Вы находитесь на странице: 1из 12

Iran J Sci Technol Trans Sci

DOI 10.1007/s40995-016-0136-2

RESEARCH PAPER

Analytical Solutions of Fractional Order Diffusion


Equations by Natural Transform Method
Kamal Shah1 • Hammad Khalil2 • Rahmat Ali Khan1

Received: 28 June 2016 / Accepted: 5 October 2016


Ó Shiraz University 2016

Abstract In this article, we develop an analytical method Anastassiou 2009; Atangana 2016). It is frequently
for solving fractional order partial differential equations. observed in many situations that the classical differential
Our method is the generalizations of homotopy perturba- operator, being local in nature, often cannot well explain
tions Laplace transform method and homotopy perturba- certain problems, specially problems having hereditary
tions Sumudu transform method. The solutions obtained properties. Fractional order operators are nonlocal in nat-
using the proposed method implies that the method is ure; in other words, they have memory effects and have the
highly accurate and easy to implement. The proposed embedded ability to describe these systems in best way as
method can be easily applied to a large variety of problems, compared with classical derivatives. The nonlocal property
which are modeled in terms of fractional order partial of FDEs is the only reason of its popularity in modeling
differential equations. Some test problems are solved to certain biological, physical, chemical, psychological,
show the accuracy of the proposed method. Some results mechanical system, and thermoplasticity. One of the non-
are shown graphically also. local process is the diffusion processes and is nicely
modeled by many authors using fractional order diffusion
Keywords Fractional partial differential equations  equations. Among others, we want to bring the attention of
Natural transform  Initial conditions  Diffusion equation  the readers to Wu et al. (2015), in which the authors
Homotopy perturbation method studied the fractional diffusion in discreet case. In Gómez-
Aguilar et al. (2016), the authors investigate fractional
order diffusion and present some simulations results. In
1 Introduction Morgado and Rebelo (2015), a detailed analysis is pre-
sented for the numerical investigation of distributed order
During the last few decades, fractional order differential diffusion equations. In Gómez-Aguilar et al. (2016a, b),
equations (FDEs) are proved to be of basic importance in Gómez–Aguilar presented a detailed study on the appli-
many physical and engineering processes. These equations cation of fractional order diffusion equations to some
are successfully applied to problems in various scientific applied problems of mechanics. The fractional order
fields, such as biology, economics, polymer chemistry, operators are nonlocal operators and provides a greater
mechanics, aerodynamics, control theory, biophysics, and degree of freedom in the model as compare to classical
many more (see Hilfer 2000; Kilbas et al. 2006; integer order which is local operator and does not allow
greater degree of freedom for modeling. On the other hand,
the computation complexities involved in fractional order
& Kamal Shah models make these models relatively difficult to analyze as
kamal@uom.edu.Pk
compared with integer order models, and therefore, it is
1
Department of Mathematics, University of Malakand, some time very difficult to solve these equations. Some
Chakadara Dir(L), Chakdara, Khyber Pakhtunkhwa, Pakistan times, it becomes difficult to arrive at the exact analytic
2
Department of Mathematics, University of Poonch solution of the problem. However, there do exist a large
Rawalakot, Rawalakot, AJK, Pakistan variety of methods, which proved to be helpful in finding

123
Iran J Sci Technol Trans Sci

approximate solution of the fractional order problems. oa vðx; tÞ o2 vðx; tÞ o


Among others, we aim to bring the attention of the readers a
¼A  ðf ðxÞvðx; tÞÞ;
ot ox2 ox
to finite difference method, Fourier series method, Ado- vðx; 0Þ ¼ gðxÞ; 0\a  1; A [ 0; ð1Þ
mian decomposition method, homotopy analysis method,
method of radial base function, wavelet techniques, spec- where v(x, t) is the probability density functions and is
tral methods (Khalil et al. 2015; Khalil and Khan 2015), used to find the position of a particle at instant t and at
and many more. These method have there own merits and position x and is the unknown solution to be determined.
demerits. Some of them provide a very good approxima- The positive constant A depends on the temperature, the
tion. For instance, in He (1999), he introduced homotopy friction coefficient, the universal gas constant, and Avo-
analysis method for the solutions of linear and nonlinear gadro number, while g(x) is the initial concentration of
differential equations. Later on, this method was modified the quantity v(x, t). The proposed idea is then extended to
to homotopy perturbation method (HPM) by He (2005). In two-dimensional time-fractional sub-diffusion equation of
Khan et al. (2011), authors established a numerical method the form
for the solutions of time-fractional diffusion equations. In
oa vðx; y; tÞ o2 vðx; y; tÞ o2 vðx; y; tÞ
Kexue and Jigen (2011), Kexue and Jigen applied Laplace ¼ þ ; 0\a  1;
ota ox2 oy2
transform method for the solutions of FDEs with constant
coefficients. Similarly, a combination of Laplace and a  x  b; c  y  d; ð2Þ
homotopy method is very nicely applied to diffusion subject to the initial condition vðx; y; 0Þ ¼ gðx; yÞ: The
equation of fractional order in Morales-Delgado et al. equation mentioned above have enormous applications in
(2016). Khan and Khan (2008) introduced the Natural mathematical physics, such as thermoelasticity, thermal
transform (NT) and applied them to solved some fluid stress theory, heat conduction process population dynam-
problems. Shah et al. (2015) extracted Laplace, Sumudu, ics, vibrational process, and nuclear reactors dynamics.
Fourier, and Mellin transform from the Natural transform, Analogous extension is used to study the dynamical
and showed that all these mentioned transforms converge behavior of three-dimensional time-fractional diffusion
to Natural transform. equation given as
In literature, one can see that different mathematical
methods have been utilized for the solutions of different oa vðx; y; z; tÞ o2 vðx; y; z; tÞ o2 vðx; y; z; tÞ
¼ þ
classes of FDEs. Silambarasn and Belgacem (2011) and ota ox2 oy2
Belgacem and Silambarasan (2012) used NT for the solu- o2 vðx; y; z; tÞ
tion of Maxwell’s equations. Shah and Khan (2015) þ ; 0\a  1; ð3Þ
oz2
applied NT for the analytical solutions of some nonho-
mogenous FDEs. In the context of numerical techniques subject to the initial condition vðx; y; z; 0Þ ¼ gðx; y; zÞ.
based on operational matrices, the work of Khalil and Khan The rest of the article is organized as: In Sect. 2, we
is of particular interest. We refer the readers to some of the recall some basic properties and definition form fractional
recent articles (Khalil et al. 2016; Khalil and Khan 2014; calculus and Natural transform. In Sect. 3, we present the
Mohebbi et al. 2013; Momani and Yildirim 2010), in idea of Natural-homotopy perturbation method. In Sect. 4,
which these authors established different efficient tech- we solve various problems to support the accuracy and
niques for solving problems from fractional calculus. efficiency of the proposed method, while the last section is
Kumar et al. (2012) applied Laplace homotopy perturba- devoted to a short conclusion.
tion method to solve diffusion equation of fractional order.
In Rathore et al. (2012), the authors applied homotopy
analysis Sumudu transform method to solve nonlinear 2 Preliminaries
Folkar–Plank equations. Moreover, for some basic methods
for the numerical solutions of diffusions equations, see We recall some basic definitions and known results from
Pang et al. (2015), Fu et al. (2013), Sun et al. (2013) and fractional calculus and applied analysis. For more details
Abdel-Salam et al. (2016). study in this area of research, we refer the reader to the
Motivated by the above-mentioned articles, we proposed studies Hilfer (2000), Kilbas et al. (2006), Anastassiou
a new numerical method Natural-homotopy perturbation (2009), Khan and Khan (2008), Silambarasn and Belgacem
method (NHPM). This method is a combination of Natural (2011), Belgacem and Silambarasan (2012).
transform and homotopy perturbation method, and is Definition 2.1 The Rieman–Liouville fractional integral
applied to find approximate solutions of fractional order of order a 2 Rþ of a function hðtÞ 2 Lð½0; 1; RÞ is defined
diffusion equation of the form by

123
Iran J Sci Technol Trans Sci

Z t   n1 nk1  
1 oa f ðx; tÞ sa X s oa f ðx; tÞ
I0a hðtÞ ¼ ðt  sÞa1 hðsÞ ds; ð4Þ N ¼ a Rðx; s; uÞ  lim :
CðaÞ 0 ota u unk t!0 ota
k¼0
provided that the integral on the right side converges. ð10Þ
Definition 2.2 (Kumar et al. 2012) For l 2 R, a function Lemma 2.9 (Belgacem and Silambarasan 2012) The NT
f : R ! Rþ is said to be in the space Cl if it can be written of a order partial derivative of f(x, t) w.r.t x is defined by
as f ðxÞ ¼ xq f1 ðxÞ with q [ l; f1 ðxÞ 2 C½0; 1Þ and it is in  a 
o f ðx; tÞ da
space f ðxÞ 2 Cln if f ðnÞ 2 Cl for n 2 N [ f0g. N a
¼ a Rðx; s; uÞ: ð11Þ
ox dx
Definition 2.3 Caputo’s fractional order derivative of a Lemma 2.10 (Shah et al. 2015) The dual relationship
n
function h 2 C1 with n 2 N [ f0g is defined by between Natural transform and Laplace transform is given
8 by
< I na f ðnÞ ; n  1\a  n; n 2 N; Z
a
Dt hðtÞ ¼ dn ð5Þ 1 1 st 1
: N ½ f ðx; tÞ ¼ Rðx; s; uÞ ¼ e u f ðx; tÞdt ¼ Lff ðx; tÞg;
n
hðtÞ; a ¼ n; n 2 N: u 0 u
dt
ð12Þ
where L is the laplace transform.
Note Through out in this paper, we use fractional order
derivative in Caputo’s sense. From the above Lemma, we can see that the Natural
transform is the generalization of Sumudu transform and
Definition 2.4 A two-parameter Mittag–Leffler function
Laplace transform. It is to be noted that for u ¼ 1, the
is defined by the following:
Natural transform becomes the Laplace transform, and for
X1
tk the case s ¼ 1, the generalization leads us to Sumudu
Ea;b ðtÞ ¼ : ð6Þ
Cðka þ bÞ transform.
k¼0

For a ¼ b ¼ 1, E1;1 ðtÞ ¼ et and E1;1 ðtÞ ¼ et .


Definition 2.5 The NT of a function v(x, t) for t  0 is 3 Natural Transform Homotopy Perturbation
defined by Method
Z 1
N ½vðx; tÞ ¼ Rðx; s; uÞ ¼ est vðx; utÞ dt; ð7Þ In this section, we describe the main idea of Natural-ho-
0 motopy perturbation method. Consider the following gen-
where s and u are the transform parameters and are eral class of fractional order partial differential equations:
assumed to be real and positive. Dat vðx; tÞ þ L½xvðx; tÞ þ N½xvðx; tÞ
Definition 2.6 The NT of Mittag–Leffler function Ea;b is ¼ hðx; tÞ; t [ 0; 0\a  1; ð13Þ
defined by vðx; 0Þ ¼ gðxÞ; x 2 R:
Z X1 a
1
ukþ1 Cðk þ 1Þ where Dat ¼ oto a is the derivative defined in Caputo’s sense,
N ½vðx; tÞ ¼ est vðx; utÞdt ¼ :
0 k¼0
skþ1 Cðka þ bÞ L[x] is linear operator, and N[x] is the nonlinear operator in
x, and h(x, t) is a given source terms of the system. By the
ð8Þ
definition of Natural transform, we can write (13) as
Definition 2.7 The NT of Da f ðtÞ in Miller and Ross sense  
N Dat vðx; tÞ þ L½xvðx; tÞ þ N½xvðx; tÞ ¼ N ½hðx; tÞ;
is defined by the following:
gðxÞ ua ua
X
n1 nk1 Rðx; s; uÞ ¼ þ a N ½hðx; tÞ  a N ½L½xvðx; tÞ þ N½xvðx; tÞ:
sa s s s s
N ð Da f ðtÞÞ ¼ Rðs; uÞ  f ðkÞ ð0Þ; n  1\a  n:
u a u nk ð14Þ
k¼0
ð9Þ Now, implementing inverse Natural transform on both
sides of above estimates, we can write
Lemma 2.8 (Silambarasn and Belgacem 2011; Belgacem  a 
oa f ðx;tÞ 1 u
and Silambarasan 2012) The NT of ota w.r.t t can be vðx; tÞ ¼ Fðx; tÞ  N N ½L½xvðx; tÞ þ N½xvðx; tÞ ;
calculated as: sa
ð15Þ

123
Iran J Sci Technol Trans Sci

where X
M

a
 a
 vðx; tÞ ¼ lim vk ðx; tÞ: ð22Þ
gðxÞ u u M!1
Fðx; tÞ ¼ N 1 ½ þ a N ½hðx; tÞ ¼ gðxÞ þ N 1 a N ½hðx; tÞ : k¼1
s s s
ð16Þ The obtained solution is rapidly convergent series and
converges uniformly to the exact solution of the problem.
Now, we define formal perturbation technique in the form We present detailed analysis of the convergence using
of power series in the parameter e as classical ratio or root test. These techniques can be found in
X
1 Rudin (1964). However, for sufficient condition of con-
vðx; tÞ ¼ ek vk ðx; tÞ; ð17Þ vergence of this method, we give the following theorem
k¼0 Biazar and Aminikhah (2009) and Ayati and Biazar (2015).
where e is perturbation parameter and e 2 ½0; 1. Theorem 3.1 Let X and Y be two Banach spaces and
The nonlinear term can be decomposed as T : X ! Y be a contractive nonlinear operator, such that
X1
Nvðx; tÞ ¼ ek Hn ðvÞ; ð18Þ for all u; u 2 X; kT ðuÞ  T ðu Þk  jku  u k; 0\j\1:
k¼0
ð23Þ
where Hn are He’s polynomials of the form
v0 ; v1 ; v2 ; . . .; vn ; and can be calculated as Then, in view of Banach contraction theorem, T has a
" !# unique fixed point u, such that T u ¼ u: Let us write the
1 X 1
generated series (22), by the aforementioned Natural
Hn ðv0 ; v1 ; . . .; vn Þ ¼ Dk N ei v i :
Cðn þ 1Þ e k¼0
transform homotopy perturbation method as
e¼0
ð19Þ X
n1
xn ¼ T ðxn1 Þ; xn1 ¼ ui ; n ¼ 1; 2; 3; . . .; ð24Þ
ok
where Dke
¼ : Using relation (17) and (18) in (14) and
oek
i¼1

constructing the Homotopy similar to that introduced by and assumed that x0 ¼ u0 2 Sr ðuÞ; where Sr ðuÞ ¼ fu 2
He Ray and Bera (2006), we obtain the couple of the X : ku  uk \r g; then, we have
Natural transform and homotopy perturbation method as
ðA1 Þ xn 2 Sr ðuÞ
X
1
ðA2 Þ limn!1 xn ¼ u:
k
e vk ðx; tÞ ¼ Fðx; tÞ  e
k¼0
" #!
1 ua X1 X1 Proof
 N a
N fL ek vk ðx; tÞ þ ek Hk ðvÞg :
s k¼0 k¼0 ðA1 Þ In view of mathematical induction for n ¼ 1, we
ð20Þ have

By equating coefficient of e on both sides, we get kx1  u1 k ¼ kT ðx0 Þ  T ðuÞk  jku0  uk:
0
e : v0 ðx; tÞ ¼ Fðx; tÞ; Let the result is true for n  1, then
 a 
1 u
e1 : v1 ðx; tÞ ¼ N N ð L½xv 0 ðx; tÞ þ H0 ðvÞ Þ ; kxn1  uk  jn1 ku0  uk:
sa
 a 
u We have
e1 : v2 ðx; tÞ ¼ N 1 a N ðL½xv1 ðx; tÞ þ H1 ðvÞÞ ;
s kxn  uk ¼ kT ðxn1 Þ  T ðuÞk  jkxn1  uk  jn kx0  uk:
..
.
 a  Hence, using ðA1 Þ, we have
k 1 u
e : vn ðx; tÞ ¼ N N ðL½xvk1 ðx; tÞ þ Hk1 ðvÞÞ ; kxn  uk  jn ku0  uk  jn r\r;
sa
k [ 0; k 2 N: ð21Þ which implies that xn 2 Sr ðuÞ:
ðA2 Þ: Since kx  uk  jn ku0  uk and as limn!1 jn ¼ 0:
The component vk ðx; tÞ can be easily calculated, which
Therefore; we have limn!1 kxn  uk ¼ 0 ) limn!1 xn ¼ u:
leads us to the determination of analytical solution in terms
of rapidly convergent series. Considering the limiting value
h
e ! 1, we get

123
Iran J Sci Technol Trans Sci

We carry out some numerical tests and calculate the e0 : v0 ðx; tÞ ¼ gðxÞ;
absolute error at different levels of truncation. We observe  a 
1 u
that the absolute error for the studied cases decreases e1 : v1 ðx; tÞ ¼ N N ½ðv Þ
0 xx þ ðxv Þ
0 x  ;
sa
rapidly.
..
It is to be noted that the proposed algorithm is gener- .
alization of Homotopy-Sumudu perturbation method. This  a 
k 1 u
framework works well to find the analytical solution of e : v1 ðx; tÞ ¼ N N ½ðvk Þxx þ ðxvk Þx  ;
sa
fractional order partial differential equations only under
some initial condition in time. To the best of our knowl- k  0; k ¼ 1; 2; 3; . . .: ð29Þ
edge, all methods coupling of homotopy perturbation and Case I. Consider gðxÞ ¼ 1 and applying the proposed
some integral transforms are mostly applicable for solving method, we have
initial value problems and these are not applicable to
problems involving some boundary constrains. Therefore, v0 ðx; tÞ ¼ 1;
this frame work is designed only for transient state problem  a  2 
1 u o v0 ðx; tÞ o
where only the initial distribution of the quantity v(x, 0) is v1 ðx; tÞ ¼ N N þ ðxv0 ðx; tÞÞ
sa ox2 ox
given. The extension of the proposed method to high-di- a
ta
1 u
mensional fractional order partial differential equations is ¼ N ð aþ1 Þ ¼
s Cða þ 1Þ
analogous.  a  2 
1 u o v1 ðx; tÞ o
v2 ðx; tÞ ¼ N N þ ðxv1 ðx; tÞÞ
sa ox2 ox
2a 2a
4 Test Problems 1 u t
¼ N ð 2aþ1 Þ ¼
s Cð2a þ 1Þ
The following test problem are solved to show the effi- ..
ciency of proposed method. .
 a  2 
Example 1 Consider the following one-dimensional time 1 u o vk1 ðx; tÞ o
vk ðx; tÞ ¼ N N þ ðxvk1 ðx; tÞÞ
diffusion equations as: sa ox2 ox
ka ka
1 u t
oa vðx; tÞ o2 vðx; tÞ o ¼ N ð kaþ1 Þ ¼ : ð30Þ
¼ þ ðxvðx; tÞÞ; s Cðka þ 1Þ
ota ox2 ox ð25Þ
vðx; 0Þ ¼ gðxÞ; 0\a  1: By mean of these terms, the solution v(x, t) is given by

Taking Natural transform of (25), we have X


1 X
1
tka
 2  vðx; tÞ ¼ lim ek vk ðx; tÞ ¼ ¼ Ea ðta Þ:
gðxÞ ua o vðx; tÞ o e!1
k¼0 k¼0
Cðka þ 1Þ
Rðx; s; uÞ ¼ þ aN þ ðxvðx; tÞÞ : ð26Þ
s s ox2 ox ð31Þ
1
Taking inverse Natural transform, we get If we take a ¼ 12, then vðx; tÞ ¼ E1=2 ðt Þ which has complete
2

 a  2  agreement to the solution obtain in Das (2009).


1 u o vðx; tÞ o
vðx; tÞ ¼ gðxÞ þ N N þ ðxvðx; tÞÞ :
sa ox2 ox Case II. If we take the initial condition v0 ðx; tÞ ¼ gðxÞ ¼ x,
ð27Þ and follow the above procedure as in case I, then we obtain
solution in the form:
Now, applying the aforementioned homotopy perturbation
X
1
ð2ta Þk
method as in (20), we have vðx; tÞ ¼ x ¼ xEa ð2ta Þ: ð32Þ
    k¼0
Cðka þ 1Þ
X
1
ua 2
o vðx; tÞ o
ek vk ðx; tÞ ¼ gðxÞ þ e N 1 N þ ðxvðx; tÞÞ : 1
k¼1
sa ox2 ox If we put a ¼ 12, then our solution is vðx; tÞ ¼ xE1=2 ð2t2 Þ
ð28Þ which has complete agreement to the solution obtain in Das
(2009).
Now, equating the coefficient of the same power of e, we
get Case III. When gðxÞ ¼ x2 , we obtain

123
Iran J Sci Technol Trans Sci

"
v0 ðx; tÞ ¼ x2 ; X
1
k ua 1 o2 X 1
 a  e vk ðx; tÞ ¼ x þ e N N ½ ek vk ðx; tÞ
u
1 o2 v0 ðx; tÞ o k¼1
sa ox 2
k¼1
v1 ðx; tÞ ¼ N N½ þ ðxv0 ðx; tÞ #!!
sa ox2 ox X1 X
1
a k o k
t þ e vk ðx; tÞ ½ e vk ðx; tÞ :
¼ ð2 þ 3x2 Þ ; k¼1
ox k¼1
Cða þ 1Þ
 a  ð37Þ
1 u o2 v1 ðx; tÞ o
v2 ðx; tÞ ¼ N N½ þ ðxv1 ðx; tÞ
sa ox2 ox Now, equating the coefficient of the same power of e of
t 2a both sides, we get
¼ ð8 þ 9x2 Þ ;
Cð2a þ 1Þ e0 : v0 ðx; tÞ ¼ x;
..  
ua
. e1 : v1 ðx; tÞ ¼ N 1 N ½ðv0 Þxx þ v0 ðv0 Þx 
 a  sa
1 u o2 vk1 ðx; tÞ o ta
vk ðx; tÞ ¼ N N½ þ ðxvk1 ðx; tÞ ¼x ;
sa ox2 ox Cða þ 1Þ
 
t ka ua
¼ ½x2 þ ð1 þ x2 Þð3k  1Þ ; e2 : v2 ðx; tÞ ¼ N 1 N ½ðv1 Þxx þ v0 ðv1 Þx þ v1 ðv0 Þx 
Cðka þ 1Þ sa
Therefore; theexactsolutionisgivenby t2a
¼ 2x ;
Cð2a þ 1Þ
X
1  a
vðx; tÞ ¼ lim ek vk ðx; tÞ u
e3 : v3 ðx; tÞ ¼ N 1 a N ½ðv2 Þxx þ v0 ðv2 Þx þ v1 ðv1 Þx
e!1
k¼0 s

X
1
½x2 þ ð1 þ x2 Þð3k  1Þtka þv2 ðv0 Þx 
¼ : ð33Þ  
Cðka þ 1Þ Cð2a þ 1Þ t3a
k¼0 ¼ 4x þ x 2 ;
C ða þ 1Þ Cð3a þ 1Þ
If we put a ¼ 12, then our solution has complete agreement and so on:
to the solution obtain in Das (2009). Thus, in all of the
above cases, the solutions obtain by NTHPM are the same ð38Þ
as obtain by LTHPM and Adomian decomposition method Thus, the solution in the form of convergent series by
(ADM). The Natural transform homotopy perturbation taking e ! 1 is given by
method solution is almost valid for wide range of times.
The solution obtain by this method is highly accurate. vðx; tÞ ¼ v0 þ v1 þ v2 þ v3 þ   

ta t2a
Example 2 Consider the nonlinear Burger equation given ¼ x 1  þ2
Cða þ 1Þ Cð2a þ 1Þ
as   
Cð2a þ 1Þ t3a
oa vðx; tÞ o2 vðx; tÞ ovðx; tÞ  4þ 2 þ  : ð39Þ
¼ þv ; C ða þ 1Þ Cð3a þ 1Þ
ota ox2 ox
1 Particularly, putting a ¼ 1 in series(39), we get the exact
vðx; 0Þ ¼ x; vð0; tÞ ¼ ; vð1; tÞ ¼ 0; 0\a  1:
1þt solution
ð34Þ x
vðx; tÞ ¼ xð1  t þ t2  t3 þ   Þ ¼ ; ð40Þ
1þt
Taking Natural transform of (34), we have
 2  which is the exact solution obtained in Jafari and Ami-
x ua o vðx; tÞ ovðx; tÞ
Rðx; s; uÞ ¼ þ aN þv : ð35Þ nataei (2010), using improved homotopy perturbation
s s ox2 ox method.
Taking inverse Natural transform, we get Example 3 In this example, we consider the following
 a  2  two-dimensional diffusion equation given by:
1 u o vðx; tÞ ovðx; tÞ
vðx; tÞ ¼ x þ N N þv :
sa ox2 ox oa vðx; y; tÞ o2 vðx; y; tÞ o2 vðx; y; tÞ
¼ þ ;
ð36Þ ota ox2 oy2 ð41Þ
0\a  1; a  x  b; c  y  d;
Now, applying the aforementioned homotopy perturbation
method as in (20), we have

123
Iran J Sci Technol Trans Sci

subject to the initial condition vðx; y; 0Þ ¼ sin x sin y: Tak- M ¼ 10; 12; 14: The absolute error is approximated at three
ing the Natural transform of (41), we have different values of time t. It is observed that as the trun-
sin x sin y ua cation level increases, the solution becomes more and more
Rðx; y; s; uÞ ¼ þ a N ½vxx þ vyy : ð42Þ accurate. This phenomena is illustrated in Fig. 1. It can also
s s
be observed that the solution is more accurate for small
Taking inverse Natural transform, we have values of t relatively to larger values of t. Similarly, as the
 a  value of t increases, the solution looses its accuracy, which
1 u
vðx; y; tÞ ¼ sin x sin y þ N N ½v xx þ v yy  : ð43Þ means that more terms are required to get approximate
sa
solution for a desired accuracy. The details of these anal-
Now, using the aforementioned homotopy perturbation ysis are shown in Fig. 1. Absolute error at fractional value
method, we get of a; that is, a ¼ 0:5 is shown in Fig. 2. It is observed that
X
1 the absolute error for fractional value of a is not enough
ek vk ðx; y; tÞ ¼ sin x sin y accurate as compared with integer value of a: We calculate
k¼0
" !#!! the approximate solution of the problem at fractional val-
 X
1
1 ua o2 o2 ues of a ¼ 0:5, as shown in Fig. 3, at different points of the
þe N N þ ek vk ðx; y; tÞ : space, and observe that at each point, the solution
sa ox2 oy2 k¼0
approaches uniformly to the solution at integer values of a
ð44Þ as the order of derivative increases from fractional to
Comparing coefficients of the same power of e integer.

e0 : v0 ðx; y; tÞ ¼ sin x sin y; Example 4 Consider the following three-dimensional


 a  2   diffusion equation:
1 u o o2
e1 : N N þ v 0 ðx; y; tÞ
sa ox2 oy2 oa vðx; y; z; tÞ o2 vðx; y; z; tÞ o2 vðx; y; z; tÞ
t a ¼ þ
¼ 2 sin x sin y ; ota ox2 oy2
Cða þ 1Þ ð47Þ
 a  2   o2 vðx; y; z; tÞ
o o2 þ ; 0\a  1;
1 u oz2
e2 : N N þ v 1 ðx; y; tÞ
sa ox2 oy2
ð45Þ subject to the initial condition
t2a
¼ ð2Þ2 sin x sin y ;
Cð2a þ 1Þ vðx; y; z; 0Þ ¼ sin x sin y sin z; a  x  b; c  y  d; m  z  n:
..
.
 a  2   Taking Natural transform of (47) of both sides, we have
1 u o o2
ek : N N þ v k1 ðx; y; tÞ sin x sin y sin z ua
sa ox2 oy2 Rðx; y; z; s; uÞ ¼ þ a N ½vxx þ vyy þ vzz :
s s
tka ð48Þ
¼ ð2Þk sin x sin y :
Cðka þ 1Þ
Taking inverse Natural transform, we have
From these estimates, we have  a 
1 u
X
M vðx; y; z; tÞ ¼ sin x sin y sin z þ N N ½v xx þ vyy þ v zz  :
sa
vðx; y; tÞ ¼ vk ðx; y; tÞ: ð46Þ
k¼0 ð49Þ
The exact solution of the above problem is given as Now, applying the aforementioned homotopy perturbation
sinðxÞsinðyÞEa ð2ta Þ: It is also evident that as M ! 1, the method, we get
approximate solution becomes equal to the exact solution. X
1
In practical, we can only find the finite sum, therefore, to ek vk ðx; y; z; tÞ ¼ sin x sin y sin z
k¼0
check the accuracy, convergence, and efficiency of pro- "  X !#!!
1
posed method, we calculate the maximum absolute error at 1 ua o2 o2 o2 k
þe N N þ 2þ 2 e vk ðx; y; z; tÞ :
different levels of truncation M, using different choices of sa ox 2 oy oz k¼0
parameters a: In particular, if a ¼ 1, the exact solution is ð50Þ
vðx; y; tÞ ¼ e2t sin x sin y. We fix a ¼ 1 and measure the
absolute difference between the exact solution and the Now, equating coefficients of the same power of e on both
series truncated at three different values of M, that is, for sides

123
Iran J Sci Technol Trans Sci

Fig. 1 Absolute error of Example 3 at a ¼ 1 for various values of t and at different values of M

Fig. 2 Absolute error of Example 3 at a ¼ 0:5 for various values of t and at different values of M

123
Iran J Sci Technol Trans Sci

Table 1 Example 4, absolute


t a¼1 a ¼ 0:6 a ¼ 0:5 a ¼ 0:4
error at different values of a at
point ðx; y; zÞ ¼ ð0:3; 0:3; 0:3Þ 0.1 2:4147e015 6:3838e015 1:1546e014 1:6875e014
0.2 2:4347e012 6:4080e012 1:1570e011 1:7106e011
0.3 1:3796e010 3:6310e010 6:5562e010 9:6933e010
0.4 2:4080e009 6:3375e009 1:1443e008 1:6919e008
0.5 2:2048e008 5:8027e008 1:0477e007 1:5491e007
0.6 1:3424e007 3:5332e007 6:3795e007 9:4321e007
0.7 6:1686e007 1:6235e006 2:9314e006 4:3341e006
0.8 2:3069e006 6:0715e006 1:0963e005 1:6208e005
0.9 7:3718e006 1:9402e005 3:5032e005 5:1795e005
1 2:0810e005 5:4769e005 9:8891e005 1:4621e004

Table 2 Example 4, absolute


t a¼1 a ¼ 0:6 a ¼ 0:5 a ¼ 0:4
error at different values of a at
point ðx; y; zÞ ¼ ð0:9; 0:9; 0:9Þ 0.1 6:3838e015 1:4211e014 2:1871e014 2:6867e014
0.2 6:4080e012 1:4347e011 2:2143e011 2:7074e011
0.3 3:6310e010 8:1293e010 1:2547e009 1:5341e009
0.4 6:3375e009 1:4189e008 2:1900e008 2:6776e008
0.5 5:8027e008 1:2992e007 2:0052e007 2:4517e007
0.6 3:5332e007 7:9103e007 1:2209e006 1:4928e006
0.7 1:6235e006 3:6348e006 5:6101e006 6:8593e006
0.8 6:0715e006 1:3593e005 2:0980e005 2:5652e005
0.9 1:9402e005 4:3438e005 6:7043e005 8:1972e005
1 5:4769e005 1:2262e004 1:8926e004 2:3140e004

Fig. 3 Approximate solution of x 10


−3

Example 3 at different values of 0.2


x=y=0.5
8 x=y=0.1
a ¼ 0:5 evaluated at different 0.15
α=0.5
points of the space. Note that 6
here, we use the truncation level 0.1
4
M ¼ 10 α=0.6
2 0.05

0 0
α=0.7
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
t t
α=0.8
0.6 x=y=1
0.4 x=y=0.8
0.3 0.4 α=0.9
0.2
0.2
0.1
α=1
0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

t t

123
Iran J Sci Technol Trans Sci

Fig. 4 Approximate solution of


Example 4 at different values of
a and t ¼ 0:5 evaluated at
different points of the space.
Note that here, we use the
truncation level M ¼ 16

e0 : v0 ðx; y; z; tÞ ¼ sin x sin y sin z; approximate solution. We approximate solution, and for
 a  2   illustration, we fix z. We observe that the solution
1 u o o2 o2
e1 : N N þ þ v 0 ðx; y; z; tÞ approaches uniformly to the exact solution at integer
sa ox2 oy2 oz2 value as the order of derivative approaches from frac-
ta tional to integer. These results at two different points of
¼ 3 sin x sin y sin z ;
Cða þ 1Þ z are shown in Figs. 3 and 4.
 a  2  
1 u o o2 o2
e2 : N N þ þ v 1 ðx; y; z; tÞ
sa ox2 oy2 oz2
t2a 5 Conclusion and Future Work
¼ ð3Þ2 sin x sin y sin z ;
Cð2a þ 1Þ
We have developed a new modified method which is the
..
. couple of Natural transform and Homotopy perturbation
 a  2   together with He’s polynomials. The proposed method is
1 u o o2 o2
ek : N N þ þ v k1 ðx; y; z; tÞ an effective technique which is the generalization of the
sa ox2 oy2 oz2
Laplace homotopy perturbation method and Sumudu
tka transform Homotopy perturbations method. If we replace
¼ ð3Þk sin x sin y sin z : ð51Þ
Cðka þ 1Þ the transform parameter u by 1, we get Laplace transform,
and if we replace s by 1, we get Sumudu transform. The
Using the above estimates, we can write the approximate
method presented in this paper is a simple and high accu-
solution as
rate method to find approximate and analytical solutions to
X
M
both fractional order ordinary as well as fractional order
vðx; y; z; tÞ ¼ lim ek vk ðx; y; z; tÞ: ð52Þ partial differential equations. Using few iterations produces
e!1
k¼0
highly accurate solution. From the test examples, it is
In particular, if we have a ¼ 1 in (52), the exact solution evident that uni- or multi-dimensional problems can be
is vðx; y; z; tÞ ¼ e3t sin x sin y sin z. To check the accuracy easily solved using the proposed method. The proposed
of the proposed method, we calculate absolute error method has the ability to find the solution of a given partial
between exact and approximate solution. For illustration differential equations only with the initial conditions, and
purpose, we display the absolute difference at a specific is, therefore, only valid for transient problems. The
point of the space using different choices of a, and adjustment of this method for solution of boundary value
observe that the solution is highly accurate for all problem is still an open problem at the time of writing. Our
t 2 ½0; 1. These results are displayed in Tables 1 and 2. It future work is related to establish this technique for solving
is one of the basic properties of fractional order problems boundary value problems of fractional order differential
that the solution of fractional order problems approaches equations. We hope that in future, we will apply this
to solution at integer order value as the order of derivative technique for solving others fractional order nonlinear
approaches from fractional to integer. To verify this models of partial differential equations which, up to date,
property, we also study the effect of parameter a on the have not solved numerically or analytically.

123
Iran J Sci Technol Trans Sci

References Khalil H, Rahmat Ali Khan, Mohammed Al Smadi, Asad Freihat


(2015) Approximation of solution of time fractional order three-
Abdel-Salam, Emad AB, Eltayeb A, Yousif, Mostafa A, El-Aasser dimensional heat conduction problems with Jacobi Polynomials.
(2016) Analytical solution of the space-time fractional nonlinear J Math 47(1):35–56
Khan NA, Ayaz M, Khan NU, Jin L (2011) On approximate solutions
Schrdinger equation. Rep Math Phys 77(1):19–34
Anastassiou GA (2009) On right fractional calculus. Chaos Solit Frac for the time-fractional reaction-diffusion equation of Fisher type.
42(1):365–376 Int J Phys Sci 6(10):2483–2496
Atangana A (2016) On the new fractional derivative and application Khan ZH, Khan WA (2008) N-transform properties and applications.
NUST J Eng Sci 1(1):127–133
to nonlinear Fisher’s reaction-diffusion equation. Appl Math
Comput 273:948–956 Kilbas AA, Srivastava HM, Trujillo JJ (2006) Theory and applica-
Ayati Z, Biazar J (2015) On the convergence of homotopy tions of fractional differential equations, 24. North-Holland
perturbation method. J Egypt Math Soc 23:424–428 Mathematics Studies, Amsterdame
Belgacem FBM, Silambarasan R (2012) Theory of natural transform, Kumar S, Yildirim A, Khan Y, Wei L (2012) A fractional model of
aerospace (MESA). Math Eng Sci 3(1):99–124 the diffusion equation and its analytical solution using Laplace
Biazar J, Aminikhah H (2009) Study of convergence of homotopy transform. Sci Iran B 19(4):1117–1123
perturbation method for system of partial differential equations. Liao SJ (2003) Beyond perturbation introduction to homotopy
Comput Math Appl 58:2221–2230 analysis method. CRC Press Company, Boca Raton
Das S (2009) Analytical solution of a fractional diffusion equation by Luchko YU, Gorenflo R (1999) An operational method for solving
variational iteration method. Comput Math Appl 57(3):483–487 fractional differential equations with the Caputo derivatives.
Acta Math Vietnam 24:207–233
Das S (2009) Analytical solution of a fractional diffusion equation
by variational iteration method. Comput Math Appl Madani M, Fathizadeh M, Khan Y, Yildirim A (2011) On the
57(3):483–487 coupling of the homotopy perturbation method and Laplace
Fu Z, Chen W, Yang H (2013) Boundary particle method for Laplace transformation. Math Comput Model 53(9):1937–1945
transformed time fractional diffusion equations. J Comput Phys Moaddy K, Momani S, Hashim I (2011) The non-standard finite
235(15):52–66 difference scheme for linear fractional PDEs in fluid mechanics.
Gómez-Aguilar JF, López–López MG, Alvarado-Martnez VM, Comput Math Appl 61:1209–1216
Reyes-Reyes J, Adam-Medina M (2016) Modeling diffusive Mohebbi A, Abbaszadeh M, Dehghan M (2013) A high-order and
transport with a fractional derivative without singular kernel. unconditionally stable scheme for the modified anomalous
Phys A Stat Mech Appl 447:467–481 fractional sub-diffusion equation with a nonlinear source term.
Gómez-Aguilar JF, Miranda-Hernández M, López-López MG, J Comput Phys 240:36–48
Momani S, Yildirim A (2010) Analytical approximate solutions of the
Alvarado-Martı́nez VM, Baleanu D (2016) Modeling and
simulation of the fractional space-time diffusion equation. fractional convection-diffusion equation with nonlinear source
Commun Nonlinear Sci Numer Simul 30(1):115–127 term by He’s homotopy perturbation method. Int J Comp Math
Gómez-Aguilar JF, Torres L, Ypez-Martnez H, Baleanu D, Reyes JM, 87(5):1057–1065
Morales-Delgado VF, Gmez-Aguilar JF, Yepez-Martnez H, Baleanu
Sosa IO (2016) Fractional Linard type model of a pipeline within
the fractional derivative without singular kernel. Adv Diff Equ D, Escobar-Jimenez RF, Olivares-Peregrino VH (2016) Laplace
2016(1):1–13 homotopy analysis method for solving linear partial differential
He JH (1999) Homotopy perturbation technique. Comput Methods equations using a fractional derivative with and without kernel
Appl Mech Eng 178(3):257–262 singular. Adv Diff Equ 2016(1):1–12
He JH (2000) Coupling method of a homotopy technique and a Morgado ML, Rebelo M (2015) Numerical approximation of
perturbation technique for non-linear problems. J Nonlinear distributed order reaction-diffusion equations. J Comput Appl
Mech 35(1):37–43 Math 275:216–227
He JH (2005) Application of homotopy perturbation method to Moustafa OL (2003) On the Cauchy problem for some fractional
nonlinear wave equations. Chaos Solitons Fract 26(3):695–700 order partial differential equations, Chaos Solitons Fractals.
Hilfer R (ed) (2000) Application of fractional calculus in physics. Chaos Solitons Fract 18(1):135–140
Pang G, Chen W, Fu Z (2015) Space-fractional advection-
World scientific publishing Co., Singapore
Jafari MA, Aminataei A (2010) Improved homotopy perturbation dispersion equations by the Kansa method. J Comput Phys
method. Int Math Forum 5(32):1567–1579 293:280–296
Javidi M, Ahmad B (2013) Numerical solution of fractional partial Rathore S, Kumar D, Sing J, Gupta S (2012) Homotopy analysis
differential equations by numerical Laplace inversion technique. Sumudu transform method for nonlinear equations. Int J Ind
Adv Diff Equ 2013(1):1–17 Math 4(4):301–314
Kexue L, Jigen P (2011) Laplace transform and fractional differential Ray SS, Bera RK (2006) Analytical solution of a fractional diffusion
equations. Appl Math Lett 24:2019–2023 equation by Adomian decomposition method. Appl Math
Khalil H, Khan RA (2014) A new method based on Legendre Comput 174(1):329–336
polynomials for solutions of the fractional two-dimensional heat Rudin W (1964) Principles of mathematical analysis, 2nd edn.
conduction equation. Comput Math Appl 67:1938–1953 McGraw-Hill, New York
Shah K, Junaid M, Ali N (2015) Extraction of Laplace, Sumudu,
Khalil H, Mohammed Al-Smadi, Khaled Moaddy, Rahmat Ali Khan,
Ishak Hashim (2016) Toward the approximate solution for Fourier and Mellin transform from the Natural transform. J Appl
fractional order nonlinear mixed derivative and nonlocal Envron Biol Sci 5(9):1–10
boundary value problems. Discrete Dyn Nat. Soc 2016:12 Shah K, Khan RA (2015) The applications of natural transform to the
analytical solutions of some fractional order ordinary differential
(Article ID 5601821)
Khalil H, Rahmat Ali Khan (2015) Extended spectral method for equations. Sindh Univ Res J 47(4):683–686
fractional order three-dimensional heat conduction problem. Silambarasn R, Belgacem FBM (2011) Applications of the natural
Prog Fract Diff Appl 1(3):165–185 trans-form to Maxwell’s equations, progress. In: electromagnet-

123
Iran J Sci Technol Trans Sci

ics research symposium proceedings, Suzhou, China, Sept., Wu GC, Baleanu D, Zeng SD, Deng ZG (2015) Discrete
12(16), 11 pages fractional diffusion equation. Nonlinear Dyn 80(1—-2):
Sun H, Chen W, Sze K (2013) A semi-discrete finite element method 281–286
for a class of time-fractional diffusion equations. Philos Trans R
Soc A 371:1471–1485

123

Вам также может понравиться