Two conservative difference schemes for Rosenau-Kawahara equation

Jinsong Hu1,2 ,Youcai Xu1 ,Bing Hu1 ,Xiaoping Xie1

1. School of Mathematics, Sichuan University, Chengdu 610064, P.R. China
2.School of Mathematics and Computer Engineering, Xihua University, Chengdu 610039, P.R. China

Abstract: Two conservative finite difference schemes for the numerical solution of the initial-
boundary value problem of Rosenau-Kawahara equation are proposed. The difference schemes
simulate two conservative quantities of the problem well. The existence and uniqueness of the
difference solution are proved. It is shown that the finite difference schemes are of second-order
convergence and unconditionally stable. Numerical experiments verify the theoretical results.

KeyWords: Rosenau-Kawahara equation; Difference scheme; Conservative rule; Convergence;

Stability

1 Introduction
In the study of compact discrete systems, the wave-wave and wave-wall interactions cannot be described
by the well-known KdV equation. To overcome this shortcoming of KdV equation, Rosenau proposed the
following called Rosenau equation in [1,2]

ut + u xxxxt + u x + uu x = 0. (1.1)

which is usually used to describe the dense discrete system and simulate the long-chain transmission model
through an L-C flow in radio and computer fields. Park proved the existence and uniqueness of solution to (1.1)
in [3]. However, it is difficult to find its analytical solution. Therefore, the numerical study of equation (1.1) is
very significant and attract many scholars (see e.g. [4, 5, 6, 7, 8, 9, 10]).
As the further consideration of nonlinear wave, Zuo obtained Rosenau-Kawahara equation by adding vis-
cous term +u xxx and u xxxxx to Rosenau equation(1.1) and studied the solitary solution and periodic solution
of this equation in [10]. In [11], Labidi and Biswas got the integral of Rosenau-Kawahara equation by using
Hes principle. Then the solitary solution and two invariance of a generalized Rosenau-Kawahara equation are
investigated in [12]. To our best knowledge, there is no study on the numerical method of Rosenau-Kawahara
equation. Therefore, we will study the difference approximate solution of Rosenau-Kawahara equation with
the initial data and boundary conditions as following.

ut + u xxxxt + u x + uu x + u xxx u xxxxx = 0. (1.2)

u(x, 0) = u0 (x), x [xL , xR ] (1.3)

u(xL , t) = u(xR , t) = 0, u x (xL , t) = u x (xR , t) = 0, u xx (xL , t) = u xx (xR , t) = 0, t [0, T ] (1.4)
where is a known constant and u0 (x) is a smooth function. When = 1, the solitary wave solution of(1.2)is
(see [10])
35 1
q
35 4 1
u(x, t) = ( + 205)sech [ 13 + 205(x 205t)]. (1.5)
12 156 2 13

1
The initial boundary value problem (1.2)-(1.4) is in accordance with the Cauchy problem of (1.2) whenxL 
0, xR  0.Hence, the boundary condition(1.4)is reasonable. It is easy to verify that (1.2)-(1.4) satisfy the
following conservative quantities (see [12]).
Z xR Z xR
Q(t) = u(x, t)dx = u0 (x)dx = Q(0), (1.6)
xL xL

E(t) = kuk2L2 + ku xx k2L2 = E(0). (1.7)

where Q(0), E(0) are both constants only depending on initial data.
It is well known that a reasonable difference scheme has not only high-accuracy but also can maintain
some physical properties of original problem. Lots of numerical experiments show that conservative difference
scheme can simulate the conservative law of initial problem well since it could avoid the nonlinear blow-up[13-
24]. Moreover, it is suitable to compute for long-time. Li and Vu-Quoc pointed in [14] that in some areas,
the ability to preserve some invariant properties of the original differential equation is a criterion to judge the
success of a umerical simulation. Therefore, constructing the conservative difference scheme is a significant
job.
The rest of this paper is organized as follows. We respectively propose a two-level nonlinear Crank-
Nicolson difference scheme and three-level linear difference scheme for initial boundary value problem (1.2)-
(1.4) in section 2 and in section 3. We analyze its two discrete conservative laws. The existence and uniqueness
of the difference solution are proved. It is shown that the finite difference schemes are of second-order conver-
gence and unconditionally stable. In section 4, we verify our theoretical analysis by numerical experiments.

2 Nonlinear Crank-Nicolson conservative difference scheme

In this section, we propose a two-level nonlinear Crank-Nicolson difference scheme and give the theoreti-
cal analysis.
In the rest of this paper, C denotes a general positive constant which may denote different value in different
occurrence.
2.1 Nonlinear difference scheme and its conservative law
Let h and be the uniform step size in the spatial and temporal direction, respectively. The interval
[xL , xR ] is divided into J equal parts,where J is a fixed positive integer. Denote h = xR x J , x j = xL + jh( j =
L

2, 1, 0, , J, J + 1, J + 2), tn = n(n = 0, 1, 2, , N, N = [ ]).Let u j be the difference approximation of

T n

u(x, t) at (x j , tn ),that is,unj u(x j , tn ). Denote Zh0 = {u = (u j )|u2 = u1 = u0 = u J = u J+1 = u J+2 = 0, j =

2, 1, 0, 1, , J, J + 1, J + 2} We define the difference operators, inner product, and norms that will be used
in this paper as follows:

unj+1 unj unj unj1 unj+1 unj1 un+1

j unj
(unj ) x = , (unj ) x = , (unj ) x = , (unj )t = ,
h h 2h
un+1
j un1
j n+ 12 un+1
j + unj un+1
j + un1
j
(unj )t = , uj = , u nj = ,
2 2 2
J1
X
hun , vn i = h unj vnj , kun k2 = hun , un i, kun k = max |unj |.
1 jJ1
j=1

Consider the following finite difference scheme for problem (1.2)-(1.4).

n+ 21 n+ 12 n+ 12 1 n+ 1 n+ 1 n+ 1
(unj )t + (unj ) xxxxt + (u j ) x + (u j ) xx x (u j ) xxxx x + {u j 2 (u j 2 ) x + [(u j 2 )2 ] x } = 0,
3

2
 ( j = 1, 2, , J 1, n = 1, 2, , N 1) (2.1)
u0j = u0 (x j ), ( j = 0, 1, 2, , J) (2.2)
n
u Zh0 , (un0 ) x = (unJ ) x = 0, (un0 ) xx = (unJ ) xx = 0, (n = 0, 1, 2, , N) (2.3)

The discrete boundary condition (2.3) is reasonable from the boundary condition (1.4).

Lemma 2.1 For any two discrete functions u, v Zh0 , we have

hu x , vi = hu, v x i, hu xx , vi = hu x , v x i

from summation by parts (see [22]). Thus,

hu x , ui = hu, u x i, hu xx , ui = hu x , u x i = ku x k2 (2.4)

And if (un0 ) xx = (unJ ) xx = 0,then

hu xxxx , ui = ku xx k2 .
The following theorem shows how the difference scheme (2.1)-(2.3) simulates the conservative law numeri-
cally.

Theorem 2.1Suppose that u0 H02 [xL , xR ], then the difference scheme (2.1)-(2.3) is conservative for dis-
crete energy, that is,
J1
X
Q =hn
unj = Qn1 = = Q0 , (2.5)
j=1

E n = kun k2 + kunxx k2 = E n1 = = E 0 . (2.6)

Proof:Multiplying h in the two sides of (2.1) and summing up for j from 1 to J 1,from boundary (2.3)
and Lemma 2.1, we obtain
J1 un+1 un
j j
X
h =0 (2.7)
j=1

From the definition of Qn , (2.5) is deduced from (2.7).

1
Taking the inner of (2.1) with 2un+ 2 (i.e., un+1 + un ), from boundary (2.3) and Lemma 2.1, we get
n+ 21 1 n+ 1 1 n+ 1 1 1 1
kun k2t + kunxx k2t + 2hu x , un+ 2 i + 2hu xx x2 , un+ 2 i 2hu xxxx
2
x
, un+ 2 i + 2h(un+ 2 ), un+ 2 i = 0, (2.8)

n+ 21 n+ 21 n+ 21 n+ 12 2
where (u j ) = 13 {u j (u j ) x + [(u j ) ] x }. On the other hand,

n+ 12 1
hu x , un+ 2 i = 0 (2.9)
n+ 1 1
hu xx x2 , un+ 2 i = 0 (2.10)
n+ 12 1
hu xxxx x
, un+ 2 i = 0 (2.11)

J1
1 1 1 X n+ 21 n+ 12 n+ 1 n+ 1
h(un+ 2 ), un+ 2 i = h {u j (u j ) x + [(u j 2 )2 ] x }u j 2
3 j=1
J1 J1
1 X n+ 12 2 n+ 21 1 X n+ 1 n+ 1
= h (u j ) (u j ) x + h [(u j 2 )2 ] x u j 2
3 j=1 3 j=1

3
 J1 J1
1 X n+ 1 n+ 1 1 X n+ 1 n+ 1
= h (u j 2 )2 (u j 2 ) x h (u j 2 )2 (u j 2 ) x = 0 (2.12)
3 j=1 3 j=1

Substituting (2.9)-(2.12) into(2.8), we get

(kun+1 k2 kun k2 ) + (kun+1

xx k ku xx k ) = 0.
2 n 2
(2.13)

From the definition of E n , we get (2.6) by deducing (2.13).

2.2 Solvability of the difference scheme

In order to prove the solvability of difference scheme, we present the following Brouwer fixed point
theorem [25].

Lemma 2.2 (Brouwer fixed point Theorem) Let H be a finite dimensional inner product space, suppose that
g : H H is continuous and there exists an > 0 such that hg(x), xi > 0 for all x H with kxk = . Then
there exists x H such that g(x ) = 0 and kx k .

Theorem 2.2 There exists un Zh0 (1 n N) that satisfies difference scheme (2.1)-(2.3).

Proof:We use the mathematical induction to prove our theorem.

Suppose that there exist u0 , u1 , , un satisfy difference scheme (2.1) for n N 1.Next we prove there
exists un+1 satisfies difference scheme.
Let g be a operator on Zh0 defined by

g(v) = 2v 2un + 2v xxxx 2unxxxx + v x + v xx x v xxxx x + (v), (2.14)

Taking the inner product of (2.14) with v , and noticing that

hv x , vi = 0, hv xx x , vi = 0, hv xxxx x , vi = 0, h(v), vi = 0,

we have

hg(v), vi = 2kvk2 2hun , vi + 2kv xx k2 2hunxx , v xx i

2kvk2 2kun k kvk + 2kv xx k2 2kunxx k kv xx k
2kvk2 (kun k2 + kvk2 ) + 2kv xx k2 (kunxx k2 + kv xx k2 )
kvk2 (kun k2 + kunxx k2 ) + kvnxx k2
kvk2 (kun k2 + kunxx k2 )

Therefore, for any v Zh0 ,if kvk2 = kun k2 + kunxx k2 + 1,then hg(v), vi > 0.From Lemma 2.2, there exists
v Zh0 such that g(v ) = 0. Choose un+1 = 2v un ,It is easy to verify that un+1 satisfies difference scheme
(2.1).
2.3 Convergence, stability and uniqueness of solution

Suppose v(x, t) is the solution to (1.2)-(1.4). Denote vnj = v(x j , tn ). Then the truncation error of
difference scheme (2.1)-(2.3) is :
n+ 12 n+ 21 n+ 21 n+ 21
rnj = (vnj )t + (vnj ) xxxxt + (v j ) x + (v j ) xx x (v j ) xxxx x + (v j ),

4
 ( j = 1, 2, , J 1, n = 1, 2, , N 1) (2.15)
v0j = u0 (x j ), ( j = 0, 1, 2, , J) (2.16)
n
v Zh0 , (vn0 ) x = (vnJ ) x = 0, (vn0 ) xx = (vnJ ) xx = 0, (n = 0, 1, 2, , N) (2.17)

By Taylor expansion, we know that rnj = O(2 + h2 ) as h, 0.

Lemma 2.3 Suppose u0 H02 [xL , xR ].Then the solution of problem (1.2)-(1.4) satisfies

kukL2 C, ku x kL2 C, ku xx kL2 C, kukL C, ku x kL C,

Proof:It follows from (1.7) that

kukL2 C, ku xx kL2 C.
Using Holder
inequality and Cauchy-Schwarz inequality, we can get
Z xR Z xR Z xR
kukL2 =
2
u x u x dx = uu x | xL
xR
uu xx dx = uu xx dx
xL xL xL

1
kukL2 ku xx kL2 (kuk2L2 + ku xx k2L2 ).
2
Thus,
ku x kL2 C,
which also yields kukL C, ku x kL C from Sobolev inequality.

Lemma 2.4 (Discrete Sobolev inequality [22]) There exist constants C1 and C2 such that

kun k C1 kun k + C2 kunx k.

Lemma 2.5(Discrete Gronwall inequality [22]) Assume that w(k), (k) re nonnegative gridding function-
sand (k) is increasing. If C > 0, and for any k
k1
X
w(k) (k) + C w(l),
l=0

then for any k

w(k) (k)eCk .

Theorem 2.3 Suppose that u0 H02 [xL , xR ].Then the solution of difference scheme (2.1)-(2.3) satisfies
kun k C, kunx k C, kunxx k C, thus kun k C, kunx k C(n = 1, 2, , N).

Proof:From (2.6), we have

kun k C, kunxx k C,

5
From (2.4) and by Cauchy-Schwarz inequality,then
1
kunx k2 kun k kunxx k (kun k2 + kunxx k2 ) C, (2.18)
2
Thus,
kunx k C,
Therefore, from Lemma 2.4, we obtainkun k C, kunx k C.
Remark 2.1: Theorem 2.3 imply that the solution of difference scheme (2.1)-(2.3) is unconditionally stable in
the sense of norm k k .
Theorem 2.4 Suppose u0 H02 [xL , xR ]. Then the solution of difference scheme (2.1)-(2.3) un converges to
the solution of (1.2)-(1.4) in the sense of norm k k , and the convergent rate is O(2 + h2 ).

Proof: Subtracting (2.1)-(2.3) from (2.15)-(2.17) and denoting enj = vnj unj , we have

n+ 12 n+ 12 n+ 21 n+ 21 n+ 21
rnj = (enj )t + (enj ) xxxxt + (e j ) x + (e j ) xx x (e j ) xxxx x + (v j ) (u j ),

( j = 1, 2, , J 1, n = 1, 2, , N 1) (2.19)
e0j = 0, ( j = 0, 1, 2, , J) (2.20)
e Zh0 ,
n
(en0 ) x = (enJ ) x = 0, (en0 ) xx = (enJ ) xx = 0, (n = 0, 1, 2, , N) (2.21)
1
Taking the inner product of (2.19) with 2en+ 2 , we get
1 n+ 12 1 n+ 1 1 n+ 1 1 1 1 1
hrn , 2en+ 2 i = ken k2t +kenxx k2t +he x , 2en+ 2 i+he xx x2 , 2en+ 2 ihe xxxx
2
x
, 2en+ 2 i+h(vn+ 2 )(un+ 2 ), 2en+ 2 i. (2.22)

Similarly to (2.9)-(2.11), we have

n+ 21 1
he x , 2en+ 2 i = 0
n+ 1 1
he xx x2 , 2en+ 2 i = 0
n+ 1 1
2
he xxxx x
, 2en+ 2 i = 0

So, (2.22) can be rewritten as

1 1 1 1
(ken+1 k2 ken k2 ) + (ken+1
xx k ke xx k ) = hr , 2e
2 n 2 n n+ 2
i h(vn+ 2 ) (un+ 2 ), 2en+ 2 i. (2.23)

As
1 1 1
h(vn+ 2 ) (un+ 2 ), 2en+ 2 i
J1 J1
2h X n+ 12 n+ 12 n+ 12 n+ 12 n+ 12 2h X n+ 21 2 n+ 1 n+ 1
= [v j (v j ) x u j (u j ) x ]e j [(v j ) (u j 2 )2 ] x e j 2
3 j=1 3 j=1
J1 J1
2h X n+ 12 n+ 12 n+ 1 n+ 1 n+ 1 2h X n+ 12 n+ 12 n+ 1 n+ 1
= [v j (e j ) x + e j 2 (u j 2 ) x ]e j 2 + [e j (v j + u j 2 )](e j 2 ) x ,
3 j=1 3 j=1

from Lemma 2.3 and Theorem 2.3, we obtain

n+ 21 n+ 12 n+ 21
|v j | C, |u j | C, |(u j ) x | C, ( j = 0, 1, 2, , J).

6
Using Cauchy-Schwarz inequality, we have
J1 J1
1 1 1 2 X n+ 1 n+ 1 n+ 1 2 X n+ 1 n+ 1
h(vn+ 2 ) (un+ 2 ), 2en+ 2 i C||h (|(e j 2 ) x | + |e j 2 |)|e j 2 | + C||h |e j 2 | |(e j 2 ) x |
3 j=1
3 j=1

1 n+ 12 2
C(ken+ 2 k2 + |e x k ) C(ken+1 k2 + ken k2 + ken+1
x k + ke x k ).
2 n 2
(2.24)
On the other hand,
1 1
hrn , 2en+ 2 i = hrn , en+1 + en i krn k2 + (ken+1 k2 + ken k2 ). (2.25)
2
Substituting (2.24) and (2.25) into (2.23), we get

(ken+1 k2 ken k2 ) + (ken+1 2 n 2

xx k ke xx k )

C(ken+1 k2 + ken k2 + ken+1

x k + ke x k ) + kr k .
2 n 2 n 2
(2.26)
Similarly to the deduction process of (2.18), one can get
1 n+1 2 1 n 2
ken+1 2
x k (ke k + ken+1
xx k ), ke x k (ke k + ke xx k ).
2 n 2 n 2
(2.27)
2 2
Therefore,(2.26) is changed into

(ken+1 k2 ken k2 ) + (ken+1 2 n 2

xx k ke xx k )

C[ken+1 k2 + ken k2 + ken+1

xx k + ke xx k ] + kr k .
2 n 2 n 2
(2.28)
Denoting Bn = ken k2 + kenxx k2 ,(2.28) is equivalent to

Bn+1 Bn C(Bn+1 + Bn ) + krn k2 ,

that is
(1 C)(Bn+1 Bn ) 2CBn + krn k2 .
Choosing sufficiently small such that 1 C > 0,we get

Bn+1 Bn CBn + Ckrn k2 . (2.29)

Summing up (2.29) from 0 to n 1, we get

n1
X n1
X
Bn B0 + C krl k2 + C Bl .
l=0 l=0

Noticing that
n1
X
krl k2 n max krl k2 T O(2 + h2 )2 ,
0ln1
l=0

and B0 = O(2 + h2 )2 . we have

n1
X
Bn O(2 + h2 )2 + C Bl .
l=0

From Lemma 2.5, we get

Bn O(2 + h2 )2 ,

7
that is
ken k O(2 + h2 ), kenxx k O(2 + h2 ).
Noticing equality (2.27), one can obtain
kenx k O(2 + h2 ).
From Lemma 2.4, we have
ken k O(2 + h2 ).
Similarly to Theorem 2.4, we have the following theorems.
Theorem 2.5 The solution to difference scheme( 2.1)-(2.3) is unique.

3 A linear conservative difference scheme

In this section, we propose a three-level linear conservative difference scheme for (1.2)-(1.4)and give the
theoretical analysis.
3.1 Linear difference scheme and its conservative law
Consider the following finite difference scheme for problem (1.2)-(1.4).
1
(unj )t + (unj ) xxxxt + (unj ) x + (unj ) xx x (unj ) xxxx x + [unj (unj ) x + (unj u nj ) x ] = 0,
3

( j = 1, 2, , J 1, n = 1, 2, , N 1) (3.1)
u0j = u0 (x j ), ( j = 0, 1, 2, , J) (3.2)
u Zh0 ,
n
(un0 ) x = (unJ ) x = 0, (un0 ) xx = (unJ ) xx = 0, (n = 0, 1, 2, , N) (3.3)

The discrete boundary condition (3.3) is reasonable from the boundary condition (1.4).
The following Theorem shows how the difference scheme (3.1)-(3.3) simulates the conservative law as
follows.
Theorem 3.1Suppose that u0 H02 [xL , xR ], then the difference scheme (3.1)-(3.3) is conservative for dis-
crete energy, that is,
J1 J1
h X n+1 h X n n+1
Qn = (u j + unj ) + u (u ) x = Qn1 = = Q0 , (3.4)
2 j=1 6 j=1 j j

1 1
E n = (kun+1 k2 + kun k2 ) + (kun+1 k2 + kunxx k2 ) = E n1 = = E 0 . (3.5)
2 2 xx
Proof:Multiplying h in both sides of (3.1) and summing up for j ,from boundary (3.3) and Lemma 2.1, we
obtain
J1 un+1 un1 J1
X j j 1 X n n
h + h u (u ) x = 0. (3.6)
j=1
2 3 j=1 j j

Noticing that
J1 J1 J1 J1 J1
X h X n n+1 h X n n1 h X n n+1 h X n1 n
h unj (unj ) x = u j (u j ) x + u j (u j ) x = u j (u j ) x u (u j ) x , (3.7)
j=1
2 j=1 2 j=1 2 j=1 2 j=1 j

From the definition of Qn , (3.4) is deduced from (3.6) and (3.7).

8
 Taking the inner product of (3.1) with 2un , from boundary (3.3) and Lemma 2.1, we obtain

kun k2t + kunxx k2t + 2hunx , u n i + 2hunxx x , u n i 2hunxxxx x , u n i + 2h(un , u n ), un i = 0, (3.8)

where (unj , u nj ) = 13 [unj (unj ) x + (unj u nj ) x ]. On the other hand,

hunx , u n i = 0 (3.9)
hunxx x , u n i = 0 (3.10)
hunxxxx x , u n i = 0 (3.11)

J1
1 X n n
h(un , u n ), u n i = h [u (u ) x + (unj u nj ) x ]unj
3 j=1 j j
J1 J1
1 X n n n 1 X
= h u j u j (u j ) x + h (unj u nj ) x u nj
3 j=1 3 j=1

J1 J1
1 X n n n 1 X n n n
= h u j u j (u j ) x h u u (u ) x = 0 (3.12)
3 j=1 3 j=1 j j j
Substituting (3.9)-(3.12) into(3.8), we obtain

(kun+1 k2 kun1 k2 ) + (kun+1

xx k ku xx k ) = 0.
2 n1 2
(3.13)

From the definition of E n , we get (3.5) by deducing (3.13).

3.2 Solvability of the difference scheme
Theorem 3.2 There exists unique solution for difference scheme(3.1)-(3.3).

Proof:Use the mathematical induction to prove it. It is obvious that u0 is uniquely determined by the initial
condition (3.2).We also can get u1 by (2.1)-(2.3). Now suppose u0 , u1 , , un (n N 1) is solved uniquely.
Consider the equation of (3.1) for un+1
1 n+1 1 n+1 1 1 n+1 1 n+1 1
u j + (u j ) xxxx + (un+1
j ) x + (u j ) xx x (u j ) xxxx x + (u j , u j ) = 0,
n n+1
(3.14)
2 2 2 2 2 2
Taking the inner product of (3.14) with un+1 ,from boundary condition (3.3)and Lemma 2.1, we get
1 n+1 2 1 n+1 n+1 1 1 1
ku k + kun+1
xx k + hu x , u
2
i + hun+1 , un+1 i hun+1 , un+1 i + h(un , un+1 ), un+1 i = 0, (3.15)
2 2 2 xx x 2 xxxx x 2
Similarly to(3.9)-(3.11), we have

x , u
hun+1 i = 0
n+1
(3.16)
xx x , u
hun+1 = 0
n+1
i (3.17)
xxxx x , u
hun+1 = 0
n+1
i (3.18)

Notice that
J1
1 1 X n n+1
h(un , un+1 ), un+1 i = h [u (u ) x + (unj un+1 n+1
j ) x ]u j
2 6 j=1 j j
J1 J1
1 X n n+1 n+1 1 X
= h u j u j (u j ) x + h (unj un+1 n+1
j ) x u j
6 j=1 6 j=1

9
 J1 J1
1 X n n+1 n+1 1 X n n+1 n+1
= h u j u j (u j ) x h u u (u j ) x = 0 (3.19)
6 j=1 6 j=1 j j

Substituting(3.16)-(3.19)into(3.15),we get

kun+1 k2 + kun+1
xx k = 0
2

that is, (3.14) only admit zero solution. Therefore, there exists an unique un+1
j that satisfies (3.1).

3.3 Convergence and stability of the difference scheme

Suppose v(x, t) is the solution to (1.2)-(1.4). Denote vnj = v(x j , tn ). Then the truncation error of difference
scheme (3.1)-(3.3) is as follows:

rnj = (vnj )t + (vnj ) xxxxt + (vnj ) x + (vnj ) xx x (vnj ) xxxx x + (vnj , v nj ),

( j = 1, 2, , J 1, n = 1, 2, , N 1) (3.20)
v0j = u0 (x j ), ( j = 0, 1, 2, , J) (3.21)
v Zh0 ,
n
(vn0 ) x = (vnJ ) x = 0, (vn0 ) xx = (vnJ ) xx = 0, (n = 0, 1, 2, , N) (3.22)

By Taylor expansion, we know that rnj = O(2 + h2 ) as h, 0.

Theorem 3.3 Suppose that u0 H02 [xL , xR ].Then the solution of difference scheme (3.1)-(3.3) satisfies
kun k C, kunx k C, kunxx k C, thus kun k C, kunx k C(n = 1, 2, , N).

Proof:From (3.5), we know

kun k C, kunxx k C,
By (2.18), one can get
kunx k C.
Then, from Lemma 2.4, we get kun k C, kunx k C.
Remark 3.1: Theorem 3.3 imply that the solution of difference scheme (3.1)-(3.3) is unconditionally stable in
the sense of norm k k .
Theorem 3.4 Suppose u0 H02 [xL , xR ]. Then the solution of difference scheme (3.1)-(3.3) un converges to
the solution of (1.2)-(1.4) in the sense of norm k k , and the convergent order is O(2 + h2 ).

Proof: Subtracting (3.1)-(3.3) from (3.20)-(3.22) and letting enj = vnj unj , we have

rnj = (enj )t + (enj ) xxxxt + (enj ) x + (enj ) xx x (enj ) xxxx x + (vnj , v nj ) (unj , u nj ),

( j = 1, 2, , J 1, n = 1, 2, , N 1) (3.23)
e0j = u0 (x j ), ( j = 0, 1, 2, , J) (3.24)
e Zh0 ,
n
(en0 ) x = (enJ ) x = 0, (en0 ) xx = (enJ ) xx = 0, (n = 0, 1, 2, , N) (3.25)

Taking the inner product of the two sides of(3.23)with 2en ,and using boundary condition(3.25)and Lemma 2.1,
we obtain

hrn , 2en i = ken k2t + kenxx k2t + henx , 2en i + henxx x , 2en i henxxxx x , 2en i + h(vn , v n ) (un , u n ), 2en i. (3.26)

10
Similarly to (3.9)-(3.11), we have

henx , 2en i = 0
henxx x , 2en i = 0
henxxxx x , 2en i = 0

Then(3.26)is changed into

(ken+1 k2 ken1 k2 ) + (ken+1

xx k ke xx k ) = 2hr , 2
2 n1 2 n
en i + 2h(vn , v n ) (un , u n ), 2en i. (3.27)

By Lemma 2.3 and Theorem 3.3, it show that

|vnj | C, |unj | C, |(unj ) x | C, ( j = 0, 1, 2, , J).

Using Cauchy-Schwarz inequality, we have

h(vn , v n ) (un , u n ), 2en i

J1 J1
2h X n n 2h X
= [v j (v j ) x unj (unj ) x ]enj [(vnj v nj ) x (unj u nj ) x ]enj
3 j=1
3 j=1
J1 J1
2h X n n 2h X
= [v j (v j ) x unj (unj ) x ]enj + (vnj v nj unj u nj )(enj ) x
3 j=1
3 j=1
J1 J1
2h X n n 2h X
= [e j (v j ) x + unj (enj ) x ]enj + (enj v nj + unj e nj )(enj ) x
3 j=1
3 j=1
C(ken k2 + ken k2 + |enx k2 )

C(ken+1 k2 + ken k2 + ken1 k2 + ken+1

x k + ke x k ).
2 n1 2
(3.28)
On the other hand,
1
hrn , 2en i = hrn , en+1 + en1 i krn k2 + (ken+1 k2 + ken1 k2 ). (3.29)
2
Substituting (3.28) -(3.29) into (3.27), we could obtain

(ken+1 k2 ken1 k2 ) + (ken+1 2 n1 2

xx k ke xx k )

C(ken+1 k2 + ken k2 + ken1 k2 + ken+1

x k + ke x k + ke x k ) + 2kr k .
2 n 2 n1 2 n 2
(3.30)
Using the deduce process which is similar to (2.18), we have
1 n+1 2 1 n 2 1 n1 2
ken+1 2
x k (ke k + ken+1
xx k ), ke x k (ke k + ke xx k ), ke x k (ke
2 n 2 n 2 n1 2
k + ken1 2
xx k ). (3.31)
2 2 2
Then(3.30)is changed into

(ken+1 k2 ken1 k2 ) + (ken+1 2 n1 2

xx k ke xx k )

C[ken+1 k2 + ken k2 + ken1 k2 + ken+1

xx k + ke xx k + ke xx k ] + 2kr k .
2 n 2 n1 2 n 2
(3.32)
Let Bn = ken+1 k2 + ken+1
xx k + ke k + ke xx k ,(3.32)can be rewritten as follows:
2 n 2 n 2

Bn Bn1 C(Bn + Bn1 ) + 2krn k2 ,

11
that is
(1 C)(Bn Bn1 ) 2CBn1 + 2krn k2 .
Choosing small enough such that 1 C > 0,then

Bn Bn1 CBn1 + Ckrn k2 . (3.33)

Summing up (3.33) from 1 to n, we obtain

n
X n1
X
Bn B0 + C krl k2 + C Bl .
l=1 l=0

Noticing that
n1
X
krl k2 n max krl k2 T O(2 + h2 )2 ,
0ln1
l=0
Firstly selecting other two-level scheme with second-order convergence (such as the scheme (2.1)-(2.3)) to get
u1 , we have
B0 = O(2 + h2 )2 .
Therefore
n1
X
B O( + h ) + C
n 2 2 2
Bl .
l=0
Using Lemma 2.5,we obtain
Bn O(2 + h2 )2 ,
that is
ken k O(2 + h2 ), kenxx k O(2 + h2 ).
Combining with(3.31),we get
kenx k O(2 + h2 ).
Then by Lemma 2.4, it shows that
ken k O(2 + h2 ).
Similarly to Theorem 3.4, we have the following theorem.

Theorem 3.5 Under the conditions of Theorem 3.4, the solution of difference scheme (3.1)-(3.3) un is stable
in the sense of norm k k .

4 Numerical simulations
In this section,let = 1, xL = 40, xR = 100, T = 40, and
35
q
35 1
u0 (x) = ( + 205)sech4 ( 13 + 205x)
12 156 12
We denote the nonlinear Crank-Nicolson conservative difference scheme (2.1)-(2.3) as Scheme I and the linear
three-level conservative difference scheme (3.1)-(3.3) as Scheme II. For some different values of and h,we list
errors of Scheme I and Scheme II at several different time in Table 1 and Table 2 respectively.We verify the two
order accuracy of the difference scheme in Table 3 and Table 4 by using the method of [26,27]. The numerical
simulation of two conservative quantities (1.6) and (1.7) is listed in Table 5 and 6.

12
 Table 1: The errors of Scheme I at various time step with various h and
= h = 0.1 = h = 0.05 = h = 0.025
n
ke k n
ke k n
ke k n
ke k n
ke k ken k
t=10 2.159730e-4 7.520810e-5 5.401591e-5 1.880684e-5 1.354230e-5 4.702168e-6
t=20 4.160331e-4 1.421127e-4 1.040460e-4 3.554032e-5 2.603381e-5 8.886828e-6
t=30 5.977491e-4 1.996835e-4 1.494989e-4 4.994370e-5 3.739276e-5 1.248752e-5
t=40 7.633356e-4 2.502663e-4 1.909241e-4 6.260150e-5 4.774838e-5 1.565301e-5

Table 2: The errors of Scheme II at various time step with various h and
= h = 0.1 = h = 0.05 = h = 0.025
n
ke k n
ke k n
ke k n
ke k n
ke k ken k
t=10 4.302763e-4 1.477136e-4 1.076811e-4 3.697042e-5 2.693880e-5 9.259288e-6
t=20 8.320777e-4 2.820308e-4 2.082744e-4 7.060528e-5 5.207086e-5 1.767754e-5
t=30 1.199482e-3 3.985869e-4 3.002761e-4 9.979291e-5 7.504636e-5 2.497799e-5
t=40 1.535826e-3 5.014481e-4 3.845124e-4 1.255601e-4 9.606941e-5 3.141830e-5

Table 3: The verification of the second convergence of Scheme I

ken (h,)k ken (h,)k
ke2n (h/2,/2)k ke2n (h/2,/2)k
= h = 0.1 = h = 0.05 = h = 0.025 = h = 0.1 = h = 0.05 = h = 0.025
t=10 3.998318 3.988313 3.998969 3.999234
t=20 3.998548 3.996548 3.998630 3.999538
t=30 3.998359 3.998072 3.998171 3.999489
t=40 3.998109 3.998457 3.997769 3.999324

Table 4: The verification of the second convergence of Scheme II

ken (h,)k ken (h,)k
ke2n (h/2,/2)k ke2n (h/2,/2)k
= h = 0.1 = h = 0.05 = h = 0.025 = h = 0.1 = h = 0.05 = h = 0.025
t=10 3.995835 3.997182 3.995455 3.993009
t=20 3.995103 3.999826 3.994473 3.994248
t=30 3.994599 4.001208 3.994140 3.995232
t=40 3.994217 4.002444 3.993690 3.996400

Table 5: When = 1,numerical simulations on the conservation invariant Qn

= h = 0.1 = h = 0.05 = h = 0.025
Scheme I Scheme II Scheme I Scheme II Scheme I Scheme II
t=0 4.120893241330 4.120918187462 4.120893214999 4.120899452450 4.120893201770 4.120894761190
t=10 4.120892312503 4.120918835632 4.120892200166 4.120898816174 4.120892129735 4.120894227434
t=20 4.120893031520 4.120921893266 4.120892362862 4.120899447408 4.120892167548 4.120894662524
t=30 4.120893107802 4.120925120669 4.120892370454 4.120900136940 4.120892149140 4.120895106600
t=40 4.120892997828 4.120924996792 4.120892342248 4.120900073781 4.120892155018 4.120895399389

13
 Table 6: When = 1,numerical simulations on the conservation invariant E n
= h = 0.1 = h = 0.05 = h = 0.025
Scheme I Scheme II Scheme I Scheme II Scheme I Scheme II
t=0 0.836201094485 0.836201094485 0.836201181724 0.836201181724 0.836201203535 0.836201203535
t=10 0.836201095943 0.836201096547 0.836201194492 0.836201202559 0.836201310312 0.836201469358
t=20 0.836201095943 0.836201096662 0.836201194483 0.836201206568 0.836201308626 0.836201594118
t=30 0.836201095952 0.836201096780 0.836201194484 0.836201210387 0.836201309421 0.836201718674
t=40 0.836201095955 0.836201096906 0.836201194369 0.836201214174 0.836201309604 0.836201843455

Finally, a numerical simulation figure comparison of u(x, t) at various time step with = h = 0.05 is as fol-
low:(see Figures 1,2.)

t=0
t=20
0.3
t=40

0.25

0.2

0.15

0.1

0.05

40 20 0 20 40 60 80 100
Figure 1:When = 1 and = h = 0.05, a numerical simulation figure on u(x, t) of Scheme I at various times

14
 t=0
t=20
0.25 t=40

0.2

0.15

0.1

0.05

0.05
40 20 0 20 40 60 80 100

Figure2:When = 1 and = h = 0.05, a numerical simulation figure on u(x, t) of Scheme II at various times

Numerical simulations show that the finite difference scheme I and II in this paper are efficient. The calcu-
lation results of scheme I are slightly better than scheme II. But iterative numerical calculation is not required,
for the scheme II is linear, so it saves computing time.

Acknowledgments: This work is supported by Scientific Research Found of Sichuan Provincial Education
Department (no. 11ZB009).

References
[1] P. Rosenau, A quasi-continuous description of a nonlinear transmission line , Physica Scripta, 34(1986):
827-829.

[2] P. Rosenau, Dynamics of dense discrete systems, Progr.Theoret. Phys. 79 (1988) 1028-1042.

[3] M.A. Park, On the Rosenau equation, Mat. Apl. Comput. 9 (1990) 145-152.

[4] S.K.Chung,S.N.Ha.Finite element galerkin solution for the Rosenau equation.Applicable Analysis.
54(1994): 39-56.

[5] Khaled Omrani,Faycal Abidi,Talha Achouri and Noomen Khiari.A new conservative finite difference
scheme for the Rosenau eqation.Appl Math Comput,201(2008):35-43.

[6] S. K. Chung. Finite difference approximate solutions for the Rosenau equation . Applicable Analysis,
1998,69(1-2):149-156.

[7] S. K. Chung and A. K. Pani, Numerical Methods for the Rosenau Equation, Appl. Anal., 77(2001): 351-
369.

[8] S. A. Manickam, A. K. Pani and S. K. Chung, A second-order splitting combined with orthogonal cubic
spline collocation method for the Rosenau equation, Numer. Meth. Partial Diff. Eq. 14 (1998): 695-716.

15
 [9] Y. D. Kim and H. Y. Lee, The convergence of finite element Galerkin solution for the Rosenau equation.
Korean J. Comput and Appl. Math 5 (1998) 171-180.

[10] J..M. Zuo. Solitons and periodic solutions for the Rosenau-KdV and Rosenau-Kawahara equations, Appl.
Math. Comput. 215 (2009) 835-840.

[11] M. Labidi, A. Biswas. Application of Hes principles to Rosenau-Kawahara equation. Math. Eng. Sci.
Aerosp. MESA 2, 2 (2011), 183-197.

[12] A. Biswas, H. Triki, and M. Labidi. Bright and Dark Solitons of the Rosenau Kawahara Equation with
Power Law Nonlinearity. Physics of Wave Phenomena. 19 (2011), 24-29.

[13] Tingchun Wang, Luming Zhang, Fangqi Chen. Conservative schemes for the symmetric regularized long
wave equations. Appl Math Comput, 190(2007): 1063-1080.

[14] S. Li and L. Vu-Quoc, Finite difference calculus invariant structure of a class of algorithms for the non-
linear Klein-Gordon equation, SIAM Journal on Numerical Analysis, vol. 32, no. 6, pp. 1839-1875,1995.

[15] Q. Chang, E. Jia, and W. Sun, Difference schemes for solving the generalized nonlinear Schr?dinger
equation, Journal of Computational Physics, vol. 148, no. 2, pp. 397-415, 1999.

[16] T-C.Wang and L.-M. Zhang, Analysis of some new conservative schemes for nonlinear Schr?dinger
equation with wave operator, Applied Mathematics and Computation, vol. 182, no. 2, pp. 1780-1794,
2006.

[17] T. Wang, B. Guo, and L. Zhang, New conservative difference schemes for a coupled nonlinear Schr?ding
er system, Applied Mathematics and Computation, vol. 217, no. 4, pp. 1604-1619, 2010.

[18] L. Zhang, A finite difference scheme for generalized regularized long-wave equation, Applied Mathe-
matics and Computation, vol. 168, no. 2, pp. 962-972, 2005.

[19] F. Zhang and L. V?azquez, Two energy conserving numerical schemes for the sine-Gordon equation,
Applied Mathematics and Computation, vol. 45, no. 1, pp. 17-30, 1991.

[20] Y. S.Wong, Q. Chang, and L. Gong, An initial-boundary value problem of a nonlinear Klein-Gordon
equation, Applied Mathematics and Computation, vol. 84, no. 1, pp. 77-93, 1997.

[21] Q. S. Chang, B. L. Guo, and H. Jiang, Finite difference method for generalized Zakharov equations,
Mathematics of Computation, vol. 64, no. 210, pp. 537-553, 1995.

[22] J. S. Hu and K. L. Zheng, Two conservative difference schemes for the generalized Rosenau equation,
Boundary Value Problems, vol. 2010, Article ID 543503, 18 pages, 2010.

[23] X. Pan, L. Zhang. On the convergence of a conservative numerical scheme for the usual Rosenau-RLW
equation , Applied Mathematical Modelling 36 (2012) 3371-3378.

[24] X. Pan, L. Zhang. Numerical Simulation for General Rosenau-RLW Equation: An Average Linearized
Conservative Scheme Mathematical Problems in Engineering Volume 2012, Article ID 517818, 15 pages.

[25] Browder F E.Existence and uniqueness theorems for solutions of nonlinear boundary value problems.
Proceedings of Symposia in Applied Mathematics,1965,17:24-49.

[26] T. Wang, B. Guo, A robust semi-explicit difference scheme for the Kuramoto-Tsuzuki equation, J. Com-
put. Appl. Math. 233 (2009) 878-888.

16
[27] Jinsong Hu, Bing Hu, and Youcai Xu.C-N Difference Schemes for Dissipative Symmetric Regularized
Long Wave Equations with Damping Term, Mathematical Problems in Engineering, 2011 (2011), Article
ID 651642, 16 pages.

17