- Research
- Open Access
- Published:
Constantin’s inequality for nabla and diamond-alpha derivative
Journal of Inequalities and Applications volume 2015, Article number: 167 (2015)
Abstract
Calculus for dynamic equations on time scales, which offers a unification of discrete and continuous systems, is a recently developed theory. Our aim is to investigate Constantin’s inequality on time scales that is an important tool used in determining some properties of various dynamic equations such as global existence, uniqueness and stability. In this paper, Constantin’s inequality is investigated in particular for nabla and diamond-alpha derivatives.
1 Introduction
To study the boundedness of solutions for some nonautonomous second order linear differential equations, Ou-Iang [1] used a nonlinear integral inequality. This type of integral inequality had been also used to obtain global existence, uniqueness and stability properties of various nonlinear differential equations. Pachpatte [2] gave the generalized Ou-Iang type integral inequality. In 1996, Constantin [3] established the following interesting alternative result for this generalization.
Theorem 1.1
If for some \(k,T>0\), \(u\in C(\mathbb{R}_{0}^{+},\mathbb{R}_{0}^{+})\) satisfies
\(\forall t\in[0,T]\), where \(f,g,h \in C(\mathbb{R}_{0}^{+},\mathbb{R}_{0}^{+})\) and w belongs to the class of continuous nondecreasing functions on \(\mathbb{R}_{0}^{+}\) such that \(w(r)>0\) if \(r>0\) and satisfies \(\int_{1}^{\infty} \frac{ds}{w(s)}=\infty\), then
where \(K(t)=k+\int_{0}^{t} h(s)\,ds\), \(G(r)=\int_{1}^{r} \frac{ds}{s+w(s)}\), \(r>1\). \(G^{-1}\) denotes the inverse function of G.
Applying above Theorem 1.1 and a topological transversality theorem, he showed that, under some suitable assumptions, the integrodifferential equation
has a solution and gave bounds on that solution. Additionally, Yang and Tan [4] gave the generalization of Constantin’s inequality, and they also presented the discrete analogue of this inequality which is stated as follows.
Theorem 1.2
Let the function \(w\in C( \mathbb{R}_{+}, \mathbb{R}_{+})\) be nondecreasing, \(w(r)>0 \) for \(r>0\), \(\phi\in C^{1}( \mathbb{R}_{+}, \mathbb{R}_{+})\) with \(\phi'\) being nonnegative and nondecreasing. \(u,c \in C( \mathbb{N}_{M}, \mathbb{R}_{+})\) with \(c(n)\) nondecreasing. Further, let
be nondecreasing with respect to n for every s fixed. Then the discrete inequality
implies
Here \(G(r)=\int_{r_{0}}^{r} \frac{ds}{s+w(s)}\), \(r\leq r_{0}\), \(1>r_{0}>0\), \(\lim_{x\rightarrow\infty}G(x)=\infty\), \(L(n)=\phi^{-1}[c(n)]+\sum_{s=0}^{n-1} h(n,s)\), \(\mathbb{N}_{M}=\{n\in\mathbb{N}: n\leq M , M \in\mathbb{N}\}\).
2 Some basic definitions related to time scales
Using [5–7] we give the following information. By a time scale, denoted by \(\mathbb{T}\), we mean a nonempty closed subset of \(\mathbb{R}\). The theory of time scales gives a way to unify continuous and discrete analysis.
The set \(\mathbb{T}^{\kappa}\) is defined by \(\mathbb{T}^{\kappa}=\mathbb{T}/ (\rho(\sup\mathbb{T}),\sup\mathbb{T}]\) and the set \(\mathbb{T}_{\kappa}\) is defined by \(\mathbb{T}_{\kappa}=\mathbb{T} / [\inf\mathbb{T},\sigma(\inf \mathbb{T}))\). The forward jump operator \(\sigma:\mathbb{T}\rightarrow \mathbb{T}\) is defined by \(\sigma(t):=\inf(t,\infty)_{\mathbb{T}}\), for \(t\in \mathbb{T}\). The backward jump operator \(\rho:\mathbb{T}\rightarrow \mathbb{T} \) is defined by \(\rho(t):=\sup(-\infty,t)_{\mathbb{T}}\), for \(t\in\mathbb{T}\). The forward graininess function \(\mu:\mathbb{T}\rightarrow\mathbb{R}_{0}^{+}\) is defined by \(\mu(t):=\sigma(t)-t\), for \(t\in\mathbb{T}\). The backward graininess function \(\nu:\mathbb{T}\rightarrow\mathbb{R}_{0}^{+}\) is defined by \(\nu(t):=t-\rho(t)\), for \(t\in\mathbb{T}\). Here it is assumed that \(\inf\emptyset= \sup\mathbb{T}\) and \(\sup\emptyset=\inf\mathbb{T}\).
For a function \(f:\mathbb{T} \rightarrow\mathbb{T}\), we define the Δ-derivative of f at \(t\in\mathbb{T}^{\kappa}\), denoted by \(f^{\Delta}(t)\) for all \(\epsilon>0\). There exists a neighborhood \(U\subset\mathbb{T}\) of \(t\in\mathbb{T}^{\kappa}\) such that
for all \(s\in U\).
For the same function define the ∇-derivative of f at \(t\in \mathbb{T}_{\kappa}\), denoted by \(f^{\nabla}(t)\), for all \(\epsilon>0\). There exists a neighborhood \(V\subset\mathbb{T}\) of \(t\in\mathbb{T}_{\kappa}\) such that
for all \(s\in V\).
We define \({\diamond_{\alpha}}\)-derivative of f at \(t\in\mathbb{T}^{\kappa}_{\kappa}\), denoted by \(f^{\diamond_{\alpha}}(t)\) for all \(\epsilon>0\). There is a neighborhood \(U\subset\mathbb{T}\) such that for any \(s\in U\),
A function \(f:\mathbb{T} \rightarrow\mathbb{R}\) is rd-continuous if it is continuous at right-dense points in \(\mathbb{T}\) and its left-sided limits exist at left-dense points in \(\mathbb{T}\). The class of real rd-continuous functions defined on a time scale \(\mathbb{T}\) is denoted by \(C_{\mathrm{rd}}(\mathbb{T},\mathbb{R})\). If \(f\in C_{\mathrm{rd}}(\mathbb{T},\mathbb{R})\), then there exists a function \(F(t)\) such that \(F^{\Delta }(t) = f (t)\). The delta integral is defined by \(\int_{a}^{b} f(x) \Delta x=F(b)-F(a)\).
Similarly, a function \(g:\mathbb{T} \rightarrow\mathbb {R}\) is ld-continuous if it is continuous at left-dense points in \(\mathbb{T}\) and its right-sided limits exist at right-dense points in \(\mathbb{T}\). The class of real ld-continuous functions defined on a time scale \(\mathbb{T}\) is denoted by \(C_{\mathrm{ld}}(\mathbb{T},\mathbb{R})\). If \(g\in C_{\mathrm{ld}}(\mathbb {T},\mathbb{R})\), then there exists a function \(G(t)\) such that \(G^{\nabla }(t) =g (t)\). The nabla integral is defined by \(\int_{a}^{b} g(x) \nabla x=G(b)-G(a)\).
By [8], if a function \(h(t):\mathbb{T} \rightarrow \mathbb{R}\) is continuous, then it is diamond-alpha integrable, and the fundamental theorem of calculus is not true for \({\diamond_{\alpha }}\)-derivative. We know that
Ferreira [9] generalized Constantin’s inequality involving delta derivatives on an arbitrary time scale.
Theorem 2.1
Assume that \(u\in C_{\mathrm{rd}} ( [ a,b ] _{ \mathbb{T}}, \mathbb{R}_{0}^{+} ) \), \(c\in C_{\mathrm{rd}} ( [ a,b ] _{ \mathbb{T}} , \mathbb{R} ^{+} ) \) is nondecreasing, \(\Phi\in C ( \mathbb{R}_{0}^{+}, \mathbb{R}_{0}^{+} ) \) is a strictly increasing function such that
Let \(f(t,\xi),h(t,\xi)\in C_{\mathrm{rd}} ( [ a,b ] _{ \mathbb{T}}\times [ a,b ] _{ \mathbb{T}^{\kappa}}, \mathbb{R} _{0}^{+} ) \) and \(g(t,\xi)\in C_{\mathrm{rd}} ( [ a,b ] _{ \mathbb{T}^{\kappa}}\times [ a,b ] _{ \mathbb{T}^{{\kappa }^{2}}}, \mathbb{R}_{0}^{+} ) \) be nondecreasing for every fixed ξ. Further, let \(w,\phi,\Psi\in C ( \mathbb {R}_{0}^{+}, \mathbb{R}_{0}^{+} ) \) be nondecreasing such that \(\{ w,\phi,\Psi \} (x)>0\) for every \(x>0\). Define
on \(\mathbb{R} _{0}^{+}\) and assume that the following function
with \(x>c(a)>x_{0}\geq0\) if \(\int_{0}^{x}\frac{ds}{M\circ \Phi ^{-1}(s)}<\infty\) and \(x>c(a)>x_{0}>0\) if \(\int_{0}^{x}\frac{ds}{M\circ \Phi ^{-1}(s)}=\infty\) satisfies \(\lim_{x\rightarrow\infty}F(x)=\infty\).
Also assume that the function
where \(r\geq0\), \(r_{0}\geq0\) if \(\int_{0}^{r}\frac {ds}{w(s)}<\infty\) and \(r>0\), \(r_{0}>0\) if \(\int_{0}^{r}\frac {ds}{w(s)}=\infty\) satisfies \(\lim_{r\rightarrow\infty}P(r)=\infty\).
If \(\Phi^{-1} [ F^{-1}(x) ] \leq x\), for all \(x\geq0\), the inequality
for \(t\in [ a,b ] _{T}\) implies
Here \(K(t)=:F [ c(t) ] + \int_{a}^{t}h(t,s)\Delta s \) and \(G(r)=\int_{r_{0}}^{r}\frac{ds}{w(s)+s}\).
3 Some new results
We try to generalize Constantin’s inequality containing nabla and diamond-alpha derivatives and present the results we have obtained.
First we look for the discrete analogue of Constantin’s inequality involving nabla derivatives.
Theorem 3.1
Let the function \(w\in C( \mathbb{R}_{+}, \mathbb{R}_{+})\) be nondecreasing, \(w(r)>0 \) for \(r>0\), \(\phi\in C^{1}( \mathbb{R}_{+}, \mathbb{R}_{+})\) with \(\phi'\) being nonnegative and nondecreasing. \(u,c \in C( \mathbb{N}_{M}, \mathbb{R}_{+})\) with \(c(n)\) nondecreasing. Further, let
be nondecreasing with respect to n for every s fixed. Then there exist fixed constants \(k,l>0\) such that the discrete inequality
implies
Here \(G(r)=\int_{r_{0}}^{r} \frac{ds}{s+w(s)}\), \(r\leq r_{0}\), \(1>r_{0}>0\), \(\lim_{x\rightarrow\infty}G(x)=\infty\), \(L(n)=\phi^{-1}[c(n)]+k\sum_{s=1}^{n} h(n,s)\), \(\mathbb{N}_{M}=\{n\in N: n\leq M , M \in\mathbb{N}\}\). Assume that \(\mathbb{N}\) starts with 1.
Proof
Fixing an arbitrary positive integer \(m\in[1,M]\).
We denote the set \(J=\{1,2,\ldots,m\}\) and we define a positive function \(z(n)\in K(J, \mathbb{R}_{+})\) such that
Then \(u(n)\leq\phi^{-1}[z(n)]\). Using (1) we get
Since we have finitely many elements in the domain, functions are bounded and, since each of them goes to \(\mathbb{R}_{+}\), it never takes 0. Then
exists.
If we multiply (2) by k and use the mean value theorem, we get
Substituting n with ξ in the last assertion and summing over \(\xi =1,\ldots,n\), we have
Define the right-hand side of the last inequality as \(v(n)\), then we get \(\phi^{-1}[z(n)]\leq v(n)\) for \(n\in J\). Taking the nabla derivative of \(v(n)\), we obtain
We define \(y(n)=v(n)+\sum_{s=1}^{n}g(m,s)w[v(s)]\) and take the nabla derivative of \(y(n)\), then we get
For inequality (3), substituting n with s and summing over \(s=1,2,\ldots, n\), we obtain
Let \(l=\max_{n\in[0,m]}\frac{\sum_{s=1}^{n}\frac{\nabla y(s)}{y(s-1)+w(y(s-1))}}{\sum_{s=1}^{n}\frac{\nabla y(s)}{y(s)+w(y(s))}}\). Multiply (4) by l and by the mean value theorem, we get
Since \(\nabla v(n)\leq kf(m,n)y(n)\), we obtain
Substituting n with ξ and summing over \(\xi=1,2,\ldots,n\) and using \(u(n)\leq\phi^{-1}[z(n)]\leq v(n)\), we get
Hence we get the desired result. □
Example 3.1
Let
Here \(n\in[1,M]\). The unique positive solution for equation (5) can be obtained by successive substitution. For instance, by letting \(n=1,2\) we obtain
By using solutions of quadratic equations, we can find \(u(0),u(1),u(2),\ldots ,u(M)\) successively. If we use the theorem above, then the bound for u will be \(u(n)\leq \sqrt{L}+ k\sum_{s=1}^{n}H(n-s)\). With the help of the proof of Theorem 3.1 here
Here \(f(n,s)=g(n,s)=0\), \(h(n,s)=H(n-s)\), \(c(n)=L>0\), L is constant, \(u,H\in K( \mathbb{N}_{M}, \mathbb{R}_{+})\), P is nondecreasing, and \(\phi (x)=x^{2}\).
If we choose \(H(n,s)\), L good enough, k does not exceed a number c when M tends to infinity. For instance, if \(H(n,s)=1\), \(L=3\), then k does not exceed 3 when M tends to infinity. Therefore, for the equality \(u^{2}(n)=3+2\sum_{s=1}^{n}u(s)\), by using the theorem above, we have the bound for
Lemma 3.1
Let \(f: \mathbb{R}\rightarrow\mathbb{R}\) be continuously differentiable and \(g: \mathbb{T}\rightarrow\mathbb{R}\) be nabla differentiable, then \(f \circ g\) is nabla differentiable and the formula is given by
Proof
Apply an ordinary substitution rule from calculus and get
Let \(t\in \mathbb{T}_{\kappa}\) and \(\epsilon>0\) be given. Since g is nabla differentiable at t, there exists a neighborhood \(U_{1}\) of t such that
where \(\epsilon^{\ast}=\frac{\epsilon}{1+2\int_{0}^{1} \vert f^{\prime}(hg(s)-(1-h)g(\rho(t)))\vert \, dh}\).
By assumption we have
where \(\forall\delta>0\) \(\exists U_{2}\) is a neighborhood of t such that \(\vert g(s)-g(t)\vert <\delta\), \(\forall s\in U_{2}\). Here let us define a neighborhood U of t such that \(U=U_{1}\cap U_{2}\). Then we have
where \(\alpha=hg(s)+(1-h)g(\rho(t))\) and \(\beta=hg(t)+(1-h)g(\rho(t))\). Hence \(f \circ g\) is nabla differentiable and its derivative is as claimed above. □
This lemma first occurred in the article of Atici et al. [10] as
With a counter example, we can show that their version of the formula is not true.
Example 3.2
Let \(g(x): \mathbb{Z}\rightarrow \mathbb{R}\) such that \(g(n)=\frac {1}{n}\), \(f(x): \mathbb{R}\rightarrow \mathbb{R}\) such that \(f(x)=x^{2}\). Therefore the first derivative of \(f(x)\) is continuous. If we apply the above formula to find the nabla derivative of the function \((f \circ g)(t)\), we get
Lemma 3.2
Let \(a,b\in \mathbb{T}\), consider the time scale \([ a,b ] _{ \mathbb{T}}\) and a function \(r\in C_{\mathrm{ld}}^{1} ( [ a,b ] _{ \mathbb{T}}, \mathbb{R} ) \) with \(r^{\nabla}(t)\geq 0\). Suppose that a function \(g\in C ( \mathbb{R} _{0}^{+}, \mathbb{R} _{0}^{+} )\) is positive and nondecreasing on \(\mathbb{R}\). Then, for each \(t\in [ a,b ] _{ \mathbb{T}}\), we have
where \(G(x)=\int_{x_{0}}^{x}\frac{ds}{g(s)}\), where \(x\geq0\), \(x_{0}\geq0\) if \(\int_{0}^{x}\frac{ds}{g(s)}<\infty\) and \(x>0\), \(x_{0}>0\) if \(\int_{0}^{x}\frac{ds}{g(s)}=\infty\).
Proof
Since g is positive and nondecreasing on \(( 0,\infty )\), we have
By using Lemma 3.1, we get
and
Hence we get the desired result. □
The delta derivative version of Lemma 3.1 was proved by Ferreira and Torres [11]. By using these results, we proved the following theorem.
Theorem 3.2
Assume that \(u\in C_{\mathrm{ld}} ( [ a,b ] _{ \mathbb{T}}, \mathbb{R}_{0}^{+} ) \), \(c\in C_{\mathrm{ld}} ( [ a,b ] _{ \mathbb{T}} , \mathbb{R} ^{+} ) \) is nondecreasing, \(\Phi\in C ( \mathbb{R}_{0}^{+}, \mathbb{R}_{0}^{+} ) \) is a strictly increasing function such that
Let \(f(t,\xi),h(t,\xi)\in C_{\mathrm{ld}} ( [ a,b ] _{ \mathbb{T}}\times [ a,b ] _{ \mathbb{T}_{\kappa}}, \mathbb{R} _{0}^{+} ) \) and \(g(t,\xi)\in C_{\mathrm{ld}} ( [ a,b ] _{ \mathbb{T}_{\kappa}}\times [ a,b ] _{ \mathbb{T}_{{\kappa }^{2}}}, \mathbb{R}_{0}^{+} ) \) be nondecreasing in t for every fixed ξ. Further, let \(w,\phi,\Psi\in C ( \mathbb {R}_{0}^{+}, \mathbb{R}_{0}^{+} ) \) be nondecreasing such that \(\{ w,\phi,\Psi \} (x)>0\) for every \(x>0\). Assume that the following function
with \(x>c(a)>x_{0}\geq0\) if \(\int_{0}^{x}\frac{ds}{M\circ \Phi ^{-1}(s)}<\infty\) and \(x>c(a)>x_{0}>0\) if \(\int_{0}^{x}\frac{ds}{M\circ \Phi ^{-1}(s)}=\infty\); \(\lim_{x\rightarrow\infty}F(x)=\infty\), where \(M(x)=\max \{ \phi(x),\Psi(x) \}\) on \(\mathbb{R} _{0}^{+}\).
Also assume that the function
where \(r\geq0\), \(r_{0}\geq0\) if \(\int_{0}^{r}\frac {ds}{w(s)}<\infty\) and \(r>0\), \(r_{0}>0\) if \(\int_{0}^{r}\frac {ds}{w(s)}=\infty\); \(\lim_{r\rightarrow\infty}P(r)=\infty\).
Then there exist fixed constants \(\alpha,\beta>0\) such that \(\Phi ^{-1} [ F^{-1}(x) ] \leq x\) for all \(x\geq0\), then
for \(t\in [ a,b ] _{\mathbb{T}}\) implies
Here \(K(t):=F [ c(t) ] +\alpha\int_{a}^{t}h(t,s)\nabla s \) and \(G(r)=\int_{r_{0}}^{r}\frac{ds}{w(s)+s}\).
Proof
If \(t=a\), then, obviously, theorem holds. Let us fix an arbitrary number \(t_{0}\in \left.(a,b]\right._{ \mathbb{T}}\), we define \(z(t)\) on \([ a,t_{0} ] _{ \mathbb{T}}\) such that
Since \(\Phi(u(t))\leq z(t)\), then \(u(t)\leq\Phi^{-1}(z(t))\), so we have
Define \(\check{z}(t)\) on \([a,b]_{ \mathbb{T}}\) such that
It is obvious that \(z(t)\leq\check{z}(t)\) on \([ a,t_{0} ] _{ \mathbb{T}}\). Using \(\check{z}(t)\) we define α such that
From the definition of the functions \(\check{z}(t)\), \(z(t)\), M, \(\Phi ^{-1}\) and by Theorem 1.65 from [12] α exists.
Multiply (6) by α and use Lemma 3.2 to get
where
By differentiating the function \(y(t)\) and using \(\Phi ^{-1}(F^{-1}(y(t)))\leq y(t)\), we obtain \(y^{\nabla}(t)\leq\alpha f(t_{0},t) [ y(t)+\int_{a}^{t}g(t_{0},\xi)w [ y(\xi) ] \nabla\xi ]\). Let us define \(\Omega(t)=y(t)+\int_{a}^{t}g(t_{0},\xi)w [ y(\xi) ]\nabla\xi\), so \(\Omega(t)\geq y(t)\). Then
Let us define \(\hat{y}(t)\) and \(\hat{\Omega} (t)\) on \([ a,b ]_{\mathbb{T}} \) such that
It is obvious that \(y(t)\leq\hat{y}(t)\) and \(\Omega(t)\leq\hat{\Omega }(t)\) on \([ a,t_{0} ] _{ \mathbb{T}}\). Using \(\hat{\Omega}(t)\) we define β such that
Again from the definition of the functions \(\Omega(t)\), \(y(t)\), \(\hat {y}(t)\), \(\hat{\Omega}(t)\), \(w(\hat{\Omega}(t))\) and by Theorem 1.65 from [12] β exists, where
If we multiply (7) by β and use Lemma 3.2, we have
Since \(y^{\nabla}(t)\leq\alpha f(t_{0},t)\Omega(t)\), then \(y(t)\leq K(t_{0})+\alpha\int_{a}^{t}f(t_{0},s)\Omega(s)\nabla s\). Also using the information \(u(t)\leq\Phi^{-1}(z(t))\leq\Phi ^{-1}(F^{-1}(y(t)))\), we get
Since \(t_{0}\) is arbitrary, we can set \(t=t_{0}\) in the above inequality, and we get the desired result. □
Remark 3.1
The function \(G(r)\) defined above satisfies \(\lim_{r\rightarrow\infty }G(r)=\infty\) by Constantin [13]. This was discussed in [14–16].
Example 3.3
If we take \(\phi(x)=\psi(x)=\Phi'(x)\) with \(\Phi'\) being nondecreasing, then \(M(x)=\Phi'(x)\), and this implies
Choose \(x_{0}=\Phi(0)\geq0\). Then \(\Phi^{-1}(\Phi(0))=0\), hence \(F(x)=\Phi^{-1}(x)\). For the particular case \(\mathbb{T}= \mathbb{Z}\), an application of Theorem 3.2 gives Theorem 3.1. For the particular case \(\mathbb{T}= \mathbb{R}\) and \(\alpha= \beta= 1\), an application of Theorem 3.2 gives the generalization of Constantin’s inequality which is done by Yang and Tan [4].
Example 3.4
Let us take \(\mathbb{T}=h \mathbb{Z}\) such that \(x,t \in[h,Mh]\), \(h(x,t)=H(x-t)\), \(g(x,t)=f(t,x)=0\), \(\Phi(x)=x^{2}\), \(\phi(x)=0\), \(\psi (x)=\frac{x}{2}\) defined for \(x\geq0\) and \(c(n)=L\), \(L\geq0 \) is constant. Then
If we set \(x_{0} =0\), we obtain
Thus, \(\lim_{x\rightarrow\infty}F(x)=\infty\). \(F^{-1}(x)={ (\frac {x}{4} )}^{2}\) and \(\Phi^{-1}(F^{-1}(x))= (\frac{x}{4} )\leq x\), \(\forall x\geq0\).
If we get the equality, we have
By letting \(n=1,2\), we get
If we apply Theorem 3.2, we find an upper bound for \(u(hn)\) as
Here
Lemma 3.3
Let \(\mathbb{T}\) be a regulated time scale, \(a,b \in\mathbb{T}\), and consider the time scale \([a,b]_{\mathbb{T}}\) such that \(\sigma(a)=a\). Let \(r \in C^{1}([a,b]_{\mathbb{T}}, \mathbb{R})\) with \(r^{\nabla }(t),r^{\Delta}(t)\geq0\). Suppose that a function \(g\in C(\mathbb {R}_{0}^{+},\mathbb{R}_{0}^{+})\) is positive and nondecreasing on \(\mathbb{R}\).
Define \(G(x)=\int_{x_{0}}^{x} \frac{ds}{g(s)}\), where \(x\geq0\), \(x_{0} \geq 0\) if \(G(x)=\int_{0}^{x} \frac{ds}{g(s)} < \infty\) and \(x> 0\), \(x_{0} >0\) if \(G(x)=\int_{0}^{x} \frac{ds}{g(s)}=\infty\). Then, for each \(t\in[a,b]_{\mathbb{T}}\), we have
Proof
By assumption we have
and
For \(\alpha\in(0,1)\), we get
and
By using [8, 17] and the assumption on a, we have the desired result
□
Theorem 3.3
\(u\in C([a,b]_{\mathbb{T}}, \mathbb{R}_{+})\) satisfies for some \(k>0\) such that
Here our time scale is regulated and \(\sigma(a)=a\).
\(\forall t\in[a,b]_{\mathbb{T}}\), \(f,g,h\in C([a,b]_{\mathbb{T}}, \mathbb{R}^{0}_{+})\), \(w(t)\in C( \mathbb{R}^{0}_{+}, \mathbb{R}^{0}_{+})\) and \(w(t)\) is nondecreasing, there exist fixed constants \(m,c>0\) such that
where \(E(r)=\int_{1}^{r} \frac{ds}{w(s)+s}\), \(r>0\).
Proof
If we take \(a=t\), then the inequality obviously holds true. Let \(t_{0}\in (a,b]\) and define \(z(t)\) in \([a,t_{0}]\) such that
Therefore, for \(u(t)\leq\sqrt{z(t)}\), \(t\in[a,t_{0}]_{\mathbb{T}}\). Since \(z^{\Delta}(t), z^{\nabla}(t)\geq0\), if we use [8], we obtain
Then \(z(t)\) is nondecreasing. In other words, \(z(\sigma(t))\geq z(t) \geq z(\rho(t))\), so we have
Functions that compose \(z(t)\) are from \(C([a,b]_{\mathbb{T}}, \mathbb {R}^{0}_{+})\), then they are regulated on \([a,b]_{\mathbb{T}}\), then \(z(t)\) is bounded and \(z(t)\) never takes zero. Therefore there exists c such that \(c=\max_{t\in[a,t_{0}]_{\mathbb{T}}}\frac{\sqrt{z(\sigma(t))}}{\sqrt {z(\rho(t))}}\).
Multiply (8) by c and obtain
Now we use Lemma 3.3 and we get
Let us say the right-hand side of the above inequality is \(V(t)\), then \(\sqrt{z(t)}\leq V(t)\). So we get
Here \(V^{\diamond_{\alpha}}(t)\geq0\). Similarly, if we take the delta and nabla derivative of \(V(t)\), we also see that \(V^{\Delta}(t), V^{\nabla}(t)\geq0\).
Therefore \(V(\sigma^{2}(t))\geq V(\sigma(t))\geq V(t)\geq V(\rho (t))\geq V(\rho^{2}(t))\). Then
Let us take \(V(\sigma^{2}(t)) +\int_{a}^{\sigma^{2}(t)} g(\tau)w( V(\tau ))\diamond_{\alpha}\tau=\Omega(t)\). Taking the nabla derivative of \(\Omega (t)\) and using (9), we have
\(\Omega^{\diamond_{\alpha}}(t)\geq0\). Similarly, \(\Omega^{\Delta}(t)\), \(\Omega^{\nabla}(t)\geq0\). It is obvious that \(V(\sigma^{2}(t))\leq \Omega(t)\)and since w is nondecreasing, we also have \(w(\Omega(\sigma (t)))\geq w(\Omega(t))\geq w(\Omega(\rho(t)))\). Therefore \(w(\Omega (\sigma(t)))\geq w( V(\sigma^{3}(t)))\),
By using the same information while finding the existence of c, there exists m such that \(m=\max_{t\in[a,t_{0}]_{\mathbb{T}}}\frac{\Omega (t)+w(\Omega(\sigma(t)))}{\Omega(\rho(t))+w(\Omega(\rho(t)))}\). If we multiply (10) by m, we have
Now we use Lemma 3.3 and get
Then we have
Using [8] we obtain
Since \(u(t)\leq\sqrt{z(t)}\leq V(t)\), we get the desired result. □
Example 3.5
If we choose \(\alpha=1\), \(\mathbb{T}=\mathbb{R}\), \(m=c=1/2\), then Theorem 3.3 becomes Theorem 1.1.
Example 3.6
Let \(\mathbb{T}=\{0\}\cup\{\frac{1}{n+1}:n\in\mathbb{N}\}\cup \mathbb{N}\). Here we assume that \(\mathbb{N}\) starts with 1. It is obvious that our time scales is regulated and \(0=\sigma(0)\). Let us choose \(M\in\mathbb{N}\) and investigate the following equality on the time scales \([0,M]_{\mathbb{T}}\):
\(h(t)=0\) if \(t\in\{0\}\cup\{\frac{1}{n+1}:n\in\mathbb{N}\}\) and \(h(t)=P(t)\) if \(t\in \mathbb{N}\). Therefore \(u(t)=k\) if \(t\in\{0\} \cup\{\frac{1}{n+1}:n\in\mathbb{N}\}\), \(u^{2}(n)=k^{2}+P(1)u(1)+P(2)u(2)+\cdots+P(n-1)u(n-1)+(1-\alpha)P(n)u(n)\) if \(n \in[0,M]\). If we apply Theorem 3.3, we get the bound for \(u(t)\) such that
where
If we take \(\alpha=\frac{1}{2}\), \(k^{2}=2\), then for the equation \(u^{2}(t)=2+2\int_{0}^{t} h(s)u(s) \diamond_{\alpha}s\), where \(h(s)=0\), if \(t\in\{0\}\cup\{\frac{1}{n+1}:n\in\mathbb{N}\}\) and \(h(s)=1\) if \(t\in\mathbb{N}\), then \(c=\frac{1}{2}\) and \(u(t)\) is bound with \(u(t)\leq\sqrt{2}\) if \(t\in\{0\}\cup\{\frac{1}{n+1}:n\in\mathbb {N}\}\) and \(u(n)\leq4\sqrt{2}+n-\frac{1}{2} \) if \(n\in\mathbb{N}\).
References
Ou-Iang, L: The boundedness of solutions of linear differential equations \(y''+A(t)y = 0\). Shuxue Jinzhan 3, 409-418 (1957)
Pachpatte, BG: Some new inequalities related to certain inequalities in the theory of differential equations. J. Math. Anal. Appl. 189(1), 128-144 (1995)
Constantin, A: Topological transversality: application to an integrodifferential equation. J. Math. Anal. Appl. 197(3), 855-863 (1996)
Yang, EH, Tan, MC: A generalization of Constantin’s integral inequality and its discrete analogue. J. Inequal. Pure Appl. Math. 8(2), Article 57 (2007)
Agarwal, RP, Bohner, M: Basic calculus on time scales and some of its applications. Results Math. 35(1-2), 3-22 (1999)
Anderson, DR: Taylor polynomials for nabla dynamic equations on time scales. Panam. Math. J. 12(4), 17-27 (2002)
Bohner, M, Peterson, A: Advances in Dynamic Equations on Time Scales. Birkhäuser Boston, Boston (2003)
Sheng, Q, Fadag, M, Henderson, J, Davis, JM: An exploration of combined dynamic derivatives on time scales and their applications. Nonlinear Anal., Real World Appl. 7, 395-413 (2006)
Ferreira, RAC: Constantin integral inequalities on time scales. Arch. Math. 93, 153-163 (2009)
Atici, FM, Biles, CD, Lebedinsky, A: An application of times scales to economics. Math. Comput. Model. 43(7-8), 718-726 (2006)
Ferreira, RAC, Torres, DFM: Generalizations of Gronwall Bihari inequalities on time scales. J. Differ. Equ. Appl. 15(6), 529-539 (2009)
Bohner, M, Peterson, A: Dynamic Equations on Times Scales: An Introduction with Applications. Birkhäuser, Basel (2001)
Constantin, A: Solutions globales d’équations différentielles perturbées. C. R. Acad. Sci. Paris 320, 1319-1322 (1995)
Bernfeld, SR: The extendability of solutions of perturbed scalar differential equations. Pac. J. Math. 42, 277-288 (1972)
Constantin, A: Global existence of solutions for perturbed differential equations. Ann. Mat. Pura Appl. (4) 168, 237-299 (1995)
Hara, T, Yoneyama, T, Sugie, J: Continuability of solutions of perturbed differential equations. Nonlinear Anal. 8(8), 963-975 (1984)
Atici, FM, Guseinov, GS: On Green’s functions and positive solutions for boundary value problems on time scales. J. Comput. Appl. Math. 141(1-2), 75-99 (2002)
Author information
Authors and Affiliations
Corresponding author
Additional information
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
BK gave the first idea that is studying about the nabla and diamond-alpha version of Constantin’s inequality, and she also made the literature review. AFG designed the pattern of the study and all findings were controlled by her in each step with BK. NNP participated in the contribution in the mathematical analysis part. AFG and NNP drafted the manuscript together. BK gave the final approval. All authors read and approved the final manuscript.
Rights and permissions
Open Access 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.
About this article
Cite this article
Güvenilir, A.F., Kaymakçalan, B. & Pelen, N.N. Constantin’s inequality for nabla and diamond-alpha derivative. J Inequal Appl 2015, 167 (2015). https://doi.org/10.1186/s13660-015-0681-9
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s13660-015-0681-9
Keywords
- Constantin’s inequality
- nabla differentiation
- nabla integration
- diamond-alpha differentiation
- diamond-alpha integration