Electronic Journal of Mathematical Analysis and Applications,
Vol. 4(1) Jan 2016, pp. 197-204.
ISSN: 2090-729(online)
http://fcag-egypt.com/Journals/EJMAA/
————————————————————————————————
COMPOSITE FINITE DIFFERENCE SCHEME APPLIED TO
SOME NONLINEAR EVOLUTION EQUATIONS
S. A. EL MORSY
Abstract. This paper is concerned with the application of the composite finite difference scheme (CFDS) to some classes of evolution equations. Three
models of evolution equations are studied. The first model is the nonlinear reaction-diffusions equation (NRD) with a reaction term, while the second is modified Korteweg de Vries equation (mKdV) and the third is the
Fitzhugh?Nagumo equation (FN). Numerical examples showed that the CFDS
give high accuracy.
1. Introduction
Nonlinear evolution equations are widely used to describe many important phenomena and dynamic processes in physics, mechanics, chemistry, biology, etc. The
study of nonlinear partial differential equations is very important. Many methods,
exact, approximate and purely numerical are available for solution of nonlinear
partial differential equations [1]-[30].
Reaction-diffusion equations (RD) are mathematical models which explain how the
concentration of one or more substances distributed in space changes under the
influence of two processes: local chemical reactions in which the substances are
transformed into each other, and diffusion which causes the substances to spread
out over a surface in space. A great deal of research work has been published on
the development of numerical and analytical solutions of NRD equations [1]-[10].
In recent years, many physicists and mathematicians have paid much attention
to the Fitzhugh? Nagumo (FN) equation due to its importance in mathematical
physics. The Fitzhugh?Nagumo equation has various applications in the fields of
flame propagation, logistic population growth, neurophysiology, branching Brownian motion process, autocatalytic chemical reaction and nuclear reactor theory;
see, e.g. [11]-[16]. This equation is an important nonlinear reaction diffusion equation and usually used to model the transmission of nerve impulses [16]. Numerical
schemes for FN equations [17]-[19] by collocation method and the ?Hopscotch?
finite difference scheme first proposed by Gordon [19], and further developed by
Gourlay [20]-[21].
2010 Mathematics Subject Classification. 34A12.
Key words and phrases. Partial differential equations, nonlinear reaction-diffusions equation,
modified Korteweg de Vries equation and Fitzhugh?Nagumo equation.
Submitted April 2, 2014. Revised April 22, 2014.
197
198
S. A. EL MORSY
EJMAA-2016/4(1)
A great deal of research work has been invested during the past decades in the study
of the mKdV equation [22]-[30]. The main goal of these studies was its analyticalzz
and numerical solutions. Several different approaches, such as Backland transformation, a bilinear form, and a Lax pair, have been used independently, by which
Anjan et al. [22]-[25] obtain soliton and multi-soliton solutions for this equation.
The aim of this paper is to apply the CFDS to obtain the solutions for the three
different types of nonlinear partial differential equations such as, nonlinear Reaction?diffusion equation (NRD), Fitzhugh? Nagumo (FN) and modified Kortewege
de Vries equation (mKdV) which are important equations.
2. The nonlinear reaction-diffusion equation with reaction term
An example of practical interest is known as the nonlinear reaction-diffusions
equation (NRD) with a reaction term [7, 8].this equation takes the form [8].
ut − u2xx = pu − qu2 , (x, t) ∋ QT
(1)
Here QT = Ω × I, Ω ≡ (a, b), I = (0, T ), a and b are real positive constants. We
consider equation (1) associated with initial condition u(x, 0) = u0 (x). In Finite
difference method (FDM) the domain is discretized to a finite number of points
forming a mesh with horizontal step size h = b−a
N , N is the number of intervals,
0 ≺ i ≼ N and k is the time step such that T = k ∗j, 0 ≼ j ≼ M . The derivatives are
replaced by difference formulas [31]-[32] as follows, for i = 1, 2 we use the forward
formula
−3uji + 4uji+1 − uji+2
(ux )ji =
2h
(uxx )ji =
2uji − 5uji+1 + 4uji+2 − uji+3
h2
(2)
−5uji + 18uji+1 − 24uji+2 + 14uji+3 − 3uji+4
2h3
while for i = 3, N − 2 we use the central formulas
(uxxx )ji =
(ux )ji =
(uxx )ji =
uji+1 − uji−1
2h
uji+1 − 2uji − uji−1
2h2
(3)
uji+2 − 2uji+1 + 2uji−1 − uji−2
2h3
And for i = N − 1, N we use the backward formulas
(uxxx )ji =
(ux )ji =
(uxx )ji =
(uxxx )ji =
3uji − 4uji−1 + uji−2
2h
2uji − 5uji−1 + 4uji−2 − uji−3
h2
5uji − 18uji−1 + 24uji−2 − 14uji−3 + 3uji−4
2h3
(4)
EJMAA-2016/4(1)
COMPOSITE FINITE DIFFERENCE
199
3. Application of CFDS to the nonlinear reaction-diffusion equation
Consider the nonlinear PDE (1), we can rewrite:
ut = u2xx + pu − qu2
multiply both sides of (5)
by ∂F
∂u
(5)
, we have
∂F ∂u
∂F 2
=
(u + pu − qu2 )
∂u ∂t
∂u xx
(6)
or
∂F 2
∂F
=
(u + pu − qu2 )
(7)
∂t
∂u xx
where F , is any continuous and differentiable function. If we choose F (u) = ln u,
we obtain the Exponential finite difference method (Exp. FDM) [33]-[36]. The
Logarithmic finite difference method (Log. FDM) [32] is obtained when we set
F (u) = exp u.
3.1. Exponential finite difference method applied to the nonlinear reactiondiffusion equation. In this sub-section we will apply the Exponential finite difference method to (7). The usual forward difference formula leads to, ∂F
∂t =
F (uj+1
)−F (uji )
i
k
is the time step. Substitute in (7)), we have
F (uj+1
) − F (uji )
∂F 2
i
=
(u + pu − q ∗ u2 )
k
∂u xx
(8)
or
∂F 2
(u + pu − q ∗ u2 ))
∂u xx
setting F (u) = ln u , then (9) takes the form,
F (uj+1
) = F (uji ) + k(
i
ln uj+1
= ln uji +
i
k
uji
((uji )2xx + puji − q(uji )2 )
(9)
(10)
so,
uj+1
= exp(
i
k
uji
((uji )2xx + puji − q(uji )2 )
(11)
Applying the difference formulas (2)-(4) to (11), we have the following recurrence
relations:
uj+1
= (uji ) exp(
i
j
j
j
j
k 2(ui )2 − 5(ui+1 )2 + 4(ui+2 )2 − (ui+3 )2
+p(uji )−q(uji )2 ), i = 1, 2
h2
uji
(12)
uj+1
= (uji ) exp(
i
uj+1
= (uji ) exp(
i
j
j
j
k (ui+1 )2 − 2(ui )2 + (ui−1 )2
+p(uji )−q∗(uji )2 ), i = 3 : N −2 (13)
j
2
2h
ui
j
j
j
j
k 2(ui )2 − 5(ui−1 )2 + 4(ui−2 )2 − (ui−3 )2
+p(uji )−q∗(uji )2 ), i = N −1, N
h2
uji
(14)
200
S. A. EL MORSY
EJMAA-2016/4(1)
3.2. Logarithmic finite difference method applied to the nonlinear reactiondiffusion equation. In (Log. FDM) we assume F (u) = exp u, equation (9) transformed to
exp(uj+1
) = exp(uji ) + k exp(uji )((uji )2xx + puji − q(uji )2 )
i
(15)
uj+1
= uji + ln(1 + k((uji )2xx + puji − q(uji )2 )
i
(16)
we have,
Similarly applying the difference formulas (2)-(4) to (16), we have the following
recurrence relations:
uj+1
= uji +ln(1+k(
i
(2(uji )2 − 5(uji+1 )2 + 4(uji+2 )2 − (uji+3 )2 )
+puji −q(uji )2 )), i = 1, 2
h2
(17)
uj+1
= uji +ln(1+k(
i
uj+1
= uji +ln(1+k(
i
(uji+1 )2 − 2(uji )2 + (uji−1 )2
+puji −q(uji )2 )), i = 3 : N −2 (18)
2h2
(2(uji )2 − 5(uji−1 )2 + 4(uji−2 )2 − (uji−3 )2 )
+puji −q(uji )2 )), i = N −1, N
h2
(19)
4. Fitzhugh?Nagumo equation
The classical Fitzhugh?Nagumo equation [16]-[17],is given by
ut = uxx + u(1 − u)(p − u)
(20)
where 0 ≼ p ≼ 1 and u(x, t) is the unknown function depending on the temporal
variable t and the spatial variable x. This equation combines diffusion, and nonlinearity which is controlled by the term u(1 − u)(p − u). When p = 1, (20) reduces
to the real Newell?Whitehead equation. To apply CFDS, we reset (20) as follows,
F (uj+1
) − F (uji )
∂F
i
=
(uxx + u(1 − u)(p − u))
k
∂u
(21)
∂F
(uxx + u(1 − u)(p − u))
∂u
(22)
ie.
F (uj+1
) = F (uji ) + k
i
4.1. Exp. FDM method applied to FN equation. Replacing the derivatives
in (22) by the difference formulas (2)-(4), we obtain
uj+1
= (uji ∗exp
i
j
j
j
j
k 2ui − 5ui+1 + 4ui+2 − ui+3
(
+uji (1−uji )(p−uji )), i = 1, 2 (23)
h2
uji
uj+1
= (uji ∗ exp
i
j
j
j
k ui+1 − 2ui + ui−1
(
+ uji (1 − uji )(p − uji )), i = 3, N − 2 (24)
2h2
uji
EJMAA-2016/4(1)
COMPOSITE FINITE DIFFERENCE
201
and
uj+1
= (uji ∗ exp
i
j
j
j
j
k 2ui − 5ui−1 + 4ui−2 − ui−3
(
+ uji (1 − uji )(p − uji )), i = N − 1, N
h2
uji
(25)
4.2. Log. FDM method applied to FN equation. For the Log. FDM, we
have the following iterative formulas,
uj+1
= (uji + ln(1 + k(
i
2uji − 5uji+1 + 4uji+2 − uji+3
+ uji (1 − uji )(p − uji ))), i = 1, 2
h2
(26)
uj+1
= (uji + ln(1 + k(
i
uji+1 − 2uji + uji−1
+ uji (1 − uji )(p − uji ))), i = 3, N − 2 (27)
2h2
and and for i = N − 1, N we have
uj+1
= uji + ln(1 + k(
i
2uji − 5uji−1 + 4uji−2 − uji−3
+ uji (1 − uji )(p − uji ))
h2
(28)
5. Application of CFDS to the modified KdV equation
Consider the modified Korteweg?de Vries equation (mKdV), which takes the
form
ut + 6u2 ux + uxxx = 0
(29)
F (uj+1
) = F (uji ) − k(6(uji )2 (uji )x + (uji )xxx )
i
(30)
Simliarly, we obtain
5.1. Exp. FDM method applied to mKdv equation. The recurrence relations
of Exp. FDM are,
uj+1
= (uji ∗exp
i
uj+1
= (uji ∗exp
i
−k
uji
−k
uji
(6(uji )2
−3uji + 4uji+1 − uji+2 −5uji + 18uji+1 − 24uji+2 + 14uji+3 − 3uji+4
+
), i = 1, 2
2h
2h3
(31)
(6(uji )2
uji+1 − uji−1 uji+2 − 2uji+1 + 2uji−1 − uji−2
+
), i = 3, N −2
2h
2h3
(32)
(6(uji )2
3uji − 4uji−1 + uji−2 5uji − 18uji−1 + 24uji−2 − 14uji−3 + 3uji−4
+
), i = N −1, N
2h
2h3
(33)
and
uj+1
= (uji ∗exp
i
−k
uji
202
S. A. EL MORSY
EJMAA-2016/4(1)
Table 1. Numerical results of solving NRD equation at different
times,absolute errors for CFDS
x
2
3
4
5
6
7
8
9
10
11
12
t = 0.01
0.0001351
0.0000716
0.0000363
0.0000695
0.0000416
0.0000249
0.0000150
0.0000091
0.0000055
0.0000028
0.0000017
t = 0.1
0.0011230
0.0008931
0.0002172
0.0006197
0.0003921
0.0002358
0.0001420
0.0000857
0.0000516
0.0000258
0.0000184
t = 0.5
0.0304806
0.0159891
0.0016654
0.0016458
0.0014175
0.0009075
0.0005527
0.0003337
0.0001966
0.0000844
0.0001753
5.2. Log. FDM method applied to mKdV equation. In case of Log. FDM
equation (30) transformed to
uj+1
= uji + ln(1 − k(6(uji )2 (uji )x + (uji )xxx ))
i
(34)
Similarly applying the difference formulas (2)-(4) to (34), we have the following
recurrence relations:
uj+1
= uji +ln(1−k(6(uji )2
i
−3uji + 4uji+1 − uji+2 −5uji + 18uji+1 − 24uji+2 + 14uji+3 − 3uji+4
+
)), i = 1, 2
2h
2h3
(35)
uj+1
= uji +ln(1−k(6(uji )2
i
uji+1 − uji−1 uji+2 − 2uji+1 + 2uji−1 − uji−2
+
)), i = 3, N −2
2h
2h3
(36)
and
uj+1
= uji +ln(1−k(6(uji )2
i
3uji − 4uji−1 + uji−2 5uji − 18uji−1 + 24uji−2 − 14uji−3 + 3uji−4
+
)), i = N −1, N
2h
2h3
(37)
6. Numerical examples
In this section, we apply preceding algorithm to three numerical examples associated with the appropriate initial conditions
.
Example.1 consider the reaction-diffusions equation (1), when 2 ≼ x ≼ 12 in case
of p = 1, q = 1, at h = 1and k = 0.00001, we start with the initial approximation,
−x
x+t
u(x, 0) = 13 (3 + e 2 ). The exact solution is u(x, t) = 31 (3 + e 2 ).
.
Example.2 consider the mKdV equation (29), when 10 ≼ x ≼ 20 in case of h = 1and
k = 0.00001, we start with the initial approximation, u(x, 0) = sech(x). The exact
solution is u(x, t) = sech(x − t).
EJMAA-2016/4(1)
COMPOSITE FINITE DIFFERENCE
203
Table 2. Numerical results of solving mKdV equation at different times
x
10
11
12
13
14
15
16
17
18
19
20
t = 0.01
4.7 ∗ 10−7
1.7 ∗ 10−7
6.4 ∗ 10−8
1.2 ∗ 10−8
4.4 ∗ 10−9
1.7 ∗ 10−9
6.3 ∗ 10−10
2.3 ∗ 10−10
8.1 ∗ 10−11
1.0 ∗ 10−9
4.7 ∗ 10−10
t = 0.1
5.2 ∗ 10−6
1.6 ∗ 10−6
4.9 ∗ 10−7
1.3 ∗ 10−7
2.5 ∗ 10−8
2.0 ∗ 10−8
6.5 ∗ 10−9
2.8 ∗ 10−9
5.8 ∗ 10−10
9.8 ∗ 10−9
7.5 ∗ 10−9
t = 0.5
0.00009056
0.00001313
0.00001414
2.8 ∗ 10−6
1.1 ∗ 10−6
3.9 ∗ 10−7
5.6 ∗ 10−8
5.3 ∗ 10−8
1.6 ∗ 10−8
1.8 ∗ 10−7
1.3 ∗ 10−7
Table 3. Numerical results of solving FN equation at different times
x
0
1
2
3
4
5
6
7
8
9
10
t = 0.01
0.0002393
0.0001758
0.0000301
0.0000025
0.0000034
0.0000036
0.0000024
0.0000013
0.0000007
0.0000086
0.0000042
t = 0.1
0.0022600
0.0018887
0.0003588
0.0000086
0.0000312
0.0000350
0.0000237
0.0000133
0.0000027
0.0000931
0.0000238
t = 0.5
0.0045646
0.0126318
0.0032443
0.0004057
0.0001697
0.0001613
0.0001158
0.0000565
0.0000676
0.0006180
0.0005093
Example.3 consider the Fitzhugh?Nagumo equation (20), when 0 ≼ x ≼ 10 in
case of h = 1and k = 0.00001, we start with the initial approximation, u(x, 0) =
2p−1
1
x
1
x
√
√
2 (1 + tanh 2 2 ). The exact solution is u(x, t) = 2 (1 + tanh( 2 2 − 4 )), p = 0.75.
.
Tables (1- 3) illustrate the numerical results of solving RD equation, mKdV equation and FN equation using CFDS at different times.
7. Conclusion
The CFDS is effective for solving linear and nonlinear partial differential equations especially for small time intervals. The numerical results show that the solution using CFDS give high accuracy and no more conditions or restrictions are
needed.
References
[1] N. Britton, Reaction-Diffusion Equations and Their Applications to Biology, Academic Press,
1986.
[2] A. Biswas, 1-Soliton Solution of the Nonlinear Reaction-Diffusion Equation, 6, (2008), 1-5.
204
S. A. EL MORSY
EJMAA-2016/4(1)
[3] G. de Vries, T. Hillen, M. Lewis, J. Muller, and B. Schonfisch, A Course in Mathematical
Biology, SIAM, Philadelphia, 2006.
[4] P. Grindrod, The Theory and Applications of Reaction-Diffusion Equations, Oxford University Press, 1996.
[5] W. Hundsdorfer and J. Verwer, Numerical Solution of Time-Dependent Advection-Di?usionReaction Equations, Springer, 2003.
[6] D. Jones, M. Plank, and B. Sleeman, Differential Equations and Mathematical Biology, CRC
Press, 2010.
[7] ] A. Huber, On an improved method for solving evolution equations of higher order importantly in science and technology, International Journal of Engineering Science and Technology,
2, 5, (2010), 1-12.
[8] S. A. El Morsy, M. S. El-Azab and I. L. El-Kalla , (G’/G) and extended (G’/G) methods for
solving the nonlinear reaction-diffusion equation and KdVB equation, Electronic Journal of
Mathematical Analysis and Applications,. 3, (1), ,2015, pp. 97-110.
[9] J. Smoller, Shock Waves and Reaction-Diffusion Equations, Springer, 1994.
[10] E. Sontag, Lecture Notes on Mathematical Biology, Rutgers University, 2005.
[11] C. Collins, Length dependence of solutions of ?tzhugh-nagumo equations, Trans. Amer. Math.
Soc., 280:809?832, 1983.
[12] J. Rauch and J. Smoller, ?Qualitative theory of the FitzHugh-Nagumo equations,? Advances
in Mathematics, 27, 1, (1978), 12-44.
[13] H. P. McKean, Nagumo?s equation, Advances in Mathematics, 4, 3, (1970), 209-223.
[14] H. Li and Y. Guo, New exact solutions to the Fitzhugh-Nagumo equation, Appl. Math.
Comput., 180, (2006), 524-528.
[15] A. M. Wazwaz, A sine-cosine method for handling nonlinear wave equations, Math. Comput.
Modelling, 40, (2004), 499-508.
[16] R. Fitzhugh, Impulse and physiological states in models of nerve membrane, Biophys. J., 1,
(1961), 445-466.
[17] A.H. Bhrawy, A Jacobi?Gauss?Lobatto collocation method for solving generalized Fitzhugh?
Nagumo equation with time-dependent coefficients, Applied Mathematics and Computation,
222, (2013), 255?264.
[18] A. K. A. Khalifa, Theory and application of the collocation method with splines for ordinary
and partial differential equations [Ph.D. thesis], Heriot-Watt University, (1979).
[19] P. Gordon, Nonsymmetric difference equations, Journal of the Society For Industrial and
Applied Mathematics, 13, 3,(1965), 667?673.
[20] A. R. Gourlay, Hopscotch: a fast second-order partial differential equation solver, Journal of
the Institute of Mathematics and Its Applications, 6, (1970), 375?390.
[21] A. R. Gourlay, Some recent methods for the numerical solution of time-dependent partial
differential equations, Proceedings of the Royal Society, 323, (1971), 219-235.
[22] A. Biswas, Solitary wave solution for KdV equation with power law nonlinearity and time
dependent coefficients, Nonlinear Dynamics, 58, 1-2, (2009), 345-348.
[23] M. S. Ismail and A. Biswas, 1- Soliton solution of the generalized KdV equation with generalized evolution, Appl. Math. and Comput., 216, 5, (2010), 1673-1679.
[24] A. Biswas, M. D. Petkovic, D. Milovic and F. Majid, An exact solution of perturbed solitary
waves due to KdV equation, Australian Journal of Basic and Applied Sciences, 4, 8, (2010),
3154-3158.
[25] H. Triki, T. Hayat, O. M. Aldossary and A. Biswas, Solitary wave and shock wave solutions
to a second order wave equation of Kortewege ?de- Vries, Appl. Math. and Comput., 217, 21,
(2010), 8860-8863.
[26] E. V. Krishnan, Y. Peng, Exact solutions to the combined KdV-mKdV equation by the
extended mapping method, Phys. Sc. 73, (2006), 405-409.
[27] A. M. Wazwaz, New sets of solitary wave solutions to the KdV, mKdV and generalized KdV
equations, Commun. Nonl. Sci. Numer. Simulat., 13, (2008), 331-339.
[28] G. Johnpillai, Ch. M. Khalique and A. Biswas, Exact solutions of the mKdV equation with
time- dependent coefficients, Math. Commun. 16, (2011), 509-518.
[29] A. A. Soliman and M. A. Abdou, The decomposition method for solving the coupled modified
KdV equations, Math. and Comp. Model., 47, 9-10, (2008), 1035?1041.
EJMAA-2016/4(1)
COMPOSITE FINITE DIFFERENCE
205
[30] H. Tari, D. D. Ganji, and M. Rostamian, Approximate solutions of K (2,2), KdV and modified KdV equations by variational iteration method, homotopy perturbation method and
homotopy analysis method, Inter. J. Nonl. Sc. Numer. Simul., 8, 2, (2007), 203?210.
[31] J. H. Mathews and K. D. Fink, Numerical methods using matlab, Fourth Edition, Pearson
Education International (2004).
[32] M. S. El-Azab, I. L. El-Kalla and S. A. El Morsy, Composite finite difference scheme applied
to a couple of nonlinear evolution equations, Electronic Journal of Mathematical Analysis
and Applications,. 2, (2), 2014, pp. 254-263.
[33] M. C. Bhattacharya, A new improved finite difference equation for heat transfer during
transient change, Appl. Math. Model., 10, 1, (1986), 68?70.
[34] M. C. Bhattacharya, M. G. Davies, The comparative performance of some finite difference
equations for transient heat conduction problems, Int. J. Numer. Meth. Eng., 21, 7, (1987),
1317? 1331.
[35] R. F. Handschuh, T. G. Keith, Applications of an exponential finite-difference technique,
Numer. Heat Transfer Part A, 22, (1992), 363?378.
[36] A. R. Bahadir, Exponential finite difference method applied to Kortewge de Vries equation
for small times, Appl. Math. and Comp. ,160, 3, (2005), 675- 682.
S. A. El Morsy
Basic Science Departement, Higher Institute of Engineering and Technology, Nile
Academy for Science, Mansoura, Egypt
E-mail address: salwazaghrot@yahoo.com
Purchase answer to see full
attachment