JOURNAL OF CHEMICAL PHYSICS VOLUME 109, NUMBER 23 15 DECEMBER 1998

Subband analysis of molecular electronic transitions by nonlinear

polarization spectroscopy in the frequency domain
W. Beenken and J. Ehlert
Max-Born-Institute for Nonlinear Optics and Short Pulse Spectroscopy, Rudower Chaussee 6,
12489 Berlin, Germany
~Received 2 February 1998; accepted 9 September 1998!
Nonlinear polarization spectroscopy in the frequency domain ~NLPF! allows a detailed analysis of
substructure behind a molecular optical absorption band. Using the x (3) -approach here we deduce
theoretically line shapes of NLPF spectra for systems with more than one optical transition, which
contribute to linear as well as excited state absorption. It is possible from such line shapes to decide
whether these transitions are independent from each other or belong to a common ground state. We
demonstrate further how energy transfer and relaxation paths can be revealed and quantified by a
global analysis of NLPF spectra. This is useful to get a comprehensive understanding of ultrafast
processes in complex molecular and supermolecular systems. © 1998 American Institute of
Physics. @S0021-9606~98!00647-3#

INTRODUCTION first model, we will discuss, describes a set of different two-

Nonlinear polarization spectroscopy in the frequency do-
main ~NLPF! was introduced in the early 1970’s by Song,
Lee, and Levenson1 to determine the homogeneous linewidth
and relaxation rates for inhomogeneously broadened elec-
tronic transitions. After that it was the idea of Neef and
Mory,2 as well as Marcano et al.,3 to use NLPF to reveal the
type of spectral broadening for absorption bands of organic
molecules in solution. The fundamental differences in NLPF
spectra between inhomogeneous and homogeneous broaden-
ing were pointed out. These authors founded their theoretical
description of the NLPF spectra, however, on the model of
independent two-level systems. In the case of inhomoge-
neous broadening the distribution of transition frequencies
was assumed to be a single Lorentzian for sake of simplicity.
This does not allow a sufficient description of NLPF spectra
with a complex substructure, which may be caused by dif-
ferent absorbing species as well as by a multi-level system,
although that is the general experimental situation. A first
move toward a description of NLPF spectra for multi-level
systems was made by Andrews and Hochstrasser, who inves-
tigate a three-level system with two well separated
transitions.4
The aim to apply NLPF on molecular and supermolecu-
lar systems as a successful experimental method with high
temporal and spectral resolution leads us to deduce theoreti-
cal models for analyzing the line shapes of NLPF spectra,
which are more structured than the accompanying absorption
spectra for such systems.5 The nonlinear response, which
results in NLPF signal, was calculated from the generalized
Bloch equations in third order perturbation theory. The
x (3) -approach is a sufficient approximation, since experi-
mental improvements, above all a new arrangement of the
pump and probe beam, as shown in Fig. 1, allows the use of
low pump intensities.6 For all models that follow the NLPF
spectra are deduced in this frame.
After reminding the results for the two-level system, the
level systems, which do not interact with each other and have
different but overlapping absorption bands. This case is im-
portant, since it contains the effect of different sample con-
stituents, for example, cis-trans isomers of polymethin dyes,7
on the NLPF spectra as well as the influence of variable
environment resulting in static inhomogeneous broadening.
Second, we will introduce an incoherent energy transfer
between the different two-level systems, like Foerster trans-
fer, into the model. With that it will be possible to describe
the NLPF spectra of supermolecular aggregates with weak
excitonic coupling. These systems are discussed for photo-
synthetic antennae, which collect, convert, and transfer sun-
light energy to the reaction centers. NLPF promises to give
decisive hints on the photophysical background of these pro-
cesses, above all on the underlying energetic structure.5
Third, we will demonstrate the case of a multi-level sys-
tem, where all transitions belong to the same ground state.
This model is not only the standard model for molecules, but
also for supermolecular aggregates with strong excitonic
coupling.8,9 On this level of modeling, for example, one can
consider the vibrational progression in the excited electronic
states. A thermal distribution in the electronic ground state,
however, cannot be taken into account. The fourth model
shows the influence of excited state absorption on NLPF
spectra.
Finally it will be shown that all models discussed above
can be simply combined. This results in the description of a
general NLPF spectrum by only one analytical line shape
function containing as parameters the transition frequencies
and dipole moments, the dephasing, energy relaxation, and
transfer rates. Importantly, it also includes the relations be-
tween the transitions, i.e., whether they are independent,
coupled by energy transfer or belonging to a common ground
state.
The ensuing considerations on orientation averaging
takes different directions of the transition dipoles into ac-

0021-9606/98/109(23)/10126/12/$15.00 10126 © 1998 American Institute of Physics

J. Chem. Phys., Vol. 109, No. 23, 15 December 1998
FIG. 1. Improved setup for NLPF: The indices p, t, and s indicate the pump,
probe~test!, and signal beams. The E¢ assign their polarization vectors while
the k¢ assign their wave vectors. s't means the component of signal field
perpendicular to the probe field, which is measured behind the analyzer.
count. This is important since NLPF depends on the coupling
of two different, linearly polarized, external fields with all
different oriented transition dipole moments in the system. If
the transitions are not independent that may have an dramatic
effect on the NLPF spectra. Examples of such systems are
the porpherins with Q x and Q y transition10 and strong exci-
tonic coupled aggregates with different allowed transitions
into the first excitonic band.8
Consequently a global line shape analysis will be intro-
duced, which allows revelation of the term scheme as well as
the energy relaxation path of complex molecular and super-
molecular systems from a set of NLPF spectra, which are
measured at different probe frequencies.
LINE SHAPE FUNCTIONS OF NLPF SPECTRA
The NLPF signal Es is generated by the nonlinear polar-
ization of the sample under the influence of the external
pump and probe field. For samples without optical activity
the crossed polarizer and analyzer blocked the probe field Et
and any linear and nonlinear response of it. Therefore the
measured NLPF signal field E s is given only by nonlinear
polarization P NL which is perpendicular to E t and depends
s
on both external fields E t and E p . In the slowly varying
amplitude approximation one obtains
]Es a~ vt! v t NL
1 E s 52 p i P ,
]xs 2 c s
where a ( v t ) is the absorption coefficient for the signal
field.6 In the lowest order of external fields, the nonlinear
polarization is given by
P ~s3 ! 5 x ~s3p!pt ~ 2 v t , v p ,2 v p , v t ! u E p u 2 E t ,
with the element of the third order nonlinear response tensor
x (3)
sp pt (2 v t , v p ,2 v p , v t ), which is taken for probe ~t!,
pump ~p!, and signal ~s! direction, and depends only on
pump frequency v p and probe frequency v t . 1,11 The line
shape function s( v p , v t ) will be used here to describe this
dependence of nonlinear response x (3) (2 v t , v p ,2 v p , v t )
on pump and probe frequency and u s( v p , v t ) u 2 v 2t gives the
line shape of the NLPF spectra.6,11 In the following, line
shape functions and the resulting NLPF spectra are demon-
strated for different quantum mechanical models of molecu-
lar systems. In all cases the perturbational approach of den-
sity matrix theory as described in Ref. 11 was used for
calculating nonlinear response up to third order.
W. Beenken and J. Ehlert 10127
FIG. 2. Two-level system: Square module of line shape function ~2! for a
two-level system is shown in arbitrary units. It is plotted over the detuning
between pump and transition frequency for v 1052G and energy relaxation
rate g 50.01G.
Two-level system
The fundamental model used by most authors2,3,12,13 up
to now is the isolated two-level system. It was shown, e.g., in
the Appendix of Ref. 11, that for this case, in the rotating
wave approximation the x (3) -approach results in the line
shape function
s ~ v p0 , v t0 ! 5
i
S
2 2G
G2i v t0 g G 1 v 2p0
2
1
2
S 1
1
1
g 1i v pt G2i v t0 G1i v p0 DD , ~3!
which depends on the detuning v p0 5 v p 2 v 0 between pump
and transition frequency as well as on detuning v t0 5 v t
2 v 0 between probe and transition frequency, respectively.
For convenience we introduce v pt 5 v p 2 v t . G51/T 2 and
g 51/T 1 denote the dephasing and the energy relaxation rate,
respectively.
Figure 2 shows the resulting line shape for a fixed probe
frequency v t . One can see two peaks: The T 1 -peak in the
~1!
vicinity of v pt '0 and the T 2 -peak around v p0 50. While
the width of the T 1 -peak is determined by the energy relax-
ation rate g, the width of the T 2 -peak is given by the dephas-
ing rate G. Since the system is isolated it is homogeneously
broadened, and thus 2G is the FWHM of the absorption line.
~2! The T 2 -peak is symmetric for all probe frequencies and the
T 1 -peak shows only a small asymmetry for v t0 Þ0, caused
by the increasing or decreasing wing of the T 2 -peak lying
underneath.
Heterogeneous substructure
Heterogeneous substructures in NLPF spectra result
from superposition of independent transitions taking place in
an ensemble of different two-level systems, with generally
different transition frequencies, dephasing, and energy relax-
ation rates. To obtain the corresponding NLPF spectrum one
has first to sum up the line shape functions s m ( v p m , v t m ) of
the individual two-level system defined by ~3!, where G and
10128 J. Chem. Phys., Vol. 109, No. 23, 15 December 1998 W. Beenken and J. Ehlert

FIG. 4. T 1 -peak for heterogeneous substructure: Square module of line

FIG. 3. Heterogeneous substructure: Square module of line shape function
~12! for two systems with equal weight factors, energy relaxation, and
dephasing rates ( a 1 5 a 2 , g 1 5 g 2 , G 1 5G 2 5G) is shown in arbitrary units.
It is plotted over the distance of pump and probe frequency to the average of
the two transition frequencies (D v p 5 v p 2( v 1 1 v 2 )/2, D v t 5 v t 2( v 1
1 v 2 )/2). The distance between the transition frequencies is arbitrarily
given as double linewidth ( v 1 2 v 2 52G).
g have to be replaced by the relaxation rates of the m-th
two-level system G m and g m , respectively. Thus the square
shape function ~13! plotted over v pt 5 v p 2 v t at v t 5( v 1 1 v 2 )/2 is shown
in arbitrary units for ~a! g 1 5 g 2 and ~b! g 1 5100g 2 , where a 21 / g 1 5 a 22 / g 2 is
constant and g 2 50.01G. All other parameters as in Fig. 3.
g m and a m . In addition, the asymmetry of the T 1 -peak may
be useful in revealing a substructure, if the T 2 -peaks are
merged to a single peak.7
In the vicinity of the probe frequency, i.e., where u v pt u
!G m , the line shape function ~4! can be approximated by

S D
module of the total line shape function
1 1
s het~ v p , v t ! 'i (p a m2 1
g m g m 1i v pt
s het~ v p , v t ! 5 (m a m2 s m~ v p m , v t m ! ~4!
4G m ~ G m 1i v t m !
gives one the line shape of the NLPF spectrum. The weight- 3 , ~6!
~ G m2 1 v t2m ! 2
ing factors
a m } An mol, m u m
W mu 2. whose square module describes the profile of the T 1 -peak
well for g m !G m .
are used to consider differing transition dipole moments m Wm In Fig. 4 an example is given showing an extremely
and concentrations n mol, m of the individual two-level sys- asymmetric T 1 -peak for g 1 Þ g 2 , which belongs to the same
tems. broadband spectra given in Fig. 3 as the symmetric T 1 -peak
To describe the resulting NLPF spectrum outside of the for g 1 5 g 2 . To describe the asymmetry of the line shape one
T 1 -peak, i.e., for v pt @ g m , the line shape function ~4! can be uses the function
approximated by
D ~ v pt ! 5 u s ~ v t 1 v pt , v t ! u 2 2 u s ~ v t 2 v pt , v t ! u 2 , ~7!
2 a m2 2G m
s het~ v p , v t ! 'i (m G m 2i v t m g m G m2 1 v 2p m
. ~5! which results for the approximated line shape function ~6! to

The square module of this function shows a series of

T 2 -peaks centered at the transition frequencies v m with
D ~ v pt ! 52 (
m,n
c m2 c 2n S 1
2
1
v
g m g n ptD
S D
widths given by G m . It is important to notice that the height
of the T 2 -peak decreases with increasing distance between 2 g m g n 2 v 2pt v tm v tn
3 2 ~8!
the transition frequency and the probe frequency v t m . Mea- ~ g m2 1 v 2pt !~ g 2n 1 v 2pt ! Gm Gn
suring the T 2 -peak maxima for different probe frequencies
with the coefficients
allows determination of a m2 / g m . In Fig. 3 the two dimen-
sional line shape of the NLPF spectrum, which depends on 2G m
both pump and probe frequency, is shown for two two-level c m5 a m . ~9!
G m 1 v t2m
2
systems with equal dephasing and energy relaxation rates.
For a heterogeneous substructure, of course, the T 1 -peak From an analysis of the T 2 -peaks the prefactors c m2 / g m can
remains close to the probe frequency, but the maximum can be obtained as shown above. Therefore ~8! in principle al-
be shifted out of the center at v p 5 v t as in the case of lows a calculation of the relaxation rates g m and with them
spectral diffusion.2 The asymmetry is mostly much stronger the weight factors a m in principle.
than for a single two-level system, which may be caused by If the profile of the T 1 -peak cannot be resolved in detail,
different energy relaxation rates. Therefore it will be useful for example, because of the limitation by laser bandwidth, it
to analyze the T 1 -peak in detail to determine the parameters is possible to infer a substructure by the asymptotic of
J. Chem. Phys., Vol. 109, No. 23, 15 December 1998 W. Beenken and J. Ehlert 10129

D( v pt ). Away from the T 1 -peak but still in the vicinity of ~13! describe the coupling between probed and pumped tran-
the probe frequency, i.e., where u g m u ! u v pt u ! u G m u , one can sitions as mentioned above. In Appendix A it is shown that
approximate they are given by the equation system

D ~ v pt ! '22 (
m,n
c m2 c 2n
v pt S
T 1 ~ m ! 2T 1 ~ n ! v t m v t n
Gm Gn
2 . D (m ~ i v pt d km 2 g km ! b m n~ v pt ! 52 d k n for all n , ~14!
~10!
which represents the continuity equation for excitation en-
If the energy-relaxation times T 1 are large, so that the ergy flow in the whole system. While the off-diagonal ele-
T 1 -peak is narrow compared to the laser bandwidth, ~10! ments of g m n are defined as energy transfer rates, the diago-
offers the only possibility to estimate the differences be- nal elements, which are given by
tween them and set them in relation to a global relaxation
g mm 52 g 0 m 2 ( g n m ,
rate that is measured by lifetime based methods, e.g., fluo-
~15!
rescence decay. nÞm
Inhomogeneous broadening can be described as a het-
erogeneous substructure generated by a wealth of two-level contain the energy relaxation rates g 0 m , which describe the
systems, which varies continuously in transition frequency. irreversible flow of excitation energy to the environment of
For this the weighting factors a m are replaced by a distribu- no optical relevance. According to this point of view, the
tion function ‘‘lifetime’’ of excitation on state m can be defined as

a ~ v ! } An mol~ v ! u m
W u2, ~11! T 1 ~ m ! 5 u g mm u 21 . ~16!

where n mol( v ) describes the number of two-level systems for the broadband spectrum outside of the T 1 -peak, i.e., for
with transition frequency v per unit volume. Thus for an u v pt u @ u g m n u , one obtains b m n (0)@ b m n ( v pt ) and the line
inhomogeneously broadened system line shape function is shape function is given approximately by
given by a ma n 2G n
(
Ea
s EET~ v p , v t ! 'i b ~0! 2 . ~17!
m,n G m 2i v t m m n G n 1 v 2p n
s inh~ v p , v t ! 5 2
~ v !s~ v p2 v , v t2 v !d v , ~12!
With that the NLPF spectrum results in a series of T 2 -peaks
according to the results of Marcano et al.3 The description according to the heterogeneous substructure. The heights of
given here also allows the use of a distribution of the relax- these peaks, however, depend differently on changes in the
ation rates g and G. For that the distribution function has to probe frequency. We will demonstrate this specific behavior
be expanded to a~v,g,G!. of the T 2 -peaks using two two-level systems, which are
coupled by irreversible incoherent energy transfer with rate
g 12.0, while g 2150. Here index 1 assigns the acceptor and
Excitation energy transfer index 2 the donor of the excitation energy. For this system
the matrix equation ~14! results in
If two-level systems are coupled by weak dipole–dipole
interactions, the excitonic states are incoherent and localized 2
b 11~ v pt ! 5 ,
on molecular sites. In this case the excitation energy transfer i v pt 1 g 01
~EET! becomes Foerster type14 and can be described as hop-
ping of the excitation from one two-level system to the other. g 12 2
b 12~ v pt ! 5 ,
We will describe the hopping from system n to m by the i v pt 1 g 01 i v pt 1 g 021 g 12
energy transfer rate g m n . If the system n has a transition
frequency that is far from the probe frequency while m has b 21~ v pt ! 50,
one nearby, one can expect that the transition belonging to n
2
will contributes to the NLPF spectrum, while in the reversed b 22~ v pt ! 5 . ~18!
case it will not. Therefore, the line shape function for sys- i v pt 1 g 021 g 12
tems with excitation energy transfer must contain a double Inserting this into the line shape function for the broadband
sum over the probed and the pumped transitions, and one approximation ~17! one obtains

S
obtains

S
ia1 a 1 4G 1
a ma n 2G n s EET~ v p , v t ! '
s EET~ v p , v t ! 5i ( G m 2i v t m
b mn~ 0 ! 2
G n 1 v 2p n
G 1 2i v t1 g 01 v 2p1 1G 21

D
m,n

D
a 2 g 12 4G 2
2G n 1i v pt 1
1 b m n ~ v pt ! . g 01~ g 021 g 12! v p2 1G 22
2
~ G n 1i v p n !~ G n 2i v t n !

For the sake of simplicity we have assumed that all transition

~13! 1
ia2
S a2 4G 2
G 2 2i v t2 g 021 g 12 v 2p2 1G 22 D
. ~19!

dipole moments are parallel. We will show later on, that this The first row dominates ~19! for probing the acceptor, the
is no considerable constriction. The functions b m n ( v pt ) in second for probing the donor. The square module of this line
10130 J. Chem. Phys., Vol. 109, No. 23, 15 December 1998 W. Beenken and J. Ehlert

FIG. 5. Substructure with energy transfer: Square module of line shape
function ~25! with g 015 g 0252 g 12 is shown in arbitrary units. All other
parameters as in Fig. 3.
FIG. 6. T 1 -peak for substructure with energy transfer: Square module of line
shape function ~26!, ~a! for g 125 g 0150.01G and g 0250 is shown in com-
parison to the heterogeneous case, ~b! for g 1250 and g 025 g 0150.01G.
Both functions are normalized to one at v pt 50. All other parameters and
definitions as in Figs. 3 and 4.

shape function is shown in Fig. 5 for identical weighting transfer in contrast to the heterogeneous substructure. This is
factors a 1 5 a 2 , dephasing rates G 1 5G 2 , and energy relax- caused by the additional term c 1 c 32 g 12 / g 11g 22 between the
ation rates g 015 g 0252 g 12 . braces in ~22!, which does not appear in the heterogeneous
Fitting of the square module of ~17! to NLPF measure- case. Figure 6 shows this behavior by comparison of both
ments allows the determination of the values of a m a n b m n (0) cases, where the T 1 -time of the second transition is equal to
analogous to the case of separated systems. Since ~17! con- that of the first. The asymmetric T 1 -peak results from the
tains a sum over all probed transitions, however, a set of energy transfer with rate g 125 g 0151/T 1 ( g 0250), while for
NLPF spectra at varied probe frequencies is required. We the symmetric T 1 -peak according to the heterogeneous sub-
will come back to this point when demonstrating global structure the T 1 -times are only determined by relaxation
analysis of NLPF spectra. rates g 025 g 0151/T 1 . Apart from that, the dependence of the
To describe the T 1 -peak for u v pt u !G m line shape func- asymmetry on v pt is the same as for the heterogeneous sub-
tion ~13! can be approximated to structure. In Fig. 6~a! the T 1 -peak belonging to the broad-
a ma n band spectrum given in Fig. 5 for v t2 52 v t1 is shown. In
s EET~ v pt , v t ! 'i (
m,n G m 2i v t m
this case both different T 1 -times and energy transfer rate g 12
results in an asymmetric T 1 -peak. The asymmetry function
2G n D( v pt ), which belongs to this case, is shown in Fig. 9.
3 ~ b m n ~ 0 ! 1 b m n ~ v pt !! . ~20!
v t n 1G 2n
2
Homogeneous substructure
Since b m*n ( v pt )5 b m n (2 v pt ) follows from matrix equation
A homogeneous substructure results from identical
~14!, the asymmetry defined as in ~7! results in
multi-level systems, with several excited states affiliated to

D ~ v pt ! 52
1
2 ( c mc nc kc l
m , n , k ,l
S v tm v tk
2
Gm Gk D
3Im$ b m n ~ v pt ! b k l ~ 0 ! 2 b m n ~ 0 ! b k l ~ v pt !
1 b m n ~ v pt ! b k l ~ 2 v pt ! % , ~21!
with coefficients c m defined as in ~9!, which can be obtained
from the broadband spectra. For the model system described
by ~18! asymmetry is given by

H S D
D ~ v pt ! 54 c 21 c 22
1
2
1
u g 11u u g 22u
1c 1 c 32
g 12
g 11g 22 J
3 S D
v t1 v t2
2
2 g 11g 222 v 2pt
v ,
G 1 G 2 ~ g 111 v 2pt !~ g 2221 v 2pt ! pt
2

~22!
with u g 11u 5 g 01 and u g 22u 5 g 021 g 12 as given by ~15!. Even FIG. 7. Homogeneous substructure: Square module of line shape function
if the T 1 -times defined as in ~16! are equal, i.e., u g 11u ~34! with g 015 g 0252 g 12 and g 2150 shown in arbitrary units with the same
5 u g 22u , an asymmetry of the T 1 -peak can occur for energy parameters as in Fig. 4.
J. Chem. Phys., Vol. 109, No. 23, 15 December 1998 W. Beenken and J. Ehlert 10131

it results in qualitatively different spectra, since here the

functions b m n ( v pt ) are given by the equation system

(m ~ i v pt d km 2 g km ! b m n~ v pt ! 5 d k n 11 for all n , ~24!

with a different expression right of the equality. This repre-

FIG. 8. T 1 -peak for homogeneous substructure: Square module of line
shape function ~26! with g 125 g 0150.01G and g 0250 shown for ~a! the
homogeneous case and ~b! the incoherent energy transfer case. Both func-
tions are normalized to one at v pt 50. All other parameters as in Fig. 3.
one common ground state. For simplification we will assume
that all transition dipole moments are parallel to each other.
Therefore the part of the NLPF signal that depends on the
orientation of the dipole moments can be separated, together
with the external fields as in the two level system case, and
for the rest, as shown in Appendix B, one obtains the line
shape function
sents the effects of bleaching of the common ground state.
The terms in the third and fourth rows of ~23! contain the
functions
12 d m n
B m n ~ v pt ! 5 . ~25!
G m n 1i v pt 1i v m n
These result from virtual transitions, which are only possible
between levels coupled via optical transitions to a common
third level. In ~25!, besides the well known dephasing rates
G m and G n , corresponding to the transversal T 2 -times in the
Bloch model for optical transitions, the dephasing rate G m n
appears. This does not describe the decay of polarization but
rather the decay of the coherence between the quantum states
m and n, which may or may not be connected with a direct
optical transition.
If the energy-relaxation rates are much smaller than the
dephasing rates, b m n ( v pt )@B m n ( v pt ) follows and the vir-
tual transitions can be neglected. Therefore away from the
T 1 -peak, i.e., b m n (0)@ b m n ( v pt ), the NLPF spectrum can be
approximately derived from the broadband line shape func-

S
tion given as in ~17! using, however, the b m n (0) resulting
a ma n 2G n
s hom~ v p , v t ! 5i (
m,n G m 2i v t m
b mn~ 0 ! 2
G n 1 v 2p n
from ~24!.
To demonstrate the specific behavior of the T 2 -peak se-
ries in the case of homogeneous substructure, a three-level
2G n 1i v pt
1 b m n ~ v pt ! system with energy relaxation from the upper to the lower
~ G n 1i v p n !~ G n 2i v t n ! excited state ( g 12) and from both to the ground state ( g 01 ,
G m 1G n 1i v m n g 02) can be used. With this matrix Eq. ~24! results in
1B m n ~ 0 !
~ G m 2i v p m !~ G n 1i v p n ! 2
b 11~ v pt ! 5
D
,
G m 1G n 1i v m n 1i v pt i v pt 1 g 01

S D
1B m n ~ v pt ! .
~ G m 2i v t m !~ G n 1i v p n ! 1 2 g 12
b 12~ v pt ! 5 11 ,
~23! i v pt 1 g 021 g 12 i v pt 1 g 01
~26!
The first row looks identical to the line shape function for the 1
heterogeneous substructure with energy transfer in ~13!, but b 21~ v pt ! 5 ,
i v pt 1 g 01

b 22~ v pt ! 5
1
i v pt 1 g 021 g 12
21S g 12
i v pt 1 g 01
. D
Inserting these functions in ~23! and neglecting the virtual
transitions, the approximated line shape function ~17! ap-
pears as

s hom~ v p , v t ! '
ia1
S
2 a 1 2G 1
G 1 2i v t1 g 01 v 2p1 1G 21

1
a2
g 021 g 12
11S2 g 12
D
2G 2
g 01 v p2 1G 22
2 D
1
ia2
S
a 1 2G 1
G 2 2i v t2 g 01 v 2p1 1G 21
FIG. 9. T 1 -peak asymmetry: Asymmetry D( v p f ) as given in ~22! for the
T 1 -peaks of ~a! the homogeneous substructure and ~b! the incoherent energy
transfer case as given in Fig. 8.
1
a2
g 021 g 12
21Sg 12
D2G 2
g 01 v 2p2 1G 22
. D ~27!
10132 J. Chem. Phys., Vol. 109, No. 23, 15 December 1998 W. Beenken and J. Ehlert

The first term dominates for probing in the 0-1 transition, the
second for probing in the 0-2 transition. Even if a relaxation
between both excited states may not exist ( g 1250) and the
spectral overlap of the transitions is low ( u v 21u @G 1 ,G 2 ),
terms which describe the pumping of the transition other
than that probed appear, resulting in T 2 -peaks far away from
the probe frequency. They are caused by the bleaching of the
common ground state. This allows one to distinguish be-
tween heterogeneous and homogeneous substructure.
Since it can be assumed in most cases that the energy
transfer and relaxation rates from higher to lower states are
higher than those in the reverse direction, the homogeneous
substructure can also be well distinguished from the incoher-
ent energy transfer case. The T 2 -peaks at lower frequencies
than the probe frequency are considerably higher for the ho-
mogeneous substructure than for all other cases of substruc-
ture. This can be seen in comparing Fig. 7 with Fig. 5, for
both of which the same parameters are used. Nevertheless
function ~17! allows determination of the product
a m a n b m n (0) for both cases and we will come back to this
point on demonstrating the case of mixed substructures.
Figure 8 shows the T 1 -peak of a homogeneous substruc-
ture as well as the comparable T 1 -peak in the case of inco-
herent energy transfer, both for the same parameters as used
for the broadband spectra in Fig. 7 and Fig. 5, respectively.
For the homogeneous substructure the T 1 -peak can also be
described by the approximated line shape function as given
in ~20!, using matrix equation ~24! for definition of the func-
tions b m n ( v pt ). Thus the asymmetry of the T 1 -peak can be
obtained from ~21! as well. In the case of two transitions it is
identical to ~22!, except prefactor 4 changes to 3. The
equivalence of the form of D( v pt ) is demonstrated in Fig. 9.
Peaks that result from virtual transitions do not contribute to
the T 1 -peak, since they are centered at v pt 5 v m n , except
when they belong to degenerated states.
Excited state absorption
Excited state absorption cannot be observed in linear ab-
sorption spectroscopy, but NLPF spectra contain information
about it. The reason is that in NLPF the first excited state is
populated by pumping, which allows probing of the transi-
tions from there to higher excited states. This even happens
in third order perturbation theory for NLPF, i.e., even for the
lowest pump intensity. As shown in Appendix C, for a three-
level system with ground state absorption g and excited state
transition e, the line shape function results in

s exc~ v p , v t ! 5i S 2 a 2g
G g 2i v tg
2
a ga e
G e 2i v te
DSb gg ~ 0 !
2G g
v pg 2G 2g
2 1 b gg ~ v pt !
2G g 1i v pt
~ G g 1i v pg !~ G g 2i v tg ! D
1i S a ga e
2
a ga e
D 1 2G g 2i ~ v pg 1 v tg !
G g 2i v tg G e 2i v te 2G 202i ~ v pg 1 v te ! ~ G g 2i v pg !~ G g 2i v tg !
, ~28!

with the function Mixed substructures

If the virtual transitions can be neglected, all theoretical
1 line shape functions shown above have a unique form given
b gg ~ v pt ! 5 b eg ~ v pt ! 5 ~29!
g 011i v pt by

containing only the energy relaxation rate g 01 from first ex-

cited state. G 20 is the dephasing rate between the second
excited state and the ground state. Neglecting the contribu-
tion of virtual transition for G 20@ g 01 in the second row,
there is no difference from the two-level case except the
additional term in the first bracket

a ga e
2 . ~30!
v te 2iG e

This has the effect that a NLPF signal appears for probing
the exited state absorption as well as for the ground state
absorption. The position and shape of the T 2 -peak, however,
depend only on the transition from ground state to first ex-
cited state. Examples are shown in Figs. 10 and 11.
In the frame of a x (3) -approach it is not possible to de-
FIG. 10. Excited state transition: Square module of line shape function ~50!
termine the energy relaxation rate from the second excited shown in arbitrary units. The transition with the higher transition frequency
state by NLPF. The T 1 -peak then only gives information starts from the excited state, the other from the ground state. Remaining
about the energy relaxation from first excited to ground state. parameters as in Fig. 3 ( a g 5 a e , G g 5G e 5G, v e 2 v g 52G).
J. Chem. Phys., Vol. 109, No. 23, 15 December 1998 W. Beenken and J. Ehlert 10133

s ~ v p , v t ! 5i (
m,n
a ma n
G m 2i v t m S
g mnb mn~ 0 ! 2
2G n
G n 1 v 2p n

1g m8 n b m n ~ v pt !
2G n 1i v pt
D
~ G n 1i v p n !~ G n 2i v t n !
.

~33!
The two geometry factors g m n and g m8 n depend on the angle
between the dipole moments and are given by

g mn5
15
16p 2
E p

0
sin u d u E
0
2p
df

3 E
0
2p
d c ~ cos2 u m 2sin2 u m sin2 f m ! cos2 u n ,

FIG. 11. Excited state transition: Square module of line shape function ~50! ~34!
shown for the case that the oscillator strength of the transition that starts
from the excited state is twice as high as that of the other ( a e 52 a g ).
Remaining parameters as in Fig. 10. g m8 n 5
15
16p 2
E p

0
sin u d u E
0
2p
df

3 E 2p
d c ~ cos u m 2sin u m sin f m !
s ~ v p , v t ! 5i (
m,n
a ma n
G m 2i v t m S
b mn~ 0 ! 2
2G n
G n 1 v 2p n
0

3cos u m ~ cos u n 1sin u n sin f n ! cos u n , ~35!

1 b m n ~ v pt !
2G n 1i v pt
D
~ G n 1i v p n !~ G n 2i v t n !
. ~31!
where u, f, and c are the Euler angles. u m is the angle
between mW m and EW p , while f m is that between m
W
W m and kW s
counting in direction of E t . For index n they are defined
11

The equation for the functions b m n ( v pt ) can be generalized
by
(m ~ i v pt d km 2 g km ! b m n~ v pt ! 5d k n for all n .
with matrix d k n , which determines the relation between the
transition m and k. If different transition m and k have a
common ground state, one obtains d k n 51; otherwise d k n
50. The diagonal elements are given by d mm 52. Especially,
if d mk 50 and g mk 50, m and k are independent transitions
as in the heterogeneous case. Furthermore, excited state ab-
sorption can be taken into account by functions b m 8 n ( v pt ),
where m 8 is the excited state transition belonging to the
ground state transition m. These functions are given by
21/2b m n ( v pt ).
Therefore it is clear that most NLPF spectra can be ana-
lyzed by fitting the parameters a m , v m , G m in line shape
function ~31! and parameters g m n , d m n in Eq. ~32!. Before
we will show how to do this, we will expand the model to
the case of different orientations of transition dipoles, which
we have neglected till now.
Different orientations of transition dipoles
For all the systems whose line shape functions were de-
duced above, parallel transition dipole moments were as-
sumed. If, however, the transition dipole moments are not
oriented parallel, the line shape function has to contain addi-
tional geometry factors. With these line shape function ~31!
changes to
correspondingly. From spherical trigonometry results
cos~ u n ! 5cos~ u m ! cos~ h m n ! 1sin~ u m ! sin~ h m n ! cos~ c ! ,
~36!
~32! with h m n as the angle between m
W m and mW n . Setting the arbi-
trary Euler angles u 5 u m and f 5 f m from ~34! one obtains
113 cos~ 2 h m n !
g mn5 , ~37!
4
and from ~35!
71cos~ 2 h m n !
g m8 n 5 . ~38!
8
Notice that g m n vanishes at a ‘‘magic’’ angle of about
54,74°. Nevertheless, a system with energy transfer between
two transitions, whose transition dipole moments include the
magic angle, can be told apart from a heterogeneous system
by the T 1 -peak, since g m8 n is always greater than zero.
GLOBAL LINE SHAPE ANALYSIS
Global line shape analysis of NLPF spectra sheds light
on the term scheme, as well as on the energy relaxation and
transfer path of an unknown molecular or supermolecular
system. This will be done by fitting line shape function ~33!
under constrictions ~32! to a set of NLPF spectra, which are
recorded over the pump frequency v p at different probe fre-
quencies v t . The fitting parameters to be used are the
weighting factors a m , the transition frequencies v m , the
dephasing rates G m , the energy transfer and relaxation ma-
trix g m n , the angles between the transition dipoles h m n , and,
10134 J. Chem. Phys., Vol. 109, No. 23, 15 December 1998
last but not least, matrix d m n , which describes the term
scheme. The quantity to be minimized by fitting can be
Q ~ $ a m , v m ,G m , g m n ,d m n , h m n % !
(S
S̄ pt 2 u s ~ v p , v t ; $ a m , v m ,G m , g m n ,d m n , h m n % ! u 2
5 ,
p,t s pt
for example, where S̄ pt is the average and s pt the statistical
error of measurements of the NLPF signal at the pump fre-
quency v p and probe frequency v t . The line shape function
s( v p , v t ) does not depend linearly on the parameter set
$ a m , v m ,G m , g m n , h m n % and so algorithms for nonlinear op-
timizing have to be used, e.g., the simplex algorithm follow-
ing Nelder and Mead.15 Only the discrete parameters d m n
cannot be obtained from this optimizing algorithm and must
be tested for any possible combination of common ground
states separately. Since the number of parameters might be
too large, it is useful to make global analysis step by step and
to determine some parameters by linear absorption spectros-
copy and other methods than NLPF. Furthermore, it is nec-
essary to make assumptions about the term scheme, first as
mentioned above, about the discrete parameters d m n , i.e.,
about homogeneous or heterogeneous substructure, but also
to reduce the number of undetermined elements in the energy
relaxation matrix g m n . The validity of these assumptions can
be proven by a final competitive global line shape analysis of
the various models.
Broadband analysis
For a stepwise global analysis, first one analyzes the
broadband part of NLPF spectra, i.e., the spectrum outside
the vicinity of v p 5 v t . For that one uses the broadband line
shape function
X mn
s( v p , v t ; $ v m ,G m ,X m n % )5i (
m,n G m 2i v t m G n2 1 v 2p n
which is deduced from the first term of ~33! for fitting. Here
the fitting parameters are the transition frequencies v m , the
dephasing rates G m , and additionally the quantities X m n
5 a m a n g m n b m n (0). It is important to note that no assump-
tions about the transition dipole moments, the term scheme
or the energy relaxation matrix are necessary for this fit.
Only the numbers of subbands is to be assumed.
In the second step one determines the weighting factors
a m by fitting the function
2G
A~ v !5 (m An mol, m G m2 1 ~ v mm2 v ! 2 a m
to the line shape of the linear absorption spectrum, using the
transition frequencies v m and the dephasing rates G m ob-
tained by NLPF broadband analysis in the first step. If one
does not obtain a good fit it can be necessary to change the
number subbands or to correct the v m and G m in the NLPF
analysis and to repeat the first step with these values. After
an iterative fitting of NLPF spectra and linear absorption
spectra one obtains values for the parameters v m , G m , and
W. Beenken and J. Ehlert
a m , which are more accurately determined than those ob-
tained only by decomposition of the absorption spectrum.
Furthermore, one obtains values for the matrix X m n , which
can used in the third step of global line shape analysis to
D
2 obtain information about the energy relaxation and transfer
matrix g m n .
If the relative orientation of the transition dipole mo-
~39! ments are known, for example, from measurements of linear
dichroism or from the steric structure, one can take the geo-
metrical prefactors g m n and g m8 n from ~37! and ~38!, respec-
tively. Otherwise it can be possible to determine them by the
method shown in Appendix D. Now it is possible to deter-
mine the value of b m n (0) from X m n and to use these values
to determine the energy relaxation matrix g m n by solving the
system of equations given in ~32! for a given model. This
model has to contain matrix d m n , which gives the principal
scheme of terms, and assumptions about the energy-
relaxation path, which allows reduction of the number of
undetermined matrix elements g m n . For example, it may be
sensible to use only a triangle matrix g m n , i.e., to allow only
downhill energy transfer and relaxation. Since the broadband
NLPF spectra, as shown in Figs. 3, 5, and 7, are well distin-
guishable for the different models, it may be possible to
make these assumptions by a first glance at the NLPF spec-
tra. One should, however, always use different models for
fitting in order to find the most significant. Finally, we note
that from the b m n (0) one obtains only ratios between the
g m n by solving ~32!. To obtain quantitative values for all g m n
the T 1 -peak has to be taken into account.
Narrow band analysis
The methods of narrow band analysis can be used to
determine the T 1 -times without any knowledge of the
weighting factors a m , i.e., without the need to use the ab-
2G n
, sorption spectrum. Furthermore, they results in absolute val-
ues of the matrix elements g m n . However, they can only be
~40! applied on NLPF spectra that are well resolved in the vicin-
ity of v pt 50 and allow determination of any possible sub-
structure of the T 1 -peaks precisely.
The method of narrow band analysis, which will be dem-
onstrated here, uses the asymmetry of the T 1 -peaks as it is
defined in ~7!. For one-directional energy transfer and relax-
ation, i.e., g is a triangle matrix, any kind of substructure
results in a asymmetry given by
2 g mm g nn 2 v 2pt
D ~ v pt ! 5 (
m.n
f mn~ v t !
~ g mm
2
1 v 2pt !~ g 2nn 1 v 2pt !
v pt ,
~42!
~41!
where the functions f m n ( v t ) containing information on sub-
structure and energy transfer. For example, in the case of two
transition only the function f 21( v t ) appears in ~42!, which is
given by
H S D
1 1
f 21~ v t ! 5 ~ 42d 21! c 21 c 22 2
u g 11u u g 22u
JS D
g 12 v t1 v t2
1c 1 c 32 2 . ~43!
u g 11g 22u G1 G2
J. Chem. Phys., Vol. 109, No. 23, 15 December 1998
Since the functions f m n ( v t ) do not depend on v pt , for fixed
probe frequency v t they are independent free parameters in
~42!. This can be used for fitting a measured asymmetry to
the function D( v pt ) and allows determination of the g mm ,
which gives the T 1 -times defined in ~16!.
Of course it may also be possible to determine the matrix
g m n directly from the T 1 -peak by using ~32! and ~33!. This
method, however, needs a solution of the system of Eq. ~32!
for every optimization step of the fitting routine, while the
asymmetry is given by the explicit expression in ~43!.
SUMMARY
Standard line shape analysis of a single NLPF spectrum
already gives information about a spectral substructure with
a frequency resolution of some tenths of cm21 and a deter-
mination of dephasing rates from some tens of fs up to ps.
This can be obtained from decomposition of the absorption
spectrum only with some effort. For a set of NLPF spectra,
recorded over the same range of pump frequency but at dif-
ferent probe frequencies, the complex NLPF line shape func-
tion, which depends on both pump and probe frequency, al-
lows one to obtain information about the energy relaxation
and transfer rates, the kind of substructure, and probably the
transition dipole orientations by a global line shape analysis.
This is an important advance, since such information cannot
be obtained from an analysis of the absorption spectrum,
which, however, is useful to reduce the number of free pa-
rameters in global line shape analysis of NLPF spectra.
Compared to time resolved techniques NLPF has the advan-
tage of being energy selective, which allows the relaxation
processes to be directly assigned to the spectral substruc-
tures.
To use the global analysis for evaluating NLPF experi-
ments successfully, it is necessary to measure the NLPF
spectra under conditions where third order perturbation
theory is still valid, i.e., the intensity of the NLPF signal
depends quadratically on pump intensity. On the one hand
this calls for low intensities, on the other hand for good
fittings the signal–noise ratio must be high. In Ref. 6 a setup
was shown that improves the NLPF measurements qualita-
tively and quantitatively.
For the heterogeneous substructure the first experimental
results are shown for pinacyanol in Ref. 7. There the analysis
of the broadband part of the NLPF spectra of pinacyanol at
different probe wavelengths results in three subbands at 556
nm, 565 nm, and 607 nm according to the absorption spec-
trum. These subbands could have been interpreted as vibra-
tional progression. The asymmetry of the T 1 -peak, however,
revealed that the energy relaxation time of the transition at
565 nm is nearly two times longer than that of the transition
at 607 nm. This leads us to the conclusion that these sub-
bands are not part of one vibrational progression, but belong
to different spectral species, probably two isomers of pina-
cyanol. The successful global line shape analysis, in the
frame of a heterogeneous substructure, confirmed this inter-
pretation.
We expect also that global line shape analysis can be
used on measurements of photosynthetic antennae. For the
Q y -absorption band of light harvesting complex II at low
W. Beenken and J. Ehlert 10135
temperatures, the first step in global line shape analysis is
demonstrated in Ref. 5, resulting in a direct observation of
six spectral subbands and a determination of the dephasing
times belonging to them. From a first visual sight of this
spectrum, one can conclude that these six subbands must be
connected by energy transfer or, if the excitonic coupling is
stronger, by a common ground state resulting in homoge-
neous substructure. Further measurements at a different
probe wavelength will give the answer to this question.
ACKNOWLEDGMENT
This work has been supported by a grant from the Deut-
sche Forschungsgemeinschaft, Bonn ~Le 729/2-2!.
APPENDIX A
For deducing the line shape function of a combination of
various two-level systems in a sample one starts with the
microscopic representation of the polarization
n
P~ t,r! } (m mW m r m~ t,rW ! ~A1!
as given in Ref. 11, where index m indicates the different
W m is the transition dipole moment and r m (t,rW )
transitions. m
the accompanying element of the reduced density matrix.
From the phase matching condition for NLPF one ob-
tains the Fourier component of polarization
n
P~ v s ,kW t ! } (m mW m r ~m0,1!d ~ v s 2 v t ! , ~A2!
which generates an electromagnetic wave propagating in the
direction of the probe beam kW t , and oscillating with the fre-
quency v s 5 v t .
Matrix element r m(0,1) is defined by expansion of the re-
duced density matrix into a Fourier series of carrier waves of
the external pump and probe field
r m ~ t,r! 5 (
n 1 ,n 2
~ n ,n t !
rm p exp@ i ~ n p ~ kW p rW 2 v p t !
1n t ~ kW t rW 2 v t t !!# . ~A3!
For the off-diagonal elements one obtains, as shown in Ref.
11, the equation
W m ~ EW t N ~m0,0! 1EW p N ~m21,1! ! ,
\ ~ G m 2i v t m ! r ~m0,1! 5 m ~A4!
where N m(2n,n)
is the n-th Fourier component of the inversion
for the two-level system the transition m belongs to, in the
following simply called system m. If the systems are coupled
by incoherent energy transfer, the inversions are given by
] tN m~ t ! 2 (n g m n N n~ t ! 52F m~ t ! , ~A5!
where g m n is the energy transfer rate from system m to n and
g mm is defined in ~15!. F m (t) is the driving term caused by
the electromagnetic fields. In the representation of carrier
waves, ~A5! appears as
10136 J. Chem. Phys., Vol. 109, No. 23, 15 December 1998 W. Beenken and J. Ehlert

off-diagonal density matrix element r n m . With the accompa-

in v pt N ~m2n,n ! 2 (n g m n N ~n2n,n !52F ~mn ! ~A6! nying dephasing rate G m n and in second order of the external
fields it is given by

S D
and the driving terms in second order of external fields are
given by !
W m EW p u 2
um 1 1
r ~n0,0
m 5 1 . ~B3!
\ ~ G m n 1i v m n ! G n 1i v p n G m 2i v p m
2
1 2G
W m EW p u 2 2 m 2
F ~m0 ! 5 2 u m ~A7! and
\ G m1 v pm
and W n EW p ! * ~ m
~m W m EW t !
r ~n21,1
m
!
5
\ ~ G m n 1i v pt 2i v m n !
2
1 2G m 1i v pt

S D
F ~m1 ! 5 W m EW p ! * ~ m
~m W m EW t ! .
\2 ~ G m 1i v p m !~ G m 1i v t m ! 1 1
~A8! 3 1 . ~B4!
G n 1i v p n G m 2i v t m
From the linear equation system ~A6!, it follows that the Inserting these terms into the last sum of ~B1! results in the
Fourier components of inversion in ~A4! are given by linear third and fourth rows of ~23!.
combination of the external driving terms. Therefore one can
write

(k b n k~ n v pt ! F ~kn ! .
APPENDIX C
N ~n2n,n ! 5 ~A9!
Here the line shape function of a three-level system,
Inserting these in ~A6!, one obtains where one transition occurs between the ground state and the
W 01), the other between the first excited
first excited state ( m
(n ~ in v pt d m n 2 g m n ! b n k~ v pt ! F ~kn !52F ~mn ! . ~A10! and the second excited state ( m W 12), and no further transition
W 0250), will be deduced. For the off-diagonal den-
appears ( m
From ~A10! Eq. ~14! follows immediately. Inserting ~A7! in sity matrix elements in Ref. 11 one obtains
~A4! one obtains W 01~ EW 1 N ~100,0! 1EW p N ~1021,1! !
\ ~ G g 2i v tg ! r ~010,1! 5 m
1
r ~m0,1! }
G m 2i v t m (n ~ b m n~ 0 ! F ~n0 !1 b m n~ v pt ! F ~n1 ! ! , W 21EW *
2m ~ 1,1!
r r 02 ,

~A11! W 12~ EW t N ~210,0! 1EW p N ~2121,1! !

\ ~ G e 2i v te ! r ~120,1! 5 m
and inserting this into ~A2! results in the formula for the line
W 10EW *
1m ~ 1,1!
p r 02 , ~C1!
shape function ~13!. p

where N (2n,n)
10 is given by
APPENDIX B
in v pt N ~102n,n ! 2 (n g m n N ~102n,n !52F ~10n ! ~C2!
For deducing the line shape function of a homogeneous
substructure, i.e., a multi-level system where all transitions according to the inversion of a two-level system. In third
belong to the same ground state, from the formula for off- order perturbation theory the N (2n,n)
21 are given by the occu-
diagonal density matrix elements in Ref. 11 one obtains pation of the first excited state, since occupation of the sec-
W m ~ EW t N ~m0,0! 1EW p N ~m21,1! !
\ ~ G m 2i v t m ! r ~m0,1! 5 m
ond excited state is only possible by two photon absorption,
resulting in a x (5) -response. Therefore one obtains

1 (
nÞm
W n ~ EW t r ~n0,0
m ! W ~ 21,1! ! .
m 1E p r n m N ~212n,n ! 52
1 ~ 2n,n !
N . ~C3!
2 10
~B1! From that one can obtain the first row in ~28!.
The first term of the right side is identical with the right The terms containing r (1,1)
02 in ~C1! describe the virtual
side of ~A4!. To determine the inversion, however, the transitions where second excited state is involved. With the
bleaching of the common ground state by driving other tran- accompanying dephasing rate G 20 and in second order of the
sitions has to be taken into account. Therefore, instead of external fields one obtains

S
~A5! one must write
\ 22 W 12EW t ! * ~ m
~m W 01EW p !
r ~021,1! 5
] tN m~ t ! 2 (n g m n N n~ t ! 52F m~ t ! 1 n( F n~ t ! . ~B2! G 201i ~ v p 1 v t 2 v 20! G 101i ~ v p 2 v 10!

D
Þm
W 12EW p ! * ~ m
~m W 01EW t !
For the rest ~24! follows from ~B2! in the same way as ~14! 1 . ~C4!
G 101i ~ v t 2 v 10!
follows from ~A5!.
The second term of ~B1! describes the virtual transitions From inserting these terms in ~C1!, summing up over the two
of the multi-level system. The virtual transition between the transitions as in ~A1!, and replacing the index pair 01 by
upper states of the transitions m and n is controlled by the index g and 12 by index e, the second row of ~28! follows.
J. Chem. Phys., Vol. 109, No. 23, 15 December 1998 W. Beenken and J. Ehlert 10137

APPENDIX D spectra of homogeneously and inhomogeneously broadened two-level sys-

If no information about the parameters g m n are available
from other methods, a analysis of a special kind of NLPF
spectrum, which is recorded for v p 5 v t , i.e., under simulta-
neous variation of both the pump and the probe frequency,
can be used to determine them. For this special case the
general line shape function is given by
tems,’’ J. Opt. Soc. Am. B 7, 2145–2149 ~1990!.
4
J. R. Andrews and R. M. Hochstrasser, ‘‘Femtosecond relaxation of an
iron porphyrin observed with polarization spectroscopy in a three-level
system,’’ Proc. Natl. Acad. Sci. USA 77, 3110–3114 ~1980!.
5
A. Schubert, B. Voigt, D. Leupold, W. Beenken, J. Ehlert, P. Hoffmann,
and H. Lokstein, ‘‘Direct observation of spectral substructure in the
Q y -absorption band of light harvesting complex II by nonlinear polariza-
tion spectroscopy in the frequency domain at low temperature,’’ Biochim.

S D
Biophys. Acta 1321, 195–199 ~1997!.
g m8 n X mn 6

(
B. Voigt, F. Nowak, and W. Beenken, ‘‘New setup for nonlinear polar-
s vp ,vp ; 5i
g mn m,n G m 2i v p m ization spectroscopy in the frequency domain. Experimental examples and
theoretical background,’’ Meas. Sci. Technol. ~to be published!.

S D
7
B. Voigt, F. Nowak, J. Ehlert, W. Beenken, D. Leupold, and W. Sandner,
g m8 n 2G n ‘‘Substructures and different energy relaxation time within the first elec-
3 11 . ~D1!
g mn G n 1 v 2p n
2
8
tronic transition of pinacyanol,’’ Chem. Phys. Lett. 278, 380–390 ~1997!.
F. C. Spano, ‘‘Multiple-photon absorption and saturation in Frenkel-
Since X m n , v m , and G m are known from global line shape exciton chains,’’ Phys. Rev. B 46, 13017–13034 ~1992!.
9
analysis of the broadband NLPF spectrum, the angles be- H. Fidder, K. Knoester, and D. A. Wiersma, ‘‘Optical properties of disor-
dered molecular aggregates: A numerical study,’’ J. Chem. Phys. 95,
tween the transition dipoles h m n are the only free fitting pa- 7880–7890 ~1991!.
rameters in ~41! appearing in the ratio 10
M. Gouterman, ‘‘Optical spectra and electronic Structure of porphyrins
and related rings,’’ in The Porphyrins Vol. III, edited by M. Dolphin
g m8 n 71cos~ 2 h m n ! ~Academic, New York, 1978!, pp. 1–156.
5 . ~D2! 11
g mn 216 cos~ 2 h m n ! W. Beenken and V. May, ‘‘Strong field theory of nonlinear polarization
spectroscopy in the frequency domain. Fundamentals and the two level
From that both g m8 n and g m n can be determined simulta- 12
system,’’ J. Opt. Soc. Am. B 14, 2804–2810 ~1997!.
S. Saikan and J. Sei, ‘‘Experimental studies of polarization spectroscopy
neously. in dye solutions,’’ J. Chem. Phys. 79, 4146–4153 ~1983!.
13
A. Marcano and F. Garcı́a Golding, ‘‘Power effects in polarization spec-
1
J. J. Song, J. H. Lee, and M. D. Levinson, ‘‘Picosecond relaxation mea- troscopy on a water solution of malachite green,’’ J. Chem. Phys. 82,
surements by polarization in spectroscopy in condensed phases,’’ Phys. 1242–1245 ~1985!.
Rev. A 17, 1439–1447 ~1978!. 14
Th. Förster, in Modern Quantum Chemistry, edited by O. Sinannoglu
2
E. Neef and S. Mory, ‘‘Content of information of nonlinear polarization ~Academic, New York, 1965!, p. 93.
spectroscopy,’’ Exp. Tech. Phys. ~Berlin! 39, 385–388 ~1991!. 15
J. A. Nelder and R. Mead, ‘‘A simplex method for function minimiza-
3
A. Marcano O., L. Marquez, L. Aranguren, and M. Salazar, ‘‘Polarization tion,’’ Comput. J. 7, 308–313 ~1965!.