RESEARCH PAPER

ON THE OSCILLATION OF
FRACTIONAL DIFFERENTIAL EQUATIONS

Said R. Grace 1 , Ravi P. Agarwal 2 ,

Patricia J.Y. Wong 3 , Ağacık Zafer 4

Abstract
In this paper we initiate the oscillation theory for fractional differential
equations. Oscillation criteria are obtained for a class of nonlinear fractional
differential equations of the form
Daq x + f1 (t, x) = v(t) + f2 (t, x), lim Ja1−q x(t) = b1 ,
t→a+
where Daq denotes the Riemann-Liouville differential operator of order q,
0 < q ≤ 1. The results are also stated when the Riemann-Liouville differ-
ential operator is replaced by Caputo’s differential operator.
MSC 2010 : Primary 34A08: Secondary 34C10, 26A33
Key Words and Phrases: fractional differential equation, oscillation,
Riemann-Liouville operators, Caputo derivative

1. Introduction
Fractional differential equations (FDE) have gained considerable im-
portance due to their various applications in viscoelasticity, electroana-
lytical chemistry, control theory, many physical problems, etc, see e.g.
[2, 5, 7, 10, 11]. A rigorous theory of FDE has been started quite recently,
see for example - the books [4, 9, 11] and to mention only few papers related
to ordinary FDE as [3, 6, 8]. Separately, there are many recent works related

c 2012 Diogenes Co., Sofia
pp. 222–231 , DOI: 10.2478/s13540-012-0016-1
 ON THE OSCILLATION OF . . . 223

to partial FDE. Differential equations involving the Riemann–Liouville and

Caputo differential operators of fractional order 0 < q < 1 appear to be
important in modeling several physical phenomena, and therefore deserve
an independent study parallel to the well-developed theory of the ordinary
differential equations of integer order.
In this paper we initiate the study of the oscillation theory for fractional
differential equations by considering equations of the form
Daq x + f1 (t, x) = v(t) + f2 (t, x), lim Ja1−q x(t) = b1 , (1.1)
t→a+

where Daq denotes the Riemann-Liouville differential operator of order q

with 0 < q ≤ 1. The operator Jap defined by
 t
p 1
Ja x(t) = (t − s)p−1 x(s) ds, Ja0 x := x
Γ(p) a
is called the Rieman-Liouville fractional integral operator. By using the in-
tegral operator Jap , the Riemann-Liouville differential operator Daq of order
q for 0 < q ≤ 1 is defined by
d 1−q
Daq f (t) :=
J f (t),
dt a
and more generally if m ≥ 1 is an integer and m − 1 < q ≤ m, then
dm m−q
Daq f (t) :=
J f (t).
dtm a
For more details on the Riemann-Liouville type operators of fractional cal-
culus, see for example [12].
We assume in this paper that the functions f1 , f2 , and v are continuous.
It is known that in this case, in fact under much weaker assumptions, the
initial value problem (1.1) is equivalent to the Volterra fractional integral
equation
 t
b1 (t − a)q−1 1
x(t) = + (t − s)q−1 [v(s) + f2 (s, x(s)) − f1 (s, x(s))]ds.
Γ(q) Γ(q) a
(1.2)
That is, every solution of (1.2) is also a solution of (1.1) and vice-versa, see
[4, Lemma 5.3].
We only consider those solutions which are continuous and continuable
to (a, ∞), and are not identically zero on any half-line (b, ∞) for some b ≥ a.
The term “solution” henceforth applies to such solutions of equation (1.1) or
(1.2). For an introduction to the theory of fractional differential equations,
we refer to [4, 11, 12]. A solution is said to be oscillatory if it has arbitrarily
large zeros on (0, ∞); otherwise, it is called nonoscillatory.
224 S.R. Grace, R.P. Agarwal, P.J.Y. Wong, A. Zafer

2. Main Results
We will make use of the conditions:
xfi (t, x) > 0 (i = 1, 2), x = 0, t ≥ a (2.1)
and
|f1 (t, x)| ≥ p1 (t)|x|β and |f2 (t, x)| ≤ p2 (t)|x|γ , x = 0, t ≥ a, (2.2)
where p1 , p2 ∈ C([a, ∞), R+ ) and β, γ > 0 are real numbers.

We will employ the following lemma [1, Lemma 3.2.1].

Lemma 2.1. For X ≥ 0 and Y > 0, we have

X λ + (λ − 1)Y λ − λXY λ−1 ≥ 0, λ > 1 (2.3)
and
X λ − (1 − λ)Y λ − λXY λ−1 ≤ 0, λ < 1, (2.4)
where equality holds if and only if X = Y.

Now we may present our first theorem when f2 = 0.

Theorem 2.1. Let f2 = 0 and condition (2.1) hold. If

 t
1−q
lim inf t (t − s)q−1 v(s)ds = −∞, (2.5)
t→∞ a
and  t
1−q
lim sup t (t − s)q−1 v(s)ds = ∞, (2.6)
t→∞ a
then every solution of equation (1.1) is oscillatory.

P r o o f. Let x(t) be a nonoscillatory solution of equation (1.1) with

f2 = 0. Suppose that T > a is large enough so that x(t) > 0 for t ≥ T .
Let F (t) = v(t) + f2 (t, x(t)) − f1 (t, x(t)), then we see from (1.2) that
 T
(t − a)q−1 1
x(t) ≤ |b1 | + (t − s)q−1 |F (s)|ds
Γ(q) Γ(q) a
 t
1
+ (t − s)q−1 v(s)ds, t ≥ T,
Γ(q) T
and hence
 t
1−q 1−q
Γ(q) t x(t) ≤ c(T ) + t (t − s)q−1 v(s)ds, t ≥ T, (2.7)
T
 ON THE OSCILLATION OF . . . 225

where and in the proof of the following theorems:

 1−q  T 1−q
T T
c(T ) = |b1 | + |F (s)| ds.
T −a a T −s
Note that the improper integral on the right is convergent. Taking the limit
inferior of both sides of inequality (2.7) as t → ∞, we get a contradiction
to condition (2.5). In case x(t) is eventually negative, a similar argument
leads to a contradiction with (2.6). 2
Next we have the following results.

Theorem 2.2. Let conditions (2.1) and (2.2) hold with β > 1 and
γ = 1. If
 t
1−q
lim inf t (t − s)q−1 [v(s) + Hβ (s)]ds = −∞, (2.8)
t→∞ a
and  t
lim sup t1−q (t − s)q−1 [v(s) + Hβ (s)]ds = ∞, (2.9)
t→∞ a
where
1/(1−β) β/(β−1)
Hβ (s) = (β − 1)β β/(1−β) p1 (s)p2 (s),
then every solution of equation (1.1) is oscillatory.

P r o o f. Let x(t) be a nonoscillatory solution of equation (1.2), say,

x(t) > 0 for t ≥ T > a. Using (2.2) in equation (1.2) with γ = 1 and β > 1
and t ≥ T , we find
 t
Γ(q) t1−q x(t) ≤ c(T ) + t1−q (t − s)q−1 v(s)ds
T
 t 
+ (t − s)q−1 [p2 (s)x(s) − p1 (s)xβ (s)]ds . (2.10)
T
We apply (2.3) in Lemma 2.1 with
 1/(β−1)
1/β −1/β
λ = β, X = p1 x and Y = p2 p1 /β
to obtain
1/(1−β) β/(β−1)
p2 (t)x(t) − p1 (t)xβ (t) ≤ (β − 1)β β/(1−β) p1 (t)p2 (t). (2.11)
Using (2.11) in (2.10), we have
 t
1−q 1−q
Γ(q) t x(t) ≤ c(T ) + t (t − s)q−1 [v(s) + Hβ (s)]ds, t ≥ T.
T
The rest of the proof is the same as in that of Theorem 2.1. 2
226 S.R. Grace, R.P. Agarwal, P.J.Y. Wong, A. Zafer

Theorem 2.3. Let conditions (2.1) and (2.2) hold with β = 1 and
γ < 1. If
 t
1−q
lim inf t (t − s)q−1 [v(s) + Hγ (s)]ds = −∞, (2.12)
t→∞ a
and  t
1−q
lim sup t (t − s)q−1 [v(s) + Hγ (s)]ds = ∞, (2.13)
t→∞ a
where
γ/(γ−1) 1/(1−γ)
Hγ (s) = (1 − γ)γ γ/(γ−1) p1 (s)p2 (s),
then every solution of equation (1.1) is oscillatory.

P r o o f. Let x(t) be a nonoscillatory solution of equation (1.2), say,

x(t) > 0 for t ≥ a > 1. Using condition (2.2) in equation (1.2) with β = 1
and γ < 1, we obtain
 t
1−q 1−q
Γ(q) t x(t) ≤ c(T ) + t (t − s)q−1 v(s)ds
a
 t 
+ (t − s)q−1 [p2 (s)xγ (s) − p1 (s)x(s)]ds . (2.14)
a
Now we use (2.4) in Lemma 2.1 with
 1/(γ−1)
1/γ −1/γ
λ = γ, X = p2 x and Y = p1 p2 /γ
to get
γ/(γ−1) 1/(1−γ)
p2 (t)xγ (t) − p1 (t)x(t) ≤ (1 − γ)γ γ/(1−γ) p1 (t)p2 (t). (2.15)
Using (2.15) in (2.14) then yields
 t
Γ(q) t1−q x(t) ≤ c(T ) + t1−q (t − s)q−1 [v(s) + Hγ (s)]ds, t ≥ T.
T
The rest of the proof is the same as in that of Theorem 2.1. 2

Finally we present the following more general result.

Theorem 2.4. Let conditions (2.1) and (2.2) hold with β > 1 and
γ < 1. If
 t
1−q
lim inf t (t − s)q−1 [v(s) + Hβ,γ (s)]ds = −∞, (2.16)
t→∞ a
and  t
1−q
lim sup t (t − s)q−1 [v(s) + Hβ,γ (s)]ds = ∞, (2.17)
t→∞ a
 ON THE OSCILLATION OF . . . 227

where
1/(1−β)
Hβ,γ (s) = (β − 1)β β/(1−β) ξ β/(β−1) (s)p1 (s)
1/(1−γ)
+(1 − γ)γ γ/(1−γ) ξ γ/(γ−1) (s)p2 (s)
with ξ ∈ C([a, ∞), R+ ), then every solution of equation (1.1) is oscillatory.

P r o o f. Let x(t) be a nonoscillatory solution of equation (1.1), say,

x(t) > 0 for t ≥ T > a. Using (2.2) in equation (1.2) one can easily write
that
 t
1−q 1−q
Γ(q) t x(t) ≤ c(T ) + t (t − s)q−1 v(s)ds
T
 t  
1−q
+t (t − s)q−1 ξ(s)x(s) − p1 (s)xβ (s) ds
T
 t
+ t1−q (t − s)q−1 (p2 (s)xγ (s) − ξ(s)x(s)) ds, t ≥ T.
T

We may bound the terms (ξx − p1 xβ ) and (p2 xγ − ξx) by using the in-
equalities (2.11) (with p2 = ξ) and (2.15) (with p1 = ξ) respectively, to
get
 t
1−q 1−q
Γ(q) t x(t) ≤ c(T ) + t (t − s)q−1 [v(s) + Hβ,γ (s)]ds, t ≥ T.
T
The rest of the proof is the same as in that of Theorem 2.1. 2

Remark 2.1. The results obtained for (1.1) are with different nonlin-
earities and one can observe that the forcing term v is unbounded, and its
oscillatory character is inherited by the solutions.

Remark 2.2. It is not difficult to see that the results remain valid
for fractional differential equations involving Riemann-Liouville differential
operator Daq of order q with m − 1 < q ≤ m, where m ≥ 1 is an integer, of
the form
Daq x + f1 (t, x) = v(t) + f2 (t, x),
(2.18)
Daq−k x(a) = bk (k = 1, 2, . . . , m − 1), limt→a+ Jam−q x(t) = bm .
It suffices to note that the initial value problem (2.18) is equivalent to the
Volterra fractional integral equation
m  t
bk (t − a)q−k 1
x(t) = + (t−s)q−1 [v(s)+f2 (s, x(s))−f1 (s, x(s))]ds.
Γ(q − k + 1) Γ(q) a
k=1
(2.19)
228 S.R. Grace, R.P. Agarwal, P.J.Y. Wong, A. Zafer

We now give an example to show that the condition (2.8) cannot be

dropped.

Example 2.1. Consider the Riemann-Liouville fractional differential

equation
t1−q
D0q x + et x3 = + tet (t2 − 1) + et x, lim J01−q x(t) = 0, (2.20)
Γ(2 − q) t→0+

where 0 < q < 1.

Clearly, in the context of (1.1) we have
t1−q
a = b1 = 0, f1 (t, x) = et x3 , f2 (t, x) = et x, and v(t) = +tet (t2 −1).
Γ(2 − q)
We see that the conditions (2.1) and (2.2) are satisfied with β = 3, γ = 1
and p1 (t) = p2 (t) = et . But the condition (2.8) is not satisfied, since v(t) ≥ 0
and
 t  t
1−q q−1 β/(β−1) 1/(1−β) 1−q 1 t
t (t − s) p2 (s)p1 (s) ds ≥ t 1−q
ds = .
0 0 (t − s) q
Indeed, one can easily verify that x(t) = t is a nonoscillatory solution of
(2.20). Note that the initial condition is satisfied in view of J01−q x(t) =
t2−q /Γ(3 − q) → 0 as t → 0+ .

3. Caputo’s derivative approach

We may replace the Riemann-Liouville differential operator Daq by the
Caputo differential operator c Daq with m − 1 < q ≤ m, which is defined by
(see [2, 11])
q
c Da f (t) := Jam−q f (m) (t).
Notice that the Caputo differential operator c Daq demands functions to be
m times differentiable.
In this case the initial value problem (2.18) should be replaced by
q
c Da x+f1 (t, x) = v(t)+f2 (t, x), x(k) (a) = bk (k = 0, 1, . . . , m−1). (3.1)
The corresponding Volterra fractional integral equation, see [4, Lemma 6.2],
becomes
m−1  t
bk (t − a)k 1
x(t) = + (t − s)q−1 [v(s) + f2 (s, x(s)) − f1 (s, x(s))]ds.
k! Γ(q) a
k=0
(3.2)
 ON THE OSCILLATION OF . . . 229

The oscillation criteria obtained for the Riemann-Liouville case read

almost the same. The only change is the appearance of t1−m instead of t1−q
in the conditions. This difference becomes more distinctive when m = 1.

Theorem 3.1. Let f2 = 0 and condition (2.1) hold. If

 t
1−m
lim inf t (t − s)q−1 v(s)ds = −∞, (3.3)
t→∞ a
and  t
1−m
lim sup t (t − s)q−1 v(s)ds = ∞, (3.4)
t→∞ a
then every solution of equation (3.1) is oscillatory.

Theorem 3.2. Let conditions (2.1) and (2.2) hold with β > 1 and
γ = 1. If
 t
lim inf t1−m (t − s)q−1 [v(s) + Hβ (s)]ds = −∞, (3.5)
t→∞ a
and  t
1−m
lim sup t (t − s)q−1 [v(s) + Hβ (s)]ds = ∞, (3.6)
t→∞ a
where
1/(1−β) β/(β−1)
Hβ (s) = (β − 1)β β/(1−β) p1 (s)p2 (s),
then every solution of equation (3.1) is oscillatory.

Theorem 3.3. Let conditions (2.1) and (2.2) hold with β = 1 and
γ < 1. If
 t
1−m
lim inf t (t − s)q−1 [v(s) + Hγ (s)]ds = −∞, (3.7)
t→∞ a
and  t
1−m
lim sup t (t − s)q−1 [v(s) + Hγ (s)]ds = ∞, (3.8)
t→∞ a
where
γ/(γ−1) 1/(1−γ)
Hγ (s) = (1 − γ)γ γ/(γ−1) p1 (s)p2 (s),
then every solution of equation (3.1) is oscillatory.

Theorem 3.4. Let conditions (2.1) and (2.2) hold with β > 1 and
γ < 1. If
 t
1−m
lim inf t (t − s)q−1 [v(s) + Hβ,γ (s)]ds = −∞, (3.9)
t→∞ a
230 S.R. Grace, R.P. Agarwal, P.J.Y. Wong, A. Zafer

and
 t
1−m
lim sup t (t − s)q−1 [v(s) + Hβ,γ (s)]ds = ∞, (3.10)
t→∞ a
where
1/(1−β)
Hβ,γ (s) = (β − 1)β β/(1−β) ξ β/(β−1) (s)p1 (s)
1/(1−γ)
+(1 − γ)γ γ/(1−γ) ξ γ/(γ−1) (s)p2 (s)
with ξ ∈ C([a, ∞), R+ ), then every solution of equation (3.1) is oscillatory.

References

[1] R.P. Agarwal, S.R. Grace, and D. O’Regan, Oscillation Theory for
Second Order Linear, Half-linear, Superlinear and Sublinear Dynamic
Equations. Kluwer, Dordrecht, 2002.
[2] M. Caputo, Linear models of dissipation whose Q is almost fre-
quency independent II. Geophys. J. R. Astr. Soc., 13 (1967), 529–539;
Reprinted in: Fract. Calc. Appl. Anal. 11, No 1 (2008), 3–14.
[3] K. Diethelm and N.J. Ford, Analysis of fractional differential equations.
J. Math. Anal. Appl., 265 (2002), 229–248.
[4] K. Diethelm, The Analysis of Fractional Differential Equations.
Springer, Berlin, 2010.
[5] K. Diethelm and A.D. Freed, On the solution of nonlinear fractional
differential equations used in the modelling of viscoplasticity. In: F.
Keil, W. Mackens, H. Vob and J. Werther (Eds.), Scientific Computing
in Chemical Engineering II: Computational Fluid Dynamics, Reaction
Engineering and Molecular Properties, Springer, Heidelberg, 1999, 217–
224.
[6] N.J. Ford, M. Luisa Morgado, Fractional boundary value prob-
lems: Analysis and numerical methods. Fract. Calc. Appl. Anal.
14, No 4 (2011), 554–567; DOI: 10.2478/s13540-011-0034-4; at
http://www.springerlink.com/content/1311-0454/14/4/
[7] W.G. Glöckle and T.F. Nonnenmacher, A fractional calculus approach
to self similar protein dynamics. Biophys. J. 68 (1995), 46–53.
[8] J.R. Graef, L. Kong, B. Yang, Positive solutions for a semipositone
fractional boundary value problem with a forcing term. Fract. Calc.
Appl. Anal. 15, No 1 (2012), 8–24; DOI: 10.2478/s13540-012-0002-7;
at: http://www.springerlink.com/content/1311-0454/14/4/
[9] A.A. Kilbas, H.M. Srivastava, J.J. Trujillo, Theory and Applications
of Fractional Differential Equations. North-Holland Math. Studies 204,
Elsevier, Amsterdam, 2006.
 ON THE OSCILLATION OF . . . 231

[10] R. Metzler, S. Schick, H.G. Kilian and T.F. Nonnenmacher, Relaxation

in filled polymers: A fractional calculus approach. J. Chem. Phys., 103
(1995), 7180–7186.
[11] I. Podlubny, Fractional Differential Equations. Academic Press, San
Diego, 1999.
[12] S.G. Samko, A.A. Kilbas and O.I. Marichev, Fractional Integrals and
Derivatives: Theory and Applications. Gordon and Breach, Yverdon,
1993.

1 Dept. of Engineering Mathematics, Faculty of Engineering

Cairo University, Oman, Giza – 12221, EGYPT
e-mail: srgrace@eng.cu.edu.eg
2 Dept. of Mathematics, Texas A & M University – Kingsville
Kingsville, TX – 78363, USA
e-mail: agarwal@tamuk.edu Received: November 2, 2011
3 School of Electrical and Electronic Engineering
Nanyang Technological University, 50 Nanyang Avenue
Singapore – 639798, SINGAPORE
e-mail: ejywong@ntu.edu.sg
4 Dept. of Mathematics, Middle East Technical University
06800 – Ankara, TURKEY
zafer@metu.edu.tr

Please cite to this paper as published in:

Fract. Calc. Appl. Anal., Vol. 15, No 2 (2012), pp. 222–231;
DOI: 10.2478/s13540-012-0016-1