1

2
3
4
5
6
7
8 Mixed finite element method based on the Crank–Nicolson
9 scheme for Darcy flows in porous media
10
11 Akshaya Poonepalle, Brent Shi, Elly Kang, Joseph Chin,
12 Mannan Singh, Prakhar Gupta, Riya Malhotra, Sanjana Kadiyala
Sirihansika Thadiparthi, Thinh Kieu, Yeeun Kim
13
14
15
16 Abstract - This article is devoted to the a priori error estimates of the fully discrete Crank-Nicolson ap-
proximation for the Darcy flows. Optimal order error estimates in both L1 (0, T ; H 1 (⌦)) and L1 (0, T ; L2 (⌦))
17 norms are established for the lowest order. Numerical experiments confirm the theoretical analysis regarding
convergence rates.
18
Keywords : Mixed finite element method; Crank-Nicolson; error estimates, optimal order.
19
Mathematics Subject Classification (2020) : 65M12; 65M15; 65M60; 35Q35; 76S05; 65N15; 65N30;
20 35K15; 35K05.
21
22 1 Introduction
23
The paper is dedicated to the analysis of mixed finite element approximations of the solutions of the
24 system of equations modeling the flows of compressible fluid in porous media subject to the Darcy law.
25 This phenomenon generated a lot of interest in the research community such as engineering, environmental
26 and groundwater hydrology and in medicine.
Darcy’s law is commonly related to viscous fluid laminar flows in porous media and is characterized by
27 the permeability coefficient, which is obtained empirically in order to match the linear relation between
then the
Ig
28 velocity vector and pressure gradient. Darcy’s equation has also been obtained rigorously within the context
of homogenization and other averaging/upscaling techniques [25,31]. Fromahydrodynamic point of view, the
29 X
30 X
Darcy’s equation is interpreted as the momentum equation. The Darcy’s equation, the continuity equation,
and the equation of state serve as the framework to model processes in reservoirs [9, 23]. For a slightly
31 compressible fluid, the original PDE system reduces to a scalar linear second order parabolic equation for
32 the density only. is
O
The popular numerical method for modeling flow in porous media are the mixed finite element ap-

É Oita
33 proximations (e.g. [10, 13, 21, 27, 28]). This method is widely used because of their inherent conservation

o
34 properties and because they produce accurate flux even for highly homogeneous media with large jumps
35 in the conductivity (permeability) tensor [12]. Since the pioneering work of Raviart and Thomas [34], the
method has become a standard way of deriving high order conservative approximations. We recommend to
36 the reader [6] for general accounts of the mixed method. Douglas et al. [11] proposed semidiscrete mixed
37 finite element methods to approximate the solution of the system (3.11a) – (3.11b) and obtained optimal
38 order error estimates for the pressure in L2 under reasonable assumptions. In [29], one of the authors has
further analyzed the method and obtained optimal order error estimates for the flux variable in several
39
40
norms of interest.
There exist several time-discretization methods to deal with the parabolic equations such as backward
the
the the one
41 fuel Euler method, Crank–Nicolson method and Runge–Kutta method [17]. As is known to all, Crank–Nicolson
42
43
scheme [8]was
X
is firstly proposed by Crank and Nicolson for the heat-conduction equation in 1947, and it is
unconditionally stable with second-order accuracy. Moreover, because of its high accuracy and uncondi-
tional stability, the scheme has been widely used in many PDEs. So we use Crank–Nicolson scheme and
44 prove optimal order of convergence.
Ene
45 M
46 the pump journal of undergraduate research 0 (2017), 000–000 1
47
48
49
50
1
2
3
4

r
5
The paper is organized as follows. In section 2, we introduce notations and some of relevant results. We
6 present a semi discrete mixed finite element approximation for the problem. Existence and uniqueness are
7 discussed and some known results are recalled. In section 4, we derive error estimates for the two relevant
8 functions. We consider the fully discrete mixed finite element method based on the Crank-Nicolson scheme
to approximate the solution of the system (3.2). The optimal order estimates are established in L2 -norms
9 for density and momentum under reasonable assumptions on the regularity of solutions. In section 5,
10 the results of a few numerical experiments using the lowest Raviart-Thomas mixed finite element in the
11 two-dimensions are reported. These results support our theoretical analysis regarding convergence rates.
12
2 Preliminaries and auxiliaries
13
u
14
15
y We consider a fluid in porous medium occupying a bounded domain ⌦ ⇢ Rd , d 2 with boundary @⌦.
Let x 2 Rd , 0 < T < 1 and t 2 (0, T ] be the spatial and time variables respectively. The fluid flow has
velocity v(x, t) 2 Rd , pressure p(x, t) 2 R and density ⇢(x, t) 2 R+ = [0, 1).
16
17 8
The Darcy equation, which is considered as a momentum equation, is studied in [4, 30] of the form

18 v(x, t) = rp(x, t) (2.1)
µ
19

if
with the permeability tensor  and the viscosity coefficient µ, describes the linear relationship between
20 the velocity v of the creep flow and the gradient of the pressure p, which is valid when the velocity v is
21 extremely small [3] observed in oil and water wells due to low permeability. A theoretical derivation of
22 Darcy’s law can be found in [16, 24, 36].
Multiplying both sides of the equation (2.1) to ⇢, we find that
23

24 ⇢(x, t)v(x, t) = ⇢(x, t)rp(x, t). (2.2)
µ
25
26 We recall that the fluid’s compressibility for isothermal conditions is

27 1 dV 1 d⇢
$= = ,
28 V dp ⇢ dp
29 where V , here, denotes the fluid’s volume. In many cases such as (isothermal) compressible liquids, $
30 is assumed to be a constant [4, 23]. In particular, it is a small positive constant for (isothermal) slightly
compressible fluids such as crude oil and water. This condition is commonly used in petroleum and reservoir
31
engineering [1,9], where the fluid dynamics in porous media have important applications. The current paper
32 is focused on (isothermal) slightly compressible fluids, hence, we study the following equation of state
33
1 d⇢
34 = $, where the constant compressibility $ > 0 is small. (2.3)
⇢ dp
35
Hence
36 r⇢ = $⇢rp, or ⇢rp = $ 1
r⇢. (2.4)
37
Combining (2.2) and (2.4) implies that
38

39 ⇢(x, t)v(x, t) = r⇢(x, t). (2.5)
µ$
40
The continuity equation is
41
42 ⇢t (x, t) + r · (⇢(x, t)v(x, t)) = f (x, t), (2.6)
43
where 2 (0; 1) is the constant porosity, f is external mass flow rate.

In
44
45
46 the pump journal of undergraduate research 0 (2017), 000–000 2
47
48
49
50
1
2
3
4
5
By combining (2.5) and (2.6) we have
6
7 m(x, t) + r⇢(x, t) = 0,
8 ⇢t (x, t) + r · m(x, t) = f (x, t),

fue
9 
where m(x, t) = ⇢(x, t)v(x, t), = µ$ > 0.
10
11
By rescaling the variables ⇢ ! ⇢, ! 1
que
m(x, t) + r⇢(x, t) = 0,
. We can assume = 1 to obtain system of equations

12 (2.7)
⇢t (x, t) + r · m(x, t) = f (x, t).
13
14 We recall some elementary inequalities that will be used in this paper.
For all a, b 0,
15 2 1 (ap + bp )  (a + b)p  2|p 1| (ap + bp ) for all p > 0. (2.8)
16
We recall a discrete version of Gronwall Lemma in backward di↵erence form, which is useful later. It
17 can be proven without much difficulty by following the ideas of the proof in Gronwall Lemma.
18
Lemma 2.1 Assume ` 0, 1 `⌧ > 0 and the nonnegative sequences {an }1 1
n=0 , {gn }n=0 satisfying
19
20 an an 1
`an  gn , n = 1, 2, 3 . . .
⌧
21
22 then
⇣ n
X ⌘
n
23 an  (1 `⌧ ) a0 + ⌧ (1 `⌧ )i 1
gi . (2.9)
i=1

E
24
25 Proof. Let ān = (1 `⌧ )n an . Simple calculation shows that
26 ān ān 1 ⇣a an 1 ⌘
n
= (1 `⌧ )n 1 `an  (1 `⌧ )n 1
gn .
27 ⌧ ⌧
28 Summation over n leads to
n
X
ān ā0
29  (1 `⌧ )i 1
gi ,
⌧ i=1
30
31 and hence (2.9) holds true. ⇤
Notations. Throughout this paper, we assume that ⌦ is an open, bounded subset of Rd , with d =
32 2, 3, . . ., and has C 1 -boundary @⌦. For s 2 [1, 1), we denote Ls (⌦) be the set of s-integrable functions on
33 ⌦ and (Ls (⌦))d the space of d-dimensional vectors which have R all components in
s
R L (⌦). We denote (·, ·)
s s d
34 the inner product in either L (⌦) or (L (⌦)) that is (⇠, ⌘) = ⌦ ⇠⌘dx or (⇠, ⌘) = ⌦ ⇠ · ⌘dx and kukLs (⌦) =
R 1/s
35 ⌦
|u(x)|s dx for standard Lebesgue norm of the measurable function. The notation h·, ·i will be used
for the L2 (@⌦) inner-product. For m 0, s 2 [1, 1], we denote the Sobolev spaces by W m,s (⌦) = {v 2
36 ⇣P R ⌘1/s
37 Ls (⌦), : D↵ v 2 Ls (⌦), |↵|  m} and the norm of W m,s (⌦) by kvkW m,s (⌦) = |↵|m ⌦ |D ↵ s
u| dx ,
P ↵ s
and kvkW m,1 (⌦) = |↵|m ess sup⌦ |D u|. Finally we define L (0, T ; X) to be the space of all measurable
38 ⇣R ⌘1/s
T s
39 functions v : [0, T ] ! X with the norm kvkLs (0,T ;X) = kv(t)kX dt , and L1 (0, T ; X) to be the
40 space of all measurable functions v : [0, T ] ! X such that v : t ! kv(t)kX is essentially bounded on [0, T ]
with the norm kvkL1 (0,T ;X) = ess supt2[0,T ] kv(t)kX .
41
Throughout this paper, we use short hand notations, k·kk = k·kk,2 and k·k = k·k0,2 and k·kr,p = k·kW r,p
42
k⇢(t)k = k⇢(·, t)kL2 (⌦) , 8t 0 and ⇢0 (·) = ⇢(·, 0).
43 Throughout this paper, we use C, C1 , C2 , . . . to denote a generic positive constant whose value may
44 change from place to place but and independent of the parameters of the discretization.
45
46 the pump journal of undergraduate research 0 (2017), 000–000 3
thereferee
This was requestedby
47
48 instead of and
Did therefereemean are makesense
Doesthis
49
50
1
2
3
4
5
3 The mixed finite element method
6
7 We study the initial– boundary value problem or IBVP
8 m(x, t) + r⇢(x, t) = 0 (x, t) 2 ⌦ ⇥ (0, T ), (3.1a)
9 ⇢t (x, t) + r · m(x, t) = f (x, t) (x, t) 2 ⌦ ⇥ (0, T ). (3.1b)
10 The initial and boundary conditions:
11 ⇢(x, 0) = ⇢0 (x) in ⌦, ⇢(x, t) = (x, t) on @⌦ ⇥ (0, T ),
12 we also require at t = 0: ⇢0 (x) = (x, 0) on boundary @⌦. The mixed formulation of (3.1a)–(3.1b) reads
13 as follows. Find (m, ⇢) : [0, T ] ! H(div, ⌦) ⇥ L2 (⌦) ⌘ M ⇥ R such that
14 (m, v) (⇢, r · v) = h , v · ⌫i 8v 2 M, (3.2a)
15 (⇢t , q) + (r · m, q) = (f, q) 8q 2 R, (3.2b)
16 where h·, ·i is the inner product in L2 (@⌦) and ⌫ denotes the unit outer normal vector to @⌦.
17 Let {Th }h be a quasi-regular polygonalization of ⌦ (by triangles, rectangles, tetrahedron or possibly

①is
18 hexahedron), with max⌧ 2Th diam ⌧  h. The discrete subspace Mh ⇥ Rh ⇢ M ⇥ R are defined as

19 Mh = {v 2 H(div, ⌦)|v = a + bx for all T 2 Th },

2
20 Rh = {q 2 L (⌦)|q is constant on each element T 2 Th }.
21
22 constant functions.
of
So Mh denotes RT0 space (the Raviart-Thomas-Nedelec [7, 33]) and Rh is the space of piecewise

For momentum, let ⇧ : M ! Mh be the Raviart-Thomas projection [32], which satisfies

23
(r · (⇧m m), q) = 0, for all m 2 M, q 2 Rh . (3.3)
24
2
25 For density, we use the standard L -projection operator, see in [7], ⇡ : R ! Rh , satisfying

26 (⇡⇢ ⇢, q) = 0, for all ⇢ 2 R, q 2 Rh ,

(3.4)
27 (⇡⇢ ⇢, r · mh ) = 0, for all mh 2 Mh , ⇢ 2 R.
28 This projection has well-known approximation properties, e.g. [5, 6, 18].
29 k⇧mk  C kmk + h kr · mk , 8m 2 M \ (W 1,2 (⌦))d . (3.5)
30 1,2 d
k⇧m mk  Ch kmk1 , 8m 2 M \ (W (⌦)) . (3.6)
31 2
k⇡⇢k  C k⇢k , 8⇢ 2 L (⌦). (3.7)
32 2 2,2
k⇡⇢ ⇢k + h k⇡⇢ ⇢k1  Ch k⇢k2 , 8⇢ 2 W (⌦). (3.8)
33 The two projections ⇡ and ⇧ preserve the commuting property div ⇧ = ⇡ div : V ! Rh .
34 We shall also find useful the following inequalities valid for each T 2 Th
35 1
kr · mkL2 (T )  Ch kmkL2 (T ) , m 2 Mh . (3.9)
36 km · ⌫kL2 (@T )  Ch
1
2 kmkL2 (T ) , m 2 Mh . (3.10)
37
38
(mh , v)
①
The mixed finite element problem are stated as: Find a pair (mh , ⇢h ) : [0, T ] ! Mh ⇥ Rh such that
(⇢ is
, r · v) =
h h , v · ⌫i 8v 2 Mh , (3.11a)
39
(⇢ht , q) + (r · mh , q) = (f, q) 8q 2 Rh . (3.11b)
40
Initially we take ⇢0h = ⇡⇢0 . With this choice, we obtain for all q 2 Rh
41
42 ⇢0h , q = (⇡⇢0 (x), q) .
43 We assume that f 2 L2 (0, T ; L2 (⌦)), 2 L2 (0, T ; L2 (@⌦)) and ⇢0 2 L2 (⌦). Then, see for instance
[15, 19, 20, 35], pages 156–158, for more details, there exists a unique weak solution for (3.2a)–(3.2b) in the
44
following sense: there exists a function ⇢ 2 L2 (0, T ; H01 (⌦)) \ C(0, T ; L2 (⌦)).
45
46 the pump journal of undergraduate research 0 (2017), 000–000 4
47
48
49
50
1
2
3
4
5
4 Fully discrete problem based on Crank-Nicolson scheme
6
7 The discretization scheme we want to consider is implicit and it is based on the use of the Crank-Nicolson
method as discretization in time and on the use of the finite element mesh described above.
8
We first divide the interval [0, T ] into N equally-spaced subintervals by the following points
9
0 = t0 < t1 < . . . < tN +1 = T
10
1
11 with ti = i⌧ , ti 1
2
= (i 2 )⌧ , for time step ⌧ = T /N . For a smooth function ' on [0, T ], we define
1
i i
12 ' = '(·, ti ) and ' = '(·, ti 12 ). We shall denote by '¯i the following arithmetic mean value, when
2

i N +1
13 {' }i=0 is a discrete function, between the two time levels i 1 and i:
14 'i + 'i 1
'¯i = .
15 2
16 We also define
17 'i 'i 1
'i+1 2'i + 'i 1
'i = , and 2
'i+1 = ( 'i+1 ) = 8i = 1, 2, . . . , N + 1.
18 ⌧ ⌧2
19 N +1
The fully discrete time mixed finite element approximation to (3.2) is defined as follows: Given f i i=1
2
20 2 i N +1 2
L (⌦), i=1
2 L (@⌦). Find a pair (mih , ⇢ih )
in Mh ⇥ Rh , i = 1, 2, . . . , N + 1 such that
21 ⌦ i ↵
m̄ih , v ⇢¯ih , r · v = ¯ , v · ⌫ , 8v 2 Mh , (4.1a)
22
23 ⇢i , q + r · m̄i , q = f¯i , q , 8q 2 Rh .
h h (4.1b)
24 Initially, we take (m0h , ⇢0h ) = (⇡r⇢0 , ⇡⇢0 ). With this choice we obtain
25 ⌦ 0 ↵
m0h , v ⇡⇢0 , r · v = , v · ⌫ , 8v 2 Mh , (4.2)
26
⇢0h , q = (⇡⇢0 , q) , 8q 2 Rh . (4.3)

i
27
Remark 4.1 We can use f i 2 in place of f¯i in the second equation of (4.1). If f is continuous function
1
28
in time then we defined f i = f (·, ti ). If f is less regularity then we define
29
30 Teti I Z
1 ti
fi =
31 ⌧ ti 1
f (·, t)dt, i = 1, . . . N + 1.
aidthygtegar
32
Lemma 4.2 (Stability) Let (mnh , ⇢nh ) solve the fully discrete finite element approximation (4.1) for each
33 time step n, n = 1, . . . , N +1. Suppose that (m, ⇢) 2 M⇥R, f 2 L1 (0, T ; L2 (⌦)), and 2 L1 (0, T ; L2 (@⌦)),
34 ⇢0 2 L2 (⌦). Then, there exists a positive constant C independent of ⌧ such that for ⌧ sufficiently small
35 (i) For all n = 1, 2, . . . , N + 1,
36 n
X 2
⇣ n
X 2 2
⌘
2 2
k⇢nh k + ⌧ m̄ih  C(h) k⇢0 k + ⌧ f¯i + ¯i L2 (@⌦)
. (4.4)
37
i=1 i=1
38
(ii) For all n = 1, 2, . . . , N + 1,
39
⇣ n X
X i ⌘
40 2 2 0 2 2 2
kmnh k  C(h) k⇢0 k + L2 (@⌦)
+ ⌧ f¯j + ¯j L2 (@⌦)
. (4.5)
41 i=1 j=1
42
Proof.
43 ⇢ih , m̄ih ) in (4.1a) and (4.1b). Adding the resultant equations gives
(i) Let (q, v) = (¯
44 ⌦ i i ↵
m̄ih , m̄ih + ⇢ih , ⇢¯ih = f¯i , ⇢¯ih ¯ , m̄ · ⌫ .
h
45
46 the pump journal of undergraduate research 0 (2017), 000–000 5
47
48
49
50
1
2
3

y
4
5
Using Hölder’s inequality, triangle inequality and inverse inequality (3.10), we find that
6
⇣ ⌘
7 2 2 1 2
m̄ih + ⇢ih ⇢ih  f¯i ⇢¯ih + ¯i L2 (@⌦)
m̄ih L2 (@⌦)
2⌧
8 1 ¯i 1
 f ( ⇢ih + ⇢ih 1
) ⇢¯ih + Ch 2 ¯i m̄ih .
9 L2 (@⌦)

y
2
10 It follows from Young inequality that
11
2 1 2
12 2 ⇢ih ⇢ih 2 2 2
m̄ih + (1 + ⌧ ) 2 ⇢ih 2 1
f¯i + Ch 1 ¯i
L2 (@⌦)
.
⌧
13

X
14 By discrete Gronwall inequality (2.9) then for n = 1, 2, . . . , N + 1,
15 ⌧ X n
2 ⌧⌘ ⇣1 n⇣ 2
n
X 2 2
⌘
2
16 k⇢nh k + m̄ih C ⇢0h + ⌧ f¯i +h 1 ¯i
L2 (@⌦)
(1 + ⌧ ) i=1 1+⌧ i=1
17 ⇣ n
X ⌘
0 2 2 2
18  C(h)en⌧ ⇡⇢ + ⌧ f¯i + ¯i L2 (@⌦)
i=1
19 ⇣ n
X ⌘
2 2 2
20  C(h) k⇢0 k + ⌧ f¯i + ¯i L2 (@⌦)
.
i=1
21
22 We complete the proof of (4.4).
(ii) By taking v = 2(mih mih 1 ) in the Eq. (4.1a) we find that
23
24 2 1 2
⌦ ↵
mih mih = 2 ⇢¯ih , r · (mih mih 1 ) 2 ¯i , (mih mih 1 ) · ⌫ .

Using Hölder’s inequality, triangle inequality andthe

25
26
tie inverse inequality (3.9)-(3.10), we find that

27 2 1 2
mih mih  ⇢ih + ⇢ih 1
r · (mih mih 1 ) + 2 ¯i L2 (@⌦)
mih mih 1
L2 (@⌦)
28 1
 Ch 1
⇢ih + ⇢ih 1 mih + mih 1 + Ch 2 ¯i mih + mih 1
.
29 L2 (@⌦)

30 Applying Cauchy- Schwartz and (2.8) inequality gives

31 ⇣ ⌘
3 2 1 2 2 2 1 2 2
32 ( mih mih ) mih  C(h) ⇢ih + ⇢ih + ¯i L2 (@⌦)
. (4.6)
2
33
Inserting (4.4) into (4.6) leads to
34
2 ⇣ i
X ⌘
35 2 1 2 2 2 2 2
( mih mih ) mih  C(h) k⇢0 k + ⌧ f¯j + ¯j L2 (@⌦)
. (4.7)
36 3 j=1
37
Due to the discrete Gronwall’s inequality (2.9) with ⌧ = 1 and ` = 23 , we find that
38
n
X i
X
39 2 2 2 2 2
kmnh k  C m0h + C(h) k⇢0 k + ⌧ f¯j + ¯j L2 (@⌦)
. (4.8)
40 i=1 j=1
41 ⌦ ↵
Since m0h , v ⇡⇢0 , r · v = 0
, v · ⌫ , let v = m0h then
42
43 m0h
2
 k⇡⇢0 k r · m0h + 0
m0h
L2 (@⌦) L2 (@⌦)
44 (4.9)
0
 C(h)(k⇡⇢0 k + L2 (@⌦)
) m0h .
45
46 the pump journal of undergraduate research 0 (2017), 000–000 6
47
48
49
50
1
2
3
4
5
Combining (4.8) and (4.9) yields
6
n X
X i
7 2 2 0 2 2 2
kmnh k  C(h) k⇢0 k + L2 (@⌦)
+ ⌧ f¯j + ¯j L2 (@⌦)
.
8 i=1 j=1
9
The proof is complete. ⇤
10
11 Lemma 4.3 Let (mnh , ⇢nh ) solve the fully discrete finite element approximation (4.1) for each time step n,
n = 2, . . . , N + 1. Suppose that f 2 L1 (0, T ; L2 (⌦)), 2 L1 (0, T ; L2 (@⌦)), ⇢0 2 L2 (⌦). Then, there exists
12
a positive constant C independent of ⌧ such that for ⌧ sufficiently small
13
⇢nh ⇢nh 1 ⇣ n
X ⌘
14 2 0 2 2 2
 C(h) k⇢0 k + L2 (@⌦)
+ f¯1 + ¯1 L2 (@⌦)
+ f¯i + ¯i . (4.10)
⌧
15 i=2

16 Proof. Acting the discrete operator on (4.1a) and (4.1b) we get, for all i = 1 . . . , N + 1
17 ⌦ i ↵
m̄ih , v ⇢¯ih , r · v = ¯ , v · ⌫ , 8v 2 Mh , (4.11a)
18
⇢ , q + r · m̄i , q = f¯i , q , 8q 2 Rh .
2 i
h h (4.11b)
19
20 Taking v = m̄ih in (4.11a) and q = ⇢¯ih in (4.11b), adding the resultant equations we obtain

21 2 1 2
⇢ih ⇢ih 2 ⌦ ↵
+ m̄ih = f¯i , ⇢¯ih + ¯i , m̄i · ⌫ .
h
22 2⌧
23 Thanks to the use of Hölder’s inequality, triangle inequality and inverse inequality (3.10), we obtain
24 that
2 2
⇢ih ⇢ih 1 2
25 + m̄ih  f¯i ⇢¯ih + ¯i L2 (@⌦) m̄ih L2 (@⌦)
2⌧
26 1
 f¯i ⇢¯ih + Ch 2 ¯i 2
L (@⌦)
m̄ih .
27
It follows from Young’s inequality that
28
2 1 2
29 ⇢ih ⇢ih 2 2 2 2
(1 + ⌧ ) 2 ⇢ih + m̄ih 2 1
f¯i + Ch 1 ¯i
L2 (@⌦)
,
30 ⌧
31 By discrete Gronwall’s inequality (2.9)
32 n
2 2 X 2 2
33 k ⇢nh k  C ⇢1h + C(h) ⌧ f¯i + ¯i
L2 (@⌦)
. (4.12)
i=2
34
2
35 Let us estimate ⇢1h . At the step i = 1, taking q = ⇢1h in Eq.(4.1b) we find that
36
⇢1h , ⇢1h + r · m̄1h , ⇢1h = f¯1 , ⇢1h .
37
38 Thus
2 2 2 2 2
⇢1h C r · m̄1h + f¯1  Ch 2
m̄1h + C f¯1
39 2 2
(4.13)
 C(h)( m1h + m0h )2 + f¯1 ).
40
41 By (4.13) and (4.5), it implies that
42 2 2 0 2 2 2
⇢1h  C(h) k⇢0 k + L2 (@⌦)
+ (⌧ + 1) f¯1 + ¯1 L2 (@⌦)
. (4.14)
43
44 Substituting (4.14) into (4.12) shows (4.10) holds true.
The proof is complete. ⇤
45
46 the pump journal of undergraduate research 0 (2017), 000–000 7
47
48
49
50
1
2
3
4
5
4.1 Error analysis for fully discrete method
6
In this section we derive an error estimate for the fully discrete scheme. First, we give some results that
7
are crucial in getting the convergence results.
8
9 Lemma 4.4 For n 1 if ⇢tt , ⇢ttt 2 L2 (0, T ; L2 (⌦)), then
10 2
Z tn
1 2
n n 3
11 (i) ⇢¯ ⇢ 2  C⌧ k⇢tt k dt. (4.15)
tn 1

12 2
Z tn
1
n 2
(ii) ⇢n ⇢t 2
 C⌧ 3 k⇢ttt k dt. (4.16)
13 tn 1

14
Proof. (i) By Taylor expansion with integral remainder
15
Z tn
16 1 ⌧ n 1
⇢n = ⇢n 2 + ⇢ 2
+ ⇢tt (t)(tn t)dt.
17 2 t tn 1
2
Z tn
18 1 ⌧ n 1 1

⇢n 1
= ⇢n 2 ⇢ 2
+ ⇢tt (t)(t tn 1 )dt.
19 2 t tn 1
2

20
This implies that
21
2 Z Z tn Z tn 1
22 ⇢n + ⇢n 1
1 1 2
⇢n 2 = ⇢tt (t)(tn t)dt + ⇢tt (t)(t tn 1 )dt dx
23 2 2 ⌦ t 1 tn 1
n
2 2
Z ⇣ Z tn ⌘2 ⇣ Z t n 1 ⌘2 (4.17)
24
 ⇢tt (t)(tn t)dt + ⇢tt (t)(t tn 1 )dt dx.
25 ⌦ tn 1 tn 1
2 2
26
We estimate the right hand side by Höder inequality
27
Z ⇣ Z tn ⌘2 ⇣ Z t n 1 ⌘2
28
⇢tt (t)(tn t)dt + ⇢tt (t)(t tn 1 )dt dx
29 ⌦ tn 1
2
tn 1
2
Z ⇣Z tn Z tn ⌘ Z ⇣Z tn Z tn ⌘
30 1
2
1
2
 |⇢tt |2 dt (tn t)2 dt dx + |⇢tt |2 dt (tn t)2 dt dx (4.18)
31 ⌦ tn 1 tn 1 ⌦ tn 1 tn 1
2 2
Z Z Z Z Z
32 ⌧3 ⇣ ⌘
1
tn tn tn
2
2
2 ⌧3 2
33  |⇢tt | dtdx + |⇢tt | dtdx  k⇢tt k dt.
24 ⌦ tn 1 ⌦ tn 1
12 tn 1
2
34
35 Then (4.15) follows directly from inserting (4.18) into (4.17).
(ii) Similar proof for (4.16). By Taylor expansion with integral remainder
36
Z
37 1 ⌧ n 1 ⌧2 n 1 1 tn
⇢n = ⇢n 2 + ⇢ 2
+ ⇢ 2
+ ⇢ttt (t)(tn t)2 dt.
38 2! t 3! tt 3! tn 1
2
Z
39 1 ⌧ n 1 ⌧2 n 1 1 tn 1
⇢n 1
= ⇢n 2 ⇢ 2
+ ⇢ 2
⇢ttt (t)(t tn 1 )2 dt.
40 2! t 3! tt 3! tn 1
2
41
42
43
44
45
46 the pump journal of undergraduate research 0 (2017), 000–000 8
47
48
49
50
1
2
3
4

0
5
Using (2.8) and Höder inequality shows that
6
2 Z Z tn Z tn
7 ⇢n ⇢n 1
n 1 ⌧ 2 2
1
2
2
⇢t 2
= ⇢ttt (t)(tn t) dt + ⇢ttt (t)(t tn 1) dt dx
8 ⌧ 36 ⌦ tn 1 tn 1
2 2
Z ⇣Z tn ⌘2 ⇣Z tn ⌘2
9 ⌧ 2 1

 ⇢ttt (t)(tn t)2 dt + ⇢ttt (t)(t tn 1)

2
dt dx
10 36 ⌦ tn 1 tn 1
2 2
11 Z Z tn Z tn Z Z tn Z tn
1 2 2 4
1
2 2
1
2
 ⌧ |⇢ttt | dt (tn t) dtdx + |⇢ttt | dt (tn t)4 dtdx
12 36 ⌦ tn 1 tn 1 ⌦ tn 1 tn 1
2 2
13 Z Z Z Z Z tn
⌧3 ⇣ ⌘
1
tn tn 2 ⌧3 2
14  |⇢ttt |2 dtdx + |⇢ttt |2 dtdx  k⇢ttt k dt.
5760 ⌦ tn 1 ⌦ tn 1
2880 tn 1
15 2

16 Then (4.16) follows. ⇤

17
Lemma 4.5 For n 2, suppose ⇢tt , ⇢ttt 2 L2 (0, T ; L2 (⌦)). Then there is a positive constant C such that
18
2
Z tn
19 (¯
⇢ n
⇢ n 1
2 )  C⌧ k⇢tt k dt.
2
(4.19a)
20 tn 2

2
Z tn
1
21 n 2
( ⇢n ⇢t 2
)  C⌧ k⇢ttt k dt. (4.19b)
22 tn 2

23 Proof. Each estimate is a result of using the Taylor theorem with the integral remainder and the Hölder
24 inequality. ⇤
First, we derive and error estimate.
25 Subtract (3.2a) from (4.1a) to obtain
26
27 m̄i m̄ih , v ⇢¯i ⇢¯ih , r · v = 0. (4.20)

28 From (3.2b) we have

29 ⇢¯it , q + r · m̄i , q = f¯i , q . (4.21)
30 Subtract (4.1b) from (4.21) to obtain
31 ⇣ 1
⌘ ⇣ 1
⌘
i i
(⇢i ⇢ih ), q + r · (m̄i m̄ih ), q = ⇢i ⇢t 2
,q ⇢¯it ⇢t 2
,q . (4.22)
32
33 Then from (4.20) and (4.22) we have
34
m̄i m̄ih , v ⇢¯i ⇢¯ih , r · v = 0, 8v 2 Mh , (4.23a)
35 ⇣ 1
⌘ ⇣ ⌘
i i 1
36 (⇢i ⇢ih ), q + r · (m̄i m̄ih ), q = ⇢i ⇢t 2
,q ⇢¯it ⇢t 2 , q , 8q 2 Rh (4.23b)
37 Let
38 ⇢i ⇢ih = ⇢i ⇡⇢i + ⇡⇢i ⇢ih = ⇣ i + ✓i = i
.
i
39 m mih =m i
⇧m + ⇧m i i
mih =⇠ +# =⌘ .i i i

40 From (3.8) and (3.6) we have

41 ⇣ Z ti ⌘
42 ⇣ i = ⇢i ⇡⇢i  Ch2 ⇢i 2
 Ch ⇢0 2
+ k⇢ t k 2 dt . (4.24)
0
43 ⇣ Z ti ⌘
44 ⇠ i = mi ⇧mi  Ch mi 1
 Ch m0 1
+ km k
t 1 dt . (4.25)
0
45
46 the pump journal of undergraduate research 0 (2017), 000–000 9
47
48
49
50
1
2
3
4
5
Theorem 4.6 Let (mn , ⇢n ) solve problem (3.2a)–(3.2b) and (mnh , ⇢nh ) solve the fully discrete finite el-
6 ement approximation (4.1a)–(4.1b) for each time step n, n = 1, . . . , N + 1. Suppose that (m0 , ⇢0 ) 2
7 (W 1,2 (⌦))d , W 2,2 (⌦)), (mt , ⇢t ) 2 (L1 (0, T ; W 1,2 (⌦))d , L1 (0, T ; W 2,2 (⌦))), and ⇢tt , ⇢ttt 2 L2 (0, T ; L2 (⌦)).
8 Then, there exists a positive constant C independent of h and ⌧ such that, for ⌧ sufficiently small,

9 ⇣ Z tn ⌘ 12 ⇣ Z tn ⌘
n n 2 2 2 2 0
(i) k⇢ ⇢h k  C⌧ k⇢tt k + k⇢ttt k dt + Ch ⇢ 2+ k⇢t k2 dt . (4.26)
10 0 0
⇣ Z tn ⌘ 12 ⇣ Z tn ⌘
11 (ii) kmn mnh k  C⌧ 2
2 2
k⇢tt k + k⇢ttt k dt + Ch m0 1 + kmt k1 dt . (4.27)
12 0 0

13 Proof.
(i) For any q 2 Rh , v 2 Mh , then from (4.23a) and (4.23b), recalling the projectors in (3.4) and (3.3),
14
we end up with
15
#̄i , v ✓¯i , r · v = 0, 8v 2 Mh , (4.28a)
16 ⇣ ⌘ ⇣ ⌘
1
i i 1
17 ✓i , q + r · #̄i , q = ⇢i ⇢t 2
,q ⇢¯it ⇢t 2 , q , 8q 2 Rh . (4.28b)
18
19
and
Now choosing (q, v) = (✓¯i , #̄i ) in (4.28a) and (4.28b), adding the resulting equations we obtain
⇣ ⌘ ⇣ ⌘
2 i 1 i 1
20 ✓i , ✓¯i + #̄i = ⇢i ⇢t 2 , ✓¯i ⇢¯it ⇢t 2 , ✓¯i .
21
22
it
Applying Young inequality and (2.8) yields
2 1 2
23 ✓i ✓i 2 i 1 2 i 1 2 2
+ #̄i C ⇢i ⇢t 2
+ C ⇢¯it ⇢t 2
+ 2 ✓¯i
24 2⌧
⇣ 1 2 1 2⌘ ⇣ 2
⌘
i i 1 2
25 C ⇢i ⇢t 2
+ C ⇢¯it ⇢t 2
+ ✓i + ✓i .
2
26
Simple calculation gives
27
2 1 2
28 ✓i ✓i 2 2 i 1 2 i 1 2
✓i C ⇢i ⇢t 2
+ ⇢¯it ⇢t 2
.
⌧ 1+⌧
29
30 According to Lemma 4.4, we have
Z ti
31 i 1 2 i 1 2
2 2
⇢i ⇢t 2
+ ⇢¯it ⇢t 2
 C⌧ 3 k⇢tt k + k⇢ttt k dt.
32 ti 1

33 By the discrete Gronwall inequaliy (Lemma 2.1 with ` = 2/(1 + ⌧ ) and noting that ✓0 = 0),
34
n Z
X ti
2 2 2
35 k✓n k  C⌧ 4 k⇢tt k + k⇢ttt k dt. (4.29)
i=1 ti 1
36
37 The result (4.26) follows straightforwardly using (4.29), (4.24) and the triangle inequality.
38 (ii) For any q 2 Rh , v 2 Mh , from (4.20) and (4.22), using L2 -project and elliptic projection, we find
that
39
40 #̄i , v ✓¯i , r · v = 0, 8v 2 Mh , (4.30a)
⇣ ⌘ ⇣ ⌘
i 1 i 1
41 ✓i , q + r · #̄i , q = ⇢i ⇢t 2 , q ⇢¯it ⇢t 2 , q , 8q 2 Rh . (4.30b)
42
From the sum of Eq. (4.30b) with q = ✓¯i and Eq. (4.30a) with v = #̄i , applying Young inequality, we
43
obtain
44 2 1 ⇣ i 2 2
⌘
i 1 2 i 1 2 2
✓¯i + # #i 1  ⇢i ⇢t 2 + ⇢¯it ⇢t 2 + ✓¯i .
45 2⌧ 2

46 the pump journal of undergraduate research 0 (2017), 000–000 10

47
48
49
50
1
2
3
4
5
By Lemma 4.4, (4.15)-(4.16) we find that
6
2 1 2 Z ti
7 #i #i 3 2 2
 C⌧ k⇢tt k + k⇢ttt k dt.
8 ⌧ ti 1

9 The discrete Gronwall lemma (Lemma 2.1 with ` = 0) yields

10
n Z
X ti
11 n 2 0 2 4 2 2
k# k  C # + C⌧ k⇢tt k + k⇢ttt k dt, (4.31)
ti
12 i=1 1

13 since
14 m0 m0h , v + ⇢0 ⇡⇢0 , r · v = ⇧m0 m0 , v 8v 2 Mh .
15 Let v = ⇧m0 m0h , then
16
⇧m0 m0h  ⇧m0 m0  Ch1 m0 1
. (4.32)
17
18 Inserting (4.32) into (4.31) we find that
19 ⇣Z tn ⌘ 21
2 2
k#n k  Ch m0 1
+ C⌧ 2 k⇢tt k + k⇢ttt k dt . (4.33)
20 0
21
Combining (4.33), (4.25) and the triangle inequality we complete the proof (4.27). ⇤
22
23 Theorem 4.7 Let (mn , ⇢n ) solve problem (3.2) and (mnh , ⇢nh ) solve the fully discrete finite element ap-
proximation (4.1) for each time step n, n = 1, . . . , N + 1. Suppose that ⇢tt , ⇢ttt 2 L2 (0, T ; L2 (⌦)). Then,
24
there exists a positive constant C independent of h and ⌧ such that, for ⌧ sufficiently small,
25
26 ⇢n ⇢n 1 ⇢nh ⇢nh 1
C h+⌧ . (4.34)
⌧ ⌧
27
28 Proof. Taking the di↵erence in time of (4.28a) and (4.28b) we find that
29
30 #̄i , v ✓¯i , r · v = 0, 8v 2 Mh , (4.35a)
⇣ 1
⌘ ⇣ ⌘
31 2 i i i 1
✓ , q + r · #̄i , q = ( ⇢i ⇢t 2
), q ⇢it ⇢t 2 ), q , 8q 2 Rh .
(¯ (4.35b)
32
33 From the sum of Eq. (4.35b) with q = ✓¯i and Eq. (4.35a) with v = #̄i , applying Young inequality, we
obtain ✓ ◆
34 2 2
✓i ✓i 1 2 2 i 1 2 i 1 2
35 ✓i  C ( ⇢i ⇢t 2 ) + (¯ ⇢it ⇢t 2 ) .
⌧ 1+⌧
36 Applying the discrete Gronwall’s inequality implies
37
n
X 2 2
2 1 1
38 2 i i
k ✓n k  ✓1 + C⌧ ( ⇢i ⇢t 2
) + ⇢it
(¯ ⇢t 2
)
39 i=2
(4.36)
1 2 Xn 2 2
40 ✓ i i 1
i 1
= + C⌧ ( ⇢ ⇢t 2
) + ⇢it
(¯ ⇢t 2
) .
41 ⌧ i=2

42 Thanks to (4.29),
2 Z t1
43 ✓1 2 2 2
 C⌧ k⇢tt k + k⇢ttt k dt. (4.37)
44 ⌧ 0

45
46 the pump journal of undergraduate research 0 (2017), 000–000 11
47
48
49
50
1
2
3
4
5
Thanks to (4.19a) and (4.19b),
6
n ⇣
X 2 2 ⌘ n Z
X ti
1 1
7 ( ⇢i ⇢t
i 2
) + ⇢it
(¯
i
⇢t 2
)  C⌧
2
k⇢tt k + k⇢ttt k dt.
2
(4.38)
8 i=2 i=2 ti 2

9 Combining (4.36) with (4.37) and (4.38) gives

10 Z t1 n Z
X ti
n 2 2 2 2 2 2 2
11 k ✓ k  C⌧ k⇢tt k + k⇢ttt k dt + C⌧ k⇢tt k + k⇢ttt k dt
0 i=2 ti 2
12 n Z ti
X 2 2
13  C⌧ 2 k⇢tt k + k⇢ttt k dt
i=2 ti 2
14
15 for all n = 2, . . . , N + 1.
Then
16 n Z
⇣X ti ⌘ 12
n 2 2
17 k ✓ k  C⌧ k⇢tt k + k⇢ttt k dt . (4.39)
i=1 ti 2

18
By the triangle inequality, (4.39) and (4.24),
19
n Z
⇣X ti ⌘ 12
20 2 2
k (⇢n ⇢nh )k  k ✓n k + k ⇣ n k  C⌧ k⇢tt k + k⇢ttt k dt + Ch ⇢i 1
.
ti
21 i=2 2

22 This proves (4.34). ⇤

23
24 5 Numerical results
25 In this section we carry out numerical experiments using mixed finite element based on the Crank- Nicol-
26 son scheme to solve problem (4.1a)–(4.1b) in two dimensional regions. For simplicity, the region of ex-
27 amples are unit square ⌦ = [0, 1]2 . We used FEniCS [22] to perform our numerical simulations. We
divided the unit square into an N ⇥ N mesh of squares, each then subdivided into two right triangles
28 using the UnitSquareMesh class in FEniCS. The triangularization in region ⌦ is uniform subdivision in
29 each dimension. Our problem is solved at each time level starting at t = 0 until the final time T = 1.
30 At time T = 1, we measured the L2 -errors of the density and the momentum. We obtain the con-
ln(ei /ei 1 )
vergence rates r = ln(h i /hi 1 )
of finite approximation at eight levels with the discretization parameters
31 p
h 2 {1/2, 1/4, 1/8, 1/16, 1/32, 1/64, 1/128, 1/256}(the mesh size is actually h 2) respectively.
32 To test the convergence rates of the proposed algorithm, we choose the the true solution of the problem
33 (3.1a)–(3.1b) by
34 ⇢(x, t) = et x21 (1 x1 )x2 (1 x2 ) and
 t
35 e x1 (2 3x1 )x2 (1 x2 )
m(x, t) = 8(x, t) 2 [0, 1]2 ⇥ (0, 1].
et x21 (1 x1 )(1 2x2 )
36
For simplicity, we take = 1 on ⌦. The forcing term f is determined from equation ⇢t + r · m = f .
37
Explicitly, ⇥ ⇤
38 f (x, t) = et x21 (1 x1 )x2 (1 x2 ) 2et (1 3x1 )x2 (1 x2 ) x21 (1 x1 ) .
39 The initial condition and boundary condition are determined according to the analytical solution as follows:
40
⇢0 (x) = x21 (1 x1 )x2 (1 x2 ), and ⇢(x, t) = 0.
41 @⌦

42 The numerical results are listed in Table 1 below.

43 For the given problem, nearly second order and first order convergence are observed respectively in L2
for the density and momentum. Slightly better than second order convergence is observed for the density,
44
but as the mesh is refined, the error ratio approaches one in accordance with the theory.
45
46 the pump journal of undergraduate research 0 (2017), 000–000 12
47
48
49
50
1
2
3
4
5
N k⇢ ⇢h k Rates km mh k Rates
6 2 1.2462e-2 – 6.2211e-2 –
7 4 3.5764e-3 1.80 3.1202e-2 0.99
8 8 8.8489e-4 2.01 1.5252e-2 1.03
16 2.1890e-4 2.02 7.4188e-3 1.03
9 32 5.4645e-5 2.00 3.6974e-3 1.00
10 64 1.3666e-5 2.00 1.8452e-3 1.00
11 128 3.4149e-6 2.00 9.2200e-4 1.00
256 8.5373e-7 2.00 4.6035e-4 1.00
12
13 Table 1: Results of the Crank-Nicolson scheme for the density and momentum with ⌧ = h/20.

14
In the second example we consider the non-homogeneous Dirichlet boundary condition. We test the
15
stability of the method with di↵erent time step. We take the true solution of the problem (3.1a)–(3.1b) to
16 be 
17 ⇡e t cos ⇡x1 sin x2
⇢(x, t) = e t sin ⇡x1 sin x2 , m(x, t) = (x, t) 2 [0, 1]2 ⇥ (0, 1].
e t sin ⇡x1 cos x2
18
The forcing term f , initial condition and boundary condition accordingly are
19
20 f (x, t) = ⇡ 2 e t
sin ⇡x1 sin x2 , (x, t) 2 [0, 1]2 ⇥ [0, 1],

21 and
(
22 0 if (x1 , x2 ) 2 {0, 1} ⇥ (0, 1],
⇢0 (x) = sin ⇡x1 sin x2 , ⇢(x, t)|@⌦ = t
23 e sin 1 sin ⇡x1 if (x1 , x2 ) 2 (0, 1) ⇥ {1}.
24 Table 2 presents the results for ⌧ = h.
25
N k⇢ ⇢h k Rates km mh k Rates
26
2 1.6323e-2 – 4.1600e-01 –
27 4 4.3628e-3 1.90 2.1611e-01 0.94
28 8 1.0634e-3 2.04 1.1040e-01 0.97
29 16 2.6451e-4 2.01 5.5712e-02 0.99
32 6.5917e-5 2.00 2.7833e-02 1.00
30 64 1.6453e-5 2.00 1.3856e-02 1.00
31 128 4.1139e-6 2.00 6.9205e-03 1.00
32 256 1.0288e-6 2.00 3.4545e-03 1.00

33 Table 2: Results of the Crank-Nicolson scheme for the density and momentum with ⌧ = h.
34
Tables 1-2 represent the numerical solution errors and convergence rates in L2 -norm. In both cases,
35 errors are calculated at time T = 1 and clearly demonstrates the second order of convergence for density
36 variable and first order of convergence for momentum variable in L2 -norm.
37
38 6 Conclusion
39 In this paper, we have established a new fully discrete mixed finite element method based on the Crank
40 Nicolson scheme for Darcy flows. The spatial discretization is mixed and based on the lowest-order Raviart–
Thomas finite elements, whereas the time discretization is based on the Crank–Nicolson scheme. We have
41
proven the convergence of the scheme by estimating the error in term of discretization parameters. It
42 has been shown that our method has the optimal convergence rate for the density and momentum and
43 this scheme has stability with the di↵erent time steps. The numerical experiments agree with the estimates
derived theoretically. Obviously, this method can be expanded to the case of many dimensions easily. There
44
are some open questions including the possible extension of the method to non-Darcy fluid flows.
45
46 the pump journal of undergraduate research 0 (2017), 000–000 13
47
48
49
50
1
2
3
4
5
Acknowledgments
6
7 The authors thank the editor and referees for their valuable comments and suggestions which helped us to
improve the results of this paper.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46 the pump journal of undergraduate research 0 (2017), 000–000 14
47
48
49
50
1
2
3
4
5
References
6
[1] T. Ahmed, Reservoir Engineering Handbook, Chemical, Petrochemical & Process, Gulf Professional Pub., 2001.
7
[2] T. Arbogast, M. F. Wheeler, and I. Yotov, Mixed finite elements for elliptic problems with tensor coefficients as
8 cell-centered finite di↵erences, SIAM J. Numer. Anal., 34 (1997), pp. 828–852.
9 [3] K. Aziz and A. Settari, Petroleum Reservoir Simulation, Springer Netherlands, 1979.
[4] J. Bear, Dynamics of Fluids in Porous Media, Dover, New York, 1972.
10
[5] J. H. Bramble, J. E. Pasciak, and O. Steinbach, On the stability of the l2 -projection in h1 (!), Mathematics of
11 Computation, 71 (2002), pp. pp. 147–156.

12 [6] F. Brezzi and M. Fortin, Mixed and hybrid finite element methods, vol. 15 of Springer Series in Computational
Mathematics, Springer-Verlag, New York, 1991.
13 [7] P. G. Ciarlet, The finite element method for elliptic problems, North-Holland Publishing Co., Amsterdam, 1978.
14 Studies in Mathematics and its Applications, Vol. 4.
[8] J. Crank and P. Nicolson, A practical method for numerical evaluation of solutions of partial di↵erential equations
15 of the heat-conduction type, Mathematical Proceedings of the Cambridge Philosophical Society, 43 (1947), p. 50–67.
16 [9] L. P. Dake, Fundamentals of reservoir engineering / L. P. Dake, Elsevier Scientific Pub. Co. ; distributors for the
U.S. and Canada Elsevier North-Holland Amsterdam ; New York : New York, 1978.
17
[10] C. N. Dawson and M. F. Wheeler, Two-grid methods for mixed finite element approximations of nonlinear parabolic
18 equations, in Domain decomposition methods in scientific and engineering computing (University Park, PA, 1993),
vol. 180 of Contemp. Math., Amer. Math. Soc., Providence, RI, 1994, pp. 191–203.
19
[11] J. J. Douglas, P. J. Paes-Leme, and T. Giorgi, Generalized Forchheimer flow in porous media, in Boundary value
20 problems for partial di↵erential equations and applications, vol. 29 of RMA Res. Notes Appl. Math., Masson, Paris,
1993, pp. 99–111.
21
[12] R. E. Ewing, R. D. Lazarov, J. E. Pasciak, and A. T. Vassilev, Mathematical modeling, numerical techniques, and
22 computer simulation of flows and transport in porous media, in Computational techniques and applications: CTAC95
(Melbourne, 1995), World Sci. Publ., River Edge, NJ, 1996, pp. 13–30.
23
[13] V. Girault and M. F. Wheeler, Numerical discretization of a Darcy-Forchheimer model, Numer. Math., 110 (2008),
24 pp. 161–198.
[14] L. Hoang and A. Ibragimov, Structural stability of generalized Forchheimer equations for compressible fluids in porous
25
media, Nonlinearity, 24 (2011), pp. 1–41.
26 [15] L. Hoang, A. Ibragimov, T. Kieu, and Z. Sobol, Stability of solutions to generalized Forchheimer equations of any
degree, J. Math. Sci., (2015). accepted.
27
[16] S. Irmay, On the theoretical derivation of Darcy and Forchheimer formulas, Transactions, American Geophysical Union,
28 39 (1958), pp. 702–707.
29 [17] C. Johnson, Numerical Solution of Partial Di↵erential Equations by the Finite Element Method, Dover Books on
Mathematics Series, Dover Publications, Incorporated, 2012.
30
[18] C. Johnson and V. Thomée, Error estimates for some mixed finite element methods for parabolic type problems,
31 RAIRO Anal. Numér., 15 (1981), pp. 41–78.
[19] T. Kieu, Existence of a solution for generalized Forchheimer flow in porous media with minimal regularity conditions,
32
Journal of Mathematical Physics, 61 (2020), p. 013507.
33 [20] T. Kieu, Solution of the mixed formulation for generalized Forchheimer flows of isentropic gases, Journal of Mathe-
matical Physics, 61 (2020), p. 081501.
34
[21] M.-Y. Kim and E.-J. Park, Fully discrete mixed finite element approximations for non-Darcy flows in porous media,
35 Comput. Math. Appl., 38 (1999), pp. 113–129.
36 [22] A. Logg, K.-A. Mardal, and G. N. Wells, eds., Automated Solution of Di↵erential Equations by the Finite Element
Method, vol. 84 of Lecture Notes in Computational Science and Engineering, Springer, 2012.
37
[23] M. Muskat, The flow of homogeneous fluids through porous media, McGraw-Hill Book Company, inc., 1937.
38 [24] S. Neuman, Theoretical derivation of darcy’s law, Acta Mechanica, 25 (1977), pp. 153–170.
39 [25] E. Palencia, Non-Homogeneous Media and Vibration Theory, Lecture Notes in Physics, Springer-Verlag, 1980.
[26] H. Pan and H. Rui, A block-centered finite di↵erence method for the Darcy-Forchheimer model, SIAM J. Numer. Anal.,
40 50 (2012), pp. 2612–2631.
41 [27] H. Pan and H. Rui, Mixed element method for two-dimensional Darcy-Forchheimer model, J. Sci. Comput., 52 (2012),
pp. 563–587.
42
[28] E.-J. Park, Mixed finite element methods for generalized Forchheimer flow in porous media, Numer. Methods Partial
43 Di↵erential Equations, 21 (2005), pp. 213–228.
44 [29] E.-J. Park, Mixed finite element methods for generalized Forchheimer flow in porous media, Numer. Methods Partial
Di↵erential Equations, 21 (2005), pp. 213–228.
45
46 the pump journal of undergraduate research 0 (2017), 000–000 15
47
48
49
50
1
2
3
4
5
[30] P.Basak, Non-darcy flow and its implications to seepage problems, Journal of the Irrigation and Drainage Division,
6 103 (1977), pp. 459–473.
7 [31] K. R. Rajagopal, On a hierarchy of approximate models for flows of incompressible fluids through porous solids,
Mathematical Models and Methods in Applied Sciences, 17 (2007), pp. 215–252.
8 [32] P. Raviart and J. Thomas, A Mixed Finite Element Method for Second Order Elliptic Problems, vol. 606, 11 2006,
9 pp. 292–315.
[33] P. A. Raviart and J. M. Thomas, A mixed finite element method for 2-nd order elliptic problems, in Mathemati-
10 cal Aspects of Finite Element Methods, I. Galligani and E. Magenes, eds., Berlin, Heidelberg, 1977, Springer Berlin
11 Heidelberg, pp. 292–315.
[34] P. A. Raviart and J. M. Thomas, A mixed finite element method for 2nd order elliptic problems, in Mathematical
12 aspects of finite element methods (Proc. Conf., Consiglio Naz. delle Ricerche (C.N.R.), Rome, 1975), Springer, Berlin,
13 1977, pp. 292–315. Lecture Notes in Math., Vol. 606.
[35] P.-A. Raviart and J. M. Thomas, Introduction à l’analyse numérique des équations aux dérivées partielles. Collection
14
Mathématiques Appliquées pour la Maı̂trise. Paris etc.: Masson. 224 p. (1983)., 1983.
15 [36] S. Whitaker, Flow in porous media i: A theoretical derivation of darcy’s law, Transport in Porous Media, 1 (1986),
pp. 3–25.
16
17
Akshaya Poonepalle
18
Lambert High School
19 805 Nichols Rd, Suwanee
20 Georgia 30024, U.S.A
21 E-mail: akshaya310512@gmail.com

22 Brent Shi
23 Atlanta International School
24 2890 N Fulton Dr NE, Atlanta
Georgia 30305, U.S.A
25 E-mail: brentshi3@gmail.com
26
27 Elly Kang
Marist School
28 3790 Ashford Dunwoody Rd NE, Atlanta
29 Georgia 30319, U.S.A
30 E-mail: ellyskang@gmail.com

31
Joseph Chin
32 North Gwinnett High School
33 20 Level Creek Rd Suwanee
Georgia 30024 , U.S.A
34
E-mail: ab94033693@gmail.com
35
36 Mannan Singh
Lambert High School
37
805 Nichols Rd, Suwanee
38 Georgia 30024, U.S.A
39 E-mail: mannanraj13@gmail.com

40
Prakhar Gupta
41 South Forsyth High School
42 585 Peachtree Pkwy, Cumming
Georgia 30041, U.S.A
43
E-mail: prakharg2805@gmail.com
44
45
46 the pump journal of undergraduate research 0 (2017), 000–000 16
47
48
49
50
1
2
3
4
5
Riya Malhotra
6
South Forsyth High School
7 585 Peachtree Pkwy, Cumming
8 Georgia 30041, U.S.A
E-mail: RiyaMalhotra3970@gmail.com
9
10 Sanjana Kadiyala
11 Milton High school
13025 Birmingham Hwy, Milton
12
Georgia 30004, U.S.A
13 E-mail: sanjana.kadiyalas@gmail.com
14
Sirihansika (Hansi) Thadiparthi
15
South Forsyth High School
16 585 Peachtree Pkwy, Cumming
17 Georgia 30041, U.S.A
E-mail: sirithadiparthi@gmail.com
18
19 Thinh Kieu
20 University of North Georgia
Gainesville Campus, 3820 Mundy Mill Rd., Oakwood
21
Georgia 30566, U.S.A
22 E-mail: thinh.kieu@ung.edu
23
Yeeun Kim
24
Gwinnett School of Mathematics, Science, and Technology
25 970 McElvaney Ln NW, Lawrenceville
26 Georgia 30044, U.S.A
E-mail: yeeunkimyeeunkim@gmail.com
27
28
29
Authors contributions
30
Thinh Kieu involved in Conceptualization (lead), Formal analysis (lead), Investigation (lead), Method-
31
ology (lead), Project administration (lead), Suppervision (lead), Validation (lead), Review and editing
32 (lead).
33 Yeeun Kim, Prakhar Gupta, Elly Kang, Joseph Chin involved in
34
Software (equal): Programming, software development; designing computer programs; implementation of
the computer code and supporting algorithms; testing of existing code components.
35 Visualization (equal): Preparation, creation and/or presentation of the published work, specifically visual-
36 ization/ data presentation.
37
Validation (equal): Verification, whether as a part of the activity or separate, of the overall replication/
reproducibility of results/experiments and other research outputs.
38 Investigation (equal): Conducting a research and investigation process, specifically performing the experi-
39 ments, or data/evidence collection.
40
Akshaya Poonepalle, Brent Shi, Mannan Singh, Riya Malhotra, Sanjana Kadiyala Sirihansika
Thadiparthi involved in
41 Resources (equal): Provision of study materials, computing resources.
42 Writing - Original Draft (equal): the derivations of the estimates presented in the paper, preparation,
43
creation and/or presentation of the published work, specifically writing the initial draft
Validation (equal): Verification, whether as a part of the activity or separate, of the overall replication/
44
45
46 the pump journal of undergraduate research 0 (2017), 000–000 17
47
48
49
50
 46
47
48
49
50
reproducibility of results/experiments and other research outputs.
Received: April 31, 2017 Accepted: June 31, 2017
Communicated by Some Editor

the pump journal of undergraduate research 0 (2017), 000–000 18