Professional Documents
Culture Documents
Two Color Cross Correlation
Two Color Cross Correlation
U U
ing beams direction.
E
2
S D SD
In practice, for a given radius r 1 , we restrict the calculation
of g I to those values of Dr and D f for which a significant 2x 2t xp
u P ~ x! u 2 }exp 2 rect , ~3!
degree of correlation is expected, i.e., when the green and red w d
q vectors approximately coincide.
It is to be pointed out that, with the arrangement shown in where xt and xp are the x components transverse and parallel
Fig. 1, it is not possible to collect exactly the same q vector to the incidence direction and
for the two colors, as the two beams impinge onto the cell
along the same direction ~see Fig. 2!. At low angle, however,
the difference d q between the two scattering wave vectors is
rect x5 H 1, u x u <1
0, u x u .1.
very small. In particular, d q can be considered negligible if it By substituting Eq. ~3! in Eq. ~2!, one obtains
S D
is less than the uncertainty associated to each q mode. In
fact, there is a natural uncertainty in the measure of q—due d q 2t w 2 d q pd
to the finite size of the scattering volume—that is associated g I ~ q 1 ,q 2 ,D f ! 5exp 2 sinc2 , ~4!
4 2
with the finite speckles size @18#. It is useful to decompose
d q into two components d qp and d qt parallel and transverse where sincx5sinx/x, and d q t and d q p depend on q1 and q2
to the incident beam, respectively ~see Fig. 2!. We will show only through their magnitudes and their relative azimuthal
in Sec. III that, with the present arrangement, to maximize orientation, due to the rotational symmetry mentioned above.
chromatic correlation the transverse components of the red From Eq. ~4! it is now possible to determine under which
and green wave vectors must be the same, i.e., d q t 50. conditions the maximum intensity cross-correlation may be
Moreover, as will be demonstrated in the following, to ob- observed. First of all, from simple geometrical arguments it
serve a significant degree of cross-correlation, d qp must be follows that for a given pair of scattering wave vectors q1
less than the typical uncertainty in the parallel component of and q2 , d q is minimum @and therefore g I is maximum, see
q, which is inversely proportional to the sample thickness Eq. ~4!# when both wave vectors have the same azimuthal
@18#. As a consequence, in order to have small enough d qp direction, i.e., for D f 50. Second, for a typical small-angle
one needs to work with thin enough samples. setup the Gaussian term in Eq. ~4! is always narrower than
The samples are microporous membrane filters ~Sarto- the other one, and it rapidly decays to zero. This is evident in
rius!, which have been characterized in a previous work @19#, Fig. 3, where the calculated behavior of the two factors in
and whose features fit very well the experimental require- Eq. ~4! is sketched as a function of the scattering angle for
ments. First of all, the membranes are quite thin, their thick- the red light u r , for three different values of u g , the green
ness being 14065 mm. Moreover, while in air they look light scattering angle. As a consequence, the Gaussian term
perfectly opaque, their transmittivity can be increased by fixes the position of the maximum of the cross-correlation
permeating them by a properly chosen, quasi-index-matching function at d q t 50, and its width at d q t '1/w, which is the
solvent. Therefore, it is possible to gradually vary the sample typical uncertainty in q associated to the speckle size @1,18#.
turbidity t by slightly changing the solvent index of refrac- By working out the dependence of d q t on the scattering
tion. Finally, membranes are static samples, so that the scat- angles, it can be easily shown that, at low angle, the condi-
tered speckle pattern does not change in time. This allows tions D f 50 and d q t 50 yield tan umax r 5(lr /lg)tanug @see
3488 LUCA CIPELLETTI, MARINA CARPINETI, AND MARZIO GIGLIO 57
g max
I 5sinc
2
S DU
d q pd
2 d q t 50
. ~5!
same sample as in Fig. 4!. As already pointed out, the cross-
correlation depends both on r g and on r r and this is evident
by noticing that the peak position moves towards larger r r as
r g increases. The peak position of the data in Fig. 5 has been
As a consequence, it follows that the cross-correlation is compared with the theoretical prediction and the result is
gradually lost when increasing the sample thickness d, for a shown in Fig. 6, where a plot of tanumax r versus tan ug is
given pair of scattering angles u r and u g —i.e.,for d q p fixed, presented. Note the remarkably good agreement between the
see Appendix A, Eq. ~A1!. Conversely, given d, g max I de- experimental data and the theoretical curve tanumax r
creases moving towards larger angles, due to the fact that 5(lr /lg)tanug , where no adjustable parameters have been
d q p u d q t 50 grows with u g @see Appendix A, Eq. ~A4!#. This used. With reference to Fig. 5, there is another interesting
sets a limit on the angular range and sample thickness for feature to notice, namely that the peak height decreases as r g
57 TWO-COLOR CROSS-CORRELATION IN SMALL-ANGLE . . . 3489
E
the space domain for light scattered by 2D samples. It is
d 3 xu P ~ x! u 2 exp~ ix• d qt ! quite remarkable that the two descriptions that apply to quite
m ~ d qt ! 5 different electromagnetic spectral regions have the same
E
,
physical root.
d 3 xu P ~ x! u 2 We will now compare the loss of cross-correlation with
~6b! the spread of the photon path length distribution, calculated
F h~ d q p ! 5 E dhp h ~ h ! exp~ ih d q p ! .
by means of a Monte Carlo simulation of the photon propa-
gation. The Monte Carlo multiple-scattering code is de-
scribed in detail in Ref. @8#. We simply recall here that the
simulation consists in tracking the path of a great number of
@Incidentally, we note that for a Gaussian incident beam, the
photons through the sample. For each launched photon, track
u m ( d qt ) u 2 term in Eqs. ~6a! and ~6b! is exactly the same as
is kept of the length of the path in getting out of the sample,
the Gaussian factor in Eq. ~4!.# From Eqs. ~6a! and ~6b!, it so that the photon path length distribution p l (l) can be cal-
follows that g I depends on the diffuser only through the culated for all q vectors of interest. The main required input
characteristic function F h ( d q p ). In the hypothesis of a parameters are the single scattering differential cross section
Gaussian distribution of the diffuser height, it can be shown and the turbidity t of the sample or, equivalently, its trans-
that at small angles F h does not depend on d q p @20#: mission T. The differential scattering cross section was that
obtained, for the same sample, in Ref. @8#. The measurement
F h 5exp@ 2Dk 2 ~ n21 ! 2 s 2h # , ~7! of the sample transmission deserves a brief comment. At low
turbidity, T can be directly measured thanks to the photodi-
Since F h is independent from q1 and q2 , it can be easily ode PD ~see Fig. 1!, by dividing the transmitted beam power
understood that for 2D samples m ( d qt ) fixes the peak posi- in the presence of the sample by that with the cell filled with
tion of g I (q1 ,q2 ) at d q t 50, while the second factor in Eq. the solvent alone. For highly turbid samples, however, the
~6a!, F h , is responsible of its height. The validity of Eqs. transmitted power can be comparable to, or even lower than,
~6a! and ~7! has been verified both qualitatively @4,6# and the power of the light scattered at extremely low angle, col-
quantitatively @7# in the past. lected together with the transmitted beam by the tiny mirror
It is to be noted that it is impossible to distinguish M , due to its finite size. Therefore, there is a comparatively
whether the wave front deformations of the scattered field high spurious signal on PD that leads to a significant over-
immediately after the cell have been introduced by a 2D or estimation of T. The correct value of the transmission can be
by a 3D sample. What is actually different is the physical recovered by iteratively running the simulation program.
mechanism that generates the phase variations, although in One starts with a guess for T to obtain the simulated total
both cases they depend on the differences in photon optical light collected by M ~scattered light plus transmitted beam!.
path lengths for crossing the sample. For 3D samples, the The initial estimate is then refined by comparing the result of
different path length of the photons is determined by the the simulation with the experimental data, and a corrected
amount of multiple scattering. In fact, photons that undergo a value of T is input for a second run. The entire process is
few scattering events travel a shorter path when compared to repeated until the simulated power converges to the mea-
those scattered many times. Moreover, we point out that, as sured one. Thanks to this procedure, it has been possible to
the turbidity grows, multiple scattering of higher orders be- estimate the transmission even for the more turbid samples,
57 TWO-COLOR CROSS-CORRELATION IN SMALL-ANGLE . . . 3491
ns g max
I I s /I t g max
I /(Is /It)
2
scribed above and is reported in Table I, together with scattered at small angles by thin 3D samples. We have used
g max
I /(Is /It) . It can be noted that the ratio g I /(Is /It) enor-
2 max 2 an experimental setup that allows to record the intensity dis-
mously increases with the amount of multiple scattering, thus tribution of the speckle fields, thanks to a CCD sensor. We
demonstrating that single scattering contributions are not the have shown that, in order to observe a significant degree of
only correlated ones. cross-correlation, it is necessary for the difference between
The main differences between the two techniques are the the scattering wave vectors to be small compared with the
following. First, we do not collect light so that the scattering natural uncertainty in q associated with the finite sample
q vector is exactly the same for the two wavelengths, as in thickness. Furthermore, we have observed that cross-
the two-color dynamic light scattering apparatus. Second, we correlation is gradually lost when the sample turbidity is in-
collect light at small angles and use samples with a differen- creased. We have suggested that this effect is associated with
tial scattering cross section strongly peaked near the forward the spread in photon path lengths caused by multiple scatter-
direction @19#, while in a typical two-color dynamic light ing. We have tested this guess thanks to a Monte Carlo simu-
scattering setup measurements are performed at much larger lation of photon propagation, and a qualitative good agree-
angles @12#, and using small scatterers, which diffuse light ment between the model and the experimental data has been
almost isotropically. As a consequence, for the dynamic light found.
scattering experiment the two-color wave vectors associated
to each intermediate scattering event are generally much dif-
ferent, and the only significantly correlated are the singly ACKNOWLEDGMENTS
scattered ones. Conversely, in the present experiment the dif-
The authors thank D. A. Weitz for useful discussions and
ferences between the intermediate scattering vectors for the
for bringing to their attention the papers by A. Z. Genack
two colors are so small that they turn out to be comparable to
et al. They also wish to thank F. Ferri for suggestions in
the unavoidable difference d q between the final wave vec- designing parts of the instrument. This work was partially
tors. Therefore, even light scattered more than once probes supported by the Ministero dell’Università e della Ricerca
essentially the same fluctuation Fourier modes for the two Scientifica e Tecnologica ~MURST! and by the Agenzia Spa-
colors. Note that the same reason for which we are able to ziale Italiana ~ASI!.
observe correlation with the present configuration—i.e., that
d q is smaller than the natural uncertainty in q—is respon-
sible for the correlation between multiply scattered contribu- APPENDIX A
tions. Only when a large number of scattering events occurs,
the small differences in the intermediate wave vectors may In this appendix we work out the dependence of d q p and
sum up and finally overcome d q, giving rise to a significant d q t on the wave vectors k 1 and k 2 of the two-color incident
loss of cross-correlation. light, and on the scattering angles u 1 and u 2 for the two
To summarize, we have performed measurements of two- colors. We will refer to the case D f 50 shown in Fig. 2.
wavelength spatial cross-correlation between the intensities From simple geometrical calculations, one gets
^ A ~ q1 ! A * ~ q2 ! & I s~ q 1 ! I s~ q 2 !
g A ~ q,q2 ! 5 , ~B2! g I ~ q,q2 ! 5 u g A,s ~ q1 ,q2 ! u 2 . ~B4!
AI ~ q 1 ! I ~ q 2 ! I t~ q 1 ! I t~ q 2 !
and A(q) is the scattered field. If we assume, as in Ref. @11#, Since the maximum intensity cross-correlation is observed
that only single scattering contributes to g A , then in the pres- for q 1 'q 2 [q, from Eq. ~B4! we obtain
ence of multiple scattering
In Eq. ~B3! subscripts s and t refer as usual to single and For a given sample thickness and a given scattering wave
total scattering, respectively, and g A,s is the field cross- vector q, g A,s is a constant and does not depend on the
correlation that would have been observed if single scatter- amount of multiple scattering. Therefore, Eq. ~B5! shows
ing alone were present @see Eq. ~B2!#. Equations ~B1!, ~B2!, that the cross-correlation peak height is proportional to the
and ~B3! yield square of the fraction of single scattering intensity.
@1# J. W. Goodman, in Laser Speckle and Related Phenomena, @11# K. Schätzel, J. Mod. Opt. 38, 1849 ~1991!.
edited by J. C. Dainty ~Springer-Verlag, Berlin, 1975!; J. W. @12# P. N. Segrè, W. Van Megen, P. N. Pusey, K. Schätzel, and W.
Goodman, Statistical Optics ~Wiley, New York, 1985!. Peters, J. Mod. Opt. 42, 1929 ~1995!.
@2# See, for example, B. J. Berne and R. Pecora, Dynamic Light @13# A. Z. Genack, Phys. Rev. Lett. 58, 2043 ~1987!.
Scattering ~Wiley, New York, 1976!. @14# N. Garcia and A. Z. Genack, Phys. Rev. Lett. 63, 1678 ~1989!.
@3# A. P. Y. Wong and P. Wiltzius, Rev. Sci. Instrum. 64, 2547 @15# A. Z. Genack and M. Drake, Europhys. Lett. 11, 331 ~1990!.
~1993!. @16# A. Z. Genack, N. Garcia, and W. Polkosnik, Phys. Rev. Lett.
@4# G. Parry, Opt. Commun. 12, 75 ~1974!. 65, 2129 ~1990!.
@5# G. Parry, in Laser Speckle and Related Phenomena ~Ref. @1#!. @17# F. Ferri, Rev. Sci. Instrum. 68, 2265 ~1997!.
@6# G. Tribillon, Opt. Commun. 11, 172 ~1974!. @18# See, for example, G. B. Benedek, in Statistical Physics, Phase
@7# M. Giglio, S. Musazzi, and U. Perini, Opt. Commun. 28, 166 Transition and Superfluidity, edited by M. Chretien, E. P.
~1978!. Gros, and S. Deser ~Brandeis University, Waltham, MA,
@8# L. Cipelletti, Phys. Rev. E 55, 7733 ~1997!. 1966!, Vol. 2.
@9# G. D. J. Phillies, J. Chem. Phys. 74, 260 ~1981!. @19# L. Cipelletti, M. Carpineti, and M. Giglio, Langmuir 12, 6446
@10# J. K. G. Dhont and C. G. De Kruif, J. Chem. Phys. 79, 1658 ~1996!.
~1983!. @20# L. Cipelletti, Ph.D. thesis, University of Milan, Italy, 1997.