Optical Multidimensional Coherent Spectroscopy Hebin Li Full Chapter

Optical Multidimensional Coherent
Spectroscopy Hebin Li
Visit to download the full and correct content document:
https://ebookmass.com/product/optical-multidimensional-coherent-spectroscopy-hebi
n-li/
Optical Multidimensional Coherent
Spectroscopy
Optical Multidimensional
Coherent Spectroscopy
H EB I N L I
Florida International University
B A C H A N A L O M S A DZ E
Santa Clara University
G A L A N MO O D Y
University of California Santa Barbara
C H R IST O P H ER L . S M A LLW O O D
San Jose State University
ST E V EN T. C U N D I F F
University of Michigan
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Hebin Li, Bachana Lomsadze, Galan Moody, Christopher L. Smallwood,
and Steven T. Cundiff 2023
The moral rights of the authors have been asserted
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2023930434
ISBN 978–0–19–284386–9
DOI: 10.1093/oso/9780192843869.001.0001
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
We dedicate this book to our families who had to put up with our absence while we
wrote it.
Preface
Ultrashort optical pulses, with duration ranging from a few picoseconds down to a few
femtoseconds, have been used to study dynamics in matter since pulsed lasers were
first developed in the 1960s. Indeed, this research area has been one of the drivers
for improvements in the field ultrafast optics, such as reductions in pulse duration.
The field of using ultrashort light pulses to probe dynamical processes in matter is
generally known as “ultrafast spectroscopy.”
Beginning in the late 1990s, the field of ultrafast spectroscopy underwent a rev-
olution due to the introduction of multidimensional coherent spectroscopy based on
concepts originally developed in nuclear magnetic resonance spectroscopy. Given their
power, multidimensional coherent methods are becoming the dominant ultrafast spec-
troscopic techniques.
This book presents optical multidimensional coherent spectroscopy methods and
their application to systems and materials that fall primarily within the field of physics.
The systems include atomic vapors and solids—particularly semiconductors and semi-
conductor nanostructures. Multidimensional coherent spectroscopy in the infrared and
visible spectral regions has been more extensively used to study molecules. As the ap-
plication of multidimensional coherent spectroscopy to molecular systems has been
covered by other books, we have chosen to not repeat that coverage here. Rather, we
seek to broaden the coverage by addressing applications that are largely not covered
elsewhere.
We begin by providing an introduction of multidimensional coherent spectroscopy
for researchers in all fields, whether or not they have a background in ultrafast spect-
roscopy, or even in optical spectroscopy more generally. We then focus on the use of
the technique to probe systems that are primarily of interest in the fields of physics
and materials science. Our goal is to illustrate the information that multidimensional
coherent spectroscopy can provide and its advantages over other methods. To do so,
we focus on several exemplary materials, but also aim to illustrate the technique’s
broader applicability.
Acknowledgements
We acknowledge all of those individuals who have worked or collaborated with us on de-
veloping and using multidimensional coherent spectroscopy including Diogo Almeida,
Travis Autry, Wan Ki Bae, Manfred Bayer, Camelia Borca, Alan Bristow, Xingcan
Dai, Matthew Day, Denis Karaiskaj, Irina Kuznetsova, Xiaoqin (Elaine) Li, Albert
Liu, Torsten Meier, Richard Mirin, Eric Martin, Shaul Mukamel, Gaël Nardin, Lazaro
Padilha, Marten Richter, Mark Siemens, Kevin Silverman, Rohan Singh, Bo Sun,
Takeshi Suzuki, Peter Thomas, Andreas Wieck and Tianhao Zhang. STC would like
to acknowledge the sustained funding for research on this topic from the DOE Atomic,
Molecular and Optical Science program and NSF funding through the JILA Physics
Frontier Center.
Table of symbols
Symbol Description Section

A cross-sectional interaction area 3.1
a0 Bohr radius 5.3
aB exciton Bohr radius 9.1
B̂/B̂ † exciton annihilation/creation operator 5.3
C probability amplitude 1.4
Ci ith London dispersion coefficient 5.3
c speed of light in vacuum 1
D spatial diffusion coefficient 1.3
Dα (x, y, z) double-sided Feynman diagram equation 3.1
E(t) total electric field in the time domain 1.2
Ê(t) complex-valued, slowly varying component of E(t) 1.2
Ẽ(t) phasor description of E 2.3.3
E0 real-valued electric field amplitude 1.2
EB biexciton energy 9.1
Ee single-particle electron energy 9.1
Eh single-particle hole energy 9.1
En energy difference between two quantum states 1
Esignal radiated electric field from polarization 2.3.3
EX exciton energy 9.1
F{...} Fourier transform operation 1.2
E(ω) total electric field in the frequency domain 1.2
E0 electric field pulse area 1.2
f0 comb offset frequency 6.1
fCE carrier-envelope offset frequency 4.4
fn frequency of nth comb tooth 6.1
frep laser repetition rate 4.4
G(ω) Fourier transform of g(t) into the frequency domain 3.2
g(t) Gaussian distribution function in the time domain 3.10
H Hamiltonian operator of a system 1.4
H0 unperturbed Hamiltonian operator of a system 1.4
h Planck’s constant, h = 6.62607015 × 10−34 J·s 1
ℏ reduced Planck constant, ℏ = h/2π 1.4
I optical intensity 1.2
ISI spectral interferogram intensity 4.3
I electric field power spectrum in the frequency domain 1.2
Jmn one-exciton coupling strength between excitons m and n 5.3
Kmn two-exciton coupling parameter 5.3
x Table of symbols
k wavevector of light, k = 2π/λ 1.1

L sample length 1.3
LB potential barrier width 9.1
Lc laser cavity length 6.1
Lmin minimum potential barrier width for confinement 9.1
mc conduction band effective mass 7.1
me electron mass 7.1
mhh heavy hole band effective mass 7.1
mlh light hole valence band effective mass 7.1
N integer number pulses, data points, etc. 4.4
N particle density 5
Np number of particles 5.4
n refractive index 1.1
⟨Ô⟩ expectation value of operator Ô 1.4
P dielectric polarization density 1.1
Pψ probability of being in state |ψ⟩ 1.4
Q(x) probability distribution function 3.2
R inter-atomic separation distance 5.3
rc correlation parameter 6.4
T delay between second and third pulses 1.3
T0 optical transmission 1.3
T1 excited state population lifetime 3.1
T2 excited/ground state coherence time 3.1
T2∗ ensemble dephasing time 3.2
Trep laser repetition period 4.4
t signal emission time 1.3
V interaction Hamiltonian 1.4
V (R) long-range atomic interaction potential 5.4
V0 potential barrier height 9.1
vgr group velocity 6.1
vph phase velocity 6.1
vs translation stage velocity 4.4
W population inversion between quantum states 1.4
Xhh heavy hole exciton 7.1
Xlh light hole exciton 7.1
α0 absorption coefficient 1.3
β optical frequency chirp parameter 1.2
∆1 singly excited state interaction shift 6.4
∆2 doubly excited state interaction shift 6.4
∆B biexciton binding energy 9.1
∆T change in optical transmission 1.3
Table of symbols xi
∆ϕce carrier-envelope phase-slip 6.1

∆X exciton binding energy 7.1
δ(t) Dirac delta function 2.3.1
δt temporal intensity pulse duration 1.2
δte temporal intensity e−1 half width 1.2
δte2 temporal intensity e−2 half width 1.2
δtFWHM temporal intensity full-width at half maximum 1.2
δU spectral intensity bandwidth in energy 1.2
δα0 change in absorption coefficient 1.3
δκ spectral intensity bandwidth in wavenumbers 1.2
δλ spectral intensity bandwidth in wavelength 1.2
δν spectral intensity bandwidth in frequency 1.2
δω spectral intensity bandwidth in angular frequency 1.2
δωe spectral intensity e−1 half width 1.2
δωe2 spectral intensity e−2 half width 1.2
δωFWHM spectral intensity full-width at half maximum 1.2
ϵ0 vacuum permittivity 3.1
Γ relaxation operator 1.4
Γ10 excited state population decay rate 3.1
Γgr grating relaxation rate 1.3
Γpop population relaxation rate 1.3
γ dephasing rate / homogeneous linewidth 1.3
γ10 excited/ground state decoherence rate 3.1
γij imaginary part of resonance frequency between states i and j 2.3.1
γ ph pure dephasing rate 1.4
γt single quantum dephasing rate 6.4
γτ double quantum dephasing rate 6.4
Θ(x) Heaviside step function 1.3
θ angle between beams 1 and 2 1.3
ϑ Bloch sphere polar angle 1.5
Λ spatial grating period 1.3
λ wavelength of light in vacuum 1
µ electric dipole moment 1.4
ν frequency of light 1
ν∗ difference frequency between signal and CW reference laser 4.4
π mathematical constant π ≈ 3.1415 1.1
ρ density matrix operator 1.4
σ N -particle spectral distribution width 3.2
τ delay between first and second pulses 1.3
τex two-level transition excited-state lifetime 1.3
χ electric susceptibility 1.1
xii Table of symbols
ϕ optical phase 1.2

ϕS,R phase of signal/reference electric field 4.3
φ Bloch sphere azimuthal angle 1.5
|ψ⟩ quantum mechanical wave function 1.4
Ωij complex resonance frequency between states i and j 2.3.1
Ωr Rabi frequency 1.4
ω optical angular frequency 1.2
ω0 resonant angular frequency of an optical transition 3.2.1
ωc carrier angular frequency of a laser 1.2
ωij real part of resonance frequency between states i and j 2.3.1
ωt frequency conjugate to emission time, t 3.1
ωτ frequency conjugate to coherence time, τ 3.1
Table of acronyms
Acronym Description
2DCS two-dimensional coherent spectroscopy
2DES two-dimensional electronic spectroscopy
AOM acousto-optic modulator
BS beamsplitter
◦
C degrees Celsius
c.c. complex conjugate
CAD computer-aided design
CB conduction Band
CCD charge-coupled device
CP compensation plate
CQD colloidal quantum dot
CVD chemical vapor deposition
CW continuous wave
DBR distributed Bragg reflector
DCM dichroic mirror
DCS dual-comb spectroscopy
DDS direct-digital synthesizer
DFT discrete Fourier transform
DQW double quantum well
DSP digital signal processor
EID excitation induced dephasing
EIS excitation induced shift
ESE excited-state emission
FFT fast Fourier transform
FPGA field programmable gate array
FTIR Fourier-transform infrared
FWHM full-width at half-maximum
GNE gold nanoelectrode
GSB ground-state bleaching
HeNe helium-neon
hh heavy hole
HWP half-wave plate
IFQD interface fluctuation quantum dots
IR infrared
lh light hole
LO local oscillator
LP lower polariton
xiv Table of acronyms
MAPI Methylammonium Lead Iodide

MBE molecular beam epitaxy
MDCS multi-dimensional coherent spectroscopy
MONSTR multidimensional optical nonlinear spectrometer
MQC multi-quantum coherence
NA numerical aperture
NMR nuclear magnetic resonance
NW narrow well
OD optical density
PBS polarizing beamsplitter
PL photoluminescence
PLE photoluminescence excitation
PZT piezoelectric transducer
QD quantum dot
QW quantum well
RBM radial breathing mode
RF radio frequency
SAQD self-assembled quantum dot
SERS surface-enhanced Raman spectroscopy
SNR signal-to-noise ratio
SR-TFWM spectrally resolved transient four-wave mixing
STM scanning tunneling microscope
SWNT single-walled carbon nanotube
TBP time-bandwidth product
TC lock-in amplifier time constant
TCS tri-comb spectroscopy
TFWM transient four-wave mixing
TI-TFWM time-integrated transient four-wave mixing
TMD Transition Metal Dichalcogenide
Tr trace of a linear operator
TR-TFWM time-resolved transient four-wave mixing
UP upper polariton
VB valence band
WW wide well
ZPL zero phonon line
Contents
1 Basics of ultrafast spectroscopy 1

1.1 Basics of spectroscopy: linear versus nonlinear 2
1.2 Ultrashort pulses 6
1.3 Ultrafast nonlinear/coherent spectroscopy 10
1.4 The density matrix 15
1.5 Bloch sphere representation of quantum states 20
2 Introduction to multidimensional coherent spectroscopy 23
2.1 Concepts of multidimensional coherent spectroscopy 24
2.2 Spectrum classification 29
2.3 Density matrix formalism and double-sided Feynman diagrams 32
2.4 Phase matching 43
2.5 Two-dimensional infrared (2D IR) spectroscopy 44
3 Interpretation of multidimensional coherent spectra 45
3.1 Isolated two-level system 45
3.2 Inhomogeneously broadened ensemble of two-level systems 49
3.3 Coherent coupling signatures 55
3.4 Incoherent coupling signatures 57
3.5 Doubly excited states and many-body interactions 58
3.6 Double-quantum spectra 62
3.7 Zero-quantum spectra 65
3.8 Three-dimensional coherent spectroscopy 67
3.9 Nonrephasing pathways and purely absorptive spectra 67
3.10 Finite-pulse effects 68
4 Experimental implementations 77
4.1 Experimental requirements and considerations 77
4.2 Overview of experimental approaches 80
4.3 Actively stabilized box geometry 81
4.4 Phase modulated collinear geometry 89
4.5 Comparison of different approaches 97
4.6 Data analysis 97
5 Multidimensional coherent spectroscopy of atomic ensembles 101
5.1 Single- and zero-quantum 2D spectra of atomic vapors 102
5.2 MDCS in optically thick samples 108
5.3 Probing many-body interactions with double-quantum 2D spect-
roscopy 112
xvi Contents
5.4 Probing many-body correlations with multi-quantum 2D spect-

roscopy 121
6 Frequency comb-based multidimensional coherent
spectroscopy 127
6.1 Introduction to frequency combs and dual-comb spectroscopy 128
6.2 Frequency comb-based four-wave-mixing spectroscopy 130
6.3 Frequency comb-based single-quantum 2D spectroscopy 135
6.4 Frequency comb-based double-quantum 2D spectroscopy 139
6.5 Tri-comb spectroscopy 144
7 Two-dimensional spectroscopy of semiconductor quantum
wells 151
7.1 Introduction to semiconductor optics 152
7.2 Many-body signatures in one-quantum 2D spectra 154
7.3 Many-body signatures in double- and multi-quantum 2D spectra 161
7.4 Two-dimensional spectroscopy of coupled quantum wells 165
7.5 Quantum well exciton-polaritons in microcavities 168
8 Three-dimensional coherent spectroscopy 171
8.1 Fifth-order 3D infrared spectroscopy 172
8.2 Fifth-order 3D electronic spectroscopy 173
8.3 Third-order 3D electronic spectroscopy 174
8.4 3D coherent frequency domain spectroscopy 191
9 Two-dimensional spectroscopy of semiconductor quantum
dots 193
9.1 Optical and electronic properties of quantum dots 193
9.2 2D coherent spectroscopy of GaAs quantum dots 197
9.3 2D spectroscopy of self-assembled In(Ga)As quantum dots 202
9.4 Coherent control of quantum dots 205
9.5 Two-dimensional spectroscopy of colloidal quantum dots 207
10 Two-dimensional spectroscopy of atomically thin 2D
materials 213
10.1 Introduction to 2D materials 213
10.2 Homogeneous linewidth in 2D materials 214
10.3 Valley coherence and coupling in 2D materials 217
11 Other applications of multi-dimensional coherent spectroscopy
in Physics 223
11.1 Semiconducting carbon nanotubes 223
11.2 Color centers in diamond 225
11.3 Perovskite materials 228
12 New trends in multidimensional coherent spectroscopy 235
12.1 Broadening the spectral range: from terahertz to x-rays 235
12.2 Improving the spatial resolution 237
Contents xvii
12.3 Multidimensional spectroscopy with quantum light 239

12.4 Photoemission-detected MDCS 240
Figure Credits 243
References 245
Index 279
1
Basics of ultrafast spectroscopy
Starting from Issac Newton’s use of a prism to observe the spectrum of sunlight, optical
spectroscopists have been striving to further our understanding of matter by studying
how it absorbs and emits light. Spectroscopic techniques remained fundamentally un-
changed for centuries compared to Newton’s method. Light was dispersed by a prism
or diffraction grating and the intensity was measured as a function of angle, which can
be mapped into wavelength. Changes in the spectral intensity can be related to either
inherent properties of the source or the light passing through a medium that absorbs
at specific wavelengths. The absorption or emission wavelength can be converted to a
frequency through c = λν where c is the speed of light, λ is the wavelength of the light
and ν is its frequency. Using the quantum mechanical relation between energy, En , and
frequency, En = hν, where h is Planck’s constant, the frequency can be understood
as the energy difference between two quantum states such as electronic or vibrational
states.
Before the invention of the laser, spectroscopic measurements were all performed in
the linear regime where the material properties are independent of the intensity of the
light. In this regime the electric field of the light is weak compared to the internal fields
of the atom or molecule. A laser can produce light that is no longer weak compared to
the internal fields of an atom or molecule, thus the invention of the laser brought the
field of optical spectroscopy into a new era of nonlinear spectroscopy. In the nonlinear
regime, the material properties are no longer independent of the light intensity, signals
scale with a higher power of the laser intensity, and two laser beams can interact in a
sample. For instance, a strong pump beam can saturate an absorption resonance and
thus increase the transmission of a weaker probe beam. Also, wave-mixing of multiple
light fields in the sample can lead to a signal beam with an entirely new direction and
frequency.
Optical spectroscopic measurements can also be made not as a function of wave-
length but rather of time. Time-domain spectroscopy is analogous to the idea of a stro-
boscope, in that short flashes of light can capture stop-action images of ultrafast dy-
namics such as a chemical reaction or charge carriers relaxing in a solid. Measurements
made in the time domain using laser pulses can be converted into frequency-domain
spectra using a Fourier transform. In general, the time-domain spectra can be func-
tions of multiple time delays, so the resulting frequency-domain spectra are functions
of multiple frequencies and thus are multidimensional. The concept of multidimen-
sional Fourier transform spectroscopy was developed in nuclear magnetic resonance
(NMR) [108] and is now transforming the field of ultrafast laser spectroscopy.
2 Basics of ultrafast spectroscopy
Multidimensional coherent spectroscopy has a number of advantages over other

types of spectroscopy, including one-dimensional methods and multidimensional meth-
ods that are not coherent. At the same time, multidimensional coherent spectroscopy
is challenging to experimentally implement in the optical portion of the electromag-
netic spectrum due to the need to use phase-related light pulses to excite the sample
and to measure the phase of the emitted light signals.
To set the stage before discussing multidimensional coherent methods, in this chap-
ter we will review several ultrafast spectroscopic methods, both because they serve as
the foundation from which multidimensional coherent methods were developed, and
they provide context for describing the advantages of multidimensional coherent meth-
ods. In Chapter 2 we will introduce the basic concepts involved in multidimensional
coherent spectroscopy, followed by an in-depth discussion of how to interpret multidi-
mensional coherent spectra in Chapter 3. Chapter 4 will review several experimental
implementations, describing how each overcomes the aforementioned challenges.
1.1 Basics of spectroscopy: linear versus nonlinear

The field of spectroscopy involves measuring a spectrum that displays the frequency
(spectral) dependence of the interaction between matter and electromagnetic radiation.
The electromagnetic radiation may be incident on the matter from an external source,
or it may be emitted by the matter. In this book, we will discuss the former case.
When an electromagnetic field is incident on matter, it displaces the electrons or
ions in the matter from their equilibrium positions, producing a polarization in the
matter that in turn radiates a new electromagnetic field. Treating the polarization as a
driving term in Maxwell’s equations and taking the far-field limit gives the result that
the phase of the reradiated field lags the phase of the polarization by 90◦ (a factor of i
in complex phasor notation). The interference of this reradiated field with the incident
field results in modification of the field due to propagation through the matter, which
is usually attributed to material properties such as an index of refraction or absorption.
In linear spectroscopy, the incident electromagnetic field is weak and the induced
polarization is linearly proportional to the incident field. For a continuous wave (CW)
incident field, the polarization will have the same frequency and wavevector as the
incident field. However, the phase of the polarization with respect to the incident field
depends on frequency if resonances are present in the material. First, we consider only
a single resonance as the simplest example.
If the frequency of the incident field is significantly below the resonant frequency,
then the induced polarization will be in-phase with the incident field and the reradiated
field will be 90◦ out of phase in the far field due to aforementioned far-field phase lag.
Since the polarization has the same wavevector as the incident field, the reradiated
field will propagate in the same direction, so a detector placed after the sample will
detect the sum of the incident field and the reradiated field. If the reradiated field is
weak compared to the incident field, the dominant result will simply be a phase shift
of the transmitted field compared to the incident field, as illustrated in Fig. 1.1(a)
[note that eikx + iδeikx ≈ ei(kx+δ) for small values of δ, where k = 2πn/λ]. This phase
shift is consistent with a transparent material with an index of refraction n.
Basics of spectroscopy: linear versus nonlinear 3
(a) Incident field Radiated field ∝ iP

(d) Polarization
P Amplitude
Sample (a) (c)
(b)
0
π
(b) Phase
π/2
P 0
Sample
Real
Imaginary
(c)
0
P
Sample
ω0
Fig. 1.1 Sketch of linear spectroscopy showing incident fields and reradiated fields in the
vicinity of an optical resonance. (a) Below resonance. (b) On resonance. (c) Above resonance.
(d) Frequency dependence of the complex polarization, in terms of amplitude and phase (top)
and real and imaginary quadratures (bottom).
If the frequency of the incident field is tuned so that it is on the resonance in

the sample, the phase of the polarization will lag that of the incident field by 90◦ .
Together with the far-field phase shift, the net result will be that the reradiated field
will be 180◦ out of phase with the incident field and they will experience destructive
interference when measured together, as sketched in Fig. 1.1(b). This destructive in-
terference means that a lower transmitted intensity is measured, as expected, because
the incident field is now being absorbed by the sample since it is on resonance.
Tuning the frequency of the incident field to be above resonance results in a situ-
ation similar to below resonance, just a phase shift of the transmitted field, although
the phase shift is in the opposite direction, as shown in Fig. 1.1(c).
Taken as a whole, this simplified picture of absorption spectroscopy is useful be-
cause it emphasizes the importance of the relative phase between the incident field and
the induced polarization: it is all the difference between a material being transparent
but causing a phase shift of the transmitted light and the material absorbing the light
and resulting in less light being transmitted. The ability to measure the phase of the
polarization with respect to the incident fields can be implemented in multidimensional
coherent spectroscopy, and it yields important information about how the material is
responding to the incident light.
The complete frequency dependence of the material polarization in the vicinity of
the resonance is depicted in Fig. 1.1(d), divided into complex amplitude and phase
components at top and into real and imaginary quadratures on bottom. The quadra-
ture depiction, in particular, exhibits a number of important features that will serve as
useful reference points for the nonlinear multidimensional spectra that will be discussed
in much of the rest of this book. As mentioned above, when the polarization signal is
both small and primarily real-valued, the dominant effect on the emitted signal is a
change in phase. In fact, the real part of the polarization remains closely connected
to phase shifts in the optical field regardless of what imaginary part if the overall
reradiated signal amplitude remains small in comparison to the incident field. Because
these phase shifts are different for light at different frequencies, the quadrature is very
often referred to in spectroscopy literature as the polarization’s “dispersive” compo-
nent. Likewise, the polarization’s imaginary component bears a strong resemblance
to the material’s absorption properties and as such is frequently termed the signal’s
“absorptive” component. In terms of the reradiated field—which is the most com-
monly examined quantity in MDCS—the dispersive component shifts into the signal’s
imaginary quadrature and the absorptive component shifts into the real quadrature.
This discussion has described a simple spectroscopy experiment, namely sweep-
ing the frequency of an electromagnetic field incident on a sample and measuring the
transmitted intensity. If a dip in the transmitted intensity is observed at a specific
frequency it indicates the presence of resonance; the width of the resonance character-
izes the damping of the resonance. However, there are important ambiguities in linear
spectra, namely the inability to distinguish between homogeneous and inhomogeneous
broadening and the inability to determine if two resonances are coupled or uncoupled.
Typically the sample is an ensemble of many systems, whether they are atoms un-
dergoing electronic transitions, molecules undergoing vibrational transitions, or nuclei
flipping their spins. If all the systems in the ensemble are identical, i.e., they have the
same resonant frequency and same linewidth, the ensemble is designated as being “ho-
mogeneously broadened”; however, this may not necessarily be the case. In particular,
there may be a distribution of resonant frequencies due to effects such as the Doppler
shift in a vapor, random crystal fields in an ion-doped solid or structural disorder in a
nanostructure. In this case, the linewidth of the measured resonance may have nothing
to do with the linewidth of the individual members of the ensemble, but rather re-
flects the distribution of resonance frequencies. This case is known as “inhomogeneous
broadening.”
The linear spectrum of an inhomogeneously broadened ensemble will have a res-
onance feature (the absorption “line”) that has a width that is characteristic of the
inhomogeneous distribution, not the damping of the individual members of the en-
semble. While both are useful to know, they provide quite different information. The
width in the absence of the inhomogeneous broadening, often called the “homoge-
neous width” provides information about processes that interrupt the oscillations, for
example collisions and radiative decay.
There is also an ambiguity in linear spectroscopy if two resonances are observed in
a linear spectrum. There are two possible situations. One possibility is that the sample
is heterogeneous, i.e., a mixture of two species with different resonance frequencies.
The other possibility is that it is pure, i.e., a single substance, but that substance has
two transitions. A good example of this latter case would be the D1 and D2 lines in
the alkali metals, which correspond to the single outer electron making a transition
Basics of spectroscopy: linear versus nonlinear 5
to ground S1/2 state the P1/2 and P3/2 states. Linear spectroscopy cannot distinguish
between these two possibilities.
These ambiguities can be resolved by using some form of nonlinear spectroscopy. In
nonlinear spectroscopy, as the intensity of the excitation field is increased, the dielectric
polarization P (ω) of a material is no longer linearly proportional to the incident field,
but rather higher-order terms must be considered. In the frequency domain, we can
describe this in terms of a power series expansion of P (ω) as function of electric field
strength E(ω) as
h i
P [E(ω)] = ϵ0 χ(1) E + χ(2) E 2 + χ(3) E 3 + χ(4) E 4 + . . . , (1.1)
where the constant ϵ0 is the vacuum permittivity and the coefficients χ(n) (ω) describe
electric susceptibility parameters of the material at each of the different orders n.
Linear spectroscopy corresponds to the situation where all the terms χ(n) E n for n > 1
are small enough to be neglected, resulting in the relationship P (ω) = ϵ0 χ(1) (ω)E(ω).
Of course, this approximation depends on the strength of E, because for large enough
E, the E n factor can make χ(n) E n > χ(1) E, no matter how small the ratio χ(n) /χ(1) . It
is possible to show that for inversion symmetric systems, the second-order term in the
expansion of Eq. (1.1), and indeed all of the even-valued higher-order terms, must be
identically equal to zero.1 Hence, in delving into the world of nonlinear spectroscopy, it
is often the χ(3) term that is actually the most important element governing nonlinear
corrections to the polarization as a whole, and so it is upon this term that we will
most heavily concentrate our attention in this book.
To understand how nonlinear spectroscopy can resolve the ambiguities in a linear
spectrum, it is easiest to consider a simple frequency-domain method known as “spec-
tral hole burning.” In spectral hole burning, a continuous wave (CW) “pump” laser
excites the sample, saturating its absorption. A second laser is then scanned to measure
the absorption of the sample. If the sample is homogeneously broadened, the absorp-
tion of the entire line simply decreases. However, if it is inhomogeneously broadened,
then the sub-ensemble that is resonant with the pump laser is most strongly saturated.
In this case, the measured absorption spectrum is unchanged, except in the spectral
region close to the pump, where the absorption is decreased, known as “burning a
hole.” The width of the spectral hole is proportional to the homogeneous width. Thus
the observation of spectral hole burning shows that the system is inhomogeneously
broadened and the width gives the homogeneous width.
Similarly, if two resonances are present in the spectrum, tuning the pump laser to
one resonance and probing the other can determine if they are coupled. If they are
coupled, then this situation will result in a change in the absorption, whereas if they
are uncoupled it will not. This example was based on using CW lasers. Although there
are some implementations of optical multidimensional coherent spectroscopy based
on this approach [55–57, 443], most are based on using mutually coherent pulses and
scanning their delays.
A time-domain multidimensional coherent spectroscopy (MDCS) measurement is
made by illuminating a sample with a series of light pulses and measuring a signal
1 See, for example, Nonlinear Optics, by Robert W. Boyd [40].
from the sample as a function of the delays between the pulses. Typically the pulses
have duration of a few picoseconds or less, which is considered the domain of “ultrafast
optics,” where traditional photodetectors are too slow to directly measure the pulse
duration. Due to their short duration, such pulses intrinsically have broad spectral
bandwidth, thus spectral features can be measured without tuning them spectrally,
but rather by spectrally resolving them. While this can be done using traditional
spectrometers, it can also be realized using Fourier transform methods. Some MDCS
approaches only use Fourier transforms, whereas others use a combination of Fourier
transforms and a spectrometer to spectrally resolve the signal.
Before we introduce MDCS, we need to briefly review the properties of ultrashort
pulses and introduce less related forms of spectroscopy that are based on ultrashort
optical pulses.
1.2 Ultrashort pulses

A short optical pulse passing through a fixed point in space can be described by its
electric field in the time domain
E(t) = Ê(t) cos (−ωc t + ϕ(t))

1
= Ê(t)e−iωc t + c.c., (1.2)
2
where ωc is the carrier frequency, and where ϕ(t) is a time-dependent phase. The
second line of the equation is expressed in phasor notation, with the complex-valued
amplitude Ê(t) = |Ê(t)|eiϕ(t) , and with the abbreviation “c.c.” standing for “complex
conjugate.” For the discussion throughout the first three chapters of this book, we
ignore the the fact that light has a polarization, and thus treat the electric field as
a scalar. Note that the choice of ωc is in principle arbitrary; the same pulse could
be described using a different ωc by adjusting the time dependence of the amplitude
coefficients to include linearly ramping phase factors. Typically, however, ωc is chosen
to eliminate a linear ramp in ϕ(t). Throughout this book we use a caret (X̂) placed
over the top of a function X to indicate a value assumed to vary “slowly” in time (i.e.,
to be nearly constant over several periods of the optical frequencies).
Although slow in comparison to the carrier oscillations, time dependence of the
amplitude factor Ê(t) as introduced in Eq. (1.2) is still required to shape the light into
a pulse. Writing out the amplitude and phase components of Ê(t) explicitly and then
expressing them in terms of optical intensity I(t) ≡ nϵ0 c⟨E(t)2 ⟩ leads to
s
1 2I(t) −i[ωc t−ϕ(t)]
E(t) = e + c.c., (1.3)
2 nϵ0 c
where n is refractive index, ϵ0 is the vacuum permittivity, c is the speed of light,

and the angle brackets in the definition of intensity specify the time average over
an optical period. As shown in Fig. 1.2, the pulse described using Eq. (1.3) can be
visualized
p as an oscillating carrier, of frequency ωc , under an envelope proportional
to I(t). The constant in the phase factor ϕ(t) allows the possibility of a shift in
Ultrashort pulses 7
1 Electric field
Envelope
Electric field (arb. units)
0.5
-0.5
-1
-60 -40 -20 0 20 40 60

Time (fs)
Fig. 1.2 The electric field (solid red line) and the envelope (blue line) of an ultrafast pulse.
alignment between the ripples of the carrier and the overall envelope position, known
as the “carrier-envelope phase.”
One reasonable mathematical description of an ultrafast pulse is Gaussian. The
envelope of a Gaussian pulse is
2
− (δtt 2
Ê(t) = E0 e e2 ) (1.4)
where E0 is the (real-valued) amplitude and the parameter δte2 describes pulse dura-
tion. The pulse intensity associated with Eq. (1.4) is
2 t2√
− (δt2t − (δt
I(t) ∝ |Ê(t)|2 = E02 e e2 )
2
= E02 e e2 / 2)2
, (1.5)
which demonstrates that for a Gaussian pulse, the duration parameter δte2 corresponds
to the half width at which the pulse intensity drops to 1/e2 (13.5%) of its peak. Beyond
this,
√ the equation shows that the pulse duration of the intensity function is a factor of
2 shorter than the pulse duration of the amplitude function. Although the 1/e2 half-
width definition of pulse duration is common, the duration of a pulse can be defined
according to any number of different conventions, including the intensity half width
1/e value (δte ), the intensity full width at half maximum (FWHM) value (δtFWHM ),
and various definitions associated with the intensity autocorrelation function. Thus, it
is important to define the pulse duration carefully when introducing it as a parameter
in order to avoid ambiguities. For Gaussian pulses, the duration parameters δte , δte2 ,
and δtFWHM are related to each other by
√ √
(δtFWHM ) = 2 ln 2 (δte2 ) = 2 ln 2 (δte ). (1.6)
However, the conversion factors relating the various width definitions are different for
different pulse shapes.
An ultrashort pulse can also be described in the frequency domain where the electric
field is written as a function of the angular frequency or frequency. The conversion
between the time and frequency domains is performed by the Fourier Transform of
the electric field (not the intensity), where the forward Fourier Transform
Z ∞
E(ω) = E(t)eiωt dt (1.7)
−∞
converts from a time-domain signal, E(t), to its frequency-domain representation,

E(ω), and the inverse Fourier transform
Z ∞
1
E(t) = E(ω)e−iωt dω. (1.8)
2π −∞
converts the other direction, giving the time-domain waveform from the frequency-
domain representation. Note that there differing conventions for the definition of the
Fourier Transform with regards to the sign of the exponent in the kernel (eiωtqversus
e−iωt ) and normalization ( 2π
1 1
in front of the inverse Fourier transform versus 2π in
front of both the forward and inverse transforms). Throughout this book we will use
the conventions given in Eqs. (1.7) and (1.8). The Fourier transform of Eq. (1.3) gives
s s
1 2I(ω − ωc ) iϕ(ω−ωc ) 1 2I(−ω − ωc ) −iϕ(−ω−ωc )
E(ω) = e + e . (1.9)
2 nϵ0 c 2 nϵ0 c
The field in the frequency domain has both positive and negative frequencies. The two
frequency components are actually equivalent, but required to make the signal real.
We usually take only the positive frequency and center the pulse at zero frequency
(i.e., expressed in the rotating frame) in complex phasor notation as
s
1 2I(ω) iϕ(ω)
E(ω) = e . (1.10)
2 nϵ0 c
Here I(ω) is the spectral intensity function and the frequency-domain phase is

Im[E(ω)]
ϕ(ω) = arctan . (1.11)
Re[E(ω)]
The phase calculated from Eq. (1.11) can only vary from −π to π so there could be 2π
phase jumps. We can unwrap the phase by adding or subtracting 2π to avoid phase
jumps and have a continuous phase. The phase defined by Eq. (1.11) has no meaning
when the intensity is zero. The phase is noisy when the intensity is small at the wings
of spectrum, in which case the phase is usually not plotted. According to the Fourier
Ultrashort pulses 9
shift theorem, a delay in the time domain is a linear phase ramp in the frequency
domain, i.e., F{E(t − t0 )} = E(ω)eiωt0 .
Applying the Fourier transform to the time-domain Gaussian pulse described by
Eq. (1.4) gives the frequency-domain pulse in a frame rotating at the carrier frequency
as
√ (δte2 )2 ω 2
E(ω) = π(δte2 )E0 e− 4 . (1.12)
The intensity spectrum of the pulse is
2
nϵ0 c (δte2 )2 ω 2 − 2ω
I(ω) = π(δte2 )2 E02 e− 2 ∝ E02 e (2/δte2 )2 , (1.13)
2
which demonstrates that the spectral profile of a time-domain Gaussian pulse is also
a Gaussian. The 1/e2 half width intensity spectral bandwidth of this Gaussian is
(δωe2 ) = 2/(δte2 ).
As in the time-domain representation of the electric field, there are different kinds
of conventions for describing spectral bandwidth, including the spectral intensity 1/e2
half width (δe2 ), the spectral intensity 1/e half width (δωe ), and the spectral intensity
FWHM value (δωFWHM ). In similar fashion to the ways that time-domain pulse dura-
tions are related to each other, for Gaussian pulses these frequency-domain bandwidth
descriptions are connected through the relationship
√ √
(δωFWHM ) = 2 ln 2 (δωe2 ) = 2 ln 2 (δωe ). (1.14)
Different research fields also use different units in the frequency domain. The spectrum
and bandwidth can be presented as a function of frequency (THz), angular frequency
(rad/s), wavelength (nm), energy (eV), or wavenumber (cm−1 ). The bandwidth and
even the lineshape are different when different units are used. The bandwidth can be
properly converted into different units. As an example, we consider a short pulse that
has a spectral bandwidth in frequency as
δν = 5 THz. (1.15)
This bandwidth can be converted, for example, into units of angular frequency,
δω = 2π(δν) = 3.14 × 1013 rad/s, (1.16)
wavenumber,
(δν)
δκ = = 167 cm−1 , (1.17)
c
energy,
δU = h(δν) = 20.7 meV, (1.18)
or wavelength,
λ2
δλ = (δν), (1.19)
c
where the last of these expressions is wavelength dependent. Assuming a wavelength
of 800 nm, we have δλ ≈ 10 nm.
√
Considering the intensity pulse duration δtFWHM = 2 ln 2 (δte2 ), the pulse dura-
tion and the spectral bandwidth are related as δtFWHM = 4 ln 2/(δωFWHM ). Therefore,
a shorter pulse in the time domain requires a broader spectrum in the frequency do-
main and vice versa. The shortest possible pulse for a given bandwidth, known as the
Fourier transform-limited pulse, can be achieved when the spectral phase is constant
across the spectrum. The product of the bandwidth and the temporal duration of a
pulse is a dimensionless number known as the time-bandwidth product (TBP). The
TBP has a minimum value for a transform-limited pulse and the exact value depends
on the pulse shape. If consistent units are used for time and bandwidth, the TBP
is a unitless number. For example, a transform-limited Gaussian pulse has a TBP
of (δtFWHM )(δνFWHM ) = 0.441 while the TBP is (δtFWHM )(δνFWHM ) = 0.315 for
a transform-limited sech2 -shaped pulse. These relations imply that 100-fs Gaussian
pulses must have a bandwidth of 4.41 THz while 100-fs sech2 -shaped pulses need a
minimum bandwidth of 3.15 THz.
A pulse can have a carrier frequency that varies in time, in which case the pulse is
called a chirped pulse. A linearly chirped Gaussian pulse can be written as
2
− (δtt 2
E(t) = E0 e e2 )
2
e−i(ωc t+βt ) , (1.20)
where ωc is the carrier frequency and βt2 is the chirp. The term βt2 modifies the carrier
frequency and varies with time. It can be considered as a second-order phase. When
β is positive, the pulse increases its frequency linearly in time (from red to blue) and
is positively chirped. When β is negative, the pulse decreases its frequency linearly in
time (from blue to red) and is negatively chirped.
Fourier transforming Eq. (1.20) gives the frequency-domain expression of the chirped
pulse,
(δωe2 /2)/4 β/4
− (δω
1/4
(ω−ωc )2 − (δω 2 2 (ω−ωc )2 −i (δω 2 2 (ω−ωc )2
E(ω) = E0 e e2 /2)−iβ = E0 e e2 /2) +β , e e2 /2) +β
(1.21)
with δωe2 = 2/(δte2 ). Adding a chirp in the time domain changes the spectral width
but not the temporal width of the pulse, while adding a chirp in the frequency domain
changes the temporal width but not the spectral width of the pulse. In an experiment,
a chirp is usually created by propagating through a dispersive medium, which is to say
that the chirp is added in the frequency domain. As a result, in experiments, adding
chirp typically results in increasing the temporal width of the pulses.
1.3 Ultrafast nonlinear/coherent spectroscopy

Ultrafast pulses enable unique capabilities in spectroscopy. First, the time resolution
provided by ultrafast pulses can probe events that occur on fs to ps timescales. Sec-
ond, the high instantaneous power and hence the strong electric field in ultrafast
pulses can lead to more efficient nonlinear effects for nonlinear spectroscopy. Finally,
a proper pulse sequence can be used to perform coherent spectroscopy. Incoherent
spectroscopy, such as time-resolved fluorescence/luminescence and transient absorp-
tion spectroscopy, is only sensitive to population relaxation and the results can be
interpreted by modeling with rate equations. In contrast, coherent spectroscopy, such
Ultrafast nonlinear/coherent spectroscopy 11
Fig. 1.3 Schematic of a typical pump-probe setup with two pulses and lock-in detection.
as transient four-wave mixing and multidimensional coherent spectroscopy, also probes

phase relaxation and interpreting coherent spectra requires optical Bloch equations.
The simplest ultrafast technique is time-resolved fluorescence/luminescence spect-
roscopy. The technique uses only one pulse to excite the sample from the ground state
to a high-lying excited state. The sample then relaxes to a lower excited state from
which a fluorescence signal is spontaneously emitted. The fluorescence has a longer
wavelength than the pump pulse so the signal can be distinguished from scattered
pump photons. The signal can be recorded as a function of time by using time-resolved
detection such as time-correlated photon counting or a streak camera, which has a
typical time resolution of a few ps. The time resolution can be improved by cross
correlating the fluorescence signal and a reference pulse through an upconversion non-
linear process, in which case the time resolution is only limited by the pulse duration.
The measured fluorescence signal usually rises initially and then decays exponentially.
The rise time of fluorescence is related to the relaxation time from the high-lying state
to the lower excited state, while the decay dynamics are determined by the relaxation
from the lower excited state to the ground state.
Pulse duration-limited time resolution can also be achieved by using a slow detector
in the two-pulse transient absorption technique (also known as pump-probe). A typical
pump-probe setup is depicted in Fig. 1.3. The sample is first excited by a pump pulse
and the change in absorption due excitation created by the pump pulse is measured
using a subsequent probe pulse. Typically the absorption decreases after the pump
pulse due to bleaching of the transition by the pump-induced population of excited
states and depletion of the ground state population. As the population relaxes from the
excited state back into the ground state, the absorption will recover, which is monitored
by changing the delay between pump and probe pulses. The time delay between the
pump and probe pulses can be varied by translating a mirror or retroreflector using a
delay stage, thereby changing the path length and hence time delay due to the changed
time-of-flight. The average power of the transmitted probe pulse is measured by a slow
detector. The change in the probe power due to the excitation by the pump is detected
by using a lock-in amplifier as the pump beam is modulated by a optical chopper. The
signal is recorded as a function of the time delay between the pump and probe pulses.
The time resolution is determined by the minimum increment of the delay stage and
the pulse duration. Assume that the unexcited sample has an absorption coefficient
α0 and the absorption coefficient decreases by ∆α0 immediately after the excitation.
For an excited state that decays exponentially with an excited-state lifetime of τex ,
the change in the absorption coefficient at a delay time τ is given by
∆α(τ ) = ∆α0 e−τ /τex . (1.22)
So the transmitted probe intensity and hence the average power depend on the delay
τ and the lifetime τex as
−τ /τex
Itran (τ ) = Iinc e−[α0 −∆α0 e ]L
≈ Itran (0)(1 + ∆α0 e−τ /τex L), (1.23)
where Iinc and Itran are the incident intensity and the transmitted intensity, respec-
tively, L is the sample length, and we assume ∆α0 L ≪ 1. The differential transmission,
i.e., the relative change in transmitted intensity, or differential transmission, is
∆T (τ ) Itran (τ ) − Itran (0)

=
T0 Itran (0)
≈ ∆α0 e−τ /τex L, (1.24)
where T0 is the transmission of unexcited sample and ∆T (τ ) is the change in the

transmission at a delay τ . In the simplest case, the pump-probe signal features an
exponential decay and measures the population decay dynamics of the pump-induced
excitations. The pump-probe signal can be complicated in samples with more complex
decay dynamics, for example, systems with multiple decay channels or intermediate
states. The temporal behavior can deviate from a single exponential decay function and
the differential transmission can even be negative. In the frequency-resolved version
of pump-probe spectroscopy, known as spectrally resolved transient absorption, the
transmitted probe is analyzed with a spectrometer to measure the wavelength depen-
dence of the change in probe pulse. Spectral resolution can give more insight into the
origin of the signal itself as well as helping to identify the reason for non-exponential
dynamics.
A common coherent spectroscopic technique is transient four-wave mixing (TFWM),
which can be performed with two or three pulses. The basic geometry for a two-pulse
TFWM experiment is sketched in Fig. 1.4(a). Two pulses with wavevectors k1 and k2
are incident on the sample. The delay between the two pulses is τ and it is defined
to be positive when pulse k1 arrives first. The nonlinear interaction gives rise to a
TFWM signal in the direction ks = 2k2 − k1 . A three-pulse TFWM experiment can
be configured in different geometries. A planar geometry is shown in Fig. 1.4(b). The
generated TFWM signal can be emitted in the direction ks = −k1 + k2 + k3 . In the
two-pulse and planar three-pulse geometries, it might be difficult to align the detector
Ultrafast nonlinear/coherent spectroscopy 13
Fig. 1.4 Different geometries for TFWM experiments. (a) Two-pulse TFWM. (b) Three-
-pulse TFWM. (c) Three-pulse TFWM in the phase conjugate geometry. (d) Three-pulse
TFWM in the box geometry.
to the signal direction when the TFWM signal is too weak to be visible. Neither of
these geometries are fully phase-matched, i.e., the induced polarization has a different
wavevector than the emitted signal. For thin samples the accumulated phase mis-
match is small for small angles, and thus the lack of phase matching is not significant.
However, for thick samples or large angles using a phase-matched geometry can be
important.
A three-pulse TFWM experiment can be arranged in special geometries that are
fully phase-matched and assist the alignment. One approach is the phase conjugate
geometry, as shown in Fig. 1.4(c), where pulses k2 and k3 counter propagate, i.e.,
k2 = −k3 , so the emission direction is ks = −k1 + k2 + k3 = −k1 , i.e., the signal
counter-propagates with respect to k1 . Another approach is the box geometry, as
shown in Fig. 1.4(d). In this geometry, three parallel beams with wavevectors k1 , k2 ,
and k3 are aligned to three corners of a square and focused by a lens to converge on the
sample. The transmitted excitation beams point to three corresponding corners of the
box on the other side of the sample. The TFWM signal direction ks = −k1 + k2 + k3
points to the fourth corner. For convenience, a fourth beam aligned to the fourth corner
of the input box can be used as a reference beam to place the detector. The reference
beam can be blocked during measurements. The box geometry is also commonly used
in optical MDCS experiments. It is worth pointing out that the TFWM signal can
emit in multiple phase-matched directions in addition to the direction ks denoted in
each geometry. However, the signals in other phase-matched directions contain the
same information. We note that TFWM experiments can also be performed in the
collinear geometry in which the excitation pulses co-propagate in one beam. In this
case, different linear/nonlinear signals are not spatially separated and other methods
such as frequency tagging are needed to isolate specific TFWM signals. The collinear
geometry will be discussed in more detail in Chapter 4.
Based on how the signal is detected, TFWM experiment has three main vari-
ants: time-integrated (TI-TFWM), time-resolved (TR-TFWM), and spectrally re-
solved (SR-TFWM) TFWM. In TI-TFWM, a slow detector is used to integrate the
signal over time. In TR-TFWM, the signal is measured with a sufficient time resolu-
tion to be resolved in time, using a reference pulse and up-conversion in a nonlinear
crystal. The signal pulse is mapped out as a function of “real” time as the reference
pulse is scanned across the signal in time. In SR-TFWM, a spectrometer is used to
record the spectrum of the signal.
For a two-level system, the three-pulse TFWM experiment can be physically de-
scribed as follows: The first pulse creates a coherence (also called a polarization) that
oscillates at the absorption frequency. The second pulse converts the coherence to a
population in either the excited or ground state, depending on the relative phase be-
tween the two pulses. For two crossing pulses, the relative phase changes with position
to produce a population that is spatially modulated with a wavevector k2 −k1 , forming
a population grating, which in turn results in a grating in the sample absorption, anal-
ogous to the change in absorption for a pump-probe experiment. The third pulse then
diffracts off the grating to generate the TFWM signal. Two-pulse TFWM is similar
since the second pulse is considered as two pulses with a zero delay.
In two-pulse TFWM, the TFWM signal decays as a function of time delay τ be-
tween the pulses and the dynamics are determined by the decoherence rate in a sample
of homogeneously broadened two-level systems [449]. In the approximation that the
excitation pulses are infinitely short (i.e., a delta function in time) and within the
Markovian approximation, the intensity of the TI-TFWM signal is a function of τ
Is (τ ) ∝ Θ(τ )e−2γτ , (1.25)
where γ is the dephasing rate (otherwise called the homogeneous linewidth) and Θ(τ )
is the Heaviside step function; however, in an inhomogeneously broadened system, the
TFWM signal is a photon echo formed at a delay of τ after the second pulse. In the
limit of strong inhomogeneous broadening, the TI-TFWM signal is
Is (τ ) ∝ Θ(τ )e−4γτ . (1.26)
Therefore, the TFWM technique can measure γ even in the presence of inhomoge-
neous broadening; however, to properly interpret the results, one must know whether
or not the sample is inhomogeneously broadened. In some circumstances, comparing
to the linear absorption linewidth can reveal whether Eq. (1.25) or (1.26) should be
used, since the linear absorption spectrum is a convolution of the homogeneous and
inhomogeneous lineshapes (a Voigt profile), which converges to the inhomogeneous
lineshape, and hence a width determined by the inhomogeneous broadening, in the
limit of strong inhomogeneous broadening. However, it can be ambiguous when the
homogeneous and inhomogeneous linewidths are comparable. Additionally, the homo-
geneous linewidth can depend on the excitation intensity, making the comparison to an
The density matrix 15
absorption linewidth more difficult. The TFWM signal can be temporally or spectrally
resolved to help remove the ambiguity.
In three-pulse TFWM, the delay between the second and third pulse, T , can be
varied. This gives the advantage to measure both the dephasing rate γ by scanning
τ and the relaxation rate of the grating Γgr by scanning T [436]. For a homogeneous
two-level system, the three-pulse TI-TFWM signal is
Is (τ, T ) = Θ(τ )Θ(T )e−2γτ e−Γgr T . (1.27)
The factor of 2 multiplying γ becomes a 4 if the sample is homogeneously broad-

ened. However, homogeneously and inhomogeneously broadened systems can be dis-
tinguished in three-pulse TFWM by measuring the signal around τ = 0 for a fixed
T where T > 0. The system is homogeneously broadened if the signal is symmetric
about zero, whereas the system is inhomogeneously broadened if there is only signal
for τ > 0. In the presence of inhomogeneous broadening, three-pulse TFWM is also
sensitive to spectral diffusion, a process by which an excitation initially at one energy
can shift to another energy. Clear signatures of spectral diffusion can be observed by
varying both T and τ . Three-pulse TFWM can also be used to study non-Markovian
behaviors in a system where the Markovian approximation is not valid. Three pulse
echo peak shift spectroscopy [85, 172, 436] can be used to extract the correlation
function of the frequency fluctuations that give rise to dephasing.
The decay of the grating occurs at the rate Γgr is determined by the excited-state
population decay rate Γpop and spatial diffusion, which washes out the population
grating and leads to an exponential decay of the signal as a function of T . The grating
relaxation rate is
Γgr = 2Γpop + 8π 2 DΛ−2 , (1.28)
where D is the spatial diffusion coefficient and Λ = nλ/2 sin θ is the grating spacing
for an angle θ between the beams k1 and k2 . If there is no spatial diffusion or very slow
diffusion compared to the population decay, γ gives a direct measurement of Γpop . In
the presence of spatial diffusion, the effects of spatial diffusion and population decay
can be separated by varying the angle between the first two pulses. Both Γpop and D
can be determined by fitting the angle dependence of Γgr with Eq. (1.28).
1.4 The density matrix

Interpretation of laser spectroscopy relies on the theory of light-matter interaction.
The semi-classical approach is usually used for laser spectroscopy in which matter
is treated quantum mechanically while light is classical. Within this approximation,
rudimentary scenarios can be treated using a wavefunction formalism in which the
atoms are assumed to be coherent states at all times. Very quickly, however, it becomes
helpful to expand this into a density matrix formalism capable of simulating incoherent
states in addition to coherent ones. Here we illustrate the salient features of these two
different formalisms in turn, using a two-level atom interacting with a monochromatic
field as an illustrative model system.
We begin with the wavefunction formalism. As shown in Fig. 1.5, we consider a
two-level atom with the ground state |0⟩ and the excited state |1⟩. An electromagnetic
|1〉
E(t)
|0〉
Fig. 1.5 Energy level diagram of a two-level atom interacting with a monochromatic field.
field with the angular frequency ω interacts with the atom. The electric field is E(t) =
|Ê(t)| cos(ωc t). The Hamiltonian of the system is H = H0 + V , where H0 is the
unperturbed part,
H0 = ℏω0 |0⟩⟨0| + ℏω1 |1⟩⟨1|, (1.29)
here ω0 and ω1 are atomic frequencies such that ℏω0 and ℏω1 are the energies of states
|0⟩ and |1⟩, respectively.
Under the dipole approximation, the atom is considered to be small compared
to the wavelength of light, thus only its dipole moment needs to be considered. The
interaction between the atom and the field is described by the interaction Hamiltonian

V = − µ10 |1⟩⟨0| + µ01 |0⟩⟨1| |Ê(t)| cos(ωc t)
= −ℏΩr cos(ωc t)|1⟩⟨0| − ℏΩr cos(ωc t)|0⟩⟨1|, (1.30)
where µ10 = µ∗01 = e⟨1|x|0⟩ is the matrix element of the electric dipole moment, and
Ωr = µ10 |Ê(t)|/ℏ is the Rabi frequency.
The system dynamics are governed by the Schrödinger equation
∂
iℏ |ψ⟩ = H|ψ⟩. (1.31)
∂t
The state of the system can be expressed in the basis of the eigenstates of H0 as
|ψ(t)⟩ = C0 (t)|0⟩ + C1 (t)|1⟩. (1.32)
Substituting Eqs. 1.29, 1.30, and 1.32 into Eq. 1.31, we obtain the equations of motion
Ċ0 = −iω0 C0 + iΩr cos(ωc t)C1 , (1.33)

Ċ1 = −iω1 C1 + iΩr cos(ωc t)C0 . (1.34)
Defining C0 = c0 e−iω0 t and C1 = c1 e−iω1 t , the equations of motion can be rewritten
Ωr i(ωc −ω10 )t
ċ0 = i (e + e−i(ωc +ω10 )t )c1 , (1.35)
2
Ωr
ċ1 = i (e−i(ωc −ω10 )t + ei(ωc +ω10 )t )c0 , (1.36)
2
where ω10 = ω1 − ω0 . Applying the rotating-wave approximation in which the counter-

rotating terms including e±i(ωc +ω10 )t are ignored yields
Ωr i(ωc −ω10 )t
ċ0 = i e c1 , (1.37)
2
Ωr
ċ1 = i e−i(ωc −ω10 )t c0 . (1.38)
2
The solutions to these equations describe the dynamics of the system. For simplicity,
we consider the resonant case (ωc = ω10 ) and assume that the atom is initially in the
ground state, i.e., c0 (t = 0) = 1 and c1 (t = 0) = 0. The solutions are

Ωr Ωr
c0 (t) = −i cos t and c1 (t) = sin t . (1.39)
2 2
The population inversion between the excited and ground states is W (t) = |c1 (t)|2 −
|c0 (t)|2 = − cos(Ωr t). The inversion oscillates between −1 and 1 at the Rabi frequency
Ωr , meaning that the population oscillates between the excited and ground states as
a result of the field-atom interaction. Note that this solution has ignored any kind
of dissipation or damping and thus assumes that the coherent superposition states
induced in the atom do not experience any decoherence, either with respect to the
driving field or within members of an ensemble. The inclusion of decoherence causes
the amplitude of oscillations to decay with time after the field is turned on.
If the excitation field is a pulse instead of a CW field, the resulting inversion can
be either 1 or -1 depending on the pulse area. The pulse area theorem states that the
final state of the system does not depend on the shape of the pulse, but only its area.
For pulses that are short compared to the system’s relaxation time, the area of a pulse
Ωr (t) is
Z +∞
A= Ωr (t)dt. (1.40)
−∞
A pulse with A = π, known as π-pulse, drives the system into the excited state and
results in an inversion of 1. A 2π-pulse with A = 2π drives the system into the excited
state and then back to the ground state, leaving an inversion of -1.
We may ask ourselves at this point where the wavefunction formalism breaks down.
The answer lies in the fact that state vectors such as |ψ⟩ are ultimately incapable of
capturing the physics of many practical situations involving dissipation or coupling to
a thermal reservoir. Beyond this, we may want to consider an ensemble, and thus must
take into account the possibility of finding different members of the ensemble existing
within in a statistical mixture of different wavefunctions (this is a fundamentally dif-
ferent object from a quantum superposition), where the probability of being in state
|ψ⟩ is Pψ . To capture these effects, it is necessary to expand our formalism into the
language of density operators.
The density operator is defined as
X
ρ= Pψ |ψ⟩⟨ψ|. (1.41)
ψ
The state is called a pure state if the density operator can be written as ρ = |ψ0 ⟩⟨ψ0 |,
that is the probability Pψ is nonzero only for state |ψ0 ⟩. With the density operator,
the expectation value of an operator O can be calculated as
X
⟨O⟩ = Pψ ⟨ψ|O|ψ⟩ = Tr(Oρ). (1.42)
ψ
The density operator can be written in a matrix form for a given basis |n⟩, in which
case the matrix elements are ρnm = ⟨n|ρ|m⟩ and the product Oρ indicates matrix
multiplication.
To write the Schrödinger equation, Eq. (1.31), in terms of the density operator, we
take the derivative of Eq. 1.41
X
ρ̇ = Pψ (|ψ̇⟩⟨ψ| + |ψ⟩⟨ψ̇|). (1.43)
ψ
Substituting Eq. (1.31) into Eq. (1.43) leads to the Liouville-von Neumann equation,
i
ρ̇ = − [H, ρ], (1.44)
ℏ
which determines the density matrix time evolution. The density matrix theoretical
framework allows a greater breadth of phenomena to be captured than can be described
using wave functions and the Schrödinger equation alone. For example, relaxation
processes can be considered by including phenomenological decay terms in the equation
of motion
i 1
ρ̇ = − [H, ρ] − {Γ, ρ}, (1.45)
ℏ 2
where {Γ, ρ} = Γρ + ρΓ and Γ is the relaxation operator.
As an example of how the formalism associated with density matrices be concretely
put into practice, we consider a two-level system in the state
|ψ⟩ = C0 |0⟩ + C1 |1⟩. (1.46)
The density operator can be written as
ρ = |ψ⟩⟨ψ| = |C0 |2 |0⟩⟨0| + |C1 |2 |1⟩⟨1| + C0 C1∗ |0⟩⟨1| + C1 C0∗ |1⟩⟨0|. (1.47)
The matrix elements are
ρ00 = |C0 |2 , ρ11 = |C1 |2 , ρ10 = C1 C0∗ , ρ01 = ρ∗10 = C0 C1∗ . (1.48)
In the matrix form, we can write the state vector as

C0
|ψ⟩ = . (1.49)
C1
The density matrix is correspondingly
|C0 |2 C0 C1∗

C0 ∗ ∗
ρ00 ρ01
ρ= C0 C1 = = . (1.50)
C1 C1 C0∗ |C1 |2 ρ10 ρ11
We can see that ρ00 and ρ11 are the probabilities of being in states |0⟩ and |1⟩, re-
spectively, while ρ10 and ρ01 are related to the coherence or polarization. In general,
in a density matrix the diagonal matrix elements ρnn represent the probabilities being
in state |n⟩ and the off-diagonal matrix elements ρnm (n ̸= m) represent a coherent
superposition between |n⟩ and |m⟩.
In matrix form, the Hamiltonian given by Eqs. (1.29) and (1.30) for the interaction
between light and a two-level system can be written
−ℏΩ∗r cos(ωc t)

ℏω0 0 0
H= + (1.51)
0 ℏω1 −ℏΩr cos(ωc t) 0
| {z } | {z }
H0 V
−ℏΩ∗r cos(ωc t)

0
= . (1.52)
−ℏΩr cos(ωc t) ℏω1
Substituting Eq. (1.52) into Eq. (1.45), we obtain equations of motion for the individual
density matrix elements,
i
ρ̇00 = −γ00 ρ00 − µ01 E(t)(ρ01 − ρ10 ), (1.53)
ℏ
i
ρ̇11 = −γ11 ρ11 + µ10 E(t)(ρ01 − ρ10 ), (1.54)
ℏ
i
ρ̇10 = −γ10 ρ10 − iω10 ρ10 − µ10 E(t)(ρ11 − ρ00 ), (1.55)
ℏ
ph
where ω10 = ω1 − ω0 , and the relaxation matrix elements are γij = 21 (Γi + Γj ) + γij
ph ph
with γij being the pure dephasing rate and γij = 0 for i = j.
There are several approaches to solving Eqs. (1.53)–(1.55). In cases where the
electromagnetic field interactions are weak, the problem can be tackled using methods
of perturbation theory. To illustrate, we show the first-order solution to the off-diagonal
density matrix element ρ10 (t) as described by Eq. (1.55). Introducing the parameter λ
to keep track of bookkeeping (we can set λ = 1 when all is said and done to reproduce
physical reality) and then replacing V by λV in the Hamiltonian matrix (1.51), we
look to express ρij (t) in terms of a power series expansion in λ of the form
(0) (1) (2)
ρij = ρij + λρij + λ2 ρij + . . . (1.56)
Substituting Eq. (1.56) back into Eqs. (1.53)–(1.55) leads to a set of equations
(0) (0) (0) (eq)
ρ̇ij = −iωij ρij − γij (ρij − ρij ), (1.57)
(n) (n) i
ρ̇ij = −i(ωij + γij )ρij − [V, ρ(n−1) ]ij for n ∈ Z+ , (1.58)
ℏ
that can be iteratively solved if the initially unperturbed states of the system are
known. Assuming that the steady-state value of ρ is the same as its unperturbed value
(0) (eq)
such that ρij = ρij , and that the system relaxes entirely into the ground state
(0)
under equilibrium conditions at time t = ∞, we have ρ00 = 1 and all other elements
(0)
of ρij = 0. This being the case, we can substitute the zeroth-order values of ρ into the
first-order version of Eq. (1.58) and integrate. Writing E(t) as two phasor components,
1h i
E(t) = Ê(t)e−iωc t + Ê ∗ (t)e+iωc t , (1.59)
2
and defining
ρ10 (t) = ρ̂10 (t)e−iω10 t , (1.60)
we obtain
t
−iN µ10 e−γ10 t
Z h
(1) ′
ρ̂10 (t) = Ê(t′ )e[γ10 −i(ωc −ω10 )]t
2ℏ −∞
′
i
+Ê ∗ (t′ )e[γ10 +i(ωc +ω10 )]t dt′ . (1.61)
For a CW field turning on at time t = 0, we can set Ê(t) = Ê ∗ (t) = E0 Θ(t) and
then analytically solve. The integral gives
" ′ ′ ′ ′
#t
(1) −iN µ10 E0 e−γ10 t eγ10 t −i(ωc −ω10 )t eγ10 t +i(ωc +ω10 )t
ρ̂10 (t) = + . (1.62)
2ℏ γ10 − i(ωc − ω10 ) γ10 + i(ωc + ω10 ) ′
t =0
In the optical regime the resonance widths are generally narrow in comparison to the
resonance frequencies, such that γ10 ≪ ω10 . At excitation frequencies near resonance
then, the denominator of the second term is much larger than that of the first term,
and the second term can be ignored. This is known as the rotating-wave approximation
(RWA). In the case of a lower frequency regime such as infrared or microwave, the
rotating-wave approximation begins to fail, and the second term causes a shift of the
apparent resonance, known as Bloch-Siegert shift.
In the opposite limit of an infinitesimally short optical pulse arriving at time t = 0,
we can set Ê(t) = Ê ∗ (t) = E0 δ(t). The solution to Eq. (1.61) becomes
(1) −iN µ10 E0 e−γ10 t

ρ̂10 (t) = [Θ(t) + Θ(t)] , (1.63)
2ℏ
which decays exponentially with time after the initial perturbation at time t = 0. In the
optical regime, the second of these two terms is typically suppressed down to unobserv-
able values by the finite bandwidth of the laser pulses (an effect that delta-function
pulse envelopes are unable to capture), and so the rotating-wave approximation is
typically invoked by discarding this second term by hand.
1.5 Bloch sphere representation of quantum states

The quantum state of a two-level system and the state’s evolution can be conveniently
discussed by visualizing the state with a Bloch vector on the Bloch sphere. The Bloch
sphere provides a geometrical visualization of any pure state in a two-level system.
The north and south poles each correspond to one of the orthogonal basis states
of a two-level system. For the two-level system shown in Fig. 1.5, the Bloch vector
Bloch sphere representation of quantum states 21
|1〉
|1〉
Fig. 1.6 Bloch spheres with (a) the Bloch vector representing state |0⟩, (b) the Bloch vector
representing state |1⟩, (c) the Bloch vector representing state √12 (|0⟩ + |1⟩), and (d) the Bloch
vector representing an state |ψ⟩.
pointing to the south pole corresponds to state |1⟩, as shown in Fig. 1.6(a). The Bloch
vector pointing to the north pole, as shown in Fig. 1.6(b), corresponds to state |1⟩. A
vector pointing to any other point on the surface of the sphere represents a coherent
superposition of states |0⟩ and |1⟩. Particularly, a point on the equator corresponds to
state √12 (|0⟩ + eiφ |1⟩), with φ being the azimuthal angle. In these states, the system
has the maximum coherence or polarization. When φ = 0, as shown in Fig. 1.6(c),
the state is √12 (|0⟩ + |1⟩), representing an in-phase polarization. In general, a point
defined by the polar angle ϑ and the the azimuthal angle φ, as shown in Fig. 1.6(d),
corresponds to state
ϑ ϑ
|ψ⟩ = sin |0⟩ + eiφ cos |1⟩. (1.64)
2 2
As the quantum state evolves, the Bloch vector rotates and leaves a trajectory on
the surface of the sphere. For example, if the system is initially in state |0⟩ (south
pole), a π2 -pulse drives the system from the south pole to the equator, i.e., from |0⟩ to
√1 (|0⟩ + |1⟩), creating a coherence/polarization. A π-pulse can excite the system from
2
the south pole to the north pole, i.e., from |0⟩ to |1⟩, or vice versa. A 2π-pulse can
drive the system from the south (north) pole to the north (south) pole and then back
to the south (north) pole. A Bloch vector can also precess about z-axis, corresponding
to the phase evolution of a coherence/polarization.
2
Introduction to multidimensional
coherent spectroscopy
Multidimensional coherent spectroscopy was first developed in nuclear magnetic res-

onance (NMR) spectroscopy [108], where it was known as “multidimensional Fourier
transform spectroscopy.” Key enabling technological advances were the development
of digital computers with sufficient computing power and the development of the fast
Fourier transform (FFT) algorithm [67]. This pair of advances meant that data could
be acquired in the time domain and the spectrum calculated in a reasonable amount
of time.
NMR relies on generating and measuring radio frequency (RF) waveforms with
full phase information, which is routine. The wavelengths of the signals are typically
on the order of a meter, thus the stability of the time delays does not need to be
stringently controlled. In contrast, implementing similar methods at optical frequencies
is challenging. In optics, detectors typically measure the intensity of the light, not its
underlying electric field waveform. Thus, heterodyne/homodyne methods are needed
to retrieve phase information. In addition, the orders-of-magnitude shorter wavelength
means that fluctuations in path lengths must be tightly controlled to avoid corruption
of the phase information.
The idea of implementing 2D coherent spectroscopy in the optical regime was first
proposed by Tanimura and Mukamel in 1993 [389]. The technical challenges hindered
experimental implementation initially and it took over six years to demonstrate the
first realizations. Since then, many different approaches of optical 2D coherent spect-
roscopy have been developed. The technique is now well established in many ultrafast
spectroscopic laboratories and even became commercially available in recent years.
The initial proposal by Tanimura and Mukamel [389] uses a Raman excitation
scheme in which molecular vibrations are excited by a five-pulse sequence. The experi-
ments were initially complicated by lower-order signals that emit in the same direction
as the desired nonlinear signal, making it difficult to separate the two [32]. This diffi-
cult led to using IR to access molecular vibrations directly for 2D spectroscopy, which
was demonstrated independently by the groups of Hochstrasser in 1999 [146] and Tok-
makoff in 2001 [136]. In the IR regime, the requirement of the timing precision and
stability can be satisfied with only passive stability due to the longer wavelength.
Implementing 2D coherent spectroscopy to access electronic transitions, often called
“two-dimensional electronic spectroscopy (2DES),” requires visible or near-IR light
and thus poses greater challenges with respect to path length stability due to the
shorter wavelengths. The first demonstration of 2DCS in these wavelength regimes
24 Introduction to multidimensional coherent spectroscopy
relied on measuring the delay for each time-step and resampling the data [161]. To
eliminate the need for resampling, methods based on diffractive optics were demon-
strated by Miller [68] and Fleming [45] in 2004. Subsequently, many implementations
extended the technique into the near-IR and visible regime, realizing 2D spectroscopy
for electronic transitions in atoms, molecules, or semiconductors. These experiments
used more elaborate approaches to control the timing at shorter wavelengths. Subse-
quently approaches based on active stabilization of the path lengths were demonstrated
[34, 461].
Multidimensional coherent spectra can be difficult to understand and interpret
when they are first encountered. This chapter will provide an introduction into the
basic concepts of multidimensional coherent spectroscopy and provide a primer for
interpreting the spectra. A common tool for analyzing the spectra, double-sided Feyn-
man diagrams, will also be introduced.
2.1 Concepts of multidimensional coherent spectroscopy

A multidimensional coherent spectrum is a map of coherent light-matter interactions
plotted across a domain of two or more frequencies. Such a spectrum is obtained
by illuminating a material with a series of electromagnetic pulses and analyzing the
frequency-dependent way in which excitation of the sample by the initial or interme-
diate pulses affects the material’s response to the subsequent pulses.
Perhaps the simplest way to understand the technique is as an extension of transient
absorption spectroscopy, as illustrated by Fig. 2.1. In a transient absorption experiment
[Fig. 2.1(a)], a sample is illuminated by two pulses. The first of these (the “pump”)
drives the sample into a nonequilibrium state, which is then measured by subtracting
the sample’s response to the second pulse (the “probe”) in the pump’s presence from
the response to the probe in the pump’s absence. The resulting differential signal can
be measured as a spectrally and temporally integrated quantity using a photodetector
or, by sending the transmitted probe into a spectrometer, as a spectrally resolved
quantity plotted against the probe frequency [Fig. 2.1(b)].
In order for a signal to be observed, the optical response of the sample must be
nonlinear, i.e., it depends on the intensity of light. If the response is not nonlinear,
then the presence of the pump pulse will not affect the interaction of the sample with
the probe pulse. The most straightforward of these is saturation, in which the pump
pulse decreases the sample’s net absorption of the probe pulse, thereby increasing the
transmission of the probe and generating a positive signal in differential transmission.
For interacting many-body systems, there are also other possibilities. For example,
the pump can broaden the linewidth of the sample’s response to the probe pulse
(excitation-induced dephasing, or EID), which tends to produce positive central peaks
with negative wings. It can also shift the center frequency of the resonance (excitation-
induced shift, or EIS), creating an asymmetric pump-probe spectrum. The effects of
both EIS and EID are visible in Fig. 2.1(b), which corresponds to an asymmetric
In0.05 Ga0.95 As double quantum well where many-body effects are known to dominate
the optical response.
Although it improves upon the information available through linear absorption,
transient absorption is still incomplete because spectral dependence of the sample’s
Concepts of multidimensional coherent spectroscopy 25
(a) ωprobe (b)

Pump
transmission
Differential
Probe
0
tprobe T
ωprobe
Delay T
(c) (d) +
ωt
Signal strength
ωτ
t T τ -
Delay ωt T
Fig. 2.1 Comparison between transient absorption spectroscopy and MDCS. (a) In transient
absorption spectroscopy, the differential transmission of an optical pulse (the “probe”) is
measured in response to the application of an earlier pulse (the “pump”). The pulses are
delayed with respect to each other by a time T . (b) For a given value of T , the response may
be plotted as a spectrally resolved function of the probe frequency ωprobe . (c) In a simple
implementation of MDCS, the pump pulse from (a) is divided into sub-pulses, defining an
additional time delay τ . d) Data may be acquired as a function of varied τ and/or T , and
Fourier transformed to produce spectra plotted against a multidimensional frequency domain.
Adapted from Ref. 363.
response to the pump pulse remains unknown even as its differential probe-pulse re-
sponse is spectrally resolved. In part because of this limitation, results are often ob-
scured in the presence of inhomogeneity resulting, for example, from Doppler broaden-
ing in atomic vapors, sample imperfections, size variations in nanostructures mapping
into variations in confinement potentials, or uneven strain. To remedy the situation,
the pump may be divided into a pair of sub-pulses [Fig. 2.1(c)]. The inter-pulse delay
between these, (τ ), can then be systematically varied, and the data can be numer-
ically Fourier transformed to obtain a spectrally resolved excitation axis. The delay
between the second and third pulses (T ) tracks the pump-probe delay. The third pulse
continues to act as a probe, leading to a differential response that is either emitted
from the sample over an elapsed time (t) as coherent four-wave mixing, or converted
by a fourth pulse (not shown) after time t into a signal in photoluminescence, pho-
tocurrent, or photoemission. The resulting MDCS plot is a time-resolved signal that is
also frequency-resolved as a function of both excitation (pump) and emission (probe)
frequency [Fig. 2.1(d)]. As presented throughout this book, unfolding the spectral
26 Introduction to multidimensional coherent spectroscopy
response across these multiple frequency dimensions facilitates a wealth of new ca-
pabilities, including the ability to disentangle microscopic dephasing in the midst of
sample inhomogeneity, the ability to identify coherent coupling mechanisms between
resonances, and the ability to more thoroughly constrain the mechanisms behind many-
body interactions.
2.1.1 Coherent spectroscopy

One might wonder, based on the schematic depicted in Fig. 2.1(c), how coherent
information related to the sample’s response to the first pulse can possibly be extracted
from the final measurement. After all, the sample’s response to the probe pulse is
ultimately the only signal measured. The issue is resolved by noting that if time
delays between pulses are controlled with sub-wavelength precision, then coherent
interactions from the earlier pulses will be written into the phase of measured signal.
A simple illustration of the effect is shown in Fig. 2.2, depicting a Bloch sphere
illustration [8] of a two-pulse correlation measurement in which π/2 pulses are applied
to a two-state system. As illustrated by Fig. 2.2(a), the system is initially in the ground
(a) Pulse 1 Pulse 2
or or
τ
(b)
Pulse: 1 2
1/2
τ)
y(
ela
ed
u ls
r-p
e
Int
Population
response
0
Polarization
Time response
Fig. 2.2 Coherent spectroscopy of a two-level system in the π/2-pulse limit. (a) Bloch
sphere illustration of the quantum state’s evolution. The system is initialized in the ground
state (left), driven into a superposition between ground and excited states by a π/2 pulse
(center), and converted into a potential variety of population states and superposition states
by a second π/2 pulse (right). (b) Varying the time delay between the two pulses maps the
intermediate quantum state’s coherent evolution (horizontal red oscillations, center) onto the
final state’s excited-state population (vertical blue projections and red oscillations, right).
Adapted from Ref. 363.
Another random document with
no related content on Scribd:
Le roi Louis XVIII était remonté sur le trône; tout espoir de rester
en France était donc perdu pour Murat; il fallait partir. Son neveu
Bonafoux fréta un brick pour les Etats-Unis, sous le nom du prince
Rocca Romana. Toute la suite se rendit à bord, et l’on commença d’y
faire transporter les objets précieux que le proscrit avait pu sauver
dans le naufrage de sa royauté. D’abord ce fut un sac d’or pesant
cent livres à peu près, une garde d’épée sur laquelle étaient les
portraits du roi, de la reine et de ses enfans, et les actes de l’état
civil de sa famille, reliés en velours et ornés de ses armes. Quant à
Murat, il avait gardé sur lui une ceinture dans laquelle était, entre
quelques papiers précieux, une vingtaine de diamans démontés qu’il
estimait lui-même à une valeur de quatre millions.
Tous ces préparatifs de départ arrêtés, il fut convenu que le
lendemain, 1er août, à cinq heures du matin, la barque du brick
viendrait chercher le roi dans une petite baie distante de dix minutes
de chemin de la maison de campagne qu’il habitait. Le roi passa la
nuit à tracer à monsieur Marouin un itinéraire à l’aide duquel il devait
arriver jusqu’à la reine, qui alors était, je crois, en Autriche. Au
moment de partir il fut terminé, et en quittant le seuil de cette maison
hospitalière, où il avait trouvé un refuge, il le remit à son hôte avec
un volume de Voltaire que son édition stéréotype rendait portatif. Au
bas du conte de Micromégas, le roi avait écrit:[2]
«Tranquillise-toi, ma chère Caroline; quoique bien

malheureux, je suis libre. Je pars sans savoir où je vais; mais
partout où j’irai mon cœur sera à toi et à mes enfans.
»J. M.»
Dix minutes après, Murat et son hôte attendaient sur la plage de

Bonette l’arrivée du canot qui devait conduire le fugitif à son
bâtiment.
Ils attendirent ainsi jusqu’à midi, et rien ne parut; et cependant ils
voyaient à l’horizon le brick sauveur qui, ne pouvant tenir l’ancre à
cause de la profondeur de la mer, courait des bordées, au risque,
par cette manœuvre, de donner l’éveil aux sentinelles de la côte. A
midi, le roi, écrasé de fatigue, brûlé par le soleil, était couché sur la
plage, lorsqu’un domestique arriva portant quelques
rafraîchissemens que madame Marouin, inquiète, envoyait à tout
hasard à son mari. Le roi prit un verre d’eau rougie, mangea une
orange, se releva un instant pour regarder si, dans l’immensité de
cette mer, il ne verrait pas venir à lui la barque qu’il attendait. La mer
était déserte, et le brick seul se courbait gracieusement à l’horizon,
impatient de partir comme un cheval qui attend son maître.
Le roi poussa un soupir et se recoucha sur le sable. Le
domestique retourna à Bonette avec l’ordre d’envoyer à la plage le
frère de monsieur Marouin. Un quart d’heure après il arrivait, et
presque aussitôt il repartait à grande course de cheval pour Toulon,
afin de savoir de monsieur Bonafoux la cause qui avait empêché la
barque de venir prendre le roi. En arrivant chez le capitaine, il trouva
la maison envahie par la force armée; on faisait une visite
domiciliaire dont Murat était l’objet. Le messager parvint enfin au
milieu du tumulte jusqu’à celui auprès duquel il était envoyé, et là il
apprit que le canot était parti à l’heure convenue, et qu’il fallait qu’il
se fût égaré dans les calangues de Saint-Louis et de Sainte-
Marguerite. C’est en effet ce qui était arrivé. A cinq heures, monsieur
Marouin rapportait ces nouvelles à son frère et au roi. Elles étaient
embarrassantes. Le roi n’avait plus le courage de défendre sa vie,
même par la fuite; il était dans un de ces momens d’abattement qui
saisissent parfois l’homme le plus fort, incapable d’émettre une
opinion pour sa propre sûreté, et laissant monsieur Marouin maître
d’y pourvoir comme bon lui semblerait. En ce moment un pêcheur
rentrait en chantant dans le port. Marouin lui fit signe de venir, il
obéit.
Marouin commença par acheter à cet homme tout le poisson qu’il
avait pris; puis, après qu’il l’eut payé avec quelques pièces de
monnaie, il fit briller de l’or à ses yeux, et lui offrit trois louis s’il
voulait conduire un passager au brick que l’on apercevait en face de
la Croix-des-Signaux. Le pêcheur accepta. Cette chance de salut
rendit à l’instant même toutes ses forces à Murat; il se leva,
embrassa monsieur Marouin, lui recommanda d’aller trouver sa
femme et de lui remettre le volume de Voltaire, puis il s’élança dans
la barque, qui s’éloigna aussitôt.
Elle était déjà à quelque distance de la côte, lorsque le roi arrêta
le rameur et fit signe à Marouin qu’il avait oublié quelque chose. En
effet, sur la plage était un sac de nuit dans lequel Murat avait
renfermé une magnifique paire de pistolets montés en vermeil, qui
lui avait été donnée par la reine, et à laquelle il tenait
prodigieusement. A peine fut-il à la portée de la voix, qu’il indiqua à
son hôte le motif de son retour. Celui-ci prit aussitôt la valise, et,
sans attendre que Murat touchât terre, il la lui jeta de la plage dans
le bateau; en tombant, le sac de nuit s’ouvrit, et un des pistolets en
sortit. Le pêcheur ne jeta qu’un coup d’œil sur l’arme royale, mais ce
fut assez pour qu’il remarquât sa richesse et qu’il conçût des
soupçons. Il n’en continua pas moins de ramer vers le bâtiment.
Monsieur Marouin, le voyant s’éloigner, laissa son frère sur la côte,
et, saluant une dernière fois le roi, qui lui rendit son salut, retourna
vers la maison pour calmer les inquiétudes de sa femme et prendre
lui-même quelques heures de repos dont il avait grand besoin.
Deux heures après, il fut réveillé par une visite domiciliaire; sa
maison, à son tour, était envahie par la gendarmerie. On chercha de
tous les côtés sans trouver trace du roi. Au moment où les
recherches étaient le plus acharnées, son frère rentra; Marouin le
regarda en souriant, car il croyait le roi sauvé; mais à l’expression du
visage de l’arrivant, il vit qu’il était advenu quelque nouveau malheur.
Aussi, au premier moment de relâche que lui donnèrent les visiteurs,
il s’approcha de son frère:
—Eh bien! dit-il, le roi est à bord, j’espère?
—Le roi est à cinquante pas d’ici, caché dans la masure.
—Pourquoi est-il revenu?
—Le pêcheur a prétexté un gros temps, et a refusé de le
conduire jusqu’au brick.
—Le misérable!
Les gendarmes rentrèrent.
Toute la nuit se passa en visites infructueuses dans la maison et
ses dépendances; plusieurs fois ceux qui cherchaient le roi
passèrent à quelques pas de lui, et Murat put entendre leurs
menaces et leurs imprécations. Enfin, une demi-heure avant le jour,
ils se retirèrent. Marouin les laissa s’éloigner, et aussitôt qu’il les eut
perdus de vue, il courut à l’endroit où devait être le roi. Il le trouva
couché dans un enfoncement et tenant un pistolet de chaque main.
Le malheureux n’avait pu résister à la fatigue et s’était endormi. Il
hésita un instant à le rendre à cette vie errante et tourmentée; mais il
n’y avait pas une minute à perdre. Il le réveilla.
Aussitôt ils s’acheminèrent vers la côte; le brouillard matinal
s’étendait sur la mer. On ne pouvait distinguer à deux cents pas de
distance: ils furent obligés d’attendre. Enfin les premiers rayons du
soleil commencèrent à attirer à eux cette vapeur nocturne; elle se
déchira, glissant sur la mer, pareille aux nuages qui glissent au ciel.
L’œil avide du roi plongeait dans chacune des vallées humides qui
se creusaient devant lui, sans y rien distinguer; cependant il espérait
toujours que derrière ce rideau mobile il finirait par apercevoir le
brick sauveur. Peu à peu l’horizon s’éclaircit; de légères vapeurs,
semblables à des fumées, coururent encore quelque temps à la
surface de la mer, et dans chacune d’elles le roi croyait reconnaître
les voiles blanches de son vaisseau. Enfin la dernière s’effaça
lentement, la mer se révéla dans toute son immensité; elle était
déserte. Le brick, n’osant attendre plus long-temps, était parti
pendant la nuit.
—Allons, dit le roi en se retournant vers son hôte, le sort en est
jeté, j’irai en Corse.
Le même jour, le maréchal Brune était assassiné à Avignon.
II.
LA CORSE.
C’est encore sur cette même plage de Bonette, dans cette même
baie où nous l’avons vu attendre inutilement le canot de son brick,
que toujours accompagné de son hôte fidèle, nous allons retrouver
Murat le 22 août de la même année. Ce n’était plus alors par
Napoléon qu’il était menacé, c’est par Louis XVIII qu’il était proscrit:
ce n’était plus la loyauté militaire de Brune qui venait, les larmes aux
yeux, lui signifier les ordres qu’il avait reçus, c’était l’ingratitude
haineuse de monsieur de Rivière, qui mettait à prix[3] la tête de celui
qui avait sauvé la sienne[4]. Monsieur de Rivière avait bien écrit à
l’ex-roi de Naples de s’abandonner à la bonne foi et à l’humanité du
roi de France, mais cette vague invitation n’avait point paru au
proscrit une garantie suffisante, surtout de la part d’un homme qui
venait de laisser égorger, presque sous ses yeux, un maréchal de
France porteur d’un sauf-conduit signé de sa main. Murat savait le
massacre des Mameluks à Marseille, l’assassinat de Brune à
Avignon; il avait été prévenu la veille par le commissaire de police de
Toulon[5] que l’ordre formel avait été donné de l’arrêter: il n’y avait
donc pas moyen de rester plus longtemps en France. La Corse,
avec ses villes hospitalières, ses montagnes amies et ses forêts
impénétrables, était à cinquante lieues à peine; il fallait gagner la
Corse, et attendre dans ses villes, dans ses montagnes ou dans ses
forêts, ce que les rois décideraient relativement au sort de celui
qu’ils avaient appelé sept ans leur frère.
A dix heures du soir, le roi descendit sur la plage. Le bateau qui
devait l’emporter n’était pas encore au rendez-vous; mais, cette fois,
il n’y avait aucune crainte qu’il y manquât; la baie avait été reconnue,
pendant la journée, par trois amis dévoués à la fortune adverse:
c’étaient messieurs Blancard, Langlade et Donadieu, tous trois
officiers de marine, hommes de tête et de cœur, qui s’étaient
engagés sur leur vie à conduire Murat en Corse, et qui en effet
allaient exposer leur vie pour accomplir leur promesse. Murat vit
donc sans inquiétude la plage déserte: ce retard, au contraire, lui
donnait quelques instans de joie filiale. Sur ce bout de terrain, sur
cette langue de sable, le malheureux proscrit se cramponnait encore
à la France, sa mère, tandis qu’une fois le pied posé sur ce bâtiment
qui allait l’emporter, la séparation devait être longue, sinon éternelle.
Au milieu de ces pensées, il tressaillit tout-à-coup et poussa un
soupir: il venait d’apercevoir, dans l’obscurité transparente de la nuit
méridionale, une voile glissant sur les vagues comme un fantôme.
Bientôt un chant de marin se fit entendre; Murat reconnut le signal
convenu, il y répondit en brûlant l’amorce d’un pistolet, et aussitôt la
barque se dirigea vers la terre; mais, comme elle tirait trois pieds
d’eau, elle fut forcée de s’arrêter à dix ou douze pas de la plage;
deux hommes se jetèrent aussitôt à la mer, et gagnèrent le bord, le
troisième resta enveloppé dans son manteau et couché près du
gouvernail.
—Eh bien! mes braves amis, dit le roi en allant au-devant de
Blancard et de Langlade jusqu’à ce qu’il sentît la vague mouiller ses
pieds, le moment est arrivé, n’est-ce pas? Le vent est bon, la mer
calme; il faut partir.
—Oui, répondit Langlade, oui, sire, il faut partir, et peut-être
cependant serait-il plus sage de remettre la chose à demain.
—Pourquoi? reprit Murat.
Langlade ne répondit point; mais, se tournant vers le couchant, il
leva la main, et, selon l’habitude des marins, il siffla pour appeler le
vent.
—C’est inutile, dit Donadieu, qui était resté dans la barque, voici
les premières bouffées qui arrivent, bientôt tu en auras à n’en savoir
que faire... Prends garde, Langlade, prends garde, parfois en
appelant le vent on éveille la tempête.—Murat tressaillit, car il
semblait que cet avis, qui s’élevait de la mer, lui était donné par
l’esprit des eaux; mais l’impression fut courte, et il se remit à
l’instant.
—Tant mieux, dit-il, plus nous aurons de vent, plus vite nous
marcherons.
—Oui, répondit Langlade, seulement Dieu sait où il nous
conduira, s’il continue à tourner ainsi.
—Ne partez pas cette nuit, sire, dit Blancard, joignant son avis à
celui de ses deux compagnons.
—Mais enfin, pourquoi cela?
—Parce que, vous voyez cette ligne noire, n’est-ce pas? eh bien!
au coucher du soleil elle était à peine visible, la voilà maintenant qui
couvre une partie de l’horizon; dans une heure il n’y aura plus une
étoile au ciel.
—Avez-vous peur? dit Murat.
—Peur! répondit Langlade, et de quoi? de l’orage? il haussa les
épaules. C’est à-peu-près comme si je demandais à votre majesté si
elle a peur d’un boulet de canon... Ce que nous en disons, c’est pour
vous, sire; mais que voulez-vous que fasse l’orage à des chiens de
mer comme nous?
—Partons donc! s’écria Murat en poussant un soupir. Adieu,
Marouin... Dieu seul peut vous récompenser de ce que vous avez
fait pour moi. Je suis à vos ordres, messieurs.
A ces mots, les deux marins saisirent le roi chacun par une
cuisse, et l’élevant sur leurs épaules, ils entrèrent aussitôt dans la
mer; en un instant il fut à bord, Langlade et Blancard montèrent
derrière lui, Donadieu resta au gouvernail; les deux autres officiers
se chargèrent de la manœuvre et commencèrent leur service en
déployant les voiles. Aussitôt, comme un cheval qui sent l’éperon, la
petite barque sembla s’animer; les marins jetèrent un coup d’œil
insoucieux vers la terre, et Murat, sentant qu’il s’éloignait, se
retourna du côté de son hôte et lui cria une dernière fois:
—Vous avez votre itinéraire jusqu’à Trieste. N’oubliez pas ma
femme!... Adieu!... Adieu.
—Dieu vous garde, sire, murmura Marouin.—Et quelque temps
encore, grâce à la voile blanche qui se dessinait dans l’ombre, il put
suivre des yeux la barque qui s’éloignait rapidement; enfin elle
disparut. Marouin resta encore quelque temps sur le rivage, quoiqu’il
ne vît plus rien; alors un cri affaibli par la distance parvint encore
jusqu’à lui: ce cri était le dernier adieu de Murat à la France.
Lorsque monsieur Marouin me raconta un soir, au lieu même où
la chose s’était passée, les détails que je viens de décrire, ils lui
étaient si présens, quoique vingt ans se fussent écoulés depuis lors,
qu’il se rappelait jusqu’aux moindres accidens de cet embarquement
nocturne. De ce moment, il m’assura qu’un pressentiment de
malheur l’avait saisi, qu’il ne pouvait s’arracher de cette plage, et
que plusieurs fois l’envie lui prit de rappeler le roi; mais, pareil à un
homme qui rêve, sa bouche s’ouvrait sans laisser échapper aucun
son. Il craignait de paraître insensé; et ce ne fut qu’à une heure du
matin, c’est-à-dire deux heures et demie après le départ de la
barque, qu’il rentra chez lui avec une tristesse mortelle dans le cœur.
Quant aux aventureux navigateurs, ils s’étaient engagés dans
cette large ornière marine qui mène de Toulon à Bastia, et d’abord
l’événement parut, aux yeux du roi, démentir la prédiction de nos
marins: le vent, au lieu de s’augmenter, tomba peu à peu, et deux
heures après le départ, la barque se balançait sans reculer ni
avancer sur des vagues qui, de minute en minute, allaient
s’aplanissant. Murat regardait tristement s’éteindre, sur cette mer où
il se croyait enchaîné, le sillon phosphorescent que le petit bâtiment
traînait après lui: il avait amassé du courage contre la tempête, mais
non contre le calme; et, sans même interrompre ses compagnons de
voyage, à l’inquiétude desquels il se méprenait, il se coucha au fond
du bateau, s’enveloppa de son manteau, et fermant les yeux comme
s’il dormait, il s’abandonna au flot de ses pensées, bien autrement
tumultueux et agité que celui de la mer. Bientôt les deux marins,
croyant à son sommeil, se réunirent au pilote, et, s’asseyant près du
gouvernail, commencèrent à tenir conseil.
—Vous avez eu tort, Langlade, dit Donadieu, de prendre une
barque ou si petite ou si grande: sans pont nous ne pouvons résister
à la tempête, et sans rames nous ne pouvons avancer dans le
calme.
—Sur Dieu! je n’avais pas le choix. J’ai été obligé de prendre ce
que j’ai rencontré, et si ce n’était pas l’époque des madragues[6], je
n’aurais pas même trouvé cette mauvaise péniche, ou bien il me
l’aurait fallu aller chercher dans le port, et la surveillance est telle
que j’y serais bien entré, mais que je n’aurais probablement pas pu
en sortir.
—Est-elle solide au moins? dit Blancard.
—Pardieu! tu sais bien ce que c’est que des planches et des
clous qui trempent depuis dix ans dans l’eau salée. Dans les
occasions ordinaires on n’en voudrait pas pour aller de Marseille au
château d’If; dans une circonstance comme la nôtre on ferait le tour
du monde dans une coquille de noix.
—Chut! dit Donadieu. Les marins écoutèrent: un grondement
lointain se fit entendre, mais si faible, qu’il fallait l’oreille exercée d’un
enfant de la mer pour le distinguer.
—Oui, oui, dit Langlade; c’est un avertissement pour ceux qui ont
des jambes ou des ailes de regagner le nid qu’ils n’auraient pas dû
quitter.
—Sommes-nous loin des îles? dit vivement Donadieu.
—A une lieue environ.
—Mettez le cap sur elles.
—Et pour quoi faire? dit Murat en se soulevant.
—Pour y relâcher, sire, si nous le pouvons...
—Non, non! s’écria Murat, je ne veux plus remettre le pied à terre
qu’en Corse; je ne veux pas quitter encore une fois la France.
D’ailleurs, la mer est calme, et voilà le vent qui nous revient...
—Tout à bas! cria Donadieu.
Aussitôt Langlade et Blancard se précipitèrent pour exécuter la
manœuvre. La voile glissa le long du mât, et s’abattit au fond du
bâtiment.
—Que faites-vous? cria Murat; oubliez-vous que je suis roi et que
j’ordonne?
—Sire, dit Donadieu, il y a un roi plus puissant que vous ici, c’est
Dieu; il y a une voix qui couvre la vôtre, c’est celle de la tempête...
Laissez-nous sauver votre majesté, si la chose est possible, et
n’exigez rien de plus...
En ce moment un éclair sillonna l’horizon, un coup de tonnerre,
plus rapproché que le premier, se fit entendre, une légère écume
monta à la surface de l’eau, la barque frissonna comme un être
animé. Murat commença à comprendre que le danger venait; alors il
se leva en souriant, jeta derrière lui son chapeau, secoua ses longs
cheveux, aspira l’orage comme il aspirait la fumée; le soldat était
prêt à combattre.
—Sire, dit Donadieu, vous avez bien vu des batailles; mais peut-
être n’avez-vous point vu une tempête: si vous êtes curieux de ce
spectacle, cramponnez-vous au mât et regardez, car en voilà une
qui se présente bien.
—Que faut-il que je fasse? dit Murat; ne puis-je vous aider en
rien?
—Non! pas pour le moment, sire; plus tard nous vous
emploierons aux pompes...
Pendant ce dialogue, l’orage avait fait des progrès; il arrivait sur
les voyageurs comme un cheval de course, soufflant le vent et le feu
par ses naseaux, hennissant le tonnerre et faisant voler l’écume des
vagues sous ses pieds. Donadieu pressa le gouvernail, la barque
céda comme si elle comprenait la nécessité d’une prompte
obéissance, et présenta sa poupe au choc du vent; alors la
bourrasque passa laissant derrière elle la mer tremblante, et tout
parut rentrer dans le repos. La tempête reprenait haleine.
—En sommes-nous donc quittes pour cette rafale? dit Murat.
—Non, votre majesté, dit Donadieu, ceci n’est qu’une affaire
d’avant-garde; tout-à-l’heure le corps d’armée va donner.
—Et ne faisons-nous pas quelques préparatifs pour le recevoir?
répondit gaîment le roi.
—Lesquels? dit Donadieu. Nous n’avons plus un pouce de toile
où le vent puisse mordre, et tant que la barque ne fera pas eau nous
flotterons comme un bouchon de liége. Tenez-vous bien, sire!...
En effet, une seconde bourrasque accourait, plus rapide que la
première, accompagnée de pluie et d’éclairs. Donadieu essaya de
répéter la même manœuvre, mais il ne put virer si rapidement que le
vent n’enveloppât la barque; le mât se courba comme un roseau; le
canot embarqua une vague.
—Aux pompes, cria Donadieu! Sire, voilà le moment de nous
aider...
Blancard, Langlade et Murat saisirent leurs chapeaux et se
mirent à vider la barque. La position de ces quatre hommes était
affreuse, elle dura trois heures. Au point du jour le vent faiblit;
cependant la mer resta grosse et tourmentée. Le besoin de manger
commença à se faire sentir; toutes les provisions avaient été
atteintes par l’eau de mer, le vin seul avait été préservé du contact.
Le roi prit une bouteille, en avala le premier quelques gorgées; puis il
la passa à ses compagnons, qui burent à leur tour: la nécessité avait
chassé l’étiquette. Langlade avait par hasard sur lui quelques
tablettes de chocolat, qu’il offrit au roi. Murat en fit quatre parts
égales et força ses compagnons de manger; puis, le repas fini, on
orienta vers la Corse; mais la barque avait tellement souffert qu’il n’y
avait pas probabilité qu’elle pût gagner Bastia.
Le jour se passa tout entier sans que les voyageurs pussent faire
plus de dix lieues; ils naviguaient sous la petite voile; de foque,
n’osant tendre la grande voile, et le vent était si variable, que le
temps se perdait à combattre ses caprices. Le soir une voie d’eau se
déclara; elle pénétrait à travers les planches disjointes; les
mouchoirs réunis de l’équipage suffirent pour tamponner la barque,
et la nuit, qui descendit triste et sombre, les enveloppa pour la
seconde fois de son obscurité. Murat écrasé de fatigue, s’endormit;
Blancard et Langlade reprirent place près de Donadieu; et ces trois
hommes, qui semblaient insensibles au sommeil et à la fatigue,
veillèrent à la tranquillité de son sommeil.
La nuit fut, en apparence, assez tranquille; cependant
quelquefois des craquemens sourds se faisaient entendre. Alors les
trois marins se regardaient avec une expression étrange; puis leurs
yeux se reportaient vers le roi, qui dormait au fond de ce bâtiment,
dans son manteau trempé d’eau de mer, aussi profondément qu’il
avait dormi dans les sables de l’Egypte et dans les neiges de la
Russie. Alors l’un d’eux se levait, s’en allait à l’autre bout du canot
en sifflant entre ses dents l’air d’une chanson provençale... puis,
après avoir consulté le ciel, les vagues et la barque, il revenait
auprès de ses camarades, et se rasseyait en murmurant:—C’est
impossible; à moins d’un miracle, nous n’arriverons jamais.—La nuit
s’écoula dans ces alternatives. Au point du jour on se trouva en vue
d’un bâtiment:—Une voile! s’écria Donadieu, une voile! A ce cri le roi
se réveilla. En effet, un petit brick marchand apparaissait, venant de
Corse et faisant route vers Toulon. Donadieu mit le cap sur lui,
Blancard hissa les voiles au point de fatiguer la barque, et Langlade
courut à la proue, élevant le manteau du roi au bout d’une espèce de
harpon. Bientôt les voyageurs s’aperçurent qu’ils avaient été vus; le
brick manœuvra de manière à se rapprocher d’eux; au bout de dix
minutes ils se trouvèrent à cinquante pas l’un de l’autre. Le capitaine
parut sur l’avant. Alors le roi le héla, lui offrant une forte récompense
s’il voulait le recevoir à bord avec ses trois compagnons et les
conduire en Corse. Le capitaine écouta la proposition; puis aussitôt,
se tournant vers l’équipage, il donna à demi-voix un ordre que
Donadieu ne put entendre, mais qu’il saisit probablement par le
geste, car aussitôt il commanda à Langlade et à Blancard une
manœuvre qui avait pour but de s’éloigner du bâtiment. Ceux-ci
obéirent avec la promptitude passive des marins; mais le roi frappa
du pied:
—Que faites-vous, Donadieu? que faites-vous? s’écria-t-il; ne
voyez-vous pas qu’il vient à nous?
—Oui, sur mon âme! je le vois... Obéissez, Langlade; alerte,
Blancard. Oui, il vient sur nous, et peut-être m’en suis-je aperçu trop
tard. C’est bien, c’est bien; à moi maintenant. Alors il se coucha sur
le gouvernail, et lui imprima un mouvement si subit et si violent, que
la barque, forcée de changer immédiatement de direction, sembla se
raidir contre lui, comme ferait un cheval contre le frein; enfin elle
obéit. Une vague énorme, soulevée par le géant qui venait sur elle,
l’emporta avec elle comme une feuille; le brick passa à quelques
pieds de sa poupe.
—Ah! traître! s’écria le roi, qui commença seulement à
s’apercevoir de l’intention du capitaine; en même temps il tira un
pistolet de sa ceinture, en criant: A l’abordage, à l’abordage! et
essaya de faire feu sur le brick; mais la poudre était mouillée et ne
s’enflamma point. Le roi était furieux, et ne cessait de crier: A
l’abordage, à l’abordage!
—Oui, oui, le misérable, ou plutôt l’imbécile, dit Donadieu, il nous
a pris pour des forbans, et il a voulu nous couler, comme si nous
avions besoin de lui pour cela.
En effet, jetant les yeux sur le canot, il était facile de s’apercevoir
qu’il commençait à faire eau. La tentative de salut que venait de
risquer Donadieu avait effroyablement fatigué la barque, et la mer
entrait par plusieurs écartemens de planches; il fallut se mettre à
puiser l’eau avec les chapeaux; ce travail dura dix heures. Enfin
Donadieu fit, pour la seconde fois, entendre le cri sauveur:—Une
voile! une voile!...
Le roi et ses deux compagnons cessèrent aussitôt leur travail; on
hissa de nouveau les voiles, on mit le cap sur le bâtiment qui
s’avançait et l’on cessa de s’occuper de l’eau, qui, n’étant plus
combattue, gagna rapidement.
Désormais c’était une question de temps, de minutes, de
secondes, voilà tout; il s’agissait d’arriver au bâtiment avant de
couler bas. Le bâtiment, de son côté, semblait comprendre la
position désespérée de ceux qui imploraient son secours, il venait au
pas de course; Langlade le reconnut le premier, c’était une
balancelle du gouvernement, un bateau de poste qui faisait le
service entre Toulon et Bastia. Langlade était l’ami du capitaine, il
l’appela par son nom avec cette voix puissante de l’agonie, et il fut
entendu. Il était temps, l’eau gagnait toujours; le roi et ses
compagnons étaient déjà dans la mer jusqu’aux genoux; le canot
gémissait comme un mourant qui râle; il n’avançait plus et
commençait à tourner sur lui-même. En ce moment, deux ou trois
câbles, jetés de la balancelle, tombèrent dans la barque; le roi en
saisit un, s’élança et saisit l’échelle de corde: il était sauvé. Blancard
et Langlade en firent autant presque aussitôt; Donadieu resta le
dernier, comme c’était son devoir de le faire, et au moment où il
mettait un pied sur l’échelle du bord, il sentit sous l’autre s’enfoncer
la barque qu’il quittait; il se retourna avec la tranquillité d’un marin,
vit le gouffre ouvrir sa vaste gueule au-dessous de lui, et aussitôt la
barque dévorée tournoya et disparut. Cinq secondes encore, et ces
quatre hommes, qui maintenant étaient sauvés, étaient à tout jamais
perdus!...[7]
Murat était à peine sur le pont, qu’un homme vint se jeter à ses
pieds; c’était un mameluk qu’il avait autrefois ramené d’Egypte, et
qui s’était depuis marié à Castellamare; des affaires de commerce
l’avaient attiré à Marseille, où, par miracle, il avait échappé au
massacre de ses frères; et, malgré le déguisement qui le couvrait et
les fatigues qu’il venait d’essuyer, il avait reconnu son ancien maître.
Ses exclamations de joie ne permirent pas au roi de garder plus
longtemps son incognito; alors le sénateur Casablanca, le capitaine
Oletta, un neveu du prince Baciocchi, un ordonnateur nommé
Boërco, qui fuyaient eux-mêmes les massacres du Midi, se trouvant
sur le bâtiment, le saluèrent du nom de majesté et lui improvisèrent
une petite cour: le passage était brusque, il opéra un changement
rapide; ce n’était plus Murat le proscrit, c’était Joachim Ier, roi de
Naples. La terre de l’exil disparut avec la barque engloutie; à sa
place, Naples et son golfe magnifique apparurent à l’horizon comme
un merveilleux mirage, et sans doute la première idée de la fatale
expédition de Calabre prit naissance pendant ces jours d’enivrement
qui suivirent les heures d’agonie. Cependant le roi, ignorant encore
quel accueil l’attendait en Corse, prit le nom de comte de Campo
Melle, et ce fut sous ce nom que le 25 août il prit terre à Bastia. Mais
sa précaution fut inutile; trois jours après son arrivée, personne
n’ignorait plus sa présence dans cette ville. Des rassemblemens se
formèrent aussitôt, des cris de: Vive Joachim! se firent entendre, et
le roi, craignant de troubler la tranquillité publique, sortit le même
soir de Bastia avec ses trois compagnons et son mameluk. Deux
heures après il entrait à Viscovato, et frappait à la porte du général
Franceschetti, qui avait été à son service tout le temps de son règne,
et qui, ayant quitté Naples en même temps que le roi, était revenu
en Corse habiter avec sa femme la maison de monsieur Colona
Cicaldi, son beau-père. Il était en train de souper lorsqu’on vint lui
dire qu’un étranger demandait à lui parler: il sortit et trouva Murat
enveloppé d’une capote militaire, la tête enfoncée dans un bonnet
de marin, la barbe longue, et portant un pantalon, des guêtres et des
souliers de soldat. Le général s’arrêta étonné; Murat fixa sur lui son
grand œil noir; puis, croisant les bras:—Franceschetti, lui dit-il, avez-
vous à votre table une place pour votre général qui a faim? avez-
vous sous votre toit un asile pour votre roi qui est proscrit?...
Franceschetti jeta un cri de surprise en reconnaissant Joachim, et ne
put lui répondre qu’en tombant à ses pieds et en lui baisant la main.
De ce moment, la maison du général fut à la disposition de Murat.
A peine le bruit de l’arrivée du roi fut-il répandu dans les environs
que l’on vit accourir à Viscovato des officiers de tous grades, des
vétérans qui avaient combattu sous lui, et des chasseurs corses que
son caractère aventureux séduisait; en peu de jours la maison du
général fut transformée en palais, le village en résidence royale, et
l’île en royaume. D’étranges bruits se répandirent sur les intentions
de Murat; une armée de neuf cents hommes contribuait à leur
donner quelque consistance. C’est alors que Blancard, Langlade et
Donadieu prirent congé de lui; Murat voulut les retenir; mais ils
s’étaient voués au salut du proscrit, et non à la fortune du roi.
Nous avons dit que Murat avait rencontré à bord du bateau de
poste de Bastia un de ses anciens mameluks nommé Othello, et que
celui-ci l’avait suivi à Viscovato: l’ex-roi de Naples songea à se faire
un agent de cet homme. Des relations de famille le rappelaient tout
naturellement à Castellamare; il lui ordonna d’y retourner, et le
chargea de lettres pour les personnes sur le dévoûment desquelles il
comptait le plus. Othello partit, arriva heureusement chez son beau-
père, et crut pouvoir lui tout dire; mais celui-ci, épouvanté, prévint la
police: une descente nocturne fut faite chez Othello et sa
correspondance saisie.
Le lendemain, toutes les personnes auxquelles étaient adressées
des lettres furent arrêtées et reçurent l’ordre de répondre à Murat
comme si elles étaient libres, et de lui indiquer Salerne comme le
lieu le plus propre au débarquement: cinq sur sept eurent la lâcheté
d’obéir, les deux autres, qui étaient deux frères espagnols, s’y
refusèrent absolument: on les jeta dans un cachot.
Cependant, le 17 septembre, Murat quitta Viscovato, le général
Franceschetti, ainsi que plusieurs officiers corses, lui servirent
d’escorte; il s’achemina vers Ajaccio par Cotone, les montagnes de
Serra et Bosco, Venaco, Vivaro, les gorges de la forêt de Vezzanovo
et Bogognone; partout il fut reçu et fêté comme un roi, et à la porte
des villes il reçut plusieurs députations qui le haranguèrent en le
saluant du titre de majesté; enfin le 25 septembre il arriva à Ajaccio.
La population tout entière l’attendait hors des murs; son entrée dans
la ville fut un triomphe; il fut porté jusqu’à l’auberge qui avait été
désignée d’avance par les maréchaux-de-logis: il y avait de quoi
tourner la tête à un homme moins impressionnable que Murat: quant
à lui, il était dans l’ivresse; en entrant dans l’auberge, il tendit la main
à Franceschetti.—Voyez, lui dit-il, à la manière dont me reçoivent les
Corses, ce que feront pour moi les Napolitains.—C’était le premier
mot qui lui échappait sur ses projets à venir, et dès ce jour même il
ordonna de tout préparer pour son départ.
On rassembla dix petites felouques: un Maltais, nommé Barbara,
ancien capitaine de frégate de la marine napolitaine, fut nommé
commandant en chef de l’expédition; deux cent cinquante hommes
furent engagés et invités à se tenir prêts à partir au premier signal.
Murat n’attendait plus que les réponses aux lettres d’Othello; elles
arrivèrent dans la matinée du 28. Murat invita tous les officiers à un
grand dîner, et fit donner double paye et double ration à ses
hommes.
Le roi était au dessert lorsqu’on lui annonça l’arrivée de monsieur
Maceroni: c’était un envoyé des puissances étrangères qui apportait
à Murat la réponse qu’il avait attendue si longtemps à Toulon. Murat
se leva de table et passa dans une chambre à côté. Monsieur
Maceroni se fit reconnaître comme chargé d’une mission officielle, et
remit au roi l’ultimatum de l’empereur d’Autriche. Il était conçu en
ces termes:
«Monsieur Maceroni est autorisé par les présentes à

prévenir le roi Joachim que sa majesté l’empereur d’Autriche
lui accordera un asile dans ses États, sous les conditions
suivantes:
«1o Le roi prendra un nom privé. La reine ayant adopté
celui de Lipano, on propose au roi de prendre le même nom.
»2o Il sera permis au roi de choisir une ville de la Bohême,
de la Moravie, ou de la Haute-Autriche, pour y fixer son
séjour. Il pourra même, sans inconvénient, habiter une
campagne dans ces mêmes provinces.
»3o Le roi engagera sa parole d’honneur envers S. M. I. et
R. qu’il n’abandonnera jamais les Etats autrichiens sans le
consentement exprès de l’empereur, et qu’il vivra comme un
particulier de distinction, mais soumis aux lois qui sont en
vigueur dans les Etats autrichiens.
»En foi de quoi, et afin qu’il en soit fait un usage
convenable, le soussigné a reçu l’ordre de l’empereur de
signer la présente déclaration.
Donné à Paris le 1er septembre 1815.
Signé le prince de Metternich.»
Murat sourit en achevant cette lecture, puis il fit signe à monsieur

Maceroni de le suivre. Il le conduisit alors sur la terrasse de la
maison, qui dominait toute la ville et qui était dominée elle-même par
sa bannière qui flottait comme sur un château royal. De là on pouvait
voir Ajaccio toute joyeuse et illuminée, le port où se balançait la
petite flottille et les rues encombrées de monde, comme un jour de
fête. A peine la foule eut-elle aperçu Murat, qu’un cri partit de toutes
les bouches: Vive Joachim! vive le frère de Napoléon! vive le roi de
Naples! Murat salua, et les cris redoublèrent, et la musique de la
garnison fit entendre les airs nationaux. Monsieur Maceroni ne savait
s’il devait en croire ses yeux et ses oreilles. Lorsque le roi eut joui de
son étonnement, il l’invita à descendre au salon. Son état-major y
était réuni en grand uniforme: on se serait cru à Caserte ou à
Capodimonte. Enfin, après un instant d’hésitation, Maceroni se
rapprocha de Murat.
—Sire, lui dit-il, quelle réponse dois-je faire à sa majesté
l’empereur d’Autriche?
—Monsieur, lui répondit Murat avec cette dignité hautaine qui
allait si bien à sa belle figure, vous raconterez à mon frère François
ce que vous avez vu et ce que vous avez entendu; et puis vous
ajouterez que je pars cette nuit même pour reconquérir mon
royaume de Naples.
III.
LE PIZZO.
Les lettres qui avaient déterminé Murat à quitter la Corse lui

avaient été apportées par un Calabrais nommé Luidgi. Il s’était
présenté au roi comme un envoyé de l’Arabe Othello, qui avait été
jeté, comme nous l’avons dit, dans les prisons de Naples, ainsi que
les personnes auxquelles les dépêches dont il était porteur avaient
été adressées. Ces lettres, écrites par le ministre de la police de
Naples, indiquaient à Joachim le port de la ville de Salerne comme
le lieu le plus propre au débarquement; car le roi Ferdinand avait
rassemblé sur ce point trois mille hommes de troupes autrichiennes,
n’osant se fier aux soldats napolitains, qui avaient conservé de
Murat un riche et brillant souvenir. Ce fut donc vers le golfe de
Salerne que la flottille se dirigea; mais, arrivée en vue de l’île de
Caprée, elle fut assaillie par une violente tempête, qui la chassa
jusqu’à Paola, petit port situé à dix lieues de Cosenza. Les bâtimens
passèrent en conséquence la nuit du 5 au 6 octobre dans une
espèce d’échancrure du rivage qui ne mérite pas le nom de rade. Le
roi, pour ôter tout soupçon aux gardes des côtes et aux scorridori[8]
siciliens, ordonna d’éteindre les feux et de louvoyer jusqu’au jour;
mais, vers une heure du matin, il s’éleva de terre un vent si violent,
que l’expédition fut repoussée en haute mer, de sorte que le 6, à la
pointe du jour, le bâtiment que montait le roi se trouva seul. Dans la
matinée il rallia la felouque du capitaine Cicconi, et les deux navires
mouillèrent à quatre heures de l’après-midi en vue de Santo-Lucido.
Le soir, le roi ordonna au chef de bataillon Ottoviani de se rendre à
terre pour y prendre des renseignemens. Luidgi s’offrit pour
l’accompagner, Murat accepta ses bons offices. Ottoviani et son
guide se rendirent donc à terre, tandis qu’au contraire, Cicconi et sa
felouque se remettaient en mer avec mission d’aller à la recherche
du reste de la flotte.
Vers les onze heures de la nuit, le lieutenant de quart sur le
navire royal distingua au milieu des vagues un homme qui s’avançait
en nageant vers le bâtiment. Dès qu’il fut à la portée de la voix, il le
héla. Aussitôt le nageur se fit reconnaître: c’était Luidgi. On lui
envoya la chaloupe et il remonta à bord. Alors il raconta que le chef
de bataillon Ottoviani avait été arrêté, et qu’il n’avait échappé lui-
même à ceux qui le poursuivaient qu’en se jetant à la mer. Le
premier mouvement de Murat fut d’aller au secours d’Ottoviani; mais
Luidgi fit comprendre au roi le danger et l’inutilité de cette tentative;
néanmoins Joachim resta jusqu’à deux heures du matin agité et
irrésolu. Enfin, il donna l’ordre de reprendre le large. Pendant la
manœuvre qui eut lieu à cet effet, un matelot tomba à la mer et
disparut avant qu’on eût eu le temps de lui porter secours.
Décidément les présages étaient sinistres.
Le 7 au matin, on eut connaissance de deux bâtimens. Le roi
ordonna aussitôt de se mettre en mesure de défense; mais Barbara
les reconnut pour être la felouque de Cicconi et la balancelle de
Courrand, qui s’étaient réunies et faisaient voile de conserve. On
hissa les signaux, et les deux capitaines se rallièrent à l’amiral.
Pendant qu’on délibérait sur la route à suivre, un canot aborda le
bâtiment de Murat. Il était monté par le capitaine Pernice et un
lieutenant sous ses ordres. Ils venaient demander au roi la
permission de passer à son bord, ne voulant point rester à celui de
Courrand, qui, à leur avis, trahissait. Murat l’envoya chercher; et,
malgré ses protestations de dévoûment, il le fit descendre avec
cinquante hommes dans une chaloupe, et ordonna d’amarrer la
chaloupe à son bâtiment. L’ordre fut exécuté aussitôt, et la petite
escadre continua sa route, longeant, sans les perdre de vue, les
côtes de la Calabre; mais, à dix heures du soir, au moment où l’on
se trouvait à la hauteur du golfe de Sainte-Euphémie, le capitaine
Courrand coupa le câble qui le traînait à la remorque, et, faisant
force de rames, il s’éloigna de la flottille. Murat s’était jeté sur son lit
tout habillé: on le prévint de cet événement. Il s’élança aussitôt sur le
pont, et arriva à temps encore pour voir la chaloupe, qui fuyait dans
la direction de la Corse, s’enfoncer et disparaître dans l’ombre. Il
demeura immobile, sans colère et sans cris; seulement il poussa un

Optical Multidimensional Coherent Spectroscopy Hebin Li Full Chapter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Optical Multidimensional Coherent Spectroscopy Hebin Li Full Chapter

Uploaded by

Copyright:

Available Formats

Optical Multidimensional Coherent

Symbol Description Section

k wavevector of light, k = 2π/λ 1.1

∆ϕce carrier-envelope phase-slip 6.1

ϕ optical phase 1.2

MAPI Methylammonium Lead Iodide

1 Basics of ultrafast spectroscopy 1

5.4 Probing many-body correlations with multi-quantum 2D spect-

12.3 Multidimensional spectroscopy with quantum light 239

Multidimensional coherent spectroscopy has a number of advantages over other

1.1 Basics of spectroscopy: linear versus nonlinear

(a) Incident field Radiated field ∝ iP

If the frequency of the incident field is tuned so that it is on the resonance in

1.2 Ultrashort pulses

E(t) = Ê(t) cos (−ωc t + ϕ(t))

where n is refractive index, ϵ0 is the vacuum permittivity, c is the speed of light,

-60 -40 -20 0 20 40 60

converts from a time-domain signal, E(t), to its frequency-domain representation,

δω = 2π(δν) = 3.14 × 1013 rad/s, (1.16)

δU = h(δν) = 20.7 meV, (1.18)

1.3 Ultrafast nonlinear/coherent spectroscopy

as transient four-wave mixing and multidimensional coherent spectroscopy, also probes

∆α(τ ) = ∆α0 e−τ /τex . (1.22)

≈ Itran (0)(1 + ∆α0 e−τ /τex L), (1.23)

∆T (τ ) Itran (τ ) − Itran (0)

where T0 is the transmission of unexcited sample and ∆T (τ ) is the change in the

Is (τ ) ∝ Θ(τ )e−2γτ , (1.25)

Is (τ ) ∝ Θ(τ )e−4γτ . (1.26)

Is (τ, T ) = Θ(τ )Θ(T )e−2γτ e−Γgr T . (1.27)

The factor of 2 multiplying γ becomes a 4 if the sample is homogeneously broad-

1.4 The density matrix

= −ℏΩr cos(ωc t)|1⟩⟨0| − ℏΩr cos(ωc t)|0⟩⟨1|, (1.30)

|ψ(t)⟩ = C0 (t)|0⟩ + C1 (t)|1⟩. (1.32)

Ċ0 = −iω0 C0 + iΩr cos(ωc t)C1 , (1.33)

Defining C0 = c0 e−iω0 t and C1 = c1 e−iω1 t , the equations of motion can be rewritten

where ω10 = ω1 − ω0 . Applying the rotating-wave approximation in which the counter-

(1) −iN µ10 E0 e−γ10 t

1.5 Bloch sphere representation of quantum states

Multidimensional coherent spectroscopy was first developed in nuclear magnetic res-

2.1 Concepts of multidimensional coherent spectroscopy

(a) ωprobe (b)

2.1.1 Coherent spectroscopy

(a) Pulse 1 Pulse 2

«Tranquillise-toi, ma chère Caroline; quoique bien

Dix minutes après, Murat et son hôte attendaient sur la plage de

«Monsieur Maceroni est autorisé par les présentes à

Murat sourit en achevant cette lecture, puis il fit signe à monsieur

Les lettres qui avaient déterminé Murat à quitter la Corse lui

You might also like