Revista Brasileira de Ensino de Fsica, v. 33, n.

4, 4302 (2011)
www.sbfisica.org.br

Fractional order calculus: historical apologia,

basic concepts and some applications
(C
alculo de ordem fracion
aria: apologia hist
orica, conceitos b
asicos e algumas aplicac
oes)

S.A. David1 , J.L. Linares e E.M.J.A. Pallone

Departamento de Ciencias B
asicas, Faculdade de Zootecnia e Engenharia de Alimentos,
Universidade de Sao Paulo, SP, Brasil
Recebido em 22/10/2010; Aceito em 23/6/2011; Publicado em 21/11/2011

Fractional order calculus (FOC) deals with integrals and derivatives of arbitrary (i.e., non-integer) order, and
shares its origins with classical integral and dierential calculus. However, until recently, it has been investigated
mainly from a mathematical point of view. Advances in the eld of fractals have revealed its subtle relationships
with fractional calculus. Nonetheless, fractional calculus is generally excluded from standard courses in mathe-
matics, partly because many mathematicians are unfamiliar with its nature and its applications. This area has
emerged as a useful tool among researchers. One of the objectives of this paper is to discuss the usefulness of
fractional calculus in applied sciences and engineering. In view of the increasing interest in the development of
the new paradigm, another objective is to encourage the use of this mathematical idea in various scientic areas
by means of a historical apologia for the development of fractional calculus.
Keywords: fractional order calculus, non-integer order systems, dynamical systems.

O calculo de ordem fracion aria (COF) lida com derivadas e integrais de ordem arbitr aria (ou seja, nao-
inteiros) e tem as mesmas origens do cl assico c
alculo diferencial e integral. No entanto, ate recentemente, tem
sido investigado principalmente do ponto de vista matem atico. Progressos nas areas de fractais revelaram suas
sutis relaco
es com o c alculo fracionario. No entanto, o c alculo fracionario e geralmente excludo dos cursos tradi-
cionais de Matem atica, em parte porque muitos matem aticos n
ao estao familiarizados com a sua natureza e suas
aplicaco
es. O c alculo fracion ario tem se mostrado uma ferramenta u til entre os pesquisadores. Um dos objetivos
deste trabalho e discutir a utilidade do c alculo fracionario nas ciencias aplicadas e na engenharia. Alem disso,
outro prop osito e incentivar o uso dessa ideia matem atica como ferramenta para pesquisadores das mais diversas
a
reas por meio de uma apologia hist orica envolvendo o desenvolvimento do c alculo fracion
ario.
Palavras-chave: c alculo fracionario, sistemas de ordem n ao-inteira, sistemas din amicos.

1. Introduction years, advances in the theory of chaos and fractals re-

The theory of fractional calculus dates back to the
birth of the theory of dierential calculus, but its in-
herent complexity delayed the application of its associ-
ated concepts. In fact, fractional calculus is a natural
extension of classical mathematics. Since the inception
of the theory of dierential and integral calculus, math-
ematicians such as Euler and Liouville developed their
ideas about the calculation of non-integer order deriva-
tives and integrals. Perhaps the subject would be more
aptly called integration and dierentiation of arbitrary
order.
Despite the work that has been done in this area,
the application of fractional derivatives and integrals
has been infrequent until recently. However, in recent
1 E-mail: sergiodavid@usp.br.
vealed relationships with fractional derivatives and in-
tegrals, leading to renewed interest in this eld [1-3].
The basic aspects of the theory of fractional calculus
are outlined in [4]. Insofar as it concerns the application
of its concepts, we can cite research in dierent areas
such as viscoelastic damping [5], robotics and control
[6-8], signal processing [9], and electric circuits [10,11].
As for the adoption of this concept in other scientic
areas, several researchers have been inspired to examine
this new possibility.
Some work has been carried out in the eld of dy-
namical systems theory, but the proposed models and
algorithms are still in the preliminary stage.
With these ideas in mind, this work introduces the
fundamentals of fractional order calculus (FOC) and its

Copyright by the Sociedade Brasileira de Fsica. Printed in Brazil.

4302-2 David et al.

applications. This paper is organized as follows. Sec- gamma integral, the integral formula can be calculated
tion 2 outlines the origins of fractional calculus, provid- for x-a . Note that in the integral
ing the background for a historical apologia for its de-
velopment, and discussing several approaches to math-
ematical formulation. Section 3 oers examples of ap- ua1 exu du, (4)
0
plications. Finally, section 4 presents a discussion, con-
clusions and outlook. if we change the variables t = xu, then

2. Historical apologia and dierent ap- ( )a1

proaches xu t 1
u a1
e du = et dt =
0 x x
Historically, fractional order calculus (FOC) has been
unexplored or its applications delayed in engineering ta1 1 1
et dt = a ta1 et dt (5)
because of its inherent complexity, the apparent self- xa x1 x x 0
suciency of integer order calculus (IOC), and the fact
that it lacks a fully acceptable geometric or physical Thus,
interpretation. The intuitive idea of FOC is as old
as IOC, as indicated in a letter written by Leibniz to

LHopital in 1695. Leibniz, when asked about what if xu 1
dn y u a1
e du = a ta1 et dt. (6)
n were 12 in dx n , said [12]: Someday it would lead to 0 x 0
useful consequences.
In 1730 Euler mentioned interpolating between in- However, in accordance with Eq. (1), this yields the
tegral orders of a derivative. In 1812 Laplace dened integral formula
a fractional derivative by means of an integral, and the

rst discussion of a derivative of fractional order ap- 1
peared in a calculus written by Lacroix in 1819. xa = ua1 exu du. (7)
(a) 0

2.1. Lacroix d ax
Consequently, by assuming that dx (e ) = a eax for
Lacroix expressed the nth derivative (for n m) in any > 0, then
terms of Legrendes symbol for the generalized facto-
rial. Recalling that
d xa (a + ) a
= x =
dx (a)
(a) = ta1 et dt (1)
0 (a + ) a
(1) x . (8)
and starting, for instance, with the function y = xm , (a)
Lacroix expressed it as follows
The (1) term in the latter equation suggests the
need to expand the theory to include complex numbers.
dn y m! (m + 1)
= xmn = xmn . (2) Indeed, in terms of contemporary denitions, the
dxn (m n)! (m n + 1) modern theory of fractional calculus is intimately con-
nected with the theory of operators. In classical calcu-
Thus, replacing n with 12 and letting m = 1, one obtains
lus, the symbol Dxn is often used for the nth derivative
the derivative or order 12 of the function x
operator (for n 0) while, less commonly, Dx1 is used
for the anti-derivative (or integral) operator.
d1/2 y (2) 2
1/2
= x1/2 = x. (3) A convenient notation described by Davis [12] was
dx (3/2)
the following: if is a positive real number, c Dx f (x)
denotes dierentiation of order of the function f along
2.2. Liouville
the x-axis. Similarly, the operator c Dx f (x) will de-
It was Liouville who engaged in the rst major study note integration of order of the function f along the
of fractional calculus. Liouvilles rst denition of a x-axis.
derivative of arbitrary order involved an innite se- Fractional calculus still lacks a geometric interpreta-
ries. Here, the series must be convergent for some . tion of integration or dierentiation of arbitrary order.
Liouvilles second denition succeeded in giving a frac- Hence, the subscripts c and x are called here termi-
tional derivative of xa whenever both x and are pos- nals of integration instead of limits of integration. This
itive. Based on the denite integral related to Eulers avoids unnecessary confusion.
Fractional order calculus: historical apologia, basic concepts and some applications 4302-3

2.3. Laurent where the function () is the generalization of facto-

rial function and is dened in Eq. (1).
In 1884 Laurent published what is now recognized as
the denitive paper on the fundaments of fractional Indeed, since the birth of dierential and integral
calculus. Using Cauchys integral formula for complex calculus, the generalization of the derivative and inte-
valued analytical functions and a simple change of no- gral concept to a non-integer order has been the subject
tation to employ a positive rather than a negative of several approaches. For this reason, there are vari-
will now yield Laurents denition of integration of ous denitions that are provenly equivalent, and their
arbitrary order > 0 use can (and should) be encouraged by researchers in
dierent scientic areas.
x
1
c Dx f (x) = = (x t)1 f (t) dt. (9)
() c 2.5. Caputo
The appropriate denition of dierentiation of arbi-
trary order is to integrate it up to a point from which In order to present Caputos fractional derivative, let
the desired result can be obtained by conventional dif- m be the smallest integer that exceeds , thus enabling
ferentiation. Caputos fractional derivative of order > 0 to be de-
Let nu = m rho where, for convenience, m is con- ned as follows
sidered the smallest integer larger than and 0 < 1.
Observe that
t
1 f (m) ( )
m D f (t)
= d,
c Dx f (x) = c Dx f (x). (10) (m ) 0 (t )+1m
Thus m 1 < < m. (14)

m dm [ ]
c Dx f (x) = c Dx f (x) , (11) Clearly, numerous mathematicians have contributed to
dxm the history of fractional calculus by attempting to solve
and consequently a fundamental problem to the best of their understand-
[ x ] ing.
dm [ ] dm 1 Each researcher sought a denition and therefore
c D x f (x) = (x t) 1
f (t) dt .
dxm dxm () c dierent approaches, which has led to various deni-
(12) tions of dierentiation and antidierentiation of non-
integer orders that are provenly equivalent. Some of the
2.4. Cauchy aforementioned denitions of non-integer order deriva-
tives are summarized and listed in Table 1. A table
Cauchys denition, which is recognized as one that pre-
can also be written for the denition of dierent and
serves some important frequency properties [13], is ex-
equivalent integrals.
pressed as follows
Although all these denitions may be equivalent,

() (t )1 from one specic standpoint, i.e., for a specic applica-
f+ = f ( ) d, (13) tion, some denitions seem more attractive.
()

Table 1 - Definition of derivatives of arbitrary order.

dn y (m + 1)
Lacroix = xmn
dxn (m n + 1)
d xa (a + ) a
Liouville = (1) x
dx (a)
[ ]
dm 1 x
Laurent D
c x

f (x) = D
c x
m
f (x) = (x t) 1
f (t) dt
dxm () c

() (t )1
Cauchy f+ = f ( ) d
()
1 t f (m) ( )
Caputo D f (t) = d
(m ) 0 (t )+1m
4302-4 David et al.

3. Applications in applied sciences and Taking the nite sine transform of Eq. (16), in-
engineering tegrating the second term of the resulting equation by
parts, and applying the boundary conditions, we obtain
Fractional order calculus can represent systems with
high-order dynamics and complex nonlinear phenom- d u

+ (ban)2 u
= 0, (18)
ena using few coecients, since the arbitrary order of dt
the derivatives provides an additional degree of freedom and L
to t a specic behavior. Another important charac-
u
=u
(n, t) = u(x, t) sin(anx)dx, (19)
teristic is that fractional order derivatives depend not 0
only on local conditions of the evaluated time but also
is the nite sine transform of u(x, t), where a = /L,
on the entire history of the function. This fact is often
and n is a wave number, Eq. (18) may be called a
useful when the system has a long-term memory and
diusion-wave equation in a wave number domain. Tak-
any evaluation point depends on the past values of the
ing the nite sine transform of Eq. (17), we obtain
function.
At this point we therefore consider it relevant to L
present some applications involving the implementa- u
(n, 0) = f (x) sin(anx)dx, (20)
0
tion of FOC-based models in dierent physical sys-
tems, namely: the diusion equation, food engineering, and ut (n, 0) = 0 for 1 < 2.
robotics and control theory, and econophysics. Taking the Laplace transform of Eq. (16) and using
the initial conditions and the properties of the Caputo
3.1. Applications in the diusion equation derivative, we obtain

It is well known that, in the classical case, the diusion s1 u(n, 0)

equation is given by U (n, s) = (21)
s + (ban)2
u 2u where s is the Laplace transform parameter, and
= b2 2 . (15) U (n, s) is the Laplace transform of u
(n, t).
t x
Taking rst the inverse Laplace transform of
A general solution is given for a fractional diu- Eq. (21) and then the inverse nite sine transform of
sion equation dened in a bounded space domain. The the resulting equation, we obtain
fractional time derivative is described in the Caputo
sense (Table 1). The Caputo fractional derivative is

considered here because it allows for the standard in- 2
clusion of traditional initial and boundary conditions in u(x, t) = E (b2 a2 n2 t ) sin(anx)
L n=1
the formulation, whereas models based on other frac- L
tional derivatives may require the values of the frac- f (r) sin(anr)dr, (22)
tional derivative terms at the initial time. Keeping this 0
denition in mind, the fractional diusion equation [14]
of interest can be written as where E is the Mittag-Leer function.
The Mittag-Leer function has several interesting
u 2u properties.

= b2 2 , (16) In particular, we have E1 (z) = ez and
t x
where is a parameter describing the order of the frac- E2 (z 2 ) = cos(z). Using these identities, the solutions
tional derivative, b denotes a constant coecient with of Eq. (16) for = 1 and 2 are given as
dimension (Time)(Length)/2 , x and t are the space

and time variables, and u = u(x, t) is the eld dened 2 (ban)2 t
in the space domain [0,L], considering the following u(x, t) = e sin(anx)
L n=1
boundary conditions
L
f (r) sin(anr)dr, (23)
0
u(0, t) = u(L, t) = 0, t > 0,
u(x, 0) = f (x), 0 < x < L, and
ut (x, 0) = 0, 0 < x < L,

for 1 < 2. (17) 2
u(x, t) = cos(ban t) sin(anx)
L n=1
The last boundary condition is assumed to ensure
L
the continuous dependence of the solution on the pa-

rameter in the transition from = 1 to = 1+ [2]. f (r) sin(anr)dr. (24)
0
Fractional order calculus: historical apologia, basic concepts and some applications 4302-5

Equations (23) and (24) represent the diusion and
the wave solutions. These are special cases of the so-
lution (Eq. (22)) of the fractional diusion-wave equa-
tion.
3.2. Applications in food engineering
Food gums are complex carbohydrates and can serve
a wide variety of functions, ranging from stabilizers to
fat replacers. In many foods, these ingredients are also
useful for building and/or modifying the products tex-
ture.
In the special case where the material and thickness
are uniform throughout, there will be bending stresses
that can be obtained from the normal relationship
3.3. Applications in robotics and control the-
ory
In industrial environments, robots have to execute
their tasks quickly and precisely, minimizing produc-
tion time. This requires exible robots working in large
workspaces, which means that they are inuenced by
nonlinear and fractional order dynamic eects.
Ferreira et al.[17]analyzed the eect of a hybrid force
and position fractional controller applied to two robotic
arms holding the same object, as illustrated in Fig. 1.

d2
(t) = k. (25)
dt2
For a material that is neither a Hookean solid nor
a Newtonian uid, Ma and Barbosa-Canovas [15] an-
alyzed experimental results from elastoviscous bodies
obtained by Bosworth [16] and proposed the following
relationship
d
(t) = k. (26)
dt
with k = constant, 0 1, represents shear stress
and is the shear strain.
Applying the Boltzmann superposition principle in Figure 1 - A cooperative cell of robots performing a given task.
combination with the fractional derivative concept, Eq.
(26) has can be written as follows The load of the object varied and some disturbances
were applied as references of force and position. In-

N
dn stead of using the classic PID (Proportional-Integral-
(t) = Kn . (27)
n=1
dtn Derivative) controller, they used a PI D controller
which was calibrated by trial and error. The result-
Equation (27) rewritten in terms of a fractional op-
ing controller proved to be robust in handling variable
erator (Laurent sense, Table 1) is
loads and small disturbances at the reference.

N
Another interesting problem in robotics which can
(t) = Kn Dn [(t)] . (28)
be treated with FOC is the control of exible robots, as
n=1
this kind of light robot uses low-power actuators, with
Equation (28) is an expression for the linear vis-
no self-destruction eects in response to high impacts.
coelasticity of materials in terms of a fractional deriva-
Nevertheless, signicant vibrations over exible links
tive. From a practical point of view, the material func-
make a position control dicult to design, because it
tions, such as dynamic viscosity ( ) and complex vis-
reveals a complex behavior that is dicult to approx-
cosity ( ), need to be derived implicitly from stress-
imate by linear dierential equations [18]. However,
strain relations. Therefore, the theory of fractional
Monje et al. [19] propose a PD for a exible robot
derivatives, when employed to manipulate these ma-
with one degree of freedom for variable loads, resulting
terial functions based on stress-strain relations, results
in a system with static phase and constant overshoot,
in
independent of the applied load.
An application in a robot with legs was presented
= k1 1 1 sin(1 ) + k2 2 1 sin( 2 ), (29)
2 2 by Silva and Machado [20], who wrote a set of PD
1 1 2 1 algorithms to control position and force, which they
= k1 cos( 1 ) + k2 cos( 2 ). (30)
2 2 applied to a hexapod robot with 12 degrees of freedom.
Equations (29) and (30) can then be used to sim- The authors dened two performance metrics, one for
ulate the linear viscoelasticity of food gums, once the quantity of energy and the other for position error. The
constants in the equations have been determined from controllers with = 0.5 showed the best performance
the experimental data. in this robot.
4302-6 David et al.

3.4. Applications in econophysics If we change variables y0 and y to x and t, and re-

place F with f , the integral equation becomes
Econophysics is a new interdisciplinary area in which
concepts and analysis techniques commonly used for x
1/2
describing physical systems are applied to investigate K = (x t) f (t) dt. (37)
nancial and economic problems. 0
The dynamics of global markets require full time
From this point on the problem is to determine the
complete and highly accurate modeling.
function f . This can be done by multiplying the last
Although numerous studies about nancial markets
equation by 1/(1/2) in order to obtain
have been published [21-27], only a few of them use
fractional calculus as a tool. Nevertheless, some inves-
tigations in nance using fractional calculus equations x
K 1 1/2
were made by Scalas et al. [28]. = (x t) f (t) dt =
David [29] proposed a very simple model based on (1/2) (1/2)
0
the following discussion: It is well- known that a sig- 1/2
0 Dx f (x) . (38)
nicant change in capital ow commonly occurs when
investors, at any moment, perceive a minor change in Consequently,
risk.
It seems to me reasonable to imagine that the square 1/2

( )2 0 Dx K = f (x) . (39)
of the capital ow invested, denoted as d dt , can be
Based on Laurents general denition (Table 1) of
proportional to the dimensionless perception of this risk
derivatives, let us now consider the derivative of order
variation, denoted as (y y0 ).
1/2 of the constant K. Using D = Dm we nd
Mathematically,
that
( )2
d
(y y0 ) , (31)
dt m x
d 1 1
can be written as

0 Dx K = (x t) K dt =
dxm ()
0
( )2 [ ]
d dm K x K p! x m
= C (y y0 ) . (32) = =
dt dxm () ( m)! ()
K
An increase in risk perception stimulates a reduc- x . (40)
tion in capital injection. That is the meaning of the (1 )
minus signal above. Thus
In particular, when = 1/2, we have
d 1/2
= C 1/2 (y0 y) , (33)
dt K K
1/2
0 Dx K = x 1/2 = x 1/2 . (41)
and (1/2)

d Therefore, from Eqs. (39) and (41) it can be con-

1/2
= C 1/2 dt. (34) cluded that
(y0 y)

Now, integrating both sides, K = x f (x). (42)
This model involving risk and capital return based
y0
T
1/2 on the non-integer order calculus concepts is simple to
C dt = (y0 y) d, (35)
implement and oers an interesting alternative way for
0 0
investigation and possibly for predictions in nancial
or markets.
y0
1/2
K = (y0 y) d, (36) 4. Discussion and conclusions
0
Fractional Order Calculus (FOC) dates back to the
1/2
where K = C T . birth of the theory of dierential calculus or Integer
Here, = F (y), where is the amount of return Order Calculus (IOC). However, FOC only began to be
on capital and y represents the risk perception. applied in the last two decades as a result of advances
Bearing in mind this fact, we can note that d = in the area of chaos, which revealed subtle relationships
F (y)dy. with the FOC concepts.
Fractional order calculus: historical apologia, basic concepts and some applications
Recent progress in the area of fractional derivatives
and integrals implies a promising potential for future
developments and application of the theory in various
scientic areas. Some basic concepts of FOC and sev-
eral applications in applied sciences and engineering
have been presented.
The treatment of fractional order calculus in this
paper is suggestive rather than rigorous in order to cap-
ture the readers interest while simultaneously oering
a hint of its potential as a research tool. A few appli-
cations have begun to appear, but they are still in the
incipient stage of development.
This article presented several case studies involving
the implementation of FOC-based models, whose re-
sults demonstrate the importance of Fractional Order
Calculus. We strongly hope they will serve as motiva-
tion for the development of new applications.
Refer
encias
[1] B.M. Hambly and M.L. Lapidus, Trans. Amer. Math.
Soc. 358, 28 (2006).
[2] F. Mainardi, Chaos, Solitons & Fractals 7, 1461 (1996).
[3] A.L. Mehaute, New Journal of Chemistry 14, 207
(1990).
[4] K.B. Oldham and J. Spanier, The Fractional Calculus:
Theory and Applications of Dierentiation and Integra-
tion to Arbitrary Order (Dover Publication, Mineola,
2006).
[5] R.L. Bagley and P.J. Torvik, Journal of Rheology 3,
133 (1986).
[6] J.A.T. Machado, Journal Systems Analysis, Modelling,
Simulation 27, 107 (1997).
[7] J.A.T. Machado and A. Azenha, in: IEEE Int. Work-
shop on Advanced Motion Control, Coimbra, p. 548-553
(1998).
[8] J.A.T. Machado, Nonlinear Dynamics, Chaos, Control
and Their Applications to Engineering Sciences 4, 99
(2002).
[9] H.M. Ozaktas, O. Arikan, M.A. Kutay and G. Bozdagi,
IEEE Trans. on signal Processing 44, 2141 (1996).
[10] A. Oustaloup, IEEE Trans. on Cicuits and Systems 28,
1007 (1981).
[11] B. Mathieu, L.L. Lay and A. Oustaloup, in: IEEE-
SMC/IMACS Symposium on Control, Optimization
and Supervision, Lille, p. 952-956 (1996).
4302-7
[12] B. Ross, Mathematics Magazine 50, 115 (1977).
[13] M.D. Ortigueira, J.A.T. Machado and J.S. da Costa,
IEEE Proceeedings: Vision, Image and Signal Process-
ing 152, 846 (2005).
[14] O.P. Agrawall, Nonlinear Dynamics 29, 145 (2002).
[15] L. Ma and G.V. Barbosa-C
anovas, Journal of Texture
Studies 27, 307 (1996).
[16] R.C.L. Bosworth, Nature 137, 447 (1946).
[17] N.M.F. Ferreira, J.A.T. Machado, A.M.S.F. Galhano
and J.B. Cunha, in: Proceedings of the 2 nd IFAC
Workshop on Fractional Dierentiation and its Appli-
cations,p. 374-379, v. 2 (2006).
[18] M.F.M. Lima, J.A.T. Machado and M. Cris, in: Pro-
ceedings of the 2 nd IFAC Workshop on Fractional Dif-
ferentiation and its Applications, p. 12-18, v. 2 (2006).
[19] C.A. Monje, F. Ramos, V. Feliu and B.M. Vinagre, in:
IET Control Theory and Applications, v. 2 (2007).
[20] M.F. Silva and J.A.T. Machado, Journal of Vibration
and Control 12, 1483 (2006).
[21] B.M. Hambly, S. Howison and T. Kluge, Quantitative
Finance 9, 937 (2009).
[22] S.A. David and E.L. Ursini, in: 6th Brazilian Con-
ference on Dynamics, Control and Their Applications,
Sao Jose do Rio Preto, v. 1, p. 999-1003 (2007).
[23] M.B. Garman and M.J. Klass, Journal of Business 53,
67 (1980).
[24] W.T. Shaw, Modelling Financial Derivatives with
Mathematica (Cambridge University Press, Cam-
bridge, 1998).
[25] M.B. Garman, Descomposici on de los Elementos Com-
ponentes del VeR Dentro de una Cartera de In-
version (Financial Engineering Associates, Inc., Berke-
ley, 1997), p. 2-7.
[26] L. Akdeniz and W.D. Dechert, in: Proc: Second Inter-
national Conference on Computing in Economics and
Finance, Geneva (1996).
[27] E.W. Anderson and L.P. Hansen, in: Second Interna-
tional Conference on Computing in Economics and Fi-
nance, Geneva, p. 1-13 (1996).
[28] E. Scalas, R. Goreno and F. Mainardi, Physica A 284,
376 (2000).
[29] S.A. David, in: Proceddings of the 2007 Iternational
Conference on Computational and Mathematical Meth-
ods in Science and Engineering, Chicago, p. 148-158,
v. 1 (2007).