Theoretical and computational study of colloid transport and retention in saturated soil porous media
Date
2015
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
In this dissertation, we conducted numerical simulations of pore-scale
flows and
colloid transport in both 2D and 3D porous media. The motivation for this work
was to better understand the mechanisms of colloid retention and transport in soil
porous media at pore scale, which is of relevance to many environmental processes and
applications. While bulk measurements of colloids retention at large scales are feasible,
it remains very diffcult to track the movement of individual colloid particles and their
interactions with grain surfaces experimentally.
In general, the pore-scale simulations of colloid transport could be divided into
two parts: the first part is the simulation of pore-scale
flow in a porous medium,
the second one is the modeling of transport of sub-micron colloid particles in the
simulated pore-scale
flow field. In the pore-scale
flow simulation, a mesoscopic lattice
Boltzmann method was applied in a modeled geometry of a porous medium. To cross
validate our
flow simulation, we also applied a hybrid Navier-Stokes based approach
Physalis developed by Professor Prosperetti
s group at the Johns Hopkins University.
This is necessary as it is very difficult to measure the pore-scale
flow field in physical
experiments. The results show that the lattice Boltzmann method better captures
flow around curved surfaces when compared to the Physalis simulation. Thus, the
flow
field obtained from the lattice Boltzmann method was then used to study the colloid
transport.
As a first step, a Lagrangian particle tracking approach was developed to simulate colloid transport and retention in a two-dimensional (2D) model porous medium.
Stokes drag force, Brownian diffusion, and colloid-collector interaction forces were con-
sidered in the simulation. The primary goal was to quantify the mechanisms of colloid
retention under unfavorable grain surface conditions where both colloid and collector
surfaces are negatively charged. Even in 2D, the simulation was computationally expensive as there exist large local gradients in the interaction forces between colloid
and collector surfaces. The 2D model was developed to mimic key parameters (e.g.,
porosity, mean
flow speed, and other physical and chemical conditions) in a 3D porous
medium, but with an advantage of lower computational cost.
The results from the 2D model quantitatively agreed with the limited results
from a 3D model. For the high repulsive energy barrier considered in the simulations
(e.g., larger than 1000 kT), it is found that colloids could only be retained at the
secondary energy minimum (SEM). The retention at SEM was found to be dynamically
irreversible when the SEM depth reached about 4 kT or less. The fraction of colloids
that could move into the SEM well and be subsequently retained by the attractive
van der Waals force was controlled by the competition of hydrodynamic along-the-streamline transport and the Brownian cross-the-streamline diffusion. The tangential
hydrodynamic force could slowly drive retained colloids towards the rear stagnation
region along the grain surface, leading to accumulation of retained colloids there. These
mechanistic insights explain well the dependence of retention ratio on
flow speed at a
given ionic strength as well as the saturation of retention ratio with ionic strength at a
prescribed
flow speed. The simulated colloid retention was also found to be consistent
with limited experimental observations.
To further validate our simulation model, colloid transport under favorable surface conditions was also briefly studied in our simulations. The results were found to be
in qualitative agreement with results from both experiments and theoretical prediction.
In order to speed up the colloid transport simulation and expand the capability for 3D model investigation, parallel computation using MPI (message passing
interface) was developed for both the
flow and colloid transport simulations. Domain
decompositions in 1D, 2D and 3D have been implemented in the
flow simulation and the
resulting parallel scalability (i.e., wallclock time versus the number of processors used)
was studied. It was found that a speedup of about 350 could be achieved with 1024
processors using 3D domain decomposition. In Lagrangian colloid transport modeling,
both domain and particle decomposition schemes were implemented and compared.
It was found that the particle decomposition provides better computational eficiency
than the domain decomposition for our specific problem of colloid transport in porous
media, due to its better load balance and less data communication.
With the parallel codes, 3D porous media with three different geometric congurations were used to study the effects of packing and size distribution of grains on
the pore-scale
flow and colloid transport. It was demonstrated that the permeability
of porous media was affected by both the packing and size distribution of the grains.
The permeability of porous media with the three different geometric congurations in
our simulation was compared to the values predicted by the empirical Kozney-Carman
equation. Simulation results show that colloid transport and retention are directly
affected by the geometric conguration of the porous media. At the same mean
flow
speed and porosity, the randomly packed glass beads yield a much higher colloid retention (about 170%) than regular packed glass beads because of the more irregular
streamlines. It appears that the size distribution of grains does not have an obvious
effect on the colloid retention. These results should be viewed as preliminary and more
porous-medium geometry models should be tested to further corroborate the above
results.