1. Introduction
Fractional calculus deals with integrals and derivatives of arbitrary order, which unifies and extends integer-order differentiation and
n-fold integration. Up to now, fractional operators have been applied in various areas, such as anomalous diffusion, long-range interactions, long-memory processes and materials, waves in liquids, and physics. Recently, Zimbardo et al. [
1] generalized the Parker equation, which describes the acceleration and transport of energetic particles in astrophysical plasmas, to the case of anomalous by Caputo fractional derivatives. As far as we know, fractional calculus is one of best tools to construct certain electro-chemical problems and characterizes long-term behaviors, allometric scaling laws, nonlinear operations of distributions [
2] and so on.
An integral equation is an equation containing an unknown function under an integral sign. Integral equations are useful and powerful mathematical tools in both pure and applied mathematics. They have various applications in numerous physical problems, chemistry, biology, electronics and mechanics [
3,
4,
5]. Many initial and boundary value problems associated with ordinary (or partial) differential equations can be transformed into problems of solving integral equations [
6]. For example, Gorenflo and Mainardi [
7] provided interesting applications of Abel’s integral equations of the first and second kind in solving the partial differential equation which describes the problem of the heating (or cooling) of a semi-infinite rod by influx (or efflux) of heat across the boundary into (or from) its interior. There have been lots of approaches, including numerical analysis, thus far to studying fractional differential and integral equations, including Abel’s equations, with many applications [
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21]. Recently, Li et al. [
22,
23] studied integral equations associated with Abel’s types in the distributional (Schwartz) sense, based on new fractional calculus of distributions and derived fresh results which are not achievable in the classical sense. On the other hand, the development of science has led to formation of many physical and engineering problems that can be mathematically represented by differential equations. For instance, problems from electric circuits, chemical kinetics, and transfer of heat can all be characterized as differential equations [
24].
We briefly introduce the necessary concepts and definitions of fractional calculus of distributions in
in
Section 2, where we demonstrate several examples of computing fractional derivatives as well as applications to solving Abel’s integral equations of the first kind for arbitrary
. In
Section 3, we present Babenko’s approach to solving several fractional differential and integral equations based on new convolution and product of generalized functions (an active area in distribution theory). We often obtain an infinite series, related to the Mittag-Leffler function, as the solution of integral or differential equation. We imply a number of novel results, which cannot be derived or approximated by numerical analysis methods or by the Laplace transform, since solutions are in the distributional sense in general.
2. Fractional Calculus in
In order to investigate fractional integral and differential equations in the generalized sense, we briefly introduce the following basic concepts, with several interesting examples of solving Abel’s integral equations in distribution. Let
be the Schwartz space (testing function space) [
25] of infinitely differentiable functions with compact support in
R, and
the (dual) space of distributions defined on
. A sequence
goes to zero in
, if and only if these functions vanish outside a certain fixed bounded set, and converge to zero uniformly together with their derivatives of any order. Clearly,
is not empty since it contains the following function
Evidently, any locally integrable function
on
R is a (regular) distribution in
as
is well defined. Hence
f is linear and continuous on
. Furthermore, the functional
on
given by
is a member of
, according to the topological structure of the Schwartz testing function space.
Let
. The distributional derivative
(or
), is defined as
for
. Therefore,
As an example, we will find the distributional derivative of
g (note that this function is not differentiable at
in the classical sense). Indeed, using integration by parts, we derive
which infers that
where
is the Heaviside function.
Let Re
,
. Then the distribution
[
25] is defined by
Clearly, the right-hand side regularizes the integral on the left. This defines the distribution for Re.
Assume that
f and
g are distributions in
. Then the convolution
is well defined by the equation [
25]
for
. This also implies that
Let
be the subspace of
with support contained in
. It follows from [
25,
26,
27] that
is an entire analytic function of
on the complex plane, and
which plays an important role in solving fractional differential equations by using the distributional convolutions. Let
and
be arbitrary numbers, then the following identity
is satisfied [
23].
Let
be an arbitrary complex number and
be a distribution in
. We define the primitive of order
of
g as the distributional convolution
Note that this is well defined since the distributions
g and
are in
. We shall write the convolution
as the fractional derivative of the distribution
g of order
if Re
, and
is interpreted as the fractional integral if Re
.
We should add that Gorenflo and Mainardi [
7] formally presented the derivative of order
n (non-negative integer) of
g by the generalized convolution between
and
g,
based on the well known properties
Then, a formal definition of the fractional derivative of order
could be
Note that this convolution is in the distributional sense (and it does not exist in the classical sense as the kernel is not locally integrable).
Example 1. LetThen,whereIn fact, we have for as the measure of rational numbers is zero. This indicates thatObviously, by the sequential fractional derivative law, we come toandusingFurthermore,wherewhich are regular distributions (locally integrable functions on R). In general, we can getdistributionally by noting thatThen, we imply that for any complex number In particular,by Now we are ready to present the following theorem.
Theorem 1. Let be an infinitely differentiable function on and f be an unknown distribution in . Then the generalized Abel’s integral equation of the first kindhas the solutionwhere α is any real number in R and . In particular, we have four different cases depending on the value of α. - (i)
If for , thenfor . - (ii)
- (iii)
If , then .
- (iv)
If , then for which is well defined.
Proof. Clearly, we can convert Equation (
4) into
which infers that
by using Equations (
1) and (
2). Assuming
for
, we derive that
where
is a locally integrable function on
R since
. Hence,
In particular, choosing
(then
) and
we come to
for a differentiable function
(in the classical sense). This is the solution for the classical Abel’s integral equation.
Obviously, we get from integration by parts,
where
is the testing function. This implies that
By mathematical induction,
This infers that for
The rest of the proof follows easily. This completes the proof of Theorem 1. ☐
Example 2. Let . Then the generalized Abel’s integral equationhas the solution in Proof. Clearly, we can write Equation (
5) to
which deduces that by Theorem 1
as
Obviously,
which claims that
By mathematical induction, we come to
This completes the proof by noting that
☐
Example 3. Then the generalized Abel’s integral equationhas the solution in where is the Mittag-Leffler function, defined by the series expansionfor . Proof. Evidently, we can convert Equation (
6) to
Hence from Theorem 1, we get
Setting
, we arrive at
This completes the proof by using
Furthermore, the generalized Abel’s integral equation
has the solution in
for
and
.
Similarly, the generalized Abel’s integral equation
has the solution in
for
. ☐
To end off this section, we would add that many applied problems from physical, engineering and chemical processes lead to integral equations, which at first glance have nothing in common with Abel’s integral equations, and due to this perception, additional efforts are undertaken for the development of analytical or numerical procedure for solving these equations. However, their transformations to the form of Abel’s integral equations will speed up the solution process [
24], or, more significantly, lead to distributional solutions in cases where classical ones do not exist [
22].
As an example, the following integral equation with a moving integration limit
can be converted into Abel’s integral equation to solve in the distributional sense. The interested readers are referred to [
22] for the detailed methods.
3. Babenko’s Approach in Distribution
In this section, we shall extend the method used by Yu. I. Babenko in his book [
28], for solving various types of fractional differential and integral equations in the classical sense, to generalized functions. The method itself is close to the Laplace transform method in the ordinary sense, but it can be used in more cases, such as solving fractional differential equations with variable coefficients. Clearly, it is always necessary to show convergence of the series obtained as solutions by other analytical tools, although it is a hard job in general [
24]. We point out that Babenko’s method can also be used to solve certain partial differential equations for heat and mass transfer theory, and suggest the interested readers are referred to [
24] for the detailed arguments.
We must add that Oliver Heaviside (1850–1925) introduced an ingenious way of solving ordinary or partial differential equations arising from electromagnetic problems through algebration [
29,
30,
31]. He considered, for example, the factor
as product of the differential operator
with
f itself. Therefore, the differential equation for a given
can be converted to
which implies that the solution
if it converges. This method is identical to Babenko’s Approach.
Let
be given. We now study Abel’s integral equation of the second kind
with demonstrations of examples in the distributional space
, where
and
is a constant. Further, we will investigate and solve several integral equations with variable coefficients by products of distributions and fractional operations of generalized functions. The results derived here cannot be achieved by the Laplace transform in general, or numerical analysis methods since distributions are undefined at points in
R.
Clearly, Equation (
8) is equivalent to the convolutional equation
in
, although it is undefined in the classical sense for
. We should point out it becomes the differential equation
if
.
Example 4. Let λ be a nonzero constant and . Then the fractional differential equationhas the solution in the space Proof. Equation (
9) can be written into
which is equivalent to
By Babenko’s method we get
which is convergent, and hence well defined, by noting that
is a locally integrable function on
R.
In particular, the differential equation
has the solution in the space
Similarly, the fractional differential equation
has the solution
where
.
Using the same argument, the fractional differential equation
has the solution
where
and
is a regular distribution.
More generally, the fractional differential equation
has the solution
☐
Remark 1. We begin by using Example 4 as a simple demonstration of Babenko’s Approach. Clearly, all the equations presented above can be easily solved by the Laplace transform. For example, applying the Laplace transform to the equationgivesThe inverse transform implies However, we must mention that Babenko’s approach in the distributional sense is much more general than that of the Laplace transform. As an example, the Laplace transform does not work for the equationas the distribution is not locally integrable. As indicated below, Babenko’s approach provides an efficient method of dealing with this kind of equation. Example 5. Let . Then the fractional differential and integral equation (mixed type)has the solution in the space Proof. Equation (
10) can be changed to
which is equivalent to
Applying
to both sides of the above equation, we only need to study the following integral equation
Therefore,
which is convergent. ☐
Example 6. The generalized Abel’s integral equationhas the solution in the space Proof. Equation (
11) can be written into
By Babenko’s method we get
Clearly,
where
and erfc
is the error function complement defined by
The result follows from the sum of and . ☐
Remark 2. Equation (11) cannot be discussed in the classical sense, including the Laplace transform, since the fractional integral or derivative of , does not exist in the normal sense. Clearly, the distribution in the solution is a singular generalized function in , while the restis regular (locally integrable). Assume f is a distribution in and g is a function in . Then the product is well defined byfor all functions as . Therefore, the product, for ,makes sense since is in andis a distribution in (subspace of ) for arbitrary if , according to Section 2. Example 7. Let be an integer. Then the integral equationhas a solution in the space where m is a non-negative integer and are arbitrary constants. Proof. First, we show that
for
On the other hand, we have
if
.
Therefore,
which implies that
where
are arbitrary constants.
Furthermore, we let
be an integer. Then the integral equation
has a solution in the space
where
m is a non-negative integer and
are arbitrary constants.
It generally follows that the integral equation
has a solution
where
. ☐
Example 8. Let be an integer. Then the generalized integral equationhas a solution in the space where and are arbitrary constants. Proof. To solve this integral equation, we require the following more complicated product of distributions, as
is not an infinitely differentiable function (but it is a locally integrable function). It follows from Theorem 2.3 in [
32] that
for
. We note that this product can be derived directly from the complex analysis approach based on the Laurent series of
,
as well as
, given below
Hence,
by noting that
where
and
are arbitrary constants. This implies that Equation (
12) has a solution
Similarly, we let
be an integer. Then, the integral equation
has a solution in the space
where
m is a non-negative integer, and
are arbitrary constants.
Clearly, the integral equation
has a solution
where
. ☐
Gorenflo and Mainardi [
7] presented the applications of Abel’s integral equations of the first and second kind to solve the following partial equation of heat flow:
in the semi-infinite intervals
and
of space and time, respectively. Our results on the distributional Abel’s integral equations have potential applications to dealing with differential equations in distribution and hence finding distributional (weak) solutions.
Remark 3. Generally speaking, there is the lack of definitions for nonlinear operations, such as product and composition in distribution theory, although it is of great demand in the areas of differential equations and quantum field theory. As an example, it seems hard to define the distribution , asis undefined. Fisher, with his coauthors [33,34,35,36,37,38,39,40,41], has actively used the δ-sequence and neutrix limit due to van der Corput since 1969, to deduce numerous products, powers, convolutions, and compositions of distributions by several workable definitions. From the above examples, it is clear to see the relations between products of generalized functions and integral equations with variable coefficients in distribution.