Mathematics Letters

2021; 7(2): 19-24

http://www.sciencepublishinggroup.com/j/ml
doi: 10.11648/j.ml.20210702.11
ISSN: 2575-503X (Print); ISSN: 2575-5056 (Online)

Convergence Analysis for Wave Equation by Explicit Finite

Difference Equation with Drichlet and Neumann Boundary
Condition
Kedir Nebi Habib
Department of Mathematics, College of Natural Science, Arba Minch University, Arba Minch, Ethiopia

Email address:

To cite this article:

Kedir Nebi Habib. Convergence Analysis for Wave Equation by Explicit Finite Difference Equation with Drichlet and Neumann Boundary
Condition. Mathematics Letters. Vol. 7, No. 2, 2021, pp. 19-24. doi: 10.11648/j.ml.20210702.11

Received: October 24, 2020; Accepted: May 3, 2021; Published: May 26, 2021

Abstract: There are many problems in the field of science, engineering and technology which can be solved by differential
equations formulation. The wave equation is a second order linear hyperbolic partial differential equation that describes the
propagation of variety of waves, such as sound or water waves. In this paper we consider the convergence analysis of the
explicit schemes for solving one dimensional, time-dependent wave equation with Drichlet and Neumann boundary condition.
Taylor's series expansion is used to expand the finite difference approximations in the explicit scheme. We present the
derivation of the schemes and develop a computer program to implement it We use spectral radius of Matrix obtained from
discretization and Von Neumann stability condition to determine stability, and consistence of the method from truncated error
from discretized method. Using Lax Equivalence Theorem, convergence of the methods was described by testing consistency
and stability of the methods. And it is found out that the scheme is stable with the Drichlet boundary and conditionally stable
with Derivative boundary condition.
Keywords: Wave Equation, Explicit Method, Convergence, Stability

, satisfies the equation. =

0 < < , > 0 the equation is one dimensional wave
then On domain
1. Introduction
The wave equation is a second order linear hyperbolic equation. [11]
partial differential equation that describes the propagation of
variety of waves, such as sound or water waves. It arises in
different fields such as an acoustics, electromagnetic or fluid
dynamics [6, 7]. In many situations finding analytic solutions
to partial differential equation is unrealistic or even
impossible. Numerical methods that utilize computer
algorithms are then used to find approximate solution [2, 3,
10]. The focus of this paper is to determine the stability and
convergence of finite difference schemes that approximates a
solution of wave equation. Suppose that an elastic string of Figure 1. A vibrating string.

= where T is the
length L is tightly stretched between two supports at the same
horizontal level. So that the x-axis lies along the string. The The coefficient in the equation is
elastic string may be thought of as violin sting, guy wire or tension (force) in the string and m is the mass per unit length

,
possibly an electric power line. Suppose that the string is set of the string material. To describes the motion of the string

, . The ends are

in motion so it vibrates in vertical plane and let completely it is necessary also to specify suitable initial and
denote the vertical displacement experienced by the string at boundary conditions for displacement

condition are 0, =0 , = 0 ≥ 0 since the wave

the point at time if damping effects, such as air resistance assumed to remain fixed and therefore the boundary
are neglected. If the amplitude of the motion is not too large,
20 Kedir Nebi Habib: Convergence Analysis for Wave Equation by Explicit Finite Difference
Equation with Drichlet and Neumann Boundary Condition

=
$,'%& ( $,' ) $,'*&
,
equation is of second order with respect to t is plausible to (5)
,0 =
prescribe two initial conditions those are the initial position

0 ≤ ≤ And its initial velocity ,0 = 0 ≤

of the string . Substituting equations (4) and (5) into the equation (1) it is
≤ where and are given functions. The finite
approximated by
$%&,' ( $,' ) $*&,' $,'%& ( $,' ) $,'*&
+ ,
difference methods are the techniques for numerical solution = (6)
to the wave equation by the discretization of space and time
=
$,'%& ( $,' ) $,'*& $%&,' ( $,' ) $*&,'
interval [0, ] on the x axis, and let ; be the
Let the string in the deformed state coincide with the
, +
displacement at time in the y direction of a point initially at
(7)

=2 − + 3 − 2 -,. + -(/,. 4
2 ,
. the displacement function is governed by the -,.)/ -,. -,.(/ + -)/,. (8)

Making -,.)/ the subject and substituting 5 =

2 ,
mathematical model

= + , ∈ 0, , ∈ 0,
+
we
(1) obtain
(Governing equation) -,.)/ = 2 − 25 -,. +5 -)/,. +5 -(/,. − -,.(/ (9)

,0 = , = ∈ 0,
,!
(Initial condition) (2) 2.2. Matrix Form of Explicit Scheme

0, = , = 0 (Boundary conditions) ∈ 0,
nodes the temperature at time j ∆ t is given by [7j =
(3) Referring to equation (1), we discretize in space, using n

2, n, 3, n, 4, n, … n − 1, n] since 1, > = 0 for all j

,0 = When grouped in values in rows and using r= Equation 9
Since this PDE Contains a second-order derivative in time,
?,
specifying the initial shape of the string , , and +
we need two initial conditions, here

= In addition, PDEs need boundary conditions,

,! can be rearranged to obtain

here 0, = , = 0, specifying that the string is fixed 7 .)/ = 5 -(/,. + 2 − 25 -,. +5 -)/,. − -,.(/
at the ends, that ist the Displacement is zero. The solution
(10)

The parameter 5 = > 0 depends up on wave speed and

?,
, varies in space and time and describes waves that are +
moving with velocity to the left and right. the ratio of space and time step size the boundary condition
(1.) Require that (2) this allow us to rewrite the system in
matrix form
7 .)/ = A 7 . − 7 .(/ + BC
2. Finite Difference Methods
(11)

2 1 − 5 5 … 0
The finite difference techniques are based up on the

5 ⋱ ⋱ ⋮
approximations that permit replacing differential equation by

Where A = D G and
⋮ ⋱ ⋱ 5
finite difference equation. There finite difference

0 … 5 2 1 − 5
approximations are algebraic in form, and the solutions are
related to grid points. Thus, a finite difference solution
/,.
7. = H ⋮ I
basically involves three steps:
Dividing the solution into grids of notes.
Approximating the given differential equation by finite -(/,.
difference equivalence that relates the solutions to grid
points. 2.3. One Dimensional Wave Equation with Derivative
Boundary Condition
, = makes change
Solving the difference equations subject to the prescribed

, =
boundary conditions and or Initial conditions. [2, 5, 9] The boundary condition

0 perfectly reflects the wave. Our next task is to determine

2.1. Explicit Finite Difference Method sign at the boundary, while the condition

The numerical solution of one dimensional wave equation the stability with boundary condition , = 0, which is

, = 0 in a finite
using explicit scheme is obtained and the error calculated. To more complicated to express numerically. We shall present
determine the stability and convergence, we will consider the two methods for implementing

= + , With the boundary condition

simplified form of the wave equation (1) Common form. difference scheme, one based on deriving a modified stencil
at the boundary, and another one bead on extending the mesh

0, = , = 0 And the initial condition u (x, 0)

with ghost cells and ghost points.

= , and "# , 0 = . Using finite difference we

Neumann boundary condition:

= J. ∇ = 0
.
have (12)

=
$%&,' ( $,' ) $*&,'
The derivative is in the outward normal direction from

+
.
(4)
a general boundary.
 Mathematics Letters 2021; 7(2): 19-24 21

For one dimension domain [0; ], we have that: tridiagonal matrix of the form [1, 4, 13, 15]
L L L L
M = , M =−
Lax Equivalence theorem: The Lax-Richtmyer

LJ L LJ L
Equivalence Theorem is often called the Fundamental
NO N! Theorem of Numerical Analysis, even though it is only
Boundary conditions that specify the value of . are applicable to the small subset of linear numerical methods for
well-posed, linear partial di　erential equations. Along with
of . When ( = 0 or = 0 ) it is Dirichler boundary
known as Neumann conditions, while refer to specifications
Dahl Quist’s equivalence theorem for ordinary di　erential
.
stability ⇒ convergence always holds has caused a great deal
equations, the notion that the relationship consistency +
condition
The Discretization of derivatives at the boundary in the of confusion in the numerical analysis of di　erential
finite difference scheme is used central differences in all the equations. In the case of PDEs, mathematicians are most
tempting to implement (12) at = 0 and = . by the
other approximations to derivatives in the scheme, it is
often interested in nonlinear phenomena, for which Lax-
Richtmyer does not apply. More damningly, the forward
difference
*&,' ( &,'
=0
implication that

∆
(13) Consistent +stability ⇒ convergence
The problem is that (/,. is not a value that is being
matrix A = (_C- ), for row i the disk `C have centre _CC and
Theorem: Gerschgorin’s theorem: Consider a square

combine 13 with the scheme for j = 0, -N/a_C- a. Then the theorem states that;
radius∑b
computed since the point is outside the mesh. However, if we

-cC
Every eigenvalue of A lies in some `C .
If S is the union of s disks `C such that S is disjoint from
2.4. Fourier Method (Von Neumann Stability)

Fourier stability analysis allows determining appropriate

all other disks of this type, then S contains precisely m
step sizes for an accurate solution when the wavelength or
eigenvalues of A.
decay constant (which is given in terms of parameters such as
a diffusion constant or wave velocity in the (PDE) has a
certain value. Fourier stability analysis does not take 3. Analysis of Convergence for Explicit
boundary conditions for a specific problem into account. And
it performed by substituting the analytic solution to a partial
Scheme
differential equation into the numerical finite difference 3.1. Consistence of the Explicit Method

, = ∑, Q t exp ikx
equation. [8, 12-14]
Assume The truncation error of the approximation of the time
derivative is / given by
just one Fourier
The sum is over X, the Fourier frequencies. Now take for

= + / Where − ,e
$,'%& ( $,' ) $,'*& , d
Term , = Q t exp ikx And evaluate it at ( -, ., ) to , /= / d

Where t−X < e < + X from the Taylor series expansion

-,. = Q . exp iki∆x , -(/,. = Q . exp ik j − 1 ∆x ,
get

-)/,.)/ = Q .)/ exp ik j + 1 ∆x -,.(/ = =

$%&,' ( $,' ) $*&,$
+
Q .(/ exp ik j + 1 ∆x These expressions can be plugged
and Similarly, for the space derivative,

Where =− f, −ℎ ≤
+
+

stability. The growth rate G is defined as G = Y Z'%& Y for

Z
directly into any finite difference scheme to check for
/
is the truncation error
' f ≤ + ℎ Combining the two terms above −_ =0

− 2 -,. + -,.(/ − 2 -,. +

stability we need G < 1 for all frequencies k. Conditional

Usually the condition limits ∆t in function of ∆x −_ +

-,.)/ -)/,. -(/,.
stability means we only have stability on a certain condition.
X ℎ h

2.5. Matrix Method to Determine Stability =0

− =− ,e
,
finding the spectral radiu which is [ \ =max (]C ) where ]C h the h= /
The condition for stability of methods is determined by
/
Where truncation error
f,
+

(i) If [ \ <1 then the system is stable. /

is an eigenvalue of matrix A, as illustrated below.
Where − ℎ ≤ f ≤ +h and t −X < e < + X It then
+

(ii) If [ \ < 1then the system is stable.

(iii) If [ \ >1 then the system is unstable.
follows that

| h| ≤ j + j X j +ℎ j
Tridiagonal matrices are often found in connection with , + /
/ / /
=

Where j is a bound for | , e |andj

finite differences. Tridiagonal matrices are easy to deal with

for | f, |
since there exists ancient numerical methods both for solving is a bound
their linear systems of equations and eigenvalue problem.
Here we consider the eigenvalue problem for a general
22 Kedir Nebi Habib: Convergence Analysis for Wave Equation by Explicit Finite Difference
Equation with Drichlet and Neumann Boundary Condition

Hence h → 0 as k, h → 0 that is the difference This places a restriction on the relative sizes of the time
approximation is consistent and space steps. We conclude that the numerical scheme is
conditionally stable. The stability criterion (23) is known as
3.2. Stability Analysis of Explicit Method the Courant condition, The Courant condition requires that
the mesh slope, which is defined to be the ratio of the space
3.2.1. Stability For Explicit with Drichlet Boundary
step size to the time step size, namely ℎ/X, must be strictly
Condition
greater than the characteristic slope c
The stability of this numerical scheme proceeds as follows

-,.)/ = 2 − 25 -,. +5 -)/,. +5 -(/,. − -,.(/

3.2.2. Stability of the Explicit Method with Derivative
(14)
Boundary Condition
We first need to recast the second order iteration system The difference equation that we will use to time stepping
(14) in to a fist order system the numerical scheme

7 .)/ = A 7 . − 7 .(/ + BC (15) 1, J = 0 Left boundary > = 1, J = 1, … , j (24)

7 .)/ A −o 7 ~, J = 0 Right boundary > = ~, J = 1, … , j (25)

l m=n p l . m + BC ,
7. o 0 7 .(/
= Derivative boundary for > = 0 (26)
(16)

7 7
-)/,. -(/,.
A −o
q̅.)/ = l .)/ m , q.̅ = l . m s=n p = Derivative boundary for > = ~ (27)
7. 7 .(/ o 0 -(/,. -)/,.
Let and

= Row 1 > = 2, … , ~ − 1 J = 1
Implies that
̅ = sq̅. + BC
-,/ -
q.)/
(28)

= + -X + 3 −2 -+ -(/ 4 Row 2 > = 2, ~ − 1

h
(17)
-, - -)/
n= 2
Therefore the stability of the method will be determined by

equation st = ]t where t = n p can be written out in its

the Eigen value of the coefficient matrix C the Eigen vector (29)

u -,.)/ = 2 − 25 -,. +5 3 -)/,. + -(/,. 4 − -,.(/ (30)

individual components
A −o
We can eliminate the fictitious value -(/,. we see that
n pn p = ]n p -(/,. = -)/,. from (16), which can be used in (21) to arrive
o 0 u u
(18)
at a modified scheme for the boundary point !,.)/
A −u =]
= 2 − 25 + 25 − -,.(/ For > = 0 (31)
(19)
-,.)/ -,. -)/,.
= ]u
Similarly, if it applied at = ~ is discretized by using
(20)

Subtitling (20) into (19) we find (31) we have

]A − ] − 1 u = 0 Or Au = ] + u = 2 − 25 + 25 − -,.(/ For > = ~ (32)

/
-,.)/ -,. -(/,.
v
(21)

The latter equation implies that u is the Eigen vector of A

with ] + ](/ the corresponding Eigen values of the
The modification of the scheme at the boundary is also

tridiagonal matrix A
required for the special formula for the first time step Then
we have the explicit scheme with derivative boundary is

n] + p = 2 1 − 5 + 5 cos
/ z, -,.)/ = 2 − 25 -,. + 25 -)/,. − -,.(/ For > = 0 (33)
v .
= 2 − 25 -,. + 5 3 -)/,. + -(/,. 4 − -,.(/ > =
k= 1, 2. N-1 (22)

Multiplying both sides of equation (22) by ] leads to a -,.)/

quadratic equation for the Eigen values let _, = 1 − 5 + 1, … , ~ − 1 (34)

5 cos = 2 − 25 + 25 − -,.(/ For > = ~
z,
. -,.)/ -,. -(/,.
] − 2_, ] + 1 = 0 Where 1 − 25 < _, < 1 Each pair of (35)

], = _, ± |_, − 1 Yields two Eigen values of matrix C

solutions to this n-1 quadratic equation namely
Its matrix form is as follows
If _, < −1 then one of the two eigenvalues will be both 7 .)/ = \ 7 . + €• + BC
< −1, which means that the linear iterative system has an
(36)

exponentially growing mode If _, < 1, then the Eigen values 2 − 25 25 ⋯ †

„ 5 2 − 25 ‰̂
Where \ = ƒ ⋮ ⋱ ⋮
are complex number of modules 1, indicated stability of the

Therefore in view of 1 − 25 < _, < 1 we should require †… ⋯ ⋱ 5

matrix C [7] ,

that ‚ 25 2 − 25 ‡

5= < 1 Or X <
2, +
+ 2
(23)
 Mathematics Letters 2021; 7(2): 19-24 23

/,. 0, = , = 0 Initial
7. = H ⋮ I , 0 = sin • ,0 = 0
condition distribution is

-,. − 1
and
stable and instability of explicit method with derivative boundary

„ + X+5
4

‰̂
-,/
−2 -+
r=1

„ ‰̂ ƒ
-,/ - -
2 -)/ -(/
r=0.5

€• = ƒ ⋮
3

=ƒ ⋮ ˆ
⋮ ƒ ⋮ ˆ
2

‚ ‡
-,/ 5
‚ + X + −2 -+ ‡
- -
2 -)/ -(/
1

solution
7 .)/ = \ 7 . + • where • = €• + BŠ•
0

(37)

This is significant result in that it shows that the vector •

-1

For stability [ \ ≤ 1 . Consider the matrix \ =

-2
does not affect stability

2 − 25 25 ⋯ 0
-3

„ 5 2 − 25 ‰̂
ƒ ⋮ ⋱ ⋮
-4

⋯ ⋱ 5
0 1 2 3 4 5 6 7 8 9 10

0… Figure 2. Stability of explicit method for 5 = 0.5, X = 0.05 _J€ ℎ = 0.1 at

x

‚ 25 2 − 25 ‡
// = 0.5 And it’s instability for r = 1 k=0.25 and. ℎ = 0.5 at // = 0.25.
The disks for row 2 to ~ − 1 have center 2 − 25 and
radii 25 so the eigen value lies in 2 − 25 − 25 ≤ ] ≤ 2 −
25 + 25
Table 1. Maximum absolute error of example 1 for r=0.75 explicit method at
different time t.

2 − 45 ≤ ] ≤ 2 Time step ∆# # Explicit Method

2 − 45 ≥ −1 which implies that 5 ≤

√Œ
2 0.0475 0.0475 0.8254
7 0.0475 0.2850 0.0455

Consider the first and last row with center 2 − 25 and

11 0.0475 0.475 5.4487

radii 25 The disk is

15 0.0475 0.665 10.2516
19 0.0475 .0.855 18.4482

Table 2. Maximum absolute error of example 1 for r=1.05 Explicit method at

different time.

Time step ∆# # Explicit Method

2 0.0525 0.0525 0.7662
7 0.0525 0.315 2.1107
11 0.0525 0.525 3.7374
15 0.0525 0.7350 6.2255

• \ ≤ 1 If 2 − 45 ≥ −1 and 2 so the explicit scheme

19 0.0525 .0.9450 32.2921

is stable for 5 ≤
√Œ
stability and instability of explicit method

Ž
8
r=0.75
This shows that with derivative boundary condition the r=1.05
6
explicit scheme stability depends on the values of r.
4
4. Numerical Experiment
2
Numerical examples are presented to verify stability and
solution

convergence of the method

0
Example 1 Use the explicit scheme to solve the one
dimensional wave equation
=4 0 ≤ ≤ 1,0 ≤ ≤ 1
-2

0, = cos 2 , •, = − cos 2 and ,0 =

sin , ,0 = 0
-4

-6
Example 2. Use the explicit scheme to solve the one 0 1 2 3 4 5 6 7 8 9 10

=4 ∈ [0, ] ∈ [0, ] With boundary Figure 3. Stability of explicit method for 5 = 0.75, X = 0.0475 _J€ ℎ =
x
dimensional wave equation

0.1 at // = 0.475 and it’s instability for r = 1.05 k=0.0525 and. h=0.1 at
For
24 Kedir Nebi Habib: Convergence Analysis for Wave Equation by Explicit Finite Difference
Equation with Drichlet and Neumann Boundary Condition

// = 0.525. the same way as the problem variable, so an unstable process

can be observed by the solution growing beyond any bounds.
5. Discussion and Conclusion With small r the method performed well. The explicit scheme
has an advantage that it is easy to set up, and disadvantage

unstable for 5 ≤
5.1. Discussion that it is unstable for r greater than one with drichilet and
√Œ

= +
I used one dimensional wave equation by considering the with derivative boundary

, with drichelet and derivative boundary conditions.

space domain to have a length of M that is
5.2. Conclusion
The explicit finite differences were considered. To determine This study has considered the explicit finite difference
its stability by Von Neumann stability condition and Eigen schemes for solving one dimensional time dependent wave
value of tridiagonal matrix obtained from discretized scheme equation with drichlet and Neumann boundary conditions.
of the equation was used to develop the analysis. To produce
The difference schemes are derived. Using Lax Equivalence
values for the problem variable finite difference mesh point
Theorem convergence of the method was described by
were used for the space domain. A mat lab code was written
testing consistency and stability of the methods. Stability was
for explicit methods with drichlet and derivative boundary
condition discussed by using Gerschgorin’s Theorem and Von
The numerical scheme is said to be stable when an error is Neumann stability condition. And the stability of the Explict
introduced at certain stage, then remains bounded as time method is shown by the table below
approaches infinity. It so happen that the error propagates in
Table 3. Summary of the finding.
“
"## = ’ "”” + •(”, #) Matrix method and Fourier method
Methods Drichilet Derivative
Explicit scheme stable 5 ≤ 1 conditionally stable 5 ≤ √3/2

In the above table the Drichilet boundary condition are [8] 2012. Rishu SinglaThesis paper of finite difference equation
(0, ) = ( , ) = 0 and the diivative boundary condition for parabolic equationHans De and Paul Ullrich (2007-2009).
are ( , ) = ( ), ( , ) = ( ) 1 is the critical value Introductionto Computational PDEs Department of Applied
Mathematics University of Waterloo.
such that for 5 ≤ 1 scheme is stable.
A systematic study was applied to the two test numerical [9] Randall J. Leveque (2006). A Finite difference Methods for
problems and the schemes have been successfully applied. Differrential Equations. AMath 585, Winter Quarter 2006
The performance of the schemes for the considered problems University of Washington.
was measured by calculating the error. [10] Rishu Singla (2012). Numerical solution of Some Parabolic
Partial Differantial Equations Using Finite Difference
Methods.
References [11] S. Larsson and V. Theme. Partial Differential Equations with
Numerical Methods, volume 45 of Texts in Applied
[1] AUTUMN (2009) the eigen vaue of the tridiagonal matrix. Mathematics. Springer, 2003.
[2] D. M. Causon; Professor C. G. Mingham (2010). Introductory [12] Abel Kurura T. and Charles Otieno N. (2018) Stability and
Finite Differences for PDEs. Consistency Analysis for Central Difference Scheme for
Advection Diffusion Partial Differential Equation Kisii
[3] Harwinder Kaur (2012). MSc. Thesis on Numerical Solutions
university, Kenya.
of Some Differantial Equations Using B-Spline Collection
Method. School of Mathematics and Computer Applications, [13] Doyo K. and Gofe G. (2016). Convergence Rates of Finite
Thapar University. Difference Schemes for the Diffusion Equation with Neumann
Boundary Conditions. American Journal of Computational and
[4] J. W. Thomas (1998) numerical partia equation: finite
Applied Mathematics, 6 (2): 92102.
difference methods.
[14] Azad, T. M. A. K., M. Begum and L. S. Andallah. (2015). An
[5] Michael T. Heath (2002). Scientific Computing: An
explicit finite difference scheme for advection diffusion
Introductotry Survey Chapter 11- Partial
equation (2015). Jahangirnagar J. Mathematics and
DifferantialEquations.
Mathematical Sciences 24: 2219-5823.
[6] Michael P. Lamoureux (2006). The mathematics of PDEs and
[15] Chan, T. F. (1984). Stability analysis of finite difference
the wave equation.
schemes for the advection diffusion equation. SIAM J. Numer.
[7] 2012 Peter J. Olver Vibration and Diffusion in One– Anal. 21: 272-284.
Dimensional Media.