Annals of Physics 323 (2008) 2756-2778

Fractional Vector Calculus and

Fractional Maxwell’s Equations
arXiv:0907.2363v3 [math-ph] 23 Jul 2011

Vasily E. Tarasov

Skobeltsyn Institute of Nuclear Physics,

Moscow State University, Moscow 119991, Russia
E-mail: [email protected]

Abstract

The theory of derivatives and integrals of non-integer order goes back to Leibniz, Liou-
ville, Grunwald, Letnikov and Riemann. The history of fractional vector calculus (FVC)
has only 10 years. The main approaches to formulate a FVC, which are used in the physics
during the past few years, will be briefly described in this paper. We solve some problems
of consistent formulations of FVC by using a fractional generalization of the Fundamen-
tal Theorem of Calculus. We define the differential and integral vector operations. The
fractional Green’s, Stokes’ and Gauss’s theorems are formulated. The proofs of these the-
orems are realized for simplest regions. A fractional generalization of exterior differential
calculus of differential forms is discussed. Fractional nonlocal Maxwell’s equations and
the corresponding fractional wave equations are considered.

PACS: 45.10.Hj; 03.50.De; 41.20.-q

1
1 Introduction
The fractional calculus has a long history from 30 September 1695, when the derivative of
order α = 1/2 has been described by Leibniz [1, 2] (see also [6]). The theory of derivatives and
integrals of non-integer order goes back to Leibniz, Liouville, Grunwald, Letnikov and Riemann.
There are many interesting books about fractional calculus and fractional differential equations
[1, 2, 3, 4, 5] (see also [7, 8]). Derivatives and integrals of fractional order, and fractional
integro-differential equations have found many applications in recent studies in physics (for
example, see books [9, 10, 11, 12], and reviews [13, 14, 15]).
The history of fractional vector calculus (FVC) is not so long. It has only 10 years and
can be reduced to the papers [16]-[27]. The main approaches to formulate a FVC, which are
used in the physics during the past few years, will be briefly described in this paper. There are
some fundamental problems of consistent formulations of FVC that can be solved by using a
fractional generalization of the Fundamental Theorem of Calculus. Fractional vector calculus
is very important to describe processes in fractal media (see for example [10]). A consistent
FVC can be used in fractional electrodynamics [18, 19, 20, 25] and fractional hydrodynamics
[21, 29].
In Section 2, we describe different approaches to formulate FVC, which are used in the
physics during the past 10 years. The problems of consistent formulation of FVC are described
in Section 3. A fractional generalization of the Fundamental Theorem of Calculus is considered
in Section 4. In Section 5, the differential and integral vector operations are defined. In Sections
6-8, the fractional Green’s, Stokes’ and Gauss’s theorems are formulated. The proofs of these
theorems are realized for simplest regions. In Section 9, a fractional generalization of exterior
calculus of differential forms is discussed. In Section 10, fractional nonlocal Maxwell’s equations
and the corresponding fractional wave equations are considered.

2
2 Approaches to fractional vector calculus
For Cartesian coordinates, fractional generalizations of the divergence or gradient operators
can be defined by
gradα f (x) = es Dsα f (x), (1)

div α F(x) = Dsα Fs (x), (2)

where Dsα are fractional (Liouville, Riemann-Liouville, Caputo, etc.) derivatives [1, 2, 3, 4, 5]
of order α with respect to xs , (s = 1, 2, 3). Here es (s = 1, 2, 3) are orthogonal unit vectors,
and Fs (x) are components of the vector field

F(x) = Fs (x)es = Fx ex + Fy ey + Fz ez . (3)

The main problem of formulation of FVC appears, when we try to generalize the curl
operator and the integral theorems. In Cartesian coordinates, the usual (integer) curl operator
for the vector field (3) is defined by

curlF = el εlmn Dm Fn , (4)

where Dm = ∂/∂xm , and εlmn is Levi-Civita symbol, which is 1 if (i, j, k) is an even permutation
of (1, 2, 3), (−1) if it is an odd permutation, and 0 if any index is repeated. The Fourier
transform of the curl operator is

F (curlF(x)) = el εlmn (ikm )F̃n (k), (5)

where Z +∞
F̃n (k) = F (Fn (x)) = d3 x e−ikx Fn (x). (6)
−∞

To define a generalization of (4), we can use a fractional integro-differentiation instead of the

derivative Dm .

3
2.1 Ben Adda’s fractional vector calculus
In the paper [16] (see also [17]), fractional generalizations of gradient, divergence and curl
operator for analytic functions have been suggested in the form

1
gradα f (x) = es Dsα f (x), (7)
Γ(α + 1)
1
div α F(x) = Dsα Fs (x), (8)
Γ(α + 1)
1
curlα F = α
el εlmn Dm Fn (x), (9)
Γ(α + 1)
where Γ(α + 1) is the Gamma function. In these definitions, the Nishimoto fractional derivative
[8] (see also Section 22 of [2]) is used. This derivative is a generalization of the Cauchy’s
differentiation formula.
Fractional generalizations of integral operations (flux and circulation), and generalizations
of Gauss’s, Stokes’, Green’s integral theorems are not considered.

2.2 Engheta’s fractional vector calculus

In the paper [18] (see also [18, 19, 20]), a fractional generalization of curl operator has been
suggested in the form
curlα F = el εlmn Dm
α
Fn (x), (10)
α
where Dm are fractional Liouville derivatives [5] of order α with respect to xm , (m = 1, 2, 3),
that are defined by
α α
Dm f (x) := lim a Dxm f (xm ). (11)
a→−∞

Here a Dxα is the Riemann-Liouville derivative

Z x
α 1 ∂n f (x′ )
D
a x f (x) = dx′ , (n − 1 < α < n). (12)
Γ(n − α) ∂xn a (x − x′ )α−n+1

The fractional Liouville derivative (11) can be defined through the Fourier transform by
Z +∞
α −1

α
 1
Dm Fn (x) = F (ikm ) F̃n (k) = d3 k eikx (ikm )α F̃n (k), (13)
(2π)3 −∞

4
where F̃n (k) is defined by (6) and iα = exp{iαπ sgn(k)/2}. For this fractional curl operator,
the fractional integral Stokes’ and Green’s theorems are not suggested. The problems of a
generalization of these theorems will be considered in the next section.
In general, the fractional vector calculus must include generalizations of the differential oper-
ations (gradient, divergence, curl), the integral operations (flux, circulation), and the theorems
of Gauss, Stokes and Green.

2.3 Meerschaert-Mortensen-Wheatcraft fractional vector calculus

In the paper of Meerschaert, Mortensen and Wheatcraft [21], a fractional generalization of curl
operator has been suggested as

curlα F = el εlmn Dm In1−α Fn , (14)

where In1−α are fractional integrals of order (1 − α) with respect to xn , (n = 1, 2, 3). Note that
the integration In1−α in (14) is considered with the index n as the component Fn . The derivative
Dm = ∂/∂xm in Eq. (14) is considered with respect to xm , where m 6= n. Therefore expression
(14) can be presented as the usual (integer) curl operator

curlα F = curlF(α) (15)

for the field

F(α) = en In1−α Fn . (16)

Equation (15) allows us to use the usual (integer) integral Stokes’ and Green’s theorems.
In Eq. (14), the fractional integral Inα and the integer derivative Dm have antisymmetric
indices, and the components of (14) are

(curlα F)x = Dy Iz1−α Fz − Dz Iy1−α Fy , (17)

(curlα F)y = Dz Ix1−α Fx − Dx Iz1−α Fz , (18)

(curlαF)z = Dx Iy1−α Fy − Dy Ix1−α Fx . (19)

5
It is easy to see that operator (14) has no fractional derivatives with respect to xm , (m = 1, 2, 3),
α 1−α
like as Dm = Dm Im or C Dm
α 1−α
= Im Dm .
As a result, we have the usual (integer) vector calculus for new type of fields as in (16). The
suggested approach cannot be considered as a fractional generalization of vector calculus. It is
important to define a curl operator with fractional derivatives in such a form that fractional
generalizations of the integral theorems exist.

2.4 Other approaches to fractional vector calculus

In the papers [25, 29], fractional generalizations of integral operations and Gauss’s, Stokes’,
Green’s theorems have been suggested. These generalizations are considered to describe frac-
tional media by a continuous medium model. The differential operations are defined with
respect to fractional powers of coordinates. These operations are connected with fractional
derivatives only by Fourier transforms (see [28]). As a result, an ”ideal” fractional vector
calculus is not suggested.
In the papers [26, 27], fractional differential vector operations are considered by using frac-
tional generalizations of differential forms that are suggested in [22] (see also [23, 24]). A frac-
tional gradient is defined by an exact fractional 1-form. A fractional curl operator is described
by a fractional exterior derivative of a fractional differential 1-form. The Riemann-Liouville
derivatives are used in [26], and the fractional Caputo derivatives are used in [27]. We have

gradα f (x) = es C α
0 Dxs f (x), (20)

curlα F = el εlmn C α
0 Dxm Fn , (21)

where C α
0 Dxm is a fractional Caputo derivative with respect to xm :
Z x n ′
C α 1 1 ′ ∂ f (x )
a Dx f (x) = dx , (n − 1 < α < n). (22)
Γ(n − α) a (x − x′ )α−n+1 ∂(x′ )n

The fractional generalizations of integral theorems (Gauss’s, Stokes’, Green’s theorems) are not
considered and the fractional integrals for differential forms are not defined.

6
3 Problems of fractional generalization of Green’s for-
mula
Let us describe a main problem that appears when the curl operator and integral formulas are
generalized on a fractional case. For simplification, we consider a rectangular domain on R2
and integral formulas in Cartesian coordinates.
The Green’s formula in Cartesian coordinates is
Z Z Z
(Fx dx + Fy dy) = dxdy [Dy Fx − Dx Fy ] , (23)
∂W W

where Fx = Fx (x, y) and Fy = Fy (x, y) are functions defined for all (x, y) in the region W .
Let W be the rectangular domain

W := {(x, y) : a ≤ x ≤ b, c ≤ y ≤ d}

with the sides AB, BC, CD, DA, where the points A, B, C, D have coordinates

A(a, c), B(a, d), C(b, d), D(b, c).

These sides form a boundary ∂W of W . Then

Z Z Z Z Z
(Fx dx + Fy dy) = Fx dx + Fx dx + Fy dy + Fy dy =
∂W BC DA AB CD
Z b Z a Z d Z c
= Fx (x, d)dx + Fx (x, c)dx + Fy (a, y)dy + Fy (b, y)dy =
a b c d
Z b Z d
= dx [Fx (x, d) − Fx (x, c)] + dy [Fy (a, y) − Fy (b, y)]. (24)
a c
The main step of proof of Green’s formula is to use the Newton-Leibniz formula
Z b
dx Dx f (x) = f (b) − f (a). (25)
a

The function f (x) in (25) is absolutely continuous on [a, b]. As a result, expression (24) can be
presented as
Z b Z d  Z d  Z b 
dx dy Dy Fx (x, y) + dy − dx Dx Fy (x, y) =
a c c a

7
 Z b Z d Z Z
= dx dy [Dy Fx (x, y) − Dx Fy (x, y)] = dxdy [Dy Fx − Dx Fy ] .
a c W

To derive a fractional generalization of Green’s formula (23), we should have a generalization

of the Newton-Leibniz formula (25) in the form

α α
a Ib a Dx f (x) = f (b) − f (a), (26)

where some integral and derivative of noninteger order are used. This generalization exists for
specified fractional integrals and derivatives, and does not exist for arbitrary taken type of the
fractional derivatives.
For the left Riemann-Liouville fractional integral and derivative (Lemma 2.5. of [5]), we
have n
α α
X (b − a)α−j
a Ib a Dx f (x) = f (b) − (Dxn−j a Ixn−α f )(a), (27)
j=1
Γ(α − j + 1)

where Dxn−j = dn−j /dxn−j are integer derivatives, and n−1 < α < n. In particular, if 0 < α < 1,
then
α α (b − a)α−1 1−α
a Ib a Dx f (x) = f (b) − a Ib f (x), (28)
Γ(α)
Obviously that Eq. (28) cannot be considered as a realization of (26). The left Riemann-
Liouville fractional integral for x ∈ [a, b] is defined by
Z x
α 1 dx′
I
a x f (x) := (α > 0). (29)
Γ(α) a (x − x′ )1−α

The left Riemann-Liouville fractional derivative for x ∈ [a, b] and n − 1 < α < n is defined by
Z x
α n n−α 1 ∂n f (x′ )dx′
a Dx f (x) := Dx a Ix f (x) = . (30)
Γ(n − α) ∂xn a (x − x′ )α−n+1

Note that Eq. (27) is satisfied if f (x) is Lebesgue measurable functions on [a, b] for which
Z b
f (x) dx < ∞,
a

and a Ibn−α f (x) of the right-hand side of (27) has absolutely continuous derivatives up to order
(n − 1) on [a, b].

8
 Properties (27) and (28) are connected with the definition of the Riemann-Liouville frac-
tional derivative, where the integer-order derivative acts on the fractional integral:

α
a Dx = Dxn a Ixn−α , (n − 1 < α < n). (31)

This definition gives that the left–hand side of (28) is

α α α
a I x a Dx = a Ix Dxn a Ixn−α , (32)

where the integer derivative Dxn is located between the fractional integrals. Since the operations
Dxn and α
a Ix are not commutative

n−α n
a Ix Dx − Dxn a Ixn−α 6= 0,

we get the additional terms, which cannot give the right-hand side of (26). This noncommu-
tativity can be presented as a nonequivalence of Riemann-Liouville and Caputo derivatives
[4, 5],
n−1
C α α
X (x − a)j−α
a Dx f (x) = a Dx f (x) − (Dxj f )(a), (n − 1 < α < n). (33)
j=0
Γ(j − α + 1)
The left Caputo fractional derivative is defined by the equation (compare with (31))

C α n−α n
a Dx f (x) := a Ix Dx f (x), (n − 1 < α < n). (34)

The noncommutativity of Dxn and a Ixα in (32) does not allow us to use semi-group property (see
Lemma 2.3 of [5] and Theorem 2.5 of [2]) of fractional integrals

α β α+β
a Ix a Ix = a Ix , (α > 0, β > 0). (35)

Note that equation (35) is satisfied at almost every point x ∈ [a, b] for f (x) ∈ Lp (a, b) and
α, β > 0. We denote by Lp (a, b) (1 < p < ∞) the set of those Lebesgue measurable functions
on [a, b] for which
Z b 1/p
p
dx |f (x)| < ∞.
a
In general, the semi-group property

α β α+β
a Dx a Dx = a Dx , (α > 0, β > 0). (36)

9
is not satisfied for fractional derivatives (see Property 2.4 in [5]). For some special cases, Eq.
(36) can be used (see Theorem 2.5. in [2]). For example, the property (36) is satisfied for the
functions
f (x) ∈ a Ixα+β (L1 (a, b)),

i.e., equation (36) is valid for f (x) if there exists a function g(x) ∈ L1 (a, b) such that

f (x) = a Ixα+β g(x).

The semi-group property for fractional derivatives is also valid if a = 0, b = ∞ and f (x) is
infinitely differentiable (generalized) function on [0, ∞) (see Sec.1.4.5. of [30] and Sec.8.3. of
[2]).
In order to have a fractional generalization of the Newton-Leibniz formula of the form (26),
we must replace the left Riemann-Liouville derivative a Dbα in Eq. (26), where

α α
a I x a Dx = a Ibα (Dxn a Ixn−α )

C α
by the left Caputo derivative a Dx , such that the left-hand side of (26) is

α C α α n−α n
a I x a Dx = a Ix ( a Ix Dx ).

Then, we can use the semi-group property (35), and

α C α α n−α n
a Ix a Dx f (x) = a Ix a Ix Dx f (x) = a Ixn Dxn f (x).

In particular, if n = 1 and 0 < α < 1, then

Z b
α C α 1
a Ib a Dx f (x) = a Ib Dx1 f (x) = dx Dx1 f (x) = f (b) − f (a). (37)
a

As a result, to generalize Gauss’s, Green’s and Stokes’ formulas for fractional case, we can
use the equation with the Riemann-Liouville integral and the Caputo derivative:

α C α
a Ib a Dx f (x) = f (b) − f (a). (38)

This equation can be considered as a fractional analog of the Newton-Leibniz formula.

10
4 Fractional Generalization of the Fundamental Theo-
rem of Calculus
The fundamental theorem of calculus (FTC) is the statement that the two central operations
of calculus, differentiation and integration, are inverse operations: if a continuous function is
first integrated and then differentiated, the original function is retrieved

Dx1 a Ix1 f (x) = f (x). (39)

An important consequence, sometimes called the second fundamental theorem of calculus, al-
lows one to compute integrals by using an antiderivative of the function to be integrated:

1
a Ib Dx1 f (x) = f (b) − f (a). (40)

If we use the Riemann-Liouville integrals and derivatives [2, 5], we cannot generalize (40) for
fractional case, since
α α
a Ib a Dx f (x) 6= f (b) − f (a), (41)

In this case, we have equation (27).

The FTC states that the integral of a function f over the interval [a, b] can be calculated by
finding an antiderivative F , i.e., a function, whose derivative is f . Integral theorems of vector
calculus (Stokes’, Green’s, Gauss’s theorems) can be considered as generalizations of FTC.
The fractional generalization of the FTC for finite interval [a, b] can be realized (see remarks
after proof of the theorem and Section 3.) in the following special form.

Fundamental Theorem of Fractional Calculus

(1) Let f (x) be a real-valued function defined on a closed interval [a, b]. Let F (x) be the function
defined for x in [a, b] by
F (x) = a Ixα f (x), (42)
α
where a Ix is the fractional Riemann-Liouville integral
Z x
α 1 f (x′ )
a Ix f (x) := ′ 1−α
dx′ , (43)
Γ(α) a (x − x )

11
then
C α
a Dx F (x) = f (x) (44)

for x ∈ (a, b), where C α

a Dx is the Caputo fractional derivative
Z x
C α n−α n 1 dx′ dn F (x′ )
a D x F (x) = a I x D x F (x) = , (n − 1 < α < n). (45)
Γ(n − α) a (x − x′ )1+α−n d(x′ )n
(2) Let f (x) be a real-valued function defined on a closed interval [a, b]. Let F (x) be a
function such that
C α
f (x) = a Dx F (x) (46)

for all x in [a, b], then

α
a Ib f (x) = F (b) − F (a), (47)

or, equivalently,
α C α
a Ib a Dx F (x) = F (b) − F (a), (0 < α < 1). (48)

As a result, we have the fractional analogs of equations (39) and (40) in the form

C α α
a Dx a Ix f (x) = f (x), (α > 0), (49)

α C α
a Ix a Dx F (x) = F (x) − F (a), (0 < α < 1), (50)

where a Ixα is the Riemann-Liouville integral, and C α

a Dx is the Caputo derivative.

Proof.
The proof of this theorem can be realized by using the Lemma 2.21 and Lemma 2.22 of [5].
(1) For real values of α > 0, the Caputo fractional derivative provides operation inverse to
the Riemann-Liouville integration from the left (see Lemma 2.21 [5]),

C α α
a Dx a Ix f (x) = f (x), (α > 0) (51)

for f (x) ∈ L∞ (a, b) or f (x) ∈ C[a, b].

(2) If f (x) ∈ AC n [a, b] or f (x) ∈ C n [a, b], then (see Lemma 2.22 [5])
n−1
α C α
X 1
a Ix a Dx f (x) = f (x) − (x − a)j (Dxj f )(a), (n − 1 < α ≤ n), (52)
j=0
j!

12
where C n [a, b] is a space of functions, which are n times continuously differentiable on [a, b]. In
particular, if 0 < α ≤ 1 and f (x) ∈ AC[a, b] or f (x) ∈ C[a, b], then

α C α
a Ix a Dx f (x) = f (x) − f (a). (53)

This equation can be considered as a fractional generalization of the Newton-Leibniz formula

in the form (26).

Remark 1. In this theorem (see Eqs. (42-48)), the spaces L1 [a, b] and AC[a, b] are used.
(a) Here AC[a, b] is a space of functions F (x), which are absolutely continuous on [a, b]. It
is known that AC[a, b] coincides with the space of primitives of Lebesgue summable functions
and therefore an absolutely continuous function F (x) has a summable derivative Dx1 (x) almost
everywhere on [a, b]. If F (x) ∈ AC[a, b], then the Caputo derivative (0 < α < 1) exists almost
everywhere on [a, b] (see Theorem 2.1 of [5]).
(b) We denote Lp (a, b) the set of those Lebesgue measurable functions f on [a, b] for which
Z b 1/p
p
kf kp = |f (x)| dx < ∞. (54)
a

If f (x) ∈ Lp (a, b), where p > 1, then the fractional Riemann-Liouville integrations are bounded
in Lp (a, b), and the semi-group property

α β α+β
a Ix a Ix f (x) = a Ix f (x), (α > 0, β > 0) (55)

are satisfied at almost every point x ∈ [a, b]. If α + β > 1, then relations (55) holds at any
point of [a, b] (see Lemma 2.1 and Lemma 2.3 in [5]).

Remark 2. For the Riemann-Liouville derivative a Dxα , the relation

α α
a Dx a Ix f (x) = f (x), (α > 0) (56)

holds almost everywhere on [a, b] for f (x) ∈ Lp (a, b) (see Lemma 2.4 of [5]).
Remark 3. The Fundamental Theorem of Fractional Calculus (FTFC) uses the Riemann-
Liouville integration and the Caputo differentiation. The main property is that the Caputo

13
fractional derivative provides us an operation inverse to the Riemann-Liouville fractional in-
tegration from the left. It should be noted that consistent fractional generalizations of the
FTC, the differential vector operations and the integral theorems for other fractional integro-
differentiation such as Riesz, Grunvald-Letnikov, Weyl, Nishimoto are open problems.

Remark 4. In the theorem, we use 0 < α ≤ 1. As a result, we obtain the fractional

Green’s, Stokes’ and Gauss’s theorems for 0 < α < 1. Equation (49) is satisfied for α ∈ R+ .
The Newton-Leibniz formula (50) holds for 0 < α ≤ 1. For α > 1, we have (52). As a result,
to generalize the Green’s, Stokes’ and Gauss’s theorems for α ∈ R+ , we can use Eq. (52) in the
form
n−1
α C α
X 1
f (b) − f (a) = a Ib a Dx f (x) + (b − a)j f (j) (a), (n − 1 < α ≤ n), (57)
j=1
j!

where f (j) (x) = Dxj f (x). In particular, if 1 < α ≤ 2, then n = 2 and

f (b) − f (a) = a Ibα C α ′

a Dx f (x) + (b − a)f (a). (58)

Remark 5. In the FTFC, we use the left fractional integrals and derivatives. The Newton-
Leibniz formulas can be presented for the right fractional Riemann-Liouville integrals and the
right fractional Caputo derivatives in the form
n−1
αC α
X (−1)j f (j) (b)
x Ib x Db f (x) = f (x) − (b − x)j . (59)
j=0
j!

In particular, if 0 < α ≤ 1, then

αC α
x Ib x Db f (x) = f (x) − f (b). (60)

For α > 0, f (x) ∈ L∞ (a, b) or f (x) ∈ C[a, b], then

C α α
x Db x Ib f (x) = f (x). (61)

As a result, fractional generalization of differential operations and integral theorems can be

defined for the right integrals and derivatives as well as for the left ones.

14
5 Definition of fractional vector operations

5.1 Fractional operators

To define fractional vector operations, we introduce the operators that correspond to the frac-
tional derivatives and integrals.
We define the fractional integral operator
Z x
α ′ 1 dx′
I
a x [x ] := , (α > 0), (62)
Γ(α) a (x − x′ )1−α

which acts on a real-valued function f (x) ∈ L1 [a, b] by

Z x
α ′ ′ 1 f (x′ )dx′
a Ix [x ]f (x ) = . (63)
Γ(α) a (x − x′ )1−α

The Caputo fractional differential operator on [a, b] can be defined by

Z x
C α ′ 1 dx′ ∂n
a D x [x ] := , (n − 1 < α < n), (64)
Γ(n − α) a (x − x′ )1+α−n ∂x′ n

such that the Caputo derivatives for f (x) ∈ AC n [a, b] is written as

Z x
C α ′ ′ 1 dx′ ∂ n f (x′ )
a D x [x ]f (x ) = , (n − 1 < α < n). (65)
Γ(n − α) a (x − x′ )1+α−n ∂x′ n

It is easy to see that

C α ′
a Dx [x ] =a Ixn−α [x′ ]D n [x′ ], (n − 1 < α < n).

Using these notations, formulas (49) and (50) of the FTFC can be presented as

C α ′ α ′′ ′′
a Dx [x ] a Ix′ [x ]f (x ) = f (x), (α > 0), (66)

α C α ′ ′
a Ib [x] a Dx [x ]f (x ) = f (b) − f (a), (0 < α < 1). (67)

This form is more convenient than (49) and (50), since it allows us to take into account the
variables of integration and the domain of the operators.

15
5.2 Definition of fractional differential vector operations
Let us define the fractional differential operators (grad, div, curl) such that fractional gener-
alizations of integral theorems (Green’s, Stokes’, Gauss’) can be realized. We use the Caputo
derivatives to defined these operators and we use the Riemann-Liouville integrals in the gener-
alizations of the integral theorems.
Let W be a domain of R3 . Let f (x) and F(x) be real-valued functions that have continuous
derivatives up to order (n−1) on W , such that the (n−1) derivatives are absolutely continuous,
i.e., f, F ∈ AC n [W ]. We can define a fractional generalization of nabla operator by

∇αW = C
DαW = e1 C DW
α
[x] + e2 C DW
α
[y] + e3 C DW
α
[z], (n − 1 < α < n). (68)

Here, we use the fractional Caputo derivatives C DW

α
[xm ] with respect to coordinates xm . For
the parallelepiped
W := {a ≤ x ≤ b, c ≤ y ≤ d, g ≤ z ≤ h},

we have
C α C α C α C α C α C α
DW [x] = a Db [x], DW [y] = c Dd [y], DW [z] = g Dh [z]. (69)

The right-hand sides of these equations the Caputo derivatives are used.

(1) If f = f (x, y, z) is (n − 1) times continuously differentiable scalar field such that Dxn−1
l
f
is absolutely continuous, then we define its fractional gradient as the following

GradαW f = C
DαW f = el C DW
α
[xl ]f (x, y, z) =

= e1 C DW
α
[x]f (x, y, z) + e2 C DW
α
[y]f (x, y, z) + e3 C DW
α
[z]f (x, y, z). (70)

(2) If F(x, y, z) is (n − 1) times continuously differentiable vector field such that Dxn−1
l
Fl are
absolutely continuous, then we define its fractional divergence as a value of the expression
 
α C
DivW F= DαW , F = C α
DW [xl ]Fl (x, y, z) =

C α
= DW [x]Fx (x, y, z) + C DW
α
[y]Fy (x, y, z) + C α
DW [z]Fz (x, y, z). (71)

16
 (3) The fractional curl operator is defined by
h i
α C
DαW , F = el εlmk C DW
α C α C α


CurlW F= [xm ]Fk = e1 DW [y]Fz − DW [z]Fy +

C α C α C α C α



+ e2 DW [z]Fx − DW [x]Fz + e3 DW [x]Fy − DW [y]Fx , (72)

where Fk = Fk (x, y, z) ∈ AC n [W ], (k = 1, 2, 3).

Note that these fractional differential operators are nonlocal. As a result, the fractional
gradient, divergence and curl depend on the region W .

5.3 Relations for fractional differential vetor operations

(a) The first relation for the scalar field f = f (x, y, z) is

α
CurlW GradαW f = el εlmn C DW
α
[xm ] C DW
α
[xn ]f = 0, (73)

where εlmn is Levi-Civita symbol, i.e. it is 1 if (i, j, k) is an even permutation of (1, 2, 3), (−1)
if it is an odd permutation, and 0 if any index is repeated.
(b) The second relation,
3
X
α
DivW GradαW f (x, y, z) = C α
DW [xl ] C DW
α
[xl ]f (x, y, z) = ( C DW
α
[xl ])2 f (x, y, z). (74)
l=1

Using notation (68),

 
α
DivW GradαW = ( C DαW )2 = C
DαW , C DαW . (75)

In the general case,

( C DW
α
[xl ])2 6= C 2α
DW [xl ]. (76)

It is obvious from
(C α 2 n−α n
a Dx ) = a I x Dx a Ixn−α Dxn =
n−α n−α n n
= a Ix a Ix Dx Dx + a Ixn−α [Dxn , a Ixn−α ]Dxn = 2α
a Dx + a Ixn−α [Dxn , a Ixn−α ]Dxn ,

where
[Dxn , a Ixn−α ] := Dxn a Ixn−α − a Ixn−α Dxn = α
a Dx −C α
a Dx 6= 0.

17
 (c) It is easy to prove the following relation,

α α C α
DivW CurlW F(x, y, z) = DW [xl ]εlmn C DW
α
[xm ]Fn (x, y, z) =

= εlmn C DW
α
[xl ] C DW
α
[xm ]Fn (x, y, z) = 0, (77)

where we use antisymmetry of εlmn with respect to m and n.

(d) There exists a relation for the double curl operation in the form

α α
CurlW CurlW F(x, y, z) = el εlmn C DW
α
[xm ]εnpq C DW
α
[xp ]Fq (x, y, z) =

= el εlmn εnpq C DW
α
[xm ] C DW
α
[xp ]Fq (x, y, z). (78)

Using
εlmn εlpq = δmp δnq − δmq δnp , (79)

we get
α α
CurlW CurlW F(x, y, z) = GradαW DivW
α
F(x, y, z) − ( C DαW )2 F(x, y, z), (80)

(e) In the general case,

     
C α ′
a Dx [x ] f (x′ )g(x′ ) 6= C
a D α ′
x [x ]f (x′
) g(x)+ C α ′
a D x [x ]g(x′
) f (x). (81)

For example (see Theorem 15.1. from [2]),

∞
α ′

′ ′
 X Γ(α + 1) 
α−j ′
 
a Dx [x ] f (x )g(x ) = a Dx [x ]f (x′ ) Dxj g(x) , (82)
j=0
Γ(j + 1)Γ(α − j + 1)

if f (x) and g(x) are analytic functions on [a, b]. As a result, we have
     
GradαW f g 6= GradαW f g + GradαW g f, (83)
   
α
DivW f F 6= GradαW f, F + f DivW
α
F. (84)

These relations state that we cannot use the Leibniz rule in a fractional generalization of the
vector calculus.

18
5.4 Fractional integral vector operations
In this section, we define fractional generalizations of circulation, flux and volume integral.
Let F = F(x, y, z) be a vector field such that

F(x, y, z) = e1 Fx (x, y, z) + e2 Fy (x, y, z) + e3 Fz (x, y, z).

If Fx , Fy , Fx are absolutely integrable real-valued functions on R3 , i.e., Fx , Fy , Fx ∈ L1 (R3 ),

then we can define the following fractional integral vector operations of order α > 0.
(1) A fractional circulation is a fractional line integral along a line L that is defined by
 
EL (F) = IL , F = ILα [x]Fx + ILα [y]Fy + ILα [z]Fz .
α α
(85)

For α = 1, we get
  Z   Z
EL1 (F) = I1L , F = dL, F = (Fx dx + Fy dy + Fz dz), (86)
L L
where dL = e1 dx + e2 dy + e3 dz.

(2) A fractional flux of the vector field F across a surface S is a fractional surface integral
of the field, such that
 
ΦαS (F) = IαS , F = ISα [y, z]Fx + ISα [z, x]Fy + ISα [x, y]Fz . (87)

For α = 1, we get
  Z Z   Z Z
Φ1S (F) = I1S , F = dS, F = (Fx dydz + Fy dzdx + Fz dxdy), (88)
S S
where dS = e1 dydz + e2 dzdx + e3 dxdy.

(3) A fractional volume integral is a triple fractional integral within a region W in R3 of a

scalar field f = f (x, y, z),

VWα (f ) = IW
α α
[x, y, z]f (x, y, z) = IW α
[x]IW α
[y]IW [z]f (x, y, z). (89)

For α = 1, we have
Z Z Z Z Z Z
VW1 (f ) := dV f (x, y, z) = dxdydz f (x, y, z). (90)
W W
This is the usual volume integral for the function f (x, y, z).

19
6 Fractional Green’s formula
Green’s theorem gives the relationship between a line integral around a simple closed curve ∂W
and a double integral over the plane region W bounded by ∂W . The theorem statement is the
following. Let ∂W be a positively oriented, piecewise smooth, simple closed curve in the plane
and let W be a region bounded by ∂W . If Fx and Fy have continuous partial derivatives on an
open region containing W , then
Z   Z Z  
Fx dx + Fy dy = Dy Fx − Dx Fy dxdy. (91)
∂W W

A fractional generalization of the Green’s formula (91) is presented by the following statement.

Theorem (Fractional Green’s Theorem for a Rectangle)

Let Fx (x, y) and Fy (x, y) be absolutely continuous (or continuously differentiable) real-valued
functions in a domain that includes the rectangle

W := {(x, y) : a ≤ x ≤ b, c ≤ y ≤ d}. (92)

Let the boundary of W be the closed curve ∂W . Then

α α α C α
[y]Fx (x, y) − C D∂W
α


I∂W [x]Fx (x, y) + I∂W [y]Fy (x, y) = IW [x, y] D∂W [x]Fy (x, y) , (93)

where 0 < α ≤ 1.

Proof
α
To prove equation (93), we change the double fractional integral IW [x, y] to the repeated frac-
α α
tional integrals IW [x] IW [y], and then employ the Fundamental Theorem of Fractional Calculus.
Let W be the rectangular domain (92) with the sides AB, BC, CD, DA, where the points
A, B, C, D have coordinates

A(a, c), B(a, d), C(b, d), D(b, c).

These sides form the boundary ∂W of W .

20
 For the rectangular region W defined by a ≤ x ≤ b, c ≤ y ≤ d, the repeated integral is

α α
IW [x] IW [y] = a Ibα [x] c Idα [y],

and equation (93) is

α
a Ib [x] (Fx (x, d) − Fx (x, c)) + c Idα [y] (Fy (a, y) − Fy (b, y)) =

α α C α ′ ′ C α ′ ′


= a Ib [x] c Id [y] c Dy [y ]Fx (x, y ) − a Dx [x ]Fy (x , y) . (94)

To prove of the fractional Green’s formula, we realize the following transformations

 
Iα∂W , F = I∂W
α α
[x]Fx + I∂W α
[y]Fy = IBC α
[x]Fx + IDA α
[x]Fx + IAB α
[y]Fy + ICD [y]Fy =

= a Ibα [x]Fx (x, d) − a Ibα [x]Fx (x, c) + c Idα [y]Fy (a, y)dy − c Idα [y]Fy (b, y) =

= a Ibα [x] [Fx (x, d) − Fx (x, c)] + c Idα [y] [Fy (a, y) − Fy (b, y)]. (95)

The main step of the proof of Green’s formula is to use the fractional Newton-Leibniz formula

Fx (x, d) − Fx (x, c) = c Idα [y] cC Dyα [y ′ ] F (x, y ′),

Fy (a, y) − Fy (b, y) = − a Ibα [x] Ca Dxα [x′ ] F (x′ , y). (96)

As a result, expression (95) can be presented as

n o n o
α α C α ′ ′ α α C α ′ ′
a Ib [x] I
c d [y] c D y [y ] Fx (x, y ) + I
c d [y] − I
a b [x] a D x [x ] Fy (x , y) =
 
α α C α ′ ′ C α ′ ′
= a Ib [x] c Id [y] c Dy [y ] Fx (x, y ) − a Dx [x ] Fy (x , y) =
 
α C α ′ ′
= IW [x, y] c Dy [y ] Fx (x, y ) −C
a D α ′
x [x ] Fy (x′
, y) .

This is the left-hand side of Eq. (94). This ends the proof.

Remark 1. In this fractional Green’s theorem, we use the rectangular region W . If the
region can be approximated by a set of rectangles, the fractional Green’s formula can also be
proved. In this case, the boundary ∂W is presented by paths each consisting of horizontal and

21
vertical line segments, lying in W .

Remark 2. To define the double integral and the theorem for nonrectangular regions R, we
can consider the function f(x, y), that is defined in the rectangular region W such that R ⊂ W
and 
F(x, y), (x, y) ∈ R;

f(x, y) = (97)
0,

(x, y) ∈ W/R.
As a result, we define a fractional double integral over the nonrectangular region R, through
the fractional double integral over the rectangular region W :

IαR [x, y] F(x, y) = IαW [x, y] f(x, y). (98)

To define double integrals over nonrectangular regions, we can use a fairly general method
to calculate them. For example, we can do this for special regions called elementary regions.
Let R be a set of all points (x, y) such that

a ≤ x ≤ b, ϕ1 (x) ≤ y ≤ ϕ2 (x).

Then, the double integrals for such regions can be calculated by

IαR [x, y] F (x, y) = a Iαb [x] α

ϕ1 (x) Iϕ2 (x) [y] F (x, y). (99)

It is easy to consider the following examples.

1) ϕ1 (x) = 0, y = ϕ2 (x) = x2 , F (x, y) = x + y.
2) ϕ1 (x) = x3 , ϕ1 (x) = x2 , F (x, y) = x + y.
3) ϕ1 (x) = 0, y = ϕ2 (x) = x, F (x, y) = xy.
The fractional integrals can be calculated by using the relations
α Γ(β + 1)
a Ix [x](x − a)β = (x − a)β+α , (100)
Γ(β + α + 1)
where α > 0, β > 0. For other relations see Table 9.1 in [2]. To calculate the Caputo derivatives,
we can use this table and the equation
n−1
C α ′ ′ α ′ ′
X f (k) (a)
a Dx [x ]f (x ) = a Dx [x ]f (x ) − , n − 1 < α ≤ n. (101)
k=0
Γ(k − α + 1)

22
Note that the Mittag-Leffler function Eα [(x′ − a)α ] is not changed by the Caputo derivative

C α ′ ′
a Dx [x ] Eα [(x − a)α ] = Eα [(x − a)α ]. (102)

This equation is a fractional analog of the well-known property of exponential function of the
form Dx1 exp(x − a) = exp(x − a). Therefore the Mittag-Leffler function can be considered as
a fractional analog of exponential function.

7 Fractional Stokes’ formula

We shall restrict ourselves to the consideration of a simple surface. If we denote the boundary
of the simple surface W by ∂W and if F is a smooth vector field defined on W , then the Stokes’
theorem asserts that Z   Z  
F, dL = curlF, dS . (103)
∂W W

The right-hand side of this equation is the surface integral of curlF over W , whereas the left-
hand side is the line integral of F over the line ∂W . Thus the Stokes’ theorem is the assertion
that the line integral of a vector field over the boundary of the surface W is the same as the
integral over the surface of the curl of F.
For Cartesian coordinates, Eq. (103) gives
Z  
Fx dx + Fy dy + Fz dz =
∂W
Z Z  
= dydz [Dy Fz − Dz Fy ] + dzdx [Dz Fx − Dx Fz ] + dxdy [Dx Fy − Dy Fx ] . (104)
W

Let F = F(x, y, z) be a vector field such that

F(x, y, z) = e1 Fx (x, y, z) + e2 Fy (x, y, z) + e3 Fz (x, y, z).

where Fx , Fy , Fx are absolutely continuous (or continuously differentiable) real-valued functions

on R3 . Then the fractional generalization of the Stokes’ formula (104) can be written as
   
α
Iα∂W , F = IαW , Curl∂W F . (105)

23
Here we use the notations

IαL = Iα∂W = em I∂W

α α
[xm ] = e1 I∂W α
[x] + e2 I∂W α
[y] + e3 I∂W [z], (106)

such that
 
α α α α
I∂W , F = I∂W [x]Fx + I∂W [y]Fy + I∂W [z]Fz . (107)

The integral (106) can be considered as a fractional line integral.

In the right-hand side of (105), IαW is a fractional surface integral over S = W such that

IαS = IαW = e1 IW
α α
[y, z] + e2 IW α
[z, x] + e3 IW [x, y]. (108)

The fractional curl operation is

α
F = el εlmn C DW
α C α C α


CurlW [xm ]Fn = e1 DW [y]Fz − DW [z]Fy +

C α C α C α C α



+ e2 DW [z]Fx − DW [x]Fz + e3 DW [x]Fy − DW [y]Fx . (109)

For α = 1, equation (109) gives the well-known expression

1
CurlW F = curlF = el εlmn Dxm Fn = e1 (Dy Fz − Dz Fy ) +

+ e2 (Dz Fx − Dx Fz ) + e3 (Dx Fy − Dy Fx ) . (110)

The right-hand side of Eq. (105) means

 
IαW , CurlW
α α C α C α


F = IW [y, z] DW [y]Fz − DW [z]Fy +

α C α C α α C α C α



+ IW [z, x] DW [z]Fx − DW [x]Fz + IW [x, y] DW [x]Fy − DW [y]Fx . (111)

This integral can be considered as a fractional surface integral.

8 Fractional Gauss’s formula

Let us give the basic theorem regarding the Gauss’s formula in a fractional case.

24
 Theorem (Fractional Gauss’s Theorem for a Parallelepiped)
Let Fx (x, y, z), Fy (x, y), Fz (x, y, z) be continuously differentiable real-valued functions in a do-
main that includes the parallelepiped

W := {(x, y, z) : a ≤ x ≤ b, c ≤ y ≤ d, g ≤ z ≤ h}. (112)

If the boundary of W be a closed surface ∂W , then

 
Iα∂W , F = IW
α α
DivW F. (113)

This equation can be called the fractional Gauss’s formula.

Proof.
For Cartesian coordinates, we have the vector field F = Fx e1 + Fy e2 + Fz e3 , and

α α
IW = IW [x, y, z], Iα∂W = e1 I∂W
α α
[y, z] + e2 I∂W α
[x, z] + e3 I∂W [x, y]. (114)

Then
 
Iα∂W , F = I∂W
α α
[y, z]Fx + I∂W α
[x, z]Fy + I∂W [x, y]Fz , (115)

and
 
α α α C α C α C α
IW DivW F = IW [x, y, z] D∂W [x]Fx + D∂W [y]Fy + D∂W [z]Fz . (116)

If W is the parallelepiped

W := {a ≤ x ≤ b, c ≤ y ≤ d, g ≤ z ≤ h}, (117)

then the integrals (114) are

α
IW [x, y, z] = a Ibα [x] c Idα [y] g Ihα [z], (118)

and
α
I∂W [y, z] = c Idα [y] g Ihα [z], (119)
α
I∂W [x, z] = a Ibα [x] g Ihα [z], (120)

25
 α
I∂W [x, y] = a Ibα [x] c Idα [y]. (121)

As a result, we can realize the following transformations

 
Iα∂W , F = I∂W
α α
[y, z]Fx + I∂W α
[z, x]Fy + I∂W [x, y]Fz =
n o n o
α α α α
= c Id [y] g Ih [z] Fx (b, y, z) − Fx (a, y, z) + a Ib [x] g Ih [z] Fy (x, d, z) − Fy (x, c, z) +
n o
+ a Ibα [x]c Idα [y] Fz (x, y, g) − Fz (x, y, h) =
n o
= a Ibα [x] c Idα [y] g Ihα [z] Ca D α ′
x [x ]F x (x ′
, y, z) + C α ′
c D y [y ]Fy (x, y ′
, z) + C α ′
g D z [z ]Fz (x, y, z ′
) =
 
α C α α α
= IW D W , F = IW DivW F.

This ends the proof of the fractional Gauss’s formula for parallelepiped region.

Remark. To define the triple integral and the theorem for non-parallelepiped regions R,
we consider the function f (x, y, z), that is defined in the parallelepiped region W such that
R ⊂ W , such that 
F (x, y, z), (x, y, z) ∈ R;

f (x, y, z) = (122)
0,

(x, y, z) ∈ W/R.
Then we have
IαR [x, y, z]F (x, y, z) = IαW [x, y, z]f (x, y, z). (123)

As a result, we define a fractional triple integral over the non-parallelepiped region R, through
the fractional triple integral over the parallelepiped region W .

9 Fractional differential forms

9.1 Brief description of different approaches

A fractional generalization of differential has been presented by Ben Adda in [16, 17]. A
fractional generalization of the differential forms has been suggested by Cottrill-Shepherd and

26
Naber in [22] (see also [23, 24]). The application of fractional differential forms to dynamical
systems are considered in [26, 27]. Fractional integral theorems are not considered.
(1) In the papers [16, 17], the fractional differential for analytic functions is defined by
1
dα f = dxj Nxαj f (x), (124)
Γ(1 + α)
where Nxαj are Nishimoto fractional derivatives [8] (see also Section 22 of [2]), which is a gener-
alization of the Cauchy’s differentiation formula.
(2) In the paper [22] (see also [23, 26, 24]), an exterior fractional differential is defined
through the Riemann-Liouville derivatives by
n
X
α
d = (dxj )α 0 Dxαj . (125)
j=1

In two dimensions n = 2,
dα = (dx)α 0 Dxα + (dy)α 0 Dyα ,

suhc that
x1−α xy −α
dα x = (dx)α + (dy)α , (126)
Γ(2 − α) Γ(1 − α)
yx−α y 1−α
dα y = (dx)α + (dy)α , (127)
Γ(1 − α) Γ(2 − α)
where we use
α
x−α
j
0 Dx j 1 = . (128)
Γ(1 − α)
(3) In the paper [27], an exterior fractional differential is defined through the fractional
Caputo derivatives in the form
n
X
dα = (dxj )α C α
0 Dx j . (129)
j=1

For two dimensions (x, y), we have

dα = (dx)α C α αC α
0 Dx + (dy) 0 Dy ,

such that
x1−α
dα x = (dx)α , (130)
Γ(2 − α)

27
 α y 1−α α
d y = (dy) , (131)
Γ(2 − α)
(compare with (126) and (127)). Equation (129) can be rewritten as
n
X
α
d = Γ(2 − α)xα−1
j dα xj C α
0 Dx j . (132)
j=1

This relation is used in [27] as a fractional exterior differential.

9.2 Definition of a fractional exterior differential

A definition of fractional differential forms must be correlated with a possible generalization of
the fractional integration of differential forms. To derive fractional analogs of differential forms
and its integrals, we consider a simplest case that is an exact 1-form on the interval L = [a, b].
It allows us to use the fractional Newton-Leibniz formula.
In order to define an integration of fractional differential forms, we can use the fractional
Riemann-Liouville integrals. Then a fractional exterior derivative must be defined through the
Caputo fractional derivative.
Equation (67) of FTFC means that
Z x
dx′ C α
′ )1−α a
Dx′ [x′′ ]f (x′′ ) = f (x) − f (a), (0 < α < 1). (133)
a Γ(α)(x − x
Using
dx′ = sgn(dx′)|dx′ | = sgn(dx′ )|dx′ |1−α |dx′ |α , (0 < α < 1),

equation (133) can be presented in the form

Z x
|dx′ |1−α 
′ ′ αC α ′′ ′′

′ 1−α
sgn(dx )|dx | a Dx′ [x ]f (x ) = f (x) − f (a), (0 < α < 1). (134)
a Γ(α)(x − x )

The expression in the big brackets of (134) can be considered as a fractional differential of the
function f (x). As a result, we have

IˆLα [x] a dαx f (x) = f (b) − f (a), (0 < α < 1), (135)

where L = [a, b], and the fractional integration for differential forms is defined by the operator
Z b
ˆα |dx|1−α
IL [x] := 1−α
. (136)
a Γ(α)(b − x)

28
The exact fractional differential 0-form is a fractional differential of the function

α
a dx f (x) := (dx)α C α ′ ′
a Dx [x ]f (x ). (137)

Equation (135) can be considered as a fractional generalization of the integral for differential
1-form.
As a result, the fractional exterior derivative is defined as

α
a dx := [dxm ]α C α ′
a Dxm [xm ]. (138)

where
[dxm ]α = sgn(dxm )|dxm |α

Then the fractional differential 1-form is

ω(α) = [dxm ]α Fm (x). (139)

The exterior derivative of this form gives

α
a dx ω(α) = [dxm ]α ∧ [dxn ]α aC Dxαn [x′ ]Fm (x′ ). (140)

To prove the proposition (140), we use the rule

∞
X
Dxα (f g) = (αk ) ( C α−s s
a Dx f )Dx g,
s=0

and the relation [5] D s [x](dx)α = 0 (s ≥ 1), for integer s, where

(−1)k−1 αΓ(k − α)
(αk ) = .
Γ(1 − α)Γ(k + 1)

For example, we have

∞
X
α α
d [[dxm ] Fm ] = [dxn ]α ∧ (αk ) ( aC Dxα−s
n
[x′ n ]Fm (x′ ))D s [xn ][dxm ]α =
s=0

= [dxn ]α ∧ [dxm ]α (α0 ) C Dxαn [x′ n ]Fm (x′ ) = C

Dxαn [x′ n ]Fm (x′ ) [dxn ]α ∧ [dxm ]α .

29
 Using the equation (see Property 2.16 in [5])

α ′ ′ Γ(β + 1)
a Dx [x ](x − a)β = (x − a)β−α , (141)
Γ(β + 1 − α)

where n − 1 < α < n, and β > n − 1, and

α ′ ′
a Dx [x ](x − a)k = 0 (k = 0, 1, 2, ..., n − 1), (142)

we obtain
α
a dx (x − a)α = [dx]α C α ′ ′ α
a Dx [x ]x = (dx) Γ(α + 1) (x > a). (143)

Then
1
[dx]α = α α
a d (x − a) , (144)
Γ(α + 1) x
and the fractional exterior derivative (138) is presented as

α 1 α αC α ′
a dx := a d (xm − am ) a Dxm [xm ]. (145)
Γ(α + 1) x

The fractional differential 1-form (139) can be written as

1 α α
ω(α) = a dx (xm − am ) Fm (x). (146)
Γ(α + 1)

Remark. Using the suggested definition of fractional integrals and differential forms, it is
possible to define a fractional integration of n-form over the hypercube [0, 1]n . Unfortunately,
a generalization of this fractional integral, which uses the mapping φ of the region W ⊂ Rn
into [0, 1]n , has a problem. For the integer case, we use the equation

Dx1 f (φ(x)) = (Dφ1 f )(Dx1 φ). (147)

For the fractional case, the chain rule for differentiation (the fractional derivative of composite
functions) is more complicated (see Section 2.7.3. [4]). As a result, a consistent definition of
fractional integration of differential form for arbitrary manifolds is an open question.

30
9.3 Differential vector operations through the differential forms
To define a fractional divergence of the field F, we can consider the 2-form

ω2 = Fz dx ∧ dy + Fy dz ∧ dx + Fx dy ∧ dz. (148)

Then the fractional exterior derivative of this form is

dω2 = (Dx Fx + Dy Fy + Dz Fz )dx ∧ dy ∧ dz = divF dx ∧ dy ∧ dz. (149)

To define a fractional generalization of the curl operation for F, we can use the 1-form

ω1 = Fx dx + Fy dy + Fz dz. (150)

Then the fractional exterior derivative of this 1-form is

dω1 = (Dx Fy − Dy Fx )dx ∧ dy + (Dy Fz − Dz Fy )dy ∧ dz + (Dx Fz − Dz Fx )dx ∧ dz. (151)

To define the fractional gradient, we consider the 0-form

ω0 = f (x, y, z) (152)

Then the fractional exterior derivative of f gives

3
X
dω0 = Dx f dx + Dy f dy + Dz f dz = (gradf )k dxk . (153)
k=1

It is not hard to obtain fractional generalizations of these definitions.

10 Fractional nonlocal Maxwell’s equations

10.1 Local Maxwell’s equations

The behavior of electric fields (E, D), magnetic fields (B, H), charge density (ρ(t, r)), and
current density (j(t, r)) is described by the Maxwell’s equations

div D(t, r) = ρ(t, r), (154)

31
 curl E(t, r) = −∂t B(t, r), (155)

div B(t, r) = 0, (156)

curl H(t, r) = j(t, r) + ∂t D(t, r). (157)

Here r = (x, y, z) is a point of the domain W . The densities ρ(t, r) and j(t, r) describe an
external sources. We assume that the external sources of electromagnetic field are given.
The relations between electric fields (E, D) for the medium can be realized by
Z +∞
D(t, r) = ε0 ε(r, r ′)E(t, r ′)dr ′ , (158)
−∞

where ε0 is the permittivity of free space. Homogeneity in space gives ε(r, r ′) = ε(r − r ′ ).
Equation (158) means that the displacement D is a convolution of the electric field E at other
space points. A local case corresponds to the Dirac delta-function permittivity ε(r) = εδ(r).
Then Eq. (158) gives D(t, r) = ε0 εE(t, r).
Analogously, we have nonlocal equation for the magnetic fields (B, H).

10.2 Caputo derivative in electrodynamics

Let us demonstrate a possible way of appearance of the Caputo derivative in the classical
electrodynamics. If we have
Z +∞
D(t, x) = ε(x − x′ )E(t, x′ )dx′ , (159)
−∞

then
Z +∞ Z +∞
Dx1 D(t, x) = [Dx1 ε(x ′ ′
− x )]E(t, x )dx = − ′
[Dx1′ ε(x − x′ )]E(t, x′ )dx′ . (160)
−∞ −∞

Using the integration by parts, we get

Z +∞
Dx1 D(t, x) = ε(x − x′ )Dx1′ E(t, x′ ) dx′ . (161)
−∞

Consider the kernel ε(x − x′ ) of integral (161) in the interval (0, x) such that

e(x − x′ ), 0 < x′ < x;

′
ε(x − x ) = (162)
0, x′ > x, x′ < 0,


32
with the power-like function
1 1
e(x − x′ ) = , (0 < α < 1). (163)
Γ(1 − α) (x − x′ )α
Then Eq. (161) gives the relation

Dx1 D(t, x) = C α
0 Dx E(t, x), (0 < α < 1) (164)

C α
with the Caputo fractional derivatives 0 Dx .

10.3 Fractional nonlocal Maxwell’s equations

Fractional nonlocal differential Maxwell’s equations have the form

α1
DivW E(t, r) = g1 ρ(t, r), (165)

α2
CurlW E(t, r) = −∂t B(t, r), (166)
α3
DivW B(t, r) = 0, (167)
α4
g2 CurlW B(t, r) = j(t, r) + g3−1∂t E(t, r), (168)

where αs , (s = 1, 2, 3, 4), can be integer or fractional.

Fractional integral Maxwell’s equations, which use integrals of noninteger orders, have been
suggested in [25] to describe fractional distributions of electric charges and currents.
In the general form, the fractional integral Maxwell’s equations can be presented in the form
 
α1 α1
I∂W , E(t, r) = g1 IW ρ(t, r), (169)
  d  α2 
Iα∂S2 , E(t, r) = − IS , B(t, r) , (170)
dt
 
Iα∂W
3
, B(t, r) = 0, (171)
    d  α4 
g2 Iα∂S4 , B(t, r) = IαS4 , j(t, r) + g3−1 IS , E(t, r) . (172)
dt
These fractional differential and integral equations can be used to describes an electromag-
netic field of media that demonstrate fractional nonlocal properties. The suggested equations
can be considered as a special case of nonlocal electrodynamics (see [31, 32, 33, 34, 35]).

33
 Fractional coordinate derivatives are connected with nonlocal properties of the media. For
example, a power-law long-range interaction in the 3-dimensional lattice in the continuous limit
can give a fractional equation [40].

10.4 Fractional conservation law for electric charge

Let us derive a conservation law equation for density of electric charge in the region W from
the fractional nonlocal Maxwell’s equations.
The time derivative of (169) is
α1
DivW ∂t E(t, r) = g1 ∂t ρ(t, r). (173)

Substitution of (168) into (173) gives

 
α1 α4
g3 DivW g2 CurlW B(t, r) − j(t, r) = g1 ∂t ρ(t, r). (174)

If α1 = α4 , then
α1 α4
DivW CurlW B(t, r) = 0, (175)

and we have the law

α1
g1 ∂t ρ(t, r) + g3 DivW j(t, r) = 0. (176)

This fractional equation is a differential form of charge conservation law for fractional nonlocal
electrodynamics.
If α1 = α4 , we can define the fractional integral characteristics such as
α1
QW (t) = g1 IW [x, y, z]ρ(t, x, y, z), (177)

which can be called the total fractional nonlocal electric charge, and
   
J∂W (t) = g3 Iα∂W
1 α1
, j = g3 I∂W α1
[y, z]jx + I∂W α1
[z, x]jy + I∂W [x, y]jz (178)

is a fractional nonlocal current. Then the fractional nonlocal conservation law is

d
QW (t) + J∂W (t) = 0. (179)
dt
This integral equation describes the conservation of the electric charge in the nonlocal electro-
dynamics for the case α1 = α4 .

34
10.5 Fractional waves
Let us derive wave equations for electric and magnetic fields in a region W from the fractional
nonlocal Maxwell’s equations with j = 0 and ρ = 0.
The time derivative of Eq. (166) is

α2
∂t2 B = −CurlW ∂t E (180)

Substitution of (172) and j = 0 into (180) gives

α2 α4
∂t2 B = −g2 g3 CurlW CurlW B(t, r). (181)

Using (80) and (167) for α2 = α3 = α4 , we get

∂t2 B = g2 g3 ( C DαW )2 B. (182)

As a result, we obtain
∂t2 B − v 2 ( C DαW )2 B = 0, (183)

where v 2 = g2 g3 . This is the fractional wave equation for the magnetic field B. Analogously,
Eqs. (166) and (172) give the fractional wave equation for electric field

∂t2 E − v 2 ( C DαW )2 E = 0. (184)

The solution B(t, r) of equation (183) is a linear combination of the solutions B+ (t, r) and
B− (t, r) of the equations
∂t B+ (t, r) − v C DαW B+ (t, r) = 0, (185)

∂t B− (t, r) + v C DαW B− (t, r) = 0. (186)

As a result, we get the fractional extension of D’Alembert expression that is considered in [36].
For the boundary conditions

lim B(t, r) = 0, B(t, 0) = G(t), (187)

|t|→∞

the general solution of equations (185) and (186) is given [5] by

Z +∞
1
Bm± (t, r) = dωEα,1 [∓ivωxαm ] G̃m (ω)e−iωt , (188)
2π −∞

35
where G̃m (ω) = F [Gm (t)], and Eα,β [z] is the biparametric Mittag-Leffler function [5]. Here
B±m (t, r), and Gm (t) are components of B± (t, r) and G(t).
For one-dimensional case, Bx (x, y, z, t) = u(x, t), By = Bz = 0, and we can consider the
fractional partial differential equation

Dt2 u(x, t) − v 2 0 Dx2α u(x, t) = 0, x ∈ R, x > 0, v > 0, (189)

with the conditions

Dxk u(0, t) = fk (t), (190)

where k = 0 for 0 < α ≤ 1/2, and k = 1 for 1/2 < α ≤ 1. If 0 < 2α < 2 and v > 0, the system
of equations (189), (190) is solvable (Theorem 6.3. of [5]), and the solution u(x, t) is given by
n−1 Z +∞
X
u(x, t) = G2α
k (y, t)fk (y)dy, (n − 1 < α ≤ n), (191)
k=0 −∞

where
1 k−α
G2α
k (x, t) = vx φ(−α, k + 1 − α, v|t|x−α ). (192)
2
Here φ(−α, k + 1 − α, v|t|x−α) is the Wright function [5].
Note that the solutions of equations as (185) and (186) are based primary on the use
C α
of Laplace transforms for equations with the Caputo 0 Dx derivatives. This leaves certain
C α
problems [5] with the fractional derivatives a Dx for a ∈ R.

11 Conclusion
Let us note some possible extensions of the fractional vector calculus.

(1) It is very important to prove the suggested fractional integral theorems for a general
form of domains and boundaries.
(2) It is interesting to generalize the formulations of fractional integral theorems for α > 1.
(3) A proof of fractional theorems for differential forms can be interesting to formulate a
fractional generalization of differential geometry.

36
 In the fundamental theorem of fractional calculus (FTFC) we use the Riemann-Liouville
integration and the Caputo differentiation. The main property is that the Caputo fractional
derivative provides us an operation inverse to the Riemann-Liouville fractional integration from
the left. Note that a fractional generalization of the differential vector operations and the
integral theorems for the fractional integro-differentiation of Riesz, Grunvald-Letnikov, Weyl,
Nishimoto is an open problem.
There are the following possible applications of the fractional variational calculus (FVC).

(a) A fractional nonlocal electrodynamics that is characterized by the power law non-locality
can be formulated by using the FVC.
(b) Nonlocal properties in classical dynamics can be described by the FVC and by possible
fractional generalizations of symplectic geometry and Poisson algebra. In general, fractional
differential forms and fractional integral theorems for these forms can be used to describe
classical dynamics.
(c) A possible dynamics of fractional gradient and Hamiltonian dynamical systems can be
described by the FVC.
(d) The continuum mechanics of fluids and solids with nonlocal properties (with a nonlocal
interaction of medium particles) can be described by the FVC.

The fractional derivatives in equations can be connected with a long-range power-law inter-
action of the systems [37, 38, 40]. The nonlocal properties of electrodynamics can be considered
[39] as a result of dipole-dipole interactions with a fractional power-law screening that is con-
nected with the integro-differentiation of non-integer order. For noninteger derivatives with
respect to coordinates, we have the power-like tails as the important property of the solutions
of the fractional equations.

37
References
[1] K.B. Oldham, J. Spanier, The Fractional Calculus: Theory and Applications of Differen-
tiation and Integration to Arbitrary Order (Academic Press, New York, 1974).
[2] S.G. Samko, A.A. Kilbas, O.I. Marichev, Fractional Integrals and Derivatives Theory and
Applications (Gordon and Breach, New York, 1993).
[3] K. Miller, B. Ross, An Introduction to the Fractional Calculus and Fractional Differential
Equations (Wiley, New York, 1993).
[4] I. Podlubny, Fractional Differential Equations (Academic Press, San Diego, 1999).
[5] A.A. Kilbas, H.M. Srivastava, J.J. Trujillo, Theory and Application of Fractional Differ-
ential Equations (Elsevier, Amsterdam, 2006).
[6] B. Ross, ”A brief history and exposition of the fundamental theory of fractional calculus”
Lect. Notes Math. 457 (1975) 1-36.
[7] V. Kiryakova, Generalized Fractional Calculus and Applications (Longman, Harlow, 1993).
B. Rubin, Fractional Integrals and Potentials (Longman, Harlow, 1996).
A.C. McBride, Fractional Calculus and Integral Transforms of Generalized Functions (Pit-
man Press, San Francisco, 1979).
[8] K. Nishimoto, Fractional Calculus: Integrations and Differentiations of Arbitrary Order
(University of New Haven Press, New Haven, 1989).
[9] G.M. Zaslavsky, Hamiltonian Chaos and Fractional Dynamics (Oxford University Press,
Oxford, 2005).
[10] A. Carpinteri, F. Mainardi, (Eds.), Fractals and Fractional Calculus in Continuum Me-
chanics (Springer, Wien, 1997).
[11] B. West, M. Bologna, P. Grigolini, Physics of Fractal Operators (Springer, New York,
2003).
[12] R. Hilfer (Ed.), Applications of Fractional Calculus in Physics (World Scientific, Singapore,
2000).
[13] G.M. Zaslavsky, ”Chaos, fractional kinetics, and anomalous transport” Phys. Rep. 371
(2002) 461-580.

38
[14] E.W. Montroll, M.F. Shlesinger, ”The wonderful world of random walks” in: Studies in
Statistical Mechanics, Vol. 11. J. Lebowitz, E. Montroll (Eds.), (North-Holland, Amster-
dam, 1984) pp.1-121.
[15] R. Metzler, J. Klafter, ”The random walk’s guide to anomalous diffusion: a fractional
dynamics approach” Phys. Rep. 339 (2000) 1-77;
”The restaurant at the end of the random walk: recent developments in the description of
anomalous transport by fractional dynamics” J. Phys. A 37 (2004) R161-R208.
[16] F. Ben Adda, ”Geometric interpretation of the differentiability and gradient of real order”
Comptes Rendus de l’Academie des Sciences - Series I - Mathematics 326 (1998) 931-934
(in French);
”Geometric interpretation of the fractional derivative” Journal of Fractional Calculus 11
(1997) 21-52.
[17] F. Ben Adda, ”The differentiability in the fractional calculus” Nonlinear Analysis 47 (2001)
5423-5428;
”The differentiability in the fractional calculus” Comptes Rendus de l’Academie des Sci-
ences - Series I - Mathematics 326 (1998) 787-791.
[18] N. Engheta, ”Fractional curl operator in electromagnetics” Microwave and Optical Tech-
nology Letters 17 (2) (1998) 86-91.
[19] E.I. Veliev, N. Engheta, ”Fractional curl operator in reflection problems” 10th International
Conference on Mathematical Methods in Electromagnetic Theory, Sept. 14-17, Ukraine,
IEEE (2004) 228-230;
M.V. Ivakhnychenko, E.I. Veliev, ”Fractional curl operator in radiation problems” 10th
International Conference on Mathematical Methods in Electromagnetic Theory. Sept. 14-
17, Ukraine, IEEE (2004) 231-233.
[20] Q.A. Naqvi, M. Abbas, ”Complex and higher order fractional curl operator in electromag-
netics” Optics Communications 241 (2004) 349-355;
A. Hussain, Q.A. Naqvi, ”Fractional curl operator in chiral medium and fractional non-
symmetric transmission line” Progress in Electromagnetics Research 59 (2006) 199-213;

39
 S.A. Naqvi, Q.A. Naqvi, A. Hussain, ”Modelling of transmission through a chiral slab
using fractional curl operator” Optics Communications 266 (2006) 404-406;
A. Hussain, S. Ishfaq, Q.A. Naqvi, ”Fractional curl operator and fractional waveguides”
Progress in Electromagnetics Research 63 (2006) 319-335.
[21] M.M. Meerschaert, J. Mortensen, S.W. Wheatcraft, ”Fractional vector calculus for frac-
tional advection-dispersion” Physica A 367 (2006) 181-190;
New Zealand Mathematics Colloquium, Massey University, Palmerston North, New
Zealand, December 2005, Available at
http://www.stt.msu.edu/∼mcubed/MathsColloq05.pdf
[22] K. Cottrill-Shepherd, M. Naber, ”Fractional differential forms” J. Math. Phys. 42 (2001)
2203-2212 (math-ph/0301013);
”Fractional differential forms II” Preprint math-ph/0301016.
[23] Chen Yong, Yan Zhen-ya, Zhang Hong-qing, ”Applications of fractional exterior differential
in three-dimensional space” Applied Mathematics and Mechanics 24 (2003) 256-260.
[24] K.K. Kazbekov, ”Fractional differential forms in Euclidean space” Vladikavkaz Mathemat-
ical Journal 7 (2005) 41-54. (in Russian) http://www.vmj.ru/articles/2005 2 5.pdf
[25] V.E. Tarasov, ”Electromagnetic field of fractal distribution of charged particles” Physics
of Plasmas 12 (2005) 082106 (physics/0610010) ;
”Multipole moments of fractal distribution of charges” Mod. Phys. Lett. B 19 (2005)
1107-1118 (physics/0606251) ;
”Magnetohydrodynamics of fractal media” Physics of Plasmas 13 (2006) 052107; ”Elec-
tromagnetic fields on fractals” Mod. Phys. Lett. A 21 (2006) 1587-1600.
[26] V.E. Tarasov, ”Fractional generalization of gradient and Hamiltonian systems” J. Phys.
A 38 (2005) 5929-5943 (math.DS/0602208);
”Fractional generalization of gradient systems” Lett. Math. Phys. 73 (2005) 49-58
(nlin.CD/0604007).
[27] V.E. Tarasov, ”Liouville and Bogoliubov equations with fractional derivatives” Mod. Phys.
Lett. B 21 (2007) 237-248 (arXiv:0711.0859);
”Fractional statistical mechanics” Chaos 16 (2006) 033108 (arXiv:0710.1807).

40
[28] V.E. Tarasov, G.M. Zaslavsky ”Fractional Ginzburg-Landau equation for fractal media”
Physica A 354 (2005) 249-261.
[29] V.E. Tarasov, ”Continuous medium model for fractal media” Phys. Lett. A 336 (2005)
167-174; ”Fractional hydrodynamic equations for fractal media” Ann. Phys. 318 (2005)
286-307 (arXiv:physics/0602096); ”Possible experimental test of continuous medium model
for fractal media” Phys. Lett. A 341 (2005) 467-472.
[30] I.M. Gelfand, G.E. Shilov, Generalized functions Vol.1. (Academic Press, New York and
London, 1964).
[31] E.H. Brandt, ”Non-local electrodynamics in a superconductor with spacially varying gap
parameter” Phys. Lett. A 39 (3) (1972) 227-228.
[32] J.T. Foley, A. J. Devaney, ”Electrodynamics of nonlocal media” Phys. Rev. B 12 (1975)
3104-3112.
[33] L. Belleguie, S. Mukamel, ”Nonlocal electrodynamics of weakly confined excitons in semi-
conductor nanostructures” J. Chem. Phys. 101 (1994) 9719-9735.
[34] Z.D. Genchev, ”Generalized nonlocal electrodynamics of distributed tunnel Josephson
junctions” Supercond. Sci. Technol. 10 (1997) 543-546.
[35] B. Mashhoon, ”Vacuum electrodynamics of accelerated systems: Nonlocal Maxwell’s equa-
tions” Annalen der Physik (Leipzig) 12 (10) (2003) 586-598;
”Nonlocal electrodynamics of linearly accelerated systems” Phys. Rev. A 70 (2004) 062103
(hep-th/0702074);
”Nonlocal electrodynamics of rotating systems” Phys. Rev. A 72 (2005) 052105.
[36] T. Pierantozzi, L. Vazquez, ”An interpolation between the wave and diffusion equations
through the fractional evolution equations Dirac like” J. Math. Phys. 46 (2005) 113512.
[37] N. Laskin, G.M. Zaslavsky, ”Nonlinear fractional dynamics on a lattice with long-range
interactions” Physica A 368 (2006) 38-54 (nlin.SI/0512010).
[38] V.E. Tarasov, G.M. Zaslavsky, ”Fractional dynamics of coupled oscillators with long-range
interaction” Chaos 16 (2006) 023110;
”Fractional dynamics of systems with long-range interaction” Commun. Nonlin. Sci. Nu-
mer. Simul. 11 (2006) 885-898 (nlin.PS/0512013).

41
[39] V.E. Tarasov, ”Universal electromagnetic waves in dielectric” Journal of Physics: Con-
densed Matter 20 (2008) 175223 (arXiv:0907.2163).
[40] V.E. Tarasov, ”Map of discrete system into continuous” J. Math. Phys. 47 (2006) 092901
(arXiv:0711.2612); ”Continuous limit of discrete systems with long-range interaction” J.
Phys. A 39 (2006) 14895-14910 (arXiv:0711.0826).

See also V.E. Tarasov, Fractional Dynamics: Application of Fractional Calculus to Dy-
namics of Particles, Fields and Media (Springer, HEP, 2011) 504 pages.

42