Lax-Wendroff Difference Scheme with Richardson Extrapolation Method for

One Dimensional Wave Equation Subjected To Integral Condition

Kedir Aliyi Koroche

Department of Mathematics, College of Natural and Computational Sciences, Ambo University,
Ambo, Ethiopia
Email address: kediraliyi39@gmail.com (Kedir A.)

Abstract

In this paper, the Lax-Wendroff difference scheme has been presented for solving the one-
dimensional wave equation with integral boundary conditions. First, the given solution domain
is discretized and the derivative involving the spatial variable is replaced by the central finite
difference approximation of functional values at each grid point by using Taylor series
expansion. Then, for solving the resulting second-order linear ordinary differential equation, the
displacement function is discretized in the direction of a temporal variable by using Taylor series
expansion and the Lax-Wendroff difference scheme is developed, then it gives a system of
algebraic equations. The derivative of the initial condition is also discretized by using the central
finite difference method. Then the obtained system of algebraic equations is solved by the matrix
inverse method. The stability and convergent analysis of the scheme are investigated. The
established convergence of the scheme is further accelerated by applying the Richardson
extrapolation which yields fourth-order convergent in spatial variable and sixth-order
convergent in a temporal variable. To validate the applicability of the proposed method, three
model examples are considered and solved for different values of the mesh sizes in both
directions. Numerical results are presented in tables in terms of maximum absolute error, L2
and L∞ norm. The numerical results presented in tables and graphs confirm that the
approximate solution is in good agreement with the exact solution.
Keywords: Hyperbolic equation, one-dimensional wave equation, Lax-Wendroff difference
scheme, Taylor series methods, Richardson extrapolation, stability, and convergence
1. Introduction

The hyperbolic partial differential equation with an integral condition arises in many physical
phenomena [10]. Over the last few years, it has become increasingly apparent that many physical
phenomena can be described in terms of hyperbolic partial differential equations with an integral
condition replacing the classic boundary condition [4]. These types of equation appear in a
variety of physical problems such as in the study of thermoelasticity and plasma physics, and
chemical heterogeneity [2], [4], acoustic waveguides [20], heterogeneous physical properties
and/or complex geometry, seismology (study of earthquakes, regional and global seismology,
accurate calculation of synthetic seismograms [23], water waves, sound waves, radio waves,
lightwave and seismic waves [14]. Hyperbolic partial differential equations play a very important
role in modern applied mathematics due to their deep physical background. This hyperbolic
differential equation subject to an integral conservation condition in one space dimensional,
feature in the mathematical modeling of many phenomena [7]. It also arises in a broad spectrum
of applications where wave motion is involved, for example; optics, acoustics, oil and gas
dynamics, and vibrating string to name but a few [14]. Waves have distinct properties specific to
their type but also exhibit characteristics in common with more abstract waves such as
sound waves and light waves [6],[8], [13]. Seismic waves may be simulated by a viscous type of
wave equation which is known to be difficult to obtain the analytical solution [14]. This
is where the numerical solution to such types of equations is needed to solve
practical problems, [14],[ 19].

Due to the wide range of the application of the wave equation, several numerical methods have
been developed. Even though many numerical methods were applied to solve these types of
equations. Accordingly, more efficient and simpler numerical methods are required to solve
wave equations. The FDM is becoming increasingly more important in the seismic industry and
structural modeling due to its relative accuracy and computational efficiency [14]. Numerical
modeling of wave propagation in the presence of surface topography requires the incorporation
of non-flat boundaries. Embedding irregular boundaries into a finite-difference method is an old
problem [11]. It can be avoided altogether with finite-element or finite-volume or mesh-free
methods, but usually at a higher computational cost. Given the existing, highly optimized finite-
difference code base, a local modification of finite-difference stencils close to the irregular
boundary is more attractive [9]. An obvious approach for dealing with topography is the
introduction of an air layer. The extreme density contrast may require some smoothing to
preserve stability [3], [12]. The presence of an airwave can be suppressed by setting the sound
speed in the air equal to that at the surface, using constant extrapolation [9]. Again the numerical
methods have been widely implemented to model acoustic and hydrodynamic waves [1]. This
method of solving for the elastic wave eigenmodes in acoustic waveguides of an arbitrary cross-
section is presented. Operating under the assumptions of linear, isotropic materials, it utilizes a
finite-difference method on a staggered grid to solve for the acoustic Eigenmodes (field and
frequency) of the vector-field elastic wave equation with a given propagation constant [20].
Therefore growing attention is being paid to the development; analysis and implementation of
numerical methods for the solution of hyperbolic partial differential equations. Most of the
researchers have studied the numerical solutions of the 1D wave equation. Shazalina et al [24]
presented numerical techniques based Cubic Trigonometric B-spline Approach solving the 1D
wave equation. Ang [22] solved the same problem using a scheme based on an
integrodifferential equation and local interpolating functions. Dehghan [21] presented numerical
techniques based on the three-level explicit finite difference schemes for solving the 1D
hyperbolic partial differential equation. He used three-level techniques are based on two second-
order (one explicit and one weighted) schemes and a fourth-order technique (a weighted
explicit). Also. Saadatmandi [27] is developed the shifted Legendre Tau technique for the
solution of the studied model. The approach in their work consists of reducing the problem to a
set of algebraic equations by expanding the approximate solution as a shifted Legendre function
with unknown coefficients. The integral and derivative operational matrices are given. These
matrices together with the tau method are then utilized to evaluate the unknown coefficients of
shifted Legendre functions. Mehdi Dehghan and Ali Shokri [10] are presented numerical
techniques based on the mesh less schemes for solving the one-dimensional wave equation with
an integral condition using radial basis functions. Wang et. al. [17] presented the solution of the
one-dimensional wave Equation by using the Lagrangian meshfree finite difference particle
method with variable smoothing length. Ang [22] developed a numerical technique based on an
integrodifferential equation and local interpolating functions for solving the one-dimensional
wave equation subject to a nonlocal conservation condition and suitably prescribed initial
boundary conditions. Ramezani et.al. [28] Combined finite difference and spectral methods to
solve the one-dimensional wave equation with an integral condition. They used that the time
variable is approximated using a finite difference scheme. But the spectral method is employed
for discretizing the space variable. The main idea behind their approach is that high-order results
can be obtained. Hikmet et.al [7] present a numerical solution of the One-dimensional wave
equation with an Integral conservation condition is presented by the method of the non-
polynomial cubic spline. Beilin [29] presented the existence and uniqueness of the solution of the
one-dimensional wave equation with integral boundary conditions. Recently, much attention has
been paid in the literature to the development, analysis, and implementation of accurate methods.
However, still, the accuracy and stability of the method need attention because the treatment of
the method used to solve one-dimensional wave equation is not trivial distribution. Even though
the accuracy and stability of the aforementioned methods need attention, they require large
memory and long computational time. So the treatments this method presents severe difficulties
that have to be addressed to ensure the accuracy and stability of the solution. To this end, this
paper aims to develop an accurate and stable Lax-Wendroff difference scheme with Richardson
extrapolation numerical method that is capable of solving a one-dimensional wave equation with
integral boundary conditions and approximate the exact solution. The convergence present
method has been shown in the sense of maximum absolute error, L∞∧L2 norm . So that the local
behavior of the solution is captured and the agreement of numerical solution with the analytical
solution is showed by table and graph. The consistency, stability, and convergent analysis of the
present method are also established.

Statement of the problem

Consider that the following linear one-dimensional wave equation with integral boundary
condition considered in [2, 5, 10, 16] given by:
utt −uxx =q( x , t),( x ,t ) ∈ ( a , b ) × ( 0 ,T ) (1)
with initial and boundary condition respectively
u ( x , 0 )=ω 1 ( x ) ,ut ( x , 0 )=ω2 ( x ) , a ≤ x ≤ b,
b
u ( a , t )=ϕ 1 (t) , ∫ u ( x ,t )=ϕ 2(t),0 ≤ t ≤ T (2)
a

Whereq (x ,t) ,ω 1 ( x ) , ω 2 ( x ) , ϕ 1 ¿ ) and ϕ 2 (t) are smooth function on[ a , b ] × [ 0 , T ]. Now we define
a mesh size h and k and the constant grid point by drawing equidistance horizontal and vertical
line of distance ‘ h ’ and ‘ k ’ respectively in ‘ x ’ and ‘ t ’ direction. These lines are called gridlines
and the point at which they interacting is known as the mesh point. The mesh point that lies at
end of the domain is called the boundary point. The points that lie inside the region are called
interiors points. The goal is to approximate the solution ‘ u j n ’ at the interior mesh points. Hence
we discretized the solution domain as:
a=x 0 < x 1< x 2 <…< x M =b
0=t 0< t 1 <t 2< …<t N =T (3)
Where x j +1=x j + jh and t n+1=t n +nk , ¿ 0 ( 1 ) M ,n=0 ( 1 ) N . M and N are the maximum numbers
of grid points respectively in the x and t direction.
Then the present paper is organized as follows. Section two is Methodology (Methods), section
three is Stability and convergence analysis, section four is Numerical results. Section five is
Discussion; section six is the conclusion of the present framework.
2. Methods
In this paper, the Lax-Wendroff difference scheme with Richardson extrapolation numerical
method is developed to solve a one-dimensional wave equation with integral boundary
conditions given in Eq. (1). Hyperbolic equations such as the wave equation in Eq. (1) have two
derivatives in time and two in space. An initial condition (both function and its derivative) is
required and boundary conditions on both sides of the domain. Contrast this with Laplace’s
equation for the gravitational potential (∇ .H = 0) which is an elliptic partial differential equation.
In this case, conditions on the entire boundary are needed and specify the solution everywhere.
Either the function or its derivative must be specified on each of the boundaries and changing the
conditions at one point will change the solution everywhere.
Lax Wendorff Scheme
A numerical technique was proposed in 1960 by P.D. Lax and B. Wendroff [30],[31], [32] for
solving partial differential equations and systems numerically. Despite the impressive
developments in numerical methods for partial differential equations from the 1970s onwards, in
which the Lax Wendroff method has played a historic role, they are presently (1998) substantial
research activities aimed at further improvements of methods [32]. Lax Wendroff’s method is
also explicit but needs improvement in accuracy in time. This method is an example of explicit
time integration where the function that defines governing equation is evaluated at the current
time [32]. Purpose of this method to achieve good enough accuracy in time.

Thus we apply Taylor series expansion for both spatial and time derivative of Eq. (1) to develop the present

numerical method. Assuming that u(x , t) has continuous higher order partial derivative on the
p
∂ u p
region [a , b] x [0 ,T ]. For the sake of simplicity, we useu ( x j , t n ) =u jn, p
=∂x u j n and
∂x
p
∂ u p
p
=∂t u j n for p ≥1 is pth order derivatives.
∂t
Spatial Discretization
By using Taylor series expansion, we have

2 3
h 2 h 3
u j+1 n=u j n +h ∂ x u j n+ ∂x u j n + ∂ x u j n +…
2! 3!

2 3
h 2 h 3
u j−1n =u j n −h ∂ x u j n + ∂ x u j n− ∂x u j n +… (4)
2! 3!

Subtract the second equation from the first equation, and adding the first equation to the second
equation of Eq.(4), respectively we obtain the central finite difference scheme of the first and
second-order of u(x , t) concerning spatial variable given by:

u j +1−u j−1
∂ x u x= +τ 2
2h

2 u j +1−2 u j n +u j−1 n
∂ x u j n= 2
+ τ2 (5)
h

2 2
−h 3 −h 4
Where τ 1 = ∂ x u j n τ 2= ∂ x u j n are respectively their local truncation error terms. Now
6 12
substituting the second equation of Eq. (5) into Eq. (1) we obtain the system of second-order
ordinary differential equation in the temporal variable given in the form:

2
d u( x j ,t ) u j+1−2 u j n+u j−1 n
2
= 2
+ q ( x j , t )+ τ 2 (6)
dt h

Subject to the initial and boundary conditions

du ( x j ,0 )
u ( x j , 0 ) =ω1 ( x j ) , =ω2 ( x ) , a ≤ x j ≤b ,
dt
b
u ( a , t )=ϕ 1 (t) , ∫ u ( x j , t )=ϕ2 (t ), 0 ≤ t ≤T (7)
a

Temporal Discretization
Since Lax Wendroff’s method is an explicit method that uses to improvement in accuracy in
time. This method is an example of explicit time integration. So to discretize the derivative
involve with the temporal variable and develop the lax Wendorff scheme, first we expand the
functional value u ( x , t +k ) =u j n +1 by using Taylor series expansion at ( x j , t n )as follow:

2 3 4
k 2 k 3 k 4
u j n+1=u j n +k ∂t u j n + ∂t u j n + ∂t u j n + ∂ t u j n +… (8)
2! 3! 4!

Since the governing hyperbolic partial differential equation in (1) is utt =uxx + q(x , t). Then
assuming thatq ( x , t ) =0,and utt =uxx , this implies thatut =u x .Thus by using this idea and
truncating third-order derivative and above, from Eq. (8) we obtain:

2
k 2
u j n+1=u j n +k ∂ x u j n+ ∂x u j n + τ 3 (9)
2!

3
k 3
where τ 3 = ∂t u j n is the local truncation of the expiation series. Now substituting the first
3!
equation of Eq. (5) and Eq.(6) into Eq.(9) we obtain,

2
u j n+1=u j n +α ( u j+1 −u j−1 ) +2 α ( u j+1 −2u j n +u j−1 n ) +G j n + τ j n (10)
k 2
where α = ,G jn =2 α q ( x j , t n) and τ j n=α τ 1 +2 α 2 τ 2 +τ 3 is local truncation error term of
2h
obtained difference equation for the full discretization of the one-dimensional hyperbolic partial
differential equation given in (1). This local truncation error is:

( )
2 3
−α 3 α 4 k 3
∂x u j n− ∂ x u j n + ∂t u j n=Ο ( h + k )
2 2 3
τ j n=h (11)
6 6 6
Thus the desired lax-Wendorff scheme that we use to solve second order linear one-dimensional
wave equation in Eq.(1) is obtained by truncating the truncation error term τ j n, from difference
equation in Eq.(10), the scheme is:
2
u j n+1=u j n +α ( u j+1 n −u j−1 n ) +2 α ( u j+1 n−2 u j n+ u j−1 n ) +G j n
Or
u j n+1=α ( 1+ 2 α ) u j +1 n+ ( 1−4 α 2) ∂ 3t u j n+ α ( 2 α −1 ) u j−1 n +G j n (12)

with the order of accuracy isΟ ( h2 +k 3 ) .

Implementation of Initial Condition
Since the initial condition in Eq. (7) is existing as the derivative in the temporal variable. so we
discretize this initial condition by using central finite difference approximation for n=0. It is
given by.
u j , 1−u j ,−1
=ω2 ( x j )
2k
u j+1 ,1=u j ,−1+2 k ω 2 ( x j ) (13)
In Eq.(13) both u j+1 ,1 and u j ,−1 are unknown value. So it is difficult to use this discrete scheme
of the initial condition. Therefore to use this discrete scheme we must avoid unknown old value
u j ,−1 from Eq. (13). To avoid this value let us substitute n=0 into central finite difference
discretization equation of one-dimensional wave equation in Eq. (1) we obtain:
2 2
u j ,1=2 u j ,0−u j ,−1 +α ( u j +1 ,0−2 u j ,0 +u j−1 ,0 ) +k q (x j , t n ) (14)

Now adding Eq.(13) into Eq.(14) and usingu ( x j , 0 ) =u j , 0=ω 1 ( x j ) we obtain :

u j ,1=ω 1 ( x j ) +k ω 2 ( x j ) + 0.5 α ( ω 1 ( x j+1 )−2 ω 1 ( x j )+ ω1 ( x j−1 ) ) + 0.5 k q(x j , t n)
2 2
(15)
Hence the grid point at j−1 and j+1 are indicated the beyond of boundary grid point of the
solution domain. Hence the scheme in Eq.(15) is only used the initial condition at the interior
grind point, because due to the existence of boundary conditions in Eq.(7). Therefore by using
the finite discrete scheme given in Eqs. (12) and (15) give the M-1 system of the equation that
gives an accurate numerical solution of the one-dimensional wave equation given in Eq.(1)
implicitly using the matrix inverse method.
Richardson Extrapolation method
Extrapolation is an extremely powerful tool available to numerical analysts for improving the
performance of a wide variety of mathematical methods. It is an incredibly powerful technique for
increasing speed and accuracy in various numerical tasks in scientific computing [33]. It is also used to
speed up the rate of convergence for numerical methods.

Theorem: Let D p−1 ( h , k ) and D p−1 ( 2h , 2 k ) are two finite differences approximate value of the partial

derivative of u ( x , t ) with an order of accuracy is Ο ( h2 p +k 2 p ) such that for operator L u ( x 0 , t 0 ) define as

L u ( x 0 , t 0 )=D p−1 ( h , k ) +c 1 ( h2 p+ k 2 p ) +c 2 ( h2 p+2 + k 2 p+2 ) +c 3 ( h2 p+ 2+ k 2 p+2 ) +…

L u ( x 0 , t 0 )=D p−1 ( 2h , 2 k ) +22 p c 1 ( h2 p + k 2 p ) +22 p+2 c 2 ( h2 p+2 +k 2 p +2 ) +… (16)

Where p is the order of the differential equation. Then from the difference scheme in (16), we can drive
the improved central finite difference scheme for the solution of the first-order partial derivative of
u(x , t) at¿ ¿) which is given by:

L u ( x 0 , t 0 )=D p ( h , k ) +Ο ( h2 p+ 2+ k 2 p+2 )

p
4 D p −1 ( h , k )−D p−1 ( 2 h , 2k )
+Ο ( h +k )
2p 2p
¿ p
4 −1

Thus from this general, the central finite difference approximation of second-order partial derivative of
u(x , t) at¿ ¿) for j ≠ 0 and n ≠ 0 given by:

L u ( x j ,t n )=D p ( h , k )+Ο ( h 2 p +2+ k 2 p+ 2)

−p
u ( x−2¿¿− p h ,t ) −p u(x ¿ ¿ ,t −2 k )
¿ u(x +2¿¿− p h , t)−2 u ( x , t ) + −p 2
+u( x ¿ ¿ , t+2 k)−2u ( x , t ) + −p 2
¿¿¿¿
2 ×2 h 2× 2 k
(17)

Until|D p +1−D p|>| D p−D p−1| or |D p +1−D p|<tolerance where tolerance is provided for p=1 ( 1 ) N

with the order of accuracy isΟ ( h2 p +k 2 p ).

Now the truncation error terms of our formulated Lax-Wendorff scheme in Eq. (12) is Ο ( h2 +k 3 ) . So the

absolute error between two solutions at the grid point ( x j , t n ) is satisfied:

|u ( x j , t n )−u j ,n|<C ( h2 +k 3 ) (18)

where u ( x j , t n ) and u j ,n are respectively exact and numerical solution of one-

dimensional wave equation given in Eq. (1) and C is constant that independent
from step length ‘h’ and time step ‘k .’

Let us consider that Ω j n is the set of grid points that we obtain by using the mesh
size h and k in Eq. (3). Consider that Eq.(18) work for any mesh size h , k ≠ 0, which
implies that for ( x j , t n ) ∈ Ω j n:
 u ( x j , t n ) −u j ,n <C ( h 2+ k 3 ) + R j ,n (19)

Where R j ,n is reminder term in this interval for which u j ,n is approximated at each grid pointΩ j n .
Let Ω2 j2 n is the set of grid point that obtained by bisecting each mesh point

in Ω j n and let u2 j ,2 n is approximated numerical value at each bisected grid point. Then, Eq.
h k
(18) also work for , ≠ 0 which implies that:
2 2

(( ) ( ) )
2 3
h k
u ( x j , t n ) −u2 j , 2 n<C + + R 2 j ,2 n
2 2

8 u ( x j , t n ) −8 u2 j , 2n <C ( 2 h2+ k 3 ) +8 R2 j ,2 n (20)

Where R2 j , 2 n is reminder term in this interval for which u2 j ,2 n is approximated at each grid points
Ω2 j2 n . Now subtracting Eq.(19) from Eq.(20) we obtain:

2
7 u ( x j , t n ) −8 u2 j , 2n +u j ,n <C h + R

exact 1 1
− ( 8 u2 j , 2 n−u j ,n ) < ( C h + R )
2
(u ( x j , t n ) ) 7 7

Where R=8 R 2 j ,2 n−R j , n total reminder terms of two schemes. Negating the right-hand side terms
of the above difference equation, we obtain:

1
ǔ j n= ( 8 u2 j ,2 n−u j ,n ) (20)
7

exact
Where ǔ j n=( u ( x j , t n ) ) is also an approximation of the exact solution ( x j , t n ) .
Therefore using the above theorem, the order of truncation error for the scheme in
Eq. (20) is O( h 4 +k 6 ). Thus, the approximate solution in Eq. (20) is satisfies

|u ( x j , t n )−ǔ j n|≤ C ( h4 + k 6 ) (21)

Thus the extrapolation method accelerates the convergence (accuracy) of the present method to
approximate the exact solution with fourth-order in special variable six orders in the time
variable.

3. Stability and convergent Analysis

Because hyperbolic equations often describe the motion and development of waves, Fourier
analysis is of great value in studying the accuracy of methods as well as their stability [34]. The
modulus of λ (k ) describes the damping and the argument describes the dispersion in the scheme,
i.e., the extent to which the wave speed varies with the frequency [34]. The Fourier analysis
(Von-Neumann) stability analysis technique is applied to investigate the stability of the proposed
method. Such an approach has been used by many researchers like [15], [18], [25], [26], [35].
Now assume that the solution of the given problem at the points ( x j , t n )is
n ijθ
u j n=λ e (22)

Where i=√ −1 ,θ=kπ /N , k ∈ R and λ ∈ C which is R set of a real number and C set of a
complex number. Substituting Eq. (22) into Eq. (12) , we obtain:

λ n+1 eij θ=α ( 1+ 2 α ) λn e i ( j +1) θ + ( 1−4 α 2 ) λn eij θ +α ( 2 α −1 ) λn e i ( j−1 )θ +G j n

λ=α (1+2 α ) ei θ + ( 1−4 α 2 ) +α ( 2α −1 ) e−i θ +G j n

λ=1−4 α + α ( 1+2 α ) [ cosθ +isinθ ] +α (2 α −1 ) [ cosθ−isinθ ] +G j n
2

2
λ ( k )=1−4 α ( 1−cos θ−β G j n ) +i2 αsinθ ¿ (23)
Where β=1/2 α and λ ( k ) is an amplification factor of the proposed scheme. From this, we see
that the modules of this amplification factor are less than one(| λ ( k )|<1).

Theorem 1. Let Z has a power series expansion in the power of variable δ as follow

2 3 4
Z ≈ c 1 δ + c2 δ + c3 δ + c 4 δ +…

forδ → 0, then tan−1 ( Z ) is equal to the power series expansion of Z, then Z lies in the unit circle.

Proof: Let u j ,n be the finite difference approximation of first-order partial differential equation
obtained by using the finite discrete scheme of the form:
 u j n+1=u j n−α ∆−¿u jn ¿

where ∆−¿¿ indicate backward difference operator andα =∆ x /∆ t . Then the Von Neumann
stability analysis gives

Z ( k )=1−α ( 1−e iθ )=1−α (1−cosθ +isinθ)

( Z ( k ) )2=1−4 α ( 1−α ) sin 2 ( θ /2 )

Thus from this for all ‘k’ and 0 ≤ α ≤ 1 we have |Z ( k )|<1. Hence the argument of amplification
factor Z ( k ) is defined by

Arg(Z ( k ) )≈ tan
−1
( 1−α(1−cosθ)
α sin ⁡(θ)
)
≈−αθ ( 1−1/6 ( 1−α ) ) θ + …
2 2

Hence the theorem is proved. Now using the idea of this theorem, the arguments of the
amplification factor in Eq.(23) is

( )
2 α sin ⁡(θ)
Arg Arg (λ ( k ) )≈ tan −1 2
1−4 α ( 1−cos θ−β G j n )

( ( )
)
2 4 6
θ θ θ
2 α 1− + − +…
−1 2 24 720
≈ tan
1−4 α + 4 α (θ− + +… )+4 α β G
3 5 7
2 θ
2 θ θ 2
− jn
6 120 5040

( )
2 4 6
θ θ θ
2 α 1− + − +…
2 24 720
≅
1−4 α + 4 α (θ− + +… )+ 4 α β G
3 5 7
2 θ2 θ θ 2
− jn
6 120 5040

7 α ( 720−360 θ2 +30 θ4 −θ6 +… )

¿
2520−2 α ( 5040−5040 θ+840 θ3 −42 θ5 +θ7−… ) + 4 α 2 β G j n
Since 0 ≤ α =k /h ≤ 1 andθ=kπ /N . Hence for k → 0, θ → 0. still, we have seen that |λ ( k )|<1. Thus
λ ( k ) are lies inside the unit circle. Therefore, the lax-Wendorff finite difference scheme
given in Eq. (12) is stable for the wave equation.

Theorem 2: The difference equation given in the form of Eq.(12) is stable if for which the
eigenvalues of the coefficient matrix of the system of the differential equation are satisfied
Real ( λ j ) < 0. Proof: See reference [15]

Since from the principal part of the local truncation error, the derived local
truncation error for the proposed scheme is

( )
2 3
−α 3 α 4 k 3
∂x u j n− ∂ x u j n + ∂t u j n +Ο ( h +k )
2 2 3
τ j n=h
6 6 6

Thus,τ j n → 0 ash , k → 0. So that, the scheme is consistent with the order of.
Hence the scheme is convergent.

Criteria for Investigating the Accuracy of the Method

This section presented the criteria that the accuracy of the present method is investigated. Since
there are two types, Round-off errors and Truncation errors occur when differential equations
are solved numerically. Rounding errors originate from the fact that computers can only
represent numbers using a fixed and limited number of significant figures. Thus, such numbers
or cannot be represented exactly in computer memory. The discrepancy introduced by this
limitation is to call a Round-off error. Truncation errors in numerical analysis arise when
approximations are used to estimate some quantity. The accuracy of the solution will depend on
how small we make the step size, h, and time step k. To test the performance of the proposed
method, maximum absolute error, L2 and L∞ norms are used to measure the accuracy of the
method. These norms are calculated by:

√
N
L∞= max |u ( x j ,t n )−u j n|, L = 1 ∑|u ( x j , t n) −u j n|
2
2
1≤ n ≤N
N j=0
 where M is the maximum number of step,u ( x j , t n ) is the exact solution and u j n approximation
solution of the wave equation in Eq.(1) at the grid point ( x j , t n ).

4. Results

To test the validity of the proposed method, we have considered the following three model
problem considered in [2], [5], [10], [16]. Numerical results and errors are computed and the
outcomes are represented tabular and graphically.
Example 1: consider the classical wave equation considered in [2, 5]
utt =uxx , ( x , t ) ∈ ( 0 , 1 ) × ( 0 ,T )
Subjected to initial and boundary condition
u ( x , 0 )=cos ( πx ) ,ut ( x , 0 )=0 ,0 ≤ x ≤ 1 ,
1
u ( 0 , t ) =cos ( t ) ,∫ u ( x ,t ) dt =0 , 0 ≤t ≤ T
0

1
An analytical solution is u ( x , t )=
2
( cos ( π ( x +t ) ) +cos ( π ( x−t ) ) )
Example 2: consider the wave equation considered in [5]

( )
−t
1
utt −uxx = π 2 + e 2
sin ⁡(πx), ( x , t ) ∈ ( 0 , 1 ) × ( 0 ,T )
4
Subjected to initial and boundary condition
1
u ( x , 0 )=sin ( πx ) , ut ( x ,0 )= sin ( πx ) , 0 ≤ x ≤ 1,
2
1 −t
2
u ( 0 , t ) =cos ( πt ) ,∫ u ( x , t ) dt= e 2 , 0 ≤ t ≤T
0 π
−t
An analytical solution is u ( x , t )=e 2 sin ⁡( πx)

Example 3: consider the wave equation considered in [2]

utt −uxx =2t−6 x −2 , ( x ,t ) ∈ ( 0 , 1 ) × ( 0 , T )
Subjected to initial and boundary condition
2 2
u ( x , 0 )=x , ut ( x , 0 )=x−x , 0≤ x ≤ 1,
1
2 t 1
u ( 0 , t ) =−t ,∫ u ( x , t ) dt =−t + + ,0 ≤ t ≤ T
2

0 6 4
 An analytical solution is u ( x , t )=( x 2 +t ) ( x−t )
Table 1 Comparison of pointwise absolute error for problem give in example one with
computations carried out until final time T = 5 with mesh size h=0.01 and k=0.1.
x By the previous method By the present method
Goh Joan et.al.[16] S. M. Zin. et.al. [5] maximum absolute error

0.2 1.21E-04 1.12E-04 1.7865E-04

0.3 1.15E-04 1.07E-04 1.6803e-05
0.4 6.88E-05 6.40E-05 1.436E-05
0.5 2.03E-13 5.05E-15 4.5452EE-15
0.6 6.88E-05 6.40E-05 2.0236E-05
0.7 1.15E-04 1.07E-04 1.7205E-05
0.8 1.21E-04 1.12E-04 3.1439E-05
0.9 7.97E-05 7.39E-05 3.1214E-05

a) surface of numerical solution of wave test b) surface of exact solution of wave test
1

0.8 0.8

1 1
0.6 0.6

0.4 0.4
U(x,t)-Amplitude

U(x,t)-Amplitude

0.5 0.5

0.2 0.2
0 0
0 0

-0.5 -0.2 -0.5 -0.2

-0.4 -0.4
-1 -1
6 -0.6 6 -0.6
1 1
4 4
-0.8 -0.8
2 0.5 2 0.5
-1
time X time X
0 0 0 0

Figure 1: Surface graphs of example 1 showing the physical behavior of the one-dimensional wave
equation when h=0.02andk =0.1
 c) numerical versus Analytical solutionfor 1D wave
1

0.8

0.6

0.4
U(x,t)-Amplitude

0.2

U-app at t= 0.8
0
U-exact at t= 0.8
U-app at t= 2.5
-0.2
U-exact at t= 2.5
-0.4

-0.6

-0.8

-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x

Figure 2: The physical behavior of the solution for example one for Comparison of the Approximate
and Exact solution
 graph representation of Absolute error
0.1 1

0.09 Abslute error at t= 0.8 0.9

Abslute error at t= 0.9
Abslute error at t= 2.3
0.08 Abslute error at t= 2.5 0.8

0.07 0.7

0.06 0.6
Abslute error

0.05 0.5

0.04 0.4

0.03 0.3

0.02 0.2

0.01 0.1

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x

Figure 3: The physical behavior of Absolute error for solution of example one when h=0.01and

k =0.03 .

Table 2. Comparison of maximum pointwise absolute error for problem give in example two
with computations carried out until final time T = 1 with mesh size h=0.01 and k=0.0001.

t My previous method By present method

Meghan [10] S. M. Zin. et.al. [5] Maximum abs. Error
0.5 1.3371E-03 3.9515E-05 1.7005E-05
1 2.3794E-03 2.008E-04 4.1897E-06
 d) surface of numerical solution of wave test e) surface of exact solution of wave test

0.9 0.9

1 1
0.8 0.8

0.8 0.7 0.8 0.7

U(x,t)-Amplitude

U(x,t)-Amplitude
0.6 0.6 0.6 0.6

0.4 0.5 0.4 0.5

0.4 0.4
0.2 0.2

0.3 0.3
0 0
1 1
0.2 0.2
1 1
0.5 0.1 0.5 0.1
0.5 0.5
0 0
time X time X
0 0 0 0

Figure 4: Surface graphs of example 2 showing the physical behavior of the one-dimensional wave
equation when h=0.02andk =0.1

f) numerical versus Analytical solutionfor 1D wave

1
U-app at t= 0.9
0.9 U-exact at t= 0.9

0.8

0.7

0.6
U(x,t)-Amplitude

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x
Figure 5: The physical behavior of the solution for example 2 for Comparison of the Approximate
and Exact solution for step length h=0.01 and time step k =0.001

graph representation of Absolute error

0.12

60
Abslute error at t= 0.4
0.1 Abslute error at t= 0.6
Abslute error at t= 0,8
Abslute error at t= 0.9 50

0.08

40
Abslute error

0.06
30

0.04
20

0.02
10

0
-2 -1 0
10 10 10
x

Figure 6: The physical behavior of Absolute error for example two when h=0.0125andk =0.01 .

Table 3. Representation of L2 and L∞ norm to show the accuracy of the method for problem give in
example three with computations carried out until final time T = 1 with mesh size h and k.
i) Representing of performance of the present method with example 3 for decreasing
step length ( h → 0 ) with the fixed time step ( k =0.01 ) .
h k L∞ L2
0.1 0.01 3.2076E-04 1.0289E-05
0.05 0.01 7.8760E-05 1.2406E-06
0.025 0.01 1.5531E-05 9.6480E-08
0.0125 0.01 1.3273E-06 1.4094E-09
ii) Representing of performance of the present method with example 3 for decreasing
 time step k → 0 with fixed step length ( h=0.1 )
h k L∞ L2
0.1 0.0667 1.1049e-04 8.1385e-07
0.1 0.025 1.6940E-04 1.4348E-06
0.1 0.02 2.6235E-04 1.7207E-06
0.1 0.0125 3.1089E-04 1.2081E-06
g) surface of numerical solution of wave test h) surface of exact solution of wave test

1 1
U(x,t)-Amplitude

U(x,t)-Amplitude
0.5 0.5

0 0

-0.5 -0.5

-1 -1
1 1
0.8 1 0.8 1
0.6 0.6
0.4 0.5 0.4 0.5
0.2 0.2
time 0 0 X time 0 0 X

Figure 7: Surface graphs of example 3 showing the physical behavior of the one-dimensional wave equation when

h=0.1andk =0.01
 i) numerical versus Analytical solutionfor 1D wave
0.8
U-app at t= 0.6
0.7 U-exact at t= 0.6

0.6

0.5
U(x,t)-Amplitude

0.4

0.3

0.2

0.1

-0.1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x

Figure 8: The physical behavior of the solution for example 3 for Comparison of the Approximate
and Exact solution for step length h=0.1 and time step k =0.01

graph representation of Absolute error

0.12 1

Abslute error at t= 0.4 0.9

Abslute error at t= 0.6
0.1
Abslute error at t= 0,8
0.8
Abslute error at t= 0.9

0.7
0.08

0.6
Abslute error

0.06 0.5

0.4

0.04
0.3

0.2
0.02

0.1

0 0
-2 -1 0
10 10 10
x
Figure 9: Graph representation for Absolute error of example three when h=0.0125andk =0.01

5. Desiccations.

In this paper, the lax-Wendorff difference scheme with the Richardson extrapolation method is
presented for solving a one-dimensional wave equation subjected to integral conditions. To
demonstrate the competence of the method, three model examples are solved by taking
different values for step size h, and time step k. Numerical results obtained by the present
method have been associated with numerical results obtained by the method in [2],[5], [10],
16] and the results are summarized in Tables and graph. Moreover, the maximum absolute
errors decrease rapidly as the number of mesh points M and N increases. Further, as shown in
Figs. 2, 5, and 8 the proposed method approximates the exact solution very well for different
values of step length h and time step k as given above for which most of the current methods fail
to give good results. To further verify the applicability of the planned method, graphs were
plotted aimed at Examples 1, 2, and 3 for exact solutions versus the numerical solutions
obtained. As Figs. 1, 2, 4, 5, 7, and 8 indicate good agreement of the results, presenting exact as
well as numerical solutions, which proves the reliability of the method. Also, Figs. 3, 6, and 9
specify the absolute error of obtained numerical solution by the effects of mesh sizes on the
solution domain. Further, the numerical results presented in this paper validate the improvement
of the proposed method over some of the existing methods described in the literature. Both the
theoretical and numerical error bounds have been established . Hence, the Richardson extrapolation
method accelerates second-order into fourth-order convergent in spatial variable and third-order into
sixth-order convergent in the temporal variable. The results in Tables 1, 2, and 3 further confirmed that
the computational rate of convergence and theoretical estimates are in agreement .

As it can also be seen from table 3 when the value ofh is decreased with fixed time step k , the
maximum absolute error and L2-norm also decreased the accuracy of the proposed method
increase. But for k → 0 with fixed step sizeh both maximum absolute error and L2-norm also
increases. So that accuracy of the proposed method decreases. It concludes that the smaller value
of h and with fixed time step k gives a better approximation to the exact solution. However
decreasing both the value of step size h and time step k , affects the accuracy of the present
method that developed this paper. Comparison among Table 1-Table 3 and the graphs of the
numerical and exact solution of 1D wave equation shows that the present method generates a
more accurate result and it is superior to the method developed in [5], [10], [16] and It is
approximate the exact solution very well.
6. Conclusion
A new approach, lax-Wendorff difference scheme with Richardson extrapolation method is using
to solve 1D wave equation numerically is presented in this study. The comparison of the results
obtained by the present method with other methods reveals that the present method is more
convenient, reliable, and effective. An error analysis based on the Fourier series is also
developed in this study. As it can be seen that, the accuracy improves when M ∧N are increased.
Tables and figures indicate that as h decreases with a fixed value of k , the errors decrease more

rapidly. Another considerable advantage of the method is that the Richardson extrapolation
techniques are well approximate and give better accuracy of the numerical solution using
bisected grid point in the domain and searching accurate solution and improve the performance
of methods. In a summary, the lax-Wendorff difference scheme with the Richardson
extrapolation method is a reliable method that is capable to solve the one-dimensional wave
equation. Based on the findings, this method is well approximate and gives better accuracy of
the numerical solution with a fixed time step, k , and smaller step sizeh .
Acknowledgments
First of all, I would like to express my thanks to Allah, for giving me life and time to accomplish my
tasks. In the name of Allah, The Most Greatest and Most Merciful, there is no power except by the power
of Allah and I humbly return my acknowledgment that all knowledge belongs to Allah. I thank Allah for
granting me this opportunity to broaden my knowledge in this field. Nothing is possible unless He made it
is possible.
Secondly, greatly appreciate the help of my mother Shamo Busho, and my father Aliyi Koroche, without
the help of them I would not be here today.
Last but not least, thanks a lot to all my beloved friends; your kindness and helps will be a great memory
for me.
References

[1] Nečasová G., Šátek V., Kunovský J. Numerical solution of wave equation using

higher-order methods. In AIP Conference Proceedings, AIP Publishing PLC, 1978(1) (2018) 360005.

https://doi.org/10.1063/1.5043964.

[2] Awawdeh F. , Alomari A. K., A numerical scheme for the one-dimensional wave

Equation subject to an integral conservation condition. In AIP Conference Proceedings, AIP

Publishing LLC, 1739(1) (2016) 020019. https://doi.org/10.1063/1.4952499.

[3] L.C. Bartel, N.P. Symons, D.F. Aldridge. Graded boundary simulation of air/Earth

interfaces in finite-difference elastic wave modeling. In SEG Technical Program Expanded,

Society of Exploration Geophysicists Abstracts 2000 (2000) 2444-2447.

https://doi.org/10.1190/1.1815958.

[4] F. Shakeri, M. Dehghan. The method of lines for the solution of the - Dimensional wave equation

subject to an integral conservation condition. Computers & Mathematics with Applications, 56 (9)

(2008): 2175-2188. https://doi.org/10.1016/j.camwa.2008.03.055.

[5] S. M. Zin, A. A. Majid, A.I.M. Ismail, M. Abbas. Cubic Trigonometric B-spline Approach to

Numerical Solution of Wave Equation. International Journal of Mathematical, Computational,

Physical and Quantum Engineering, 8 (10) (2014) 1212-1216.

[6] S. N. Kihara. Heat Equations and their Applications, Kenya, School of Mathematics University of

Nairobi, Master’s Thesis, (2009).

[7] H.I.K.M.E.T. Caglar, S. Yilmaz, N.A.Z.A.N. Caglar , M. Iseri. A non- Polynomial spline solution

of the one-dimensional wave equation subject to an integral Conservation condition. In Proceedings

of the 9th Wseas international conference on applied computer and applied computational science,

Hangzhou. (2010) 27-30.

[8] G.D. Dean. Solutions of Partial Differential Equations, India, New Delhi:
CBS Publishers and Distributors, 1988.

[9] W. A. Mulder, M. J. Huskies. A simple finite-difference scheme for handling topography with

the first-order wave equation. Geophysical Journal International. 210 (1) (2017) 482-499.

https://doi.org/10.1093/gji/ggx178.

[10] M. Meghan, A. Shokri. A meshless method for the numerical solution of the one-dimensional

wave equation with an integral condition using radial basis functions. Numerical Algorithms, 52

(3) (2009) 461.

[11] G. H. Shortley, R. Weller. The numerical solution of Laplace's equation. Journal of

Applied Physics 9(5) (1938) 334-348.

[12] M.M. Haney. Generalization of von Neumann analysis for a model of two discrete Half-spaces: The

acoustic case. Geophysics. 72(5) (2007) SM35-SM46.

[13] J. Everink. Numerically solving the wave equation using the finite element
the method, Faculty of Science, Utrecht University, 2018.

[14] E. C. Rop, O. M. Oduor, O. M. Oduor. Solution of Third Order Viscous Wave Equation Using

Finite Difference Method, Nonlinear Analysis and Differential Equations, 7(1) (2019)53 – 68.

[15] K. A. Koroche. Numerical Solution for One Dimensional Linear Types of Parabolic Partial

Differential Equation and Application to Heat Equation. Mathematics and Computer Science. 5(4)

(2020) 76-85. https://doi: 10.11648/j.mcs.20200504.12.

[16] J. Goh, A. A. Majid &, A. I. M. Ismail. Numerical method using cubic B-spline for the heat and

wave equation. Computers & Mathematics with Applications. 62(12) (2011) 4492-4498.

https://doi.org/10.1016/j.camwa.2011.10.028.

[17] S. Wang, Y. O. Zhang, J. P. Wu. Lagrangian meshfree finite difference particle

method with variable smoothing length for solving wave equations. Advances in

Mechanical Engineering, 10(7) (2018). https://doi.org/10.1177%2F1687814018789248.

[18] M.K. Jain, S.R.K. Iyengar, R. K. Jain. computational methods for partial
differential
equations, New age international publishers, 2007.

[19] M.W. Oday. Stability and Convergence for nonlinear partial differential
equations, Master of Science in Mathematics thesis, Boise State University, (2012).

[20] N. Dostart, Y. Liu, M. A. Popović. Acoustic Waveguide Eigenmode Solver based on a

Staggered-Grid Finite-Difference Method. Scientific reports, 7(17509)(2017) 1-11.

https://doi.org/10.1038/s41598-017-17511-x.

[21] M. Dehghan. On the solution of an initial-boundary value problem that combines Neumann

and integral condition for the wave equation. Numerical Methods for Partial Differential

Equations, 21(1) (2005) 24-40. https://doi.org/10.1002/num.20019

[22] W.T. Ang. A numerical method for the wave equation subject to a non-local conservation Condition,

Appl. Numer. Math. 56(8) (2006) 1054–1060. https://doi.org/10.1016/j.apnum.2005.09.006.

[23] G. Seriani , S. P. Oliveira. Numerical modeling of mechanical wave propagation. La

Rivista del Nuovo Cimento, 43(2020)459–514. https://doi.org/10.1007/s40766-020-00009-0

[24] J. Rashidinia, M. Mohsenyzade. Numerical Solution of One-Dimensional Heat and

Wave Equation by Non-Polynomial Quintic Spline, International Journal of
Mathematical Modeling & Computations 5(4(FAAL))( 2015)291- 305.
.

[25] M. Dugassa, G. File, T. Aga. Numerical Solution of Second-Order One Dimensional Linear

Hyperbolic Telegraph Equation. Ethiopian Journal of Education and Sciences, 14(1) (2018)

39-52.

[26] S. Shokofeh, J.Rashidinia. Numerical solution of the hyperbolic telegraph the Equation by cubic B-

spline collocation method. Applied Mathematics and Computation 281(2016) 28–38.

https://doi.org/10.1016/j.amc.2016.01.049

[27] A. Saadatmandi, M. Dehghan, Numerical solution of the one-dimensional wave equation

with an integral condition, Numerical. Method Partial Differential Equations 23 (2007) 282–292

[28] M. Ramezani, M. Dehghan, M. Razzaghi, Combined finite difference and spectral methods

for the numerical solution of the hyperbolic equation with an integral condition. Numerical Methods

for Partial Differential Equations: An International Journal, 24(1) (2008) 1–8.

https://doi.org/10.1002/num.20230.

[29] S.A. Beilin, Existence of solutions for one-dimensional wave equations with nonlocal conditions.

Electronic Journal of Differential Equations (EJDE)[electronic only].76(2001) 1–8.

[30] Noye, J.N. Nonlinear Partial Differential Equations in Engineering. Conference in

Queens

College, University of Melbourne, North-Holland Publishing Company, Australia,

1981.

[31] R.C. Mittal, S. K. Kumar. Numerical Study of Fisher's Equation by Wavelet Galerkin

Method. International Journal of Computer Mathematics, 83(3) (2006) 287-298.

https://doi.org/10.1080/00207160600717758

[32] M. Bastani, D. K. Salkuyeh. A highly accurate method to solve Fisher’s equation. Pramana, 78(3)

(2012) 335-346. https://doi.org/10.1007/s12043-011-0243-8

[33] Liu E. H. L.,. Fundamental methods of numerical extrapolation with

Applications. MIT OpenCourseWare, Massachusetts Institute of Technology 209 (2006).

[34] K.W. Morton, D. F. Mayers. Numerical Solution of Partial Differential Equations. An

Introduction, Second Edition, Cambridge University Press, New York, 2005.

[35] Z.U.Zafar, M. Rafiq, A. Pervaiz, M. O. Ahmed. Finite Element Model for Linear Second
Order One Dimensional Inhomogeneous Wave Equation. Pakistan Journal of Engineering

and Applied Sciences. 17(2016)58–63.