数学杂志  2017, Vol. 37 Issue (5): 1007-1012   PDF    
扩展功能
加入收藏夹
复制引文信息
加入引用管理器
Email Alert
RSS
本文作者相关文章
ZOU Min
CHEN Rong-san
LIU An-ping
OSCILLATION OF NONLINEAR IMPULSIVE DELAY HYPERBOLIC EQUATION WITH FUNCTIONAL ARGUMENTS VIA RICCATI METHOD
ZOU Min, CHEN Rong-san, LIU An-ping    
School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China
Abstract: In this paper, we mainly deal with the oscillation problems of nonlinear impulsive hyperbolic equation with functional arguments. By using integral averaging method and a generalized Riccati technique, a sufficient condition for oscillation of the solutions of nonlinear impulsive hyperbolic equation with functional arguments is obtained. We can make better use of some existing conclusions about oscillation of the solutions of impulsive ordinary differential equations with delay.
Key words: oscillation     impulsive     delay     hyperbolic equation     Riccati inequality    
里卡蒂方法研究带泛函参数的非线性脉冲时滞双曲方程的振动性
邹敏, 陈荣三, 刘安平    
中国地质大学(武汉)数学与物理学院, 湖北武汉 430074
摘要:本文研究了带泛函参数的非线性脉冲时滞双曲方程的振动性问题.利用积分平均法和里卡蒂方法得到了这类方程解的振动性的一个充分条件,对非线性时滞双曲方程解的震动性进行了推广,能更好地利用一些现有的脉冲时滞常微分方程解的振动性的结论.
关键词振动    脉冲    时滞    双曲方程    Riccati不等式    
1 Introduction

The theories of nonlinear partial functional differential equations are applied in many fields. In recent years the research of oscillation to impulsive partial differential systems caught more and more attention. In this paper, we study the oscillation properties of the solutions to impulsive delay hyperbolic equation

$ \begin{eqnarray} \frac{\partial}{\partial t}((r(t)\frac{\partial}{\partial t}u(x, t))&=&a(t)h(u(x, t))\triangle u(x, t) -\sum\limits_{i=1}^{n}b_{i}(t)h_{i}(u(x, \tau_{i}(t)))\triangle u(x, \tau_{i}(t))\nonumber\\ &&+\sum\limits_{j=1}^{m}q_{j}(x, t)\varphi_{j}(u(x, t)), \ t\neq t_{k}, (x, t)\in\Omega\equiv G\times (0, +\infty), \end{eqnarray} $ (1.1)
$ u(x, t_{k}^{+})-u(x, t_{k}^{-})=\alpha_{k}u(x, t_{k}), \ \ \ t=t_{k}, k=1, 2, \cdots, $ (1.2)
$ u_{t}(x, t_{k}^{+})-u_{t}(x, t_{k}^{-})=\beta_{k}u_{t}(x, t_{k}), \ \ \ t=t_{k}, k=1, 2, \cdots. $ (1.3)

The following is the boundary condition

$ \frac{\partial u(x, t)}{\partial n}+\gamma(x, t)u(x, t)=0, ~~ (x, t)\in \partial G\times[0, +\infty). $ (1.4)

where $G$ is a bounded domain of $R^{n}$ with the smooth boundary $\partial G$ and $n$ is the unit exterior normal vector to $\partial G$.

Following are the basic hypothesis

(H1) $r(t)\in C([0, +\infty);(0, +\infty))$, $a(t), b_{i}(t)\in PC([0, +\infty);[0, +\infty)), \ i=1, 2, \cdots, n$. $\gamma(x, t)\in C(R_{+}\times \partial G, R_{+})$. $q_{j}(x, t)\in C(\bar{\Omega};[0, +\infty)), \ j=1, 2, \cdots, m$, where PC denotes the class of functions which are piecewise continuous in $t$ with discontinuities of the first kind only at $t=t_{k}, k=1, 2, \cdots$.

(H2) $\tau_{i}(t)\in C([0, +\infty);R), \ \lim\limits_{t\rightarrow+\infty}\tau_{i}(t)=+\infty, \ i=1, 2, \cdots, n$.

(H3) $h(u), h_{i}(u)\in C(R, R)$, $uh(u)\geq0$, $uh'(u)\geq0$, $uh'_{i}(u)\geq0$, $i=1, 2, \cdots, n$; $\varphi_{j}(s)\in C(R, R)$, $\frac{\varphi_j(s)}{s}\geq C_{j}={\rm const.}>0$ for $s\neq0$. $\alpha_{k}, \beta_{k}={\rm const.}>-1$, $0<t_{1}<t_{2}<\cdots<t_{k}<\cdots$, $\lim\limits_{t\rightarrow+\infty}t_{k}=+\infty$, $k=1, 2, \cdots$.

We introduce the notations $U(t)=\displaystyle\int_{G}u(x, t)dx$ and $q_{j}(t)= \min\limits_{x\in \bar{G}} q_{j}(x, t)$.

Definition 1.1   The solution $u(x, t)$ of the problems (1.1)-(1.4) is said to be nonoscillatory in domain $\Omega$ if it is either eventually positive or eventually negative. Otherwise, it is called oscillatory.

Definition 1.2   We say that functions $H_{i}, \ i=1, 2, $ belong to a function class $H$, if $H_{i}\in C(D;[0, +\infty)), \ i=1, 2$, satisfy

1. $H_{i}(t, s)=0, \ i=1, 2\ {\rm for}\ t=s, $

2. $ H_{i}(t, s)>0, \ i=1, 2\ {\rm for}\ t>s, $

where $D=\{(t, s):0<s\leq t<+\infty\}$. Moreover, the partial derivatives $\partial H_{1}/\partial s$ and $\partial H_{2}/\partial s$ exist on $D$ such that

$ \frac{\partial H_{1}}{\partial s}(t, s)=h_{1}(t, s)H_{1}(t, s)\ {\rm and}\ \frac{\partial H_{2}}{\partial s}(t, s)=-h_{2}(t, s)H_{2}(t, s), $

where $h_{1}, h_{2}\in C_{loc}(D;\mathbb{R}).$

In recent years, there was much research activity concerning the oscillation theory of nonlinear hyperbolic equations with functional arguments by employing Riccati technique. Riccati techniques were used to obtain various oscillation results. Recently, Shoukaku and Yoshida [2] derived oscillation criteria by using oscillation criteria of Riccati inequality. In this work, we study the hyperbolic equation with impulsive.

2 Main Results

Theorem 2.1   If for each $T\geq0$, there exist $(H_{1}, H_{2})\in H$ and $a, b, c\in \mathbb{R}$ such that $T\leq a<c<b$ and

$ \begin{eqnarray}\begin{aligned} &\frac{1}{H_{1}(c, a)}\int_{a}^{c}H_{1}(s, a)\prod\limits_{t_{1}\leq t_{k}< s}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}(C_{l}q_{l}(s) -\frac{1}{4}r(s)\lambda_{1}^{2}(s, a))\psi(s)ds\\ +&\frac{1}{H_{2}(b, c)}\int_{c}^{b}H_{2}(b, s)\prod\limits_{t_{1}\leq t_{k}< s}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1} (C_{l}q_{l}(s) -\frac{1}{4}r(s)\lambda_{2}^{2}(b, s))\psi(s)ds>0, \end{aligned} \end{eqnarray} $ (2.1)

then every solution of the problems (1.1)-(1.4) oscillates in $\Omega$, where

$ \psi(t)\in C^{1}((T_{0}, +\infty);(0, +\infty)) $

for some $t_{1}>0$ and

$ \lambda_{1}(s, t)=\frac{\psi'(s)}{\psi(s)}+h_{1}(s, t), ~~\lambda_{2}(t, s)=\frac{\psi'(s)}{\psi(s)}-h_{2}(t, s). $

Proof  Suppose to the contrary that there is a nonoscillatory solution $u(x, t)$ of the problems (1.1)-(1.4). Without loss of generality we may assume that $u(x, t)>0$ in $G\times[t_{0}, +\infty)$ for some $t_{0}>0$ because the case where $u(x, t)<0$ can be treated similarly. Since (H2) holds, we see that $u(x, \tau_{i}(t))>0$ $(i=1, 2, \cdots n)$ in $G\times[t_{1}, +\infty)$ for some $t_{1}\geq t_{0}$.

(1) For $t\geq t_{1}, \ t\neq t_{k}, \ k=1, 2, \cdots$, integrating (1) with respect to $x$ over $G$, we obtain

$ \begin{eqnarray*} \frac{d}{dt}(r(t)\int_{G}\frac{\partial}{\partial t}u(x, t)dx)&= & a(t)\int_{G}h(u(x, t))\triangle u(x, t)dx-\sum\limits_{j=1}^{m}\int_{G}q_{j}(x, t)\varphi_{j}(u(x, t))\\ &&+\sum\limits_{i=1}^{n}b_{i}(t)\int_{G}h_{i}(u(x, \tau_{i}(t))) \triangle u(x, \tau_{i}(t))dx. \end{eqnarray*} $

By Green's formula and the boundary condition, we have

$ \begin{eqnarray*} &&\int_{G}h(u(x, t))\Delta u(x, t)dx =\int_{\partial G}h(u(x, t))\frac{\partial u(x, t)}{\partial n}ds-\int_{G}h'(u)|grad u|^{2}dx\\ &=&-\int_{G}\gamma(x, t)u(x, t)h(u(x, t))ds-\int_{G}h'(u)|grad u|^{2}dx\leq0, \\ &&\int_{G}h_{i}(u(x, \tau_{i}(t)))\Delta u(x, \tau_{i}(t))dx\leq0. \end{eqnarray*} $

For condition (H3) we can easily obtain

$ \int_{G}q_{j}(x, t)\varphi_{j}(u(x, t))dx\geq C_{j}q_{j}(t)\int_{G}u(x, t)dx, $

then $U(t)>0$, and it follows that

$ (r(t)U'(t))'+\sum\limits_{i=1}^{m}C_{i}(t)q_{i}(t)U(t)\leq0. $

For some $l\in\{1, 2, \cdots, m\}$, we can get

$ (r(t)U'(t))'+C_{l}q_{l}(t)U(t)\leq0, \ \ t\geq t_{1}, t\neq t_{k}. $

(2) For $t=t_{k}$, $k=1, 2, \cdots$. From (1.2)-(1.3), we have that

$ \begin{eqnarray*} &&\int_{G}u(x, t_{k}^{+})dx-\int_{G}u(x, t_{k}^{-})dx=\alpha_{k}\int_{G}u(x, t_{k}), \\ &&\int_{G}u_{t}(x, t_{k}^{+})dx-\int_{G}u_{t}(x, t_{k}^{-})dx=\beta_{k}\int_{G}u_{t}(x, t_{k}), \end{eqnarray*} $

that is

$ U(t_{k}^{+}) =(1+\alpha_{k})U(t_{k}), \ U'(t_{k}^{+}) =(1+\beta_{k})U'(t_{k}). $

Thus we obtain that the functions $U(t)$ is a eventually positive solution of the impulsive differential inequality

$ \begin{eqnarray}\begin{aligned} &(r(t)y'(t))'+ C_{l}q_{l}(t)y(t)\leq 0, \\ &y(t_{k}^{+}) =(1+\alpha_{k})y(t_{k}), \\ &y'(t_{k}^{+}) =(1+\beta_{k})y'(t_{k}).\end{aligned} \end{eqnarray} $ (2.2)

Set $w(t)=\frac{r(t)U'(t)}{U(t)}$ for $t\geq t_{1}$. From (2.2), we obtain that

$ w'(t)+\frac{1}{r(t)}w^{2}(t)\leq-C_{l}q_{l}(t), ~~w(t_{k}^{+})=\frac{1+\beta_{k}}{1+\alpha_{k}}w(t_{k}). $

Define $v(t)=(\prod\limits_{t_{1}\leq t_{k}<t}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1})w(t)$. In fact, $w(t)$ is continuous on each interval $(t_{k}, t_{k+1}]$, and in view of $w(t_{k}^{+})=(\frac{1+\beta_{k}}{1+\alpha_{k}})w(t_{k})$, it follows that for $t\geq t_{1}$,

$ v(t_{k}^{+})=\prod\limits_{t_{1}\leq t_{j}\leq t_{k}}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}w(t_{k}^{+})=\prod\limits_{t_{1}\leq t_{j}< t_{k}}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}w(t_{k})=v(t_{k}), $

and for all $t\geq t_{1}$,

$ v(t_{k}^{-})=\prod\limits_{t_{1}\leq t_{j}\leq t_{k-1}}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}w(t_{k}^{-})=\prod\limits_{t_{1}\leq t_{j}< t_{k}}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}w(t_{k})=v(t_{k}), $

which implies that $v(t)$ is continuous on $[t_{1}, +\infty)$,

$ \begin{eqnarray*} & &v'(t)+(\prod\limits_{t_{1}\leq t_{k}< t}\frac{1+\beta_{k}}{1+\alpha_{k}})\frac{1}{r(t)}v^{2}(t)+(\prod\limits_{t_{1}\leq t_{k}<t}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1})C_{l}q_{l}(t)\\ &=&(\prod\limits_{t_{1}\leq t_{k}< t}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1})w'(t)+\frac{1}{r(t)}(\prod\limits_{t_{1}\leq t_{k}< t}\frac{1+\beta_{k}}{1+\alpha_{k}})(\prod\limits_{t_{1}\leq t_{k}< t}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1})^{2}w^{2}(t)\\ &&+(\prod\limits_{t_{1}\leq t_{k}<t}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1})C_{l}q_{l}(t)\\ &=&\prod\limits_{t_{1}\leq t_{k}< t}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}[w'(t)+\frac{1}{r(t)}w^{2}(t)+C_{l}q_{l}(t)]\leq0. \end{eqnarray*} $

That is to say

$ v'(t)+(\prod\limits_{t_{1}\leq t_{k}\leq t}\frac{1+\beta_{k}}{1+\alpha_{k}})\frac{1}{r(t)}v^{2}(t)\leq-(\prod\limits_{t_{1}\leq t_{k}<t}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1})C_{l}q_{l}(t). $ (2.3)

Multiplying (2.3) by $\psi(s)$, we obtain

$ (\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1})\psi(s)C_{l}q_{l}(s)\leq-\psi(s)v'(s)-(\prod\limits_{t_{1}\leq t_{k}<s}\frac{1+\beta_{k}}{1+\alpha_{k}})\frac{\psi(s)}{r(s)}v^{2}(s). $ (2.4)

Multiplying (2.4) by $H_{2}(t, s)$ and integrating over $[c, t]$ for $t\in[c, b)$, we have

$ \begin{eqnarray*} & &\int_{c}^{t}(\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1})H_{2}(t, s)\psi(s)C_{l}q_{l}(s)ds\\ &\leq &-\int_{c}^{t}H_{2}(t, s)\psi(s)v'(s)ds-\int_{c}^{t}H_{2}(t, s)(\prod\limits_{t_{1}\leq t_{k}<s}\frac{1+\beta_{k}}{1+\alpha_{k}})\frac{\psi(s)}{r(s)}v^{2}(s)ds \\ &=&H_{2}(t, c)v(c)\psi(c)-\int_{c}^{t}H_{2}(t, s)(\sqrt{\frac{\prod\limits_{t_{1}\leq t_{k}<s}\frac{1+\beta_{k}}{1+\alpha_{k}}}{r(s)}}v(s)-\frac{1}{2}\lambda_{2}(t, s)\sqrt{\frac{r(s)}{\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}}}})^{2}\psi(s)ds\\ &&+ \frac{1}{4}\int_{c}^{t}\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})H_{2}(t, s)r(s)\psi(s)\lambda^{2}_{2}(t, s)ds\\ &\leq&H_{2}(t, c)v(c)\psi(c)+\frac{1}{4}\int_{c}^{t}\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})H_{2}(t, s)r(s)\psi(s)\lambda^{2}_{2}(t, s)ds, \end{eqnarray*} $

and so

$ \frac{1}{H_{2}(t, c)}\int_{c}^{t}H_{2}(t, s)\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}(C_{l}q_{l}(s)-\frac{1}{4}r(s)\lambda^{2}_{2}(t, s))\psi(s)ds\leq v(c)\psi(c). $

Let $t\rightarrow b^{-}$ in the above, we obtain

$ \frac{1}{H_{2}(b, c)}\int_{c}^{b}H_{2}(b, s)\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}(C_{l}q_{l}(s)-\frac{1}{4}r(s)\lambda^{2}_{2}(b, s))\psi(s)ds\leq v(c)\psi(c). $ (2.5)

On the other hand, multiplying (2.4) by $H_{1}(s, t)$ and integrating over $[t, c]$ for $t\in(a, c]$, we obtain

$ \begin{eqnarray*} & &\int_{t}^{c}(\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1})H_{1}(s, t)\psi(s)C_{l}q_{l}(s)ds \\ & \leq & -\int_{t}^{c}H_{1}(s, t)\psi(s)v'(s)ds-\int_{t}^{c}H_{1}(s, t)(\prod\limits_{t_{1}\leq t_{k}<s}\frac{1+\beta_{k}}{1+\alpha_{k}})\frac{\psi(s)}{r(s)}v^{2}(s)ds \\ &=&-H_{1}(c, t)v(c)\psi(c)-\int_{t}^{c}H_{1}(s, t)(\sqrt{\frac{\prod\limits_{t_{1}\leq t_{k}<s}\frac{1+\beta_{k}}{1+\alpha_{k}}}{r(s)}}v(s)\\&&-\frac{1}{2}\lambda_{1}(s, t)\sqrt{\frac{r(s)}{\prod\limits_{t_{1}\leq t_{k}<s}\frac{1+\beta_{k}}{1+\alpha_{k}}}})^{2}\psi(s)ds\\ &&+ \frac{1}{4}\int_{t}^{c}\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})H_{1}(s, t)r(s)\psi(s)\lambda^{2}_{1}(s, t)ds\\ &\leq&-H_{1}(c, t)v(c)\psi(c)+\frac{1}{4}\int_{t}^{c}\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})H_{1}(s, t)r(s)\psi(s)\lambda^{2}_{1}(s, t)ds, \end{eqnarray*} $

and so

$ \frac{1}{H_{1}(c, t)}\int_{t}^{c}H_{1}(s, t)\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}(C_{l}q_{l}(s)-\frac{1}{4}r(s)\lambda^{2}_{1}(s, t))\psi(s)ds\leq -v(c)\psi(c). $

Let $t\rightarrow a^{+}$ in the above, we get

$ \frac{1}{H_{1}(c, a)}\int_{a}^{c}H_{1}(s, a)\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}(C_{l}q_{l}(s)-\frac{1}{4}r(s)\lambda^{2}_{1}(s, a))\psi(s)ds\leq -v(c)\psi(c). $ (2.6)

Adding (2.5) and (2.6), we easily obtain the following

$ \begin{eqnarray*} & &\frac{1}{H_{1}(c, a)}\int_{a}^{c}H_{1}(s, a)\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}(C_{l}q_{l}(s)-\frac{1}{4}r(s)\lambda_{1}^{2}(s, a))\psi(s)ds\\ &+&\frac{1}{H_{2}(b, c)}\int_{c}^{b}H_{2}(b, s)\prod\limits_{t_{1}\leq t_{k}<s}(\frac{1+\beta_{k}}{1+\alpha_{k}})^{-1}(C_{l}q_{l}(s)-\frac{1}{4}r(s)\lambda_{2}^{2}(b, s))\psi(s)ds\leq0, \end{eqnarray*} $

which contradicts condition (2.1).

References
[1] Lakshmikantham V, Bainov D, Simeonov P S. Theory of impulsive differential equations[M]. Singapore: World Scientific, 1989.
[2] Yutaka Shoukaku, Norio Yoshida. Oscillations of nonlinear hyperbolic equations with functional arguments via Riccati method[J]. Appl. Math. Comput., 2010, 217: 143–151.
[3] Luo Zhiguo, Shen Jianhua. Oscillations of second linear differential equations with impulses[J]. Appl. Math. Lett., 2007, 20: 75–81. DOI:10.1016/j.aml.2006.01.019
[4] Bainov D D, Minchev E. Oscillation of the solutions of impulsive parabolic equations[J]. J. Comput. Appl. Math., 1996, 69: 207–214. DOI:10.1016/0377-0427(95)00040-2
[5] Liu Anping, Liu Ting, Zou Min. Oscillation of nonlinear impulsive parabolic differential equations of neutral type[J]. Rocky Mount. J. Math., 2011, 41: 833–850. DOI:10.1216/RMJ-2011-41-3-833
[6] Chen Rongsan, Zou Min, Liu Anping. Comparison of several numerical schemes for scalar linear advaction equation[J]. J. Math., 2015, 35(4): 977–982.