数学物理学报  2017, Vol. 37 Issue (6): 1094-1104   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
李海侠
一类食物链模型正解的稳定性和唯一性
李海侠     
宝鸡文理学院数学与信息科学学院 陕西 宝鸡 721013
摘要:该文研究了一类带有Leslie-Gower项和Crowley-Martin反应函数的食物链模型.运用不动点指数理论给出了正解存在的充分条件, 进而利用椭圆型方程正则理论讨论了正解的不存在性、稳定性和唯一性.结果表明, 当参数$c$充分大或有界时系统在不同条件下均存在唯一且线性稳定的正解.最后通过数值模拟对分析结果进行了验证和补充.
关键词食物链模型    Crowley-Martin反应函数    不动点指数    稳定性    唯一性    
Stability and Uniqueness of Positive Solutions for a Food-Chain Model
Li Haixia     
Institute of Mathematics and Information Science, Baoji University of Arts and Sciences, Shanxi Baoji 721013
Abstract: A food-chain model with Leslie-Gower and Crowley-Martin functional response is investigated in this paper. The sufficient conditions for the existence of positive solutions are given by means of the fixed point index. Furthermore, use the regularity theory of elliptic equations, the nonexistence, stability and uniqueness of positive solutions are discussed. The results show that there exists a unique linearly stable positive solution under different conditions when the parameter $c$ is large or bounded. Finally, some numerical simulations are presented to verify and complement the theoretical results.
Key words: Food-chain model     Crowley-Martin functional response     Fixed point index     Stability     Uniqueness    
1 引言

本文在齐次Dirichlet边界条件下讨论如下食物链模型

$\begin{eqnarray}\label{pde1} \left\{\begin{array}{ll}-\Delta u=u(r_{1}-u-a_{1}v),&x\in\Omega,\\ -\Delta \displaystyle v=v\Big(r_{2}-v+\frac{a_{2}u}{(1+cu)(1+dw)}\Big),&x\in\Omega,\\ \displaystyle -\Delta w=w\Big(r_{3}-\frac{a_{3}w}{1+ev}\Big),&x\in\Omega,\\ u=v=w=0,&x\in\partial\Omega,\end{array}\right. \end{eqnarray}$ (1.1)

其中 $\Omega$ ${\Bbb R}^{N}$ 中带有光滑边界 $\partial\Omega$ 的有界区域, $\!\!u,v$ $w$ 分别表示食饵、第一个捕食者和第二个捕食者的浓度. $\!\!r_{1},r_{2}$ $r_{3}$ 分别是 $u,v$ $w$ 的最大增长率. $\!\!a_{1}$ 表示捕食者捕获食饵的转化率. $\!\!a_{2},c$ $d$ 分别代表了捕食者的捕获率、捕食者对食饵的处理时间和捕食者间的干扰度.系统(1.1)中参数 $c,d,e,a_{i},r_{i}(i=1,2,3)$ 都是正常数. $\!\!a_{2}u/[(1+cu)(1+dw)]$ 是Crowley-Martin (简记C-M)反应函数, 它认为无论某个捕食者目前是否寻找或处理食饵, 捕食者间的干扰总存在, 这是比其它反应函数的优越之处, 也更加符合很多生物现象.对于C-M反应函数更多的生物意义及背景, 感兴趣的读者可见文献[1-7].

两物种捕食-食饵模型一直以来被许多学者所关注[8-12].近年来, 三物种捕食-食饵模型的动力学行为也得到了一些研究, 见文献[13-21].文献[13]利用全局分歧理论得到了系统正解的局部唯一性和稳定性, 文献[14, 16]讨论了抛物系统的渐近行为以及物种增长率远离 $-\Delta$ $\Omega$ 上关于齐次Dirichlet边界条件下的主特征值 $\lambda_{1}$ 时正平衡态解的唯一性和稳定性, 文献[20]利用扰动理论、分歧理论和度理论分析了系统正解的存在性和多重性.然而, 当物种增长率接近主特征值 $\lambda_{1}$ 时模型的动力学分析非常复杂, 因此目前这种情况下对正解唯一性和稳定性的研究工作甚少.鉴于此, 本文主要探讨食饵 $u$ 的增长率 $r_{1}$ 接近主特征值 $\lambda_{1}$ 时参数 $c$ 对系统(1.1)正解的影响, 从而得到系统(1.1)正解的唯一性和稳定性条件.

这节最后给出一些基础知识.令 $\lambda_{1}(q)$ $-\Delta+q(x)$ $\Omega$ 上关于齐次Dirichlet边界条件下的主特征值.记主特征值 $\lambda_{1}$ 所对应的特征函数为 $\varphi_{1}$ $\|\varphi_{1}\|_{\infty}=1$ .

众所周知, 问题

$\begin{eqnarray}\begin{array}{ll} -\Delta u=u(r-au),x\in\Omega;\ \ u=0,x\in\partial\Omega \end{array}\end{eqnarray}$ (1.2)

$r>\lambda_{1}$ 时有唯一正解, 记为 $\Theta_{(r,a)}$ .而且 $\Theta_{(r,a)}<r/a$ .特别地, 记 $\Theta_{(r,1)}=\Theta_{r}$ .

显然系统(1.1)可能有如下四类形式的非负解

(ⅰ) 平凡解 $(0,0,0)$ ;

(ⅱ)弱半平凡解 $(u,0,0),(0,v,0)$ $(0,0,w)$ ;

(ⅲ)强半平凡解 $(u,v,0),(0,v,w)$ $(u,0,w)$ ;

(ⅳ)正解 $(u,v,w)$ .

2 共存解的存在性

本小节利用度理论给出系统(1.1)正解存在的条件.易看出当 $r_{1}>\lambda_{1},r_{2}>\lambda_{1},r_{3}>\lambda_{1}$ 时, 系统(1.1)分别存在弱半平凡解 $(\Theta_{r_{1}},0,0),(0,\Theta_{r_{2}},0),(0,0,\hat{w})$ , 这里 $\hat{w}=\Theta_{(r_{3},a_{3})}.$ $r_{2}>\lambda_{1},r_{3}>\lambda_{1}$ , 记方程

$\begin{eqnarray} -\Delta w=w\Big(r_{3}-\frac{a_{3}w}{1+e\Theta_{r_{2}}}\Big),\ x\in\Omega;\ \ w=0,x\in\partial\Omega \end{eqnarray}$

的唯一正解为 $\tilde{w}$ .令 $\bar{r}_{2}=\lambda_{1}(-\frac{a_{2}\Theta_{r_{1}}}{(1+c\Theta_{r_{1}})(1+d\hat{w})})$ .

首先给出系统(1.1)正解存在的必要条件.设 $(u,v,w)$ 是系统(1.1)的正解, 则根据特征值的比较原理知 $r_{1}=\lambda_{1}(u+a_{1}v)>\lambda_{1}$ .同理得 $r_{2}>\bar{r}_{2},r_{3}>\lambda_{1}$ .于是有

引理1 如果 $(u,v,w)$ 是系统(1.1)的正解, 则 $r_{1}>\lambda_{1},r_{2}>\bar{r}_{2},r_{3}>\lambda_{1}$ .

接着利用上下解方法给出系统(1.1)正解的先验估计.证明方法是基本的, 故我们不予以证明.

引理2 系统(1.1)的任意正解 $(u,v,w)$ 满足

$\begin{eqnarray}u\leq\Theta_{r_{1}}<r_{1},v\leq\hat{v}<Q_{1},\hat{w}\leq w<Q_{2}.\end{eqnarray}$

进而, 若 $r_{2}>\lambda_{1}$ , 则 $\Theta_{r_{2}}\leq v$ , 这里 $Q_{1}=r_{2}+a_{2}r_{1},Q_{2}=\frac{r_{3}(1+eQ_{1})}{a_{3}},$ $\hat{v}$ 是如下问题的唯一正解

$\begin{eqnarray} -\Delta v=v\Big(r_{2}-v+\frac{a_{2}\Theta_{r_{1}}}{1+c\Theta_{r_{1}}}\Big), \ x\in\Omega;\ \ v=0,x\in\partial\Omega. \end{eqnarray}$

其次, 类似文献[2, 引理2.5]的证明给出如下表明 $r_{2}$ 有界的结果.

引理3 如果 $(u,v,w)$ 是系统(1.1)的正解且 $r_{2}>\lambda_{1}$ , 则存在正常数 $M$ 使得 $r_{2}\leq M$ .

其次, 为了利用度理论引入如下空间

$\begin{eqnarray}\begin{array}{lll}C_{0}(\overline{\Omega})=\{u\in C(\overline{\Omega}):u(x)=0,x\in\partial\Omega\};K=\{u\in C(\overline{\Omega}):u(x)\geq0,x\in\overline{\Omega}\}; \\X=C_{0}(\overline{\Omega})\times C_{0}(\overline{\Omega})\times C_{0}(\overline{\Omega}); W=K\times K\times K; \\D=\{(u,v,w)\in W: u< r_{1}+1,v< Q_{1}+1,w<Q_{2}+1\}.\end{array}\end{eqnarray}$

定义算子 ${\rm A}_{\tau}:X\rightarrow X$

$\begin{eqnarray}{\rm A}_{\tau}=(-\Delta+q)^{-1}\left( \begin{array}{c} \tau u(r_{1}-u-a_{1}v)+qu \\ \displaystyle \tau v\Big(r_{2}-v+\frac{a_{2}u}{(1+cu)(1+dw)}\Big)+qv \\ \displaystyle \tau w\Big(r_{3}-\frac{a_{3}w}{1+ev}\Big)+qw \end{array} \right), \end{eqnarray}$

其中 $ \tau\in[0, 1],q$ 为充分大的正常数.显然 ${\rm A}_{\tau}$ 是紧算子.记 ${\rm A}={\rm A}_{1}$ , 则易知 ${\rm A}:D\rightarrow W$ 是连续可微算子.显然, 系统(1.1)有正解当且仅当 ${\rm A}$ $D$ 中有正不动点.

利用Dancer指数定理[22]可得如下引理, 由于其计算过程繁琐, 故在此省略其证明.

引理4 (ⅰ)设 $r_{1}\neq\lambda_{1},r_{2}\neq\lambda_{1},r_{3}\neq\lambda_{1}.$ $r_{1}>\lambda_{1}$ $r_{2}>\lambda_{1}$ $r_{3}>\lambda_{1}$ , 则 ${\rm index}_{W}({\rm A},(0,0,0))=0$ ; 若 $r_{1}<\lambda_{1}, r_{2}<\lambda_{1}$ $r_{3}<\lambda_{1}$ , 则 ${\rm index}_{W}({\rm A},(0,0,0))=1$ .

(ⅱ) ${\rm index}_{W}({\rm A},D)=1$ .

(ⅲ)设 $r_{1}>\lambda_{1},r_{2}\neq\lambda_{1}(-\frac{a_{2}\Theta_{r_{1}}}{1+c\Theta_{r_{1}}}),r_{3}\neq\lambda_{1}$ .若 $r_{2}>\lambda_{1}(-\frac{a_{2}\Theta_{r_{1}}}{1+c\Theta_{r_{1}}})$ $r_{3}>\lambda_{1}$ , 则 ${\rm index}_{W}({\rm A},$ $(\Theta_{r_{1}},0,0))=0$ ; 若 $r_{2}<\lambda_{1}(-\frac{a_{2}\Theta_{r_{1}}}{1+c\Theta_{r_{1}}})$ $r_{3}<\lambda_{1}$ , 则 ${\rm index}_{W}({\rm A},(\Theta_{r_{1}},0,0))=1$ .

(ⅳ)设 $r_{2}>\lambda_{1},r_{1}\neq\lambda_{1}(a_{1}\Theta_{r_{2}}),r_{3}\neq\lambda_{1}$ .若 $r_{1}>\lambda_{1}(a_{1}\Theta_{r_{2}})$ $r_{3}>\lambda_{1}$ , 则 ${\rm index}_{W}({\rm A},(0, \Theta_{r_{2}},0))=0$ ; 若 $r_{1}<\lambda_{1}(a_{1}\Theta_{r_{2}})$ $r_{3}<\lambda_{1}$ , 则 ${\rm index}_{W}({\rm A},(0,\Theta_{r_{2}},0))=1$ .

(ⅴ) 设 $r_{3}>\lambda_{1},r_{1}\neq\lambda_{1},r_{2}\neq\lambda_{1}$ .若 $r_{1}>\lambda_{1}$ $r_{2}>\lambda_{1}$ , 则 ${\rm index}_{W}({\rm A},(0,0,\hat{w}))=0$ ; 若 $r_{1}<\lambda_{1}$ $r_{2}<\lambda_{1}$ , 则 ${\rm index}_{W}({\rm A},(0,0,\hat{w}))=1$ .

(ⅵ)设 $r_{1}>\lambda_{1},r_{3}>\lambda_{1}$ .若 $r_{2}>\bar{r}_{2}$ , 则 ${\rm index}_{W}({\rm A},(\Theta_{r_{1}},0,\hat{w}))=0$ ; 若 $r_{2}<\bar{r}_{2}$ , 则 ${\rm index}_{W}({\rm A},(\Theta_{r_{1}},0,\hat{w}))=1$ .

(ⅶ)设 $r_{2}>\lambda_{1},r_{3}>\lambda_{1}$ .若 $r_{1}>\lambda_{1}(a_{1}\Theta_{r_{2}})$ , 则 ${\rm index}_{W}({\rm A},(0,\Theta_{r_{2}},\tilde{w}))=0$ ; 若 $r_{1}<\lambda_{1}(a_{1}\Theta_{r_{2}})$ , 则 ${\rm index}_{W}({\rm A},(0,\Theta_{r_{2}},\tilde{w}))=1$ .

$M_{1}=\{u\in C(\overline{\Omega}):u(x)<r_{1}+1\},M_{2}=\{v\in C(\overline{\Omega}):v(x)<Q_{1}+1\},S=\{(\bar{u},\bar{v},0)\in M_{1}\oplus M_{2} \oplus \{0\}\}$ , 这里 $(\bar{u},\bar{v})$ 是如下问题

$\begin{eqnarray} \left\{\begin{array}{ll} -\Delta u=u(r_{1}-u-a_{1}v),&x\in\Omega,\\ \displaystyle -\Delta v=v\Big(r_{2}-v+\frac{a_{2}u}{1+cu}\Big),&x\in\Omega,\\ u=v=0,&x\in\partial\Omega \end{array}\right. \end{eqnarray}$

的正解.类似文献[21, 引理4.11]的证明可得如下引理.

引理5 设 $r_{1}>\lambda_{1}(a_{1}\Theta_{r_{2}}),r_{2}>\lambda_{1}$ $r_{1}>\lambda_{1},\lambda_{1}(-\frac{a_{2}\Theta_{r_{1}}}{1+c\Theta_{r_{1}}})<r_{2}<\lambda_{1}$ .若 $r_{3}>\lambda_{1}$ , 则 ${\rm index}_{W}({\rm A},S)=0$ .

最后由度的可加性得系统(1.1)正解存在的充分条件.

定理1 (ⅰ)如果 $r_{1}>\lambda_{1}(a_{1}\Theta_{r_{2}}),r_{2}>\lambda_{1},r_{3}>\lambda_{1}$ , 则系统(1.1)至少存在一个正解;

(ⅱ)如果 $r_{1}>\lambda_{1},\bar{r}_{2}<r_{2}<\lambda_{1},r_{3}>\lambda_{1}$ , 则系统(1.1)至少存在一个正解.

3 共存解的不存在性、稳定性和唯一性

本节将重点讨论当 $r_{1}$ 接近于 $\lambda_{1}$ $r_{2}\leq\lambda_{1}$ 时系统(1.1)正解的唯一性和稳定性.首先给出唯一性引理.

引理6 设 $r_{1}>\lambda_{1},r_{3}>\lambda_{1},\hat{r}_{2}>\lambda_{1}$ 定义为: $r_{1}=\lambda_{1}(a_{1}\Theta_{\hat{r}_{2}})$ .如果 $r_{2}\in(\bar{r}_{2},\hat{r}_{2})$ 且系统(1.1)的任意正解非退化且线性稳定, 则系统(1.1)存在唯一正解.

 只需要证明唯一性.由 $r_{1}>\lambda_{1},r_{3}>\lambda_{1},r_{2}\in(\bar{r}_{2},\hat{r}_{2})$ 可证系统(1.1)的平凡解和半平凡解都远离正解.又因为系统(1.1)的任意正解非退化, 所以根据线性算子的紧性理论可知系统(1.1)最多有有限个正解.记为 $\{(u_{i},v_{i},w_{i}):i=1,2,\cdots,n\}$ .又根据文献[22, 定理1]通过计算得 ${\rm index}_{W}({\rm A},(u_{i},v_{i},w_{i}))=1$ .另外, 在已知条件下由引理4可知算子 ${\rm A}$ 在系统(1.1)的平凡解和所有半平凡解处的指数为0.所以根据度的可加性得

$\begin{eqnarray}1={\rm index}_{W}({\rm A},D)=\mathop{\sum}\limits_{1\leq i\leq n}{\rm index}_{W}({\rm A},(u_{i},v_{i},w_{i}))+0=n.\end{eqnarray}$

这说明系统(1.1)存在唯一非退化且线性稳定的正解.

利用引理1和全局分歧理论可得如下引理.证明是基本的, 因此在此省略.

引理7 (ⅰ)设 $r_{3}>\lambda_{1}$ .若 $r_{2}<\lambda_{1}(-\frac{a_{2}}{c(1+d\hat{w})})$ , 则系统(1.1)无正解;

(ⅱ)设 $r_{3}>\lambda_{1}$ .若 $\lambda_{1}(-\frac{a_{2}}{c(1+d\hat{w})})<r_{2}\leq\lambda_{1}$ , 则存在唯一常数 $\bar{r}_{1}>\lambda_{1}$ , 当 $r_{1}>\bar{r}_{1}$ 时系统(1.1)至少存在一个正解; 当 $r_{1}\leq\bar{r}_{1}$ 时系统(1.1)无正解, 这里 $\bar{r}_{1}$ $r_{2}=\lambda_{1}(-\frac{a_{2}\Theta_{\bar{r}_{1}}}{(1+c\Theta_{\bar{r}_{1}})(1+d\hat{w})})$ 唯一确定.

下面给出对于任意 $c\geq0$ , 系统(1.1)正解的唯一性和稳定性条件.

定理2 设 $r_{2}\leq\lambda_{1},r_{3}>\lambda_{1}$ .则存在任意小的 $\varepsilon>0$ , 对于任意 $\lambda_{1}<r_{1}\leq\lambda_{1}+\varepsilon$ $c\geq0$ , 系统(1.1)至多存在一个正解.而且正解(如果存在)是非退化和线性稳定的.

 在定理的条件下, 根据引理6和引理7只需证明系统(1.1)的任意正解非退化且线性稳定.采用反证法.假设存在 $r_{1,i}\rightarrow\lambda^{+}_{1},c_{i}\rightarrow c\geq0$ $r_{2,i}\leq\lambda_{1}$ 使得系统(1.1)带有 $(r_{1},c,r_{2})=(r_{1,i},c_{i},r_{2,i})$ 的正解 $(u_{i},v_{i},w_{i})$ 退化或不稳定, 即存在 ${\rm Re}(\nu_{i})\leq0$ $\nu_{i}$ 以及满足 $\|\xi_{i}\|^{2}_{L^{2}}+\|\omega_{i}\|^{2}_{L^{2}}+\|\chi_{i}\|^{2}_{L^{2}}=1$ $(\xi_{i},\omega_{i},\chi_{i})\neq(0,0,0)$ 使得

$\begin{eqnarray} \left\{\begin{array}{lll}-\Delta \xi_{i}-(r_{1,i}-2u_{i}-a_{1}v_{i})\xi_{i}+a_{1}u_{i}\omega_{i}=\nu_{i}\xi_{i},&x\in\Omega,\\ \displaystyle -\Delta \omega_{i}- \Big[r_{2,i}-2v_{i}+\frac{a_{2}u_{i}}{(1+c_{i}u_{i})(1+dw_{i})})\Big]\omega_{i}-\frac{a_{2}v_{i}}{(1+c_{i}u_{i})^{2}(1+dw_{i})}\xi_{i} \\ \displaystyle\quad +\frac{a_{2}du_{i}v_{i}}{(1+c_{i}u_{i})(1+dw_{i})^{2}}\chi_{i}=\nu_{i}\omega_{i},&x\in\Omega,\\ \displaystyle -\Delta \chi_{i}-\Big(r_{3}-\frac{2a_{3}w_{i}}{1+ev_{i}}\Big)\chi_{i}-\frac{a_{3}ew_{i}^{2}}{(1+ev_{i})^{2}}\omega_{i}=\nu_{i}\chi_{i},&x\in\Omega,\\ \xi_{i}=\omega_{i}=\chi_{i}=0,&x\in\partial\Omega, \end{array}\right. \end{eqnarray}$ (3.1)

因为 $0<u_{i}\leq\Theta_{r_{1,i}}$ , 所以在 $L^{\infty}$ $u_{i}\rightarrow0$ .根据系统(1.1)中 $v_{i}$ 的方程易有

$\begin{eqnarray}\begin{array}{lll} \Theta_{r_{2,i}}\leq v_{i}\leq \Theta_{r_{2,i}+a_{2}\|\Theta_{r_{1,i}}\|_{\infty}}, \end{array}\end{eqnarray}$ (3.2)

$v_{i}\not\equiv0$ $r_{2,i}+a_{2}\|\Theta_{r_{1,i}}\|_{\infty}>\lambda_{1}$ .于是 $\lambda_{1}-a_{2}\|\Theta_{r_{1,i}}\|_{\infty}<r_{2,i}\leq\lambda_{1}$ .又因为 $r_{1,i}\rightarrow\lambda^{+}_{1}$ , 所以 $\|\Theta_{r_{1,i}}\|_{\infty}\rightarrow0$ .故 $r_{2,i}\rightarrow\lambda_{1}$ .由(3.2)式知在 $L^{\infty}$ $v_{i}\rightarrow0$ .同理由 $w_{i}$ 的方程有 $\hat{w}\leq w_{i}\leq\Theta_{(r_{3},\frac{a_{3}}{1+e\|v_{i}\|_{\infty}})}$ , 故 $w_{i}\rightarrow\hat{w}$ .令 $U_{n}=\frac{u_{i}}{\|u_{i}\|_{\infty}},$

$\begin{eqnarray} -\Delta U_{n}=U_{n}(r_{1,i}-u_{i}-a_{1}v_{i}),x\in\Omega;\ \ U_{n}=0,x\in\partial\Omega . \end{eqnarray}$

根据 $L^{p}$ 估计和Sobolev嵌入定理可设 $U_{n}\rightarrow\bar{U}\geq0,\not\equiv0$ .又由Harnack不等式得 $\bar{U}>0,x\in\Omega$ .上式取极限得在 $C^{1}$ $\frac{u_{i}}{\|u_{i}\|_{\infty}}\rightarrow\varphi_{1}$ .同理可得 $\frac{v_{i}}{\|v_{i}\|_{\infty}}\rightarrow\varphi_{1}$ .因此可令

$\begin{eqnarray} u_{i}=s_{i}{\rm cos}\alpha_{i}(\varphi_{1}+p_{i}),v_{i}=s_{i}{\rm sin}\alpha_{i}(\varphi_{1}+q_{i}) ,\end{eqnarray}$

其中在 $C^{1}$ $p_{i},q_{i}\rightarrow0,(p_{i},\varphi_{1})_{2}=(q_{i},\varphi_{1})_{2}=0,\alpha_{i}\in(0,\frac{\pi}{2})$ $s_{i}=\sqrt{\frac{\|u_{i}\|^{2}_{\infty}}{\|\varphi_{1}+p_{i}\|^{2}_{\infty}}+\frac{\|v_{i}\|^{2}_{\infty}}{\|\varphi_{1}+q_{i}\|^{2}_{\infty}}}\rightarrow0^{+}$ .记 $\bar{\xi}_{i},\bar{\omega}_{i}$ $\bar{\chi}_{i}$ 分别是 $\xi_{i},\omega_{i}$ $\chi_{i}$ 的共轭算子, 则由(3.1)式得

$\begin{eqnarray*} \nu_{i}&=&\int_{\Omega}|\nabla\xi_{i}|^{2}{\rm d}x-\int_{\Omega}(r_{1,i}-2u_{i}-a_{1}v_{i})|\xi_{i}|^{2}{\rm d}x+\int_{\Omega} a_{1}u_{i}\omega_{i}\bar{\xi}_{i}{\rm d}x\\ &&+\int_{\Omega}|\nabla\omega_{i}|^{2}{\rm d}x -\int_{\Omega}\Big[r_{2,i}-2v_{i}+\frac{a_{2}u_{i}}{(1+c_{i}u_{i})(1+dw_{i})}\Big]| \omega_{i}|^{2}{\rm d}x\\ &&-\int_{\Omega}\frac{a_{2}v_{i}}{(1+c_{i}u_{i})^{2}(1+dw_{i})}\xi_{i}\bar{\omega}_{i}{\rm d}x +\int_{\Omega}\frac{a_{2}du_{i}v_{i}}{(1+c_{i}u_{i})(1+dw_{i})^{2}}\chi_{i} \bar{\omega}_{i}{\rm d}x\\ &&+\int_{\Omega}|\nabla\chi_{i}|^{2}{\rm d}x -\int_{\Omega}\Big(r_{3}-\frac{2a_{3}w_{i}}{1+ev_{i}}\Big)|\chi_{i}|^{2}{\rm d}x-\int_{\Omega}\frac{a_{3}ew_{i}^{2}}{(1+ev_{i})^{2}}\omega_{i}\bar{\chi}_{i}{\rm d}x. \end{eqnarray*}$

因为 $u_{i},v_{i}$ $w_{i}$ 有界, 所以 ${\rm Re}(\nu_{i})$ ${\rm Im}(\nu_{i})$ 有界, 故 $\nu_{i}$ 有界.设 $\nu_{i}\rightarrow\nu$ , 则 ${\rm Re}(\nu)\leq0$ .又由 $L^{p}$ 估计可知 $\xi_{i},\omega_{i}$ $\chi_{i}$ 有界, 设 $\xi_{i}\rightarrow\xi,\omega_{i}\rightarrow\omega,\chi_{i}\rightarrow\chi$ .对(3.1)式取极限得

$\begin{eqnarray} \left\{\begin{array}{lll}-\Delta \xi-\lambda_{1}\xi=\nu\xi,&x\in\Omega,\\[2pt] -\Delta\omega-\lambda_{1}\omega=\nu\omega,&x\in\Omega,\\[1pt] -\Delta\chi-(r_{3}-2a_{3}\hat{w})\chi-a_{3}e\hat{w}^{2}\omega=\nu\chi,&x\in\Omega,\\[2pt] \xi=\omega=\chi=0,&x\in\partial\Omega. \end{array}\right. \end{eqnarray}$

$\nu$ 是实的.若 $\omega\equiv0$ , 则 $\nu\geq\lambda_{1}(-r_{3}+2a_{3}\hat{w})>0$ , 矛盾.因此 $\omega\not\equiv0$ .则 $\nu=0$ $\xi=k_{1}\varphi_{1},\omega=k_{2}\varphi_{1},\chi=(-\Delta-(r_{3}-2a_{3}\hat{w}))^{-1}(a_{3}e\hat{w}^{2}k_{2}\varphi_{1})$ , 此处 $k_{1},k_{2}$ 都是实数且 $(k_{1},k_{2})\neq(0,0)$ .由Kato's不等式和(3.1)式的第二个式子得

$\begin{eqnarray*}-\Delta|\omega_{i}| &\leq&-{\rm Re}\Big(\frac{\bar{\omega}_{i}}{|\omega_{i}|}\Delta\omega_{i}\Big)\\ &\leq& \Big[r_{2,i}-2v_{i}+\frac{a_{2}u_{i}}{(1+c_{i}u_{i})(1+dw_{i})}\Big]|\omega_{i}|+{\rm Re}(\nu_{i})|\omega_{i}|+\frac{a_{2}v_{i}}{(1+c_{i}u_{i})^{2}(1+dw_{i})}|\xi_{i}|. \end{eqnarray*}$

上式乘以 $v_{i}$ 并在 $\Omega$ 上积分得

$\begin{eqnarray*}-\int_{\Omega}\Delta|\omega_{i}|v_{i}{\rm d}x &=&\int_{\Omega}\Big[r_{2,i}-v_{i}+\frac{a_{2}u_{i}}{(1+c_{i}u_{i})(1+dw_{i})}\Big]|\omega_{i}|v_{i}{\rm d}x\\ &\leq&\int_{\Omega}\Big[r_{2,i}-2v_{i}+\frac{a_{2}u_{i}}{(1+c_{i}u_{i})(1+dw_{i})} \Big]|\omega_{i}|v_{i}{\rm d}x\\ &&+\int_{\Omega}{\rm Re}(\nu_{i})|\omega_{i}|v_{i}{\rm d}x +\int_{\Omega}a_{2}v^{2}_{i}|\xi_{i}|{\rm d}x. \end{eqnarray*}$

因此

$\begin{eqnarray}\int_{\Omega}v^{2}_{i}|\omega_{i}|{\rm d}x\leq a_{2}\int_{\Omega}v^{2}_{i}|\xi_{i}|{\rm d}x+{\rm Re}(\nu_{i})\int_{\Omega}|\omega_{i}|v_{i}{\rm d}x\leq a_{2}\int_{\Omega}v^{2}_{i}|\xi_{i}|{\rm d}x, \end{eqnarray}$

$\begin{eqnarray}\int_{\Omega}\Big(\frac{v_{i}}{\|v_{i}\|_{\infty}}\Big)^{2}|\omega_{i}|{\rm d}x\leq a_{2}\int_{\Omega}\Big(\frac{v_{i}}{\|v_{i}\|_{\infty}}\Big)^{2}|\xi_{i}|{\rm d}x. \end{eqnarray}$

上式取极限有 $\int_{\Omega}\varphi^{2}_{1}|\omega|{\rm d}x\leq a_{2}\int_{\Omega}\varphi^{2}_{1}|\xi|{\rm d}x$ .故 $k_{1}\neq0$ $|k_{2}|\leq a_{2}|k_{1}|$ .令 $k_{1}=1$ , 则 $\xi_{i}\rightarrow\varphi_{1},\omega_{i}\rightarrow h\varphi_{1},\chi_{i}\rightarrow(-\Delta-(r_{3}-2a_{3}\hat{w}))^{-1}(a_{3}e\hat{w}^{2}h\varphi_{1})$ , 这里 $|h|\leq a_{2}$ .于是可假设 $\xi_{i}=\varphi_{1}+\hat{\xi}_{i},\omega_{i}=h_{i}(\varphi_{1}+\hat{\omega}_{i})$ , 其中 $(\varphi_{1},\hat{\xi}_{i})_{2}=(\varphi_{1},\hat{\omega}_{i})_{2}=0,\hat{\xi}_{i}\rightarrow0,\hat{\omega}_{i}\rightarrow0,h_{i}\rightarrow h$ .给(3.1)式的第一个方程乘以 $\varphi_{1}$ , 在 $\Omega$ 上积分, 并结合 $\xi_{i}=\varphi_{1}+\hat{\xi}_{i},\omega_{i}=h_{i}(\varphi_{1}+\hat{\omega}_{i})$

$\begin{eqnarray}\lambda_{1}\int_{\Omega}\varphi^{2}_{1}{\rm d}x=r_{1,i}\int_{\Omega}\varphi^{2}_{1}{\rm d}x+\nu_{i}\int_{\Omega}\varphi^{2}_{1}{\rm d}x-\int_{\Omega}(2u_{i}+a_{1}v_{i})\xi_{i}\varphi_{1}{\rm d}x -\int_{\Omega}a_{1}u_{i}\omega_{i}\varphi_{1}{\rm d}x, \end{eqnarray}$

$\begin{eqnarray}(r_{1,i}-\lambda_{1})\int_{\Omega}\varphi^{2}_{1}{\rm d}x+\nu_{i}\int_{\Omega}\varphi^{2}_{1}{\rm d}x=\int_{\Omega}(2u_{i}+a_{1}v_{i})\xi_{i}\varphi_{1}{\rm d}x+\int_{\Omega}a_{1}u_{i}\omega_{i}\varphi_{1}{\rm d}x. \end{eqnarray}$

$\alpha_{i}\rightarrow\alpha\in[0,\frac{\pi}{2}]$ .将 $u_{i}=s_{i}{\rm cos}\alpha_{i}(\varphi_{1}+p_{i}),v_{i}=s_{i}{\rm sin}\alpha_{i}(\varphi_{1}+q_{i})$ 带入上式, 两边除以 $s_{i}$ 并取极限, 得

$\begin{eqnarray} \mathop{\lim}\limits_{i\rightarrow\infty} \Big(\frac{r_{1,i}-\lambda_{1}}{s_{i}}+\frac{\nu_{i}}{s_{i}}\Big)\int_{\Omega}\varphi^{2}_{1}{\rm d}x =2{\rm cos}\alpha\int_{\Omega}\varphi^{3}_{1}{\rm d}x+a_{1}{\rm sin}\alpha\int_{\Omega}\varphi^{3}_{1}{\rm d}x+a_{1}h{\rm cos}\alpha\int_{\Omega}\varphi^{3}_{1}{\rm d}x.\end{eqnarray}$ (3.3)

$u_{i}$ 的方程除以 $s_{i}{\rm cos}\alpha_{i}$ , 再乘以 $\varphi_{1}$ 并在 $\Omega$ 上积分, 得

$\begin{eqnarray}\lambda_{1}\int_{\Omega}\varphi^{2}_{1}{\rm d}x=r_{1,i}\int_{\Omega}\varphi^{2}_{1}{\rm d}x-\int_{\Omega}(\varphi_{1}+p_{i})(u_{i}+a_{1}v_{i})\varphi_{1}{\rm d}x. \end{eqnarray}$

$u_{i}=s_{i}{\rm cos}\alpha_{i}(\varphi_{1}+p_{i}),v_{i}=s_{i}{\rm sin}\alpha_{i}(\varphi_{1}+q_{i})$ 带入上式, 两边除以 $s_{i}$ 并取极限, 得

$\begin{eqnarray} \mathop{\lim}\limits_{i\rightarrow\infty} \Big(\frac{r_{1,i}-\lambda_{1}}{s_{i}}\Big)\int_{\Omega}\varphi^{2}_{1}{\rm d}x =({\rm cos}\alpha+a_{1}{\rm sin}\alpha)\int_{\Omega}\varphi^{3}_{1}{\rm d}x.\end{eqnarray}$ (3.4)

由(3.3)和(3.4)式可知

$\begin{eqnarray} \mathop{\lim}\limits_{i\rightarrow\infty}\frac{\nu_{i}}{s_{i}}\int_{\Omega}\varphi^{2}_{1}{\rm d}x =({\rm cos}\alpha+a_{1}h{\rm cos}\alpha)\int_{\Omega}\varphi^{3}_{1}{\rm d}x.\end{eqnarray}$ (3.5)

$c_{i}s_{i}\rightarrow\bar{c}\in[0,\infty]$ .同理给 $v_{i}$ 的方程除以 $s_{i}{\rm sin}\alpha_{i}$ , 再乘以 $\varphi_{1}$ 并在 $\Omega$ 上积分, 得

$\begin{eqnarray}\lambda_{1}\int_{\Omega}\varphi^{2}_{1}{\rm d}x=r_{2,i} \int_{\Omega}\varphi^{2}_{1}{\rm d}x-\int_{\Omega}(\varphi_{1}+q_{i}) \Big(v_{i}-\frac{a_{2}u_{i}}{(1+c_{i}u_{i})(1+dw_{i})}\Big)\varphi_{1}{\rm d}x. \end{eqnarray}$

$u_{i}=s_{i}{\rm cos}\alpha_{i}(\varphi_{1}+p_{i}),v_{i}=s_{i}{\rm sin}\alpha_{i}(\varphi_{1}+q_{i})$ 带入上式, 两边除以 $s_{i}$ 并取极限, 得

$\begin{eqnarray}\mathop{\lim}\limits_{i\rightarrow\infty} \Big(\frac{r_{2,i}-\lambda_{1}}{s_{i}}\Big)\int_{\Omega}\varphi^{2}_{1}{\rm d}x =\Big({\rm sin}\alpha-\frac{a_{2}{\rm cos}\alpha}{(1+\bar{c}\varphi_{1}{\rm cos}\alpha)(1+d\hat{w})}\Big)\int_{\Omega}\varphi^{3}_{1}{\rm d}x.\end{eqnarray}$ (3.6)

因为 $r_{2,i}\leq\lambda_{1}$ , 所以(3.6)式暗含了 $\alpha\neq\frac{\pi}{2}$ , 即 $\alpha\in[0,\frac{\pi}{2})$ .而且, 如果 $\alpha\neq0$ , 则 $\bar{c}\neq\infty$ .若 $h_{i}\rightarrow h=0$ , 则由(3.5)式有 $\mathop{\lim}\limits_{i\rightarrow\infty}\frac{\nu_{i}}{s_{i}}\int_{\Omega}\varphi^{2}_{1}{\rm d}x ={\rm cos}\alpha\int_{\Omega}\varphi^{3}_{1}{\rm d}x>0$ , 这与 ${\rm Re}(\nu_{i})\leq0$ 矛盾.因此 $h\neq0$ .给(3.1)式第二个方程乘以 $v_{i}$ 并在 $\Omega$ 上积分, 得

$\begin{eqnarray*}\int_{\Omega}\omega_{i}v^{2}_{i}{\rm d}x &=&a_{2}\int_{\Omega}\frac{v^{2}_{i}}{(1+c_{i}u_{i})^{2}(1+dw_{i})}\xi_{i}{\rm d}x -a_{2}d\int_{\Omega}\frac{u_{i}v^{2}_{i}}{(1+c_{i}u_{i})(1+dw_{i})^{2}}\chi_{i}{\rm d}x\\ &&+\nu_{i}\int_{\Omega}\omega_{i}v_{i}{\rm d}x. \end{eqnarray*}$

给上式除以 $h_{i}(s_{i}{\rm sin}\alpha_{i})^{2}$ , 取实部并取极限, 得

$\begin{eqnarray}\int_{\Omega}\varphi^{3}_{1}{\rm d}x\leq {\rm Re} \Big(\frac{1}{h}\Big)a_{2}\int_{\Omega}\frac{\varphi^{3}_{1}}{(1+\bar{c}\varphi_{1}{\rm cos}\alpha)^{2}(1+d\hat{w})}{\rm d}x. \end{eqnarray}$

于是根据(3.5)式有 $\mathop{\lim}\limits_{i\rightarrow\infty}\frac{{\rm Re}(\nu_{i})}{s_{i}}\int_{\Omega}\varphi^{2}_{1}{\rm d}x>0$ , 又与 ${\rm Re}(\nu_{i})\leq0$ 矛盾.

最后给出当有 $c$ 界时, 系统(1.1)正解的不存在性、唯一性和稳定性的更一般结果.

定理3 设 $r_{3}>\lambda_{1}$ .对于任意小的 $\varepsilon>0$ , 存在 $\hat{c}>0$ , 如果 $\lambda_{1}<r_{1}\leq\lambda_{1}+\varepsilon$ $0\leq c\leq\hat{c}$ , 则当 $r_{2}\not\in(\bar{r}_{2},\hat{r}_{2})$ 时系统(1.1)无正解; 当 $r_{2}\in(\bar{r}_{2},\hat{r}_{2})$ 时系统(1.1)存在唯一且渐近稳定的正解.

 首先证明对于 $\lambda_{1}<r_{1}\leq\lambda_{1}+\varepsilon, 0\leq c\leq\hat{c}$ 和任意 $r_{2}$ , 系统(1.1)的任意正解非退化且线性稳定.采用反证法.假设存在 $r_{1,i}\rightarrow\lambda^{+}_{1},c_{i}\in[0,\hat{c}]$ $r_{2,i}$ 使得系统(1.1)带有 $(r_{1},c,r_{2})=(r_{1,i},c_{i},r_{2,i})$ 的正解 $(u_{i},v_{i},w_{i})$ 退化或不稳定, 即存在 ${\rm Re}(\nu_{i})\leq0$ $\nu_{i}$ 以及 $(\xi_{i},\omega_{i},\chi_{i})\neq(0,0,0)$ 使得(3.1)式成立.同定理2可得在 $L^{\infty}$ $u_{i}\rightarrow0,\lambda_{1}-a_{2}\| \Theta_{r_{1,i}}\|_{\infty}<r_{2,i}$ .设 $r_{2,i}\rightarrow\breve{r}_{2} \in[\lambda_{1},+\infty)$ .由(3.2)式知在 $L^{\infty}$ $v_{i} \rightarrow\Theta_{\breve{r}_{2}}$ .另一方面, $r_{1,i}=\lambda_{1} (u_{i}+a_{1}v_{i})\rightarrow\lambda_{1}(a_{1}\Theta_{\breve{r}_{2}})$ .即 $\lambda_{1}(a_{1}\Theta_{\breve{r}_{2}})=\lambda_{1}$ , 这意味着 $\breve{r}_{2}=\lambda_{1}$ .因此 $r_{2,i}\rightarrow\lambda_{1}$ .再由(3.2)式得在 $L^{\infty}$ $v_{i}\rightarrow0$ .于是在 $L^{\infty}$ $w_{i}\rightarrow\hat{w}$ .

易看出定理2的证明过程仍然成立, 只不过没有条件 $r_{2,i}\leq\lambda_{1}$ 的限制.此时, $c_{i}s_{i}\rightarrow\bar{c}=0,$ 而且 ${\rm Re}(h)>0$ 也成立.另一方面, 给 $u_{i}$ 的方程乘以 $\xi_{i}$ $\Omega$ 上积分, 得

$\begin{eqnarray*}\int_{\Omega}\xi_{i}u_{i}(r_{1,i}-u_{i}-a_{1}v_{i}){\rm d}x &=&\int_{\Omega}(-\Delta u_{i})\xi_{i}{\rm d}x=\int_{\Omega} u_{i}(-\Delta\xi_{i}){\rm d}x\\ &=&\int_{\Omega}\xi_{i}u_{i}(r_{1,i}-2u_{i}-a_{1}v_{i}){\rm d}x \\ && -a_{1}\int_{\Omega}u^{2}_{i}\omega_{i}{\rm d}x+\nu_{i}\int_{\Omega}u_{i}\xi_{i}{\rm d}x. \end{eqnarray*}$

于是

$\begin{eqnarray}\int_{\Omega}\xi_{i}u^{2}_{i}{\rm d}x=-a_{1}\int_{\Omega}u^{2}_{i}\omega_{i}{\rm d}x+\nu_{i}\int_{\Omega}u_{i}\xi_{i}{\rm d}x. \end{eqnarray}$

$u_{i}=s_{i}{\rm cos}\alpha_{i}(\varphi_{1}+p_{i})$ 带入上式, 除以 $s^{2}_{i}({\rm cos}\alpha_{i})^{2}$ , 得

$\begin{eqnarray}\int_{\Omega}(\varphi_{1}+p_{i})^{2}\xi_{i}{\rm d}x=-a_{1}\int_{\Omega}(\varphi_{1}+p_{i})^{2}\omega_{i}{\rm d}x +\frac{\nu_{i}}{s_{i}{\rm cos}\alpha_{i}}\int_{\Omega}(\varphi_{1}+p_{i})\xi_{i}{\rm d}x. \end{eqnarray}$

上式取实部并取极限, 得

$\begin{eqnarray}\int_{\Omega}\varphi^{3}_{1}{\rm d}x\leq -a_{1}{\rm Re}(h)\int_{\Omega}\varphi^{3}_{1}{\rm d}x,\end{eqnarray}$

这与 ${\rm Re}(h)>0$ 矛盾.因此系统(1.1)的任意正解非退化且线性稳定.故根据引理6可知当 $r_{2}\in(\bar{r}_{2},\hat{r}_{2})$ 时系统(1.1)存在唯一非退化渐近稳定的正解.

最后证明当 $r_{2}\not\in(\bar{r}_{2},\hat{r}_{2})$ 时系统(1.1)无正解.由引理1知当 $r_{2}\leq\bar{r}_{2}$ 时系统(1.1)无正解, 故只需证明 $r_{2}\geq\hat{r}_{2}$ 时系统(1.1)无正解即可.反证.假设对于 $\lambda_{1}<r'_{1}\leq\lambda_{1}+\varepsilon,0\leq c'\leq\hat{c}$ , 存在 $r'_{2}\geq\hat{r}_{2}$ 使得系统(1.1)存在带有 $(r_{1},c,r_{2})=(r'_{1},c',r'_{2})$ 的正解.令

$\begin{eqnarray}\tilde{r}_{2}={\rm sup}\{r''_{2}: \mbox{ 系统(1.1)存 在$(r_{1},c,r_{2})=(r'_{1},c',r''_{2})$的正解}\ \}.\end{eqnarray}$

显然 $\tilde{r}_{2}\geq r'_{2}\geq \hat{r}_{2}$ .而且 $\tilde{r}_{2}<\infty$ .因此只有两种情况.

(ⅰ) $\tilde{r}_{2}=r'_{2}=\hat{r}_{2}$ .则系统(1.1)存在带有 $r_{2}=\tilde{r}_{2}$ 的正解 $(\tilde{u},\tilde{v},\tilde{w})$ .根据前面讨论 $(\tilde{u},\tilde{v},\tilde{w})$ 非退化, 所以由隐函数定理和最大值原理得当 $r_{2}$ 充分接近 $\tilde{r}_{2}$ 时, 在 $(\tilde{u},\tilde{v},\tilde{w})$ 附近系统(1.1)存在一直延伸到 $\tilde{r}_{2}$ 右侧的正解 $(u,v,w)$ , 这与 $\tilde{r}_{2}$ 的定义矛盾.

(ⅱ) $\tilde{r}_{2}>\hat{r}_{2}>\lambda_{1}$ .由紧性理论可知系统(1.1)存在 $r_{2}=\tilde{r}_{2}$ 的非负解 $(\breve{u},\breve{v},\breve{w})$ .若 $(\breve{u},\breve{v},\breve{w})$ 是正解, 则同(ⅰ)推出矛盾.而 $\breve{v}\geq\Theta_{\tilde{r}_{2}},\breve{w}\geq\hat{w}$ , 因此可设 $\breve{u}\equiv0$ , 则 $\breve{v}=\Theta_{\tilde{r}_{2}}$ .于是 $r_{1}=\lambda_{1}(a_{1}\Theta_{\tilde{r}_{2}})$ .然而 $r_{1}=\lambda_{1}(a_{1}\Theta_{\hat{r}_{2}})<\lambda_{1} (a_{1}\Theta_{\tilde{r}_{2}})$ 矛盾.因此 $r_{2}\geq\hat{r}_{2}$ 时系统(1.1)无正解.

4 数值模拟

本节通过数值模拟对前面的理论分析结果进行补充和说明.我们采用Matlab软件及有限差分法在一维情形 $\Omega=(0,l)$ 下模拟系统(1.1)的抛物系统.取 $l=10$ , 则 $\lambda_{1}=0.0987$ .

(1) 三物种共存与不共存. 图 1(b)(c)和(d)除某些参数值以外,其余与图 1(a)相同. (a)取 $r_{1}=2,r_{2}=1,r_{3}=2,a_{1}=0.5,a_{2}=1,a_{3}=1,c=0.5,d=0.5,e=1$ ; (b)取 $r_{2}=0.08,c=0.2$ .我们发现只要参数在一定范围内则三物种共存, 见图 1(a)(b), 这与定理1的结论吻合. (c)取 $r_{3}=0.02$ ; (d)取 $r_{1}=0.2,r_{2}=0.01,a_{2}=0.01,c=2$ .当物种的增长率充分小时 $(r_{3}<\lambda_{1}$ $r_{2}<\lambda_{1}-\frac{a_{2}r_{1}}{1+cr_{1}}<\bar{r}_{2})$ , 三物种不共存, 见图 1(c)(d).

图 1 三物种共存与不共存模拟图(a) $r_{1}=2,r_{2}=1,r_{3}=2,a_{1}=0.5,a_{2}=1,a_{3}=1,c=0.5,d=0.5,e=1$ ; (b) $r_{2}=0.08,c=0.2$ ; (c) $r_{3}=0.02$ ; (d) $r_{1}=0.2,r_{2}=0.01,a_{2}=0.01,c=2$

(2) 参数 $c$ 对正解的影响. 图 2(b)(c)和(d)除了某些参数值以外, 其余与图 2(a)相同. (a)取 $r_{1}=0.101,r_{2}=0.08,r_{3}=2,a_{1}=0.5,a_{2}=20,a_{3}=5,c=500,d=0.05,e=1$ ; (b)取 $c=20$ ; (c)取 $r_{1}=0.099$ ; (d)取 $r_{2}=5,c=20$ .当物种 $u$ 的增长率 $r_{1}$ 接近于 $\lambda_{1}$ 时, 对于任意 $c$ 或有界 $c$ , 只要物种 $v$ 的增长率 $r_{2}$ 满足一定条件, 则系统(1.1)最多有一个正解, 见图 2(a)-(d), 恰好与定理2和定理3的结论一致.而且, 观察发现随着 $c$ 的增大, $u$ 的浓度增大, $v$ 的浓度快速减少, $w$ 的浓度慢慢减少.

图 2 参数 $c$ 的影响(a) $r_{1}=0.101,r_{2}=0.08,r_{3}=2,a_{1}=0.5,a_{2}=20, a_{3}=5,c=500,d=0.05,e=1$ ; (b) $c=20$ ; (c) $r_{1}=0.099$ ; (d) $r_{2}=5,c=20$

(3) 参数对 $d$ 正解的影响. 图 3中的参数值除 $d$ 不同外, 其余参数值与图 1(a)相同.参数 $d$ 的取值见图 3.模拟显示随着 $d$ 的增大, 物种 $u$ 的浓度慢慢增大, 而物种 $v$ $w$ 的浓度减少.

图 3 参数 $d$ 的影响(a) $d=0.5$ ; (b) $d=1$ ; (c) $d=2$ ; (d) $d=5$
参考文献
[1] Wang M X, Qiang W. Positive solutions of a prey-predator model with predator saturation and competition. Journal of Mathematical Analysis and Applications, 2008, 345(2): 708–718. DOI:10.1016/j.jmaa.2008.04.054
[2] Wei M H, Wu J H, Guo G H. The effect of predator competition on positive solutions for a predator-prey model with diffusion. Nonlinear Analysis, 2012, 75(14): 5053–5068.
[3] Dong Q L, Ma W B, Sun M J. The asymptotic behavior of a Chemostat model with Crowley-Martin type functional response and time delays. Journal of Mathematical Chemistry, 2013, 51(5): 1231–1248. DOI:10.1007/s10910-012-0138-z
[4] Zhou J. Positive solutions of a diffusive Leslie-Gower predator-prey model with Bazykin functional response. Zeitschrift für Angewandte Mathematik und Physik, 2014, 65(1): 1–18. DOI:10.1007/s00033-013-0315-3
[5] Li H X. Asymptotic behavior and multiplicity for a diffusive Leslie-Gower predator-prey system with Crowley-Martin functional response. Computers and Mathematics with Applications, 2014, 68(7): 693–705. DOI:10.1016/j.camwa.2014.07.018
[6] Tripathi J P, Tyagi S, Abbas S. Global analysis of a delayed density dependent predator-prey model with Crowley-Martin functional response. Communications in Nonlinear Science and Numerical Simulation, 2016, 30(1/3): 45–69.
[7] Li S B, Wu J H. Qualitative analysis of a predator-prey model with predator saturation and competition. Acta Applicandae Mathematicae, 2016, 141(1): 165–185. DOI:10.1007/s10440-015-0009-2
[8] Du Y H, Shi J P. A diffusive predator-prey model with a protection zone. Journal of Differential Equations, 2006, 229: 63–91. DOI:10.1016/j.jde.2006.01.013
[9] Peng R, Shi J P. Non-existence of non-constant positive steady states of two Holling type-Ⅱ predator-prey systems:strong interaction case. Journal of Differential Equations, 2009, 247: 866–886. DOI:10.1016/j.jde.2009.03.008
[10] Li Y L, Wu J H. Convergence of solutions for Volterra-Lotka prey-predator systems with time delays. Applied Mathematics Letters, 2009, 22: 170–174. DOI:10.1016/j.aml.2008.03.005
[11] Jia Y F, Xu H K, Agarwal R P. Existence of positive solutions for a prey-predator model with refuge and diffusion. Applied Mathematics and Computation, 2011, 217: 8264–8276. DOI:10.1016/j.amc.2011.02.111
[12] Xiao Q Z, Dai B X, Xu B X, Bao L S. Homoclinic bifurcation for a general state-dependent Kolmogorov type predator-prey model with harvesting. Nonlinear Analysis:Real World Applications, 2015, 26: 263–273. DOI:10.1016/j.nonrwa.2015.05.012
[13] 陈滨, 王明新. 一类三种群捕食模型的正解. 数学物理学报, 2008, 28A(6): 1256–1266.
A(6):1256-1266 Chen B, Wang M X. Positive solutions to a three-species predator-prey model. Acta Mathematica Scientia, 2008, 28A(6): 1256–1266.
[14] Ko W, Ryu K. Analysis of diffusive two-competing-prey and one-predator systems with Beddington-DeAngelis functional response. Nonlinear Analysis:Theory, Methods and Applications, 2009, 71(9): 4185–4202. DOI:10.1016/j.na.2009.02.119
[15] Xie Z F. Cross-diffusion induced Turing instability for a three species food chain model. Journal of Mathematical Analysis and Applications, 2012, 388(1): 539–547. DOI:10.1016/j.jmaa.2011.10.054
[16] Yang W B, Li Y L, Wu J H, Li H X. Dynamics of a food chain model with ratio-dependent and modified Leslie-Gower functional responses. Discrete and Continuous Dynamical Systems Series B, 2015, 20(7): 2269–2290. DOI:10.3934/dcdsb
[17] Dawit H, Xie Z F. Long-time behavior and Turing instability induced by cross-diffusion in a three species food chain model with a Holling type-Ⅱ functional response. Mathematical Biosciences, 2015, 267: 134–148. DOI:10.1016/j.mbs.2015.07.001
[18] Pao C V. Dynamics of food-chain models with density-dependent diffusion and ratio-dependent reaction function. Journal of Mathematical Analysis and Applications, 2016, 433(1): 355–374. DOI:10.1016/j.jmaa.2015.05.075
[19] Ma Z P, Yue J L. Competitive exclusion and coexistence of a delayed reaction-diffusion system modeling two predators competing for one prey. Computers and Mathematics with Applications, 2016, 71(9): 1799–1817. DOI:10.1016/j.camwa.2016.02.025
[20] 李海侠. 一类带有比率依赖型反应函数的捕食-食饵模型正解的存在性和多重性. 浙江大学学报(理学版), 2016, 43(2): 156–163.
LI H X. The existence and multiplicity of positive solutions for a predator-prey model with ratio-dependent type functional response. Journal of Zhejiang University (Science Edition), 2016, 43(2): 156–163. DOI:10.3785/j.issn.1008-9497.2016.02.006
[21] Li H X, Li Y L, Yang W B. Existence and asymptotic behavior of positive solutions for a one-prey and two-competing-predators system with diffusion. Nonlinear Analysis:Real World Applications, 2016, 27: 261–282. DOI:10.1016/j.nonrwa.2015.07.010
[22] Dancer E N. On the indices of fixed points of mapping in cones and applications. Journal of Mathematical Analysis and Applications, 1983, 91(1): 131–151. DOI:10.1016/0022-247X(83)90098-7