Nonlinear Dyn (2013) 71:641–652

DOI 10.1007/s11071-012-0443-x

O R I G I N A L PA P E R

Fractional boundary value problems with singularities

in space variables
Donal O’Regan · Svatoslav Staněk

Received: 10 March 2012 / Accepted: 26 April 2012 / Published online: 17 May 2012
© Springer Science+Business Media B.V. 2012

Abstract We are concerned with the existence of chanics, chemistry, engineering, economy, and so on.
solutions for the singular fractional boundary value For examples and details, see [1–9], and references
problem cD α u = f (t, u), u(0) + u(1) = 0, u (0) = 0, therein. There has been a significant development in
where α ∈ (1, 2), f ∈ C([0, 1] × (R \ {0})) and the study of regular fractional boundary value prob-
limx→0 f (t, x) = ∞ for all t ∈ [0, 1]. Here, cD is the lems on compact intervals [10–22] and noncompact
Caputo fractional derivative. Increasing solutions of intervals [23–25].
the problem vanish at points of (0, 1), that is, they Singular fractional boundary value problems hav-
“pass through” the singularity of f inside of (0, 1). ing nonlinearities singular in space variables are dis-
The results are based on combining regularization cussed in [26–32]. Paper [26] investigates the Dirichlet
and sequential techniques with a nonlinear alternative. problem
In limit processes, the Vitali convergence theorem is  
used. D α u + f t, u, D μ u = 0,
u(0) = u(1) = 0,
Keywords Fractional differential equation · Singular
problem · Caputo fractional derivative · Nonlinear where 1 < α < 2, 0 < μ ≤ 1 and f is a Carathéodory
alternative · Vitali convergence theorem function on [0, 1] × (0, ∞) × R. Here, D α denotes the
Riemann–Liouville fractional derivative. In [27], the
problem
1 Introduction  
D α u + f t, u, u , D μ u = 0,
Fractional differential equations are valuable tools in
the modeling of many phenomena in physics, me- u(0) = 0, u (0) = u (1) = 0,
is considered, where 2 < α < 3, 0 < μ ≤ α − 1 and
f is a Carathéodory function on [0, 1] × (0, ∞)3 . The
D. O’Regan
Department of Mathematics, National University of
mixed boundary value problem
 
Ireland, Galway, Ireland
D u + f t, u, u , D μ u = 0,
c α
e-mail: donal.oregan@nuigalway.ie
u (0) = 0, u(1) = 0,
S. Staněk ()
Department of Mathematical Analysis, Faculty of Science, is studied in [28]. Here, 1 < α < 2, 0 < μ < 1 and
Palacký University, 17. listopadu 12, 771 46 Olomouc,
f is a Lq -Carathéodory function on [0, 1] × D, D =
Czech Republic
e-mail: svatoslav.stanek@upol.cz (0, ∞) × (−∞, 0) × (−∞, 0), where q > α−1 1
. cD α is
642 D. O’Regan, S. Staněk

the Caputo fractional derivative. In [26–28], the exis- at points inside of (0, 1). Note that such problems for
tence of positive solutions is proved by combining reg- differential equations are investigated, for example, in
ularization and sequential techniques with fixed-point [33, 34], and the references therein.
theorems on cones. In this paper, we discuss the singular fractional
The problem boundary value problem
 
D α u + f (t, u) = 0, D u(t) = f t, u(t) ,
c α
(1.1)
u(0) + u(1) = 0, u (0) = 0, (1.2)
u(0) = u (0) = 0, u (1) = 0
where α ∈ (1, 2), f ∈ C([0, 1] × (R \ {0})) and
is investigated in [29]. Here, 2 < α < 3 and f ∈
limx→0 f (t, x) = ∞ for all t ∈ [0, 1]. Here, cD is the
C((0, 1) × (0, ∞)). By the Gatica–Oliker–Waltman
Caputo fractional derivative. Since f (t, x) is singular
fixed point theorem, the existence of a positive solu- at the value 0 of the variable x, it follows from the
tion is proved. condition u(0) + u(1) = 0 that increasing and contin-
Paper [30] deals with the problem uous solutions of problem (1.1), (1.2) vanish inside of
D α u = f (t, u), (0, 1) and, therefore, “pass through” the singularity of
f inside of (0, 1).
u(0) = u (0) = 0, u(1) = u (1) = 0, We recall that the Caputo fractional derivative
cD γ x of order γ > 0 of a function x : [0, 1] → R is
where 3 < α ≤ 4 and f ∈ C([0, 1] × (0, ∞)). The ex-
istence of at least one and at least two positive solu- defined as (see, e.g., [2, 3])
⎧ 
tions is proved by a Leray–Schauder nonlinear alter- ⎪ dn t n−γ −1 (x(s)
⎨ (n−γ ) dt n 0 (t − s)
1
⎪
native and by a fixed-point theorem on cones, while x (k) (0) k
D x(t) =
c γ
− n−1 k=0 k! s ) ds if γ ∈ N,
uniqueness is established via a fixed- point theorem ⎪
⎪
⎩ x (γ ) (t) if γ ∈ N,
for mixed monotone operators.
The problem where n = [γ ] + 1 and [γ ] means the integral part of
D α u = μa(t)f (t, u), γ and where  is the Euler gamma function.
We say that an increasing function u : [0, 1] →
u(0) = u (0) = 0, u(1) = u (1) = 0, R is an increasing solution of the singular problem
is discussed in [31]. Here, 3 < α ≤ 4, μ > 0, a ∈ (1.1), (1.2) if u satisfies the boundary conditions (1.2),
L1 [0, 1] is positive, f is continuous, and f : [0, 1] × u(ξ ) = 0 for some ξ ∈ (0, 1), u ∈ C[0, 1] ∩ C 1 ([0, 1] \
(0, ∞) → (0, ∞) (the positone case) or f : [0, 1] × {ξ }), cD α u ∈ C([0, 1] \ {ξ }) and equality (1.1) holds
(0, ∞) → R (the semipositone case). The existence of for t ∈ [0, 1] \ {ξ }.
positive solutions is proved by the Guo–Krasnosel’skii Throughout the paper R0 = R \ {0} and x =
fixed-point theorem on cones.  1 : t ∈ [0, 1]} is the norm in C[0, 1] while
max{|x(t)|
Zhang [32] considered the problem x L = 0 |x(t)| dt is the norm in L1 [0, 1]. For M ⊂
  [0, 1], meas (M) denotes the Lebesgue measure of M.
D α u + q(t)f t, u, u , . . . , u(n−2) = 0, We work with the following conditions on the func-
tion f in (1.1).
u(0) = u (0) = · · · = u(n−2) (0) = 0,
(H1 ) f ∈ C([0, 1] × R0 ), limx→0 f (t, x) = ∞ for all
u(n−2) (1) = 0,
t ∈ [0, 1] and
where α ∈ (n − 1, n), n ≥ 2, f ∈ C((0, ∞)n−1 ) and at 2−α ≤ f (t, x) for (t, x) ∈ [0, 1] × R0 ,
q ∈ Lq [0, 1] (q > 0). By a combination of regulariza-
where a > 0.
tion and sequential techniques with a fixed-point the-
(H2 ) For (t, x) ∈ [0, 1] × R0 the estimate
orem for mixed monotone operators on normal cones,
  A
the existence results for positive solutions are proved. f (t, x) ≤ g |x| + ν
In all these papers, solutions “start” and/or “finish” |x|
at singular points of nonlinearities f . No contribution is fulfilled with g ∈ C[0, ∞) positive and non-
exists, as far as we know, concerning singular frac- decreasing, A > 0, ν ∈ (0, 2(3−α)
α−1
), and
tional boundary value problems on the interval [0, 1] g(x)
lim = 0.
for which solutions “pass through” singularities of f x→∞ x
Fractional boundary value problems with singularities in space variables 643

The existence result for problem (1.1), (1.2) is The Riemann–Liouville fractional integral I γ x of
proved by regularization and sequential techniques. To order γ > 0 of a function x : [0, 1] → R is given as
this end, for n ∈ N we define a function fn acting on [1–3]
[0, 1] × R as 1 t
⎧ I γ x(t) = (t − s)γ −1 x(s) ds.
⎪
⎪ f (t, x) if x ≥ n1 , (γ ) 0
⎪
⎪
⎨ n [f (t, ( 1 ))( 1 + x)
fn (t, x) = 2 n n Lemma 2.1 ([2, 3]) I γ : C[0, 1] → C[0, 1] for γ ∈
⎪ + −( n ))( n − x)] if |x| ≤ n ,
1 1 1
⎪
⎪ f (t, (0, 1), and I β I δ x(t) = I β+δ x(t) for t ∈ [0, 1], where
⎪
⎩
f (t, x) if x ≤ − n1 . x ∈ L1 [0, 1], β + δ ≥ 1.
Then condition (H1 ) gives fn ∈ C([0, 1] × R) and
Note that the equality I β I δ x(t) = I β+δ x(t) in
at 2−α
≤ fn (t, x) for (t, x) ∈ [0, 1] × R. (1.3) Lemma 2.1 can be written in the form
t s
It follows from condition (H2 ) that (t − s)β−1 (s − ξ )δ−1 x(ξ ) dξ ds
  0 0
fn (t, x) ≤ g |x| + 1 + Anν t
(β)(δ)
for (t, x) ∈ [0, 1] × R, (1.4) = (t − s)β+δ−1 x(s) ds. (2.1)
(β + δ) 0
  A
fn (t, x) ≤ g |x| + 1 + ν By [35, 36], a sequence {ϕn } ⊂ L1 [0, 1] is called
|x|
uniformly integrable on [0, 1] if for any ε > 0 there ex-
for (t, x) ∈ [0, 1] × R0 . (1.5)
ists δ > 0 such that if M ⊂ [0, 1] and meas (M) < δ,
We begin by investigating the auxiliary regular then
fractional differential equation
  ϕn (t) dt < ε for n ∈ N.
D u(t) = fn t, u(t)
c α
(1.6) M

and define a solution of the regular problem (1.6), An immediate consequence of the definition of uni-
form integrability is the following result. If ϕn , ϕ ∈
(1.2) as a function u ∈ C 1 [0, 1] such that cD α u ∈
L1 [0, 1] and |ϕn (t)| ≤ ϕ(t) for a.e. t ∈ [0, 1] and all
C[0, 1], u satisfies the boundary conditions (1.2), and
n ∈ N, then {ϕn } is uniformly integrable on [0, 1].
equality (1.6) holds for t ∈ [0, 1].
The following lemma gives the condition for uni-
The paper is organized as follows. Section 2 con-
form integrability (see, e.g., [34, 37])
tains results from fractional calculus and from the uni-
form integrability of sequences {ϕn } ⊂ L1 [0, 1] which
Lemma 2.2 Let {ϕn } ⊂ L1 [0, 1]. Suppose that for ev-
we need in the next sections. Section 3 deals with the
ery ε > 0 there exists δ > 0 such that for any at most
regular fractional boundary value problem (1.6), (1.2).
countable set {(ai , bi )}i∈J of mutually disjoint inter-
We prove its solvability and give properties of its so-
vals (ai , bi ) ⊂ (0, 1), i∈J (bi − ai ) < δ, the inequal-
lutions un . Here, in addition, two technical lemmas
ity
are stated that are used in Theorem 4.1 for proving
the smoothness of increasing solutions to the singular  bi
ϕn (t) dt < ε for n ∈ N
problem (1.1), (1.2), and for applying the Vitali con- ai
i∈J
vergence theorem in limit processes. Note that there
is no Lebesgue majorant function for the sequence holds. Then {ϕn } is uniformly integrable on [0, 1].
{fn (t, un (t))} and so we cannot apply the Lebesgue
The following result is the Vitali convergent theo-
dominated convergence theorem. The existence result
rem [35, 36].
for problem (1.1), (1.2) is given in Sect. 4.
Lemma 2.3 Let {ϕn } ⊂ L1 [0, 1] and let
limn→∞ ϕn (t) = ϕ(t) for a.e. t ∈ [0, 1]. Then the fol-
2 Preliminaries
lowing statements are equivalent:
This section contains results from fractional calcu- (i) ϕ ∈ L1 [0, 1] and limn→∞ ϕn − ϕ L = 0,
lus and from the uniformly integrability of sequences (ii) the sequence {ϕn } is uniformly integrable on
{ϕn } ⊂ L1 [0, 1] that we need in the next sections. [0, 1].
644 D. O’Regan, S. Staněk

3 Auxiliary regular problem

t1  
− (t1 − s)α−1 fn s, x(s) ds
0
This section deals with the auxiliary regular problem 1 t1  
≤ (t2 − s)α−1 − (t1 − s)α−1
(1.6), (1.2). We prove its solvability and give the prop- (α) 0
erties of its solutions. We also state two technical Lem-  
× fn s, x(s) ds
mas 3.5 and 3.6.
1 t2  
In order to prove the solvability of problem (1.6), + (t2 − s)α−1 fn s, x(s) ds
(1.2), we define an operator Hn : C[0, 1] → C[0, 1] (α) t1
by the formula K α 
≤ t − t1α , x ∈ Ω,
(α + 1) 2
1 t  
(Hn x)(t) = (t − s)α−1 fn s, x(s) ds we see that the set Hn (Ω) is bounded in C[0, 1] and
(α) 0
equicontinuous on [0, 1]. Hence, Hn (Ω) is relatively
1 1  
− (1 − s)α−1 fn s, x(s) ds. compact in C[0, 1] by the Arzelà–Ascoli theorem. As
2(α) 0 a result Hn is completely continuous.
(3.1)
(b) (⇐) Let us first assume that x is a fixed point
of Hn . Then x ∈ C[0, 1] and
Lemma 3.1 Let (H1 ) hold. Then
1 t  
(a) Hn is a completely continuous operator, x(t) = (t − s)α−1 fn s, x(s) ds
(b) x is a solution of problem (1.6), (1.2), if and only (α) 0
if it is a fixed point of Hn . 1 1  
− (1 − s)α−1 fn s, x(s) ds
2(α) 0
Proof (a) Let {xm } ⊂ C[0, 1] be convergent and let for t ∈ [0, 1]. Since fn (t, x(t)) is continuous on [0, 1],
limm→∞ xm = x, x ∈ C[0, 1]. Let we have
 
ρm (t) = fn t, xm (t) , 1 t  
  x  (t) = (t − s)α−2 fn s, x(s) ds,
ρ(t) = fn t, x(t) for t ∈ [0, 1], m ∈ N. (α − 1) 0
t ∈ [0, 1],
Then ρm , ρ ∈ C[0, 1] and limm→∞ ρm − ρ = 0.
From the relation and so x  ∈ C[0, 1] by Lemma 2.1. Hence, x ∈
C 1 [0, 1], x  (0) = 0, x  > 0 on (0, 1] and x(0) +
(Hn xm )(t) − (Hn x)(t) x(1) = 0. By the definition of the Caputo fractional
1 t derivative, we have (cf. (2.1))
≤ (t − s)α−1 ρm (s) − ρ(s) ds
(α) 0 1 d2
1 1 D α x(t) =
c
+ (1 − s)α−1 ρm (s) − ρ(s) ds (2 − α) dt 2
2(α) 0 t  
ρm − ρ 1 × (t − s)1−α x(s) − x(0) − x  (0)s ds
≤ tα + 0
(α + 1) 2 1 d2 t
3 ρm −ρ
= (t − s)1−α
we conclude that Hn xm − Hn x ≤ for (2 − α) dt 2 0
2(α+1)
m ∈ N. Hence, limm→∞ Hn xm − Hn x = 0. Con- 1 s  
× (s − ξ )α−1 fn ξ, x(ξ ) dξ ds
sequently, Hn is a continuous operator. (α) 0
Let Ω ⊂ C[0, 1] be bounded and let K = d 2 t  
sup{ fn (t, x(t)) : x ∈ Ω}. Then K < ∞. Since = 2 (t − s)fn s, x(s) ds
dt 0
|Hn x(t)| ≤ 2(α+1)
3K
for t ∈ [0, 1], x ∈ Ω, and, for  
= fn t, x(t) , t ∈ [0, 1].
0 ≤ t1 < t2 ≤ 1,
Consequently, x is a solution of problem (1.6), (1.2).
(Hn x)(t2 ) − (Hn x)(t1 )
(⇒) Let now x be a solution of problem (1.6),
1 t2   (1.2). Then x ∈ C 1 [0, 1], x(0) + x(1) = 0, x  (0) = 0,
= (t2 − s)α−1 fn s, x(s) ds
(α) 0 and the equality
Fractional boundary value problems with singularities in space variables 645

1 d2 t   λ t  
(t − s)1−α x(s) − x(0) ds x(t) = (t − s)α−1 fn s, x(s) ds
(2 − α) dt 2 0 (α) 0
 
= fn t, x(t) λ 1  
− (1 − s)α−1 fn s, x(s) ds,
2(α) 0
is fulfilled for t ∈ [0, 1]. Integrating the last equal-
and, therefore,
ity twice we get (note that I γ denotes the Riemann–
λ t  
Liouville fractional integral of order γ > 0) x  (t) = (t − s)α−2 fn s, x(s) ds,
    (α − 1) 0
I 2−α x(t) − x(0) = I 2 fn t, x(t) + c1 + c2 t, t ∈ [0, 1].
t ∈ [0, 1], (3.2) Hence, by (1.3) and (1.4), x  > 0 on (0, 1] and
1 t
where c1 , c2 ∈ R. Since I 2−α (x(t) − x(0))|t=0 = 0 x  (t) ≤ (t − s)α−2
and I 2 fn (t, x(t))|t=0 = 0, we have c1 = 0. Applying (α − 1) 0
   
I α−1 to both sides of (3.2) we obtain, by Lemma 2.1, × g x(s) + 1 + Anν ds, t ∈ [0, 1].
    Since x(0) = −x(1), the equality x(ξ ) = 0 holds for
I 1 x(t) − x(0) = I α+1 fn t, x(t) + c2 I α−1 t,
some ξ ∈ (0, 1). Therefore, x ≤ x  and
t ∈ [0, 1]. g( x  + 1) + Anν t
x  (t) ≤ (t − s)α−2 ds
Differentiating the last equality yields (α − 1) 0
g( x  + 1) + Anν
1 t   ≤ , t ∈ [0, 1].
x(t) − x(0) = (t − s)α−1 fn s, x(s) ds (α)
(α) 0
In particular, the inequality
c2 α−1
+ t , t ∈ [0, 1], g( x  + 1) + Anν
(α) 1≤ (3.3)
(α) x 
t α−1
since d α−1
dt I t = (see, e.g., [2, Property 2.5]). The ν

(α)
t
is fulfilled. By (H2 ), limv→∞ = g(v+1)+An
(α)v = 0, and
functions x(t) and − s)α−1 fn (s, x(s)) ds belong
0 (t so there exists S > 0 such that
to the class C [0, 1] and, therefore, c2 = 0. Now it fol-
1
g(v + 1) + Anν
lows from the condition x(0) = −x(1) that < 1 for all v ≥ S.
(α)v
1 1   The last inequality together with (3.3) give x  < S
x(0) = − (1 − s)α−1 fn s, x(s) ds. and, therefore, x < S. We have proved that the
2(α) 0
set {x ∈ C[0, 1] : x = λHn x for some λ ∈ (0, 1)} is
Consequently, x is a fixed point of Hn .  bounded. Since, by Lemma 3.1, Hn is completely
continuous, Lemma 3.2 (with L = Hn , X = C[0, 1])
We are now in the position to prove that problem shows that Hn admits a fixed point un . Lemma 3.1
(1.6), (1.2) has a solution. We use the following non- now guarantees that un is a solution of problem
linear alternative [38]. (1.6), (1.2). 

Lemma 3.2 Let X be a Banach space and let L : X → The properties of solutions to problem (1.6), (1.2)
X be completely continuous. Then the following alter- are collected in the following lemma.
native holds: Either the equation x = λLx has a solu-
tion for every λ ∈ [0, 1] or the set A = {x ∈ X : x = Lemma 3.4 Let (H1 ) and (H2 ) hold. Let un be a so-
λLx for some λ ∈ (0, 1)} is unbounded. lution of problem (1.6), (1.2). Then {un } is bounded
in C 1 [0, 1], un (ξn ) = 0 for some ξn ∈ [Δ, 1) ⊂ (0, 1),
where Δ ∈ (0, 1) is independent of n, and
Theorem 3.3 Let (H1 ) and (H2 ) hold. Then for each
n ∈ N problem (1.6), (1.2) has a solution. a(3 − α) 2
un (t) ≥ t − ξn2
2
for t ∈ [0, 1], n ∈ N, (3.4)
Proof Let us choose n ∈ N. Suppose that x = λHn x
for some x ∈ C[0, 1] and λ ∈ (0, 1). Then un (t) ≥ a(3 − α)t for t ∈ [0, 1], n ∈ N. (3.5)
646 D. O’Regan, S. Staněk

t t
Proof Since un = Hn un for n ∈ N by Lemma 3.1, we (t − s)(α−2)q ds = (t − s)−
3−α
2 ds
have 0 0
2 α−1 2
1 t   = t 2 ≤ ,
un (t) = (t − s)α−2 fn s, un (s) ds, α−1 α−1
(α − 1) 0
we have
t ∈ [0, 1], n ∈ N, (3.6)  
t 2
− s)(α−2)q ds ≤
q q
and, therefore, by (1.3), (t . (3.9)
0 α−1
a t It follows from the estimate
un (t) ≥ (t − s)α−2 s 2−α ds
(α − 1) 0 t ds
= a(3 − α)t, t ∈ [0, 1], n ∈ N. 0 |s − ξn |
2 2 pν
1 ds
Hence, (3.5) holds and the condition un (0)+un (1) = 0 ≤ ds
implies un (ξn ) = 0 for some ξn ∈ (0, 1). Conse- 0 |s 2 − ξn2 |pν
quently, un ≤ un and ξn ds 1 ds
= +
t t 0 |s 2 − ξn2 |pν ξn |s 2 − ξn2 |pν
un (t) = un (s) ds ≥ a(3 − α) s ds ξn ds 1 ds
ξn ξn ≤ +
0 (ξn − s)pν s νp ξn (s − ξn )νp (s + ξn )νp
a(3 − α) 2
= t − ξn2 and from the relations (note that B(x, y) is the beta
2
function and 1 − 2νp > 0)
for t ∈ [0, 1] and n ∈ N, which proves (3.4). From
ξn ds
(1.5) and (3.6), it follows that the relation
0 (ξn − s)pν s νp
1 t
un (t) ≤
1
(t − s)α−2 1−2νp ds
(α − 1) 0 = ξn
0 (1 − s)νp s νp
  A 1−2νp
× g |un (s)| + 1 + ds = ξn B(1 − νp, 1 − νp),
|un (s)|ν 1 ds
g( un + 1) α−1
≤ t ξn (s − ξn )νp (s + ξn )νp
(α) 1−ξn dv
A t (t − s)α−2 =
+ ds (3.7) 0 v νp (v+ 2ξn )νp
(α − 1) 0 |un (s)|ν 1−ξn dv (1 − ξn )1−2νp
is satisfied for t ∈ [0, 1] and n ∈ N. We now give an ≤ =
t α−2 0 v 2νp 1 − 2νp
upper bound for the function 0 (t−s)|un (s)|ν ds. Let q = that
3−α
2(2−α) . By (3.4) and the Hölder inequality, t ds 1−2νp
≤ ξn B(1 − νp, 1 − νp)
0 |s 2 − ξn2 |pν
t (t − s)α−2
ds (1 − ξn )1−2νp
0 |un (s)|ν + ≤ K,
ν 1 − 2νp
2 t (t − s)α−2
≤ ds where K = B(1 − νp, 1 − νp) + 1
a(3 − α) 0 |s 2 − ξn2 |ν 1−2νp . Hence,
 
2 ν t t ds √
p
≤
q
(t − s)(α−2)q ds
p
≤ K.
0 |s − ξn |
2 2 pν
a(3 − α) 0
 The last inequality together with (3.8) and (3.9) yield
t 1
× p
ds, (3.8) 
|s − ξn2 |pν
2 t (t − s)α−2 2 ν
2 √ p
ds ≤
0 q
K
0 |un (s)|ν a(3 − α) α−1
where p = 3−α
α−1 (that is, 1
p + 1
q = 1). Since =: Q1 , t ∈ [0, 1], n ∈ N. (3.10)
Fractional boundary value problems with singularities in space variables 647

Consequently (cf. (3.7)),     A

0 ≤ f t, u(t) ≤ g u +
|u(t)|ν
g( un + 1)
0 ≤ un (t) ≤ + Q, t ∈ [0, 1], n ∈ N, for t ∈ [0, 1] \ {ξ }.
(α)
Hence,
AQ1
where Q = (α−1) . Hence,   t t (t − s)α−2
p(t) ≤ g u (t − s)α−2 ds + A ds
g( un + 1) |u(s)|ν
un ≤
0 0
+ Q, n ∈ N. (3.11)
(α) g( u ) t (t − s)α−2
≤ +A ds, t ∈ [0, 1].
α−1 |u(s)|ν
Since limv→∞ g(v)v = 0 by (H2 ), there exists S > 0 0
(3.13)
such that g(v+1)
(α) + Q < v for all v ≥ S. The last in- Since (cf. (3.10) with un replaced by u)
equality and (3.11) give un < S for n ∈ N and, t(t − s)α−2
therefore, un < S, n ∈ N. We have shown that {un } ds ≤ L for t ∈ [0, 1],
is bounded in C 1 [0, 1]. 0 |u(s)|ν
It remains to verify that ξn ∈ [Δ, 1) for n ∈ N, where L is a positive constant, (3.13) shows that p :
where Δ > 0. Suppose that there exists a subse- [0, 1] → [0, ∞).
quence {ξkn } of {ξn } such that limn→∞ ξkn = 0. Since The continuity of p on [0, ξ ) follows from Lem-
|ukn (0)| = |ukn (0) − ukn (ξkn )| = |ukn (νn )|ξkn ≤ Sξkn , ma 2.1 and from the fact that f (t, u(t)) is continuous
where νn ∈ (0, ξkn ), we see that limn→∞ ukn (0) = 0. on [0, ξ ).
Therefore, limn→∞ ukn (1) = 0 because ukn (0) = We now prove that p is continuous on (ξ, 1]. To this
end choose t∗ ∈ (ξ, 1]. In order to show that p is con-
−ukn (1), which contradicts (cf. (3.5))
tinuous at t∗ , let us choose ρ ∈ (ξ, t∗ ) and throughout
1 1 the proof let t1 , t2 be such that
ukn (1) − ukn (0) = ukn (s) ds ≥ a(3 − α) s ds t∗ + ρ
0 0 ≤ t1 ≤ t∗ ≤ t2 ≤ 1, t2 − t1 > 0.
a(3 − α) 2
= , n ∈ N. We need to prove that for every ε > 0 there exists
2
δ > 0 such that
Hence, inf{ξn : n ∈ N} > 0. As a result ξn ∈ [Δ, 1) for
n ∈ N with some Δ > 0.  t2 − t1 ≤ δ ⇒ p(t2 ) − p(t1 ) ≤ ε. (3.14)
By (H2 ) and (3.12), the relation
In order to prove that solutions u of the singular
p(t2 ) − p(t1 )
problem (1.1), (1.2) belong to the set C 1 ([0, 1] \ {ξ })
for some ξ = ξ(u) ∈ (0, 1), we need the following
t2  
= (t2 − s)α−2 f s, u(s) ds
technical result in the proof of Theorem 4.1. 0
t1  
− (t1 − s)α−2 f s, u(s) ds
Lemma 3.5 Let (H1 ) and (H2 ) hold. Let u ∈ C[0, 1], 0
u(ξ ) = 0 for some ξ ∈ (0, 1), and let t1    
≤ (t1 − s)α−2 − (t2 − s)α−2 f s, u(s) ds
u(t) ≥ W t 2 − ξ 2 for t ∈ [0, 1], (3.12) 0
t2  
where W > 0. Then the function + (t2 − s)α−2 f s, u(s) ds
t1
t   t1    
(t − s)α−2 f s, u(s) ds ≤ (t1 − s)α−2 − (t2 − s)α−2 g u(s) ds
0 0
is continuous on the set [0, 1] \ {ξ }. t2  
+ (t2 − s)α−2 g u(s) ds
t t1
Proof Let p(t) = 0 (t − s)α−2 f (s, u(s)) ds. We start
A t1 (t1 − s)α−2 − (t2 − s)α−2
by proving that p is well defined on [0, 1]. Since + ν ds
|u(t)| > 0 for t ∈ [0, 1] \ {ξ }, we conclude from (H1 ) W 0 |s 2 − ξ 2 |ν
and (H2 ) that the function f (t, u(t)) is continuous on A t2 (t2 − s)α−2
+ ν ds (3.15)
[0, 1] \ {ξ } and W t1 |s 2 − ξ 2 |ν
648 D. O’Regan, S. Staněk

holds. It is easy to verify that The function q(s) = (t1 − s)α−2 − (t2 − s)α−2 is
t1     increasing on [0, ρ] since q  (s) = −(α − 2)((t1 −
(t1 − s)α−2 − (t2 − s)α−2 g u(s) ds s)α−3 − (t2 − s)α−3 ) > 0 on this interval. Therefore,
0
t2   (t1 − s)α−2 − (t2 − s)α−2
+ (t2 − s)α−2 g u(s) ds
t1 ≤ (t1 − ξ )α−2 − (t2 − ξ )α−2 for s ∈ [0, ξ ],
  t1  
≤g u (t1 − s)α−2 − (t2 − s)α−2 ds and
0
t2 (t1 − s)α−2 − (t2 − s)α−2
+ (t2 − s) α−2
ds ≤ (t1 − ρ)α−2 − (t2 − ρ)α−2 for s ∈ [ξ, ρ].
t1
(3.19)
g( u )  α−1 
≤ t1 + 2(t2 − t1 )α−1 − t2α−1 . Hence,
α−1
The function r(t) = t α−1 is uniformly continuous on (t1 − s)α−2 − (t2 − s)α−2
ξ
ds
[0, 1] and, therefore, for every ε > 0 there exists δ1 > 0 0 |s 2 − ξ 2 |ν
such that   ξ ds
≤ (t1 − ξ )α−2 − (t2 − ξ )α−2
t 2 − t 1 ≤ δ1 ⇒ 0 (ξ − s) ν (ξ + s)ν

g( u )  α−1  ε (t1 − ξ )α−2 − (t2 − ξ )α−2 ξ ds

t1 + 2(t2 − t1 )α−1 − t2α−1 ≤ . ≤
α−1 3 ξν 0 (ξ − s)
ν
In particular, (t1 − ξ )α−2 − (t2 − ξ )α−2 1−2ν
t1  ⎫ = ξ .
   ⎪ 1−ν
(t1 − s) α−2
− (t2 − s)α−2
g u(s) ds ⎪
⎪
⎪
0 ⎪
⎬ Since r1 (t) = (t − ξ )α−2 is uniformly continuous on
t2   ε (3.16) [ρ, 1], for every ε > 0 there exists δ3 > 0 such that
+ (t2 − s)α−2 g u(s) ds ≤ ⎪
⎪
3 ⎪
⎪
t1 ⎪
⎭ t 2 − t 1 ≤ δ3 ⇒
for t2 − t1 ≤ δ1 .
W ν (1 − ν) ε
Since (t 2 − ξ 2 )ν ≥ (t12 − ξ 2 )ν > (ρ 2 − ξ 2 )ν for t ∈ (t1 − ξ ) α−2
− (t2 − ξ )α−2 ≤ .
Aξ 1−2ν 9
(t1 , t2 ), we have
In particular,
t2 (t2 − s)α−2 1 t2
ds < 2 (t2 − s)
α−2
ds (t1 − s)α−2 − (t2 − s)α−2
ξ Wν ε
t1 (s 2 − ξ 2 )ν (ρ − ξ 2 )ν t1 ds ≤
0 |s − ξ |
2 2 ν A 9
(t2 − t1 )α−1 for t2 − t1 ≤ δ3 . (3.20)
=.
(α − 1)(ρ 2 − ξ 2 )ν
The uniform continuity of r2 (t) = (t − ρ)α−2 on
Hence, for every ε > 0 there exists δ2 > 0 such that
[ ρ+t ∗
2 , 1] implies that for every ε > 0 there exists
t 2 − t 1 ≤ δ2 ⇒ δ4 > 0 such that
A t2 (t2 − s)α−2 ε t 2 − t 1 ≤ δ4 ⇒
ds ≤ . (3.17)
Wν t1 (s 2 − ξ 2 )ν 3 W ν (2ξ )ν (1 − ν) ε
(t1 − ρ) α−2
− (t2 − ρ)α−2 ≤ .
Keeping in mind (3.15), we now investigate A(ρ − ξ )1−ν 9
(t1 − s)α−2 − (t2 − s)α−2
t1 The last relation together with (3.19) and with the in-
ds equality
0 |s 2 − ξ 2 |ν
ξ (t − s)α−2 − (t − s)α−2 ρ (t1 − s)α−2 − (t2 − s)α−2
1 2
= ds ds
0 |s 2 − ξ 2 |ν
ξ |s 2 − ξ 2 |ν
ρ (t − s)α−2 − (t − s)α−2
1 2 (t1 − ρ)α−2 − (t2 − ρ)α−2 ρ ds
+ ds ≤
ξ |s 2 − ξ 2 |ν
(2ξ )ν ξ (s − ξ )ν
t1 (t1 − s)α−2 − (t2 − s)α−2 (t1 − ρ)α−2 − (t2− ρ)α−2
+ ds. (3.18) = (ρ − ξ )1−ν
ρ |s 2 − ξ 2 |ν (2ξ )ν (1 − ν)
Fractional boundary value problems with singularities in space variables 649

give Lemma 3.6 Let (H1 ) and (H2 ) hold. Let un be a so-
lution of problem (1.6), (1.2). Then the sequence
ρ (t1 − s)α−2 − (t2 − s)α−2 Wν ε   
ds ≤
ξ |s − ξ |
2 2 ν A 9 fn t, un (t) ⊂ C[0, 1] (3.24)
for t2 − t1 ≤ δ4 . (3.21) is uniformly integrable on [0, 1].
Next, it follows from the uniform continuity of r2 (t) =
(t −ρ)α−1 on [ ρ+t ∗ Proof Lemma 3.4 guarantees that un (ξn ) = 0 for n ∈
2 , 1] that for every ε > 0 there exists
δ5 > 0 such that N, where ξn ∈ [Δ, 1) ⊂ (0, 1), L := sup{ un : n ∈
N} < ∞ and that inequality (3.4) holds. Let M ⊂
t 2 − t 1 ≤ δ5 ⇒ [0, 1] be measurable. Then, by (1.5),
(t1 − ρ) α−1
+ (t2 − t1 )α−1 − (t2 − ρ)α−1  
fn s, un (s) ds
W ν (ρ 2 − ξ 2 )ν (α − 1) ε M
≤ .
A 9   A
≤ g un (s) + 1 + ds
Combining the last relation with the relation M |un (s)|
ν

ds
t1 (t1 − s)α−2 − (t2 − s)α−2 ≤ g(L + 1) ds + A
ds M |u
M n (s)|
ν
ρ |s 2 − ξ 2 |ν
t1 
ds
1  = g(L + 1)meas (M) + A .
≤ (t1 − s)α−2 − (t2 − s)α−2 ds |u
M n (s)|
ν
(ρ 2 − ξ 2 )ν ρ
Hence, if the sequence
(t1 − ρ)α−1 + (t2 − t1 )α−1 − (t2 − ρ)α−1  
= , 1
(ρ 2 − ξ 2 )ν (α − 1) ⊂ L1 [0, 1] (3.25)
|un (t)|ν
we obtain
is uniformly integrable on [0, 1], then the sequence in
t1 (t1 − s)α−2 − (t2 − s)α−2 Wν ε
ds ≤ (3.24) is uniformly integrable on [0, 1]. We now show
ρ |s 2 − ξ 2 |ν A 9 that the sequence in (3.25) is uniformly integrable on
for t2 − t1 ≤ δ5 . (3.22) [0, 1]. Let {(ai , bi )}i∈J be at most countable set of mu-
tually disjoint intervals (ai , bi ) ⊂ [0, 1].
Now, we conclude from (3.18), (3.20), (3.21), and
Since |t 2 − ξn2 | = |t − ξn |(t + ξn ) ≥ |t − ξn |Δ, we
(3.22) that the inequality
have (cf. (3.4))
(t1 − s)α−2 − (t2 − s)α−2
t1 Wν ε  
ds ≤ 1 K
|s 2 − ξ 2 |ν A 3 ≤ = KΔ−ν ω |t − ξn |
0
|un (t)|ν |t − ξn | Δ
ν ν
for t2 − t1 ≤ δ6 = min{δ3 , δ4 , δ5 } (3.23)
for t ∈ [0, 1] \ {ξn }, n ∈ N,
holds. Finally, from (3.17) and (3.23), we obtain
where K = ( a(3−α)
2
)ν and ω(x) = 1/x ν for x ∈
t1 (t1 − s)α−2 − (t2 − s)α−2 (0, ∞). Therefore,
ds
0 |s 2 − ξ 2 |ν bi dt bi  
≤ KΔ−ν ω |t − ξn | dt.
t2 (t2 − s)α−2 W ν 2ε |un (t)|ν
+ ds ≤ ai ai
(s − ξ )
2 2 ν A 3
t1
If ξn ≤ ai , then
for t2 − t1 ≤ min{δ2 , δ6 }. The last inequality and
(3.15) and (3.16) imply that (3.14) holds with δ =
bi   bi
ω |t − ξn | dt = ω(t − ξn ) dt
min{δ1 , δ2 , δ6 }.  ai ai
bi −ξn
In order to apply the Vitali convergence theorem = ω(s) ds.
ai −ξn
in the proof of Theorem 4.1, we need the following
result. If ai0 < ξn < bi0 for some i0 ∈ J, then
650 D. O’Regan, S. Staněk

bi0   4 The main result

ω |t − ξn | dt
ai0
ξn bi0 We now give the existence result for the singular frac-
= ω(ξn − t) dt + ω(t − ξn ) dt tional problem (1.1), (1.2).
ai0 ξn
ξn −ai0 bi0 −ξn Theorem 4.1 Let conditions (H1 ) and (H2 ) hold.
= ω(s) ds + ω(s) ds. Then problem (1.1), (1.2) has at least one increasing
0 0
solution.
Finally, if bi ≤ ξn , then
Proof By Theorem 3.3 and Lemma 3.4, problem
bi   bi
ω |t − ξn | dt = ω(ξn − t) dt (1.6), (1.2) has a solution un such that un (ξn ) = 0
ai ai for some ξn ∈ [Δ, 1) ⊂ (0, 1), un satisfies inequali-
ξn −ai ties (3.4), (3.5), and the sequence {un } is bounded in
= ω(s) ds. C 1 [0, 1]. Hence, by the Arzelà–Ascoli theorem and
ξn −bi
the Bolzano–Weierstrass theorem, there exist subse-
To summarize, quences {ukn } and {ξkn } and u ∈ C[0, 1], ξ ∈ [Δ, 1]
 bi dt  bi   such that limn→∞ ukn − u = 0, limn→∞ ξkn = ξ .
≤ KΔ−ν ω |t − ξn | dt Then u(ξ ) = 0, u(0) + u(1) = 0, and letting n → ∞
ai |un (t)|ν
ai
i∈J i∈J in the inequality
= KΔ−ν ω(s) ds, (3.26) a(3 − α) 2
ukn (t) ≥ t − ξk2n , t ∈ [0, 1], n ∈ N,
M 2
where M ⊂ [0, 1], meas (M) ≤ i∈J (bi − ai ). Since we obtain
ω ∈ L1 [0, 1], for every ε > 0 there exists δ > 0 such a(3 − α) 2
u(t) ≥ t − ξ 2 , t ∈ [0, 1].
that 2
εΔν Hence, f (t, u(t)) ∈ C([0, 1] \ {ξ }) and the last in-
ω(t) dt < equality together with u(0) + u(1) = 0 give u(1) = 0,
M K
that is, ξ ∈ [Δ, 1).
whenever M ⊂ [0, 1] is measurable and meas (M) < Since
δ. Hence, for every ε > 0, there exists δ > 0 such that    
for any at most countable set {(ai , bi )}i∈J of mutually lim fkn t, ukn (t) = f t, u(t)
n→∞
disjoint intervals (ai , bi ) ⊂ [0, 1], i∈J (bi − ai ) < δ, for t ∈ [0, 1] \ {ξ }, (4.1)
we have (cf. (3.26))
and
 bi dt ν
−ν εΔ t  
≤ KΔ = ε. 1
ai |un (t)|ν K ukn (t) = (t − s)α−1 fkn s, ukn (s) ds
i∈J (α) 0
Consequently, the sequence in (3.25) is uniformly in- 1 1  
− (1 − s)α−1 fkn s, ukn (s) ds,
tegrable on [0, 1] by Lemma 2.2.  2(α) 0
it follows from Lemma 2.3 (for ϕn (t) = fkn (t, ukn (t))
Corollary 3.7 Let the assumptions of Lemma 3.6 be and for ϕn (t) = (t0 − t)α−1 fkn (t, ukn (t))) and from
satisfied. Then for each t0 ∈ (0, 1] the sequence Lemma 3.6 and Corollary 3.7 that f (t, u(t)) ∈ L1 [0, 1]
   and that the equality
(t0 − t)α−1 fn t, un (t) ⊂ C[0, t0 ]
1 t  
is uniformly integrable on [0, t0 ]. u(t) = (t − s)α−1 f s, u(s) ds
(α) 0
1 1  
Proof Choose t0 ∈ (0, 1]. Then Lemma 3.6 guarantees − (1 − s)α−1 f s, u(s) ds (4.2)
that {fn (t, un (t))} is uniformly integrable on [0, t0 ]. 2(α) 0

Since 0 ≤ (t0 − t)α−1 fn (t, un (t)) ≤ fn (t, un (t)) for holds for t ∈ [0, 1]. Since f (t, u(t)) ∈ L1 [0, 1], we ob-
t ∈ [0, t0 ] and n ∈ N, the result follows.  tain from Lemma 2.1 and (4.2) that
Fractional boundary value problems with singularities in space variables 651

t s   q(t, u)
(s − τ )α−2 f τ, u(τ ) dτ ds D α u = p(t, u) + r(t)φ(u) +
c
,
0 0 |u|ν
(α − 1) t   u(0) + u(1) = 0, u (0) = 0,
= (t − s)α−1 f s, u(s) ds
(α) 0 has at least one increasing solution.
= (α − 1)
Acknowledgements The research of the second author was
1 1  
× u(t) + (1 − s)α−1 f s, u(s) ds , supported by grants PrF-2011-022 and PrF-2012-017.
2(α) 0
t ∈ [0, 1].
References
In particular,
t s   1. Podlubny, I.: Fractional differential equations. In: Math-
1
u(t) = (s − τ )α−2 f τ, u(τ ) dτ ds ematics in Science and Engineering, vol. 198. Academic
(α − 1) 0 0 Press, San Diego (1999)
2. Kilbas, A.A., Srivastava, H.M., Trujillo, J.J.: Theory and
1 1  
− (1 − s)α−1 f s, u(s) ds Applications of Fractional Differential Equations. Elsevier,
2(α) 0 Amsterdam (2006)
for t ∈ [0, 1] (4.3) 3. Diethelm, K.: The Analysis of Fractional Differential Equa-
t tions. Lectures Notes in Mathematics. Springer, Berlin
Since, by Lemma 3.5, 0 (t − s)α−2 f (s, u(s)) ds ∈ (2010)
4. Baleanu, D., Diethelm, K., Scalas, E., Trujillo, J.J.: Frac-
C([0, 1] \ {ξ }), equality (4.3) yields tional Calculus: Models and Numerical Methods. Series on
1 t   Complexity, Nonlinearity and Chaos, vol. 3. World Scien-
u (t) = (t − s)α−2 f s, u(s) ds tific, Singapore (2012)
(α − 1) 0 5. Abbas, S., Banerjee, M., Momani, S.: Dynamical analysis
for t ∈ [0, 1] \ {ξ }. of fractional-order modified logistic model. Comput. Math.
Appl. 62, 1098–1104 (2011)
Therefore, u (0) = 0, u ∈ C([0, 1] \ {ξ }), and u > 0 6. Daftardar-Gejji, V., Bhalekar, S.: Boundary value problems
on [0, 1] \ {ξ }. Arguing as in the proof of Lemma 3.1, for multi-term fractional differential equations. J. Math.
Anal. Appl. 345, 754–765 (2008)
we have 7. Dalir, M., Bashour, M.: Applications of fractional calculus.
Appl. Math. Sci. 4(21), 1021–1032 (2010)
1 d2
D α u(t) =
c 8. Kilbas, A.A., Trujillo, J.J.: Differential equations of frac-
(2 − α) dt 2 tional order: methods, results and problems I. Appl. Anal.
t   78(1–2), 153–192 (2001)
× (t − s)1−α u(s) − u(0) − u (0)s ds 9. Kilbas, A.A., Trujillo, J.J.: Differential equations of frac-
0 tional order: methods, results and problems II. Appl. Anal.
d2 t   78(2), 435–493 (2002)
= 2 (t − s)f s, u(s) ds 10. Agarwal, RP., Ahmad, B.: Existence theory for anti-
dt 0 periodic boundary value problems of fractional differential
 
= f t, u(t) , t ∈ [0, 1] \ {ξ }. equations and inclusions. Comput. Math. Appl. 62, 1200–
1214 (2011)
Consequently, u is an increasing solution of problem 11. Bai, Z.: On positive solutions of a nonlocal fractional
boundary value problem. Nonlinear Anal. 72, 916–924
(1.1), (1.2). 
(2010)
12. Caballero, J., Harjani, J., Sadarangani, K.: Existence and
Example 4.2 Let α ∈ (1, 2) and ν ∈ (0, 2(3−α)α−1
). Let uniqueness of positive solution for a boundary value
p, q ∈ C([0, 1] × R) be nonnegative and bounded, 0 < problem of fractional order. Abstr. Appl. Anal. (2011).
doi:10.1155/2011/165641
a < p(t, x) on [0, 1] × R. Suppose that r ∈ C[0, 1], 13. Benchohra, M., Hamani, S., Ntouyas, S.K.: Boundary value
φ ∈ C(R) are nonnegative, and lim|x|→∞ φ(x) x = 0. problems for differential equations with fractional order
The function f (t, x) = p(t, x) + r(t)φ(x) + q(t,x)
|x|ν
and nonlocal conditions. Nonlinear Anal. 71, 2391–2396
(2009)
satisfies conditions (H1 ) and (H2 ). To check it we take 14. Guo, Y.: Nontrivial solutions for boundary-value problems
A = sup{q(t, x) : (t, x) ∈ [0, 1] × R} and g(y) = B + of nonlinear fractional differential equations. Bull. Korean
C max{φ(s) : s ∈ [−y, y]} for y ∈ [0, ∞), where B = Math. Soc. 47(1), 81–87 (2010)
sup{p(t, x) : (t, x) ∈ [0, 1] × R}, C = max{r(t) : t ∈ 15. Jiang, D., Yuan, C.: The positive properties of the Green
function for Dirichlet-type boundary value problems of
[0, 1]}. nonlinear fractional differential equations and its applica-
Hence, Theorem 4.1 guarantees that the problem tion. Nonlinear Anal. 72, 710–719 (2010)
652 D. O’Regan, S. Staněk

16. Li, C.F., Luo, X.N., Zhou, Y.: Existence of positive solu- 27. Staněk, S.: The existence of positive solutions of singular
tions of the boundary value problem for nonlinear fractional fractional boundary value problems. Comput. Math. Appl.
differential equations. Comput. Math. Appl. 59, 1363–1375 62, 1379–1388 (2011)
(2010) 28. Agarwal, R.P., O’Regan, D., Staněk, S.: Positive so-
17. Lin, X., Sun, W., Jiang, D.: Existence and uniqueness lutions for mixed problems of singular fractional dif-
of solutions for boundary value problems to the singular ferential equations. Math. Nachr. 285(1), 27–41 (2012).
one-dimensional p-Laplacian. Bound. Value Probl. (2008). doi:10.1002/mana.201000043
doi:10.1155/2008/194234 29. Cui, Y.: Existence results for singular boundary value prob-
18. Qiu, T., Bai, Z.: Positive solutions for boundary value prob- lem of nonlinear fractional differential equation. Abstr.
Appl. Anal. (2011). doi:10.1155/2011/605614
lem of nonlinear fractional differential equation. J. Nonlin-
30. Xu, X., Jiang, D., Yuan, C.: Multiple positive solutions for
ear Sci. Appl. 1(3), 123–131 (2008)
the boundary value problem of a nonlinear fractional differ-
19. Rehman, M., Khan, R.A., Asif, N.A.: Three point boundary ential equation. Nonlinear Anal. 71, 4676–4688 (2009)
value problems for nonlinear fractional differential equa- 31. Yuan, C., Jiang, D., Xu, X.: Singular positone and semi-
tions. Acta Math. Sci. 31, 1337–1346 (2011) positone boundary value problems of nonlinear frac-
20. Wang, J., Xiang, H., Liu, Z.G.: Existence of concave posi- tional differential equations. Math. Probl. Eng. (2009).
tive solutions for boundary value problem of nonlinear frac- doi:10.1155/2009/535209
tional differential equation with p-Laplacian operator. Int. 32. Zhang, S.: Positive solutions to singular boundary value
J. Math. Math. Sci. (2010). doi:10.1155/2010/495138 problem for nonlinear fractional differential equation.
21. Wang, J., Xiang, H.: Upper and lower solutions method Comput. Math. Appl. 59, 1300–1309 (2010)
for a class of singular fractional boundary value prob- 33. Rachůnková, I., Staněk, S., Tvrdý, M.: Singularities and
lems with p-Laplacian operator. Abstr. Appl. Anal. (2010). Laplacians in boundary value problems for nonlinear or-
doi:10.1155/2010/971824 dinary differential equations. In: Cañada, A., Drábek, P.,
22. Wang, J., Xiang, H., Zhao, Y.: Monotone and concave posi- Fonda, A. (eds.) Handbook of Differential Equations. Ordi-
tive solutions to a boundary value problem for higher-order nary Differential Equations, vol. 3, pp. 607–723. Elsevier,
fractional differential equation. Abstract and Applied Anal- Amsterdam (2006)
ysis (2011). doi:10.1155/2011/430457 34. Rachůnková, I., Staněk, S., Tvrdý, M.: Solvability of Non-
23. Arara, A., Benchohra, M., Hamidi, N., Nieto, J.J.: Frac- linear Singular Problems for Ordinary Differential Equa-
tional order differential equations on an unbounded do- tions. Hindawi, New York (2009)
main. Nonlinear Anal. 72, 580–586 (2010) 35. Bartle, R.G.: A Modern Theory of Integration. AMS, Prov-
idence, Rhode Island (2001)
24. Baleanu, D., Mustafa, O.G., Agarwal, R.P.: An existence
36. Natanson, I.P.: Theorie der Funktionen Einer Reellen
result for a superlinear fractional differential equation.
Veränderlichen, Herausgegeben Von Karl Bögel, 4. Aufl.
Appl. Math. Lett. 23, 1129–1132 (2010) (German). Mathematische Lehrbücher und Monographien.
25. Zhao, X., Ge, W.: Unbounded solutions for a frac- Akademie, Berlin (1975). English transl.: Theory of Func-
tional boundary value problems on the infinite inter- tions of Real Variable. Ungar, New York (1955)
val. Acta Appl. Math. 109, 495–505 (2010). doi:10.1007/ 37. Agarwal, R.P., O’Regan, D., Staněk, S.: Positive solutions
s10440-008-9329-9 of singular boundary value problems for delay differential
26. Agarwal, R.P., O’Regan, D., Staněk, S.: Positive solutions equations. Dyn. Syst. Appl. 16, 755–770 (2007)
for Dirichlet problems of singular nonlinear fractional dif- 38. Deimling, K.: Nonlinear Functional Analysis. Springer,
ferential equations. J. Math. Anal. Appl. 371, 57–68 (2010) Berlin (1985)