Proceedings of the 2nd IFAC

Workshop on Fractional Differentiation and its Applications

Porto, Portugal, July 19-21, 2006

CAPUTO LINEAR FRACTIONAL

DIFFERENTIAL EQUATIONS

A. A. Kilbas ,1 M. Rivero ,1
L. Rodrguez-Germa ,1 J. J. Trujillo ,1

Department of Mathematics and Mechanics,
Belarusian State University, Minsk 220050,
Belarus. Email: [email protected]

Departamento de Analisis Matematico,
Universidad de la Laguna, 38271-La Laguna. Tenerife ,
Spain. Email: [email protected]

Departamento de Matematica Fundamental,
Universidad de la Laguna, La Laguna-Tenerife 38271,
Spain. Email: [email protected]; [email protected]

Abstract: This paper is devoted to the study of nonsequential linear fractional

differential equations with constant coefficients involving the Caputo fractional
derivatives. The Laplace transform is applied to obtain the general explicit
solutions for the equations being studied in terms of Mittag-Leffler functions
and generalized Wright functions. Conditions are given for obtaining linearly
independent solutions which form a fundamental system of solutions. Some
examples are presented.

Keywords: fractional differential equations, Caputo fractional derivative, Cauchy

type problems, Mittag-Lefler function, generalized Wright function

1. INTRODUCTION Such a derivative C D0+

y of order > 0 is defined
by
Our paper is devoted to the study of the explicit " #!
l1 (k)
X
general solution to the nonhomogeneous linear C y (a) k
fractional differential equation (LFDE) ( D0+ y)(x) = D0+ y(t) t (x),
k!
m k=0
X k
(1.2)
Ak C D0+ y (x) + A0 y(x) = f (x), (1.1)
with l = [Re() + 1], for 6 N0 = {0, 1, 2, };
k=1
and l = for N0 . Here D0+ y is the Riemann-
where x > 0; m N; 0 < 1 < < m , with
Liouville fractional derivative of order > 0 (see,
constant real coefficients Ak R (k = 0, , m),
for example, (Samko et al., 1993, (5.8))):
Am 6= 0, and a given function f (x) on the half-
axis R+ = (0, ) involving the so-called Caputo l Z x
1 d y(t)dt
fractional derivatives C D0+k
y (k = 1, , m). D0+ y (x) = l+1
,
(l ) dx 0 (x t)
(1.3)
1 This investigation was partially funded by the Belaru- with l = [] + 1 and x > 0.
sian Fundamental Research Fund, by ULL, by DGUI of
G.A.CC. (PI2003/133), and by MEC of Spain (MTM2004- If l 1 < < l (l N) and y(x) is continuously
00327). differentiable up to order l N on any finite
 X n
interval [0, d] (0 < d < ), then (1.2) can be (n + i + 1) z i
represented by = E, (z). (1.7)
(n + + i) i! z
Z x i=0
1 y (i) (t)dt
(C D0+

y)(x) = , (1.4) We note that the Laplace transform method has
(n 0 (x t)l+1
been used by several authors, including Miller,
with x > 0 and l 1 < < l. The derivative Ross, Gorenflo, Mainardi, Podlubny, and so on,
(C D0+

y)(x) in the form (1.4) was introduced in to solve some particular cases of (1.1), and also
(Caputo, 1967) and other monographs. See in to solve Cauchy type problems for such an equa-
this regard (Gorenflo and Mainardi, 1997) and tion, [see, for example, (Miller and Ross, 1993,
(Podlubny, 1999, Section 2.4.1). Sections V.5-V.9), (Gorenflo and Mainardi, 1997),
(Podlubny, 1999, Sections 4.1.1 and 4.1.2)), and
First, we establish a fundamental system of solu-
(Kilbas and Trujillo, 2001, Section 6)]. In some
tions for the homogeneous equation corresponding
cases, the authors focused their attention in solv-
to (1.1), then we obtain an explicit particular
ing the same problem we deal with here, but
solution to equation (1.1) by applying the classical
through the use of the Riemann-Liouville frac-
Laplace transform.
tional operator in the LFDE. See, for example,
Without losing generality we can consider equa- (Miller and Ross, 1993, Sections V.5-V.9), where
tion (1.1) with Am = 1. Our investigation is based the authors obtained a particular solution to equa-
on the following formula for the Laplace trans- tion (1.1) with the Riemann-Liouville fractional
form of the Caputo fractional derivative, which derivative.
is well known, see, for example, (Podlubny, 1999,
(2.252)): Our paper is organized as follows: Section 2 is ded-
icated to obtaining an explicit solution to equation
l1
X
C
(1.1), and we present several nontrivial particular
L D0+ y (s) = s (Ly) (s) dj sj1 ,
cases as well; in Section 3 we present some ap-
j=0
plications of such results to obtain the particular
where l 1 < l, l N, and with solution for certain Cauchy type problems, and we
dj = y (j) (0) (j = 0, 1, , l 1). give selected examples.
We also indicate that equations of the form (1.1)
The general solutions of (1.1) can be repre- arise in many applications; see, in this regard,
sented in terms of the well known generalized (Spanier, 1974), (Podlubny, 1999), (Kilbas and
Mittag-Leffler entire function, which, for complex Trujillo, 2001), and (Kilbas, A. A. and Trujillo,
z, , C (Erdelyi et al., 1954, Section 18.1), is 2006).
given by

We have omitted presenting the proof of the re-
X zk sults obtained in this paper because the tech-
E, (z) = , (1.5)
j=0
(j + ) niques used are similar to those used in the classi-
cal case. For more details see the preprints (Kilbas
or, in terms of the generalized Wright function and Trujillo, 2005a, 2005b).
p q (z) defined for z C, complex ai , bj C, and
real i , j R (i = 1, 2, p; j = 1, 2, , q), by
the series 2. GENERAL SOLUTION OF NON-
SEQUENTIAL LFDE
(ai , i )1,p
p q (z) p q
| z =
(bi , i )1,q In this Section we establish some theorems nec-
essary to obtain an explicit general solution to
X Qp
(ai + i k) z k equation (1.1), as well as some especially relevant
Qqi=1 . (1.6) particular cases of equation (1.1) through the use
j=1 (bj + j k) k!
k=0
of the Laplace transform.
We only indicate here the following particular
cases: first, when p = q = 1 and a1 = 1 = 1, By way of introduction we present first the follow-
then ing two theorems, related to the simplest form of

(1, 1) the particular equation (1.1):
1 1
| z = E,b (z) (b, R),
(b, ) Theorem 1. Let l 1 < l (l N) and R.
Then the functions
and the second case, when , R, n N and
p = q = 1, then yj (x) = xj E,j+1 (x ) ,
with j = 0, , l1, yield the fundamental system
(n + 1, 1)
| z := of solutions for the equation
1 1 c
(n + , ) D0+ y (x) y(x) = 0
Theorem 2. Let R and let f (x) be a given with
m2
X
function on R+ . Then the general solution of the
equation q = ( )n + + ( )k ,
C =0
D0+ y (x) y(x) = f (x)
provided that the series (2.3) and the integral
with 0 < 1 and 1 < 2 has the forms on the right-hand side of (2.2) are convergent.
Z x
A summation in the inner sum is taken for all
y(x) = (x t)1 E, [(x t) ] f (t)dt+ k0 , , km2 N0 such that k0 + + km2 = n.
0
c0 E (x ) ,
Theorem 4. Let m 3, and let , , m2 , , 0
and and l, lm1 , , l0 N0 be such that
Z x
y(x) = (x t)1 E, [(x t) ] f (t)dt+ 0 = 0 < 1 < m2 < < ,
0
+c0 E (x ) + c1 xE,2 (x ) , 0 = l0 l1 lm1 l,
respectively, where c0 and c1 are arbitrary real l 1 < l, lm1 1 < lm1 ,
constants.
lk 1 < k lk (k = 1, , m 2),
As in the ordinary case, if we know a fundamental and let , A0 , , Am2 R. Then a set of
system of solutions for the homogeneous equation solutions for equation (2.1), with f (x) = 0, is
associated with (1.1), then we need only obtain given by the formulas
one particular solution to (1.1) to explicitly ex-
X X
press the general solution for (1.1). We should 1
yj (x) =
point out that an existence and uniqueness the- n=0
k 0 ! km2 !
k0 ++km2 =n
orem for the corresponding Cauchy problem as-
"m2 #
sociated with (1.1) has yet to be established, and Y Pm2
thus we cannot assure, although it is probable, (A ) k
x()n+j+ =0 ( )k
that the general solution to (1.1) contains all the =0

solutions to said equation. (n + 1, 1)
1 1
| x
For our purposes it is convenient to rewrite equa-
tion (1.1) in the form (R1 , )

(C D0+

y)(x) (C D0+ y)(x) (n + 1, 1)
x 1 1 | x
m2
X k (R2 , )
Ak (C D0+ y)(x) A0 y(x) = f (x) (2.1)
k=1 m2
X (n + 1, 1)
with x > 0, m 3, 0 < 1 < < m2 < < Ak xk 1 1 | x ,

, and , A0 , , Am2 R. k=i (R3 , )
(2.4)
The following two theorems bring together some
when 0 j lm2 1 with li j li+1 1 for
general and interesting results concerning equa-
i = 0, 1, , m 3; by
tion (2.1).
X X
1
Theorem 3. Let m 3, > > m2 > > yj (x) =
k0 ! km2 !
1 > 0 = 0, let , A0 , , Am2 R, and let n=0 k0 ++km2 =n
f (x) be a given function on R+ . Then equation "m2 #
Y Pm2
(2.1) has the following particular solution:
Z x (A )k x()n+j+ =0 ( )k
=0
y(x) = (xt)1 G1 , ,m2 ,,; (xt)f (t)dt,
0 (n + 1, 1)
(2.2) 1 | x
where 1
(R1 , )
G1 , ,m2 ,,; (z) =
"m2 # (n + 1, 1)
X X Y
1 x 1 1 | x ,
(A )k
n=0 k0 ++km2 =n
k 0 ! k m2 ! =0
(R2 , )
Pm2 (2.5)
()n+ ( )k
z =0 for lm2 j lm1 1, and by

(n + 1, 1) X X
1
1 1 | z , (2.3) yj (x =
k0 ! km2 !
(q, ) n=0 k0 ++km2 =n
 "m2 #
Y Pm2 (n + 1, 1)
(A )k x()n+j+ =0
( )k
1 1
| x ,
=0 (n + j + 1, )

(n + 1, 1)
1 1
| x , (2.6)
(R1 , ) The next theorem is a particular case of the
previous one for = 0
for lm1 j l 1, R1 = ( )n + 1 + j +
Pm2
( )k , R2 = ( )(n + 1) + 1 + j +
P=0
m2 Corollary 1. Let x > 0 and alpha > > 0. The
=0 ( )k , and R3 = R1 + k , provided general solution of the equation
that the series in (4.4)-(4.6) are convergent. A C

summation in the inner sums is taken for all D0+ y (x) C D0+ y (x) = f (x)
k0 , , km2 N0 such that k0 + + km2 = n.
has the form
If l + 1 , then yj (x) in (2.4)-(2.6) are Z x

linearly independent solutions of equation (2.1). y(x) = (xt)1 E, (x t) f (t)dt+
0
In particular, for l + 1 > they yield the
fundamental system of solutions. l1
X
cj yj (x),
The following theorems and corollary summarize j=0
results that, in a certain way, can be considered where yj (x) (j = 0, , m 1) are given by
as specific cases of Theorems 3 and 4.
yj (x) = xj E,j+1 x
Theorem 5. Let l1 < l and m1 < m
x+j E,+j+1 x ,
(l, m N, m l) be such that l + 1 , let
, R and let f (x) be a given real function on but, when j = m, , l 1

R+ . Then the equation yj (x) = xj E,j+1 x .
C

D0+ y (x) C D0+ y (x) y(x) = f (x), Theorem 6. The equation
C
with x > 0, l 1 < l, l N, 0 < < and
D0+ y (x) C D0+ y (x)
, R, has the following general solution:
Z x
C D0+ y (x) y(x) = f (x),
y(x) = (x t)1 G,;, (x t)f (t)dt+
0 with x > 0, , , R, l 1 < l, l N, and
l1 0 < < < , has the particular solution
X Z x
cj yj (x),
j=0
y(x) = (x t)1 G,,; (x t)f (t)dt,
0
where cj (j = 1, , l 1) are arbitrary real
where
constants, and G,;, (z) is given by

X X i

X
n (n + 1, 1) G,,; (z) = z ()n+i()
z n 1 1 | z , n=0 i+=n
i!!
n!
n=0 (n + , )
(n + 1, 1)
provided that such a series is convergent, and 1 1
| z ,
yj (x) (j = 0, , m 1) are given by (Q, )
X n
n+j with Q = ( )n + + i + ( ).
yj (x) = x
n=0
n! Remark 1. We should point out that some formal
results analogous to those set out here, but in-
(n + 1, 1)
volving LFDE with Riemann-Liouville fractional
1 1 | x
derivatives, were presented in (Podlubny, 1999).
(n + j + 1, )
X n
n+j+
x
n=0
n! 3. EXAMPLES

(n + 1, 1)
| x , In this section we present some examples which
1 1
(n + j + 1 + , ) apply the results obtained in the previous section.
but, when j = m, , l 1 Example 1. Let x > 0 and , , R. Then
X n Z x
n+j
yj (x) = x y(x) = (x t)3/2 G1/2,3/2,5/2; (x t)f (t)dt,
n=0
n!
0
where Example 4. Let x > 0, 0 < < 1, , R. The


XX i Cauchy problem
3i
G1/2,3/2,5/2; (z) = z n+ 2
n=0 i+=n
i!!
y 00 (x) C D0+ y (x) y(x) = f (x),

(n + 1, 1) with the initial conditions y(0) = b0 and


1 1
| z , y 0 (0) = b1 (b0 , b1 R), has the solution
5 3i Z x
n+ + + , 1
2 2 y(x) = (x t)G2,;, (x t)f (t)dt+
is a particular solution of the equation 0

C 5/2 3/2
D0+ y (x) C D0+ y (x) X (n + 1, 1)
n 2n
b0 x 1 1 | x2
n!
1/2
C D0+ y (x) y(x) = f (x). n=0 (2n + 1, 2 )

X (n + 1, 1)
n 2n+2
Example 2. Let x > 0, l N, and R. The x 1 1 | x2 +
general solution of equation n!
n=0 (p, 2 )

C l1/2
D0+ y (x) y(x) = f (x), X n (n + 1, 1)
2n+1
+b1 x 1 1 | x2 ,
has the form n!
Z x h i
n=0 (2n + 2, 2 )
y(x) = (x t)l3/2 El1/2,l1/2 (x t)l1/2 where p = 2n + 3 and G2,;, (z) is given by
0
l1
X X (n + 1, 1)
f (t)) dt + j
cj x El1/2,j+1 x l1/2
, n 2n
z 1 1 | z 2 ,
j=0 n!
n=0 (2n + 2, 2 )
where cj (j = 0, 1, , l 1) are arbitrary real
constants.
In particular, REFERENCES
Z x h i
y(x) = (xt)1/2 E3/2,3/2 (x t)3/2 f (t)dt+ A. A. Kilbas, H. M. Srivastava and J. J. Trujillo
0
(2006). Theory and Applications of Fractional
+c0 E3/2 x3/2 + c1 xE3/2,2 x3/2 , Differential equations. North-Holland Math-
ematics Studies, 204 , Elsevier. New York-
with arbitrary real c0 and c1 , is the general solu- London.
tion of equation Caputo, M. (1967). Lineal model of dissipation
whose q is almost frequency independent II.
Geophys. J. Astronom. Soc. 13, 529539.
C 3/2
D0+ y (x) y(x) = f (x) (x > 0; R).
Erdelyi, A., W. Magnus, F. Oberhettinger and
F.G. Tricomi (1954). Higher Transcendental
Example 3. Let x > 0, 0 < < 1, and , R. Functions Vol.3. McGraw-Hill Book Corp.
The Cauchy problem New York.
Gorenflo, R. and F. Mainardi (1997). Frac-

y 0 (x) C D0+ y (x) y(x) = f (x), tional calculus, integral and differential equa-
with the initial condition y(0) = b R has the tions of fractional order. In: Fractals and
solution Fractional Calculus in Continuum Mechanics
Z x (A. Carpinteri and F. Mainardi, Eds.), 223
y(x) = G1,;, (x t)f (t)dt+ 276. Springer Verlag. New York.
0 Kilbas, A. A. and J. J. Trujillo (2001). Differential

X (n + 1, 1) equations of fractional order: methods, results
n n
b x 1 1 | x1 and problems. Appl. Anal. 78(1-2), 153192.
n!
n=0 (n + 1, 1 ) Kilbas, A. A. and J. J. Trujillo (2005a). Solutions
of homogeneous linear differential equations
X n (n + 1, 1)

xn+1 1 1 with Caputo derivatives and constant coeffi-
| x1 ,
n! cients. Preprint. University of La Laguna.
n=0 (P, 1 )
Kilbas, A. A. and J. J. Trujillo (2005b). Solutions
where P = n + 2 , and G1,;, (z) is given by
of nonhomogeneous linear fractional differen-

X (n + 1, 1) tial equations with Caputo derivatives and
n n
z 1 1 | z 1 . constant coefficients. Preprint. University of
n! La Laguna.
n=0 (n + 1, 1 )
Miller, K. S. and B. Ross (1993). An Introduc-
tion to the Fractional Calculus and Fractional
Differential Equations. John Wiley and Sons.
New York.
Podlubny, I. (1999). Fractional Differential Equa-
tions. Academic Press. San-Diego.
Samko, S. G., A. A. Kilbas and O. I. Marichev
(1993). Fractional Integrals and Deriva-
tives. Theory and Applications. Gordon and
Breach. Yverdon.
Oldham, K. B. J. Spanier (1974). The Fractional
Calculus. Academic Press. New York-London.