Microscale Simulation of Nanoparticles Transport in Porous Media for

Groundwater Remediation
Francesca Messina*, Matteo Icardi*, Daniele Marchisio**, Rajandrea Sethi*

* Politecnico di Torino DIATI C.so Duca degli Abruzzi 24, 10129 - Torino, Italy
** Politecnico di Torino DISAT C.so Duca degli Abruzzi 24, 10129 - Torino, Italy
reagent for the remediation of contaminated

Abstract
Micro and Nanoscale zerovalent iron (MZVI and
NZVI) is one of the most promising reagent for
the remediation of contaminated groundwater;
these particles, in fact, can efficiently degrade,
through

redox

reactions,

recalcitrant

and

carcinogenic compounds.
The aim of this study is to simulate at the
microscale the transport of iron nanoparticles,
their interaction with the porous media and their
deposition on the aquifer material.
The simulations have been carried out with a
Langrangian

approach

implemented

in

COMSOL Multiphysics 4.2a. The model under

study includes the relevant forces acting on the
single particles such as drag, Brownian, gravity,
Van der Waals and electric double layer force.
The simulation results can deliver, thanks to this
microscale description, an estimation of the
attachement efficiency that can be used for
macroscale simulations and compared with the
relationships obtained by the clean bed filtration
approach.

groundwater [15].
These particles can degrade, through redox
reactions,

recalcitrant

and

carcinogenic

compounds such as perchloroethylene and

trichloroethylene.
Due to his very high specific surface NZVI is
characterized by a high reactivity. Furthermore it
can be injected in the subsurface more easily if
compared to the emplacement of granular iron
commonly used in permeable reactive barriers
[3] and it can be used directly to treat the source
of contamination itself and not only on the plume
(the dissolve part of contamination).
The iron nanoparticles are normally injected into
the soil as an aqueous slurry, which has to come
into contact with the source of contamination or
reach the polluted groundwater [10].
The main issue related to the field application of
this technology is the reduced mobility of NZVI.
In order to enhance the mobility of iron
nanoparticles on the subsurface, it is important to
(i) prevent the formation of large aggregates that
tend to be easily filtered and (ii) reduce the

Keywords: Zerovalent iron, microscale, aquifer

attachment of nanoparticles to the soil grains. To

remediation, Lagrangian particle tracing.

date, the best strategy to achieve these goals has

been to modify the surface of the iron

1. Introduction
Micro and nanoparticles of zerovalent iron
(MZVI and NZVI) are one of the most promising

nanoparticles using polymers or surfactants [10].

A number of other studies [12, 13] have
investigated the MZVI and NZVI mobility using
transport tests in order to develop a powerful

Excerpt from the Proceedings of the 2012 COMSOL Conference in Milan

2. Methods

tools for the design and the implementation of

full scale zerovalent iron applications [13].
The aim of this study is to simulate the transport
of iron nanoparticles and their interaction with
the porous media at the microscale.
The deposition of iron particles on the porous
media is an important mechanism controlling the
mobility of colloids in aquifer systems and the
effectiveness of the technology. A microscale
domain (few micron) was reconstructed through
SEM images of a sandy aquifer material and the
flow field was calculated using Navier-Stokes
equations over a wide range of flow velocities,
but always in laminar conditions.
In porous media the dynamic of the particles

2.1 Geometry and flow field

In a granular medium, the flow field around each
grain

is

influenced

by

the

neighbouring

collectors and so studying the flow field around

an

isolated

sphere

could

be

not

much

representative. Therefore the first approach used

to construct a computational domain is the
Happels geometry [6]. The Happels sphere-incell model (Figure 1), in fact, simulates the flow
interaction due to the presence of other particles
through solve the fluid flow in a fluid shell
around each solid grain considering that its
affected by the neighbouring grains.

inside of a pore is governed mainly by drag,

gravity and Brownian forces, but close to the
sand surface also the electric double layer (EDL)
and the Van der Waals forces (VdW) come into
play and become prevailing. In order to simulate
the trajectories of the MZVI and NZVI particles
a Lagrangian approach was implemented by
using the COMSOL Particle Tracing for Fluid
Flow module.
The aim is to derive a correlation between the
parameters characterizing the iron particles, the
sand grains and the flow field and the capacity of
MZVI and NZVI to cover long distance without
be captured from the porous media grains.

Figure 1: schematic description of Happels

sphere-in-cell flow model for a homogeneous porous media
(adapted from [5]).

This correlation would be useful for future

In the Happels model, the porous medium is

application in field and for the project of a real

treated as an assemblage of identical, spherical

reclamation activity.

collectors, each of which is enveloped by a shell

of fluid. The thickness of the shell, b, is
determined so that the overall porosity n of the

Excerpt from the Proceedings of the 2012 COMSOL Conference in Milan

porous medium is maintained for the single

been investigated (Figure 5) and an inlet velocity

collector of radius ac [4]:

equal to U, the approach velocity,

1

b ac (1 n) 3

eq.1

has been

imposed.

In this case an analytic solution for the flow field

exists and it is reported in literature (Elimelech
M. [4], Chi Tien, B. V. Ramarao [9]) and
represented in Figure 2.

Figure 3: Original SEM photograph of a sandy porous media.

Figure 2: Flow field in Happels model.

Grain radius ac = 250 m, porosity n = 0.30, approach
velocity U = 10-6 m/s.

The second step consists in considering a more

realistic 2D domain. Starting from some sands
samples, scanning electron microscope (SEM)
was used in order to obtain some digital image of
them (Figure 3).
Then,

through

CAD

manipulation,

computational grids have been obtained and used

in COMSOL Multiphysics 4.2a (Figure 4).
The domain has dimension of 2.69 10-3 x 2.23
10-3 m.
The flow field was found by solving the NavierStokes equations in laminar conditions, using the
Laminar

flow

interface

of

COMSOL

Multiphysics 4.2a.

Figure 4: Definitive geometry used in COMSOL

Multiphysics 4.2a.

The grains boundaries are all set as wall, so that

the fluid velocity on their surface is zero.
On the rest of the boundary sites a symmetry
boundary

condition,

that

prescribes

no

penetration and vanishing shear stresses, was

used.

With a parametric sweep, a range of approach

velocity U between 10-6 m/s and 10-3 m/s has

Excerpt from the Proceedings of the 2012 COMSOL Conference in Milan

FEDL the Electric Double Layer force, FVdW the

Van der Waals force and FB the Brownian

force.
Their complete mathematical formulations are
reported in Table 2. In Table 1 there are the
constants and parameters numeric value used in
the simulations.
The gravity force, is constant in all domain. Its
direction is always vertical, perpendicular to the
earth surface.
The drag force depends from the flow field, in
general it is bigger in the centre of each pore and
becomes

less

significant

near

the

grains

boundaries where the velocity of the flow fields

is zero.
Figure 5 : Flow field in the 2D porous medium
domain, a) approach velocity U = 10-6 m/s, b) approach
velocity U = 1 m/s.

Water density

998 kg/m3

Iron particles density

5500 kg/m3

Dynamic viscosity

110-3 Pas

0r

7.083210-10F/m

k -1

3.2108 1/m

Hamaker constant

1.5110-20 J

particles trajectories in the flow field and their

Gravity acceleration

9.81 m/s2

interaction with the sand grains at a microscale

Happels collector radius

ac

250 m

Iron particles radius

ap

40 nm,400 nm,4
m

Boltzmann constant

1.3806510-23 J/K

Absolute temperature

298 K

Collector zeta potential

-7 mV

Iron particle zeta

potential

-58 mV

2.2 Particle tracing

Particles transport was simulated by using a
Lagrangian approach exploiting the Particle
Tracing for Fluid Flow module.
The Lagrangian approach was chosen because it
gives the possibility to describe accurately the

by implementing all the forces acting on them.

The basic equation of this approach is the
classical Newtons low:

dU
i Fi mp dt FD FG FEDL FVdW FB eq.2

U and mp are the particle velocity and

mass, FD is the drag force, FG the gravity force,

Where

Water electric
permittivity
Invers of Debye
parameter

Table 1: Constants and parameters used in the

simulations.

Excerpt from the Proceedings of the 2012 COMSOL Conference in Milan

Forces

Mathematical expression

Drag force

FD 6a p u v

Gravity force

4
FG a 3p p f g
3

Electric double layer force

p2 c2 e h p c
h
FEDL 0 r a p
2

n
21 e 2h p2 c2

Van der Waals force

Ha p ( 28h)
FVdW 2
n
6h 14h 2
FB R

Brownian force

2kT
t

The parameters (see also Table 1) are: u fluid velocity, v particle velocity, ap particle
radius, p particle density, f fluid density, g acceleration of gravity, H Hamaker
constant, h particle collector distance, average wavelength of electron oscillation (100
nm), R random normal

distribution number, k Boltzmans constant, T absolute

temperature, friction coefficient (equal to 6ap), k Debye - Huckel parameter, P iron

particle zeta potential, C collector zeta potential
Table 2: Mathematical Expression of all forces considered in the Particle Tracing for fluid flow module.

FVdW is the Van der Waals force, its an attractive

the water pH and from the water ionic strength.

force between particles and aquifer grains. It is

These two forces are active in all domain but

due to the dipole moments generated from the

significant only in a nanoscale range of distance

temporary asymmetry of the electrons cloud

between particle and sand grains, far away they

distribution around the nucleus of atoms. It is

quickly tends to zero.

always negative.

The Brownian force is a random force and so its

FEDL is the electric double layer force: the

implementation requires the generation of a

particles and the collectors (the aquifer grains)

random number. Its formulation includes also

are provided with a surface charge which is

some functions that take in consideration the

balanced by ions presented in the solution

presence of the solid walls [8]. These function

creating so an electric double layer. It depends

are described by Elimelech (1995) [5].

from both particle and grains zeta potential (a

Being the Brownian force a stochastic term, its

measured of the surface charge of them), from

explicit formula depends also on the time step

Excerpt from the Proceedings of the 2012 COMSOL Conference in Milan

used by the software in order to solve the

interception (when the streamline along which

problem.

the particle is moving crosses a sand grain),

The particle tracing equations were solved by a

gravitational sedimentation (due to the settling of

time dependent solver and the time step,

the particles when their density is higher than

automatically chosen by the software, is used in

waters),

the force formulation.

Brownian thermal movement diverge particles

The time step, however, must be greater than the

from a streamline and brings them to the grain

particle momentum relaxation time MRT = mp/,

surface) [11].

and

Brownian

diffusion

(when

where mp is the particle mass and friction

coefficient (equal to 6ap) [8].
This sets the lower limit, but it is necessary also
to consider a higher limit to assume the forces
constant in that time step.
The largest force gradients are present near the
grains surface, where the distance between grains
and NZVI particles became small and the FVdW
and FEDL could have big oscillations.
Since in COMSOL we can use only a single time
step for the whole domain, without refining it in
these regions, we were obliged to use in all
domain a very small time step in order to have a
very small particle displacement (1-2 nm) able to
feel the high gradient of forces.
This is a very critical point because the
computational

time

significantly

increases,

therefore a limited number of particles can be

simulated.

Figure 6: Different mechanisms of particle adhesion to the

grain.

Conversely, the properly called attachment

step is controlled by surface interaction forces.
According to the CFT theory, the transport step
is described by the single-collector contact
efficiency 0, while the attachment step is
defined by the attachment efficiency .
Many equations have been proposed to date for
the single collector efficiency [Rajagopalan and

2.3 Collector efficiency

The mechanisms leading to colloid deposition
can be divided into two steps: transport and
deposition. The term transport refers to the
mechanisms that bring the colloidal particles
from the bulk of the fluid to the proximity of the

Tien, 1976; Tufenkji and Elimelech, 2004]. The

most complete one [Tufenkji and Elimelech,
2004] accounts for the superposition of the
effects of hydrodynamic forces, van der Waals
interactions and gravity effects, and proposes the
following correlation equation [11]:

grain surface. These mechanisms (Figure 6) are

Excerpt from the Proceedings of the 2012 COMSOL Conference in Milan

central grains and all the released particles.

0.052
0 2.4 As1/ 3 N R0.081N Pe0.715N vdW

0.053
0.55 As N R1.675N A0.125 0.22 N R0.24 N G1.11N vdW

eq.3
where AS is a porosity-dependent parameter, NR
is the aspect ratio, NPe is the Peclet number, NvdW
is the Van der Waals number (describing the
entity of the Van der Waals interaction energy
with respect to the thermal energy of the system),
NA is the attraction number (characterizing the
effects of fluid velocity and Van der Waals
interaction on the particle deposition due to
NG

interception),

is

the

gravity

number

(representing the ratio between the Stokes

particle settling velocity to the effective velocity
of the pore water). These are non-dimensional
parameters. For more information see [14].
In eq.3 the first term represents the effects of
diffusion, the second one the contribution due to
interception, the third one the gravity effects.

3. Results

In the Happels model the particle release was

from the top semicircle.
Different simulations increasing each time the
number of particles released have been done.
For the cases with particle radius equal to 4 m
and 400 nm, simulations with 20, 100, 500, 1000
and 2000 particles were performed. For the
smaller particles (40 nm of radius) only
simulations with 20, 100, 150 and 200 particles
have been performed.
Decreasing the radius of the particles, in fact, the
particle mass decreases and the role of the
Brownian

force

increases.

It is

therefore

necessary to reduce considerably the time step of

calculation,

but

this

increases

lot

the

computational efforts, not allowing a significant

number of particles release.
In the second geometry (that taken from the
porous media image) the particles were released
from a line in the top of the domain. In this case
the particle trajectories calculation results to be

Both the geometries introduced above have been

problematic. The complexity of the geometry

used in the simulations.

and of the flow field, in fact, increases a lot the

The approach velocity U was varied in order to

computational efforts, limiting the numbers of

simulate different flow field and also the particle

particles for which is possible to plot a complete

radius ap was modified between different

trajectory.

simulations (see Table 1).

In order to calculate the collector efficiency in

All the forces listed in Table 2 has been

COMSOL we can use the global evaluation tool

implemented.

to evaluate the total number of particles released

So, with respect to the work of Tufenkji and

and the number of particles that exit from the

Elimelech, we consider also the effect of the

domain. The difference gives us the number of

electric double layer and the Brownian motion.

particles captured by the grains. This number

The collector efficiency is calculated as the ratio

divided by the total number of particles released

between the number of particles captured by the

is the collector efficiency .

Excerpt from the Proceedings of the 2012 COMSOL Conference in Milan

As a general observation, the trajectory is much

3 formulation.

more vertical for the bigger particles that are

Moreover in eq. 3 the Brownian effect is

heavier and that are so less mobile.

consider as a diffusion coefficient, in our case

Furthermore,

the

smaller

particles

have

we implement the mathematical formulation of a

trajectories more similar to the streamlines and

force, this could bring to different results.

we can also observe the Brownian vibration.

Finally,

The flow field velocity also has an influence on

approach is different from an Eulerian approach.

the particles mobility.

The most critical point of our study its the time

The simulations carried out in the Happels

step interval, the need of a very small time step

domain can be summarized in terms of collector

interval in all our domain brings to very high

efficiency, as showed in Table 3.

computational efforts.

In the realistic domain instead, it was not

For this reason we are not able to release a lot

possible to track a significant number of particles

numbers of particles and this is a limitation in

because of computational limits. So we were not

our study.

able to calculate reliable statistics.

Furthermore a smaller time step interval could

more in general,

the

Lagrangian

perhaps lead to more precise results. But the

COLLECTOR EFFICIENCY [%]

MZVI and NZVI particle radius
[m]

software and the computer at our dispositions, in

spite of its 32 Gb of RAM, couldnt allow us to
use it.
In future works a better time discretization will

Velocity
field [m/s]

ap= 410-6

ap=410-7

ap=410-8

U = 10-6

71

46

59

in our study. The possibility to use different

-5

66

19

19

the bulk of the fluid, small time step close to the

U = 10-4

42

10

-3

18

U = 10

U = 10

Table 3: Happels model result of collector efficiency.

The results (Table 3) show some differences with

respect to the literature curves of collector
efficiency (eq. 3).
This can be explained by different reasons.
First of all we have also consider the electric
double layer effect that it is not present in the eq.

be considered in order to have a better precision

criteria for particles motion (i.e. high time step in
sand grains) would also be desired.
A further development would be to implement
also particle-particle interactions.

4. Conclusions
Micro and nanoscale zerovalent iron is a
promising reagent for the remediation of
contaminated groundwater.
The aim of this study was to simulate, from a
microscale point of view, the transport of iron
nanoparticles, their interaction with the porous

Excerpt from the Proceedings of the 2012 COMSOL Conference in Milan

media and their deposition on the porous matrix

of the aquifer in order to calculate a collector
efficiency and to derive an empirical formula
useful for future field applications.
A Langrangian approach was used through the
Particle Tracing for fluid flow module of
COMSOL Multiphysics 4.2a.
This allowed us to calculate the particle
trajectories in (i) a simplified domain, the
Happels model, and (ii) in a more realistic
domain obtained from a SEM image.
The drag, Brownian, gravity, Van der Waals and
electric

double

layer

forces

have

been

implemented.
Several limitations linked to the time step
interval used from COMSOL to solve the model,
did not allow us to calculate the complete
trajectories of a lot numbers of particles.
For this reason we have no results from the
simulations carried out in the realistic geometry,
that obtained from a SEM image.
The Happels model geometry was simpler and
this allowed to conclude the simulations with a
significance numbers of particles. The collector
efficiency results show some difference from
literature, this can be due to the choice of a
Langrangian approach, to the implementation of
the electric double layer force, not considered in
other studies, and to the time step interval used.
Future developments of this work will include a
better discretization and optimization of the
model together with additional physical models
such as the particle-particle interactions.

5. References
1. Bensaid S., Marchisio D.L., Fino D. Numerical simulation of soot filtration and
combustion within diesel particulate filters Chem. Eng. Sci. 65(1), 357-363 (2010)
2. Della Vecchia E., Luna M., & Sethi R. (2009
- Transport in porous media of highly
concentrated iron micro- and nanoparticles
in

the

presence

of

xanthan

gum

Environmental Science & Technology, 43,

89428947
3. Di

Molfetta

A;

Sethi

R.,

Clamshell

excavation of a permeable reactive barrier,

Environmental Geology, 2006, ISSN: 09430105, DOI: 10.1007/s00254-006-0215-3
4. Elimelech M. - Particle deposition on ideal
collectors from dilute flowing suspensions:
Mathematical

formulation,

numerical

solution, and simulations - Sep. Technol.,

1994, vol. 4, October
5. Elimelech M. (1995) - Particle deposition
and aggregation: measurement, modelling,
and simulation - xv, 441 p. pp., ButterworthHeinemann, Oxford [England] ; Boston
6. Happel J. Viscous Flow in Multiparticle
System: Slow Motion of Fluid Relative to
Beds of Sherical Particles A.I.Ch.E
Journal, Vol.4, No.2 p.197
7. Happel J. Viscous Flow Relative to Arrays
of Cylinders A.I.Ch.E Journal, 1959, p.174
8. Ma H., Pedel J., Fife P., Johnson W.P. Hemispheres-in-Cell Geometry to Predict
Colloid Deposition in Porous Media Environmental

Science

&

2009, 43, 8573-8579

Excerpt from the Proceedings of the 2012 COMSOL Conference in Milan

Technology,

9. Tien C., B. V. Ramarao - Granular filtration

of aerosol and hydrosols 1989
10. Tiraferri A., Sethi R. - Enhanced transport of
zerovalent iron nanoparticles in saturated
porous media by guar gum - Nanopart Res

6. Acknowledgements
The authors thank Dott. Gianluca Boccardo for
the SEM image and grids, for his helps and
advices.

(2009) 11:635645 DOI 10.1007/s11051008-9405-0

11. Tosco T., Modelling the transport of ironbased colloids in saturated porous media,
Tesi di dottorato, Politecnico di Torino,
2010
12. Tosco T. , Tiraferri A., Sethi R. - Ionic
Strength

Dependent

Transport

of

Microparticles in Saturated Porous Media:

Modeling Mobilization and Immobilization
Phenomena

under

Transient

Chemical

Conditions - Environ. Sci. Technol. 2009,

43, 44254431
13. Tosco T., Sethi R. - Transport of NonNewtonian

Suspensions

of

Highly

Concentrated Micro- And Nanoscale Iron

Particles in Porous Media: A Modeling
Approach - Environ. Sci. Technol. 2010, 44,
90629068
14. Tufenkji, N., and Elimelech M. - Correlation
equation for predicting single-collector
efficiency in physicochemical filtration in
saturated porous media - Environmental
Science & Technology, 38(2), 529-536
15. Zhang W. X. - Nanoscale iron particles for
environmental remediation: An overview - J.
Nanopart. Res. 2003, 5 (3-4), 323332

Excerpt from the Proceedings of the 2012 COMSOL Conference in Milan