AAA Abstract and Applied Analysis 1687-0409 1085-3375 Hindawi Publishing Corporation 705126 10.1155/2013/705126 705126 Research Article New Gronwall-Bellman Type Inequalities and Applications in the Analysis for Solutions to Fractional Differential Equations http://orcid.org/0000-0002-3358-8328 Zheng Bin http://orcid.org/0000-0002-7435-718X Feng Qinghua Huang Chengming School of Science Shandong University of Technology Zibo, Shandong 255049 China sdut.edu.cn 2013 28 11 2013 2013 29 05 2013 17 10 2013 22 10 2013 2013 Copyright © 2013 Bin Zheng and Qinghua Feng. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Some new Gronwall-Bellman type inequalities are presented in this paper. Based on these inequalities, new explicit bounds for the related unknown functions are derived. The inequalities established can also be used as a handy tool in the research of qualitative as well as quantitative analysis for solutions to some fractional differential equations defined in the sense of the modified Riemann-Liouville fractional derivative. For illustrating the validity of the results established, we present some applications for them, in which the boundedness, uniqueness, and continuous dependence on the initial value for the solutions to some certain fractional differential and integral equations are investigated.

1. Introduction

It is well known that the Gronwall-Bellman inequality [1, 2] and their generalizations can provide explicit bounds for solutions to differential and integral equations as well as difference equations. Recently, many authors have researched various generalizations of the Gronwall-Bellman inequality (e.g., see  and the references therein). These Gronwall-Bellman type inequalities established have proved to be useful in the research of boundedness, global existence, uniqueness, stability, and continuous dependence of solutions to differential and integral equations as well as difference equations. However, in the research for the properties of solutions to some fractional differential and integral equations, the earlier inequalities established are inadequate to fulfill such analysis, and it is necessary to establish new Gronwall-Bellman type inequalities so as to obtain the desired result.

On the other hand, recently, Jumarie presented a new definition for the fractional derivative named the modified Riemann-Liouville fractional derivative (see [27, 28]). The modified Riemann-Liouville fractional derivative is defined by the following expression.

Definition 1.

The modified Riemann-Liouville derivative of order α is defined by the following expression: (1)Dtαf(t)={1Γ(1-α)ddt0t(t-ξ)-α·(f(ξ)-f(0))dξ,0<α<1,(f(n)(t))(α-n),nα<n+1,n1.

Definition 2.

The Riemann-Liouville fractional integral of order α on the interval [0,t] is defined by(2)Iαf(t)=1Γ(1+α)0tf(s)(ds)α=1Γ(α)0t(t-s)α-1f(s)ds.

Some important properties for the modified Riemann-Liouville derivative and fractional integral are listed as follows (the interval concerned below is always defined by [0,t]):

Dtαtr=(Γ(1+r)/Γ(1+r-α))tr-α.

Dtα(f(t)g(t))=g(t)Dtαf(t)+f(t)Dtαg(t).

Dtαf[g(t)]=fg[g(t)]Dtαg(t)=Dgαf[g(t)](g(t))α.

Iα(Dtαf(t))=f(t)-f(0).

Iα(g(t)Dtαf(t))=f(t)g(t)-f(0)g(0)-Iα(f(t)Dtαg(t)).

The modified Riemann-Liouville derivative for a constant is zero.

The modified Riemann-Liouville derivative has many excellent characters in handling many fractional calculus problems. Many authors have investigated various applications of the modified Riemann-Liouville fractional derivative. For example, in , the authors seeked exact solutions for some types of fractional differential equations based on the modified Riemann-Liouville fractional derivative, and in , the modified Riemann-Liouville fractional derivative was used in fractional calculus of variations, where the authors considered the fractional basic problem with free boundary conditions as well as problems with isoperimetric and holonomic constraints in the calculus of variations. In , Khan et al. presented a fractional homotopy perturbation method (FHPM) for solving fractional differential equations of any fractional order based on the modified Riemann-Liouville fractional derivative. In , fractional variational iteration method based on the modified Riemann-Liouville fractional derivative was concerned. In , a fractional variational homotopy perturbation iteration method was proposed.

Motivated by the wide applications of the modified Riemann-Liouville fractional derivative, in this paper, we use this type of fractional derivative to establish some fractional Gronwall-Bellman type inequalities. Based on these inequalities and some basic properties of the modified Riemann-Liouville fractional derivative, we derive explicit bounds for unknown functions concerned in these inequalities. As for applications, we apply these inequalities to research qualitative properties such as the boundedness, uniqueness, and continuous dependence on initial data for solutions to some certain fractional differential and integral equations.

We organize the rest of this paper as follows. In Section 2, we present the main inequalities, and based on them derive explicit bounds for unknown functions in these inequalities. Then in Section 3, we apply the results established in Section 2 to research boundedness, uniqueness, and continuous dependence on initial data for the solution to some certain fractional differential and integral equations.

2. Main Results Lemma 3 (see [<xref ref-type="bibr" rid="B38">38</xref>]).

Assume that a0,pq0; and p0, then, for any K>0, one has (3)aq/pqpK(q-p)/pa+p-qpKq/p.

Lemma 4.

Let α>0, a,b,u be continuous functions defined on t0. Then for t0, (4)Dtαu(t)a(t)+b(t)u(t) implies (5)u(t)u(0)exp[0tα/Γ(1+α)b((sΓ(1+α))1/α)ds]+1Γ(α)0t(t-τ)α-1a(τ)×exp[-τα/Γ(1+α)tα/Γ(1+α)b((sΓ(1+α))1/α)ds]dτ.

Proof.

By the properties (a),(b), and (c) we have the following observation: (6)Dtα{u(t)exp[-0tα/Γ(1+α)b((sΓ(1+α))1/α)ds]}=exp[-0tα/Γ(1+α)b((sΓ(1+α))1/α)ds]×Dtαu(t)+u(t)Dtα×{exp[-0tα/Γ(1+α)b((sΓ(1+α))1/α)ds]}=exp[-0tα/Γ(1+α)b((sΓ(1+α))1/α)ds]×Dtαu(t)-b(t)u(t)×exp[-0tα/Γ(1+α)b((sΓ(1+α))1/α)ds]×Dtα(tαΓ(1+α))=exp[-0tα/Γ(1+α)b((sΓ(1+α))1/α)ds]  ×[Dtαu(t)-b(t)u(t)]a(t)exp[-0tα/Γ(1+α)b((sΓ(1+α))1/α)ds].

Substituting t with τ, fulfilling fractional integral of order α for (6) with respect to τ from 0 to t, we deduce that (7)u(t)exp[-0tα/Γ(1+α)b((sΓ(1+α))1/α)ds]u(0)+1Γ(α)0t(t-τ)α-1a(τ)×exp[-0τα/Γ(1+α)b((sΓ(1+α))1/α)ds]dτ, which implies (8)u(t)exp[0tα/(Γ(1+α))b((sΓ(1+α))1/α)ds]×{b((sΓ(1+α))1/α)dsexp[-0τα/(Γ(1+α))]dτ}u(0)+1Γ(α)×0t(t-τ)α-1a(τ)×exp[-0τα/(Γ(1+α))b((sΓ(1+α))1/α)ds]dτ}.

The desired result can be obtained subsequently.

Lemma 5.

Suppose α>0, the functions u, a, b, g, and h are nonnegative continuous functions defined on t0, and a,b are nondecreasing, p,q, and r are constants with pq>0,  pr>0. If the following inequality holds: (9)up(t)a(t)+b(t)Γ(α)×0t(t-s)α-1×[g(s)uq(s)+0sh(ξ)ur(ξ)dξ]ds,  t0, then one has the following explicit estimate for u(t): (10)u(t){[0tα/(Γ(1+α))H2((sΓ(1+α))1/α)ds]a(t)b(t)×exp[0tα/Γ(1+α)H2((sΓ(1+α))1/α)ds]+b(t)Γ(α)×0t(t-τ)α-1H1(τ)×exp[τα/(Γ(1+α))tα/(Γ(1+α))-b(t)×τα/Γ(1+α)tα/Γ(1+α)H2×((sΓ(1+α))1/α)dsτα/(Γ(1+α))tα/(Γ(1+α))]dτ}1/p,t0,

where (11)H1(t)=g(t)(p-qpKq/p)+0th(ξ)(p-rpKr/p)dξ,H2(t)=g(t)qpK(q-p)/p+0th(ξ)rpK(r-p)/pdξ, and K>0 is an arbitrary constant.

Proof.

Fix T0, and let t[0,T]. Denote (12)v(t)=a(T)+b(T)Γ(α)×0t(t-s)α-1[0sh(ξ)ur(ξ)dξ]g(s)uq(s)+0sh(ξ)ur(ξ)dξ]ds.

Then we have (13)up(t)v(t),t[0,T].

Since u,g, and h are continuous, then there exists a constant M such that |g(t)uq(t)+0th(ξ)ur(ξ)dξ|M for t[0,ε], where ε>0. So for t[0,ε], we have |0t(t-s)α-1[g(s)uq(s)+0sh(ξ)ur(ξ)dξ]ds|M0t(t-s)α-1ds=(M/α)tα. Then one can see v(0)=a(T). Furthermore, (14)Dtαv(t)=b(T)[g(t)uq(t)+0th(ξ)ur(ξ)dξ]b(T)[g(t)vq/p(t)+0th(ξ)vr/p(ξ)dξ].

By Lemma 3, we have (15)Dtαv(t)b(T)[g(t)(qpK(q-p)/pv(t)+p-qpKq/p)+0th(ξ)(rpK(r-p)/pv(ξ)+p-rpKr/p)dξ]=b(T)[g(t)(p-qpKq/p)+0th(ξ)(p-rpKr/p)dξ]+b(T)[0t]g(t)qpK(q-p)/pv(t)+0th(ξ)rpK(r-p)/pv(ξ)dξ]b(T)[g(t)(p-qpKq/p)+0th(ξ)(p-rpKr/p)dξ]+b(T)[0t]g(t)qpK(q-p)/p  +0th(ξ)rpK(r-p)/pdξ]v(t)=b(T)H1(t)+b(T)H2(t)v(t),t[0,T], where H1(t),H2(t) are defined in (11). Applying Lemma 4 to (15), considering v(0)=a(T), we get that (16)v(t)a(T)b(T)×exp[0tα/Γ(1+α)H2((sΓ(1+α))1/α)ds]+b(T)Γ(α)×0t(t-τ)α-1H1(τ)×exp[τα/(Γ(1+α))tα/(Γ(1+α))H2-b(T)τα/Γ(1+α)tα/Γ(1+α)H2×((sΓ(1+α))1/α)ds[-b(T)τα/(Γ(1+α))tα/(Γ(1+α))H2]dτ,t[0,T].

Letting t=T in (16) and considering T0 is arbitrary, after substituting T with t, we get that (17)v(t)a(t)b(t)×exp[0tα/Γ(1+α)H2((sΓ(1+α))1/α)ds]+b(t)Γ(α)0t(t-τ)α-1H1(τ)×exp[-b(t)τα/Γ(1+α)tα/Γ(1+α)H2×((sΓ(1+α))1/α)ds[-b(t)τα/(Γ(1+α))tα/(Γ(1+α))H2]dτ,t0.

Combining (13) and (17), we get (10).

Remark 6.

In Lemma 5, if a,b are not necessarily nondecreasing, then one can let v(t)=(1/Γ(α))0t(t-s)α-1[g(s)uq(s)+0sh(ξ)ur(ξ)dξ]ds instead in the proof, and following in a similar process, obtain another explicit bound for u(t): (18)u(t){a(t)+b(t)Γ(α)×0t(t-τ)α-1H^1(τ)×exp[-τα/Γ(1+α)tα/Γ(1+α)H^2×((sΓ(1+α))1/α)ds[-τα/(Γ(1+α))tα/(Γ(1+α))H^2]dτ}1/p,t0, where (19)H^1(t)=g(t)(p-qpKq/p+qpK(q-p)/pa(t))+0th(ξ)(p-rpKr/p+rpK(r-p)/pa(ξ))dξ,H^2(t)=b(t)[g(t)qpK(q-p)/p+0th(ξ)rpK(r-p)/pdξ].

Remark 7.

We note that if we take g(t)1,  h(t)0, and p=q=1, then the inequality (9) in Lemma 5 reduces to the inequality in [39, Theorem 1]. So the present inequality is of more general form than that in . Furthermore, the explicit bounds obtained for the function u(t) above are essentially different from that in .

Theorem 8.

Suppose that α>0, the functions u,  a,  b,  g,  p,  q, and r are defined as in Lemma 5, and p1, m is a nonnegative continuous function defined on t0. If the following inequality holds: (20)up(t)a(t)+0tm(s)up(s)ds+b(t)Γ(α)0t(t-s)α-1×[+0sh(ξ)ur(ξ)dξ]g(s)uq(s)+0sh(ξ)ur(ξ)dξ]ds,t0,

then we have (21)u(t)exp(1p0tm(s)ds)×{a(t)b(t)exp[0tα/Γ(1+α)H~2×((sΓ(1+α))1/α)ds[0tα/(Γ(1+α))H~2]+b(t)Γ(α)  ×0t(t-τ)α-1H~1(τ)×exp[-b(t)τα/Γ(1+α)tα/Γ(1+α)H~2×((sΓ(1+α))1/α)ds]dτ{a(t)b(t)exp[0tα/(Γ(1+α))H~2}1/p,t0, where (22)H~1(t)=g~(t)(p-qpKq/p)+0th~(ξ)(p-rpKr/p)dξ,(23)H~2(t)=g~(t)qpK(q-p)/p+0th~(ξ)rpK(r-p)/pdξ,(24)g~(t)=g(t)exp(qp0sm(τ)dτ),h~(t)=h(t)exp(rp0tm(τ)dτ).

Proof.

Denote zp(t) by a(t) + (b(t)/Γ(α))0t(t-s)α-1[g(s)uq(s)+0sh(ξ)ur(ξ)dξ]ds. Then (25)up(t)zp(t)+0tm(s)up(s)ds.

Since a(t),b(t) are both nondecreasing, then z(t) is also nondecreasing, and subsequently we can deduce that (26)up(t)zp(t)exp(0tm(s)ds).

Furthermore,(27)zp(t)a(t)+b(t)Γ(α)×0t(t-s)α-1×[rp0ξm(τ)dτdξ]g(s)zq(s)×exp(qp0sm(τ)dτ)+0sh(ξ)zr(ξ)exp(rp0ξm(τ)dτ)dξ]ds=a(t)+b(t)Γ(α)0t(t-s)α-1×[+0sh~(ξ)zr(ξ)dξ]g~(s)zq(s)+0sh~(ξ)zr(ξ)dξ]ds, where g~(t),  h~(t) are defined in (24). Then applying Lemma 5 to (23) yields (28)z(t){[0tα/(Γ(1+α))H~2((sΓ(1+α))1/α)ds]a(t)b(t)×exp[0tα/Γ(1+α)H~2((sΓ(1+α))1/α)ds]+b(t)Γ(α)×0t(t-τ)α-1H~1(τ)×exp[-b(t)τα/Γ(1+α)tα/Γ(1+α)H~2×((sΓ(1+α))1/α)ds-b(t)τα/Γ(1+α)tα/Γ(1+α)H~2]dτ[0tα/(Γ(1+α))H~2((sΓ(1+α))1/α)ds]}1/p,t0,

where H~1(t),H~2(t) are defined in (22) and (23). Combining (26) and (28), we obtain the desired result.

Lemma 9.

Suppose α>0, the functions u,a,b, and g are defined as in Lemma 5, and ω is a nonnegative continuous function defined on t0 being nondecreasing, and ω(r)>0 for r>0. Define G(v)=0v(1/ω(r))dr, and assume G(v)< for v<. If the following inequality holds: (29)u(t)a(t)+b(t)Γ(α)0t(t-s)α-1  ×g(s)ω(u(s))ds,t0,

then we have the following explicit estimate for u(t): (30)u(t)G-1×[G(a(t))+b(t)Γ(α)0t(t-τ)α-1g(τ)dτ],t0.

Proof.

Fix T0, and let t[0,T]. Denote (31)v(t)=a(T)+δ+b(T)Γ(α)0t(t-s)α-1g(s)ω(u(s))ds, where δ>0. Then we have (32)u(t)v(t),t[0,T].

Since u,g, and ω are continuous, then there exists a constant M such that |g(t)ω(u(t))|M for t[0,ε], where ε>0. So for t[0,ε], we have |0t(t-s)α-1g(s)ω(u(s))ds|  M0t(t-s)α-1ds=(M/α)tα. Then one can see v(0)=a(T)+δ, and (33)Dtαv(t)=b(T)g(t)ω(u(t))b(T)g(t)ω(v(t)),

which implies (34)Dtαv(t)ω(v(t))b(T)g(t).

That is, (35)DtαG(v(t))b(T)g(t).

Substituting t with τ, fulfilling fractional integral of order α for (35) with respect to τ from 0 to t, we deduce that (36)G(v(t))-G(v(0))b(T)Γ(α)0t(t-τ)α-1g(τ)dτ,

which implies (37)v(t)G-1×[G(a(T)+δ)+b(T)Γ(α)0t(t-τ)α-1g(τ)dτ],

and furthermore, (38)u(t)G-1[×0t(t-τ)α-1g(τ)dτ]G(a(T)+δ)+b(T)Γ(α)×0t(t-τ)α-1g(τ)dτ],t[0,T].

Letting t=T in (38), we get that (39)u(T)G-1[G(a(T)+δ)+b(T)Γ(α)×0T(T-τ)α-1g(τ)dτ].

Since T>0 is arbitrary, substituting T with t in (39) and after letting δ0, we can obtain the desired result.

Theorem 10.

Suppose α>0, the functions u,a,b, and g are defined as in Lemma 5, m is a nonnegative continuous function defined on t0, and ω is defined as in Lemma 9, and furthermore, assume that ω is submultiplicative; that is, ω(αβ)ω(α)ω(β),  α,  β0. Define G(v)=0v(1/ω(r))dr, and assume G(v)< for v<. If the following inequality holds: (40)u(t)a(t)+0tm(s)u(s)ds+b(t)Γ(α)0t(t-s)α-1g(s)ω(u(s))ds,t0,

then we have the following explicit estimate for u(t): (41)u(t)G-1[0τm(ξ)dξdτ]G(a(t))+b(t)1Γ(α)  ×0t(t-τ)α-1g(τ)  ×exp(0τm(ξ)dξ)dτ],t0.

Proof.

Let z(t)=a(t)+(b(t)/Γ(α))0t(t-s)α-1g(s)ω(u(s))ds.

Then we have (42)u(t)z(t)+0tm(s)u(s)ds.

Since z(t) is nondecreasing, then furthermore we have (43)u(t)z(t)exp(0tm(s)ds).

So (44)z(t)a(t)+b(t)Γ(α)0t(t-s)α-1g(s)ω(z(s))×exp(0sm(ξ)dξ)ds.

Since ω is submultiplicative, then furthermore we get that (45)z(t)a(t)+b(t)Γ(α)×0t(t-s)α-1g(s)ω(z(s))×ω(exp(0sm(ξ)dξ))ds.

Noticing that the structure of (45) is similar to that of the inequality (29), a suitable application of Lemma 9 to (45) yields that (46)z(t)G-1[G(a(t))+b(t)1Γ(α)×0t(t-τ)α-1g(τ)ω(exp(0τm(ξ)dξ))dτ].

Combining (43) and (46) we can obtain the desired result.

3. Applications

In this section, we apply the inequalities established above to research of boundedness, uniqueness, continuous dependence on the initial value for solutions to certain fractional differential and integral equations. Let us first consider the following IVP of fractional differential equation: (47)Dt0.5u3(t)=L(t,u(t),0tM(ξ,u(ξ))dξ),t0,u(0)=C, where uC([0,),R),  MC(R×R,R), and LC([0,)×R2,R).

Theorem 11.

Suppose that u(t) is a solution of the IVP (47). If |L(t,u,v)|g(t)|u|3+|v|, and |M(t,u)|h(t)|u|3, where g,h are nonnegative continuous functions on [0,), then we have the following estimate for u(t): (48)|u(t)||C|exp[0t/Γ(1.5)H2((sΓ(1.5))2)ds]3,t0, where H2(t)=g(t)+0th(ξ)dξ.

Proof.

Similar to [28, Equation (5.1)], we can obtain the equivalent integral form of the IVP (47) as follows: (49)u3(t)=C+1Γ(0.5)0t(t-s)-0.5×L(s,u(s),0sM(ξ,u(ξ))dξ)ds.

So (50)|u(t)|3|C|+1Γ(0.5)×0s(t-s)-0.5×|L(s,u(s),0sM(ξ,u(ξ))dξ)|ds|C|+1Γ(0.5)×0s(t-s)-0.5×[g(s)|u(s)|3+|0sM(ξ,u(ξ))dξ|]ds|C|+1Γ(0.5)×0s(t-s)-0.5[0sg(s)|u(s)|3+0s|M(ξ,u(ξ))|dξ]ds|C|+1Γ(0.5)×0s(t-s)-0.5[0sg(s)|u(s)|3+0sh(ξ)|u(ξ)|3dξ]ds.

Then a suitable application of Lemma 5 to (50) (with α=0.5,  p=q=r=3) yields the desired result.

Remark 12.

At the end of the proof of Theorem 11, if we apply the result of Remark 6 instead of Lemma 5 to (50), then we obtain the following estimate: (51)|u(t)|(|C|{((sΓ(1+α))1/α)ds[-τα/(Γ(1+α))tα/(Γ(1+α))H2]dτ}1+1Γ(α)×0t(t-τ)α-1H2(τ)×exp[-τα/Γ(1+α)tα/Γ(1+α)H2×((sΓ(1+α))1/α)ds[-τα/(Γ(1+α))tα/(Γ(1+α))H2]dτ})1/3,t0, where H2(t) is defined as in Theorem 11.

Remark 13.

In Theorem 11, if we change the conditions by |L(t,u,v)|g(t)|u|+|v|, and |M(t,u)|h(t)|u|, where g,  h are nonnegative continuous functions on [0,), then we can obtain the following estimate for u(t): (52)|u(t)|{|C|exp[0t/Γ(1.5)H2((sΓ(1.5))2)ds]+1Γ(0.5)0t(t-τ)-0.5H1(τ)×exp[-τ/Γ(1.5)t/Γ(1.5)H2×((sΓ(1.5))2)ds[-τ/(Γ(1.5))t/(Γ(1.5))H2]dτ}1/3,t0, where H1(t)=(2/3)K1/3[g(t)+0th(ξ)dξ],H2(t)=(1/3)K-2/3[g(t)+0th(ξ)dξ], and K>0 is an arbitrary constant.

Theorem 14.

If |L(t,u1,v1)-L(t,u2,v2)|g(t)|u13-u23|+|v1-v2|, |M(t,u1)-M(t,u2)|h(t)|u13-u23|, where g,  h are nonnegative continuous functions defined on [0,), then the IVP (47) has at most one solution.

Proof.

Suppose that the IVP (47) has two solutions u1(t),u2(t). Then similar to Theorem 11, we can obtain that (53)u13(t)=C+1Γ(0.5)×0t(t-s)-0.5×L(s,u1(s),0sM(ξ,u2(ξ))dξ)ds,u23(t)=C+1Γ(0.5)×0t(t-s)-0.5×L(s,u2(s),0sM(ξ,u2(ξ))dξ)ds.

Furthermore, (54)u13(t)-u23(t)=1Γ(0.5)0t(t-s)-0.5×[L(0sM(ξ,u1(ξ))dξ)s,u1(s),0sM(ξ,u1(ξ))dξ)-L(0sM(ξ,u2(ξ))dξ)s,u2(s),0sM(ξ,u2(ξ))dξ)]ds,

which implies (55)|u13(t)-u23(t)|1Γ(0.5)0t(t-s)-0.5×|L(0sM(ξ,u1(ξ))dξ)s,u1(s),0sM(ξ,u1(ξ))dξ)-L(0sM(ξ,u2(ξ))dξ)s,u2(s),0sM(ξ,u2(ξ))dξ)|ds1Γ(0.5)0t(t-s)-0.5  ×[0sM(ξ,u2(ξ))dξ|]g(s)|u13(s)-u23(s)|+|0sM(ξ,u1(ξ))dξ-0sM(ξ,u2(ξ))dξ|]ds1Γ(0.5)0t(t-s)-0.5×[0sh(ξ)|u13(ξ)-u23(ξ)|dξ]g(s)|u13(s)-u23(s)|+0sh(ξ)|u13(ξ)-u23(ξ)|dξ]ds.

Treating |u13(t)-u23(t)| as one whole function, a suitable application of Lemma 5 to (55) (with α=0.5) yields |u13(t)-u23(t)|0, which implies u1(t)u2(t). So the proof is complete.

Now we research the continuous dependence on the initial value for the IVP (47).

Theorem 15.

Suppose that u(t) is the solution of the IVP (47), and u~(t) is the solution of following IVP of fractional differential equation: (56)Dt0.5u~3(t)=L(t,u~(t),0tM(ξ,u~(ξ))dξ),t0,u~(0)=C~, where u~C([0,),R),  MC(R×R,R), and LC([0,)×R2,R). If |L(t,u1,v1)-L(t,u2,v2)|g(t)|u13-u23|+|v1-v2|, |M(t,u1)-M(t,u2)|h(t)|u13-u23|, where g,  h are nonnegative continuous functions on [0,) and |C-C~|<ε, we have the following estimate: (57)|u(t)|ε{exp[0t/Γ(1.5)H2((sΓ(1.5))2)ds]},t0,

where H2(t)=g(t)+0th(ξ)dξ.

Proof.

For the IVP (56), we have the following integral form: (58)u~3(t)=C~+1Γ(0.5)×0t(t-s)-0.5  ×L(s,u~(s),0sM(ξ,u~(ξ))dξ)ds.

So by (49) and (58), we deduce that (59)u3(t)-u~3(t)=C-C~+1Γ(0.5)×0t(t-s)-0.5×[L(0sM(ξ,u(ξ))dξ)s,u(s),0sM(ξ,u(ξ))dξ)-L(0s)s,u~(s),0sM(ξ,u~(ξ))dξ)]ds.

Furthermore, (60)|u3(t)-u~3(t)||C-C~|+1Γ(0.5)×0t(t-s)-0.5×|L(s,u(s),0sM(ξ,u(ξ))dξ)-L(s,u~(s),0sM(ξ,u~(ξ))dξ)|dsε+1Γ(0.5)×0t(t-s)-0.5×[0sM(ξ,u~(ξ))dξ|]g(s)|u3(s)-u~3(s)|  +|0sM(ξ,u(ξ))dξ          -0sM(ξ,u~(ξ))dξ|]dsε+1Γ(0.5)×0t(t-s)-0.5×[+0sh(ξ)|u3(ξ)-u~3(ξ)|dξ]g(s)|u3(s)-u~3(s)|+0sh(ξ)|u3(ξ)-u~3(ξ)|dξ]ds.

A suitable application of Lemma 5 to (60) yields the desired result.

Remark 16.

Theorem 15 indicates that the solution of the IVP (47) depends continuously on the initial value u(0)=C.

Example 17.

Consider the following fractional integral equation: (61)u3(t)=C+0tL1(s,u(s))ds+IαL2(t,u(t),0tM(ξ,u(ξ))dξ),0<α<1.

Theorem 18.

Suppose that u(t) is a solution of the fractional integral equation (61). If |L1(t,u)|m(t)|u|3, |L2(t,u,v)|g(t)|u|+|v|, and |M(t,u)|h(t)|u|, where m,g, and h are nonnegative continuous functions on [0,), then we have the following estimate for u(t): (62)|u(t)|exp(130tm(s)ds)×{|C|exp[0tα/Γ(1+α)H~2((sΓ(1+α))1/α)ds]  +1Γ(α)0t(t-τ)α-1H~1(τ)  ×exp[-τα/Γ(1+α)tα/Γ(1+α)H~2×((sΓ(1+α))1/α)ds[-τα/(Γ(1+α))tα/(Γ(1+α))H~2]dτ}1/3,t0, where (63)H~1(t)=2K1/33[g~(t)+0th~(ξ)dξ],H~2(t)=K-2/33[g~(t)+0th~(ξ)dξ],g~(t)=g(t)exp(130sm(τ)dτ),h~(t)=h(t)exp(130tm(τ)dτ).

Proof.

From (61), we have (64)|u3(t)||C|+0t|L1(s,u(s))|ds+|IαL2(t,u(t),0tM(ξ,u(ξ))dξ)||C|+0tm(s)u3(s)ds+1Γ(α)0t(t-s)α-1×|L2(s,u(s),0sM(ξ,u(ξ))dξ)|ds|C|+0tm(s)u3(s)ds+1Γ(α)0t(t-s)α-1×[g(s)|u(s)|+0sh(ξ)|u(ξ)|dξ]ds.

Then a suitable application of Theorem 8 (with p = 3, q = r = 1) to (64) yields the desired estimate (62).

Theorem 19.

For the fractional integral equation (62), if |L1(t,u1)-L1(t,u2)|m(t)|u13-u23||L2(t,u1,v1)-L2(t,u2,v2)|g(t)|u13-u23|+|v1-v2|, |M(t,u1)-M(t,u2)|h(t)|u13-u23|, where m,g, and h are nonnegative continuous functions on [0,), then the solution of (61) depends continuously on the initial value C.

Proof.

Let u~(t) be the solution of the following fractional integral equation (65)u~3(t)=C~+0tL1(s,u~(s))ds+IαL2(t,u~(t),0tM(ξ,u~(ξ))dξ),0<α<1.

A combination of (61) and (65) yields (66)u3(t)-u~3(t)=C-C~+0t[L1(s,u(s))-L1(s,u~(s))]ds+Iα[L2(0tM(ξ,u(ξ))dξ)t,u(t),0tM(ξ,u(ξ))dξ)-L2(0tM(ξ,u~(ξ))dξ)t,u~(t),0tM(ξ,u~(ξ))dξ)]  =C-C~+0t[L1(s,u(s))-L1(s,u~(s))]ds+1Γ(α)0t(t-s)α-1×[L2(0sM(ξ,u(ξ))dξ)s,u(s),×0sM(ξ,u(ξ))dξ)-L2(s,u~(s),0sM(ξ,u~(ξ))dξ)]ds.

Furthermore, we have (67)|u3(t)-u~3(t)||C-C~|+0t|L1(s,u(s))-L1(s,u~(s))|ds+1Γ(α)0t(t-s)α-1×|L2(s,u(s),0sM(ξ,u(ξ))dξ)-L2(s,u~(s),0sM(ξ,u~(ξ))dξ)||C-C~|+0tm(s)|u3(s)-u~3(s)|ds+1Γ(α)0t(t-s)α-1×{|0s|}g(s)|u3(s)-u~3(s)|  +|0s[M(ξ,u(ξ))-M(ξ,u~(ξ))]dξ|0s|}|C-C~|+0tm(s)|u3(s)-u~3(s)|ds+1Γ(α)0t(t-s)α-1×{0sh(ξ)|u3(ξ)-u~3(ξ)|dξg(s)|u3(s)-u~3(s)|+0sh(ξ)|u3(ξ)-u~3(ξ)|dξ}.

Then treating |u3(t)-u~3(t)| as one whole function, applying Theorem 8 to (67), we get that (68)|u3(t)-u~3(t)||C-C~|exp(0tm(s)ds)×{exp[0tα/Γ(1+α)H~2  ×((sΓ(1+α))1/α)ds[0tα/(Γ(1+α))H~2]},t0, where (69)H~2(t)=g~(t)+0th~(ξ)dξ,g~(t)=g(t)exp(0sm(τ)dτ),h~(t)=h(t)exp(0tm(τ)dτ).

So the continuous dependence on the initial value C for the solution of (61) can be obtained from (68).

Remark 20.

From the two examples presented above, one can see that the main results established in Section 2 are mainly used in the qualitative analysis as well as quantitative analysis of the solutions to some certain fractional differential or integral equations, such as the bound estimate, the number of the solutions, and the continuous dependence on the initial value for unknown solutions. On the other hand, by the variational iteration method and the homotopy perturbation method, approximate solutions for some fractional differential equations can be obtained (see the examples in , e.g.), while in few cases, the closed form of these approximate solutions can be obtained. So to this extent, we note that the starting point of establishing the main results in this paper is different from the variational iteration method and the homotopy perturbation method.

Acknowledgments

This work was partially supported by the Natural Science Foundation of Shandong Province (in China) (Grant no. ZR2013AQ009) and the Doctoral Initializing Foundation of Shandong University of Technology (in China) (Grant no. 4041-413030).

Gronwall T. H. Note on the derivatives with respect to a parameter of the solutions of a system of differential equations Annals of Mathematics 1919 20 4 292 296 10.2307/1967124 MR1502565 Bellman R. The stability of solutions of linear differential equations Duke Mathematical Journal 1943 10 643 647 MR0009408 10.1215/S0012-7094-43-01059-2 ZBL0061.18502 Li L. Z. Meng F. W. Ju P. J. Some new integral inequalities and their applications in studying the stability of nonlinear integro-differential equations with time delay Journal of Mathematical Analysis and Applications 2010 377 853 862 Feng Q. Zheng B. Generalized Gronwall-Bellman-type delay dynamic inequalities on time scales and their applications Applied Mathematics and Computation 2012 218 15 7880 7892 10.1016/j.amc.2012.02.006 MR2900122 ZBL1244.26038 Ma Q.-H. Pečarić J. The bounds on the solutions of certain two-dimensional delay dynamic systems on time scales Computers & Mathematics with Applications 2011 61 8 2158 2163 10.1016/j.camwa.2010.09.001 MR2785575 ZBL1219.34118 Wang W. S. A class of retarded nonlinear integral inequalities and its application in nonlinear differential-integral equation Journal of Inequalities and Applications 2012 2012, article 154 Gallo A. Piccirillo A. M. About some new generalizations of Bellman-Bihari results for integro-functional inequalities with discontinuous functions and applications Nonlinear Analysis. Theory, Methods & Applications 2009 71 12 e2276 e2287 10.1016/j.na.2009.05.019 MR2672000 ZBL1239.26014 Ma Q.-H. Estimates on some power nonlinear Volterra-Fredholm type discrete inequalities and their applications Journal of Computational and Applied Mathematics 2010 233 9 2170 2180 10.1016/j.cam.2009.10.002 MR2577756 ZBL1184.26024 Sun Y. G. On retarded integral inequalities and their applications Journal of Mathematical Analysis and Applications 2005 301 2 265 275 10.1016/j.jmaa.2004.07.020 MR2105670 ZBL1057.26022 Lipovan O. Integral inequalities for retarded Volterra equations Journal of Mathematical Analysis and Applications 2006 322 1 349 358 10.1016/j.jmaa.2005.08.097 MR2239243 ZBL1103.26018 Agarwal R. P. Deng S. F. Zhang W. N. Generalization of a retarded Gronwall-like inequality and its applications Applied Mathematics and Computation 2005 165 3 599 612 10.1016/j.amc.2004.04.067 MR2138904 ZBL1078.26010 Pachpatte B. G. Inequalities for Differential and Integral Equations 1998 197 New York, NY, USA Academic Press Mathematics in Science and Engineering MR1487077 Saker S. H. Some nonlinear dynamic inequalities on time scales Mathematical Inequalities & Applications 2011 14 3 633 645 10.7153/mia-14-54 MR2850178 ZBL1222.26032 Feng Q. Meng F. Zheng B. Gronwall-Bellman type nonlinear delay integral inequalities on time scales Journal of Mathematical Analysis and Applications 2011 382 2 772 784 10.1016/j.jmaa.2011.04.077 MR2810830 ZBL1223.26051 Ferreira R. A. C. Torres D. F. M. Generalized retarded integral inequalities Applied Mathematics Letters 2009 22 6 876 881 10.1016/j.aml.2008.08.022 MR2523598 ZBL1171.26328 Feng Q. H. Meng F. W. Zhang Y. M. Generalized Gronwall-Bellman-type discrete inequalities and their applications Journal of Inequalities and Applications 2011 2011, article 47 10.1186/1029-242X-2011-47 MR2837902 ZBL1272.26017 Zheng B. Feng Q. H. Meng F. W. Zhang Y. M. Some new Gronwall-Bellman type nonlinear dynamic inequalities containing integration on infinite intervals on time scales Journal of Inequalities and Applications 2012 2012, article 201 Saker S. H. Some nonlinear dynamic inequalities on time scales and applications Journal of Mathematical Inequalities 2010 4 4 561 579 10.7153/jmi-04-50 MR2777272 ZBL1207.26034 Agarwal R. Bohner M. Peterson A. Inequalities on time scales: a survey Mathematical Inequalities & Applications 2001 4 4 535 557 10.7153/mia-04-48 MR1859660 ZBL1021.34005 Wang W. S. Some retarded nonlinear integral inequalities and their applications in retarded differential equations Journal of Inequalities and Applications 2012 2012, article 75 Pachpatte B. G. Explicit bounds on certain integral inequalities Journal of Mathematical Analysis and Applications 2002 267 1 48 61 10.1006/jmaa.2001.7743 MR1886815 ZBL0996.26008 Li W. N. Some delay integral inequalities on time scales Computers & Mathematics with Applications 2010 59 6 1929 1936 10.1016/j.camwa.2009.11.006 MR2595967 ZBL1189.26046 Kim Y.-H. Gronwall, Bellman and Pachpatte type integral inequalities with applications Nonlinear Analysis. Theory, Methods & Applications 2009 71 12 e2641 e2656 10.1016/j.na.2009.06.009 MR2672035 ZBL1239.26017 Li W. N. Han M. Meng F. W. Some new delay integral inequalities and their applications Journal of Computational and Applied Mathematics 2005 180 1 191 200 10.1016/j.cam.2004.10.011 MR2141494 ZBL1067.26019 Feng Q. H. Meng F. W. Zhang Y. M. Zheng B. Zhou J. C. Some nonlinear delay integral inequalities on time scales arising in the theory of dynamics equations Journal of Inequalities and Applications 2011 2011, article 29 Cheung W.-S. Ren J. L. Discrete non-linear inequalities and applications to boundary value problems Journal of Mathematical Analysis and Applications 2006 319 2 708 724 10.1016/j.jmaa.2005.06.064 MR2227933 ZBL1116.26016 Jumarie G. Modified Riemann-Liouville derivative and fractional Taylor series of nondifferentiable functions further results Computers & Mathematics with Applications 2006 51 9-10 1367 1376 10.1016/j.camwa.2006.02.001 MR2237634 ZBL1137.65001 Jumarie G. Table of some basic fractional calculus formulae derived from a modified Riemann-Liouville derivative for non-differentiable functions Applied Mathematics Letters 2009 22 3 378 385 10.1016/j.aml.2008.06.003 MR2483503 ZBL1171.26305 Zheng B. ( G ' / G ) -expansion method for solving fractional partial differential equations in the theory of mathematical physics Communications in Theoretical Physics 2012 58 5 623 630 10.1088/0253-6102/58/5/02 MR3089091 ZBL1264.35273 Feng Q. H. Exact solutions for fractional differential-difference equations by an extended Riccati Sub-ODE Method Communications in Theoretical Physics 2013 59 521 527 Zhang S. Zhang H.-Q. Fractional sub-equation method and its applications to nonlinear fractional PDEs Physics Letters A 2011 375 7 1069 1073 10.1016/j.physleta.2011.01.029 MR2765013 ZBL1242.35217 Almeida R. Torres D. F. M. Fractional variational calculus for nondifferentiable functions Computers & Mathematics with Applications 2011 61 10 3097 3104 10.1016/j.camwa.2011.03.098 MR2799834 ZBL1222.49026 Khan Y. Wu Q. Faraz N. Yildirim A. Madani M. A new fractional analytical approach via a modified Riemann-Liouville derivative Applied Mathematics Letters 2012 25 10 1340 1346 10.1016/j.aml.2011.11.041 MR2947405 ZBL1251.65101 Faraz N. Khan Y. Jafari H. Yildirim A. Madani M. Fractional variational iteration method via modified RiemannCLiouville derivative Journal of King Saud University—Science 2011 23 413 417 Khan Y. Faraz N. Yildirim A. Wu Q. Fractional variational iteration method for fractional initial-boundary value problems arising in the application of nonlinear science Computers & Mathematics with Applications 2011 62 5 2273 2278 10.1016/j.camwa.2011.07.014 MR2831688 ZBL1231.35288 Merdan M. Analytical approximate solutions of fractionel convection-diffusion equation with modified Riemann-Liouville derivative by means of fractional variational iteration method Iranian Journal of Science and Technology A 2013 37 1 83 92 MR3033306 Guo S. Mei L. Li Y. Fractional variational homotopy perturbation iteration method and its application to a fractional diffusion equation Applied Mathematics and Computation 2013 219 11 5909 5917 10.1016/j.amc.2012.12.003 MR3018433 ZBL06214556 Jiang F. C. Meng F. W. Explicit bounds on some new nonlinear integral inequalities with delay Journal of Computational and Applied Mathematics 2007 205 1 479 486 10.1016/j.cam.2006.05.038 MR2324854 ZBL1135.26015 Ye H. P. Gao J. M. Ding Y. S. A generalized Gronwall inequality and its application to a fractional differential equation Journal of Mathematical Analysis and Applications 2007 328 2 1075 1081 10.1016/j.jmaa.2006.05.061 MR2290034 ZBL1120.26003