Accelerat ing t he world's research.

Pore size distribution from challenge

coreflood testing by colloidal flow
Pavel Bedrikovetsky

Chemical Engineering Research and Design

Need to cite this paper? Want more papers like

this?
Get the citation in MLA, APA,
or Chicago styles Download a PDF Pack of
related papers

Search Academia's catalog of

22 million free papers

Downloaded from Academia.edu 
 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Research and Design

journal homepage: www.elsevier.com/locate/cherd

Pore size distribution from challenge coreflood testing

by colloidal flow

P. Chalk, N. Gooding, S. Hutten, Z. You, P. Bedrikovetsky ∗

Australian School of Petroleum, The University of Adelaide, SA 5005, Australia

a b s t r a c t

The transport of colloidal and suspension particles and the resultant particle retention occur in a wide range of
porous media. The micro scale pore throat size distribution is an important characteristic of porous media, allowing
for evaluation of important transport properties. An effective method based on micro scale modelling for the deter-
mination of overall pore throat size distribution (PSD) by injection of colloidal particle suspensions into engineered
porous media with monitored inlet and breakthrough particle concentrations is developed. The treatment of inlet
and outlet colloidal particle concentrations obtained in coreflooding results in a good agreement between the mod-
elling and experimental data. Yet, some deviation was observed between the obtained PSD and that calculated by
the Monte Carlo simulation based on the Descartes’ theorem.
© 2011 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Keywords: Pore size distribution; Porous media; Suspension; Colloid; Size exclusion; Monte Carlo; Challenge testing

1. Introduction
The pore size distribution (PSD) is an important character-
istic of the porous media allowing prediction of transport
and volumetric properties (Amix et al., 1964; Dullien, 1992;
Bedrikovetsky, 1993; Selyakov and Kadet, 1996). Different
methods for determination of PSD have been applied exten-
sively in many industrial sectors: chemical and environmental
engineering, petroleum, natural gas, food, medical and phar-
maceutical, etc.
Presently, there are two commonly used methods in the
industry for the determination of PSD: porosimetry and chal-
lenge testing. The method of porosimetry suggests injection
of non-wetting phase in the core under increasing pressure.
During the displacement, the injected phase fills the pores in
decreasing order of their radii and the curve “injection pres-
sure versus phase saturation” allows calculating PSD (Amix
et al., 1964; Brakel et al., 1981; Dullien, 1992; Yortsos, 1999).
Different options of the method use mercury, gas and other
non-wetting displacing fluids. Disadvantages of the method
include: it underestimates concentration of thin pores because
high pressures must be applied in order to force the non-
wetting fluid into thin pores; it can be destructive under
high pressure hence not applicable for deformable and fragile
porous materials; the last but not the least is environmental
unfriendliness of mercury for the case of mercury porosimetry.
The challenge testing method, which was recently signifi-
cantly improved, utilises the injection of particle suspension
into porous media (Purchas and Sutherland, 2002; Rideal,
2007, 2009). Since particle cannot pass the pore throat with
the size smaller than the particle size, the maximum size
of passed-through particle determines the maximum pore
throat size. Besides the maximum pore throat size, the inlet
and breakthrough concentrations for different size particles
allow determining the overall PSD curve, but only for thin
filters or membranes (Aimar et al., 1990; Frising et al., 2003;
Rideal, 2009). The method is based on calculation of the parti-
cle capture probability in a single sieve and cannot be applied
for deep bed filtration. Yet, usually the reservoir rock sam-
ple sizes vary as 0.5–10 cm; a thin slice can be submitted to
flow only for highly consolidated cores. To the best of our
knowledge, there is no available in the literature method to
determine the overall PSD of core plugs from particle size dis-
tribution in the injected and produced suspensions.
Determination of PSD from challenge testing data, i.e. using
inlet and breakthrough concentrations of different size par-
ticles, is an inverse problem for suspension flow in porous
media. Since pore and particle sizes affect the passage of

∗
Corresponding author. Tel.: +61 8 83033082.
E-mail address: pavel@asp.adelaide.edu.au (P. Bedrikovetsky).
Received 2 February 2011; Received in revised form 7 August 2011; Accepted 18 August 2011
0263-8762/$ – see front matter © 2011 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.cherd.2011.08.018
64 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77

Inlet particle distribution

Nomenclature
Outlet particle distribution
C suspended particle concentration distribution
(m−4 ) Pore size distribution
C0 inlet (injected) suspended particle concentra-
tion distribution (m−4 )
CL outlet (effluent, breakthrough) suspended par-
ticle concentration distribution (m−4 ) rpmin
rs rp rpmax
c total suspended particle concentration (m−3 )
fa accessible flow fraction Fig. 1 – Schema of determination of pore size distribution
fns inaccessible flow fraction from inlet and effluent particle size distributions.
k1 pore conductance (m4 )
k1a accessible pore conductance (m4 ) becomes the only particle capture mechanism. The population
H pore concentration distribution (m−3 ) balance model is chosen in this paper because it contains an
h total pore concentration (m−2 ) explicit relationship between particle and pore sizes as a cri-
j jamming ratio terion for size exclusion. In the case of particle–rock repulsion,
L length scale (m) the population balance model becomes simpler and contains
PSD pore throat size distribution the PSD along with the inter-chamber distance as the only
p fraction of pores thinner than particle input parameters. The inverse problem is solved for the case
q1 flow rate in a single pore (m3 s−1 ) of short duration injection with low particle concentrations,
rg grain radius (m) where the initial PSD remains constant. Five laboratory tests
rp pore radius (m) on injection of latex particles into engineered porous media
<rp > mean pore radius (m) have been performed. The PSD is determined from the labora-
rs particle radius (m) tory data using solution of the inverse problem. Good match
s1 cross-sectional area of pore throat (m) between the experimental and modelling data was observed.
t time (s) Yet, some deviation between PSD as obtained from the Monte
U total flux (m s−1 ) Carlo simulation using Descartes’ theorem and that calculated
V column volume (m3 ) from breakthrough concentrations was observed.
x coordinate (m) The structure of this paper is as follows: Section 2 starts
with the general description of the challenge coreflood test
Greek letters for PSD determination. Section 3 presents the derivation of
 filtration coefficient (m−1 ) the analytical micro model for suspension flow in porous
 flux reduction factor in a single pore media with size exclusion particle capture. Subsequently, the
˙ retained particle concentration distribution laboratory procedure for injection of latex particles into a
(m−4 ) transparent plastic box with one layer of glass bead grains and
 total retained particle concentration (m−3 ) into a column packed with these glass beads is presented (Sec-
0 standard deviation for pore size distribution (m) tion 4). Determination of PSD from grain size distribution by
 porosity the Descartes’ model and Monte Carlo simulations is given in
a accessible porosity Section 5. The data treatment is performed in Section 6. The
 area accessibility fraction discussions presented in Section 7 and conclusions in Section
8 complete the paper.

particle through porous media, the pore size is an ade- 2. Description of the problem
quate length scale for modelling in order to determine PSD
(Payatakes et al., 1973, 1974). Presently used mathematical Let us consider the challenge testing, which includes sus-
models for transport of particulated suspensions in porous pension flow in porous specimen measuring inlet and outlet
materials on micro scale include random walk equations concentrations of each size particles (Fig. 1). Fig. 2 illustrates
(Cortis et al., 2006; Shapiro, 2007; Yuan and Shapiro, 2010), the definition of the pore throat size as a radius of the min-
population balance models (Sharma and Yortsos, 1987a,b,c; imum sphere inscribed between three neighbouring grains.
Santos and Bedrikovetsky, 2006; Bedrikovetsky, 2008), equa-
tions for resident time distributions (Fan et al., 2008; Lin et al.,
2009), different versions of Boltzmann model (Shapiro and
Wesselingh, 2008; Tang et al., 2010) and direct micro scale
modelling (Biggs et al., 2003; Rong et al., 2010). These models
are widely used in chemical, environmental and petroleum
engineering (Bergendahl and Grasso, 2000; Mays and Hunt,
2005; Gitis et al., 2010).
In the present work, the PSD is determined from the chal-
lenge testing for “long” engineered porous columns where the
deep bed filtration occurs. In order to avoid cumbersome and
expensive characterisation of particle–rock interaction, it was
proposed to use the liquid and particle materials that promote
net repulsion between particles and rock, so size exclusion Fig. 2 – Formation of the pore throat by three tangent grains.
 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77
Particle passes the pore if the particle size does not exceed
the pore size:
rs < rp .
Further in the text, the term “thin pore” means that it
is smaller than the passing particle, i.e. inequality (1) is ful-
filled.
The water composition (salinity, pH) for a given particle
and grain material is selected in such a way, that particles
repulse from the rock grains, so no particles become attached
to the rock and size exclusion mechanism determines particle
capture by the rock.
The particles larger than the maximum pore do not pass.
Therefore, their breakthrough concentration equals zero. It
allows calculating the maximum pore size as the minimum
size of the particle with zero breakthrough concentration
(Fig. 1). The particles smaller than the minimum size pore,
pass through the specimen without being captured. It permits
calculation of the minimum pore size as the maximum size of
particles with coinciding inlet and breakthrough concentra-
tions. So, two end points of the PSD curve can be determined
directly from the pair of concentration curves for size exclu-
sion deep bed filtration.
Fig. 1 shows that the larger is the particle the higher is the
capture probability, i.e. the larger is the difference between
inlet and outlet concentration. This phenomenon is described
by the population balance model for suspension transport
with size exclusion in porous media. The direct problem deliv-
ers breakthrough concentration distribution C(rs , x = L, t) for
known injected concentration distribution C(rs , x = 0, t) and
pore concentration distribution H(rp ). The inverse problem
determines the PSD from the inlet and effluent concentration
curves.
Since the majority of the models uses a cylindrical pore
shape, where the pore body and pore throat sizes coincide, the
term “pore size distribution” is commonly used instead of the
“pore throat size distribution”. As follows from inequality (1),
the challenge method determines distribution of pore throats.
Keeping it in mind and for the sake of simplicity, we use both
terms.
3. Micro scale model for suspension flow in
porous media with particle straining
In this section, derivation of the system of governing equa-
tions for suspension flow in porous media with size exclusion
particle capture mechanism is presented for triangular pore
throat form. The parallel tube model intercalated by mix-
ing chambers is used to describe the geometry of the porous
medium. The analytical model for low retention filtration
and the solution of the inverse problem of PSD determina-
tion are provided. The definitions and nomenclature follow
Bedrikovetsky (2008).
3.1. Definition of variables for micro scale model
Let us consider suspension flow in granulated porous media.
The triangular shape capillaries (Fig. 2) more precisely describe
pore space between spherical grains than the circular capillary
(Mason and Morrow, 1991; Patzek and Silin, 2001).
The main features of size exclusion deep bed filtration
of suspension in three dimensional porous space are as
follows:
65
(1)
Fig. 3 – Schematic of suspension flow towards the cross
section of porous media represented by the PTM model: (a)
flow through capillaries and mixing chambers, (b) size
distribution of triangular capillaries.
particle flows in a pore with the throat larger than the parti-
cle, (1);
the particle has choice of entering different pores in the cap-
illary junction; the probability of entering the pore is equal
to the fractional flow via this pore;
if the particle is larger than the pore, the particle capture
occurs upon the entrance into the pore throat.
Fig. 3 presents the parallel bundle of different size capillary
intercalated by the mixing chambers as a geometric model of
porous media, where the size exclusion suspension flow takes
place. A suspended particle may only be transported to the
next chamber if it is sufficiently small enough to pass through
the capillary, (1). Full mixing of the inlet particle fluxes occurs
inside the chambers. The particles enter the capillary from
the chamber according to the carrier water fluxes via the cap-
illary. Depending on whether the particle is larger or smaller
than the entrance capillary, it will be either captured or passed.
The particle capture occurs at the chamber exits. So, particles
are effectively sieved at the entrance to each capillary bundle
according to their sizes. The geometric model of parallel cap-
illary intercalated by the mixing chambers fulfils three above
mentioned features. The model is determined by the concen-
tration distribution H(rp , x, t) and the distance between the
chambers l.
A dimensionless term, the jamming ratio j, is introduced to
quantify the ratio between particle size and pore throat size:
rs
j= . (2)
rp
The total concentration of all sized pores h(x, t) is defined
as the total number of pores per unit surface area at certain
distance x and time t:
∞

h(x, t) = H(rp , x, t)drp , (3)
0
where H(rp , x, t)drp is the concentration of pores with sizes
between rp and rp + drp .
The population of particles in porous media is described in
similar manner. The concentration of all suspended particles
in porous media, c(x, t), is defined as the total number of parti-
cles per unit volume at certain distance x and time t, resulting
from the integration of concentration of particles having radii
between rs and rs + drs :
 ∞
c(x, t) = C(rs , x, t)drs , (4)
0
where C(rs , x, t) is the particle concentration distribution. Sim-
ilarly, the concentration distribution of retained particles ˙(rs ,
x, t) is defined as the number of retained particles ˙(rs , x, t)drs
66 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77

having radii between rs and rs + drs . The total concentration of For triangular pores, the conductance is calculated from
retained particles (x, t) is obtained as exact solution of the one dimensional viscous steady state
flow of Newtonian fluid (see Landau and Lifshitz, 1987): k1 =
 ∞ √
9 3rp4 /20.
(x, t) = ˙(rs , x, t)drs . (5)
0 Likewise, the inaccessible flow fraction through thin pores
(Fig. 4), fns , is defined by
The retained particle concentration can also be distributed
according to rp since both rp and rs take effect in particle cap-  rs
Uns k1 (rp )H(rp , x, t)drp
ture. Therefore, we introduce ˙ - (rp , rs , x, t) as the distribution fns = =  o∞ . (11)
U k1 (rp )H(rp , x, t)drp
of retained particle concentration over the pore and particle o
radii:
The flow through inaccessible areas in large pores, fnl , is
  ∞

(rs , x, t) = (rp , rs , x, t)drp . (6)
- ∞
o
rs
k1 (1 − (rs /rp ))H(rp , x, t)drp
fnl [H, r] = ∞ . (12)
The curvilinear pore throat presented in Fig. 2 may be o
k1 H(rp , x, t)drp
approximated by a circle or a triangle, leading to two defini-
tions of a single pore throat’s cross-sectional area: The accessible conductance formula obtained from k1
replacing rp by rp -rs is described below in Eq. (13) for the two
S1,o (rp ) = rp2

considered pore throat approximations:
S1,i (rp ) = √ . (7)
S1, (rp ) = 3 3rp2

⎨ k1a,o (rp ) = 8 rp4 (1 − j)2 (1 + 2j − j2 )
⎧
Here index i = o corresponds to a circles for cylindrical capillary √
k1a,i (rp ) = .
⎩ k1a, (rp ) = 9 3 r4 (1 − j)2 1 + 2j − 1 j2 − 4 j3

in PTM; i =  stands for capillary with triangular shape.

p
The accessibility of pore space to a particular sized particle 20 3 9
(13)
is limited by the particle radius. This leads to the notions of
accessible and inaccessible area as shown in Fig. 4. The acces-
sibility factor, (j), describes the ratio between the accessible
and total area. As it follows from (7), the accessibility factor The flux reduction factors for circular and triangular pore
for circular pores, where the radius of accessible circle is rp -rs , throats follow from (13):
is
⎧
k
2 ⎨ o (rp ) = 1a,o = (1 − j)2 (1 + 2j − j2 )
⎪
(j) = (1 − j) . (8) k1,o
i = . (14)
⎩  (rp ) = k1a, = (1 − j)2 1 + 2j − 1 j2 − 4 j3

⎪
The radius of inscribed circle in the accessible area for tri- k1, 3 9
angular pore is equal to rp -rs , which also results in expression
(8) for accessibility factor (Ilina et al., 2008; Panfilov et al., 2008; 3.2. Governing equations for suspension flow in
Bedrikovetsky, 2008). porous media
The particle capture kinetics requires the definition of the
accessible and inaccessible flow fractions fa and fn , respec- In this section, the governing equations of suspension
tively. For the PTM, the accessible flow fraction is defined by transport in porous media with brief derivations follow
∞ Bedrikovetsky (2008).
Ua rs
k1 (rp )(j)H(rp , x, t)drp The main assumptions of the model for transport of the
fa = = . (9)
colloidal material in porous media include: no particle coagu-
∞
U k1 (rp )H(rp , x, t)drp
o
lation or attrition, negligible external filter cake, particle–rock
where  = (j) is the flux reduction factor, which is the ratio repulsion leading to size exclusion mechanism of the particle
between fluxes via the accessible and the overall pore cross capture, no diffusion, one pore is blocked by just one particle
section and k1 = rp4 /8 is the pore conductance determined (no bridging), homogeneous porous media (Herzig et al., 1970;
from the Poiseuille’s flow in a circular tube Sharma and Yortsos, 1987a,b,c; Grenier et al., 2008; Rebai et al.,
2010; Daniel et al., 2011). Low retention process is considered,
r4 dp k1 dp i.e. the amount of retained particles is negligibly smaller that
q1 = − =− (10)
8 dx  dx the number of vacant pores available for the particle capture.

Fig. 4 – Accessible and inaccessible pore space.

 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77 67

The accessible porosity is the total of accessible pore cross Substituting the expression for retention rate (19) into mass
sectional areas per unit of the porous media cross section balance equation (16) yields

 ∞ ∂ ∂
{a [H, rs ]C(rs , x, t)} + U {C(rs , x, t)fa [H, rs ]}
a [H, rs ] = s1,i (rp )(j)H(rp , x, t)drp . (15) ∂t ∂x
rs
1
= − fns [H, rs ]C(rs , x, t)fa [H, rs ]U, (21)
l
where s1,i (rp ) is the cross-sectional area of a single pore throat
having radius rp and (j) is the area accessibility fraction. Here
round brackets are used for functions of scalar variables and System of two Eqs. (20) and (21) for two unknowns describes
square brackets are used for the functionals of functions. size exclusion deep bed filtration. The initial conditions for the
The population balance equation for colloidal particles in clean bed process include zero suspension concentration and
porous media for the above assumptions accounts for sus- initial uniform PSD for pores:
pended particles flowing via accessible area and retained
particles distributed throughout the overall porous space: t = 0 : C(rs , x, 0) = 0, H(rp , x, 0) = H0 (rp ). (22)

∂  ∂ The boundary conditions include the model for particle

{a [H, rs ]C(rs , x, t) + (rs , x, t)} + U {C(rs , x, t)fa [H, rs ]} = 0.
∂t ∂x
(16)
Here the flow velocity U is independent of co-ordinate x due
to incompressibility of particulate suspension.
The derivation of the kinetic equation for particle cap-
ture and pore plugging is based on the aforementioned PTM
model. Since C(rs , x, t) is particle concentration in suspen-
sion, i.e. number of particles per volume of accessible area, the
suspension concentration in chambers is fa C(rs , x, t); suspen-
sion dissolves in the clean water flowing into the chambers
from inaccessible pores. The assumption of perfect mixing
between capillary bundles indicates that colloidal particles are
randomly presented to a pore throat and there exists no pref-
erence for a colloidal particle to a particular throat. Therefore
the number of rs particles arriving at pores with size rp from
chambers by the flow per unit volume during the time interval
t is
capture and percolation at thin layer with thickness of a few
pore lengths. Consider a porous column of the packed glass
beads, since this porous media will be used in laboratory tests
further described in the paper. It is assumed that particles,
which enter thin pores, will remain there after capturing,
i.e. the inlet particle flux Cfns U is completely captured in the
inlet cross section. The particles entering thick pores via the
inaccessible area are not captured due to the particle–grain
repulsion, i.e. the entrance flux Cfnl U is not captured. These
two assumptions about the particle capture at the inlet x = 0
correspond to the particle retention expression (19) for x > 0.
The inlet flux via the accessible area Cfa U is not captured.
Mass conservation at the inlet cross section is the equal-
ity of the particle flux, entering larger pores with the injected
concentration, and the flux of particles, transported by water
in accessible pore space with boundary value concentration
x = 0 : C0 (rs , t)(fa [H, rs ] + fnl [H, rs ])U = C(rs , 0, t)fa [H, rs ]U, (23)

1
q1 (rp )H(rp , x, t)C(rs , x, t)fa [H, rs ]t. (17) Since the injected carrier water flux exceeds the post-
l
inlet water flux carrying particles, the post-inlet concentration
It is assumed that particles are not captured in inaccessible exceeds the injected particle concentration. Eq. (23) deter-
areas of larger pores (1) under the particle–grain repulsion. The mines the post-inlet concentration and provides the boundary
suspended particles are captured by smaller pores only. condition for population balance equation (21).
The rate of retained rs particles being removed by smaller rp
pores, rp < rs , is equal to the overall particle flux from chambers 3.3. The analytical solution
into these pores
In this section we show that the assumption of low retention
∂

(rp , rs , x, t) filtration results in linearization of the governing system (16),
1
- = q1 (rp )H(rp , x, t)C(rs , x, t)fa [H, rs ]. (18) (19) and (20) causing separate independent filtration of particle
∂ l
populations with different sizes. This feature finally enables
Integration of (2) in rp for all smaller pores yields the expres- determining the pore size distribution from the effluent and
sion for the retention rate of rs particles inlet particle size distributions (Section 6.2). The linearized
flow can be described by explicit analytical formulae.
Let us consider steady state of suspended particles along

∂ (rs , x, t) 1
- = UC(rs , x, t)fa [H, rs ]fns [H, rs ]. (19) the core, where the difference between the number of parti-
∂t l
cles entering and exiting the reference volume is captured by
From the assumption that one retained particle plugs one the porous media. Eq. (21) becomes an ordinary differential
pore, it follows that the rp pore plugging rate due to particle equation:
capture is equal to the overall flux of larger particles flowing
d 1
into rp pores from chambers U {C(rs , x)fa [H, rs ]} = − fns [H, rs ]C(rs , x)fa [H, rs ]U, (24)
dx l
 ∞
∂H(rp , x, t) ∂  Let us discuss the initial stage of size exclusion deep bed
=− (rs , x, t)drs
∂t 0
∂t filtration, where retained concentration is negligible if com-
 ∞ pared with the vacant pore concentration. It happens either
1
=− UC(rs , x, t)fa [H, rs ]fns [H, rs ]drs , (20) at small injection times, or at injection of diluted suspensions
l 0 or at small filtration coefficient (small particle size). Since it
68 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77
is assumed that one pore is plugged by one particle and vice
versa, the conditions of particle capture do not change during
low retention filtration, i.e. the pore concentration remains
intact. The same conclusion can be drawn from Eq. (20) for
pore plugging rate – the time variation of the PSD tends to zero
when either injected concentration tends to zero, the correla-
tion length l (inter chamber distance) tends to infinity or the
particle size does not exceed the minimum pore size.
The initial PSD for pores is uniform. So, the assumption of
low retention filtration leads to constant PSD during the injec-
tion, i.e. H(rp , x, t) = H0 (rp ). Therefore, flow fractions become
functions of particle size only and will be denoted as fj (rs ),
j = a, ns and nl further in the text. In this case, Eq. (24) is solved
independently for each particle size, i.e. the particles of the dif-
ferent sizes filter independently from each other and deposit
as a bunch of mono disperse suspensions.
Boundary condition (23) becomes
C0 (rs )(fa (rs ) + fnl (rs ))
C(rs , 0) = , (25)
fa (rs )
Eq. (24) becomes
dC(rs , x) 1 other than straining, in particular retention via attachment.
= − fns (rs )C(rs , x),
dx l The objective of this preliminary study is to design and
Solving ordinary differential equation (26) subject to ini-
tial condition (25) yields the value of the pre-outlet particle
concentration
C0 (rs )(fa (rs ) + fnl (rs ))
x

C(rs , x) = exp −fns (rs )
fa (rs ) l
The outlet upstream particle flux is equal to the down-
stream particle flux immediately after the core
C(rs , x)fa (rs )U = CL (rs )U,
where CL (rs ) is the particle concentration measured in pro-
duced suspension sample. Further in the text CL (rs ) is called
the effluent (breakthrough, outlet) concentration. Since the
water flux carrying particles upstream the outlet is lower than
the downstream flux, the pre-outlet concentration is higher
than the breakthrough concentration. The pre-outlet particle
flux is dissolved in the overall water flux after passing the core
outlet.
From (27) and (28) follows that
CL (rs ) = C(rs , x)fa (rs ) = C0 (rs )(fa (rs ) + fnl (rs )) exp −fns (rs )
Formula (29) connects the injected concentration with that
measured after the core outlet. It will be used further in the
text for determining PSD H0 (rs ) from the values of the break-
through concentration CL (rs ) as obtained in laboratory tests for
different injected particle sizes rs .
The formula show that the larger are the injected particles,
the higher is the fractional flow through small inaccessi-
ble pores fns , the lower is the fractional flow via large pores
fa + fnl = 1 − fns and the lower is the exponent term in (29).
Finally, the lower is the effluent concentration CL . The efflu-
ent concentration vanishes when the particle size reaches the
maximum pore size, where the fractional flow via large pores
fa + fnl becomes zero.
The steady state distribution of suspended particles in the
overall core is established after arrival of the suspension front
to the core outlet (Herzig et al., 1970). As it follows from (21),
the velocity of the front is Ufa /a , so the constant value of the
effluent concentration (29) is established for t > La /Ufa .
4. Laboratory procedure for forcing
suspensions through porous media
In this section, the conditions of particle–grain repulsion and
the consequent absence of particle retention due to attach-
ment are established by tests in a simplified one-grain-layer
engineered porous media. Then the laboratory procedures
developed for the flow of colloidal suspensions through glass-
bead-packed porous media are discussed.
4.1. Preliminary study
An experimental method is developed where particle reten-
tion in porous media occurs exclusively due to particle strain-
ing. Operational conditions are experimentally determined for
the purpose of eliminating particle retention via mechanisms
(26)
construct a micro model unit that provides the visual obser-
vation of colloidal suspension flow through porous media.
This experiment allows the determination of the solution
chemistry that promotes the particle–surface repulsion forces
between porous structure and suspended colloids and, there-
, (27) fore, minimises the particle attachment.
In the developed experimental method, a colloidal suspen-
sion of spherical, fluorescent carboxyl latex microspheres is
injected through an engineered porous medium. The yellow-
green (rs = 4.9 ␮m) and blue (rs = 2.3 ␮m) fluorescent latex
(28) microspheres (Polysciences Inc., Warrington, PA) are selected
as the colloidal particles held in suspension. The surfaces
of these colloids are grafted with carboxyl functionalised
groups by the manufacturer. This creates a negatively charged
hydrophilic colloidal surface possessing a net negative charge
when in an alkaline solution. The net charge of the sur-
face prevents coagulation of colloids and reduces electrostatic
attraction to the porous media. The colloid concentration is
held constant at 20 ppm in all the preliminary tests.
A single layer of sieved and cleaned spherical glass
beads (Ballotini Bead, Potters Industries Pty. Ltd., Australia) is
adopted as a 2D porous medium. The main component of the
L porous medium, silica (SiO2 ), has a net negative surface charge
,
l in alkaline solutions (SiO4 4− ). Silica is also selected because
(29) it represents the fundamental material of sandstone reser-
voirs. According to the manufacturer, the glass beads have
a reported chemical composition: 72.0% SiO2 , 15.0% Na2 O,
7.0% CaO, 4.2% MgO, 0.4% Fe2 O3 and 0.3% Al2 O3 . These glass
beads are packed homogeneously with a theoretical porosity
of 39.6%. Two sizes of beads are chosen for the tests with mean
radii 150 ␮m and 300 ␮m, respectively.
In order to examine the attachment and straining of col-
loids moving through a single layer of glass beads, the micro
model housings are designed to support the observation
under an optical microscope. The housings are milled out
of polyvinyl chloride (PVC) plastic and are designed in such
a way that two glass slides are held in place to contain the
porous media in a single layer. The deionised ultrapure MilliQ
water (resisitivity of 18.2 M × cm at 25 ◦ C) after degassing in
 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77
Fig. 5 – Images of particles strained in porous media: (a) different salinities and (b) different pH levels.
vacuum at pressure ≈ 10−2 Pa is used for the preparation of
a colloidal suspension. The salinity, acidity and alkalinity of
the prepared suspensions are adjusted with the addition of
NaCl(aq), HCl(aq) and NaOH(aq), respectively.
The effect of colloidal suspension salinity on the parti-
cle retention is investigated under three different NaCl(aq)
concentrations (0 mM, 10 mM and 100 mM) at constant acid-
ity/alkalinity. A proportional increase in particle attachment
with salinity is observed throughout these experiments, as
shown in Fig. 5(a), where the images of colloid particles are
small bright “points” while the images of glass beads are semi-
transparent large circles. Simultaneously, the particle–grain
interaction force was calculated for three salinity values using
DLVO theory (Landau and Lifshitz, 1980; Khilar and Fogler,
1998). Three plots of the force versus the particle–surface
separation distance show that the increase in retention is
due to an increased depth of the secondary minimum and
a reduced energy barrier to the primary minimum (Tufenkji
and Elimelech, 2004, 2005). The same results were obtained by
Kuznar and Elimelech (2007).
Colloidal particle retention by glass beads is investigated
using suspensions with five different pH levels (2.79, 4.26, 7,
8.48 and 10.38) at constant salinity. The monotonic decrease
in attachment is found to occur with increased alkalinity. The
colloid retention in the unit is comparatively low in all alka-
line resident solutions, as shown in Fig. 5(b). The high particle
retention in acidic conditions is due to the reduction in the
strength of the electrical charge on the surface of the glass
beads and the colloidal particles.
In order to minimise the colloidal particle attachment,
the resident solutions with high pH and pure water must
be employed so that the physical capture mechanisms
would dominate. The above conclusions indicate that net
repulsion exists between the porous media and the colloidal
suspensions.
4.2. Experimental procedure
Using the outcomes of the preliminary study, an experimen-
tal setup and procedure simulating the flow of suspended
69
particles through a 3D pore structure is developed. A plas-
tic column with a packing space of 47 mm in diameter and
L = 50 mm in length to simulate the dimensions of core sam-
ples is constructed. The column design has incorporated
homogenised sections to negate the end effects, allowing for
the particle suspension to be injected as a piston-like front
into the packing matrix. A 30 ␮m sized stainless steel mesh is
utilised to contain the porous media within the column. The
application of distribution plates ensures the rigidity of the
packed glass beads. An exploded diagram of the resulting col-
umn with its components is demonstrated in Fig. 6(a). Fig. 6(b)
shows the photo of the column.
Monodisperse suspensions of yellow-green fluorescent
polystyrene latex microspheres used in the preliminary study
are taken as colloidal particles in these experiments. Five sizes
of microspheres are used in these experiments. Table 1 sum-
marises the specific physical characteristics of the colloids,
where rs denotes the mean radius of the colloid particle,  is
the standard deviation, and s is the particle density.
Glass beads similar to those used in the preliminary exper-
iments are utilised as the porous media in all experimental
procedures. The glass bead sizes, as reported by supplier, are
10–80 ␮m. A thorough sieving process is performed to reduce
the size range of the glass beads; the size variation after the
sieving becomes 10–66 ␮m. A complete washing procedure is
also performed to remove organic impurities using acetone,
hexane and hydrochloric acid.
Colloids are held in suspension solution made up with
0.1 M sodium hydroxide and degassed ultrapure MilliQ water
Table 1 – Size distributions (mean radius <rs > and
standard deviation  0 ) and densities of the colloidal
particles.
Type <rs >, ␮m  0 , ␮m s , g cm−3
1 1.789 0.339 1.055
2 2.297 0.457 1.055
3 2.976 0.342 1.055
4 4.915 0.796 1.055
5 9.844 0.059 1.055
70 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77
Fig. 6 – The schematic of the column used for experiment:
(a) exploded diagram, (b) photo of the column.
to create a resident solution with a pH of 10. The effect
of pH variation on particle retention is investigated by
repeating the same experiment in acidic and neutral pH
environment.
In order to exclude the possibility of air pockets ingress to
the packed glass bead bed, the column is wet packed with glass
beads for all experiments. The packed column was placed
in the sonic bath for 30 min, allowing achieving the dense

2
and close packing of glass beads. Afterwards, the column
is connected to a syringe piston pump (AdelabNE-1000). A
steady state bottom-up flow of the colloidal solution with lin-
ear velocity of 10−5 m s−1 is established through a vertically
placed packed column to minimise sedimentation due to the
density difference between the colloids and the resident solu-
tion. Five pore volumes of the resident solution are flushed
through the column to remove any particulate matter from
the porous media that may have contaminated the effluent
colloid sample. Samples of the 50 mL effluent colloidal solu-
tion are collected and passed through the PAMAS S4031 GO
portable particle counter to determine the effluent particle
concentration.
The effluent solution sampling should be stopped after a
stable concentration of colloidal particles in the outlet stream
is observed. A standard deviation (in %) from the mean value
for all colloidal particles ranging from 1.789 to 9.844 ␮m was
introduced by repetitive measurement (up to 5 measurements)
of the same particle suspension. This standard deviation
may be adopted as a measure of experimental uncertainty
of PAMAS particle counter. Therefore, if the concentration of
effluent colloidal particles for the next sample differs from
the previous one by the value less than the above stan-
dard deviation, then the sampling process and flow-through
test should be stopped, otherwise the experiment should
continue.
5. Determination of pore size distribution
from grain sizes
A predictive method for the PSD from the measured grain
size distribution is presented in this section. The validation
of Monte Carlo simulation for PSD using Descartes’ theorem
is performed by treating the literature data. Afterwards, this
method is used for calculation of PSD using known size dis-
tribution of glass beads in the porous columns described in
Section 4.
An idealised porous medium can be represented as a gran-
ular packing with distribution of grain sizes. The typical for
petroleum reservoirs sandstone rocks are often treated as the
packing of spherical grains (Amix et al., 1964). This simplifi-
cation is commonly used to model and simulate petroleum
systems (Dullien, 1992).
Assume that the spherical grains are densely and closely
packed, and each pore body is formed by 4 grains. This type
of packing is the densest for the case of equal radii beads.
Four throats exit the same pore body. Therefore, the throats
have a form of curvilinear triangle (Fig. 2). Consider a particle
approaching the pore throat. The throat is formed by 3 neigh-
bouring grains. The radius of a sphere touching these 3 grains
reaches minimum if the sphere centre is located on the plane
that crosses centres of 3 grains. Therefore, the radius of a pore
throat is effectively defined as the maximum inscribed radius
of three mutually tangent grains in the plane formed by the
centres of those spherical grains (Fig. 2). Any larger spherical
particle does not pass the throat. As a result, analysis of the
pore throat size is dissolved to a 2D problem whose solution
can be determined by the Descartes’ theorem, which was first
published in 1936 by Soddy (Descartes, 1901; Pedoe, 1967). The
theorem states that if four circles in a plane touch each other
externally, with the radii r1 , r2 , r3 and rp , then the following
relationship holds:
1 1 1 1 1 1 1 1
2 + + + = + + + . (30)
r12 r22 r32 rp2 r1 r2 r3 rp
Because of the random nature of the grain packing, it is
beneficial to employ a statistical algorithm for the simulation.
In this work, the Monte Carlo method (Binder and Heermann,
1988), which relies on repeated random sampling to calculate
 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77 71

(a) (b)
7 8

6 7

6
5
5
f(rp ) 4

f(rp)
4
3
3
2
2
1 1

0 0
0 0.5 1 1.5 0 0.5 1 1.5
rp (µm) rp (µm)

Nolan & Kavanagh (close packed c) Nolan & Kavanagh (close packed d)

MC (Descarte's theorem) MC (Descarte's theorem)

Fig. 7 – Validation of the Monte Carlo model for pore size distribution using Descartes’ theorem: (a, b) treatment of data by
Nolan and Kavanagh (1994) (♦); results of modelling ().

the results, is applied. The method is suitable for modelling
phenomena with significant input uncertainty and studies
systems with a large number of coupled degrees of freedom.
Applying the Monte Carlo simulation with Latin Hypercube
selection criteria utilising Descartes’ theorem, the PSD curve
can be determined from the grain size distribution.
Let us validate the proposed method for prediction of PSD.
Nolan and Kavanagh (1994) determined the PSD from a packed
column of spherical particles with a given size distribution
using geometric 3D simulation. The pore throat sizes were
described by the maximum inscribed radius, geometrically
restricted by the confining spheres in a packed system. Fig. 7
presents the results of Monte Carlo simulation for the case
when three grains are chosen according to their probability
distribution to generate a single pore radius by Descartes’ the-
orem presented in Section 2. The process is repeated 100,000
times using Latin-hypercube sampling in a Monte Carlo simu-
lator to generate a PSD, and the predicted curves are smooth.
On the contrary, the curves obtained by Nolan and Kavanagh
and shown in Fig. 7 (a) and (b) exhibit oscillations due to small
number of Monte Carlo runs. Despite some deviation between
the prediction results by the Descartes’ method and those
presented by Nolan and Kavanagh (1994), the quality of the
agreement allows for utilisation of the proposed method for
the evaluation of PSD from known size distribution of packed
glass beads.
Several other cases (Rouault and Assouline, 1998; Al-
Raoush et al., 2003; Reboul et al., 2008) have been investigated
as well and show the same order of magnitude in deviation
between the Descartes’ prediction and geometric simulation.
The proposed method aims at an approximate evaluation and
Section 6 to validate the method for PSD calculation from the
suspension flow test.
6. Treatment of experimental results
In this Section, it is verified whether the conditions of low
retention filtration leading to a simple analytical model (29)
have been fulfilled during the laboratory tests. Afterwards, the
laboratory data are treated using the analytical model.
6.1. Validation of the low retention assumption
Let us check whether the assumption of the analytical model
(Section 3.3) that the PSD during particle straining in the
porous media is invariant have been fulfilled in the laboratory
tests. This occurs when the concentration of pore vacancies
(thin throats) is significantly higher than the concentration of
retained particles, i.e. the particle retention and the conse-
quent consumption of vacant throats do not affect the total
number of vacancies:
(x, t) ≪ h(x, t). (31)
This condition can be satisfied if either the injected con-
centration is very low, or the injection time in the test is short,
0.8 0.06
0.6
0.04
f(r )

f(r )

0.4
cannot substitute high precision geometric simulation.
0.02
After validation of the Descartes’ method, the PSD is
0.2
determined from the grain size distribution for conditions
of laboratory experiments. The size distribution of the glass 0 0
beads adopted as the porous media is determined using a 0 10 20 30 40 50 60 70
Malvern Mastersizer (2000). The measured grain size distribu- r , r (µm)
tion and the calculated PSD by Monte Carlo method are given
Pore Radius from Monte Carlo SimulaƟon Grain Radius
in Fig. 8. The PSD can be approximated by lognormal distri-
bution with the mean of 3.52 ␮m and the standard deviation Fig. 8 – Determination of pore size distribution from grain
of 0.53 ␮m. The obtained PSD curve will be used further in size distribution using Monte Carlo simulation.
72 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77
or the filtration coefficient is small resulting in low retention
concentration even for high suspension concentration.
Let us examine whether the inequality Eq. (31) is fulfilled
under the conditions of the performed tests. The concen-
tration of vacant (thin) pores is h = pNp /V, where Np is the
number of pores, V is the volume of the column and p is the
fraction of vacant pores. The number of grains for spherical
grain packing is Ng = (1 − )V/(4rg3 /3). For assumed close and
dense packing in the column, each pore is formed by four
grains and each grain is surrounded by eight pores. Hence,
Np = 2Ng :
(1 − )V
Np = , (32)
(2rg3 /3)
The suspended concentration is smaller than the injected
one c(x,t) < c0 , the retained particle concentration is limited
by (x, t) = c(x, t)Ut ≤ c0 Ut. Let us define the dimensionless
time tD = tU/L calculated in pore volumes of injected sus-
pension. Recalculate the injected concentration c0 measured
by the number of particles per volume of suspension into vol-
umetric concentration measured in ppm multiplying by the
colloid particle volume (4rs3 × 106 /3). Accounting for (32), the
condition (31) is transformed to the following form:
c0 tD L
r 3
g
× 10−6 ≪ 1, (33)
2p(1 − ) rs
The product c0 tD in the numerator of (33) supports the
above statement that low retention case corresponds to either
small injected concentration, or short injection time, or small
,0 ,l/L l C (rsi )
filtration coefficient.
Substituting the values of parameters, which are typical
for the coreflooding tests described in Section 4 of the present
work (c0 = 8 ppm,  = 0.4, the measured particle size rs = 5.0 ␮m
and the grain size rg = 40 ␮m, the fraction of thinner pores
p = 0.3, dimensionless filtration coefficient L = 3.0 as calcu-
lated from experimental data and tD = 7 PVI) yields the value
of left hand side 0.093, i.e. the condition (33) is fulfilled. Yet,
the breakthrough concentrations in 5 tests stabilize after 4–5
PVI. Also, the maximum retention concentration at the inlet
Fig. 9 – Breakthrough concentration curves for five sized particles.
exponentially decreases along the core. It allows concluding
that the blocking is negligible under conditions of the con-
ducted tests (Section 4).
6.2. Results of the coreflood tests
Let us determine PSD for the glass bead packed porous column
from breakthrough concentrations as obtained from the injec-
tion of different size suspended particles (Section 4.2) using
the analytical model (29)
CL (rs ) L
= (1 − fns (rs )) exp −fns (rs ) , (34)
C0 (rs ) l
Here the fractional flux via thick pores fa + fnl is substituted by
1 − fns .
Consider n challenge tests using particles with radii rs1 ,
rs2 , . . ., rsn . The inlet and breakthrough concentrations C0 (rsi )
and CL (rsi ) are known from experimental data for each test.
Assume lognormal distribution for pore sizes as determined
by mean pore radius <rp > and standard deviation  0 . It allows
expressing the fractional flow via thin pores fns via <rp > and  0
using formula (11). Finally, n tests yield system of n transcen-
dental equations (34) for three unknowns – mean pore radius,
standard deviation and dimensionless correlation length l/L.
The system is solved by the least square method, i.e. the
solution minimises the total quadratic deviation between the
experimental data and those predicted by modelling
n  2
L CL (r )

min (fa (rsi ) + fnl (rsi )) exp −fns (rsi ) − 0 si . (35)
rp
i=1
The results of minimisation (35) are shown in Figs. 9–11
and in Table 2: mean pore size is 3.13 ␮m, distance between
chambers l = 0.25 mm. The solid piecewise lines 1,2, . . ., 5 in
Fig. 9 correspond to breakthrough curves during 5 injections of
mono sized particles with different sizes presented in Table 1.
The concentrations are constant over the time intervals where
the effluent samples were taken. The stabilised relative break-
through values CL /C0 for type 1 particles almost coincide with
 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77 73

1
model
Table 2 – Summary of pore size distribution results from
0.9 1 exp the model and Monte Carlo simulation.

0.8
MC Modelling Monte Carlo
2 data simulation
0.7
Mean radius (␮m) 3.13 3.52
0.6 Standard deviation (␮m) 2.76 0.53
Average relative error 1.94e−4 0.11
c L/c 0

0.5
3
0.4

0.3 calculations presented in Fig. 10 was taken from the results

of matching (35) and rp0 = <rp >. The stabilised values of the
0.2
breakthrough concentrations in all tests were adopted as CL
0.1 values.
4 5
The solution (<rp >,  0 ) of optimisation problem (35) is pre-
0
0 2 4 6 8 10 12 sented in Table 2 for circular pore shape (first column) and
rs for triangular pore shape (second column). Fig. 11(a) and (b)
presents the cumulative PSD curve as calculated from the
Fig. 10 – Normalized breakthrough concentration versus
solution (<rp >,  0 ) for triangular pore shapes.
jamming ratio for five size particles.
Dashed horizontal straight lines in Fig. 9 correspond to
line CL /C0 = 0.93, i.e. it is expected that rs1 slightly exceeds min- stabilised breakthrough concentrations as predicted by ana-
imum pore throat size. The relative breakthrough values CL /C0 lytical model (34) for solution (<rp >,  0 ) for five particle sizes
for types 4 and 5 particles almost coincide with line CL /C0 = 0, for the triangular pore shape using optimisation procedure
i.e. it is expected that maximum pore size is less than rs4 and (35), showing good agreement between the experimental and
rs5 . Fig. 10 presents the normalized breakthrough concentra- predicted values of stabilised breakthrough concentrations.
tion versus the jamming ratio j = rs /rp0 , where rp0 is a typical Fig. 10 shows that all 5 experimental points match very well
pore throat size. In the sake of forthcoming analysis, rp0 for with the modelling curve as obtained by formula (34).
Dashed curve in Fig. 10 corresponds to the normalized
breakthrough curve as obtained by the analytical model (34)
(a) 1 0.25
using the PSD calculated by the Monte Carlo simulations
according to Descartes theorem (30). Some deviation is evi-
0.8 0.2 dent. Despite the Descartes model predicts the averaged pore
throat size with a good accuracy, the difference in standard
deviations is significant (Table 2).
0.6 0.15
The test would be more complete with the injection of par-
F(rp)

fns (rs)

ticles smaller than the minimum pore radius resulting in full

0.4 0.1 particle recovery. The minimum pore radius is equal, accord-
ing to first column of Table 2, to 0.37 ␮m. Yet, such small
particles were unavailable. Therefore, the test with the injec-
0.2 0.05
tion of particles smaller than the minimum pore radius was
performed for larger particles and glass beads under the same
0 0 salinity, pH, etc. The glass bead size range was 20–31.5 ␮m cor-
0 2 4 6 8 10 12
rp, rs responding to 2.46–4.58 ␮m range of pore sizes; particle size
was 0.886 ␮m. The recovery (normalized breakthrough con-
(b) 1 1
centration) was 0.98.
Besides the cumulative PSD plot shown in Fig. 11 as a
0.8 0.8 continuous curve, dashed curve shows the cumulative dis-
tribution of flow rates via different size pores fns , (11). Since
the rate via the pore is proportional to rp4 , dashed curve lays
0.6 0.6 significantly to the right from the continuous curve. It means
that even for large pores, where the fraction of smaller pores
F(rp)

fns (rs)

is large, the flux via the smaller pores is relatively small.

0.4 0.4
For large particles, where fns has order of magnitude of
unity, the first term in right hand side of Eq. (34) is small.
Since L/l ≫ 1, the second term is also small. So, the effluent
0.2 0.2
concentration is negligibly small if compared with the inlet
concentration. Therefore, the proposed method allows deter-
0 0 mining PSD only up to a maximum value rpmax , which is
0 20 40 60 80 100 120
rp, rs
determined by the accuracy of concentration measurements.
Let us calculate the maximum pore radius, which the method
Fig. 11 – Cumulative pore size F(rp ) (continuous curve) and allows to determine.
pore flux fns (rs ) (dashed curve) distributions obtained by For porous media under consideration, L/l = 200. Consider
matching with the model assuming the triangular pores (a) three cases of the particle counter accuracy – 0.03, 0.02 and
full interval of pore throat sizes; (b) zoom. 0.01. The maximum value 0.03 corresponds to the particle
74 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77
0.2 0.08
0.15 0.06
c L/c 0
0.1 0.04
fns
0.05 0.02
0 0
0 1 2 3 4 5 6
rs
Fig. 12 – Maximum pore size that can be determined from
challenge coreflood test.
counter Pamas (see website). The minimum value 0.01 corre-
sponds to Malvern Mastersizer (Mastersizer, 2000). These are
the values for left hand side of Eq. (34). The modelling curve
CL /C0 (34) allows determining the corresponding particle sizes,
which are 4.39, 4.53 and 4.75 ␮m, respectively. Fig. 12 presents
the plot of the fractional flow via smaller pores fns (rs ) versus
particle radius and the normalized breakthrough concentra-
tion CL /C0 versus the particle radius. The graph fns versus rs
in Fig. 12 allows calculating the fractional flow values fns for
the above the accuracy values, which are 0.0174, 0.0195 and
0.0229, respectively.
The medium pore size, as determined above by tuning pro-
cedure (35) is 3.13 ␮m (Table 2). It corresponds to 50% of pores
smaller than 3.13 ␮m. Formula (11) and Fig. 12 allows calcu-
lating that only 0.0048 of the total flux moves via the smaller
pores. The product in exponent of (34) is −0.96, the normal-
ized breakthrough concentration is 0.383, i.e. the value which
can be measured with high accuracy. For points 1, 2 and 3, the
fractional flow fns is equal 0.0004, 0.0013 and 0.0037; despite
large value of L/l = 200, the products in exponent of (34) are
0.08, 0.26 and 0.74. The corresponding values of normalized
breakthrough concentrations are 0.93, 0.79 and 0.47, i.e. quite
“measurable”. The distribution curve for pore throat sizes is
located to the left of the distribution curves for fluxes via these
pores. So, even for large pore radii rp , the corresponding fns is
small enough for the corresponding particle to be detected
at the core outlet. It explains why the challenge coreflood test
allows determining PSD for pores which are significantly larger
than the mean value.
7. Discussions
The proposed method of PSD determining involves matching
the analytical model (29) for deep bed filtration with exper-
imental breakthrough concentrations, as obtained from size
exclusion deep bed filtration tests, and uses PSD as calculated
from grain sizes to validate the proposed method.
The analytical model uses geometrical description of the
porous media as a set of parallel capillary intercalated by mix-
ing chambers which is widely used for suspension transport
modelling. The main assumption of the model – low reten-
tion filtration – can be fulfilled in laboratory challenge tests
on colloid transport in porous media. Yet, the model (19)–(21)
presents a simplified description of the pore space geome-
try. The network models better represent the pore and throat
shapes. The percolation and effective media models better
represent the topology of the pore network (Bedrikovetsky,
1993; Selyakov and Kadet, 1996; Panfilov et al., 2008). The
low retention assumption limits the injection period; the
numerical population balance models can simulate long term
injection processes. The above shows the ways of the improve-
ment of the proposed method.
The key point of the proposed laboratory method is cre-
ation of the environment that provides the grain-particle
repulsion. It excludes various particle capture mechanisms
driven by the particle–grain attraction like attachment, bridg-
ing, internal filter cake and avoids cumbersome and expensive
physico-chemical characterisation of grain and particle sur-
faces, interactions, geometry of retained matter, etc. The only
particle capture mechanism under the repulsion is size exclu-
sion. The straining mechanism is characterised mainly by pore
and particle size distributions, which suits to the goal of this
research.
Several filtration tests with latex particles in packed glass
bead medium with simplified geometry show that low salinity
and high pH of water promote the grain-particle repulsion.
Therefore, the same glass beads, particle and water have been
used further for column experiments in 3D porous media.
The preliminary repulsion tests have been performed for
latex colloids and glass beads. More research is required
to determine the injected water composition providing the
colloid–grain repulsion for the natural reservoir rocks (sand-
stones with clays, dolomites, carbonates, etc.), filters and
membranes.
The breakthrough concentration curves for five tests, as
plotted in Fig. 9, stabilise after injecting 5 pore volumes at the
most. The experiments have been conducted with a minimum
of 250 mL (7 pore volumes) of injected fluid, i.e. the stabilised
values have been reached in all tests.
The restriction of the colloid concentration measurement
is that the minimum volume required for an accurate test is
50 mL. Therefore, the rig used makes it possible to determine
only a stepwise breakthrough curve – the concentration values
are constant over the time intervals where a single sample
was taken. Yet, this does not affect the quality of the results
because only the stabilised outlet concentrations are used for
PSD determination, see (34).
The Descartes’ model for pore sizes, used for validation
of the proposed method for PSD determination, is based on
the assumption of dense and close packing, where four grains
form one pore body. This assumption effectively defines the
radius of a pore throat as the maximum inscribed radius of
three mutually tangent grains in the plane formed by the cen-
tres of these spherical grains. If the assumption of densely
packed media is incorrect, more than three grains may con-
struct a pore throat and hence the calculation of PSD using
Descartes’ theorem becomes inaccurate. In the case of close
packing not occurring, high permeability streaks may trans-
port the colloidal suspension through the column without
exposure to the matrix pore throats. If the assumption of
spherical grains is inaccurate, the shape of a pore throat
becomes more complex and is less likely to be defined by
a curvilinear triangle. In order to ensure that the matrix
is closely and densely packed, an extensive sieving process
is performed to narrow the size distribution. It is reasoned
that similar sized particles will reduce the occurrence of
unconstrained matrix particles. Furthermore, the experimen-
tal procedure includes the wet packing of the column, an
extended settling period and sonification of the glass beads to
 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77
Fig. 13 – Visual observation of glass bead packing matrix.
increase the likelihood that the matrix is closely and densely
packed. The assumption of sphericity is confirmed through
analysis of the packing matrix under an Olympus BX51 opti-
cal microscope. From observation of the glass beads under the
microscope, it is evident that any irregularities of the glass
beads are within a tolerable range.
Despite wet packing have been performed in the tests and
the column was immersed into a sonic bath, the ideally dense
packing have not been achieved. Fig. 13 shows the cross sec-
tion of the column under SEM, where the complete contact
between grains in some places does not occur. Therefore,
despite the results of the Descartes’ modelling are in a good
agreement with that of the precise geometrical 3D modelling
of the porous space (Fig. 7), the latter is more preferable for
validation of the proposed PSD determination method.
Another limitation of the Monte Carlo simulation employed
in this work is that the grain size distribution is stepwise
(Fig. 8). A more accurate size distribution for the glass beads
could not be obtained due to the constraints of the particle
sizer data (Malvern Mastersizer, 2000).
The ratio of total number of injected colloids to the pore
vacancies is kept below 0.1 throughout the tests (see (33)) to
ensure that there is no competition between colloidal particles
for pore throats. It shows that the assumption of low retention
filtration is fulfilled during the injection process in the carried
out tests. Such a low ratio also prevents the occurrence of con-
centration decline due to accumulation of particles around the
blocked pore throats.
The adjustment of PSD using 5 experimental points shows
that the experimental points are tuned well by the predicted
curve as calculated for pores with triangular shape – five points
are located exactly on the dashed curve in Fig. 10. Yet, very
limited sizes for particles and glass beads, available on the
market, narrowed the amount of experimental data. Just five
points have been recovered from the flow tests, where just
three points are clearly located between the minimum and
maximum pore sizes. More points must be utilised for reliable
determination of three model parameters (mean radius and
standard deviation of PSD, correlation length).
The deviation between the Monte Carlo and the model gen-
erated curve, presented in Fig. 10, can be explained by the
existence of four-grain patterns in the media. Despite a com-
pact structure of grains in the bed, some inter-grain spaces are
75
formed by four neighbouring grains instead of three. However,
Fig. 13 presents the photo of glass beads under SEM micro-
scope; it shows that the concentration of four-grain patterns is
significantly lower than that of three-grains. The Monte Carlo
simulation based on Descartes’ theorem (Section 5) assumes
that the porous space consists of three-grain patterns only,
which leads to a deviation of PSD from the real media.
Although the validation of the proposed PSD calculation
method was successful, more extensive research is required to
validate the proposed method and recommend it for practical
applications in porous media technologies. First, the proposed
Monte Carlo method for calculation of the PSD from the grain
size distribution must be improved and further validated. The
method assumes the triangular throat shape only. Yet, there
may be some caverns which cannot be plugged by particles
in hexagonal mode due to particle sizes and quantities, avail-
able from the given set of grains. More complex throat shapes,
formed by four and more particles, may be considered. Since
the data on known grain and PSD for the same porous media
are almost not available in the literature, the laboratory stud-
ies have to be carried out for validation of the method. Second,
the number of “experimental points”, as obtained from our
lab tests is very limited. The good match between the PSD
curve, as predicted by analytical modelling based on the test
data from suspension flow, with that calculated from the grain
size distribution, do not validate the method completely – on
account of the limited number of points. Since very restricted
sizes of the used latex particles are available in the market, it
would be a solution to find other sources for particles.
The proposed model (Eqs. (20)–(23)) describes long time
filtering where the retained particle accumulation affect the
filtration coefficient, while in the present paper the short
time injections with constant filtration coefficient are dis-
cussed. Further research may include treatment of the long
term data using optimisation procedure for the general model
(Eqs. (20)–(23)). This option is particularly attractive since the
time variation of pressure drop across the core can be used
as additional information to characterise the filtration system
(Bedrikovetsky et al., 2001; Grenier et al., 2008; Rebai et al.,
2010; Daniel et al., 2011).
The particle pulse injection could be considered for the use
instead of the continuous injection. It would shorten the test
period and decrease the particle consumption. Yet, the high
particle concentration during a short time pulse injection may
be restricted by the low retention assumption, besides the
higher accuracy of concentration measurement is required
if compared with continuous injection. The pulse option of
laboratory tests and its combination with already applied con-
tinuous injection would also demand an additional research.
8. Summary and conclusions
The challenge test on colloidal flow with different size parti-
cles under the particle–grain repulsion allows for determining
the pore throat size distribution in long cores. Creation of the
particle–grain repulsion environment allows avoiding the par-
ticle capture by attachment and expensive and cumbersome
characterisation of particle–grain interaction in order to model
the particle attachment. The particle–grain repulsion leaves
size exclusion as a single particle retention mechanism. This
mechanism is described by pore space geometry and particle
sizes only. The fact that size exclusion deep bed filtration is
determined by pore space only is in accordance with the goal
76 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77
of determining the PSD from transport of different size colloids
through porous media.
The particle–grain repulsion can be achieved by the correct
choice of the material and coating of the particles and of water
composition (salinity, pH).
The geometrical model of a porous space as a set of parallel
capillaries intercalated by mixing chambers, which is widely
used for colloidal flows in porous materials, is fully deter-
mined by PSD and the inter-chamber distance. Therefore, this
model was chosen for determining PSD.
The colloidal flow test includes injection of different
size particles with measurements of the inlet and sta-
bilised breakthrough concentrations. The analytical model for
size exclusion deep bed filtration relates the inlet and sta-
bilised breakthrough concentrations with PSD. Finally, PSD
is determined from the inlet and stabilised breakthrough
concentrations by least square method using the analytical
model.
The laboratory tests performed in packed glass bead col-
umn with 5 types of particles with different sizes show that
the determined PSD curve tunes well 5 experimental points.
The curve is also in a good agreement with that obtained
by the Descartes’ method. The analytical model for triangu-
lar shape capillary tunes the experimental points and agrees
with the Descartes’ method better than that for circular pore
shape.
The proposed method needs more substantial validation.
More experimental points for particle radii between the max-
imum and minimum pore sizes must be obtained. It is
preferable to use high accuracy geometric simulation of the
packed column for validation of the results.
Acknowledgements
Many thanks are due to Prof. Y. Yortsos (University of Southern
California), Prof. A. Shapiro (Technical University of Denmark)
and Prof. P. Currie (Delft University of Technology) for long-
term cooperation in suspension flow studies. Prof. A. Shapiro
critically revised the text and provided valuable comments.
Special thanks go to Dr. A. Badalyan for his helpful suggestions
and comments. The authors are grateful to A. Zeinijahromi
for calculation of DLVO forces. Fruitful discussions with T.
Rodrigues and I. Abbasy (Santos Pty Ltd., Australia) are grate-
fully acknowledged.
Financial supports from the Australian Research Council
(ARC) Discovery Project 1094299, Linkage Project 100100613
and Santos Pty Ltd are gratefully acknowledged.
References
Aimar, P., Meireles, M., Sanchez, V., 1990. A contribution to the
translation of retention curves into pore size distributions for
sieving membranes. J. Membr. Sci. 54, 321–338.
Al-Raoush, R., Thompson, K., Willson, C.S., 2003. Comparison of
network generation techniques for unconsolidated porous
media. Soil Sci. Soc. Am. J. 67, 1687–1700.
Amix, B., Bass, R., Whiting, A., 1964. Applied Reservoir
Engineering. McGraw Hill Book Co, New York.
Bedrikovetsky, P., 1993. Mathematical Theory of Oil and Gas
Recovery: with Applications to ex-USSR Oil and Gas Fields.
Kluwer Academic, London.
Bedrikovetsky, P.G., Marchesin, D., Checaira, F., Serra, A.L.,
Resende, E., 2001. Characterization of deep bed filtration
system from laboratory pressure drop measurements. J.
Petrol. Sci. Eng. 64, 167–177.
Bedrikovetsky, P., 2008. Upscaling of stochastic micro model for
suspension transport in porous media. Transport Porous Med.
75, 335–369.
Bergendahl, J., Grasso, D., 2000. Prediction of colloid detachment
in model porous media: hydrodynamics. Chem. Eng. Sci. 55,
1523–1532.
Biggs, M.J., Humby, S.J., Buts, A., Tüzün, U., 2003. Explicit
numerical simulation of suspension flow with deposition in
porous media: influence of local flow field variation on
deposition processes predicted by trajectory methods. Chem.
Eng. Sci. 58, 1271–1288.
Binder, K., Heermann, D.W., 1988. Monte Carlo Simulation in
Statistical Physics: an Introduction. Springer-Verlag, Berlin.
Brakel, J.V., Modry, S., Svata, M., 1981. Mercury porosimetry: state
of the art. Powder Technol. 29, 1–12.
Cortis, A., Harter, T., Hou, L., Atwill, E.R., Packman, A.I., Green,
P.G., 2006. Transport of Cryptosporidium parvum in porous
media: long-term elution experiments and continuous time
random walk filtration modeling. Water Resour. Res. 42,
W12S13.
Daniel, R.C., Billing, J.M., Russell, R.L., Shimskey, R.W., Smith,
H.D., Peterson, R.A., 2011. Integrated pore blockage-cake
filtration model for crossflow filtration. Chem. Eng. Res. Des.
89, 1094–1103.
Descartes, R., 1901. Oeuvres de Descartes par Adam et Tannery,
vol. 4, Paris.
Dullien, F.A.L., 1992. Porous Media: Fluid Transport and Pore
Structure, 2nd ed. Academic Press, New York.
Fan, L., Reti, H., Wang, W., Lu, Z., Yang, Z., 2008. Application of
computational fluid dynamic to model the hydraulic
performance of subsurface flow wetlands. J. Environ. Sci. 20,
1415–1422.
Frising, T., Thomas, D., Contal, P., Bemer, D., Leclerk, D., 2003.
Influence of filter fiber size distribution on filter efficiency
calculations. Chem. Eng. Res. Des. (former Trans IChemE) 81A,
1179–1184.
Gitis, V., Rubinstein, I., Livshits, M., Ziskind, G., 2010. Deep-bed
filtration model with multistage deposition kinetics. Chem.
Eng. J. 163, 78–85.
Grenier, A., Meireles, M., Aimar, P., Carvin, P., 2008. Analysing flux
decline in dead-end filtration. Chem. Eng. Res. Des. 86,
1281–1293.
Herzig, J.P., Leclerc, D.M., Le Goff, P., 1970. Flow of suspensions
through porous media – application to deep filtration. Ind.
Eng. Chem. 62, 8–35.
Ilina, T., Panfilov, M., Buès, M., Panfilova, I., 2008. A pseudo
two-phase model for colloid facilitated transport in porous
media. Transport Porous Med. 71, 311–329.
Khilar, K., Fogler, S., 1998. Migration of Fines in Porous Media.
Kluwer Academic Publishers, Dordrecht.
Kuznar, Z.A., Elimelech, M., 2007. Direct microscopic observation
of particle deposition in porous media: role of the secondary
energy minimum. Colloid Surf. A 294, 156–162.
Landau, L.D., Lifshitz, E.M., 1980. Statistical Physics. Pergamon
Press, Oxford.
Landau, L.D., Lifshitz, E.M., 1987. Fluid Mechanics, 2nd ed.
Pergamon Press, Oxford.
Lin, H.-K., Pryadko, L.P., Walker, S., Zandi, R., 2009. Attachment
and detachment rate distributions in deep-bed filtration.
Phys. Rev. E 79, 046321.
Mason, G., Morrow, N.R., 1991. Capillary behavior of a perfectly
wetting liquid in irregular triangular tubes. J. Colloid Interface
Sci. 141, 262–274.
Mastersizer, 2000. http://www.malvern.com/LabEng/products/
Mastersizer/MS2000/mastersizer2000.htm (23.02.11).
Mays, D., Hunt, J., 2005. Hydrodynamic aspects of particle
clogging in porous media. J. Environ. Sci. Technol. 39, 577–584.
Nolan, G.T., Kavanagh, P.E., 1994. The size distribution of
interstices in random packings of spheres. Powder Technol.
78, 231–238.
Panfilov, M., Stepanyants, Y., Panfilova, I., 2008. Mechanisms of
particle transport acceleration in porous media. Transport
Porous Med. 74, 49–71.
 chemical engineering research and design 9 0 ( 2 0 1 2 ) 63–77
Patzek, T.W., Silin, D.B., 2001. Shape factor and hydraulic
conductance in noncircular capillaries. I. One-phase creeping
flow. J. Colloid Interface Sci. 236, 295–304.
Payatakes, A.C., Tien, C., Turian, R.M., 1973. A new model for
granular porous media. I. Model formulation. AIChE J. 19,
58–76.
Payatakes, A.S., Rajagopalan, R., Tien, C., 1974. Application of
porous medium models to the study of deep bed filtration.
Can. J. Chem. Eng. 52, 722–731.
Pedoe, D., 1967. On a theorem in geometry. Am. Math. Mon. 74,
627–640.
Purchas, D.B., Sutherland, K., 2002. Handbook of Filter Media, 2nd
ed. Elsevier Advanced Technology, Oxford.
Rebai, M., Prat, M., Meireles, M., Schmitz, P., Baclet, R., 2010.
Clogging modeling in pleated filters for gas filtration. Chem.
Eng. Res. Des. 88, 476–486.
Reboul, N., Vincens, E., Cambou, B., 2008. A statistical analysis of
void size distribution in a simulated narrowly graded packing
of spheres. Granul. Matter 10, 457–468.
Rideal, G., 2007. Physical measurement of pores by glass bead
challenge testing. Filtration 7, 132–137.
Rideal, G., 2009. The filtration society: the importance of testing
and standards. Filtr. Separat. 46, 28–30.
Rong, F., Guo, Z., Chai, Z., Shi, B., 2010. A lattice Boltzmann model
for axisymmetric thermal flows through porous media. Int. J.
Heat Mass Transfer. 53, 5519–5527.
Rouault, Y., Assouline, S., 1998. A probabilistic approach towards
modeling the relationships between particle and pore size
distributions: the multicomponent packed sphere case.
Powder Technol. 96, 33–41.
Santos, A., Bedrikovetsky, P., 2006. A stochastic model for
particulate suspension flow in porous media. Transport
Porous Med. 62, 23–53.
77
Selyakov, V.I., Kadet, V.V., 1996. Percolation Models for Transport
in Porous Media with Applications to Reservoir Engineering.
Kluwer Academic Publishers, Dordrecht.
Shapiro, A.A., 2007. Elliptic equation for random walks.
Application to transport in microporous media. Physica A 375,
81–96.
Shapiro, A.A., Wesselingh, J.A., 2008. Gas transport in tight
porous media. Gas kinetic approach. Chem. Eng. J. 142,
14–22.
Sharma, M.M., Yortsos, Y.C., 1987a. Transport of particulate
suspensions in porous media: model formulation. AIChE J. 33,
1636–1643.
Sharma, M.M., Yortsos, Y.C., 1987b. Network model for deep bed
filtration processes. AIChE J. 33, 1644–1653.
Sharma, M.M., Yortsos, Y.C., 1987c. Fines migration in porous
media. AIChE J. 33, 1654–1662.
Tang, G.H., Ye, P.X., Tao, W.Q., 2010. Pressure-driven and
electroosmotic non-Newtonian flows through microporous
media via lattice Boltzmann method. J. Non-Newton. Fluid
Mech. 165, 1536–1542.
Tufenkji, N., Elimelech, M., 2004. Deviation from the classical
colloid filtration theory in the presence of repulsive DLVO
interactions. Langmuir 20, 10818–10828.
Tufenkji, N., Elimelech, M., 2005. Breakdown of colloid filtration
theory: role of the secondary energy minimum and surface
charge heterogeneities. Langmuir 21,
841–852.
Yortsos, Y.C., 1999. Probing pore structures by sorption isotherms
and mercury porosimetry. In: Wong, P.-z. (Ed.), Methods in the
Physics of Porous Media. Academic Press, New York.
Yuan, H., Shapiro, A., 2010. A mathematical model for
non-monotonic deposition profiles in deep bed filtration
systems. Chem. Eng. J. 162, 974–988.