Journal of Colloid and Interface Science 222, 83–89 (2000)
doi:10.1006/jcis.1999.6599, available online at http://www.idealibrary.com on
Structure of Hydrous Ferric Oxide Aggregates
Bill Lo and T. David Waite1
School of Civil and Environmental Engineering, The University of New South Wales, Sydney, New South Wales 2052, Australia
Received June 4, 1999; accepted October 19, 1999
The small angle light scattering behavior of hydrous ferric ox-
ide flocs is examined here and found to provide useful insights into
the nature of the aggregates formed despite the large size of these
aggregates at later times. The flocs appear to exhibit fractal prop-
erties over a significant size range though the aggregates appear
to be easily disrupted through mixing effects resulting in breakup
and/or restructuring to denser assemblages. Background electrolyte
concentrations also have some impact on floc structure but mixing
effects and apparent destabilization by ferric ions limit the effect
of added electrolytes on the stability and structure of ferric oxyhy-
droxides. Similar estimates of fractal dimensions of these hydrous
ferric oxide flocs are obtained both by static light scattering anal-
ysis and by a cluster mass scaling approach. The choice of density
distribution cutoff function has some impact on derived size and
structure parameters and further refinement in this area is needed.
°
C 2000 Academic Press
Key Words: iron oxide; ferrihydrite; fractal; aggregate; light scat-
tering; structure.
INTRODUCTION
Iron oxides occur widely in nature and are used extensively
in water and wastewater treatment because of their ability to ad-
sorb contaminants. The so-called “amorphous” iron oxide (also
known as hydrous ferric oxide and ferrihydrite) is of particular
interest as it is the form that typically occurs initially in natural
aqueous environments prior to aging to more crystalline forms
and is the form that occurs in treatment systems when ferric
salts are hydrolyzed under neutral to alkaline conditions. The
crystallinity (as determined by X-ray diffractometry) of amor-
phous iron oxide may vary from the highly disordered 2-line
ferrihydrite to the more ordered 6-line form (1).
Considerable attention has been given to the structure of col-
loidal amorphous iron oxide in its nonflocculated form through
application of small angle X-ray scattering studies (2, 3). Thus,
Tchoubar et al. (2) report the formation of colloidal aggregates
whose size (at 400 s after commencement of hydrolysis) ranged
from 10 nm at low OH/Fe ratios to around 700 nm at high OH/Fe
ratios (but still below the flocculation threshold). The colloid
1 To whom correspondence should be addressed: Fax: +61-2-9385 6139.
E-mail: d.waite@unsw.edu.au.
83
size was observed to decrease during the first 30 min of hydrol-
ysis/nucleation with the final mean size centered around 10 nm.
Highly linear structures were observed at low OH/Fe ratios (ap-
parently as a result of long-range magnetic dipolar interactions)
but became more branched (exhibiting fractal dimensions in the
1.7 to 2.0 range) at higher OH/Fe ratios.
Once flocculation of these “precursor” colloids is induced (for
example, by reduction in surface charge through pH increase),
large aggregates are rapidly formed. These large aggregates are
typical of those formed in water and wastewater treatment where
the large size is a factor in their ease of sedimentation or their
removal by filtration. The structure of these aggregates remains
of interest since it will be a determinant (together with cluster
size) of their permeability—a factor of obvious importance in
size separation processes.
Given the large (relative) size of these aggregates, small angle
light scattering represents one possible means of probing their
structure. The light scattering behavior of aggregating amor-
phous iron oxides under various system conditions (electrolyte
concentration, mixing regime) is reported here and, based on the
results presented, conclusions are drawn on their structure as a
function of system conditions including background electrolyte
concentration and mixing regime.
THEORETICAL
In a scattering experiment, a beam of light (or X rays or neu-
trons) is directed onto a sample and the scattered intensity I
(in photon counts) is measured as a function of wave number
q = (4π n/λ) sin(θ/2) where θ is the scattering angle, λ is the
wavelength of the incident beam, and n is the refractive index of
the medium. If N (M) denotes the number of clusters of mass M
in the scattering volume and if interactions between the clusters
are negligible, the total scattering intensity is given by (4)
X
I (q) = N (M)I M (q), [1]
M
where I M (q) is the scattered intensity from clusters of mass M.
The choice of the scattering angle sets a length scale, given
by q −1 , on which the clusters are probed. This length scale must
be compared with a suitably defined length scale of a cluster of
mass M. Choosing the radius of gyration Rg , one may consider
0021-9797/00 $35.00
Copyright °C 2000 by Academic Press
All rights of reproduction in any form reserved.
84 LO AND WAITE
the two limits q Rg ¿ 1 and q Rg À 1. In the first case, the cluster
appears to the light probe essentially as a point particle and thus
scatters isotropically, independent of q. Therefore, I M (q) ∼ M 2
(4). In the second case, q Rg À 1, one may decompose the clus-
ter of radius Rg into smaller subunits (termed “blobs” by Lin
et al. (4)) of size q −1 . As long as these subunits contain many
colloidal particles and these particles exibit a fractal structure
within the subunit, the subunit mass will scale as m su ∼ (qa)−Df ,
where “a” denotes the radius of an individual colloidal particle
and Df is the fractal dimension. The phase differences between
the scattered fields from particles within a single subunit are al-
ways less than one. Thus, a single subunit scatters coherently,
and the scattered intensity of the subunit will be proportional
to m 2su . By contrast, the phase differences between the scattered
fields from different subunits are always greater than one. Thus,
the phases of the electric fields scattered from different subunits
add incoherently and the total intensity will be the sum of the
intensities scattered from each subunit. The number of subunits
in a cluster of mass M is M/m su ∼ M(qa) Df . Therefore, the
total scattered intensity is I M (q) ∼ (M/m su )m 2su ∼ M(qa)−Df or
I M (q) ∼ M 2 (q Rg )−Df (4).
Since the factor M 2 appears in both limiting cases, it is con-
venient to write I M (q) = M 2 SM (q), where S M (q) denotes the
“structure factor” of clusters of mass M (4). The total scattered
intensity (from Eq. [1]) is therefore given by
X
I (q) = N (M)M 2 S M (q), [2]
M EXPERIMENTAL
which is a measure of both the cluster mass distribution N (M)
and the structure of the clusters.
The structure factor reflects, in q space, the distribution of
scattering material in real space; S M (q) is the Fourier trans-
form of g(r ), the density correlation function. For a three-
dimensional, self-similar object with fractal dimension Df , this
function scales as g(r ) ∼ r Df −3 . For colloidal aggregates, how-
ever, self-similar scaling occurs only over a limited range of
length scales. The upper limit for the scaling form of g(r ) oc-
curs when r reaches the radius of gyration, Rg . The lower limit
of the scaling is given by the primary particle radius, a. The ex-
istence of the upper limit introduces a cutoff function h(r/Rg )
such that
g(r ) ∼ r Df −3 h(r/Rg ), [3]
with h(x) = 1 for x ¿ 1 and h(x) → 0 for x À 1. A variety of
cutoff functions have been used (5–7) including exponential,
Gaussian, “overlapping sphere,” and stretched exponential ex-
pressions. The exponential cutoff function h(r/ξ ) = e−(r/ξ ) is
the simplest and has the correct asymptotic behavior but has no
physical basis. It has been extensively used though Sorensen and
co-workers (6, 7) have shown that this function is inappropriate
if the system exhibits some polydispersity and argue that the
Gaussian cutoff function h(r/ξ ) = e−(r/ξ ) is a better choice. A
2
number of other authors including Jullien (8) and Yanwei and
Meriani (9) have also argued that the Gaussian cutoff function
provides a better description of light scattering by finite-sized
fractal aggregates. Expressions for ξ 2 and S(q) for the Gaussian
cutoff function are given below
4Rg2
ξ2 = [4]
Df
µ ¶
−(q Rg )2 /Df ∗ 3 − Df 3 (q Rg )2
S(q) = e 1 F1 , ; , [5]
2 2 Df
where 1 F1 is the hypergeometric function.
It should be noted that the assumption that the phase difference
between scattered waves depends only upon their location in the
cluster and is independent of any material property of the cluster
(the Rayleigh-Gans-Debye approximation) is valid as long as
|m − 1| ¿ 1 [6a]
4πa
|m − 1| ¿ 1, [6b]
λ
where m is the refractive index of the particle relative to the sus-
pending medium (10). It is this condition that requires |m − 1|
to become smaller and smaller as the RGD theory is applied to
progressively larger particles.
Ferrihydrite Preparation and Aggregation
The amorphous iron oxide, ferrihydrite, was prepared by
adding FeCl3 from a stock solution of 0.5 wt% to a 500-mL
mixture of 1 mM NaHCO3 and either 1 or 100 mM NaCl. Homo-
geneity was achieved by gentle swirling. Solutions of 10 mg/L
FeCl3 were typically prepared which, at the pH induced by the
carbonate/CO2 buffer (pH 7.8), rapidly hydrolyzed then nucle-
ated to colloidal amorphous iron oxide. Aliquots (8 mL) of these
suspensions were immediately transferred to the small cell of the
Malvern Mastersizer where the amorphous iron oxides contin-
ued to aggregate. This small cell either was not stirred or was
stirred at one of two stirrer speeds (“moderately” or “rapidly”)
using a magnetic stirring bar.
Aggregate Size and Structure Measurement
Size distributions and structural information of amorphous
iron oxide aggregates were determined as a function of aggre-
gation time using a Malvern Mastersizer/E which ascertains size
by analysis of forward scattered light. The Mastersizer consists
of a 5 mW He-Ne laser (632.8-nm wavelength) with 18-mm
beam expansion and collimation. The direction of the polarized
laser is parallel to the detector axis (vertical). The particles pass
through the expanded and collimated laser beam in front of the
optic lens in focal plane of which are positioned 31 photosensi-
tive detectors. The Fourier optics can be used with 45-, 100-, or
 STRUCTURE OF HYDROUS FERRIC OXIDE AGGREGATES
300-mm lenses which allow the collection of light scattered at
angles from 0.01 to 32.5◦ . In this work, the flocs were analyzed
using the 300-mm lens, enabling the collection of light scattered
from 0.03 to 6.25◦ .
Size distribution information was obtained using the software
supplied and involved calculation of the size distribution by an
iterative algorithm which matches the measured and calculated
energy distributions over the detectors. The fitting procedure is a
model-independent technique (i.e., it does not constrain the vol-
ume distribution to follow some common analytical expression
such as the log normal distribution function) and uses Fraun-
hoffer theory to develop a scattering pattern assuming a popu-
lation of equivalent spheres. Information on distribution size is
presented in this paper principally as volume average diameter
(VAD) where
X ÁX
VAD = (Vi di ) Vi , [7]
i i
where Vi is the relative volume in size class i with mean class
diameter di .
Information on assemblage structure was obtained by mea-
suring the intensity of light scattered at all available detectors
and plotting log intensity versus log q. Information required to
compute the angle of each detector to the incident beam as well
as size and responsiveness of each detector (the so-called “magic
numbers”) was supplied by Malvern Instruments.
RESULTS AND DISCUSSION
With gentle mixing in the Malvern cell, the (equivalent spher-
ical) size of amorphous iron oxide aggregates evolve over time
from distributions exhibiting modal diameters in the low mi-
crometer range at early times (within the first minute after addi-
tion to the Malvern cell) to modal diameters in excess of 100 µm
after a 15-min aggregation time. As can be seen from Fig. 1, the
FIG. 1. Amorphous iron oxide size distributions obtained using the Malvern
Mastersizer as a function of aggregation time.
85
FIG. 2. Conventional static light scattering plots (log I versus log q) for
amorphous iron oxides as a function of aggregation time.
distributions obtained from the Malvern fits to the light scat-
tering data are reasonably log normal in shape with relatively
narrow, consistent spread in aggregate size over time.
The log I versus log q light scattering plots shown in Fig. 2
reflect increasing power law behavior with aggregation time
with at least one decade of linearity (10−4 nm−1 < q < 2 ×
10−3 nm−1 ) observed after about 6 min of aggregation. Sim-
ilar behavior, albeit over considerably longer aggregation times,
has previously been reported for silica particles (11).
Given the large size of particles developed in these amor-
phous iron oxide aggregation studies, the question of whether
the primary particles maintain independent scattering must be
raised before interpreting these scattering plots further (i.e., is
the Rayleigh-Gans-Debye approximation valid). As discussed
earlier, the RGD approximation is valid when both |m − 1| ¿ 1
and 4πa/λ|m − 1| ¿ 1 are satisfied. Of prime importance here
is the nature of the primary scatterers. As discussed in the intro-
duction, Bottero and co-workers (2, 3) have shown that colloidal
ferrihydrite particles are initially formed which exhibit fractal di-
mensions in the 1.7–2.0 range and, as such, are relatively porous
and would be expected to exhibit low refractive indices (most
likely m < 1.1). The size of these aggregates (which may be
considered our primary scatterers at the wave vector (q) mag-
nitudes being used here) is unclear and, from Tchoubar et al.’s
work (2), could range from 10’s to 100’s of nanometres. Elec-
tron microscopy studies reported by Cornell and Schwertman
(1) suggest that these “core” aggregates are in the 50 to 300 nm
size range. Given that these primary particles are themselves
aggregates of considerable porosity, their refractive index will
be much closer to unity than that of crystalline iron oxides. For
example, at a refractive index of 1.1 and a primary scatterer size
(a) of (say) 300 nm, the RGD conditions may be considered
to be satisfactorily met. It should also be noted that early work
by Kerker (12) and more recent studies by Farias et al. (13)
86 LO AND WAITE

(average) radius of gyration of the clusters. In reality, of course,

a distribution in cluster sizes exists and, as might be expected,
improved fits of the scattering model to the data are obtained by
introducing some polydispersity in cluster size (as defined by
a log normal distribution in radius of gyration) into the fitting
process (see Fig. 3). As shown in Table 1, increasing “spread”
in Rg is observed to have only a small effect on the estimated
Df value.
Care should be taken in interpreting the derived Df values as a
definitive measure of the fractal dimension of the assemblage as
the choice of cutoff function clearly has an impact on the values
obtained by fitting of the scattering equation to the obtained
I versus q data. While we have adopted (for good reason) a
Gaussian cutoff function in all data analysis presented here, it
is clear that some shift in parameter values might be obtained
if alternate descriptions were adopted. For this reason, we will
consistently refer to Rg and Df values as “apparent” radii of
gyration and fractal dimensions, respectively.
Fits of the full light scattering equation to aggregates devel-
oped in the Malvern cell with moderate mixing are shown in
Fig. 2. A Gaussian cutoff function (assuming an average Rg
FIG. 3. Light scattering plot (at t = 6 min) showing Gaussian fit assuming value in all cases) has been used in all fits. The time evolution
both a single (average) Rg value and a log normal distribution of Rg values. in selected measured (Imax and VAD) and fitted (Rg and Df )
parameters is given in Table 2.
indicate that these conditions may be relaxed considerably, par- The results reported in Table 2 suggest an increase in “com-
ticularly for the case of low angles of scatter. Indeed, Heimenz pactness” of amorphous iron oxide aggregates over the course
and Rajagopalan (14) note that for a scattering angle of 10◦ , the of the aggregation process and may reflect the fragility of these
RGD approximation is good to within 10% for spheres with a aggregates and their susceptibility to breakup and/or restructur-
radius that is about 62, 37, and 25 times that of the wavelength ing on mixing. The effects of mixing appear to be even more
for refractive indices of 1.1, 1.2, and 1.3, respectively. Increasing distinct in the results shown in Fig. 4. In this case, flocs have
effects of multiple scattering and interaction between particles,
however, are to be expected as the aggregate size and compact-
TABLE 2
ness increase.
Measured and Derived Parameter Values for Amorphous Iron
As outlined under theoretical, the full light scattering data
Oxides as a Function of Aggregation Time (in Small Malvern Mea-
set can be used to obtain an estimate of the fractal dimension surement Cell with Moderate Mixing)a
(Df ) and the size cutoff value (ξ ) (or radius of gyration, Rg )
provided an appropriate expression for the density correlation Time VAD Imax Rg
cutoff function is used. The best fit result for t = 6 min data using (min) (µm) (arbitrary units) (µm) Df
a Gaussian cutoff function is shown in Fig. 3 and fit parameters
0 5.2 5.4 3.0 1.48
shown in Table 1. These results are obtained assuming a single 1 6.3 7.5 3.1 1.85
2 8.2 21.6 3.8 2.09
3 11.1 66.4 4.8 2.22
TABLE 1 4 14.9 89.1 6.7 2.27
Average Radius of Gyration (Rg ) and Apparent Fractal Dimen- 5 19.9 221.5 9.4 2.27
sions (D f ) Obtained Using a Gaussian Cutoff Function in the Light 6 27.2 654.3 12.4 2.33
Scattering Equation Used to Fit Log I Versus Log q Data Obtained 7 36.4 981.8 15.2 2.36
after Aggregation with Moderate Stirring for 6 mina 8 48.6 1782.0 20.0 2.39
9 60.5 2953.0 24.0 2.41
Rg 10 67.8 4338.9 27.3 2.45
(µm) Df 11 80.0 6523.2 32.1 2.46
12 92.9 9078.4 33.0 2.47
Single (average) Rg 12.35 2.33 13 111.7 14603.4 33.5 2.54
Distribution of Rg 12.35 ± 1σ 2.42 14 121.3 17258.0 33.9 2.56
15 128.3 19953.5 34.0 2.58
a Also shown is the effect of accounting for the presence of polydispersity

in cluster size by using a log normal distribution of Rg values (a log σ = 0.2 is a Radius of gyration and apparent fractal dimension values are obtained by

assumed). fitting the complete light scattering equation using a Gaussian cutoff function.
 STRUCTURE OF HYDROUS FERRIC OXIDE AGGREGATES
FIG. 4. Effect of mixing conditions on volume average diameter and ap-
parent fractal dimension of amorphous iron oxide flocs over time.
been gently added to the measurement cell and either the cell
not stirred or the cell stirred via a magnetic stirring bar rotating
either slowly (“moderate” stirring) or rapidly (“rapid” stirring).
As can be seen from the floc size and structure results over time,
mixing condition has a dramatic impact upon the size and appar-
ent structure of the resulting flocs. With no stirring, flocs grow
very slowly (<20 µm in 20 min) and light scattering intensity
is low. The apparent fractal dimension of these flocs eventually
approaches a relatively constant value of 2.1–2.2. With “moder-
ate” stirring, the VAD exhibits an exponential increase over time
to about 100 µm after only 10 min and the apparent fractal di-
mension increases from under 1.9 to approximately 2.5 in about
10 min. With more rapid stirring, the flocs commence growing
very quickly (as measured by the VAD) reaching almost 50 µm
in about 2 min. After this time, however, floc growth is curtailed
with a reduction in volume average diameter observed on con-
tinued stirring. The apparent fractal dimension is observed to
rapidly increase on this 2- to 3-min time scale and plateaus at
just over 2.5 after about 5 min of mixing.
Another factor that has a significant influence on aggregation
kinetics and resulting aggregate structure in many colloidal sys-
tems is the background electrolyte concentration. In the case of
ferrihydrite, increasing electrolyte (NaCl) concentrations from
87
TABLE 3
Effect of Background Electrolyte Concentration on Fractal Di-
mension of Resulting Aggregates under Conditions of No Mixing
and “Moderate” Mixing in the Small Malvern Measurement Cell
Df
[NaCl] Analysis time No mixing Moderate mixing
10 mM 15 min 2.07 2.58
100 min 2.29
100 mM 15 min 1.99 2.50
100 min 2.08
10 to 100 mM (the concentration range normally found to induce
colloid destabilization) did not have a large effect on aggrega-
tion rate, presumably because sufficient ferric ions are already
present in solution to induce rapid aggregation. As shown in
Table 3, a small effect of salt in inducing a more open struc-
ture is observed though the results are by no means as dramatic
as those in standard cases of reaction-limited versus diffusion-
limited aggregation where fractal dimensions of 1.7–1.8 are typi-
cal of DLA while fractal dimensions of 2.2–2.3 are more typical
of RLA. The obtained results are also complicated by the in-
crease in fractal dimension over the course of aggregation that
was described above.
While an observable difference in fractal dimensions of flocs
is observed under low salt compared to high salt conditions when
aggregates are formed with no stirring, the presence of mixing
in the cell leads to rapid formation of more compact aggregates
for which the Df , once in the 2.5–2.6 range, appears relatively
stable.
A number of additional insights may be gleaned from the in-
formation presented in Table 2. For example, a strongly linear
correlation is observed between the measured volume average
diameter and the estimated radius of gyration (Fig. 5). Such a
relationship is to be expected if the flocs maintain a uniform
shape as aggregation occurs. Additionally, at low q, where the
FIG. 5. Radius of gyration (Rg ) as deduced from modeling of static light
scattering data for hydrous ferric oxide aggregates using Gaussian cutoff function
versus volume average diameter (VAD) of the aggregates.
88 LO AND WAITE
scattered intensity reflects scattering from the complete aggre-
gates and the scattered intensity reaches a q-invariant maximum,
Imax ∝ MN (M)M. Now for an aggregating system where we are
not creating or losing mass (assuming no sedimentation is oc-
curring), we would expect N (M)M to be constant. If this is the
case, then we would expect to see Imax ∝ M. For a fractal ag-
gregate for which M ∝ RgDf , Imax should increase according to
RgDf ; that is, a plot of log Imax versus log Rg should provide a
measure of the fractal dimension. Such plots are shown in Fig. 6
and, apart from results at early times when the flocs are small
and still establishing their fractal character, do exhibit linear re-
lationships between log Imax and log Rg over a significant size
range. The fact that the log Imax versus log Rg plots are linear and
not showing a gradual increase in Df with time (as found from
the log I versus log q fits) may suggest that the change in Df val-
ues derived from the log I versus log q plots is a function of the
fitting process (a possibility, especially when using a Gaussian
FIG. 6. Log-log plot of maximum scattering intensity (Imax ) versus Rg for
the flocculation of amorphous iron oxides in the presence of (a) 10 mM NaCl
and (b) 100 mM NaCl. Data points used in regression analysis shown as solid
symbols.
cutoff function) or may indicate that the value obtained in the
low q region using the mass scaling behavior (Mα Imax α RgDf ) is
an averaged value with small scale change indiscernible.
Results for low salt conditions (10 mM NaCl) are shown in
Fig. 6a while those for a high salt concentration (100 mM NaCl)
are given in Fig. 6b. Satisfyingly, the fractal dimensions esti-
mated from the slope of the linear portion of these plots (2.68
and 2.61 for the 10 and 100 mM NaCl cases, respectively) are
not too dissimilar to the values derived from analysis of the static
light scattering data (I vs q plots) using the full scattering equa-
tion incorporating the Gaussian cutoff function (2.58 and 2.50
for the 10 and 100 mM NaCl cases, respectively). Even closer
agreement could no doubt be obtained by incorporating some
degree of polydispersity into the radius of gyration term used in
fitting the full scattering equation as discussed earlier.
CONCLUSIONS
Static light scattering appears to be a useful tool for investi-
gation of the structure of ferric oxyhydroxide flocs despite the
large size of aggregates formed. The finite size of ferric oxide
flocs examined necessitated the use of a density distribution cut-
off function in fitting the complete light scattering data with the
choice of function having some impact on size and structure
parameters determined.
The flocs appear to exhibit fractal properties over a signifi-
cant size range though this fractal structure appears to be easily
disrupted through mixing effects resulting in breakup and/or
restructuring to dense aggregates. Background electrolyte con-
centrations also have some impact on apparent floc structure but
mixing effects and apparent destabilization by ferric ions limit
the effect of added electrolytes on the stability and structure of
ferric oxyhydroxides. Similar estimates of fractal dimension of
hydrous ferric oxide flocs are obtained by analysis of both the
static light scattering (I vs q) data and the cluster mass scaling
(Imax α M vs Rg ) behavior.
ACKNOWLEDGMENTS
The CRC for Waste Management and Pollution Control is thanked for the
provision of a postgraduate research scholarship to Mr. Lo.
REFERENCES
1. Cornell, R. M., and Schwertmann, U., “The Iron Oxides–Structure,
Properties, Reaction, Occurrences and Uses.” VCH Publ., Weinheim,
1996.
2. Tchoubar, D., Bottero, J.-Y., Quienne, P., and Arnaud, M., Langmuir 7, 398
(1991).
3. Bottero, J.-Y., Manceau, A., Villieras, F., and Tchoubar, D., Langmuir 10,
316 (1994).
4. Lin, M. Y., Klein, R., Lindsay, H. M., Weitz, D. A., Ball, R. C., and Meakin,
P., J. Colloid Interface Sci. 137, 263 (1990).
 STRUCTURE OF HYDROUS FERRIC OXIDE AGGREGATES
5. Schmidt, P. W., in “The Fractal Approach to Heterogeneous Chemistry: Sur-
faces, Colloids, Polymers” (D. Avnir, Ed.), p. 67. Wiley, New York, 1989.
6. Sorensen, C. M., Cai, J., and Lu, N., Langmuir 8, 2064 (1992).
7. Sorensen, C. M., in “Surface and Colloid Chemistry” (K. S. Birdi, Ed.),
Chap. 13, p. 533. CRC Press, Boca Raton, FL, 1997.
8. Jullien, R., J. Phys. I (France) 2, 759 (1992).
9. Yanwei, Z., and Meriani, S., J. Appl. Cryst. 27, 782 (1994).
10. Bohren, C. F., and Huffman, D. R., “Absorption and Scattering of Light by
Small Particles.” Wiley, New York, 1983.
89
11. Cannell, D. S., and Aubert, C., in “On Growth and Form: Fractal and Non-
Fractal Patterns in Physics” (H.E. Stanley and N. Ostrowsky, Eds.), p. 187.
Martinus Nijhoff Publ., Dordrecht, 1986.
12. Kerker, M., “The Scattering of Light and Other Electromagnetic Radiation.”
Academic Press, New York, 1969.
13. Farias, T. L., Koylu, U. O., and Carvalho, M. G., Appl. Optics 35, 6560
(1996).
14. Heimenz, P. C., and Rajagopalan, R., “Principles of Colloid and Surface
Chemistry,” 3rd ed., p. 218. Marcell Dekker, New York, 1997.