Accepted Manuscript

Numerical simulation of displacement characteristics of CO2 injected in pore-scale

porous media

Qianlin Zhu, Qianlong Zhou, Xiaochun Li

PII: S1674-7755(15)00107-9
DOI: 10.1016/j.jrmge.2015.08.004
Reference: JRMGE 194

To appear in: Journal of Rock Mechanics and Geotechnical Engineering

Received Date: 18 May 2015

Revised Date: 18 August 2015
Accepted Date: 24 August 2015

Please cite this article as: Zhu Q, Zhou Q, Li X, Numerical simulation of displacement characteristics of
CO2 injected in pore-scale porous media, Journal of Rock Mechanics and Geotechnical Engineering
(2015), doi: 10.1016/j.jrmge.2015.08.004.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Numerical simulation of displacement characteristics of CO2 injected in pore-scale porous

media

Qianlin Zhua, b, *, Qianlong Zhoua, Xiaochun Lib

PT
a. Key Laboratory of Coal-based CO2 Capture and Geological Storage, China University of Mining and Technology, Xuzhou, 221008, China
b. State Key Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics, Chinese Academy of Sciences, Wuhan, 430071,
China

RI
Received 18 May 2015; received in revised form 18 August 2014; accepted 24 August 2015

SC
Abstract: Pore structure of porous media, including pore size and topology, is rather complex. In immiscible two-phase displacement process, the capillary force
affected by pore size dominates the two-phase flow in the porous media, affecting displacement results. Direct observation of the flow patterns in the porous
media is difficult, and therefore knowledge about the two-phase displacement flow is insufficient. In this paper, a two-dimensional (2D) pore structure was
extracted from a sandstone sample, and the flow process that CO2 displaces resident brine in the extracted pore structure was simulated using the Navier-Stokes

U
equation combined with the conservative level set method. The simulation results reveal that the pore throat is a crucial factor for determining CO2 displacement
process in the porous media. The two-phase meniscuses in each pore throat were in a self-adjusting process. In the displacement process, CO2 preferentially broke
through the maximum pore throat. Before breaking through the maximum pore throat, the pressure of CO2 continually increased, and the curvature and position
AN
of two-phase interfaces in the other pore throats adjusted accordingly. Once the maximum pore throat was broken through by the CO2, the capillary force in the
other pore throats released accordingly; subsequently, the interfaces withdrew under the effect of capillary fore, preparing for breaking through the next pore
throat. Therefore, the two-phase displacement in CO2 injection is accompanied by the breaking through and adjusting of the two-phase interfaces.
Key words: level set method; displacement; pore-scale porous media; numerical simulation
M

scale viewpoint, which is important but very complicated due to a large

1. Introduction number of factors influencing the flow, such as fluid density, fluid
D

viscosity, surface tension, flow rates, surface wettability, pore geometry

The increase of greenhouse gas emissions to the atmosphere during and medium heterogeneity (Pan, 2003; Liu et al., 2013).
the 20th century is considered to have contributed to global warming. A Several approaches have been applied to simulate multiphase flow at
TE

reduction in the released rate of CO2 to the atmosphere is essential for
mitigating greenhouse effects. One way of achieving this is to inject
CO2 into deep formations (Yang et al., 2015). Such formations include
oil and gas reservoirs, saline aquifers and unminable coal seams, of
EP
pore-scale in porous media, in which the pore-network model is a
conventional method. Based on two-dimensional (2D) model that is
simplified by ball and column connection, Lenormand and Touboul
(1988) studied the immiscible two-phase flow using the pore-network

which the deep saline aquifers are widely distributed with the largest
CO2 storage capacity (IPCC, 2005). In China, the storage capacity of
deep saline aquifers accounts for 88% of the total geological storage
capacity (Li et al., 2009).
C
model, and established a phase-diagram to identify the flow patterns. In
this phase-diagram, the flow patterns were identified by two
dimensionless parameters (i.e. capillary number and viscosity ratio) that
are determined by fluid viscosity, flow velocity, contact angle and

CO2 that is injected into a saline aquifer displaces the resident brine. interfacial tension. The capillary number does not refer to the pore
Different to single-phase flow, the flow process that CO2 displaces structure. Therefore, the phase-diagram is difficult to be applied but
AC

brine is an immiscible two-phase flow, in which the capillary force
between the CO2 and brine plays an important role. The capillary force
is inversely proportional to interface curvature and directly proportional
to interface tension. Pore diameter of rock is commonly smaller than
millimeter-scale, and the rock is commonly hydrophilic. Therefore, the
capillary force is an important factor that may control the displacement
of the immiscible two-phase flow. However, tempo-spatial variation of
two-phase interfaces between the immiscible two-phase fluids is
difficult to be observed in porous media (Liu et al., 2012). In addition,
methods based on Darcy’s law are effective in investigating macroscale
two-phase flow but cannot reveal the flow dynamics in the immiscible
displacement (Christensen, 2006). Therefore, the dynamics of the
immiscible two-phase flow process needs to be understood from a pore-
plays an instructive role in analysis of two-phase flow in many natural
porous media. Blunt et al. (2002) reviewed the pore-network simulation
in pore-scale flow. The pore-network model simplifies porous media by
connection model of pore body and pore throat with idealized
geometries, acquiring certain prediction effects. However, it is difficult
to extract a pore-network that topologically and geometrically
represents the complex pore geometry and physics, and therefore the
pore-network model cannot reveal the fluid dynamics mechanism
(Ramstad et al., 2012). In addition to pore-network model,
computational fluid dynamics methods, including the lattice Boltzmann
method (Liu et al., 2012), level set method (Gunde et al., 2010),
volume-of-fluid method (Raeini et al., 2012) and phase field method
(Anderson et al., 1998), have been developed in pore-scale simulation.

*Corresponding author. Tel: +86 56 83883501; E-mail: zhuql@cumt.edu.cn

 ACCEPTED MANUSCRIPT
Simulating pore-scale two-phase flow by computational fluid dynamics
methods can reveal flow characteristics.
In this paper, a computational fluid dynamics method with the level
set method is used to simulate movement characteristics of the two-
phase interfaces between the CO2 and resident brine in the displacement
process at pore-scale, and the displacement mechanisms are analyzed,
which may improve our knowledge on CO2 displacement flow in
porous media.
2. Methodology
simulation that uses fluid velocity u, the level sets close to the zero
level set move with velocities that considerably distort the level set
function. In this case, the level set function will cease to be an exact
signed distance function after a few of time steps, affecting numerical
calculation at the next time step (Sussman and Puckett, 2000).
Therefore, a process called re-initialization is needed, which is used to
stop the level set calculation at some points in time and to rebuild a
level set function corresponding to the signed distance function. There
are several ways to accomplish this re-initialization. The most
commonly used method among them is the method introduced by
Sussman et al. (1994). Its virtue is that the level set function is re-

PT
2.1. Geometric model initialized without explicitly finding the zero level set. This re-
A sandstone sample is chosen from a shallow saline aquifer at a initialization method is used to solve Eq. (3) to steady state ( τ → ∞ ):
depth of about 200 m, which is located about 41 km northeast of φ0
φτ = (1 − ∇φ ) (3)
Tongliao, Inner Mongolia, China. In this saline aquifer, a small-scale φ + h2
2

RI
0

CO2 shallow injection experiment is conducted (Zhu et al., 2015). The where φ0 is the level set function distribution calculated by Eq. (2)
pore structure (Fig. 1) is extracted from a microgram of sandstone slice. before the re-initialization; h is a parameter of amplitude that is a
function of grid size; the subscript τ denotes a partial derivative with

SC
respect to the re-initialization time τ that is different from the time t.
Calculating Eq. (3) to steady state will provide a new value for φ with
a property that ∇φ = 1 . In addition, this re-initialization calculation
ensures that the zero level set does not move (Sussman et al., 1994).

U
However, a main defect of this level set method is that the advection of
φ , including the re-initialization steps, is not done in a conservative
AN
way, not even for divergence-free velocity fields. This implies that the
area bounded by the zero level set is not conserved (Sussman and
Puckett, 2000; Olsson and Kreiss, 2005). In order to ensure mass
conserved, Olsson and Kreiss (2005) and Olsson et al. (2007) modified
M

this level set function by replacing the signed distance function with a
v
Fig. 1. Pore structure extracted from a sandstone slice. smeared-out Heaviside function φ% ( x , t ) :
v
0 [φ ( x , t ) < −ε ]
2.2. Numerical method v 
 φ πφ v
φ% ( x , t ) =  1 + + 1 sin [ − ε < φ ( x, t ) < ε ] (4)
D

According to the continuum assumption of fluid, a crucial issue for

 2 2ε 2π ε
v
numerically simulating immiscible two-phase displacement is to track 1 [φ ( x , t ) > ε ]
the movement of interfaces in the two-phase flow. The level set method, where ε corresponds to half of the interface thickness, which can be
TE

an interface capture method, was firstly proposed by Osher and Sethian
(1988). In this method, the flow field is covered by a level set function,
and the interfaces are expressed with contour lines or surfaces of the
level set function. The level set function moves with the flow just as the
EP
taken from half of the characteristic dimension of the smallest cell in
the mesh of computational geometry (Gunde et al., 2010). After the
modification, the interface location is represented by the contour
v
φ% ( x , t ) = 0.5 . In this paper, this modified level set method is chosen to

interfaces do, and the level set function remains a smooth function
when it evolves. The level set function performs naturally for problems
with singularities, large distortion or interface breakup, etc. Unlike the
volume-of-fluid method, the level set function is a continuous function.
C
be the interface tracking method in the CO2 displacement simulation.
The normal vector n and curvature k of the interface can easily be
obtained from the level set function field as (the subscript of modified
function φ% will be omitted in the following part):

Therefore, there are more choices for numerical discretization. In ∇φ 

n= 
φ = 0.5
addition, some geometric parameters such as the interface normal ∇φ
AC

 (5)
k = −∇ ⋅ n φ = 0.5 
vector, tangent vector and curvature can be derived from simple
derivation.
v The momentum equation of fluid is described by the incompressible
In standard level set method, the level set function φ ( x , t ) is defined
to be a signed distance function: Navier-Stokes equation as
v v v v
φ ( x , t ) = d ( x ) = min( x − xi ) (1) ρ ut + ρ (u ⋅ ∇)u = ∇ ⋅ {− pI +η[∇u+(∇u)T ]} + Fsv (6)
xi∈i
v and the continuity equation as
where x is the point coordinates, t is the time, i is the interface domain,
v v ∇⋅u = 0 (7)
φ ( x , t ) >0 on one side of the interface, φ ( x , t ) <0 on the other side,
v v where ρ is the fluid density, η is the dynamic viscosity, I is the identity
and φ ( x , t ) =0 on the interface. The level set function φ ( x , t ) evolves
tensor, p is the pressure field, and Fsv is the term representing
under advection velocity field as
interfacial tension. The level set function evolving under the effect of
φt + u ⋅ ∇φ = 0 (2)
fluid velocity can be described as
where u is the velocity field, and therefore is the movement velocity on 1
φt + u ⋅ ∇φ = ∇ ⋅ [ε∇φ − φ (1 − φ )n] (8)
the interface; the subscript t, which denotes the time, represents a µ
partial derivative with respect to time; ∇ is the Laplace operator. In a
 ACCEPTED MANUSCRIPT
where µ denotes the amount of re-initialization or stabilization of the
level set function φ , and ε determines the interface thickness between
the two fluids. The density ρ and viscosity η are related to the level set
φ in the region of the two-phase flow:
ρ = ρl + ( ρg − ρl )φ 
 (9)
η = ηl + (ηg − ηl )φ 
where the subscripts l and g refer to the liquid and gas phases,
respectively. The interfacial tension Fsv is calculated according to the
continuum surface tension model proposed by Brackbill et al. (1992) as
follows:
v
Fsv ( x ) = σδ (−∇ ⋅ n)n (10) Fig. 2. Porous media with a buffer region.

PT
where σ is the surface tension coefficient; and δ is the Dirac delta
function, which can be written as The inlet velocity is set to 0.5 cm/s. According to conditions of the
δ = 6 ∇φ φ (1 − φ ) (11) shallow saline aquifer, the density and viscosity of the water are set to
1×103 kg/m3 and 1.34×10−3 Pa s, respectively. The density and viscosity

RI
This function ensures the following equation:
1 of the CO2 are set to 38.5 kg/m3 and 1.47×10−5 Pa s, respectively. The
∫0 Fsvdφ = σn (12)
interfacial tension between water and CO2 is set to 44.57×10−3 N/m
In addition, the continuum surface tension model is independent of (Chalbaud et al., 2010).

SC
grid size. The finite element method is used to solve the coupled equation
system. The geometric model is discretized with 2D unstructured
3. Numerical solution and discussion triangle meshes, and boundaries with large curvature are discretized
with finer meshes as shown in Fig. 3. The number of the mesh is

U
The left and right sides of the porous media shown in Fig. 1 are taken
as inlet and outlet boundary conditions, respectively; the lower and
AN
upper boundaries are taken as a symmetric boundary condition. Pores
are considered as water wettable with a contact angle of 180°. The CO2
injected into porous media will automatically choose flow channels.
Therefore, a buffer region, where the CO2 can automatically choose
M
251,327. A linear system solver PARDISO (parallel direct solver) is
used. A relative tolerance of 0.01 and an absolute tolerance of 0.1 are
assumed for the time discretization scheme. The time-stepping method
uses the generalized alpha method to determine each time step.
An initial condition given by an assumed phase distribution may not
be reasonable. Therefore, before the displacement simulation, the flow

flow channels, is connected to the inlet boundary of the geometric
model. The modified geometric model is shown in Fig. 2.
D
field is calculated with zero inlet velocity within a short time to make
the assumed phase distribution to reach a reasonable pattern. The
calculated initial phase distribution is shown in Fig. 4.
At the beginning of the injection, the injected CO2 displaces the

resident water in the porous media through a relatively large channel, as

shown by the arrow in Fig. 5. At this time, the CO2 in relatively narrow
TE

pores does not show obvious flow. When the CO2 in the relatively large
channel reaches a narrow pore throat as shown by the arrow in Fig. 6a,
the CO2 will stop displacing through this channel, and the CO2 in
relatively narrow pores begins to displace the water as shown by the
EP

arrows in Fig. 6b.

C
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 3. Pore geometry discretized with unstructured 2D meshes.

Fig. 5. Injected CO2 flows through the relatively large pore.

M

Comparing CO2 distribution in Fig. 6b and c, we may see that

although CO2 firstly flows through a relatively large pore, a relatively
narrow throat restricts the CO2 to flow through this channel. Therefore,
D

pore throat, rather than pore body, is determinant that determines

channel of the CO2 flow. This result indicates that capillary force is
TE

determinant in CO2 flow in porous media.

Distribution of the two-phase interface in Fig. 6c shows that, when
reaching a pore throat, CO2 will displace through another channel.
Consequently, there will be a moment when all two-phase interfaces
Fig. 4. Initial CO2-water distribution.
EP

reach respective pore throats of potential flow channels. At this time,

once the maximum pore throat is broken through by the CO2,
meniscuses at the other pore throats will retreat quickly; this process
can be shown by comparing Fig. 6c and d. Therefore, the two-phase
C

interfaces are in a self-adjusting process in the displacement process.

This result also indicates that CO2 may have ever reached a position in
AC

porous media where although the CO2 is not seen at a certain time, such
as the positions shown with pink arrows in Fig. 6d. This push-pull
movement of interfaces may improve the dissolution and reaction of
CO2 in porous media (Li, 2011).
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 6. Variations of two-phase interface during displacement.
M

Fig. 7 shows the variation of injection pressure during the
displacement process. The pressure variation indicates that injection
pressure fluctuates throughout the displacement process. The variations
of two-phase distribution and injection pressure reveal that, when CO2
D
Fig. 8 shows the finally simulated results of two-phase distribution.
This result shows that 47.71% of the water is displaced. Therefore,
more than half of the water remains in the porous media.

reaches a pore throat and needs to break through it, the injection
pressure needs to increase high enough. Once the CO2 breaks through
the pore throat under effect of the increased injection pressure, the
TE

injection pressure will release quickly, and the stored pressure by the
capillary pressure of meniscus will release with the meniscuses
retreating. Subsequently, the injection pressure gradually increases to
prepare for breaking through the next pore throat.
EP

Fig. 8. Simulated results of CO2 distribution.

C

4. Conclusions
AC

A 2D pore structure is extracted from a sandstone sample, and CO2

displacement in the sandstone is simulated using the Navier-Stokes
equation combined with the conservative level set method. The
simulation results show that pore throat, rather than pore body, is a
crucial factor that determines flow channel of the CO2 displacement in
porous media. The injected CO2 firstly flows through the largest pore.
When reaching a pore throat, the pore throat will restrict the CO2 from
flowing in that channel, and the CO2 will choose to flow through the
second-largest pore, and so on. In this way, CO2 displacement needs to
break through one or more pore throats. At this time, the CO2 will
Fig. 7. Variation of injection pressure in the injection process.
choose the relatively large pore throat to break through, and therefore
this pore throat determines flow channel of the CO2.
The two-phase meniscuses in each pore throat are in a self-adjusting
 ACCEPTED MANUSCRIPT
process during the displacement. Before breaking through the relatively
large pore throat, the injection pressure of the CO2 increases gradually,
and the curvature and the position of interfaces in other pore throats
adjust accordingly. Once the maximum pore throat is broken through by
the CO2, the capillary force in other pore throats releases subsequently,
and the interfaces retreat accordingly, preparing for breaking through
the next pore throat. The two-phase displacement in CO2 injection is
accompanied by the breaking through and adjusting of the two-phase
interfaces.
Conflict of interest
The authors wish to confirm that there are no known conflicts of
interest associated with this publication and there has been no
significant financial support for this work that could have influenced its
outcome.
Acknowledgements
This work is funded by Key Laboratory of Coal-based CO2 Capture
and Geological Storage, Jiangsu Province, China and US Advanced
Coal Technology Consortium (No. 2013 DFB60140-08). We also thank
U
Prof. Qi Li for useful discussions.
References
AN
Anderson DM, McFadden GB, Wheeler AA. Diffuse-interface models in fluid
mechanics. Annual Review of Fluid Mechanics 1998; 30(1): 139-65.
Blunt MJ, Jackson MD, Piri M, Valvatne PH. Detailed physics, predictive capabilities
M
and macroscopic consequences for pore-network models of multiphase flow.
Advances in Water Resources 2002; 25(8-12): 1069-89.
Brackbill JU, Kothe DB, Zemach C. A continuum method for modeling surface tension.
D
Journal of Computational Physics 1992; 100(2): 335-54.
Chalbaud C, Robin M, Lombard JM, Bertin H, Egermann P. Brine/CO2 interfacial
properties and effects on CO2 storage in deep saline aquifers. Oil & Gas Science and
TE
Technology 2010; 65(4): 541-55.
Christensen BSB. Using X-ray tomography and lattice boltzmann modeling to evaluate
pore-scale processes in porous media. PhD Thesis. Lyngby, Denmark: Institute of
Environment & Resources, Technical University of Denmark, 2006.
EP
Gunde A, Bera B, Mitra SK. Investigation of water and CO2 (carbon dioxide) flooding
using micro-CT (micro-computed tomography) images of Berea sandstone core
using finite element simulations. Energy 2010; 35(12): 5209-16.
C
Qianlin Zhu obtained his M.Sc. and Ph.D. degrees in geotechnical engineering from Hohai University in 2008 and Institute of Rock and
Soil Mechanics, Chinese Academy of Sciences in 2011, respectively. He is a research assistant at Institute of Low Carbon Energy, China
AC
University of Mining and Technology (Xuzhou). His research interests include CO2 geological storage, permeability theory in
geotechnical engineering and isotopic hydrology. In recent years, he is striving to study two-phase displacement flow in porous media
and coupled fluid flow-geomechanics characteristics in CO2 saline aquifer storage.
Intergovernmental Panel on Climate Change (IPCC). IPCC special report on carbon
dioxide capture and storage. Cambridge, UK: Cambridge University Press, 2005.
Lenormand BR, Touboul E. Numerical model and experiments on immiscible
displacements in porous media. Journal of Fluid Mechanics 1988; 189: 165-87.
Li Q. Coupled reactive transport model for heat and density driven flow in CO2 storage
in saline aquifers. Journal of Hazardous, Toxic, and Radioactive Waste 2011; 15(4):
251-8.
Li XC, Fang ZM, Wei N, Bai B. Discussion on technical roadmap of CO2 capture and
storage in China. Rock and Soil Mechanics 2009; 30(9): 1674-8 (in Chinese).
Liu HH, Valocchi AJ, Kang QJ. Pore-scale simulation of high-density-ration multiphase
flows in porous media using lattice Boltzmann method. In: Proceedings of the 19th
PT
International Conference on Computational Methods in Water Resources. University
of Illinois at Urbana-Champaign, 2012.
Liu HH, Valocchi AJ, Kang Qj, Werth C. Pore-scale simulations of gas displacing liquid
in a homogeneous pore network using the lattice Boltzmann method. Transport in
RI
Porous Media 2013; 99(3): 555-80.
Olsson E, Kreiss G. A conservative level set method for two phase flow. Journal of
Computational Physics 2005; 210(1): 225-46.
Olsson E, Kreiss G, Zahedi S. A conservative level set method for two phase flow II.
SC
Journal of Computational Physics 2007; 225(1): 785-807.
Osher S, Sethian JA. Fronts propagating with curvature-dependent speed: Algorithms
based on Hamilton-Jacobi formulations. Journal of Computational Physics 1988;
79(1): 12-49.
Pan CX. Use of pore-scale modeling to understand flow and transport in porous media.
PhD thesis. Chapel Hill, USA: University of North Carolina, 2003.
Raeini AQ, Blunt MJ, Bijeljic B. Modelling two-phase flow in porous media at the pore
scale using the volume-of-fluid method. Journal of Computational Physics 2012;
231(17): 5653-68.
Ramstad T, Idowu N, Nardi C, Øren PE. Relative permeability calculations from two-
phase flow simulations directly on digital images of porous rocks. Transport in
Porous Media 2012; 94(2): 487-504.
Sussman M, Puckett EG. A coupled level set and volume-of-fluid method for computing
3D and axisymmetric incompressible two-phase flows. Journal of Computational
Physics 2000; 162(2): 301-37.
Sussman M, Smereka P, Osher S. A level set approach for computing solutions to
incompressible two-phase flow. Journal of Computational Physics 1994; 114(1):
146-59.
Yang DX, Li Q, Zhang LZ. Propagation of pore pressure diffusion waves in saturated
porous media. Journal of Applied Physics 2015; 117(13): 134902. Doi:
10.1063/1.4916805.
Zhu QL, Li XC, Jiang ZB, Wei N. Impacts of CO2 leakage into shallow formations on
groundwater chemistry. Fuel Processing Technology 2015; 135: 162-7.
 ACCEPTED MANUSCRIPT
The authors wish to confirm that there are no known conflicts of interest associated with this publication and there
has been no significant financial support for this work that could have influenced its outcome.

PT
RI
U SC
AN
M
D
TE
C EP
AC