Eur. Phys. J.

Special Topics 193, 193204 (2011)

c EDP Sciences, Springer-Verlag 2011
DOI: 10.1140/epjst/e2011-01391-5
THE EUROPEAN
PHYSICAL JOURNAL
SPECIAL TOPICS
Regular Article
A fractional calculus approach to nonlocal
elasticity
A. Carpinteri
1
, P. Cornetti
1,a
, and A. Sapora
1
Department of Structural Engineering and Geotechnics, Politecnico di Torino, Corso Duca
degli Abruzzi 24, 10129 Torino, Italy
Received 01 December 2010 / Received in nal form 27 January 2011
Published online 4 April 2011
Abstract. If the attenuation function of strain is expressed as a
power law, the formalism of fractional calculus may be used to handle
Eringen nonlocal elastic model. Aim of the present paper is to provide
a mechanical interpretation to this nonlocal fractional elastic model
by showing that it is equivalent to a discrete, point-spring model.
A one-dimensional geometry is considered; the static, kinematic and
constitutive equations are presented and the governing fractional
dierential equation highlighted. Two numerical procedures to solve
the fractional equation are nally implemented and applied to study
the strain eld in a nite bar under given edge displacements.
1 Introduction
One of the most attractive properties of fractional calculus is that, by varying the
order of derivation, it is possible to describe the transition between completely dif-
ferent phenomena. In the last decades, many fractional dierential equations have
been proposed, solved and used to interpret experimental data that cannot be caught
by usual dierential models. They include relaxation equations, wave equations and
diusion equations (see e.g. [1, 2]). For what concerns solid mechanics, most of the
applications refer to rheological models, where the fractional derivative is taken with
respect to the time variable. More recently, two research directions [3, 4] have been set
that make use of fractional derivative with respect to the space variable. The former
one [5] explores the connection between fractal sets and fractional calculus and applies
the (local) fractional calculus formalism to address the problem of deformation and
damage in solid mechanics. Fractal patterns often arise in heterogeneous materials
and developing mathematical models able to catch the fractality of such phenomena
is a matter of primary concern. The strength of such approach are the non-integer
physical dimensions provided by the fractal geometry and fractional operators. This
property has proven to be very eective in the description of the size-scale eects in
solid mechanics. For a review, see [6].
The latter research direction [710] aims to model nonlocal continua, i.e. solids
characterized by nonlocal interactions [1113]. The novelty is that internal forces are
a
e-mail: [email protected]
194 The European Physical Journal Special Topics
described by fractional derivatives. One of the most remarkable achievements of this
approach is that, by exploiting the Marchaud denition of fractional derivative, the
fractional operators have a clear mechanical interpretation, i.e. springs connecting
non-adjacent points of the body [8]. The related stiness decays along with the
distance among the material points. However, since only the integral part of the
Marchaud derivative is retained in the equilibrium equation, the model developed by
Di Paola and co-workers [8, 14, 15] does not coincide with the corresponding Eringen
fractional nonlocal elastic model; also the order of fractional derivation is dierent.
On the other hand, aim of the present paper is to provide a mechanical interpretation
to Eringen fractional nonlocal model and an ecient algorithm for its solution.
2 Fractional integrals and derivatives
There are various denitions of fractional dierintegral operators, not necessarily
equivalent to each other. A complete list of these denitions can be found in the
fractional calculus treatises, e.g. [16]. These denitions have dierent origins. The
most frequently used denition of fractional integral of order ( R
+
) is due to
Riemann-Liouville and is a straightforward generalization to non-integer values of
Cauchy formula for repeated integrations:
I

a+
f(x) =
1
()

x
a
f(y)
(x y)
1
dy. (1)
The Riemann-Liouville fractional derivative of order is dened as the (integer)
derivative of order n (n N and n 1 < < n) of the fractional integral of order
(n ). That is:
D

a+
f(x) = D
n
[I
n
a+
f(x)]. (2)
However, it is also possible to dene the fractional derivative as the fractional integral
of order (1 ) of the rst derivative. In such a case we obtain the Caputo denition
of fractional derivative,
c
D

a+
:
c
D

a+
f(x) = I
n
a+
[D
n
f(x)]. (3)
It is worth observing that the Riemann-Liouville derivative of a constant is not zero,
whereas it is null the corresponding Caputo derivative. Since Caputo denition gener-
alizes this well-known property of the derivatives of integer orders, Caputo fractional
derivative is usually more practical for applications.
Equations (1)(3) represent the so-called left (or forward) fractional integrals and
derivatives. Analogously, it is possible to dene the right (or backward) operators as:
I

b
f(x) =
1
()

b
x
f(y)
(y x)
1
dy (4)
D

b
f(x) = (D)
n
[I
n
a+
f(x)] (5)
c
D

b
f(x) = I
n
a+
[(D)
n
f(x)]. (6)
A general result in fractional analysis states that the Caputo fractional derivatives
(either forward or backward) of a function f(x) are equal to the Riemann-Liouville
Perspectives on Fractional Dynamics and Control 195
derivatives provided that the polynomial of order n 1 (evaluated either in x = a or
x = b) is subtracted from the function itself:
D

a+

f(x)
n1

k=0
f
(k)
(a)
k!
(x a)
k

=
c
D

a+
f(x) (7)
D

f(x)
n1

k=0
f
(k)
(b)
k!
(x b)
k

=
c
D

b
f(x). (8)
By recalling the fractional derivatives of the power functions (x a)
k
and (b x)
k
,
Eqs. (7)(8) provide:
D

a+
f(x) =
n1

k=0
f
(k)
(a)
(1 + k )
(x a)
k
+
c
D

a+
f(x) (9)
D

b
f(x) =
n1

k=0
(1)
k
f
(k)
(b)
(1 + k )
(b x)
k
+
c
D

b
f(x) (10)
that will be exploited for the numerical implementation in Sec. 5.
In the case 0 < < 1, by applying the formula of integration by parts to the
Caputos denition and after some analytical manipulations, it is possible to give an
alternative form to the Riemann-Liouville fractional derivative [16]:
D

a+
f(x) =
f(x)
(1 )(x a)

+

(1 )

x
a
f(x) f(y)
(x y)
1+
dy (11)
D

b
f(x) =
f(x)
(1 )(b x)

+

(1 )

b
x
f(x) f(y)
(y x)
1+
dy. (12)
Equations (11)(12) are the so-called Marchaud denitions of fractional derivative.
Since, for 0 < < 1, they coincide with the Riemann-Liouville denitions for a wide
class of functions, we will indicate them with the same symbol.
Finally, it is possible to introduce the Riesz fractional integrals and derivatives,
dened as the sum of the forward and backward fractional operator up to a multi-
plicative factor:
I

a,b
f(x) =
1
2

a+
f(x) + I

b
f(x)

=
1
2()

b
a
f(y)
|x y|
1
dy (13)
D

a,b
f(x) =
1
2

a+
f(x) + (1)
n
D

b
f(x)

. (14)
Note that the multiplicative constant, here taken simply equal to 1/2 following,
e.g., [17], can vary according to the dierent fractional calculus treatises.
For 1 < < 2, we proved that the Riesz fractional derivative (14) can be expressed
in a Marchaud-like form as:
D

a,b
f(x) =
1
2(1 )

f(x)
(x a)

+
f(x)
(b x)

b
a
f(x) f(y)
|x y|
1+
dy

. (15)
It is important to emphasize that Eq. (15) is not trivial, since the Marchaud denitions
of fractional derivative (11) and (12) hold true only for 0 < < 1 (otherwise the
196 The European Physical Journal Special Topics
integrals at the right-hand side diverge). Equation (15) plays a fundamental role in the
subsequent part of the paper, since it allows us to provide a mechanical interpretation
to the Eringen non-local fractional model. Details of the proof of Eq. (15) will be
given elsewhere; however note that a formula analogous to Eq. (15), but dened
on an innite domain, has been proved by Goreno and Mainardi [18] extending a
previous result by Samko [16]. For fractional operators dened on innite domains
see also [19, 20].
3 Eringen nonlocal fractional model
According to Eringen nonlocal elasticity, the stress at a given point depends on the
strain in a neighborhood of that point by means of a convolution integral. This de-
pendence is described by a proper attenuation function g, which decays along with
the distance. In the case of a one-dimensional domain (i.e. a bar):
(x) = E

(x) +

b
a
(y)g(x y)dy

(16)
where is the stress, x the longitudinal coordinate, x = a and x = b the bar extreme
coordinates, E the Youngs modulus, the strain dened as the derivative of the
longitudinal displacement u and

is a material constant. The bar length is l (l =

ba). Note that the Eringen model [11] is sometime referred to as strong (or integral)
nonlocality, to distinguish it from the weak (or gradient) nonlocal elastic model, where
the stress depends on the strain and its derivatives [12].
Let us now assume the following form for the attenuation function g:
g() =
1
2(1 )||

(17)
with 0 < < 1. With the choice of Eq. (17), the constitutive relationship becomes:
(x) = E

(x) +

2(1 )

b
a
(y)
|x y|

dy

. (18)
The presence of the Riesz integral (13) can be recognized in Eq. (18), which can thus
be rewritten as:
(x) = E[ +

(I
1
a,b
)]. (19)
Note that

has anomalous physical dimensions [L]

1
. Since = du/dx, it is
easy to highlight in Eq. (19) the presence of the Caputo fractional derivatives of the
displacement. Therefore, by exploiting Eqs. (3) and (6), the dependence of the stress
upon the displacement becomes:
(x) = E

du
dx
+

2
(
c
D

a+
u
c
D

b
u)

. (20)
Before proceeding, it is worth observing that the fractional nonlocal constitutive law
expressed by Eq. (19) or (20) has been originally proposed by the present authors
in [9] and, independently, by Atanackovic and Stankovic in [10], where its eect on
the wave propagation in an innite bar was investigated.
In order to get the equilibrium equation in terms of the displacement function
u(x), we simply need to substitute Eq. (20) into the static equation d/dx+f(x) = 0,
Perspectives on Fractional Dynamics and Control 197
where f(x) is the longitudinal force per unit volume. By means of Eqs. (7) and (8)
and some more analytical manipulations, we get:
d
2
u
dx
2
+

2
{D
1+
a+
[u(x) u(a)] + D
1+
b
[u(x) u(b)]} =
f(x)
E
. (21)
Equation (21) is a fractional dierential equation. Note that, while the left frac-
tional derivative coincides always with its integer order counterpart when the order
of derivation is an integer number, the right fractional derivative coincides with the
corresponding integer derivative only when the order of derivation is even; when the
order of derivation is an odd number, it is equal to its opposite. Therefore, the term
in curly brackets is equal to 2u

(x) when = 1, and vanishes when = 0. In order

to highlight the presence of the Riesz fractional derivative, Eq. (21) can be set in the
following form:
d
2
u
dx
2
+

D
1+
a,b
[u(x)] +

2(1 )

u(a)
(x a)
1+
+
u(b)
(b x)
1+

=
f(x)
E
. (22)
4 Equivalent point-spring model
A useful interpretation of the governing equation (22) for the nonlocal elastic bar
is sought. To this aim, it is convenient to express the Riesz fractional derivatives
in the Marchaud-like form [8]. This goal can be achieved by means of Eq. (15), since
the order of the fractional derivative in Eq. (22) is comprised between 1 and 2 (while
the order of fractional derivation considered in [8, 14, 15] is less than unity). Hence,
by setting = 1 + in Eq. (15), we get:
d
2
u
dx
2

2(1 )

u(x) u(a)
(x a)
1+
+
u(x) u(b)
(b x)
1+
+ (1 + )

b
a
u(x) u(y)
|x y|
2+
dy

=
f(x)
E
(23)
where the gamma function property (1 ) = () has been used. In this
form it is evident that the rst term at the left-hand side rules the local interactions,
whereas the terms in the square brackets rule the nonlocal interactions by means of
linear elastic springs and can be seen as an extra-force per unit of volume acting at
the point of abscissa x. More in detail, the rst two terms in the brackets refer to
long-range interactions between the inner (a < x < b) and outer (x = a, x = b) points,
whereas the integral term takes into account the interaction between two inner generic
points. To make the concept even clearer, it is useful to write Eq. (23) in discrete form.
To this purpose, let us introduce a partition of the interval [a, b] on the x axis made
of n (n N) intervals of length x = l/n. The generic point of the partition has the
abscissa x
i
, with i = 1, . . . , n + 1 and x
1
= a, x
n+1
= b; that is, x
i
= a + (i 1)x.
Hence, for the inner points of the domain (i = 2, . . . , n), the discrete form of Eq. (23)
reads:
u
i+1
+ 2u
i
u
i1
(x)
2
+

2(1 )

u
i
u
1
(x
i
x
1
)
1+
+
u
i
u
n+1
(x
n+1
x
i
)
1+

(1 + )
2(1 )
n+1

j=1,j=i
(u
i
u
j
)x
|x
i
x
j
|
2+
=
f
i
E
(24)
198 The European Physical Journal Special Topics
where u
i
u(x
i
) and f
i
f(x
i
). Multiplying both the sides by EAx (A being the
area of the bar cross section), Eq. (24) may be rewritten as:
k
l
i,i+1
(u
i
u
i+1
) + k
l
i,i1
(u
i
u
i1
) + k
vs
i,1
(u
i
u
1
) + k
vs
i,n+1
(u
i
u
n+1
)
+
n+1

j=1,j=i
k
vv
i,j
(u
i
u
j
) = f
i
Ax. (25)
It is evident how the nonlocal fractional model is equivalent to a point-spring model
where three kinds of springs appear: the local springs, ruling the local interactions,
whose stiness is k
l
; the springs connecting the inner material points with the bar
edges, ruling the volume-surface long-range interactions, with stiness k
vs
; the springs
connecting the inner material points each other, describing the nonlocal interactions
between non-adjacent volumes, whose stiness is k
vv
. Provided that the indexes are
never equal one to the other, the following expressions for the stinesses hold:
k
l
i,i+1
= k
l
i+1,i
= EA/x (26)
k
vs
i,1
= k
vs
1,i
= EA

2(1 )
x
(x
i
x
1
)
1+
(27)
k
vs
i,n+1
= k
vs
n+1,i
= EA

2(1 )
x
(x
n+1
x
i
)
1+
(28)
k
vv
i,j
= k
vv
j,i
= EA

(1 + )
2(1 )
(x)
2
|x
i
x
j
|
2+
. (29)
Furthermore, by exploiting the Principle of Virtual Work to derive the proper either
kinematic or static boundary conditions, it is possible to show that a fourth set of
springs has to be introduced: it is composed by a unique spring connecting the two
bar extremes with stiness:
k
ss
1,n+1
= k
ss
n+1,1
=
EA

2(1 )
1
(x
n+1
x
1
)

. (30)
The superscript ss for the stiness (30) is used since the spring connecting the bar
edges can be seen as modelling the interactions between material points lying on
the surface, which, in the simple one-dimensional model under examination, reduce
to the two points x = a, b. Note that the presence of such a spring was implicitly
embedded in the constitutive equation (19). However, since it provides a constant
stress contribution throughout the bar length, its presence was lost by derivation
when inserting the constitutive relationship into the dierential equilibrium equation,
i.e. when passing from Eq. (20) to Eq. (21).
To summarize, the constitutive fractional relationship (19) is equivalent to a point-
spring model with four sets of springs, one local (26) and three nonlocal (27)(30).
Note that their stinesses all decay with the distance, although the decaying velocity
diers from one kind to the other. The equivalent point-spring model is drawn in g. 1.
Each internal point is connected to the adjacent points by two local springs, to the bar
extremes by two volume-surface nonlocal springs and to all the other material points
by the volume-volume nonlocal springs. Finally a surface-surface nonlocal spring con-
nects the bar edges. Turning the attention to the whole bar, the number of the local
springs is n, the number of the volume-surface springs is 2n 1 , the number of the
volume-volume springs is n(n + 1)/2.
Perspectives on Fractional Dynamics and Control 199
1 2 3 4 5
k
14
+ k
14
vv vs
k
13
+ k
13
vv vs
k
25
+ k
25
vv vs
k
15
+ k
15
+ k
15
vv vs ss
k
12
+ k
12
+ k
12
vv vs l
k
23
+ k
23
vv l
k
34
+ k
34
vv l
k
45
+ k
45
+ k
45
vv vs l
k
35
+ k
35
vv vs
k
24
vv
Fig. 1. Pont-spring model equivalent to the nonlocal fractional elastic bar (n = 4).
For what concerns the limit cases, if 0
+
, the volume-volume and the volume-
surface spring interactions ruled by Eqs. (27)(29) vanish, and only the contribution
(30) remains (together with the local springs (26)): the nonlocal model corresponds
to a classical (local) elastic bar in parallel with a spring of stiness EA

/2. The
governing equation reverts to the classical case: u

= f/E. On the other hand, in the

limit case 1

, since (0
+
) = +, the surface-surface (Eq. (30)) and the volume-
surface (Eqs. (27)(28)) contributions disappear. For what concerns the interactions
between inner material points (Eq. (29)), only the interactions between adjacent
material points are retained (the Gamma function tends to innity, but the integral
in Eq. (23) diverges). Correspondingly, the additive term in Eq. (19) has the same
form as the classical (local) one, the model representing a bar with a stiened (by
a factor of (1 +

)) Youngs modulus, while the governing equation (23) becomes

u

= f/[E(1 +

)].
5 Numerical analysis
Based on Eq. (15), the equivalence between the fractional model and the point-spring
model proved in the previous section provide also a straightforward numerical algo-
rithm to implement the fractional governing equation (22)(23). In fact, by using the
same partition previously introduced, Eq. (23) can be discretized as:
n+1

j=1
k
i,j
u
j
= F
i
, i = 1, . . . , n + 1 (31)
where the right hand side F
i
is equal to f
i
Ax for the inner points (i = 2, . . . , n)
and to the external forces F
a
and F
b
acting at the bar edges for i = 1 and i = n + 1,
respectively; k
i,j
is the generic element of the stiness (square) matrix K, which is
the sum of four stiness matrices:
K = K
l
+K
vv
+K
vs
+K
ss
(32)
200 The European Physical Journal Special Topics
whose non-diagonal terms are provided by the opposite of the corresponding sti-
nesses (2630). Furthermore, the diagonal terms k
i,i
of each matrix is provided by
the relationship:
k
i,i
=
n+1

j=1,j=1
k
i,j
i = 1, . . . , n + 1 . (33)
Note that all the four matrices are symmetrical, with positive elements on the diagonal
and negative outside. More in detail, the local matrix K
l
is tridiagonal; the nonlocal
matrix K
vv
ruling the long-range interactions between inner points is fully populated;
the nonlocal matrix related to the inner-outer interactions K
vs
has only border and
diagonal elements dierent from zero; nally, the nonlocal matrix K
ss
describing the
interaction between the bar edges is empty except for the four corner elements.
Despite the clear physical-mechanical meaning, however, the discretization (31) is
not the most ecient way to solve the fractional dierential equation (22). Particu-
larly, it is not able to catch the solution for approaching unity, when the weight
function in the integral in Eq. (23) behaves as a Dirac function. Since the order of
fractional derivation is comprised between 1 and 2 (i.e. 0 < 1), we chose to im-
plement the so-called L2 algorithm rstly proposed by Oldham and Spanier [21] and
later applied to discretize the Riesz derivative by Yang et al. [22]. The L2 algorithm
is based on the formulae (9) and (10), which now read:
D
1+
a+
f(x) =
1
(1 )

f(a)
(x a)
1
+
f

(a)
(x a)

x
a
f

(y)
(x y)

dy

(34)
D
1+
b
f(x) =
1
(1 )

f(b)
(b x)
1

f

(b)
(b x)

b
x
f

(y)
(y x)

dy

. (35)
By approximating the second order derivatives by means of the usual nite dierences
and evaluating analytically the remaining part of the integrals in Eqs. (34)(35), we
get the following approximate discrete expressions of the fractional derivatives in the
internal points of the interval [a, b], i.e. for i = 2, . . . , n:
D
1+
a+
f(x
i
)
(x)
(1+)
(2 )

(1 )
(i 1)
1+
f
1
+
1
(i 1)

(f
2
f
1
)+
+
i2

k=0
(f
ik+1
2f
ik
+ f
ik1
)[(k + 1)
1
k
1
]

(36)
D
1+
b
f(x
i
)
(x)
(1+)
(2 )

(1 )
(n i + 1)
1+
f
n+1

1
(n i + 1)

(f
n+1
f
n
)
+
ni

k=0
(f
i+k+1
2f
i+k
+ f
i+k1
)[(k + 1)
1
k
1
]

. (37)
Developing the sums in Eqs. (36)(37), the Riesz fractional derivative can hence be
approximated as:
D
1+
a,b
f(x
i
)
(x)
(1+)
2(2 )
n+1

j=1
r
i,j
f
j
(38)
Perspectives on Fractional Dynamics and Control 201
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.9
1
1.1
1.2
1.3
x 10
3
Longitudinal coordinate, x (m)
S
t
r
a
i
n
,

Fig. 2. Strain eld for a bar under given edge displacements: comparison between the spring
algorithm (continuous line) and the L2 algorithm (dotted line) for = 0.5.
where the terms r
i,j
of the matrix R are provided in Appendix A. By exploiting
Eq. (38), we may discretize the fractional dierential equation (22) in a form alter-
native to Eq. (24) (i.e. (31)) as:
u
i+1
2u
i
+ u
i1
(x)
2
+

2(2 )
1
(x)
1+

n+1

j=1
r
i,j
u
j
+ (1 )

u
1
(i 1)
1+
+
u
n+1
(n i + 1)
1+

=
f
i
E
(39)
holding for i = 2, . . . , n.
We applied the developed fractional nonlocal model to analyze the strain eld in
a nite bar with given edge displacement: u
a
= u
0
= 0 m and u
b
= u
n+1
= 10
3
m.
The bar length l is 1 m and the material constant

is assumed equal to 1 m
1
.
Note that, since no external force is present, the Youngs modulus E does not aect
the result.
We solved the governing equation (22) in the case = 0.5 both with the physically-
based algorithm (31), i.e. the point-spring model, and with the mathematically-based
algorithm (39). The strain elds corresponding to the two numerical procedures are
plotted in Fig. 2. From a mathematical point of view, the almost perfect coincidence
between the two solutions is an important result since it can be seen as an indirect
proof that Eq. (15) holds true also for order of fractional derivation comprised between
1 and 2, as we claimed in Sec. 2. On the other hand, from a physical point of view,
it is interesting to observe that, with respect to the classical case represented by a
uniform strain eld, the strain localizes near the bar ends. This eect can be explained
observing that the zones close to the borders are less sti because of lower presence of
the long-range interactions. Finally, it is interesting to note that the same behaviour
is provided also by gradient elasticity [23].
As already observed, algorithm (31) is much less ecient than algorithm (39)
for any , becoming completely unreliable as 1

. Therefore, in Fig. 3, we used

algorithm (39) to solve the same geometry considered in Fig. 2 but for dierent values.
202 The European Physical Journal Special Topics
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.9
1
1.1
1.2
1.3
x 10
3
Longitudinal coordinate, x (m)
S
t
r
a
i
n
,

=0.6
=0.7
=0.8
=0.9
=1.0
Fig. 3. Strain eld for a bar under given edge displacements: strain elds for dierent
values obtained by the L2 algorithm.
Note that the area beneath each curve is constant, being equal to the assigned relative
displacement between the bar extremes (i.e. 10
3
m). As 1

, the solution tends

to the homogeneous one since the nonlocal term in the constitutive equation (39)
acquires the same form as the local one. On the other hand, up to some extent, the
strain localization increases as decreases.
6 Conclusions
As is well-known, in solid mechanics, temporal fractional derivatives can be used to
develop improved visco-elastic models. On the other hand, in the present paper we
showed that spatial fractional derivative (namely the Riesz fractional derivative) can
be used to take into account nonlocal eects in the constitutive material behaviour.
It has been shown that, in the one-dimensional case, the nonlocal fractional bar is
equivalent to a point-spring model with four sets of springs. It is argued that these
springs may describe the long-range interactions between volume and surface elements
of the solid, that, as is well-known, play an important role at the smaller scales. A
key role in deriving the present model is played by the proof of a suitable formula
expressing the Riesz fractional derivative of order comprised between 1 and 2 in a
Marchaud-like form.
The nancial support of the Italian Ministry of Education, University and Research (MIUR)
to the Project Advanced applications of Fracture Mechanics for the study of integrity
and durability of materials and structures within the Programmi di ricerca scientica di
rilevante interesse nazionale (PRIN) program for the year 2008 is gratefully acknowledged.
Appendix A: Numerical approximation of the Riesz derivative
In this appendix we provide the explicit expressions of the terms r
i,j
(i = 2, . . . , n
and j = 1, . . . , n + 1) of the square (n + 1) (n + 1) matrix R (see Eq. (38)). The
terms on the rst and last row depend on the assigned boundary conditions and are
Perspectives on Fractional Dynamics and Control 203
not given here. We have:
r
2,1
= 1 +
2
(A.1)
r
2,2
= 2
1
4 = r
n,n
(A.2)
r
3,1
= 1 + (2 + +
2
)/2
1+
(A.3)
r
3,2
= 4 (3 + )/2

. (A.4)
The remaining diagonal terms are (i = 3, . . . , n 1):
r
i,j
= 2(3 2
1
) (A.5)
while the terms on the rst and second columns are, respectively (i = 4, . . . , n):
r
i,1
= (1 )(i 1)
(1+)
(1 )(i 1)

+(i 1)
1
(i 2)
1
(A.6)
r
i,2
= (1 )(i 1)

2(i 1)
1
+ 3(i 2)
1
(i 3)
1
. (A.7)
The elements on the diagonal close to the main one have the following values (i =
4, . . . , n):
r
i,i1
= 4 + 3
1
3 2
1
. (A.8)
All the other terms beneath the main diagonal are equal to (i = 5, . . . , n; j =
3, . . . , i 2):
r
i,j
= (i j + 2)
1
3(i j + 1)
1
+ 3(i j)
1
(i j 1)
1
. (A.9)
The remaining elements, i.e. those placed above, are immediately ob-
tained by observing that the matrix R fulls a sort of polar symmetry, i.e.
(i = 2, . . . , n; j = i + 1, . . . , n + 1):
r
i,j
= r
n+2i,n+2j
(A.10)
and the matrix R is completely dened.
References
1. A. Carpinteri, F. Mainardi, Fractals and Fractional Calculus in Continuum Mechanics
(Springer-Verlag, Wien, 1997)
2. F. Mainardi, Fractional Calculus and Waves in Linear Viscoelasticity: An Introduction
to Mathematical Models (Imperial College Press, London, 2010)
3. A. Carpinteri, P. Cornetti, A. Sapora, Z. Angew. Math. Mech. 89, 207 (2009)
4. A. Carpinteri, P. Cornetti, A. Sapora, M. Di Paola, M. Zingales Phys. Scr. T136, 014003
(2009)
5. A. Carpinteri, P. Cornetti, Chaos Solitons Fractals 13, 85 (2002)
6. A. Carpinteri, P. Cornetti, S. Puzzi, Appl. Mech. Rev. 59, 283 (2006)
7. K.A. Lazopoulos, Mech. Res. Commun. 33, 753 (2006)
8. M. Di Paola, M. Zingales, Int. J. Sol. Struct. 45, 5642 (2008)
9. A. Carpinteri, P. Cornetti, A. Sapora, M. Di Paola, M. Zingales in Proceedings of the XIX
Italian Conference on Theoretical and Applied Mechanics, Ancona, Italy, 2009, edited
by S. Lenci (Aras Edizioni, Fano/Italy, 2009), p. 315
204 The European Physical Journal Special Topics
10. T.M. Atanackovic, B. Stankovic, Acta Mech. 208, 1 (2009)
11. A.C. Eringen, D.G.B. Edelen, Int. J. Eng. Sci. 10, 233 (1972)
12. E.C. Aifantis, Int. J. Eng. Sci. 30, 1279 (1992)
13. C. Polizzotto, Int. J. Sol. Struct. 38, 7359 (2001)
14. G. Cottone, M. Di Paola, M. Zingales, Physica E 42, 95 (2009)
15. G. Failla, A. Santini, M. Zingales, Mech. Res. Commun. 37, 13 (2010)
16. S.G. Samko, A. Kilbas, O.I. Marichev, Fractional Integrals and Derivatives (Gordon and
Breach Science Publisher, Amsterdam, 1993)
17. O.P. Agrawal, J. Phys. A 40, 6287 (2007)
18. R. Goreno, F. Mainardi, in Problems in Mathematical Physics (Siegfried Prossdorf
Memorial Volume) edited by J. Elschner, I. Gohberg, B. Silbermann (Birkh auser Verlag,
Basel/Switzerland, 2001), p. 120
19. F. Mainardi, R. Goreno, D. Moretti, G. Pagnini, P. Paradisi, Chem. Phys. 284, 521
(2002)
20. M.D. Ortigueira, J. Vib. Control 14, 1255 (2008)
21. K.B. Oldham, J. Spanier, The Fractional Calculus (Academic Press, New York, 1974)
22. Q. Yang, F. Liu, I. Turner, Appl. Math. Model 34, 200 (2010)
23. I. Vardoulakis, G. Exadaktylos, E.C. Aifantis, Int. J. Sol. Struct. 33, 4531 (1996)