ZWO
LIBRARY
Michigan State
University
This is to certify that the
dissertation entitled
Nonlinear imaging with femtosecond laser pulses
presented by
Yves Coello
has been accepted towards fulfillment
of the requirements for the
Doctoral degree in Chemistry
%
Major Professor’s Signature
6’17-40
Date
MSU is an Afiinnative Action/Equal Opportunity Employer
nn-a—n-nan-o-n-v—---c--o----n--o-—--n--a-a--o--.--u-.-.---—---o-¢-n--u—----u-u--u-o--n----n_--o-----o-
PLACE IN RETURN Box to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.
DATE DUE DATE DUE DATE DUE
5/08 KzlProilecauprelelRC/Dateouo.indd
3
. .-
1... .21....
ISM .
.... I: .f .w.
NONLINEAR IMAGING WITH FEMTOSECOND LASER PULSES
By
Yves Coello
A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Chemistry
2010
ABSTRACT
NONLINEAR IMAGING WITH FEMTOSECOND LASER PULSES
By
Yves Coello
The high peak powers delivered by femtosecond laser pulses easily induce nonlinear
optical processes which are useful for a variety of applications, including biological
imaging. This dissertation presents various contributions to the development of nonlinear
imaging with femtosecond laser pulses.
The spectral phase distortions that femtosecond laser pulses suffer due to dispersion
as they transmit through microscope objectives have hindered the application of sub—50 fs
pulses in nonlinear imaging. Here, accurate spectral phase characterization of such pulses
was accomplished using multiphoton intrapulse interference phase scan (MIIPS). As a
result, pulse compression to the theoretical transform-limited duration at the focal plane
was demonstrated, including the challenging case of 4.3 fs pulses. These MIIPS
developments will allow taking advantage of sub-50 fs pulses for nonlinear imaging. In
addition, dispersion measurements of optical media with unprecedented accuracy were
obtained using MIIPS. This information is critical for pulse propagation models.
Two-photon spectra of fluorescent dyes are necessary for the most popular nonlinear
imaging method, two-photon laser scanning fluorescence microscopy. In this dissertation,
a fast and automated approach able to measure two-photon spectra of fluorophores by
PUISe shaping ultrabroad-bandwidth femtosecond laser pulses is demonstrated. The
approach is particularly valuable given that it is suitable for non-laser expert use.
Finally, direct non-resonant femtosecond laser desorption and ionization were
applied in the development of a new atmospheric pressure mass-spectrometric imaging
approach using amplified femtosecond pulses. These results, which show unprecedented
spatial resolution, open new possibilities for the use of femtosecond laser pulse in
biological imaging.
ACKNOWLEDGEMENTS
The work presented here is the result of a learning process that has been greatly enriched
through interactions with many people that shared with me their experience and
knowledge and offered me their help and support. First of all, I want to thank my advisor
Prof. Marcos Dantus for providing me with interesting and challenging research projects,
offering me useful suggestions through frequent scientific discussions and financially
supporting me during my graduate studies. I am also thankful to the members of my
committee, Profs. John McCracken, James McCusker and Katharine Hunt for their time
and advice. I also want to thank past and present members of the Dantus research group:
Dr. Vadim Lozovoy, Dr. Igor Pastirk, Dr. Tissa Gunaratne, Dr. Haowen Li, Dr. D. Ahmasi
Harris, Dr. Dmitry Pestov, Dr. Bingwei Xu, Dr. Scott Smith, Jess Gunn, Xin Zhu, Paul
Wrzesinski, Christine Kalcic, and Marshall Bremer. The regular interaction with them in
group meetings, conferences, labs and offices has represented a continuous learning
eXperience. Especially, I would like to thank Vadim for his endless suggestions and his
lucid perspective. I am also very grateful to Bingwei, with whom I was fortunate to start
working with when I just entered the group. He was very kind and patient answering
innumerable questions about femtosecond lasers, optics and pulse shapers at that time. As
a re>Sult of a fruitful collaboration, Vadim, Bingwei and I coauthored a number of research
Publications. I was also very lucky to have the advice of two mass spectrometry experts,
iv
Profs. A. Daniel Jones and Gavin Reid. In particular, help and suggestions provided by
Prof. Jones were critical for the progress of the imaging mass spectrometry project, which
I concentrated on during the last years of my graduate studies. I would also like to thank
the College of Natural Sciences for providing me the Dissertation Completion Fellowship
in the summer of 2010.
TABLE OF CONTENTS
LIST OF TABLES ................................................................................... viii
LIST OF FIGURES ................................................................................... ix
LIST OF ABBREVIATIONvau
INTRODUCTION ..................................................................................... 1
CHAPTER 1
PRELIMINARY CONCEPTS ......................................................................... 4
1.1. Ultrashort laser pulses and the spectral phase4
1.2. Multiphoton intrapulse interference (MII)6
CHAPTER 2
EXPERIMENTAL TOOLS ............................................................................ 13
2.1. F emtosecond laser systems .................................................................. 13
2.1.1. Ultrabroad-bandwidth Ti:Sa femtosecond laser oscillator
2.1.2. Ti:Sa regenerative amplifier
2.2. Pulse shapers .................................................................................. 16
2.3. Multiphoton intrapulse interference phase scan (MIIPS) .................................... 21
CHAPTER 3
FEMTOSECOND LASER PULSE CHARACTERIZATION AND COMPRESSION
USING MIIPS .................................................................................................................... 29
3.]. Spectral phase measurements .............................................................. 3O
3. 2. Pulse compression. . . 36
3.2.1. Compression of ultrabroad- bandwidth femtosecond laser pulses
3.2.2. Compression of regeneratively amplified femtosecond laser pulses
3.3. Conclusions .......................................................................................................... 44
CHAPTER 4
CHROMATIC DISPERSION MEASUREMENTS WITH MIIPS. . . . . . . . . . . . . . . . ............46
4.1. Introductlon47
4.2. Experimental secttonSO
4.3. Results ......................................................................................... 52
4.3.1. Chromatic dispersion of deionized water
4.3.2. Chromatic dispersion of seawater
vi
4.3.3. Chromatic dispersion of the vitreous humor and the comea-lens complex
4.4. Discussion ............................................................................................................ 60
4.4.1. Chromatic dispersion of deionized water and seawater
4.4.2. Chromatic dispersion of the vitreous humor and the comea-lens complex
4.5. Conclusions .......................................................................................................... 62
CHAPTER 5
HIGH-RESOLUTION TWO-PHOTON FLUORESCENCE EXCITATION
SPECTROSCOPY BY PULSE SHAPING ULTRABROAD-BANDWIDTH
FEMTOSECOND LASER PULSES ................................................................................. 63
5.1 . Introduction .......................................................................................................... 63
5.2. Experimental part ................................................................................................. 65
5.2.1. Optical setup
5.2.2. Sample preparation
5.3. Results and discussion ......................................................................................... 67
5.4. Conclusions .......................................................................................................... 77
CHAPTER 6
ATMOSPHERIC PRESSURE FEMTOSECOND LASER DESORPTION IONIZATION
IMAGING MASS SPECTROMETRY .............................................................................. 78
6.1. Introduction 78
6.2. Experimental section ............................................................................................ 82
6.2.1. Mass spectrometer and laser system
6.2.2. Imaging
6.2.3. Materials and sample preparation
6.2.4. Metabolite identification
6.3. Optimization of ion source parameters and preliminary results .......................... 85
6.4. Results and discussion ......................................................................................... 89
6.5. Conclusions .......................................................................................................... 96
REFERENCES ........................................................................................... 98
vii
LIST OF TABLES
Table 4.1. Parameters of the Sellmeier formula for water at 21 .5°C ................................. 54
Table 4.2. Seawater parameters for Equation 4.3 in the range 660-930 nm ...................... 57
Table 4.3. Vitreous humor parameters for Equation 4.4 in the range 660-930 nm ........... 58
Table 4.4. Experimental dispersion measurements for water, seawater and eye
components ........................................................................................................................ 59
viii
LIST OF FIGURES
Figure 1.1. Multiphoton intrapulse interference for SHG (a) illustrates MII for the case of
a TL pulse, with a constant spectral phase (p(co). Frequency pairs (when) and (col-Q,
031+Q) in the fundamental spectrmn S((o) interfere constructively to produce maximum
intensity in the SHG spectrum S‘z)(2co) at 20);. Constructive interference occurs for any
frequency (01 within the bandwidth of the pulse because all frequency pairs are in phase.
As a consequence the SHG signal is maximized for all frequencies and the spectrum
5(2)(2(0) shown at the top is obtained. The uniform filling under 8(a)) indicates that the
spectral phase is constant. Note that frequencies in the unfilled part of the spectrum do
not contribute to the signal at 2001 because no frequencies in the lower frequency part of
the spectrum can pair with them. (b) illustrates MII for a sinusoidal spectral phase
=Wexpii¢L (1.4)
which provides the electric field E(co) in terms of the spectrum S(a)) and the spectral
phase rp(co). While the magnitude of E(co) can be easily measured with a spectrometer, the
spectral phase has been traditionally much more difficult to measure. This interesting
issue will be discussed in section 2.3. Note that according to Equation 1.4, the spectrum
and the spectral phase together are sufficient to completely characterize an ultrashort
pulse.
The spectral phase is the relative phase that different frequency components of the
pulse carry. A variation in War) is directly reflected in the temporal intensity of the pulse,
as will be illustrated in the next section. It can be shown that for a given spectrum the
pulse achieves its minimum time duration when War) is a linear firnction of frequency.
Such a pulse is called transform-limited (TL). For a TL pulse, the equality in the time
bandwidth product is satisfied (At Av=K).
Spectral phase distortions are introduced by dispersion [11]. Material dispersion is
the most usual and well-known class of dispersion and occurs because the refractive
index of materials is frequency-dependent. As a consequence, different frequency
components of a laser pulse travel with different velocities and the pulse changes its
temporal intensity as it propagates through the medium. A more detailed discussion about
material dispersion and its measurement will be presented in chapter 4. Material
dispersion introduces mainly quadratic phase distortions, which correspond to a linear
variation of frequency with time. In this case, the pulse is said to be linearly chirped.
Higher-order (>2) spectral phase distortions introduced by material dispersion become
more important as the pulse duration decreases (<50 fs). Spectral phase distortions are
introduced by material dispersion in the laser system itself as well as by propagation of
the laser pulse through different optical elements present in the experimental setup such
as lenses, microscope objectives, filters, beamsplitters and cuvettes. Introducing or
changing any of these elements from the optical setup affects the spectral phase of the
pulses and therefore the outcome and reproducibility of an experiment unless a method
able to characterize and correct the spectral phase distortions is used. Multiphoton
intrapulse interference phase scan (MIIPS), which will be described in section 2.3, is a
method useful for such purpose. MIIPS is based on multiphoton intrapulse interference.
1.2 Multiphoton intrapulse interference (MII)
Nonlinear optical (NLO) processes depend on the spectral phase of the laser pulses
that generated them. By NLO processes we refer here to multiphoton excitation of
chemical systems (e. g. two-photon absorption in molecules) or to generation of nonlinear
optical signals in optical media (e. g. second-harmonic generation, SHG, in a crystal). The
dependence of NLO processes on the spectral phase was recognized in a number of
studies thanks to the development of pulse shaping techniques, which allow controlling
the amplitude and phase of ultrashort pulses (section 2.2). In 1992, amplitude modulation
and different amounts of chirp (quadratic spectral phase modulation) were used to
manipulate the SHG spectrum of the laser pulses and two-photon absorption in Rb atoms
[12, 13]. Control of two-photon transitions in Cs atoms was demonstrated using a step
spectral phase function (11: jump) in 1998 [14, 15]. This observation was explained in
terms of constructive or destructive interference between pairs of frequencies within the
bandwidth of the laser, which depended on the spectral phase of the pulses, that
determined the multiphoton transition probability (by enhancing or supressing it) at a
particular frequency corresponding to the transition. The Dantus group realized that the
spectral phase modulates the probability for NLO processes at all frequencies in the nth-
order nonlinear spectrum of the pulses Swan), where n is the order of the NLO process
being considered. They demonstrated control of two- and three-photon transitions in large
molecules in solution and control of the SHG spectrum with sinusoidal spectral phases
[16]. The phenomenon was named MII and was further studied in subsequent papers [17,
18].
To understand MII let us consider SHG, a two-photon optical process relevant for
MIIPS. In SHG, an input wave interacting with a nonlinear medium, typically a crystal,
generates an output wave with twice the optical frequency (thus, SHG is also known as
frequency doubling). An ultrashort pulse contains a number of different frequencies and
thus different pathways exist for the frequency doubling process. In general, a pair of
photons with frequencies (0 -Q and co 42, where Q is a frequency detuning, can combine
and generate a photon at frequency 2(1). Mathematically, The SHG spectrum Smear) is
given by
5(2) (20)) cc le(w+r2)E* (co—r2)arr2|2 (1.5),
where E((n) is the field of the pulses [17]. Multiple different frequency pairs (1) -Q and or -
Q in the spectrally broad pulses provide multiple pathways that give rise to interference.
This interference is accounted for by integration over all possible Q. All three words in
M11 are essential: it is the interference of the field with itself (intrapulse) taking place in a
multiphoton process. To make evident that such interference depends on the spectral
phase we can rewrite E(or) in terms of its amplitude and phase and obtain
2
5(2) (20)) cc UlE(a)+.(2)”E(w—.Q)lexp[i((o(w+.(2) +(o(co— 12))] ml (1.6).
Figure 1.1 illustrates the M11 process for SHG. In the left panel the spectral phase
(a( a1) is constant (TL pulse). Consequently, for any frequency co in the bandwidth of the
pulse the complex exponential term in Equation 1.6 is constant and all frequency pairs
interfere constructive to maximize the SHG spectrum Smear). In the right panel, the
spectral phase War) is a sine function with its inversion center at (01. Therefore, the
complex exponential term in Equation 1.6 is constant for all frequency pairs around to]
and all frequency pairs around to; interfere constructive to maximize the SHG spectrum
Smear) at 2(01. In contrast, for any other frequency cigar-“(1)1 the exponential term in
Equation 1.6 is frequency dependent and different degrees of destructive interference take
place giving rise to low signal.
(a) 8(2)(2w) (b) 8(2)(2a,)',"
- - - ‘ ‘ 2a)] 6% 260]
(01+!) 01+!)
- qt nuwl
i
I 02,—!) I I (01—!)
I I |
Figure 1.1. Multiphoton intrapulse interference for SHG. (a) illustrates MII for the case of
a TL pulse, with a constant spectral phase (p(co). Frequency pairs ((01,011) and (ml-Q,
(01+Q) in the fundamental spectrum S(u)) interfere constructively to produce maximum
intensity in the SHG spectrum 8(2)(2a)) at 2001. Constructive interference occurs for any
frequency to] within the bandwidth of the pulse because all frequency pairs are in phase.
As a consequence the SHG signal is maximized for all frequencies and the spectrum
5(2)(20)) shown at the top is obtained. The uniform filling under S(co) indicates that the
spectral phase is constant. Note that frequencies in the unfilled part of the spectrum do
not contribute to the signal at 2031 because no frequencies in the lower frequency part of
the spectrum can pair with them. (b) illustrates MII for a sinusoidal spectral phase qr((o)
with its inversion center at col. In this case, frequency pairs ((01,031) and (col-Q, (01+Q) in
the fundamental spectrum S((n) interfere constructively to produce maximum intensity in
the SHG spectrum Smear) at 2031. For any other frequency (02:0); the SHG is not
maximized because destructive interference takes place. As a consequence the spectrum
Smear) shown at the top is obtained. Note that the sum of the phases of frequency pairs
under 5(a)) which show the same pattern is a constant. Thus, constructive interference
takes place at col.
Let us now consider the effect of simple spectral phases on the SHG spectrum and
the temporal intensity of a pulse. A linear phase function only advances or delays a pulse
in time, but it does not have any effect on the SHG spectrum. The effect of quadratic,
cubic and sinusoidal spectral phases is simulated in Figure 1.2. The fundamental
spectrum (same for all cases) and phase are shown in the lefi column, the simulation of
the temporal intensity of the pulses [(0, calculated using Equation 1.3, is shown in the
center column, and the SHG spectrum is shown in the right column. The SHG spectra
were simulated using
2
8(2) (00) = ”520) exp[ia)t]dtl , (1.7)
which allows to calculate the SHG spectrum by Fourier transforming E2(t) [19]. The thin
lines in the center and right columns correspond to the TL case and are shown for
comparison. A quadratic spectral phase (or chirp, Figure 1.2a) temporally broadens the
pulse and reduces the intensity of Sa)(2w) without changing its shape. A cubic spectral
phase. (Figure 1.2b) temporally broadens the pulse and creates subpulses. It has a
significant effect on the shape of Smear) and generates a maximum in the spectrum that
reaches the intensity of the TL case at the frequency corresponding to inversion center of
the phase. The intensity at other frequencies is relatively low. A sinusoidal spectral phase
(Figure 1.2c) has an effect similar to that of a cubic phase. However, given that it does
not diverge toward infinity its experimental implementation is easier (a cubic phase may
require extremely high retardations at the edges of the spectrum). Again, the maximum in
5(2)(2(0) occurs at the frequency corresponding to inversion center of the phase and the
signal is minimized elsewhere. In contrast, TL pulses maximize SHG (and any NLO
10
process) at all frequencies, but without any selectivity. MII principles have been used to
design shaped pulses able to control NLO processes [19].
In general, any spectral phase with inversion center at (1) will generate SHG signal as
intense as the corresponding to a flat phase (TL case) at frequency 201) in a similar way to
the sinusoidal and cubic spectral phases described before. However, an approximation
that will become useful later states that the condition rp"(or)=0 is enough to have a
maximum at 20). In fact, if we make a Taylor series expansion of the phases at positive
and negative detuning in Equation 1.6 we obtain
(p(a)+.(2) = ¢(w)+¢'(w)0+%¢"(w){22 +%(p"'(w){23 +... (1.8)
rp(a)—.Q) = (p(co)+(o’(a))(—.Q)+ %(o'(w)(—.Q)2 + %(o"(w)(—{2)3 +... (1.9)
Therefore,
(p(a)+.Q)+¢(w—.Q)=2go(w)+¢"(w).(22 +... (1.10)
According to Equation 1.6, Smear) is maximized when (0(ar+.(2)+(a(a>-.Q) is constant.
Equation 1.10 states that (p(a)+.(2)+rp(a)-.(2) is a constant when ¢’(co)=0, if we ignore
terms of even order n24. Therefore, to first approximation, 5(2)(2(n) is maximized when
¢"(03)=0-
The simulations shown in Figure 1.2 illustrate the fact that the spectral phase has a
critical effect on the outcome of NLO processes. For this reason, a method able to
characterize and correct the spectral phase distortions of femtosecond laser pulses at the
target position is necessary to achieve optimal and reproducible experimental results. MII
principles have been the basis for the development of such a method (section 2.3).
11
l I I I
I— —1 r-— —1 r—
1 1 \ 1 1 r
l I : I
l 1 l
I I
.—.q
b— “ _ —-q 1—- —n
— -——t
l
770 800 830 -500 O 500 390 400 410
A (nm) t (fs) 1. (nm)
Figure 1.2. Effect of the spectral phase on the temporal intensity and the SHG spectrum
of a pulse. The left column shows the spectrum (thin line) and spectral phase of the pulse
(thick line). The center and right columns show simulations of the temporal intensity and
SHG spectrum of the pulse (thick line). Simulations for the TL case are shown for
comparison (thin lines). (a) A quadratic spectral phase broadens the pulse and reduces the
SHG spectral intensity without altering the spectral shape. (b) A cubic spectral phase
broadens the pulse and creates subpulses, and alters the SHG spectral shape creating a
maximum at the frequency corresponding to inversion center of the phase. (c) A
sinusoidal spectral phase has an effect similar to a cubic phase. The spectral phase
functions simulated were ¢(w)=3000fsz(co-coo)2, ¢(a))=16667fs3((n-oro)3, and
rp(a))=1.57rcos[35fs(co-c)o)+ 71/3], respectively, where coo is the center frequency of the
spectrum.
Chapter 2
Experimental Tools
2.1 F emtosecond laser systems
The first sub-picosecond laser pulses were produced in the mid-1970’s using organic
dyes as the gain medium [20], more than a decade after the first experimental
demonstration of the laser [21]. These dye-lasers dominated ultrashort pulse generation
research until the late 1980’s and achieved their shortest pulse duration of 6 fs afier
external pulse compression in 1987, a world record not surpassed for about 10 years [22].
The dominance of dye lasers ended soon after the discovery of Titanium-doped sapphire
(Ti3+:A1203), usually abbreviated as Ti:Sa, a new solid-state laser material that had the
necessary broad gain bandwidth to support femtosecond laser pulses and provided long
term stability, unlike the dye-based laser systems that preceded it [23]. Since the early
1990’s ultrashort pulse generation with Ti:Sa lasers has progressed significantly [24].
Pulses as short as 5 fs have been produced directly from a Ti:Sa laser oscillator with
dispersion compensating mirrors in 2001 [25], and with external compreSsion these
pulses reached the world record duration of 4.3 fs in 2005 [26]. Even until today, no other
laser material has produced pulses shorter than 6 fs.
Two different Ti:Sa femtosecond lasers systems, which are described in the
following sections, were employed for the experiments presented in this dissertation.
2.1.1 Ultrabroad-bandwidth Ti:Sa femtosecond laser oscillator
This is a commercial laser system (V enteon Pulse 1, Nanolayers GmbH) whose
design is based on the Ti:Sa oscillator with dispersion compensating mirrors mentioned
earlier [25]. Intracavity temporal pulse broadening is controlled by a combination of
13
chirped mirrors and a Ban wedge pair. This system, pumped by a NszVO4 laser
(V erdi-5, Coherent), generates ~1 nJ laser pulses at 78 Mhz with a spectrum spanning
more than 400 nm (from 620 to 1050 nm). This frequency bandwidth is broad enough to
support a time duration of 4.3 fs. Figure 1.1 shows a scheme of the laser oscillator setup.
oc P
/)
‘ 1
M5
L cm
Pump laser
CM2
M4 M3
ll
M2 W1 W2
IM1
Figure 2.1. Ultrabroad-bandwidth Ti:Sa femtosecond laser oscillator setup. L: lens, CMl
and CM2: curved dispersion compensating mirrors, Ml-M5: dispersion compensating
mirrors, Ti:Sa: Ti:Sa crystal, W1- W2: dispersion compensating Ban wedges, P: Ban
plate, and OC: output coupler.
After external pulse compression, as described in section 3.2, 4.3 fs pulses were
produced with this laser system matching the shortest pulse duration ever obtained
directly from a laser oscillator [27]. These compressed pulses were then used for the
experiments reported in chapters 4 and 5.
2.1.2 Ti:Sa regenerative amplifier
Femtosecond laser pulses generated by a Ti:Sa laser oscillator, typically with a few
ml of energy per pulse, are suitable for weak-field studies such as multiphoton
spectroscopy (see chapter 4) or two-photon laser scanning fluorescence microscopy [7].
However, strong-field experiments such as laser-induced ionization and fragmentation
14
require substantially higher pulse energies and therefore laser pulse amplification is
required [28].
The regenerative amplifier is a device used for laser pulse amplification in which
multiple passes of the pulse through the gain medium, a Ti:Sa crystal in this case, are
achieved by placing the gain medium in an optical resonator together with an optical
switch that lets the laser pulse out once the desired number of round trips in the resonator
(possibly hundreds) is completed. The peak intensity of the laser pulse being amplified
can become so high that detrimental nonlinear effects may distort the laser pulse or
destruction of the gain medium or other optical element can take place. These problems
can be prevented by the use of the chirped-pulse amplification (CPA) technique [29]. In
CPA the pulses are temporally stretched to a much longer duration (typically from under
100 fs to ~100 ps) using a strongly dispersive device, such as a grating pair, before
passing through the amplifier. As a consequence of the substantial temporal broadening
produced by the stretcher, the peak intensity of the pulses is severely reduced (~3 orders
of magnitude) and the detrimental effects mentioned earlier are avoided. The stretched
pulses enter the amplifier where several passes through the amplification medium
(typically 4-50 passes with a gain of 2-100 per pass) provide 6-9 orders of magnitude
increase in the energy [30]. After the amplifier, a compressor, typically a grating pair, is
used to recompress the pulses to a duration similar to that of the initial input pulses (seed
pulses) and therefore a very high peak intensity is produced (Figure 2.2a).
For the experiments described in chapter 5 we used a regenerative amplifier
(Legend USP, Coherent) seeded by a Ti:Sa oscillator (Micra, Coherent). This amplified
15
femtosecond laser system, depicted in Figure 2.2, generates 45 fs pulses at 1 kHz with
energies up to ~1 mJ/pulse.
(a)
Seed pulse
J‘- 8 EM
PC1
Figure 2.2. Ti:Sa regenerative amplifier setup. The amplifier is composed by the
following three elements: (a) Grating-based stretcher, in which the seed pulses are
stretched to a much longer duration, (b) regenerative amplifier, where the stretched pulses
are amplified, and (c) grating-based compressor, where the amplified pulses are
recompressed and a high peak intensity is produced. The curves illustrate the laser pulse
shapes at different stages of the amplification process. FI: Faraday isolator, G1 and G2:
grating, CM: curved mirror, LM: long mirror, PCI and PC2: Pockels cells, and M2: half-
wave plate.
2.2 Pulse shapers
Pulse shapers are devices able to manipulate laser pulses according to the user
specification and thus have become an important tool to control laser-driven processes.
While some simple pulse shaping tasks such imposing a quadratic spectral phase (chirp)
on a pulse can be carried out with passive optics such as prism- or grating-based
compressors, more complex pulse shaping tasks -such as imposing higher order (>2)
polynomial or sinusoidal spectral phases- require the use of programmable pulse shapers.
16
Applications of programmable pulse shapers include femtosecond laser pulse
characterization and compression (chapter 2), control of nonlinear optical processes [19,
31], nonlinear microscopy [7, 32], and coherent control of physicochemical processes
[33, 34]. Programmable pulse shapers can be divided in two categories according to the
domain, time or frequency, in which the laser pulses are manipulated. Acousto-optic
programmable dispersive filters (AOPDF) shape pulses in the time domain using a time-
dependent acoustic signal in a crystal (typically TeOz) in which the pulses propagate to
control both the amplitude and phase of the optical pulses [35]. Fourier-transform pulse
shapers manipulate pulses in the frequency domain by spatial masking of the spatially
dispersed frequency spectrum of the pulses [36]. In simple terms these devices advance
or retard individual frequency components within the pulse. A more detailed description
of the Fourier-transform pulse shaper, the one used throughout this dissertation, is given
in the following paragraphs.
Figure 2.3 shows the Fourier-transform pulse shaper apparatus, which consists of a
pair of diffraction gratings and a pair of lenses arranged in a configuration known as “4f
configuration” or “zero dispersion pulse compressor”, and a spatial light modulator
(SLM) placed at the focal plane. The frequency components of the input beam are
angularly dispersed by the first diffraction grating (a prism can also be used) and then
focused by the first lens (a focusing mirror can also be used). Note that the frequency
components of the pulse are spatially separated along one dimension at the focal plane of
the lens, known as the Fourier plane of the pulse shaper. The second lens and grating
recombine the frequency components and a single collimated output beam is obtained. In
the absence of the SLM the output should be identical to the input pulse (therefore the
17
name “zero dispersion pulse compressor”). In a F ourier-transform pulse shaper a SLM is
placed at the Fourier plane to manipulate the spatially dispersed frequency components of
the beam. Three kinds of SLM have been the most widely used in Fourier-transform
pulse shapers. A purely reflective phase shaper can be implemented with a deformable
f Fourier plane f f
<—> .<—> <——>.
<—£—>
Mirror
Figure 2.3. Scheme of a Fourier—transform pulse shaper. The frequency components of
the input beam are angularly dispersed by a grating and focused by a lens. An SLM is
placed at the focal plane (Fourier plane), where frequency components are spatially
separated along one dimension, to manipulate them and obtain user-designed shaped
output pulses. The second half of the shaper recombines the frequency components of the
light and a collimated output beam is obtained. The setup is known as 4f because the
optical components are separated a distance equal to the focal length of the lens (f).
mirror, a device consisting of a number of independently controlled mirrors, placed at the
Fourier plane [37, 38]. This kind of SLM has a small loss, but the number of optical
elements is typically small and the device is not able to provide amplitude control. An
acousto-optic crystal can also be used as an SLM in a 4f setup [39, 40]. In this device an
acoustic wave produces changes in the refractive index of the crystal that manipulate the
spectral phase of the pulses. This kind of SLM is not limited by the number of pixels, but
its efficiency can be lower than 40%. The most popular kind of SLM is the liquid crystal
(LC) SLM, the device used throughout this dissertation. A LC SLM is basically a thin
18
layer of nematic liquid crystal sandwiched between two pieces of glass, with the inside
surface of each piece coated with a thin and transparent electrically conducting material
patterned into a number of separate electrodes, called pixels. When an electric field is
applied to a pixel, the liquid crystal molecules tilt causing a change of the refractive
index. The liquid crystal is birefringent, therefore the applied voltage can introduce pure
phase retardation or a combination of phase retardation plus polarization rotation
depending on the polarization of the incoming light with respect to the liquid crystal axis.
Both situations are useful for pulse shaping purposes. In a phase-only LC SLM a single
LC layer is used and the polarization of the input light is such that pure phase retardation
is introduced [41, 42]. In a phase-amplitude LC SLM two layers of LC with
perpendicular optical axes at 45° respect to the polarization of the input light are
. combined with an output polarizer to manipulate both the amplitude and phase of the
input beam [43]. Amplitude control is then provided by the attenuation that light suffers
after travelling through the output polarizer, which depends on the degree of polarization
rotation that took place. If no output polarizer is used the polarization of the output pulse
can be shaped and pulses with a frequency-dependent polarization can be synthesized
[44]. Such pulses have been applied for coherent control applications [45 , 46].
The transmission dependence on polarization described before is also useful for
calibration purposes of both phase-only and phase-amplitude pulse shapers. Detailed
calibration procedures have been described elsewhere [33, 36]. Very briefly, a voltage
scan on the SLM results on a transmission curve for each pixel, which can be used for
calibration purposes. Due to the broad frequency bandwidth of the laser systems used in
19
this dissertation and the frequency dependence of the refractive index, an automated pixel
by pixel calibration was required for the pulse shapers.
Venteon mm] """"""""""" NI """"" c {ll/I E
OSC'Hator ! Collimation Telescope l I
M, : M "7' ' '5
i s
: E ‘t M L M;
I M ;
To SHG detection ECM :
Pulse Shaper :
s ; :
CM 5 :
' G I
Figure 2.4. Scheme of pulse shaper I. The beam was first directed to a 1:2.5 telescope
consisting of two spherical mirrors (CM) and a pinhole (P) placed at the focal point of the
first spherical mirror. The all-reflective folded pulse shaper consisted of a spherical
mirror (CM), mirrors (M), grating (G) and SLM.
In this dissertation a folded all-reflective pulse shaper design was used, as shown in
Figure 2.4. In this case a retro-reflection mirror is placed behind the SLM. This design
has two advantages with respect to the linear 4f setup depicted in Figure 2.3. First, it
occupies half of the space. Second, it doubles the phase retardance that the shaper can
introduce because light goes through the SLM twice. The output beam needs to be
displaced vertically so that input and output beams can be separated. Two different pulse
shapers based on this design were used:
Pulse shaper 1, shown in Figure 2.4, was used with the ultrabroad-bandwidth source
described in section 2.1.1. Before entering the pulse shaper, a 1:2.5 telescope was used to
20
collimate and expand the beam to 4 mm diameter for optimal pulse shaping resolution. In
addition, a 150 um diameter pinhole was placed at the focal point of the first spherical
mirror to clean up the beam. The main components of the pulse shaper were an enhanced-
aluminum coated lSO-lines-per-mm grating (Newport), a 762-mm—focal-length gold-
coated spherical mirror, and a 640-pixel dual-mask SLM (CRI, SLM-640).
Pulse shaper II (not shown), was used with the amplified described in section 2.1.2
between the Micra oscillator and the Legend USP amplifier. The main components of the
pulse shaper were an enhanced-aluminum coated 300-lines-per-mm grating (Newport), a
508-mm—focal-length spherical mirror, and a 128-pixel SLM (CR1, SLM-128).
Commercial SLM’s such as those described here provide a retardance of ~41t. When
a larger retardance is required the phase is wrapped or folded into separate 21: segments.
The pulse shapers described before were used throughout this dissertation to
characterize and compress pulses (chapters 3 and 6), to perform chromatic dispersion
measurements of materials (chapter 4) and to generate shaped pulses to measure two-
photon fluorescence excitation spectra of fluorophores (chapter 5).
2.3 Multiphoton intrapulse interference phase scan (MIIPS)
The ability to measure and correct femtosecond laser phase distortions is critical to
obtain Optimal and reproducible experimental results in the sub-50-fs regime. MIIPS is a
single-beam method based on phase shaping that provides both capabilities with
unprecedented simplicity and accuracy. The following description of MIIPS has been
adapted from Y. Coello, V. V. Lozovoy, T. C. Gunaratne, B. Xu, 1. Borukhovich, C. H.
Tseng, T. Weinacht and M. Dantus. “Interference without an interferometer: a different
21
approach to measuring, compressing, and shaping ultrashort laser pulses”. J. Opt. Soc.
Am. B 25, A140 (2008) [27].
Pulse compression, shaping, and characterization at the laser target are of critical
importance to ensure reproducible femtosecond laser applications, which now include:
biomedical imaging, metrology, micromachining, analytical chemistry, material
processing, photodynamic therapy, surgery, and even dentistry. In principle, the Fourier
transform of the ultrashort electromagnetic pulse spectrum provides its temporal duration.
This statement is accurate when the pulse is transform-limited (TL); i.e., all fiequency
components in its bandwidth have the same phase. The actual pulse duration of ultrashort
pulses is always greater than that of the TL pulse because of phase distortions that arise
from optics and from transmission through any medium other than vacuum. Here we use
the ratio ‘t/TTL as a parameter to characterize the quality of the pulse, where t and In are
the time durations of the measured and TL pulses, respectively. Strictly speaking, root-
mean-square time durations should be used for 1:; however, the generalized approach is to
use the FWHM time duration for simplicity. The Thin, parameter is similar to the M2
parameter used in optical design, giving the ratio between the measured value versus the
theoretical optimum. Typical values for ‘t/‘ETL range from 1.1 for well-tuned systems to
less than 1.5 for most advertised commercial systems, and finally from 10 to 100 when
pulses are broadened by optics such as high-numerical-aperture microscope objectives.
Ultrashort pulse broadening is a serious problem affecting every application. One
can divide approaches to dealing with it into two broad categories: direct compression
and phase measurement followed by compensation. For the former approach, phase
distortions are minimized without being measured; the latter depends on accurate phase
22
measurement followed by accurate compensation. The most common approaches to pulse
compression are illustrated schematically in Figure 2.5a through 2.5c. The early
incorporation of compressors consisting of gratings, prisms, and their combination led to
the great advancements in femtosecond technology during the early eighties, culminating
in the production of 6 fs pulses [22]. This approach, which requires one or more laser
experts, is illustrated in Figure 2.5a. A second characterization-free approach uses a
computer-controlled pulse shaper and an optimization algorithm that takes the integrated
second harmonic generation (SHG) intensity from the laser pulses as the feedback in a
closed loop [47, 48], as it is illustrated in Figure 2.5b. In both of these characterization-
free cases, success depends on the noise level of the laser system. The pulse-to-pulse
stability of the SHG output is typically 2-6%, assuming laser fluctuations of 1-3% in the
fundamental. Because T/TTL=ISHG.T[/ISH0, measurement-free approaches could reach t/TTL
values as low as 1.02-1.06 provided the algorithm is given sufficient time to converge.
For many cases this level of performance is sufficient, and using a prism/grating
compressor or even a simple uncalibrated pulse shaper with feedback will accomplish the
task.
If T/‘CTL< 1.1 is consistently required, such as when the ultrashort pulses are used to
study optical properties of materials, an actual measurement of the pulses is required. The
simplest situation arises when well-characterized pulses, such as TL pulses, are used for
measuring phase-distorted pulses. In this situation, the unknown phase distortions can be
calculated from the interferogram between the unknown and the reference pulses.
Unfortunately, well-characterized pulses are not usually available. It has also been shown
theoretically and experimentally that, through an iterative algorithm, one can determine
23
the pulse field from a fiinge resolved autocorrelation and the spectrum of the pulse [49].
However, this algorithm is rarely used. A more common approach is to retrieve the
unknown phase using an autocorrelator/interferometer—based technique such as frequency
resolved optical gating (FROG) [50, 51] or spectral phase interferometry for direct
electric-field reconstruction (SPIDER) [52, 53], and to use the knowledge of the retrieved
phase and a calibrated pulse shaper for pulse compression [35, 54-56]. This approach is
illustrated in Figure 2.5c.
(a) Prism or grating compressor (D) Pulse shaper and optimization algorithm
-> Shaper b SHG ->
Prism/grating m
+ compressor SHG * T ,'
w‘ __________ SLM Voltage Integrated intensity
Integrated intensity ‘\ ,
I
Angles and distances \‘x\ ‘z’
#4
(0) Pulse shaper and FROG or SPIDER (d) MIIPS
—> Shaper 1‘ > Shaper > NLO ->
I
i 4. I
“¢(w): —(p(a)) signal modulation
.\ p I
El Wm) \\ lMIlPS t/
a - — - —- \
i : FROG or SPIDER ,—:
Figure 2.5. Pulse compression approaches. (a) Manual prism/grating compressor
adjustment. (b) Optimization algorithm using the SHG signal as feedback. These two
approaches do not require spectral phase measurements. (c) Measurement and correction
using FROG or SPIDER as the characterization technique. ((1) MIIPS. Measurement and
correction are seamlessly integrated in a compact setup. NLO, nonlinear optical medium.
Here we discuss a different approach, called multiphoton intrapulse interference
phase scan (MIIPS) [18, 57-59], to accurately measure and correct the unknown phase
24
distortions of the pulses, while avoiding the use of autocorrelation or interferometry, as is
illustrated in Figure 2.5d.
We start our discussion by remembering the effect of the different terms of a Taylor
expansion of the spectral phase ¢( 0)) on the time profile of an ultrashort pulse.
_ 1 7- 1 3
(0(0)) - (p0 +(ol (w—w0)+2(p2 (co—r00) + 3(03 (co—coo) (2.1)
The zeroth order phase (00 (sometimes called absolute phase) determines the relative
position of the carrier wave with respect to the pulse envelope. In most cases, the (po term
is of little interest. This is due to the fact that when the pulse is many carrier—wave cycles
long, which is the most common situation, a change in (pa has a very small effect on the
pulse field. None of the pulse characterization methods mentioned in this paper is able to
measure the zeroth order phase. The first order phase (01 corresponds to a shift of the
pulse envelope in time. Given that the interest is typically centered on the pulse shape,
and not on the arrival time of the pulse, the (pl term is also of little interest. The second
and higher order terms do have an effect on the time profile of the pulses. From the above
discussion, it becomes clear that it is the second derivative of the spectral phase
(p'((:)) = $2 + $3 (a) — coo ) - - , the parameter that determines the pulse shape.
MIIPS measures (0"(w) by successively imposing a set of parameterized (p)
reference spectral phases flew) to the pulses with unknown phase distortion (pan) and
acquiring the corresponding nonlinear optical (N LO) spectra, for example SHG. The total
second-derivative of the spectral phase is then (p"(a))-f'(a), p) and maximum SHG
25
intensities will take place at frequencies that satisfy rp"(a))-f'(a),p)=0, as demonstrated
using multiphoton intrapulse interference (MII) principles in Section 1.2. In second
derivative space, the set of reference functions f'(a), p) can be visualized as a grid used to
map the unknown (0"(w), i.e. to find which f’(a),p) intersects (p"(a)) at any desired
fiequency 60,-. The required reference function is simply the one that maximizes the NLO
local signal, i. e.
w"( (01):!" '(wr,Pmax( 601)), (2-2)
where'pmaxmri) is the parameter in the reference phase function for which the NLO signal
is maximized at 0),.
The simplest grid for mapping the unknown second derivative of the phase consists
of constant functions f' (a), p)=p (Figure 2.6a) [60], which correspond to different amounts
of linear chirp. In this case, different amounts of linear chirp can be imposed on the
pulses using passive or adaptive optics. For each reference phase, a NLO spectrum is
plotted as a function of p in a two dimensional contour map (Figure 2.6c). The feature of
interest is pmfio), which can be visualized drawing a line through the maxima in the
contour plot (solid curve in Figure 2.60). The spectral phase information is directly
obtained by finding pmaxflo) and using Equation 2.2. In the case of chirp MIIPS, Equation
2.2 reads (o"(w)=f'(a),pmax(w))=pmax(a)). Therefore, the unknown (p"(a)) is directly
obtained from the contour plot without any mathematical retrieval procedure, as shown in
Figure 2.6c [60].
26
If an adaptive pulse shaper is used, the number of possible reference functions that
can be used is unlimited. Sinusoidal reference spectral phases flu), 6)=asin[)(co-wo)-é],
where 6 is a parameter scanned across a 47: range, have been extensively used [18, 58,
59]. When the NLO signal is plotted as a firnction of a) and f’(w, p), the results obtained
from applying any type of reference phase function reveal the unknown rp"(a1) by finding
the line that goes through the maxima in the contour plot. Figures 2.6d through 2.6f
illustrate the case of sinusoidal reference functions. The dashed lines in Figure 2.6d
correspond to the second derivative of the reference frmctions, f'(a), 6)= -a}}sin[}(a)-wo)-
a]. The sinusoidal MIIPS approach has been described in great detail elsewhere [19, 59].
Experimental measurements using both chirp and sinusoidal MIIPS are presented in this
dissertation.
A MIIPS scan takes between 5 and 15 3 depending on the device used to introduce
the reference phases and the number of phases used. Although not necessary in all cases,
an iterative measurement-compensation routine can be used to achieve the maximum
possible accuracy, especially in the case of complex spectral phases [19, 59, 60]. Double
integration of the measured (0"(co) results in Mar). Once (p(a)) is obtained, the introduction
of -¢(ar) by the shaper eliminates the measured phase distortions to achieve TL pulses. A
comprehensive analysis of the precision and accuracy of MIIPS was carried out in 2006
[59]. Using MIIPS, t/tTL values routinely reach the 1.01 level and in some cases are even
lower than 1.001.
MIIPS has been used in this dissertation to characterize and compress pulses
(chapters 3, 5 and 6) and to perform chromatic dispersion measurements of materials
(chapter 4). Other reported applications of MIIPS include standoff chemical detection
27
[61], coherent control of molecular fragmentation [28], two-photon laser scanning
fluorescence microscopy [7, 32], and micromachining [62].
Chirp MIIPS Sinusoidal MIIPS
k3) 'v """""""" 1 : 5 :‘L ji('d),""~‘
§
Figure 2.6. Principle of MIIPS. A set of reference functions f'(a), p) provides a reference
grid that is used to map the unknown phase ¢"(a)). (a)-(c) and (d)-(f) illustrate the case of
a horizontal and sinusoidal reference grid, respectively. (a) The unknown (0"(a1) (solid
line) is probed using a horizontal reference grid (dashed horizontal lines). (b) shows the
SHG spectra corresponding to four reference phases. The maximum SHG intensity for
every frequency indicates that the corresponding reference chirp value intersected the
unknown function at such frequency. (c) shows the corresponding MIIPS trace, a 2D
contour plot showing the SHG intensity as a function of a) and p. The line drawn through
the maxima in the countour plot directly reveals the unknown ¢"(w). (d) The unknown
¢"(a)) (solid line) is probed using a sinusoidal reference grid (dashed curves). (b) shows
the SHG spectra corresponding to three reference phases. The maximum SHG intensity
for every frequency indicates that the corresponding reference sine function intersected
the unknown function at such frequency. (c) shows a 2D contour plot showing the SHG
intensity as a function of a) and f’(ar p). When the SHG intensity is plotted this way, the
line drawn through the maxima in the contour plot directly reveals the unknown qr”(a)).
28
Chapter 3
Femtosecond laser pulse characterization and compression using MHPS
The advantages of using femtosecond laser pulses in the sub-50 fs regime cannot be
realized if an accurate pulse compression method is not employed. For instance, if 50 and
10 fs TL pulses propagated through 1 cm of quartz, their durations would increase to 54
and 100 fs, respectively. The initially shortest pulse (10 fs) would suffer more extensive
pulse broadening (a 10-fold increase in its temporal duration) and would end up being the
longest. This example illustrates the fact that pulse broadening becomes more severe as
the time duration decreases. This phenomenon occurs because shorter pulses have
broader spectral bandwidths and therefore experiment bigger spectral phase distortions
due to material dispersion. The lack of an accurate method able to compress the pulses at
the sample position by correcting the phase distortions introduced by optical media
present in the experimental setup explains the fact that most multiphoton imaging
experiments, in which signal is inversely related to the temporal duration of the pulses,
still employ ~100 fs pulses [6] even though laser systems generating substantially shorter
pulses (<10fs) durations have been commercially available for several years.
This chapter presents results on femtosecond laser pulse characterization and
compression using MIIPS, a method that satisfies the above requirements and allows
achieving optimal and reproducible results in applications using sub-SO fs pulses. An
illustrative example of the potential of MIIPS to enhance nonlinear imaging applications
was reported in 2008, when numerous advantages resulting from MIIPS pulse
compression were demonstrated for two-photon laser scanning fluorescence microscopy
(TPLSM) using 10 fs pulses, including higher fluorescence intensity, deeper sample
29
penetration, improved signal-to-noise ratio, and less photobleaching. These advantages
were not observed if only quadratic phase distortions were compensated for while higher
order phase distortions were not [7].
The chapter is organized as follows. Section 3.1 describes selected spectral phase
measurements, while section 3.2 demonstrates pulse compression for the two laser
systems employed in this dissertation. All SHG spectra simulations were calculated using
Equation 1.7.
3.1 Spectral phase measurements
The spectral phase measurement examples presented here were performed using the
ultrabroad-bandwidth laser oscillator described in section 2.1.1. Unfortunately, there was
no available commercial filter able to separate the SHG from the firndamental spectrum
due to very broad bandwidth (almost one octave) of both spectra. A light separation
device was built for this purpose. Figure 3.1 shows the setup used for these experiments,
which includes the device for SHG generation, separation and detection. Alter pulse
shaper I (section 2.2) the beam is focused by a ZOO—mm-focal-length silver-coated
spherical mirror (SM) onto a 20m KDP (potassium dihydrogen phosphate) crystal (C) to
generate the SHG signal. Both the fundamental (dashed line) and SHG (solid line) beams
are collimated by lens L1 and then directed to the separation device, a quartz-prism-based
folded 4f Fourier Transform pulse shaper. A razor blade (R) was placed at the Fourier
plane to block fundamental frequency components while allowing the reflection of the
SHG light by the retro-reflection mirror (RM). The output of the separation device was
focused by lens L3 into the spectrometer (QE65000, Ocean Optics Inc.). No optical fiber
was used to optimize the transmission of SHG light. All lenses were made of quartz and
30
all mirrors after the crystal were coated with protected aluminum to avoid losing SHG
light.
Ultrabroad-bandwidth _ _ Pulse
laser oscillator shaper I
I
M s
l s
__ L15 :3
M : f‘g’
g C t i g
Spectrometer I ‘1‘: .5,
SHG separation and \i a)
detection S'Lj‘,
Figure 3.1. MIIPS optical setup for the ultrabroad-bandwidth femtosecond laser system.
The beam (dashed line) from pulse shaper l is focused by a spherical mirror (SM) onto a
KDP crystal (C) to generate the SHG beam (solid line). Both beams are collimated by
lens L1 and directed a quartz-prism-based folded 4f Fourier Transform pulse shaper (P:
prism, L2: lens, RM: retro-reflection mirror) with a razor blade (R) placed at the Fourier
plane to block all fundamental frequency components while allowing the reflection of the
SHG light by RM. The output beam from the separation device is focused into the
spectrometer by lens L3. M: mirror.
Some of the results presented in this section were originally published in V. V.
Lozovoy, B. Xu, Y. Coello and M. Dantus. “Direct measurement of spectral phase for
ultrashort laser pulses”. Opt. Express 16, 592 (2008).
Quadratic and cubic spectral phases are the most commonly encountered phase
distortions because they are introduced by material dispersion (section 1.1). Both
correspond to a linear function in second-derivative space. To demonstrate the ability of
MIIPS to measure this kind of spectral phase, the cubic function (p(cu)=500fs3((o-co0)3 was
31
added to TL pulses (how to obtain TL pulses is described in next section) using the pulse
shaper and then measured using chirp MIIPS. Figure 3.2 shows experimental (Figure
3.2a) and simulated (Figure 3.2b) chirp MIIPS traces of the pulses. For the simulated
spectra, the cubic spectral phase was added to perfectly TL pulses (rp(a))=0).
2000 . . r - . - . - . - 1
.(a) . ,(b)
1000- - - .
NA 1 if ‘ i
.2
‘V o. q . .
‘8 ,x’ h .3/
-1000-1 --------------- I - :. ------------------
i l
-2000 . . . - . T - . - , - .
350 400 450 500 350 400 450 500
Wavelength (nm) Wavelength (nm)
1 . - r - I -
(C)
SHG intensity (arb. u.) '
0
0'!
O
r
350 400 450 500
Wavelength (nm)
Figure 3.2. Experimental and simulated chirp MIIPS traces demonstrating accurate pulse
shaping. The cubic spectral phase defined by (p((1))=500fs3(co-n)0)3 was introduced to TL
pulses. (a) Experimental chirp MIIPS trace. (b) Simulated chirp MIIPS trace. The dashed
line indicates the chirp reference value used for (c). (c) Experimental (dotted curve) and
simulated (gray solid line) SHG spectra corresponding to the dashed line in (a) and (b).
The excellent agreement between experiment and simulation demonstrates accurate pulse
shaping capability.
32
For both the experiment and simulation, the chirp reference functions were varied in steps
of 5fsz. Figure 3.2c shows the experimental and simulated SHG spectra corresponding to
the dashed line in Figs. 3.2a and b. The excellent agreement between experimental and
simulated results demonstrates accurate pulse shaping. If the cubic and chirp phases were
not introduced accurately, this agreement would not be possible.
Figure 3.3a shows the normalized chirp MIIPS trace, in which the maximum SHG
intensity for each frequency was set to the same value, to visualize the measured 1;)" more
clearly. As explained in section 2.3, (0" is directly obtained from the curve drawn through
the maxima in the trace, without any mathematical treatment. Figure 3.3b shows the
resulting measured rp" (dotted line) together with the calculated second-derivative of the
introduced phase (solid gray line). Note the agreement between both. The only points that
deviate from the expected values correspond to the edges of the fundamental spectrum,
where insufficient SHG light is generated. These deviations have little or no effect in the
calculated spectral phase, which is obtained after double integration. Figure 330 shows
the resulting spectral phase (9(a)) (dotted line) together with the introduced one (solid gray
line). Note the excellent agreement between both across the whole bandwidth. An
independent cross-check can be performed by comparing the experimental and simulated
SHG spectrum corresponding to the introduced phase. Figure 3.3d shows that
comparison, which confirms the accuracy of the measurement.
The availability of automated pulse shapers has made possible the generation of
complex spectral phases, which are used in areas such as coherent control [19, 34] and
nonlinear microscopy [7, 18, 32]. To demonstrate the ability of MIIPS to measure
33
2000 1 1 150
1000 3: 13 moo
0"; E 3 L50
3”; 0 a 3 -
x 2 ...... a ~0
2 9; .
- C I
1000 _ , "50
-2000 . . . f . o ' . 1 . 1 . . . -100
350 400 450 500 700 800 900 1000
2000 I I ' l ' I ' I 1 ' I ' ‘ ' I ' I
5 (d) ,,
3
2' 1
3
> 1
'17? . r .
C
2 1
.E t
(D
:1:
(D
-2000 1 . , . 1 a , 0-4;
700 800 900 1000 350 400 450 500
Wavelength (nm) Wavelength (nm)
Figure 3.3. MIIPS measurement of a cubic spectral phase. The introduced phase function
was ¢(w)=500fs3(co-oro)3. (a) Normalized chirp MIIPS trace. (b) Measured second-
derivative of the phase (dotted line), which directly corresponds to the line drawn through
the maxima in (a). The calculated second-derivative of the introduced phase (solid gray
line) is shown for comparison. (c) Measured (dotted line) and introduced (solid gray line)
cubic spectral phase. The agreement across the whole bandwidth is excellent. ((1)
Measured (dotted line) and simulated (solid gray line) SHG spectrum corresponding to
the introduced phase. The agreement confirms the accuracy of the spectral phase
measurement.
complex spectral phases, the sinusoidal function ¢(w)=51tsin[7fs (or-030)] was added to
TL pulses using the pulse shaper and then measured using chirp MIIPS. The chirp
reference functions were varied in steps of 5fsz. Figure 3.4a shows the normalized chirp
MIIPS trace, from which (0" is directly obtained, as shown in Figure 3.4b. Figure 3.4c
shows the measured (dotted line) and the introduced (solid gray line) spectral phases.
34
Finally, Figure 3.4d shows the experimental and simulated SHG spectrum corresponding
to the introduced sinusoidal phase. The agreement between the introduced and measured
phases illustrates the performance of MIIPS for the case of complex spectral phases.
Intensity (arb. u.)
.0
3'
(D
M
(D
A
'fi
9)
O.
V
an F
_> "_\'
O
760 ' 860 ' 960 '1o'oo
(d)
SHG intensity (arb. u.) _.
L
o
o
o
o
l
700 800 900 1000 350 400 450 500
Wavelength (nm) Wavelength (nm)
Figure 3.4. MIIPS measurement of a complex spectral phase. The introduced phase
function was ¢(w)=51tsin[7fs (or-coon. (a) Normalized chirp MIIPS trace. (b) Measured
second-derivative of the phase (dotted line), directly obtained from the trace in (a). The
calculated second-derivative of the introduced phase (solid gray line) is shown for
comparison. (c) Measured (dotted line) and introduced (solid gray line) cubic spectral
phase show an excellent agreement. ((1) Measured (dotted line) and simulated (solid gray
line) SHG spectrum corresponding to the introduced phase. The agreement confirms the
accuracy of the spectral phase measurement.
35
3.2 Pulse compression
Pulse compression is a straightforward task provided the spectral phase of the pulses
is known. As demonstrated in the previous section, MIIPS is able to accurately measure
the spectral phase (0(a)) of femtosecond laser pulses. After the measurement, the
introduction of 10(0)) by the shaper eliminates the measured phase distortions to achieve
TL pulses.
Wavelength (nm)
Angular frequency (rad/fs)
l
S” .01 .
o- o
A L A
5 (rad)
Figure 3.5. Characteristic sinusoidal MIIPS traces. The experiments were performed
using the setup depicted in Figure 3.1. (a) Sinusoidal MIIPS trace for TL pulses. The four
diagonal features are equally spaced and have the same slope. (b) Sinusoidal MIIPS trace
showing the change in spacing between the diagonal features caused by a rp"=120 fs2
quadratic phase distortion in the frequency domain. (c) Sinusoidal MIIPS trace showing
the slope change caused by a p" =336 fs3 cubic phase distortion in the frequency domain.
36
Pulse compression has been performed routinely on the two laser systems employed
in this work using sinusoidal MIIPS. Due to historical reasons, the SHG intensity has
been plotted as a function of frequency and the parameter 6 rather than as a function of
frequency and the reference function in sinusoidal MIIPS traces, i. e. ISHG(0),8) instead of
Isyg(c),f(w,5)). Consequently, the measured rp" is not directly visualized in the sinusoidal
MIIPS trace, as was described in section 2.3. Instead, diagonal parallel lines separated by
1t are obtained for Jmax(a)) when the pulses are transform-limited (Figure 3.5a). Quadratic
phase distortions cause a change in the spacing between the lines (Figure 3.5b), and cubic
phase distortions cause a change in the inclination of these lines (Figure 3.5c). In these
two last cases cases, the changes are proportional to the magnitude and sign of the
distortion.
3.2.1. Compression of ultrabroad-bandwidth femtosecond laser pulses
This section describes pulse compression carried out with MIIPS for the laser
system described in 2.1.1. The setup depicted in Figure 3.1 was employed. A
representative fundamental spectrum of this system is shown in Figures 3.3 and 3.4.
The ability to measure and correct the spectral phase of a femtosecond laser
becomes more challenging as the spectral bandwidth increases. Furthermore, when
dealing with sub—5 fs pulses, the task represents a significant challenge. Actually, there
was a single report on compression of sub-5 fs pulses before 2005 [63] and a couple more
appeared in 2005-2006 together with our first report on pulse compression of this laser
system [26, 64, 65]. Since that first demonstration [65], pulse compression for this system
was achieved routinely in our laboratory.
37
A c, _____________ 1:593 fs " A
m- .
E v t/ TTL=1 36 2,
E18 Atp=160 rad é
a " » a
an l 2
3 ,.~. '1 2
($3 . ____________ > ‘ ‘4“.",A‘l( ““311"; gm)“ ,) “ <
.1
Ao ( ) t=47 fs 1 A
E ‘3 t/rTL—11 l =
E I f.
D v
c ’1 ' ‘
% '8 r \ r \ (ll 5
5 A(p=10 rad g
3 8 1 E
to ‘ .
A c, (0) =4.5 fs 1 A
g 3 T/TTL=1.02 i =:
g ‘3
5 §- ’3 g
o .
a Acp=2 rad i E
' E
3 33 l
m 1
Ac, (d) t=4.4fs i 1 ‘. ‘ A
E3 ‘E/‘ETL=1.0021 1 >. ,< 1 :1
z -2 1%
E g, F < E
2 V x \ ‘ a
2 ’ s
g § F Atp=0.5 rad l g
, 11
o n 211: 311: 41! 700 800 900 1000 -200 o 200
5 (rad) Wavelength (nm) 1' (f3)
Figure 3.6. MIIPS compression process of ultrabroad-bandwidth femtosecond laser
pulses from ~600fs to sub-5fs TL time duration. In this experiment, ultrabroad-bandwidth
femtosecond laser pulses with severe spectral phase distortions were characterized and
compressed to their TL pulse duration of 4.4 fs using MIIPS. The left, middle and right
columns show the sinusoidal MIIPS traces, measured spectral phases and calculated
temporal profiles for each MIIPS scan, respectively. (a), (b), (c) and ((1) correspond to the
first, second, third and fourth MIIPS scans. Note that the first two MIIPS measurement-
correction iterations, (a) and (b), are enough to compress the pulses, as revealed by the
diagonal parallel features in the MHPS trace corresponding to the third scan (c), which
indicate TL pulses. Quantitatively, the pulses were compressed from T/TTL=136 to
t/tn,=1 .02 in the first two scans. After the inverse of the phase distortion measured in the
third scan was applied to the pulses, further compression to T/tTL=l.002 was obtained, as
indicated by the fourth scan (d).
38
Figure 3.6 illustrates an example pulse compression procedure after initially TL
pulses of 4.4 fs duration were temporally broadened by propagating through 2 cm of
water. The left, middle and right columns in Figure 3.6 show the MIIPS traces, measured
spectral phases and calculated temporal profiles for each scan, respectively. Figure 3.6a
corresponds to the first scan. Note that the trace (left panel) indicates the presence of both
quadratic and cubic components in the phase distortion because of the spacing and slope
of the diagonal features, respectively. The measured spectral phase (middle panel) reveals
a distortion spanning 160 rad across the spectrum. The calculated temporal profile (right
panel) indicates severe temporal broadening to 593 fs, corresponding to ‘t/TTL=136. The
inverse of the measured phase is added to the pulses after each scan. Consequently, the
MIIPS features are much less distorted in the second scan, corresponding to Figure 3.6b,
but still show altered spacing and curvature (left panel). The corresponding spectral phase
now spans 10 rad (middle panel) and the FWHM time duration of the pulses is 47 fs
(right panel). The resulting diagonal parallel features in the third scan (Figure 3.6c, left
panel) indicate that the pulse compression procedure is complete. Note that the MIIPS
procedure could have been stopped in the second scan and 4.5 fs pulses, corresponding to
T/‘CTL=1.02, would have been obtained after applying the inverse of the measured phase.
This level of pulse compression is enough for most applications. However, after a third
scan (Figure 360) further pulse compression corresponding to T/‘tTL=l.002 is achieved.
The fourth scan was performed only to measure the compression level after the third
MIIPS scan.
Slight variations in the fundamental spectrum of the pulses due to different
oscillator settings lead to consequent slight variations of the duration of the pulses in the
39
range 4.3-4.6 fs. The shortest pulse duration achieved with this laser system (4.3fs)
matched the shortest pulse duration ever obtained directly from a laser oscillator [26].
.0
l
a s a.
Phase (rad)
Intensity (arb. u.)
.0
.3
.0
O
.1.
llllll
SHG intensity (arb. u.)
E 3 . fl . . . . . .
300 350 400 450 500
Wavelength (nm)
Figure 3.7. Results of MIIPS compression for ultrabroad-bandwidth femtosecond laser
pulses. (a) Spectrum of the ultrabroad—bandwidth femtosecond laser pulses. The longer
wavelength edge of the spectrum looks noisier than rest because of the lower spectral
response of the spectrometer in that region. The spectral phase was corrected within 0.1
rad accuracy across the whole bandwidth (top panel). The resulting FWHM time duration
was 4.3 fs (inset), the shortest ever obtained directly from a laser oscillator. (b) Measured
SHG spectrum (solid curve) of the pulses after compression, the broadest UV spectrum
ever obtained directly from an oscillator and a nonlinear crystal. The simulated SHG
spectrum is also shown (dashed curve). The response function of the crystal was not
considered in the calculation.
40
The results shown in Figure 3.7 were originally published in Y. Coello, V. V. Lozovoy,
T. C. Gunaratne, B. Xu, 1. Borukhovich, C. H. Tseng, T. Weinacht and M. Dantus.
“Interference without an interferometer: a different approach to measuring, compressing,
and shaping ultrashort laser pulses”. J. Opt. Soc. Am. B 25, A140 (2008) [27]. Figure
3.7a shows the fundamental spectrum of the pulses together with the residual spectral
phase after MIIPS compression (top panel). The inset shows the calculated temporal
profile corresponding to 4.3fs FWHM. The compressed pulses generated the SHG
spectrum shown in Figure 3.7b (solid line), the broadest UV spectrum obtained directly
from an oscillator and a nonlinear crystal.
The broad UV spectrum generated by the compressed pulses (Figure 3.7b)
represents additional solid evidence of successful pulse compression. Even though, an
extra independent cross-check of the compression result was performed by comparing the
experimental interferometric autocorrelation (IAC, also known as fringe-resolved
autocorrelation) [51] of the pulses with the simulated IAC of perfectly TL pulses
(¢(a))=0). IAC experiments require splitting the pulse in two, variably delaying one with
respect to the other, and spatially overlapping both collinearly propagating pulses in a
nonlinear optical medium such as an SHG crystal. This is typically achieved by placing a
SHG crystal at the output of a Michelson interferometer. The IAC, obtained plotting the
SHG intensity versus time delay, can be used to estimate the time duration of the pulses.
Using the pulse shaper, two identical replicas of a pulse separated by variable time delays
can be generated and thus an IAC experiment can be performed without the
' complications that an interferometric setup involves. The pulse shaping functions
required for this task have been described elsewhere [27]. Here, the experimental and
41
simulated IAC of the pulses are shown in Figure 3.8 to further demonstrate successful
pulse compression.
8 I l I I I l I f T f
. 0 Experiment 1
— Simulation
O)
l
l
l
A
I
v
Intensity (arb. u.)
A
l
N
l
l
0
q
I 1 I I I I l I
1
-30 -20 -10 O 10 2O 30
Time delay (fs)
Figure 3.8. Pulse-shaper assisted and simulated IAC of TL sub-5fs laser pulses. The
dotted curve shows the experimental pulse-shaper assisted IAC of the compressed pulses,
while the solid gray line corresponds to the simulated IAC of perfectly TL pulses
(¢(w)=0). The excellent agreement indicates successful and accurate spectral phase
correction.
3.2.2. Compression of regeneratively amplified femtosecond laser pulses
This section describes MIIPS spectral phase correction of the amplified laser system
described in section 2.1.2. Pulse shaper 11 (section 2.2) was placed between the laser
oscillator and the regenerative amplifier to avoid damage to the SLM due to the high
peak intensity of the amplified pulses (Figure 3.9). The goal in this case was to achieve
TL pulses after focusing the beam with a 5X a microscope objective, as will be required
for the experiments described in chapter 6. The SHG crystal was placed ~l cm before the
42
focal plane of the objective to avoid damage to the crystal. In contrast to the SHG
separation device required for pulse compression of ultrabroad-bandwidth pulses
described previously, in this case a BG40 filter was enough to block the fundamental
light while allowing the transmission of SHG light, as shown in the setup depicted in
Figure 3.9. The SHG light was collected with an optical fiber and directed to a
spectrometer (U SB4000, Ocean Optics Inc.).
Regenerative Pulse Laser
amplifier shaper H oscillator
, _ [Spectrometer]
5X objective *
Mirror
SHG crystal Filter
Figure 3.9. Optical setup for MIIPS spectral phase correction of the Ti:Sa regenerative
amplifier laser system. The pulse shaper was placed between the laser oscillator and the
regenerative amplifier. The amplified pulses were focused with a 5X microscope
objective ~1 cm after the SHG crystal to avoid damaging the crystal. The fundamental
light was blocked with a filter and the SHG light was collected with an optical fiber and
directed to the spectrometer.
Pulse compression for this laser system was performed routinely. Typically, the
built-in grating-based compressor in the amplifier was used to precompress the quadratic
component of the spectral phase distortions by varying the grating position until the SHG
spectrum was maximized at ~400nm. After this precompression step, MIIPS was used to
measure and correct the spectral phase. Figure 3.10 shows a representative measured
spectral phase together with the spectrum of this laser system. Note that the measured
43
phase is a cubic function as expected because the quadratic component of the phase
distortion introduced by the microscope objective was previously compensated with the
grating compressor. After phase correction, ~45 fs TL pulses were obtained and used for
the experiments described in chapter 6.
I 1 ' 1 1 r 1
10- »
.3
A F.-
‘o 5- 4— (DD
(0
s- 2.
V u-o-
ar 0 ‘<
(I) " ..
co 1.;
i .0
-5, c
.L/
-10..
_ 0
I ' I ' l ' I ' l ' I
770 780 790 800 810 820 830 840
Wavelength (nm)
Figure 3.10. MIIPS measurement of the spectral phase distortion introduced by a 5X
microscope objective on amplified femtosecond laser pulses. The fundamental spectrum
and measured spectral phase are shown. The phase is close to a cubic firnction because
the grating-based compressor in the amplifier was used to correct the quadratic
component of the overall phase distortion introduced by the 5X microscope objective.
After compensating the measured phase, 45 fs TL pulses were obtained.
3.3 Conclusions
Spectral phase distortions on ultrabroad-bandwidth femtosecond laser pulses,
including those introduced by material dispersion, were accurately measured using
MIIPS. As a result, these pulses were compressed to their TL duration of 4.3 fs. This
challenging example illustrates MIIPS compression performance, which allows taking
44
advantage of the use of sub-50 fs pulses in a variety of applications, including nonlinear
imaging.
45
Chapter 4
Chromatic dispersion measurements with MIIPS
Knowledge of the dispersion characteristics of optical media is necessary for pulse
propagation models. For instance, the dispersion parameters of certain biological tissue
allow calculating the amount of temporal broadening that a pulse will suffer after
traveling through the tissue. Similarly, the amount of pulse precompression required to
achieve TL pulses after propagation through the tissue can be calculated if the dispersion
parameters are known. This chapter demonstrates how accurate dispersion measurements
of optical media can be obtained using MIIPS. The utility of this kind of measurement is
not only important for nonlinear imaging applications of femtosecond laser pulses. For
this reason, example measurements relevant for other areas such as surgery are
considered here.
The content of this chapter has been adapted from Y. Coello, B. Xu, T. L. Miller, V.
V. Lozovoy and M. Dantus. “Group-velocity dispersion measurements of water,
seawater, and ocular components using multiphoton intrapulse interference phase scan
(MIIPS)”. Appl. Optics 46, 8394 (2007) [66].
The increased use of femtosecond lasers requires more accurate measurements of
the dispersive properties of media. Here we measure the second and third order
dispersion of water, seawater, and ocular components in the range 660-930 nm using
MIIPS. Our direct dispersion measurements of water have the highest precision and
accuracy to date. We found that the dispersion for seawater increases proportionally to
the concentration of salt. The dispersion of the vitreous humor was found to be close to
46
that of water. The chromatic dispersion of the comea-lens complex was measured to
obtain the full dispersive properties of the eye.
4.1 . Introduction
The growing number of femtosecond lasers in industry, medicine, and
communications has increased the need for measuring the dispersive properties of media
beyond that of glass and quartz. Because of their broad bandwidth, femtosecond lasers
are particularly sensitive to chromatic-dispersion characteristics of materials, in particular
second-order (k") and third-order dispersion (k’"), which typically cause pulse
broadening.
Pulse duration is a very important parameter in femtosecond laser applications, for
example laser micromachining and laser eye surgery, because it determines the peak
power density available to ablate the material. If substantial broadening takes place, the
ability of the laser to achieve consistent ablation is greatly diminished. In this article we
carry direct measurements on the chromatic dispersion of water, seawater, and ocular
components.
Femtosecond lasers are routinely used for opening the corneal flap in the bladeless
LASIK technique. A number of additional procedures are presently under investigation.
In vitro experiments have shown that femtosecond laser ablation may be useful for the
treatment of glaucoma by making channels through the trabecular meshwork in the eye
without damaging the surrounding tissues. These channels provide a pathway for the
release of fluid and may result in a significant intraocular pressure reduction in vivo [67].
Femtosecond laser surgery on retinal lesions appears to be a promising treatment for
macular degeneration [68]. More recently, intratissue multiphoton ablation in the cornea
47
hr
[166
C01
01
C1
has been demonstrated opening the possibility of treating visual disorders without the
need of corneal flaps as used for LASIK [69]. Femtosecond laser cuts in the lens without
damaging adjacent tissue is being developed for the treatment of presbyopia, a very
common disease with no satisfactory treatment presently available [70, 71].
All of the applications above can be greatly improved using the shortest
femtosecond laser pulses, taking advantage of the reduced energy required to achieve a
specific peak power density. Less energy implies less collateral damage to healthy tissue.
However, sub-SO-fs pulses undergo significant broadening by transmitting through
optical media including ocular components. Therefore, accurate dispersion measurements
and a means to eliminate phase distortions, as presented here, will be required to
consistently achieve the best results. Conversely, from the point of view of laser eye
safety, femtosecond pulses that achieve their shortest duration at the retina pose the
greatest risk [72].
We start our discussion by remembering that the wavenumber is defined by
kEaM(w)/52nn(/t)/A, where a) is the angular frequency of light, 11 is the refractive index
of the medium, c is the speed of light in vacuum, and l is the free-space wavelength of
light. The phase retardation that light with frequency a) experiences is given by
(0(w)=k(a))z, where z is the path length traveled by the light. As a broad-bandwidth
femtosecond laser pulse propagates, it undergoes chromatic dispersion ((p'édzp/dwz),
second-order dispersion (k"stk/dwz), and third- order dispersion (k’ "25113 k/dw3). For a
Gaussian pulse the second-order dispersion k" is equal to the group-velocity dispersion
(GVD=_d(1/ug)/da), where 08 is the group velocity) [73]. The chromatic dispersion qr" for
a Gaussian pulse is also known as group-delay dispersion (GDD).
48
Another common definition of group-velocity dispersion is the derivative of the
group velocity with respect to wavelength dug/d1. This is sometimes a source of
confusion because the strictest definition of group-velocity dispersion is dug/dam
According to either of these definitions the group-velocity and group-delay dispersion are
not only functions of the media but also functions of the shape of envelope of the field.
Here, we measure qr" and k", which depend on the media only.
While the refractive index of materials can be directly measured using a hollow
prism arrangement [74], the measurement of k” is more difficult. It can be calculated
using an analytical formula for the index of refraction according to
2
“(fizz/132 d n(1).
(4.1)
7rc (12.2
However, because derivatives are very sensitive to noise and more importantly
depend on the phenomenological formula used to fit the data, this indirect method has
unpredictable precision and accuracy. One of the most accurate methods used to measure
dispersive properties is white-light interferometry, where an interferometer is constructed
using a broad-bandwidth source of light and the material is introduced in one of the arms
[75]. In this case, the phase distortions introduced in the sample arm can be measured
directly, and usually after decomposition in a Taylor series, the second derivative (0" can
be extracted. There is also a variety of time-of-flight, phase-shift, interferometric
measurement techniques [76]. These methods are limited by their temporal resolution and
are time-consuming because they depend on wavelength and/or time scans.
Here we report on the use of MIIPS to measure the second-order dispersion k" of
deionized water and compared it to literature values. We then measured the dependence
49
of k" on the concentration of sea salt in water and k" of the ocular vitreous humor. Finally
we measured (0" of the comea-lens complex. These two eye components together, the
vitreous humor and the comea-lens complex, account for all the transparent parts of the
eye.
4.2. Experimental section
For these experiments we used the ultrabroad-bandwidth femtosecond laser
oscillator (section 2.1.1), pulse shaper I (section 2.2) and the SHG generation, separation
and detection device described in section 3.1. The detailed setup is shown in Figure 3.1.
The laser system used provides measurements of chromatic dispersion in the very broad
spectral range without tuning the laser or realigning the optical system.
Before making the measurements, it is important that phase distortions, including
those introduced by empty glass cuvettes or slides, are eliminated using MIIPS. Once the
phase distortions of the system are eliminated, the desired medium with thickness 2 is
introduced, as shown in Figure 4.1, and its chromatic dispersion rp" as function of
wavelength or frequency is measured using MIIPS.
Femtosecond I 7 Pulse shaper I > Medium
osc1|lator
I
F-——J v
Computer - - - - Spectrometer L - - - SHG crystal
Figure 4.1. Block diagram of the experimental setup for MIIPS GVD measurements. The
solid and dashed arrows indicate the propagation of the laser and SHG beam,
respectively; the dashed lines indicate computer communication.
50
An example measurement is presented in Figure 4.2a for the case of different path
lengths of water. The lines in Figure 4.2a correspond to a fit of the experimental data set
500 -
250 .
760 1 750 I 800 ' 860 . 960
,1 (nm)
800
03) 700i r‘
(0"(2, 800nm) = k"z, *
600-3
II 2: ." i
500_ R 0.9995
4009
300!
2001
1009 -""
<0” (r32)
2 (mm)
Figure 4.2. Chromatic dispersion as a function of medium thickness. (a) (p"(7\.) data points
measured by MIIPS and second-order polynomial fits for water samples of 5, 10, 20 and
30 mm thickness (ascending order). (b) Linear regression of (p "(1.) at 800 nm.
containing hundreds of points to rp"(l)=a+bl+c/12. We confirmed that the results are
independent from the fitting function selected. In the MIIPS measurements of ¢"we used
51
7= 6 fs and a=21t (see section 2.3). For large phase distortions, the first MIIPS iterations
were carried out with larger or values (up to 101:), and as the distortions were eliminated,
or was reduced. The 6 parameter was scanned across the 4n rad range in 128 steps.
Given that the chromatic dispersion varies linearly with the sample thickness, p" =
k"z, the second-order dispersion k"(l) was calculated from the $10pe of the dependence at
each wavelength (see Figure 4b). The error bars reported for k" correspond to the
standard deviation of the slopes for three sets of independent experiments.
Deionized water with a room-temperature mean conductivity of 2uS/cm was used in
all cases. Artificial seawater was prepared by dissolving Instant Ocean© sea salt in
deionized water with concentrations of 35.8 g/L and 107.4 g/L for the 1x and 3x samples
respectively. Four glass cuvettes with path lengths of 5, 10, 20 and 30 mm were used for
the water and seawater measurements. Fresh (uncured) adult about 18-month-old
Holstein cow eyes, that would have otherwise been discarded, were obtained from a
slaughterhouse. The vitreous humor was extracted and three glass cuvettes with path
lengths of 5, 10, and 20 mm were used for the measurements. A cow comea-lens
complex was extracted and squeezed to approximately 5 mm of thickness with glass
slides for the measurements.
4.3. Results
4.3.1. Chromatic dispersion of deionized water
Measurements of k"(l) are presented in Figure 4.3b together with a comparison to
earlier measurements and results of calculations based on the knowledge of the index of
refraction as a function of frequency. The black dots represent our measurements, the
Open dots represent measurements using white-light interferometry [77], the solid line
52
corresponds to the calculated k"(/i.) values based on a Sellmeier’s approximation for the
refractive index of distilled water [74], and the dashed line corresponds to the calculated
k"(/t) values based on a polynomial formula for the refractive index of water adopted by
the National Institute of Standards and Technology (NIST) of the USA [78].
Figure 4.3a shows the deviation of the experimental measurements from the
calculated dispersion based on the Sellmeier formula (line). Below we provide the
Sellmeier formula (Equation 4.2) and the parameters provided by Daimon and Masumura
[74] used for our calculations of k” using Equation 4.1.
I\J
.-
“
‘
Ak" (stImm)
403 (b) - MIIPS
----- NIST formula
5} —- Sellmeier formula
o---..,....,..-.,....,--..,....
650 700 750 800 850 900 950
Wavelength (nm)
Figure 4.3. Comparison of k” for water measured by MIIPS and white-light
interferometry and calculated using the Sellmeier and NIST dispersion formulas. The
Upper graph shows the difference of the corresponding values with respect to those
calculated using the Sellmeier dispersion formula. For both calculations a temperature of
21.5°C was used.
53
n2 —1 = 23—4—— (4.2)
where A.- and ,1,- can be associated with effective resonances and are collected in
Table 4.1.
Table 4.1. Parameters of the Sellmeier formula for water at 21 .5°C
i A: ,1.- (11m)
1 0.5689093832 71 .486374884
2 0.1719708856 135.099228532
3
4
0.02062501582 161 .992558595
0.] 123965424 3267.269774598
Experimental values of k” for water at selected wavelengths are shown in table 4.4.
Based on the measured second-order dispersion we can calculate the third-order
dispersion as k’"(a))=dk"/da). The result of this calculation is presented in Figure 4.4
705
60:: ' MIIPS X,
50:2 O White-light interferometry %I,
40::
30-?
k '" (fs3/mm)
20 - ' ----- NIST formula
10%
03.......
650 700 750 800 850 900 950
Wavelength (nm)
— Sellmeier formula
Ilffi' rITTTY
Figure 4.4. Comparison of water third-order dispersion calculated from our
measurements, the Sellmeier and NIST dispersion formulas, and obtained using white-
light interferometry.
54
together with the corresponding measurements by white-light interferometry [77] and the
calculation based on the Sellmeier [74] and NIST dispersion formulas [78].
4.3.2. Chromatic dispersion of seawater
A difference between the k"s of water and seawater was measured, and the results
are shown in Figure 4.5. Furthermore, we found that the increase in k” when salt is added,
Ak”(S) =k”(S)-k”(S=0), is directly proportional to the salinity (S) and independent from
wavelength. This direct proportionality can also be derived from the formula for the
refractive index of seawater proposed by Quan and Fry [79]. They fit the experimental
refractive index data for seawater measured by Austin and Halikas [80] to a ten-term
empirical formula in the wavelength range 400-700 nm and salinity range 0-35 g/L. The
40; =c‘:~. _'- water
35j \-‘\~~..‘;';o-.. .. seawater
\: o seawater 3x
J
T I r l I I I I V V I I T TV I V
650 760' 7350 800 850 . '960' ' ' 950
Wavelength (nm)
Figure 4.5. Experimental measurements of k” of water, seawater, and water with 3 times
the concentration of salt in seawater (3x). Only a few experimental points were plotted
for clarity.
55
A 5. r3 700nm ,.,
E 4 o 750 nm ,'—
g 4- <> 800nm I
‘3, v 850nm
‘5 31 A 900 nm
9": 2i -« Calculated/I ,
it : , tr
[I 1‘ /
2:: . /“
0—§// I I I ' I ' I I I
O 20 40 60 80 100
Concentration (glL)
Figure 4.6. Increase in k" of seawater with respect to deionized water as a function of the
concentration of sea salt. The symbols correspond to MIIPS measurements and the line to
a calculation based on the refractive index formula for seawater proposed by Quan and
Fry. For the calculation the temperature was set at 21.5°C. Note that the calculation has
been extended beyond the original range of validity for S (~35 g/L).
resulting formula contains only one term proportional to S and A (n(l)ocS/A). Since NM)
9: 13d: n(/i)/d/12 the formula predicts that Ak”(S) is directly proportional to S and
independent from A. In Figure 4.6, we compare the effect of salinity on k” measured by
MIIPS with that calculated using the refractive index formula for seawater in [79].
There seems to be a need for a formula that correctly predicts the wavelength and
salinity dependence of k". We propose that k”(7s.) can be expressed as follows
k"().) = C0 +Clflt+szt2 +CSS, (4.3)
where the coefficients C0, C1 and C 2 were calculated from experimental data for
deionized water (see Figure 4.5) and the coefficient CS was calculated using a linear
regression for Ak"(S) (see Figure 4.6). The parameters of the fit are given in Table 4.2.
These values and the errors were calculated using the Origin 6.1 software (OriginLab
56
Corporation). Experimental values of k” for seawater at selected wavelengths are shown
in table 4.4.
Table 4.2. Seawater parameters for Equation 4.3 in the range 660-930 nm.
C0 102.42174 1: 0.46809 fszmm'l
Cl -0.09476 1 0.00118 fszmm'lnm'l
C2 -2.88686x10‘° i 0.74132x10'6 fszmm'lnm’z
Cs 0.04008 i 0.00157 fszmm" g" L
4.3.3. Chromatic dispersion of the vitreous humor and the comea-lens complex
Ocular dispersion data are particularly scarce for near-infrared wavelengths [81, 82].
To help remedy this lack of information we measured the dispersive properties of all the
transparent components of the eye, the comea-lens complex and the vitreous humor
(Figure 4.7).
We found that the second-order dispersion k" of the vitreous humor is very close to
that of water, as it is shown in Figure 4.8. The data were fit to a second-order polynomial
(Equation 4.4). The parameters of the fit are shown in Table 4.3. Experimental
Retina
Vitreous humor
Cornea
Figure 4.7. Diagram of an eye. The dashed square corresponds to the comea—lens
complex.
57
measurements of k” for the vitreous humor at selected wavelengths are shown in Table
4.4.
HM
1-0-1
DOC
to!
'0!
L...
0—0
H
1-—-.—4
40«E . MIIPS
Sellmeir formula
650 ' ' 700T 750 ' '80'0' ' '85'0' ' ' 900 ' ' 950
Wavelength (nm)
Figure 4.8. Comparison of k" for the vitreous humor measured by MIIPS and k" for water
calculated using the Sellmeier dispersion formula for distilled water. The upper graph
shows the deviation of the vitreous humor measurements with respect to the calculated
values for water.
k"().) = C0 + C11 + C222, (4.4)
Table 4.3. Vitreous humor parameters for Equation 4.4 in the range 660-930 nm.
C0 76.04648 i 1.52198 fszmm'l
Cl 002954 :1: 0.00382 fszmm'lnm"
c2 -4.27553x10'5 i 0.239476x10'5 fszmm'lnm'z
58
The measured chromatic dispersion 01" of a comea-lens complex is shown in Figure
4.9. For this experiment we were not able to use different path lengths. The data
presented come from a set of three measurements on the same tissue. Values of 01" for the
comea-lens complex at selected wavelengths are shown in table 4.4. From our data we
obtain a value of 33 fsz/mm for k”(/1=800 nm)
250
200 ~
1504
t
100—
50-
(1)" (ts?)
o a---a.
650 700 750 800 850 900 950
Wavelength (nm)
Figure 4.9. Measured chromatic dispersion (0" of a cornea-lens complex. The dots
correspond to the experimental points. The line corresponds to a third-order polynomial
fit of the data.
Table 4.4. Experimental dispersion measurements for water, seawater and eye
components.
Wavelength Water Seawater Vitreous humor Cornea-Lens
(nm) k" (fsz/mm) k" (fsz/mm) k” (fsz/mm) g)" (st)
660 38.62 :t 0.33 39.85:l:0.31 - -
675 37.14 :1: 0.26 38.42 i 0.26 36.63 i 0.42 -
700 34.67 :1: 0.15 36.03 i 0.19 34.42 3: 0.31 203 i 7
725 32.20 :1: 0.06 33.61 :t 0.14 32.16 :t 0.22 183 :t 2
750 29.72 i 0.02 31.18 i 0.09 29.84 :t 0.16 172 :t 1
775 27.25 :1: 0.08 28.72 :1: 0.06 27.47 :t 0.12 167 :1: 2
800 24.76 i 0.13 26.25 i 0.04 25.05 3: 0.11 164 :t 1
825 22.28 :1: 0.16 23.75 i 0.03 22.58 i 0.13 161 :1: 1
850 19.79 i 0.18 21.23 :1: 0.03 20.05 :I: 0.18 153 :1: l
875 17.29 i 0.20 18.69 :t 0.05 17.47 :I: 0.25 137 i 1
900 14.80 i 0.21 16.14 i 0.09 14.83 i 0.35 111 i 4
930 11.79zt0.24 13.04zt0.14 116010.51 -
59
4.4. Discussion
4.4.1. Chromatic dispersion of deionized water and seawater
Analysis of our experimental results for water clearly shows excellent agreement
with the calculations based on the Sellmeier model (Equation 4.2). The deviation between
our measurements and the calculated values based on this model are smaller than
0.2 fsz/mm within the measured wavelength range, and for some of the points, the
deviation is even lower than 0.1 fsz/mm. Note that using the refractive index formula
adopted by NIST to calculate the second-order dispersion leads to values which
considerably deviate from experimental measurements and calculations based on
Equation 5. We conclude that MIIPS provides precision and accuracy at least two times
better than white-light interferometry.
There is also a surprisingly good agreement between our experimental
measurements for seawater and the calculations based on the formula for the refractive
index of seawater in [79], considering that this was derived for a different range of
wavelengths and for a much smaller range of salinity.
4.4.2. Chromatic dispersion of the vitreous humor and the cornea-lens complex
Having measured the chromatic dispersion of the ocular components, we can now
predict pulse broadening. If an initial TL Gaussian pulse of time duration to and centered
at 10 acquires a chromatic dispersion (p ”=01 ”(10) after propagating through a dispersive
medium, then its final pulse duration is
2 0.5
n —2
r = To [1+(4go r0 ln2) ] , (4.4)
60
[73]. For a human eyeball we estimate that (o ”= kh”z +01 ”c, where ”h is the second-order
dispersion of the vitreous humor, 2 = 20 mm is the path length that light travels inside the
vitreous humor, and We is the chromatic dispersion of the cornea-lens complex. For
initial lO-fs and 50-fs TL pulses centered at 800 nm, using Table 4.4 and Equation 4.4,
we find that (0”21665 fs2 and that the pulse durations at the retina would be 185 fs and 62
fs respectively. From this, we can also learn that sub-SO-fs pulses with 01": -665 fs2 pose
the greatest eye laser safety risk.
Pulse broadening is important because peak power densities at the target are
inversely proportional to the time duration of the pulses. For example, if one focuses a
10-fs 1-uJ pulse to 10 pm”, the peak power density would be 1015 W/cmz. This extreme
power density causes the desired localized ablation without collateral damage. If the time
duration increases ten times, the peak intensity is reduced in an order of magnitude, and
the ablation ability of the pulses decreases.
We want to point out that we performed calculations of k" for human vitreous
humor based on several proposed refractive index formulas [82-84]. These calculations
lead to values of k”(/1=800nm) that deviate from 7 fsz/mm to 20 fsz/mm from our
experimental measurements at 800 nm. The difference between measured and calculated
values comes from the fact that the k" calculation is highly dependent on the formula
used to fit the refractive index data (see formula 1).
61
4.5. Conclusions
We introduced a new direct method to measure the chromatic dispersion of
transparent media. Comparison of our data with other measurements suggests that the
accuracy and precision of our method are the most reliable to date. Knowledge of the
reported chromatic dispersion measurements of water and seawater is necessary for laser
propagation models in these media. The chromatic dispersion for ocular components will
be useful for laser eye surgery and laser safety data guidelines. The measurement and
elimination of chromatic dispersion (second and higher orders) as shown here, using
MIIPS, is highly recommended for the highest reliability and reproducibility of
applications using femtosecond pulses.
62
Chapter 5
High-resolution two-photon fluorescence excitation spectroscopy by
pulse shaping ultrabroad-bandwidth femtosecond laser pulses
A fast and automated approach to measure two-photon fluorescence excitation (TPE)
spectra of fluorophores with high resolution (~5 nm) by pulse shaping ultrabroad-
bandwidth femtosecond laser pulses is demonstrated. Selective excitation in the range
680-990 nm was achieved imposing a series of specially designed phase-amplitude masks
on the pulses using a Fourier-transforrn pulse shaper. The method eliminates the need of
laser wavelength tuning and is thus suitable for non-laser-expert use. The TPE spectra of
the pH-sensitive dye 8-hydroxypyrene-1,3,6-trisulfonic acid (HPTS) in acidic and basic
environments were measured for the first time using this approach.
5.1 . Introduction
Simultaneous two-photon absorption (TPA) by an atom or molecule was first
predicted by Maria Goeppert-Mayer in 1931 [85], although the first experimental
demonstration of this phenomenon had to wait 30 years until the development of lasers
[3]. Two-photon (TP) spectroscopy has been of interest to study the electronic structure
of molecular excited states [86] because TPA selection rules are different than those for
the one-photon case. TP induced fluorescence resulted very valuable in the biological
imaging field. In 1990, Denk and Webb developed two-photon laser scanning
fluorescence microscopy (TPLSM) [5], a technique that takes advantage of the high
photon density required to overcome the low probability for the simultaneous absorption
of two photons. In TPLSM, TPA and fluorescence emission occur only at the beam
fOcus. Consequently, intrinsic three-dimensional resolution is obtained, out-of-focus
63
background fluorescence is eliminated, and photobleaching is reduced, among other
advantages over wide-field and one-photon confocal laser scanning fluorescence
microscopy [6, 87]. TPLSM has become a widely used tool for medical and biological
research and as a result commercial TP microscopes have been available for a number of
years. The TP spectra of fluorophores are required to determine which are suitable for
this technique and for quantitative TPLSM studies. More recently, TPA is also becoming
relevant for photodynamic therapy [88-90]. The development of new compounds with
tailored TPA properties and well-characterized TP spectra is of outrnost importance for
the development of these applications. Consequently, the measurement of TP cross
sections has become critically important. For instance, a high TP cross section indicates
that the compound absorbs a relatively high fraction of the light focused on the sample,
minimizing possible photodarnage to the surroundings.
The techniques commonly used to measure TP cross-sections of materials can be
divided in two groups. One group is based on nonlinear transmission measurements [91-
93]. These techniques directly yield the TPA cross-section arm and can be applied to
non-fluorescent materials, but their implementation is often difficult given that only a
very small fraction of photons fi'om the excitation beam is absorbed as it passes through
the sample. The second group relies on TP induced fluorescence measurements and
provides better sensitivity [9, 94-98]. In most cases, these techniques yield the TPE cross-
section cm, which is directly proportional to the TPA cross-section oTpA with the
constant of proportionality being the fluorescence quantum efficiency 11 (i.e.,
Grpe=noTpA). Although Fourier-transforrn methods have been reported [99, 100], the
measurement of TPE spectra has been typically performed by selectively exciting the
64
sample with a narrow-bandwidth laser source, recording the resulting TP induced
fluorescence, tuning the laser wavelength and repeating the process for all the desired
wavelengths. Using this approach, a valuable database of TPE spectra of several
commercial organic dyes widely used in TPLSM was reported in the late 1990’s using a
tunable femtosecond laser for selective excitation from 690 to ~1000 nm [8, 94]. More
recently, an extended collection of spectra have been measured in a broader spectral
range using an optical parametric amplifier (OPA) [9]. Determination of TPE spectra can
be significantly simplified using a standard calibration sample with well-known TPE
cross-sections. In this case, several measurements of experimental parameters required
for absolute cross-section measurements are avoided [8, 101, 102].
Here, a fast and automated approach able to measure TPE spectra of fluorophores
with high resolution is demonstrated. The method employs shaped ultrabroad-bandwidth
femtosecond laser pulses to achieve selectively excitation of the sample. The approach
eliminates the need of laser wavelength tuning and is thus suitable for non-laser-expert
use, especially now that F ourier-transform pulse shapers are commercially available.
5.2. Experimental part
5.2.1 . Optical setup
For these experiments we used the ultrabroad-bandwidth femtosecond laser
oscillator (section 2.1.1) together with pulse shaper I (section 2.2). This laser system
covers the most relevant wavelength region for TPLSM (620-1050 nm).
The second-order nonlinear spectrum Swan) of the shaped pulses was characterized
by measuring the corresponding second-harmonic generation (SHG) spectrum [16]. For
65
this purpose, the SHG generation, separation and detection device described in section
3.1 was used.
To measure the TP induced fluorescence, the beam from the pulse shaper was
focused on the sample solution, which was placed in a 2-mm-path-length quartz cell,
using a 50-mm-focal-length spherical mirror. The fluorescence was collected at 90° by a
40x, 0.6 NA microscope objective and was then focused on a silicon-avalanche-
photodiode-based single-photon-counting module (SPCM-AQR-IZ, PerkinElmer)
connected to a gated photon counter (Model SR400, Stanford Research Systems). A
bandpass filter allowing the transmission of light in the fluorescence wavelength region
(400-600 nm) was also used to filter light from other sources, including scattering of the
fundamental pulses (Figure 5.1).
From pulse shaper
IAPD
SM
Figure 5.1. Two-photon induced fluorescence generation, collection and detection setup.
The beam from the pulse shaper is focused onto the fluorescent solution (S), placed in a
2-mm-path-length quartz cell, using a spherical mirror (SM). The fluorescence signal was
collected with a 40x objective (0) at 90° and focused onto the avalanche photodiode
(APD) detection unit with a lens (L). A bandpass filter (F) allowing the transmission of
fluorescence light was placed before the detection system.
66
For all the experiments, spectral phase distortions of the optical system were
measured and corrected at the sample position using MIIPS before the designed phase-
amplitude masks were imposed on the pulses using the shaper. Uncorrected phase
distortions, such as those introduced by the 40x microscope objective, would reduce the
generated TP induced fluorescence due to the effect of phase modulation on Swan) (see
Section 1.2). Furthermore, variations in the phase distortions-, for instance, due to
different oscillator settings or shaper alignment, require frequent phase correction
procedures to ensure reproducibility in the measurements.
5.2.2. Sample preparation
HPTS 100 uM solutions at pH 6, 7, 8, 9 and 10, and a Fluorescein 100 uM solution
at pH 13 were prepared by diluting ~10 mM stock solutions of the dye (Fluorescein and
HPTS sodium salts, Fluka) in a buffer at the corresponding pH. Sodium tetraborate (EM
Science) was used to prepare 50 mM buffers at pH 9 and 10(i0.1), and potassium
phosphate monobasic (Mallinckrodt) was used to prepare 50 mM buffers at pH 6, 7, 8,
and 13(i0.1). The pH of the solutions was measured with a pH-meter (Accumet Basic,
Fisher Scientific) and adjusted using hydrocloric acid or sodium hydroxide solutions.
Deionized water was used in all cases.
5.3. Results and discussion
Selective TP excitation in the range 680-990 nm was achieved with narrow-
bandwidth laser pulses obtained by amplitude shaping. For this purpose, the pulse shaper
was used to block all the wavelengths in the fundamental laser spectrum except those in a
narrow spectral window around the desired excitation wavelength. Such a spectral
67
SHG intensity (arb. u.) Intensity (arb. u.)
Amplitude mask number
1.0
0.8i
0.6;
6.45
0.2:
0.0
q
I
vvvvv
------1-d-—-
'900; "
. (c)
Y r U V V
350
I I Y
400
I V 7"
450
fir
500
Wavelength (nm)
Figure 5.2. Selective two-photon excitation by amplitude shaping of ultrabroad-
bandwidth femtosecond laser pulses. In this experiment the goal was to generate narrow-
bandwidth second-order nonlinear spectra suitable for selective TP excitation.
Experimental SHG spectra were used to characterize the second-order nonlinear spectra.
Sixty amplitude masks evenly spaced across the fundamental spectrum of the pulses were
used to generate the same number of narrow-bandwidth fundamental spectra. (a) Four of
such spectra. (b) Eight of the narrow-bandwidth SHG spectra. The dashed lines indicate
the correspondence between four of the fundamental and SHG spectra. (c) 2D contour
plot showing the SHG spectra corresponding to the whole set of amplitude masks.
68
window is called an amplitude mask. Figure 5.2 illustrates the amplitude shaping
approach. In this experiment, sixty amplitude masks evenly spaced across the available
fundamental spectrum bandwidth were applied to the pulses. Figure 5.2a shows four of
the resulting arnplitude-shaped fundamental spectra, while Figure 5.2b shows eight
experimental SHG spectra, which characterize the second-order nonlinear spectra of the
pulses. Dashed lines indicate the correspondence between four of the fundamental and
SHG spectra. The SHG spectra corresponding to the whole set of amplitude shaped
pulses is shown in Figure 5.2c. A very high contrast ratio (CR) of ~13 (the ratio between
the integral of the signal peak and the integral over the entire background) was achieved
with this approach.
For the TP induced fluorescence experiments, narrower amplitude masks were used
to obtain ~5 nm F WHM peaks in 5(2)(w) that ensure high spectral resolution. For each
amplitude mask centered at wavelength A, the resulting TP induced fluorescence intensity
F(A) was recorded. Figure 5.2 clearly shows that the intensity of Swan) varies
significantly due to the different fundamental spectral intensities and shapes. For this
reason, a normalization procedure was used to obtain the relative cross-section oTpEOt)
according to oTpE(A)=F(A)/.S(2)(A), where 8‘30») is the integrated intensity of the
corresponding SHG spectrum. Absolute cross-sections, which are typically expressed in
Goeppert-Mayer units (1 GM=10—50 cm4 s/photon), can be obtained by comparison with
a calibration standard [8].
First, fluorescent dyes with reported TPE spectra were studied. Figure 5.3 shows a
comparison of TPE spectra of two commonly used fluorescent dyes measured using this
69
method and independently measured with a tunable femtosecond laser [8, 103]. The good
agreement between the measurements reveals the accuracy of this approach.
A 50 ' l I v v r l u v t u
E (a) Fluorescein
SD, 40.. —o—This method -
E l‘ fl —o—Reported
t3 30-\ , f
3 --. .7, i:
Q .
a 20‘ go I
e o’
g; to- j
0.
700 800 900 1000
3300- ' I (b) RhodamineB 4
9250; o This method _
g . —o—Reported
w .. O
2100- ), ,o'f -
. (\VO/ 0
< 50- 0'0, \O\O. o -
'E 6‘ . . Vt"
700 800 900 1000
Wavelength (nm)
Figure 5.3. TPE spectra of Fluorescein and Rhodamine B. TPE spectra of (a) Fluorescein
at pH 13 and (b) Rhodamine B in methanol measured with the method described here
(black circles) and measured with a tunable femtosecond laser system as reported in [8]
and [103], respectively (white circles). The agreement demonstrates the accuracy of this
approach.
70
The fluorescent dye 8-hydroxypyrene-1,3,6-trisu1fonic acid, commonly referred to
as HPTS or pyranine (Figure 5.4a), exhibits (one-photon) absorption spectra highly
dependent on pH (Figure 5.4b). Interestingly, the fluorescence spectra maximum occurs
at 515 nm regardless of the pH because the pKa of the excited state decreases
dramatically upon photocxcitation resulting in fast deprotonation. Thus, emission occurs
only from the ionized form of the molecule. HPTS is stable, commercially available,
highly soluble in various solvents and its pKaz7.7 is conveniently near the pH of neutral
aqueous solutions. These properties, in addition to its large Stokes’ shift and high
fluorescence quantum yield, make HPTS a useful pH-sensitive probe molecule [104,
105]. Nevertheless, HPTS pH dependent TPE spectra have not been reported yet.
In Figure 5.4c, the TPE spectra of HPTS at pH’s 6, 7, 8 and 10 are shown. The data
show very good agreement with a preliminary independent study [31] that measured the
TPE spectra of HPTS at pH’s 6 and 10, but in a much narrower spectral range (770-840
nm). Some correlations can be made between the peaks in the one-photon absorption
spectra and those in the TPE spectra. The transition at 375 nm, characteristic the acidic
form of HPTS, clearly appears at 750 nm in the TPE spectra. The transition at 405 nm,
also characteristic the acidic form, can be observed at 810 nm in the TPE spectrum at pH
6. Finally, the transition at 455 nm, characteristic of the basic form of HPTS, clearly
appears at 910 nm in the TPE spectra.
71
"O33
‘ ' lpH 6 '—'0— [3H 10 ./I—o\' I
8 0.15" ' ‘ “ pH 7 [ ./'/ _ _ \. (b);
c . +PH 8 ./ . - ' .
“5 ' ’ ' “s /
e 0.10- a A “ ‘ A ‘ / .\ _.
O
(D .. _ .
.Q /' \
< 0.05- _ -,..... . H ..
0.00- 1 , . . ‘ .___.__.__. _
375 400 425 450 475
1.00“ ' ' ' l ‘ ,i , ' ..
——¢— pH 6 ~o\r\ c
1 :5, * L pH 7 «(J "’ ( )
_ ’1', q“ ——0— pH 8 [\o‘. 0‘. _
0 75 via}. “‘1’” 10:" m \-.
- figm‘.’ J , 35:35 0‘
a I ‘ k 0 PDHDCF \.
050* fifi‘o‘ 1" flab mo
c] ..
1 ' '0, ”do as}? i
0.25 - - -
* 4.
fl 5"
'1 W» A
O M‘ “V” ‘ :‘r—“w‘uw
TPE cross section (arb. u.)
7'50' 800' 3'56 9'00' 9'50
Wavelength (nm)
Figure 5.4. TPE spectra of HPTS in acidic and basic aqueous environments. (3) Acid-
base equilibrium reaction of HPTS. (b) UV-visible absorption spectra of HPTS at pH’s 6,
7, 8 and 10. The absorption maximum of HPTS changes from ~400 to ~450 nm upon
deprotonation. (c) TPE spectra of HPTS at pH’s 6, 7, 8 and 10. The one-photon
transitions at ~375 and ~450 nm are clearly observed in the TPE spectra at 750 and 900
nm, respectively.
72
The spectrum acquisition time of this method is significantly shorter than that
required by conventional approaches based on laser wavelength tuning. At the
wavelength resolution used for these measurements (~5 nm) the acquisition of a TPE
spectrum takes ~2 min, including the time required for SHG characterization.
The wavelength resolution of the presented measurements (~5 nm) is already higher
than that of most reported TPE spectra measurements (10-20 nm). However, if a higher
resolution were required, narrower peaks in S‘z)(co) would have to be generated. These
can be obtained using narrower amplitude masks. However, the amplitude shaping
approach will become experimentally impractical when peaks in S(2)(m) with insufficient
intensity are obtained. This will necessarily happen for all amplitude masks narrower
than some critical spectral width given that the fundamental spectral intensity outside the
amplitude mask is blocked and thus do not contribute to generation of S(Z)(co). A more
efficient approach, although also more involved, is to apply specially designed spectral
phase functions able to generate constructive interference at the desired excitation
frequency and simultaneously generate destructive interference everywhere else. Such
Phase fimctions can be designed guided by MII principles. For instance, sinusoidal
spectral phases generate a peak in Sm(w) at the frequency corresponding to the inversion
Center of the phase and low spectral intensity elsewhere (Section 1.2). llnfortunately,
Sinusoidal phase modulation cannot provide CRs greater than 0.5, and as the phase is
tIlned away from the central frequency the CR drops below 0.1 [106]. This CR may be
inSufficient for high-resolution TPE spectroscopy. A powerful phase shaping approach
Suitable for high-resolution nonlinear spectroscopy is briefly described next.
73
Narrow peaks in 5mm) can be generated using binary phase (BP) shaping, in which
the spectral phase is limited to 0 and 1: values [31, 106, 107]. For an arbitrary phase, a
peak in the SHG spectrum reaching maximum intensity will occur at a frequency 2wc
such that ga(a)c—w)=-(p(a)c+w), as was demonstrated in Section 1.2. For BPs, the
conditions (p(wc—w)=—(p(wc+co) (antisymmetric phase) or (0(wc-co)=(p(wc+w) (symmetric
phase) are enough to maximize constructive interference and the SHG spectrum at ch
according to Equation 1.6. While any symmetric or antisymmetric BP will maximize the
SHG spectrum at 2coc, the BPs required for selective TP excitation have to
simultaneously maximize destructive interference in order to minimize the SHG spectrum
at all frequencies different than ch. Destructive interference can be maximized using
BPs with minimum autocorrelation [31]. To understand this point, assume that the
magnitude of the field is unity for all fi'equencies and that the BP contains N bits. Then,
the electric field E]- (i=1,2,...N) can only take —1 or +1 values and Equation 1.6 can then
be rewritten as
2
5(2)): °C ZEk—iEk+i (5'1)
i=0
for k=1,2,...N, k-zZl and k+i_<_N. Minimum autocorrelation BPs minimize the sum simply
by introducing a similar number of +1 and —1 values into it. As a consequence, 5(2)(w) is
minimized at all frequencies. While the design of minimum-autocorrelation binary
sequences requires significant effort, they are freely available thanks to research in the
fields of communications engineering and statistical mechanics [108].
74
Figure 5.5 illustrates this BP approach applied to ultrabroad-bandwidth femtosecond
laser pulses for the first time. First, a binary sequence with minimum autocorrelation is
selected. In this case, the 13-bit binary sequence 1:01:01m00mm1m was used [108]. After
symmetrization, the 26-bit binary sequence 1:01c0mt00mm1mmumrc001m01r0n is obtained.
By imposing this 26-bit binary sequence over certain wavelength range in the
fimdamental spectrum and blocking all other wavelengths outside such a range via
amplitude shaping, maximum intensity at the frequency corresponding to the center of
symmetry of the phase and minimum intensity everywhere else in 50%)) are obtained,
exactly as required for selective TP excitation. Shifting the center of symmetry of the
phase across the spectrum, the peak in Swan) can be tuned. Following this approach,
sixty BP-amplitude masks evenly spaced across the available fundamental spectrum
bandwidth were imposed to the pulses to generate the same number of SHG spectra.
Figure 5.5a shows two of the applied phases, while Figure 5.5b shows five SHG spectra.
Dashed lines indicate the correspondence between two of the BPs and SHG spectra. Note
that the peaks occur at the wavelength corresponding to the symmetry point of the BP.
The SHG spectra corresponding to the whole set of BPs are shown in Figure 5.5c. The
CR is ~l.5, but can be further improved following the procedure described in [31]. The
FWHM of the peaks is ~1.8 nm. Therefore, the wavelength resolution in TPE spectra
measurements is expected to increase using the BP shaping approach. By using a BP with
more bits, the resolution can be reduced to ~l nm, which corresponds to the optical
resolution of a single SLM pixel [31]. Without MIIPS spectral phase correction, accurate
delivery of the BPs and thus high-resolution second-order nonlinear excitation would be
impossible.
75
A n | '
E (a) : W
7.: a :
(D l I
N l I
.c I '
n. ' '
o . ..... : -. z
,5 70d 800 [900 1000
:3. (b) I 1 i I
910 :
e ‘ : ‘
a I
'Z’
.9 5‘ ‘
.s
3:” o
<0 350 400 450 500
60—.(0) ' ' ' “
E 50- .-" -
40- _
g . .. .
C 30- +<—FWHM=1.3nm-
x .l ' ' .
E10- _
o fifi.,....,T--..--f-
350 400 450 500
Wavelength (nm)
Figure 5.5. Selective two-photon excitation using binary phase shaping on ultrabroad-
bandwidth femtosecond laser pulses. In this experiment the goal was to generate narrow-
bandwidth second-order nonlinear spectra suitable for selective TP excitation. Sixty BP-
amplitude masks evenly spaced across the fundamental spectrum of the pulses were used
for such a purpose. (a) Two of the BPs. (b) Five of the narrow-bandwidth SHG spectra.
The dashed lines indicate the correspondence between BPs and SHG spectra. (c) 2D
contour plot showing the SHG spectra corresponding to the whole set of BP-amplitude
masks
76
5.4. Conclusions
Selective TP excitation by F ourier-transform pulse shaping ultrabroad-bandwidth
femtosecond laser pulses was successfully applied to TPE spectroscopy. This pulse
shaping approach represents a valuable alternative to the conventional laser wavelength
tuning method commonly employed to measure TPE spectra of fluorophores. The
presented approach is fully automated given that no source wavelength tuning and hence
no laser tweaking are required. Consequently, it is significantly faster and suitable for
most potential interested users, who are not laser experts.
TPE spectra with ~5 nm resolution were obtained using amplitude shaping.
Although this resolution is higher than that of most reported TPE spectra and sufficient
for the majority of purposes, a phase-amplitude shaping approach based on minimum-
autocorrelation binary phases expected to provide ~1 nm resolution was presented.
77
Chapter 6
Atmospheric pressure femtosecond laser desorption ionization
imaging mass spectrometry
The content of this chapter, except section 6.3, has been adapted from Y. Coello, A.
Daniel Jones, T. C. Gunaratne and M. Dantus. “Atmospheric pressure femtosecond laser
imaging mass spectrometry”. Anal. Chem. 82, 2753 (2010) [109].
A novel atmospheric pressure imaging mass spectrometry approach that offers
improved lateral resolution (10 um) using near-infrared femtosecond laser pulses for non-
resonant desorption and ionization of sample constituents without the need of a laser-
absorbing matrix is demonstrated. As a proof of concept the method was used to image a
two-chemical pattern in paper. To demonstrate the ability of the approach to analyze
biological tissue, a monolayer of onion epidermis was imaged allowing the chemical
visualization of individual cells using mass spectrometry at ambient conditions for the
first time. As the spatial resolution is currently limited by the limit of detection of the
setup (~500 finol limit of detection for citric acid), improvements in sensitivity will
increase the achievable spatial resolution.
6.1 . Introduction
Imaging mass spectrometry (IMS) [110, 111] has become an important tool in the
life sciences because of its ability to localize specific analytes, from small metabolites to
proteins, in biological samples. There are two different ways to obtain the spatial
information in an IMS experiment. Typically a tightly focused ionization beam is used to
examine a small region of the sample from where a mass spectrum is obtained. This
process is repeated until the whole sample area has been analyzed and a mass spectrum
78
for each position has been stored. Chemical images can then be obtained from the set of
mass spectra and the corresponding spatial coordinates. This approach, called microprobe
mode, requires the sample to be probed point by point and therefore is relatively slow
because it is limited by the rate of data acquisition and/or repetition rate of the laser
beam. In addition, the spatial resolution is limited by the size of the focused ionization
beam. Although much less popular than the microprobe mode a powerful approach that
overcomes the previous limitations has been demonstrated. In the microscope mode [112,
113] the tightly focused ionization beam is replaced by one that illuminates a relatively
large area of the sample (~200 um), and ion detection is spatially resolved. However, the
microscope mode can only be applied in vacuum conditions to preserve the ion
trajectories from multiple locations in the sample to the detector.
Secondary ion mass spectrometry (SIMS) [114, 115] and matrix-assisted laser
desorption/ionization (MALDI) [116, 117] are currently the most popular techniques
used for obtaining chemically resolved images. In SIMS a beam of primary ions is used
to bombard the sample surface and generate secondary ions. SIMS provides the highest
spatial resolution available (typically >50 nm), however it has only proved useful for
identifying elements and low mass molecules (ca. <1000 Da) because the ionization
method leads to fragmentation that is more pronounced for higher mass analytes. SIMS
requires vacuum conditions and is therefore, incompatible with the analysis of live cells
and tissues. To analyze the distribution of macromolecules such as proteins (1000 < m/z <
50 000), ultraviolet (UV) MALDI remains the method of choice. This technique requires
treating the sample with an external matrix that absorbs the radiation, which makes
sample preparation a critical step. Most UV MALDI imaging experiments have been
79
performed under vacuum conditions providing typical spatial resolutions of 25-200 pm as
limited by laser spot size and perturbation of analyte localization during matrix addition.
Several atmospheric pressure (AP) ionization techniques have been developed in the past
years to overcome incompatibility with the analysis of live tissues and other limitations
imposed by a vacuum environment [118]. Some of these AP ionization techniques, laser
ablation inductively coupled plasma (LA-ICP) [119, 120], laser ablation electrospray
ionization (LAESI) [121, 122], infrared (IR) MALDI [123, 124], and desorption
electrospray ionization (DESI) [125, 126] have already been applied to IMS. In contrast
to the rest of these methods, LA-ICP does not provide molecular information and can
only be used for elemental analysis of the sample because ICP is an atomic ion source. IR
MALDI has employed a 2940 nm wavelength laser for both desorption and ionization of
chemicals using the inherent water content present in biological samples as a matrix. LA-
ICP and LAESI use a laser to ablate the sample while a postionization method, ICP and
electrospray ionization (ESI) respectively, is employed to generate the ions. In LAESI the
use of a postionization process following laser desorption (ablation) leads to higher
ionization efficiencies compared to IR MALDI because laser ablation typically produces
a significant proportion of neutral species in addition to ions and clusters. Although ink
mock patterns have been analyzed with 40 um spatial resolution using AP IR MALDI
[123] and DESI [127], imaging biological samples at AP with a spatial resolution better
than ~200 um has not been reported yet.. A higher spatial resolution would be desirable,
as it would allow studying, for instance, the distribution of chemicals in cellular and
subcellular structures.
80
The spatial resolution of a laser ablation IMS experiment depends on the laser
spot size and the step size [111], which is the distance the laser focal spot moves to
analyze an adjacent sample location. Typically, the step size is larger than the laser spot
and thus is the limiting factor in determining the lateral resolution of the imaging
experiment. However, a step size smaller than the laser spot can be used in an approach
known as oversampling [128]. This method requires complete removal of the analyte
within the laser ablation volume by the desorption process or the use of data processing
algorithms. With oversampling the step size becomes the limiting factor in determining
the spatial resolution and is itself referred to as the pixel resolution of the experiment. In
this situation, decreasing the step size leads to a higher spatial resolution but also to a
smaller sampled volume and thus to lower signal. Using oversampling, an AP IR-MALDI
chemical image of a dye mock pattern with 40 um resolution has been demonstrated
[123]. Recently, significant progress towards chemical imaging with cellular resolution
has been reported using LAESI [129]. This experiment demonstrated in situ metabolic
profiling of single large cells (~50 um width) with a 2940 nm laser beam coupled to a
glass fiber. However, no actual chemical image was presented.
Near IR (N IR) femtosecond laser pulses coupled with mass spectrometry have
been used to demonstrate improved gas-phase molecular identification, including isomer
differentiation [130, 131], and laser-controlled molecular fragmentation [28]. More
recently, femtosecond-laser induced ionization/dissociation (fs-LID) of protonated
peptides has been shown to provide greater sequence information than conventional ion
activation methods [132]. Femtosecond laser pulses have also been used for ablation with
ion-trap MS [133], LA-ICP [134], ESI [135] and as a postionization method for
81
molecular imaging after ion beam desorption [136] and laser ablation [137] of molecular
thin films. MALDI experiments using femtosecond laser pulses in different wavelength
regions have showed very similar results to those using nanosecond pulsed lasers [138].
NIR femtosecond laser (800 nm) MALDI mass spectra using standard matrices with
absorption bands in the UV spectrum have been recently demonstrated [139]. However,
due to the very high peak power densities achieved by focused femtosecond laser pulses
(~10l4 W/cmz) direct non-resonant ablation and ionization of the analytes can occur
[140]. Here we use such an approach for IMS at AP conditions. Focused NIR
femtosecond laser pulses are used to ablate and ionize the sample without using a laser-
absorbing matrix, either native or external, making sample preparation and handling
significantly simpler. Due to its AP implementation the method is a promising imaging
technique for in vivo studies. Finally, the spatial resolution provided by our AP
femtosecond laser desorption ionization (fs-LDI) IMS approach is significantly higher
than that of other AP IMS techniques. Here we demonstrate 10 um spatial resolution in a
biological tissue sample (onion epidermis monolayer) allowing the chemical visualization
of individual cells using mass spectrometry under atmospheric pressure conditions for the
first time.
6.2. Experimental section
6.2.1. Mass spectrometer and laser system
For these experiments we used the Ti:Sa regenerative amplifier and pulse shaper 11,
described in sections 2.1.2 and 2.2, respectively. The output pulses (1 kHz, centered at
800 nm) were focused on the sample using a 5X objective. The pulses were previously
compressed using MIIPS, as described in section 3.2.2, to ensure efficient and
82
reproducible ablation and ionization within focal volume. As a result of MIIPS
compression, 45 fs TL pulses were obtained at the focal plane alter the objective.
For the “S” character, onion tissue and limit of detection experiments 3, 15 and 3 [LI
pulses were used, respectively. The focused laser pulses had a spot diameter of ~20 um,
determined from the optical image of the ablation craters produced by the laser on the
onion epidermis tissue. The ion source of a time-of-flight mass spectrometer (Micromass
LCT, Waters) was replaced with a custom made AP femtosecond laser ion source
containing a motorized XY stage (MAX200, Thorlabs) which holds the sample ~5 mm
from the sample cone of the mass spectrometer. For the “S” character and limit of
detection experiments a copper surface with a -200 V DC offset was used as a sample
holder. For the onion tissue experiments the sample was free standing, only held to the
XY stage from the upper edge. For all the experiments, the sample was positioned
between the sample cone (i65V DC) and a repeller electrode (:1 kV DC) so that the
potential difference helped to direct the ion plume toward the inlet (Figure 6.1), where the
+ and - signs apply for positive and negative ion mode experiments, respectively. The
mass scale was calibrated periodically using MassLynx 4.1 software (Waters Corp.) and
electrospray ionization of a commercial NaCsI solution (Waters Corp.). The accuracy of
measured m/z values was better than 100 mDa over the studied range (m/z<500) over
multiple days between calibrations.
6.2.2. Imaging
The motorized stage was computer-controlled to move the sample surface laterally.
Mass spectra were averaged for ~23 for every spot on the sample and were stored as a
fimction of time. Data acquisition for each imaging experiment took ~80 minutes. A
83
computer program then converted spectrum acquisition time to the corresponding spatial
coordinates. Finally, chemical images were obtained by plotting two-dimensional
distributions of the different chemical species. To obtain the optical images, the sample
was illuminated with a USB-powered diode and the light collected by the objective was
reflected by an 800 nm-transmitter / 400 mn-reflector and directed to a monochromatic
CCD camera (Apogee Alta, Apogee Instruments, Inc.). A BG40 filter was placed before
the CCD camera to block the remaining scattered laser light (Figure 6.1).
XY stage Electrode ,
5X objective \I—
Reflector \
Filteil Light
CCD camera
MS
Oscillator
Amplifier
Figure 6.1. Atmospheric pressure femtosecond laser desorption ionization imaging mass
spectrometry (AP fs-LDI IMS) setup. Femtosecond laser pulses from a Ti:Sa oscillator
were regeneratively amplified and focused on the sample by a 5X objective. The pulse
shaper was used to compress the pulses at the focus. The sample was placed on a
motorized XY stage close to the sample cone of the mass spectrometer (MS) and a
counter-electrode was used to direct the ions to the sample cone. To obtain the optical
images the light illuminating the sample was collected by the objective and directed to a
CCD camera. A filter before the camera was used to block the scattered laser light.
6.2.3. Materials and sample preparation
All solutions were prepared in deionized water. Citric acid (monohydrate) was
purchased from IT Baker. The iodide/iodine dye solution was prepared by dissolving 10
mg of iodine and 175 mg of potassium iodide (Mallinckrodt) in 1 mL of water. For the
“S” experiment the handwritten marks were produced on bond paper using a 0.25 mm
84
diameter wire. The paper sample was then glued to the copper sample holder and
transferred to the translation stage. Fresh red onions were purchased from a local
supermarket. Onion epidermis tissue sections were obtained with a razor blade,
transferred to the translation stage and analyzed without any pretreatment. In the
partially-stained tissue experiment, iodine/iodide dye was deposited on a region of an
onion tissue section by a 0.25 mm diameter wire before transferring the sample to the
translation stage. For the limit of detection experiment, 1 uL of an aqueous 10 mM
solution of citric acid was deposited on the copper sample holder and the solution was
allowed to dry under ambient conditions. The deposited material covered an area of ~6
mmz.
6.2.4. Metabolite identification
Confirmations of metabolite annotation were performed on extracts
(methanolzwater, 1:1 v/v) of onion tissue (1.0 mL of solvent per 100 mg of tissue, wet
weight) using an LCT Premier (Waters) mass spectrometer that was coupled to a
Shimadzu LC-ZOAD ternary pump. Extracts were analyzed using negative mode ESI
following separation on an Ascentis Express C18 column (2.1 x 50 mm, Supelco) using a
gradient described previously [141]. Accurate mass assignments were aided by use of a
lock mass reference (N-butylbenzenesulfonamide) and comparisons of retention times
and ion masses to authentic standards (Sigma-Aldrich).
6.3. Optimization of ion source parameters and preliminary results
Initial experiments were aimed at detecting molecular or fragment ions from
molecules in the gas phase. The pure solid or liquid was placed below the focal spot of
the objective (Figure 6.1) so that the vapor diffused toward the focal volume. Several
85
compounds were used for this purpose, including 2-pentanone, 4-nitrotoluene and 2,4-
dinitrotoluene. The intensity of the molecular ion peak was monitored to optimize the ion
source geometry and several parameters. For instance, the alignment of the electrode and
location of the focal spot were found to have a significant effect on ion transport from ion
source to mass spectrometer. Optimal signal was observed when the electrode was
aligned with the axis of the sample cone and the focal spot was placed on this axis. The
following ion source parameters were also found to have a significant effect on ion
transport. The set of values that provided optimal signal are indicated next.
Electrode: (:l:1000 V)
Sample cone: (:65 V)
Extraction cone: :5 V
RF lens: 100 V,
where the + and - signs apply for positive and negative ion mode experiments,
respectively. The values in parentheses were measured with a multimeter with respect to
ground, but do not correspond to those indicated by the instrument, which had a different
reference.
The source temperature was set to 80°C to reduce the formation of water clusters.
The microchannel plate (MCP) detector voltage was set to the minimum voltage that
provided optimum sensitivity, typically ~2750 V. In addition, the noise threshold (Stop)
for the MCP detector was set to the maximum possible value that did not filter any
chemical signal, typically ~20 mV.
86
28’
87.04
,3 - (a) 2-pentanone
23 80..
2:1
E
2 60'
.E 1
a)
.3, 40' 173.06
5 .
0)
DC 20-
. 105.04
qu. .LL . -- .~ -
80
100 120 140 ' 100 180
100.
A . (b)2,4-nitrotoluene
2% 80- 138.11
.é‘ .
8 60.
E ‘¢ x30 ’
g 40-91.“ 103.11
% 92'" 107.10
a: 20.] 109.11 121.10
Ol ': .1- .- l' - 'l-1 r. _. lI
90 100 110 120 130 140
m/z
Figure 6.2. Atmospheric pressure femtosecond laser ionization mass spectra of molecules
in the gas phase. (a) Mass spectrum of 2-pentanone in positive ion mode. The annotated
peaks at m/z 87.04, 105.04 and 173.06 correspond to [M+H+]+, [M+H++H20]+ and
[2M+H+]+ of pentanone. (b) Mass spectrum of 4-nitrotoluene in positive ion mode. The
annotated peaks at m/z 91.11, 92.11, 107.10, 108.11, 109.11, 121.10 and 138.11
correspond to [M-NozT‘, [M-N02+H*]+, [M-NOT’, [M-NO+H+]+, [M—NOz'+HzO]+, [M-
OT and [M+H+]+ of nitrotoluene.
The positive ion mode mass spectra of 2-pentanone and 4-nitrotoluene are shown as
an example in Figures 6.2a and 6.2b, respectively. For these experiments, data were
averaged over ~l min. The annotated peaks were identified as molecular or fragment ions
of the corresponding molecules. In Figure 6.2a the peaks at m/z 87.04, 105.04 and 173.06
87
correspond to [M+H+]+, [M+H++H20]+ and [2M+H+]+ of pentanone (theoretical
monoisotopic m/z 87.09, 105.09 and 173.15, respectively). In Figure 6.2b the peaks at m/z
91.11, 92.11, 107.10, 108.11, 109.11, 121.10 and 138.11 correspond to [M-NOzT, [M-
N02+H+]+, [M-NoT, [M-NO+H+]+, [M-NOz'+HzO]+, [M-OT and [M+H+]+ of
nitrotoluene (theoretical monoisotopic m/z 91.05, 92.06, 107.05, 108.06, 109.07, 121.05
and 138.06, respectively).
A number of unexpected peaks fiequently appeared in mass spectra (not shown)
taken in positive ion mode even in the absence of a sample. Many of these peaks,
appearing at m/z 149, 279 and 391, were found to come from the plasticizer bis-(2-
ethylhexyl) phthalate, a common laboratory contaminant.
A sample holder was used for solid samples, as shown in Figure 6.1. No significant
signal was observed unless a potential was applied to the holder. For this reason, a copper
sample holder was used. For the ion source parameters given above, the optimal sample
holder potential was found to be :65 V, where the + and - signs apply for positive and
negative ion mode experiments, respectively. Shorter distances between the tip of the
electrode and the sample cone vertex (or longer electrodes, see Figure 6.1) were also
found to increase the signal intensity. A ~1 cm distance was used to let sufficient space to
move the sample during imaging experiments. Finally, different repulsive electrodes were
evaluated including sharp pins and flat electrodes. While the optimal electrode voltage
changed for the different electrodes, the maximum signal level did not vary significantly.
A first imaging experiment, shown in Figure 6.3, was performed with a set of
tyrosine needles crystallized from a saturated solution of tyrosine in aqueous ammonia.
The femtosecond laser desorption ionization mass spectrum showed the protonated
88
molecular ion [M+H]+ at m/z 182.06 almost exclusively (theoretical monoisotopic m/z
182.08). The optical image of the sample was not taken given that no camera was
available at that time.
Figure 6.3. Chemical image of tyrosine crystals. The plot shows the distribution of m/z
182, the protonated molecular ion [M+H]+. A step size of 50m was used. The area
imaged is about 1.7mmx0.9mm.
6.4. Results and discussion
The motivation for developing improved IMS instrumentation is to obtain chemical-
species resolved images. In order to demonstrate this capability we analyzed an “S”
character drawn with iodide/iodine dye. An optical image of the sample is shown in
Figure 6.4a. Also present, although not visible in the optical image, is a diagonal mark
across the “S” character produced with an aqueous ~5% (w/v) solution of citric acid. The
mass spectrum of the dye showed the presence of peaks at m/z 126.91, 253.77 and 380.64
corresponding to I', 12' and 13', respectively (theoretical monoisotopic m/z 126.90, 253.81
and 380.71, respectively). The spatial distribution of triiodide (m/z 380.64) shows
excellent agreement with the “S” character, as shown in Figure 6.4b. The distribution of
[M-H]' from citric acid (m/z 191.09, theoretical monoisotopic m/z 191.02) shows the
89
optically invisible diagonal mark drawn on the sample. The imaging experiment was
performed using a step size of 25 pm under atmospheric conditions.
(a) Optical (b) m/z 380.7 (c) m/z 191
200 pm
Figure 6.4. Chemical image of a dye pattern obtained under atmospheric conditions. (a)
Optical image of the sample. The “S” character was drawn with iodine/iodide dye.
Although not visible, a diagonal line drawn across the “S” with citric acid is also present
in the sample. (b) The distribution of triiodide (m/z 380.64), which shows an excellent
agreement with the “S” character. (c) The distribution of citrate (m/z 191.09), invisible in
the optical image, shows the diagonal line drawn across the “S” character with citric acid.
The step size is 25 um.
Imaging biological samples is one of the most promising applications of IMS. We
selected onion epidermis tissue, a classic sample in optical microscopy, to demonstrate
the ability of our method to image biological samples with unprecedented spatial
resolution at atmospheric pressure using mass spectrometry. This tissue was selected
because it contains cells of appropriate sizes (~50 um width) to be resolved with the
current spatial resolution of our system. Figure 6.5 shows the laser-induced mass spectra
in negative and positive ion modes obtained afler a 100 um step of the sample across the
laser focal spot. The spectra show the presence of common plant metabolites. The peaks
at m/z 179.05, 225.05, 341.09 and 387.10 in Figure 6.5a correspond to the [M-H]' of
glucose (with possible contribution from other isomeric hexoses), [M+formate]‘ of
90
glucose, [M-H]' of sucrose, and [M+forrnate]' of sucrose. These assignments were
179.05 225.05
A 8‘
e: . (a)
2:
.6 6.1
C
.23
.E 4.
G)
.2
E 24
9’ ' 341.09 387.10
[K . l
04 v““".‘“=‘ L ‘~la
100' 198.08
(b)
2\: 301
z: .
'7) 6 145.04
$3 163.05
.E
0
.2
E
0)
tr
100 200 300 400
m/z
Figure 6.5. Mass spectra of onion epidermis tissue generated using femtosecond laser
desorption ionization. Common plant metabolites were identified. (a) Negative ion mode.
The annotated peaks at m/z 179.05, 225.05, 341.09 and 387.10 correspond to the [M-H]'
of glucose, [M+formate]' of glucose, [M-H]' of sucrose, and [M+formate]' of sucrose,
respectively. (b) Positive ion mode. The annotated peaks at m/z 127.03, 145.04, 163.05,
180.07, 198.08 and 325.07 are consistent with [M+H+-3HzO]+ of glucose, [M+H+-2HzO]+
of glucose, [M+H+-H20]+ of glucose, [M+NH4+-H20]+ of glucose, [M+NH.{']+ of glucose
and [M+H+-HZO]+ of sucrose, respectively.
confirmed by coincident retention times and accurate mass measurements from LC/MS
analyses of onion extracts (see Metabolite identification) which detected glucose as a
91
minor constituent (m/z 179.0553, theoretical monoisotopic m/z 179.0561) and sucrose as
an abundant metabolite (m/z 341.1078, theor. m/z 341.1089). In Figure 6.5b the peaks at
m/z 127.03, 145.04, 163.05, 180.07, 198.08 and 325.07 are consistent with [M+H+-
3H20]+ of glucose, [M+H+-2H20]+ of glucose, [M+HJ'-HZO]+ of glucose, [M+NH4+-
H20]+ of glucose, [M+NH4+]+ of glucose and [M+H+-H20]+ of sucrose (theor. m/z 127.04,
145.05, 163.06, 180.09, 198.10 and 325.11, respectively).
Imaging experiments were performed in the negative ion mode. Given that the
width of the tissue cells in our experiments was ~50 pm, a resolution higher than ~20 um
was necessary to resolve individual cells. At these resolutions, obtained by decreasing the
step size, only the peaks at m/z 179.06 and 225.06 were detected and both showed similar
spatial distributions. To enhance chemical contrast, a portion of the sample was stained
with an iodine/iodide dye which is commonly used to stain starch. Figure6.4a shows the
optical image (false color) of the sample. The stained region of the sample appears at the
lower right region and is slightly darker than the rest of the tissue. The horizontal mark at
the lower left region was intentionally produced by ablating the tissue with the laser.
Figures 6.6b and 6.6c correspond to chemical images obtained using a 15 um step size.
Figure 6.6b shows the spatial distribution of deprotonated glucose ions (m/z 179.06).
Note that higher m/z 179 regions (darker blue) match the location of the cell walls in the
tissue, and thus the glucose ions are probably produced by fragmentation of cellulose
from the cell walls. The ablated region also appears clearly in the chemical image. The
spatial distribution of triiodide (m/z 380.64), from the dye solution, is shown in Figure
6.6c and agrees well with the location of the stained region.
92
(3) Optical (b) m/z 179.05 (c) m/z 380.7
. ’ 150 pm
Figure 6.6. Chemical image of onion epidermis cells obtained under atmospheric
conditions in negative ion mode. (a) Optical image (false color) of the tissue analyzed.
The lower right region was partially stained with an iodine/iodide dye and appears
slightly darker than the rest of the tissue. The horizontal mark was produced by ablating
the tissue with the laser to determine the sampling width of the laser spot. (b) Chemical
image of the same region showing the spatial distribution of deprotonated glucose (m/z
179.06). Note the excellent agreement with the optical image. (c) Chemical image of the
same region showing the spatial distribution of triiodide (m/z 380.64). The step size for
both chemical images was 15 mm.
The single pixel resolution of an experiment (step size) does not necessarily agree
with the experimental spatial resolution of an image (the length scale that can be
distinguished), which depends also on other factors such as the spatial distribution of
analytes in the sample and the signal intensity per pixel [111]. A way to estimate the
experimental spatial resolution is by examining a line across a feature of interest and
measuring the distance required to move from 20% to 80% of the maximum intensity
value of the feature [142]. To estimate the experimental spatial resolution of our system,
we recorded another chemical image of onion epidermis using a 10 um step size. Smaller
step sizes compromised reproducibility of signal intensities across the sample in the
93
present configuration of our setup. Figure 6.7a shows the chemical image of the tissue
showing the spatial distribution of m/z 179 (deprotonated glucose). The inset shows the
corresponding optical image (false color). The experimental spatial resolution was
calculated as ~10 pm from the analysis of several line scans across the image. An
example of such line scans is shown as a red dashed line in Figure 6.7a and its
corresponding intensity profile is shown in Figure 6.7b. To our knowledge, this is the
highest spatial resolution chemical image obtained at AP conditions.
1.0- . - . - 1 .
- (b)
"T 0.8 .
3 O
.o'
5; 0.6- .
3*
'5 0.4- -
C
e /
s 0.2. \ ° \/ / \
—C\. O
0.0 , . -, .\/. . , . . . . .
0 100 200
Distance (pm)
Figure 6.7. Chemical image of onion epidermis cells demonstrating the highest spatial
resolution under atmospheric pressure conditions. (a) Chemical image generated in
negative ion mode showing the spatial distribution of m/z 179 generated by probing the
onion epidermis tissue. The inset shows the corresponding optical image. The scale bar in
the inset is 100 um. (b) Intensity distribution of m/z 179 corresponding to the red dashed
line shown in (a). The analysis of several line scans such as the one shown indicated an
experimental spatial resolution of ~10 pm.
The cell monolayers analyzed previously were completely ablated during the
imaging experiment. However, the damage inflicted by the laser on thicker biological
samples is superficial and most plant and animal tissue samples can survive the analysis.
In vivo chemical imaging experiments are therefore possible with AP fs-LDI IMS.
94
While the identified peaks in the mass spectra of onion skin epidermis (Figure 6.5)
likely correspond to cellulose fragments, the molecular ion is typically present in the AP
fs-LDI mass spectrum of low molecular weight (<400 Da) solid samples of metabolite
standards. The [M-H]' of the analyte is observed for acidic metabolites analyzed in
negative ion mode such as in the case of citric acid. Similarly, the [M+H]+ of the analyte
is observed for molecules analyzed in positive ion mode such as tyrosine and 2,4-
dinitrotoluene (not shown). The ionization of heavier molecules with AP fs-LDI has not
been studied thoroughly, but molecular ions of molecules heavier than 400 Da have not
been observed so far. As it is also suggested by the mass spectra shown in Figure 6.5 the
ion yield of AP fs-LDI seems to decrease with increasing mass probably due to
inefficient ablation for heavier fragments or inefficient transport of ions from the sample
surface into the mass spectrometer using the present source configuration.
The limit of detection (LOD) of the method was calculated analyzing a layer of
citric acid deposited on the sample holder (see Materials and sample preparation). The
mass spectra corresponding to ten different laser spots (20 um diameter) were averaged
yielding a signal-to-noise ratio (S/N) of 5:1 for the citrate ion (m/z 191). Assuming that
all the deposited material was ablated from the illuminated area, each laser spot would
provide ~500 frnol of citric acid molecules. Other AP IMS techniques including DESI,
LAESI and IR MALDI have limits of detection of a few finol [121]. Note that in IR
MALDI a laser-absorbing matrix present in high concentration resonantly absorbs the
laser radiation [123]. In contrast, non-resonant laser-analyte interaction with no matrix
occurs in fs—LDI. This difference may explain the higher LOD observed for fs-LDI. Laser
desorption experiments have shown to produce a significant amount of neutrals together
95
with ions.[l 18, 121] Therefore, the introduction of a secondary ionization method such as
ESI after laser desorption[121, 135] is expected to increase the ionization efficiency of fs-
LDI. Additional improvements in the sensitivity of AP fs-LDI can be expected by
optimizing several of the AP ion source parameters including the potentials on the sample
cone, sample holder and electrode; and the distances between the electrode, sample
holder, laser focal spot and sample cone. No effort was made here to synchronize the ion
packets generated by the laser with the pusher pulses in the mass spectrometer. Such
synchronization together with the use of an analog-to-digital converter, rather than the
time-to-digital electronics in the current detection system offer opportunities to
significantly increase the sensitivity and dynamic range of the method.
6.5. Conclusions
A novel IMS approach using near-IR femtosecond laser pulses for direct sample
desorption and ionization at AP conditions has been presented. Given that ablation and
ionization occur via nonlinear laser-analyte interactions the presence of a laser-absorbing
matrix is not required. Consequently, sample preparation and handling are significantly
simplified compared to AP MALDI IMS techniques.
In its current level of development AP fs-LDI IMS offers a limited m/z range
(m/z<400) and sensitivity compared to other AP IMS techniques. Both figures of merit
can be improved by adding a secondary ionization method following laser desorption to
improve the ionization efficiency, by optimizing several of the AP ion source parameters
to enhance ion collection, and by introducing ion packet-pusher pulse synchronization
with new ADC detectors.
96
In contrast to the established vacuum IMS techniques MALDI and SIMS, AP fs-
LDI IMS allows the analysis of biological samples in their natural state. Improvements in
the sensitivity of the setup, as described before, will minimize damage to the sample and
make in viva investigations more feasible.
While no AP IMS technique can compete with SIMS imaging in terms of spatial
resolution yet, the 10 um spatial resolution for biological tissue demonstrated here with
AP fs-LDI IMS represents an improvement over other AP IMS techniques and a step
towards mass spectrometric chemical imaging at the cellular level. Efforts to increase
resolution will also require improvements in the sensitivity in order to maintain an
acceptable S/N. The resolution of the system could then be improved by reducing the
laser focal spot diameter and the step size. The laser pulses used here can be focused to
~1 um using a higher magnification objective. In theory, the smallest possible focal spot
diameter would be determined by the diffi'action limit ~M2, 400 nm in this case. Because
the ionization and ablation processes produced by focused femtosecond pulses are highly
nonlinear, it is conceivable that sub-diffraction-limit focal spot diameters could be
ablated. This would allow, for instance, imaging subcellular structures.
97
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(3)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(17)
References
Zewail, A. H. J. Phys. Chem. A 2000, 104, 5660-5694.
Franken, P. A.; Weinreich, G.; Peters, C. W.; Hill, A. E. Phys. Rev. Lett. 1961, 7,
118-&.
Kaiser, W.; Garrett, C. G. B. Phys. Rev. Lett. 1961, 7, 229 LP - 231.
Carnpagnola, P. J.; Loew, L. M. Nat. Biotechnol. 2003, 21, 1356-1360. 1
Denk, W.; Strickler, J.; Webb, W. Science 1990, 248, 73-76.
Diaspro, A.; Chirico, G.; Collini, M. Quarterly Reviews of Biophysics 2005, 38, L
97-166. ' "'
Xi, P.; Andegeko, Y.; Weisel, L. R.; Lozovoy, V. V.; Dantus, M. Opt. Commun.
2008, 281, 1841-1849.
Albota, M. A.; Xu, C.; Webb, W. W. Appl. Opt. 1998, 37, 7352-7356.
Makarov, N. S.; Drobizhev, M.; Rebane, A. Opt. Express 2008, 16, 4029-4047.
Wollenhaupt, M.; Assion, A.; Baumert, T. In Springer Handbook of Lasers and
Optics; Trager, F., Ed.; Springer: New York, 2007, pp 937-983.
Walmsley, I.; Waxer, L.; Dorrer, C. Rev. Sci. Instrum. 2001, 72, 1-29.
Broers, B.; Vandenheuvell, H. B. V.; Noordam, L. D. Optics Communications
1992, 91, 57-61.
Broers, B.; Noordam, L. D.; Vandenheuvell, H. B. V. Physical Review A 1992,
46, 2749-2756.
Meshulach, D.; Silberberg, Y. Nature 1998, 396, 239-242.
Meshulach, D.; Silberberg, Y. Physical Review A 1999, 60, 1287-1292.
Walowicz, K. A.; Pastirk, 1.; Lozovoy, V. V.; Dantus, M. J. Phys. Chem. A 2002,
106, 9369-9373.
Lozovoy, V. V.; Pastirk, 1.; Walowicz, K. A.; Dantus, M. Journal of Chemical
Physics 2003, 118, 3187-3196.
98
(18)
(19)
(20)
(21)
(22)
(23)
(24)
(25)
(26)
(27)
(28)
(29)
(30)
(31)
(32)
(33)
(34)
(35)
Dela Cruz, J. M.; Pastirk, I.; Lozovoy, V. V.; Walowicz, K. A.; Dantus, M. J.
Phys. Chem. A 2004, 108, 53-58.
Lozovoy, V. V.; Dantus, M. Chemphyschem 2005, 6, 1970-2000.
Shank, C. V.; Ippen, E. P. Appl. Phys. Lett. 1974, 24, 373-375.
Maiman, T. H. Nature 1960, 187, 493-494.
Fork, R. L.; Cruz, C. H. B.; Becker, P. C.; Shank, C. V. Opt. Lett. 1987, 12, 483-
485.
Moulton, P. F. J. Opt. Soc. Am. B 1986, 3, 125-133.
Spence, D. E.; Kean, P. N.; Sibbett, W. Opt. Lett. 1991, 16, 42-44.
Ell, R.; Morgner, U.; Kartner, F. X.; Fujimoto, J. G.; Ippen, E. P.; Scheuer, V.;
Angelow, G.; Tschudi, T.; Lederer, M. J .; Boiko, A.; Luther-Davies, B. Opt. Lett.
2001, 26, 373-375.
Binhammer, T.; Rittweger, E.; Ell, R.; Kartner, F. X.; Morgner, U. IEEE J.
Quantum Elect. 2005, 41, 1552-1557.
Coello, Y.; Lozovoy, V. V.; Gunaratne, T. C.; Xu, B. W.; Borukhovich, I.; Tseng,
C. H.; Weinacht, T.; Dantus, M. J. Opt. Soc. Am. B 2008, 25, A140-A150.
Lozovoy, V. V.; Zhu, X.; Gunaratne, T. C.; Harris, D. A.; Shane, J. C.; Dantus,
M. J. Phys. Chem. A 2008, 112, 3789-3812.
Strickland, D.; Mourou, G. Optics Communications 1985, 56, 219-221.
Backus, S.; Durfee, C. G.; Mumane, M. M.; Kapteyn, H. C. Rev. Sci. Instrum.
1998, 69, 1207-1223.
Lozovoy, V. V.; Xu, B. W.; Shane, J. C.; Dantus, M. Physical Review A 2006, 74.
Ogilvie, J. P.; Debarre, D.; Solinas, X.; Martin, J. L.; Beaurepaire, E.; Joffre, M.
Opt. Express 2006, 14, 759-766.
Dantus, M.; Lozovoy, V. V. Chemical Reviews 2004, 104, 1813-1859.
Lozovoy, V. V.; Dantus, M. Annu. Rep. Prog. Chem, Sect. C: Phys. Chem. 2006,
102, 227 - 258.
Verluise, F.; Laude, V.; Cheng, Z.; Spielmann, C.; Toumois, P. Opt. Lett. 2000,
25, 575-577.
99
(36)
(37)
(38)
(39)
(40)
(41)
(42)
(43)
(44)
(45)
(46)
(47)
(48)
(49)
(50)
(51)
(52)
(53)
Weiner, A. M. Rev. Sci. Instrum. 2000, 71, 1929-1960.
Zeek, E.; Bartels, R.; Mumane, M. M.; Kapteyn, H. C.; Backus, S.; Vdovin, G.
Opt. Lett. 2000, 25, 587-589.
Zeek, E.; Maginnis, K.; Backus, S.; Russek, U.; Mumane, M.; Mourou, G.;
Kapteyn, H.; Vdovin, G. Opt. Lett. 1999, 24, 493-495.
Hillegas, C. W.; Tull, J. X.; Goswarni, D.; Strickland, D.; Warren, W. S. Opt.
Lett. 1994, 19, 737-739.
Dugan, M. A.; Tull, J. X.; Warren, W. S. J. Opt. Soc. Am. B 1997, 14, 2348-2358.
Weiner, A. M.; Leaird, D. E.; Patel, J. S.; Wullert, J. R. IEEE J. Quantum Elect.
1992, 28, 908-920.
Weiner, A. M.; Leaird, D. E.; Patel, J. S.; Wullert, J. R. Opt. Lett. 1990, I5, 326-
328.
Wefers, M. M.; Nelson, K. A. Opt. Lett. 1995, 20, 1047-1049.
Brixner, T.; Gerber, G. Opt. Lett. 2001, 26, 557-559.
Oron, D.; Dudovich, N.; Silberberg, Y. Phys. Rev. Lett. 2003, 90.
Brixner, T.; Krampert, G.; Pfeifer, T.; Selle, R.; Gerber, G.; Wollenhaupt, M.;
Graefe, O.; Horn, C.; Liese, D.; Baumert, T. Phys. Rev. Lett. 2004, 92.
Yelin, D.; Meshulach, D.; Silberberg, Y. Opt. Lett. 1997, 22, 1793-1795.
Baumert, T.; Brixner, T.; Seyfried, V.; Strehle, M.; Gerber, G. Appl. Phys. B-
Lasers O. 1997, 65, 779-782.
Naganuma, K.; Mogi, K.; Yamada, H. IEEE J. Quantum Elect. 1989, 25, 1225-
1233.
Trebino, R.; Kane, D. J. J. Opt. Soc. Am. A 1993, 10, 1101-1111.
Trebino, R. Frequency-resolved optical gating .' the measurement of ultrashort
laser pulses; Kluwer Academic: Boston, 2000.
Iaconis, C.; Walmsley, I. A. Opt. Lett. 1998, 23, 792-794.
Iaconis, C.; Walmsley, I. A. IEEE .1. Quantum Elect. 1999, 35, 501-509.
100
(54)
(55)
(56)
(57)
(58)
(59)
(60)
(61)
(62)
(63)
(64)
(65)
(66)
(67)
(68)
(69)
(70)
Baum, P.; Lochbrunner, S.; Gallrnann, L.; Steinmeyer, G.; Keller, U.; Riedle, E.
Appl. Phys. B-Lasers O. 2002, 74, 8219-8224.
Yamane, K.; Zhang, Z. G.; Oka, K.; Morita, R.; Yamashita, M.; Suguro, A. Opt.
Lett. 2003, 28, 2258-2260.
Miao, H. X.; Weiner, A. M.; Langrock, C.; Roussev, R. V.; Fejer, M. M. Opt.
Lett. 2007, 32, 424-426.
Dantus, M.; Lozovoy, V. V.; Pastirk, I. OE Magazine 2003, 9, 15-17.
Lozovoy, V. V.; Pastirk, 1.; Dantus, M. Opt. Lett. 2004, 29, 775-777.
Xu, B. W.; Gunn, J. M.; Dela Cruz, J. M.; Lozovoy, V. V.; Dantus, M. J. Opt.
Soc. Am. B 2006, 23, 750-759.
Lozovoy, V. V.; Xu, B.; Coello, Y.; Dantus, M. Opt. Express 2008, 16, 592-597.
Li, H.; Harris, D. A.; Xu, B.; Wrzesinski, P. J .; Lozovoy, V. V.; Dantus, M. Opt.
Express 2008, 16, 5499-5504.
Zhu, X.; Gunaratne, T. C.; Lozovoy, V. V.; Dantus, M. Opt. Express 2007, 15,
16061-16066.
Baltuska, A.; Kobayashi, T. Appl. Phys. B-Lasers O. 2002, 75, 427-443.
Yamashita, M.; Yamane, K.; Morita, R. IEEE J. Sci. T 0p. Quant. 2006, 12, 213-
222.
Xu, B. W.; Coello, Y.; Lozovoy, V. V.; Harris, D. A.; Dantus, M. Opt. Express
2006, 14, 10939-10944.
Coello, Y.; Xu, B. W.; Miller, T. L.; Lozovoy, V. V.; Dantus, M. Appl. Opt. 2007,
46, 8394-8401.
Toyran, S.; Liu, Y. M.; Singha, S.; Shan, S.; Cho, M. R.; Gordon, R. J.; Edward,
D. P. Exp. Eye Res. 2005, 81, 298-305.
Johnson, T. M.; Glaser, B. M. Graefes Arch. Clin. Exp. Ophthalmol. 2005, 243,
570-575.
Wang, B. G.; Riemann, I.; Schubert, H.; Halbhuber, K. J .; Koenig, K. Cell T iss.
Res. 2007, 328, 515-520.
Blum, M.; Kunert, K.; Nolte, S.; Riehemann, S.; Palme, M.; Peschel, T.; Dick,
M.; Dick, H. B. Ophthalmologe 2006, 103, 1014-1019.
101
(71)
(72)
(73)
(74)
(75)
(76)
(77)
(73)
(79)
(30)
(31)
(32)
(83)
(34)
(85)
(86)
(87)
(88)
Gerten, G.; Ripken, T.; Breitenfeld, P.; Krueger, R. R.; Kerrnani, O.;
Lubatschowski, H.; Oberheide, U. Ophthalmologe 2007, 104, 40-46.
Cain, C. P.; Thomas, R. J .; Noojin, G. D.; Stolarski, D. J .; Kennedy, P. K.;
Buffington, G. D.; Rockwell, B. A. Graefes Arch. Clin. Exp. Ophthalmol. 2005,
243,101-112.
Siegrnan, A. E. Lasers; University Science Books: Mill Valley, CA, 1986.
Daimon, M.; Masumura, A. Appl. Opt. 2007, 46, 3811-3820.
Diddams, S.; Diels, J. C. J. Opt. Soc. Am. B 1996, 13, 1120—1129.
Cohen, L. G. J. Lightwave T echnol. 1985, 3, 958-966.
Van Engen, A. G.; Diddams, S. A.; Clement, T. S. Appl. Opt. 1998, 37, 5679-
5686.
Harvey, A. H.; Gallagher, J. S.; Sengers, J. J. Phys. Chem. Ref Data 1998, 27,
761-774.
Quan, X. H.; Fry, E. S. Appl. Opt. 1995, 34, 3477-3480.
Austin, R. W.; Halikas, G.; Scripps Institution of Oceanography: San Diego,
1976.
Hammer, D. X.; Welch, A. J .; Noojin, G. D.; Thomas, R. J .; Stolarski, D. J .;
Rockwell, B. A. J. Opt. Soc. Am. A 1999, 16, 2092-2102.
Atchison, D. A.; Smith, G. J. Opt. Soc. Am. A 2005, 22, 29-37.
Navarro, R.; Santamaria, J.; Bescos, J. J. Opt. Soc. Am. A 1985, 2, 1273-1281.
Le Grand, Y. F arm and space vision, Rev. ed.; Indiana University Press:
Bloomington, 1967.
Goppert-Mayer, M. Ann. Phys. -Berlin 1931, 9, 273-294.
Ashfold, M. N. R.; Howe, J. D. Annual Review of Physical Chemistry 1994, 45,
57-82.
Helmchen, F.; Denk, W. Nat. Methods 2005, 2, 932-940.
Fisher, W. G.; Partridge, W. P.; Dees, C.; Wachter, E. A. Photochem. Photobiol.
1997, 66, 141-155.
102
(89)
(90)
(91)
(92)
(93)
(94)
(95)
(96)
(97)
(98)
(99)
(100)
(101)
(102)
(103)
(104)
(105)
Kim, S.; Ohulchanskyy, T. Y.; Pudavar, H. E.; Pandey, R. K.; Prasad, P. N. J. Am.
Chem. Soc. 2007, 129, 2669-2675.
Kuimova, M. K.; Collins, H. A.; Balaz, M.; Dahlstedt, E.; Levitt, J. A.; Sergent,
N.; Suhling, K.; Drobizhev, M.; Makarov, N. S.; Rebane, A.; Anderson, H. L.;
Phillips, D. Org. Biomol. Chem. 2009, 7, 889-896.
Oulianov, D. A.; Tomov, I. V.; Dvomikov, A. S.; Rentzepis, P. M. Optics
Communications 2001, 191, 235-243.
Sengupta, P.; Balaji, J .; Banerjee, S.; Philip, R.; Krunar, G. R.; Maiti, S. Journal
of Chemical Physics 2000, 112, 9201-9205.
Tian, P. F.; Warren, W. S. Opt. Lett. 2002, 27, 1634-1636.
Xu, C.; Webb, W. W. J. Opt. Soc. Am. B 1996, 13, 481-491.
Xu, C.; Guild, J .; Webb, W. W.; Denk, W. Opt. Lett. 1995, 20, 2372-2374.
Kaatz, P.; Shelton, D. P. J. Opt. Soc. Am. B 1999, 16, 998-1006.
Kapoor, R.; Friend, C. S.; Patra, A. J. Opt. Soc. Am. B 2003, 20, 1550-1554.
Kauert, M.; Stoller, P. C.; Frenz, M.; Ricka, J. Opt. Express 2006, 14, 8434-8447.
Ogilvie, J. P.; Kubarych, K. J .; Alexandrou, A.; Joffre, M. Opt. Lett. 2005, 30,
911-913.
Isobe, K.; Suda, A.; Tanaka, M.; Kannari, F.; Kawano, H.; Mizuno, H.;
Miyawaki, A.; Midorikawa, K. Physical Review A 2008, 77.
Blab, G. A.; Lommerse, P. H. M.; Cognet, L.; Harms, G. S.; Schmidt, T. Chem.
Phys. Lett. 2001, 350, 71-77.
Huang, S. H.; Heikal, A. A.; Webb, W. W. Biophysical Journal 2002, 82, 2811-
2825.
h_ttp://www. drbio. cornell. edu/cross_sections. html.
Dela Cruz, J. M.; Pastirk, 1.; Comstock, M.; Lozovoy, V. V.; Dantus, M. Proc.
Natl. Acad. Sci. USA 2004, 101, 16996-17001.
Spry, D. B.; Goun, A.; Bell, C. 8.; III; Fayer, M. D. J. Chem. Phys. 2006, 125,
144514.
103
(106)
(107)
(108)
(109)
(110)
(111)
(112)
(113)
(114)
(115)
(116)
(117)
(118)
(119)
(120)
(121)
(122)
(123)
Comstock, M.; Lozovoy, V. V.; Pastirk, I.; Dantus, M. Opt. Express 2004, 12,
1061-1066.
Lozovoy, V. V.; Shane, J. C.; Xu, B. W.; Dantus, M. Opt. Express 2005, 13,
10882-10887.
http://www. chemistry. msu. eduffacultv/dantus/merit_factor_records. html.
Coello, Y.; Jones, A. D.; Gunaratne, T. C.; Dantus, M. Anal. Chem. 2010, 82,
2753-2758.
Heeren, R. M. A.; Smith, D. F .; Stauber, J .; Kukrer-Kaletas, B.; MacAleese, L. J.
Am. Soc. Mass Spectrom. 2009, 20, 1006-1014.
McDonnell, L. A.; Heeren, R. M. A. Mass Spec. Rev. 2007, 26, 606-643.
Luxembourg, S. L.; Mize, T. H.; McDonnell, L. A.; Heeren, R. M. A. Anal.
Chem. 2004, 76, 5339-5344.
Luxembourg, S. L.; McDonnell, L. A.; Mize, T. H.; Heeren, R. M. A. J. Proteome
Res. 2005, 4, 671-673.
Boxer, S. G.; Kraft, M. L.; Weber, P. K. Ann. Rev. Biophys. 2009, 38, 53-74.
Winograd, N. Anal. Chem. 2005, 77, l42A-149A.
Franck, J .; Arafah, K.; Elayed, M.; Bonnel, D.; Vergara, D.; Jacquet, A.; Vinatier,
D.; Wisztorski, M.; Day, R.; Fournier, I.; Salzet, M. Mol. Cell. Proteomics 2009,
8, 2023-2033.
Cornett, D. S.; Reyzer, M. L.; Chaurand, P.; Caprioli, R. M. Nat. Methods 2007,
4, 828-833.
Van Berkel, G. J .; Pasilis, S. P.; Ovchinnikova, O. J. Mass Spectrom. 2008, 43,
1161-1180.
Becker, J. S.; Zoriy, M. V.; Pickhardt, C.; Palomero-Gallagher, N.; Zilles, K.
Anal. Chem. 2005, 77, 3208-3216.
Wu, B.; Zoriy, M.; Chen, Y. X.; Becker, J. S. T alanta 2009, 78, 132-137.
Nemes, P.; Vertes, A. Anal. Chem. 2007, 79, 8098-8106.
Nemes, P.; Barton, A. A.; Vertes, A. Anal. Chem. 2009, 81, 6668-6675.
Li, Y.; Shrestha, B.; Vertes, A. Anal. Chem. 2007, 79, 523-532.
104
(124)
(125)
(126)
(127)
(128)
(129)
(130)
(131)
(132)
(133)
(134)
(135)
(136)
(137)
(138)
(139)
Li, Y.; Shrestha, B.; Vertes, A. Anal. Chem. 2008, 80, 407-420.
Wiseman, J. M.; Ifa, D. R.; Song, Q. Y.; Cooks, R. G. Angew. Chem. -1nt. Edit.
2006, 45, 7188-7192.
Wiseman, J. M.; Ifa, D. R.; Zhu, Y. X.; Kissinger, C. B.; Manicke, N. E.;
Kissinger, P. T.; Cooks, R. G. Proc. Natl. Acad. Sci. USA 2008, 105, 18120-
18125.
Kertesz, V.; Van Berkel, G. J. Rapid Commun. Mass Spectrom. 2008, 22, 2639-
2644.
Jurchen, J. C.; Rubakhin, S. S.; Sweedler, J. V. J. Am. Soc. Mass Spectrom. 2005,
16, 1654-1659.
Shrestha, B.; Vertes, A. Anal. Chem. 2009, 81, 8265-8271.
Pastirk, I.; Zhu, X.; Lozovoy, V. V.; Dantus, M. Appl. Opt. 2007, 46, 4041-4045.
Dela Cruz, J. M.; Lozovoy, V. V.; Dantus, M. J. Mass Spectrom. 2007, 42, 178-
186.
Kalcic, C. L.; Gunaratne, T. C.; Jones, A. D.; Dantus, M.; Reid, G. E. J. Am.
Chem. Soc. 2009, 131, 940-942.
Samek, O.; Margetic, V.; von Wiren, N.; Michels, A.; Niemax, K.; Hergenroder,
R. Appl. Phys. A-Mater. Sci. Process. 2004, 79, 957-960.
Fernandez, B.; Claverie, F.; Pecheyran, C.; Donard, O. F. X. Trends Anal. Chem.
2007, 26, 951-966.
Brady, J. J .; Judge, E. J .; Levis, R. J. Rapid Commun. Mass Spectrom. 2009, 23,
3 151-3 157.
Willingham, D.; Kucher, A.; Winograd, N. Chem. Phys. Lett. 2009, 468, 264-269.
Savina, M. R.; Lykke, K. R. Anal. Chem. 1997, 69, 3741-3746.
Papantonakis, M. R.; Kim, J .; Hess, W. P.; Haglund, R. F. J. Mass Spectrom.
2002, 37, 639-647.
Wichmann, J. M.; Lupulescu, C.; Woste, L.; Lindinger, A. Rapid Commun. Mass
Spectrom. 2009, 23, 1105-1108.
105
(140) Gamaly, E. G.; Rode, A. V.; Luther-Davies, B.; Tikhonchuk, V. T. Phys. Plasmas
2002, 9, 949-957.
(141) Schilmiller, A.; Shi, F.; Kim, J .; Charbonneau, A. L.; Holmes, D.; Jones, A. D.;
Last, R. L. The Plant Journal, "Accepted Article ",' doi: 10.1 1 1 1/j.1365-
313X. 2010. 04154.x.
(142) Colliver, T. L.; Brummel, C. L.; Pacholski, M. L.; Swanek, F. D.; Ewing, A. G.;
Winograd, N. Anal. Chem. 1997, 69, 2225-2231.
106