You are on page 1of 9

~ Pergamon Wat Sci. Tuh. Vol. 36. No 4. pp. 15-23.1997.

1997 IA WO. Published by Elsev,er Science LId


Printed In Great Bntain.
0273-1223197 $17'00 + 0'00
PH: S0273- I 223(97)004 I 4-9

PARTICLE SIZE DISTRIBUTIONS IN


TREATMENT PROCESSES: THEORY AND
PRACTICE

Desmond F. Lawler
Department afCivil Engineering, ECl8.6, University afTexas. Austin, TX 78712,
USA
ABSlRACf

The removal of particles from water and wastewater streams is essential and is usually accomplished through
flocculation, sedimentation, and filtration. Understanding particle removal requires understanding of particle
heterodispersity, especially with respect to size. Ajoint mathematical and experimental approach to studying
changes in size distributions in these processes has proven quite insightful. Recommendations for analysts
and manufacturers of particle size distribution analyzers based on mathematical principles are elucidated.
flocculation modeling is quite advanced and can be used predictively under certain well defmed conditions,
but a full description of changes in the size distribution is not available for all conditions encountered.
Insights from flocculation modeling in the last ten years might make significant design and operational
differences in the next ten years. Filtration modeling is not as advanced, inasmuch as filtration is a far more
complex treatment process, but size distribution measurements have increased understanding enormously,
and modeling has enabled design and operational changes in common practice to be understood.
1997 IA WQ. Published by Elsevier Science Ltd

KEYWORDS

Particles; particle size distributions; flocculation; mtration.

INTRODUCfION

The removal of particles from water and wastewater has been of major interest to environmental engineers
throughout the twentieth century. Many pollutants in water and wastewater are particles, or are made into
particles prior to ultimate removal Particles have several properties that influence their behavior in water and
therefore influence cur ability to remove them. These properties include size, shape, density, surface charge,
settling velocity, and, perhaps, porosity. In suspensions of interest in environmental engineering practice,
these properties all vary from particle to particle; there is a distribution of particle sizes, a variety of shapes, a
range of densities, etc. The heterodispersity with respect to some of these properties is still beyond our
ability to measure and account for; however, size heterodispersity is reasonably understood and is of primary
interest in this paper.

The objectives in this paper are as follows: (1) to show how understanding panicle size heterodispersity
yields insight into panicle treatment processes, (2) to show how understanding based on both mathematical
modeling and experimental behavior often leads to greater insight than either one alone, and (3) to both draw
conclusions about our current state of understanding and indicate directions in which that understanding
needs to increase in a few selected areas of particle removal processes. The paper draws heavily on the work
of my former students and focuses on two particle removal processes: flocculation and filtration.
IS
16 D. F. LAWLER

RESEARCH APPROACH
Scientific or engineering research is undertaken to answer specific questions. Two approaches to answering
the questions can be considered. One approach is to translate the question into mathematical terms and solve
the mathematics; the mathematical results constitute an answer to the question being raised. For the answer to
be correct, all three steps must be correct-the formulation of the question, its translation into mathematical
terms, and the solution. Assumptions are often made to simplify the problem; if those simplifications are
erroneous, the answer will be erroneous. The second approach is to translate the question into an
experimental investigation, at bench, pilot, or full scale. Experimental results also constitute an answer to the
question. Again, the translation to an experimental setup, the carrying out of the experiments, and the
interpretation of the results all must be done with precision to obtain a proper answer to the question.

If both the mathematical and experimental approaches are taken, then the two answers can be compared. If
the answers agree, the understanding of the process or question under investigation is probably correct,
especially if that agreement is robust under a variety of conditions. If the two answers do not agree, the
source of the disagreement can often be determined through the comparision. The comparison can lead to the
insight about how the mathematical model or the experiments have to be changed to more properly answer the
question. If only one approach is taken, the answer is presumed correct; if both approaches are taken, they
form an internal check: on the correctness of the answers, making the research tighter. or course, not all
problems can be translated well into one type of investigation or the other. IDustrations of this duel approach
are discussed subsequently for flocculation and filtration. In both cases, the questions asked relate to particle
size distributions. Prior to that, concerns about the measurement and mathematical representation of particle
size distributions are considered.

PARTICLE SIZE DISTRIBUTIONS


Suspensions in water typically have particle size distributions that are essentially continuous; i.e., no matter
how finely one divides the overall size range, particles exist in every size increment On the other hand, there
is presumably an upper limit to the size of particles in any suspension. A cumulative number concentration
(N(d, distribution would always be increasing as the particle size (diameter, or equivalent spherical diameter
for non-spherical particles) increased, until some maximum size is reached. The slope or derivative at any
. 0 f suc h a dis'b'
pomt tn Ullon, dN(d p ) , is known as teh Ie SlZC
paruc . dis'b
tn utl'on fun'
ctlon, usu all'
y gIven the
dd p
symbol n(d,). As shown in a variety of sources (e.g., Stumm and Morgan. 1996), the number, surface area,
and volume distributions can be related to the particle size distribution function.

Several investigators have reported size distributions in terms of the power law distribution, described as
follows:
n(d,) = A d,~ or log n(d,) = log A -\3 log d, (Eq. 1)
The power law distribution is convenient mathematically but also leads to some difficulties at any constant
value of p. Two assumed distributions are shown in Figure la; one follows the power law distribution with a
constant value of \3 3, while the other has an increasing value of \3 with increasing size. Both distributions
yield a total suspended solids concentration of approximately 25 mYL if the particles have a density of 2
g/cm'. The associated number and volume distributions are shown in Parts b and C of Figure I, respectively.
For the case of the constant \3 = 3, the number distribution increases with decreasing size and the volume
distribution increases with increasing size. Recognizing that the total number concentration in any size range
is the integral under the number distribution in that range, one can show mathematically and see graphically
that the total particle number concentration would approach infinity as the lower size limit decreased toward
zero (log d, =-00), if the particle size distribution function continued to lower sizes at that slope indefinitely.
Similarly, the total volume concentration (the area under the volume distribution) would approach infinity as
the upper size limit increased, if the panicle size distribution function continued to higher sizes at that slope
indefmitely. Both situations are impoSSible. For closure of the number and volume distributions, the slope
of the particle size distribution function (i.e., the value of \3) must be less than one at the lower end of the size
range and greater than four at the upper end of the size range.
Particle size distributions in treatment processes 17

The second distribution shown on Figure 1 meets these criteria. The distribution is assumed for the sake of
illustration here; it fits the power law distribution given in Eq. 1 but the exponent ~ varies with diameter:
~=~~~ ~~
This distribution meets the criteria that ~ is less than one at the lower end of the distribution and greater than
four at the upper end. The associated number and volume distributions clearly indicate "closure" of the total
number and volume concentrations; i.e., these values are finite for the range shown and would be finite if the
range were extended and the particle size distribution function continued with the same characteristics (or
even a constant slope at each end at the value taken on at the ends of the range shown).

These criteria. that the particle size distribution function should be measured to small enough sizes that ~ < 1
at the lower end and ~ > 4 at the upper end, should be adopted as a goal by equipment manufacturers and
water analysts. At the lower end, this goal is lofty; available electronic particle counters are not capable of
measuring accurately the size distribution at small enough sizes to meet this goal for most water suspensions.
Most instruments are capable of meeting the upper end criterion if used carefully; at the upper end. the particle
concentrations are so low that it is difficult to measure the distribution accurately.

State o/the art. In suspensions of water and wastewater, particle size distributions are usually
measured with particle counters in which an electronic signal is created for every particle in a sensing zone.
The signal is related to the particle size; as each particle passes through the sensing zone. the signal is created,
sized, and counted. The sizing is done in increments--all signals within a certain range are counted as equal,
resulting in a discrete rather than continous distribution. The distribution (number of counts in each range of
signal sizes) is interpreted as the particle size distribution, when the relationship between signal size and
particle size is known. That relationship depends on the methodology of the instrument; light scattering, light
blockage, electrical resistance, and imaging are all used, and each has a different relationship between signal
size and particle size. In some cases, the signal size is dependent on more than particle size (e.g., refractive
index), leading to some possible fuzziness or error in translating signal size to particle size.

In many instruments, the discrete size increments are large. The resulting information is sufficient for plant
operation, but the small number of data points on a distribution curve might be easily misinterpreted. If an
instrument measured six equally spaced points from 2 to 20 JlID (log dp-o.3 to 1.3) from the curved
distribution in Figure la. the few points would not allow one to see the curvature, and the resulting
distribution might be characterized as fitting the power law distribution with constant~. Several distributions
reported in the literature as fitting the power law distribution seem to have this characteristic.

PARTICLE SIZE DISlRIBlTI10NS IN fLOCCULATION


The mathematical description of changes in particle size distributions in flocculation was first given by
Smoluchowski (1917) who wrote a population balance for particles in suspension. In discrete form, one
equation must be written for each particle size (size class) considered. The number concentration of each
particle size increases by the formation of a floc of that size from two smaller particles and decreases by the
formation of a floc of larger size from two particles, one (or both) of which is the size under consideration.
Only two-particle collisions are considered, and they can occur by Brownian motion, fluid motion (shear),
and differential sedimentation. Olanges in the particle size distribution through flocculation can be predicted
by simultaneous numerical integration of the set ofrate equations (Lawler et al., 1980).

Collision frequency functions for each of these three collision mechanisms were developed by Smoluchowski
and others many years ago. These original collision frequencies were based on consideration of long range
forces only; they are referred to as rectilinear collision frequency functions because, for the non-Brownian
mechanisms, the particles are assumed to follow straight paths. Particles that are heading toward one another
are considered to continue on a straight path until they collide. This view ignores hydrodynamic effects (i.e.,
water between the particles tends to push the particles away from each other). When hydrodynamic effects
and van der Waals attraction are accounted for, the paths are curvilinear and far fewer collisions are predicted
to occur. The ramifications of the difference between rectilinear and curvilinear flocculation modeling are
explained in Han and Lawler (1992).
18 D.F.LAWLER

Differences in the curvilinear and rectilinear models are dramatic in at least two ways. First, collisions an:
considerably less frequent in the curvilinear than in the rectilinear model In particular, collisions between
large and small particles by fluid shear and differential sedimentation are predicted to be far less. The
curviIinearcorrection to the rectilinear model (i.e., the fraction of collisions predicted in the rectilinear model
that is predicted to occur in the curvilinear model) for fluid shear is shown in Figure 2. This figure is
corrected from the one originally published in Han and Lawler (1992) in which the ordinate values were low
by a factor varying from 0.5 to 0.125 depending on the size ratio (smaller to larger) of the two particles; the
calculational table for these values in the original publication is correct. The results on the figure show that as
few as one out of ten thousand (i.e., 10--) collisions predicted by the rectilinear model really occur according
to the curvilinear model This extreme condition is at low values of HA and low values of the size ratio.

Second, the dominant mechanism (i.e., the one responsible for more collisions than either of the other two)
for collisions for different combinations of particle sizes is quite different in the two models, as shown in
Figure 3. Brownian collisions dominate in the curvilinear model when one particle is small, whereas both
must be small in the rectilinear model. Differential sedimentation also has an expanded region of dominance
in the curvilinear model. Auid shear is far less important as a collision mechanism in the curvilinear model
than in the rectilinear model. This result calls into question the primacy of the velocity gradient as the most
imponant design variable in flocculation, as explained more fully in Han and Lawler (1992).

The substantial difference in the two models prompted experimental research to test which model was more
correct. Ii (1996) performed numerous flocculation experiments under a wide variety of chemical conditions
and with several suspensions. These experiments were done in an Couette flow apparatus, i.e., two
concentric cylinders separated by a small annular gap, in which the outer cylinder rotates while the inner
cylinder is stationary. This reactor has a nearly uniform and well-known Velocity gradient, although
secondary flows detract a bit from the ideality. In batch experiments, samples were removed occasionally
and the complete particle size distribution measured on a Coulter Counter Multisizer (Hialeah, Fl.).

Results from one of U's experiments with a suspension of PVC particles destabilized with a low dose of
alum are shown along with predictions of both models in Figure 4. At this dose, alum destabilizes by charge
neutralization, and no large precipitates of Al(OHh are formed. Both models require calibration; the values of
the collision efficiency factor, a, were found by fitting the model to results from another experiment with the
same suspension and chemical conditions but a different velocity gradient (0). The higher value of (1 for the
cmvilinear model reflects the fact that the curvilinear model predicts fewer collisions. The input to the models
was the same as the experimental results at time zero, except that the particle size distribution was extended
below the measured lower limit in recognition that smaller particles flocculate into the measureable range.
Althought not perfect, the curvilinear model clearly fits the experimental results after 150 minutes of
flocculation far better than the rectilinear modeL These results are typical of all results in this research under
charge neutralization conditions; complete results will be published in a future paper.

Model testing against experiments using alum at higher (sweep floc) doses were not nearly so encouraging.
In this case (not shown), neither model fit well. Experimental results showed clearly the formation of large
aluminum hydroxide floes (with original particles incorporated) and a dramatic reduction of small particles.
The models, which do not inCOIpOrate a mathematical description of precipitation kinetics, are incapable of
predicting these changes. The rectilinear model appears to fit the experimental results better, but this better fit
is apparently caused by the offset of two errors: oveIpfedicting collisions of small and large particles, and
ignoring formation of precipitates.
State a/the an. flocculation is the only process designed to change the size distribution without actual
removal of the particles from the water. Capabilities to measure and model these changes in the size
distribution have increased dramatically in recent years. The experimental and mathematical progress, of
which the author has only been a small pan, has increased understanding of flocculation. The developing
understanding of flocculation kinetics is likely to influence design and operation of water and wastewater
flocculation facilites. For example, if the curvilinear model proves to be generally more correct than the
rectilinear model, flocculators could be designed with lower values of velocity gradient than common today.
Particle size distributions in treatment processes 19

However, the learning progress must continue, as so much is still unknown. The impact of rapid mixing and
chemical addition on the subsequent flocculation in slow mix conditions, the kinetics of chemical precipitation
(particularly the size distribution of precipitates and interaction with the existing particles in suspension), the
fractal characteristics of flocs, the ability of water to pass through porous flocs and mitigate the need for the
curvilinear model: all these topics influence the changes in the size distribution in flocculation and need far
more exploration before theoretical predictions will match experimental and full-scale results.
FILTRATION
Deep bed filtration is commonly employed in both drinking water and wastewater treatment and generally
works quite well. Nevertheless, our understanding of particle behavior in filters is remarkably incomplete
due to the complexity of the process. Filters are never at steady state; as particles are captured, they become
part of the media, leading to increased removal efficiency (ripening) and/or decreased removal (breakthrough
caused by lack of capture or breakoff of previously captured particles).

Filtration dynamics was studied in our laboratory, beginning with a focus on ripening with suspensions of
latex spheres. Later experiments were performed with drinking water and wastewater suspensions for long
times so that not only ripening but breakthrough was evident. Complete size distributions of samples from
several different depths were measured. Several articles (e.g., Darby and Lawler, 1990, on latex; Clark, et
al., 1992, on contact filtration; Kau and Lawler, 1995, on softening plant water; and Darby, et aI., 1991, on
wastewater filtration) have been published on this extensive set of experiments.

Ripening and breakthrough were quite evident in the measured size distributions in vitually all experiments,
with an example from one softening plant experiment shown in Figure 5 (taken from Kau and Lawler, 1995).
From Figure Sa, it is obvious that substantial removal of particles of all sizes occurred over the 746 mm depth
and that greater removal occurred at 66 minutes than at 17 minutes, i.e., ripening occurred. Removal
generally increased with increasing size for the range measured at both times shown, although the effect of
particle size was not as dramatic as expected from existing models (Rajagopalan and Tien, 1976, 1982). In
Figure 5b, the data clearly indicate that, except for the smallest sizes measured, less removal occurred at 2682
minutes (almost two days) than at 66 minutes; this reduced removal represents breakthrough.

Special experiments were undertaken to determine whether breakthrough was caused by breakoff (detachment
of previously captured particles or flocs) or direct passage of particles from the influent. A regular
experiment was run for an extended period (approximately two days) to load the filter bed with captured
particles. Before the end of the run, effluent was captured and re-filtered through a cartridge filter, creating a
suspension chemically identical to but with far fewer particles than the standard suspension. The influent was
switched to this new reduced concentration suspension. Particle size distribution measurements of samples
taken shonly after that switch (not shown) indicated quite clearly that breakoff was occurring--many particle
sizes, particularly in the mid-range of the sizes measured had higher particle concentration than the influent.

The clear evidence of breakoff led to yet another question: were the particles that broke off ones that had come
into the filter at that size, or were they flocs that had formed on the surface of the media as individual particles
were captured? That question could not be answered with data from experiments with the natural
suspensions. However, data from monodisperse latex experiments were useful, with an example shown in
Figure 6. Results from an experiment with a monodisperse suspension of (nominal) 2 I.1ID particles are
presented as the ratio of the concentration at a depth of 10 mm to that of the influent Primary particles (i.e.,
Window I, the original latex spheres) were removed throughout the experiment. as indicated by the ordinate
values being less than one. Larger flocs. represented by Window 3 as dermed on the figure, were removed at
first but then consistently had values greater than one--evidence of the breakoff of flocs formed on the media
surface by the capture of primary particles. The scatter in the data indicates the random nature of breakoff.

Mathematical modeling of filtration was undertaken in a few ways. First, ideas from the Rajagopalan and
Tien (1976, 1983) and O'Melia and Ali (1978) models were used to plot the data in novel ways to conflrm
that the effects of independent variables such as depth and velocity were as expected from those models.
Second, a new model for clean bed (time zero) removal was developed by Cushing (1993); this model
N
o

PARTICLE DIA~ETER (~)

z
o e "
1
,~~~---r~-----'---'" ~,.-------r--
to
r-" ---, J2
"
~
8z .s.., ~
:i
(a) u
..: 10. 1
;;.. _________________n--.B f-LI=3 I yan~ ~~
u'"
.
~

~ i Z:;
'"
_ c ........ 101", 10"
Q - ........ , 2a
'"~ z~ ~~
~: 10'"
'" -....
~
g~
II A
A
-:::

~o
A/IH1tj.ld,\'
lIamak!!f constant
d~ , ~1 Yhl..:oslty

~
:=

;: i2 10"
d, ::- brgt'r partkle dwmeter
..:
"
Co.
-
<l
~
I. 1- .
::;
lop-til,,)
(; = veloClty _gradient
20
(b) 10" I I I
7. 0,0 0.2 0.4 0,6 0.8 1.0
o SIZE RATIO lid

.... -
...l
- IE
S.;- Il
Figure 2. Curvilinear Modeling of Flocculation:
~ Collision Efficiency Factor for Fluid Shear.
i2 c
... ; o
~
Q Ol
.,.,
100 r

3=z~
"~ .E
<'l ;l-
~
r
.,. m
;;0

, , ," ,-, , I ,--, , , I


~
., ...:l
" (c) ~ 10
zs
~! 1=
0:
"" :ie '0 ~
;'Q
..

...;!l -.E
" . / ....._----.\.. '"00:
..'"
af
Q -
10 ~
;.

...l <l S"


o -
;. ; ..'
or _ ac:aL21_ ~. j ~ _. __ ,_ .
-0.5 0.0 5 1.0 I.!'
LOG OF PARTICLE D1AMEn:R (d, In ~m) (1.1 V I
0.1 10'-
Figure I. Representations of Partlde Size Distributions:
Consequenl"Cs or Pow~r La,,'. UIA.\lETER 01' PARTIl1,[ i tiJllll

Figure 3. Dominant Regions for CoIlisiom by Each


Mechanism in Redilinear and CurvUinear Models.
~
z 7
t o up (Omin)
1= . __ .. ~!;~"SO min)
~ ~ 6

~.
~ '.
... lIlearModel
i;1: " ~
..........
__ CurvtboearModel(ISOmin)
(ISOmin) 6
z ~ 5 ~ . . r~~ntSM4 l
o .. , 5 ~........ , Media = Sand, 1.85 mm i
iVelocity = 1.8 mmls I
-........-."""'".......
~4 , '....
-~
-":::~....
..."" , Port G at 746 mm deplll. .
;z
1tI~ 4 :. _----- '-_.-

....."~'
...

t; ~ 3
3 '- "
Port G ~.....-.
S--.. ........ A
~
-
roI ... 1 ........ '"" Port
~~ S';spen"o;;;pVC Ii! SO m 2 r ~nfluent
z'" ~e
"'z
J 17 min ......
~~~,.........
u~
Alum Dose ~ J mgIL
roI <l pH=6.0
.....""...:-:-. ' ........ ~
..:1-
,.. .."." :..'.--.
;l
-0= I81's
U~ - -- ---- -- -- ~i:
0"
1=c.:..:1 z .... n'
0 0 ~, - I u o Port G / .
.....
..;r
."i' ..... ...v.,':
:
.,D
;:: ..
-...
:; 0.5 0.0 0.5 1.0 1.5
66 min .... "" ...~_ _ ~.

...~~...
LOG OF PARTICLE DIAMETER (d. in JIID) ;l'. 1
co:! .. (a) ~
Figure 4. Testing Calibrated Models Against Experimental
g:
:;'
.1 I 1
Results. leM = 0.9; (laM = 0.4) t; ~ 2
Q- 6 -, I ~.
5
o WiDl I.SS<d <2.l2
1>1;0.
......
N<1
",z
5 "~'....
,....... .., s
a
1>1$ 4 ~
~~~o~,. . . . .
......
Win 3 3.44 < d < 4.23 ..l~
4 uO ' ~ sa'"
~ .~..
!;l- 5'" 3 "v_
.:\.:;. ........~ Influent
Port A
~
.....
~.;
3
~ 2 Port G ~ ..........: .".,._~,~ '"
z=
~
",
z'"
... .
66 min .... ~ 'Sb"
. ........ "'..
~
role
OJ"
-
..:Ii!:
...
1
... o
Port G ,............--......
2682 min
...~ ........
' . ''':. :" ...... ~:
1!i ........ ... c .. ~
.... -- -0 - a 1 a
0 ~) ..... - .. co
0 0 0 00 0 0 0 0 00000
~~-J~~L'-L-L-L~J-~~~L-L-~~
2
41 .0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
o 60 120 180 240
TIME (mIDuIes) LOG OF PARTICLE DIAMETER (dp in 11m)
Figure 6. Attachment of Primary Particles and Detachment Figure 5. Ripening and Breakthrough During
of Floes in Filtration of Monodisperse Latex Spheres. Filtration.

~
22 o. F. LAWLER

accounts for the contact points between filter grains in determining the flow field and therefcre the particle
behavior. The model (not shown) shows far less sensitivity of removal to particle size than earlier models. a
result that is consistent with all of the experimental data found for heterodisperse suspensions in this research.

State o/the art. Knowledge of particle behavior in filtration. as in flocculation, has grown
tremendously in recent years. Phenomena that were once impossible to measure are now well documented.
Understanding of the role of chemistry in filtration, the phenomena of ripening and breakoff. the mechanics
and hydrodynamics of baclcwashing. and other areas of filtration has all grown. Fllters are often designed
and operated differently than they were just 20 years agc>-with deeper and larger media, higher filtration
velocities, and different baclcwash strategies. These advances have been brought about by the combination of
experimental (laboratory and full-scale) research, mathematical modeling. and the creativity of design
engineers. Nevertheless, what happens to particles as they pass through a filter is not completely understood.
The need for filters to capture Giardia cysts andCryptosporidium oocysts has rekindled the drive to improve
filtration removal efficiency. Some particles are not captured. Is this because they are relatively stable. or is
there some other systematic reason for certain particles not to be caught (or to detach after capture). or is it
simply a random phenomenon? Further research is necessary.

CONCLUSIONS
Size distributions change in systematic ways through each particle treatment process. The effectiveness of
each process depends in part on the influent size distribution. Otanges in particle size distributions can now
be measured analytically and modeled mathematically to an unprecedented extent These sophisticated
methodologies bring new responsibilties and challenges-to understand panicle processes from the
microcsopic to the macroscopic level Progress is being made in this quest, but many questions remain
unanswered. Fwtherresearch is necessary.

REFERENCES
Clark, S.C., Lawler. D.P and Cushing. RS (1992). "Contact Filtration: Particle Size and Ripening."
Journal. American Water Works Association. VoL 84. No. 12.61-71.
Cushing. RS . "Mathematical Modeling of Depth Filtration: Investiga~on of Initial Re~val through 'Jbree
Dimensional Trajectory Analysis," PhD. Dissertation. Univemty of Texas, Austin, TX. 1993.
Darby. I. L . and Lawler. D. F . (1990). "Ripening in Depth Filtration: Effect of Particle Size on Removal
and Head Loss," Environmental Science and Technology. Vol. 24. No.7, 1069-1079.
Darby, I.L. Lawler. D.F. and WIlshusen. T.P., (1991). "Depth Futration <?fWastewater: Particle Size and
Ripening". Research Journal a/the Water Pollution Co1l1701 Federatton. Vol 63, No.3. 228-238.
Han, M.Y. and Lawler, D.F., "The (Relative) Insignificance ofG in Flocculation," Journal o/the American
Water Works Association. 84.10.79-91. October. 1992.
Kau. S.M. and Lawler. D.P "DynaInics of Deep-Bed Futration: Velocity, Depth, and Media,"Journal 0/
Environmental Engineering. ASCE. 121. 12.850-859. 1995.
Lawler. D.F. O'Melia. C.R. and Tobiason, I.E . "In~egral W.ater Treatment Plant Design: J:Tom ?article
Size to Plant Performance." Chapter 16 in pf!'lIcuIates.1n Wate~. Kavanaugh and Leclde, editors,
Advances in Chemistry Series. #189. Amencan Cllenucal SOClety. 353-388,1980.
O'Melia, C.R. and Ali, W "The Role of Retained Particles in Deep Bed Filtration," Progress in WateT
Technology. 10.5/6. 167-182. 1978.
U. J. "Rectilinear vs. Curvilinear Models of Flocculation: Experimental Tests." PhD. Dissertation.
University of Texas, Austin, TX. 1996.
Rajagopalan, R.. and Tien. C. (1976). "Trajectory Analysis of Deep-Bed Futration with the Sphere-in-CeU
Porous Media Model." AIChE Journal, 22. 3. 523-533.
Rajagopalan. R.. Tien. C. Pfeffer. R., and Tardos. G., (1982) "Letter to the Editor." AIChE Journal. 28. 5.
871-872.
Particle size distributions in treatment processes 23

Smoluchowski, M., "Versuch Einer Mathematischen Theorie der Koagulations - Kinetik Kolloider
Losungen," Z. Physik. Chem., 92, 129, 1917.
Stumm, W., and Morgan, 1.1., Aquatic Chemistry, (Third Edition) Wiley-Interscience, New York, 1996.

You might also like