Localized Adsorption of Particles on Spherical and Cylindrical Interfaces

Z. ADAMCZYK 1 AND P. BELOUSCHEK

Institute of Physical Chemistry, University of Essen, Universitiitsstrasse 5-7,
4300 Essen, Federal Republic of Germany
Received December 21, 1989; accepted March 7, 1990

Localized, sequential adsorption of spherical particles interacting via the screened Coulomb potential
was analyzed theoretically by m e a n s of the Monte Carlo ( M C ) numerical technique. Simulations were
performed for planar, spherical, and cylindrical interfaces (collectors) for various dimensionless particle
radii (aspect ratio) A = aiR and the ra parameter characterizing the range of repulsive double-layer
forces. All results obtained for hard spheres (Ka = oo ) and various interface shapes can be reduced to
the universal relationship 7ra2Nv/S * = 0~ = 0.546 (where)Vp is the average n u m b e r of particles adsorbed
at a collector a n d S* = So ( 1 + A) m is the effective surface area available for particles ( m = 0, 1, and 2
for the planar, cylindrical, and spherical collector, respectively, and the m i n u s sign denotes adsorption
on the internal collector surface)). The m a x i m u m surface concentration of soft spheres found from our
MC simulations was always smaller than the above value of 0 ~ and can be well approximated by the
relationship 0rex = 0°~( 1 + A)m/( 1 + H* )2 (where H * is the effective range of the lateral double-layer
interactions). The pair correlation function for hard and soft spheres was also determined and a definite
tendency toward a short-range ordering was revealed for particle surface concentrations close to the
above 0rex values (analogous to a 2D hard-disk fluid). Our theoretical results were compared with the
existing experimental measurements of others and a satisfactory qualitative agreement was found. © 1991
AcademicPress,Inc.

INTRODUCTION The significance of particle deposition was

Interactions of suspended colloid-size par- recognized long ago and many works have ap-
ticles with larger ones or with boundary sur- peared concerning mostly kinetic aspects of
faces (collectors) leading to adsorption and this process (see, for example, review papers
( 1, 2)). Usually, the initial particle flux under
deposition are of vast interest for many prac-
tical processes such as filtration, water and linear deposition conditions (low surface con-
wastewater cleaning, papermaking, and cartier centration) was calculated or measured as a
function of particle size, flow intensity, ionic
flotation. A quantitative analysis of deposition
and adsorption phenomena (which can be strength of suspension, etc., for simple collec-
treated as a limiting form of heterocoagula- tor geometries. On the other hand, important
tion) is also important for colloid science itself nonlinear effects appearing at large surface
concentrations of adsorbed particles were little
and for polymer science, biology, and bio-
physics. By using monodisperse colloid sys- studied, either experimentally or theoretically,
tems one can well mimic processes occurring mostly due to difficulties in considering prop-
on a molecular scale, retaining additionally the erly the many-body problem arising. The few
possibility of direct experimental verification exceptions include the works of Feder (3) and
of theoretical predictions. Feder and Giaever (4) concerning protein ad-
sorption on planar, homogenous surfaces. An
elegant theoretical analysis (based on Monte
1 To w h o m correspondence should be addressed. O n
leave from Institute of Catalysis and Surface Chemistry,
Carlo simulations) was performed by these
Polish Academy of Sciences, 30-239 Krakow, ul. Nieza- authors which showed that for hard (nonin-
pominajek 1, Poland. teracting) spheres the maximum surface coy-

123
0021-9797/91 $3.00
Copyright© 1991by AcademicPress,Inc.
lournalof Colloidand lnteoCaceScience,Vol. i46, No. 1, October 1, 1991 All fightsof reproductionin any formreserved.
 124
erage was 54.6%. However, an experimental
confirmation of this value encountered large
difficulties (the maximum surface coverage
observed was considerably smaller) apparently
due to neglecting the lateral double-layer in-
teractions of adsorbed protein molecules.
In a series of papers (5-7) Vincent et al.
studied adsorption equilibria of small posi-
tively charged polystyrene particles onto large
negatively charged polystyrene spheres, show-
ing that the maximum surface concentration
(erroneously identified with the "equilibrium"
value) was dependent on the ionic strength of
suspensions and the presence of adsorbed
polymer layers on small particles. No theo-
retical analysis aimed at a quantitative expla-
nation of these findings was given. Recently
Adamczyk et al. (8) analyzed theoretically and
measured experimentally the influence of lat-
eral double-layer interactions (whose extent
was regulated by a change of the ionic strength
of the suspension) on the maximum surface
concentration and particle distribution over
planar collectors. Theoretically predicted Omx
values for suspensions of soft spheres (inter-
acting via the screened Coulomb potential)
were found to be much smaller for lower ionic
strength than the limiting value predicted for
hard spheres, i.e., 54.6%. This finding was
confirmed experimentally using a mica col-
lector and a monodisperse polystyrene latex
suspension of the size 0.9 #m. A distinctive
tendency toward a short-range liquid-like or-
dering was observed in the irreversibly ad-
sorbed particle layer for surface concentrations
close to 0mxvalues, analogous to ordering phe-
nomena observed previously for concentrate
latex suspensions (9-12).
The aim of this paper is to generalize theo-
retical results previously obtained for the
planar collector to other collector geometries
such as spherical and cylindrical geometries
which are particularly important from a prac-
tical viewpoint. Our main goal shall be, there-
fore, solving the important parking problem,
i.e., finding the maximum surface concentra-
tion of irreversibly adsorbed particles on these
collectors, both for hard and for soft spheres.
Iournal of Colloid and Interface Science, Vol. 146, No. 1, October 1, 1991
ADAMCZYK AND BELOUSCHEK
MONTE CARLO SIMULATION ALGORITHM
We simulated the particle deposition kinet-
ics (consecutive, irreversible, and localized
adsorption) according to a numerical algo-
rithm similar to that used before (8), whose
main features are described below.
(i) The adsorbing interface was assumed to
be either a spherical one of radius R equal to
unity (see Fig. 1 ) or a cylindrical one also of
unit radius, with length L varying between 10
and 10 3 depending on particle size (see Fig.
2). For the spherical interface, due to its sym-
metry, no boundary conditions are needed,
whereas for the cylindrical interface periodic
boundary conditions were applied at its edges.
(ii) In the calculation loop the coordinates
of a virtual particle of radius a (the dimen-
sionless particle radius a / R , often referred to
as the aspect ratio, is denoted by A) adsorbing
on these collectors were selected at random by
using the pseudorandom number generator
subroutine described below.
(iii) Then, the minimum distance between
the virtual particle and its closest neighbor H m
was found (cf. Figs. 1 and 2). If rim was found
to be smaller than 2 (overlapping of particles),
then the entire loop was repeated; if not, then
the probability density of placing the virtual
particle at the point Xv, Yv, Zv was calculated
from the Boltzmann distribution law, i.e., ac-
cording to the relationship
Pv = exp(--~(Hm)), [1]
where ~ = 4J/kT is the dimensionless energy
of lateral, double-layer interactions between
particles. As discussed previously (8) this pair
interaction energy can be approximated with
good accuracy by the formula
2
= ~0 ~ exp(-raHm), [2]
where ~0 = 8"y(kTa/e2)th2(~p/4) is the di-
mensionless energy at zero separation, 3, is the
dimensionless correction factor accounting for
the influence of the interface on particle lateral
interactions (being usually of the order of
 PARTICLE ADSORPTION 125

FIG. l. Schematic representation of localized adsorption of spherical particles onto the spherical interface
(collector).

zI

O
O
()
©

0
j~

FIG. 2. Localized adsorption of spherical particles onto the cylindrical interface (collector), a schematic
view.
Journal of Colloid and Interface ,Science, Vol. 146, No. 1, October 1, 1991
126 ADAMCZYK AND
i ( )8 ) ,
k is the Boltzmann constant, T is the
absolute temperature, e is the elementary
charge, ~p = ~ppe/kT is the dimensionless sur-
face potential of particles, H = (r~2 - 2a)/a
is the dimensionless distance between parti-
cles, and ra = (8~re2Ia2/ekT) ~/2 is the di-
mensionless double-layer parameter, where 1
is the ionic strength of the suspension and e is
the dielectric constant of the dispersing me-
dium. The Ka parameter characterizes the
"hardness" of the adsorbing particle; for Ka
= ~ the limiting hard sphere behavior is re-
flected, whereas for smaller ra values the ad-
sorbing particles become increasingly "soft."
It should be mentioned that Eq. [ 2 ], describing
lateral double-layer interactions, is based on
the superposition principle and is, therefore,
valid for Ka >> 1.
(iv) Next, a uniform random deviate was
generated within the interval 0-1. When this
number proved to be less than Pv the virtual
particle was taken as adsorbed at the point Xv,
Yv, zv and its coordinates were stored. If not,
the entire loop was repeated until the pre-
scribed surface concentration of adsorbed
particles 0 was attained. The value of 0 was
calculated according to the definition
0 - Np&
Sc ' [3]
where Np is the total number of adsorbed par-
ticles, Sp = 7ra 2 is the particle cross-section
area, and Sc is the collector geometrical surface
area. Equation [ 3 ] gives the following explicit
expressions for the spherical and cylindrical
collectors, respectively:
a 2
0 = Np-y
a 2
0 = No 2 L " [4]
(v) After recording the coordinates of an
appropriate number o f particles the pair cor-
relation function g12 (r~2) was calculated for a
discrete set of intervals by a special subroutine
using the definition
Journal of Colloid and Interface Science, Vol. I46, No. 1, October 1, 1991
BELOUSCHEK
0(rl2k) -- 4 7 r / V k
g12(r12k) - 0 NpSk' [5]
where 0(rl2g ) is the surface concentration of
particles at the distance r12k from a central one
whose center is located at r12 = 0 (averaged
over all particles) and Nk is the averaged num-
ber of particles found on the segment of the
collector contained between rl2k and r~2k+,• For
the spherical collector, due to its symmetry,
the pair correlation function g12 is isotropic
and well defined. On the other hand, for the
cylindrical interface the definition of g~2 is
ambiguous because the distance measured
along the cylinder axis differs from the distance
measured in other directions.
It should be mentioned that for the spherical
collector, in the case of A larger than 20%, the
maximum number of particles adsorbed on
an individual collector becomes relatively
small; consequently, large fluctuations in par-
ticle surface concentration appear for various
realizations (collectors). Therefore, in order
to gain proper statistics one must simulate
many individual collector particles with ad-
sorbed smaller particles. On the other hand,
for the cylindrical collector these fluctuations
can be minimized by increasing the cylinder
length proportionally to the aspect ratio pa-
rameter. Usually, for both collectors, on av-
erage 10,000 to 20,000 particles were counted,
giving a relative accuracy better than 1%. In
our numerical simulations we used the pseu-
dorandom number generator subroutine based
on the function RAN1 described in (13) which
uses three different linear, congruent random
number generators. As proved before (8) no
significant sequential correlations were found
for random number strings up to 108.
For the aspect ratio below 10%, especially
for the cylindrical collector, the above "prim-
itive" algorithm becomes less effective because
the nonoverlapping test is carried out for the
entire collector surface, although it is obvious
that it is sufficient to look for other particles
within a small surface area around the virtual
particle. In this case we used a modified al-
 PARTICLE A D S O R P T I O N 127

gorithm in which the entire collector surface RESULTS A N D DISCUSSION

was subdivided into a mesh of areas smaller
1. Spherical Collector
than particle size. A similar procedure was ap-
plied previously by Hinrichsen et al. (14). Selected results illustrating the kinetics of
Then, the nonoverlapping test was carried out irreversible, localized adsorption of hard and
by considering "boxes" adjacent to the virtual soft spheres on the spherical collector are
particle only, rather than the entire collector. shown in Figs. 3 and 4. In the first figures (3a
In this way a considerable increase in com- and 3b) the dependence of the surface con-
puting efficiency was attained because the centration of adsorbe' particles 0 is plotted
simulation time was proportional to the num- as a function of the 1 / V~-parameter for many
ber of particles adsorbed, whereas in the prim- independent MC simulations (for hard spheres
itive algorithm this time increased as the characterized by Ka = ~ andA = 10 and 20%,
quadrat of the particle number. respectively). The dimensionless quasi-time
The main advantage of our algorithm is its variable r is defined as the ratio of the cu-
simplicity and consequently high speed of cal- mulative number of attempts at placing par-
culations. Hence, large particle populations ticles of some characteristic number Nob, e.g.,
can be simulated, enabling a proper statistical the number of particles needed to complete a
analysis. It should be remembered, however, prescribed surface coverage (54.6% in our
that only pair interactions between particles case). The coordinate system 0 vs 1/V~ was
are considered and adsorption is assumed to first applied by Feder (3) in the case of con-
be fully uncorrelated (consecutive). As a con- secutive adsorptions of hard spheres onto a
sequence, our algorithm is particularly well planar surface. Feder has shown that the de-
suited for dilute suspensions and thin double pendence of 0 on 1/f~r becomes linear for r
layers. >> 1. Pomeau ( 15 ), Swendsen (16), and later

b lO 2 lO 12"
i i i
8
[%]
80

a 10 2 10
~ .... ~.~:.~
1 "'~-:.,'.
• . : "-y.,.
o~
60 60 .... :""" '
.. "? j " ....
• ~ ..? .

..... !~.;;..
• 1~-.....

40 -~'~'... 40 ..'., ,: ..
"7.~.....;.:i~. :. ''L:::. .7..,
.... !:.... . .:.: ~ '~ . -. ..
["'!L.: . ....'" ...
.- :.:.~:.... . ...

i i i i ' t ' Q ' ' ' 20 , l , ] , l l , , ,

20
Q2 0.4 os s 1,/t/~ o2 0.4 o.6 o.8 1/v"~'

FIG. 3. (a) Kinetics of adsorption of hard spheres ( r a = ~ ) onto the spherical interface. The dimensionless
surface concentration of adsorbed particles 0 = NpSp/S¢ = Npa2/4 vs 1/~rr (where Np is the n u m b e r of
particles adsorbed, r = N~,/Nch is the dimensionless quasi-time variable, Na~ is the cumulative n u m b e r of
attempts at placing particles on the collector surface, and Nc~ is the characteristic n u m b e r (in our case equal
to the n u m b e r of particles forming a 54.6% surface coverage)). Dots represent separate MC runs performed
forA = 10%. (b) The same as for (a) b u t A = 20%.

Journal of Colloid and Interface Science, Vol. 146, No. 1, O c t o b e r 1, 1 9 9 1

 128
102 10
i J
[*/*l
s0
Z.:,L .
•.: .-.~....
• " : .::?a:
1( 0.2 0.4
FIG, 4. The kinetics of localized adsorption of "soft"
spheres interacting via the screened Coulomb potential
(spherical collector),A = 10%,Ka = 10, ~o = 200.
on Hinrichsen et al. (14) presented an elegant
analytical justification of this relationship by
analyzing the targeted area distribution at high
surface concentrations of adsorbed particles.
Such a presentation of MC results proved
to be also advantageous for soft sphere ad-
sorption (8) because for r < 1 the dependence
of 0 on 1/f~z becomes linear (for Ka much
larger than unity). By extrapolating to
1/f~r --~ 0 ( r --~ ~ ) one can, therefore, de-
termine the maximum surface concentration
0rex without using prohibitively long simula-
tion times. As can be seen in Fig. 3a, for A
= 10%, the adsorption kinetics is characterized
by a "noisy" pattern; i.e., considerable fluc-
tuations appear for various simulation runs.
These fluctuations as seen in Fig. 3b increase
with the aspect ratio A when the number of
simulated particles adsorbed on the single col-
lector becomes smaller according to the gen-
eral fluctuation law.
Analogous, linear relationships between
0 and 1/ ~ are observed for localized adsorp-
tion of soft spheres interacting via the screened
Coulomb potential (cf. Fig. 4, where the
Monte Carlo kinetic data are presented for A
= 10%, Ka = 10, and ~0 = 200). The only
difference in comparison with the results for
hard spheres is observed for very small values
of the I / ~ parameter, approximately 0.05 (~
Journal of Colloid and Interface Science,
ADAMCZYK AND BELOUSCHEK
1 '/" >/400), where the "softness" of the interacting
J
potential manifests itself as a small positive
deviation from the linear dependence. Results
shown in Figs. 3 and 4 have a significant im-
plication for the kinetics of localized adsorp-
tion over homogenous surfaces when no active
centers can be distinguished; thus all inter-
particle distances are plausible (as is the case
.....S:~:.:~....
•
for adsorption of massive colloid particles).
....... ::..:
One can easily show that the relationship
0.6 0.8 1 (0+x)
1- ~~r 1
observed for r > 1 implies
dO( O) ~
d--~ 1-~ . [6]
Thus, Eq. [ 6 ] indicates that the blocking of a
collector surface due to adsorbed particles is
much more pronounced in the continuous
adsorption regime than in the classical Lang-
muir model, where localized adsorption occurs
on specifically distributed active centers only
and the blocking effect is proportional to (1
- 0/0mx). These interesting aspects of colloid
particle adsorption shall be discussed in more
detail in future papers.
By extrapolating the kinetic data discussed
above to 1 / f ~ ~ 0 one can determine the
maximum surface concentration of adsorbed
particles which is a quantity of a primary in-
terest from a practical point of view. Extensive
numerical simulations performed for hard
spheres (Ka = ~ ) have shown that the ana-
lytical relationship
0rex = 0°~(l + A) 2 [7]
describes very well our MC simulations for a
broad range of the dimensionless particle ra-
dius A varied from 0 to 70%. Equation [ 7 ] can
be reduced to a single point only by introduc-
ing in place of the geometrical surface area of
the collector equal to 4 r R 2 the "effective"
surface, which is available for particles S*
given by
S* = 4~-R2(1 + A ) 2. [8]
Vol. 146, No. 1, O c t o b e r 1, 1 9 9 1
 PARTICLE ADSORPTION 129

0rex values o b t a i n e d according to the above

procedure for soft spheres a n d various aspect
ratios are s h o w n i n Fig. 5 for a b r o a d range o f
the Ka parameter, i.e., 1-103 a n d ~o = 200
(which corresponds to a typical value o f the
zeta potential of adsorbing particles equal to
r701 m V a n d 3' = 0.25 ). R e m e m b e r i n g that
(Ka)-I has a n a t u r a l physical i n t e r p r e t a t i o n as

i
the d i m e n s i o n l e s s double-layer thickness one
can deduce from Fig. 5 that the m a x i m u m
surface c o n c e n t r a t i o n of adsorbed particles at-
. . . . . . . . . .
t a i n a b l e w i t h i n a t i m e scale of typical experi-
m e n t a l c o n d i t i o n s is decisively d e t e r m i n e d by
the lateral double-layer interactions a m o n g
particles. F o r example, for the double-layer
thickness equal to 10% o f particle radius only,
the value o f 0rex f o u n d from o u r M C simula-
q2/A
tions is a b o u t three times smaller t h a n the FIG. 6. The pair correlation function g]2 calculated for
hard sphere adsorption on the spherical collector from
closest hexagonal packing of hard spheres
MC simulations, Ka = oo, A = 10%. ( 1) 0 = 65.5%; (2) 0
= 54.6%; (3) 0 = 40%.

. . . . , , , , i L_2_'-_ equal to 90%. Results s h o w n i n Fig. 5 have

LOor great significance for practical processes, e.g.,
filtration, i n d i c a t i n g that the efficiency o f par-
80 1 .....

2 .....
g12t '

6O

4O

0 i I i i I i i i i I i i i i
10 "102 ~0

FIG. 5. The maximum surface concentration of '"soft" 0 2 4 6 q2/A

spheres 0rexvs the Ka parameter (spherical collector). Re-
sults obtained from our MC simulations are marked by FIG. 7. The pair correlation function g12 calculated for
dots, and solid lines represent results calculated from the hard spheres of various aspect ratios at the maximum
equation 0rex = 0~(1 + A)2/(1 +H*) 2, where H* was jamming limit, i.e., for 0 = 0rex(spherical collector). ( 1)
calculated from the nonlinear equation 5oh = 5o(2/(2 A = 0 (planar interface); (2) A = 10%;(3) A = 20%. The
+ H* ))exp( - KaH* ) and 5oh = 10, 5o = 200. ( 1) A = 30%; broken line denotes results of MC simulations of a 2D
(2) A = 20%; (3) A = 10%;(4) A = 0 (planar interface). hard-disk fluid (17).
Journal of Colloid and Interface Science, Vol. 146, No. 1, O c t o b e r 1, 1 9 9 1
130
gl 2
2
1 ........ 5--
2 4
FIG. 8. The g12functionfor softspheresadsorbedonto
the sphericalcollector,A = 10%,Ka= 10, and ~0 = 200.
(1)0 =41.5%;(2)0 = 37.5%;(3)0 = 30%;(4)0 = 0 (the
Boltzmanndistribution,i.e., g12 = exp(-~(r12))).
ticle retention from a stable suspension should
be very low for low ionic strength. These re-
suits also strongly suggest that obtaining the
maximum hard-sphere value of 0mx would be
very difficult by experimental measurements
involving colloid-size particles.
It is also interesting to note that by intro-
ducing the appropriate cross section of the soft
spheres based on the effective range of the lat-
eral interactions H* (8) given by the equation
Seff = ira2(1 + H * ) 2,
all results shown in Fig. 5 can be well ac-
counted for by the simple relationship
0inx = 0°°(1 + A ) 2 / ( 1 + H * ) 2.
AS discussed previously (8), H* can be de-
termined by solving the nonlinear equation
2 Thus, for higher Surface concentrations, short
~ch--~0 2 + H . e x p ( - K a H * ) = 0 ,
where ~ch is the characteristic energy of inter-
actions usually assumed to be 10 (this corre-
sponds to a dimensional ~ h value of lOkT
units). It was also shown in (8) that for Ka
>> 1 Eq. [ 11 ] can be solved iteratively, giving
the approximate analytical solution
H* =--1 l n ~ - - - 1 In(1 + - - 1 ln~),
Ka Ka 2ga
Journal of Colloid and Interface Science, Vol. 146, No. 1, October 1, 1991
ADAMCZYK AND BELOUSCHEK
where ~ = 4~0/~bch.It should be mentioned that
H* is always larger than the dimensionless
double-layer thickness (Ka)-I because usually
~>1.
The 0mx data presented in Fig. 5, although
very interesting from a practical point of view,
provide one with averaged, global information
only, giving no hints as to how adsorbed par-
ticles are distributed over the collector surface.
This additional information can be extracted
~ A
from MC-simulated particle distributions by
determining the pair correlation function g12
according to the method described previously.
The gl2 function can be calculated for an ar-
bitrary surface concentration of particles
(provided that the total number of particles
considered is large enough to assure an appro-
priate level of accuracy), giving the possibility
of a quantitative characterization of ordering
processes occurring during particle adsorption
as a function of 0. In Fig. 6 the g~2 function
of hard spheres is presented for A equal to 10%
and for various surface concentrations includ-
ing also the limiting one characteristic for a
planar interface equal to 54.7%. One can see
in Fig, 6 that a definite tendency toward the
short-range ordering is especially pronounced
for higher surface concentrations close to the
[9] 0rex value.
It seems also that for surface concentration
0 approaching 0mxthe pair correlation function
of hard spheres possesses a singularity at q2
[10] = 2A (i.e., at the contact point of spheres).
Our finding is, therefore, in accordance with
previous results obtained for hard sphere ad-
sorption onto planar interfaces (3, 8, 14-16).
[11]
separations between adsorbed particles are
largely favored due to the "steric" repulsion
a m o n g adsorbed and adsorbing spheres. This
explains also why the geometrical blocking ef-
fect is much larger in our case in comparison
with the classical Langmuirian model.
Note also that the "absolute" surface con-
centration 0 as defined above is not a proper
[12] definition of the state of the adsorbed layer for
nonplanar collectors because the pair corre-
 PARTICLE
lation function g12 calculated for the same
value of 0 but different A differs markedly. It
is the relative surface concentration 0/0mx
which seems to be a proper measure of particle
"jamming" onto a collector surface. This sup-
position is confirmed by results presented in
Fig. 7, where the g12 function is plotted for the
maximum jamming state of adsorbed particles
for A = 0 ( p l a n a r surface), A = 10%, and A
= 20%. As can be seen in Fig. 7 the g12 function
is for all these cases almost identical (within
statistical error bounds). In the same figure
theoretical results derived from MC simula-
tions of 2D hard-disk fluid (17) are also plotted
for comparison (broken line in Fig. 7). One
can see that for a small gap between particles,
i.e., for H = r~2/A - 2 < 1, our results con-
cerning localized adsorption are very similar
to the 2D fluid phase behavior. However, for
larger distances between particles the short-
range ordering is significantly larger for the
2D fluid in comparison with our results, in
which the ordering manifests itself by larger
amplitude of the g12 function oscillation. From
results shown in Fig. 7 one can, therefore, draw
the conclusion that localized adsorption onto
a spherical collector is fully analogous to ad-
sorption onto the planar interface and can be
described by the same gl2 function provided
that the reduced surface concentration is used.
The pair correlation function for soft
spheres characterized by Ka = 10, ~ ~ 200,
and A = 10% is shown in Fig. 8 for various
particle surface concentrations 0, including the
m a x i m u m jamming value of 42%. One can
see in this figure that for 0 = 0 the g~2 function
approaches the Boltzmannian distribution,
i.e., exp(-~(r12)), whereas for large surface
concentrations a definite short-range ordering
is observed quite analogous to the hard sphere
adsorption (Figs. 6 and 7), although the g12
function for soft spheres has no singularity for
H = 0 and 0 --~ 0rex. It is alSO interesting to
mention that our results obtained for localized
adsorption of particles bear a striking resem-
blance to the 3D correlation function calcu-
lated for concentrated colloid suspensions in-
ADSORPTION 131
teracting via the repulsive double-layer poten-
tial ( 11, 12) and also to the true 3D molecular
liquids at lower temperatures ( 18, 19).
2. Cylindrical Collector
The cylindrical geometry is particularly in-
teresting from a practical point of view, con-
sidering the widespread use of fibrous filter
materials. Also, in many papermaking pro-
cesses latex suspensions are usually added to
the pulp in order to promote particle adsorp-
tion on cellulose fibers. In the latter case, as
well as in many biological systems (capillary
tube flows), particle adsorption and deposition
may also take place on the inside part of the
cylindrical collector interface. Thus, for this
collector we consider both possibilities, i.e.,
particle adsorption on the outside and on the
inside parts of the interface involved (from
within the cylinder). As stated previously, by
a proper increase of cylinder length for larger
A values, one can minimize fluctuations of
MC-simulated particle distributions. Many
calculation runs were performed in which the
cylinder length was varied (for a fixed value
of A) in order to determine the significance of
would-be edge effects. As found from these
tests, the influence of edge effects can be ef-
fectively eliminated for cylinder length, larger
than 50 particle diameters.
As stated previously, the definition of the
pair correlation function g12 for particle ad-
sorption onto the cylindrical interface is rather
ambiguous, especially for larger A parameter
values and, therefore, a quantitative deter-
mination of g12 was not attempted. Instead,
we present the simulated collector surfaces
bearing adsorbed particles. Thus, in Fig. 9 the
effect of the double-layer lateral forces is shown
for A = 10% and the same particle surface
concentration equal to 25% (adsorption oc-
curred on the external cylinder surface). As
can be easily seen in this figure double-layer
forces provoke a significant short-range or-
dering of particles, due to the fact that short
interparticle distances are practically inacces-
Journal of ColloM and Interface Science, Vol. 146, No. 1, October 1, 1991
 132 ADAMCZYK AND BELOUSCHEK

1 is shown in Fig. 11 for various A values. As

k,.o'~ ~ °o°O;~°o~£ oo o o ~ o % oo ~'~OoOO~o0OO8' oo~ ooOoO o,~
can be seen these results are quite analogous
to the spherical collector (with the exception
that fluctuations in particle surface concentra-
tion are practically absent due to reasons given
2
above). The dependence of 0 on 1 / f ~
0°°%00°0°0°00000° o°ooo. ° % ~ °0o~ooo%o oo~%O%~
o,~ ooo o° o o o o ~ o o ° O % ° o o o o ° o oOoOo o ~
~ o o o o o %0,'o°0 ~ o%O°° %o%oo ~ % OOOO°OO o ~0OoO°O~oO°£ becomes always linear for z >> 1, both for hard
and for soft sphere adsorption. Therefore, pre-
viously derived Eq. [6 ] retains its validity for
3
the cylindrical collector as well; i.e., the geo-
o~o%°%o 0%° oOO oo %0°%0°00%0 o o~ o ~ oo ° oo °o%o%~o~,~ o
metrical blocking effect is m u c h larger than
o~ooo . o o. . . .o Oo
i o °~° °g0. .%.o.°.oooo o °%oo%ooOo%Oog Oo OoOO?oOog .o. .oeo. £ o~
! ioOo ooOo%O~%OgoOo g o ~o . .o. . . ~ OOOoo~%Ooo
o~ the classical Langmuirian model predicts.
Results obtained for hard spheres can be
FIG. 9. The effect of lateral double-layer interactions alternatively expressed as the dependence of
on particle distribution over the cylindrical collector. A the reduced surface concentration 0/0mx on
= 10%, 0 = 25%. ( 1) ~a = oo (hard spheres); (2) ~a = 10
(~o = 200); (3) ~a = 5 (~o = 200). 1 / f ~ , as shown in Fig. 12. As can be deduced
from this figure the analytical relationship

sible, being highly improbable. Similar 2D 0mx = 0 ° ° ( 1 -I- A ) [13]

liquid-like ordering p h e n o m e n a are observed
for adsorption on the internal cylinder surface (where the plus and minus signs denote ad-
and for larger values of the aspect ratio A. sorption on the external and internal collector
Short-range structuralization p h e n o m e n a oc- interface, respectively) is in a quantitative
curring in the adsorbed layer on the cylindrical agreement with our MC simulations for the
interface are particularly well pronounced for entire range of the A parameter studied, from
0 close to the 0 m x values, as shown in Fig. 10
( f o r A = 10%). 10; 10 I T
The kinetics of hard sphere adsorption on
the external and internal cylindrical interface [°~l"
80
. 1

1
• 2

4
60 ~ 5
6
2 • 7

oo oo O.o~o%q oo ~voooo o~o% o o?~OOOog0000°%°-°000°00 OOo Oo

o~ - oo~ o o,~ ~ o o o ~ oo
40 ; i

o2;o&oOOooogOoOoo%OoOooo % oO % ° ° ° ° ° ° oOoOo.

20
0.2 0.4 0.6 o.8 1 1/(/-~ ,
F i G . 10. T h e e f f e c t o f d o u b l e - l a y e r f o r c e s o n 0rex. M C -
s i m u l a t e d p a r t i c l e d i s t r i b u t i o n s o v e r t h e c y l i n d r i c a l col- F I G . 11. K i n e t i c s o f h a r d s p h e r e a d s o r p t i o n onto the
lector forA = 10% a n d ( 1 ) ~ a = oo, 0rex = 6 1 % ( h a r d cylindrical interface 0 vs 1/l~r for (1) A = 60%; (2) A
spheres); (2) Ka = 10, 0rex= 37.5% (~0 = 200); (3) Ka = 50%; (3)A = 40%; (4)A = 30%; (5)A = 20%; ( 6 ) A
= 5, 0rex= 27.5% (~0 = 200)(only the upper part of the = 10%; ( 7 ) A = 1%; ( 8 ) A = 10% ( i n s i d e s u r f a c e ) ; ( 9 ) A
collector is shown). = 20% (inside); (10)A = 30% (inside).

Journal of Colloid and InterfaceScience, Vol. 146, No. 1, October 1, 1991

 PARTICLE A D S O R P T I O N 133

102 10 1 I" [ 11 ] and [ 121. The small positive deviation of

~s/e,.~ ' ' MC simulations from the analytical relation-
0.8 1 ~ ship Eq. [15] observed for Ka ~ 10 is caused
by the softness of the interaction potential•
It should be mentioned, however, that the
0.6 results presented above for the spherical and
cylindrical collectors apply strictly to perfectly
0.4. random and consecutive particle adsorption
processes, i.e., when pair correlations between
0.2 particles in suspensions are absent. Thus, our
adsorption model is particularly suited for di-
0 0.2 0.4 0.6 0.8 1' 1 lute suspensions and diffusion-controlled ad-
sorption regimes. The flow can complicate the
FIG. 12. Kinetics of hard sphere adsorption onto the adsorption kinetics by enhancing the geomet-
cylindrical interface, 0/0 ~ ( 1 _+A) vs 1 / !/7 for ( 1 )A = 30% rical blocking effect (20) but is expected to
(inside); (2) A = 10% (inside); (3) A = 10% (outside); affect little 0mx values (this should be studied
(4) A = 30% (outside); (5) A = 50% (outside). systematically in our future papers). In many
practical processes, e.g., filtration or paper-
making, the flow is very slow or collector par-
0 to 0.7. As with the spherical collector, all
results can be reduced to a single point only
by introducing the effective surface area avail-
able of particles given by the equation (gr~
[~/,]
S* = 47rRL(1 + A ) . [14]
Note that for A > 1 there exists no mech-
anism for bringing particles to the internal
• 2 ......
collector surface which has a sound physical 6C

interpretation. Particle adsorption may occur

in this case on the cylinder ends only, leading
to a fast clogging of the capillary. Thus, the
4E
amount of particles adsorbed is not well de-
fined and depends strongly on the cylinder
length. These aspects of particle adsorption on
cylindrical surfaces should be addressed in a 20
separate paper.
In Fig. 13 all results obtained for the cylin-
drical collector for both soft and hard spheres
are summarized in the form of the relationship i E i i I : i r i r i r i i
10 102 "~O
Of 0mx on the Ka parameter. As with the spher-
ical collector, results obtained for the cylinder FIG. 13. The m a x i m u m surface concentration of soft
can be well described by the analytical rela- spheres 0mx vs the Ka parameter (cylindrical collector).
tionship Results obtained from MC simulations are marked by dots,
and solid lines represent results calculated from the equa-
0mx = 0~(1 _ A)/(1 +H*) 2, [15] tion 0rex= 0 o~( 1 + A)/( 1 + H * )2, where H * was calculated
as previously for ~c~ = 10 and 30 = 200. ( 1 ) A = 50%;
where the effective range of lateral double-layer (2) A = 30%; ( 3 ) A = 10%; (4) A = 0 (planar interface );
interactions H* can be calculated from Eqs. (5) A = 20% (inside).

Journal of Colloid and Interface Science, V o L 146, N o . 1, O c t o b e r 1, 1991

134
tides (fibers) rotate due to the presence of
sheared liquid motion. Thus, they are sur-
rounded by regions of closed trajectories where
particle transport is mainly controlled by dif-
fusion. In these circumstances our results dis-
cussed above are expected to describe well
particle adsorption kinetics.
COMPARISON WITH EXPERIMENTAL
RESULTS
Despite its large theoretical and practical
significance the problem of determining the
adsorption kinetics and the maximum surface
concentration of soft spheres adsorbing on
various collectors was rarely addressed exper-
imentally in a systematic way. In a previous
paper (8) Adamczyk et al. presented results
obtained for the planar adsorbing surface using
the direct microscope observation method and
a monodisperse suspension of latex spheres
having the size of 0.9 tzm. The pair correlation
function determined experimentally for var-
ious ionic strengths (Ka parameter) and for
the surface concentration of adsorbed particles
was in good agreement with the theoretical
MC simulations. However, the 0mx values for
soft spheres were not determined explicitly,
nor was the kinetics of particle adsorption
studied systematically.
Feder and Giaever (3) performed experi-
ments aimed at determining the maximum
surface concentration of the protein ferritin
(whose molecule had a diameter of 100 _+ 10
A) adsorbing onto planar surfaces. Even by
using a highly concentrated saline solution
(0.15 M NaC1) they were unable to reach the
0 °° value equal to 54.7%, obtaining instead
values of 0mx between 20 and 50%. This can
be well accounted for by our theoretical pre-
dictions because for this electrolyte concen-
tration the value of the ra parameter is equal
to 12.5 for the above size of the ferritin mol-
ecule. Thus, from Fig. 5 one obtains a value
Of 0mxequal to 35% which agrees well with the
above experimental results of Feder and
Giaever (3).
Journal of Colloid and Interface Science, Vol. 146,No. 1, October 1, 1991
ADAMCZYK AND BELOUSCHEK
As far as the spherical or cylindrical collec-
tors are concerned any direct experimental
measurements of particle adsorption kinetics
are even more difficult. Vincent et al. ( 5 - 7 )
measured the adsorption of small positively
charged polystyrene particles (0.2/zm in di-
ameter) onto 3.2-t~m negative polystyrene
spheres (thus A --- 6% in their case). Adsorp-
tion of particles occurred over 24 h from stag-
nant suspensions of concentrations varying
between 2 × 101° and 2 × 1011 cm -3 (thus
the adsorption conditions were expected to be
in accordance with our model). The amount
of particles adsorbed was determined indi-
rectly from the concentration change in the
bulk of the suspension before and after exper-
iments. This simple experimental procedure,
although very efficient, seems to be charged
with some systematic experimental error be-
cause a mechanical retention of particles could
occur during the centrifugation of large par-
ticles. This supposition seems to be supported
by data presented in Table I, where experi-
mental values of 0rex are compared with our
theoretical MC simulations. One can see that
the experimental results are systematically
larger than theoretical ones. Assuming that
15% of the surface concentration measured
TABLE I
Comparison of MC Simulations with Experimental
Results o f V i n c e n t et aL (5)
NaCt(M) ra 0~(%) 0~b(%) 0~c(%)
10 -2 32 80 65 52
10 -3 10 45 30 38
10 -4 3.2 38 23 23
10 -5 1 30 15 11 a
N o t e . a = 0.2 # m , R = 3.2 ~ m , n ~ = 2 × 1011 (cm-3).
T = 293 K, t = 8.64 × 103 s (24 h), A = 6%,
= ~raZjon~t/O.55,jo = D ~ / R , D ~ = k776~rna.
a V i n c e n t et al. e x p e r i m e n t a l results uncorrected.
b V i n c e n t et al. e x p e r i m e n t a l results w i t h a 15% correc-
t i o n for m e c h a n i c a l retention.
c O u r M C simulations.
a M C s i m u l a t i o n s are d u b i o u s in this case because the
double-layer t h i c k n e s s is very large.
 PARTICLE ADSORPTION 135

can be attributed to particle mechanical re-
tention (this effect is expected to be fairly in-
dependent of the surface concentration), an
almost quantitative agreement between ex-
perimental and theoretical results can be
brought about. However, a quantitative test
of the validity of our theoretical predictions
can be performed only by carrying out more
precise experiments, enabling a direct obser-
vation of particle adsorption on collector sur-
faces in situ.
CONCLUDING REMARKS
Monte Carlo simulations of reversible lo-
calized adsorption of hard spheres onto the
planar, spherical, and cylindrical interfaces
have revealed that the maximum jamming
surface coverage 0rexcan be quantitatively de-
scribed by the analytical equation
0mx = 0~(1 + A ) m, [16]
where 0 o~ = 54.6% and m = 0 for the planar
itnerface, m = 1 for the cylindrical interface,
and m = 2 for the spherical interface; the plus
sign denotes adsorption on the external col-
lector surface and the minus sign adsorption
on the internal surface.
When introducing, instead of the geomet-
rical collector surface, the effective surface area
available for particles S* = So( 1 + A ) m all
results obtained for various collectors can be
reduced to one number only, i.e., ~ra2No/S *
= 0.546 (where Np is the average number of
particles adsorbed under the maximum jam-
ming limit).
On the other hand, the maximum surface
concentration of soft spheres interacting via
the repulsive screened Coulomb potential can
be adequately described by the analytical
equation valid for Ka ~> 1,
0mx = 0 ~ ( 1 +__A) rn
(1 + H * ) 2 ' [17] on the A parameter (although the 0rex value
where H* is the effective range of double-layer
interactions given by Eqs. [11 ] and [ 12]. qualitative agreement with existing experi-
Knowing that H* is usually much larger than
1/Ka one can expect that the range of appli-
cability of Eq. [17 ] can be extended to non-
spherical particles as well, e.g., rough, mineral
particles. This is so because at large separations
the detailed surface morphology can little af-
fect double-layer interaction forces. For Ka
10 our MC simulations produce 0 values
slightly above those predicted from Eq. [17]
which is the result of the "softness" of adsorb-
ing spheres. Our numerical simulations proved
also that these maximum surface concentra-
tions are attained for both hard and soft
spheres (Ka ~> 1 ) according to the equation
0 1 1
1 0mx V~T N~att ' [18]
where Natt is the total number of attempts at
adsorbing particles and ~ is the dimensionless
quasi-time variable. Equation [18] implies
unequivocably that the geometrical blocking
effect in the case of localized adsorption over
homogenous surfaces is much larger than in
the classical Langmuirian model (adsorption
on active centers of prescribed configuration
only) and can be expressed by the equation
d--~ ~ l- . [19]
The pair correlation function g12 deter-
mined for the spherical collector for a broad
range of the aspect ratio parameter A (includ-
ing A = 0, i.e., planar interface) revealed a
definite tendency toward a short-range liquid-
like ordering occurring within the adsorbed
particle layer quite analogous to 3D ordering
phenomena occurring in concentrate latex
suspensions. This ordering is especially pro-
nounced for 0 close to the 0rex values given by
Eq. [17]. The g12 function for hard spheres
exhibits a singularity for H ~ 0 at the maxi-
m u m jamming state and is then independent
varies considerably).
Our theoretical predictions are in a good

Journal of Colloid and Inte~ace Science, Vol. 146, No. I, October 1, 1991
 136 ADAMCZYK AND BELOUSCHEK
m e n t a l results, a l t h o u g h a precise, q u a n t i t a t i v e
v e r i f i c a t i o n o f o u r t h e o r e t i c a l results r e q u i r e s
further experiments carried out preferably by
using direct observation methods.
ACKNOWLEDGMENTS
This work was supported by the Av Humboldt Foun-
dation and partially by Grant CPBR 03.07.
REFERENCES
1. Adamczyk, Z., Dabro~, T., van de Ven, T. G. M., and
Czarnecki, J., Adv. Colloid Interface Sci. 19, 183
(1983).
2. Adamczyk, Z., Colloids Surf 39, 1 (1989).
3. Feder, J., J. Theor. Biol. 87, 237 (1980).
4. Feder, J., and Giaever, I., J. Colloid Interface Sci. 78,
144 (1980).
5. Vincent, B., Young, L. A., and Tadros, Th. F., Far-
aday Discuss. Chem. Soc. 65, 296 (1978).
6. Vincent, B., Young, L. A., and Tadros, Th. F., J.
Chem. Soc. Faraday Trans. 176, 665 (1980).
7. Vincent, B., Jafelicci, M., Luckham, P. F., and Tadros,
Th. F., J. Chem. Soc. Faraday Trans. 1 76, 674
(1980).
lournal of Colloid and Interface Science, Vol. 146, No. I, October 1, 1991
8. Adamczyk, Z., Zembala, M., Siwek, B., and War-
szyfiski, P., J. Colloid Interface Sci. 140, 123
(1990).
9. Van Megen, W., and Snook, I., J. Colloid Interface
Sci. 53, 172 (1975).
10. Snook, I., and van Megen, W., J. Chem. Soc. Faraday
Trans. H76, 216 (1976).
11. Van Megen, W., and Snook, I., J. Chem. Phys. 66,
813 (1977).
12. Svenson, B., and J6nsson, B., Mol. Phys. 50, 489
(1983).
13. Press, W. H., Flannery, B. P., Teukolsky, S. A., and
Vetterling, W. T., "Numerical Recipes: The Art of
Scientific Computing," Chap. 7, p. 196. Cambridge
Univ. Press, London/New York (1986).
14. Hinrichsen, E. L., Feder, J., and Jossang, T., J. Sta.
Phys. 44, 793 (1986).
15. Pomeau, Y., J. Phys. A: Math. Gen. A 13, L!93
(1980).
16. Swendsen, R. H., Phys. Rev. A 24, 504 (1981).
17. Wood, W. W., J. Chem. Phys. 48, 415 (1969).
18. Kirkwood, G., "Theory of Liquids" (I. Oppenheimer,
Ed.), p. 61. Gordon & Breach, New York/London/
Paris (1968).
19. Bowles, R, J., Tildeskey, D. J., and Quirke, J. Chem.
Soc. Faraday Trans. H85, 1505 (1989).
20. Dabrog, T., Colloids Surf. 39, 127 (1989).