
3
For the sake of clarity, these two features are kept sep-
arate, and investigated independently from one another
in the two models.
Finally, in deriving the response functions, we as-
sume that the system dynamics is triggered by a Franck-
Condon transition between the electronic states, induced
by the interaction with the external electric field. This is
followed by a free evolution of the system, resulting from
the interplay between the electronic and vibrational de-
grees of freedom.
FIG. 2. Second model system (B), which includes two, non-
adiabatically coupled excited electronic states |1iand |2i.
These correspond to the excitation respectively of the first
and second monomer that form a dimer. Transitions between
the ground (g) and excited state (e) of each monomer can
be induced optically (green arrows). The excitation of each
monomer results in a displacement by −zeof the correspond-
ing vibrational mode (harmonic oscillator).
III. MAIN RESULTS
The central result of the present article is represented
by the time propagators between the excited states be-
longing to the subspace Se. This, result is then used
to derive the expressions of the linear and nonlinear re-
sponse functions. In the following, we present a brief
discursive description of the method (Subsec. III A), fol-
lowed by the presentation of the final expressions (Sub-
secs. III B–III E). The formal derivations of the results
are presented in Section IV.
A. Brief description of the method
The relevant time propagator is the matrix element
of the time evolution operator between the states |σ; 0i
and |σ0; 0i: these are given by the product of the diabatic
electronic states that are coupled by the non-adiabatic in-
teraction V(σ, σ0= 1,2), and of the vibrational ground
state of the undisplaced harmonic oscillator. The time-
evolution operator is computed by performing a Dyson
expansion with respect to V: each term in the expan-
sion corresponds to an electronic pathway,i.e. to a given
sequence of electronic states e1, . . . , eM(an alternating
sequence of |1iand |2i), being M−1 the order of the
expansion. The overall time evolution of the system that
one can associate to each electronic pathway consists of
a sequence of sudden transitions between the two dia-
batic states, interleaved by time intervals during which
the system remains in the same electronic state (Fig. 3).
For the vibrational state, each transition between the
states |1iand |2iimplies a hopping of the coherent state
from one potential energy surface to the other, being
these two relatively displaced parabolas. The resulting
time evolution resembles that induced by sequences of
delta-like laser pulses within the linearly displaced har-
monic oscillator model23. This formal analogy allows us
to use in the present case the analytical expressions that
have recently been derived for the vibrational component
of the response function in the adiabatic case (R).
The following step consists in the integration over all
the possible values of the non-adiabatic interaction times.
In order to perform such integration analytically, we per-
form a Taylor expansion of R, which can be written as
the product of M(M+1)/2 double exponential functions.
Each term in the Dyson expansion (order M−1 in the
non-adiabatic coupling constant η) thus gives rise to a
number of infinite terms, one for each set of orders ki
(i= 1, . . . , M(M+ 1)/2) of the Taylor expansions. For-
mally, each of these terms can be written as a product of
exponential functions, that oscillate during the intervals
of duration τj(j= 1, . . . , M) with a frequency Ωj. This
is given by the sum of an electronic and a vibrational
contributions. The former corresponds to the energy ¯ωj
(~≡1) of the electronic state for the relevant time in-
terval (specified by j) and electronic pathway (specified
by M−1); the latter one is the energy qjωof the qj-th
eigenstate of the undisplaced harmonic oscillator. The
values of qjresult from those of klin a one-to-many cor-
respondence. Physically, one can thus associate to each
term of the Taylor expansion a vibronic pathway, defined
by a sequence of electronic and vibrational states |ej;qji,
with j= 1, . . . , M. Besides, each of these term is propor-
tional to the displacements (z1or z2) or their difference
to the power of kT=PM(M+1)/2
i=1 ki. Being the modulus
of the displacement typically smaller than one, this se-
ries is expected to converge, even though the number of
terms increases rapidly with kT.
These functions can be analytically integrated, and
give a formally simple result, consisting - for each vi-
bronic pathway - in the sum of Mterms, each one os-
cillating at a frequency Ωj. If all these frequencies differ
from one another, the oscillating terms e−iΩjtare multi-
plied by constants Aj. If kof those frequencies coincide,
then each of the multiple e−iΩjtis multiplied by a mono-
mial ajtrj, with rj= 0, . . . , k −1 (it follows from the
calculations that the number of identical frequencies for