Xu and Meng Journal of Inequalities and Applications (2016) 2016:78 DOI 10.1186/s13660-016-1015-2

RESEARCH

Open Access

Some new weakly singular integral inequalities and their applications to fractional differential equations Run Xu* and Fanwei Meng *

Correspondence: [email protected] Department of Mathematics, Qufu Normal University, Qufu, Shandong 273165, People’s Republic of China

Abstract Some new inequalities with weakly singular integral kernel are developed, which generalize some known inequalities and can be used in the qualitative and quantitative analysis of the solutions to certain fractional differential equations. MSC: 34C10 Keywords: inequality; weakly singular integral kernel; fractional differential and integral equation; quantitative analysis

1 Introduction The well-known inequalities, such as the Gronwall type inequality, the Gronwall-Bellman type inequality, the Henry-Gronwall type inequality, the Henry-Bihari type inequality and their variants in retarded form played important roles in the research of quantitative analysis of the solutions to differential and integral equations, as well as in the modeling of engineering and science problems. Recently, with the development of fractional differential equations, integral inequalities with weakly singular kernels have drawn more attention [–]. In , Henry [] proposed a method to find solutions and proved some results concerning linear integral inequalities with a weakly singular kernel. In , Medved˘ [] presented a new method to solve integral inequalities of Henry-Gronwall type and their Bihari version, then he got the explicit bounds with a quite simple formula, similar to the classic Gronwall-Bellman inequalities. Furthermore, he also obtained global solutions of the semilinear evolutions in []. In , Ye and Gao [] presented the integral inequalities of Henry-Gronwall type, ⎧ ⎨u(t) ≤ a(t) +  t (t – s)β– [b(s)u(s) + c(s)u(s – r)] ds, ⎩u(t) ≤ ϕ(t),

t

t ∈ [t , T),

t ∈ [t – r, t ).

In , Shao and Meng [] established Gronwall-Bellman type inequalities with a weakly singular integral ⎧ ⎨u(t) ≤ a(t) +  t (t – s)β– b(s)u(s) ds +  t (t – s)β– p(s)uγ (s – r) ds, ⎩u(t) ≤ ϕ(t),

t

t

t ∈ [t , T),

t ∈ [t – r, t ).

© 2016 Xu and Meng. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 2 of 16

In [], Feng and Meng studied the following Gronwall-Bellman type inequalities: 

x

up (x) ≤ a(x) +

b(t)uq (t) dt + 

 h(x) (α)



x

  (x – t)α– L t, u(t) dt.



In , by using the modified Medved˘ method, Ma and Pečarić [] studied the inequality  up (t) ≤ a(t) + b(t)

t

t α – sα

β–

t ∈ R+ .

sγ – f (s)uq (s) ds,



The aim of this paper was to give explicit bounds to some new nonlinear HenryGronwall type retarded integral inequalities with weakly singular integral kernel of the form ⎧ ⎨up (t) ≤ a(t) +  t (t – s)β– b(s)uq (s) ds +  t (t – s)β– c(s)ul (s – r) ds, ⎩u(t) ≤ ϕ(t),

t

t

t ∈ I,

t ∈ [t – r, t ),

and Gronwall-Bellman type integral inequalities with nonlinear weakly singular integral kernel of the form  u (t) ≤ a(t) + b(t)



t

p

(t – s)

β–

m

c(s)u (s) ds + d(t)



t

t α – sα

β–

sγ – f (s)uq (s) ds,



which can be used as handy and effective tools in the study of the delay fractional differential equations. We also give some examples to illustrate applications of our results.

2 Preliminary knowledge In the following, R denotes the set of real numbers, N denotes the set of integer numbers, R+ = [, +∞), I = [t , T) ⊂ R+ , T < ∞. For convenience, we give some lemmas which will be used in the proof of the main results. Lemma . (Jensen’s inequality) Let n ∈ N , a , a , . . . , an be nonnegative real numbers. Then, for r > ,

n

r ≤ nr–

ai

n

i=

ari .

i=

Lemma . (see []) Let a ≥ , p ≥ q ≥ , p = , then q q q–p p – q pq ap ≤ K p a + K p p

for any K > .

Lemma . (see []) Let α, β, γ , and p be positive constants, then 

t

α

t –s

 α p(β–) p(γ –) s



Here B(m, n) =

 



p(γ – ) +  tθ , p(β – ) +  , ds = B α α

t ∈ R+ .

sm– ( – s)n– ds (m > , n > ) is the B-function, θ = p[α(β – ) + γ – ] + .

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 3 of 16

Definition . (see []) Let [x, y, z] be an ordered parameter group of nonnegative real numbers. The group is said to belong to the first class distribution and denoted by [x, y, z] ∈ I if the conditions x ∈ (, ], y ∈ (  , ), and z ≥  – y are satisfied; the group is said to belong to the second class distribution and denoted by [x, y, z] ∈ II if the conditions x ∈ (, ], y ∈  are satisfied. (,  ], and z > –y –y Lemma . (see []) Suppose that the positive constants α, β, γ , p , and p satisfy the conditions (a) if [α, β, γ ] ∈ I, p = β ; , then (b) if [α, β, γ ] ∈ II, p = +β +β

pi (γ – ) +  B , pi (β – ) +  ∈ (, ∞) α and   θi = pi α(β – ) + γ –  +  ≥  are valid for i = , . Lemma . ([]) Let a(t), b(t) ∈ C(I, R+ ), p ≥ q > . If u(t) ∈ C(I, R+ ), and  up (t) ≤ a(t) +

t

b(s)uq (s) ds, t

then  p   t t A (s)e s B (τ ) dτ ds , u(t) ≤ a(t) +

t ∈ I.

(.)

t

Here A (t) = b(t)

p – q pq q q–p K p a(t) + K , p p

q q–p B (t) = K p b(t). p

Lemma . Let a(t), b(t), c(t) ∈ C(I, R+ ), p ≥ q > , p ≥ l > , ϕ(t) ∈ C([t – r, t ], R+ ), a(t ) = ϕ(t ). If u(t) ∈ C(I, R+ ), and ⎧ ⎨up (t) ≤ a(t) +  t b(s)uq (s) ds +  t c(s)ul (s – r) ds, t ∈ I, t t (.) ⎩u(t) ≤ ϕ(t), t ∈ [t – r, t ). Then, for t ∈ [t , t + r),  p   t  t t l B (τ ) dτ s c(s)ϕ (s – r) ds + A (s)e ds . u(t) ≤ a(t) + t

t

Here r ∈ R+ ,    t q q–p p – q pq A (t) = K p b(t) a(t) + K , c(s)ϕ l (s – r) ds + p p t q q–p B (t) = B (t) = K p b(t). p

(.)

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 4 of 16

For t ∈ [t + r, T),   p  t t B (τ ) dτ s u(t) ≤ a(t) + A (s)e ds ,

(.)

t

where q q–p p – q pq l l–p p – l pl A (t) = K p a(t)b(t) + K b(t) + K p a(t – r)c(t) + K c(t), p p p p q q–p l l–p B (t) = K p b(t) + K p c(t). p p Proof For t ∈ [t , t + r), we get 



t

up (t) ≤ a(t) +

t

c(s)ϕ l (s – r) ds + t

b(s)uq (s) ds, t

then from Lemma ., the inequality (.) holds. t t For t ∈ [t + r, T), let z(t) = t b(s)uq (s) ds + t c(s)ul (s – r) ds, then z(t ) = , z(t) is nondecreasing, and up (t) ≤ a(t) + z(t). Then from Lemma . we obtain, for any K > , z (t) = b(t)uq (t) + c(t)ul (t – r)

  p–q q q q–p p p ≤ b(t) K a(t) + z(t) + K p p

 p–l l l l–p  + c(t) K p a(t – r) + z(t – r) + Kp p p ≤ A (t) + B (t)z(t), t

and we have z(t) ≤

t

t

A (s)e

s B (τ ) dτ



ds. Then the inequality (.) holds.

Lemma . ([]) Let a(t), b(t, s), b (t, s) be continuous and nonnegative. If u(t) ∈ C(I, R+ ) and  u(t) ≤ a(t) +

t

b(t, s)u(s) ds, t

then t

u(t) ≤ a(t) + e

t b(t,s) ds



t

 – tτ b(τ ,s) ds

e t





τ

a(τ )b(τ , τ ) + t

bτ (τ , s)a(s) ds

dτ .

(.)

3 Main results We are now to deal with a certain class of Henry-Gronwall type retarded inequalities with weak integral kernels.

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 5 of 16

Theorem . Let a(t), b(t), c(t) ∈ C(I, R+ ), p ≥ q > , p ≥ l > , ϕ(t) ∈ C([t – r, t ], R+ ), a(t ) = ϕ(t ), u(t) ∈ C(I, R+ ). Suppose ⎧ ⎨up (t) ≤ a(t) +  t (t – s)β– b(s)uq (s) ds +  t (t – s)β– c(s)ul (s – r) ds, t

⎩u(t) ≤ ϕ(t),

t

t ∈ I,

(.)

t ∈ [t – r, t ),

where β > . (i) Suppose that β >  , then ⎧  ⎨u(t) ≤ {a (t) +  t c (s)ϕ l (s – r) ds +  t A (s)e st B (τ ) dτ ds} p ,  t  t   ⎩u(t) ≤ {a (t) +  t A (s)e st B (τ ) dτ ds} p , t ∈ [t + r, T).

t ∈ [t , t + r),

(.)

t

Here a (t) = a (t)e–t ,

b (t) =

(β – )  t( pq –) b (t)e , β

(β – )  t( pl –) c (s)e , β

 t q q–p p – q pq l p K , c (s)ϕ (s – r) ds + A (t) = K b (t) a (t) + p p t c (t) =

q q–p B (t) = K p b (t), p

q q–p p – q pq l l–p p – l pl A (t) = K p a (t) + K b (t) + K p a (t – r) + K c (t), p p p p q q–p l l–p B (t) = K p b (t) + K p c (t). p p (ii) If  < β ≤  , let q =

+β , β

p =  + β, then

⎧  ⎨u(t) ≤ {a (t) +  t c (s)ϕ l (s – r) ds +  t A (s)e st B (τ ) dτ ds} p ,  t  t   ⎩u(t) ≤ {a (t) +  t A (s)e st B (τ ) dτ ds} p , t ∈ [t + r, T).

t ∈ [t , t + r),

t

Here  q – q

a (t) = 

a (t)e

 q –

c (t) = 

–q t

,

q –

b (t) = 

( – ( – β)p )

 pq



–(–β)p p

( – ( – β)p )

cq (t)e(–q +

–(–β)p p q l p )t

 pq



bq (t)e(–q +

q q p )t

,

   t q q–p p – q pq K , c (s)ϕ l (s – r) ds + A (t) = K p b (t) a (t) + p p t q q–p B (t) = K p b (t), p

p – q pq p – l pl q q–p l l–p p p A (t) = K a (t) + K b (t) + K a (t – r) + K c (t), p p p p q q–p l l–p B (t) = K p b (t) + K p c (t). p p

,

(.)

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 6 of 16

Proof (i) Using the Cauchy-Schwarz inequality by (.), we get  up (t) ≤ a(t) +



t

t

(t – s)β– es e–s b(s)uq (s) ds +

(t – s)β– es e–s c(s)ul (s – r) ds

t

t

 ≤ a(t) +

  

t

(t – s)

β– s



–s q

b (s)e

u (s) ds

t

  

t

+



e ds

t

 

t

(t – s)

β– s



–s l

c (s)e

e ds

t

 

t

u (s – r) ds

t

 ≤ a(t) + 

et (β – ) β

   

t

·



–s q

b (s)e



u (s) ds

  . u (s – r) ds

t 

+

–s l

c (s)e

t

t

Using Lemma . for n = , r = , we obtain up (t) ≤ a (t) + 

et (β – ) β





t

t

b (s)e–s uq (s) ds + t

 c (s)e–s ul (s – r) ds .

t



Let v(t) = e– p t u (t), then we get vp (t) ≤ a (t)e–t + 

t

·

(β – ) β



q

t

 l c (s)es( p –) vl (s – r) ds .

t

b (s)es( p –) vq (s) ds +

t

Using Lemma ., we get (.). (ii) By the hypothesis, we get p +  u (t) ≤ a(t) +

 q

= , using the Hölder inequality by (.), we obtain  p 

t

p

(t – s)

p (β–) p s

e

ds

u

(t – s)p (β–) ep s ds

 p 

t



t

e–q s cq (s)uq l (s – r) ds

 q

= a(t) +

ep t ( – ( – β)p )

 p



–(–β)p p

 ·

t

bq (s)e–q s uq q (s) ds

 q





t

cq (s)e–q s uq l (s – r) ds

+

t

t

Using Lemma . for n = , r = q , we have  u

(t) ≤ 

q – q

q –

a (t) + 

 ·



t



pq



(s) ds

t

t

+

–q s q q

q

b (s)e

t



 q

t



ep t ( – ( – β)p ) –(–β)p p



t

t

bq (s)e–q s uq q (s) ds + t

t

 pq



cq (s)e–q s uq l (s – r) ds .

 q 

.

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 7 of 16

q

Let v(t) = e– p t uq (t), then we get  v (t) ≤ 

q – q

p

–q t

a (t)e



t

·

q –

+

bq (s)e(–q +

( – ( – β)p )

 pq



–(–β)p p



q q p )s q

t

v (s) ds +

t

cq (s)e(–q +

v (s – r) ds .

q l p )s l

t

By Lemma . we get (.). This completes the proof of Theorem ..



Remark . If p = q = , then Theorem . becomes Theorem  in []. If p = q = l = , then Theorem . becomes Theorem . in []. If p = l = , b(t) = , then Theorem . becomes Theorem . in []. Theorem . Suppose that u(t), a(t), b(t), c(t), d(t), f (t) are nonnegative continuous functions for t ∈ R+ . Let p, m, q be constants with p ≥ m > , p ≥ q > . If  up (t) ≤ a(t) + b(t)



t

(t – s)β– c(s)um (s) ds + d(t) 

t

t α – sα

β–

sγ – f (s)uq (s) ds, (.)



then for [α, β, γ ] ∈ I,   t  t τ ¯ ¯ e–  b (τ ,s) ds u(t) ≤ a(t) + a¯  (t) + e  b (t,s) ds 

  ¯ · a¯  (s)b (τ , τ ) + 

τ

 –β  p ¯b (τ , s)¯a (s) ds dτ , τ

where β  β –β    β –β a¯  (t) =  –β b(t)et M A (t) + d(t) N t θ B (t) –β ,   –β m m–p β –  s  K p b(t)et M e –β c –β (s) p   

 θ β –β  q q–p  p –β K d(t) N t + f (s) , p β

b¯  (t, s) =  –β

M =



( – ( – β) β ) ( β )

–(–β) β

=

( – β )  –

,

ββ

 γ – +     β –  β + γ –  β –   β ·B , + = ·B , , α α β α α β   θ = α(β – ) + γ –  + , β     t  s  m m–p p – m mp –β – –β p –β K K e c (s) a(s) + ds, A (t) = p p      t  p – q pq –β q q–p p –β K f (s) K a(s) + ds; B (t) = p p  N =

(.)

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 8 of 16

for [α, β, γ ] ∈ II,   t  t τ ¯ ¯ u(t) ≤ a(t) + a¯  (t) + e  b (t,s) ds e–  b (τ ,s) ds 

  · a¯  (s)b¯  (τ , τ ) +

τ



b¯ τ (τ , s)¯a (s) ds

β 

+β p



,

dτ

(.)

where a¯  (t) = 

+β β

+β β   +β β   +β b(t)et M+β A+β (t) + d(t) N t θ +β B+β (t) β ,

 +β +β β m m–p – +β s +β K p b(t)et M+β e β c β (s) p  +β

   +β β +β q q–p θ +β p K d(t) N t f β (s) , + p

b¯  (t, s) = 

M =

+β β



) ( – ( – β) +β +β ( +β ) +β

–(–β) +β +β



β ) ( +β

=

β 

,

( +β ) +β +β

 +β (γ – ) +    + β +β , (β – ) +  α  + β   γ – β + βγ β   , , = ·B α α( + β)  + β

 N = · B α

  + β  α(β – ) + γ –  + ,  + β   +β  t β m m–p p – m mp – +β s +β e β c β (s) ds, K p a(s) + K A (t) = p p   +β   t β q–p q +β q p – q B (t) = Kp f β (s) K p a(s) + ds. p p 

θ =

Proof Define 



t

(t – s)β– c(s)um (s) ds + d(t)

v(t) = b(t) 

t

t α – sα

β–

sγ – f (s)uq (s) ds,

(.)



then   up (t) ≤ a(t) + v(t) or u(t) ≤ a(t) + v(t) p . By Lemma ., from (.), for any K > , we have

  p–m m m m–p p p ds v(t) ≤ b(t) (t – s) c(s) a(s) + v(s) + K K p p 

 t  α α β– γ –   p–q q q q–p p p + d(t) t –s a(s) + v(s) + K ds s f (s) K p p   t m m–p = A(t) + K p b(t) (t – s)β– c(s)v(s) ds p  



t

β–

(.)

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 9 of 16

 t  α α β– γ – q q–p t –s + K p d(t) s f (s)v(s) ds p  m m–p q q–p = A(t) + K p b(t) · B(t) + K p d(t) · C(t). p p

(.)

Here

m m–p p – m mp K p a(s) + K ds p p 

 t q–p  α α β– p – q pq γ – q p t –s K a(s) + K ds, + d(t) f (s)s p p   t B(t) = (t – s)β– c(s)v(s) ds, 

t

(t – s)β– c(s)

A(t) = b(t)





t

C(t) =

t α – sα

β–

sγ – f (s)v(s) ds.



Now we try to estimate A(t), B(t), and C(t).  If [α, β, γ ] ∈ I, let p = β , q = –β ; if [α, β, γ ] ∈ II, let p = +β , q = +β . Then +β β for i = , . Using Hölder’s inequality, Lemma ., and Lemma . we get  A(t) ≤ b(t)

(t – s)

e

+ qi = 

p

t pi (β–) pi s

 pi

ds

i







t

m m–p p – m mp K p a(s) + K p p   t

p  α α pi (β–) p (γ –) i i t –s + d(t) s ds ·

e–qi s cqi (s)

q

qi ds

i



 ·



t

q q–p p – q pq K p a(s) + K p p    pi t e ( – ( – β)pi ) pi f qi (s)

= b(t)

qi

q ds

i

–(–β)pi

pi



q i p – m mp qi m m–p K p a(s) + K ds p p  θi   p i pi (γ – ) +  t B , pi (β – ) +  + d(t) α α   t 

q i p – q pq qi q q–p K · f qi (s) K p a(s) + ds p p  

·



t

e–qi s cqi (s)

    q  p q = b(t)et Mi i Ai i (t) + d(t) Ni t θi pi Bi i (t).

Here   θi = pi α(β – ) + γ –  + , Mi =

( – ( – β)pi ) –(–β)pi

pi

,

   pi (γ – ) +  , pi (β – ) +  , Ni = B α α

(.)

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78



Page 10 of 16



t

m m–p p – m mp K p a(s) + K Ai (t) = e c (s) p p     t q q–p p – q pq qi qi p Bi (t) = f (s) K a(s) + ds, K p p  –qi s qi

qi ds, i = , .

Similarly, we can get 

t

B(t) ≤

(t – s)pi (β–) epi s ds

p 

t

i



e–qi s cqi (s)vqi (s) ds

q

i





 pi

t

–qi s qi

= e Mi

q

t qi

i

c (s)v (s) ds

e

(.)



and 

t

C(t) ≤

α

t –s

 α pi (β–) pi (γ –) s

p  ds



q

t

i

qi

qi

f (s)v (s) ds

i



  = Ni t θi pi



t

f qi (s)vqi (s) ds

q

i

.

(.)



From (.), (.), and (.) we get  t

q  i m m–p pi t –qi s qi qi p v(t) ≤ A(t) + K b(t)e Mi e c (s)v (s) ds p 

q  t  θ  p i q q–p qi qi i p i + K d(t) Ni t f (s)v (s) ds . p  Using Lemma . and (.) we can get 

qi  t  m m–p p e–qi s cqi (s)vqi (s) ds K p b(t)et Mi i p       qi t q  q q–p f i (s)vqi (s) ds + qi – K p d(t) Ni t θi pi p   t  t qi – qi –qi s qi qi e c (s)v (s) ds + gi (t) f qi (s)vqi (s) ds =  A (t) + ei (t)

vqi (t) ≤ qi – Aqi (t) + qi –





     q q  p q ≤ qi – b(t)et Mi i Ai i (t) + d(t) Ni t θi pi Bi i (t) i  t   ei (t)e–qi s cqi (s) + gi (t)f qi (s) vqi (s) ds. +

(.)



Here  qi –

ei (t) = 

 m m–p p K p b(t)et Mi i p

qi

 qi –

,

gi (t) = 

  q q–p K p d(t) Ni t θi pi p

qi .

By Lemma . we have t

vqi (t) ≤ a¯ (t) + e

¯



t

e–

 b(t,s) ds



τ

¯  b(τ ,s) ds



¯ ,τ) + a¯ (s)b(τ

 

τ

b¯ τ (τ , s)¯a(s) ds dτ .

(.)

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 11 of 16

Here      q q  p q a¯ i (t) = qi – b(t)et Mi i Ai i (t) + d(t) Ni t θi pi Bi i (t) i ,

b¯ i (t, s) = ei (t)e–qi s cqi (s) + gi (t)f qi (s). Combining (.) and (.) we get   t  t τ ¯ ¯ u(t) ≤ a(t) + a¯ i (t) + e  bi (t,s) ds e–  bi (τ ,s) ds 

  · a¯ i (s)b¯ i (τ , τ ) +

τ



 q  p i ¯b (τ , s)¯ai (s) ds dτ . iτ

(.)

Considering the two situations for i = ,  and using the parameters α, β, and γ to pi , qi and θi in (.), we can get (.) and (.), respectively. This completes the proof of Theorem ..  Corollary . Suppose that the conditions of Theorem . hold, if b(t) = , then 

t

up (t) ≤ a(t) + d(t)

t α – sα

β–

sγ – f (s)uq (s) ds



and we deduce the following results. For [α, β, γ ] ∈ I,   t  t τ ¯ b¯  (t,s) ds  u(t) ≤ a(t) + a¯  (t) + e e  b (τ ,s) ds 

  · a¯  (s)b¯  (τ , τ ) +

τ



 –β  p b¯ τ (τ , s)¯a (s) ds dτ ,

(.)

where    β a¯  (t) = β d(t)N t (α+)(β–)+γ –β B (t), β

b¯  (t, s) =  –β



q q–p β K p d(t)N t (α+)(β–)+γ f (s) p

  –β

,

N , B (t) are the same as in Theorem .. For [α, β, γ ] ∈ II,   t  t τ ¯ b¯  (t,s) ds  u(t) ≤ a(t) + a¯  (t) + e e  b (τ ,s) ds 

  ¯ · a¯  (s)b (τ , τ ) + 

τ

β   +β p ¯b (τ , s)¯a (s) ds dτ , τ

where a¯  (t) = 

+β β



+β

d(t)N+β t

[α(β–)+γ ](+β)+β +β

 +β β

B (t),

(.)

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

b¯  (t, s) = 

+β β



Page 12 of 16

+β [α(β–)+γ ](+β)+β q q–p +β f (s) K p d(t)N+β t p

 +β β ,

N , B (t) are the same as in Theorem .. Corollary . Suppose that the conditions of Theorem . hold, α = γ = . If  up (t) ≤ a(t) + b(t)



t

(t – s)β– c(s)um (s) ds + d(t) 

t

(t – s)β– f (s)uq (s) ds, 

then for [α, β, γ ] ∈ I,   t  t τ ¯ ¯ e  b (τ ,s) ds u(t) ≤ a(t) + a¯  (t) + e  b (t,s) ds 

  · a¯  (s)b¯  (τ , τ ) +

τ



 –β  p ¯b (τ , s)¯a (s) ds dτ , τ

(.)

where β     ¯ β –β β –β a¯  (t) =  –β b(t)et M A (t) + d(t) N¯  t θ B (t) –β ,   –β     θ¯ β –β  s  m m–p q q–p β  p p β– β– K K d(t) N t b(t)M e c (s) + f p p   β –  β –   ¯ , θ¯ = (β – ) +  = , N = B , β β β β

b¯  (t, s) =  –β



 –β

(s) ,

M , A (t), B (t) are the same as in Theorem .. For [α, β, γ ] ∈ II,   t  t τ ¯ b¯  (t,s) ds  e  b (τ ,s) ds u(t) ≤ a(t) + a¯  (t) + e 

  · a¯  (s)b¯  (τ , τ ) + 

τ

β   +β p b¯ τ (τ , s)¯a (s) ds dτ ,

where a¯  (t) = 

+β β



+β β β +β   +β ¯  b(t)et M+β A+β (t) + d(t) N¯  t θ +β B+β (t) β ,

 +β +β β m m–p – +β s +β +β t p K b(t)e M e β c β (s) p

 +β   +β β +β  q q–p θ¯ +β p β ¯ f (s) , K d(t) N t + p   β  N¯  = B , ,  + β b¯  (t, s) = 

θ¯ =

+β β



β   + β (β – ) +  = ,  + β  + β

M , A (t), B (t) are the same as in Theorem ..

(.)

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 13 of 16

Remark . If b(t) = , then Theorem . becomes Theorem . in []. Remark . If b(t) = d(t) = , p = m = α = γ = , then Theorem . becomes Theorem  in []. Due to the difference in methods, our results are not the same as in []. The results in Theorem . have the simple exponential function of Theorem  in [].

4 Applications () Consider the delay fractional differential equations with initial condition ⎧ C β p ⎪ ⎪ ⎨t Dt x (t) = f (t, x(t), x(t – r)), t ∈ I = [t , T), C β k = , , , . . . , m – , t Dt x(t ) = bk , ⎪ ⎪ ⎩ x(t) = ϕ(t), t ∈ I = [t – r, t ).

(.)

β

Here Ct Dt is the Caputo fractional derivative of order β (m –  ≤ β < m), r ∈ R+ is a real constant, f (t, y, z) ∈ C(I × R , R), ϕ is a given continuously differentiable function on [t – r, t ] up to order m (m = –[–β]). In this case, we denote ϕ k (t ) = bk , k = , , , . . . , m – . In [], the initial value problem (.) is equivalent to the Volterra fractional integral equation ⎧ ⎨xp (t) = m–

bk k= k! (t

⎩x(t) = ϕ(t),

– t )k +

 (β)

t

t (t

t ∈ I,

– s)β– f (s, x(s), x(s – r)) ds,

t ∈ I = [t – r, t ).

(.)

The next theorem deals with the estimates of the solution of (.). Theorem . Suppose that |f (t, y, z)| ≤ b(t)|y|q + c(t)|z|l , t ∈ I, m, p, q, l ∈ R+ , p ≥ q, p ≥ l, b(t), c(t) ∈ C(I, R+ ). If x(t) is the solution of initial problem (.), then the following estimations hold. (i) Suppose that β >  , then ⎧  ⎨|x(t)| ≤ {g (t) +  t k (s)ϕ l (s – r) ds +  t F (s)e st G (τ ) dτ ds} p ,  t  t   ⎩|x(t)| ≤ {g (t) +  t F (s)e st G (τ ) dτ ds} p t ∈ [t + r, T).

t ∈ [t , t + r),

(.)

t

Here –t

g (t) = e

 m–

|bk | k=

k!

 (t – t )

k

,



(β – ) t( pq –) b(t)  h (t) = e , β (β)



(β – ) c(t)  t( pl –) k (t) = e , β (β)

 t q q–p p – q pq q q–p K , k (s)ϕ l (s – r) ds + G (t) = K p h (t), F (t) = K p h (t) g (t) + p p p t

q q–p p – q pq l l–p p – l pl F (t) = K p g (t) + K h (t) + K p g (t – r) + K k (t), p p p p q q–p l l–p G (t) = K p h (t) + K p k (t). p p

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 14 of 16

(ii) If  < β ≤  , then ⎧  ⎨|x(t)| ≤ {g (t) +  t k (s)ϕ l (s – r) ds +  t F (s)e st G (τ ) dτ ds} p ,  t  t   ⎩|x(t)| ≤ {g (t) +  t F (s)e st G (τ ) dτ ds} p , t ∈ [t + r, T).

t ∈ [t , t + r),

(.)

t

Here – +β β t

 β

g (t) =  e

 m–

|bk | k=

 β

h (t) = 



(β  ) ( + β)β 

k!  β

 +β β (t – t )

k

q

,

[– +β β + p (+β)]t

e



b(t) (β)

+β β ,

 β

+β β q (β  ) [– +β + p (+β)]t c(t) β k (t) =  e ,  β (β) ( + β)

 t q q–p p – q pq q q–p l p F (t) = K h (t) g (t) + K , k (s)ϕ (s – r) ds + G (t) = K p h (t), p p p t

q–p p – q pq p – l pl q p l l–p p K g (t) + K h (t) + K g (t – r) + K k (t), F (t) = p p p p  β



q q–p l l–p G (t) = K p h (t) + K p k (t). p p Proof By (.), we derive that ⎧ t m– |bk |  p k β– q ⎪ ⎪ ⎨|x(t)| ≤ k= k! (t – t ) + (β) t (t – s) b(s)|x(s)| ds t  β– + (β) c(s)|x(s – r)|l ds, t ∈ I, t (t – s) ⎪ ⎪ ⎩ |x(t)| = |ϕ(t)|, t ∈ [t – r, t ).

(.)

Using Theorem ., we get the desired conclusion. This proves the results (.) and (.).  () In this section, based on the definition of Riemann-Liouville (R-L) and Erdélyi-Kober (E-K) fractional integral, we will study the boundedness of a certain FDE with R-L fractional operator and E-K fractional operator. The definitions of two fractional operators are given below. Definition . ([]) The R-L fractional integral and fractional derivative of order α of the function f (x) ∈ C(R+ , R) are given by  I f (x) = (α)



α

Dα f (x) =

x

(x – t)α– f (t) dt, 

d  ( – α) dx



α > ,

x

(x – t)–α f (t) dt,

 < α < ,



provided that the right side is point-wise defined on R+ .

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 15 of 16

Definition . ([, ]) The E-K fractional integral of continuous function f (x) ∈ C(R+ , R) is defined by γ ,δ

Iβ f (x) =

x–β(γ +δ) (δ)



x

xβ – t β

δ–

  t βγ f (t)d t β ,

δ, γ , β ∈ R+ ,



provided that the right side is point-wise defined on R+ . Consider the following Volterra type integral equations: b(t) u (t)– (β)



p

t

(t –s)

β–



d(t) c(s)u (s) ds – (β)



m

t

t α –sα

β–

sγ – f (s)uq (s) ds = a(t). (.)



Theorem . Let u(t), a(t), b(t), c(t), d(t), f (t) ∈ C[, ∞), p, m, q, α, β, γ be the same as in Theorem .. Then, for any K > , every solution of equation (.) has the bounds and the , |c(t)|, |d(t)| , |f (t)| instead same modality as (.), (.), in which we have |u(t)|, |a(t)|, |b(t)| (β) (β) of u(t), a(t), b(t), c(t), d(t), and f (t). Proof From (.) we have  t   m     u(t)p ≤ a(t) + |b(t)| (t – s)β– c(s)u(s) ds (β)   q |d(t)| t  α α β– γ –   – t –s s f (s) u(s) ds. (β)  Then by the difference in the selection of a(t), b(t), c(t), d(t), f (t), p, m, q, α, β, γ , we get the desired results.  Competing interests The authors declare that they have no competing interests. Authors’ contributions RX carried out the generalized weakly singular integral inequalities and completed the corresponding proof. FM participated in Section 4 - Applications. All authors read and approved the final manuscript. Acknowledgements This research is supported by National Science Foundation of China (11171178 and 11271225). Received: 26 October 2015 Accepted: 9 February 2016 References 1. Denton, Z, Vatsala, AS: Fractional integral inequalities and applications. Comput. Math. Appl. 59, 1087-1094 (2010) 2. Ye, H, Gao, J, Ding, Y: A generalized Gronwall inequality and its application to a fractional differential equation. J. Math. Anal. Appl. 328, 1075-1081 (2007) ˘ M: A new approach to an analysis of Henry type integral inequalities and their Bihari type versions. J. Math. 3. Medved, Anal. Appl. 214, 349-366 (1997) ˘ M: Integral inequalities and global solutions of semilinear evolution equations. J. Math. Anal. Appl. 267, 4. Medved, 643-650 (2002) 5. Ye, H, Gao, J: Henry-Gronwall type retarded integral inequalities and their applications to fractional differential equations with delay. Appl. Math. Comput. 218, 4152-4160 (2011) 6. Shao, J, Meng, F: Gronwall-Bellman type inequalities and their applications to fractional differential equations. Abstr. Appl. Anal. 2013, Article ID 217641 (2013). doi:10.1155/2013/217641 7. Feng, Q, Meng, F: Some new Gronwall-type inequalities arising in the research of fractional differential equations. J. Inequal. Appl. 2013, 429 (2013) 8. Ma, Q, Peˇcari´c, Q: Some new explicit bounds for weakly singular integral inequalities with applications to fractional differential and integral equations. J. Math. Anal. Appl. 341, 894-905 (2008) 9. Cheung, WS, Ma, QH, Tseng, S: Some new nonlinear weakly singular integral inequalities of Wendroff type with applications. J. Inequal. Appl. 2008, Article ID 909156 (2008)

Xu and Meng Journal of Inequalities and Applications (2016) 2016:78

Page 16 of 16

10. Wang, H, Zheng, K: Some nonlinear weakly singular integral inequalities with two variables and applications. J. Inequal. Appl. 2010, Article ID 345701 (2010) 11. Lakhal, F: A new nonlinear integral inequity of Wendroff type with continuous and weakly singular kernel and its application. J. Math. Inequal. 6(3), 367-379 (2012) 12. Henry, D: Geometric Theory of Semilinear Parabolic Equations. Springer, Berlin (1981) 13. Jiang, FC, Meng, FW: Explicit bounds on some new nonlinear integral inequalities with delay. J. Comput. Appl. Math. 205, 479-486 (2007) 14. Prudnikov, AP, Brychkov, YA, Marichev, OI: Integrals and Series: Elementary Functions, vol. 1. Nauka, Moscow (1981) (in Russian) 15. Ma, QH, Yang, EH: Estimations on solutions of some weakly singular Volterra integral inequalities. Acta Math. Appl. Sin. 25, 505-515 (2002) 16. Zhang, H, Meng, F: Integral inequalities in two independent variables for retarded Volterra equations. Appl. Math. Comput. 199, 90-98 (2008) 17. Podlubny, I: Fractional Differential Equations. Academic Press, San Diego (1999) 18. Kiryakova, VS: Generalized Fractional Calculus and Applications. Pitman Res. Notes Math. Ser., vol. 301. Longman, Harlow (1994) 19. Al-Saqabi, B, Kiryakova, VS: Explicit solutions of fractional integral and differential equations involving Erdélyi-Kober operators. Appl. Math. Comput. 95, 1-13 (1998)