrium situation where a hot dense free electron plasma is created within a
relatively cold lattice. This minimises thermal diffusion out of the focal volume
during the pulse and enables the material to be heated efficiently. This in turn
minimises the amount of energy required to modify the material, localising any
modification to the immediate focal region and minimising collateral damage.
The transfer of energy from the plasma to the lattice rapidly heats the material. This
heating can induce either no modification to the lattice, locally melt the lattice or
drive it into a gas or plasma phase. These phase changes may also be accompanied
by shock waves and high thermal gradients. Heat then diffuses out of the focal region
on the microsecond timescale [13] and the material cools and re-solidifies. After re-
solidification the irradiated material may exhibit a variety of structural changes.
13 Ultrafast Laser Inscription of Photonic Devices in Bulk Dielectrics 327
Thus far, it has been shown that there are three qualitatively different modifica-
tion regimes in glasses and crystals, examples of which are shown in Fig. 12.3.
When using pulse energies only slightly above the modification threshold a
smoothly varying change in the material structure may be observed (Fig. 12.3
(left)) [14]. This modification is characterised by a similarly smooth variation in the
refractive index across the modified region which can be either positive, negative or
a combination of both. The origin of the refractive index modification depends on
the substrate material and on the ULI parameters, but it may have contribution from
material densification, stress, chemical bond changes, colour centre formation and
photorefractivity. If slightly higher pulse energies are used, self-organised nano-
gratings can be formed in a number of materials (Fig. 12.3 (centre)) [15]. The
nano-gratings are always aligned perpendicular to the polarisation of the linearly
polarised incident light and may be either periodic variations in the stoichiometry
of the material or nano-cracks. The physical mechanisms responsible for the nano-
grating formation are still the subject of some debate and two possible explanations
have been suggested. Shimotsuma et al. [15] have suggested they form due to
interference between the incident light field and a density wave travelling through
the free electron plasma, whereas Taylor et al. [16] have suggested a mechanism
involving the growth of disc-shaped nano-plasmas and that the periodic nature is due
to waveguiding effects by these nano-plasmas. Finally, if even higher pulse energies
are used, the lattice itself can be transformed into a gas or plasma. The pressure
inside the irradiated volume can be sufficient to induce a micro-explosion, during
which hot material is driven into the surrounding cold material. The final result is
to leave a void or voxel of material that exhibits a significantly reduced density, as
shown in Fig. 12.3 (right) [6].
From the preceding section it is clear that ultrashort laser pulses can induce a variety
of localised structural modifications inside a dielectric material. From a photonic
device fabrication point of view, however, it is the manifestations of these structural
changes that is of most interest. To date, it has been shown that ULI can be used to
modify the thermal conductivity [17], Young’s modulus [18] and optical absorption
[19] characteristics of the material. It is, however, ULI’s ability to locally modify the
refractive index [4, 6] and chemical etch rate [2, 7] of materials which has attracted
most attention for photonic device fabrication.
The ULI induced refractive index modification has been attributed to many different
mechanisms, including density changes [20], colour centre formation [21] and
photorefractivity [22]. The substrate material and ULI parameters determine the
328 R.R. Thomson and G. Cerullo
Fig. 13.3 (a) Optical micrograph of the end facet of a ULI fabricated Type I optical waveguide in
a silicate glass. (b) Near field image of the 1,550 nm mode guided by the waveguide shown in (a).
(c) Optical micrograph of the end facet of a ULI fabricated double line Type II optical waveguide
in YAG ceramic. (d) Near field image of the 633 nm mode guided by the structure shown in (c).
The scale bar in (c) represents 20 m and is common to (d) (Figures (a) and (b) reproduced and
adapted with permission from Ref. [25], Figs. (c) and (d) reproduced and adapted with permission
from Ref. [28])
role, if any, that each mechanism plays. In certain polymer and crystalline materials
and many glass materials it is possible to directly modify a region of material that
then exhibits a net positive refractive index change. This is loosely known as a Type
I refractive index modification and it can be used to fabricate diffraction gratings
[23], Fresnel lenses [24] and most notably optical waveguides [4]. Figure 13.3a
presents an optical micrograph of the end facet of a Type I optical waveguide in
a silicate glass material. Figure 13.3b presents a near field image of the 1,550 nm
mode guided by the same waveguide [25].
In some materials, particularly those that are single- or polycrystalline, it may not
be possible to readily induce a Type I refractive index modification, perhaps since
the crystal structure usually minimises the free energy of the constituent atoms.
13 Ultrafast Laser Inscription of Photonic Devices in Bulk Dielectrics 329
One of the most unique and interesting capabilities of ULI is its ability to locally
increase the chemical etch rate of certain materials. This effect has only been
demonstrated in a few substrate materials, examples of which include fused silica
[2, 7] and Foturan glass [1], and the mechanisms by which the chemical etch rate is
increased is highly dependent on the substrate material.
In fused silica, Bellouard et al. have tentatively suggested that the etch rate
increase is due to the generation of planes of material that exhibit a reduction in the
Si–O–Si bond angle, and that these bonds are more readily broken by hydrofluoric
(HF) acid [17]. Hnatovsky et al. [8] have proposed a second mechanism involving
the generation of nano-cracks which enable HF acid to penetrate more easily into
the modified material than into the virgin material. Regardless of the true physical
mechanisms by which the etch rate is increased, ULI can be used to locally increase
the HF acid etch rate of fused silica by a factor of around 100. This incredible
capability has been used to fabricate structures such as micro-mechanics [3] and
micro-fluidics [31], as shown in Fig. 13.4.
ULI can also be used to locally increased the chemical etch rate of Foturan, a
photosensitive silicate glass supplied by Schott which contains cerium and silver
in the form of Ce3C and AgC ions. Although an investigation into the exact
mechanisms by which ULI increases in the chemical etch rate of Foturan is still
to be conducted, the mechanisms by which UV light induces an etch rate increase
in Foturan have been studied extensively and the mechanisms are expected to be
similar for ULI. In the case of UV light exposure, the high energy photons convert
330 R.R. Thomson and G. Cerullo
Fig. 13.4 Scanning electron microscope images of (a) a selectively etched fused silica flexure
hinge and (b) a selectively etched subsurface micro-fluidic channel in fused silica (Figures (a) and
(b) are reproduced and adapted with permission from Refs. [3] and [31] respectively)
Fig. 13.5 SEM images of a micro-optical cylindrical lens fabricated from Foturan glass using (a)
ULI induced selective etching only and (b) selective etching and a post-etch annealing process
(Both images are reproduced and adapted with permission from Ref. [1])
some of the Ce3C ions to Ce4C ions. This generates trapped electrons which, through
a controlled heat treatment up to 500 ı C, combine with the AgC ions to create
Ag atoms. These atoms then agglomerate to form silver nano-clusters which act as
nucleation sites for the growth of a lithium metasilicate crystalline phase (Li2 SiO3 )
when the substrate temperature is raised further to 600 ı C. This crystalline phase
exhibits an HF acid etch rate approximately 50 times higher than the original
Foturan material.
Unlike fused silica, which has a melting point above 1,500 ı C, Foturan has
a relatively low melting point of around 600 ı C. As a result, a relatively low-
temperature post-etch annealing process can be readily performed on selectively
etched Foturan structures to smooth out high spatial frequency surface roughness.
Figure 13.5 shows two SEM images of a micro-optical cylindrical lens that has been
selectively etched from Foturan [1]. The image on the left shows the lens before the
post-etch annealing step, the image on the right shows the lens after the annealing.
13 Ultrafast Laser Inscription of Photonic Devices in Bulk Dielectrics 331
From these images, it is clear that the annealing has significantly smoothed the
surface of the lens, removing the periodic lines that are characteristic of the ULI line-
by-line rastering process. This post-etch annealing step has been shown to enable the
fabrication of surfaces with a remarkable sub-nm surface roughness [32]. It should
also be mentioned that it has also recently been shown that high temperature flame
brushing can also be used to smoothen the surfaces of selectively etched fused silica
structures [33].
The ULI process is controlled via the ULI parameters. These include those
associated with the ultrafast laser itself, such as laser wavelength, polarisation, pulse
repetition frequency (PRF), pulse duration, pulse energy, pulse front tilt, as well as
others, such as substrate translation velocity, translation direction, focusing optics
and substrate material. All have been shown to play a role with varying degrees of
importance.
Generally speaking, the “optimum” ULI parameters for a given job are found
using a quasi trial-and-error approach, where structures are repeatedly fabricated
and tested, and with each cycle the parameters are improved. The starting parameters
for this process are of course made based on previous knowledge of how each
parameter may affect the ULI process. Although an extensive review of the effect of
each parameter on the ULI process would be outside the scope of this chapter, it is
necessary to at least briefly review the roles that the most important parameters play.
Due to its importance, a short section is included at the end of this part of the chapter
which focuses on how the cross-section of ULI fabricated optical waveguides may
be controlled.
During ULI, the substrate can either be translated primarily along, or transverse, to
the direction of propagation of the inscription laser beam – as shown in Fig. 12.7.
These writing geometries are known as the longitudinal and transverse writing
geometries (LWG & TWG) respectively. Optical waveguide inscription using the
LWG has the advantage that the cross section of the inscribed waveguide is
inherently circular due to the rotational symmetry of the inscription laser beam, but
has the disadvantage that the maximum waveguide length is limited by the working
distance of the lens. Furthermore, due to the depth dependent degree of spherical
aberration imparted on the laser beam by the substrate-air interface, the properties
of the ULI induced modification vary as a function of depth unless steps are taken
to correct for this [34]. For these reasons, the LWG has attracted progressively less
attention. In contrast, the TWG is extremely flexible and enables the inscription of
332 R.R. Thomson and G. Cerullo
structures with effectively arbitrary path and length inside the substrate. Unlike the
LWG however, the cross section of the modified region may be highly asymmetric
as discussed in the following section. This is particularly problematic for optical
waveguide inscription since an asymmetric waveguide core results in waveguide
birefringence and asymmetric waveguide modes.
The focusing optics control the E-field distribution in and around the ULI focal
region, and hence the spatial distribution of the photo-induced free electron plasma
[35]. If thermal accumulation and thermal diffusion play a minimal role (see next
section for a discussion of these effects), and if beam distortion effects due to
nonlinear refraction are ignored, the dimensions of the plasma created by each pulse
are directly related to the confocal parameter (Eq. 13.2) and beam waist diameter
(Eq. 13.3) of the focused beam inside the sample. In Eqs. (13.2) and (13.3), b is the
confocal parameter, n is the refractive index of the substrate at the laser wavelength,
D and ! 0 are the focused beam waist diameter and radius respectively, is the free
space wavelength and NA is the numerical aperture of the focusing.
2 n !02
bD (13.2)
2
D D 2 !0 D (13.3)
NA
For structures written using the LWG, the diameter of the inscribed structure is
directly related to the focused beam waist diameter. However, due to its flexibility,
it is the TWG that is of most interest. In this case, the dimensions of the structure
cross section are directly related to both the confocal parameter and the beam
waist diameter inside the material. An inspection of Eqs. (13.2) and (13.3) indicate
that when using the TWG the cross section of the modified region will be highly
asymmetric unless the NA of the focusing is close to n. Such tight focusing is, in
general, impractical for ULI applications and numerical apertures in the region of
0.6 are generally used. At the end of this section we review how the cross section of
ULI fabricated optical waveguides can be controlled.
Since ULI relies on the nonlinear absorption of light, the laser wavelength deter-
mines the number of photons required to overcome the substrate material bandgap.
The inscription laser wavelength that can be used is therefore set by the substrate
13 Ultrafast Laser Inscription of Photonic Devices in Bulk Dielectrics 333
material. For example, in order to perform ULI in silicon, the laser wavelength must
be significantly longer than the band edge of silicon (1.1 m). If the wavelength
of the laser system is not immediately suited to the application, its wavelength
can be shifted using nonlinear techniques such as second harmonic generation or
parametric amplification. For example, by pumping an optical parametric amplifier
using a Ti:sapphire laser operating at 810 nm, Nejadmalayeri et al. [36] generated
ultrashort pulses of 2.4 m light to inscribe waveguides in crystalline silicon.
Frequency conversion can also be used to enhance the absorption of the laser
radiation. For example, Shah et al. [37] generated ultrashort pulses of 522 nm light
by frequency doubling the 1,045 nm light from a Yb-doped fibre laser. Waveguides
inscribed in fused silica using the 522 nm light were found to guide infra-red light
significantly better than those inscribed using the 1,045 nm light – presumably due
to the enhanced absorption of the shorter wavelength light.
The pulse repetition frequency (PRF) is one of the most important ULI parameters
due to the role of thermal accumulation. Since heat requires 1.0 s to diffuse
out of the focal region [13], the use of laser systems operating at PRFs greater
than 1.0 MHz prevents the material in the focal volume from cooling completely
between pulses. The temperature of the irradiated material may therefore be driven
significantly higher when using high PRFs – even though the total energy fluence
on the sample may be the same.
To demonstrate the importance of heat accumulation and thermal diffusion in
ULI, Eaton et al. [38] used a finite difference model to model the effect of PRF on
the temperature in the focal region when performing ULI in borosilicate glass. The
results of this simulation are presented in Fig. 13.6 where it can be seen that at a PRF
of 100 kHz the temperature of the material is continually cycled around the material
working point of 1,225 ı C. When the PRF is raised to 500 kHz and then 1.0 MHz,
thermal accumulation is found to play an increasing role and the temperature of the
material increases significantly.
Since heat diffuses out of the focal region equally in all directions, structures
inscribed in the thermal accumulation regime may exhibit a considerably more
symmetric cross section than would be expected from the theoretical E-field
distribution in the focal region. A further effect of thermal accumulation and thermal
diffusion is to increase the size of the modified region cross section. As an example,
Fig. 13.7 presents optical micrographs of structures inscribed in borosilicate glass
using static exposures. The key point to note here is that the size of the modified
region is only weakly dependent on the exposure fluence, which is varied by
four orders of magnitude, whereas a tenfold increase in the size of the modified
region is observed when the PRF is increased from 100 kHz to 1.0 MHz. This
clearly demonstrates the role that thermal accumulation plays in determining the
temperature induced in the focal region during ULI. Finally, it should also be noted
that even when lasers operating at relatively low PRFs (200 kHz) are used, high
334 R.R. Thomson and G. Cerullo
Fig. 13.6 Modelled temperature versus exposure at a radial position of 2.0 m from the centre of
the focus. Material: AF45 borosilicate glass, absorbed pulse energy: 260 nJ, focused spot diameter:
2.0 m (Reproduced and adapted with permission from Ref. [38])
pulse energies can be used to drive strong thermal diffusion effects. This can in turn
create a structure cross section that is significantly different than would be expected
from the theoretical E-field distribution [39].
Fig. 13.7 Optical micrographs of structures inscribed in AF45 borosilicate glass using static
exposures. The 1,045 nm light was focused inside the substrate using a 1.4 NA objective lens.
Each column of structures was inscribed using the same exposure fluence, whereas each row was
inscribe using the same pulse repetition frequency (Reproduced and adapted with permission from
Ref. [38])
The substrate translation velocity is used to vary the laser fluence on the substrate –
while keeping the pulse energy and therefore the modification regime the same.
Generally speaking, a reduced translation velocity increases the magnitude of the
induced modification. If thermal accumulation and diffusion is significant in the
ULI process, the translation velocity can also be used to vary the energy fluence
on the substrate, which can in turn be used to control the size of the laser affected
zone as shown in Fig. 13.7. This effect has been shown to be particularly useful
for controlling the cross section of optical waveguides and can enable the precise
control of the guided optical mode [40].
Generally speaking, the ULI parameters that should first be chosen and optimised
are the writing geometry, focusing optics, laser wavelength, laser PRF, pulse energy,
and translation velocity, but it has also be shown that other parameters such as the
laser polarisation and pulse duration can all play important roles.
A number of studies have reported that the laser polarisation significantly affects
the magnitude of the refractive index modification that may be induced in both
amorphous and crystalline materials using ULI. In one study, Little et al. [41]
studied the effect of polarisation on the refractive index change induced in fused
336 R.R. Thomson and G. Cerullo
Fig. 13.8 (a) and (b) SEM images of ULI fabricated nano-gratings in fused silica after a weak
etch. In both (a) and (b), S represents the sample translation direction, E represents the orientation
of the E-field and k represents the direction of propagation of the incident light. (c) Optical
micrographs of sub-surface channels etched in fused silica. The top channel corresponds to (a), the
bottom channel corresponds to (b) and the channel in the middle corresponds to linear polarisation
with E at 45ı to S (All images reproduced and adapted with permission from Ref. [8])
silica using sub 120 fs pulses of 800 nm light at a PRF of 1.0 kHz. They observed
that at low irradiances linearly polarised light induces a larger refractive index
modification than circularly polarised light, and that the situation reverses at larger
irradiances. The authors suggested that at lower irradiances multi-photon ionisation
dominates over tunnelling ionisation and that linearly polarised light therefore
induces higher photoionisation frequencies than circularly polarised light. At higher
irradiances the situation may be reversed as tunnelling ionisation plays a more
significant role.
The laser polarisation has also been shown to play an important role in the
refractive index modification of crystalline materials. Nejadmalayeri et al. [42]
demonstrated that the use of circularly polarised light significantly enhanced the
magnitude of the refractive index modification that could be induced in LiNbO3 .
They suggested that the use of circular polarisation may increase the efficiency and
localisation of energy deposition into the material by reducing nonlinear absorption
and refraction effects as the pulse propagates towards the focus.
The laser polarisation can also strongly influence the magnitude of the chemical
etch-rate increase that can be induced in fused silica using ULI. As shown in
Fig. 13.8, Hnatovsky et al. [8] demonstrated that the ULI induced nano-grating
structures shown in Fig. 12.3b enable HF acid to efficiently penetrate efficiently into
the irradiated material. They also demonstrated that when ULI was used to fabricate
a subsurface micro-channel, a dramatically higher etch rate was achieved when the
nano-gratings were aligned parallel, rather than perpendicular, to the direction of
the micro-channel.
13 Ultrafast Laser Inscription of Photonic Devices in Bulk Dielectrics 337
The pulse duration has been shown to play an important role in the ULI process.
Zhang et al. [43] investigated the effect of pulse duration on the propagation loss
and birefringence of waveguides inscribed in fused silica using 800 nm pulses
at a pulse repetition frequency of 1.0 kHz. The authors varied the pulse duration
between 100 fs and 2 ps and found two windows at 100 fs and 1 ps which enabled
the inscription of waveguides with relatively low propagation losses (0.5 dB/cm).
Interestingly, they also found that waveguides inscribed using the 1 ps pulse duration
exhibited significantly reduced birefringence compared to those inscribed using
lower pulse durations. The mechanisms by which the pulse duration impacted the
properties of the waveguides are not yet understood.
The pulse duration has also been shown to play a crucial role when inscribing
waveguides in highly nonlinear materials such as LiNbO3 [42] and ZnSe [44].
Macdonald et al. [44] investigated the effect of pulse duration when inscribing
waveguides in ZnSe using the 1,047 nm pulses at a PRF of 2.0 MHz. They varied the
pulse duration between 500 fs and 2.0 ps and found that waveguides could only be
inscribed using pulse durations >1.0 ps. The authors suggested that the use of longer
pulses reduced the peak power for a given pulse energy, thus reducing nonlinear
effects (such as self focusing, pulse break up and filamentation) which reduce the
efficiency with which energy can be deposited in the focal region.
The effect of pulse duration on the chemical etch rate increase that can be induced
in fused silica using ULI has been comprehensively studied by Hnatovsky et al.
[8]. They used 800 nm pulses at a PRF of 100 kHz and observed that the pulse
duration strongly affected the magnitude of the ULI induced chemical etch rate
increase. For example, when the authors inscribed sub-surface structures using a
TWG, linearly polarised light with the E-field perpendicular to the scan direction,
and 1.0 J pulses, they observed that structures inscribed with a pulse duration of
50 fs exhibited a chemical etch rate of 2.0 m/min, whereas structures inscribed
using a pulse duration of 185 fs exhibited a chemical etch rate of 5.5 m/min.
The authors were able to demonstrate that the chemical etch rate differences were
related to the properties of the nano-grating structures formed using the two different
pulse durations. The interested reader is directed to Refs. [15] and [16] for theories
about the nano-gratings formation, and Refs. [8] and [16] which explore the effect
of nano-gratings on the chemical etch rate in fused silica.
Finally, it is also worth noting that when using the TWG, the properties of the
inscribed structure may also depend on whether the substrate is translated along
the Cve or ve direction. In crystalline LiNbO3 for example, it has been observed
that the morphology of the inscribed structure may depend on whether the sample
is translated along the Cve or ve crystallographic y-axis, and that this directional
dependence is reversed when the direction of propagation of the inscription beam
along the crystallographic z-axis is reversed [45]. It has been suggested that this
phenomena is due to a light-induced thermal current – the direction of which is
dependent on the crystal orientation with respect to the sample translation direction
and the beam propagation direction. Non-reciprocal phenomena have also been
observed in amorphous silica glass [46]. In this case, the authors demonstrated
that pulse front tilt across the beam results in an intensity gradient in the E-field
338 R.R. Thomson and G. Cerullo
across the focal region. This gradient acts, via the ponderomotive force, to trap and
displace the free-electron plasma. The non-reciprocity in the ULI process comes
from the fact the direction of the ponderomotive force is determined by the direction
of the pulse front tilt. It is reasonable therefore to expect that the properties of
the inscribed structure would depend on the relative angle between the sample
translation direction and the direction of the ponderomotive force.
The potential asymmetry of structures inscribed using the TWG can be a particular
problem for optical waveguide inscription. A number of techniques have therefore
been demonstrated which address this issue. These include the astigmatic beam
shaping [35], slit-beam shaping [47], deformable mirror [48] and multiscan [29, 49]
techniques. The first three of these techniques use beam shaping in some manner
to shape the E-field distribution in and around the focus and hence the waveguide
cross section. The multiscan technique on the other hand simply accepts that the ULI
modification created by a single scan of the material through the focus may exhibit
a high degree of asymmetry, and relies on correcting this by scanning the sample
through the focus multiple times to “build” the desired waveguide cross section
from the modified regions created by each scan. It should be noted that waveguides
inscribed using a TWG are not always asymmetric, and thermal accumulation and
diffusion effects can act to create waveguides that exhibit cross sections that are
significantly more symmetric than would be expected from the E-field distribution in
and around the focus. The techniques described in this section are therefore mostly
applicable to waveguides inscribed using low PRF laser systems.
The first technique developed to control the cross section of ULI fabricated
optical waveguides was the astigmatic beam shaping technique – first demonstrated
by Cerullo et al. [35, 50] in 2002. The idea behind the astigmatic beam shaping
technique, the experimental setup for which is shown in Fig. 13.9, is that the width of
a waveguide inscribed along the x-axis is not normally dependant on the beam waist
diameter along the x-axis. This fact can be exploited by focusing tightly in the x–z
plane to reduce the Rayleigh range of the focus in this plane. The rapid divergence of
the beam in the x–z plane reduces the depth of focus of the beam, reducing the width
of the waveguide along the z-axis. The size of the focus along the y-axis can then be
used to control the width of the waveguide along the y-axis. Although this technique
would work well in theory, in practice a beam waist ratio (! 0y /! 0x ) of 10 would
be required to achieve a symmetric waveguide cross section. This would require a
high demagnification cylindrical telescope, which would be difficult to align, would
introduce significant aberrations and would be difficult to optimise. To solve these
issues, Cerullo et al. demonstrated that by adjusting the astigmatic difference (z0 ),
the offset in the focal planes for ! 0y and ! 0x , a symmetric waveguide could be
inscribed using beam waist ratios of only 2–3. The basis of this suggestion is that
for a focused astigmatic beam, the region of highest intensity is normally positioned
13 Ultrafast Laser Inscription of Photonic Devices in Bulk Dielectrics 339
Fig. 13.10 (Left) Modelled beam profiles in the y/z plane (solid curves) and the x/z plane (dashed
curves) for focused astigmatic Gaussian beams with ! 0x D 1.0 m, ! 0y D 3.0 m and (a) Z0 D 0
(b) Z0 D 100 m. The shaded area corresponds to the waveguide shape that would be expected
according to the modelled distribution of the free electron plasma induced by the pulse. (Right) Left
column: Optical microscope images of the waveguide cross section written using an astigmatic
beam with ! 0x D 1.2 m and ! 0y D3.6 m. Right column: Corresponding simulated electron
density profiles (All images reproduced and adapted with permission from Ref. [35])
across the beam prior to entering the lens. The effect of this chirp is to reduce the
spectral bandwidth of the pulse at each point across the beam. According to Fourier
theory, this reduction stretches the pulse temporally, and its duration can only be
fully restored in the focal region where the spectral components of the pulse are
spatially reunited. In effect, the pulse is focused both spatially and temporally by
the objective lens. Since the pulse duration directly determines the peak power of
the pulse, the gradient of the temporal focusing can be used to control the Rayleigh
range of the peak E-field distribution of the focus. The STF technique therefore
uniquely enables the creation of a peak E-field distribution that is close to spherical
in structure. Under these circumstances, the cross section of the inscribed structure
would be symmetric, regardless of the sample translation direction. This technique
13 Ultrafast Laser Inscription of Photonic Devices in Bulk Dielectrics 341
is yet to be applied to the inscription of optical waveguides and has only been used
to fabricate microfluidics using selective etching. It is reasonable to expect that it
will soon be applied to waveguide inscription.
Compared to the multi-million dollar setup and high running costs of a conventional
planar waveguide fabrication plant, the few hundred thousand dollar setup costs and
low running costs of a ULI system are insignificant. Furthermore, despite the inher-
ently serial fabrication nature of ULI, the possibility of translation velocities in the
10’s of mm.s1 makes ULI potentially well suited to the mass production of simple
passive waveguide devices for applications in areas such as telecommunications and
optical sensor networks.
To date, a range of passive waveguide devices, including wavelength division
multiplexers [30], power splitters [40], Bragg-waveguides [52] and multicore fibre
couplers [53] have been fabricated using ULI. For example, Eaton et al. [40] used
ULI to fabricate evanescent field power splitters which performed a broadband
splitting operation throughout the 1,250–1,650 nm spectral region. The authors
of this study were able to achieve a range of power splitting ratios by inscribing
asymmetric splitters which were realised using different translation velocities for
each arm of the splitter.
Another ULI fabrication capability that is attractive for passive waveguide
devices is the inscription of waveguides with Bragg-gratings – Bragg-waveguides.
Bragg-waveguides can be inscribed using a number of different approaches. The
first ULI fabricated Bragg-waveguide was demonstrated by Marshall et al. in 2006
[54]. In this work, 800 nm pulses at a PRF of 1.0 kHz were used to first inscribe a
waveguide in fused silica. A second order grating operating at 1,550 nm was then
inscribed on top of the waveguide by translating the substrate through the focus a
second time, exposing the waveguide to a single pulse at each grating period. These
Bragg-waveguides exhibited a low grating strength and were slow to inscribe due to
the low pulse repetition rate of the laser system. To overcome these issues, Zhang
et al. [52] later demonstrated that high strength (>30 dB transmission attenuation)
first order gratings could be inscribed in fused silica using an externally modulated
ultrafast fibre laser operating at a PRF of 500 kHz, thus opening the way to the rapid
ULI fabrication of high quality, commercially interesting Bragg-waveguide devices
in fused silica.
342 R.R. Thomson and G. Cerullo
ULI can be used to fabricate waveguides in almost any transparent material, making
it exceptionally well suited to the fabrication of novel active waveguide devices
such as waveguide amplifiers and lasers. Much of the initial work in this area
was pioneered by the group at the Politecnico di Milano. For example, the first
demonstration of fibre-to-fibre net gain by a ULI fabricated waveguide amplifier
was in 2004 by Osellame et al. [55], who inscribed a waveguide in an ErYb-doped
phosphate glass. The 12 mm long amplifier exhibited a peak net gain of 1.2 dB at
1,533 nm under pumping at 980 nm. Later the same year, Taccheo et al. [56] used a
20 mm long ULI fabricated ErYb-doped waveguide amplifier as the gain element in
the first ULI fabricated waveguide laser. This device emitted 1.7 mW of continuous
wave 1533.5 nm radiation under 420 mW of pumping at around 980 nm. In 2006,
the same group demonstrated the first ULI fabricated mode-locked waveguide laser
by constructing an optical fibre-ring cavity which incorporated a carbon nanotube
saturable absorber and a 39 mm long ULI fabricated ErYb-doped amplifier [57].
Under 430 mW of pumping at 980 nm, the device emitted a 16.74 MHz train of
1.76 ps pulses with a central wavelength of 1,535 nm and an average output power
of 0.1 mW.
The active waveguide applications of ULI have now progressed considerably,
and waveguide lasers operating at around 1.9 m [58, 59], 1.5 m [60], 1.04 m
[61] and 1.06 m [62] have been realised by inscribing waveguides in Tm-, Er-,
Yb- and Nd-doped glasses respectively. In 2010, Thomson et al. [63] demonstrated
the ULI fabrication of a high gain Er-doped bismuthate glass waveguide amplifier
which exhibited a peak fibre-to-fibre net gain of 16.0 dB at 1,533 nm under 1.0 W of
980 nm pumping. Beecher et al. [64] later used this waveguide amplifier as the laser
gain element in a mode-locked laser cavity. Under 350 mW of 980 nm pumping, the
laser emitted a 40 MHz train of 320 fs pulses with a central wavelength of 1.56 m
and an average output power of 1.25 mW.
The level of integration that can be achieved in ULI waveguide lasers has
also been the subject of significant research attention. The group at Macquarie
University successfully demonstrated the first ULI fabricated distributed feedback
waveguide laser by inscribing a Bragg-waveguide in an ErYb-doped phosphate glass
substrate [65]. Under 710 mW of pumping at 980 nm, the 20.0 mm long device
emitted a maximum output power of 0.37 mW at a lasing wavelength of 1,537 nm.
The same group later reported the ULI fabrication of a high power and efficient
distributed feedback waveguide laser by inscribing a Bragg-waveguide in a Yb-
doped phosphate glass substrate [61]. When pumped using 726 mW of 976 nm
light, the 9.5 mm long waveguide laser emitted 102 mW of 1,033 nm light with a
slope efficiency of 17.3 %.
A number of groups have worked to increase the slope efficiencies and output
powers that can be achieved from ULI fabricated waveguide lasers. In 2008 Torchia
et al. [27] reported the ULI fabrication of a 5.0 mm long Nd-doped YAG ceramic
waveguide laser that emitted >80 mW of 1,064 nm radiation under 200 mW of
13 Ultrafast Laser Inscription of Photonic Devices in Bulk Dielectrics 343
The selective etching capability of ULI has been used to fabricate both surface [2]
and sub-surface microfluidic structures [67]. However, the ability of ULI to readily
integrate microfluidics with optical waveguides is potentially an extremely powerful
capability. In one example, Bellini et al. [68] used ULI to fabricate a monolithic
cell stretcher – a sketch of which is shown in Fig. 13.11. This device consisted
of a ULI fabricated sub-surface microfluidic and two optical waveguides which
intercepted the microfluidic. The authors demonstrated that red blood cells could
be inserted into the microfluidic and optically trapped by counter-propagating laser
beams coupled into the optical waveguides. The authors also demonstrated that the
cells could be stretched by the optical forces when the trapping laser power was
increased.
Interestingly, the optical stretching properties of cells have been shown to be a
marker for the malignant state of a cell [69], without a requirement for chemical
labelling. This simple and elegant demonstration clearly demonstrates the power of
ULI for fabricating novel and monolithic microfluidic and optofluidic devices for a
plethora of applications in biophotonics and beyond.
344 R.R. Thomson and G. Cerullo
Fig. 13.12 Computer assisted drawing of the ULI fabricated displacement sensor (Reproduced
and adapted with permission from Ref. [3])
13.6.4 Micro-Systems
13.6.5 Astrophotonics
In this chapter the basic physics of the ULI process was reviewed. We have seen
that ULI relies on the nonlinear absorption of sub-bandgap ultrashort pulses inside
a transparent dielectric, and that the deposited energy induces a local modification
to the material structure. We have seen how this structural modification may
manifest itself in a variety of ways, and that the nature of these manifestations
can be controlled via the various ULI parameters. We have discussed how the ULI
induced refractive index change and chemical etch rate modification can be used
and controlled to inscribed structures such as optical waveguides, micro-mechanics,
microfluidics and micro-optics. Finally, we have outlined a few of the many ULI
applications that are being investigated. To conclude, it is clear that ULI presents
itself as an extremely flexible and unique fabrication tool. In the future we envisage
that ULI, like many other fabrication techniques, will move towards becoming more
of an established fabrication route for both industrial and research scale applications.
Until then, there remains much to be explored in this exciting field.
Acknowledgments RRT acknowledges support through and STFC Advanced Fellowship – Grant
No. ST/H005595/1.
References
11. M. Ams, G.D. Marshall, P. Dekker, M. Dubov, V.K. Mezentsev, I. Bennion, M.J. Withford,
Investigation of ultrafast laser–photonic material interactions: challenges for directly written
glass photonics. IEEE J. Sel. Top. Quantum Electron 14, 1370–1381 (2008)
12. C.B. Schaffer, A.O. Jamison, J.F. Garcı́a, E. Mazur, Structural changes induced in transparent
materials with ultrashort laser pulses, in Ultrafast Lasers: Technology and Applications, ed. by
M.E. Fermann, A. Galvanauskas, G. Sucha (CRC Press, Boca Raton, 2002). p. 347
13. C.B. Schaffer, A. Brodeur, J.F. Garcı́a, E. Mazur, Micromachining bulk glass by use of
femtosecond laser pulses with nanojoule energy. Opt. Lett. 26, 93–95 (2001)
14. C.B. Schaffer, A.B. Jamison, E. Mazur, Morphology of femtosecond laser-induced structural
changes in bulk transparent materials. Appl. Phys. Lett. 84, 1441–1443 (2004)
15. Y. Shimotsuma, P.G. Kazansky, J. Qiu, K. Hirao, Self-organized nanogratings in glass
irradiated by ultrashort light pulses. Phys. Rev. Lett. 91, 247405 (2003)
16. R. Taylor, C. Hnatovsky, E. Simova, Applications of femtosecond laser induced self-organized
planar nanocracks inside fused silica glass. Laser Photonics Rev. 2, 26–46 (2008)
17. Y. Bellouard, E. Barthel, A.A. Said, M. Dugan, P. Bado, Scanning thermal microscopy and
Raman analysis of bulk fused silica exposed to low energy femtosecond laser pulses. Opt.
Express 16, 19520–19534 (2008)
18. Y. Bellouard, T. Colomb, C. Depeursinge, M. Dugan, A.A. Said, P. Bado, Nanoindentation
and birefringence measurements on fused silica specimen exposed to low-energy femtosecond
pulses. Opt. Express 14, 8360–8366 (2006)
19. S. Zhou, W. Lei, J. Chen, J. Hao, H. Zeng, J. Qiu, Laser-induced optical property changes
inside Bi-doped glass. IEEE Photonics Technol. Lett. 21, 386 (2009)
20. J.W. Chan, T. Huser, S. Risbud, D.M. Krol, Structural changes in fused silica after exposure to
focused femtosecond laser pulses. Opt. Lett. 26, 1726–1728 (2001)
21. A.M. Streltsov, N.F. Borrelli, Study of femtosecond-laser-written waveguides in glasses. J. Opt.
Soc. Am. B 19, 2496–2504 (2002)
22. R. Osellame, M. Lobino, N. Chiodo, M. Marangoni, G. Cerullo, R. Ramponi, H.T. Bookey,
R.R. Thomson, N.D. Psaila, A.K. Kar, Femtosecond laser writing of waveguides in periodically
poled lithium niobate preserving the nonlinear coefficient. Appl. Phys. Lett. 90, 241107
(2007)
23. R. Martinez-Vazquez, R. Osellame, G. Cerullo, R. Ramponi, O. Svelto, Fabrication of photonic
devices in nanostructured glasses by femtosecond laser pulses. Opt. Express 15, 12628–12635
(2007)
24. E. Bricchi, J.D. Mills, P.G. Kazansky, B.G. Klappauf, J.J. Baumberg, Birefringent fresnel zone
plates in silica fabricated by femtosecond laser machining. Opt. Lett. 27, 2200–2202 (2002)
25. N.D. Psaila, R.R. Thomson, H.T. Bookey, A.K. Kar, N. Chiodo, R. Osellame, G. Cerullo,
A. Jha, S. Shen, Er:Yb-doped oxyfluoride silicate glass waveguide amplifier fabricated using
femtosecond laser inscription. Appl. Phys. Lett. 90, 131102 (2007)
26. J. Burghoff, C. Grebing, S. Nolte, A. Tünnermann, Efficient frequency doubling in femtosec-
ond laser-written waveguides in lithium niobate. Appl. Phys. Lett. 89, 081108 (2006)
27. G.A. Torchia, A. Ródenas, A. Benayas, E. Cantelar, L. Roso, D. Jaque, Highly efficient laser
action in femtosecond-written Nd: yttrium aluminum garnet ceramic waveguides. Appl. Phys.
Lett. 92, 111103 (2008)
28. A. Benayas, W.F. Silva, A. Ródenas, C. Jacinto, J. Vázquez de Aldana, F. Chen, Y. Tan, R.R.
Thomson, N.D. Psaila, D.T. Reid, G.A. Torchia, A.K. Kar, D. Jaque, Ultrafast laser writing of
optical waveguides in ceramic Yb:YAG: a study of thermal and non-thermal regimes. Appl.
Phys. A – Mater 104, 301–309 (2011)
29. Y. Nasu, M. Kohtoku, Y. Hibino, Low-loss waveguides written with a femtosecond laser for
flexible interconnection in a planar light-wave circuit. Opt. Lett. 30, 723–725 (2005)
30. S.M. Eaton, W.-J. Chen, H. Zhang, R. Iyer, J. Li, M. Ng, S. Ho, J.S. Aitchison, P.R.
Herman, Spectral loss characterization of femtosecond laser written waveguides in glass with
application to demultiplexing of 1300 and 1550 nm wavelengths. J. Lightwave Technol. 27,
1079–1085 (2009)
348 R.R. Thomson and G. Cerullo
Abdullah Alharbi
(University of Ottawa)
Vid Agrež
(University of Ljubljana)
Alexandria Anderson
(Femtolasers GmbH)
Rodrigo Aviles-Espinosa
(Institut de Ciencies Fotoniques)
Cyrill Baer
(ETH Zurich)
Tom Baer
(Stanford University)
Stephen Beecher
(Heriot-Watt University)
Petra Beličev
(Vinca Institute of Nuclear Sciences)
Martynas Beresna
(University of Southampton)
Martin Blazek
(Technische Universitat Darmstadt)
Henry Bookey
(Heriot-Watt University)
Elisabeth Bothschafter
(Technische Universitat Munchen)
Daniele Brida
(Politecnico di Milano)
Phil Bucksbaum
(Stanford University)
Mantas Butkus
(University of Dundee)
Giulio Cerullo
(Politecnico de Milano)
Yong Chen
(University of Bath)
Rim Cherif
(Ecole Superiuere des Communications des Tunis)
Debaditya Choudhury
(Heriot-Watt University)
Marcus Dahlstrom
(Lund University)
Louis Daniault
(Institut d’Optique)
Julius Darginavicius
(Vilnius University)
Thomas Denis
(University of Twente)
Sven Doring
(Universitat Jena)
Chris Dorman
(Coherent Scotland)
Ghassen Dridi
(Universite de Bourgogne)
John Dudley
(Universite de Franche-Comte)
Martin Duffy
(Queen’s University Belfast)
Hanieh Fattahi
(Max-Planck-Institut fur Quantenoptik)
Ksenia Fedorova
(University of Dundee)
Participants 359
Volker Feige
(Automation Dr.)
Allister Ferguson
(University of Strathclyde)
Teresa Ferreiro
(Heriot-Watt University)
Georg Gademann
(FOM-AMOLF)
Xunyou Ge
(CEA Saclay)
Gregory Golovin
(Moscow State University)
Miguel Gonzalez-Galicia
(Universidad Nacional Autonoma de Mexico)
Itandehui Gris Sanchez
(University of Bath)
David Hadden
(University of Warwick)
Vahe Hakobyan
(Universite de Bourgogne)
Tawfique Hasan
(University of Cambridge)
Christophe Hazera
(Universite de Bordeaux)
Oliver Heckl
(ETH Zurich)
Matthias Heinrich
(Universitat Jena)
Gordon Henderson
(University of Edinburgh)
Carlos Hernandez-Garcia
(Universidad de Salamanca)
Christoph Heyl
(Lund University)
Gareth Hobbs
(University of Bath)
360 Participants
Lucy Hooper
(University of Bath)
Daniel Horke
(University of Durham)
Christopher Hutchison
(Imperial College London)
Igor Ilic Vinca
(Institute of Nuclear Sciences)
Ann James
(Imperial College London)
Christine Jappy
(University of St Andrews)
Edgaras Jelmakas
(Vilnius University)
Olof Johansson
(University of Edinburgh)
Ajoy Kar
(Heriot-Watt University)
Ursula Keller
(ETH Zurich)
Christian Kern
(Universitat Jena)
Juho Kerttula
(Tampere University of Technology)
Vikaran Khanna
(University of Hamburg)
Florian Kienle
(University of Southampton)
Markus Kohler
(Max-Planck-Institut fur Kernphysik)
Nils Krebs
(LMU Munich)
Tobias Lamour
(Heriot-Watt University)
Sebastien Landon
(Universite de Saint-Etienne)
Participants 361
Ken Ledingham
(University of Strathclyde)
Ruth Livingstone
(Heriot-Watt University)
Madis Lohmus
(University of Tartu)
Lauryna Lotscher
(LMU Munich)
John Macdonald
(Heriot-Watt University)
Guillaume Machinet
(Universite de Bordeaux)
Benoit Mahieu
(University of Nova Gorica & CEA Saclay)
Matthias Mang
(University of Oxford)
Erik Mansson
(Lund University)
Cristian Manzoni
(Politecnico di Milano)
Jon Marangos
(Imperial College London)
Rose Mary
(Heriot-Watt University)
Kamila Mazur
(University of East Anglia)
David McCabe
(Universite de Toulouse)
John McCarthy
(Heriot-Watt University)
Richard McCracken
(Heriot-Watt University)
Joanna McLenaghan
(University of St Andrews)
Niels Meiser
(KTH Stockholm)
362 Participants
Margaret Murnane
(University of Colorado)
Giovanni Nava
(University of Pavia)
Catalin Neacsu
(Femtolasers GmbH)
Danill Nikitichev
(University of Dundee)
Julien Nillon
(Universite de Bordeaux)
Stefan Nolte
(Universitat Jena)
Ruud Oldenbeuving
(University of Twente)
Malte Oppermann
(Imperial College London)
Jonathan Orchard
(University of Sheffield)
Christian Ott
(Max-Planck-Institut fur Kernphysik)
Tim Paasch-Colberg
(Max-Planck-Institut fur Quantenoptik)
Monika Pawłowska
(Freie Universitat Berlin)
Selina Pekarek
(ETH Zurich)
Peeter Piksarv
(University of Tartu)
Dario Polli
(Politecnico di Milano)
Daniel Popa
(University of Cambridge)
Philipp Raith
(Max-Planck-Institut fur Kernphysik)
Rafal Rakowski
(Lund University)
Participants 363
Cristtel Ramirez
(Heriot-Watt University)
Euan Ramsay
(Heriot-Watt University)
Jussi Rautiainen
(Tampere University of Technology)
Derryck Reid
(Heriot-Watt University)
Sandrine Ricaud
(Laboratoire Charles Fabry de l’Institut d’Optique)
Soren Richter
(Universitat Jena)
Carolina Romero
(Universidad de Salamanca)
Mattia Rossetti
(Politecnico di Torino)
Peter Roth
(University of Strathclyde)
Eleonora Rubino
(Universita degli Studi dell’Insubria)
Clara Saraceno
(ETH Zurich)
Markus Schenk
(Max-Planck-Institut fur Quantenoptik)
Andrea Schiavi
(University of Oxford)
Kai Seger
(KTH Stockholm)
Wilson Sibbett
(University of St Andrews)
Teresa Sibillano
(University of Bari)
Eran Small
(Weizmann Institute)
Richard Squibb
(Imperial College London)
364 Participants
Jeff Squier
(Colorado School of Mines)
Robert Stefanuik
(University College Dublin)
Jinghua Sun
(Heriot-Watt University)
Ayhan Tajalli
(Universite de Toulouse)
Robert Thomson
(Heriot-Watt University)
Dave Townsend
(Heriot-Watt University)
George Tserevelakis
(IESL – Foundation for Research and Technology – Hellas)
Thomas Udem
(Max-Planck-Institut fur Quantenoptik)
Mor Verbin
(Weizmann Institute)
Andreas Vernaleken
(Max-Planck-Institut fur Quantenoptik)
William Wadsworth
(University of Bath)
Ian Walmsley
(University of Oxford)
Frank Wang
(University of Cambridge)
Zhaowei Zhang
(Heriot-Watt University)
The SUSSP66 Participants
A Auger
Aberration decay, 68
chromatic, 303 electrons, 162, 164
spherical, 303 photoelectron peaks, 164, 165
Acoustic wave propagation, 166
Acousto-optic tunable filter (AOTF), 282
Adiabatic B
gas preparation for Raman conversion, 64 Backfolding, 156
pulse compression, 211 Back-projection algorithm, 14
Akhmediev breathers, 186, 187 Backscattered photons, 243, 245
Ammosov, Delone and Krainov (ADK) Ballistic correction, 166
theory, 51 Beam waist, 332, 338, 339
Analytic signal, 3–5 Beer’s law, 293, 297
frequency representation, 4 Binding energy, 155
Ancilla, 17–18 Blazed grating, 281
chirp rate, 17 Born-Oppenheimer, 106–107
Angle-resolved photoemission spectroscopy approximation, 107, 108, 113
(ARPES), 154, 156–158 surfaces, 106–107
nano ARPES, 170 Bragg
Ankylography, 168 gratings, 341
AOTF. See Acousto-optic tunable filter (AOTF) mirror, 81, 82, 88, 91
Apollon laser, 246 peak, 236
Arbitrary waveform generator, 120 Bremsstrahlung, 242
ARPES. See Angle-resolved photoemission Brillouin
spectroscopy (ARPES) scattering, 198
Arrayed waveguide gratings (AWGs), 345 zone, 156
Astrophotonics, 344–345 Brumer-Shapiro, 106, 111–113
Atomic clock, 129, 135, 136, 138, 143
Attosecond
dynamics, 61 C
generation, 45–69 Cancer therapy using protons and ions,
measurement technology, 45 235–237
pulse bursts, 152 Carrier envelope, 120, 142–143
pulse production, 56–58 offset frequency, 4, 134
pulse train (APT), 57 phase shift, 130
single pulse, 67 phase (CEP) stabilisation, 47, 58
Carrier frequency, 26, 40, 46, 47, 60, 63, 86, first-order dispersion (group delay), 80
129, 131, 142 flattened fibre, 213
chirp, 86, 131 geometrical, 80, 82
Catalysis, 162 grating pairs, 80
Cell stretcher, 343 group-delay dispersion (GDD), 15, 82, 268
Charge group velocity dispersion (GVD), 26, 98,
density wave, 156, 158 179, 202–203, 210, 215, 272
migration, 46, 47 high-order, 184
transfer, 149, 162, 163 intra-cavity, 80, 130
Chirped pulse amplification (CPA), 2, 40, 228 material, 56, 80, 203–206, 213, 303
Chronocyclic mirror, 81, 82
phase space, 6, 7 normal, 86, 184, 189, 190, 204, 205, 208,
tomography, 14 213, 215, 219
wigner function, 6 optical fibre, 202–205
Coherent second-order, 80, 268
diffractive imaging, 167–169 shifted fibre, 132
oversampling phase retrieval third-order (TOD), 80, 81, 209, 211,
technique, 169 268, 270
phase retrieval algorithm, 168 waveguide, 82, 153, 179, 204, 206–208,
wakefield emission, 65 213, 215, 220
x-ray generation, 150 zero dispersion wavelength, 132, 182, 184,
Colour centres, 301, 327 185, 187, 190, 203, 214, 215, 219
Complex field envelope, 25 Dispersive waves, 184, 186, 189
Compton scattering, 243, 246 Dream beams, 232
Confocal parameter, 332 Dressed states, 51
Conical intersections, 108–109 laser dressed state continuum state, 51
Correlated-electron materials, 154–158, 161 Dynamic gain saturation, 86, 87
Correlation function, 5, 6, 9, 10
Cosmic rays, 230
Coulomb interaction, 46, 105 E
CPA. See Chirped pulse amplification (CPA) Elbe electron beam, 244
Cross phase modulation (XPM), 197, 198, 202 Electrochemical processes, 162
Cutoff photon energy, 150, 152 Electron
beams via laser plasma interaction,
230–232
D cooling time, 295, 296
Debye sheath, 235 dynamics, 46, 47, 162
Dephasing, 106 heat capacity, 295, 296
DESY, Hamburg, 67 hot, 156, 163, 292, 297
Difference-frequency generation (DFG), 38, 39 laser
Diffraction before destruction, 149 acceleration, 228, 243
Diode-pumping, 73 production, 228–230
Dispersion lattice coupling, 294
anomalous, 86, 87, 184–186, 203, 205, 208, quasi monoenergetic, 232, 235
209, 212, 213, 217–219 quiver
characteristic dispersion, 75, 94, 180, 188 energy, 228, 229
chromatic, 178, 202 motion, 231
compensation, 81, 93, 254, 271 velocity, 231
control in optical fibres, 202–208 relaxation time, 297
control in PCF, 213 spectra, 60, 232
convex, 188–189 temperature, 295–297
double-passed prism, 80 thermal conductivity, 297
engineering, 179 wave-packet, 52
Index 369