Food Research International 27 (1994) 195-198

Particle size analysis by dynamic light scattering

F. Ross Hall&t
Guelph- Waterloo Program for Graduate Work in Physics, Department of Physics, University of Guelph, Guelph, Ontario,
Canada Nl G 2 WI

Dynamic light scattering is an established technique for measuring the average

size and size distribution of particles in a suspension. The technique has the
advantage of being fast and non-invasive, but it does require low particle con-
centrations. As well, dynamic light scattering results are often open to misinter-
pretation if one is unaware of the state of the sample and the method of data
analysis. The following discussion reviews some of the basic concepts of
dynamic light scattering and outlines some of the pitfalls that are often encoun-
tered in data interpretation. A modification of dynamic light scattering, diffus-
ing wave spectroscopy, can be used to obtain approximate size information at
higher particle concentrations. The fundamentals of this new technique are
summarized.

Keywords: dynamic light scattering, particle size analysis, diffusing wave spectro-
scopy, moments analysis, exponential sampling.

BASIC PHYSICS OF DYNAMIC LIGHT netic wave can impart or receive energy and mo-
SCATTERING mentum from the external and internal motions of
the scatterer. The frequency shifts due to Brown-
The dynamic light scattering (DLS) technique is ian motion are so small that the energy difference
based on the scattering of light by diffusing par- between incident and scattered photons can be
ticles. At any instant the suspended particles will neglected. However, the change in momentum ex-
have a particular set of positions within the scat- perienced by the photon during the scattering
tering volume. The particles scatter the radiation process is a very important parameter in DLS.
to the detector, but the relative phases of scattered The mopentum transfer vector or the scattering
wavelets differ, due to the differing incident phases vector, Q, is defined as,
that they experience at these positions and due to Q = ;, - zi
different particle-detector distances. The electric (1)
field at the detector is the superposition of the where the zs are the wave vectors of the scattered
fields due to all the scattered wavelets and will, at and the incident light. From the geometry of the
time t, have a value E(t). At the time, t + T, which scattering arrangement (see Fig. l), the magnitude
is a very small time later than t, the particles, of the scattering vector is,
which are diffusing, will have new positions,
slightly removed from those at the earlier time.
Superposition of the new slightly shifted wavelets
yields a changed electric field at the detector, E(t + T).
> Laser beam
As time progresses, the electric field and hence the
intensity at the detector, will fluctuate as the Brow-
nian processes in the sample volume continue.
During the scattering process, the electromag-

Food Research International 0963-9969/94/$07.00 Fig. 1. Scattering geometry shywing laser beam, wave vectors
0 1994 Canadian Institute of Food Science and Technology (ks), scattering vector (Q) and scattering angle 0.

195
196 F. Ross Hallett

i.0

I ---
I I

1
I I I

1
<I(t)%

I
2
..
*.
*.
0.
??
...
. .._
*...
*...
**.......*
.‘...........*
‘..............
1 <I(t)>2

” 30 60 90 120 150 160 10 20 30 40 50 60

SCATTERING ANGLE (degrees) CHANNEL NUMBER

Fig. 2. Distance scale (Q’) sampled by light (helium-neon Fig. 4. Normalized intensity autocorrelation function.
laser) as a function of scattering angle.

It is clear from Equation (2) that Q has the dimen- Smart, 1988) from which one can obtain the in-
sions of an inverse length. Indeed Q’ sets the length tensity autocorrelation function, C~(T), where r is
scale that will be probed by the light scattered at an instrumental delay time. This function has a T =
angle 0 (see Fig. 2). Thus, light scattered at low 0 limit equal to the mean square intensity ~1(t)~>
angle will contain information on the gross dynamic and an asymptotic limit as T approaches infinity
and structural properties of the scatters, whereas equal to the mean intensity squared, i.e. cI(~)>~.
light
Y--- scattered
- at high angle contains this informa- The decay time is rouehlv
o---J indicative of the tvnical
-ar-----
tion at a finer scale. Figure 3 shows a comparison fluctuation time of the signal. The normalized
of scattering techniques and the length scales that form of the intensity autocorrelation function
they probe. DLS has a greater range than other (Fig. 4) can be related to the electric field autocor-
scattering techniques because it is concerned with relation function through the relationship.
the distance (27r/Q) that a particle diffuses in a
time interval. g(2)(7) = -
Cl(T) = 1+ 1g(‘)(T)12 (3)
The main problem in data recovery in the DLS <I(t)>2
experiment is the extraction of quantitative infor-
A biock diagram of a modern DLS spectrometer
mation from a fluctuating signal. Small rapidly
is shown in Figure 5.
diffusing particles will yield fast fluctuations,
Usually, dynamic light scattering methods are
whereas larger particles and aggregates general
employed to study solutions under conditions
have relatively slow fluctuations. The rate of the
where particles are assumed to be small and
fluctuations can be determined through the tech-
spherical. Under these conditions, only the relative
nique of autocorrelation analysis (Abbiss &
phases of the scattered waves are of concern and

I
SMALL ANGLE NEUTRON SCATTERING

cl
Detector
I I II I
COLD SOURCE LIGHT SCATTERING

SMALL ANGLE X-RAY SCATTERING

Quantum
I I
Photometer

I I I
100 1000 10000

2n/Q (angstroms)

Fig. 3. Comparison of the particle sizing ranges of several

J
different scattering techniques. Fig. 5. Schematic diagram of a DLS spectrometer.
 Particle size analysis by dynamic light scattering 197
_
g(‘)(T) has the form: e-r7 = e-rTe(-(r-r)T)

1
g(l)(T) = e-W+
(4) (r-r)*T* (r-r)3T3 +
=e -” i-(r-f)T + (7)
where D is the diffusion coefficient. 21 - 31 ***
The theoretical function g(‘)(r) approaches a Substituting Equation (7) into Equation (6) and
maximum value of unity as T approaches zero.
taking the logarithm yields (Koppel, 1972)
This would also be true experimentally if the pin-
holes used to collimate the scattered light were in-
finitely small. However, no light would reach the h@‘)(T)] = 1IlU - FT + * T* - E”3
ti +. . . (8)
2! 3!
detector. In practice, an experimental compromise
is reached, with the effect that the zero delay (T = 0) The second moment,
intercept is ‘a’ where a I 1. In addition, Equation
p2 = (I*) (9)
(4) holds only for solutions of small monodisperse
particles (all particles are identical in size). The is the variance
more common situation is one where the solution
02” l-42-PI2 = EL2 (10)
contains a size distribution of scatterers. In this
situation g(l)(r) becomes a summation. Including since the first moment about the mean (pJ is al-
both these effects, Equation (4) can be written as, ways zero; o, the standard deviation, is a measure
of the width of G(I).
g(i)(r) = a (01e-D,Qz7+ ee-D2Qzr+ . . .)
A variety of mathematically more sophisticated
m (5) procedures have been developed to yield the inver-
= 0C w,e-D@T
sion of Equation (6). Most of these approaches
i=I
are variants of a discrete method in which
where the wi are weighting factors related to the
relative abundance of particles of a size indexed
by i and m is the number of sizes. For a continuous
VUr = (&T) - 5 % exP(-IJ))* (11)
distribution of particle sizes, Equation (5) can be is minimized with respect to the variables U, and
replaced by, rm. As mentioned earlier, this is notoriously
g(l)(r) = a [ G (I) eer’& unstable, and if no constraints were applied it is
(6)
essentially impossible to obtain a unique set of
where I = D@. best-fit parameters. One of the more common
In principle, a complete Laplace inversion of restraints is to specify, in advance of the fit, the
Equation 6 would yield the distribution of relax- range and values of a ‘trial’ set of I,. In the tech-
ation times, G(I), and hence the distribution of nique called exponential sampling (Ostrowsky et
particle sizes (through the Stokes-Einstein equa- al., 1981), the I, are themselves exponentially
tion). In general, such an inversion is termed ‘ill spaced according to the relationship
conditioned’ because of its mathematical instability
and because unattainably high precision in the ex-
r m+l
= r,exp(>o (12)
perimental data is required. Several alternative where x is a constant that sets the initial spacing
methods of various complexities have been devel- between the first two Is. The resulting set of OS
oped and brief treatments of two of these will be corresponds to the amplitudes, or relative weights
given. The first and most common method of pro- of each of the exponentials in Equation (11). Since
ceeding is called moments analysis or the method each I has a corresponding Y, the amplitudes are
of cumulants. A given distribution, such as G(I) usually presented as a histogram. If the data are
can be described in terms of a set of moments of sufficient quality (often runs of several hours
about the origin or moments about the mean. The are necessary), then reliable and reproducible his-
object of moments analysis is to obtain these mo- tograms of the radius distribution, G(r), can be
ments without actually performing the inversion. obtained.
Specifically one attempts to obtain the first and The histograms that are produced from this
second moments from which one can obtain the procedure are intensity-weighted distributions, i.e.
mean value of I and the variance, respectively. In the amplitudes represent the amount of light scat-
this approach the exponential, eerT from Equation tered by each particle size, r and decay constant I’.
(6) is expanded about the mean value emr7: In order to obtain number distributions, one must
198 F. Ross Hallett

centrated suspensions. The apparatus used in

DWS is the same as for DLS except that the inci-
dent light and the back-scattered light are both
transmitted and received by a fibre optic. DWS
depends upon multiple scattering being so severe
that the incident photons experience a random
walk (diffusing wave) in the sample before being
finally scattered into the returning fibre optic. Pine
et al. (1988) have shown that for back-scatter de-
tection from thick samples, the electric field auto-

0.00 0.02 0.04 0.06

I-L
RADIUS (micrometres)
0.08 0.10
correlation function is

g(l)(r) = exp [-y (67/r,,) ‘1 (13)

In this expression 7. is defined as a characteristic
Fig. 6. Size distribution of extruded vesicles obtained from DLS.
diffusion time
1
include the relative scattering ability of each size 70 = - (14)
in the distribution. This can be accomplished by Dko2
including the relative Rayleigh-Gans-Debye or and k, is the incident wave vector. The constant y
Mie scattering factors in the summation of Equa- must be determined independently by calibration
tion (11). Hallett et al. (1991) have shown how with known size scatterers or by transmission
this can be accomplished without the introduction measurements. Once this is done, however, accu-
of spurious oscillations into the histogram. The in- rate estimates of mean sizes have been obtained in
tensity and the number distribution for a set of a variety of dense scattering media.
vesicles are shown in Figure 6.
Dynamic light scattering requires that the con-
centration of scatters be sufficiently dilute that REFERENCES
only single scattering takes place. This means that
most samples must be diluted to the point where Abbiss, J. B. & Smart, A. E. (1988). Photon correlation tech-
they are optically almost clear. If multiple scatter- niques and applications. OSA Proceedings, 1.
Hallett, F. R., Watton, J. & Krygsman, P. H. (1991). Vesicle
ing occurs then multiple phase shifts to the scat- sizing: Number distributions by dynamic light scattering.
tered light occur, with the result that particle sizes, Biophys. J., 59, 357-62.
as determined by moments analysis or by discrete Koppel, D. E. (1972). Analysis of macromolecular polydis-
persity in intensity correlation spectroscopy - Method of
inversion techniques, are too small. As a result cumulants. J. Chem. Phys., 57, 481620.
standard DLS methods cannot be applied to more Ostrowsky, N., Sornette, D., Parker, P. & Pike, E. R. (1981).
concentrated suspensions such as milk. However, Exponential sampling method for light scattering polydis-
a related form of laser light scattering, called di- perse analysis. Opt. Acta., 28, 1059-63.
Pine, D. J., Weitz, D. A., Chaikin, P. M. & Herbolzheimer, E.
ffusing wave spectroscopy (DWS), is available and (1988). Diffusing-wave spectroscopy. Phys. Rev. Lett., 60,
provides estimates of mean particle sizes in con- 11367.