O.I. Marichev, E.L. Shishkina Overview of Fractional Calculus and Its Computer Implementation in Wolfram Mathematica

O.I. Marichev, E.L.
Shishkina
Overview of fractional calculus and its computer implementation in
Wolfram Mathematica
arXiv:2306.11660v1 [math.HO] 20 Jun 2023
1
Contents
Introduction 4
1 History, nowadays and applications 5

1.1 Background and concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.1 Idea of non-integer derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.2 Formal definition of fractional integro-differentiation . . . . . . . . . . . . . . . . 7
1.2 Different approaches to classical fractional calculus . . . . . . . . . . . . . . . . . . . . 8
1.2.1 Riemann-Liouville type fractional integrals and derivatives . . . . . . . . . . . . 8
1.2.2 Finite differences approach to fractional calculus . . . . . . . . . . . . . . . . . . 12
1.2.3 One dimensional Riesz and Bessel potentials and their inversions . . . . . . . . . 13
1.2.4 Fractional integro-differentiation of analytic functions . . . . . . . . . . . . . . . 14
1.2.5 Weyl fractional derivative of a periodic function . . . . . . . . . . . . . . . . . . 16
1.3 Generalizations of fractional derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3.1 Operators with power-logarithmic kernels. Fractional powers of operators . . . . 17
1.3.2 Erdelýi-Kober-type operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3.3 Local fractional derivatives and variable-order differential operators . . . . . . . 19
1.3.4 Generalized fractional integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.4 Applied aspects of fractional derivatives and integrals . . . . . . . . . . . . . . . . . . . 22
1.4.1 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.4.2 Numerical calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.4.3 Publications after 1987 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2 Fractional-order differentiation in the Wolfram language 26

2.1 Riemann-Liouville-Hadamard integro-differentiation in the Wolfram language . . . . . . 26
2.1.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.1.2 Simplest examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 Calculation of fractional derivatives and integrals by series expansion . . . . . . . . . . 30
2.2.1 Basic power–logarithmic examples . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.2.2 Hadamard finite part . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.3 Hadamard finite part for basic power-logarithmic examples . . . . . . . . . . . . 33
2.3 The Meijer G-function and fractional calculus . . . . . . . . . . . . . . . . . . . . . . . 37
2.3.1 Definitions and main properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3.2 Representability through Meijer G-functions . . . . . . . . . . . . . . . . . . . . 40
2.3.3 Fractional integro-differentiation of Meijer G–functions . . . . . . . . . . . . . . 41
2
2.3.4 Fractional integro-differentiation of ez and Kν (z) . . . . . . . . . . . . . . . . . 42
2.3.5 Riemann-Liouville integral of generalized Meijer G–function . . . . . . . . . . . 43
2.3.6 Big O representations for asymptotics of Meijer G-functions . . . . . . . . . . . 47
2.3.7 Conditions of convergence for Riemann-Liouville integrals with Meijer G-functions 48
2.4 Support of differential constants. Generic formulas for fractional differentiation . . . . 50
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
References 54
3
Introduction
In this article we would like to consider some approaches to non-integer integro-differentiations and
its implementation in computer algebra system Wolfram Mathematics.
In the Section 1 some historical moments and applications are briefly discussed. The ideas that
started the era of fractional calculus are presented in Subsection 1.1. Some different approaches to
d
non-integer power of dx are given in Subsection 1.2. Generalizations of fractional derivatives such as
d
operators with a power-logarithmic kernel, non-integer power of x dx and x1γ dx
d γ d
x dx , fractional derivative
of a function f with respect to another function g, Erdelýi-Kober type operators, local fractional
derivatives, variable-order differential operators and sequential fractional derivatives are described in
Subsection 1.3. Fractional calculus has diverse and widespread applications. Some references to works
where fractional calculus applies to physics, engineering, medicine are presented in Subsection 1.4 as
well as some approaches to numerical calculation of fractional derivatives.
In Section 2 some essential details of the fractional integro-diffirentiation, which has been realized
in Wolfram Mathematica are presented. This Section form a new picture of fractional differentiation
for analytic functions of complex variables. The Hadamard regularized Riemann-Liouville operator was
taken as the basis for creating the integro-differential operator in Wolfram Mathematica. The Riemann-
Liouville-Hadamard integro-diffirentiation of an arbitrary function to arbitrary symbolic order α in the
Wolfram Language is described in Subsection 2.1. In Subsection 2.2 the approach in which fractional
integro-diffirentiation applies to each term of Taylor series expansions of all functions near zero using
Hadamard regularization was presented. For evaluation derivative of arbitrary complex order Meijer
G-function can be used. Quite a lot of functions can be represented as a Meijer G-function, and
it is easy to find a α-derivative of the Meijer G-function. Such approach is described in Subsection
2.3. In Subsection 2.4 fractional integro-diffirentiation applies to ”differential constants”, which are
not constants everywhere, but they are different constants on some domains of complex plane with
discontinuity on some lines. Also in this last Subsection includes description of generic formulas for
fractional differentiation.
The list of references contains numerous of items, but of course the total number of works on
fractional calculus is immense.
4
1. History, nowadays and applications
1.1 Background and concept

1.1.1 Idea of non-integer derivative
Calculus teaches us how to compute derivatives of any integer order. We can interpret differentiation
of negative integer order as a repeated integration. The zero order of differentiation gives the function
itself. The question is how to generalize derivatives to non-integer order? The first attempt to discuss
such an idea recorded in history was contained in the correspondence of Leibniz. In one of his letters
to Leibniz, Bernoulli asked about the meaning of one theorem in the case of non-integer order of
differentiation. Leibniz in his letters to L’Hôpital in 1695 and to Wallis in 1697 made some remarks on
the possibility of considering differentials and derivatives of order 1/2 (see [49, 92] for to get acquainted
with the main milestones of fractional calculus).
In the 18-19th centuries, mathematicians introduced many important functions, using integral and
series representations. In 1729, Euler derived the integral representation for factorial n!, leading to the
gamma function Γ(n+1) = n!, allowing one to define n! for any complex numbers n except 0, −1, −2, ....
In addition to the gamma function, there are other ways to generalize the factorial to complex numbers
(see [63], p. 35-36, [64]). Euler introduced the gamma function by the integral, which later got name
Euler integral of the second kind:
Z∞
Γ(α) = xα−1 e−x dx, Re(α) > 0. (1)
0
Property Γ(α) = Γ(α+1)

α
can be used to uniquely extend the Γ(α) to a meromorphic function defined for
all complex numbers α, except integers less than or equal to zero. In α=0, −1, −2, ... gamma function
has simple poles. Such a generalization of the factorial allowed Euler to notice that the concept of the
n-th order derivative of the power function xp acquired meaning for a non-integer n.
Namely, let n ∈ N, x > 0 and p ∈ R. It is obviously that (xp )(n) = p(p − 1)...(p − n + 1)xp−n or
p!
(xp )(n) = xp−n . (2)
(p − n)!
The expression (2) has sense and for non-integer n if we use the gamma function (1) and the derivative
of xp of non-integer order α can be defined by
dα p Γ(p + 1)
α
x = xp−α . (3)
dx Γ(p − α + 1)
5
We should notice here that formula (3) is valid not for all values p and α. So, we have to exclude
the arguments giving the poles of the gamma function. That means that in (3)
−(p + 1), −(p − α + 1) ∈

/ N ∪ {0}.
d 0
p p
However we know that for α = 0 we just get the dx 0 x = x and for α = −1 we get integral and for
α = −2, −3, ... iterated integrals. For example, if α = −1 and p = −1 we obtain
Z
d−1 −1 dx
−1
x = = log(x) + C.
dx x
We can see that this case already gives us not a power function, but a logarithm. Let’ s find out how
to get the logarithm from the below formula, including arbitrary constant C
Z
xp+1
xp dx = +C
p+1
1
by passing to the limit. We can change this constant on the other constant − p+1 + C1 , including
parameter α and a constant C1 . After evaluation, integral and taking limit
Z
xp+1 1 xp+1 − 1
xp dx = − + C1 = + C1 ,
p+1 p+1 p+1
we get Z Z
dx xp+1 − 1
= lim xp dx = lim + C1 = log(x) + C1 .
x p→−1 p→−1 p + 1

1
Here we applied regularization using the term − p+1 . The general approach to finding an arbitrary
order derivative of a power function is considered in Subsection 2.2.
The question of the −1/2 order derivative was raised by Euler, who, in 1738, mentioned that
derivative of power function z b has a meaning for non-integer order of differentiation. But only much
later in 1820 Lacroix realized Euler’s idea and presented an exact formula for the order −1/2 derivative
of the power function z b (see [92] for details).
Therefore, if function f (x) is locally given by a convergent power series or f (x) is an analytic
function: ∞
X f (p) (0)
f (x) = ap xp , ap = ,
p=0
p!
then the derivative of order α > 0 can be formally defined as
∞
dα f (x) X Γ(p + 1)
= ap xp−α .
dxα p=0
Γ(p − α + 1)
Of course, for existing of such fractional derivative, it is necessary to verify the convergence of the
specified series. Here α > 0 and p ≥ 0 so we are avoiding the poles of the gamma function. General
case when α and p can be negative is considered in Subsection 2.2.
In 1823 first application of fractional calculus was discovered by Abel (see for details [49, 83]). Abel
was looking for a curve in the plane such that the time required for a particle to slide down the curve
6
to its lowest point under the influence of gravity is independent of its initial position on the curve.
Such curve is called the tautochrone. Abel found that for funding this curve it is necessary to solve the
equation
Zx
f (t)dt
= ϕ(x), x > 0.
(x − t)1/2
0
It should be noticed that Abel solved more general equation

Zx
f (t)dt
= ϕ(x), x > 0,
(x − t)α
0
where 0 < α < 1.

Further, we note that in 1832 Liouville formally extended the formula for the integer derivative of
dn bx dα bx
the exponent dx ne (b is a some number) to derivatives of an arbitrary order dx αe . Namely,
dα ebx
α
= bα ebx . (4)
dx
Based on the formula (4), one can formally write the derivative of the order α ∈ R of an arbitrary
function f represented by the series
∞ ∞
dα f (x) X α bk x X
= ck bk e , where f (x) = ck ebk x .
dxα
k=0 k=0
The limitation of this definition is related to the convergence of the series.
1.1.2 Formal definition of fractional integro-differentiation

So, starting from the 17th century, the need for a formal definition of fractional integro-differentiation
gradually formed. We present here such definition following [92].
Let z ∈ C, x ∈ R are variables, the starting point in fractional integro-differentiation is 0, ν ∈ C or
ν ∈ R is an order of integro-differentiation. Fractional integro-differentiation D ν acts to f (z) by z and
acts to f (x) by x. In [92] one can find the following criteria of fractional integro-differentiation.
Criteria 1. Operator D ν f (z) is the integro-differential operator of order ν ∈ C if and only if
1. If f (z) is an analytic function of the complex variable z, the derivative D ν f (z) is an analytic
function of ν and z.
2. The operation D ν f (z) must produce the same result as ordinary differentiation when ν is a positive
integer. If ν is a negative integer, say ν = −n, then D −n f (z) must produce the same result as
ordinary n-fold integration and D −n f (z) must vanish along with its (n − 1)-derivatives at z = 0.
3. The operation of order zero leaves the function unchanged:
D 0 f (z) = f (z).
7
4. The fractional operators must be linear:
D ν [af (z) + bg(z)] = aD ν f (z) + bD ν g(z).
5. The semigroup property of arbitrary order holds:

D ν D µ f (z) = D ν+µ f (z).
In the next Subsections 1.2 and 1.3 we examine different operators which realized non-integer powers
of differentiations and integrations.
1.2 Different approaches to classical fractional calculus

d α
In this Subsection we consider different classical approaches to introduce a non-integer power dx
d
of the differentiation operator dx . If an operator has two forms, left-sided and right-sided, we will
consider only the left-sided.
1.2.1 Riemann-Liouville type fractional integrals and derivatives

d
The most well known definitions of operators giving arbitrary real (complex) powers of the dx
which
fulfill Criteria 1 are
Zx
α 1 f (t)
(Ia+ f )(x) = dt, a < x ≤ b, (5)
Γ(α) (x − t)1−α
a
and
n Zx
α 1 d f (t)dt
(Da+ f )(x) = , a < x ≤ b. (6)
Γ(n − α) dx (x − t)α−n+1
a
Operators (5) and (6) are called left-sided Riemann-Liouville integral and derivative on the segment
[a, b], respectively. The definition of fractional integral is based on a generalization of the formula for
an n-fold integral
Zx Zx Zx Zx
1
dx... dx f (x)dx = (x − t)n−1 f (t)dt, x > a. (7)
(n − 1)!
|a {za a
} a
n
Then the fractional derivative (6) is obtained as a left inverse operator to (5) which is constructed
as a solution to the Abel equation.
α
If we consider Da+ xp , x > 0, p ∈ R we obtain
 
n Zx p
α 1 d t dt
Da+ xp =  =
Γ(n − α) dx (x − t)α−n+1
a
8

n
1 dp−α+n Γ(p + 1)Γ(n − α)
= x − B xa (p + 1, n − α) =
Γ(n − α) dx Γ(p + n − α + 1)
n
Γ(p + 1) p−α 1 d
= x − xp−α+n B xa (p + 1, n − α).
Γ(p − α + 1) Γ(n − α) dx
It is easy to see that in order for the resulting formula to correspond to (3), we need to set a = 0, when
last term disappear:
α Γ(p + 1)
D0+ xp = xp−α .
Γ(p − α + 1)
Because of this, in the Section 2 we will consider only fractional derivatives and integrals tied to the
origin, i.e. with a = 0.
Liouville fractional integral of order α on semiaxis (0, ∞) for f ∈ L1 (0, ∞) has the form
Zx
α 1 f (t)
(I0+ f )(x) = dt, x ∈ (0, ∞). (8)
Γ(α) (x − t)1−α
0
Liouville fractional integral of order α on the whole real axis is

Zx
1 f (t)
(I+α f )(x) = dt, x ∈ R. (9)
Γ(α) (x − t)1−α
−∞
Liouville fractional derivatives of order α on the semi-axis and on the axis are, respectively
n Zx
α 1 d f (t)dt
(D0+ f )(x) = , x ∈ (0, +∞), (10)
Γ(n − α) dx (x − t)α−n+1
0
n Zx
α 1 d f (t)dt
(D+ f )(x) = , x ∈ R, (11)
Γ(n − α) dx (x − t)α−n+1
−∞
where n = [α] + 1.
Next, we consider a modification of Riemann-Liouville integration, for which lower and upper integral
limits are symmetric. For fixed arbitrary point c ∈ R and α > 0 operator
 x
R


 (x − t)α−1 f (t)dt if x > c;
1
(Icα f )(x) = c
Rc (12)
Γ(α) 
 (t − x) α−1
f (t)dt if x < c

x
is called the Chen fractional integral.

Let α > 0 is not integer, n = [α] + 1 then Chen fractional derivative is

 Rx f (t)dt
d n

 dx (x−t)α−n+1
if x > c;
α 1 c
(Dc f )(x) = Rc f (t)dt (13)
Γ(n − α) 
 − d n
if x < c.
 dx (t−x)α−n+1
x
9
Riemann-Liouville, Liouville and Chen operators are considered in this book.
One of the most popular fractional derivative that used now is the Caputo fractional derivative
which has the form
Zx
1 f (n) (y)dy
( C D0+
α
f )(x) = , n = [α] + 1, x > 0. (14)
Γ(n − α) (x − y)α−n+1
0
Operator (14) was introduced by an Italian physicist M. Caputo in 1967 in paper [9] and studied in the
monograph [10].
In 1948 (see [34], submitted in 1947) the Soviet mechanic A.N. Gerasimov introduced fractional
derivative of the form
Zx
1 f ′ (y)dy
, y > 0, x ∈ R, 0 < α < 1. (15)
Γ(1 − α) (x − y)α
−∞
In the same work, A.N. Gerasimov studied two new problems in viscoelasticity theory. He reduced this
problem to differential equations with partial fractional derivative.
Let us notice that in the article [83] Abel’s contribution to the theory of fractional calculus was
carefully analyzed, in particular the appearance of the derivative (14) for α ∈ (0, 1).
The popularity of the fractional derivative (14) in applications is explained as follows. If we consider
the fractional differential equation with Riemann–Liouville fractional derivative of the form
α
(D0+ f )(x) = λf (x), x > 0, 0 < α < 1, λ∈R
we should add the initial condition

α−1
(D0+ f )(0+) = 1
and the solution of this problem is (see [43])
f (x) = xα−1 Eα,α (λxα ) (16)

∞
P (λxα )k
where Eα,β (λxα ) = Γ(αk+β)
is the Mittag–Leffler function. Note that we have a singularity at zero
k=0
in (16). Therefore, it is not possible to consider the classical Cauchy problem at zero initial point with
Riemann–Liouville fractional derivative.
From the other side, the solution to the Cauchy problem for fractional differential equation with
fractional derivative (14)
( C D0+
α
f )(x) = λf (x), x > 0, 0 < α ≤ 1, λ∈R
f (0+) = 1,
is (see [43])
f (x) = Eα,1 (λxα ). (17)
Now it is bounded at zero and so the classical Cauchy problem is correct.
10
Let find the Riemann–Liouville derivative of (17):
∞ ∞
α
X λk X λk Γ(αk + 1)
D0+ Eα,1 (λxα ) = α
D0+ xαk = xαk−α =
k=0
Γ(αk + 1) k=0
Γ(αk + 1) Γ(αk − α + 1)
∞ ∞
X λk xα(k−1) x−α X (λxα )k−1
= = +λ
Γ(αk − α + 1) Γ(1 − α) Γ(α(k − 1) + 1)
k=0 k=1
∞
x−α (λxα )k
X x−α
= +λ = + λEα,1 (λxα ).
Γ(1 − α) k=0
Γ(αk + 1) Γ(1 − α)
x −α
So we see that the term Γ(1−α) has a singularity at x = 0.
C α p
Let consider D0+ x , p ∈ R:
Zx
C α Γ(p + 1) y p−ndy
D0+ xp = =
Γ(p − n + 1)Γ(n − α) (x − y)α−n+1
0
Γ(p + 1) Γ(n − α)Γ(p − n + 1)xp−α Γ(p + 1)

= = xp−α .
Γ(p − n + 1)Γ(n − α) Γ(p − α + 1) Γ(p − α + 1)
Let us note that integration can only be done with p > n − 1, where n = [α] + 1. Therefore, we have
serious limitations and cannot, generally speaking, apply the derivative (14) to term by term to each
member of the power series. In order to get around this problem, we put
(
Γ(p+1)
C α p Γ(p−α+1)
xp−α if p > n − 1;
D0+ x =
0 if p ≤ n − 1
C α
then we can find D0+ Eα,1 (λxα ) by the term-by-term differentiation of the series
∞ ∞
C α
X λk X (λxα )k−1
D0+ Eα,1 (λxα ) = C α αk
D0+ x = λ = λEα,1 (λxα ).
k=0
Γ(αk + 1) k=1
Γ(α(k − 1) + 1)
So here we saw significant limitations when using a fractional derivative (14).

In addition to the Caputo derivative, various other modifications of the Riemann-Liouville derivative
appeared. We present here some of them.
Although we only consider left-sided operators, we will mention one right-sided fractional integral
of the form
Z∞
α 1
(I− f )(x) = (y − x)α−1 f (y)dy.
Γ(α)
x
Some author call it Weyl fractional order integral of function f (y) (see [123]).
Modified Riemann–Liouville fractional derivative is (see [40], formula 2.10)
Zx
1 d (f (y) − f (0))dy
( M RL D0+
α
f )(x) = , x > 0, 0 < α < 1.
Γ(1 − α) dx (x − y)α
0
11
Γ(p+1)
Since xp |x=0 = 0 for p > 0 we get M RL D0+
α
xp = Γ(p−α+1) xp−α for p > 0.
Cossar in [14] presented the following fractional derivative
ZN
α 1 d f (y)
(D− f )(x) =− lim dy, 0 < α < 1.
Γ(1 − α) N →∞ dx (y − x)α
x
Osler fractional derivative (see [75], formula 2.2)

Z
O α Γ(1 + α) f (t)
( D0+ f )(x) = dt, 0 < α < 1.
2πi (t − z)α+1
C[a,z + ]
where the contour C[a, z + ] starts and ends at t = 0.

Properties of the Riemann–Liouville and the Caputo derivatives were studied in [53].
1.2.2 Finite differences approach to fractional calculus

Now let consider how to generalize derivative of order n of the form
(∆nh f )(x)
f (n) (x) = lim , (18)
h→0 hn
where (∆nh f )(x) is a finite difference
n
k n n n!
X
n
(∆h f )(x) = (−1) f (x − kh), = (19)
k=0
k k k!(n − k)!
to the non-integer order.

Since the gamma-function is a generalization of the factorial function to non-integer values then for
α > 0 a fractional derivative can be given as a generalization of (18) by
(∆αh f )(x)
f (α) (x) = lim , (20)
h→+0 hα
where
∞
X α α (−1)k−1 αΓ(k − α)
(∆αh f )(x) = (−1) k
f (x − kh), = . (21)
k=0
k k Γ(1 − α)Γ(k + 1)
For h > 0 and α > 0 (20) is the left-hand sided Grünwald–Letnikov derivative. We can also consider
the case when h < 0 in (21). The left-hand sided Grünwald–Letnikov fractional integral in a real line is
f (−α) (x) = lim hα (∆−α
h f )(x). (22)
h→+0
Let find Grünwald–Letnikov derivative of the power function xp , x > 0, p ∈ R. Since the definition
(20) was given for whole real line, we consider a function
(
xp , x > 0;
fp (x) =
0, x 6 0.
12
Then fp can be written as an inverse Laplace transform
c+i∞
Z
Γ(p + 1)
fp (x) = s−p−1 exs ds (c > 0),
2πi
c−i∞
then
c+i∞
Z ∞
Γ(p + 1) X α −khs
fp(α) (x) = xs −p−1
e s lim h −α
(−1) k
e ds =
2πi h→+0
k=0
k
c−i∞
c+i∞
Z c+i∞
Z
Γ(p + 1) xs −p−1 −hs α Γ(p + 1) Γ(p + 1)
= e s lim h −α
(1 − e ) ds = exs sα−p−1 ds = xp−α
2πi h→+0 2πi Γ(p − α + 1)
c−i∞ c−i∞
α
that coincides with D0+ xp
for x > 0.
Grünwald–Letnikov type fractional derivative was considered in [74].
The definitions of Marchaud derivatives based on finite differences approach (21) is
Z∞ l
1 (∆lt f )(x) X l α
(Dα+ f )(x) =− dt, Al (α) = (−1) k−1
k , (23)
Γ(−α)Al (α) tα+1 k=0
k
0
where 0 < Re α < l, l ∈ N or l = [α] + 1 when α ∈ R.

Other modifications to fractional differences constructed from fractional differences are presented in
Chapter 4 of this book. For example, Grünwald–Letnikov–Riesz fractional derivative of order α > 0
was defined by
1 (∆αh f )(x) + (∆α−h f )(x)

( GLR Dxα f )(x) = lim (24)
2 cos(απ/2) h→0+ |h|α
where the difference (∆αh f )(x) of a fractional order α > 0 is defined by the series (21).
Grünwald–Letnikov operators (20) and (22), Marchaud derivative (23) and Grünwald–Letnikov–
Riesz fractional derivative (24) are considered in [52] and in this book.
1.2.3 One dimensional Riesz and Bessel potentials and their inversions
The integral
Z∞
α 1 f (t)dt
(I f )(x) = π
, Re α > 0, α 6= 1, 3, 5, ... (25)
2Γ(α) cos 2
α |t − x|1−α
−∞
is called the Riesz potential. Along with (25), we consider its modification of the form
Z∞
α 1 sgn(x − t)
H f (x) = π
f (t)dt, Re α > 0, α 6= 2, 4, 6, ... (26)
2Γ(α) sin 2
α |t − x|1−α
−∞
13
For 0 < α < 1 operators inverse to I α and H α may be constructed in the forms
Z∞
α −1 1 f (x − t) − f (x)
((I ) f )(x) = π
dt,
2Γ(−α) cos 2
α |t|1+α
−∞
Z∞
α −1 1 f (x − t) − f (x)
((H ) f )(x) = π
sgn t dt.
2Γ(−α) sin 2
α |t|1+α
−∞
Next we introduce a convolution operator

Z∞
(Gα f )(x) = Gα (x − t)f (t)dt,
−∞
which is defined using the Fourier transform by the equality

1
F [Gα f ](x) = F [f ](x), Re(α) > 0. (27)
(1 + |x|2 )α/2
The function Gα (x), whose Fourier transform is (1 + |x|2 )−α/2 , is evaluated in terms of Bessel functions.
That is why the operator Gα (x) is called an operator of Bessel fractional integration or Bessel potential.
Bessel potential or Bessel fractional integral can be written in the form
1−α Z∞
α 2 2 α−1
(G f )(x) = √ α
f (t)|x − t| 2 K 1−α (|x − t|) dt, Re(α) > 0.
πΓ 2
2
−∞
1.2.4 Fractional integro-differentiation of analytic functions

∞
P
Let the function f (z) = fk z k , be analytic in the unit disc so
k=0
∞
X Γ(k + 1)
(Dα0 f )(z) = z −α fk z k . (28)
k=0
Γ(k − α + 1)
Γ(k+1)
A natural way to generalize the (28) is to replace the factor Γ(k−α+1) in (28) by a more general one.
One of the such generalization is Gel’fond-Leont’ev differentiation of the form
∞
n
X ak−n
D (a; f ) = fk z k−n , (29)
k=n
ak
∞
P
where a(z) = ak z k . The operator in (29) is called the Gel’fond-Leont’ev operator of generalized
k=0
dn f
differentiation. It is obvious that Dn (a; f ) = dz n
in the case a(z) = ez .
14
The operator
∞
X ak+n
In (a; f ) = fk z k+n ,
k=0
ak
which is the right inverse to (29), will be called the Gel’fond-Leont’ev operator of generalized inte-
gration. Operators Dn and In are introduced as direct generalizations concerning integer order n of
integro-differentiation, they contain that for fractional order as well. To show this, let us consider the
P∞
zk
following special case when a(z) = Eα (z) = Γ(αk+1)
, α > 0 is the Mittag-Leffler function [13]. The
k=0
corresponding operator for generalized integration of order n = 1 is
∞
1
X Γ(αk + 1)
(Iα f )(z) = I (E1/α ; f ) = fk z k+1 .
k=0
Γ(αk + α + 1)
This operator admits the following integral representation

Z1
1
(Iα f )(z) = (1 − t)α−1 f (ztα )dt.
Γ(α)
0
Corresponding differential operator for 0 < α < 1 is

Zz
1 1 d f (t)g ′(t)
(Dα f )(z) = dt,
′
Γ(1 − α) g (z) dz [g(z) − g(t)]α
0
where g(z) = z 1/α . The operator Dα corresponds to the expansion

∞
X Γ(αk + 1)
(Dα f )(z) = fk z k−1 .
k=1
Γ(αk + 1 − α)
∞
P
If we consider the analytic in the unit disc function b(z) = bk z k then the Hadamard product
k=0
composition of functions b(z) and f (z) is
∞
X
D{b; f } = b ◦ f = bk fk z k . (30)
k=0
The (30) is a very wide generalization of the differentiation. Under the assumption bk → ∞ a generalized
integration is
∞
X fk
I{b; f } = zk . (31)
k=0
bk
Choosing various functions b(z) in (30) and (31) we obtain integro-differentiation operations of various
Γ(α+1)
types. Then b(z) = (1−z) α+1 (30) gives Riemann-Liouville fractional differentiation of the function
Γ(α+1)z
z α f (z). If we took b(z) = (1−z) α+1 in (30) we obtain the Ruscheweyh fractional derivative. Function
∞
P
b(z) = (ik)α z k gives the fractional differentiation by Weyl.
k=1
15
1.2.5 Weyl fractional derivative of a periodic function
Let f (x) be a 2π-periodic function on R and let
∞ Z2π
X
−ikx 1
f (x) ∼ fk e , fk = e−ikx f (x)dx
2π
k=−∞ 0
be its Fourier series. Here we will consider functions having zero mean value:
Z2π
f (x)dx = 0.
0
For periodic functions definition fractional integro-differentiation, suggested by Weyl, is

∞ Z2π
(α)
X 1
(I+ f )(x) ∼ (ik)−α fk eikx , fk = e−ikt f (t)dt, f0 = 0. (32)
k=−∞
2π
0
Similarly fractional differentiation is defined:

∞ Z2π
(α)
X 1
(D+ f )(x) ∼ (ik)α fk eikx , fk = e−ikt f (t)dt, f0 = 0. (33)
2π
k=−∞ 0
The definition (32) may be interpreted as

Z2π
(α) 1
(I+ f )(x) = f (x − t)Ψα+ (t)dt, α > 0, (34)
2π
0
where
∞
X cos(kt − απ/2) 1
Ψα+ (t) = 2 = e− 2 iπα eiπα Liα e−it + Liα eit , (35)
k=1
kα
∞
P zk
where Liα (z) = kα
is the polylogarithm function. The right-hand side in (34) is called the Weyl
k=1
fractional integral of order α.
The Marchaud–Weyl derivative is defined as
Z2π
(α) 1 d
(D+ f )(x) = (f (x) − f (x − t)) Ψα+ (t)dt, 0 < α < 1. (36)
2π dt
0
R2π
It is known that for 2π-periodic function f , such that f ∈ L1 (0, 2π) and f (x)dx = 0 the Weyl
0
(α)
fractional integrals I+ for 0 < α < 1 coincide with the Liouville integrals on the real line I+α (9):
(α)
(I+ f )(x) = (I+α f )(x). Weyl approach to fractional derivative of a periodic function was considered in
this book.
16
1.3 Generalizations of fractional derivatives
1.3.1 Operators with power-logarithmic kernels. Fractional powers of op-
erators
α α
One of the direct generalizations of the fractional integrals Ia+ and Ib− on a finite interval [a, b] of
the real axis has the form
Zx
(α,β) 1 β γ
(Ia+ f )(x) = ln (x − t)α−1 f (t)dt, α > 0, β ≥ 0, γ > b − a, (37)
Γ(α) x−t
a
Zb
(α,β) 1 β γ
(Ib− f )(x) = ln (t − x)α−1 f (t)dt, α > 0, β ≥ 0, γ > b − a, (38)
Γ(α) t−x
x
containing both logarithmic and power singularities. These constructions are called operators with
power-logarithmic kernels. Such integrals arise when investigating integral equations of the first kind
with power-logarithmic kernels.
Operators with a power-logarithmic kernel was studied in [44] for integer β and γ = 1. In the general
case these operators were considered in [45].
d α
Riemann-Liouville fractional integro-differentiation is formally a fractional power dx of the
d
differentiation operator dx and is invariant relative to translation if considered on the whole axis.
Hadamard [36] suggested
a construction of fractional integro-differentiation which is a fractional power
d α
of the type x dx . This construction is well suited to the case of the half-axis, and is invariant relative
to dilation.
Hadamard fractional integral has the form
Zx
1 f (t)dt
(Fα+ f )(x) = 1−α , x > 0, α > 0. (39)
Γ(α) t ln xt
0
It is easily seen that operator Fα+ is connected with Liouville operators I+α of the form (9)
(Fα+ f )(x) = A−1 I+α Af, (Af )(x) = f (ex ).
For 0 < α < 1 Hadamard fractional derivative has the form

Zx
1 d f (t)dt
(Dα+ f )(x) = x α .
Γ(1 − α) dx t ln xt
0
We may also consider Hadamard fractional integral and derivative on a finite segment [a, b].
Next we briefly consider the fractional powers (Bγ )α , α ∈ R of differential Bessel operator in the
form
γ 1 d d d
Bγ = D 2 + D = γ xγ , γ ≥ 0, D := . (40)
x x dx dx dx
17
For fractional powers of (40) explicit formulas were derived in [66] as compositions of simpler operators.
An important step was done in [113] in which explicit definitions were derived in terms of the Gauss
hypergeometric functions with different applications to PDE. The most general study was fulfilled by
I. Dimovski and V. Kiryakova [19–21,46] for the more general class of hyper–Bessel differential operators
related to the Obrechkoff integral transform.
−α
The left–sided fractional Bessel integral Bγ,a+ on a segment [a, b] for f ∈L1 (a, b), a, b ∈ (0, ∞), is
defined by the formula
−α α
(Bγ,a+ f )(x) = (IBγ,a+ f )(x) =
Zx γ 2 2 2α−1
1 y x −y γ−1 y2
= 2 F1 α+ , α; 2α; 1− 2 f (y)dy. (41)
Γ(2α) x 2x 2 x
a
n−α n−α
Let α > 0, n = [α] + 1, f ∈L1 (a, b), IBγ,b− f, IBγ,a+ f ∈C 2n (a, b) and even. The left-sided fractional
Bessel derivatives on a segment of the Riemann-Liouville type for α 6= 0, 1, 2, ... is defined as
α α
(Bγ,a+ f )(x) = (DBγ,a+ f )(x) = Bγn (IBγ,a+
n−α
f )(x), n = [α] + 1.
When α = n ∈ N ∪ {0}, then

0 n
(Bγ,a+ f )(x) = f (x), (Bγ,a+ f )(x) = Bγn f (x),
where Bγn is an iterated Bessel operator (40). Fractional Bessel integrals and derivatives were studied
in [105, 106].
1.3.2 Erdelýi-Kober-type operators

Let 0 ≤ a < x < b ≤ ∞ for any σ ∈ R or −∞ ≤ a < x < b ≤ ∞ for integer σ. Erdelýi-Kober-type
operators are
Zx
α σx−σ(α+η)
Ia+; σ, η f (x) = (xσ − tσ )α−1 tση+σ−1 f (t) dt, α > 0, (42)
Γ(α)
a
n
α d
Ia+; σ, η f (x) =x −σ(α+η)
xσ(α+n+η) Ia+;
α+n
σ, η f (x), α > −n, n ∈ N. (43)
σxσ−1 dx
After the change of variables xσ = y, tσ = τ (42)–(43) are reduced to the usual Riemann-Liouville
fractional integrals and derivatives
α −α−η α
Ia+; σ, η f (x) = y (Iaσ + ϕ)(y), ϕ(y) = y η f (x), xσ = y. (44)
Erdelýi-Kober operators are essential and important in transmutation theory. For example, the most
well–known transmutations of Sonine and Poisson are of this class when σ = 2 (see for details [42,112]).
Important properties of Erdelýi-Kober operators were studied in the monographs [42, 107].
18
Let Re α > 0. The fractional integral of a function f with respect to another function g is
Zx
1
α
(Ia+,g f )(x) = (g(x) − g(t))α−1 g ′(t)f (t)d t, −∞ ≤ a < b ≤ ∞. (45)
Γ(α)
a
Integral (45) is defined for every function f (t) ∈ L1 (a, b) and for any monotone function g(t), having a
continuous derivative.
α α
If g ′ (x) 6= 0, a ≤ x ≤ b, then operators Ia+,g , Ib−,g are expressed via the usual Riemann-Liouville (or
Liouville) fractional integration after the corresponding changes of variables
α α α α
Ia+,g f = QTc+ Q−1 f, Ib−,g f = QTd− Q−1 f, c = g(a), d = g(b),
where (Qf )(x) = f [g(x)].

The fractional derivative of a function f with respect to another function g of the order α ∈ (0, 1) is
Zx
α 1 1 d f (t)
(Da+,g f )(x) = g ′ (t)d t, −∞ ≤ a < b ≤ ∞. (46)
′
Γ(1 − α) g (x) dx (g(x) − g(t))α
a
It is also possible to consider the Marchaud and others forms of the fractional derivatives of a
function f with respect to another function g.
The operational calculus approach for fractional calculus with respect to functions were given in [31].
Riemann-Liouville fractional integral is obtained by choosing g(x) = x in (45). If we take in (45)
the function g(x) = xσ we obtain Erdelýi-Kober type operator; if we take g(x) = ln x in (45) we get
Hadamard fractional integral, and a choice g(x) = exp(−x) with its applications was considered in [24].
As A. M. Djrbashian pointed out to us operators of fractional derivatives of a function with respect
to another function (45) even in some more general setting were introduced and studied by his father
M.M. Djrbashian, cf. [22, 23, 25]. In these papers were studied integral representations of this operator
class, their inversion and corresponding integro-differential equations of fractional order.
1.3.3 Local fractional derivatives and variable-order differential operators

Another branch of fractional calculus is the theory of fractional derivatives and integrals adapted to
work with highly irregular non-differentiable curves and surfaces, the graphs of which are fractal sets.
Such operators are called local fractional derivative and local fractional integrals. Since local fractional
derivatives and integral refer to describe the properties of fractal objects and processes, its theory called
fractal calculus (see [39, 48]).
The local fractional derivative (D α f )(x0 ) of order α of a function f (x) at the point x0 can be defined
by
n−1 (k)
!
X f (x 0 )
(D α f )(x0 ) = lim Dxα0 f (x) − (x − x0 )k , n − 1 < α ≤ n, (47)
x→x0 k!
k=0
where Dxα0 is the left-sided Riemann-Liouville fractional derivative of order α.
19
As a result, the local fractional derivative (47) of order α at the point x = x0 can be defined via the
above Caputo fractional derivative by the equation
(D α f )(x0 ) = lim ( C Dxα0 f )(x), n − 1 < α ≤ n, (48)

x→x0
As a result, we can see that local fractional derivative is a limit of the Caputo fractional derivative.
Local fractional Yang integral is
Zb N −1
1 1 X
( a Ibα f )(x) = f (t)(dt)α = lim f (tj )(∆tj )α , (49)
Γ(1 + α) Γ(1 + α) ∆t→0 j=0
a
where {tj }Nj=0 is a partition of the segment [a, b], t0 = a, tN = b, ∆tj = tj+1 − tj , j=0, 1, 2, ..., N−1 and
N −1
∆t = max{∆tj }j=0 .
In [103] fractional integration and differentiation when the order is not a constant but a function
was presented. In order to define variable-order differential operators in [103] two ways were given: the
first way is a direct one and the second uses Fourier transforms. For the Riemann–Liouville fractional
integrals of variable order, we have (see [103], formula 2)
Zx
α(x) 1
(Ia+ f )(x) = (x − y)α(x)−1 f (y)dy, Re(α(x)) > 0, x > a. (50)
Γ(α(x))
a
If a > −∞ we have the Riemann definition and for a = −∞ we have the Liouville definition.
The Riemann-Liouville derivative is also extended to the case of variable order (see [103], formula
3):
Zx
α(x) 1 d
(Da+ f )(x) = (x − y)−α(x) f (y)dy, 0 < Re(α(x)) < 1, x > a. (51)
Γ(1 − α(x)) dx
a
We observe that the fractional operators (50) and (51) are not inverse to each other as in the case
−α(x) α(x)
of constant order, as it will be seen below. So, it will not be correct to introduce Da+ as [Da+ ]−1 .
Coimbra derivative [12] is
Zx
CV α(x) 1 f ′ (y) f (0+) − f (0−)
( D0 f )(x) = α(x)
dy + α(x) , 0 ≤ Re(α(x)) < 1.
Γ(1 − α(x)) (x − y) x (Γ(1 − α(x)))
0
Now let consider fractional order depending of two variables α(x, y). Let f : [a, b] → R be a function.
The left–sided Riemann–Liouville fractional integral of order α(·, ·) is defined by (see [62])
Zx
(x − y)α(x,y)−1
( a Ixα(·,·) f )(x) = f (y)dy, x > a. (52)
Γ(α(x, y))
a
20
Left–sided Riemann–Liouville derivative of order α(x, y) is
Zx
d (x − y)−α(x,y)
( a Dxα(·,·) f )(x) = f (y)dy, x > a, 0 < α(x, y) < 1. (53)
dx Γ(1 − α(x, y))
a
Similarly introduced Caputo and Marchaud derivatives of variable fractional order (see [102, 103]).
Next we consider so called sequential derivatives. Djrbashian–Nersesyan fractional derivatives, as-
sociated with a sequence {γ0 , γ1 , . . . , γm } of order σ, where σ = γ0 + γ1 + ... + γm , is defined by
σ
DDN = D γ0 D γ1 · · · D γm , (54)
where D γk are fractional integrals and derivatives of Riemann–Liouville with some endpoint. These
operators were introduced in [25–27] and then studied and applied in [28]. The original definitions
demand −1≤γ0 ≤0, 0≤γk ≤1, 1≤k≤m, as in the above papers integrodifferential equations under such
conditions were studied for operators (54). But Djrbashian–Nersesyan fractional operators may be
defined and considered for any parameters γk if appropriate definitions of Riemann–Liouville operators
are used.
Djrbashian–Nersesyan operators were patterns for introducing in the book of Miller and Ross [68]
more general sequential operators of fractional integrodifferentiation for which compositions in defini-
tions of the form (54) are consisted of any fractional operators, cf. an useful discussion in [79].
1.3.4 Generalized fractional integrals

There are a lot of generalizations of fractional integro–differential operators are connected with com-
binations and compositions of more standard fractional operators. For example, averaged or distributed
order fractional operator, associated with any given fractional operator Rt , is introduced by the next
formula
Zb
(a,b)
IM R f = Rt f (t)d t, (55)
a
where Rt is a given fractional operator of order t of any kind. In case then Rt is in particular the
fractional Riemann–Liouville operator names continued or distributed fractional integrals or derivatives
are often used. Such operators were studied in [87, 88].
Other fractional integrals containing the Gaussian hypergeometric function in the kernel are the
Saigo fractional integrals (see [101]) which are defined by
Z∞
1 x
Jxγ,β,η f (x) = (t − x)γ−1 t−γ−β 2 F1 γ + β, −η; γ; 1 − f (t)dt, (56)
Γ(γ) t
x
and
Zx
x−γ−β t
Ixγ,β,η f (x) = (x − t) γ−1
2 F1 γ + β, −η; γ; 1 − f (t)dt, (57)
Γ(γ) x
0
21
where γ > 0, β, θ are real numbers. Another similar class of generalizations introduced by Love and
more generalizations with special function kernels are mentioned in this book.
We consider the case of the interval [0, 1]. Following Hadamard [36] and M.M. Djrbashian [22, 23]
we introduce the operator
Z1
(ω)
(L f )(x) = − f (xt)ω ′(t)dt, (58)
0
where the function ω ∈ C([0, 1]) is supposed to satisfy the following assumptions:
1. ω(x) is monotone,
2. ω(0) = 1, ω(1) = 0, ω(x) 6= 0 as 0 < x < 1,
3. ω ′(x) ∈ L1 (0, 1).

(1−x) α
If ω(x) = Γ(1+α) , then obviously (L(ω) f )(x) = x−α (I0+
α α
f )(x), where I0+ is (8).
M.M. Djrbashian (also Mkhitar Dzhrbashjan, M. M. Jerbashian) in [22,23] considered the operators
L(ω) in a more general form
 1 
Z Z1
(ω) d  ω(x)
(L f )(x) = − x f (xt)dp(t) , p(t) = t dx.
dx x2
0 t
1.3.5 Conclusion
We mentioned different approaches to fractional integro-differentiation. It can be important to apply
them to basic simplest functions z λ , ez , etc. which wildly used in Taylor and Fourier series.
As we have seen, it is possible to find different approaches to create a fractional derivative and a
fractional integral. The method of analytic continuation is very well known, but it is not possible to
use this method directly in order to continue analytically the derivative with respect to the order of
differentiation because initially an order of derivative can be only natural. It is possible only to find
an analytic function that coincides with the n-th derivative of a function when plugging in a natural
number, and next analytically continue this function to the whole or to the almost whole complex plane.
Therefore, we can construct fractional derivative and integral in a non-unique way.
From the other side, if two different integro-differential operators satisfy the criteria 1 there is a
class of functions where these operators coincide. For example, if α > 0 and f ∈ L1 (a, b) the left-hand
sided Grünwald–Letnikov fractional integral (22) is equal to Riemann-Liouville fractional integral (5)
when a = 0 for almost all x (see [79]).
1.4 Applied aspects of fractional derivatives and integrals

1.4.1 Applications
The reasons for the large number of applications of fractional calculus are as follows.
22
1. The nonlocality of the fractional derivative makes it possible to use it for mathematical modeling
of media with memory.
2. In many models of physics, biology and medicine, differential equations of fractional orders de-
scribe the phenomenon under consideration more accurately, since the order of the fractional
derivative gives an additional degree of freedom.
3. Fractional equations describe non-Markovian processes, which opens up new horizons in proba-
bility theory and statistics.
The use of fractional calculus in theoretical physics and mechanics described in the books [1, 37, 79,
117]. The mail idea of using fractional derivatives and integrals in physics is based on nonlocal effects
which may occur in space and time. Thus, the fractional derivative by time in the model is interpreted
by physicists as the presence of the memory of the described process. A fractional derivative with
respect to a spatial variable indicates the presence of movement restrictions.
The paper [119] introduces scientists who started to apply fractional calculus to scientific and en-
gineering problems during the nineteenth and twentieth centuries. How to use the Mellin integral
transform in fractional calculus was described in [57, 58]. In [35] fundamental theorem of fractional
calculus is presented.
The use of fractional derivatives in continuum mechanics is due to the emergence of new polymeric
materials that have both the property of viscosity and the property of elasticity. Describing the behavior
of solids using viscoelastic models that contain fractional order operators and relate stresses to strains
was began in 1948 from papers of Rabotnov Yu. N. (see translation [89]) and of Gerasimov A. N. [34].
Let mention here the paper [71] where there are many biographical facts and papers of Gerasimov A. N.,
including his paper [34] were given. Models of viscoelasticity with fractional operator were described by
Rossikhin Yu. A. and Shitikova M. V. in Encyclopedia of Continuum Mechanics [94]. The mathematical
modeling of a stress-strain state of the viscoelastic periodontal membrane using fractional calculus is
given in [5]. The fractional derivative of strain with time to the ratio between stress and strain is used
in this case to describe the intermediate state of the material between elastic and liquid [96, 104, 108].
Monograph [61] provides an overview of application of fractional calculus and waves in linear viscoelastic
media, which includes fractional viscoelastic models, dispersion, dissipation and diffusion. Diethelm K.
in [18] described numerical methods for the fractional differential equations of viscoelasticity.
In [11] it was noticed that in the modelling of the energy dissipation and dispersion in the propagation
of elastic waves, introduction of memory mechanisms in their constitutive equations is needed. It turned
out that the most successful memory mechanism used to represent variance and energy dissipation is a
fractional derivative. The fractional derivative generalizing the stress strain relation of anelastic media
was introduced in [10, 11].
Models with fractional order operators used in dynamic problems of rigid body mechanics first
appeared in [33, 34, 104] and described in reviews [93, 108–110]. Fractional calculus in the problems of
mechanics of solids was given in [95, 97–99].
In problems of thermodynamics, solutions for unsteady flows near the boundary of a semi-infinite
region are expressed by combinations of fractional derivatives in [2]. In paper [67] thermodynamics
was generalized in terms of fractional derivatives, while the results of traditional thermodynamics by
Carnot, Clausius and Helmholtz were obtained in the particular case when the exponent of the fractional
derivative is equal to one. In [85] was described how fractional calculus used in thermoelasticity.
23
Application of fractional modelling in rheology providing by Scott Blair was presented in [91].
Uchaikin V. V. in [118] presented fractional models in hydromechanics.
Applications of fractional calculus in electromagnetic including fractional multipoles, fractional solu-
tions for Helmholtz equations, and fractional-image methods were studied and briefly reviewed in [29,30].
In [76] fractional-order chaotic systems were studied. Namely, the model of chaotic system considered
in [76] as three single differential equations, where the equations contain the fractional order derivatives.
Application of the fractional calculus in the control theory can be found in [78]. It was shown that
for fractional-order systems the most suitable way is to use fractional-order controllers involving an
integrator and differentiator of fractional order.
The works [121, 122] deals with the fractional generalization of the diffusion equation, fractality
of the chaotic dynamics and kinetics, and also includes material on non-ergodic and non-well-mixing
Hamiltonian dynamics.
Physical interpretation of a fractional integral can be found in [100]. Authors of [41] gave interpre-
tation of fractional-order operators in fracture mechanics.
Multidisciplinary presentation of fractional calculus for researchers in fields such as electromag-
netism, control engineering, and signal processing were given in the book [73].
In [6] was overviewed how to use fractional calculus in biomechanics. Collection of applications of
fractional calculus in science and engineering can be found in [115].
Epidemiological SIR model in which classic first order ODE’s were replaced by fractional derivatives
is gaining popularity due to Covid 19 (see, for example, [70]). The presence of fractional derivatives in
the model allows you to increase the number of degrees of freedom.
The book [120] shows how the fractional calculus can be used to model the statistical behavior of
complex systems, gives general strategies for understanding wave propagation through random media,
the nonlinear response of complex materials and the fluctuations of heat transport in heterogeneous
materials.
Review [47] is fully described advances in the analysis of the fractional and generalized fractional
derivatives from the probabilistic point of view. The theory of fractional Brownian motion and other
long-memory processes is describes in [69], where was particularly obtained different forms of the Black-
Scholes equation for fractional market. In [4] Wiener-transformable markets, where the driving process
is given by an adapted transformation of a Wiener process including processes with long memory,
like fractional Brownian motion and related processes were studied. Probabilistic interpretation of a
fractional integration is given in [114].
1.4.2 Numerical calculus

Fractional derivatives and integrals are complex objects, the calculation of which is often technically
difficult. In this case, numerical methods are applied. Since the fractional derivative is non-local
approximation of it is more complicated than the integer derivatives. Professor of Mathematics in
Shanghai University Changpin Li with coauthors made a great contribution to the development of the
numerical calculus of fractional integrals and derivatives and their applications. Almost all existing
numerical approximations for fractional integrals and derivatives were systematically examined in [54–
56].
Difference schemes for solutions to equation with fractional time derivative was first used in [111]. In
a series of articles [81, 82, 84] triangular strip matrices are used for approximating fractional derivatives
24
and solving fractional differential equations.
Monte Carlo method is introduced for numerical evaluation of fractional-order derivatives in [50,51].
A feature of these works is that the points at which the function is calculated are distributed unevenly
and their distribution depends on the order of the derivative.
Group of scientists as Igor Podlubny, Ivo Petras, Blas Vinagre and others deeply advanced in ap-
plications of fractional derivatives in physics including fractional order controller and distributed-order
dynamic systems (see [8, 38, 76–80]).
1.4.3 Publications after 1987

After 1987 till nowadays, a lot of monographs (more then 200) by authors from different countries
had been published, wherein various aspects of fractional calculus, and its applications were considered.
Nowadays, the list of publications wholly or partly devoted to fractional calculus and its applications
is endless. Information about current state of fractional calculus can be found in [7, 15–17, 59, 90, 116,
119, 123].
We should mention the great impact of Virginia Kiryakova with her coauthors. She published a
book about generalized Fractional derivatives [46] and a lot of papers, including popularizing fractional
derivatives [21, 119]. Also she is an editor-in-chief of the most popular journal specialized in fractional
calculus: ”Fractional Calculus and Applied Analysis” (FCAA; Fract. Calc. Appl. Anal.).
A more detailed list of books on fractional integro-differential calculus is presented in [60].
25
2. Fractional-order differentiation in the Wolfram language
2.1 Riemann-Liouville-Hadamard integro-differentiation in

the Wolfram language
2.1.1 Definition
We describe the Riemann-Liouville integro-differentiation of an arbitrary function to arbitrary sym-
bolic order α in the Wolfram Language. Using techniques described below, this operation, hereafter
referred to as ”fractional differentiation”, has been published in the Wolfram Function Repository as
ResourceFunction["FractionalOrderD"]. Defined via an integral transform, fractional differentia-
tion is an analytic function of α which coincides with the usual α-th order derivative when α is a
positive integer and with repeated indefinite integration for negative integer α.
dα f (z)
We will use notation z α for Riemann-Liouville integro-diffirentiation for all α ∈ C.
α
Definition. By definition of d dzf α(z) we put


 f (z), α = 0;

 (α)


 f (z), α ∈ Z and α > 0;

 z t t

 Z Z Z




 dt... dt f (t)dt, α ∈ Z and α < 0;


0
dα f (z) | {z0 0
}
= −α (59)
dz α 
 1 d n Rz
f (t)dt


 Γ(n−α) dz n (z−t)α−n+1
, n = ⌊α⌋ + 1 and Re(α) > 0;

 0


 1
Rz f (t)dt


 Γ(−α) (z−t)α+1
, Re(α) < 0 and α ∈ / Z;

 0


 1 d
Rz f (t)dt

 Γ(1−α) dz (z−t)α
, Re(α) = 0 and Im(α) 6= 0,
0
where in the case of divergent integral we use Hadamard finite part. Such integro-differentiation (59)
is called Riemann-Liouville-Hadamard derivative.
Above we separated cases of symbolic positive integer n-th order derivatives from generic result
integro-differentiation of fractional order. In particular for α = −1, −2, ... we have
Zz Zz Zt Zt
1 f (t)dt
= dt... dt f (t)dt.
Γ(−α) (z − t)α+1
0
|0 {z0 0
}
−α
26
So we can combine the third and fifth formulas. Function FractionalOrderD realized regularized
Riemann-Liouville integro-diffirentiation. That means if any of the integrals in (59) diverges we use
Hadamard regularization of this integral (see Subsection 2.2).
dα
Function FractionalOrderD which calculating dz α is presented in Wolfram Function Repository. In
order to get this function we should write

ResourceFunction["FractionalOrderD"].
2.1.2 Simplest examples

Let us consider how the function FractionalOrderD acts to simple functions. For example,
FractionalOrderD[x2 ,{x, α}] gives
2x2−α
.
Γ(3 − α)
and FractionalOrderD[sin[z],{z, α}] gives
(
sin z + πα 2 if α ∈ Z, α ≥ 0;
√
2α−1 πz 1−α α 3−α z2
Γ(1− α
F 1; 1 − 2 , 2 ; − 4
Γ 3−α 1 2
in other cases.
2) ( 2 )
′′′
If we put α = 3 in previous result we obtain (sin(z)) = − cos(z).
When α = −3, f (z) = ez we obtain
   
−3 z Zz Zz Zt3 Zt2 2
d e 1 2 t   et1 dt1  dt2  dt3 = ez − z − z − 1.
= (z − t) e dt =
dz −3 Γ(3) 2
0 0 0 0
For α = −1/2 and for α = 1/2 we can write

Zz
d−1/2 ez 1 et √
−1/2
= 1/2
dt = ez erf( z),
dz Γ(1/2) (z − t)
0
Zz !
d1/2 ez 1 d et Γ − 21 , z
1/2
= dt = ez √ +1 ,
dz Γ(1/2) dz (z − t)1/2 2 π
0
Rz 2
where erf(z) is the integral of the Gaussian distribution, given by erf(z) = √2 e−t dt, Γ (α, z) is the
π
0
R∞
incomplete gamma function, given by Γ(α, z) = tα−1 e−t dt.
z
Now we can evaluate FractionalOrderD operator for more then 100000 functions, analytical in
variable z. Let show some √ examples starting with the ”simplest” mathematical functions, involving
only one or two letters z1 , z, z b , az , ez , z z . Next, we see ”named functions” with head-titles like
√
log(z), sin(z), Jν (z), Jz (b). At last we see ”composed functions” z 2 , (z a )b , az , arcsin(z 3 ). If we apply
c
differentiation or indefinite integration

′
1 1 √ 1
= − 2, ( z)′ = √ , (z b )′ = bz b−1 , (az )′ = az log(a), (ez )′ = ez ,
z z 2 z
27
1
(z z )′ = z z (log(z) + 1), (log(z))′ = , (sin(z))′ = cos(z),
z
1 √ z
(Jν (z))′ = (Jv−1 (z) − Jv+1 (z)), ( z 2 )′ = √ ,
2 z2
c c 3z 2
((z a )b )′ = abz a−1 (z a )b−1 , (az )′ = c log(a)z c−1 az , (arcsin(z 3 ))′ = √ ,
1 − z6
Z Z Z
dz √ 2z 3/2 z b+1
= log(z) + C, zdz = + C, z b dz = + C,
z 3 b+1
Z Z Z
z az z z
a dz = + C, e dz = e , log(z)dz = −z log(z) − z + C,
log(a)
Z Z
2−ν−1 z ν+1 Γ v+1 2 v+1 v + 3 z2
sin(z) = − cos(z) + C, Jν (z)dz = 1 F2 ; v + 1, ;− ,
Γ (ν + 1) Γ ν+3 2
2 2 4
Z √ √ Z
z z2 z (z a )b
2
z dz = + C, (z a )b dz = + C,
2 ab + 1

z (log(a) (−z c ))−1/c Γ 1c , −z c log(a)
Z
zc
a dz = − + C,
c
Z
3 3 3 4 1 2 5 6
arcsin(z )dz = z arcsin(z ) − z 2 F1 , ; ;z .
4 2 3 3
We find that not each integral and even derivative can be evaluated. Here there are no results for
Z Z
′ z
(Jz (b)) , z dz, Jz (b) dz.
Also in many cases we see special functions as result of integration.

Operation FractionalOrderD evaluates repeatable α-integer order derivatives and indefinite inte-
grals and extends results for arbitrary complex or real order α. For example:

dα 1 −α 1 1,1
0
=z 2 F2 (1, 1; 2, 1 − α; −z) + G1,2 z
,
dz α z Γ(1 − α) −1, α
√ √ 1 −α
dα z πz 2 dα z b Γ(b + 1)z b−α
= , = ,
dz α 2Γ 23 − α dz α Γ(b − α + 1)

α z

 az logα (a), α ∈ Z, α ≥ 0;
d a
= az logα (a)(1 − Q(−α, z log(a))), α ∈ Z, α < 0;
dz α 
 z −α α
a z (z log(a)) (1 − Q(−α, z log(a))), in other cases,
(
dα ez ez , α ∈ Z, α ≥ 0;
α
= z
dz e (1 − Q(−α, z)), in other cases,

πα
α
d sin(z)  sin 2
+ z , α ∈ Z, α ≥ 0;
= 1−α α 3 α z2

dz α  z 1 F2 1;1− 2 , 2 − 2 ;− 4 , in other cases,
Γ(2−α)
28
√ α−2ν
dα Jν (z) π2 Γ(v + 1)z ν−α ν+1 ν+2 ν−α+1 ν −α+2 z2
= ν−α+2 2 F3 , ; , , ν + 1; − ,
dz α Γ ν−α+1
2
Γ 2
2 2 2 2 4
√ √
dα z 2 z 2 z −α dα (z a )b z −α (z a )b Γ(ab + 1)
= , = ,
dz α Γ(2 − α) dz α Γ(ab − α + 1)
dα arcsin(z 3 )
=
dz α

6z 3−α 1 1 2 5 7 4 2 α 5 α α 7 α 4 α 3 α 6
= 7 F6 , , , , 1, , ; − , − , 1 − , − , − , − ; z ,
Γ(4 − α) 2 2 3 6 6 3 3 6 6 6 6 6 6 3 6 2 6
R∞ α−1 −t
where Q(α, z) = Γ(α,z)
Γ(α)
is the regularized incomplete gamma function, Γ(α, z) = t e dt is the
z
c
dα z z dα log(z) dα Jz (b) dα az
incomplete gamma function. Values dz α
, dz α , dz α , dz α
are also calculated but formulas too
much complicated.
We would like to compare Mathematica operations FractionalOrderD, D and Integrate. The
direct computation from definition in Mathematica gives
1 40320
D7 = D[1/z2
, z, 7] = − ,
z2 z9
d7 1 40320
7 2
= ResourceFunction[”FractionalOrderD”][1/z2 , z, 7] = − 9 ,
dz z z
Zz Zt Zt Z
dt 2 137z 5 1 5
dt... dt 2
= Nest #dz&, 1/z , 7 = − z log(z),
t 7200 120
|0 {z0 0
}
7
d−7 1
= ResourceFunction[”FractionalOrderD”][1/z2 , z, −7] =
dz −7 z 2

1 5 77 137z 5 1 5
=− z log(z) − = − z log(z).
120 60 7200 120
Therefore FractionalOrderD generalizes D and Integrate.
α J (b)
We should mention, that differentiation by parameters of functions (for example, d dz z
α ) as rule can
not produce well known functions. But we can say the following: ”The first derivative with respect
to an upper ”parameter” ak , and all derivatives of symbolic integer order m with respect to a ”lower”
parameter bk of the generalized hypergeometric function p Fq (a1 , ..., ap ; b1 , ..., bq ; z), can be expressed in
terms of the Kampé de Fériet hypergeometric function of two variables” (see [140]).
dα
Let us describe how FractionalOrderD works. For evaluation of dz α we are using three approaches.
1. Generic formulas for exist for single simple functions; these are converted into pattern matching
rules. Here power series, generalized power series and Hadamard regularization are used.
2. Convert function to the Meijer G-function, using powerful new MeijerGReduce operator and
rule-based MeijerGForm (applies to hypergeometric type functions). Then use formula for the
fractional derivative of the Meijer G-function.
29
3. Generic formulas for fractional differentiation, which produce Appell function F1 of two variables
(in future Lauricella function D of several variables and Humbert function Φ of two variables will
be used).
In the following sections we consider these three approaches.
2.2 Calculation of fractional derivatives and integrals by se-

ries expansion
2.2.1 Basic power–logarithmic examples
d α
Let us consider the first approach to calculating dz α . The FractionalOrderD function allows us
to find all these and many other fractional derivatives because this operation applied to each term of
Taylor series expansions of all functions near zero. So if
∞
X
b
f (z) = z cn z n , (60)
n=0
then
∞
dα f (z) X dα z b+n
= cn . (61)
dz α n=0
dz α
Sum representations by formula (60) we meet for functions like

∞ ∞
z
X zn X (−1)n z 2n+ν
e = , Jν (z) = ,
n=0
n! n=0
n!Γ(n + ν + 1) 2
but sometimes series expansions include log(z) function as in the logarithmic case of K0 (z):
z ∞
X Hn z 2n
K0 (z) = − log + γ I0 (z) + , (62)
2 n=1
(n!)2 2
n
P 1
with n-th harmonic number Hn = k
and γ is Euler–Mascheroni constant
k=1
n
! Z∞
X 1
γ = lim − log n + =− e−x log(x)dx. (63)
n→∞
k=1
k
0
It means, that we should consider more general series

∞
X
b k
fL (z) = z log (z) cn z n , (64)
n=0
30
and evaluate for arbitrary b, α and integer k = 0, 1, 2, ... the following values
∞
dα fL (z) X dα b+n k
= cn α z log (z). (65)
dz α n=0
dz
α λ d0 z λ
Let λ = b + n. In order to calculate (61) we should find ddzzα by the formula (59). Since dz 0
= zλ
and for α ∈ Z and α ≥ 0 we have α = 0, 1, 2, ... and integer derivative gives
dα z λ
= λ(λ − 1)...(λ − α + 1)z λ−α = (λ − α + 1)α z λ−α = (−1)α (−λ)α z λ−α , (66)
dz α
where (λ)α is a Pochhammer symbol.
Therefore we obtain



 zλ , α = 0;




 (λ − α + 1)α z λ−α , α ∈ Z and α > 0;


 1 dn
Rz tλ dt
α λ
d z  Γ(n−α) dz n (z−t)α−n+1
, n = ⌊Re(α)⌋ + 1 and Re(α) > 0;
= 0 (67)
dz α 
 1
Rz tλ dt


 Γ(−α) (z−t)α+1
, Re(α) < 0;

 0


 1 d
Rz tλ dt
 Γ(1−α) dz (z−t)α ,
 Re(α) = 0 and Im(α) 6= 0.
0
d α k
λ
In order to calculate (65) we should find dz α z log (z) by the formula (59)



 z λ logk (z), α = 0;

 λ k (α)


 (z log (z)) , α ∈ Z and α > 0;

 Rz λ k
 1 d n t log (t)dt
dα λ k  Γ(n−α) dz n (z−t)α−n+1
, n = ⌊α⌋ + 1 and Re(α) > 0;
z log (z) = 0 (68)
dz α 
 1
Rz tλ logk (t)dt


 Γ(−α) (z−t)α+1
, Re(α) < 0;

 0


 1 d
Rz tλ logk (t)dt

 Γ(1−α) dz (z−t)α
, Re(α) = 0 and Im(α) 6= 0.
0
Here for α ∈ Z and α > 0

α
λ k (α)
X α Γ(λ + 1) λ−j dα−j logk (z)
(z log (z)) = z α−j
.
j=0
j Γ(λ − j + 1) dz
If some of integrals
Zz Zz Zz Zz
tλ dt tλ dt tλ logk (t)dt tλ logk (t)dt
, , , (69)
(z − t)α+1 (z − t)α−n+1 (z − t)α+1 (z − t)α−n+1
0 0 0 0
in (67) or (68) diverges we take Hadamard finite part of this integral.
31
2.2.2 Hadamard finite part
The concept of the ”finite part” of a singular integral introduced by Hadamard based on dropping
some divergent terms and keeping the finite part.
Let a function f = f (x) be integrable in an interval ε < x < A for any 0 < ε, ε < A < ∞ and the
representation
Z N
X 1
f (x)dx = ak ε−λk + h ln + Jε (70)
k=1
ε
ε<x<A
hold valid, where ak , h, λk are some constant positive numbers independent of A. If the limit lim Jε
ε→0
exists, then it is called the Hadamard finite part of the singular integral of the function f . The function
f = f (x) is said to possess the Hadamard property at the origin. The standard notation for the finite
part of the Hadamard singular integral is as follows
Z
f.p. f (x)dx = lim Jε . (71)
ε→0
x<A
In the case h = 0 in the representation (70), the function f = f (x) is said to possess the non-logarithmic-
type Hadamard property at the origin.
For example, let consider the integral
Z∞
x−1 f (x)dx,
0
where f (x) is an analytic function on the half infinite interval [0, ∞) such that f (0) 6= 0 and f (x) =
O(x−δ ), x → ∞ with δ > 0. It is easy to see, that this integral is divergent. But if we consider the
integral
Z∞
x−1 f (x)dx, ε > 0,
ε
then by integrating by part we get
Z∞ Z∞
x f (x)dx = −f (ε) log(ε) − log(x)f ′ (x)dx.
−1
ε ε
Then, the limit ∞ 

Z
lim  x−1 f (x)dx + f (ε) log(ε)
ε→0
ε
R∞
exists and is finite. This limit is the Hadamard finite-part of x−1 f (x)dx and
0
∞ 
Z∞ Z
f.p. x−1 f (x)dx = lim  x−1 f (x)dx + f (ε) log(ε) .
ε→0
0 ε
32
In general, for an analytic function f (x) such that f (0) 6= 0 and f (x) = O(xn−δ−1 ), x → ∞ with δ > 0
R∞
the f.p. of the integral x−n f (x)dx is defined by
0
∞ 
Z∞ Z n−2 k+1−n
X ε log(ε) (n−1) 
f.p. −n
x f (x)dx = lim  −n
x f (x)dx − f (k) (0) + f (0) .
ε→0
k=0
k!(n − 1 − k)! (n − 1)!
0 ε
2.2.3 Hadamard finite part for basic power-logarithmic examples

Rz tλ dt
Let consider how to take Hadamard finite part of (z−t)α+1
. For the fourth case in (67) we have
0
Re(α) < 0 so we do not have a singular point at t = z. When Re(λ) > −1 we also do not have a
singular point at t = 0 and can just to calculate integral, so
Zz
tλ dt Γ(−α)Γ(λ + 1) λ−α
α+1
= z for Re(λ) > −1.
(z − t) Γ(λ + 1 − α)
0
If −2 < Re(λ) < −1 we can consider

Zz Zz
tλ dt tλ 1 1 Γ(1 − α)Γ(λ + 1) λ−α
f.p. = − dt + z .
(z − t)α+1 (z − t)α z−t z Γ(λ + 2 − α)
0 0
Integral
Zz Zz
tλ 1 1 1 tλ+1 dt Γ(−α)Γ(λ + 2) λ−α
− dt = = z
(z − t)α z−t z z (z − t) α+1 Γ(λ + 2 − α)
0 0
converges for Re(λ) > −2, then

Zz
tλ dt Γ(−α)Γ(λ + 2) λ−α Γ(1 − α)Γ(λ + 1) λ−α
f.p. α+1
= z + z =
(z − t) Γ(λ + 2 − α) Γ(λ + 2 − α)
0
z λ−α
= [Γ(−α)Γ(λ + 2) + Γ(1 − α)Γ(λ + 1)] =
Γ(λ + 2 − α)
z λ−α
= [Γ(−α)(λ + 1)Γ(λ + 1) + (−α)Γ(−α)Γ(λ + 1)] =
Γ(λ + 2 − α)
(λ + 1 − α)Γ(−α)Γ(λ + 1) λ−α Γ(−α)Γ(λ + 1) λ−α
= z = z .
Γ(λ + 2 − α) Γ(λ + 1 − α)
For −n − 1 < Re(λ) < −n, λ 6= −1, −2, ..., −n we will use the next regularisation
Zz Zz n
! n
tλ dt tλ 1 1 X tk−1 X Γ(1 − α)Γ(λ + k) λ−α
f.p. α+1
= α
− k−1
dt + z .
(z − t) (z − t) z − t z k=1 z k=1
Γ(λ + k + 1 − α)
0 0
33
Let consider the integral
Zz n
! Zz
tλ 1 1 X tk−1 tλ 1 1 t tn+1
− dt = − − 2 − .. − n dt =
(z − t)α z − t z k=1 z k−1 (z − t)α z−t z z z
0 0
Zz
tλ z n − z n−1 (z − t) − z n−2 t(z − t) − ... − tn−1 (z − t)
= dt =
(z − t)α z n (z − t)
0
Zz Zz
tλ z n − z n + z n−1 t − z n−1 t + z n−2 t2 − ... − ztn−1 + tn 1 tλ+n dt
= dt = n =
(z − t)α z n (z − t) z (z − t)α+1
0 0
λ−α
Γ(−α)Γ(n + λ + 1)z
= .
Γ(n − α + λ + 1)
It converges for Re(λ) > −n − 1. Then
Zz n
tλ dt Γ(−α)Γ(n + λ + 1) λ−α X Γ(1 − α)Γ(λ + k) λ−α
f.p. = z + z =
(z − t)α+1 Γ(n − α + λ + 1) k=1
Γ(λ + k + 1 − α)
0
n
!
Γ(n + λ + 1) X Γ(λ + k)
= −α Γ(−α)z λ−α .
Γ(n − α + λ + 1) k=1
Γ(λ + k + 1 − α)
Calculating sum we get
n
X Γ(λ + k) 1 Γ(n + λ + 1) Γ(λ + 1)
= − .
k=1
Γ(λ + k + 1 − α) α Γ(n − α + λ + 1) Γ(−α + λ + 1)
Therefore for Re(λ) > −n − 1, λ 6= −1, −2, ..., −n

Zz
tλ dt Γ(−α)Γ(λ + 1) λ−α
f.p. α+1
= z .
(z − t) Γ(λ + 1 − α)
0
So for all α and λ such that λ 6= −1, −2, ..., −n we have
dα z λ Γ(λ + 1)
α
= z λ−α . (72)
dz Γ(λ + 1 − α)
This approach is the same as in [32], p. 69 was used.

Now let consider the case when λ ∈ Z and λ < 0 i.e. λ = −1, −2, ..., −n. Here we have two variants.
The first one it is the case when α ∈ Z and λ < α. When α ∈ Z and α < 0 we have α = −1, −2, ... and
Zz
d−1 z λ
= f.p. tλ dt.
dz −1
0
34
Since for ε > 0
Zz
z λ+1 ελ+1
tλ dt = −
λ+1 λ+1
ε
then
Zz
d−1 z λ λ z λ+1
= f.p. t dt = for λ = −2, −3, ...
dz −1 λ+1
0
Similarly, for α = −2, −3, ... and λ < α, λ ∈ Z

Zz Zt Zt
dα z λ z λ−α
= f.p. dt... dt tλ dt = .
dz α (λ + 1)(λ + 2)...(λ − α)
|0 {z0 0
}
−α
Let α = −n, λ = −m. Since m > n we can write
1 1 (−1)n (−1 − n − λ)! λ!

= = = =
(λ + 1)(λ + 2)...(λ − α) (λ + 1)(λ + 2)...(λ + n) (−1 − λ)! (λ − α)!
= (−1)α (−λ)α .
So for all α and λ such that α ∈ Z, λ ∈ Z, λ < 0 and λ < α taking into account (66) we have
dα z λ
= (−1)α (−λ)α z λ−α . (73)
dz α
In the case λ ∈ Z, λ < 0 and λ ≥ α the integration of z λ can produce log(z) (for negative integer λ
when λ ≥ α). For example, when λ = −1 and α = −1
Zz
dt
= log(z) − log(ε)
t
ε
so by (70) and (71) we get

Zz
d−1 z −1 dt
= f.p. = log(z).
dz −1 t
0
The same regularisation we can obtain if we consider the Laurent series expansion by λ at λ = −1:
Zz
d−1 z λ z λ+1 1
= tλ dt = = 1 + (λ + 1) log(z) + O (λ + 1)2 =
dz −1 λ+1 λ+1
0
1
= + log(z) + O (λ + 1) .
λ+1
35
1
Then throwing away the main part λ+1
we get
Zz
d−1 z −1 dt
= f.p. = lim [log(z) + O (λ + 1)] = log(z).
dz −1 t λ→−1
0
In general case when λ = −n, n ∈ Z, α ≤ −n the Laurent series expansion of (72) by λ at λ = −n is
Γ(λ + 1)
z λ−α =
Γ(λ + 1 − α)
(−1)λ−1 z λ−α (−1)λ−1 (ψ(−λ) − ψ(λ − α + 1) + log(z)) λ−α

= + z + O (λ + n) .
(−1 − λ)! Γ(1 + λ − α)(λ + n) (−λ − 1)!Γ(λ − α + 1)
Γ′ (z)
Here ψ is the digamma function, given by ψ(z) = Γ(z)
. Then throwing away the main part of the
Laurent series expansion we obtain
Zz
dα z −n tλ dt
= f.p. =
dz α (z − t)α+1
0

(−1)λ−1 (ψ(−λ) − ψ(λ − α + 1) + log(z)) λ−α
= lim z + O (λ + n) =
λ→−n (−λ − 1)!Γ(λ − α + 1)
(−1)−n−1 (ψ(n) − ψ(1 − n − α) + log(z)) −n−α
= z .
(n − 1)!Γ(1 − n − α)
So we put
dα z λ (−1)λ−1 (ψ(−λ) − ψ(λ − α + 1) + log(z)) λ−α

= z (74)
dz α (−λ − 1)!Γ(λ − α + 1)
under certain restrictions.

Finally, combining formulas (72), (73) and (74) we get

 (−1)α (−λ)α , α ∈ Z, λ ∈ Z, λ < 0, λ < α;
dα z λ λ−α

(−1)λ−1 (log(z)+ψ(−λ)−ψ(1−α+λ))
=z (−λ−1)!Γ(1−α+λ)
, λ ∈ Z, λ < 0; (75)
dz α 
 Γ(λ+1)
Γ(λ+1−α)
, in other cases.
Rz tλ dt
The second integral (z−t)α−n+1
in (69) is calculated in same way since Re(α − n) < 0.
0
We also need to calculate other integrals in (69). Calculation and taking the Hadamard finite part
Rz tλ logk (t)dt Rz tλ logk (t)dt Rz tλ dt
of (z−t) α+1 and (z−t) α−n+1 based on the same ideas as calculation of (z−t)α+1
but much more
0 0 0
complicated. For example, when Re(α) < 0, Re(z) > 0 and Re(λ) > −1 we have
dα λ Γ(λ + 1) (Hλ − Hλ−α + log(z))

α
z log(z) = z λ−α , (76)
dz Γ(λ − α + 1)
36
n
P 1 Γ′ (z)
where Hs = G + ψ(s + 1) is the harmonic number (Hn = k
for integer n), ψ(z) = Γ(z)
is the digamma
k=1
function, G is Euler–Mascheroni constant. However, in general
dα
α
z λ log(z) =
dz


(−1)α (−λ)α (ψ(−λ) − ψ(α − λ) + log(z)); α, λ ∈ Z, λ < 0, λ < α,
 λ+1
2

 (−1) π


 log2 (z) + − ψ (1) (λ − α + 1) − ψ (1) (−λ)+
2Γ(−λ)Γ(λ − α + 1) 3
= z λ−α λ ∈ Z, λ < 0,



 +(ψ(−λ) − ψ(λ − α + 1)) (ψ(−λ) − Hλ−α + 2 log(z) + λ) ;



 Γ(λ+1)(Hλ −Hλ−α+log(z))
Γ(λ−α+1)
; in other cases,
(77)
dm+1
Here ψ (m) (z) = dz m+1 ln Γ(z) is the polygamma function of order m.
Expanding basic functions in Mathematica to sums of the forms (60) or (64) allowed us to build
fractional derivatives of order α for more than 100000 test cases, involving basic functions and their
compositions.
2.3 The Meijer G-function and fractional calculus

2.3.1 Definitions and main properties
All known special functions can be divided on several large groups:
• basic special functions,
• hypergeometric type functions,
• Riemann zeta and related functions,
• elliptic and related functions,
• number theoretic functions,
• generalized functions and other standard special functions.
The majority of the most important special functions are the analytical functions, which by defini-
tions typically can be presented through infinite series:
∞
X
(z − z0 ) α logk (z − z0 ) cn (z − z0 )n , (78)
n=0
where k ∈ Z, k ≥ 0, cj , α ∈ C, j = 0, 1, 2.... Such series can be analytically continued (re-expanded)

or represented through integrals. Very often they are solutions of corresponding differential equations,
which came from applications and define special functions.
37
Hypergeometric type functions can be defined as the functions, which generically can be represented
through linear combinations of Meijer G–function which is a very general special function of the form
[131]
m
Q n
Q
Z Γ(bk − s) Γ(1 − ak + s)
m,n
a1 , ..., an , an+1 , ..., ap 1 k=1 k=1
Gp,q z
= p q z s ds, (79)
b1 , ..., bm , bm+1 , ..., bq 2πi Q Q
L Γ(ak − s) Γ(1 − bk + s)
k=n+1 k=m+1
where contour L represents the path to be followed while integrating. This integral is of the so-called
Mellin–Barnes type, and may be viewed as a generalization of inverse Mellin transform. There are three
different paths L of integration:
• L runs from −i∞ to +i∞ so that all poles of Γ(bi − s), i = 1, ..., m, are to the right, and all the
poles of Γ(1 − ak + s), k = 1, ..., n, to the left, of L. This contour can be a vertical straight line
(γ − i∞, γ + i∞) if Re(bi − ak ) > −1 (then Re(a k ) − 1 < γ < p+q
Re(bi )). The integral converges if
p + q < 2(m + n) and |arg z| < m + n − p+q 2
π. If m + n − 2
= 0, then z must be real and
Pq Pp
positive and additional condition (q − p)γ + Re(µ) < 0, µ = bi − ak + p−q
2
+ 1, should be
i=1 k=1
added.
• L is a loop starting and ending at +∞ and encircling all poles of Γ(bi − s), i = 1, ..., m, once in
the negative direction, but none of the poles of Γ(1 − ak + s), k = 1, ..., n. The integral converges
if q ≥ 1 and either p < q or p = q and |z| < 1 or q = p and |z| = 1 and m + n − p+q 2
≥ 0 and
Re(µ) < 0.
• L is a loop starting and ending at −∞ and encircling all poles of Γ(1 − ak + s), k = 1, ..., n, once
in the positive direction, but none of the poles of Γ(bi − s), i = 1, ..., m. The integral converges
if p ≥ 1 and either p > q or p = q and |z| > 1 or q = p and |z| = 1 and m + n − p+q 2
≥ 0 and
Re(µ) < 0.
Above definition of the Meijer G-function holds under the following assumptions:
• 0 ≤ m ≤ q and 0 ≤ n ≤ p, where m, n, p and q are integer numbers,
• ak −bj 6= 1, 2, 3, ... for k = 1, 2, ..., n and j = 1, 2, ..., m, which implies that no pole of any Γ(bj + s),
j = 1, 2, ..., m, coincides with any pole of any Γ(1 − ak − s), k = 1, 2, ..., n.
• z 6= 0.
Remark 1. A different from (79) but equivalent to it form was used in the books [3, 63, 86]:
m
Q n
Q
Z Γ(bk + s) Γ(1 − ak − s)
m,n
a1 , ..., an , an+1 , ..., ap 1 k=1 k=1
Gp,q z = p q z −s ds (80)
b1 , ..., bm , bm+1 , ..., bq 2πi Q Q
L Γ(ak + s) Γ(1 − bk − s)
k=n+1 k=m+1
where
m ∈ N, n ∈ Z, p ∈ N, q ∈ Z, m ≤ q, n ≤ p.
Both forms can be transformed each to other by changing variable of integration s → −s.
38
In the system Mathematica for standard form of the Meijer G-function the following notation is
used
MeijerG[{{a1 , ..., an }, {an+1 , ..., ap }}], {{b1 , ..., bm }, {bm+1 , ..., bq }}, z] =
 n m 
Q Q
Γ (1 − a k − s) Γ (b k + s)
1  k=1 k=1 −s

= ContourIntegrate  p q z , {s, L} ;
2πi  Q Q 
Γ (ak + s) Γ (1 − bk − s)
k=n+1 k=m+1
m ∈ Z ∧ m ≥ 0 ∧ n ∈ Z ∧ n ≥ 0 ∧ p ∈ Z ∧ p ≥ 0 ∧ q ∈ Z ∧ q ≥ 0 ∧ m ≤ q ∧ n ≤ p.
In many classical special functions (like Bessel functions) instead z the construction cz 1/r can be
used and under integrand factor z −s becomes (cz 1/r )−s which is not equal powerexpanded expression
c−s z −s/r . Correct formulas for these transformation one can find at Wolfram Function Site, for example
in [126].
Described properties are supported in Mathematica by command PowerExpand with or without
assumption option:
{PowerExpand[((z a )b ), Assumptions → True], PowerExpand[(z a )b ]}
n 1 Im(a log(z))
o
e2iπb⌊ 2 − 2π ⌋ z ab , z ab
This situation stimulated us to define generalized form of the Meijer G-function with additional real
parameter r, which allowed effectively without restrictions work in the full complex plane with functions
like Ja (z).
Remark 2. The generalized form of the Meijer G-function with additional real parameter r is defined
in [125] by similar to (79) integral

m,n
a1 , ..., an , an+1 , ..., ap
Gp,q z, r =
b1 , ..., bm , bm+1 , ..., bq
m
Q n
Q
Z Γ(bk + s) Γ(1 − ak − s)
r k=1 k=1 s
= p q z − r ds, (81)
2πi Q Q
L Γ(ak + s) Γ(1 − bk − s)
k=n+1 k=m+1
where r ∈ R, r 6= 0, m ∈ Z, m ≥ 0, n ∈ Z, n ≥ 0, p ∈ Z, p ≥ 0, q ∈ Z, q ≥ 0, m ≤ q, n ≤ p.
Evidently that for default case r = 1 we have equality

m,n
a1 , ..., an , an+1 , ..., ap m,n
a1 , ..., an , an+1 , ..., ap
Gp,q z, 1 = Gp,q z .
b1 , ..., bm , bm+1 , ..., bq b1 , ..., bm , bm+1 , ..., bq
This Meijer G-function with parameter r satisfies two important properties [127, 128]:

m,n 1 1 − b1 , ..., 1 − bm , 1 − bm+1 , ..., 1 − bq m,n
a1 , ..., an , an+1 , ..., ap
Gp,q ,r =Gp,q z, r ,z∈ / R, (82)
z 1 − a1 , ..., 1 − an , 1 − an+1 , ..., 1 − ap b1 , ..., bm , bm+1 , ..., bq

m,n
α + a1 , ..., α + an , α + an+1 , ..., α + ap α/r m,n
a1 , ..., an , an+1 , ..., ap
Gp,q z, r =z Gp,q z , (83)
α + b1 , ..., α + bm , α + bm+1 , ..., α + bq b1 , ..., bm , bm+1 , ..., bq
39
2.3.2 Representability through Meijer G-functions
Using classical and generalized Meijer G-functions we can represent Bessel function Jν (z) by formulas
2
1,0 z − π π
Jν (z) = G0,2 ν ν , − < arg(z) ≤ , (84)
4 2, −2
2 2

z 1 −
Jν (z) = G1,0 , ν . (85)
0,2
2 2 2
, − ν2
The first formula includes parameter r = 1/2 but it is valid for all complex z-plane. The second formula
including classical G-function is correct only for half plane. For finding last representation through
Meijer G–function we can use commands ResourceFunction["MeijerGForm"] and MeijerGReduce:
ResourceFunction["MeijerGForm"][BesselJ[ν, z], z], MeijerGReduce[BesselJ[ν, z], z].
For best understanding Meijer G-function we can suggest to use several internal commands
MeijerGInfo, MeijerGToSums, SlaterForm, which operate with context System‘MeijerGDump‘, for
example,
System‘MeijerGDump‘SlaterForm[MeijerG[a, b, c, d, z, 2/3], s]
gives
2z −s Γ 1 − a − 2s
3
Γ c + 2s3
.
3Γ b + 2s 3
Γ 1 − d − 2s
3
Through Meijer G-function we can represent more than 100 named well known functions such
as log, exp, arctan, Bessel, Airy and Legendre functions. In Mathematica TraditionalForm
notations these functions form the following list with 127 functions to which we can apply
ResourceFunction["MeijerGForm"]. See below List MeijerG127 in TraditionalForm
√
MeijerG127={az , ez , z, z b , Ai (z), Ai′ (z), Bi(z), Bi′ (z), J a (z), J ba (z), arccos (z), arcosh (z),
arccot (z), arcoth (z), arccsc (z), arcsch (z), arcsec (z), arsech (z), arcsin (z), arsinh (z), arctan (z),
arctan (a, z), arctan (z, a), artanh (z), Ia (z), Ja (z), Ka (z), Ya (z), Bz (a, b), B(c,z) (a, b), B(z,c) (a, b),
Iz (a, b), I(c,z) (a, b), I(z,c) (a, b), RC (x, z), RC (z, y), RE (x, z), RE (y, z), RK (x, z), RK (y, z), Ta (z), Ua (z),
cos(z), cosh(z), Chi(z), Ci(z), F (z), E(z), K(z), erf(z), erf(a, z), erf(z, b), erfc(z), erfi(z), Ea (z), Ei(z),
(b)
Fz , Fa (z), C(z), F (z), G(z), S(z), Γ(a, z), Γ(a, b, z), Γ(a, z, b), Q(a, z), Q(a, b, z), Q(a, z, b), Ca (z),
(1) (2)
Ha (z), Ha (z), hav(z), Ha (z), 0 F1 (; a; z), 0 F̃1 (; a; z), 1 F1 (a; b; z), 1 F̃1 (a; b; z), 2 F1 (a, b; c; z),
−1
2 F̃1 (a, b; c; z), p Fq (a1 , ..., ap ; b1 , ..., bq ; z), p F̃ Fq (a1 , ..., ap ; b1 , ..., bq ; z), U(a, b, z), hav (z), bei0 (z),
beia (z), ber0 (z), bera (z), kei0 (z), keia (z), ker0 (z), kera (z), La (z), Lba (z), Pν (z), Pab (z), Qν (z), Qba (z),
log(z), Lz , La (z), Da (z), Li2 (z), Lia (z), Gi(z), Gi′ (z), Hi(z), Hi′ (z), sin(z), sinc(z), sinh(z), Shi(z),
(1) (2)
Si(z), ja (z), ya (z), ha (z), ha (z), H ν (z), L ν (z), θ(z), E ν (z), E aν (z), Ma,b (z), Wa,b (z) }
Different combinations, including these functions also sometimes can be presented through Meijer
G-function. Long time work in this direction allowed to build basic collection of such functions with
about 3000 cases and more full test set with 90000 cases.
40
2.3.3 Fractional integro-differentiation of Meijer G–functions
For all above functions one can find different representations of the fractional integro-derivatives
searching Wolfram Functions Site [130]. In particular, definition formula (59) for fractional integro-
derivative of the order α of the Meijer G-function can be converted to the following representation (see
also [131]):
dα m,n a1 , ..., an , an+1 , ..., ap
G z =
dz α p,q b1 , ..., bm , bm+1 , ..., bq

m,n+1
−α, a1 − α, ..., an − α, an+1 − α, ..., ap − α
= Gp+1,q+1 z . (86)
b1 − α, ..., bm − α, 0, bm+1 − α, ..., bq − α
The right side of this formula through Fox H-function we can see if use the following input in
Mathematica:
ResourceFunction[”FractionalOrderD”][ MeijerG[{Table[Subscript[a,i], {i, 1, n}],
Table[Subscript[a,i], {i, n + 1, p}]}, {Table[Subscript[b,i], {i, 1, m}], Table[Subscript[b,i], {i, m + 1,
q}]}, z], {z, α}]
Below we demonstrate more general formula for generalized Meijer G–function with argument a z r
and parameter r1 :
dα m,n r
a1 , ..., an , an+1 , ..., ap
G az , r1 =
dz α p,q b1 , ..., bm , bm+1 , ..., bq
r1
r1

r {−α, r}, a − α, r , ..., a − α, r
−α r α/r m,n+1
1 1 p 1
= r1 z (az ) Hp+1,q+1 az r r . (87)
b1 − rr1 α, r1 , ..., bq − rr1 α, r1 , {0, r}
In general (non logarithmic) cases the Meijer G-function can be represented through one or finite
combination of the series (78) with z0 = 0 and k = 0.
Assume p ≤ q, no two of the bottom parameters bj , j = 1, ..., m, differ by an integer, and (aj − bk )
is not a positive integer when j = 1, 2, ..., n and k = 1, 2, ..., m. Then double sums of ”left” residues for
Gm,n
p,q can be written as finite sum of generalized hypergeometric functions (see [129]):

m,n
a1 , ..., an , an+1 , ..., ap
Gp,q z =
b1 , ..., bm , bm+1 , ..., bq
m
X 1 + bk − a1 , ..., 1 + bk − ap
= Am,n
p,q,k z bk
p Fq−1 ; (−1) p−m−n
z , (88)
1 + bk − b1 , ... ∗ ....1 + bk − bq
k=1
where ” ∗ ” indicates that the entry (1 + bk − bk ) is omitted. Also,

m
Q n
Q
Γ(bj − bk ) Γ(1 + bk − aj )
j=1,j6=k j=1
Am,n
p,q,k = p q .
Q Q
Γ(aj − bk ) Γ(1 + bk − bj )
j=n+1 j=m+1
The Function Site [130] includes the most complete collection of formulas for asymptotics of Meijer
G-functions. In more complicated logarithmic cases, the Meijer G-function can be represented through
41
a finite combination of the above series with z0 = 0 and n > 0. It means that we can evaluate Meijer
G-function, as infinite sums including terms z b logk (z) with already uniquely defined basic elementary
functions: power and logarithmic.
So if we have a hypergeometric type function f (z) we first can write f (z) as z b g(z) or z b logk (z)g(z)
where g(x) is supposed to be simpler than f (z). Using the series expansion of the function g(z) =
P∞
cn z n and formula (88) we can rewrite f (z) as a single G-function. In more complex cases, the
n=0
function f (z) can be written as a finite sum of G-functions. Next applying the formula (86) we can
find fractional integral or derivative of f (z) in the form of the Meijer G-function. Then we can write
the G-function as a simpler function if possible.
2.3.4 Fractional integro-differentiation of ez and Kν (z)

In Wolfram Mathematica there is a function MeijerGForm[g(z),z] which reduces g(z) to the Meijer
G–function as a function of z. In order to use function MeijerGForm we should write
ResourceFunction["MeijerGForm"].
For example, ResourceFunction["MeijerGForm"][ez , z] gives

z 1,0
− 1,0
−
e = MeijerG({{}, {}}, {{0}, {}}, −z, 1) = G0,1 −z, 1
= G0,1 −z

0 0
and ResourceFunction["MeijerGForm"][ BesselK[ν, z], z] gives

nn ν ν o o z 1
1 2,0 z 1 −
Kν (z) = MeijerG {{}, {}}, ,− , {} , , = G0,2 , ν ,
2 2 2 2 2 2 2 2
, − ν2
where formula (81) was used.

dα
Let us compare the approach of finding the fractional operator dz α through series expansion and
through the G-function representation. Direct calculation through the G-function representation gives
(
dα ez ez , α ∈ Z and α ≥ 0;
α
= z
(89)
dz e (1 − Q(−α, z)), in other cases,
Γ(a,z) R∞
where Q(a, z) = Γ(a)
is the regularized incomplete gamma function, Γ(a, z) = ta−1 e−t dt is the
z
incomplete gamma function. The exponential function ez has Taylor series at z = 0 of the form
∞
X zn
ez = .
n=0
n!
Since n ∈ N ∪ {0} using the third line in (75) we obtain

∞ ∞ ∞
dα ez X 1 dα z n X 1 Γ(n + 1)z n−α X z n−α z αΓ(−α, z)
= = = =e 1+ .
dz α n=0
n! dz α
n=0
n! Γ(n − α + 1) n=0
Γ(n − α + 1) Γ(1 − α)
42
αΓ(−α,z)
For α = 0, 1, 2, ... we have Γ(1−α)
= 0 and
αΓ(−α, z) αΓ(−α, z) Γ(−α, z)

= =− = −Q(a, z),
Γ(1 − α) −αΓ(−α) Γ(−α)
therefore, the result coincides with the formula (89).

For K0 (z) we have

dα 1 2,2 z 1 1−α 2
, − α2
K0 (z) = G2,4 , . (90)
dz α 2 2 2 − α2 , − α2 , 0, 12
From the other side by (62)

∞ ∞
!
α α z 2n
d d X 1 X Hn z 2n
α
K0 (z) = α − (log(z) − log(2) + G) + ,
dz dz n=0
(n!)2 2 n=1
(n!)2 2
where G is Euler–Mascheroni constant (63). Applying (75) and (76) we obtain

∞ ∞ z

dα 1 X 21−2k (2k)!ψ(k + 1) 2k−α X (2k)! H2k − H2k−α + log
K0 (z) = z − 2
z 2k−α . (91)
dz α 2 k=0 (k!)2 Γ(2k − α + 1) k=0
22k (k!)2 Γ(2k − α + 1)
Using Mathematica one can check that (90) equals to (91).
2.3.5 Riemann-Liouville integral of generalized Meijer G–function

Let now evaluate the following ”main” Riemann-Liouville integral
Zz
1 β−1 α−1 m,n g
a1 , ..., an , an+1 , ..., ap
(z − τ ) τ Gp,q wτ , r
dτ (92)
Γ(β) b1 , ..., bm , bm+1 , ..., bq
0
which defines fractional order β integral of generalized Meijer G–function (81), multiplied by τ α−1 , with
r ∈ R, r 6= 0, m ∈ Z, m ≥ 0, n ∈ Z, n ≥ 0, p ∈ Z, p ≥ 0, q ∈ Z, q ≥ 0, m ≤ q, n ≤ p. In formula (81)
for simplicity we use vertical contour L = {γ − i∞, γ + i∞}.
The integral (92) with particular parameters g = kℓ ∈ Q and r = 1 has value, shown at the [3], p.
535, formula 3.36.2.1. Below we derive representations of this integral through Fox H-function or Meijer
G-functions and in details demonstrate how to build corresponding sets of conditions for convergence
of this integral.
After substitution of above definition of Meijer G-function to Riemann-Liouville integral and chang-
ing order integration by s and by τ we come to the following chain operations where we met and
evaluated internal integral
Zz
1 β−1 α−1
a1 , ..., an , an+1 , ..., ap
(z − τ ) τ Gm,n
p,q
g
wτ , r
dτ =
Γ(β) b1 , ..., bm , bm+1 , ..., bq
0
43
 m
Q n
Q 
Zz γ+i∞
Z Γ(bk + s) Γ(1 − ak − s)
1 β−1 α−1
 k=1 k=1 s 
= (z − τ ) τ 
p q (wτ g )− r ds
 dτ =
2πiΓ(β)  Q Q
0 γ−i∞ Γ(ak + s) Γ(1 − bk − s)
k=n+1 k=m+1
m
Q n
Q  z 
γ+i∞
Z Γ(bk + s) Γ(1 − ak − s) Z
1 s k=1 k=1  (z − τ )β−1 τ α−g rs −1 dτ  ds =
= w− r p q
2πiΓ(β) Q Q
γ−i∞ Γ(ak + s) Γ(1 − bk − s) 0
k=n+1 k=m+1
m
Q n
Q
γ+i∞
Z Γ (bk + s) Γ (1 − ak − s) Γ α − gr s
z α+β−1 k=1 k=1
1 g −s
= p q w z
r r ds. (93)
2πi Q Q g
γ−i∞ Γ (ak + s) Γ (1 − bk − s) Γ α + β − r
s
k=n+1 k=m+1
The last Mellin-Barnes integral (93) can be written through corresponding Fox H-function by definition
of Fox H-function. As result we have the following representation of Riemann-Liouville fractional order
β integral of generalized Meijer G-function through Fox H-function
Zz
1 β−1 α−1 m,n g
a1 , ..., an , an+1 , ..., ap
dτ =
Γ(β) b1 , ..., bm , bm+1 , ..., bq
0

α+β−1 m,n
1 − α, g , (a1 , 1), ..., (an , 1), (an+1, 1), ..., (ap , 1)
1 g
=z Hp,qw z r r r . (94)
(b1 , 1), ..., (bm , 1), (bm+1 , 1), ..., (bq , 1), 1 − α − β, gr
s
1 g −s
We should mention that operations like (wτ g )− r → w r z r with complex variables demands special
conditions for their correctness.
The formula (93) with products of gamma functions includes gammas with variable of integration s
which has coefficients +1 or −1 or − gr . Definition of Meijer G-function does not include coefficients not
equal to +1 or −1 but after some special transformations above Mellin-Barnes integral can be re-written
as Meijer G-function.
Let consider the Mellin-Barnes integral from the formula (93), which we denote as MB:
m
Q n
Q
γ+i∞
Z Γ (bk + s) Γ (1 − ak − s)
z α+β−1 Γ α − gr s 1 g
−s
MB = g
· k=1
p · k=1
q wrzr ds.
2πi Γ α+β− r s Q Q
γ−i∞ Γ (ak + s) Γ (1 − bk − s)
k=n+1 k=m+1
Suppose that gr is the rational value with positive integers g and r (that do not have common factors).
Making change of variable s = rζ we obtain
m
Q n
Q
δ+i∞
Z Γ (bk + rζ) Γ (1 − ak − rζ)
z α+β−1 Γ (α − gζ) 1 g
−rζ γ
MB = · k=1
p · k=1
q wr zr rdζ, δ= .
2πi Γ (α + β − gζ) Q Q r
δ−i∞ Γ (ak + rζ) Γ (1 − bk − rζ)
k=n+1 k=m+1
44
Since r, g ∈ N for each gamma function in the previous Mellin-Barnes integral, we can apply known
Gauss multiplication formulas of the forms (see [132])
r−1
1−r
b+ζr− 21
Y b+j
Γ(b + ζr) = (2π) 2 r Γ +ζ , r ∈ N, (95)
j=0
r
g−1
1−g Y α + j
α−ζg− 21
Γ(α − ζg) = (2π) 2 g Γ −ζ , g∈N (96)
j=0
g
for each gamma function. We get

m
Q n
Q
Γ (bk + rζ) Γ (1 − ak − rζ) 1 g −rζ
Γ(α − gζ)
A(α, β, r) = · k=1
p · k=1
q · wrzr r=
Γ(α + β − gζ) Q Q
Γ (ak + rζ) Γ (1 − bk − rζ)
k=n+1 k=m+1
g−1 n
r−1
Q α+j Q 1−r
−ak −rζ+ 21
Q j−ak +1

1−g
α−gζ− 21
Γ g
−ζ (2π) 2 r Γ r
−ζ
(2π) 2 g j=0 k=1 j=0
= 1−g · α+β−gζ− 21
· g−1 · Q
q
r−1 ·
(2π) 2 g Q α+β+j 1−r
−bk −rζ+ 1
Q j−bk +1

Γ g
−ζ (2π) 2 r 2 Γ r
−ζ
j=0 k=m+1 j=0
m
r−1
Q 1−r
bk +rζ− 21
Q bk +j

(2π) 2 r Γ r
+ζ
k=1 j=0
· p
r−1 (wz g )−ζ r.
Q 1−r
+rζ− 1
Q ak +j

(2π) 2 r ak 2 Γ r
+ζ
k=n+1 j=0
Simplifying we obtain
g−1
Q n
Q r−1
Q
α+j 1−r 1 1−ak +j
−β
g Γ g
−ζ (2π) 2 r −ak −rζ+ 2 Γ r
−ζ
j=0 k=1 j=0
A(α, β, r) = g−1 · q r−1
·
Q α+β+j
Q 1−r
−bk −rζ+ 21
Q 1−bk +j

Γ g
−ζ (2π) 2 r Γ r
−ζ
j=0 k=m+1 j=0
m
Q 1−r 1
r−1
Q
bk +j
(2π) 2 r bk +rζ− 2 Γ r
+ζ
k=1 j=0
· p r−1
(wz g )−ζ r =
Q 1−r 1 Q ak +j

(2π) 2 r ak +rζ− 2 Γ r
+ζ
k=n+1 j=0
n m g−1
Q
Q 1−r
−ak −rζ+ 21
Q 1−r
bk +rζ− 21 α+j
(2π) 2 r (2π) 2 r Γ g
−ζ
k=1 k=1 j=0
= q · p · g−1 ·
Q 1−r
−bk −rζ+ 21
Q 1−r
ak +rζ− 21 Q α+β+j
(2π) 2 r (2π) 2 r Γ g
−ζ
k=m+1 k=n+1 j=0
45
n r−1
Q Q m r−1
Q Q
1−ak +j bk +j
Γ r
−ζ Γ r
+ζ
k=1 j=0 k=1 j=0
· q r−1
· p r−1
g −β (wz g )−ζ r.
Q Q 1−bk +j
Q Q ak +j

Γ r
−ζ Γ r
+ζ
k=m+1 j=0 k=n+1 j=0
Collecting powers we get
n m n m g−1
Q
Q Q bk α+j
1−r 1
(2π) r2 2
−rζ 1−r
(2π) r
2
ζr− 21 r −ak
r Γ g
−ζ
k=1 k=1 j=0
A(α, β, r) = 1−r 1
q−m · 1−r

1 p−n
· q
Q
·
Qp · g−1 ·
(2π) 2 r ζr− 2
Q
(2π) 2 r 2 −rζ r −bk r ak Γ α+β+j
−ζ
g
k=m+1 k=n+1 j=0
n r−1
Q Q m r−1
Q Q
1−ak +j bk +j
Γ r
−ζ Γ r
+ζ
k=1 j=0 k=1 j=0
· q r−1
· p r−1
g −β (wz g )−ζ r =
Q Q 1−bk +j
Q Q ak +j

Γ r
−ζ Γ r
+ζ
k=m+1 j=0 k=n+1 j=0
g−1
Q n r−1
Q Q
α+j 1−ak +j
Γ g
−ζ Γ r
−ζ
j=0 k=1 j=0
= g−1 · q r−1
·
Q α+β+j
Q Q 1−bk +j

Γ g
−ζ Γ r
−ζ
j=0 k=m+1 j=0
m r−1
Q Q q p
bk +j
Γ +ζ ak + p−q
P P
bk − +1 −ζ
r 2
k=1 j=0 r k=1 k=1 wz g
· p
· · .
Q r−1
Q ak +j 2 )gβ
(2π)(r−1)(m+n− p+q r r(q−p)
Γ r
+ζ
k=n+1 j=0
1 g
−rζ
In above transformations we changed w z to (wz g )−ζ , which is allowed under some restrictions.
r r
But in result we came to products of gamma functions with coefficients +1 or −1 before variable of
integration ζ.
This result allows us to write the following representation of Riemann-Liouville fractional order β
integral of generalized Meijer G-function through Meijer G-function in the case when gr is the rational
value with positive integers g and r:
Zz
1 β−1 α−1
a1 , ..., an , an+1 , ..., ap
(z − τ ) τ Gm,n
p,q
g
wτ , r
dτ =
Γ(β) b1 , ..., bm , bm+1 , ..., bq
0
q p
ai + p−q
P P
bk − 2
+1
r k=1 i=1 wz g ∆1 , ∆2
=z α+β−1
Grm,rn+g
rp+g,rq+g , (97)
(2π)(r−1)(m+n−
p+q
2 )gβ r r(q−p) ∆3 , ∆4
where
1−α g − α a1 a1 + r − 1 an an + r − 1
∆1 = , ..., , , ..., , ..., , ..., ,
g g r r r r
46
an+1 an+1 + r − 1 ap ap + r − 1
∆2 = , ..., , ..., , ..., ,
r r r r
b1 b1 + r − 1 bm bm + r − 1
∆3 = , ..., , ..., , ..., ,
r r r r
bm+1 bm+1 + r − 1 bq bq + r − 1 1 − α − β g−α−β
∆4 = , ..., , ..., , ..., , , ..., .
r r r r g g
2.3.6 Big O representations for asymptotics of Meijer G-functions

If we use ”Big O notation” [124] for description of asymptotic behaviour of Meijer G-function, we
get the following generic formulas, describing Big O representation of Meijer G-function near its two or
three singular points (see Wolfram Functions Site [126] and [3], p. 571):
 Pm


 z bk if p ≤ q;

 k=1

 m (−1)q−m−n
 P
 z bk
+ z χ
e z if p = q + 1;
a1 , ..., an , an+1 , ..., ap 
m,n k=1
Gp,q z ↔ m
q (98)
b1 , ..., bm , bm+1 , ..., bq  P bk χ (−1)q−m−n−1


 z + z cos 2 z
if p = q + 2;

 k=1

 m
P b 1
q−p

 z k + z χ e(p−q)(−z) if p ≥ q + 3,
k=1
in (98) |z| → 0,
q p
!
1 X X p−q+1
χ= bj − aj + . (99)
q−p j=1 j=1
2
When z → (−1)m+n−p we get

 ψ
 1 + (1 − (−1)p−m−n z) p if q = p and ψp 6= 0;
a , ..., an , an+1 , ..., ap
Gm,n
p,q z 1 ↔ 1 + log (1 − (−1)p−m−n z) if q = p and ψp = 0; (100)
b1 , ..., bm , bm+1 , ..., bq 
1 if q =
6 p,
where
p
X
ψp = (aj − bj ) − 1. (101)
j=1
Finally, for |z| → ∞

 Pn


 z ak −1 if q ≤ p;

 k=1

 Pn
p−m−n z
z ak −1 + z χ e(−1)


 if q = p + 1;
a1 , ..., an , an+1 , ..., ap
Gm,n
p,q z ↔ k=1
n (102)
b1 , ..., bm , bm+1 , ..., bq  P z ak −1 + z χ cos 2p(−1)p−m−n−1 z
 if q = p + 2;



 k=1

 Pn 1
q−p

 z ak −1 + z χ e(q−p)(−z) if q ≥ p + 3,
k=1
where χ is defined by (99).
47
2.3.7 Conditions of convergence for Riemann-Liouville integrals with Mei-
jer G-functions
Above relations include power functions like z bk or z χ or z ak −1 or 1 or log(1 − (−1)p−m−n z) which
match ”Big O terms” [124] and use here instead of O(z bk ) or O(z χ ) or O(z ak −1 ) or O(1) or O(log(1 −
(−1)p−m−n z)) inside of asymptotic expansions. It allows maximally simple establish conditions of
convergence integrals involving Meijer G-functions. For example, classical Riemann-Liouville integral
Zz
1 β−1 α−1 m,n g
a1 , ..., an , an+1 , ..., ap
dτ (103)
Γ(β) b1 , ..., bm , bm+1 , ..., bq
0
has Meijer G-function with parameter Gm,n g

p,q [wτ , r|...], which under some conditions can be written as
1 g
classical G-function without parameter r = 1 as Gm,n
p,q [w τ |...]. The interval of integration here is finite
r r
g
and we do not get conditions on convergence at infinity if gr > 0 (then τ r → ∞ for τ → ∞). But for
q < p and gr > 0 the point τ = 0 becomes essentially singular and this integral will converge at zero if
and only if the following ”Big O-equivalent” integrals will converge:
Zz m
X
τ α−1 z bk dτ, p ≤ q, (104)
0 k=1
Zz m
!
X (−1)q−m−n
τ α−1 z bk + τ α−1 z χ e z dτ, p = q + 1, (105)
0 k=1
Zz m
r !!
X (−1)q−m−n
τ α−1 z bk + τ α−1 z χ cos 2 dτ, p = q + 2, (106)
k=1
z
0
Zz m
X 1
!
α−1 bk α−1 χ (p−q)(−z) q−p
τ z +τ z e dτ, p ≥ q + 3, (107)
0 k=1
1 g
where z = w r τ r and gr > 0, χ is defined by (99). The convergence of the integral (104) takes place
when the integral (107) in below chain has convergence. It can happened under condition
g
min = Re(α) + min {Re(bk )} > 0. (108)
r 1≤k≤m
Evidently, that this condition exists in other three situation with p > q. Let

0, δ < 0;
θ(δ) =
1, δ ≥ 0.
There we see additional three integrals with exponential and cosine functions, for which one can write
and evaluate the following ”convergent equivalents” :
Z1 γ+1
γ axδ (−a)− δ γ+1 γ+1
x e dx = θ(δ)Γ −Γ , −a ,
δ δ δ
0
48
where Re(γ) > −1 when δ > 0 and Re(a) ≤ 0 when δ < 0;
Z1
xγ cos(axδ )dx =
0

1 γ+1 1 γ+1 a2 θ(−δ) π(γ + 1) γ+1 γ+1
= 1 F2 ; , + 1; − − cos Γ |a|− δ ,
γ+1 2δ 2 2δ 4 δ 2δ δ

where a ∈ R, δ ∈ R and Re γ+1 δ
< 1.
If we apply above conditions of convergence to mentioned three integrals, re-written in the form
Z1 1 −g
g q−m−n w − r
τ α+ r χ−1 e(−1) τ r
dτ,
0
Z1 q
α+ gr χ−1 g
− 1
τ cos 2τ 2r (−1)q−m−n−1 w r dτ,
0
Z1 1
1 q−p
g
α+ gr χ−1 (p−q) −w r τ r(q−p)
τ e dτ
0
we come to the following three groups of conditions

( g

Re 1α + r χ > 0, gr < 0;
(109)
Re w − r (−1)q−m−n ≤ 0, gr > 0,
q
1 g αr
w − r (−1)q−m−n−1 ∈ R and ∈ R and 2Re + χ < 1, (110)
r g

 Re α + gr χ > 0, g
r(q−p)
> 0;
1

1
q−p
g (111)
 Re (p − q) −w r ≤ 0, r(q−p)
< 0.
After incorporation these conditions into ”Big O-equivalent” integrals above we come to the following
conditions of convergence initial Riemann-Liouville integral at the point τ = 0:

 min > 0, ( p ≤ q;
 g g
Re α + r χ >1 0, < 0;


 r

 min > 0, q−m−n − g p = q + 1;


 Re (−1) w r ≤ 0, r > 0,
 q
min > 0, w − r1
(−1) q−m−n−1 ∈ R, g ∈ R, Re αr + χ < 1 , p = q + 2; (112)

 r g 2

Re α + gr χ > 0, g



  r(q−p)
> 0;
 1



 min > 0, 1
q−p g p ≥ q + 3,
  Re (p − q) −w r ≤ 0, r(q−p) < 0,
49
with gr > 0, χ is given by (99).
Evidently, that we need to add restriction Re(β) > 0 for convergence integral at point τ = z and
1 g
maybe we can have third singular point of Meijer G-function w r τ r = (−1)m+n−p , arising for q = p.
Let
1 r
τ0 = ((−1)m+n−p w r ) g . (113)
Then under conditions τ0 ∈ R and 0 < τ0 < z we should add restriction

p
!
X
Re (aj − bj ) > 0 (114)
j=1
(it is Re(ψp + 1) > 0 for q = p) which becomes weaker as

p
!
X
Re (aj − bj ) + β − 1 > 0 (115)
j=1
if point coincides with z: τ0 = z (here restriction Re(β) > 0 should be annulated). It provides
convergence at the point τ0 for 0 < τ0 ≤ z.
Using notations (99), (101), (108) and (113) for χ, ψp , min and τ0 we can write the following
conditions of convergence for integral (103):


 Re(β) > 0, min > 0, p < q;


 0 < τ0 < z, Re (ψp ) > −1 if Re(β) > 0;

 min > 0, τ0 ∈ R and p = q;


 τ0 = z
( if Re (β + ψp ) > 0,



 Re α + gr χ > 0 if gr < 0;
 Re(β) > 0, min > 0 and
 1 p = q + 1;
Re (−1)q−m−n w − r ≤ 0 if gr > 0, (116)
 q
 1 g
− αr 1


 Re(β) > 0, min > 0, (−1)q−m−n−1 w r ∈ R, r ∈ R, Re g + χ < 2 , p = q + 2;
 
Re α + gr χ >0 g



  if r(q−p) >0;
 1

 Re(β) > 0, min > 0, and  Re (p − q) −w r1 q−p ≤0 if g <0, p > q + 2,



 r(q−p)
2.4 Support of differential constants. Generic formulas for

fractional differentiation
With the advent of computer algebra systems, such as Wolfram Mathematics, Maple, Matlab, etc.,
it became necessary to revise approaches to well-known functions. Computer systems demanded to
provide correct numerical evaluations
√ of analytical functions everywhere in complex plane, including
on branch lines like (−∞, 0) for z. It stimulated developers of the Mathematica system to revise
mathematical formulas, where the behavior of functions on branch cuts was accurately described not
only theoretically, but supported by numerical evaluations in Mathematica. As a result, Wolfram
Mathematics uses a simple axiom: ”The argument of all complex numbers z satisfies the inequality
50
−π < arg(z) ≤ π.” As a result functions are internally consistent with each to other and can be
described even on branch cut lines. As a result Mathematica makes operations √ in the full complex
z2
plane with all functions involving so-called differential constants (like z , log (z 2 ) − 2 log(z)) and
piecewise construction including logarithmic situations. The question of what happens to the fractional
integro-differentiation in such cases previously ignored in the literature. Operation FractionalOrderD
fills this gap.
By definition in Mathematica we have
√
z2 1, − π2 < arg(z) ≤ π2 ;
= (117)
z −1, in other cases
and

0, − π2 < arg(z) ≤ π2 ;
2
log z − 2 log(z) = 2iπ, −π < arg(z) ≤ − π2 ; (118)

−2iπ, in other cases.
The built-in Mathematica derivative operator D misses discontinuities at branch cuts:
"√ #
z 2
D ,z = 0 and D log z 2 − 2 log(z), z = 0. (119)
z
√
FractionalOrderD provides the following results, which saves components z2 and
log(z 2 )= log(−iz)+ log(iz):
√ √
dα z 2 z 2 z −α−1
= (120)
dz α z Γ(1 − α)
and
dα 2
z −α (log(−iz) + log(iz) − 2 log(z))
[log z − 2 log(z)] = . (121)
dz α Γ(1 − α)
Classical (order 1 or 2, etc) differentiation and integration can operate with abstract functions
f (z), g(z), h(z) and some their constructions (product, composition, ratio, inverse function, series, etc.).
Similar formulas can be derived for fractional integro-differentiation. Many of such formulas we already
have in ResourceFunction["FractionalOrderD"]. For example, we have generic formulas for
1 f (z)
f (−1) (z), f (z) · g(z), f (z) · g(z) · h(z), , ,
f (z) g(z)
h(z)
(f (z))a , (f (z))g(z) , f (z)g(z) , (az)b logc (d · f (z)),
a +d a )+d z )+d
cbf (z) , cbf (z , cbf (a , f (g(z)), f (g(h(z))), etc.
(−1)
Here f (z) is an inverse for f (z) function. Below we give one rather simple example for product
f (z) · g(z):
 α
P α (α−k)
(z)g (k) (z) if z ∈ Z, α ≥ 0;

α

 f
d k=0
k
(f (z)g(z)) = α (122)
dz α 

P
α dα−k f (z) dk g(z)
in other cases.
 k dz α−k dz k
k=0
51
Of course, most formulas of fractional integro-differentiation of abstract functions are very com-
plicated and large. But in the case of specific formulas, we get simpler expressions. For example,
√ α −α
dα z 2 π2 z 1 α+1 α 2
e = F
α 2 2
, 1; 1 − ,1− ;z . (123)
dz α Γ 1 − α+1
2
Γ 1 − 2
2 2 2
In recent years Wolfram Language has added some new generic functions
• 10 Heun functions,
• 4 Lamé functions,
• 8 Carlson elliptic integrals,
• Fox H–function,
• 4 Coulomb functions.
For these functions corresponding formulas of fractional order integro-differentiation can be obtained.
For example, we consider Carlson’s elliptic integral (see [133])
Z∞
3
RD (x, y, z) = (t + x)−1/2 (t + y)−1/2 (t + z)−3/2 dt, x > 0, y > 0, z>0 (124)
2
0
and Heun G–function H(a, b, c, d, p, q, z) (see [134]). Heun G–function H(a, b, c, d, p, q, z) specializes to
2 F1 (c, d, p; z) if b = a · c · d and q = c + d − p + 1 or a = 1 and b = c · d.
We obtain
dα 1
RD , z, 1 =
dz α 2
 3(−1)α
 2α+3 21 α F1 α + 23 ; 12 , 1; α + 25 ; 21 , 1 − z , α ∈ Z, α ≥ 0;
∞ √ ∞ (125)
2 1( 2 2 )( 2 )k
1 3 1 3 1 −α P
P F ,k+ ;2; z k
(k+1)! 2 F1 ( 2 ,k+2;2; 2 )z
1 1 k
− 32Γ πz3 −α
3πz −α 2 2
 4Γ(1−α) , in other cases
k=0
(1−α)k ( 2 ) k=0 ( 3
2
−α ) k
and
∞
dα X k!ck z k−α
H(a, b, c, d, p, q, z) = , (126)
dz α k=0
Γ(k − α + 1)
b c P +c Q
where |z| < min(1, |a|), c0 = 1, c1 = ap , cj = − j−2 j−2Rj j−1 j−1 , Pj = (c + j)(d + j), Qj = −j(a(j + p +
q − 1) + c + d + j − q) − b, Rj = aj(j + p − 1).
The Heun G–function, in general, is not hypergeometric type function and its series representa-
k!ck
tion has coefficients ck = Γ(k−α+1) , α = 0, which satisfies three-terms recurrent relations, which were
described in ConditionalExpression. c f (z) z a a
Currently we have general rules for: ag(z) , abf (z) , cba +d , cbf (a )+d , cbf (z) +d , cbf (z ) , eg(z) ,
c 1 h(z)
eg(h(z)) , f (az ), (f abz )c , f abz , f (z) , zf (z), z 2 f (z), (f (z))a , (f (z))g(z) , (f (z))(g(z)) , f (z a ), f (g(z)),
f (z)
f (g(h(z))), g(z)
, f (z) · g(z), f (z) · g(z) · h(z), f (−1) (z).
52
Conclusion
Having considered various approaches to introducing an arbitrary power of a differential operator
d
dx
, we come to the conclusion that these approaches are not so different. So, when calculating various
fractional derivatives of a power function xp , we almost always get the same result. This makes the
considered derivatives coincide on the class of analytic functions. On the other hand, if there are
additional parameters in the fractional derivative, then the result of applying such an operator to a
power function will depend on these parameters. In some applied problems, this makes sense. Another
reason for introducing different definitions of fractional derivatives is to apply regularizations of divergent
integrals. Wolfram Mathematica uses the Hadamard regularization of divergent integrals. In addition,
Mathematics systematizes formulas for fractional derivatives of general form and for general functions.
Finally, returning to main question: ”How maximally naturally make extensions of differentiation
dn
dxn
from natural order n = 1, 2, 3, ... to arbitrary symbolical (or complex) order α” we can say the
following: ”Like in the case of extension factorial n! from natural n to gamma function Γ(α + 1) (α = n)
we do not have unique solution in such extension (corresponding details for n! see in [63], [64] page 43.).
As result we see in literature the numerous approaches to definitions of fractional order differentiation
by variable x. Corresponding inversions define fractional order integrations, which in majority cases
include additional point a at the end of integral from a to x, which demands additional restrictions
for convergence. Existence of this point distracts attention and it is naturally to use a = 0 as the end
of integral from 0 to x and allow x to be negative or complex. Such construction leads to Riemann-
Liouville left-sided fractional integral with beginning at 0 and Riemann-Liouville fractional derivative
described at formula (59). After involving Hadamard’s concept of the ”finite part” we avoid influence of
other ”aside” point (x = 0) on procedure of fractional integro-differentiation in our main point x, which
effectively works in all complex x-plane for any analytical functions in their regular or branch points,
including logarithmic cases (log(x), x1 , xc , (xn )1/m , etc.). It allowed to form described above Riemann-
Liouville-Hadamard fractional order integro-differentiation, which looks as maximally natural extension
of integer order differentiation and its inversion (integer order repeatable indefinite integration without
arbitrary polynomials, that arises). This statement follows from comparison of different approaches to
fractional order integro-differentiation between themself and how they realized on the basic simplest
functions z λ , ez , etc. which is widely used in Taylor and Fourier series.”
Additional information about Fractional Differentiation one can find in the talk of Oleg Marichev,
Paco Jain ”New in Fractional Differentiation” at Wolfram virtual Technology Conference (October
12-15, 2021) [135], which was extended to the next talk of Oleg Marichev ”Compositional Structure
of Classical Integral Transforms” at the Wolfram Technology Conference (October 18-21, 2022) [136].
Later these results got development at the talk of Marichev O., Shishkina E. [65] and set of Internet
presentations of Oleg Marichev [137–139].
We are eternally grateful to Paco Jain for help in compiling materials on fractional calculus in
Wolfram Mathematics and to Michael Trott, who describes the computer’s world of mathematical
functions including principal value axiomatic approach. A very special thanks to Igor Podlubny who
generously shared ideas for improving the material.
53
Bibliography
1. Atanacković T.M., Pilipović S., Stanković B., Zorica D. Fractional Calculus with Applications in
Mechanics: Wave Propagation, Impact and Variational Principles. London: Wiley, 2014.
2. Babenko Yu. I. Method of fractional differentiation in applied problems of theory heat and mass
transfer, St. Petersburg: NPO ”Professional”, 2009.
3. Brychkov Yu. A., Marichev O. I., Savischenko N. V., Handbook of Mellin transforms, Advances in
Applied Mathematics, CRC Press, Boca Raton, FL, 2019.
4. Boguslavskaya E., Mishura Y., Shevchenko G. Replication of Wiener-Transformable Stochastic
Processes with Application to Financial Markets with Memory. In: Silvestrov, S., Malyarenko,
A., Rančić, M. (eds) Stochastic Processes and Applications. SPAS 2017. Springer Proceedings in
Mathematics & Statistics, vol 271. Springer, Cham., 2018, pp. 335–361.
5. Bosiakov S., Rogosin S. Analytical modeling of the viscoelastic behavior of periodontal ligament
with using Rabotnov’ s fractional exponential function. In Lect. Notes Electr. Eng. 2015, pp. 156–
167.
6. Bosiakov S. Fractional calculus in biomechanics, in Encyclopedia of Continuum Mechanics, Springer,
Berlin, 2020, Vol. 2, pp. 946–953.
7. Butzer P.L., Westphal U. An Introduction to Fractional Calculus. In: Hilfer, R., Ed., Applications
of Fractional Calculus in Physics, World Scientific, Singapore City, 2000, pp. 1–85.
8. Caponetto R., Dongola G., Fortuna L., Petras I. Fractional Order Systems: Modeling and Control
Applications. World Scientific, Singapore, 2010.
9. Caputo M. Lineal model of dissipation whose Q is almost frequancy independent – II. Geophys. J.
Astronom. Soc. 13, 1967, pp. 529–539.
10. Caputo M. Elasticita e Dissipazione. Bologna: Zanichelli, 1969.
11. Caputo M. Mean fractional-order-derivatives differential equations and filters. Ann. Univ. Ferrara
41, 1995, pp. 73–84.
12. Coimbra C. F. M. Mechanics with variable-order differential operators, Annalen der Physik, vol.
12, no. 11-12, 2003, pp. 692–703.
13. Gorenflo R., Mainardi F., Rogozin S. Mittag-Leffler function: properties and applications. De Gry-
ter, 2019.
54
14. Cossar J. A theorem on Cesàro summability. J. London Math. Soc., 16, 1941, pp. 56–68.
15. De Oliveira, E. C., Machado J. A. T. A Review of Definitions for Fractional Derivatives and Integral,
Mathematical Problems in Engineering, 2014, pp. 1–6.
16. Debnath L. A brief historical introduction to fractional calculus // Int. J. Math. Educ. Sci. Tech.
2004. P. 35. No 4. P. 487–501.
17. Diethelm K. The Analysis of Fractional Differential Equations. An Application-Oriented Exposition

Using Differential Operators of Caputo Type. Heidelberg: Springer, 2004. 247 p.
18. Diethelm K. Numerical methods for the fractional differential equations of viscoelasticity, in Ency-
clopedia of Continuum Mechanics, Springer, Berlin, 2020, Vol. 3, pp. 1927–1938.
19. Dimovski I. Operational calculus for a class of differential operators. C. R. Acad. Bulg. Sci. 19 (12),
1966, pp. 1111–1114.
20. Dimovski I. On an operational calculus for a differential operator. C. R. Acad. Bulg. Sci. 21 (6),
1968, pp. 513–516.
21. Dimovski I.H., Kiryakova V.S. Transmutations, convolutions and fractional powers of Bessel-type
operators via Meijer’s G-function, in: “Complex Analysis and Applications ’83” (Proc. Intern. Conf.
Varna 1983), Sofia, 1985, pp. 45–66.
22. Dzhrbashjan M. M The generalized Riemann–Liouville operator and some of its applications. Dokl.
USSR Academy of Sciences 177 (4), 1967, pp. 767–770.
23. Dzhrbashjan M. M. The generalized Riemann–Liouville operator and some of its applications. Izv.
Akad. nauk SSSR, Ser. Matem. 32 (5), 1968, pp. 1075–1111.
24. Dzhrbashjan M. M., Nersesyan A. B. On the application of certain integro-differential operators.

Dokl. USSR Academy of Sciences 121 (2), 1958, pp. 210–213.
25. Dzhrbashjan M.M., Nersesyan, A.B., Fractional derivatives and Cauchy problems for fractional
differential equations. Izv. Acad. Nauk Arm. SSR 3 (1), 1968, pp. 3–28.
26. Dzhrbashjan M. M., Nersesyan A. B., Expansions in special biorthogonal systems and boundary
value problems for fractional differential equations. Dokl. USSR Academy of Sciences 132 (4), 1960,
747–750.
27. Dzhrbashjan M. M., Nersesyan A. B. Expansions in some biorthogonal systems and boundary value
problems for fractional differential equations. Tr. MMO 10, 1961, pp. 89–179.
28. Dzhrbashjan M.M. The basis property of biorthogonal systems generated by boundary value prob-
lems for fractional differential operators. Dokl. RAN SSSR 261 (5), 1981, pp. 1054–1058.
29. Engheta N., On Fractional Calculus and Fractional Multipoles in Electromagnetism, IEEE Trans-
actions on Antennas and Propagation, AP-444, April 1996, pp. 554–566.
55
30. Engheta N., On the role of fractional calculus in electromagnetic theory, IEEE Antennas and
Propagation Magazine, Vol. 39,No. 4,August 1997, pp. 35–46.
31. Fahad H. M., Fernandez A. Operational calculus for Riemann-Liouville fractional calculus with
respect to functions and the associated fractional differential equations. Fract. Calc. Appl. Anal.
24(2), 2021, pp. 518–540.
32. Gelfand I. M., Shilov G .E. Generalized functions, vol. 1: Properties and Operations. Providence:
AMS Chelsea Publishing, 2016.
33. Gemant A. A method for analyzing experimental results obtained from elasto-viscous bodies, Phys.
V. 7. 1936, pp. 311–317.
34. Gerasimov A.N. Generalization of the linear laws of deformation and their application to the prob-
lems of internal friction. Academy of Sciences of the USSR. Applied mathematics and mechanics
12, 1948., pp. 529–539.
35. Grigoletto E. C., de Oliveira E. C. Fractional versions of the fundamental theorem of calculus. Appl.
Math. 4, 2013, pp. 23–33.
36. Hadamard, J. 1892. Essai sur l’étude des fonctions données par leur développement de Taylor. J.
math. pures et appl., Ser. 4, 8, 101–186.
37. Herrmann R. Fractional Calculus: An Introduction For Physicists (Third Edition): An Introduction
for Physicists: 3rd Edition, 2018.
38. Jiao Zh., Chen Y. Q, Podlubny I. Distributed-Order Dynamic Systems: Stability, Simulation,
Applications and Perspectives. Springer, London, 2012.
39. Ji-Huan H. Fractal calculus and its geometrical explanation, Results in Physics, Volume 10, 2018,
pp. 272–276.
40. Jumarie G. Modified Riemann-Liouville derivative and fractional Taylor series of nondifferentiable
functions further results Comput. Math. Appl., 51 (2006), pp. 1367–1376.
41. Kaminsky A. A., Selivanov M. F. and Chornoivan Yu. O. Fractional-order operators in fracture
mechanics, in Encyclopedia of Continuum Mechanics, Springer, Berlin, 2020, Vol. 2, pp. 982–989.
42. Katrakhov V. V., Sitnik S. M. The transmutation method and boundary-value problems for singular
elliptic equations. Contemporary Mathematics. Fundamental Directions, 4 (2), 2018, pp. 211–426.
43. Kilbas A. A., Srivastava H. M., Trujillo J. J. Theory and applications of fractional differential
equation. Amsterdam, 2006.
44. Kilbas A. A. Power-logarithmic integrals in spaces of Hölder functions (Russian). Vestsi Akad. Nank
BSSR, Ser. Fiz.-Mat. Navuk, no I, 1975, pp. 37–43.
45. Kilbas A.A. Operators of potential type with power-logarithmic kernels in Holder spaces with weight
(Russian). Vea”i Akad. NavuJ: BSSR, Ser. Fiz.-Mat. NatIUk, no 2, 1978, pp. 29–37.
56
46. Kiryakova V. Generalized Fractional Calculus and Applications. Pitman Res. Notes Math. 301,
Longman Scientific & Technical, Harlow, Co-publ. John Wiley, New York, 1994.
47. Kolokoltsov V. N. The Probabilistic Point of View on the Generalized Fractional Partial Differential
Equations. FCAA 22, 2019, pp. 543–600 .
48. Kolwankar K. M., Gangal A. D. Fractional differentiability of nowhere differentiable functions and
dimensions, Chaos, 6 (4), 1996, pp. 505–513.
49. Lazarević M.P., Milan R.R., Tomislav B.Š. Introduction to fractional calculus with brief historical
background, Adv. Top. Appl. Fractional Calc. Control Probl. Syst. Stab. Model., 2014, 3, pp. 82–85.
50. Leonenko, N., Podlubny, I. Monte Carlo method for fractional-order differentiation. Fract Calc Appl
Anal 25, 346–361 (2022).
51. Leonenko, N., Podlubny, I. Monte Carlo method for fractional-order differentiation extended to
higher orders. Fract Calc Appl Anal 25, 841–857 (2022).
52. Letnikov A. V. Theory of differentiation of fractional order. Math. Sb. 3, 1868, 1–7.
53. Li C., Deng W . Remarks on fractional derivatives, Appl. Math. Comput. 187, 2007, pp. 777–784.
54. Li C.P., Wu, Y.J., Ye R.S. eds. Recent advances in applied nonlinear dynamics with numerical
analysis, World Scientific, 2013.
55. Li C.P., Zeng F.H. Numerical methods for fractional calculus. Chapman and Hall/CRC, 2015.
56. Li C.P., Cai M. Theory and numerical approximations of fractional integrals and derivatives. SIAM,
2019.
57. Luchko Y.F., Gorenflo R. An operational method for solving fractional differential equations. Acta
Math. Vietnamica 24, 1999, pp. 207–234.
58. Luchko Y., Kiryakova V. The Mellin integral transform in fractional calculus. Fract. Calc. Appl.
Anal. 16, 2013, pp. 405–430.
59. Machado J.T., Kiryakova V., Mainardi F. Recent history of fractional calculus // Commun. Non-
linear Sci. Numer. Simulat. 2011. V. 16. P. 1140–1153.
60. Machado J. T. A., Kiryakova V. Recent history of the fractional calculus: data and statistics,
Volume 1 Basic Theory, edited by Anatoly Kochubei and Yuri Luchko, Berlin, Boston: De Gruyter,
2019, pp. 1-22.
61. Mainardi F. Fractional Calculus and Waves in Linear Viscoelasticity: An Introduction to Mathe-
matical Models. Imperial College Press, London, 2010.
62. Malinowska A. B., Odzijewicz T., Torres D. F. M. Advanced methods in the fractional calculus of
variations, SpringerBriefs in Applied Sciences and Technology, Springer, Cham, 2015.
57
63. Marichev O. I. A method of calculating integrals of special functions. (Theory and tables of formu-
las), Nauka i Tekhnika, Minsk, 1978.
64. Marichev O. I. Handbook of Integral Transforms of Higher Transcendental functions (theory and
algorithmic tables), Ellis Horwood Ltd. 1983.
65. Marichev O., Shishkina E. Fractional order differentiation of Meijer G-functions and their cases,
book of matherials of the International Conference Polynomial Computer Algebra, 2023, St. Peter-
burg department of Steklov Institute of Mathematics, RAS, pp. 113–117.
66. McBride A. C. Fractional powers of a class of ordinary differential operators. Proc. London Math.
Soc. 3 (45), 1982, pp. 519–546.
67. Meilanov R. P., Magomedov R. A. Thermodynamics in fractional calculus. Journal of Engineering

Physics and Thermophysics, Vol. 87, No. 6, November, 2014, pp. 1521–1531.
68. Miller K. S. Ross B. An introduction to the fractional calculus and fractional differential equations.
Jon Wiley & Sons. Inc., New York, 1993.
69. Mishura Yu. S. Stochastic Calculus for Fractional Brownian Motion and Related Processes. Springer-
Verlag Berlin Heidelberg, 2008.
70. Monteiro N. Z., Mazorche S. R. Fractional Derivatives Applied to Epidemiology. Trends Comput.
Appl. Math. 22 (2), 2021, pp. 157–177.
71. Novozhenova O. G. Life and science of Alexey Gerasimov, one of the pioneers of fractional calculus
in Soviet Union, Fract. Calc. Appl. Anal. 20(3), 2017, pp. 3–14.
72. Ogata H. A numerical method of computing Hadamard finite-part integrals with an integral power
singularity at the endpoint on a half infinite interval arXiv:1910.00807v1
73. Ortigueira M. D. Fractional Calculus for Scientists and Engineers, Springer Dordrecht, 2011.
74. Ortigueira M. D., Trujillo J. J., A unified approach to fractional derivatives, Communications in
Nonlinear Science and Numerical Simulation, V. 17, Iss. 12, 2012, pp. 5151-5157.
75. Osler T. J., Leibniz rule for fractional derivatives generalized and an application to infinite series,
SIAM Journal on Applied Mathematics, vol. 18, 1970, pp. 658–674.
76. Petras I. Fractional-Order Nonlinear Systems: Modeling, Analysis and Simulation, Springer Science
and Business Media, Berlin, 2011.
77. Petras I., Podlubny I., O’Leary P., Dorcak L., Vinagre B. Analogue Realization of Fractional Order
Controllers. FBERG, Technical University of Kosice, Kosice, 2002, pp. 1–84.
78. Podlubny I. Fractional-Order Systems and Fractional-Order Controllers. UEF-03-94, Inst. Exp.
Phys., Slovak Acad. Sci. 1994, pp. 1–21.
58
79. Podlubny I. Fractional differential equations. An introduction to fractional derivatives, fractional
differential equations, to methods of their solution and some of their applications. Mathematics in
Science and Engineering, 198, Academic Press, SanDiego, 1999.
80. Podlubny I. Fractional-order systems and P I λD µ -controllers, in IEEE Transactions on Automatic
Control, 1999, vol. 44, no. 1, pp. 208–214.
81. Podlubny I. Matrix approach to discrete fractional calculus Fractional calculus and applied analysis
3(4), 2000, pp. 359–386.
82. Podlubny I, Chechkin A., Skovranek T., Chen Y.Q., Vinagre Jara B.M. Matrix approach to discrete
fractional calculus II: Partial fractional differential equations. Journal of Computational Physics V.
228, I. 8, 2009, pp. 3137–3153.
83. Podlubny I., Magin R. L., Trymorush I. Niels Henrik Abel and the birth of fractional calculus,
Fractional Calculus and Applied Analysis, vol. 20, no. 5, 2017, pp. 1068–1075.
84. Podlubny I., Skovranek T., Vinagre Jara B. M., Petras I., Verbitsky V., Chen Y. Q. Matrix approach
to discrete fractional calculus III: non-equidistant grids, variablestep length and distributed orders.
Phil Trans RSoc A 371: 20120153, 2013, pp. 1–15.
85. Povstenko Yu. Fractional calculus in thermoelasticity, in Encyclopedia of Continuum Mechanics,
Springer, Berlin, 2020, Vol. 2, pp. 953–961.
86. Prudnikov A. P., Marichev O. I., Brychkov Yu. A. Integrals and Series, Vol. 3: More Special
Functions. Newark, NJ: Gordon and Breach, 1990.
87. Pskhu A. V. Partial partial differential equations. Moscow, Nauka, 2005.
88. Pskhu A. V. Boundary value problems for fractional and continuum differential partial differential
equations. Nalchik: ed. KBSC RAS, 2005.
89. Rabotnov Yu. N. Equilibrium of an elastic medium with after-effect, Fract. Calc. Appl. Anal. 17,
2014, 684–696.
90. Rogozin S., Dubatovskaya M. Letnikov vs. Marchaud: A survey on two prominent constructions of
fractional derivatives // Mathematics. 2018. V. 6. No 1. PaperID 3.
91. Rogosin S., Mainardi F. George Scott Blair – the pioneer of fractional calculus in rheology. Com-
munications in Applied and Industrial Math., Special issue, 2014, 14 p.
92. Ross, B., 1975. A brief history and exposition of the fundamental theory of fractional calculus. In:
Ross B. (eds) Fractional Calculus and Its Applications. Lecture Notes in Mathematics, vol 457.
Springer, Berlin, Heidelberg.
93. Rossikhin Yu. A. Reflections on two parallel ways in progress of fractional calculus in mechanics of
solids, Appl. Mech. Rev. V. 63. N 1, 2010, pp. 1–12.
94. Rossikhin Yu. A., Shitikova M. V. Fractional operator models of viscoelasticity, in Encyclopedia of
Continuum Mechanics, Springer, Berlin, 2020, Vol. 2, pp. 971–982.
59
95. Rossikhin Yu. A., Shitikova M. V. Features of fractional operators involving fractional derivatives
and their applications to the problems of mechanics of solids, in Fractional Calculus: History,
Theory and Applications, Nova Science Publishers, New York, 2015, Chap. 8, pp. 165–226.
96. Rossikhin Yu. A., Shitikova M. V. Fractional calculus models in dynamic problems of viscoelasticity,
in Handbook of Fractional Calculus with Applications, Vol. 7: Applications in Engineering, Life
and Social Sciences, Part A., De Gruyter, Berlin, 2019, pp. 139–158.
97. Rossikhin Yu. A., Shitikova M. V. Applications of fractional calculus to dynamic problems of linear
and nonlinear hereditary mechanics of solids, Appl. Mech. Rev. 50, No 1, 1997, 15–67.
98. Rossikhin Yu. A., Shitikova M. V. Application of fractional calculus for dynamic problems of solid
mechanics: Novel trends and recent results, Appl. Mech. Rev. 63 (1), 2010, 010801.
99. Rossikhin Yu.A., Shitikova M.V. Fractional calculus in structural mechanics. Handbook of Frac-
tional Calculus with Applications. Volume 7: Applications in Engineering, Life and Social Sciences,
Part A, (D. Baleanu, A.M. Lopes, Eds.), De Gruyter, 2019, pp. 159–192.
100. Rutman R. S. On physical interpretations of fractional integration and differ-entiation. Theor.

Math. Phys. 105, 1995, 1509–1519.
101. Saigo M. A remark on integral operators involving the Gauss hypergeometric functions. Math.
Rep. Kyushu Univ. 11 (2), 1977, pp. 135–143.
102. Samko A. G. Fractional integration and differentiatin of variable order. Anal. Math. 21, 1995, pp.
213–236.
103. Samko S. G., Ross B. Integration and differentiation to a variable fractional order. Integral Trans-
form Spec Funct 1(4), 1993, pp. 277–300.
104. Scott Blair G.W. Analytical and integrative aspects of the stress-strain-time problem, J. Sci.
Instrum. V. 21. N 5, 1944, pp. 80–84.
105. Shishkina E. L., Sitnik S. M. On fractional powers of Bessel operators. Journal of Inequalities and
Special Functions 8 (1), 2017, pp. 49–67.
106. Shishkina E. L., Sitnik S. M. On fractional powers of Bessel operators. Journal of Inequalities
and Special Functions, Special issue To honor Prof. Ivan Dimovski’s contributions 8 (1), 2017, pp.
49–67.
107. Shishkina, E.L., Sitnik, S.M. Transmutations, Singular and Fractional Differential Equations with
Applications to Mathematical Physics, Mathematics in Science and Engineering, Elsevier. Academic
Press, 2020, 592 pp.
108. Shitikova M. V. Fractional operator viscoelastic models in dynamic problems of mechanics of solids,
A review. Mechanics of Solids. V. 57, N 1, 2022, pp. 1–33.
60
109. Shitikova M. V., Krusser A. I. Models of Viscoelastic Materials: a Review on Historical Develop-
ment and Formulation. Advanced Structured Materials, 2022, Vol. 175 (I. Giorgio, L. Placidi, E.
Barchiesi, B.E. Abali, H. Altenbach, Editors), Springer , Chapter 14, pp. 285-326
110. Shitikova M. V. Wave theory of impact and Professor Yury Rossikhin contribution in the field (A
Memorial Survey), Journal of Materials Engineering and Performance. 2019. Vol. 28, Issue 6, pp.
1–13.
111. Shkhanukov M. K., On the convergence of difference schemes for differential equations with a
fractional derivative, Dokl. Akad. Nauk, 348(6), 1996, pp. 746–748.
112. Sitnik S. M., Shishkina E. L. Transmutation operators method for differential equations with Bessel
operator. Moscow. Fizmathlit, 2019.
113. Sprinkhuizen-Kuyper I.G. A fractional integral operator corresponding to negative powers of a

certain second-order differential operator. J. Math. Analysis and Applications 72, 1979, pp. 674–
702.
114. Stanislavsky A. A. Probabilistic interpretation of the integral of fractional order. Theor. Math.
Phys. 138, 2004, pp. 418–431.
115. Sun H.G., Zhang Y., Baleanu D., Chen W., Chen Y.Q. A new collection of real world applications
of fractional calculus in science and engineering. Commun. Nonlinear Sci. Numer. Simulat. 2018.
V. 64, pp. 213–231.
116. Teodoro G. S. , Machado J.A. T., de Oliveira E. C. A review of definitions of fractional derivatives
and other operators, Journal of Computational Physics, 2019. V. 388, pp. 195–208.
117. Uchaikin V. V. Fractional Derivatives for Physicists and Engineers. Springer Berlin, Heidelberg,
2013.
118. Uchaikin V. V. Fractional models in hydromechanics, Izv. Vyssh. Uchebn. Zav. Prikl. Nelin. Din.
2019, 27, 1, pp. 5–40.
119. Valério D., Machado J. T., Kiryakova V. Some pioneers of the applications of fractional calculus.
FCAA 17, 2014, 552–578.
120. West B. J., Bologna M., Grigolini P. Physics of Fractal Operators. Springer-Verlag, New York,
2003.
121. Zaslavsky G. M. Hamiltonian Chaos and Fractional Dynamics, Oxford University Press, 2005.
122. Zaslavsky G. M. Chaos, Fractional Kinetics, and Anomalous Transport, Physics Reports, vol. 371,
2002, pp. 461–580.
123. Zhmakin A. I. A Compact Introduction to Fractional Calculus, arXiv:2301.00037v1
124. https://en.wikipedia.org/wiki/Big O notation

125. https://functions.wolfram.com/HypergeometricFunctions/MeijerG1/
61
126. http://functions.wolfram.com/01.02.16.0037.01
130. https://functions.wolfram.com
131. https://functions.wolfram.com/HypergeometricFunctions/MeijerG/20/03/01/ShowAll.html
132. https://functions.wolfram.com/GammaBetaErf/Gamma/16/02/0004/
133. https://mathworld.wolfram.com/CarlsonEllipticIntegrals.html
134. https://en.wikipedia.org/wiki/Heun function
135. https://www.wolfram.com/events/technology-conference/2021/presentations/#day3
136. https://www.wolfram.com/events/technology-conference/2022/
137. https://community.wolfram.com/groups/-/m/t/2821053
140. https://blog.wolfram.com/2016/05/16/new-derivatives-of-the-bessel-functions-have-been-
discovered-with-the-help-of-the-wolfram-language/
62

O.I. Marichev, E.L. Shishkina Overview of Fractional Calculus and Its Computer Implementation in Wolfram Mathematica

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

O.I. Marichev, E.L. Shishkina Overview of Fractional Calculus and Its Computer Implementation in Wolfram Mathematica

Uploaded by

Copyright:

Available Formats

O.I. Marichev, E.L.

1 History, nowadays and applications 5

2 Fractional-order differentiation in the Wolfram language 26

1.1 Background and concept

Property Γ(α) = Γ(α+1)

−(p + 1), −(p − α + 1) ∈

It should be noticed that Abel solved more general equation

where 0 < α < 1.

The limitation of this definition is related to the convergence of the series.

1.1.2 Formal definition of fractional integro-differentiation

Criteria 1. Operator D ν f (z) is the integro-differential operator of order ν ∈ C if and only if

3. The operation of order zero leaves the function unchanged:

5. The semigroup property of arbitrary order holds:

1.2 Different approaches to classical fractional calculus

1.2.1 Riemann-Liouville type fractional integrals and derivatives

Liouville fractional integral of order α on the whole real axis is

is called the Chen fractional integral.

we should add the initial condition

f (x) = xα−1 Eα,α (λxα ) (16)

f (x) = Eα,1 (λxα ). (17)

Now it is bounded at zero and so the classical Cauchy problem is correct.

Γ(p + 1) Γ(n − α)Γ(p − n + 1)xp−α Γ(p + 1)

So here we saw significant limitations when using a fractional derivative (14).

Osler fractional derivative (see [75], formula 2.2)

where the contour C[a, z + ] starts and ends at t = 0.

1.2.2 Finite differences approach to fractional calculus

to the non-integer order.

where 0 < Re α < l, l ∈ N or l = [α] + 1 when α ∈ R.

1 (∆αh f )(x) + (∆α−h f )(x)

Next we introduce a convolution operator

which is defined using the Fourier transform by the equality

1.2.4 Fractional integro-differentiation of analytic functions

This operator admits the following integral representation

Corresponding differential operator for 0 < α < 1 is

where g(z) = z 1/α . The operator Dα corresponds to the expansion

For periodic functions definition fractional integro-differentiation, suggested by Weyl, is

Similarly fractional differentiation is defined:

The definition (32) may be interpreted as

(Fα+ f )(x) = A−1 I+α Af, (Af )(x) = f (ex ).

For 0 < α < 1 Hadamard fractional derivative has the form

When α = n ∈ N ∪ {0}, then

1.3.2 Erdelýi-Kober-type operators

where (Qf )(x) = f [g(x)].

1.3.3 Local fractional derivatives and variable-order differential operators

where Dxα0 is the left-sided Riemann-Liouville fractional derivative of order α.

(D α f )(x0 ) = lim ( C Dxα0 f )(x), n − 1 < α ≤ n, (48)

1.3.4 Generalized fractional integrals

2. ω(0) = 1, ω(1) = 0, ω(x) 6= 0 as 0 < x < 1,

3. ω ′(x) ∈ L1 (0, 1).

1.4 Applied aspects of fractional derivatives and integrals

1.4.2 Numerical calculus

1.4.3 Publications after 1987

2.1 Riemann-Liouville-Hadamard integro-differentiation in

order to get this function we should write

2.1.2 Simplest examples

For α = −1/2 and for α = 1/2 we can write

differentiation or indefinite integration

Also in many cases we see special functions as result of integration.

In the following sections we consider these three approaches.

2.2 Calculation of fractional derivatives and integrals by se-

Sum representations by formula (60) we meet for functions like

It means, that we should consider more general series

Here for α ∈ Z and α > 0