Single Collector Analysis of Collection
Mechanisms in Water Filtration
R . RAJAGOPALAN" and CHI TZEN
Department of Chemical Engineering and Materials Science, Syracuse University,
Syracuse, New Y m k 13210
A complete trajectory analysis of particle collection on single
spheres, which includes all the known collection mechanisms
and forces, is presented here for later experimental verification.
The results show that the surface interactions between the
collector and the particles have negligible influence on col-
lection rate except to determine whether the surface condi-
tions are favorable or unfavorable for collection. It is also
evident that the hydrodynamic retardation of the particle in
the neighborhood of the collector can reduce deposition sig-
nificantly (by 50% or more), but such effects may be negli-
gible in view of the usually wide scatter of filtration data.
This paper also provides examples of how the single collector
analysis can be directly applied to modelling retention in
packed beds. To make the work easily accessible to the readers,
all the basic concepts necessary for future application of the
trajectory analysis are presented clearly along with a table
containing all the relevant forces and torques acting on the
particle. An experimental analysis of the results repdrted here
is presented in a subsequent publication.
T heoretical modelling of deep bed filtration has
often been made from a phenomenological point
of view (for example, see Ives")). Such an approach
involves the use of a macroscopic rate expression for
the rate of retention of particulate matters in the
filter bed and empirical equations to relate the in-
crease in pressure gradient with the extent of par-
ticle retention. Conservation equations for the con-
centration of the particulates and flow equations
based on the Darcy's law complete the method. The
disadvantage of this approach is that it provides little
information about the exact mechanisms of particle
collection, their relative significance and possible in-
teractions. Such details are, however, desirable or
even necessary in order to permit a rational design
of the filtration process through a proper selection
of operating variables.
An alternative and more exact way of studying
deep bed filtration is to consider a filter medium a s
an assembly of collectors and to treat the problem of
particle collection on filter grains as one of trans-
porting the suspended particles toward the grains,
where they become deposited upon contact. The prob-
lem then reduces to one of setting up the equation
of motion for the suspended particles. Such an equa-
tion of motion, in principle, can be formulated to
include contributions from all possible collection
mechanisms. We refer to this method of analysis as
trajectory analysis, since the equation of motion so
formulated yields trajectories of the suspended par-
ticles. Although this technique is conceptually simple,
the manner in which certain mechanisms should be
represented in the equation of motion remains un-
- --
_.
*Presently a t : Department of Chemical and Environmental Engineer-
ing, Rensselaer Polytechnic Institute, Troy, New York
12181.
246 The Canadian Journal of Chemical Engineering, Vol. 5 5 , June, 1977
On prbente, sujette h une verification experimentale subsk-
quente, une analyse compltSte basbe sur des trajectoires de la
collection de particules SUI de simples spheres; cette analyse
embrasse toutes les forces et mkcanismes connus de collection.
Les resultats obtenus indiquent que les interactions superfi-
cielles entre le collecteur et les particules ont une influence
negligeable sur le taux de collection; cependant, elles permet-
tent de determiner si les conditions de la surface sont favora-
bles ou non B la collection. I1 est aussi hident que le retarde-
ment hydrodynamique des particules h proximite du collecteur
peuvent diminuer leur sedimentation d'une manitre impor-
tante (50% ou plus), mais des effets de ce genre peuvent &re
negligeables &ant donnk la diffusion gknkralement considerable
des donnkes sur la filtration. On fournit aussi, dans le present
travail, des exemples de la manitre avec laquelle on peut a p
pliquer l'analyse avec simple collecteur pour simuler la rkten-
tion dans des lits garnis. Pour rendre la presente etude faci-
lement accessible aux lecteurs, on prksente clairement tous les
concepts fondamentaux qui sont requis pour toute application
future de l'analyse baske sur les trajectoires; cette presentation
comprend kgalement toutes les forces et tous les couples perti-
nents qui agissent sur les particules. On prksentera, dans une
publication subskquente, une analyse expkrimentale des rksnl-
tats rapportks dans le present travail.
settled. In particular, the exact role of the inter-
molecular forces between the particle and the col-
lector, the significance of the charge concentration of
the suspending medium, and their effects on particle
deposition all require experimental verification.
To test the validity of trajectory analysis in the
study of deep bed filtration a number of investiga-
tors"2.3,4) have compared theoretically calculated filter
coefficients (obtained through trajectory calcula-
tions) with experimental filtration data. These com-
parisons, however, do not provide definitive conclu-
sions as to the accuracy of the trajectory calculations
as they inevitably require further assumptions and
approximations ; for example, the representation of
the filter bed by porous media models as outlined
above. Consequently, it is not possible to separate
the errors due to one source from those due to the
others. Since trajectory analysis using single col-
lectors of well-defined geometry can be simulated
experimentally rather easily, we present in this paper
results of such trajectory calculations. In a subse-
quent publication an experimental examination will
be presented.
Trajectory analysis based on single collectors serves
other purposes as well. F o r instance, i t provides valu-
able information on the proper selection of porous
media models to represent the filter and on the pro-
jected behavior of the collection mechanisms in a n
actual filter, In view of these, we have also presented
in this paper, whenever appropriate, comparisons with
the trends shown by filtration experiments so t h a t

CONCENTRATION Co
ATr-m
SE D IM ENTATI 0 N
Figure ISingle (spherical) collector model (not to scale).
( i ) the proper way to account for surface interac-
tions can be studied, (ii) the relative influence of
the various collection mechanisms may be analyzed,
and (iii) the desirable features of the necessary
porous media models can be identified.
The trajectory concept used here has been common
in a i r filtration studies for some time (see Lang-
m u i P ) ; its introduction into liquid filtration is
largely due to the success of Yao’s analysis‘” of deep
bed filtration using this technique, The trajectory
equation used in this paper is in fact an extension
of Yao’s work; more specifically, the analysis pre-
sented here incorporates all the important forces into
the trajectory equation including London attraction,
double layer interaction, and non-Stokesian behavior
of drag forces on the particle in the vicinity of the
collector. Furthermore, the effect of electromagnetic
retardation on the London attractive force is also
considered.
Before presenting the theoretical study and the
discussion of the results, we introduce definitions of
certain terms that will be used in the subsequent
discussions. Some of these, such as ‘limiting trajec-
tory’ (or critical trajectory), are already well-estab-
lished in the literature, but others are less common
and might require clarification.
Definitions of the terms used
In what follows, the term ‘particle’ will always re-
fer to the suspended particle to be filtered. Cor-
respondingly, the term ‘grain’ will be reserved for
the material which makes up the filter bed. When
the reference is to the geometrical model which rep-
resents the filter bed in the theoretical analysis, we
shall use collector^ instead of ‘grain’.
By a ‘unit collector’ we shall refer to a geometrical
entity that, in representing the filter bed in the
theoretical studies on filtration, accounts for both
the porous nature of the bed and the resultant change
in the (hydrodynamic) flow field around (or, through)
the collector through appropriate means. According
The Canadian Journal of Chemical Engineering, Vol. 55, June, 1977
I
10-2
9
I O N OF THE BASIC
F PARAMETERS
I
10-2
I o-~,
NRI~
b) 7] vs. NRtd
Figure 2-Effect of London interaction on the collection ef-
ficiency.
to this definition, the Brinkman model used by
Payatakes, Rajagopalan, and Tien‘s) and the sphere-
in-cell model used by Rajagopalan and TienC4)are of
the unit collector type.
In contrast, the term ‘single collector^ will refer to
a collector of well-defined geometry, the flow field
around (or, through) which does not recognize any-
thing other than the collector itself. The spherical
collector we use in this work and also used by Yao“)
is an example of the above.
The term ‘model filtration system’ will denote an
idealized filtration system where the filter bed is
described by a unit or single collector and where the
fluid to be filtered is assumed to be a homogeneous
suspension containing monodispersed particles of
smooth spherical geometry.
The ‘limiting trajectory’ or the ‘critical trajectory’
of the suspended particle is the trajectory of the
particle that just manages to be captured by the
collector (see Figure 1).Thus, the particles approach-
ing the collector on one side of the limiting trajec-
tory are captured, and those on the other side escape.
The ‘collection efficiency’ is defined a s the ratio of
the rate of particles captured to the rate of particles
aproaching the collector. In the case of single col-
lectors, the rate of particles approaching the collector
is usually defined as the rate of particles passing
through the projected area of the collector on a
plane perpendicular to the direction of flow. In the
case of unit collectors, however, this may depend on
the nature of the model used (see, for example,
Rajagopalan and Tien‘l’).
The trajectory mutlysis
When a suspension flows through a packed bed, the
suspended particles are transported toward the sur-
faces of the grains o f the bed. This motion is caused
by various mechanisms, the four most important
among them being Brownian motion, gravitational
sedimentation, influence of intermolecular surface
forces, and forces arising from charges on the sur-
face of the particles and grains. Apart from these,
the finite size of the particles causes interception
with the grain surface, thus promoting deposition,
247
and furthermore, when the particle is in the close TABLE1
vicinity of the grains, the hydrodynamic drag on
the particle increases due to the presence of the FORCES AND TORQUES
ACTINGON THE SUSPENDED PARTICLES
grains and hinders deposition. If the particles are
assumed to be collected on the grains on contact, I. Znerl~alForce and Torque:
the rate of deposition will be equal to the rate at
which the particles make contact with the grains
due to the influence of the above forces, Since
Brownian motion, which manifests itself as diffusion
in the macroscopic sense, is of significance only where - D =
for submicron particles, for large particles one can Dt
'follow', a t least in principle, the particle trajectories -t.' = -0
to compute what fraction of the particulate matter is
collected by the grains. Computation of the fraction 11. Gravitational Force and Torque:
impacted is simplified by the fact that the particle
-f G = (4/3)Uaig(pp - PI)
trajectories do not intersect each other in the ab-
sence of inertia, which has already been shown to = (4/3)?rai (P, - P , ) g [-cos e? + sin et81
be negligible in water filtration by Yao"). Thus the -
10 = -
0
concept of 'limiting trajectory' defined earlier can g = ILI
be used to determine the collection efficiency (see
111. Surface Force and Torque:
Figure 1 ) .
Even when the random force that induces Brownian a) Molecular Dispersion Force and Torque
motion is included in the analysis, the 'paths' of the (London Force and Torque)
particles can be determined analytically. However,
due to the random nature of the resulting differential
-Po = [ - 2 H a (6;u p , A,) 4 / 3 6 ? (20, + 6)'I1e,
equation, one can obtain the position of the particle where a(6 ; up, A,) is the retardation correction factor
only in the probabilistic sense. Although technically
the 'solution' would still be known as the 'trajectory' -
tL0 = 0
-
of the particle, such a solution is clearly not deter- b) Double Layer Interaction Force and Torque*
ministic, but is characterized by probability distribu-
tions. The appropriate setting for this analysis would -
fDL = {[L'apK(l; +
1,2)/21 x
then be the theory of diffusion processes, a s shown +
1(21clp/(13 l d ) ) - e-861
by Rajagopalan"). A detailed discussion of this is be- [e-Na/(l - e-zrb)])2
yond the scope of this paper, and for our purpose 0
here, it is expedient and sufficient to treat the -1DL = -
Brownian motion separately, That such an analysis IV. Drag Forces and Torques: #, @
is adequate enough for the consideration of diffusion
in filter beds has been shown by Yao"' and, more Due to the Translation of the Particles:
recently, by Prieve and Ruckenstein"". +
(f")' = - G?TPapbg:(6+)e, uefle(6+) 581
Figure 1 presents a schematic representation of - = 8 ~ ue 4h (a+) 3
OD)'
the single collector model used in this work. The
trajectory equation for the particle is simply a mathe- (BrennerW; Goren and O'Neill(Z1); o"ei11(22)]
matical expression of the dynamic equilibrium of the Due to the Rotation of the Particle:
suspended particle under the action of various forces.
The forces considered in the present paper are due
CLDY = 6?r~a,2w f i ( 6 + ) ee-
to gravity, buoyancy of the particle, fluid motion - = - 87rP4 w g i (6+)3
(tD)'

(convection), drag on the particle, and surface in- wherew = IwI

teractions (London attraction and double layer re-
pulsion or attraction). Apart from these, there are
two phenomena (not caused by any force) that con-
tribute to the modification of three of the above
forces. One of these is the socalled 'electromagnetic
retardation' of the London attractive force (see
Israela~hvili'~')caused by the finite propagation
velocity of the electromagnetic waves.
Of more significance, however, is the socalled 'hy-
drodynamic retardation' caused by the increase in
the drag forces acting on the suspended particle
when it is in the vicinity of the collector surface.
The analytical formulas necessary for correcting for
these increases in drag forces (due to both the
fluid velocity and the motion of the particle) are
already available in the literature and are referred
to in Table 1. (See also Happel and Brenner"")
The formation of the trajectory equation is based
on the assumption t h a t the particle concentra-
tion is sufficiently low so that the motion of any
given particle does not interfere with t h a t of the
others. We further note that the motion of the par-
ticles is axisymmetric. (This is a consequence of
the axisymmetry of the flow field of the liquid.) The
(Goldman, G x , and BrennerW)
Due to the Fluid Velocity in the Presence of the Sta-
tionary Particle:
+
ED)"'= h ~ a 1p- A Y ' ~ Y ( S + )2 [B~ f i (a+)
" ~
+D P f r e (6+)1e
e
(c)" = 87r~ai?[BgPe (6+) +
D Y gF+ (&+)I 9
(Goren and O'Neill(21);Goldman, Cox and
Brenner(24))
*The potentials, lcand lp,in the expression for fDL are actually
the socalled 'surface otentials' [see, for instance, Hogg, Healy
and Fuerstenau(W1. gut it is a common practice to equate them
to the corresponding zeta potentials. In computing the force due
to the double layer interaction, the unit of the zeta potentials
should be in electrostaticunits.
#See Rajagopalan(') for the tabulationof all correction factors.
@Thereferences following the entries may be consulted for fur-
ther details.
force and torque balance that results in the equation
of motion f o r t h e particle has already been presented
by Payatakes, Rajagopalan, and Tien"' in connection
with the capillaric and Brinkman models. Hence we

248 The Canadian Journal of Chemical Engineering, Vol. 5 5 , lune, 1977

present here only the final equation. However, 'for the of interest would be
sake of convenience, easy reference and clarification
of the notations used we have summarized the var- .......................... , (9)
ious forces and torques in Table 1.
The limiting trajectory thus obtained joins a stream-
The trajectory equation line (denoted by +*) far away from the collector -
Since we are interested in the axisymmetric flow usually 3 to 4 collector radii from the collector sur-
around the collector, the azimuthal component (i.e., face - on the upstream side as shown in Figure 1.
+-directional component; see Figure 1) of the force The collection efficiency is then given by
due to fluid motion does not exist. Thus, we have 11 = (Zm-$lC Co)l[.K(a, + UP)* v co1
only r and 6 components; If the particle inertia and
the random force due to molecular fluctuations are
= 2+rt/(a. + up)2 v.. . . . . . . . . . . . . . . . . . . . . . . . . . . (10)
ignored, one has, for the r-component of the force The initial value problem posed by Equations (7) and
balance : (9) was solved in this work with an Adams-Moulton
4 predictor-corrector scheme. The integration was
~ ;6 s ~ a p j:'
~ . K P U ,= 5 - 7 7 r ~ (i P , - PI) g cos 0 stopped when the trajectory coincided with a stream-
- 12Ha,3a(6; X e , ~ p ) / [ 3 6(2ap ~ + 61'11 + line far away on the upstream side of the collector.
As mentioned earlier, the efficiency calculated us-
I(%K (c:
f S2)/2) x ing Equation (10) does not include the deposition
+
I ( X J p / ( C l3) - e -la x] due to Brownian motion. To account for Brownian
x [e - ~ ~ / -( 1e - z ~ a ) ] } . . . . . . . . . . . . . . . . . . . . . . . . . . . . ' , (1) diffusion, one can simply add the efficiency due to
diffusion to Equation (10). The collection efficiency
In dimensionless form, this becomes: based on convective diffusion follows from Levich's
-
U:ff = Vf': - NQCOS t? - "LO a(&+;
NRtd)/ analysis"~"' and is given by
((6+)'-(2 + + NEI[NEZ-
e -NDL6+l X q ~ i f =
j 4.04 N6'I3... . . . . . . . . . . . . . . . . . . . . . . . . . . . . .(11)
le - N D L ~/ (+I - e - ~ N D L ~ +.). ]. . . . . . . . . . . . . ..(2)
where NPc is the Peclet number given by
Similarly, the &component gives : Npe = 1 2 . ~ / . i a ~ ~ , V / k. T
. ..... . . . . . . . . . . . . . . . . . . . . . .(12)
sI u: = B: sz + B$ y+ s3 + NOsin 0 . . . . . . . . . . . . . . . .(3) In Equation (12), R is the Boltzmann constant, and
where we have taken the 6'-component of the fluid T is the absolute temperature of the suspension.
velocity 2 to be The total collection efficiency is, therefore, given
by
vg = B l y + B z y Z. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
- (4) q T o l o l = VTroj + T D j f f . , . . . . . . . . . . . . . . . . . . . . . . . . . . (13)
u:=-
1 dr
-
dl . . . . . . . . . . . . . . . . .. . . . .
The trajectory equation (Equation (7)), when expressed
- V . (5) in terms of dimensionless variables, gives rise to certain
dimensionless groups which signify the various collection
ue* = 1 r d d0
- l ................................ mechanisms. Notable among these are No (gravity and
- . (6)
buoyancy), N R (ratio of particle size to collector size;
by dividing Equation (2) by Equation ( 3 ) , we get signifies the importance of interception), Nel, Nsz, and
the socalled trajectory equation : N D L(double layer interaction), and NLo and NW ( I m -
don attraction and the extent of electromagnetic retarda-
dr+/dB = (& + y + ) X { -A+y+'f? - N Q C O S ~ +
tion). The role of diffusion is represented by the Peclet
number, NP,,which, due to its exclusion from the analysis,
does not appear in the trajectory equation. Apart from the
N E INEz
~ - e -NDL6+] X [e -NDL6+ / above dimensionless numbers, certain other dimensionless
(1 - -2NDL6+)] groups such as Nol, N D ~N, D ~and, f? also enter the tra-
- NLO f f ( 6 ' ; NRtd)/[(6')' (2 + 6+)z11 jectory equation to account for the hydrodynamic re-
f { N DB: ~ y+ NDZ + B$y+' + N DNO
~ sin 0 ) . . .(7)* tardation. However, as we shall see shortly, the hydro-
dynamic retardation is of importance only when the
The solution of Equation (7) requires an initial distance of separation between the particle and the collect-
condition, namely, the starting position of the par-
ticle whose trajectory is desired. Since the trajectories TABLE2
of only those particles that are captured are of in- *
terest, it is convenient to start from the surface of BASICSET OR VALUES OF VARIABLES AND PARAMETERS
the collector. For example, use of a condition such as USED IN THEORETIDAL CALCULATIONS

d, = 1.2 cm Dimensionless Parameters

d , = 5.7 um NR = 0.475 X
P p = 1.05 g/cm3 N R ~= 3.3 (based on d )
in the integration scheme would result in the trajectory o P I = 0.9973 g/ml NQ = 0.4 X 10-2
a particle that eventually intercepts the collector a t the / I = 0.9142 mN.s/mZ N E ~= 0.475 X 10'
position ro+. Notice that Equation (8) has incorporated v = 1.5 cm/min NEZ = 0.63
collectiozdue to interception by accounting for the finite T = 24°C NDL = 0.1140 X 10'
size of the particle [i.e., ro # &/up = 1/NR,but rather
H = 0.3 x 10-20 J ivLO= 0.56 x 10-4
i c = -lOmV(= -10 x 10-31
(a, + a,)/a, = 1 l/NR]. +
However, since we need to 300 e.s.u.) N ~ t d= 0.1791 X lo3
compute only the limiting trajectory, the initial condition I , = - 4mV ( = -4 x: 10-31
300 e.s.u.) Np. = 0.366 X 10'
K = 400 pm-l
*Notice that u, is taken to be of the form ' -Ay".
-
The Cunudim Journal of Chemical Engineering, VoZ. 55, lune, 1977 249
 10
I
, 10
10
I:. 1 I I , , ( I I
'i'
$ 5
-10
;
I/ \-LONDON
ATTRACTION
t (UNRETARDEDI
I::
I
I
I I I I 1 1 1 , I I
10
S,nm
Figure 3-Surface interaction forces (for the sphere-plane sys-
tem). Corresponding to the parameter in Table 2.
or surface is about lOa,. Consequently, the relative size
group, NR, can effectively characterize the importance of
hydrodynamic retardation as well, and the above-men-
tioned groups such as Nol are necessary only to introduce
the appropirate correction in the trajectory equation.
The discussion in the previous paragraph in-
dicates that the relative influence of the various col-
lection mechanisms may be studied by changing the
pertinent dimensionless groups individually Therefore,
the results of the analysis are presented below as
functions of Nc, Ng, N E I ,NEZ,N D , NLO, and N R t d .
The basic set of the values of the above groups used
in the following discussion is presented in Table 2.
Results of the trajectory analysis
Role of Surface Interactions
The influence of particle-collector surface interac-
tions in liquid filtration has been the subject of
analysis by many investigators. Ives and Gregory"*)
have attempted to analyze this experimentally. An
excellent introduction to this area (as applied to
water filtration) may be found in the review article
by Wnek(ls). The surface interactions arise due to
the socalled London interaction between the molecules
of the particle and the collector and due to the in-
teraction between the electric double layers surround-
ing the particle and the collector. In most cases, as in
liquid filtration, the London interaction is attractive,
although under special circumstances it may be repul-
sive (see Israelach~ili'~)). On the other hand, the
double layer interaction may be repulsive or attractive
depending on whether the particle and the collector
have like charges or opposite charges, respectively. The
model used in this work to account for double layer in-
teraction is based on the socalled Derjaguin-Landau-
Verwey-Ovrebeek theory (DLVO theory), and the de-
tails on this may be found in Hogg, Healy, and
Fuerstenau"+'. The dimensionless group, NEe, which
varies from -1 to +1, determines whether the double
layer interaction is attractive (non-positive values of
NEz) or repulsive (positive N E B ) The
. group NDL sig-
nifies the relative thickness of the double layer (with
respect to the particle radius). Figure 2 presents
the collection efficiency as a function of NLO and
NRtd.* The electrokinetic repulsion and the London
*Figures similar to Figure 2 can be constructed for other dimensionless groups
( N E ~N, g z , and N D L ) to demonstrate that remains constant (see reference 7).
250
I I 1 I I l I O
ds=O36mm
dp = 3.5 p m
V =03*0.lcmhr UNFAVORABLE
+-+
M
FAWRABLE
0
,
I .
2x M NoCl h $0
I
5
;t 1
=-4mv
P M M A COLLECTOR
V = 0.029 crnh
p = 9 1 4 2 ~ 1 0 - ~s/m2
N I
i-zo I
t
I
I I 1 1 1 1 I
I00
I
I
1
I
I
I
I
I
I
I
lo-' lo-' 10'' I00
(NEI/ N D ~ )
Figure 4-Effect of unfavorable surface interaction on filter
coefficient (Data of FitzPatrick"").
attraction corresponding to the parameters in Table 2
are shown in Figure 3. I t is clear that the net force
is attractive for all 6 > 0, and thus there is no barrier
against the particle-collector contact. In the light of
this, the results of Figure 2 suggest that the surface
interactions, when considered as transport mechan-
isms, are unimportant as long as their net effect is
attractive in the vicinity of the collector, However,
when the dimensionless groups are varied, the repul-
sive force may exceed the London attraction and, in
fact, may become strong enough to overcome the
gravitational and hydrodynamic forces in the im-
mediate neighborhood of the collector. When such is
the case the predicted collection efficiency drops to
zero. The regions where this happens are shown in
the figure by the shaded portions. In fact, even when
NEz < 0 , i.e., double layer interaction is also attractive,
there is no enhancement of collection (see reference
7).
Payatakes, Tien, and Turian") observed that the
influence of London attraction and the electromag-
netic retardation may be significant. In this con-
nection, the information presented in Figure 2 is re-
levant. The variation of NLO can be considered as due
to the variation of the Hamaker constant, H, since
NLo a H. As is evident from the above figure, in
the case of single collectors even large changes in H
do not affect the collection as long as the London at-
traction dominates the double layer repulsion. I t is
true that the effect may be different in the case of
unit collectors. A case in point is the work of Rajago-
palan and Tien'4' which uses the sphere-in-cell porous
media model. In this case, an increase of six orders
of magnitude in NLo increases the collection effici-
encies by one order of magnitude. Since in practice
a change in JH by six orders of magnitude is un-
likely, the change in NLO may be viewed as a decrease
of up by 10' (notice that NLO a a:). However, when
the particle radius reaches such low values, diffu-
sional deposition increases by a factor of lo2 (see
Equation (11)) , thus out-weighing the increase due
The Canadimz Journal of Chemical Engineering, Vol. 5 5 , June, 1977

to the London attraction. Consequently, there is no
need to consider London force as a major factor.
To assess the effect of retardation (of the Lon-
don attraction) on the collection efficiency, consider
a case corresponding to NLO equal to, say 5 x lo4 in
Figure 2. A five-fold decrease in NLO (NLo = lo-') will
reduce the London force to (1/6) of the original
value for all 8 > 0. In spite of this, the change of q
in Figure 2 is extremely small. Since the retardation
correction, a, changes gradually from 1 to about 0.2
for a change in S from 0 to approximately 200 nm
(where the gravitational force usually starts dominat-
ing)', the effect of a five-fold reduction in NLOon
the collection efficiency, q, should be more severe
than the inclusion of the retardational correction.
Since, as mentioned above, the reduction in NLo
hardly affects q,i t follows that the effect of retarda-
tion should also be negligible. In fact, the theoretical
computations performed in this work confirm the
above conclusion. Moreover, since the London at-
traction itself is important only insofar as eliminating
any repulsive barrier created by the double layer in-
teraction, it is unlikely that the retardation of the
London attraction can be any more important.
In general, i t is logical to expect the contribution
due to the surface interactions to be proportional to
the increase in collection due to interception if the
collector boundary is enlarged to accommodate the
range of the surface forces. We have observed this
to be true over a wide range of N R . Nevertheless,
Wnek"" points out that there have been reports in
literature which seem to indicate a substantial en-
hancement in collection due to increased favorable
surface interactions. These increases, however, are
usually only about 200 to 300% (i.e,, by a factor of
2 or 3) and could have been caused by even minor
increases in flocculation of the particles due to the
addition of electrolytes (as we shall see shortly).
Moreover, many of these investigations have used par-
ticle-grain systems that cannot be characterized well
for mathematical analysis. Although it is true that
the peculiar hydrodynamical situation in packed beds
may affect the relative influence of the surface in-
teractions, it is interesting to note that the results
obtained by F i t z P a t r i ~ k " ~using
) well-controlled par-
ticle-grain surfaces do support our contention re-
garding the role of surface forces. As a n example,
Figure 4 presents some of FitzPatrick's data for
various Ntii/NoL. An almost sudden decrease in the
filter coefficient X ( X a q ) for high values of N ~ N D L
The Canadian Iournal of Chemical Engineering, Vol. 55, June, 1977
NG I
Figure 6-Variation of t with N G for N G > 10'.
is clear. For low values, is practically constant."
In spite of the above concurrence with experi-
mental observation, it is worthwhile remembering
that the surfacei charge distribution on the grain
surfaces (or collector surface) and the possibility of
having a wide range of zeta potentials for the par-
ticles in the suspension may affect the results ob-
served in filtration practice. Indeed, the very need
to examine the method of analysis implicit in the
formulation of the trajectory equation is motivated
by considerations such as these.
Role of Sedimentation
Figures 5 and 6 show the influence of sedimentation
on collection. For low values of N G (i.e., pp p t ) , the
collection efficiency is essentially constant a t a value
-
dictated by the combined effects of interception and
surface interactions. For larger values of N G ( N o 5
lo4), however, there is a first order dependence on
N G as evident from Figure 6 , thereby suggesting
that q due to sedimentation alone is approximately
equal to N G . The usual range of N o in typical fil-
tration systems lies above and consequently this
observation has important implications with respect
to filtration practice a s well. In particular, it may
indicate a strong influence of flocculation of the
suspended particles on filtration rates. We shall post-
pone this discussion to a later section and concentrate
here on its relevance to single collector studies.
Since NO a ag, the need for stable suspensions in single
collector experiments is clear. More importantly, the fact
that q a NO indicates that collection may be sensitive
to the exact magnitude of the density difference between
the particle and the fluid. This clearly follows from the
fact that NO cc ( p p - p j ) . For instance, consider a case
where pw = 1 g/ml and pp = 1.05 g/cm3. An increase in
p, of about 5% causes an increase of 100% in No and
consequently a 100% increase in q . Our experimental ob-
servations regarding this are discussed in a subsequent
paper.( 6 ,
*Thin comparison, however, is only approximate since it is not always feasible
to vary N s r , while keeping N E Zand N D L constant simultaneously, in ex-
periments. The data in Figure 4 have been correlated as a function of ( N ~ l l
N D L ) ,since ( N E ~ I N D Lis) a measure of the magnitude of the zeta potentials of
the particle and the collector.
251
 10' I 1 1 1 1 1 1 1 I I , , I ) , I
SET OF PARAMETERS.
,, . '--*'
10-3:
\ -
- chosen for studying deep bed filtration. Figure 7
Figure 7-Effect of interception and drag corrections on the
collection efficiency: TI vs NR.
Role of Interception and Drag Corrections
The behavior of r] as a function of N R is of con-
siderable importance since N R characterizes two of
the most important effects influencing the capture of
the particle by the collector. The first of these is
interception, and when the particle diameter is large
as compared to the collector diameter (i.e., large N R )
one would expect substantial collection due to inter-
ception. More importantly, the increase in drag re-
sistance in the neighborhood of the collector is a
function of the diameter of the particle and is also
represented by NR. This point may not be imme-
diately clear and requires the examination of the
theoretical results shown in Figure 7.
Figure 7 shows the collection efficiency a s a func-
tion of N R . The upper line corresponds to the case
when the drag corrections are ignored, and the one
immediately below represents the efficiency under
the influence of drag corrections. The efficiency due
to interception alone can be computed easily from
the flow field around the collector (since the ex-
clusion of other mechanisms implies that the particle
simply follows the streamline until intercepted by
the collector). Since the streamlines are closest to
the collector surface at 0 = n / 2 , it follows from Equa-
tion (10) that t h e efficiency due to interception
(when the drag corrections a r e not considered) * is
given by
T I =
=
[I2lr$
y+= (1 +&+)]
+ raZV
1.5 N i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..(14)
Equation (14) indicates the increasing influence of
interception for increasing values of N R , and the
total collection efficiency shown in Figure 7 (upper
curve) registers a corresponding increase. In fact,
it turns out that sedimentation and interception act
independently under these circumstances, and the
total efficiency in this case (as computed from tra-
jectory analysis) can be obtained as the sum of ef-
ficiencies due t o sedimentation and interception, i.e.,
V T =~ V Q~ + ~~1 = NQ + 1.5Nl ..... . . . . . . . . . . . . . . (15)
(Efficiency due to diffusion, r]Diff, is negligible in
this range of N R . )
'Since the drag resistance on the particle increasea indefinitely as
the distance of separation from the particle and the collector dimin-
ishes. inclusion of drag resistance in the computation would lead to
zero collection efficiency unless an attractive force (such as the
London force) is included in the computation.
252
On the other hand, when the drag corrections a r e
included in the trajectory analysis, a drastic decrease
in collection rate results for N R $ 6 x lo*. As evident
from the figure, the hydrodynamic retardation causes
a monotonic reduction in r] as NRincreases. We shall
see in a later section that such a behavior is a charac-
teristic of the single collector models and may not
necessarily be desirable for porous media models
also indicates that the influence of drag corrections
varies as a function of N R . This follows from the fact
that the hydrodynamic retardation introduced by the
presence of the collector is effectively negligible when
the particle is beyond a distance of about 10 ap from
the collector surface (see Reference 7, Appendix A ) .
To obtain an intuitive picture, notice that the flow
rate of the suspension through an annular region
of thickness ( y l a , ) around the collector (at 0 = n / 2 )
is given by
Consequently, the ratio of this rate to the rate of
suspension approaching this collector is given by
Since, as mentioned earlier, drag corrections extend
their influence only over y ? 10 ap (i.e., y + -? l o ) ,
Equation (17) demonstrates that the region over
which wall-effects are important affects decreasing
percentage of particles and, therefore, the drag cor-
rections should have less and less effect on collection
as NR decreases. Figure 7 attests to this observa-
tion.
Discussions of the results
Results as Applied to Filtration Data
The results presented above have been examined
mainly from the single collector model thus far. How-
ever, these results have larger significance than
suggested by the above discussion. Before examining
the theory from the point of view of experimental
filtration data, we notice that an important conse-
quence of the trajectory analysis is that the tra-
jectory collection efficiency can be obtained as the
sum of contributions due to sedimentation and in-
terception when the drag corrections are ignored
(see Equation (15))" and when the surface inter-
actions are favorable. For the range of N R encoun-
tered in deep bed filtration, the drag corrections re-
duce the trajectory efficiency t o approximately one-
third or one-fourth of the uncorrected value. How-
ever, since the variations in the experimental data
under 'identical' conditions are usually much higher,
the exclusion of the drag corrections may be un-
important. On the other hand, although the theory
predicts no collection under unfavorable surface in-
teractions, a s mentioned earlier, experimental data
suggest a gradual decline in r]. Thus, it appears that
in principle the total efficiency can be expressed as
V'Z'otczt = P1 [ T D z l / + P2 ( V O + VI)] .. . .. .. (18)
where PI corrects for the gradual decline in r] due t o
*In fact, we have observed this to he true for certain unit collector
models such 88 the Brtnkman model (Rajagopalan and Tien (un-
published) and the sphere-in-cell model (Rajagopalan and Tien(4)).
The Canadian Iournal of Chemical Engineering, Vol. 5 5 , June, 1977
unfavorable surface conditions, and pa accounts for
drag corrections. (Notice that PI I:1 and p8 5 1.)
It is indeed possible that the fraction pl can be em-
pirically related to the repulsive barrier in the vicinity
of the collector, the distribution of zeta potentials
of the particles in the suspension, and the distribu-
tion of potentials on the collector surface.* Before
this can be done, experimental studies examining the
exact role of surface interactions are necessary. One
such investigation will be reported in a subsequent
paper.
The filtration data are usually reported in terms
of the socalled 'filter coefficient', A, defined by
ac - XC.,
az
where C is the concentration of the suspension in
the filter a t depth z. A can be related to r] by'B37'
X ~ ~ =t ~ [3(1
- i - 4/2d,] X In (1 -
where E is the porosity of the filter bed. Since q T o t o i
is normally very low (- lo-') and c 0.4, Equation
(20) becomes
X ~ ~ ct q~~ ,i, l ~ l / d. ~. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Consequently, the individual filter coefficients (due
to diffusion, gravity, and interception) can be de-
fined similarly. Thus it follows that
XO d.-' v-' d: p-'. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
XI a d8-3 V O d," p o .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The review of Ives and Sholji"" demonstrates that
the various experimental investigations on deep bed
filtration result in the following dependence of h
on V , p, and d.:
X a va d,fl p7
a=--1
- 12 p2 - 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
,y= - 2
Thus it is clear from Equations (221, (23), and (24)
that depending on the relative importance of the
collection mechanisms the experimentally observed
results represented by Equation (25) can be modeled
well by trajectory analysis.
That the trajectory analysis presented in this paper
agrees qualitatively with the experimental observa-
tions is supported further by the results of Ison and
Ives'18'. Ison and Ives conducted filtration experi-
ments using reasonably well-defined systems con-
sisting of suspensions of kaolinite in tap water and
beds of randomly packed, nearly mono-sized ballotini
(glass beads). They conclude from their results that
the dimensionless filter coefficient defined by
X* = Xd, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
is almost linearly dependent on the gravitational
group (corresponding to our dimensionless group,
,"c). The role of sedimentation had been held under
doubt by previous investigators, and Yao") and Ison
and Ives"*' thus provided the experimental evidence
to show the significance of sedimentation, Recall that
*In fact, p, is the socalled collision efficiency defined aa number of
particles captured by the collector divided by the number colliding
with the collector surface.
The Canadian Journal of Chemical Engineering, VoZ. 55, June, 1977
Figures 5 and 6 show a linear dependence of 7 on No
and since A* is equaI to r] by Equation (21) i t fol-
lows that A* a Nc. Thus, the predictions of the
single collector model support the experimental ob-
servations.
Another important observation of Ison and Ives is
that the dimensionless (experimental) filter coef-
ficient, A*, decreased with increasing (d,ld.) (equal
to the relative size group, N R , of this work). This led
them to speculate that the reduction in h* could be
due to the wall-effects of the grains on the motion
of the suspended particles. Notice that the discussion
in the previous section following Figure 7 provides
a theoretical basis for this observation by Ison and
Ives. As we show in the above Section, the increase
in N R increases the wall-effects, and consequently
the collection efficiency and A* decrease (see the
discussion following Equation (9) ) .
When the filtration data are obtained through well-
q n t a i ) .. . . . . . . .(20) controlled experiments the single collector analysis
- does more than provide mere qualitative agreement.
For instance, FitzPatrick"" conducted filtration ex-
periments using glass beads and suspensions contain-
ing polystyrene latex particles. Figures 8 and 9
present some of his data corresponding to negligible
(21) double layer repulsions. Also shown on the figures are
the theoretical results based on Equations (131, (151,
and (21) of this paper. I t is evident that the tra-
jectory analysis presented in this work does ex-
ceptionally well in modeling the filtration experi-
(22) ments of FitzPatrick. Certain observations based on
(23) these figures seem worthwhile. For smaller collector
diameters the theoretical predictions are lower than
(24) the experimental values. This could be the result of
two important effects. Firstly, Rajagopalan"' pointed
out that the conditions used in the above experi-
ments could introduce flocculation of the suspended
particles, thus resulting in a suspension with ag-
gregates whose effective diameters would be larger
than the ones used in the figures. Consequently, re-
tention by sedimentation and interception would be
higher than predicted. This would be especially true
in the case of interception since smaller grain diame-
(25)
ters give rise to smaller pore 'diameters' in the filter.
Secondly, the role of interception is usually much
higher in packed beds than i t is for the case of single
collectors. This is in fact one of the drawbacks of
the single collector models. On the other hand, with
the grains of large diameters the effect of intercep-
tion is very low, and in this range the single col-
lector model seems to over-estimate the collection rate
(see Figure 9 ) . However, normally deep bed filters
use grains of diameter around 0.1 cm, and for d.
around this value the predictions of Figures 8 and 9
are indeed excellent.
Possible Choices of P o r k Media Models
We mentioned earlier that the results of trajectory
analysis using single collectors could be utilized in
the selection of appropriate porous media models for
(26) studying deep bed filtration. One such benefit of the
present analysis was alluded to in the previous para-
graph when we discussed the role of interception in
packed beds.
Although the single collector model predicts the
collection efficiency in filters within experimental
variations, i t suffers from two major drawbacks.
Firstly, it is ineffective in modeling the pressure drop
increase due to collection, and secondly, i t underes-
timates the influence of interception in packed beds.
Both these drawbacks result from the fact that the
253
Figure 8-Comparison with FitzPatrick's experimental data: Figure 9-Comparison with FitzPatrick's experimental data:
X vs d,. vs ds.

single collector model excludes the influence of the bed filters. The goals have been to present the re-
presence of other grains in the filter. The discussion sults for subsequent verification of the trajectory
of the results shown in Figure 7 indicated that the analysis'"' and to explore the relevance of single
single collector model predicts a monotonic reduction collector models in filtration. The paper presents the
in q as N R increases; this reduction being caused by drawbacks of the model and suggests possible rem-
increasing dominance of hydrodynamic retardation. edies.
On the other hand, in packed beds such a monotonic
reduction is unlikely since increase in N R may cause Acknowledgment
a large 'increase' in interception due to the fact that
This work was performed under Grant No. GK-33876 and ENG 76-
the particles would be unable to pass through the 08576, National Science Foundation.
pores for large d,. Needless t o say, whether the in-
crease in interception can off-set the decrease in Nomenclature
due to hydrodynamic retardation or not would depen
on the other variables; however, i t is obvious that
7 UP = particle radius
= collector radius (see Figure 1)
these two effects work against each other. In fact,
Weber"" (see Chapter 4) shows that the efficiency
2 = coefficient in the expression for v,
A+ = dimensionless A defined by A a:/ V
of a filter may in fact increase during the initial B I ,B2 = coefficients in Equation (4)
period of filtration. This is possibly due to a decrease B:, B$ = dimensionless Bi defined by BT = Blap/V and
in pore diameters as collection proceeds and can be B t = Baa,2/V
C = concentration of the Darticles in the sumension.
represented by a corresponding increase in Ng. A number/unit volume *(in general a function of
single collector model would indicate only a decrease spatial variables)
in q under these conditions. Thus i t is clear that the C O = original concentration of the suspension
d, = particle diameter
model used should be capable of being flexible in d, = collector diameter
this regard. These considerations have led the present DBM = Brownian diffusion coefficient
authors to examine a cell which has been -e,, 5 0 , c& = unit vectors in r, and @Jdirections
very successful in extending the single collector ap- f!,fe, fi",fie,fps andfi = drag correction factors in Table 1;
proach. functions of 6
Summary
L = force vector acting on the particle: superscript
specfies the source of the force; for example fDL is
We have presented in this paper a complete model the force due to double layer interaction (seeTable
1)
for the prediction of collection efficiencies in deep !i = magnitude of the vector, g

254 The C2anadian Iournal of Chemical Engineering, Vol. 5 5 , June, 1977

gi, 8,gr6 andC+ = torque correcth factors in Table 1; VTmj = efficiency computed from trajectory analysis
functions of 6 e = angular coordinate (see Figure 1)
= acceleration due to gravity eo = 6 corresponding to the start of the integration
= Hamaker constant (- 1 X 10+’0 J) K = Debye-Huckel reciprocal length
= Blotzmann constant (1.38048 X J/K) x = filter coefficient
= mass of the suspended particle x e = wavelength of electron oscillation; A, = 100 nm
= f! SdS1 cc = viscosity of the liquid
= f:sa/si T = 3.14159. . radians
= fVSl PI = density of the liquid
= double layer group; DDL= ~a~ PP = density of the particle
= electrokinetic group #l; u = dielectric constant of the medium ( w 81)

=
NE= +
~ V K ( C l’3/127tccV
electrokinetic group #2;
lk = streamfunction for creeping flow around the
collector
NEa = ~ C C { P / f( { l’’)
~ PI1 = streamfunction that coincides with the limiting
= ravity group; NG = 2a2 ( p , - pf)g/&V trajectory on the upstream side (see Figure 1)
= fondon group; NLO = fi/9rpa; V
’
w = magnitude of the angular velocity of the particle
= Peclet number: NpS = Vd#/DBM w_ = angular velocity of the particle
= relative size group; N R = a /a,
= retardation group; = &rap/As
= constants in Equation (18) References
= radial coordinate (see Figure 1) (1) Ives, K. J CRC Review in Environ. Control 2. 293 (1971).
= r/ap ( 2 ) Pavatakea.”A. C.. Tien. C. a n d Turian. R. M.. AIChE J. 20. 900
= dimensionless radial coordinate; r* = r/a, ’ (1974). .
(3) Payatakes. A. C., Rajagopalan, R. and Tien. C.. Can. J. Chem.
= r+ corresponding to the start of integration of the Eng. 62. 722 (1974).
trajectory equation ( 4 ) Rajagopalan, R. and Tien, C.. AIChE J. 22. 623 (1976).
= positionvector; r = (r, 6 ) (6) L a n m u i r , I., “Filtration of Aerosols and the Development of
Filter Materiala”, Office of Scient. Res. Develop., Rept. 866
= dimensionlessposition vector; r* = (r/a8, 8 ) (1942).
(6) Yao K-M., “Influence of Suspended Particle Size on the
= (ro/ap, 6J TraAsport Aspect of Water Filtration”, Ph.D. Diss.. Univ. North
= (figlb -fm4 Carolina, Chapel Hill, N.C. (1968).
= (fig?+ + y+flegt)/+ (7) Rajagopalan. R., “Stochastic Modeling and Experimental Anal-
ysis of Particle Transport in Water Filtration”. Ph.D. Diss.,
=
=
cf‘egP++ Y+fPeg+)/g+
time variable
Syracuse Univ., Syracuse. N.Y. (1974).
(8) Prieve D. C. and Ruckenstein E., AIChE J. 20, 1178 (1974).
(9) Israela’chivili, J. N., Contemp.’ Phys. 16, 169 (1974).
= torque vector acting on the particle; superscript (10) Happel, J and Brenner H.,“Low Reynolds Number Hydrody-
namics”, Prentice-Hall knglewood Cliffs N.J. (1966).
specifiesthe sourceof the torque (11) Levich, V. G., “Physicb-Chemical Hydrodhamics”. Prentice-Hall,
= temperature of the suspension in K Englewood Cliffs, N.J. (1962).
= particle velocity vector; subscript denotes the (12) Ives. K. J. and Gregory, J., Proc. SOC. Water Treat. Exam. 16.
93 (1966).
particular component (13) Wnek. W. J., Filtr. Sep. 11, 237 (1974).
= liquid velocity field; subscripts denote the com- (14) Hogs, R., Healy, T. W. and Fuerstenau. T. W., Trans. Faradav
ponents SOC.62, 1638 (1966).
= approach velocity of the liquid (16) FitzPatrick, J. A., “Mechanisms of Particle Capture in Water
Filtration”, Ph.D. Diss., Harvard Univ., Cambridge, Mass. (1972).
= ( r - as),see Figure 1 (16) Rajagopalan, R. and Tien, C., Can. J. Chem. Eng. 66, 246. (1977.
= y/p, (17) Ives, K. J. and Shobi, I., J. Sanit. Eng. Div., Proc. ASCE 94
= axial variable in a filter [(seeEquation (19)l (SA4) 1 (1966).
(18) Ison, b. R. and Ives, K. J., Chem. Eng. Sci. 24, 717 (1969).
= retardation correction; a function of 6, a, and A. (19) Weber, W. J., Jr. (Ed ) “Physicochemical Processes for Water
(see Table 1) Quality Control” Wile; interscience, New York. N.Y. (1972).
= surface-to-surface se aration between the collector (20) Brenner, H., Chim. Eng. Sci. 16. 242 (1961).
(21) Goren, S. L. and O”eil1, M. E., Chem. Eng. Sci. 26, 326 (1971).
and the particle (seehgure 1) (22) O’Neill, M. E., Mathemtatika 11 67 (1964).
= 6/a, (23) Goldman, A. J., Cox, R. G., an’d Brenner, H.,Chem. Eng. sci.
= initial porosity of the filter bed 22. 637 (1967a).
(24) --, Chem. Eng. Sci. 22, 663 (1967b).
= zeta potential of the collector
= zeta potential of the particle
= initial collection efficiency Manuscript received June 28, 1976 ; accepted f o r publication Jan-
= efficiency due to diffusion alone uary 31. 1977.
collection efficiencies due to sedimentation and
=
interception, respectively * * *

The Camdim Journal of Chemical Engineering, Vol. 55, June, 1977 255