Chemical Physics Letters 592 (2014) 1823

Contents lists available at ScienceDirect

Chemical Physics Letters

journal homepage: www.elsevier.com/locate/cplett

Hybrid quantumclassical study of the non-adiabatic cistrans

photoisomerization in a model polyatomic molecule
F. Rodrguez-Hernndez, A. Martnez-Mesa , L. Uranga-Pia
Department of Theoretical Physics, Faculty of Physics, University of Havana, San Lzaro y L, Vedado 10400, Havana, Cuba

a r t i c l e i n f o a b s t r a c t

Article history: The non-adiabatic dynamics of cistrans photoisomerization in polyatomic systems is investigated using
Received 2 September 2013 a model Hamiltonian, which exploits the approximate separation of the photochromic unit from the rest
In nal form 21 November 2013 of the molecular degrees of freedom. The dynamical response of the active modes upon photoexcitation is
Available online 9 December 2013
described using wavepacket propagation while the remaining modes are treated classically, explicitly
taking into account the coupling between both subsystems via the symplectic PICKABACK algorithm
[1]. The model enables a microscopic description of the non-radiative excited-state lifetime and the isom-
erization yield, as well as the time evolution of the congurational rearrangement and the internal energy
redistribution.
2013 Elsevier B.V. All rights reserved.

1. Introduction plished by introducing a rigorous approach to the absorption pro-

Cis-to-trans photoisomerization represents a paradigmatic
example of a dynamical process driven by a conical intersection
[27] and it plays an important role in several phenomena in phys-
ics, chemistry and biology. For example, it is a key step in both
photosynthesis and vertebrates vision [815]. Likewise, it is at
the heart of the design of novel nanoscale devices such as molecu-
lar switches, sensors and motors [1625], owing to the distinct
macroscopic chemical and physical properties that correspond to
each isomer [2629]. From the fundamental standpoint, the study
of photoisomerization provides considerable inside into the mech-
anisms of translation of an electronic excitation into a structural
change in the molecule.
Regarding the theoretical modelling of these photochemical
reactions, the intrinsic many-body character of these processes
makes numerically converged calculations of the isomerization
rates difcult to achieve. Photoisomerization yields are chiey
determined by the nuclear dynamics near conical intersections,
thereby ruling out the possibility to carry out a quantum treatment
of the molecular dynamics within the BornOppenheimer approx-
imation and making studies of the time evolution of this process
very challenging. Fortunately, very often there is a clear separation
between the chromophore, which actively responds to photoexci-
tation, and the rest of the system. This separation enables to apply,
to the analysis of photoisomerization, the nowadays available the-
oretical and computational tools for the accurate study of the
quantum dynamics of small molecular systems. This is accom-
Corresponding author.
E-mail address: aliezer@sica.uh.cu (A. Martnez-Mesa).
cess and the structural modication along active modes, while
approximations are made in order to simplify the integration of
the equations of motion for the remaining degrees of freedom.
A possible route to undertake the computer simulation of cis
trans photoisomerization is the introduction of hybrid quantum
classical integration schemes. The modication of the dynamical
state of a many-body system, according to the classical equations
of motion, is local with respect to phase space variables. Therefore,
the classical hypothesis leads to a signicant saving in the storage
requirements and the computational cost of such calculations,
with respect to the corresponding quantum treatment, making fea-
sible the computation of nuclear trajectories (and the subsequent
analysis of molecular restructuring) beyond the picosecond time
scale.
Since the quantum character of the isomerization process can
not be completely neglected, its numerical investigation relies, at
present, on the accuracy of the mixed quasi-classical approaches.
As a consequence, new models are continuously developed and
tested [3035]. Moreover, elaborate algorithms (e.g., the Multi-
Congurational Time-Dependent Hartree (MCTDH) [36,37], the
Full Multiple Spawning [38], the Surface-Hopping with Arbitrary
couplings [39] and the Field-Induced Surface Hopping [40,41]
methods) has been applied in order to get a deeper insight into
non-adiabatic dynamics in polyatomic systems.
Simple models are specially valuable as they provide intuitive
representations of reaction mechanisms and they constitute ideal
test grounds for new computational approaches. The purpose of
this Letter is to carry out the study of non-adiabatic cistrans
photoisomerization of a prototype polyatomic molecule, within
the framework of quantumclassical molecular dynamics, for a

0009-2614/$ - see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.cplett.2013.11.062
 F. Rodrguez-Hernndez et al. / Chemical Physics Letters 592 (2014) 1823
two-state two-mode model previously introduced in the literature
[42]. To this aim, the simultaneous resolution of the time-depen-
dent Schrdinger equation, along the active modes of the chromo-
phore, and the canonical equations of motion, for the classical
subsystem, are tackled via the symplectic algorithm introduced
in Ref. [1]. The availability of previous theoretical results, from
quasi-classical and reduced density matrix integration schemes
[43], provides an ideal benchmark to assess the performance of this
integration scheme in this specic problem. Moreover, the results
reported in Ref. [43] are complemented by the analysis of the
dynamical state of the bath.
The Letter is organized as follows. In Section 2, the theoretical
methodology employed to study the time-evolution of the nuclear
wavepacket of the active subsystem and of the dynamical state of
the remaining modes, as well as the relevant intermolecular inter-
actions, are described. The results concerning the cistrans isomer-
ization rate, the population of electronic states, together with the
probability distributions along the isomerization coordinate, are
presented and discussed in Section 3. Finally, in Section 4 some
conclusions are drawn.
2. Methodology
2.1. Model Hamiltonian
Following Refs. [4246], we assume that a two-state model
involving two nuclear modes is capable to account for the essential
aspects of the isomerization process. The model Hamiltonian
Hchrom: , in the diabatic representation, in terms of the two lowest
electronic states, jw0 iand jw1 i, is written as follows
X   sical propagation scheme:
Hchrom: jwn i T dnm V nm hwm j; 1
n;m0;1
where V nm is the diabatic potential matrix. The dependence of the
matrix elements of the kinetic energy operator T and of the diabatic
potential energy surfaces and couplings, on the dimensionless
(mass-weighted) variables q and u, has been described elsewhere
[42]. The vibrational coordinate q effectively accounts for all the
normal modes of oscillation in the chromophore, while the reaction
coordinate u parametrises the twist of the molecular fragment cor-
responding to isomerization. Assessments on the ability of the
Hamiltonian (1) to capture the essentials of the geometrical rear-
rangement of specic chromophores has been reported [7,47,48].
The potential energy surfaces along the rotational and the
deformation coordinates, as well as a schematic representation of
Figure 1. Potential energy surfaces, along the q and u coordinates, corresponding
to the two lowest electronic states. The vibrational ground state of the chromo-
phore, located around the equilibrium position corresponding to the cis geometry, is
also shown. The arrows indicate the initial (vertical) excitation and the alternative
paths leading to the product cis and trans isomers.
19
the photoisomerization process are presented in Figure 1. In this
gure, the interval  p2 6 u 6 p2 correspond to the cis conguration
while the range p2 6 u 6 32p correspond to the trans geometry. The
wavepacket created by the laser pulse excitation evolves on the
S1 potential energy surface until it reaches the conical intersection
at u p2 . At this point, the system may undergo isomerization or it
may return to the initial conguration.
The remaining N degrees of freedom, corresponding to the
vibrations of particles lying outside the region that plays an active
role in the absorption process, are represented by a bath of har-
monic oscillators. The dynamics of the bath oscillators is modeled
classically and it is governed by the Hamiltonian
X xa  
^ bath
H x2a p2a ; 2
2
a
xa and pa are, respectively, the dimensionless position and momen-
tum coordinates corresponding to the a oscillation mode. Further-
more, the coupling Hcoupl: between the chromophore and the bath
is chosen as in Eqs. (5) to (7) of Ref. [43], with symmetric cis and
trans coupling parameters wC 1  wT 12. The coupling constants
g ua and g qa are chosen according to an Ohmic spectral bath density.
The values of the parameters appearing in the different terms of the
total Hamiltonian
H Hchrom: Hbath Hcoupl: ; 3
were taken from the literature [43].
2.2. Hybrid quantumclassical calculations
There are two main assumptions leading to the quantumclas-
(i) the factorization of the total time-dependent wavefunction
in two components: one corresponding to the subset of the
most important degrees of freedom, Wq; u; t, and another
one for the bath modes;
(ii) the classical modelling of the dynamics of the bath
oscillators.
Based on these two approximations, the dynamical state of the
system evolves according to the coupled set of equations:
@ Wq; u; t h ^ i
^ coupl xt Wq; u; t;
i Hchrom: H 4
@t
@Hbath
x_ ; 5
@p
@Hbath @ D E
^ coupl: j W :
p_   WjH 6
@x @x
The vectors x x1 ; x2 ; . . . ; xN and p p1 ; p2 ; . . . ; pN span the
full phase space of the harmonic bath.
Hereafter, atomic units are chosen ( h 1). The symplectic PICK-
ABACK scheme [1] allows to update the states of the quantum and
classical subsystems, explicitly taking into account the interaction
between them. As in previous studies [43], the discretized thermal
bath is built by 200 harmonic oscillators, with homogeneously dis-
tributed frequencies in the range 0 < xa < 0.3 eV. An ensemble of
up to N copies 5000 copies for each oscillator were used in the cal-
culation of the observables, although it was found that the choice
of N copies J 100 trajectories already yield converged results for the
observables presented in Section 3. It is not surprising that conver-
gence can be achieved for such a relatively small number of quasi-
classical trajectories, since the motion along the subset of degrees
of freedom undergoing major changes is treated independently on
every diabatic surface. The initial conditions of the copies sample
20 F. Rodrguez-Hernndez et al. / Chemical Physics Letters 592 (2014) 1823
the quantum distribution, in position and momentum spaces, cor-
responding to the ground state of each oscillator.
For the quantum subsystem, we need to solve the coupled time-
dependent Schrdinger Eq. (4). In the diabatic representation, the
wavefunction reads
X
Wq; u; t /n q; u; tjwn i: 7 Z
n0;1
It is worth to notice that the present approach deviates from the
standard time-dependent self-consistent eld (or Ehrenfest) meth-
od for non-adiabatic dynamics. Indeed, it overcomes the main lim-
itation of the latter, namely the incapability to cope with the
distinct dynamical behaviour on every potential energy surface,
by treating the degrees of freedom which play a key role in the
molecular restructuring quantummechanically, within a multi-
state expansion. Since there are separated wavepackets /n q; u; t
to represent the reorganization of the chromophore on each elec-
tronic state, those nuclear modes (i.e., those being the most impor-
tant ones in the description of the isomerization dynamics)
experience a state-specic potential, instead of a weighted average
of the different contributions.
However, the set of dynamical variables which do not take part
directly in the geometrical rearrangement of the molecule (that is,
those grouped in the classical bath) will still suffer the usual limi-
tations of mean-eld approaches. Regardless, based on the results
of benchmark simulations on the absorption spectrum of a pyra-
zine molecule using a model Hamiltonian [37], this approximation
is expected to introduce minor deviations from the exact dynam-
ics, compared to the assumption of the factorizability of the molec-
ular wavefunction in a product of wavefunctions for the system
modes and the bath modes. In Ref. [37], for instance, the energy
transferred by the system to a Hartree bath was found to oscillate
around the predictions obtained for the so-called independent-
state Hartree scheme. As it will be discussed bellow, the long-term
uctuations of the energy of the bath are rather small here. Conse-
quently, no dramatic effects on the photoisomerization dynamics
are anticipated to emerge from the evaluation of the forces exerted
on the bath modes within the mean-eld approximation.
Since the laser excitation pulse is assumed to be instantaneous,
the initial wavepacket promoted to the excited electronic state is
the exact copy of the ground state wavefunction corresponding
to the (separable) potential V 00 q; u. For the vibrational ground
state, the dependence of the initial wavefunction /0 q; u; t 0
on the vibrational coordinate q is known analytically. The depen-
dence of the two-dimensional wavefunction /0 q; u; t 0 on the
isomerization angle u is obtained using a Numerovs nite-differ-
ence scheme [49].
The initial wavepacket is evolved in time by introducing a -
nite-difference representation of the Hamiltonian in a two-dimen-
sional grid. The number of points along the q and u directions of
the uniform mesh was set to 1024. The split-operator technique
was used to represent the evolution operator, in conjunction with
Cayleys form of the matrix exponential [50] to compute the action
of the kinetic operator on the state vector. Albeit more sophisti-
cated representations of the Hamiltonian exponential exist, e.g.,
the Chebyshev polynomial expansion, the total calculation time
is not a critical issue here and the Caleys form sufces for the pres-
ent purposes. Nevertheless, the more elaborate methods might
lead to a signicant reduction of the computational cost associated
with attaining a prescribed accuracy in problems where realistic
intermolecular potentials are involved.
Both quantum and classical equations of motion were inte-
grated using a time-step of D t = 0.1 a.u.. A total of 2:5  106 updates
of the molecular state x; p; Wq; u; t were carried out, equivalent
to a time interval of about 6 ps after photoexcitation. From the
time-dependent wavefunction, the diabatic populations Pdiab n t,
the probability density along the isomerization coordinate
qu; t, and the populations corresponding to the cis and trans con-
gurations,P cis t and Ptrans t, were computed after the formulas:
Z
Pdiab
n t dqdujhwn jWq; u; t j2 ; 8
qu; t dqjWq; u; t j2 ; 9
Z p=2
Pcis t duqu; t; Ptrans t 1  P cis t: 10
p=2
It should be stressed that the denition (9) implies that contri-
butions of both the n 0 and the n 1 diabatic states are taken
into account in qu; t.
On the other hand, the irreversibility of the energy transfer is
monitored through the average energy of the oscillators in the
bath.
3. Results and discussion
3.1. Dynamical state of the bath
Owing to the large number of bath modes, compared to those of
the chromophore, the redistribution of the excess energy deposited
by the laser will result in the cooling of the latter. One important
aspect to be checked is whether the energy transfer is unidirec-
tional, i.e., if it ows preferentially out of the quantum subsystem
into the set of harmonics oscillators. In Figure 2, the time evolution
of the total energy corresponding to the bath modes is represented
(only the rst 3 ps after the initial excitation are shown). As it was
mentioned in the previous Section, the data correspond to calcula-
tions carried out where each vibrational mode was represented by
a swarm of 5000 trajectories.
The overall behaviour conrms the absorption of the excess en-
ergy by the classical subsystem. As it is shown in the gure, there is
an initial increase of the bath energy by approximately 0.03 eV tak-
ing place during the rst 20 fs upon photoexcitation, followed by a
slower energy uptake (amounting to 0.015 eV in addition) over a
period of about 580 fs. The sudden increase of baths energy by
0.017 eV, happening at 600 fs, is followed by a further rise lasting
about 200300 fs. The latter takes place at an approximately linear
rate. After 1 ps, the bath have absorbed about 0.4 meV per mode on
average. Since the drop in the average energy of the chromophore
is governed by the system-bath coupling H ^ coupl: , which is chosen to
be position-dependent, its course before and after t = 600 fs sug-
gests that a marked restructuring of the molecule occurs at this
time. This reorganization will be conrmed by the examination
0.1
0.08
Ebath [eV]
0.06
0.04
0.02
0
0 1 2 3
t [ps]
Figure 2. Time evolution of the total energy of the harmonic bath during the rst
3 ps of the relaxation dynamics. The (average) asymptotic energy of the bath modes
is indicated by the dotted line.
 F. Rodrguez-Hernndez et al. / Chemical Physics Letters 592 (2014) 1823
of the time-dependent probability distribution associated to the
wavepacket Wq; u; t. From 1 ps onwards, it ceases the net energy
drift from the chromophore to the bath. It is replaced by a plateau-
like behaviour, where only relatively small uctuations (5%) oc-
cur around the asymptotic energy of the bath modes.
In Figure 3, the change in the distribution of vibrational ener-
gies of the bath oscillators is illustrated by comparing the energy
density of the bath for two selected points in time (t = 0 ps and
t = 3 ps). It can be seen that there are not major changes in the en-
ergy distribution, which is consistent with the moderate value of
the average energy uptake per mode. Indeed, the two curves are
identical for oscillation energies above 1 eV. Since the assignment
of the initial positions and momenta corresponds to the vibrational
ground state of every oscillator, the analysis of these curves also re-
veals that the energy transfer takes place preferentially for the
low-frequency vibrational modes.
3.2. Quantum subsystem
In Figure 4 it is shown the time-evolving population corre-
sponding to the cis geometry (the population of the trans isomer
mirrors this curve), during the rst 3 ps after photoexcitation.
The initial value Pcis(t = 0) = 1 follows from the initial conditions
for the wavepacket, centred around the potential energy minimum
of the cis conguration. Upon photoexcitation, the geometrical
rearrangement of the molecule takes place in two stages. During
the rst 90 fs, a fast decrease of the probability P cis t occurs. After
attaining its minimal value, it begins a gradual recovery of the cis
population. This increase occurs in a much larger time scale: the
cis and trans populations reach their asymptotic limits in about
1 ps. This early phase is characterized by strong recurrences, which
get continuously damped until the equilibrium is reached. The
asymptotic cistrans conversion efciency, P trans t ! 1  0:5, is
larger than the value of 0.4 reported by previous calculations car-
ried out using quasi-classical trajectories and a Redeld reduced
density matrix approach [43]. While the amplitude of the uctua-
tions of the cis and trans populations in Figure 4 matches that ob-
tained in Ref. [43], it becomes striking that their stabilization tends
to occur faster here. Nevertheless, comparing the corresponding
curves, an overall qualitative agreement is found between the var-
ious methods.
A more microscopical understanding of the time evolution of
the molecular rearrangement is provided by the probability distri-
bution qu; t along the reaction coordinate u. This probability
density is shown in Figure 5 for selected points in time in the inter-
val 0 6 t 6 110fs. It can be noticed how, during the rst 30 fs, the
wavefunction describing the displacement in the diabatic state
n 1 gets progressively broader, under the inuence of the V 11 po-
t = 0 ps
t = 3 ps
n(E)
0 0.25 0.5 0.75
E [eV]
Figure 3. Vibrational energy distribution nE of the bath oscillators for t = 0 ps and
for t = 3 ps. The ordinate scale is in arbitrary units.
21
1
Isomeric conformations
Pcis(t)
0.75 Ptrans(t)
0.5
0.25
0
0 1 2 3
t [ps]
Figure 4. Time evolution of the probability of the cis conguration P cis . The time-
dependent population of the trans isomer, P trans (dashed line), mirrors this curve.
tential energy surface, while it keeps its nearly Gaussian shape. At
t = 30 fs, the right front of the wavepacket reaches the position of
the conical intersection, giving rise to transitions between the elec-
tronic states, as it can be seen in the right panel. This moment also
marks the beginning of the transfer of population to the trans
geometry.
Likewise, the slight enhancement of the density qu; t 35 fs
at u 0 is a ngerprint of the onset of the non-adiabatic transi-
tions. The explanation is as follows: since the initial wavefunction
is an eigenstate of the Hamiltonian T^ V ^ 00 and, in the beginning,
the contribution of the non-adiabatic coupling terms is negligible,
then the density distribution corresponding to the fraction of the
population in the diabatic state n 0 remains fairly unaffected at
short times. It makes, however, a vanishing contribution to the
molecular wavefunction, since the initial population has being
fully promoted to the excited electronic state by the pump pulse.
Once the wavepacket arrives to the vicinity of the conical intersec-
tion (about 30 fs after the photoexcitation), the state n 0 gets re-
populated and the time-dependent wavefunction becomes a
superposition of the ground and the excited state wavepackets
/n q; u; t. For times t J 30 fs, the total wavefunction closely
resembles the component /1 q; u; t (the wavefunction Wq; u; t
is only perturbed on a small scale by the population transfer)
and the corresponding distribution essentially matches that of
the n 1 electronic state, except for a tiny contribution coming
from the population in the diabatic state n 0, which is maximal
at u 0 (see the appropriate probability distributions in Figure 5).
From 30 fs onwards, there is an increasingly larger fraction of
the wave-packet lying in the neighbourhood of the point of degen-
eracy, thereby strengthening the rate at which electronic transi-
tions take place. After 110 fs, as it is shown in the right panel of
Figure 5, as a consequence of these transitions, the diabatic state
V 00 has been repopulated up to a signicant extent
(Pdiab
0  Pdiab
1  0:5). The simultaneous population of the two dia-
batic states is reected by the further splitting of the nuclear wave-
packet into two parts which move away from the conical
intersection in opposite directions, i.e., one part going back to the
equilibrium conguration of the cis isomer while the other contin-
ues to approach the equilibrium position corresponding to the
trans geometry. Albeit exhibiting a comparatively large probability
of being around the conical intersection, the electronic transitions
tend to balance beyond 600 fs. Hence, after that moment, the pos-
sibility of a net population transfer between the electronic states to
occur is largely reduced and the displacement along the isomeriza-
tion coordinate slows down, leading to the stabilization of the cis
and trans populations presented above.
Interestingly, at short times, the diabatic populations shown in
Figure 5 agree qualitatively with the previous simulations carried
1
out for the same model, based on Redeld theory [43]. In particu-
lar, within the subpicosecond time scale, the quantumclassical
methodology employed in this Letter performs better than the
classical mapping formalism introduced in Ref. [43], despite the
22 F. Rodrguez-Hernndez et al. / Chemical Physics Letters 592 (2014) 1823

1
0 fs (,t) 30 fs 2(,t) diab
1 21 P0
0 20
P1
diab 0.8

0.6

(t)
diab
Pn
35 fs 2(,t) 110 fs 4(,t) 0.4
21 41
20 40

0.2

0
-0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5 0 0.2 0.4 0.6
[] [] t [ps]

Figure 5. Early time evolution of the probability density qu; t along the isomerization coordinate (left panel, the ordinate scales are in arbitrary units) and the population of
diabatic states P diab
n t; n 0; 1 (right panel). On the left, they are represented the initial density (0 fs) and the distributions corresponding at various times (30 fs, 35 fs and
110 fs) after the photoexcitation, together with the contributions q0 and q1 coming from the diabatic states n 0 and n 1, respectively. Some of them have being rescaled to
allow a better visualization, as indicated in the respective legends. The position of the conical intersection is indicated by the vertical dotted lines. On the right panel, the
strengthening of the nonadiabatic transition rates from 30 fs is indicated by the vertical dashed line.

Figure 6. Wavepacket dynamics (left panel): evolution of the time-dependent probability density along the isomerization coordinate qu; t following the photoexcitation of
the chromophore. On the right panel, it is shown the time evolution of the projections of the wavepacket Wq; u; t on a representative set of the vibrational eigenstates
localized in the trans well.

extent of quantum corrections brought into the latter. Still, the
amplitude of the recurrences is larger for the diabatic populations
computed employing the hybrid quantumclassical integration
scheme. Additionally, as it happens for the isomeric conformation
probabilities, the equilibrium population of the n 1 electronic
state, Pdiab
1 t ! 1  0:5, is 20% larger than the corresponding va-
lue obtained via reduced density matrix calculations. Nonetheless,
it reproduces the long-term behaviour of the electronic popula-
tions predicted by the classical mapping description in absence
of zero-point energy corrections [43]. Both factors, the overestima-
tion of the amplitude of recurrences compared to quantum
mechanical calculations, and the approaching of the asymptotic re-
sults of the classical (mean-eld) trajectory simulations, point to a
mild underestimation of the electronic relaxation rates by the pres-
ent integration scheme. In spite of this limitation, the present ap-
proach provides a good compromise between the accuracies of
the descriptions of the short- and long-time photoisomerization
dynamics.
A more detailed examination of the isomerization dynamics is
made in terms of the two-dimensional probability density
qu; t, depicted in Figure 6. At early stages, the distribution is cen-
tred at the position u 0, again reecting the initial conditions
corresponding to the excitation of the Wq; u; t 0 wavepacket
from the ground state wavefunction on the V 00 potential energy
surface. As would be expected, up to t 0.1 ps, this plot matches
the analysis made based on the one-dimensional cuts of qu; t
shown in Figure 5. For later times, the time evolution of the density
reects the second stage of the molecular rearrangement. After the
occurrence of a few fractional revivals of the wavepacket in the cis
well, consistent with the recurrences of the population P cis t ob-
served in Figure 4, the rather steady partitioning of the probability
density between the cis and trans geometries becomes striking.
From about 500600 fs after the initial excitation by the laser
pulse, there is a continuous drift of the population into the inner
regions of cis and trans potentials wells. Such behaviour is a conse-
quence of the increasing population of the less energetic vibra-
tional states of the chromophore, as more energy is released into
the bath. This is illustrated in the right panel of Figure 6, for a rep-
resentative set of vibrational states localized within the well corre-
sponding to the trans isomer, which undergo a slow but steady
increment of their occupation probability. Hence, the probability
distribution concentrated around the turning points, becomes
more spread out over time.
4. Conclusions
In this Letter, a numerical study of the dynamics of cistrans
photoisomerization in model polyatomic molecules, triggered by
 F. Rodrguez-Hernndez et al. / Chemical Physics Letters 592 (2014) 1823
the selective photoexcitation of the chromophore, have being pre-
sented. We investigate the structural modication of a generic
molecular system employing a model hamiltonian which describes
the chromophore as a quantum system with only two degrees of
freedom, while the rest of the generalized molecular coordinates
of the molecule are treated as a set of classical harmonic oscilla-
tors. Only the ground and the rst excited electronic states are con-
sidered to take part in the process, although the methodology
employed here is directly extensible to cases where more than
two electronic states are involved.
We analysed the time evolution of the wave packet that de-
scribes the dynamical state of the chromophore, via the integra-
tion of the time-dependent Schrdinger equation for the coupled
electronic states, and we simultaneously solved the Hamiltons
equations of motion for the degrees of freedom comprised in
the classical bath, explicitly taking into account the interaction
between both subsystems. The mixed quantumclassical integra-
tion scheme is based on the PICKABACK algorithm.
The main features of the mechanism of photoisomerization
are explained in terms of the probability density along the isom-
erization coordinate, i.e., the splitting of the wavepacket and the
electronic transitions induced by non-adiabatic couplings in the
vicinity of the degeneracy seam. For this model, the cis-to-trans
reaction yield is found to be about 0.5, while the isomerization
process does not ocurr in a single step. Instead, it proceeds in
two stages which occur in two different time scales: initially,
there is a rapid decrease of the population of the cis congura-
tion over a period of 90 fs, followed by a gradual recovery which
extents up to 1 ps. The early time evolutions of the populations
of the cis and trans isomers (and also those of the diabatic elec-
tronic states) exhibit rather large recurrences, which reects the
coherent dynamics of the nuclear wavepacket. The present ap-
proach appears to underestimate the electronic relaxation rates,
compared to previous density matrix calculations carried out
for the same model Hamiltonian, but the method provides a
good compromise between the accuracies of the descriptions
of the short- and long-time dynamics of cistrans
photoisomerization.
Both the early dynamics of the chromophore and its vibrational
relaxation at long times are associated with the energy drift from
the quantum mode to the bath of classical oscillators. The bath
takes approximately 0.4 meV per mode, the energy transfer to
low frequency modes being more efcient.
Summing up, from the comparison with previous reports on the
non-adiabatic dynamics of similar systems and owing to the uni-
versal features of the model employed, it can be concluded that
the present quantumclassical simulation strategy is capable to
describe the process of cis-to-trans photoisomerization in a satis-
factory manner. The methodology can be readily applied to poly-
atomic molecular systems interacting with realistic bath models,
where the coherent character of the nuclear dynamics or the large
number of degrees of freedom involved prevent, respectively, the
use of purely classical approaches or more rigorous schemes, like
the MCTDH package.
23
References
[1] P. Nettesheim, F.A. Bornemann, B. Schmidt, C. Schtte, Chem. Phys. Lett. 256
(1996) 581.
[2] F. Bernardi, M. Olivucci, M.A. Robb, Chem. Soc. Rev. 25 (1996) 321.
[3] M. Garavelli, P. Celani, F. Bernardi, M.A. Robb, M. Olivucci, J. Am. Chem. Soc.
119 (1997) 6891.
[4] W. Weber, V. Helms, J.A. McCammon, P.W. Langhoff, Proc. Natl. Acad. Sci. USA
96 (1999) 6177.
[5] M. Ben-Nun, T.J. Martnez, Chem. Phys. 259 (2000) 237.
[6] W. Domcke, D.R. Yarkony, H. Koppel (Eds.), Conical Intersections: Electronic
Structure, Dynamics and Spectroscopy, World Scientic, Singapore, 2004.
[7] B.G. Levine, T.J. Martnez, Annu. Rev. Phys. Chem. 58 (2007) 613.
[8] G. Wald, Science 162 (1968) 230.
[9] C.R. Goldschmidt, M. Ottolenghi, T. Rosenfeld, Nature 263 (1976) 169.
[10] Q. Wang, R.W. Schoenlein, L.A. Peteanu, R.A. Mathies, C.V. Shank, Science 266
(1994) 422.
[11] H. Luecke, B. Schobert, J.P. Cartailler, H.T. Richter, A. Rosengarth, R. Needleman,
J.K. Lanyi, J. Mol. Biol. 300 (2000) 1237.
[12] T. Okada, O.P. Ernst, K. Palczewski, K.P. Hofmann, Trends Biochem. Sci. 26
(2001) 318.
[13] J.E. Kim, M.J. Tauber, R.A. Mathies, Biochemistry 40 (2001) 13774.
[14] S. Hayashi, E. Tajkhorshid, K. Schulten, Biophys. J. 85 (2003) 1440.
[15] D. Polli et al., Nature 467 (2010) 440.
[16] A.B. Cubitt, R. Heim, S.R. Adams, A.E. Boyd, L.A. Gross, R.Y. Tsien, Trends
Biochem. Sci. 20 (1995) 448.
[17] C. Joachim, J.K. Gimzewski, A. Aviram, Nature 408 (2000) 541.
[18] B. Feringa, Molecular Switches, Wiley, Weinheim, 2001.
[19] A.J. Myles, N.R. Branda, Adv. Funct. Mat. 12 (2002) 167.
[20] A. Nitzan, M.A. Ratner, Science 300 (2003) 1384.
[21] H. Murakami, A. Kawabuchi, R. Matsumoto, T. Ido, N. Nakashima, J. Am. Chem.
Soc. 127 (2005) 15891.
[22] M.N. Chatterjee, E.R. Kay, D.A. Leigh, J. Am. Chem. Soc. 128 (2006) 4058.
[23] D. Gust, T.A. Moore, A.L. Moore, Chem. Commun. (2006) 1169.
[24] W.R. Browne, B.L. Feringa, Annu. Rev. Phys. Chem. 60 (2009) 407.
[25] A.A. Beharry, G.A. Woolley, Chem. Soc. Rev. 40 (2011) 4422.
[26] C. Zhang, M.H. Du, H.P. Cheng, X.G. Zhang, A.E. Roitberg, J.L. Krause, Phys. Rev.
Lett. 92 (2004) 158301.
[27] L.Y. Zhu, M.Q. Zhu, J.K. Hurst, A.D.Q. Li, J. Am. Chem. Soc. 127 (2005) 8968.
[28] M.Q. Zhu, L.Y. Zhu, J.J. Han, W.W. Wu, J.K. Hurst, A.D.Q. Li, J. Am. Chem. Soc. 128
(2006) 4303.
[29] J.M. Mativetsky, G. Pace, M. Elbing, M.A. Rampi, M. Mayor, P. Samori, J. Am.
Chem. Soc. 130 (2008) 9192.
[30] U. Muller, G. Stock, J. Chem. Phys. 107 (1997) 6230.
[31] S. Wilsey, K.N. Houk, J. Am. Chem. Soc. 124 (2002) 11182.
[32] M. Barbatti, J. Paier, H. Lischka, J. Chem. Phys. 121 (2004) 11614.
[33] M. Barbatti, M. Ruckenbauer, H. Lischka, J. Chem. Phys. 122 (2005) 174307.
[34] V.A. Morozov, Opt. Spectr. 103 (2007) 614.
[35] T. Montagnese, F. Ogliaro, S. Wilsey, M.J. Bearpark, J. Phys. Chem. A 112 (2008)
13239.
[36] M.H. Beck, A. Jckle, G.A. Worth, H.-D. Meyer, Phys. Rep. 324 (2000) 1.
[37] G.A. Worth, H.-D. Meyer, L.S. Cederbaum, J. Chem. Phys. 105 (1996) 4412.
[38] A. Toniolo, S. Olsen, L. Manohar, T.J. Martinez, Faraday Disc. 127 (2004) 149.
[39] M. Richter, P. Marquetand, J. Gonzalez-Vazquez, L. Gonzalez, J. Chem. Theory
Comp. 7 (2011) 1253.
[40] R. Mitric, J. Petersen, M. Wohlgemuth, U. Werner, V. Bonacic-Koutecky, Phys.
Chem. Chem. Phys. 13 (2011) 8690.
[41] R. Mitric, J. Petersen, U. Werner, V. Bonacic-Koutecky, in: Advances in the
Theory of Quantum Systems in Chemistry and Physics 22 (2012) 299.
[42] L. Seidner, W. Domcke, Chem. Phys. 186 (1994) 27.
[43] I. Uspenskiy, B. Strodel, G. Stock, Chem. Phys. 329 (2006) 109.
[44] S. Hahn, G. Stock, J. Phys. Chem. B 104 (2000) 1146.
[45] B. Balzer, G. Stock, Chem. Phys. 310 (2005) 35.
[46] X. Chen, V.S. Batista, J. Photochem. Photobiol. A Chem. 190 (2007) 274.
[47] R. Gonzalez-Luque, M. Garavelli, F. Bernardi, M. Merchan, M.A. Robb, M.
Olivucci, Proc. Natl. Acad. Sci. USA 97 (2000) 9379.
[48] D. Mandal, T. Tahara, S.R. Meech, J. Phys. Chem. 108 (2004) 1102.
[49] G. Delgado-Barrio, A.M. Cortina, A. Varadt, P. Mareca, P. Villarrreal, S. Miret-
Artes, J. Comput. Chem. 7 (1986) 208.
[50] W.H. Press, S.A. Teukolsky, W.T. Vetterling, B.P. Flannery, Numerical Recipes in
Fortran 90, 2nd. ed., Cambridge University Press, 1996.