F. García-Golding Vol. 73, No. 1/January 1983/J. Opt. Soc. Am.

59

Spectral-diffusion effects in polarization spectroscopy

F. García-Golding
instituto Venezolano de investigaciones Científicas, Centro de Física, Apartado 1827, Caracas 1010A, Venezuela

Received (uly 20,1981; revised manuscript received August 11,1982

The effect of cross relaxation on the third-order nonlinear susceptibility, x ( ^ 2 » « l , — u¡\, 0)2), describing polar-
(3) —

ization spectroscopy experiments is calculated with a density matrix formalism. By using a model of inhomo-
geneously broadened two-level systems with a reservoir state and allowing for cross relaxation within the inhomo-
geneous line, we find that both the ratio of the máximum signal at wi = (¿2 to the background and the broad compo-
nent resembling the absorption band depend on the cross-relaxation rate. Also, one additional component with
a characteristic width of the order of the spectral-diffusion rate r¡ is expected to appear, depending on the valué
of this parameter. Furthermore, we estímate from published data [Phys. Rev. A 17, 1439 (1978)] the product of
the transverse-relaxation time times the cross-relaxation time T0T3 to be 0.014 psec for cyanine dye and 0.0009
2

psec for malachite green oxalate.

2

Recent nonlinear optical techniques, such as resonant Ray-
leigh-type optical mixing and polarization spectroscopy
1-3
( P S ) , have made it possible to measure, in the frequency
4,5
domain, the homogeneous linewidth and decay rate of inho-
tween theory and the experimental data. In this simplified
model, we have neglected the potentially interesting phe-
nomenon of rotational diffusion.
The spectral-diffusion model introduced by Mourou and 6

mogeneously broadened electronic transitions involving states extended by Yajima is used to perform the calculations of the
1

that decay in the subpicosecond range. The absence of any
wave-vector matching condition in P S makes this technique
more attractive than the Rayleigh-type technique. In addi-
tion, the detected signal has a contribution from the steady-
appropriate third-order susceptibility for PS. In this model
the probability that a molecule changes its transition fre-
quency from oír/ to a valué withintoo~ too + dtoo in unit time
isg(wo)T3 dtoo, whereg(ton) is the molecular distribution.
-1

state change of the populations (hole burning) that permits
the extraction of the transverse relaxation time T in a simple
2
way.
In a P S experiment, the intensity and frequency-dependent
dichroism and birefringence are created by a pump wave and
probed by a second wave at frequencies a>i and (¿2, respec-
The nature and shape of this distribution is attributéd to
two random quantities obeying a Gaussian distribution, that
is, the different solvent cages and the different configurations
adopted by the solute molecule. As the work of Mourou has
shown, the adequacy of this hypothesis has already been
tested in the case of cryptocyanine dye. 6

tively. In order to explain these optical anisotropies in organic The model system considered is shown in Fig. 1, where y , a

dye solutions, Song et al. used an ensemble of two level sys-

4 7 b , and y are the population decay rates and y¿ = T ~ and
c 3 l

tems with a distribution of resonant frequencies for modeling y i = g(too)T3 (per unit to ) are the cross-relaxation rates for
_1 0

the absorbing species. In that case, the expected line shape direct spectral-diffusion and inverse processes, respectively. 7

consists of two components, the broader having a width twice Two incident light waves of frequencies toi andto2are nearly
the homogeneous linewidth (21^2 = 2 / T ) and the narrower
2 resonant with the transition (a - * 6) assumed inhomo-
having a width related to the population-decay rate ( 1 / T i ) . geneously broadened. The level (c) is far from resonance with
Furthermore, the ratio of the máximum signal observed at a>i both toi and o>. The density matrix equations describing this
2

— 0)2 = 0 to the background level that is due to the component system are
of the width 2T2 is predicted to be 4:1. However, the pre-
sented experimental data generally show lower valúes: the
4

ratio varies from 3.2:1 for malachite green oxalate in water 7 (HabPba — PabHba) + "YaPbb + y Pcc c ~ ~~
di h I3

solutions to less than 2:1 for cyanine dye in high-viscosity

solvents. It is also concluded that the optical transitions in + ~^~ I Paa(w. t)dto', (1)
these organic dyes are nearly homogeneous. This is incon- 1 3 JO
sistent with recent Raleigh-type optical-mixing experi- "Pfcfe _ l tu ij Pbb
ments. ~J~ ~ 7 \HbaPab — Pb n b) a a - ybPbb - y a Pbb ~ ~
dt h
3
T3
It has been learned from hole-burning and optical-mixing s(to) r™
experiments that cross relaxation is always present in systems
6 + ~ p (<»',t)da',
bb (2)
TQ JO
such as organic dyes, that is, a molecule can migrate freely to
any place to experience varying environmental molecular
Paa + Pbb + Pee = g(<>>), gMdto = 1, (3)
fields, which causes random shifts in transition frequen-
cies.
It will be seen that the inclusión of spectral diffusion within dpba _ dPab* 1 1 \ . • \\
u

~T~ - = 7 HbaKPaa ~ Pbb) ~ (1 2 + l<¿0)Pba, (4)

the homogeneous broadening yields a better agreement be-
di di h

Reprinted from Journal of the Optical Society of America, Vol. 73, page 59, January 1983
Copyright © 1983 by the Optical Society of America and reprinted by permission of the copyright owner.
60 J . Opt. Soc. Am./Vol. 73, No. 1/January 1983 F. García-Golding

The complex macroscopic nonlinear polarization is given

lb> 1- •Ti c; by

(i
P(3)(to ) = N 2 </i 6P6a >dto ,
a (3) 0 (8)

where N is the density of the absorbing species and the

brackets denote an ensemble average over the initially iso-
UJ0
•lc> tropic orientational distribution. Then, after integration,
2TT¿
P ( 3 ) ( 0 ) 2 ) = ~^ToN (Hab(Hba • E ) \(l 2 ba • E i | ) F(A), 2 (9)

where
la>-
F(A) = FAA) + F (A), 2

1 i
Fig. 1. Energy-level scheme. +
r + ÍA r r x s x r, + ¿A
H b a = H * = -n a b ba • ( E ^ - " " ' + E e-»* 2 + c • c), (5) i
where r is the transverse-relaxation rate, E i and E are the
2 2
( r + ¿A)(r + ¿A)
x 2r + s 2 ¿A
electric-fíeld amplitudes of the líght waves, Hba is the electric i
- r ) + •r ( r - r )
F (A) =
2
dipole matrix element, and g(o>o) is the distribution of reso- r + ¿A r r
4 x x 3 s s 3

nant frequencies describing the total absorption band.

i
All levéis except (a) are assumed unpopulated at equilib- +
rium, and, for the sake of simplicity, a Lorentzian shape of ( r + iA)(r - r + ¿A)
x x 3

width T (HWHM) will be used for g(to). Also, the following

4 I r ,3 r D 1

condition will be used: (r, + iA)(T, r + 3 ¿A) ( r + ¿A)

4

T » r ,7a,
4 2 7fc, 7 c - (6) i i
Solving Eqs. (l)-(5) in the rotating-wave approximation r (r -r )\rx r - r s 3 s x

by the method of successive approximations gives the fol- í^

lowing third-order solution of p at the probé frequency: b a ( r + ¿A)(r - r + ÍA)
x s 3

-¿Qifeq-E^lMfea-Exl 2
I i
(3)
Ph, (W ) =
2

h*T [ r + ¿(too ~ to )]
4 2 2
x \ r + ¿A ' r - r + ¿ A ;
s x 3

r + r - r
x /1 i s D where condition (6) was used to perform the calculation.
x — — — - + — + — 5
The third-order nonlinear susceptibility is defined, as usual,
r r \D DÍ x s 2

by the following relation between the complex field ampli-

i i r D
tudes:
v + ¿ A r + ¿A ( r + ¿A)(r + ¿A
x s x s

P¿< >(to ) = 6
3 2 X ( 3 W - w , to -toj, 2 1( to ) 2 EjíuJEkHwJEiiwo).
i i
- 2flT3g(to ) 0
The two components necessary for PS, i.e., X 1 2 1 2 and X 1 1 2 2 ,
r (r - r ) r(r - r ) x x 3 s 3

are proportional to the function F( A) introduced above. The

i
F i ( A ) is identical with the one obtained by the result of Song
( r + ¿A)(r - r + ¿A)
x s 3
et ai, except that the population rates are replaced by effec-
i tive rates that take into account'spéctral diffusion. This part
+ •(r + ¿A)(r r + ¿A)
s 3
broadens as cross relaxation increases. The F part origi- 2

nating from inverse spectral diffusion has resonant denomi-

1
- 2xr g(to ) 3 0
nators, such as T - T + Í A and T — T + ¿A, which are
x 3 s 3

r ( r - r >rx s 3 s independent of T . In addition, it has a background compo-

3

nent that follows the absorption line. When T increases, 3

(T x + iA)(T s - r + ¿A) 3
F ( A ) increases and reaches a plateau in contrast to contin-
2

uously decreasing F i ( A ) . Let us assume that

i i
Q— ÍCÜ2Í (7)
\r s + ¿A 'r - r + x 3 ¿A/J r »r ,
4 2 r 3 » 7 i , > 7 . > C 7 ; Q

where then
D¡ = T + i(to - 2 0 to ),£ (i = 1,2), Tz, = 7 - Q 7 c ,
r ~ r ,
x 3

r = 7c
x + y a + jb + — =r , s r ~ r , s 3
T,' 1 3
r - r =
x 3 y, e

r = —-,
3 A =to!- co2.
i3 r, - r « 3 7 6 . (10)
F. García-Golding Vol. 73, No. 1/January 1983/J. Opt. Soc. Am. 61

From Eq. (9), we can see that 1.0

2
F (0)
x c* F (0) 2 di)
r r '
2 3 r r '
4 c

then 8

F (O) ^ r r
2 2 3
(12)
>i(0) ~ r ' 4 T c

The assumption that y , 7 6 » 7 c leads to a

R _ i r r 2 3

2 r 4 7 c

Depending on the valúes of the different relaxation times in-

volved in Eq. (12), the inverse spectral-diffusion effect can
become as important as the direct one.
The appearance of the spectra for a P S experiment pa-
rameterized by T is plotted in Fig. 2, which shows that, for
3

low cross-relaxation rates, the central peak broadens. Nev-

ertheless, at larger valúes of T a narrow peak will appear in 3

the spectra with a width of the order of y . This peak is c

precisely the contribution from the inverse spectral-diffusion

process, specifically, the resonant term T — T + ¿A. x 3

Figure 3 shows the whole appearance of the spectra pa-

rameterized by T in which the following three regions can be
3
Fig. 3. Same as Fig. 2 but on a larger frequency range, showing the
intermedíate región and the broad component.
observed: the narrow peak that was mentioned before, having
a width related to the slowest population decay time; an in-
exhibit only two time scales: 7 and ( 2 r ) for T / r « C _ 1 2 _1 3 2
termedíate región related to 2r , T , , and T (more noticeable
1, and 7
2 x
and T for T a¡ T . We show below that, in the
for r / r = 0.2 in Fig. 3); and a wider component similar to
C _ 1 4 _ 1 3 2

case in which T ca T , the contrast of the central peak with

3 2

the absorption band (more noticeable for r / r = 0.2, 0.7 in

3 2
3 2
respect to the wider component can provide us with infor-
Fig. 3).
mation about the product T2T3.
For valúes of T such as T / r = 0.2 in Fig. 3, the width of
3 3 2
In light of this model, the central peak observed in P S ex-
the intermedíate región approaches T . It is the contribution 3
periments on malachite green oxalate in water solutions and
of the oscillation of the population that comes from F\, and
cyanine dye on high-viscosity solvents should be interpreted
in that case, the spectra will clearly show three characteristic
as the contribution of the inverse spectral-diffusion process.
times: 7 , ~ r , and r . For very low ( r / r « 1) or
C _ 1 3 _ 1 4 _ 1 3 2
It is consistent with the following facts:
very high ( T a* T ) spectral-diffusion rates, the spectra will
3 2

(1) The decay rate deduced from the width of the central
peak in malachite green agrees with the time-domain exper-
iment of Ippen et al., and the faster transient that is due to
9

cross relaxation should not be noticeable in these experi-

ments.
(2) The ratio of the máximum signal of A = 0 to the in-
tensity of the background component is lower than 4.
(3) There is a broad component similar to the absorption
band.

Within the approximations expressed in Eqs. (10), and with

yc « | A| « 7 r , r , T , we get
6 > 2 3 4

Re F i ( A ) F,(0),
ImFi(A) V A [(2r + r ) / r r ] F ( o ) « F M ,
8 2 3 2 3 1

Re F ( A ) 2
i / r 7 c = y F (o),
4 2 2

Im F ( A )
l / r A « F (0).
2
4 2

Then
F(0) 2 íFi(0) + F (0) ja 1 + Ry 2

= 12 (13)
F(A) Fi(0) + F (0)/2. 2 + RJ2

and
Fig. 2. Computer plot of |F(A)/F(0)| for 0 < A < I0y . V is
R = 2 S/A - 1
2 c 3

considered a parameter. The solid line is for r¿ < y ; the dashed line c
r r 2 3
(14)
is for Ta > 7,.. VA i\y c
62 J. Opt. Soc. Am./Vol. 73, No. 1/January 1983
We now analyze the P S experiment of Song et al. on 1,3'-
diethyl-2,2' quínolylthía-carbocyanine iodide (DQCI) in light
of this model. It is reasonable to assume that the width of the
central peak, i.e., Ao>/27re =Ü 3 c m (FWHM) corresponding
_ 1
to a time of ~3.4 psec, is the longest relaxation time y . The
time related to the inhomogeneous distribution l / r taken
from the absorption data of Yajima for T = AÍO/2-ITC = 818
cm- (HWHM) is 1/T = 0.0064 psec. From the P S experi-
1 4
ment, A 2; therefore, from Eq. (14), y T /T T 1.41 and
T T
2 1.4 X 10- psec .
3 2 2 we became aware of a recent paper in which the influence of
The Rayleigh-type optical-mixing experiment of Souma et
al. on DQCI in ethanol, although performed with a different
3 considered. Attention was focused on the disappearance of
solvent, is in reasonable agreement with our estímate. They
report T between 0.1 and 0.2 psec and T / T Y in the range
2 2
0.4-1 psec, where 1/7Y = VTi + 1 / T ^ 1 / T . Unfortu-
nately, for the case of malachite green, we do not have a
Rayleigh-type optical-mixing experiment to compare with,
but we can estímate the product T T from the P S experi- 2 3
ment. Repeating the same arguments that were previously
applíed to DQCI with A 3, y ~ 1.2 psec, and l / r =
c l
0.0063 psec, weobtainR ^ 8.6and T T = 9 X 10~ psec for 2 3
malachite green oxalate.
This analysis also implies that the relaxation rate extracted
from the narrow component in a P S experiment is not always
comparable with the effective relaxation rate ( 1 / T i + 1 / T )
= 1/TY measured by optical-mixing experiments, which
explains the apparent contradiction pointed out by Song et
al. when they compared both techniques. In the interpre-
tation of these experiments, the population-decreasing effect,
or direct spectral diffusion, was taken into account; the inverse
spectral diffusion is not expected to influence the line shape
except for | A | < 1/Ti (see, for example, Ref. 3).
The previous analysis was done under the hypothesis that
the homogeneous linewidth is small compared with the dis-
tribution of resonant frequencies; if we drop that restriction
we need to make a convolution with the line shape of the ab-
sorption band, and the spectra will depend not only on A but
also on the position of the pump and probé frequencies within
the inhomogeneous distribution.
In conclusión, we have made a study of the expected line
shape in a PS experiment in which cross relaxation within an
inhomogeneous line is considered. The features exhibited
by the región around the narrow component of the spectra and
the existence of a broad component similar to the absorption
band are explained. Also, one additional component with a
characteristic width of the order of T is expected to show,
depending on the valué of the spectral-diffusion rate. It was
shown that the product T T can be estimated from the
2 3
contrast of the narrow peak with respect to the broader
component of the spectra and that such estimates agree with
F. García-Golding
previous experiments that use a different technique, such as
Rayleigh-type optical-mixing spectroscopy.
Experiments are in process to detect the predicted com-
ponent, so that not only T T but also T and T can be de-
2 3 2 3
c - 1 termined. Such a search is important because it is necessary
4 to have at least two independent techniques with which to
4 measure in the subpicosecond regime in order to test a specific
model for relaxation mechanisms in the condensed phase.
c A 3 2 Note added in proof: During the writing of this paper, 11
cross relaxation for a two-level system in P S spectroscopy was
the information about T (specifically, the broadening of the
2
wing of the spectra connected with the steady-state change
3 3 of the population hole burning) for increasing spectral-dif-
fusion rate. Our emphasis is on the possibility of obtaining
information about the product T T and, under certain con-
2 3
ditions, T and T separately even in the presence of relatively
2 3
high spectral-diffusion rates.
10 4
4 2
REFERENCES
1. T. Yajima and H. Souma, "Study of ultrafast relaxation processes
by resonant Rayleigh-type optical mixing. I . Theory," Phys.
3
Rev. A 17,309-323 (1978).
2. T. Yajima, H. Souma, and Y. Ishida, "Study of ultrafast relaxation
1,2
processes by resonant Rayleigh-type optical mixing. I I . Ex-
periment on dye solutions," Phys. Rev. A 17, 324-334 (1978).
4 3. H. Souma, T. Yajima, and Y. Taira, "Ultrafast relaxation study
by resonant Raleigh-type mixing spectroscopy using picosecond
light pulses," J. Phys. Soc. Jpn. 48, 2040-2047 (1980).
4. J. J . Song, J. H. Lee, and M. D. Levenson, "Picosecond relaxation
measurements by polarization spectroscopy in condensed
phases," Phys. Rev. A 17,1439-1447 (1978).
5. J . H. Lee, J . J . Song, M. A. F. Scarparo, and M. D. Levenson,
"Coherent population oscillations and hole burning observed in
Sm :CaF using polarization spectroscopy," Opt. Lett. 5,196-198
+ 2
(1980).
6. G. Mourou, "Spectral hole burning in dye solutions," I E E E J .
Quantum Electron. QE-11,1-8 (1975).
7. The división of the spectral diffusion into direct (population-
decreasing effect) and inverse (population-increasing effect)
diffusion is not meant to imply the existence of two different
physical effects. It is rather a convenient way of separating the
population-changing effects of spectral diffusion on a selected
pair of levéis with transition frequency a>o-
8. A parallel expression was obtained in Ref. 3 in relation to Ray-
leigh-type optical mixing.
9. E . P. Ippen, C. V. Shank, and A. Bergman, "Picosecond recovery
dynamics of malachite green," Chem. Phys. Lett. 38, 611-614
3 (1976).
10. An absorption spectrum was made in malachite green with a Cary
17DI spectrophotometer in order to obtain the quantity r . 4
11. Y. Taira and T. Yajima, "Cross relaxation effect in polarization
spectroscopy of condensed phases," J . Phys. Soc. Jpn. 50,
3459-3466(1981).