Applied Mathematics and Computation 257 (2015) 119–133

Contents lists available at ScienceDirect

Applied Mathematics and Computation

journal homepage: www.elsevier.com/locate/amc

A novel expansion iterative method for solving linear partial

differential equations of fractional order
Ahmad El-Ajou a, Omar Abu Arqub a, Shaher Momani b,c, Dumitru Baleanu d,e,⇑, Ahmed Alsaedi c
a
Department of Mathematics, Faculty of Science, Al Balqa Applied University, Salt 19117, Jordan
b
Department of Mathematics, Faculty of Science, The University of Jordan, Amman 11942, Jordan
c
Department of Mathematics, Faculty of Science, King Abdulaziz University, Jeddah 21589, Saudi Arabia
d
Department of Mathematics and Computer Sciences, Faculty of Arts and Sciences, Cankaya University, Turkey
e
Institute of Space Sciences, Magurele-Bucharest, Romania

a r t i c l e i n f o a b s t r a c t

Keywords: In this manuscript, we implement a relatively new analytic iterative technique for solving
Fractional partial differential equations time–space-fractional linear partial differential equations subject to given constraints con-
Fractional power series ditions based on the generalized Taylor series formula. The solution methodology is based
Residual power series on generating the multiple fractional power series expansion solution in the form of a rap-
idly convergent series with minimum size of calculations. This method can be used as an
alternative to obtain analytic solutions of different types of fractional linear partial differ-
ential equations applied in mathematics, physics, and engineering. Some numerical test
applications were analyzed to illustrate the procedure and to confirm the performance
of the proposed method in order to show its potentiality, generality, and accuracy for solv-
ing such equations with different constraints conditions. Numerical results coupled with
graphical representations explicitly reveal the complete reliability and efficiency of the
suggested algorithm.
Ó 2015 Elsevier Inc. All rights reserved.

1. Introduction

Fractional partial differential equations (PDEs) are found to be an effective tool to describe certain physical phenomena
such as, damping laws, rheology, diffusion processes, electrostatics, electrodynamics, fluid flow, elasticity, and so on [1–7].
Problems in fractional PDEs are not only important but also quite challenging which usually involves hard mathematical
solution techniques. Anyhow, in most real-life applications, it is too complicated to obtain exact solutions to PDEs of frac-
tional order in terms of composite elementary functions in a simple manner, so an efficient, reliable numerical algorithm
for the solutions of such equations is required; it is little wonder that with the development of fast, efficient digital comput-
ers, the role of numerical methods in mathematics, physics, and engineering problem solving has increased dramatically in
recent years. The objective of the present letter is to extend the application of the iterative residual power series (RPS)
method [8–14] to provide analytic solutions for initial value problems of linear PDEs of fractional order and to make com-
parisons with some of the well-known analytic methods.
The fractional calculus is a name for the theory of integrals and derivatives of arbitrary order, which unifies and gener-
alizes the notions of integer-order differentiation and n-fold integration. Theory of fractional calculus is a significantly

⇑ Corresponding author at: Department of Mathematics and Computer Sciences, Faculty of Arts and Sciences, Cankaya University, Turkey.
E-mail address: dumitru@cankaya.edu.tr (D. Baleanu).

http://dx.doi.org/10.1016/j.amc.2014.12.121
0096-3003/Ó 2015 Elsevier Inc. All rights reserved.
120 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133

important and useful branch of mathematics having a broad range of applications at almost any branch of science. Tech-
niques of fractional calculus have been employed at the modeling of many different phenomena in mathematics, physics,
and engineering [15–21]. The most important advantage of using fractional calculus in these and other applications is their
nonlocal property. It is well known that the integer order differential operators and the integer order integral operators are
local, while on the other hand, the fractional order differential operators and the fractional order integral operators are non-
local. This means that the next state of a system depends not only upon its current state but also upon all of its historical
states. In fact, this is main reason why differential operators of fractional order provide an excellent instrument for descrip-
tion of memory and hereditary properties of various mathematical, physical, and engineering processes.
Series expansions are very important aids in numerical calculations, especially for quick estimates made in hand calcu-
lation, for example, in evaluating functions, integrals, or derivatives. Solutions to fractional PDEs can often be expressed in
terms of series expansions. Since, the advent of computers, it has, however, become more common to treat fractional PDEs
directly, using different approximation method instead of series expansions. But in connection with the development of
automatic methods for formula manipulation, one can anticipate renewed interest for series methods. These methods have
some advantages, especially in multidimensional solutions for PDEs of fractional order.
The RPS method was developed as an efficient method for determine values of coefficients of the power series solution for
the first-order and the second-order fuzzy DEs [8]. It has been successfully applied in the numerical solution of the general-
ized Lane-Emden equation, which is a highly nonlinear singular DE [9], in the numerical solution of higher-order regular DEs
[10], in the solution of composite and noncomposite fractional DEs [11], in predict and represent the multiplicity of solutions
to boundary value problems of fractional order [12], in construct and predict the solitary pattern solutions for nonlinear
time-fractional dispersive PDEs [13], and in approximate solution of the nonlinear fractional KdV-Burgers equation [14].
The RPS method is effective and easy to construct power series solution for strongly linear and nonlinear equations without
linearization, perturbation, or discretization. Different from the classical power series method, the RPS method does not need
to compare the coefficients of the corresponding terms and a recursion relation is not required. This method computes the
coefficients of the power series by a chain of algebraic equations of one or more variables. In fact, the RPS method is an alter-
native procedure for obtaining analytic solutions for PDEs of fractional order. By using residual error concept, we get a series
solutions; in practice truncated series solutions. Moreover, the obtained solutions and all their time–space-fractional deriv-
atives are applicable for each arbitrary point and each multidimensional variable in the given domain. On the other aspect as
well, the RPS method does not require any converting while switching from the low-order to the higher-order; as a result the
method can be applied directly to the given problem by choosing an appropriate initial guess approximation.
In the present paper, the RPS method will investigate to construct a new algorithm for finding analytical solutions to the
following classes of higher-order linear fractional PDEs of the general form
Dat uðx; tÞ ¼ L½uðx; tÞ; x 2 R; t P t0 ; 0 6 m  1 < a 6 m; m 2 N; ð1:1Þ
subject to the nonhomogeneous initial conditions
@ j uðx; t 0 Þ
¼ uj ðxÞ; x 2 R; j ¼ 0; 1; 2; . . . ; m  1; ð1:2Þ
@t j
and subject to the constraint linear differential operator
" !#
X
m1
@ j @ kj uðx; tÞ X
1 X
m1
hij ðxÞ
L½uðx; tÞ ¼ j
g j ðxÞt þ ðt  t 0 Þjþia ; kj 2 N; ð1:3Þ
j¼0 @t j @xkj i¼0 j¼0
Cðia þ j þ 1Þ

where Dat is the Caputo fractional derivative of order a and uj ðxÞ; g j ðxÞ; hij ðxÞ are given analytic functions on R. Throughout
this paper, N the set of integer numbers, R the set of real numbers, and C is the Gamma function.
In most cases, the higher-order linear fractional PDEs cannot be solved analytically and their solutions cannot be
expressed in closed forms, where solutions of such equations are always demand due to physical interests. Anyhow, in
the literature, a number of methods have been developed for numerical or analytical solutions for fractional PDEs. The reader
is asked to refer to [22–29] in order to know more details about these methods, including their kinds and history, their mod-
ifications and conditions for use, their scientific applications, their importance and characteristics, and their relationship
including the difference.
The outline of the letter is as follows. In the next section, we utilize some necessary definitions and results from fractional
calculus theory. In Section 3, basic idea of the RPS method is represented in order to construct and predict the multiple frac-
tional power series expansion solution. In Section 4, some physical and mathematical linear fractional PDEs of different types
and orders are performed in order to illustrate capability and simplicity of the proposed method. The conclusion is given in
the final part, Section 5.

2. Overview of fractional calculus theory: mathematical tools and theories

For the concept of fractional derivative we will adopt the Caputo’s definition which is a modification of the Riemann–
Liouville definition and has the advantage of dealing properly with initial value problems in which the initial conditions are
given in terms of the field variables and their integer order which is the case in most physical and engineering processes [21].
 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133 121

Next, some necessary definitions and essentials results from fractional calculus theory are presented, where these defi-
nitions and results are collected from the mentioned references.

Definition 2.1. A real function uðx; tÞ; x 2 R; t > 0 is said to be in the space C l ; l 2 R if there exists a real number p > l such
@ n uðx;tÞ
that uðx; tÞ ¼ tp u1 ðx; tÞ, where u1 ðx; tÞ 2 CðR  ½0; 1ÞÞ, and it is said to be in the space C nl if @t n 2 C l ; n 2 N.

Definition 2.2. The Riemann–Liouville fractional integral operator of order a > 0 of uðx; tÞ 2 C l , l P 1 is defined as
( Rt a1
1
ðt  sÞ uðx; sÞds; a > 0; x 2 I; t > s > s P 0;
J at uðx; tÞ ¼ CðaÞ s
ð2:1Þ
uðx; tÞ; a ¼ 0:

Definition 2.3. The Caputo fractional derivative of order a > 0 of uðx; tÞ 2 C n1 ; n 2 N is defined as
8  
na @ n uðx;tÞ
a @ a uðx; tÞ < J t @tn
; 0 6 n  1 < a < n; x 2 I; t > s > s P 0;
Dt uðx; tÞ ¼ a ¼ ð2:2Þ
@t : @ uðx;tÞ
n

@t n
; a ¼ n 2 N:
On the one hand, for some certain properties of the operator Dat , it is obvious that when c > 1; t > s P 0, and C 2 R, we
have Dat ðt  sÞc ¼ CCðcðþ1
cþ1Þ
aÞ ðt  sÞ
ca
and Dat C ¼ 0. On the other hand, properties of the operator J as can be summarized shortly in
the form of the following: for uðx; tÞ 2 C l ; l P 1, a; b P 0, C 2 R, and c P 1, we have J at C ¼ CðaCþ1Þ ðt  sÞa , J at J bt uðx; tÞ
¼ J at þb uðx; tÞ ¼ J bt J at uðx; tÞ, and J at ðt  sÞc ¼ CCðaðþcþ1Þ
cþ1Þ ðt  sÞ
aþc
.

Theorem 2.1. If n  1 < a 6 n, uðx; tÞ 2 C nl , n 2 N, and l P 1, then Dat Jat uðx; tÞ ¼ uðx; tÞ and Jat Dat uðx; tÞ ¼ uðx; tÞ
P @ j uðx;sþ Þ ðtsÞ j
 n1
j¼0 @t j j!
, where t > s P 0.
The reader may go through [1–7,15–21] in order to know more details about the mathematical properties of fractional
derivatives and fractional integrals, including their types and history, their motivation for use, their characteristics, and their
applications. Next, some results related to the fractional power series (FPS) in the sense of the Caputo’s definition for frac-
tional derivative are collected in order to find analytic series solution for Eqs. (1.1) and (1.2).

Definition 2.4. A power series expansion of the form

X
1 X
m1
f nk ðxÞðt  t 0 Þkþna ; 0 6 m  1 < a 6 m; t P t 0 ; ð2:3Þ
n¼0 k¼0

is called a multiple FPS about t ¼ t0 , where t is a variable and f nk ðxÞ are functions of x called the coefficients of the series.
Notice that in writing out the term corresponding to n ¼ 0 and k ¼ 0 in Eq. (2.3) we have adopted the convention that
W
ðt  t 0 Þ0 ¼ 1 even when t ¼ t 0 . Also, when t ¼ t 0 each of terms of Eq. (2.3) are vanished for n–0 k–0 and so. On the other
hand, the multiple FPS representation of Eq. (2.3) always converges when t ¼ t0 . In the next lemma by Dnt0a we mean that
Dat0  Dat0 . . . Dat0 (n -times).

Lemma 2.1. Suppose that uðx; tÞ 2 CðR  ½t0 ; t0 þ RÞÞ and Djta uðx; tÞ 2 CðR  ðt 0 ; t 0 þ RÞÞ for j ¼ 0; 1; 2; . . . ; n þ 1 where
0 6 m  1 < a 6 m. Then
a
Dðnþ1Þ uðx; fÞ
J ðnþ1Þ
t
a ðnþ1Þa
Dt uðx; tÞ ¼ t
ðt  t0 Þðnþ1Þa ; t0 6 f 6 t < t 0 þ R: ð2:4Þ
Cððn þ 1Þa þ 1Þ

Proof. From the definition of the operator J at and by using the second mean value theorem for fractional integrals, one can
find

Z a Z
1 t
Dðnþ1Þ uðx; fÞ t
J ðnþ1Þ
t
a ðnþ1Þa
Dt uðx; tÞ ¼ ðt  sÞðnþ1Þa1 Dðnþ1Þ
t
a
uðx; sÞds ¼ t
ðt  sÞðnþ1Þa1 ds
Cððn þ 1ÞaÞ t0 Cððn þ 1ÞaÞ t0
a
Dðnþ1Þ uðx; fÞ
¼ t
ðt  t0 Þðnþ1Þa : ð2:5Þ
Cððn þ 1Þa þ 1Þ
h

Theorem 2.2. Suppose that uðx; tÞ 2 CðR  ½t 0 ; t 0 þ RÞÞ, Djta uðx; tÞ 2 CðR  ðt0 ; t0 þ RÞÞ, and Djta uðx; tÞ can be differentiated
ðm  1Þ-times with respect to t on ðt 0 ; t 0 þ RÞ for j ¼ 0; 1; 2; . . . ; n þ 1, where 0 6 m  1 < a 6 m. Then
122 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133

X X Dkþja uðx; t0 Þ
n m1
Dðnþ1Þ a
uðx; fÞ
uðx; tÞ ¼ t
ðt  t 0 Þkþja þ t
ðt  t 0 Þðnþ1Þa ; t 0 6 f 6 t 6 t0 þ R: ð2:6Þ
j¼0 k¼0
Cðj a þ k þ 1Þ C ððn þ 1Þ a þ 1Þ

Proof. From the certain properties of the operator J at , we have

a ðnþ1Þa
J ðnþ1Þ
t Dt uðx; tÞ ¼ J nt a ððJ at Dat ÞDnt a uðx; tÞÞ ¼ J nt a ððJ m m na
t Dt ÞDt uðx; tÞÞ
!
na na
X @ k Dnt a uðx;tþ Þ
m 1 k
k
0
¼ J t Dt uðx; tÞ  @t
k!
ðt  t 0 Þ
k¼0
!
X
m1
Dk Dna uðx;t
¼ J nt a Dnt a uðx; tÞ  J nt a t t
k!
0Þ
ðt  t0 Þ k

k¼0 ð2:7Þ
  X
m1 kþna
D uðx;t
¼ J ðn1Þ
t
a
ðJ m m ðn1Þa
t Dt ÞDt uðx; tÞ  t
Cðnaþkþ1Þ
0Þ
ðt  t 0 Þkþna
k¼0
!
X
m1 k ðn1Þa
D D uðx;t
X
m1 kþna
D uðx;t
0Þ
¼ J ðn1Þ
t
a
Dðn1Þ
t
a
uðx; tÞ  t
k!
ðt  t0 Þk  t
Cðnaþkþ1Þ
0Þ
ðt  t0 Þkþna :
k¼0 k¼0

On the other direction, if we keep the repeating of this process, then after n -times of scientific computations, we will
obtain

Xn mX1
Dkþj a
uðx; t 0 Þ
J ðnþ1Þ
t
a ðnþ1Þa
Dt uðx; tÞ ¼ uðx; tÞ  t
ðt  t 0 Þkþja ; t 0 6 t 6 t0 þ R: ð2:8Þ
j¼0 k¼0
C ðj a þ k þ 1Þ

By using Lemma 2.1, the proof of the theorem will be complete. h

Remark 2.1. If one fixed m ¼ 1, then the series representation formula of Eq. (2.6) will leads to the following expansion of
uðx; tÞ about t 0 :
X
n
Djta uðx; t0 Þ Dðnþ1Þa
uðx; fÞ
ðt  t 0 Þja þ ðt  t0 Þðnþ1Þa ;
t0
uðx; tÞ ¼ 0
t0 6 f 6 t < t 0 þ R; ð2:9Þ
j¼0
Cðja þ 1Þ Cððn þ 1Þa þ 1Þ

which is the same as of Generalized Taylor’s series formula that obtained in [30] for 0 < a 6 1.
As with any convergent series, this means that uðx; tÞ is the limit of the sequence of partial sums. In the case of general
P P kþja
Dt uðx;t 0 Þ kþja
form of generalized Taylor’s series, the partial sums are given as T n ðx; tÞ ¼ nj¼0 m1
k¼0 Cðjaþkþ1Þ ðt  t 0 Þ . In general, uðx; tÞ is
the sum of its general form of generalized Taylor’s series if uðx; tÞ ¼ lim T n ðx; tÞ. But on the other aspect as well, if
n!1
Rn ðx; tÞ ¼ uðx; tÞ  T n ðx; tÞ, then Rn ðx; tÞ is the remainder of general form of generalized Taylor’s series. That is,
ðnþ1Þa
D uðx;fÞ ðnþ1Þa
Rn ðx; tÞ ¼ Cððnþ1Þ
t
aþ1Þ ðt  t 0 Þ ; t 0 6 f 6 t < t0 þ R.

ðnþ1Þa
Corollary 2.1. If jDt0 uðx; tÞj 6 MðxÞ on t 0 6 t 6 d, where m < a 6 m  1, then the reminder Rn ðx; tÞ of general form of
generalized Taylor’s series satisfies
MðxÞ
jRn ðx; tÞj 6 ðt  t 0 Þðnþ1Þa ; t 0 6 t 6 d: ð2:10Þ
Cððn þ 1Þa þ 1Þ

3. Construction of residual power series method

In this section, we find out series solutions for linear fractional PDEs by substituting their FPS expansions among their
truncated residual functions. From the resulting equations a recursion formulas for the computation of the coefficients
are derived, while the coefficients in the multiple FPS expansions can be computed recursively by recurrent fractional dif-
ferentiating of the truncated residual functions.
The RPS method consists in expressing the solution of Eqs. (1.1) and (1.2) as a multiple FPS expansion about the initial
point t ¼ t0 . To achieve our goal, suppose that the solution takes the following infinite expansion form
X1 mX1
ðt  t0 Þiaþj
uðx; tÞ ¼ f ij ðxÞ ; m  1 < a 6 m; x 2 R; t 0 6 t < t0 þ R: ð3:1Þ
i¼0 j¼0
Cðia þ j þ 1Þ
 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133 123

@ j uðx;t 0 Þ
Obviously, uðx; tÞ satisfy the initial conditions of Eq. (1.2); so from Eq. (3.1), we obtain, @t j
¼ j!f 0j ðxÞ ¼ uj ðxÞ;
uj ðxÞ
j ¼ 0; 1; 2; . . . ; m  1. Then f 0j ðxÞ ¼ j!
; j ¼ 0; 1; 2; . . . ; m  1 and the initial guess approximation of uðx; tÞ can be written as
X
m 1
uj ðxÞ
uð0;m1Þ ðx; tÞ ¼ ðt  t0 Þ j ; x 2 R; t 0 6 t < t0 þ R: ð3:2Þ
j¼0
j!
As a result, using Eq. (3.2) we can reformulate the expansion form of Eq. (3.1) as follows:
X
m1
uj ðxÞ X
1 X
m1
ðt  t0 Þiaþj
uðx; tÞ ¼ ðt  t0 Þ j þ f ij ðxÞ ; x 2 R; t 0 6 t < t0 þ R: ð3:3Þ
j¼0
j! i¼1 j¼0
Cðia þ j þ 1Þ
The RPS provides an analytical approximate solution in terms of an infinite multiple FPS. However, to obtain numerical
values from this series, the consequent series truncation and the practical procedure are conducted to accomplish this task.
In the following step, we will let uðk;lÞ ðx; tÞ to denote the (k, l)-truncated series of u(x, t). That is
X
m 1
uj ðxÞ Xk X l
ðt  t0 Þiaþj
uðk;lÞ ðx; tÞ ¼ ðt  t0 Þ j þ f ij ðxÞ ; x 2 R; t P t 0 ; ð3:4Þ
j¼0
j! i¼1 j¼0
Cðia þ j þ 1Þ
where the indices counters k and l whenever used mean that k ¼ 1; 2; 3; . . . and l ¼ 0; 1; 2; . . . ; m  1.
Prior to applying the RPS technique for finding form of the coefficients f ij ðxÞ in the series expansion of Eq. (3.4), we must
define the residual function concept for Eq. (1.1) as
Resðx; tÞ ¼ Dat uðx; tÞ  L½uðx; tÞ; x 2 R; t P t0 ; ð3:5Þ
and the following (k, l)-truncated residual function:
Resðk;lÞ ðx; tÞ ¼ Dat uðk;lÞ ðx; tÞ  L½uðk;lÞ ðx; tÞ; x 2 R; t P t 0 : ð3:6Þ
As in [8–14], it is clear that Resðx; tÞ ¼ 0 for each t 2 ½t0 ; t0 þ RÞ; x 2 R, where R is a nonnegative real number. In fact, this
a j
shows that Dði1Þ
t Dt Resðx; tÞ ¼ 0 for each i ¼ 1; 2; 3; . . . ; k and j ¼ 0; 1; 2; . . . ; l, since the fractional derivative of a constant
a j
function in the Caputo sense is zero. In the mean time, the fractional derivatives Dði1Þ t Dt for each i ¼ 1; 2; 3; . . . ; k and
j ¼ 0; 1; 2; . . . ; l of Resðx; tÞ and Resði;jÞ ðx; tÞ are matching at ðx; tÞ ¼ ðx; t0 Þ. Anyhow, one can write
a j a j
Dði1Þ
t Dt Resðx; t 0 Þ ¼ Dði1Þ
t Dt Resði;jÞ ðx; t 0 Þ ¼ 0; x 2 R; i ¼ 1; 2; 3; . . . ; k; j ¼ 0; 1; 2; . . . ; l: ð3:7Þ
To obtain form of the coefficients f v w in Eq. (3.4) for v ¼ 1; 2; 3; . . . ; k and w ¼ 0; 1; 2; . . . ; l, we apply the following subrou-
tine: substitute (v, w)-truncated series of u(x, t) into Eq. (3.6), find the fractional derivative formula Dðtv 1Þa Dwt of Resðv ;wÞ ðx; tÞ
at t ¼ t0 , and then finally solve the obtained algebraic equation to get the required coefficients.
To summarize the computation process of RPS method in numerical values, we apply the following steps: fixed i = 1 and
run the counter j ¼ 0; 1; 2; . . . ; l to find (1, j)-truncated series expansion of suggested solution, next fixed i = 2 and run the
counter j ¼ 0; 1; 2; . . . ; l to obtain the (2, j)-truncated series, and so on. In fact, to obtain (1, 0)-truncated series expansion
of Eqs. (1.1) and (1.2), we use Eq. (3.4) and write

ðt  t 0 Þðm1Þ ðt  t 0 Þa
uð1;0Þ ðx; tÞ ¼ u0 ðxÞ þ u1 ðxÞðt  t 0 Þ þ    þ um1 ðxÞ þ f 10 ðxÞ : ð3:8Þ
ðm  1Þ! Cða þ 1Þ
On the other aspect as well, to determine form of the first unknown coefficient f 10 ðxÞ in Eq. (3.8), we should substitute Eq.
(3.8) into both sides of the (1, 0)-residual function that obtained from Eq. (3.6), to get the following result:
" #
ðt  t 0 Þðm1Þ ðt  t 0 Þa
Resð1;0Þ ðx; tÞ ¼ f 10 ðxÞ  L u0 ðxÞ þ u1 ðxÞðt  t 0 Þ þ    þ um1 ðxÞ þ f 10 ðxÞ : ð3:9Þ
ðm  1Þ! Cða þ 1Þ
ðk Þ
Depending on the result of Eq. (3.7) for ði; jÞ ¼ ð1; 0Þ; then Eq. (3.9) gives f 10 ðxÞ ¼ L½u0 ðxÞ ¼ g 0 ðxÞu0 0 þ h00 ðxÞ, where
ðk Þ
u0 0 ðxÞ denotes to the k0 th-derivative of u0 ðxÞ. Hence, using the (1, 0)-truncated series expansion of Eq. (3.8); the (1, 0)-
RPS approximate solution for Eqs. (1.1) and (1.2) can be expressed as
ðt  t0 Þðm1Þ ðt  t0 Þa
uð1;0Þ ðx; tÞ ¼ u0 ðxÞ þ u1 ðxÞðt  t 0 Þ þ    þ um1 ðxÞum1 ðxÞ þ L½u0 ðxÞ : ð3:10Þ
ðm  1Þ! Cða þ 1Þ
Similarly, to find out the (1, 1)-truncated series expansion for Eqs. (1.1) and (1.2), we use Eq. (3.4) and write

ðt  t 0 Þðm1Þ ðt  t 0 Þa ðt  t0 Þaþ1
uð1;1Þ ðx; tÞ ¼ u0 ðxÞ þ u1 ðxÞðt  t 0 Þ þ    þ um1 ðxÞ þ L½u0 ðxÞ þ f 11 ðxÞ : ð3:11Þ
ðm  1Þ! Cða þ 1Þ Cða þ 2Þ
Again, to find out form of the second unknown coefficient f 11 ðxÞ in Eq. (3.11), we must find and formulate (1, 1)-residual
function based on Eq. (3.6) and then substitute the form of uð1;1Þ ðx; tÞ of Eq. (3.11) to find new discretized form of this residual
function as follows:
124 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133

Resð1;1Þ ðx; tÞ ¼ f 10 ðxÞ þ f 11 ðxÞðt  t 0 Þ

" #
ðt  t 0 Þðm1Þ ðt  t 0 Þa ðt  t0 Þaþ1
 L u0 ðxÞ þ u1 ðxÞðt  t 0 Þ þ    þ um1 ðxÞ þ L½u0 ðxÞ þ f 11 ðxÞ : ð3:12Þ
ðm  1Þ! Cða þ 1Þ Cða þ 2Þ

By considering Eq. (3.7) for (i, j) = (1, 1) and applying the operator Dt to the both side of Eq. (3.12), we get
Dt Resð1;1Þ ðx; t 0 Þ ¼ f 11 ðxÞ
" #
ðt  t0 Þðm1Þ ðt  t0 Þa ðt  t0 Þaþ1
 Dt L u0 ðxÞ þ u1 ðxÞðt  t 0 Þ þ    þ um1 ðxÞ þ L½u0 ðxÞ þ f 11 ðxÞ :
ðm  1Þ! Cða þ 1Þ Cða þ 2Þ
t¼t0

ð3:13Þ
a j
Using the fact that Dði1Þ
t Dt Resði;jÞ ðx; t0 Þ ¼ 0 for (i, j) = (1, 1}) from Eq. (3.7) and solving the resultant algebraic equation for
f 11 ðxÞ, we can easily obtain
ðk Þ ðk Þ
f 11 ðxÞ ¼ g 1 ðxÞu0 0 ðxÞ þ g 0 ðxÞu1 1 ðxÞ þ h01 ðxÞ: ð3:14Þ
Hence, using the (1, 1)-truncated series expansion of Eq. (3.1); the (1, 1)-RPS approximate solution for Eqs. (1.1) and (1.2)
can be expressed as
X
m 1
uj ðxÞ   ðt  t Þa   ðt  t Þ1þa
0 0
ðt  t0 Þ j þ g 0 ðxÞu0 0 þ h00 ðxÞ
ðk Þ ðk Þ ðk Þ
u1;1 ðx; tÞ ¼ þ g 1 ðxÞu0 0 ðxÞ þ g 0 ðxÞu1 1 ðxÞ þ h01 ðxÞ :
j¼0
j! Cða þ 1Þ Cða þ 2Þ
ð3:15Þ
This procedure can be repeated till the arbitrary order coefficients of the multiple FPS solution of Eqs. (1.1) and (1.2) are
obtained. Moreover, higher accuracy can be achieved by evaluating more components of the solution. As we will see later, if
there is a pattern in the series coefficients, then calculating few of terms in series is sufficient to reach the solution.

4. Applications and numerical discussions

The application problems are carried out using the proposed RPS method, which is one of the modern analytical tech-
niques because of it’s iteratively nature; it can be handle any kind of initial conditions and other constraints. In this section,
we consider six applications to show potentiality, generality, and superiority of our method to solve linear fractional PDEs as
compared with other well-known methods. In the first application we consider the time-fractional wave equation, in the
second two applications we consider two versions of the space-fractional telegraph equations, in the fourth and five appli-
cations we consider the time-fractional Poisson and Navier–Stokes equation, respectively, while in the remaining application
we construct a time-fractional equation. In the process of computation, all the symbolic and numerical computations were
performed by using MATHEMATICA 7 software package.
Throughout this work, we will try to give the results of the all applications; however, in some cases we will switch
between the results obtained for the applications in order not to increase the length of the paper without the loss of gener-
ality for the remaining applications and results. However, by easy calculations we can collect further results and discussion
for the desire application.

Application 4.1. Consider the following homogeneous time-fractional wave equation:

@ a uðx; tÞ 1 2 @ 2 uðx; tÞ
¼ x ; x 2 R; t P 0; 1 < a 6 2; ð4:1Þ
@ta 2 @x2
subject to the nonhomogeneous initial conditions

uðx; 0Þ ¼ x; ut ðx; 0Þ ¼ x2 : ð4:2Þ

2
uðx;tÞ
According to Eq. (3.3) with u0 ðxÞ ¼ x and u1 ðxÞ ¼ x2 , and considering the linear differential operator L½uðx; tÞ ¼ 12 x2 @ @x2
,
we assume that the series solution of Eqs. (4.1) and (4.2) can be written as
X
1 X
1
tjþia
uðx; tÞ ¼ x þ x2 t þ f ij ðxÞ ; ð4:3Þ
i¼1 j¼0
Cð1 þ j þ iaÞ
where uð0;1Þ ðx; tÞ ¼ x þ x2 t is the initial guess approximation which is obtained directly from Eq. (3.2). Next, according to Eqs.
(3.4) and (3.6) the (k, l)-truncated series of u(x, t) and the (k, l)-truncated residual function of Eq. (4.1) can be defined and
thus constructed, respectively, as follows:
X
k X
l
tjþia
uðk;lÞ ðx; tÞ ¼ x þ x2 t þ f ij ðxÞ ; k ¼ 1; 2; 3; . . . ; l ¼ 0; 1; ð4:4Þ
i¼1 j¼0
Cð1 þ j þ iaÞ
 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133 125

1 @ 2 uðk;lÞ ðx; tÞ
Resðk;lÞ ðx; tÞ ¼ Dat uðk;lÞ ðx; tÞ  x2 ; k ¼ 1; 2; 3; . . . ; l ¼ 0; 1: ð4:5Þ
2 @x2
To determine form of the coefficient f 10 ðxÞ, we find out the (1, 0)-truncated series of uðx; tÞ as
ta 00 ta
uð1;0Þ ðx; tÞ ¼ x þ x2 t þ f 10 ðxÞ Cð1þ 1 2
aÞ and the ð1; 0Þ -truncated residual function as Resð1;0Þ ðx; tÞ ¼ f 10 ðxÞ  2 x ð2t þ f 10 ðxÞ Cð1þaÞÞ.
On the other aspect as well, by considering Eq. (3.7) for ði; jÞ ¼ ð1; 0Þ and substituting t ¼ 0, one can obtain f 10 ðxÞ ¼ 0.
Similarly, to find out form of the coefficient f 11 ðxÞ, we evaluate the ð1; 1Þ -truncated series of uðx; tÞ as
1þa 00 1þa
uð1;1Þ ðx; tÞ ¼ x þ x2 t þ f 11 ðxÞ Cð2þ
t 1 2 t
aÞ and the ð1; 1Þ-truncated residual function as Resð1;1Þ ðx; tÞ ¼ f 11 ðxÞt  2 x ð2t þ f 11 ðxÞ Cð2þaÞÞ.
00 ta
Thus, for ði; jÞ ¼ ð1; 1Þ, we conclude that Dt Resð1;1Þ ðx; tÞ ¼ f 11 ðxÞ  12 x2 ð2 þ f 11 ðxÞ Cð1þaÞÞ. Anyhow, the substitution of t = 0 will
2
leads to f 11 ðxÞ ¼ x .
1þa 2a
To evaluate form of the coefficient f 20 ðxÞ, we need to write uð2;0Þ ðx; tÞ ¼ x þ x2 t þ x2 Cð2þ
t t
aÞ þ f 20 ðxÞ Cð1þ2aÞ and
ta t 1þa 00 t 2a
Resð2;0Þ ðx; yÞ ¼ f 20 ðxÞ Cð1þaÞ  12 x2 ð2 Cð2þaÞ þ f 20 ðxÞ Cð1þ2aÞÞ. Considering the fact that Dat Resð2;0Þ ðx; 0Þ ¼ 0, we can easily find
f 20 ðxÞ ¼ 0. Anyhow, the continuation in the same manner will leads also to f 21 ðxÞ ¼ x2 . Moreover, for k ¼ 3; 4; 5; . . . ; we
can conclude that f k0 ðxÞ ¼ 0 and f k1 ðxÞ ¼ x2 . Therefore, according to Eq. (4.3), the RPS solution of Eqs. (4.1) and (4.2) can
be written in the form of infinite series as follows:
 
t1þa t 1þ2a t 1þ3a
uðx; tÞ ¼ x þ x2 t þ þ þ þ  : ð4:6Þ
Cð2 þ aÞ Cð2 þ 2aÞ Cð2 þ 3aÞ
As a special case when a ¼ 2, the RPS solution of Eqs. (4.1) and (4.2) has the general pattern form which is coinciding with
the following exact solution in terms of infinite series:

t3 t5 t7
uðx; tÞ ¼ x þ x2 ðt þ þ þ þ   Þ: ð4:7Þ
3! 5! 7!
So, the exact solution of Eqs. (4.1) and (4.2) in a closed form of elementary function will be uðx; tÞ ¼ x þ x2 sinh t.

Remark 4.1. It is to be noted that, the (3, 1)-truncated series solution of Eqs. (4.1) and (4.2) is

t1þa t 1þ2a t 1þ3a

uð3;1Þ ðx; tÞ ¼ x þ x2 ðt þ þ þ Þ; ð4:8Þ
Cð2 þ aÞ Cð2 þ 2aÞ Cð2 þ 3aÞ
which is the same as the fourth term approximate solution that obtained by the Adomian decomposition method (ADM)
[23]. Whereas, the fourth term approximate solution for Eqs. (4.1) and (4.2) that obtained by the variational iteration method
(VIM) [31] and the homotopy perturbation method (HPM) [32] are the same and are given as
 
t3 t5 t7 3t5a 2t 7a t72a
u4 ðx; tÞ ¼ x þ x2 t þ þ þ   þ : ð4:9Þ
2 40 7! Cð6  aÞ Cð8  aÞ Cð8  2aÞ
In fact, the expansion form of Eq. (4.8) shows that it is the most natural solution, and the most fruitful one to Eqs. (4.1) and
(4.2) in comparison with the expansion form of Eq. (4.9). Whilst the solution of Eq. (4.8) when a ¼ 2 coincides with the clas-
sical solution of Eq. (4.7) in ordinary case in contrast to the solution of Eq. (4.9).
Results from numerical analysis are an approximation, in general, which can be made as accurate as desired. Because a
computer has a finite word length, only a fixed number of digits are stored and used during computations. Next, the agree-
ment between the solutions obtained using the RPS method, the ADM, the VIM, and the HPM are investigated for Application
4.1. Anyhow, Tables 1–3 show, respectively, the residual errors when a ¼ 1:5; 1:75; 2 and for various x and t. Those results are
generated from the (3, 1)-truncated series approximation for the RPS method and from the fourth term approximate solution
for the remaining mentioned method.
It is clear from the tables that, the residual errors can be measure the extent of agreement between the (3, 1)-truncated
series approximation of the RPS solutions and unknowns closed form solutions in the case of 1 < a < 2 which are inappli-
cable in general for such fractional equations. Anyhow, from the tables, it can be seen that the RPS technique provides us
with the accurate approximate solution and explicates the rapid convergence in approximating the solution of Eqs. (4.1)
and (4.2) from the other mentioned methods.

Application 4.2. Consider the following homogeneous space-fractional telegraph equation:

@ a uðx; tÞ @ 2 uðx; tÞ @uðx; tÞ

¼ þ þ uðx; tÞ; x 2 R; t P 0; 1 < a 6 2; ð4:10Þ
@xa @t 2 @t
subject to the nonhomogeneous initial conditions
uð0; tÞ ¼ ux ð0; tÞ ¼ et : ð4:11Þ
In this application, the initial conditions are on the independent variable x. So, we will replace x by t and t by x in Eq. (3.1).
Therefore, we assume that the series solution of Eqs. (4.10) and (4.11) can be written as
126 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133

Table 1
The comparison values of the residual errors for Application 4.1 at a ¼ 1:5 .

t x VIM HPM ADM RPS method

0.2 10 5.51277 5.51277 4:97102  10 5
4:97102  105
1 5:51277  102 5:51277  102 4:97102  107 4:97102  107
10 5.51277 5.51277 4:97102  10 5
4:97102  105
0.4 10 5.31407 5.31407 2:24963  10 3
2:24963  103
1 5:31407  10 2
5:31407  102
2:24963  10 5
2:24963  105
10 5.31407 5.31407 2:24963  103 2:24963  103
0.6 10 4.64626 4.64626 2:09224  102 2:09224  102
1 4:64626  102 4:64626  102 2:09224  104 2:09224  104
10 4.64626 4.64626 2:09224  10 2
2:09224  102
0.8 10 4.91059 4.91059 0:101806 0:101806
1 4:91059  102 4:91059  102 1:01806  103 1:01806  103

Table 2
The comparison values of the residual errors for Application 4.1 at a ¼ 1:75.

t x VIM HAM ADM RPS method

0.2 10 1.15719 1.15719 3:70435  106 3:70435  106
1 1:15719  102 1:15719  102 3:70435  108 3:70435  108
10 1.15719 1.15719 3:70435  106 3:70435  106
0.4 10 0.92375 0.92375 2:81935  10 4
2:81935  104
1 9:23749  103 9:23749  103 2:81935  106 2:81935  106
10 0.92375 0.92375 2:81935  104 2:81935  104
0.6 10 1.06133 1.06133 3:55402  103 3:55402  103
1 1:06133  102 1:06133  102 3:55402  105 3:55402  105
10 1.06133 1:06133 3:55402  103 3:55402  103
0.8 10 1.73713 1.73713 2:14579  10 2
2:14579  102
1 1:73713  102 1:73713  102 2:14579  104 2:14579  104

Table 3
The comparison values of the residual errors for Application 4.1 at a ¼ 2.

t x VIM HAM ADM RPSmethod

0.2 10 2:53968  107 2:53968  107 2:53968  107 2:53968  107
1 2:53968  10 9
2:53968  10 9
2:53968  10 9
2:53968  109
10 2:53968  107 2:53968  107 2:53968  107 2:53968  107
0.4 10 3:25079  105 3:25079  105 3:25079  105 3:25079  105
1 3:25079  107 3:25079  107 3:25079  107 3:25079  107
10 3:25079  10 5
3:25079  105
3:25079  10 5
3:25079  105
0.6 10 5:55429  10 4
5:55429  10 4
5:55429  10 4
5:55429  104
1 5:55429  10 6
5:55429  10 6
5:55429  10 6
5:55429  106
10 5:55429  104 5:55429  104 5:55429  104 5:55429  104
0.8 10 4:16102  103 4:16102  103 4:16102  103 4:16102  103
1 4:16102  105 4:16102  105 4:16102  105 4:16102  105

X
1 X
1
xjþia
uðx; tÞ ¼ f ij ðtÞ : ð4:12Þ
i¼0 j¼0
Cð1 þ j þ iaÞ

Using RPS method, taking f 00 ðtÞ ¼ uð0; tÞ ¼ et and f 01 ðxÞ ¼ ux ð0; tÞ ¼ et ; the initial guess approximation of uðx; tÞ is of the
form
uð0;1Þ ðx; tÞ ¼ et þ xet : ð4:13Þ
 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133 127

According to Eq. (3.3), the multiple FPS solution of Eqs. (4.10) and (4.11) will take the following infinite expansion:
X
1 X
1
xjþia
uðx; tÞ ¼ et þ et x þ f ij ðtÞ ; ð4:14Þ
i¼1 j¼0
Cð1 þ j þ iaÞ

while the ðk; lÞ-truncated series of uðx; tÞ and the ðk; lÞ-truncated residual function that are derived from Eqs. (3.4) and (3.6)
can be formulated, respectively, in form of the following:
X
k X
l
xjþia
uðk;lÞ ðx; tÞ ¼ et þ et x þ f ij ðtÞ ; ð4:15Þ
i¼1 j¼0
Cð1 þ j þ iaÞ

@ 2 uðk;lÞ ðx; tÞ @uðk;lÞ ðx; tÞ

Resðk;lÞ ðx; tÞ ¼ Dax uðk;lÞ ðx; tÞ    uðk;lÞ ðx; tÞ: ð4:16Þ
@t 2 @t
To determine form of the first unknown coefficient f 10 ðtÞ in the expansion of Eq. (4.15) we should substitute the ð1; 0Þ-
xa a
truncated series uð1;0Þ ðx; tÞ ¼ et þ et x þ f 10 ðtÞ Cð1þ aÞ into the ð1; 0Þ-truncated residual function Resð1;0Þ ðx; tÞ ¼ Dx uð1;0Þ ðx; tÞ
@ 2 uð1;0Þ ðx;tÞ @uð1;0Þ ðx;tÞ
 @t 2
 @t
 uð1;0Þ ðx; tÞ to get
0 00
xa ½f 10 ðtÞ þ f 10 ðtÞ þ f 10 ðtÞ
Resð1;0Þ ðx; tÞ ¼ et  et x þ f 10 ðtÞ  : ð4:17Þ
Cð1 þ aÞ
Depending on the result of Eq. (3.7) in the case of ði; jÞ ¼ ð1; 0Þ, the substituting of x ¼ 0 back into Eq. (4.17) will yields
f 10 ðtÞ ¼ et : Hence, the ð1; 0Þ -truncated series solution of Eqs. (4.10) and (4.11) could be expressed as
xa
uð1;0Þ ðx; tÞ ¼ et þ et x þ et : ð4:18Þ
Cð1 þ aÞ
Again, to find out form of the second unknown coefficient f 11 ðtÞ we substitute the (1, 1)-truncated series solution
Resð1;1Þ ðx; tÞ ¼ Dax uð1;1Þ ðx; tÞ
x x a 1þa
uð1;1Þ ðx; tÞ ¼ et þ et x þ et Cð1þaÞ þ f 11 ðtÞ Cð2þaÞ into the ð1; 1Þ-truncated residual function
@ 2 uð1;1Þ ðx;tÞ @uð1;1Þ ðx;tÞ
 @t 2
 @t
 uð1;1Þ ðx; tÞ to obtain
0 00
et xa x1þa ½f 11 ðtÞ þ f 11 ðtÞ þ f 11 ðtÞ
Resð1;1Þ ðx; tÞ ¼ et x þ xf11 ðtÞ   : ð4:19Þ
Cð1 þ aÞ Cð2 þ aÞ
Now, applying the operator Dx one time on the both sides of Eq. (4.19) will gives the following first partial derivative of
Resð1;1Þ ðx; tÞ with respect to x:

aet xa1 ð1 þ aÞ½f 11 ðtÞ þ f 011 ðtÞ þ f 0011 ðtÞxa

Dx Resð1;1Þ ðx; tÞ ¼ et þ f 11 ðtÞ   : ð4:20Þ
Cð1 þ aÞ Cð2 þ aÞ
According to Eq. (3.7) for ði; jÞ ¼ ð1; 1Þ, we have Dx Resð1;1Þ ð0; tÞ ¼ 0 (recall that in this application, we replace x by t and t by
x). Solving the resulting equation for f 11 ðtÞ, one obtain f 11 ðtÞ ¼ et . Therefore, the ð1; 1Þ -truncated series solution of Eqs.
(4.10) and (4.11) is obtained and one can collect the previous results to get the following expansion:
xa x1þa
uð1;1Þ ðx; tÞ ¼ et þ et x þ et þ et : ð4:21Þ
Cð1 þ aÞ Cð2 þ aÞ
As the former, by applying the same procedure for ði; jÞ, j ¼ 0; 1 and i ¼ 2; 3; 4; . . . will leads after easy calculations to
f ij ðtÞ ¼ et . Anyhow, if we employ the last results, then the RPS solution of Eqs. (4.10) and (4.11) can be constructed in
the form of infinite series as follows:
 
xa x1þa x2a x1þ2a
uðx; tÞ ¼ et 1 þ x þ þ þ þ þ  : ð4:22Þ
Cð1 þ aÞ Cð2 þ aÞ Cð1 þ 2aÞ Cð2 þ 2aÞ
As a special case when a ¼ 2, the RPS solution of Eqs. (4.10) and (4.11) has the general pattern form which is coinciding
with the exact solution in terms of infinite series
 
x2 x3 x4 x5
uðx; tÞ ¼ et 1 þ x þ þ þ þ þ    : ð4:23Þ
2! 3! 4! 5!
So, the exact solution of Eqs. (4.10) and (4.11) in a closed form of elementary function will be uðx; tÞ ¼ ext .
Although there are a lot of studies for the linear time-fractional PDEs and some profound results have been established, it
seems the detailed results and conclusions in the RPS procedure in finding and predicting the solution of Application 4.2 that,
the RPS solution is the same as the Adomian decomposition solution [22] and differential transform solution [33]. Anyhow,
Table 4 shows the full agreement between those solutions and the RPS solution that obtained from the (1, 1)-truncated series
128 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133

Table 4
The comparison values of the approximate solutions for Application 4.2.

x t a ¼ 1:25 a ¼ 1:5 a ¼ 1:75 a¼2

0.2 0 1.3343810819 1.2740803840 1.2402165526 1.2214027581
0.25 1.0392170315 0.9922548007 0.9658816223 0.9512294245
0.5 0.8093430379 0.7727688158 0.7522293638 0.7408182206
0.75 0.6303169917 0.6018329589 0.5858368176 0.5769498103
1 0.4908913667 0.4687079796 0.4562501723 0.4493289641
0.4 0 1.7667126467 1.6328207772 1.5471211015 1.4918246976
0.25 1.3759171927 1.2716420999 1.2048991253 1.1618342427
0.5 1.0715653871 0.9903558632 0.9383763823 0.9048374180
0.75 0.8345359626 0.7712899218 0.7308082613 0.7046880897
1 0.6499372611 0.6006811950 0.5691540462 0.5488116360
0.6 0 2.2986560463 2.0770503324 1.9265225531 1.8221187986
0.25 1.7901951289 1.6176084253 1.5003772729 1.4190675472
0.5 1.3942053682 1.2597947083 1.1684949950 1.1051709170
0.75 1.0858082325 0.9811291054 0.9100248171 0.8607079755
1 0.8456283017 0.7641041155 0.7087280402 0.6703200453
0.8 0 2.9473348307 2.6209313794 2.3914567305 2.2255408965
0.25 2.2953866741 2.0411834106 1.8624683744 1.7332529930
0.5 1.7876489393 1.5896752386 1.4504918284 1.3498587882
0.75 1.3922223937 1.2380403206 1.1296441718 1.0512710813
1 1.0842638904 0.9641867712 0.8797677656 0.8187307413
1 0 3.7347380599 3.2840656609 2.9594178326 2.7182815255
0.25 2.9086169256 2.5576329084 2.3047969255 2.1169997807
0.5 2.2652331393 1.9918865118 1.7949776504 1.6487210869
0.75 1.7641653427 1.5512827752 1.3979299997 1.2840252736
1 1.3739333504 1.2081402401 1.0887089784 0.9999998885

when a ¼ 1:25; 1:5; 1:75; 2 and for various x and t on the domain ½0; 1  ½0; 1. As a result, we can easily conclude that the RPS
method is an efficient tool in finding the solution.

Application 4.3. Consider the following nonhomogeneous space-fractional telegraph equation:

@ a uðx; tÞ @ 2 uðx; tÞ @uðx; tÞ

¼ þ  x2  t þ 1; x 2 R; t P 0; 1 < a 6 2; ð4:24Þ
@xa @t 2 @t
subject to the nonhomogeneous initial conditions
uð0; tÞ ¼ t; ux ð0; tÞ ¼ 0: ð4:25Þ
As the previous application, the initial conditions are on the independent variable x. So, we will replace x by t and t by x in
Eq. (3.1). Moreover, Eq. (4.24) is not exactly as the general form that suggested in Eqs. (1.1) and (1.3) because the term x2 is of
a power 2 which must be of the form j þ ia; j ¼ 0; 1; :::; m  1; i ¼ 0; 1; 2; . . ., that is either of the form ia or 1 þ ia since m ¼ 2
in this application. Therefore, we assume that the solution of Eqs. (4.24) and (4.25) as in the following form:

X
1 X
2
xjþia
uðx; tÞ ¼ f 00 ðtÞ þ f 01 ðtÞx þ f ij ðtÞ : ð4:26Þ
i¼1 j¼0
Cð1 þ j þ iaÞ

Using RPS method and according to the initial conditions of Eq. (4.25), the initial guess approximation of uðx; tÞ is
uð0;1Þ ðx; tÞ ¼ t. Therefore, the RPS solution of Eqs. (4.24) and (4.25) has the following expansion form:
X
1 X
2
xjþia
uðx; tÞ ¼ t þ f ij ðtÞ : ð4:27Þ
i¼1 j¼0
Cð1 þ j þ iaÞ

Whilst the ðk; lÞ -truncated series of uðx; tÞ and the ðk; lÞ-truncated residual function that are derived from Eqs. (3.4) and
(3.6) can be formulated, respectively, in form of
X
k X
l
xjþia
uðk;lÞ ðx; tÞ ¼ t þ f ij ðtÞ ; l ¼ 0; 1; 2; ð4:28Þ
i¼1 j¼0
Cð1 þ j þ iaÞ

@ 2 uðk;lÞ ðx; tÞ @uðk;lÞ ðx; tÞ

Resðk;lÞ ðx; tÞ ¼ Dax uðk;lÞ ðx; tÞ   þ x2 þ t  1: ð4:29Þ
@t 2 @t
 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133 129

To determine form of the first unknown coefficient f 10 ðtÞ in the expansion of Eq. (4.28), we should substitute the (1, 0)-
truncated series of uðx; tÞ into the (1, 0)-truncated residual function to get the following result:
0 00
xa ½f 10 ðtÞ þ f 10 ðtÞ
Resð1;0Þ ðx; tÞ ¼ 2 þ t þ x2 þ f 10 ðtÞ  : ð4:30Þ
Cð1 þ aÞ
Depending on the result of Eq. (3.7), we have f 10 ðtÞ ¼ 2  t. Hence, the ð1; 0Þ -truncated series solution of Eqs. (4.24) and
(4.25) takes the form
xa
uð1;0Þ ðx; tÞ ¼ t þ ð2  tÞ : ð4:31Þ
Cð1 þ aÞ
Similarly, to find out form of the second unknown coefficient f 11 ðtÞ, we substitute the (1, 1)-truncated series solution of
Eqs. (4.24) and (4.25) into the (1, 1)-truncated residual function to obtain
0 00
xa x1þa ½f 11 ðtÞ þ f 11 ðtÞ
Resð1;1Þ ðx; tÞ ¼ x2 þ xf11 ðtÞ þ  : ð4:32Þ
Cð1 þ aÞ Cð2 þ aÞ
Hence, the application of the operator Dx on both sides of Eq. (4.32) will gives the first partial derivative of Resð1;1Þ ðx; tÞ
with respect to x as
0 00
axa1 ð1 þ aÞ½f 11 ðtÞ þ f 11 ðtÞxa
Dx Resð1;1Þ ðx; tÞ ¼ 2x þ f 11 ðtÞ þ  : ð4:33Þ
Cð1 þ aÞ Cð2 þ aÞ
According to Eq. (3.7) in the case of ði; jÞ ¼ ð1; 1Þ, we have Dx Resð1;1Þ ð0; tÞ ¼ 0. Solving the resulting equation for f 11 ðtÞ, one
can obtain f 11 ðtÞ ¼ 0. To determine rule of the coefficient f 12 ðtÞ, we need to solve the algebraic equation
a j
Dði1Þ
x Dx Resði;jÞ ð0; tÞ ¼ 0 when ði; jÞ ¼ ð1; 2Þ. After some calculations, we get f 12 ðtÞ ¼ 2. Similarly, by applying the same pro-
cedure for ði; jÞ ¼ ð2; 0Þ, we have f 20 ðtÞ ¼ 1. It is to be noted that, the rest of the coefficients f ij ðtÞ ¼ 0 for i ¼ 2; 3; . . . and
j ¼ 0; 1; 2. Therefore, the RPS solution of Eqs. (4.24) and (4.25) can be constructed as follows:
xa x2þa x2 a
uðx; tÞ ¼ t þ ð2  tÞ 2  ; ð4:34Þ
Cð1 þ aÞ Cð3 þ aÞ Cð1 þ 2aÞ
which is precisely the exact solution. As a special case when a ¼ 2 ., the RPS solution of Eqs. (4.24) and (4.25) has the general
pattern form which is coinciding with the exact solution in terms of elementary function
x2 x4
uðx; tÞ ¼ t þ ð2  tÞ  : ð4:35Þ
2 8

Remark 4.2. We mention here that, the author in [22] obtained the following Adomian decomposition solution of Eqs. (4.24)
and (4.25):
xa xa x2þa x2a
uðx; tÞ ¼ t þ ð1  tÞ þ ð1 þ tÞ 2 t  ; ð4:36Þ
Cð1 þ aÞ Cð1 þ aÞ Cð3 þ aÞ Cð1 þ 2aÞ
which is not give any visualize about the form of the exact solution. Anyhow, the expansion form of Eq. (4.34) shows that it is
the most natural solution, and the most fruitful one to Eqs. (4.24) and (4.25) in comparison with the expansion form of Eq.
(4.36). Whilst the solution of Eq. (4.34) when a ¼ 2 coincides with the classical solution of Eq. (4.35) in ordinary case in con-
trast to the solution of Eq. (4.36).
The geometric behavior of the solutions of Application 4.3 are studied next in Fig. 1 by drawing the 3 -dimensional space
figures of the ð2; 1)-truncated ss solution together with the exact solution expressed by Eq. (4.34). Anyhow, the scenario of
the subfigures (a), (b), (c), and (d) is to plot uð2;0Þ ðx; tÞ when a ¼ 1:25, a ¼ 1:5, a ¼ 1:75, and a ¼ 2, respectively, on the domain
½1; 0  ½0; 3.
It is clear from the Fig. 1 that each of the subfigures are nearly coinciding and similar in their behavior. As a result, one can
achieve an excellent approximation and thus the exact solution by using a few terms only. It is to be noted from the subfig-
ures that when the value of x. moving away to the left side of zero on the specific domain of t the representation surface
graph solutions are increasing firstly and then decreases gradually.

Application 4.4. Consider the following nonhomogeneous space-fractional Poisson equation:

@ a uðx; tÞ @ 2 uðx; tÞ
þ ¼ xa1 et ; x; t P 0; 1 < a 6 2; ð4:37Þ
@xa @t 2
subject to the nonhomogeneous initial conditions
uð0; tÞ ¼ et ; ux ð0; tÞ ¼ et : ð4:38Þ
130 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133

Fig. 1. The surface graph of the RPS approximate solution uð1;0Þ ðx; tÞ for Application 4.3: (a) uð2;0Þ ðx; tÞ when a ¼ 1:25, (b) uð2;0Þ ðx; tÞ when a ¼ 1:5, (c) uð2;0Þ ðx; tÞ
when a ¼ 1:75, (d) uð2;0Þ ðx; tÞ when a ¼ 2.

As the previous applications, the initial conditions are on the independent variable x. So, we will replace x by t and t by x
in Eq. (3.1). Moreover, Eq. (4.37) is not exactly as the general form which we suggested in Eq. (1.1) because the term xa1 is of
a power a  1 which should be one of the forms f0; 1; a; a þ 1g. Therefore, we assume that the solution of Eqs. (4.37) and
(4.38) as in the following expansion form:

X X
1 m 2
xjþia X1
xi a
uðx; tÞ ¼ f 00 ðtÞ þ f 01 ðtÞx þ f ij ðtÞ ¼ f 00 ðtÞ þ f 01 ðtÞx þ f i0 ðtÞ : ð4:39Þ
i¼1 j¼0
Cð1 þ j þ iaÞ i¼1
Cð1 þ iaÞ

Using RPS method and according to the initial conditions of Eq. (4.38), the initial guess approximation of uðx; tÞ is
uð0;1Þ ðx; tÞ ¼ et þ et x. Therefore, the RPS solution of Eqs. (4.37) and (4.38) takes the form
X
1
xi a
uðx; tÞ ¼ et þ et x þ f i0 ðtÞ : ð4:40Þ
i¼1
Cð1 þ iaÞ

Whilst the (k, 0)-truncated series of uðx; tÞ and the (k, 0)-truncated residual function that are derived from Eqs. (3.4) and
(3.6) can be formulated, respectively, in form of
X
k
xi a
uðk;lÞ ðx; tÞ ¼ et þ et x þ f i0 ðtÞ ; ð4:41Þ
i¼1
Cð1 þ iaÞ

@ 2 uðk;lÞ
Resðk;lÞ ðx; tÞ ¼ Dax uðk;lÞ   xa1 et : ð4:42Þ
@t2
According to the RPS procedure and without the loss of generality for the remaining computations and results, the fol-
lowing are the first four unknown coefficients in the multiple FPS expansion of Eq. (4.39): f 10 ðtÞ ¼ et , f 20 ðtÞ ¼ et ,
f 30 ðtÞ ¼ et , and f 40 ðtÞ ¼ et . Therefore, the ð4; 0Þ -truncated series solution of Eqs. (4.37) and (4.38) can be written as
xa x2 a x3 a x4 a
uð4;0Þ ðx; tÞ ¼ et þ et x  et þ et  et þ et : ð4:43Þ
Cð1 þ aÞ Cð1 þ 2aÞ Cð1 þ 3aÞ Cð1 þ 4aÞ
 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133 131

Consequently, the components of the RPS approximate solution are obtained as far as we like. In fact, this series is exact to
the last term, as one can verify, of the multiple FPS of the exact solution can be collected to discover that the approximate
solution of Eqs. (4.37) and (4.38) has the general pattern form which is coinciding with the exact solution in terms of infinite
series
X
1
xia
uðx; tÞ ¼ et x þ et ð1Þi : ð4:44Þ
i¼0
Cð1 þ iaÞ

As a special case when a ¼ 2, the RPS solution of Eqs. (4.37) and (4.38) has the general pattern form which is coinciding
with the exact solution in terms of elementary function
uðx; tÞ ¼ et x þ et cos x: ð4:45Þ
Accuracy refers to how closely a computed or measured value agrees with the true value. To show this accuracy for Appli-
cation 4.4, we report the consecutive error which is defined by Conðk;lÞ ðx; tÞ ¼ juðkþ1;lÞ ðx; tÞ  uðk;lÞ ðx; tÞj, where x; t P 0 and uðk;lÞ
is the ðk; lÞ-truncated series of uðx; tÞ obtained from the RPS method. Anyhow, while one cannot know the error without
knowing the solution, in most cases the consecutive error can be used as a reliable indicator in the iteration progresses.
In Table 5, the numerical values of consecutive errors for the two consecutive approximate consecutive solutions have been
calculated when a ¼ 1:25; 1:5; 1:75; 2 and for various x and t on the domain ½0; 1  ½0; 1, to measure the difference between
consecutive solutions obtained from the ð8; 0Þ-RPS approximate solution.
The computational results of Table 5 provide a numerical estimate for the convergence of the RPS method. Also, it is clear
that the accuracy obtained using present method is in advanced by using only few terms approximations. In addition, we can
conclude that higher accuracy can be achieved by evaluating more components of the solution. On the other hand, based on
this heuristic, we terminate the iteration in our method.

Application 4.5. Consider the following nonhomogeneous time-fractional Navier-Stokes equation:

@ a uðx; tÞ @ 2 uðx; tÞ 1 @uðx; tÞ

a ¼pþ þ ¼ 0; p 2 R; x; t > 0; 0 < a 6 1; ð4:46Þ
@t @x2 x @x
subject to the nonhomogeneous initial condition

uðx; 0Þ ¼ 1  x2 : ð4:47Þ

Table 5
The numerical values of the consecutive errors for Application 4.4 at ðk; lÞ ¼ ð8; 0Þ.

x t a ¼ 1:25 a ¼ 1:5 a ¼ 1:75 a¼2

0.2 0 2:22045  10 16 0 0 0
0.25 2:22045  1016 0 0 0
0.5 4:44089  1016 0 0 0
0.75 4:44089  1016 0 0 0
1 4:44089  1016 0 0 0

0.4 0 4:52527  1013 1:11022  1016 0 0

0.25 5:81091  1013 2:22045  1016 0 0
0.5 7:46070  1013 2:22045  1016 0 0
0.75 9:57900  1013 2:22045  1016 0 0
1 1:23013  1012 4:44089  1016 0 0

0.6 0 4:33177  10 11

4:37428  10 14 0 0
0.25 5:56211  1011 5:61773  1014 0 0
0.5 7:14189  1011 7:21645  1014 0 0
0.75 9:17035  1011 9:25926  1014 0 0
1 1:17750  1010 1:19016  1013 0 0

0.8 0 1:10212  109 2:12941  1012 2:88658  1015 0

0.25 1:41514  109 2:73415  1012 3:66374  1015 0
0.5 1:81708  109 3:51086  1012 4:66294  1015 0
0.75 2:33318  109 4:50795  1012 6:21725  1015 0
1 2:99586  109 5:78826  1012 7:77156  1015 0

1 0 1:35664  10 8
4:33044  10 11
9:61453  10 14
1:11022  1016
0.25 1:74196  10 8
5:56040  10 11
1:23457  10 13
2:22045  1016
0.5 2:23672  108 7:13969  1011 1:58540  1013 2:22045  1016
0.75 2:87201  10 8
9:16754  10 11
2:03504  10 13
2:22045  1016
1 3:68773  108 1:17714  1010 2:61346  1013 4:44089  1016
132 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133

According to Remark 2.1 and the previous discussions, we assume that the multiple FPS solution of Eqs. (4.46) and (4.47)
takes the following expansion form:

X
1
t ia
uðx; tÞ ¼ 1  x2 þ f i ðxÞ : ð4:48Þ
i¼1
Cð1 þ iaÞ
Depending on the RPS approach, the initial guess approximation of uðx; tÞ is of the form u0 ðx; tÞ ¼ 1  x2 , while on the
other hand, the kth-truncated series of uðx; tÞ and the kth-truncated residual function are given, respectively, as follows:
X
k
t ia
uk ðx; tÞ ¼ 1  x2 þ f i ðxÞ ; ð4:49Þ
i¼1
Cð1 þ iaÞ

@ 2 uðx; tÞ 1 @uðx; tÞ
Resk ðx; tÞ ¼ Dat uk ðx; tÞ    p: ð4:50Þ
@x2 x @x
According to the RPS procedure and without the loss of generality for the remaining computations and results, we can
conclude that the unknown coefficients f k ðtÞ of Eq. (4.48) should take the following values f 1 ðxÞ ¼ ðp  4Þ and
f k ðtÞ ¼ 0; k ¼ 2; 3; . . .. Consequently, the components of the RPS approximate solution are obtained as far as we like. In fact,
this series is exact to the last term, as one can verify, of the multiple FPS of the exact solution can be collected to discover that
the exact solution of Eqs. (4.46) and (4.47) has the general pattern form which is coinciding with the exact solution
ta
uðx; tÞ ¼ 1  x2 þ ðp  4Þ ; ð4:51Þ
Cð1 þ aÞ
which is the same solution that obtained by the ADM [26]. As a special case when a ¼ 1, the RPS solution of Eqs. (4.46) and
(4.47) has the general pattern form which is coinciding with the exact solution in terms of elementary function
uðx; tÞ ¼ 1  x2 þ ðp  4Þt.

Application 4.6. Consider the following nonhomogeneous time-fractional equation:

@ a uðx; tÞ @ 2 uðx; tÞ @ 2 uðx; tÞ

a ¼ 2
þt þ uðx; tÞ  xt; x 2 R; t P 0; 1 < a 6 2; ð4:52Þ
@t @x @t@x
subject to the nonhomogeneous initial conditions
uðx; 0Þ ¼ 1; ut ðx; 0Þ ¼ x: ð4:53Þ
2
uðx;tÞ
According to Eq. (3.3) with u0 ðxÞ ¼ 1 and u0 ðxÞ ¼ x, and considering the nonlinear differential operator L½uðx; tÞ ¼ @ @x2
2
uðx;tÞ
þt @ @t@x þ uðx; tÞ  xt, we assume that the series solution of Eqs. (4.52) and (4.53) can be written as

X
1 X
1
tjþia
uðx; tÞ ¼ 1 þ xt þ f ij ðxÞ : ð4:54Þ
i¼1 j¼0
Cð1 þ j þ iaÞ

According to the RPS procedure and without the loss of generality for the remaining computations and results, we can
conclude that the unknown coefficients f ij ðxÞ of Eq. (4.54) should take the following values f ij ðxÞ ¼ 1 for all i ¼ 0; 1; 2; . . .
and j ¼ 0; 1. Consequently, the components of the RPS approximate solution are obtained as far as we like. In fact, this series
is exact to the last term, as one can verify, of the multiple FPS of the exact solution can be collected to discover that the exact
solution of Eqs. (4.52) and (4.53) has the general pattern form which is coinciding with the exact solution in terms of infinite
series
 
ta t 1þa t 2a t1þ2a
uðx; tÞ ¼ 1 þ xt þ þ þ þ þ  : ð4:55Þ
Cð1 þ aÞ Cð2 þ aÞ Cð1 þ 2aÞ Cð2 þ 2aÞ
In a particular case, as a special case when a ¼ 2, the RPS solution of Eqs. (4.52) and (4.53) has the general pattern form
which is coinciding with the exact solution in terms of elementary function uðx; tÞ ¼ t þ xt þ et .

5. Conclusion

In this letter, we have developed a new analytical iterative method, so-called RPS, for solving initial value problems of
linear PDEs of fractional order in the Caputo sense. The present technique is performed based on mainly the generation
of residual error function and then applying the generalized Taylor series formula. The RPS does not require linearization,
perturbation, or discretization of the variables, it is not affected by computation round off errors, and one is not faced with
necessity of large computer memory and time. The main advantage of this method is the simplicity in computing the coef-
ficients of terms of the series solution by using the differential operators only and not as the other well-known analytic tech-
niques such as, ADM, HAM, or VIM that need the integration operators which is difficult in the fractional case. Anyhow, using
 A. El-Ajou et al. / Applied Mathematics and Computation 257 (2015) 119–133 133

the present method we can easily discover the pattern of exact solution with less complexity and less computational
calculations.

Acknowledgment

The authors express their thanks to unknown referees for the careful reading and helpful comments.

References

[1] J. Klafter, S.C. Lim, R. Metzler, Fractional Dynamics in Physics: Recent Advances, World Scientific, Singapore, 2011.
[2] D. Baleanu, K. Diethelm, E. Scalas, J.J. Trujillo, Fractional calculus models and numerical methods, Series on Complexity, Nonlinearity and Chaos, World
Scientific, 2012.
[3] Y. Zhou, Basic Theory of Fractional Differential Equations, World Scientific, Singapore, 2014.
[4] R.L. Magin, Fractional Calculus in Bioengineering, Begerll House, West Redding, Conn, USA, 2006.
[5] B. West, M. Bologna, P. Grigolini, Physics of Fractal Operators, Springer, New York, 2003.
[6] R. Hilfer, Applications of Fractional Calculus in Physics, World Scientific, Singapore, 2000.
[7] F. Mainardi, Fractional Calculus and Waves in Linear Viscoelasticity, Imperial College Press, London, UK, 2010.
[8] O. Abu Arqub, Series solution of fuzzy differential equations under strongly generalized differentiability, J. Adv. Res. Appl. Math. 5 (2013) 31–52.
[9] O. Abu Arqub, A. El-Ajou, A. Bataineh, I. Hashim, A representation of the exact solution of generalized Lane-Emden equations using a new analytical
method, Abstract Appl. Anal. 2013 (2013) 10, http://dx.doi.org/10.1155/2013/378593. Article ID 378593.
[10] O. Abu Arqub, Z. Abo-Hammour, R. Al-Badarneh, S. Momani, A reliable analytical method for solving higher-order initial value problems, Discrete Dyn.
Nature Soc. 2013 (2013) 12, http://dx.doi.org/10.1155/2013/673829. Article ID 673829.
[11] A. El-Ajou, O. Abu Arqub, Z. Al Zhour, S. Momani, New results on fractional power series: theories and applications, Entropy 15 (2013) 5305–5323.
[12] O. Abu Arqub, A. El-Ajou, Z. Al Zhour, S. Momani, Multiple solutions of nonlinear boundary value problems of fractional order: a new analytic iterative
technique, Entropy 16 (2014) 471–493.
[13] O. Abu Arqub, A. El-Ajou, S. Momani, Construct and predicts solitary pattern solutions for nonlinear time-fractional dispersive partial differential
equations, Journal of Computational Physics, in Press.
[14] A. El-Ajou, O. Abu Arqub, S. Momani, Approximate analytical solution of the nonlinear fractional KdV-Burgers equation a new iterative algorithm,
Journal of Computational Physics, in Press.
[15] R.L. Magin, C. Ingo, L. Colon-Perez, W. Triplett, T.H. Mareci, Characterization of anomalous diffusion in porous biological tissues using fractional order
derivatives and entropy, Microporous Mesoporous Mater. 178 (2013) 39–43.
[16] S. Cifani, E.R. Jakobsen, Entropy solution theory for fractional degenerate convection-diffusion equations, Annales de l’Institut Henri Poincare (C) Non
Linear Anal. 28 (2011) 413–441.
[17] S. Zhang, H.Q. Zhang, Fractional sub-equation method and its applications to nonlinear fractional PDEs, Phys. Lett. A 375 (2011) 1069–1073.
[18] B. Mathieu, P. Melchior, A. Oustaloup, C. Ceyral, Fractional differentiation for edge detection, Signal Process. 83 (2003) 2421–2432.
[19] F. Mainardi, M. Raberto, R. Goreno, E. Scalas, Fractional calculus and continuous-time finance II: the waiting-time distribution, Physica A 287 (2000)
468–481.
[20] J.H. He, Some applications of nonlinear fractional differential equations and their approximations, Bulletin Sci. Technol. 15 (1999) 86–90.
[21] M. Caputo, Linear models of dissipation whose Q is almost frequency independent: part II, Geophys. J. Int. 13 (1967) 529–539.
[22] S. Momani, Analytic and approximate solutions of the space- and time-fractional telegraph equations, Appl. Math. Comput. 170 (2005) 1126–1134.
[23] S. Momani, Z. Odibat, Analytical approach to linear fractional partial differential equations arising in fluid mechanics, Phys. Lett. A 355 (2006) 271–279.
[24] S. Momani, Z. Odibat, Comparison between the homotopy perturbation method and the variational iteration method for linear fractional partial
differential equations, Comput. Math. Appl. 54 (2007) 910–919.
[25] A. Golbabai, K. Sayevand, Analytical treatment of differential equations with fractional coordinate derivatives, Comput. Math. Appl. 62 (2011) 1003–
1012.
[26] S. Momani, Z. Odibat, Analytical solution of a time-fractional Navier-Stokes equation by Adomian decomposition method, Appl. Math. Comput. 177
(2006) 488–494.
[27] M. Rehman, R.A. Khan, Numerical solutions to initial and boundary value problems for linear fractional partial differential equations, Appl. Math.
Modell. 37 (2013) 5233–5244.
[28] M. Inc, The approximate and exact solutions of the space- and time-fractional Burgers equations with initial conditions by variational iteration
method, J. Math. Anal. Appl. 345 (2008) 476–484.
[29] H. Jafari, A. Golbabai, S. Seifi, K. Sayevand, Homotopy analysis method for solving multi-term linear and nonlinear diffusion-wave equations of
fractional order, Comput. Math. Appl. 59 (2010) 1337–1344.
[30] Z. Odibat, N. Shawagfeh, Generalized Taylor’s formula, Appl. Math. Comput. 186 (2007) 286–293.
[31] A. Wazwaz, S. El-Sayed, A new modification of the Adomian decomposition method for linear and nonlinear operators, Appl. Math. Comput. 122 (2001)
393–405.
[32] S. Momania, Z. Odibat, Comparison between the homotopy perturbation method and the variational iteration method for linear fractional partial
differential equations, Comput. Math. Appl. 54 (2007) 910–919.
[33] J. Biazar, M. Eslami, Analytic solution for Telegraph equation by differential transform method, Phys. Lett. A 374 (2010) 2904–2906.