Getting Started on Time-Resolved Molecular Spectroscopy Jeffrey A. Cina instant download
Getting Started on Time-Resolved Molecular Spectroscopy Jeffrey A. Cina instant download
or textbooks at https://ebookmass.com
https://ebookmass.com/product/getting-started-on-time-
resolved-molecular-spectroscopy-jeffrey-a-cina/
https://ebookmass.com/product/programming-arduino-getting-started-
with-sketches-tab-monk/
https://ebookmass.com/product/programming-arduino-getting-started-
with-sketches-third-edition-simon-monk/
https://ebookmass.com/product/programming-arduino-getting-started-
with-sketches-3rd-edition-simon-monk/
Programming the Photon: Getting Started with the Internet
of Things Rush Christopher.
https://ebookmass.com/product/programming-the-photon-getting-started-
with-the-internet-of-things-rush-christopher/
https://ebookmass.com/product/programming-the-raspberry-pi-third-
edition-getting-started-with-python-simon-monk/
https://ebookmass.com/product/getting-started-with-angular-create-and-
deploy-angular-applications-1st-edition-victor-hugo-garcia/
https://ebookmass.com/product/getting-started-with-sql-and-databases-
managing-and-manipulating-data-with-sql-mark-simon/
https://ebookmass.com/product/programming-the-intel-galileo-getting-
started-with-the-arduino-compatible-development-board-christopher-
rush/
GETTING STARTED ON TIME-RESOLVED MOLECULAR
SPECTROSCOPY
Getting Started on Time-Resolved
Molecular Spectroscopy
Jeffrey A. Cina
University of Oregon
3
3
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
© Jeffrey A. Cina 2022
The moral rights of the author have been asserted
Impression: 1
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: 2021945727
ISBN 978–0–19–959031–5
DOI: 10.1093/oso/9780199590315.001.0001
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.
For Barbara, of course.
Preface
Well, it has been quite an adventure so far. It was my good fortune to be drawn
into the theory of ultrafast spectroscopy early in the femtosecond era, due in large
part to the prior influence of Rick Heller’s wave-packet descriptions of continuous-
wave spectroscopies (thank you, Laurie!); Bob Silbey’s coaching during the 1980’s,
when multi-pulse optical-phase-controlled picosecond measurements were first under
consideration; and the pleasure and long-term benefit of collaboration with Norbert
Scherer, Stuart Rice, Graham Fleming, and other co-workers at Chicago. At that
stage, one had to find one’s own way, and for me that way started with ordinary
time-dependent perturbation theory as I’d learned it in Bob Harris’s (personally life-
changing) quantum mechanics classes at Berkeley, coupled with a desire to illuminate
the dynamics underlying optical measurements in terms of the evolving nuclear wave
packets that accompany each molecular electronic state.
It hadn’t occurred to me to write this book, or any other, but in 2011 I was in a bit
of a hiatus that came at the end of a multi-year group-reading project with co-workers
at Oregon on L&L’s Electrodynamics of Continuous Media. At Bob Mazo’s suggestion,
Sonke Adlung from Oxford University Press called sometime that year and asked if I
might like to write a book. It happened that I was running a fever at the time and
said, “Sure!”
So, here you have it, the result of the succeeding decade of puzzlement, formulation,
and reformulation—aided by the patience and helpful advice of the numerous collab-
orators and colleagues whose names are among those listed in my Acknowledgements,
hopefully without any inadvertent omissions.
There are valuable treatises already available on the principles of nonlinear optical
spectroscopy of molecular and material systems, notably those by Shaul Mukamel
and Minhaeng Cho.1,2,3,4 The coverage of those works goes beyond what is presented
here. So why write another? I believe that the ultrafast community could benefit
from a deliberately accessible, stepwise treatment (which doesn’t mean an easy one),
more of a textbook than a comprehensive exposition, which can serve as a bridge
between the graduate-level training in quantum mechanics that’s standard in Physical
Chemistry programs and the advanced formulations that serve as guidebooks for those
practicing in the field. The standard approach is to start from the equilibrium density
1 S. Mukamel, Principles of Nonlinear Optical Spectroscopy (Oxford University Press, New York,
1999).
2 M. Cho, Two-Dimensional Optical Spectroscopy (CRC Press, Boca Raton, 2009).
3 D. J. Tannor, Introduction to Quantum Mechanics. A Time-Dependent Perspective (University
Science Books, Sausalito, 2007).
4 J. Yuen-Zhou, J. J. Krich, I. Kassal, A. Johnson, and A. Aspuru-Guzik, Ultrafast Spectroscopy:
Quantum Information and Wavepackets, (IOP Publishing, Bristol, 2014).
viii Preface
matrix of the target molecule and express nonlinear optical signals as a convolution
of the appropriate nonlinear optical response functions with the electric field of the
incident laser pulses. In addition, informed by very early analyses of multi-wave mixing
in nonlinear optical crystals, conventional descriptions are often couched in terms of
optical wave propagation in extended media. Everything in the present text is (or
should be!) physically consistent with the widely applied existing theoretical analyses.
Applications of existing descriptions sometimes tend, in my view, to lose track of
the molecular-level dynamics underlying ultrafast signals, in part because those de-
scriptions compel, or at least encourage one to think about both bra- and ket-sides of
the density operator simultaneously. On the other hand, informed by the superposition
principle, framing things at least initially as a sum of terms in Hilbert space having
various orders in the external-field strengths makes it easier to think about the perti-
nent contributions to the molecular state one at a time. What is developed, analyzed,
and interpreted in terms of Hilbert-space wave functions can easily be converted to a
density-matrix description through a manipulation that takes just a couple of lines.
The wave-propagation picture of nonlinear optical response, while important for
experimental purposes, tends to obscure the fact that, even in an extended sample, each
molecule can often be regarded as undergoing absorption, fluorescence, and Raman
scattering all by itself, as well as acquiring the nonlinear induced dipole moments that
give rise to time-resolved signals. One goal of this text is to formulate as much as
possible in terms of nonlinear induced molecular dipoles—these are ultimately to be
expressed as quantum mechanical overlaps between pairs of multi-pulse wave packets—
and to derive any necessary macroscopic wave-propagation aspects, such as wave-
vector matching, from that microscopic starting point.
The strength of treatments based on nonlinear optical response functions is their
generality. Nonlinear optical signals are to be derived from that fundamental under-
pinning by convolving the relevant response functions with the actual form of the in-
cident laser pulses by multidimensional integration over time variables. It is the case,
though, that such an approach encodes much quantum dynamical information that is
not of immediate use in calculating or interpreting the actual laser-driven dynamics
under consideration. Those strategies sometimes tend, in practice, to invite simplified
comprehensive but phenomenological descriptions that do not attach directly to the
specific, perhaps non-generic form of the regions of the molecular potential energy
surfaces that govern the measured signal.
The treatment put forward here reverses the order of operation between quantum
mechanical averaging and integration over time. It carries out the latter first, with
the use of pulse propagators, operators analogous to those sometimes used in magnetic
resonance spectroscopy, which encapsulate the influence of a nonzero-duration laser
pulse within an instantaneously acting quantum mechanical operator. The reduced
pulse propagators that transfer nuclear wave packets between electronic potential en-
ergy surfaces under the influence of the external fields, reshaping them in the process,
capture the elements of coherent control that are inherent in short-pulse optical spec-
troscopy. This approach puts the focus on the motion of nuclear wave packets in the
optically accessed regions of the electronic potential energy surfaces on which they
evolve. It declines the frequent practice in work-ups based on nonlinear response func-
Preface ix
tions of idealizing the characteristics of the experimentally available light pulses and,
in effect, regarding those response functions themselves as constituting signals, rather
than signal transducers. As mentioned above, the working expressions arrived at here
present signal contributions as overlaps between well characterized individual multi-
pulse nuclear wave packets.
For ease of use and at the cost of some redundancy, I’ve tried to make each chapter
of this book usable and comprehensible on its own. There are some slight inconsisten-
cies of notation between chapters, so unless the reader is advised to do so, applying a
formula from one chapter to an equation in another should be done with caution.
In lieu of end-of-chapter problems, there are many boxed exercises embedded in
the text. By and large, these represent the type of derivation or physical analysis that
I work through to consolidate my own grasp of the ideas at hand. It’s my guess that
readers who excuse themselves from these exercises will be sacrificing something in
the depth of their understanding.
The final portion of most chapters consists of illustrative signal calculations. Car-
rying these out or interpreting their form is both our reward for slogging through the
underlying theory and a pale substitute for the experiential satisfaction of performing
actual measurements. The molecular models that are targeted in these calculations
are chosen to facilitate a thorough interpretation in terms of the underlying molecu-
lar dynamics. It is hoped that these illustrations will help inform what may be the
less complete interpretations that are possible with more complicated experimental
targets.
There is a smattering of references throughout the text but no comprehensive
bibliography. The works cited are just a few of those I found helpful or inspiring.
Many others could undoubtedly have served as well or better; the ones I mention can
perhaps serve as points of entry in the search for other relevant examples.
With help from the many colleagues who have read portions of this text, I’ve tried
hard to root out conceptual misdirection and outright errors. From experience, I know
that there’s no such thing as a small mistake in a scientific text. My sincere apology
for any that may remain.
Personal taste has a significant influence in science (gasp!), and this textbook
reflects my own. I hope, though, that the treatment given here will be to the liking of
at least some readers. I presume further to hope that, after working their way through
this book, those who persevere in doing so will be equipped to make the best use of
more sophisticated methodologies, gaining as much physical insight as possible and
avoiding some of the pitfalls to which applications of those approaches occasionally
give rise.
Jeff Cina
University of Oregon
2021
Acknowledgements
My heartfelt thanks to the following individuals and organizations: y’all know what for.
John Adamovics, Paul Alivisatos, Rise Ando, Ara Apkarian, Sonke Adlung, Natasha
Aristov, Alan Aspuru-Guzik, Bridgette Barry, Mark Berg, Berkeley Chemists for
Peace, Jason Biggs, Sandra Bigtree, Isabella Bischel, Eric Bittner, Harry Bonham,
Stephen Bradforth, Bill Braunlin, Paul Brumer, Irene Burghardt, Carlos Bustamante,
Laurie Butler, Carl Bybee, Jianshu Cao, Roger Carlson, Howard Carmichael, Sister
Carola FSPA, Craig Chapman, Xiaolu Cheng, Minhaeng Cho, Gerri Hoffman Cina,
Merrill T. Cina, Zoë Cina-Sklar, Rob Coalson, Devin Daniels, Joshua Daniels, Julia
Daniels, Rebecca Daniels, Peter Dardi, Brent Davidson, Jahan Dawlaty, Peter Dewey,
Jenni Dobbins, Mike and Carol Drake, Camille and Henry Dreyfus Foundation, Tom
Dyke, Ellen Epstein, my many friends at Espresso Roma, Elaine Finley Giannone,
Graham Fleming, Joan Florsheim, Ignacio Franco, David Frank, Karl Freed, Sarah
Fuchs, Rosemary Garrison, Ibrahim Gassama, Diane Gerth, Michelle Golden, Evi
Goldfield, Eoghan Gormley, John Simon Guggenheim Memorial Foundation, John
Hardwick, Alex Harris, Charles Harris, Christine Harris, Jerry Harris, Katherine Har-
ris, Robert A. Harris (my former PhD advisor, an exemplar of science, and a lifelong
friend), Ray Heath, Rick Heller, Eric Hiller, Maddie Holst, Travis Humble, Katharine
Hunt, Heide Ibrahim, Deb Jackson, Suggy Jang, Truus Jansen, David Jonas, Tara
Jones, Taiha Joo, Chanelle Jumper, Joe Kao, Mike Kellman, Alexis Kiessling, Dmitri
Kilin, Young-Kee Kim, Phil Kovac, Peter Kovach, Cindy Larson, Larry LeSueur, Sis-
ter Leora FSPA, Susan Levine-Friedman, Mark Limont, Katja Lindenberg, Giulia Lip-
parini, Andy Marcus, Alex Matro, Karin Matsumoto, Jack Maurer, Bob Mazo, David
McCamant, Alden Mead, Horia Metiu, Stan Micklavzina, Coleen Miller, Florabelle
Moses, Shaul Mukamel, Sid Nagel, US National Science Foundation, Keith Nelson,
Kenji Ohmori, Takeshi Oka, Janine O’Guinn, Colleen O’Leary, April Oleson, Mrs.
Ruth Patterson, Barbara Perry, Fred Perry, David Picconi, David Pratt, Jim Prell,
Mike Raymer, Tom Record, Stuart Rice, Philip Richardson, Mary Rohrdanz, Victor
Romero-Rochin, Sandy Rosenthal, Sister Rosilda FSPA, Penny Salus, Marsha Saxton,
Rich Saykally, Norbert Scherer, Greg Scholes, Moshe Shapiro, Mike Sheahan, Yu-Chen
Shen, Nancy Shows, Robert J. Silbey (my postdoctoral advisor, a gem of a human be-
ing), Michael Sipe, Barbara Sklar, Martin Smith, Tim Smith, Peter Straton, Joe
Subotnik, Susie Cina Sullivan, David Tannor, Judithe Thompson, Nacho Tinoco, Dr.
Brandy Todd, Emma Tran, Daniel Turner, Rob Tycko, David Tyler, Lowell Ungar, my
fellow members of Oregon’s faculty labor union United Academics, Marian Valentine,
Steven van Enk, Hailin Wang, John Waugh, Bob Weiss, Julia Widom, Kevin Wiles,
Cathy Wong, Claude Woods, Duane C. Wrenn and Leticia Steuer of Energetic Soul,
John Wright, Joel Yuen-Zhou, Larry Ziegler.
Contents
and
V (t) = −m̂ · E(t) . (1.2)
Hg and He are the nuclear Hamiltonians governing all relevant intra- and intermolec-
ular degrees of freedom in the ground and excited electronic states, respectively, and
is the “bare” electronic transition energy. The electric dipole-moment operator1 is
m̂ = m |eihg| + |gihe| , (1.3)
and the electric field of the pulse is assumed to take the form
x,y
~ct
~cσ
Fig. 1.1 Disk-shaped region of overlap between the propagating laser pulse and the induced
dipolar field from a molecular source at the origin. Electromagnetic interference between these
fields accounts for the light pulse’s energy change due to its interaction with the molecule.
As hinted by the physically suggestive form of eqn (1.13), there is a direct route
to this formula by way of a basic one for the rate of change of the molecular energy.
In general, one has
dE d
= hΨ(t)|H(t)|Ψ(t)i = h dΨ dΨ dH
dt |H(t)|Ψi + hΨ|H(t)| dt i + hΨ| dt |Ψi . (1.14)
dt dt
Using | dΨ 1
dt i = i~ H(t)|Ψi and H(t) = H + V (t), this becomes dE/dt = hΨ|dV /dt|Ψi;
the well-known statement that the time derivative of the expectation value of the
Hamiltonian equals the expectation value of the Hamiltonian’s time derivative. In the
present instance, we have
dE dE(t)
=− · hΨ(t)|m̂|Ψ(t)i ∼
= EΩf (t) sin(Ωt + ϕ)mx (t) , (1.15)
dt dt
which, together with energy conservation, returns eqn (1.13) upon integration, irre-
spective of the order in E to which the induced dipole is approximated.
1
Rt
where P (t; τ ) ≡ i~ −∞
dτ Ṽ (τ ).2
For an electronically resonant or near-resonant field, it is a good approximation
to neglect highly oscillatory terms in P ’s integrand (i.e., to make a rotating-wave
approximation) and write
and
| ↑↓i = −|giF 2 [t]gg p(ge) (t; τ )p(eg) (τ ; τ̄ )|ng i . (1.22)
The reduced pulse propagator p(eg) is evidently responsible for shaping the e-state
nuclear wave packet generated from |ng i in the e ← g transition, while the nested
combination p(ge) p(eg) sculpts the second-order wave packet in the g-state.
Upon substituting these forms in eqn (1.9), we find
the observation time has been set to infinity because the energy increment stops chang-
ing after the end of the laser pulse. Notice that the molecular energy change is indepen-
dent of the optical phase ϕ, as we should expect under the conditions considered here.
The first term in the last member of eqn (1.23) accounts for the energy gained by pop-
ulation transfer to the electronic excited state, while the second tracks the energetic
consequence of ground-state “bleaching” (i.e., the loss of g-state population).
2 The τ appearing after the semicolon in P (t; τ ) is not an argument of the pulse propagator,
but simply identifies its integration variable. This notation proves useful in writing nested pulse
propagators like those appearing in the second-order term of eqn (1.17).
The Heller formula 5
For the change in electromagnetic energy (1.13) on the other hand, we have
Z ∞
∆U = −EΩmx dτ f (τ ) sin(Ωτ + ϕ)
−∞
which relates the field’s energy loss to the center frequency of the laser pulse and the
squared norm of the nuclear wave packet it generates in the excited electronic state.
R∞
Use p(ge) (∞; τ )( + He ) − Hg p(ge) (∞; τ ) = −∞ dτ d
iΩτ
σ i~ dτ [−τ ]gg [τ ]ee e f (τ ) along
with integration by parts to verify once again, acknowledging approximations, that
∆E = −∆U, as required by energy conservation.
In the continuous-wave (cw) limit (σ longer than the inverse of the frequency spacing
between adjacent vibronic levels), eqn (1.27) becomes Heller’s celebrated expression
for the molecular electronic absorption spectrum as a Fourier transform of the time-
dependent overlap between a propagating e-state nuclear wave packet and the g-state
wave function from which it originates.3
Equation (1.27) can be formally evaluated for shorter pulses (lower spectral reso-
lution) by using e-state nuclear eigenkets obeying He |ne i = ne |ne i to obtain
2
X
∆E = 2π~ΩF 2 |hne |ng i|2 exp{− σ~2 ( + ne − ~Ω − ng )2 } . (1.28)
ne
1. Derive eqn (1.28) from eqn (1.27) by inserting the stated vibronic completeness
relation and carrying out the resulting Gaussian integration.
2. Repeat this derivation,
P Pstarting instead
R ∞ from the final form in eqn (1.25) and
i τ~ (+ne −~Ω−ng )
using p(eg) (∞, τ ) = ne ng |ne ihng | −∞ dτ
σ f (τ )e .
(sys)
Trsys [e−Hg /kB T ] is the partition function of the system and kB is the Boltzmann
constant. In this way, any of the optical signal expressions investigated here can be
readily translated into the corresponding finite-temperature Boltzmann-weighted form.
A handy shortcut is to regard the |ng i themselves as nuclear eigenstates of a
subsystem in weak contact with a temperature reservoir. Then the thermally weighted
sum of individual signal contributions hng |Ô|ng i reduces to Tr[Ôρ̂], with
ρ̂ = Z −1 ng |ng ie−ng /kB T hng |, or Tr[Ôρ̂] = Z −1 ng e−ng /kB T hng |Ô|ng i.
P P
mω 2 2
Vg (x) = x − αx3 + βx4 (1.30)
2
and Ve (x) = Vg (x − xe ). Here α = ~ω/32x3rms and β = ~ω/860x4rms , with xrms =
p
~/2mω. We set the wave number of the harmonic vibration to 200 cm−1 , which
corresponds to a period of τx = 2π/ω = 166.8 fs. The spatial shift of the e-state
potential is assigned the value xe = 2.626xrms , so the classical inner turning point at
an energy 2e lies at x = 0, directly above the g-state minimum. It is not necessary to
specify a value for the vibrational reduced mass m. Fig. 1.2 shows the two potential-
energy curves. The bare electronic transition energy in eqn (1.1) is given an arbitrary
wave number of 20, 000 cm−1 .
1200
1000
V(x) (cm-1 )
800
Vg(x) Ve(x)
600
400
200
0
0 5 10 15
x/xrms
Fig. 1.2 Potential curves governing nuclear motion in the g- and e-states of the one-dimen-
sional model Hamiltonian. Ve does not include the bare electronic transition energy.
For this simple system, it is a straightforward numerical exercise to solve for the
eigenvalues and eigenstates of both nuclear Hamiltonians and to obtain position repre-
sentations of the latter, using a harmonic-oscillator basis for instance. Matrix elements
of the reduced pulse propagators can also be easily determined. The absorption spec-
trum can then be evaluated as a function of the carrier frequency for any incident pulse
8 Short-pulse electronic absorption
duration, using either eqn (1.25) or eqn (1.28). The spectrum for σ = 5τx starting from
the initial state |g|0g , shown in Fig. 1.3, is of sufficient resolution clearly to discern
individual vibronic transitions. The absolute value of the corresponding Heller kernel
appearing in the integrand of eqn (1.27) is plotted in the same figure, and reflects the
complicated wave-packet dynamics that ensues upon the abrupt promotion of |0g to
the e-state potential. It exhibits quasi-harmonic motion for a couple vibrational pe-
riods, after which wave-packet spreading in Ve (x) born of anharmonicity gives rise to
less regular temporal evolution in the overlap. The kernel’s overall amplitude decays
on a timescale set by the pulse-duration parameter σ.
!%( !%(
!%' i !%'
Abs[h0g | e
!%& !%&
!%$ !%$
2
t2 /4
!%! !%!
e
Fig. 1.3 On the left is the absorption spectrum for the 1D model, calculated for the case of
narrow-bandwidth σ = 5τx pulses of continuously variable center frequency. The dimension-
less field-strength parameter F is chosen so that the highest (1e ← 0g ) peak has unit intensity.
The right panel shows the absolute value of the temporal overlap function, which provides a
wave-packet-dynamical description of the absorption process.
The absorption spectrum of the same system at temperature T = 298 K, with the
same pulse duration and field strength, is plotted as a solid line in Fig. 1.4. While the
spectral resolution is unchanged, the Boltzmann-weighted distribution of population
over initial levels with ng = 0, 1, . . . leads to the appearance of new peaks at various
ne ← 1g , 2g , . . . transition frequencies. The resolution is degraded with the use of
wider-bandwidth, σ = 0.5τx pulses. The resulting linear absorption spectrum of the
model system in the ng = 0 state is shown as the dashed line in Fig. 1.4.
The spectra in Figs 1.3 and 1.4 illustrate several key ideas. The quasi-cw absorp-
tion of this relatively simple system provides detailed information on the energy eigen-
spectrum and the Franck–Condon overlaps ne |ng , and is amenable to a dynamical
description through the Heller kernel. In more complicated, multi-mode systems, how-
ever, thermal congestion and the presence of many closely spaced energy levels make it
more difficult to resolve individual vibronic transitions. Short-pulse linear absorption,
on the other hand, pays for its temporally well-localized excitation with a correspond-
ing loss of spectral resolution. In a one-pulse experiment, this loss of resolution in the
frequency domain remains uncompensated by improved temporal resolution resulting
from precise control of the timing within a sequence of individual light-matter interac-
tions. As is explored in subsequent chapters, strategies have been developed to address
Example absorption calculations 9
1.0
0.8
0.6
0.4
0.2
0.0
0 500 1000 1500 2000
- (cm-1 )
Fig. 1.4 Solid line is the absorption spectrum of the model at room temperature, again with
σ = 5.0τx . Dashed curve gives the spectrum starting from ng = 0, with σ = 0.5τx .
this conundrum by systematically varying the delay between excitation and probing
in several different forms of ultrafast nonlinear optical spectroscopy.
Fig. 1.5 Nuclear wave packets in the electronic excited state of the model system at t = 0
(top) and 0.5753τx (bottom). The legend identifies the long, intermediate, and short values
of the pulse-duration parameter σ used in these calculations
.
Figure 1.5 illustrates the dynamics initiated by pulses of differing duration in the
model molecule. It shows wave packets in the position representation,
at t = 0 and 0.5753τx (about half the period of vibrational motion at energy 2e ), with
center frequency Ω = ( + 2e − 0g )/. The upper panel plots the real part of ψe (x, 0).
For the longest pulse-length, σ = 2.0τx , this initial wave packet closely resembles the
10 Short-pulse electronic absorption
stationary wave function hx|2e i. The shortest pulse, with σ = 0.2τx , essentially copies
hx|0d i into the excited electronic state. Each of these wave packets after half a period of
evolution, along with one generated by a pulse of intermediate duration, is illustrated
in the lower panel. The longest-pulse packet remains almost unchanged, while that
generated by the shortest pulse migrates to larger distance and is distorted by motion
in the anharmonic e-state potential. It is ultrashort laser pulses’ capacity to initiate
and interrogate wave-packet dynamics that underlies their usefulness in time-resolved
nonlinear optical spectroscopy.
2
Adiabatic approximation
X P̂α · P̂α
TN =
2Mα
nuclei α
1 Some of the arguments in this section follow those made in Chap. 21 of Gordon Baym, Lectures
on Quantum Mechanics (Westview Press, New York, 1990).
12 Adiabatic approximation
Thus nuclear momenta are about ten times larger than electronic momenta, and nu-
clear speeds P/M are one thousand times smaller than electronic speeds. Finally, the
typical deviation of a nuclear position from its equilibrium average value is given by
M ω 2 δ 2 /2 ∼ ~ω/2, which leads to the estimate (δ/a)2 ∼ (m/M )1/2 , or
m 1/4
δ∼ a; (2.4)
M
the spatial range of zero-point nuclear motion is about one-tenth the range of electronic
motion (i.e., one-tenth the “molecular size”). These ballpark figures support a physical
picture of a molecule as a well-defined configuration of nuclei executing slow, small
amplitude oscillations around their equilibrium positions within a delocalized medium
of rapidly moving electrons.
The electronic eigenkets are parametrized by the nuclear coordinate operators, which
we symbolically denote by R̂ = (R̂1 , R̂2 , . . . ). The electronic eigenenergies En (R̂) like-
wise depend on the locations of the nuclei.
We can relate the nuclear-coordinate-operator-dependent electronic eigenkets to
those at some reference nuclear arrangement—say the equilibrium configuration in
the electronic ground state—by a unitary tranformation:
The electronic unitary operator U (R̂) is actually a large collection of different unitary
operators parametrized by R̂. Since the reference eigenkets |ϕn i constitute a complete
orthonormal set in the electronic state-space, we could P seek molecular eigenkets as
expansions in this fixed electronic basis of the form n |ϕn i|Θn i, where the |Θn i are
nuclear states. But it proves more useful by far to look for energy eigenstates of the
combined electronic and nuclear degrees of freedom as expansions in the adiabatic
electronic basis: X X
|Ψi = |ϕn (R̂)i|Φn i = U (R̂) |ϕn i|Φn i . (2.7)
n n
Justification for the adiabatic electronic basis as the practical choice emerges when
we substitute |Ψi in the molecular Schrödinger equation (2.1) and take account of the
small ratio of electronic to nuclear mass.
Making this substitution and availing ourselves of the electronic Schrödinger equa-
tion (2.5) lead to
Xh i X
TN + Em (R̂) |ϕm (R̂)i|Φm i = E |ϕm (R̂)i|Φm i . (2.8)
m m
Taking the inner product of eqn (2.8) with |ϕn (R̂)i yields a set of coupled equations
for the nuclear states {|Φn i}:
h i X
hϕn (R̂)|TN |ϕn (R̂)i + En (R̂) |Φn i + hϕn (R̂)|TN |ϕm (R̂)i|Φm i = E|Φn i . (2.9)
m6=n
Let us examine the operator coupling nuclear kets associated with different electronic
states, which can be rewritten as
X P̂α · P̂α
hϕn (R̂)|TN |ϕm (R̂)i = hϕn |U † (R̂) U (R̂)|ϕm i
α
2Mα
X1
= hϕn | U † (R̂)P̂α U (R̂) · U † (R̂)P̂α U (R̂)|ϕm i
α
2M α
X 1 h i h i
= hϕn | P̂α + Aα (R̂) · P̂α + Aα (R̂) |ϕm i , (2.10)
α
2Mα
14 Adiabatic approximation
where the induced vector potential Aα (R̂) = −i~U † (R̂)∇α U (R̂) is an operator in the
electronic state-space parametrized by the nuclear coordinate operators.2 Now
X 1
hϕn (R̂)|TN |ϕm (R̂)i = hϕn |ϕm i P̂α · P̂α
α
2Mα
X 1
+ P̂α · hϕn |Aα (R̂)|ϕm i + hϕn |Aα (R̂)|ϕm i · P̂α
α
2Mα
X 1
+ hϕn |Aα (R̂) · Aα (R̂)|ϕm i . (2.11)
α
2Mα
The inner product hϕn |ϕm i equals δnm . The matrix elements of the vector potential,
hϕn |Aα (R̂)|ϕm i = −i~hϕn (R̂)|∇α ϕm (R̂)i, are of estimated size ~/a as one would have
to move the αth nucleus a molecule-sized distance to produce near unit overlap between
a given adiabatic electronic state and another to which it had been orthogonal. The
second sum on the right-hand side of eqn (2.11) therefore contains terms of order
3/4
p2
~ m
Pα ∼ ,
Mα a m Mα
a thousand times smaller than electronic excitation energies. Terms in the third sum
are of similar or smaller size,
1 1 X
hϕn |Aα (R̂) · Aα (R̂)|ϕm i = hϕn |Aα (R̂)|ϕl i · hϕl |Aα (R̂)|ϕm i
2Mα 2Mα
l
2
1 X ~ m p2
∼ ∼ 10 ,
2Mα a Mα 2m
l
We are led to adopt the Born–Oppenheimer approximation, under which the ex-
pansion in eqn (2.7) is reduced to a single term,4
|Ψi ∼
= |ϕn (R̂)i|Φn i . (2.12)
In this approximation, the nuclear state that accompanies |ϕn (R̂)i obeys the equation
h i
hϕn (R̂)|TN |ϕn (R̂)i + En (R̂) |Φn i = E|Φn i , (2.13)
hϕn |Aα (R̂) · Aα (R̂)|ϕn i = hϕn |Aα (R̂)|ϕn i · hϕn |Aα (R̂)|ϕn i
X
+ hϕn |Aα (R̂)|ϕm i · hϕm |Aα (R̂)|ϕn i ,
m6=n
where
†
A(n)
α (R̂) = −i~hϕn |U (R̂)∇α U (R̂)|ϕn i = −i~hϕn (R̂)|∇α ϕn (R̂)i (2.15)
is the adiabatic vector potential experienced by the αth nucleus in the nth electronic
state, and
X 1 X
En0 (R̂) = En (R̂) + hϕn |Aα (R̂)|ϕm i · hϕm |Aα (R̂)|ϕn i . (2.16)
α
2Mα
m6=n
is imaginary. If the electronic wave function hr1 , r2 · · · |ϕn (R̂)i can be chosen real,
(n)
as it most often can be, then hϕn (R̂)|∇α ϕn (R̂)i will be real. Hence Aα (R̂) is also
imaginary and therefore zero.
4 The Born–Oppenheimer-approximated state |Ψi is not, in general, a tensor product. Rather it is
a “sum” of products d3 R1 d3 R2 · · · |ϕn (R1 , R2 , · · · )i|R1 , R2 , · · · ihR1 , R2 , · · · |Φn i, in which the
R R
electronic state is correlated with the arrangement of the nuclei and weighted by the probability
amplitude, Φn (R1 , R2 , · · · ) = hR1 , R2 , · · · |Φn i, for the nuclei to have a particular arrangement.
5 M. V. Berry, “Quantal phase factors accompanying adiabatic changes,” Proc. R. Soc. Lond.
A392, 45–57 (1984).
16 Adiabatic approximation
When the dust settles, it is often sufficient to combine a solution of the simplified
nuclear wave equation, h i
TN + En (R̂) |Φn i = E|Φn i , (2.17)
1. How much can you figure out about the eigenstates and eigenenergies of the
model Hamiltonian
P12 + P22 M ω2 2
H= + (Q1 + Q22 ) + κQ1 σx + κQ2 σy ,
2M 2
where (Q1 , Q2 ) and (P1 , P2 ) are vibrational coordinates and momenta, respectively,
σx = |eihg| + |gihe| ,
σy = −i|eihg| + i|gihe| ,
and
σz = |eihe| − |gihe| ,
√
are Pauli operators in a two-dimensional electronic state-space, and κ = k ~M ω 3 ,
with a dimensionless constant k, is a coupling coefficient between electronic and
nuclear degrees of freedom?
2. In addition to determining the stationary nuclear wave functions, the approximate
Hamiltonian appearing in eqn (2.17) governs nuclear dynamics. Consider a one-
dimensional vibrational Hamiltonian,
P2 M ω2 2
T + E(X) = + X ,
2M 2
where [X, P ] = i~. Investigate the time-dependent expectation value and mean-
square deviation of X and P in a Glauber coherent state specified by the initial
condition |Φi = e−iP̂ X◦ /~ |0i, where X◦ is some initial displacement and |0i is the
ground state of this harmonic nuclear Hamiltonian.
Two references pertinent to the first of these exercises are given below.6
the potential energy surfaces of whose nuclear Hamiltonians take the value zero at
their respective minima, and whose bare electronic energies obey f − e < e < f .
The molecule’s fixed location identifies the spatial origin. Its interaction with two
ultrashort laser pulses is governed by
The pump (u) and probe (r ) pulses have envelopes fu (t) and fr (t − td ), respectively,
of temporal width ∼ σu and ∼ σr ; their arrival times are separated by td . The two
field strengths (Eu and Er ), center frequencies (Λu and Λr ), and uncontrolled optical
phases (ϕu and ϕr ) are all independent.
The measured quantity in a transient-absorption experiment is the change in the
spectrally resolved loss of energy from the probe pulse due to the prior action of
18 Transient-absorption spectroscopy
the pump. To derive a formula for such a signal, we could calculate the change in
electromagnetic energy—in a narrow frequency range of width δω centered at ω̄—due
to interference between the spatially propagating probe field and the field radiated by
a portion of the system’s induced dipole moment,
Z
1 n
d3 R Erω̄ (R, t) + Eu2 rω̄ (R, t) · Erω̄ (R, t) + Eu2 rω̄ (R, t)
∆Uω̄ =
4π
o
− Er2ω̄ (R, t)
Z
∼ 1
= d3 R Erω̄ (R, t) · Eu2 rω̄ (R, t) . (3.3)
2π
In this equation, Eu2 r is the contribution to the induced dipolar field proportional
to pump intensity and the probe field strength. The subscript ω̄ denotes a spectrally
filtered field component; for instance, the filtered probe is given by
Z ω̄+ δω2 dω
Erω̄ (R, t) = e−iωt Ẽr (R, ω) + c.c. , (3.4)
ω̄− δω
2
2π
R∞
where Ẽr (R, ω) = −∞ dt eiωt Er (R, t).1
The integration over space in eqn (3.3) could be evaluated by a procedure analo-
gous to that in Appendix A.2 As an alternative, we pursue a direct derivation of the
transient-absorption signal as a time integral for the contribution to the change in
molecular energy, ∆Eω̄ = −∆Uω̄ , that is bilinear in the pulse intensities. As in eqn
(3.2), we assume for simplicity that both transition dipole moments and both field
polarizations are oriented along the same (say x -) axis. Temporarily writing m(t) for
mu2 r (t) and E(t) for Er (t), we can break up the probe field into pieces which would
be detected by idealized pixels sensitive to different small frequency ranges:
X
E(t) = Eω̄ (t) , (3.5)
ω̄>0
In terms of the latter two quantities, the transient-absorption signal would be written
Z ∞
dEω̄ (t)
∆Eω̄ = − dt mω̄ (t) , (3.7)
−∞ dt
in which both probe-field and the dipole-moment components belong to the spectral
segment centered at ω̄ (compare eqn (1.15)). Now,
1 Equations (3.3) and (3.4) differ from eqn (3.2) in making explicit the vector nature of the incident
and radiated fields.
2 See also J. A. Cina, P. A. Kovac, C. C. Jumper, J. C. Dean, and G. D. Scholes, “Ultrafast
transient absorption revisited: Phase-flips, spectral fingers, and other dynamical features,” J. Chem.
Phys. 144, 175102/1–18 (2016).
Model Hamiltonian and signal expression 19
Z ω̄+ δω
2 dω −iωt
Eω̄ (t) = e Ẽ(ω) + c.c. , (3.8)
ω̄− δω
2
2π
where Z ∞ Z ∞
Ẽ(ω) = dt eiωt E(t) = eiωtd dτ eiωτ E(τ + td ) . (3.9)
−∞ −∞
Since the probe field E(τ + td ) has a temporal width ∼ σ 2π/δω about τ = 0, for
frequencies in the range ω̄ − δω δω
2 < ω < ω̄ + 2 we have
Z ∞
Ẽ(ω) ∼
= eiωtd dτ eiω̄τ E(τ + td ) = ei(ω−ω̄)td Ẽ(ω̄) . (3.10)
−∞
sin δω
2 (t − td )
Eω̄ (t) ∼
= 2 Re Ẽ(ω̄)e−iω̄t
. (3.11)
π(t − td )
The molecular energy acquisition involving a particular spectral segment of the
probe field and the full induced dipole moment is
Z ∞
dEω̄ (t) δω
− dt m(t) ∼
= iω̄ Ẽ(ω̄)m̃∗ (ω̄) + c.c. (3.12)
−∞ dt 2π
Since the right-hand side of eqn (3.12) involves the ω̄ Fourier components of both
the probe field and the induced dipole, it clearly corresponds to ∆Eω̄ of eqn (3.7),
the transient-absorption signal at the given frequency. Reintroducing the appropriate
subscripts, the signal may therefore be written
Z ∞
dErω̄ (t)
∆Eω̄ = − dt mu2 r (t) ; (3.13)
−∞ dt
the transient-absorption dipole in its entirety appears in the integrand, rather than
just its ω̄ element.
For a probe pulse of the form given in eqn (3.2), the Fourier components are
Er iωtd −iϕr ∞
Z
∼
Ẽr (ω > 0) = e dt ei(ω−Λr )(t−td ) fr (t − td )
2 −∞
Er iωtd −iϕr ˜
= e fr (ω − Λr ) , (3.14)
2
whence3
sin δω
2 (t − td )
Erω̄ (t) ∼
= Er f˜r (ω̄ − Λr ) cos[ω̄(t − td ) + ϕr ] . (3.15)
π(t − td )
Inserting eqn (3.15) in eqn (3.13) gives a working formula for the transient-absorption
signal,
3 We specialize to symmetric pulse envelopes, so f˜r (ω̄ − Λr ) is real-valued.
20 Transient-absorption spectroscopy
∞
sin δω
2 (t − td )
Z
∆Eω̄ = ω̄Er f˜r (ω̄ − Λr ) dt mu2 r (t) sin[ω̄(t − td ) + ϕr ] . (3.16)
−∞ π(t − td )
The dynamics of the targeted system enters the signal through the interpulse delay-
dependence and spectral content of the induced dipole moment.
Fill in the steps that lead to eqns (3.11), (3.12), and (3.15). Show that eqn (3.11)
is consistent with eqn (3.15) for a probe pulse of the form given in eqn (3.2).
|Ψi ∼
= |0i + | ↑u i + | ↑r i + | ↑u ↑u i + | ↑u ↓u i + | ↑u ↑r i + | ↑u ↓r i + | ↑r ↑u i + | ↑r ↓u i
+ | ↑r ↑r i + | ↑r ↓r i + | ↑u ↑u ↓r i + | ↑u ↓u ↑r i + | ↑u ↑r ↓u i + | ↑u ↓r ↑u i
+ | ↑r ↑u ↓u i + | ↑r ↓u ↑u i . (3.17)
Here |0i is the time-dependent molecular state unperturbed by either laser pulse. The
arrows in the other kets signify energetically upward (g-to-e or e-to-f ) or downward
(e-to-g or f -to-e) electronic transitions driven by the pump or the probe, as indicated
by the subscript. These transitions are required to occur in the order listed from left
to right. Kets in which the pulses act “out of order” (probe before pump) vanish for
interpulse delays significantly longer than the pulse durations. From the expectation
value of the dipole-moment operator, we can isolate
The four under-braced matrix elements will be seen to carry an uncontrolled optical
phase factor exp{−2iϕu }. Over the many laser shots needed to acquire a transient-
absorption signal, the phase 2ϕu randomly samples the range from zero to 2π, leading
to a negligible net contribution from these elements.
Transient-absorption dipole 21
Provide physical explanations for each of the interpretive labels in eqn (3.18).
where tu = 0, tr = td , and
after making a rotating-wave approximation. In eqn (3.20), we have used reduced pulse
propagators of the form
Z t
(eg) dτ
pj (t; τ ) = [−τ ]ee [τ ]gg e−iΛj τ fj (τ ) , (3.21)
−∞ σj
(eg)
for example, along with Fj = Ej meg σj /2~.
Rather than generate explicit expressions for all the multi-pulse kets and bras ap-
pearing in eqn (3.18), some of which may become insignificant with particular choices
22 Transient-absorption spectroscopy
of pulse center frequencies and bandwidths, or certain ranges of interpulse delay time,
it is more instructive simply to write out a few examples. We have, for instance,
and
Gain some practice by writing expressions analogous to eqns (3.22)–(3.25) for a few
more multi-pulse kets.
From expressions of the sort just considered, we can obtain the dipole matrix
elements. For the ESA element, for example, we find
2
h↑u ↑r |m̂| ↑u i = −imf e Fu(eg) Fr(f e) eiϕr hng |p(ge) (ef )
u (τ + td ; τ̄ )[−td ]ee pr (t − td ; τ )
(eg)
× [−t + td ]ff [t − td ]ee [td ]ee pu (t − td ) + td ; τ̄¯ |ng i . (3.26)
∞
dτ sin δω
2 τ
Z
(f e)
prω̄ (∞; τ ) = δω
[−τ ]ff [τ ]ee e−iω̄τ , (3.27)
−∞ σ̄ 2 τ
Carry out this exercise for the contributions of ground-state bleach, impulsive stim-
ulated Raman scattering, and stimulated emission to transient absorption. You can
check your answers by consulting Section 3.3.2.
Exemplary calculations 23
Ω2 ωg2
Vg (X, x) = (X − Xg )2 + (x − xg )2 . (3.30)
2 2
The slow-mode frequency Ω serves as a reference, set to 40 cm−1 for plotting purposes,
while ωg = 13.5pΩ; the ground-statepminimum lies at (Xg , xg ) = (−1.8 Xrms , 0 xrms ),
where Xrms = ~/2Ω and xrms = ~/2ωg . The slow mode serves as a photochemical
reaction coordinate in the e-state, whose potential function is
3
Ve (X, x) = αe (X − Xe )4 + 4Xe (X − Xe )3 + 3Xe2 (X − Xe )2
2
2
ω
+ e (x − xe )2 + βe (X − Xe )(x − xe ) , (3.31)
2
4
with αe = 0.01613 ~Ω/Xrms , Xe = 4.0 Xrms , ωe = 11.0 Ω, xe = 1.5 xrms , and βe =
−0.4550 ~Ω/Xrms xrms . We set the bare electronic energy of this state to e = 600.0 ~Ω.
The f -state potential function is of similar form,
3 4 3 2 2
Vf (X, x) = αg (X − Xf ) + 4Xf (X − Xf ) + 3Xf (X − Xf )
2
ωf2
+ (x − xf )2 + βf (X − Xf )(x − xf ) , (3.32)
2
4
in which αf = 0.02419 ~Ω/Xrms , Xf = 3.0 Xrms , ωf = 9.0 Ω, xf = 0.9 xrms , and
βf = −0.3344 ~Ω/Xrms xrms . The bare electronic energy of the f -state is assigned
the value f = 900.0 ~Ω. Contour diagrams of the three electronic potential energy
surfaces are displayed in Fig. 3.1.
The continuous-wave electronic absorption spectrum provides a first acquaintance
with the spectroscopy and dynamics of this system. The model Hamiltonian is simple
enough to be diagonalizable (we used a discrete position basis), and the spectrum can
readily be calculated from eqn (1.28) once the transition energies and Franck–Condon
overlaps are determined. As seen in Fig. 3.2, with the incident frequency denoted as Λ
and line-widths determined by the finite duration σ = 1.6 × 2π/Ω of a Gaussian pulse,
24 Transient-absorption spectroscopy
"
#
!!!"#
!!!
!!
!#
! !!
!!!"# ! !!
!!!"# ! !!
!!!"#
$ # " ! $ # " ! $ # " ! %
!! !! !!
Fig. 3.1 Vg , Ve , and Vf are shown from left to right. Contours are separated by 2 Ω.
1.0
0.8
0.6
0.4
0.2
0.0
0 200 400 600 800
ℏΛ -ϵϵe (cm-1 )
Fig. 3.2 Linear absorption of the two-dimensional model system in arbitrary units versus
wave-number displacement from the e ← g bare electronic transition energy.
Can you rationalize the value e − 34.4 cm−1 of the energy for the very weak zero-
zero absorption transition through a consideration of the zero-point energies in the
ground and excited electronic states?
The wave-packet dynamics underlying linear absorption in this system is brought
to light by the Heller overlap,
appearing in eqn (1.27). In this formula, |0g i denotes the two-dimensional vibrational
ground state. As explained in Chapter 1, C(t) is the complex-valued overlap between
the initial, stationary vibrational wave function and the moving wave packet engen-
dered by its transfer to the electronic excited state. Fourier transformation of eqn (3.33)
times the temporal window exp{−t2 /4σ 2 } (and some constant pre-factors) yields the
spectrum. The absolute value of the overlap times the window function is plotted in
Fig. 3.3. It decays rapidly as the 2D wave packet gains slow-mode momentum P and
1.0
0.8
|exp{-t2 /4σ2 } C(t)|
0.6
0.4
0.2
0.0
0 1 2 3 4 5 6
t⨯Ω/2π
Fig. 3.3 The magnitude of the Heller overlap times the temporal window versus time in
g-state slow-mode periods. Values taken at several small-integer multiples of 5/2 times the
fast-mode vibrational period in the e-state are marked with dots.
4 4 4
2 2 2
x/xrms
x/xrms
x/xrms
0 0 0
-2 -2 -2
2 2 2
x/xrms
x/xrms
x/xrms
0 0 0
-2 -2 -2
Fig. 3.4 Real part of nuclear wave packet in the excited electronic state at t = 0, 1, ..., 5
times 2.5 (2π/ωe ), the times marked in the plot of Heller’s absorption kernel.
2
〈X〉/Xrms
-1
-2
2.0
1.5
〈x〉/xrms
1.0
0.5
0.0
0 1 2 3 4 5 6
t⨯Ω/2π
Fig. 3.5 hXi and hxi versus time during wave-packet motion in the e-state.
Exemplary calculations 27
with the g-to-e electronic transition.4 Because these pulses are non-resonant with
f ↔ e, all of the overlaps in the transient-absorption dipole of eqn (3.18) involving
amplitude transfer to or from the f -state become negligibly small, and it simplifies to
mu2 r (t) ∼
= h↑u ↓u ↑r |m̂|0i+h↑r ↓u ↑u |m̂|0i+h↑r |m̂| ↑u ↓u i+h↑u |m̂| ↑u ↓r i + c.c. (3.34)
According to the pattern established in Section 3.2.2, the first of these terms gives rise
to a ground-state-bleach signal contribution,
(compare eqn (3.28)). The second term in eqn (3.34) requires the probe to act before
the pump and will be neglected here. The third term generates the impulsive Raman
signal contribution,
The minus in front of these signals corresponds to the fact that in each case the pump
pulse tends to decrease probe absorption. In both GSB and ISRS contributions, wave-
packet evolution during the interpulse delay occurs in the electronic ground state,
whereas in SE it is in the excited state.
Adopting a simplification that affects the calculated signal only at delay times as
short as the pulse durations, we neglect the truncation of pump action by the probe.
In the stimulated-emission contribution, for instance, we make the replacement,
p(ge) (t; τ )[td ]ee p(eg) ∼ (ge) (t; τ )[td ]ee p(eg) (∞; τ̄ ) .
r u (τ + td ; τ̄ ) = pr u (3.38)
This approximation ceases to matter once td exceeds the pulse durations and the ma-
(eg)
trix elements of pu take constant values (see below). Detailed attention must be paid
to the implementation of this approximation in the case of a pump-pulse propagator
whose upper limit involves the integration variable of a reduced propagator for the
spectrally filtered probe pulse. For example, the impulsive stimulated Raman signal of
(ge) (eg) (ge)
eqn (3.36) includes the combination pr (t; τ̄¯)prω̄ (∞; t)[td ]gg pu (t + td ; τ ), in which
4 We postpone to Chapter 5 the calculation and interpretation of two-color transient-absorption
signals from a similar model system in which a pump-pulse resonant with g-to-e is accompanied by
a probe-pulse resonant with e-to-f.
28 Transient-absorption spectroscopy
the pump is truncated by the integration variable of the (temporally elongated) fil-
(ge)
tered probe. The left-most term, pr (t; τ̄¯), only takes non-negligible values once t
approaches or exceeds td ; and we are seeking an accurate treatment only for td signif-
icantly longer than the pulse durations. In this delay range, t + td in the argument of
(ge)
pu is always larger than σu , and we can write,
p(ge)
(eg)
(t;τ̄¯)prω̄ (∞;t)[td ]gg p(ge) ∼ (ge) (t;τ̄¯)prω̄ (∞;t)[td ]gg p(ge) (∞;τ ) . (3.39)
(eg)
r u (t+td ;τ ) = pr u
(ge) (eg)
Show that the same reasoning can be applied to pu (t + td ; τ̄¯)[−td ]ee prω̄ (∞; t)
(ge)
× pr (t; τ ) appearing in eqn (3.37) for the stimulated-emission signal.
Under these approximations, the three relevant contributions (3.35)-(3.37) to the one-
color transient-absorption signal take the simpler forms
and
f˜r (ω̄ − Λr ) (eg) 2 (eg) 2 (eg)
∆Eω̄SE = −~ω̄ Fu Fr h0g |p(ge) ¯
u (∞; τ̄ )[−td ]ee pr ω̄ (∞; t)
σr
× p(ge)
r (t; τ )[td ]ee p(eg)
u (∞; τ̄ )|0g i + c.c. , (3.42)
respectively.
With a choice of Gaussian pulses having envelopes fj (t) = exp{−t2 /2σj2 } for j =
u or r, the relevant first-order reduced pulse propagators have forms like
Z t
(eg) dτ τ2 iτ
hne |pj (t; τ )|0g i = hne |0g i exp − 2 + e + ne − 0g − ~Λj
σ
−∞ j 2σ j ~
q
π −s2 /2
1 t
= hne |0g i 2 e 1 + erf √2 σj − is , (3.43)
σ
where s = ~j (e + ne − 0g − ~Λj ) (i.e., the dimensionless offset of ne ← 0g from
Rx 2
resonance) and erf(x) = √2π 0 dy e−y . This quantity is plotted in Fig. 3.6 with respect
to its arguments s and t. Notice that the real part of the reduced propagator begins
to take non-negligible values as the light pulse turns on and remains non-negligible
thereafter, whereas the imaginary part is substantial only during the short episode of
pulse action. Both portions become small for transitions displaced from the carrier
frequency by more than ∼ 2π/σj .
Exemplary calculations 29
(eg)
Fig. 3.6 Real part (left) and imaginary part (right) of pj (t; τ ) divided by the Franck—
Condon overlap hne |0g i, as functions of s = (e + ne − 0g − ~Λj )σj /~ and t/σj .
The ground-state bleach signal of eqn (3.40) and the impulsive Raman signal of eqn
(ge) (eg)
(3.41) both call for the nested double pump-pulse propagator, pu (∞; τ )pu (τ ; τ̄ ).
Nuclear matrix elementsPof this operator are easily obtained by inserting a com-
pleteness relation 1 =
√ ne |ne ihn
√e
| and changing the variables of integration to
x = (τ + τ̄ )/ 2 and y = (τ − τ̄ )/ 2. This procedure leads to
2
σu 2 X 2
− 4~2 (ng −0g )
hng |p(ge) (eg)
u (∞; τ )pu (τ ; τ̄ )|0g i = πe hng |ne ihne |0g ie−ξ [1 + erf(iξ)],
ne
(3.44)
with ξ = [~Λu + (ng +0g )/2 − e − ne ] σu /~. The summand in eqn (3.44) is a prod-
uct of a term that depends only on the nuclear eigenenergies and terms that depend
2 2 2
on the form of the corresponding eigenfunctions. The overall factor e−σu (ng −0g ) /4~
enforces a requirement that the initial and final nuclear states lie within the spectral
2
bandwidth of the pump pulse. The factor e−ξ [1 + erf(iξ)] might appear similarly to
suggest that e + ne − (ng + 0g )/2 must equal ~Λu within ±2π~/σu in order for
a given e-state vibrational level to contribute to the double pulse-propagator; that
this is emphatically not the case is illustrated by Fig. 3.7, which plots the real and
2
imaginary parts of this factor as a function of ξ. While the real part is just e−ξ , the
2 ξ 2
imaginary part is e−ξ √2π 0 dz ez , which is seen to take its values of largest magni-
R
tude at ξ ∼
= ±0.924 and to fall off much more slowly with increasing |ξ| than the real
part.5 This behavior of the second-order pulse propagator accounts for the efficacy of
2 2 2 Rξ 2
5 Forlarge positive ξ, we have √2π 0ξ dz ez −ξ ∼ = √2π e−2ξ 0 dz e2ξz = √1πξ (1 − e−2ξ ), so the
R
imaginary part of the double pulse-propagator falls off inversely as the resonance offset.
30 Transient-absorption spectroscopy
1.0
0.5
e - (1+erf[i ])
2
0.0
-0.5
-10 -5 0 5 10
2
Fig. 3.7 Real (solid) and imaginary (dashed) parts of e−ξ [1 + erf(iξ)].
Under the current neglect of pump-probe overlap effects, the remaining second-
order pulse propagators are those involving the probe pulse nested inside the spectrally
filtered probe, with matrix elements such as
∞
dt sin δω
t
2 t −iω̄t dτ − 2σ
τ2
(eg) +iΛr τ
ne |prω̄ (∞; t)p(ge)
r (t; τ )|n̄e = δω
e e r2
ng −∞ σ̄ 2 t −∞ σr
× ne |[−t]ee [t]gg |ng ng |[−τ ]gg [τ ]ee |n̄e . (3.45)
The outer, t-integration takes the form of a windowed Fourier transformation; it could
easily be carried out numerically in the present calculations. But the sinc-like form
of its prolonged temporal window is just an artifact of our idealized treatment of
spectral filtration.8 It is therefore expedient—and little affects the outcome—to make
the replacement
δω 2
sin δω
2 t ∼
1 δω 3
∼ −1
δω 2 t
δω t 2 t − 6 2 t =e 6 ,
2
δω = (3.46)
2 t
which is correct through second order in t and vanishes for |t| → ∞. With this sub-
stitution, it becomes possible to evaluate eqn (3.45) explicitly, yielding
√ X 6λ2 λ̄2 σr2
(eg)
hne |prω̄ (∞; t)p(ge)
r (t; τ )|n̄ e i = 3 hn |n
e g ihn |n̄
g e i exp − −
ng
δω 2 2
i 12λ + λ̄σr2 δω 2
× 1 + erf p , (3.47)
δω 24 + 2σr2 δω 2
where ~λ = e + ne − ng − ~ω̄ and ~λ̄ = e + n̄e − ng − ~Λr . The summand in
eqn (3.47) is a product of contributing Franck–Condon overlaps with a factor that
depends on the offset of the corresponding transition frequencies from ω̄ and Λr . We
now have at our disposal representative examples of matrix elements of all the reduced
pulse-propagator combinations needed to calculate transient-absorption signals from
our model system.
1. Derive eqn (3.47) from√ eqns (3.45) √ and (3.46) by making successive variable
changes to (t̄ = tδω/2√ 6,p τ̄ = τ /σr 2), then (U = t̄ sinp θ + τ̄ cos θ, u = t̄ cos θ −
τ̄ sin θ), with cos θ = 2 3/ 12 + σr2 δω 2 and sin θ = σr δω/ 12 + σr2 δω 2 . As a check
on your√derivation (and mine) you may consider the unrealistic situation in which
δω = 2 3/σr and ω̄ = Λr , and compare the result to eqn (3.44).
(ge) (eg)
2. Obtain an approximate formula for hng |prω̄ (∞; t)pr (t; τ )|n̄g i, which is needed
for the GSB and ISRS contributions, by analogy with eqn (3.47).
Now we can take a look at the one-color transient-absorption spectrum of our model
system. The components given by eqns (3.40)–(3.42) were calculated with identical
pump and probe pulses having center frequencies Λu = Λr matching e /~ = 600 Ω and
durations σu = σr = 0.20(2π/ωe ). The dectector slit-width is set to δω = 3.33 Ω, so
that σ̄ = 250 fs remains sufficiently short to resolve motion on the 833-fs timescale of
the 40-cm−1 mode.
The calculated contributions along with the net transient-absorption signal are
shown in Figs 3.8 and 3.9. Each of these mostly negative quantities is plotted as
the hyperbolic tangent of the corresponding component divided by about 87% of its
largest magnitude in order to accentuate the lower-amplitude features.9 Those max-
imum magnitudes take relative proportions 1 : 1.07 : 1.98 : 3.70 for the ground-state
bleach, impulsive stimulated Raman, stimulated emission, and total signals, respec-
tively. Interpulse delay is reckoned in periods of the slow ground-state vibration, and
the exhibited spectral range is nearly twice the full width at half maximum of the
pulses’ power spectrum.
The central stripes in the GSB and ISRS contributions undulate at frequency
Ω, while sideband oscillations occur at the higher frequency ωg . In the stimulated-
emission portion, e-state dynamics at the round-trip time of the slower motion and
the vibrational frequency ωe both become manifest. Because the pump pulse is much
shorter than 2π/Ω, impulsive Raman excitation of the slower mode becomes inefficient,
9 The plotted quantities are independent of E , E , and the transition dipole moment m , so we
u r eg
need not specify values for these quantities.
32 Transient-absorption spectroscopy
Fig. 3.8 Ground-state bleach (left) and impulsive stimulated Raman (right) components
of the transient-absorption signal from the two-dimensional model system with degenerate,
electronically resonant pump and probe pulses. Hyperbolic-tangent scaling is used here and
in the next figure to spread the most closely spaced contours.
Fig. 3.9 Stimulated emission (left) and total signal (right) from the 2D model.
and the low-frequency features in GSB and ISRS are less prominent than the Franck–
Condon displacement-driven slow motion in the SE signal. The blue sidebands in GSB
and the red ones in ISRS are seen to be out of phase with each other. While sideband
structure is present on both blue and red sides of the SE spectrum, the latter appears
more complicated. The two sides of the SE signal more obviously exhibit anti-phasing
Other documents randomly have
different content
l'amusait et l'intéressait; elle vit jouer Andromaque, qui lui fit
répandre plus de six larmes; le Médecin malgré lui l'a fait pâmer de
rire, le Tartuffe l'intéressa [80]. Et tout cela ne l'empêche nullement de
remplir exactement ses devoirs de religion, et de demander à sa fille
toutes les fois qu'elle communie [81].
Les affaires, les divertissements et les festins ne faisaient pas oublier
les jeux d'esprit, passés en habitude dans la haute société de cette
époque. «Lavardin et des Chapelles ont rempli des bouts-rimés que
je leur ai donnés; ils sont jolis, je vous les enverrai [82].»
Madame de Sévigné, entraînée elle-même par la nécessité de
paraître aux états d'une manière conforme à son rang et à la
réception qu'on lui faisait, se pare d'un luxe qu'elle ne pouvait avoir
à la cour et à Paris, mais qui dans sa province était convenable et de
bon goût. Ainsi, quand elle rendait des visites dans ses environs, ou
quand elle allait à Vitré, elle faisait atteler six chevaux à sa voiture;
et elle témoigne naïvement à sa fille que son bel attelage et la
rapidité de ses chevaux lui plaisent beaucoup [83].
Pendant le temps que durèrent les assises des états, elle se rendait
à Vitré le moins souvent qu'elle pouvait, et préférait se tenir à la
campagne; mais elle n'était pas toujours maîtresse de suivre en cela
sa volonté. D'ailleurs on ne la laissait jamais jouir en paix de ses
champs et de ses bois; et la dépense que lui occasionnaient les
visiteurs était pour elle un motif puissant pour céder aux instances
qui lui étaient faites de sortir des Rochers.
Elle écrit de Vitré, le 12 août, à madame de Grignan [84]:
«Enfin, ma chère fille, me voilà en pleins états; sans cela, les états
seraient en pleins Rochers. Dimanche dernier, aussitôt que j'eus
cacheté mes lettres, je vis entrer quatre carrosses à six chevaux
dans ma cour, avec cinquante gardes à cheval, plusieurs chevaux de
main et plusieurs pages à cheval: c'étaient M. de Chaulnes, M. de
Lavardin [85], MM. de Coëtlogon [86], de Locmaria, le baron de Guais,
les évêques de Rennes, de Saint-Malo, les messieurs d'Argouges [87],
et huit ou dix autres que je ne connais point; j'oublie M. d'Harouïs,
qui ne vaut pas la peine d'être nommé. Je reçois tout cela. On dit et
on répondit beaucoup de choses. Enfin, après une promenade dont
ils furent fort contents, une collation, très-bonne et très-galante,
sortit d'un des bouts du mail, et surtout du vin de Bourgogne, qui
passa comme de l'eau de Forges: on fut persuadé que cela s'était
fait avec un coup de baguette. M. de Chaulnes me pria instamment
d'aller à Vitré. J'y vins donc lundi au soir.»
Quatre jours après, elle écrit de nouveau de Vitré [88]: «Je suis
encore ici; M. et madame de Chaulnes font de leur mieux pour m'y
retenir; ce sont sans cesse des distinctions peut-être peu sensibles
pour nous, mais qui me font admirer la bonté des dames de ce pays-
ci; je ne m'en accommoderais pas comme elles, avec toute ma
civilité et ma douceur. Vous croyez bien aussi que sans cela je ne
demeurerais pas à Vitré, où je n'ai que faire. Les comédiens nous
ont amusés, les passe-pieds nous ont divertis, la promenade nous a
tenu lieu des Rochers. Nous fîmes hier de grandes dévotions... Je
meurs d'envie d'être dans mon mail. La Mousse et Marphise ont
grand besoin de ma présence.»
Les lettres que madame de Sévigné recevait de sa fille lui
apprenaient que la Provence ne se montrait pas aussi facile que la
Bretagne. «Vous me ferez aimer, lui dit-elle, l'amusement de nos
Bretons plutôt que l'indolence parfumée de vos Provençaux [89]»; et
elle mande à sa fille que M. d'Harouïs souhaite que les états de
Provence donnent à madame de Grignan autant que ceux de
Bretagne ont donné à madame de Chaulnes [90]. En effet, les états de
Bretagne firent à la duchesse de Chaulnes présent de deux mille
louis d'or, qui lui furent envoyés par une députation composée de
dix-huit membres, à la tête desquels étaient les évêques de Quimper
et de Nantes, chargés de la complimenter [91].
Madame de Sévigné parle de ces dons avec un ton ironique qui
décèle sa pensée: «On a donné cent mille écus de gratifications,
deux mille pistoles à M. de Lavardin, autant à M. de Molac, à M.
Boucherat, au premier président, au lieutenant du roi; deux mille
écus au comte des Chapelles, autant au petit Coëtlogon; enfin des
magnificences. Voilà une province [92]!» Oui; mais la Bretagne, mal
défendue par ses députés contre les exactions du pouvoir, se révolta
quatre ans après; et la Provence, sous la bénigne administration du
comte de Grignan, qui se ruina en la gouvernant, fut heureuse et
tranquille.
Madame de Sévigné est exacte pour les sommes données à
Lavardin, premier lieutenant général, pour des Chapelles et
Coëtlogon; mais elle se trompe pour M. de Molac, second lieutenant
général, qui n'eut que 25,000 liv. Le marquis de Lavardin eut, en
outre des 25,000 liv., 16,000 liv. pour ses gardes et officiers; le duc
de Chaulnes, gouverneur, eut 100,000 liv., et 20,000 liv. pour ses
gardes et officiers; le duc de Rohan eut 22,000 liv.; l'évêque de
Rennes eut la même somme, et le premier président 20,000 liv. De
Colbert, intendant de Bretagne, reçut 9,000 liv.; le marquis de
Louvois, grand maître et surintendant des forêts, 8,000 liv., et tous
les autres à proportion [93].
En accordant tout ce qui leur était demandé, les états firent des
remontrances tendant à faire révoquer plusieurs édits nuisibles à la
province; mais les réponses furent faites aux états tenus deux ans
après, en 1673: elles prouvent que ces remontrances furent
illusoires. Cependant quelques-unes sont des espèces de
protestations contre certaines dispositions des édits royaux, qu'on
affirme être contraires aux coutumes de la province. Pour toutes les
demandes de cette nature, le roi promet de se faire informer de ces
coutumes: il semble ainsi reconnaître qu'il veut les respecter [94].
Les assises des états furent terminées le 5 septembre. Madame de
Sévigné, en annonçant à sa fille cette fin dans sa lettre datée de
Vitré le lendemain, s'exprime ainsi [95]: «Les états finirent à minuit;
j'y fus avec madame de Chaulnes et d'autres femmes. C'est une
très-belle, très-grande et très-magnifique assemblée. M. de Chaulnes
a parlé à tutti quanti avec beaucoup de dignité, et en termes fort
convenables à ce qu'il avait à dire. Après dîner, chacun s'en va de
son côté. Je serai ravie de retrouver mes Rochers. J'ai fait plaisir à
plusieurs personnes; j'ai fait un député, un pensionnaire; j'ai parlé
pour des misérables, et de Caron pas un mot [96], c'est-à-dire, rien
pour moi; car je ne sais point demander sans raison.»
On voit que madame de Sévigné désapprouvait les prodigalités des
états; mais son texte, pour ce qui la concerne, a besoin d'une
explication, qui n'a jamais été donnée.
La terre de Sévigné [97] avait été démembrée, ou avait depuis
longtemps cessé d'être la principale possession de la famille de ce
nom [98]. Cette famille possédait la seigneurie des Rochers depuis le
milieu du XVe siècle, par le mariage d'Anne de Mathefelon, fille et
héritière de Guillaume de Mathefelon, seigneur des Rochers, avec
Guillaume de Sévigné. Mais il restait à la famille de Sévigné de ses
anciennes possessions une terre de Sévigné près de Rennes, dans la
commune de Gevezé, consistant en deux métairies, en moulins et
quelques fiefs, dont la valeur totale est estimée par le fils de
madame de Sévigné à 18,000 livres (36,000 fr.), tandis qu'il porte le
prix de la terre des Rochers à 120,000 liv. (240,000 fr.) [99]. Parmi les
fiefs restés à la famille de Sévigné, était, dans la ville de Vitré, une
maison avec cour et jardin, qu'on appelait la Tour de Sévigné. Cette
maison était un fief qui relevait du duc de la Trémouille, baron de
Vitré [100]. Par acte passé le 2 septembre 1671 (trois jours avant la fin
des états), madame de Sévigné fit une rente de cent francs aux
bénédictins de Vitré, et hypothéqua cette rente ou pension sur la
Tour de Sévigné [101].
Ce don fut sans doute fait en reconnaissance des réparations
exécutées aux frais de la province à la grosse tour qui donnait son
nom à la maison de Vitré. Voilà pourquoi elle dit, «J'ai fait un
pensionnaire,» et qu'en même temps elle avance qu'elle n'a rien
demandé, parce que la demande qu'elle avait formée ne pouvait
souffrir aucune difficulté, puisque cette grosse tour était engagée
dans les fortifications de la ville, et en faisait partie. M. le duc de
Chaulnes, qui voulait faire venir à Vitré madame de Sévigné, prit ce
prétexte pour la forcer à quitter son château des Rochers: il fit la
plaisanterie de l'envoyer chercher par ses gardes, en lui écrivant
qu'elle était nécessaire à Vitré pour le service du roi, attendu qu'il
fallait qu'elle donnât des explications sur la demande qu'elle faisait
aux états; et qu'en conséquence madame de Chaulnes l'attendait à
souper [102].
C'est dans cet hôtel de la Tour de Sévigné que demeurait la brillante
marquise lorsqu'elle restait à Vitré. Cette année, elle en laissa la
jouissance à son fils, qui y donnait à souper à ses amis [103]. C'est
aussi dans cette maison qu'allèrent loger, lorsqu'ils arrivèrent à Vitré
pour la tenue des états, de Chesières, l'oncle de madame de
Sévigné, son parent d'Harouïs, et un député nommé de Fourche [104].
Lorsqu'elle y restait, elle était accablée de visites. «Hier, dit-elle, je
reçus toute la Bretagne à ma Tour de Sévigné [105].» Mais lorsque les
états furent terminés, que le duc de Chaulnes fut parti, elle n'alla
plus à Vitré. Son fils l'avait quittée depuis longtemps, et bien avant la
fin des états, où son âge ne lui permettait pas d'être admis. Quoique
l'été fût constamment froid et pluvieux [106], madame de Sévigné
resta aux Rochers, pour que l'abbé de Coulanges pût surveiller les
travaux de la chapelle [107], et pour avoir le temps de terminer les
embellissements de son parc [108]. Elle avait envie d'aller visiter une
autre terre qu'elle possédait en Bretagne, près de Nantes, nommée
le Buron; mais, dit-elle, «notre abbé ne peut quitter sa chapelle; le
désert du Buron et l'ennui de Nantes ne conviennent guère à son
humeur agissante [109].» Madame de Sévigné se soumet, et ne va pas
au Buron.
Cette terre, à quatre lieues de Nantes, avait un château ancien, mais
bien bâti [110]. Le marquis de Sévigné en fit abattre les arbres
séculaires qui en faisaient tout l'agrément [111], et ce beau domaine
fut ensuite dégradé et ruiné par un administrateur infidèle ou
inintelligent, et par un fermier de mauvaise foi [112].
Il n'en fut pas ainsi des Rochers, que madame de Sévigné ne cessa
jamais d'accroître et d'embellir, et qu'elle vint si souvent habiter [113].
La construction de la chapelle, de forme octogone, surmontée d'une
coupole, située au bout du château et isolée, fut achevée en cette
année 1671; mais ce ne fut qu'après quatre ans que l'intérieur fut
entièrement en état, et qu'on put enfin y célébrer la messe, pour la
première fois, le 15 décembre 1675. A cette époque si froide de
l'année, madame de Sévigné se promenait avec plaisir dans ses bois,
plus verts que ceux de Livry, et augmentés de six allées charmantes,
que madame de Grignan ne connaissait point [114]. Depuis, ce
nombre d'allées fut presque doublé [115].
Madame de Sévigné avait multiplié dans son parc les inscriptions
morales, religieuses et autres, presque toujours tirées de l'italien.
Sur deux arbres voisins elle avait inscrit deux maximes contraires:
sur l'un, La lontananza ogni gran piaga salda (L'absence guérit les
plus fortes blessures); sur l'autre, Piaga d'amor non si sana mai
(Blessure d'amour jamais ne se guérit). Une des plus heureuses
inscriptions fut sans doute ce vers du Pastor fido, qu'elle avait fait
graver au-dessus d'une petite fabrique placée au bout de l'allée de
l'Infini, afin de se garantir de la pluie:
Di nembi il cielo s'oscura indarno [116].
Une autre allée, nommée la Solitaire, longue de douze cents pas, fut
plantée plus tard, et madame de Sévigné s'en enorgueillit comme de
la plus belle [117]. Elle avait fait construire dans différents endroits du
parc un assez grand nombre de petites cabanes qu'elle appelle des
brandebourgs [118], pour lire, causer et écrire à son aise, à l'abri du
soleil, du serein, et surtout de la pluie. Quant à son mail, dont elle
parle si souvent, c'est pour elle une belle et grande galerie, au bout
de laquelle on trouvait la place Madame, d'où, comme d'un grand
belvéder, la campagne s'étendait à trois lieues, vers une forêt de M.
de la Trémouille (la forêt du Pertre). Elle n'est pas moins engouée de
son labyrinthe, que son fils aimait par-dessus tout, et où nous
apprenons qu'il se retirait souvent avec sa mère pour lire ensemble
l'Histoire des variations de l'Église protestante, de Bossuet [119]. Mais
ce fut seulement vingt-sept ans après avoir été commencé, vers la
fin de l'année 1695, que madame de Sévigné, alors à Grignan, apprit
de son fils, qui était aux Rochers, que Pilois avait enfin terminé le
labyrinthe. Ainsi, les Rochers furent pour madame de Sévigné,
comme ses lettres, l'occupation de toute sa vie [120].
De son antique manoir, des constructions qu'elle avait ajoutées, des
ombrages qu'elle avait formés, il ne reste plus rien que la
chapelle [121], où le Christ est toujours invoqué, et l'écho de la place
de Coulanges, qui répète encore le nom de madame de Sévigné [122].
CHAPITRE II.
1671.
ebookmasss.com