数学物理学报, 2024, 44(1): 140-159

具有非局部效应的时滞SEIR系统的周期行波解

张广新,, 杨赟瑞,*, 宋雪,

兰州交通大学数理学院 兰州 730070

Periodic Traveling Wave Solutions of Delayed SEIR Systems with Nonlocal Effects

Zhang Guangxin,, Yang Yunrui,*, Song Xue,

School of Mathematics and Physics, Lanzhou Jiaotong University, Lanzhou 730070

通讯作者: 杨赟瑞, E-mail:lily1979101@163.com

收稿日期: 2023-02-6   修回日期: 2023-08-16  

基金资助: 国家自然科学基金(12361038)
国家自然科学基金(117610468)
甘肃省自然科学基金(20JR5RA411)

Received: 2023-02-6   Revised: 2023-08-16  

Fund supported: National Natural Science Foundation of China(12361038)
National Natural Science Foundation of China(117610468)
Natural Science Foundation of Gansu Province(20JR5RA411)

作者简介 About authors

张广新,E-mail:z13526096865@163.com

宋雪,E-mail:sx18604126839@163.com

摘要

该文研究了一类具有非局部效应和非线性发生率的时滞 SEIR 系统的周期行波解. 首先, 定义基本再生数$\Re_{0}$并构造适当的上下解, 将周期行波解的存在性转化为闭凸集上非单调算子的不动点问题, 利用 Schauder 不动点定理结合极限理论建立该系统周期行波解的存在性. 其次, 利用反证法结合比较原理, 建立当基本再生数$\Re_{0} <1$时该系统周期行波解的不存在性.

关键词: 周期行波解; 不动点定理; 存在性

Abstract

In this paper, the periodic traveling wave solutions to a class of delayed SEIR systems with nonlocal effects and nonlinear incidence are investigated. Firstly, the existence of periodic traveling waves is transformed into the fixed point problem of an non-monotone operator defined on a closed convex set by defining the basic reproducing number$\Re_{0}$and constructing appropriate upper and lower solutions, and thus the existence of periodic traveling waves of the system is established by using Schauder fixed point theorem and limit theory. Secondly, the non-existence of periodic traveling wave solutions of the system is proved when the basic regeneration number$\Re_{0}<1$by contradictory arguments and comparison principle.

Keywords: Periodic traveling wave solutions; Fixed point theorem; Existence

PDF (660KB) 元数据 多维度评价 相关文章 导出 EndNote| Ris| Bibtex  收藏本文

本文引用格式

张广新, 杨赟瑞, 宋雪. 具有非局部效应的时滞SEIR系统的周期行波解[J]. 数学物理学报, 2024, 44(1): 140-159

Zhang Guangxin, Yang Yunrui, Song Xue. Periodic Traveling Wave Solutions of Delayed SEIR Systems with Nonlocal Effects[J]. Acta Mathematica Scientia, 2024, 44(1): 140-159

1 引言

反应扩散方程作为非线性抛物型方程的一类重要模型, 在物理、生物和传染病学领域有着广泛的应用[1-5], 而行波解作为其稳态解, 常用来描述自然界中的诸如传染病的传播、生物种群的入侵等传播现象[4-7]. 然而, 传染病的传播、生物的生长还会受气候、温度、湿度、季节等周期性因素的影响, 例如新冠病毒喜冷, 夏季时高温使病毒活性下降、冬季时低温使病毒长期存活. 因此, 对具有时间周期性的反应扩散方程时间周期行波解 (以下简称周期行波解) 的研究更具理论意义和实际价值[6-8]. 例如, Zhang 等[6]研究了经典 Laplace 扩散的 SIR 系统

$\begin{cases} \frac{\partial S(t,x)}{\partial t}=d_{1}\Delta S(t,x)-\beta (t)S(t,x)I(t,x), \\[3mm] \frac{\partial I(t,x)}{\partial t}=d_{2}\Delta I(t,x)+\beta (t)S(t,x)I(t,x)-\gamma (t)I(t,x),\\[3mm] \frac{\partial R(t,x)}{\partial t}=d_{3}\Delta R(t,x)+\gamma (t)I(t,x)\end{cases}$

周期行波解的 (不) 存在性, 其中$S(t,x),I(t,x),R(t,x)$分别表示易感者, 感染者和治愈者的密度,$\beta(t)$和$\gamma(t)$分别表示感染率和恢复率.

时间滞后 (简称时滞) 的现象普遍存在[7,10]. 例如, 人感染病毒后过一段时间才会出现症状, 这就是疾病的潜伏期. 此外, 个体的发展不只与当前位置和时间的扩散有关、还与其它位置和时间以及与其他个体产生的相互作用有关, 这就是非局部效应. 为此, 2020 年, Wang 等[7]将 Zhang 等[6]的工作推广到一类具有非局部效应的时滞 SEIR 系统

$\begin{cases}\frac{\partial S(t,x)}{\partial t}=d_{1}\Delta S(t,x)-\beta (t)S(t,x)I(t,x), \\[3mm]\frac{\partial E(t,x)}{\partial t}=d_{L}\Delta E(t,x)-\gamma_{L} (t)E(t,x)+\beta (t)S(t,x)I(t,x)\\[3mm] -\int_{\mathbb{R}} \Gamma(t,t-\tau;x-y)\beta(t-\tau)S(t-\tau,y)I(t-\tau,y){\rm d}y,\\[3mm]\frac{\partial I(t,x)}{\partial t}=d_{2}\Delta I(t,x)-\gamma(t)I(t,x)\\[3mm] +\int_{\mathbb{R}} \Gamma(t,t-\tau;x-y)\beta(t-\tau)S(t-\tau,y)I(t-\tau,y){\rm d}y,\\[3mm]\frac{\partial R(t,x)}{\partial t}=d_{R}\Delta R(t,x)+\gamma_{L} (t)E(t,x)+\gamma(t)I(t,x),\end{cases}$

利用上下解结合 Schauder 不动点定理建立了系统 (1.2) 周期行波解的 (不) 存在性. 注意到, 上述模型[6-7]中的发生率都是双线性发生率, 而更一般的非线性函数能反映个体之间的其它相互作用类型 (例如标准发生率)[8-9]. 2021 年, Wu 等[6]借助上下解和 Schauder 不动点定理建立了具有非线性发生率的无时滞 SIR 系统

$\begin{cases}\frac{\partial S(t,x)}{\partial t}=d_{1}\Delta S(t,x)-\beta (t)f(S(t,x),I(t,x)), \\[3mm]\frac{\partial I(t,x)}{\partial t}=d_{2}\Delta I(t,x)+\beta (t)f(S(t,x),I(t,x))-\gamma (t)I(t,x),\\[3mm]\frac{\partial R(t,x)}{\partial t}=d_{3}\Delta R(t,x)+\gamma (t)I(t,x)\end{cases}$

周期行波解的(不)存在性. 基于此, 受Wang等[7]和Wu等[8]工作的启发, 考虑时滞和非局部效应的共同影响, 本文研究具有非局部效应和非线性发生率的时滞 SEIR 系统

$\begin{equation}\frac{\partial S(t,x)}{\partial t}=d_{1}\Delta S(t,x)-\beta (t)f(S(t,x),I(t,x)), \\[3mm] \frac{\partial E(t,x)}{\partial t}=d_{L}\Delta E(t,x)-\gamma_{L} (t)E(t,x)+\beta (t)f(S(t,x),I(t,x))\\[3mm] \int_{\mathbb{R}} \Gamma(t,t-\tau;x-y)\beta(t-\tau)f(S(t-\tau,y),I(t-\tau,y)){\rm d}y,\\[3mm] \frac{\partial I(t,x)}{\partial t}=d_{2}\Delta I(t,x)-\gamma(t)I(t,x)\\[3mm] +\int_{\mathbb{R}} \Gamma(t,t-\tau;x-y)\beta(t-\tau)f(S(t-\tau,y),I(t-\tau,y)){\rm d}y,\\[3mm] \frac{\partial R(t,x)}{\partial t}=d_{R}\Delta R(t,x)+\gamma_{L} (t)E(t,x)+\gamma(t)I(t,x)\end{equation}$

的周期行波解, 通过构造适当的上下解, 结合 Schauder 不动点定理和极限理论建立系统 (1.4) 周期行波解的存在性, 利用反证法和比较原理建立系统 (1.4) 周期行波解的不存在性. 与系统 (1.1) 相比,$E(t,x)$表示携带者的密度, 核函数

$\Gamma(t, t-\tau ; x-y)={\rm e}^{-\int_{t-\tau}^{t} \gamma_{L}(s){\rm d}s} \frac{1}{\sqrt{4 \pi d_{L} \tau}} {\rm e}^{-\frac{(x-y)^{2}}{4 d_{L} \tau}}.$

注意到, 当$f(S(t, x), I(t, x))=S(t, x) I(t, x)$时, 系统 (1.4) 退化为系统 (1.2). 因此, 本文推广了 Wang 等[7]对具有双线性发生率的系统 (1.2) 和 Wu 等[8]对无时滞系统 (1.3) 的研究, 完善了非线性发生率的传染病系统周期行波解的研究结果. 另外, 在建立系统 (1.4) 周期行波解的存在性时, 由于时间周期性和时滞的同时出现, 常用的下一代矩阵法不再适用, 为此借助次代算子法定义基本再生数$\Re_{0}$. 而且, 由于 Schauder 不动点定理不能直接应用于周期系统, 为了克服这一缺陷, 需要定义具有周期性的闭凸集并在其上定义非单调算子, 从而将周期行波解的存在性转化为非单调算子的不动点问题, 再利用上下解结合不动点定理得到系统 (1.4) 周期行波解的存在性.

下面给出本文用到的假设条件

(A$_{1})$$f(S,I)$对$S,I\geq0$是二阶连续可微的, 并且$f(0, I)=f(S, 0)=0,(S, I>0)$.

(A$_{2})$当$S>0,I>0$时,$\partial_{2} f(S, I)=\frac{\partial f(S,I)}{\partial I}>0, \partial_{22} f(S, I)=\frac{\partial^{2} f(S,I)}{\partial I^{2}} \leq 0$; 当$I\geq0$时,$\partial_{1} f(S, I)=\frac{\partial f(S,I)}{\partial S}>0$且对所有$S\geq0$有界.

因为系统 (1.4) 中的第二个和第四个方程相对独立, 故仅需考虑系统 (1.4) 的子系统

$\begin{cases}\frac{\partial S(t,x)}{\partial t}=d_{1}\Delta S(t,x)-\beta (t)f(S(t,x),I(t,x)), \\[3mm]\frac{\partial I(t,x)}{\partial t}=d_{2}\Delta I(t,x)-\gamma(t)I(t,x)\\[3mm] +\int_{\mathbb{R}} \Gamma(t,t-\tau;x-y)\beta(t-\tau)f(S(t-\tau,y),I(t-\tau,y)){\rm d}y,\end{cases}$

系统 (1.5) 的周期行波解是指形如

$\begin{equation}(S(t,x),I(t,x)):=(\phi(t, z), \psi(t, z))=(\phi(t, x+c t), \psi(t, x+c t)), \quad c>0.\end{equation}$

(1.5) 式的解. 对任意的$(t,z)\in\mathbb{R}^2$, 满足

$(\phi(t+T,z),\psi(t+T,z))\text{=}(\phi(t,z),\psi(t,z)),$
$(\phi(t,\pm\infty),\psi(t,\pm\infty)):=\lim\limits_{z\to\pm\infty}(\phi(t,z),\psi(t,z))=(\phi_\pm(t,z),\psi_\pm(t,z)),$

这里$(\phi_+(t,z),\psi_+(t,z))$和$(\phi_-(t,z),\psi_-(t,z))$是系统 (1.5) 相应 Kinetic 系统

$\begin{cases}\frac{\partial S(t)}{\partial t}=-\beta (t)f(S(t),I(t)), \\[3mm]\frac{\partial I(t)}{\partial t}=-\gamma(t)I(t)+{\rm e}^{-\int_{t-\tau}^t\gamma_L(s){\rm d}s}\beta(t-\tau)f(S(t-\tau),I(t-\tau))\end{cases}$

的周期解. 将 (1.6) 式代入 (1.5) 式, 可得相应 (1.5) 式的行波系统为

$\begin{cases}\phi_{t}(t,z)=d_{1}\phi_{zz}(t,z)-c\phi_{z}(t,z)-\beta (t)f(\phi(t,z),\psi(t,z)), \\\psi_{t}(t,z)=d_{2}\psi_{zz}(t,z)-c\psi_{z}(t,z)-\gamma(t)\psi(t,z)\\[2mm]+\int_{\mathbb{R}} \Gamma(t,t-\tau;x-y)\beta(t-\tau)f(\phi(t-\tau,z-c\tau-y),\psi(t-\tau,z-c\tau-y)){\rm d}y.\end{cases}$

因此, 系统 (1.5) 的周期行波解也就是研究系统 (1.8) 满足渐近边界条件

$\begin{equation} \phi(t,-\infty)=S_0>0,\:\:\phi(t,+\infty)=S^\infty>0,\:\:\psi(t,\pm\infty)=0,t\in\mathbb{R}\end{equation}$

的周期解, 其中$S_0$和$S^\infty$分别是疾病开始传播时易感个体的数量和疾病传播后易感个体的数量.

2 周期行波解的存在性

本节建立系统 (1.4) 周期行波解的存在性. 首先借助 Zhao[10] 提出的周期时滞模型基本再生数理论, 通过次代算子的方法定义基本再生数并构造 (1.8) 式的上下解, 然后在截断区间上定义闭凸集, 将周期行波解的存在性转化为定义在这个闭凸集上的非单调算子的不动点问题, 再利用 Schauder 不动点定理结合极限理论建立系统 (1.4) 周期行波解的存在性.

2.1 基本再生数

记$Y=C\left([-\tau,0],\mathbb{R}\right),\:Y^+=C\left([-\tau,0],\mathbb{R}_+\right).$令$C_T\subset C(\mathbb{R})$表示所有$T$周期连续函数组成的集合, 因此,$C_T$是有序 Banach 空间.

将 (1.7) 式的第二个方程在无病平衡点$(S_0,0)$处线性化得

$\begin{equation} \frac{{\rm d} I(t)}{{\rm d}t}=-\gamma(t)I(t)+{\rm e}^{-\int_{t-\tau}^{t}\gamma_{L}(s){\rm d}s}\beta(t-\tau)\partial_{2}f(S_{0},0)I(t-\tau).\end{equation}$

记$L(t,s)\:=\:{\rm e}^{-\int_s^t\gamma(r){\rm d}r},(t,s)\:\in\:\mathbb{R}^2, \:\:t\geq s.\:\:\vartheta\:\in\:C_T\:$表示感染个体的初始分布, 定义乘积算子$F(t)\text{:}C_T\rightarrow C_T$,

$\begin{equation} F(t)\omega={\rm e}^{-\int_{t-\tau}^{t}\gamma_{L}(r){\rm d}r}\beta(t-\tau)\partial_{2}f(S_{0},0)\omega(-\tau), \omega\in C_{T}.\end{equation}$

对任意给定的$s\geq0$,$F(t-s)\vartheta_{t-s}$表示由$[t-s-\tau,t-s]$这段时间内被感染个体产生的$t-s$时刻的新感染个体的数量. 从而, 对任意的$t\geq s$,$L(t,t\:-\:s)F(t\:-\:s)\vartheta_{t-s}$表示$t-s$时刻新感染个体持续到$t$时刻仍为感染个体的数量, 则$t$时刻累积新感染个体的数量为

$\int_{\tau}^{\infty}L(t,t-s+\tau){\rm e}^{-\int_{t-s}^{t-s+\tau} \gamma_{L}(r){\rm d}r}\beta(t-s)\:\vartheta(t-s){\rm d}s=\int_{0}^{\infty}L(t,t-s)F(t-s)\:\vartheta(t-s){\rm d}s.$

从而, 可以定义次代算子$\mathcal{L}$为$\begin{bmatrix}\mathcal{L}\vartheta\end{bmatrix}(t):=\int_0^\infty L(t,t\:-\:s)F(t\:-\:s)\vartheta(t\:-\:s){\rm d}s\:.$受文献[10]的启发, 定义基本再生数$\Re_{0}:=r(\mathcal{L})$, 其中$r(\mathcal{L})$为算子$\mathcal{L}$的谱半径. 本文第 2 节总假定$\Re_{0}>1$.

2.2 上下解

首先给出系统$(1.8)$的上下解定义.

定义 2.1 假设 (A$_{1})$-(A$_{2})$成立, 若$(\phi^{+}(t,z),\psi^{+}(t,z))$和$(\phi^{-}(t,z),\psi^{-}(t,z))$满足

$\begin{cases}\phi_{t}^{+}(t,z)\geq d_{1}\phi_{zz}^{+}(t,z)-c\phi_{z}^{+}(t,z)-\beta(t)f(\phi^{+}(t,z),\psi^{-}(t,z)),\\\psi_{t}^{+}(t,z)\geq d_{2}\psi_{zz}^{+}(t,z)-c\psi_{z}^{+}(t,z)-\gamma(t)\psi^{+}(t,z)+\int_{\mathbb{R}} \Gamma(t,t-\tau;y)\beta(t-\tau) \notag\\ \times f(\phi^{+} (t-\tau,z-c\tau-y),\psi^{+} (t-\tau,z-c\tau-y)){\rm d}y,\\\phi_{t} ^{-}(t,z)\leq d_{1}\phi_{zz}^{-}(t,z)-c\phi_{z}^{-}(t,z)-\beta(t)f(\phi^{-}(t,z),\psi^{+}(t,z)),\\\psi_{t}^{-}(t,z)\leq d_{2}\psi_{zz}^{-}(t,z)-c\psi_{z}^{-}(t,z)-\gamma(t)\psi^{-}(t,z)+\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau) \notag\\ \times f(\phi^{-}(t-\tau,z-c\tau-y),\psi^{-}(t-\tau,z-c\tau-y)){\rm d}y,\end{cases}$

则称$(\phi^{+}(t,z),\psi^{+}(t,z))$和$(\phi^{-}(t,z),\psi^{-}(t,z))$分别为$(1.8)$式的上解和下解.

将 (1.5) 式的第二个方程在无病平衡点$(S_0,0)$处线性化得

$\begin{equation} \frac{\partial I(t,x)}{\partial t}=d_{2}\Delta I(t,x)-\gamma(t)I(t,x)+\int_{\mathbb{R}} \Gamma(t,t-\tau;y)\beta(t-\tau)\partial_{2}f(S_{0},0)I(t-\tau,x-y){\rm d}y,\end{equation}$

则对任意的$\mu>0$, 将$I(t,x)\text{:}=\:{\rm e}^{\mu x}\eta(t)$代入 (2.3) 式中, 可得

$\begin{equation}{\rm e}^{\mu x}\eta'(t)=d_2\mu^2{\rm e}^{\mu x}\eta(t)-\gamma(t){\rm e}^{\mu x}\eta(t)+\beta(t-\tau)\partial_2f(S_0,0)\int_\mathbb{R}\Gamma(t,t-\tau;y){\rm e}^{\mu(x-y)}\eta(t-\tau){\rm d}y.\end{equation}$

又因为$\int_{\mathbb{R}}\Gamma(t,t-\tau;y){\rm e}^{\mu y}{\rm d}y={\rm e}^{-\int_{t-\tau}^t\gamma_L(s){\rm d}s}{\rm e}^{d_L\tau\mu^2}$, 所以

$\begin{equation} \eta'(t)=d_2\mu^2\eta(t)-\gamma(t)\eta(t)+\beta(t-\tau)\partial_2f(S_0,0){\rm e}^{-\int_{t-\tau}^t\gamma_L(s){\rm d}s}{\rm e}^{d_L\tau\mu^2}\eta(t-\tau).\end{equation}$

下面定义 (2.5) 式的解半流$I_t^\mu(\varphi)(s)=I(t+s;\varphi;\mu),\forall s\in[-\tau,0),t>0$, 其中$I(t;\varphi;\mu)$是初值$\varphi\in Y^{+}$时 (2.5) 式的解. 进一步, 定义$Y^{+}\rightarrow Y^{+}$的 Poincaré映射$P_{\mu}=I^{\mu}_{T}(\varphi)$, 记$\rho(\mu)$为$P_{\mu}$的谱半径,$\rho_{0}=\rho(0)$为方程 (2.1) Poincaré映射的谱半径. 根据文献 [10,定理 2.1] 知$\Re_{0}-1$与$\rho_{0}-1$符号相同. 由于$\Re_{0}>1$, 从而$\rho(\mu)>\rho_{0}>1$. 记$\lambda(\mu):=\frac{\ln\rho(\mu)}{T}$. 根据文献[11,命题 2.1] 知, 存在$T$周期函数$K_\mu(t)$使得$\eta_\mu(t)\text{=}{\rm e}^{\lambda(\mu)t}K_\mu(t)$是 (2.5) 式的一个解.

定义函数$\Phi(\mu)=\frac{\lambda(\mu)}{\mu},\mu\in(0,\infty)$, 由文献[12,引理 3.8] 可知存在$\mu^{\ast},c^{\ast}\in(0,\infty)$, 使得$c^*=\Phi(\mu^*)=\operatorname*{inf}\limits_{\mu>0}\Phi({\mu})$. 进一步, 对任意给定的$c>c^{\ast}$, 存在$\mu_{1}(c),\mu_{2}(c)$满足$0<\mu_{1}<\mu_{2}<\infty$, 使得$\Phi(\mu_{1})=c$且$\Phi(\mu)<c, \forall \mu\in(\mu_{1},\mu_{2})$. 由上面的讨论可知,$K_{\mu_1}(t)$满足线性方程

$\begin{matrix} \ &K'_{\mu_1}\big(t\big)-d_2\mu_1^2K_{\mu_1}\big(t\big)+c\mu_1K_{\mu_1}\big(t\big)+\gamma\big(t\big)K_{\mu_1}\big(t\big) \notag\\&-\beta\big(t-\tau\big)\partial_2f\big(S_0,0\big)\int_\mathbb{R}\Gamma\big(t,t-\tau;y\big){\rm e}^{-\mu_1\left(c\tau+y\right)}K_{\mu_1}\big(t-\tau\big){\rm d}y=0.\end{matrix}$

定义函数

${\phi^+}\left(t,z\right)\colon=S_0,\quad\quad \psi^+\left(t,z\right)\colon={\rm e}^{\mu_1z}K_{\mu_1}\left(t\right),$
$\phi^-\left(t,z\right)\colon=\max\left\{0,S_0\left(1-M_1{\rm e}^{\varepsilon_1z}\right)\right\},\psi^{-}\left(t,z\right):=\text{max}\left\{0,{\rm e}^{\mu_1 z}K_{\mu_1}\left(t\right)-M_2{\rm e}^{\mu_{\varepsilon_2}z}K_{\mu_{\varepsilon_2}}\left(t\right)\right\},$

其中$\varepsilon_i>0,M_i>0,i=1,2$为待定常数, 且$\varepsilon_2\in\big(0,\mu_2-\mu_1\big),\mu_{\varepsilon_2}=\mu_1+\varepsilon_2,$则有$c^*<c_{\varepsilon_2}:=\Phi(\mu_{\varepsilon_2})<c.$令$\lambda\left(\mu_{\varepsilon_2}\right)=\frac{\ln\rho\left(\mu_{\varepsilon_2}\right)}{T}$, 则存在一个正$T$周期函数$K_{\mu_{\varepsilon_2}}(t)$使得$\eta_{\mu_{\varepsilon_2}}(t)={\rm e}^{\mu_{\varepsilon_2}z}K_{\mu_{\varepsilon_2}}(t)$是 (2.5) 式的解. 因此,$K_{\mu_{\varepsilon_2}}(t)$满足

$\begin{matrix} \ &K'_{\mu_{\varepsilon_2}}(t)-d_2\mu_{\mu_{\varepsilon_2}}^2K_{\mu_{\varepsilon_2}}(t)+c_{\varepsilon_2}\mu_{\varepsilon_2}K_{\mu_{\varepsilon_2}}(t)+\gamma(t)K_{\mu_{\varepsilon_2}}\big(t\big) \notag\\&-\beta(t-\tau)\partial_2f(S_0,0)\int_\mathbb{R}\Gamma(t,t-\tau;y){\rm e}^{-\mu_{\varepsilon_2}\left(c_{\varepsilon_2}\tau+y\right)}K_{\mu_{\varepsilon_2}}(t-\tau){\rm d}y=0.\end{matrix}$

引理 2.1 函数$\phi^{+}(t,z)=S_{0}$满足

$\begin{matrix} \phi_{t}^{+}(t,z)-d_{1}\phi_{zz}^{+}(t,z)+c\phi_{z}^{+}(t,z)+\beta(t)f(\phi^{+}(t,z),\psi^{-}(t,z))\geq0.\end{matrix}$

由条件 (A$_{1})$可知,$f(\phi^{+}(t,z),\psi^{-}(t,z))\geq0$, 从而 (2.8) 式显然成立. 证毕.

引理 2.2 假设$0<\varepsilon_{1} <\min\{\mu_{1},\frac{c}{d_{1}}\}$充分小且$M_{1} >1$充分大, 则对任意的$z\neq z_{1} :=-\frac{1}{\varepsilon_{1}}\ln M_{1}$, 函数$\phi^{-} (t,z)$满足

$\begin{matrix} \phi_{t} ^{-}(t,z)-d_{1}\phi_{zz}^{-}(t,z)+c\phi_{z}^{-}(t,z)+ \beta(t)f(\phi^{-}(t,z),\psi^{+}(t,z))\leq0.\end{matrix}$

证 若$z>-\frac{1}{\varepsilon_{1}}\ln M_{1}$, 则$\phi^{-} (t,z)=0$, 从而 (2.9) 式成立. 若$z<-\frac{1}{\varepsilon_{1}}\ln M_{1}$, 则$\phi^-(t,z)=S_0(1-M_1{\rm e}^{\varepsilon_1z})$且$\psi^+\left(t,z\right)={\rm e}^{\mu_1z}K_{\mu_1}\left(t\right).$此时, (2.9) 式等价于

$\begin{matrix} \ S_0M_1\varepsilon_1(c-d_1\varepsilon_1)\geq\frac{\beta(t)f(S_0(1-M_1{\rm e}^{\varepsilon_1z}),{\rm e}^{\mu_1z}K_{\mu_1}(t))}{{\rm e}^{\varepsilon_1z}},\end{matrix}$

由条件 (A$_{1})$和 (A$_{2})$可知,$f(S_0(1-M_1{\rm e}^{\varepsilon_1z}),{\rm e}^{\mu_1z}K_{\mu_1}(t))<f(S_0,{\rm e}^{\mu_1z}K_{\mu_1}(t))$, 再利用微分中值定理可得$f(S_0,{\rm e}^{\mu_1z}K_{\mu_1}(t))\leq\partial_{2} f(S_0,0){\rm e}^{\mu_1z}K_{\mu_1}(t)$. 又因为$0<\varepsilon_{1}<\mu_{1}$, 从而

$\begin{eqnarray*} \frac{f\left(S_{0},{\rm e}^{\mu_{1}z}K_{\mu_{1}}\left(t\right)\right)}{{\rm e}^{\varepsilon_{1}z}}\leq\frac{\partial_{2}f\left(S_{0},0\right){\rm e}^{\mu_{1}z}K_{\mu_{1}}\left(t\right)}{{\rm e}^{\varepsilon_{1}z}}\leq\partial_{2}f\left(S_{0},0\right)K_{\mu_{1}}\left(t\right).\end{eqnarray*}$

为了证明 (2.10) 式, 只需证明

$\begin{matrix} \ S_0M_1\varepsilon_1(c-d_1\varepsilon_1)\geq\beta(t)\partial_2f(S_0,0)K_{\mu_1}(t).\end{matrix}$

注意到$\beta(t),K_{\mu_1}(t)$是正周期函数, 取$0<\varepsilon_1<\min\big\{\mu_1,\frac{c}{d_1}\big\}$充分小且$M_{1}>1$充分大, 则 (2.11) 式成立. 证毕.

引理 2.3 函数$\psi^{+}(t,z)$满足

$\begin{matrix} &\psi_{t}^{+}(t,z)-d_{2}\psi_{zz}^{+}(t,z)+c\psi_{z}^{+}(t,z)+\gamma(t)\psi^{+}(t,z)-\int_{\mathbb{R}} \Gamma(t,t-\tau;y)\beta(t-\tau) \notag\\&\times f(\phi^{+} (t-\tau,z-c\tau-y),\psi^{+} (t-\tau,z-c\tau-y)){\rm d}y\geq0.\end{matrix}$

由$\psi^+\left(t,z\right)={\rm e}^{\mu_1z}K_{\mu_1}\left(t\right)$和$K'_{\mu_1}(t)$的定义式可得

$\begin{align} \psi_t^+\left(t,z\right)={\rm e}^{\mu_1z}K_{\mu_1}'\left(t\right)&=d_2\mu_1^2K_{\mu_1}\left(t\right){\rm e}^{\mu_1z}-c\mu_1K_{\mu_1}\left(t\right){\rm e}^{\mu_1z}-\gamma\left(t\right)K_{\mu_1}\left(t\right){\rm e}^{\mu_1z} \notag\\& \,+\beta(t-\tau)\partial_{2}f(S_0,0)\int_\mathbb{R}\Gamma(t,t-\tau;y){\rm e}^{\mu_1(z-c\tau-y)}K_{\mu_1}(t-\tau){\rm d}y. \notag\end{align}$

由微分中值定理和条件 (A$_{2})$可知

$f(\phi^+(t-\tau,z-c\tau-y),\psi^+(t-\tau,z-c\tau-y))\leq\partial_2f(S_0,0)\psi^+(t-\tau, z-c\tau-y)).$

所以

$\begin{align}\psi_t^+\left(t,z\right)&=d_2\psi_{zz}^+\left(t,z\right)-c\psi_z^+\left(t,z\right)-\gamma\left(t\right)\psi^+\left(t,z\right) \notag\\& +\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)\partial_2f(S_0,0)\psi^+(t-\tau,z-c\tau-y){\rm d}y \notag\\&\geq d_2\psi_{zz}^+\left(t,z\right)-c\psi_z^+\left(t,z\right)-\gamma\left(t\right)\psi^+\left(t,z\right) \notag\\& +\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\phi^+(t-\tau,z-c\tau-y),\psi^+(t-\tau,z-c\tau-y)){\rm d}y. \notag\end{align}$

从而 (2.12) 式成立. 证毕.

引理 2.4 假设$0<\varepsilon_{2}<\min\{\varepsilon_{1},\mu_{2}-\mu_{1}\}$充分小以及满足$\max\limits_{t\in [T]}\frac{1}{\varepsilon_{2}}\ln\frac{K_{\mu_{1}}(t)}{M_{2}K_{\mu_{\varepsilon_{2}}} (t)} <-\frac{1}{\varepsilon_{1}} \ln M_{1}$的正数$M_{2}$充分大, 则当$z\neq z_{2}(t):=\frac{1}{\varepsilon_{2}}\ln\frac{K_{\mu_{1}}(t)}{M_{2} K_{\mu_{\varepsilon_{2}}}(t)}$, 函数$\psi^{-}(t,z)$满足

$\begin{matrix}&\psi_{t}^{-}(t,z)-d_{2}\psi_{zz}^{-}(t,z)+c\psi_{z}^{-}(t,z)+\gamma(t)\psi^{-}(t,z)-\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau) \notag\\&\times f(\phi^{-}(t-\tau,z-c\tau-y),\psi^{-}(t-\tau,z-c\tau-y)){\rm d}y\leq0.\end{matrix}$

若$z>z_{2}(t)$, 则$\psi^{-}(t,z)=0$, 此时, (2.13) 式显然成立. 若$z<z_{2}(t)<z_{1}$, 则

$\phi^-(t,z)=S_0(1-M_1{\rm e}^{\varepsilon_1z}), \psi^{-}(t,z)={\rm e}^{\mu_1 z}K_{\mu_1}(t)-M_2{\rm e}^{\mu_{\varepsilon_2}z}K_{\mu_{\varepsilon_2}}(t).$

此时, (2.13) 式等价于

$\begin{matrix}\ & \psi_t^{-}(t,z)-d_2\psi_{zz}^{-}(t,z)+c\psi_z^{-}(t,z)+\gamma(t)\psi^{-}(t,z) \notag\\& -\int_\mathbb{R}\Gamma(t,t-\tau;y)\beta\big(t-\tau)\partial_2f(S_0,0)\psi^-(t-\tau,z-c\tau-y){\rm d}y \notag\\&\leq\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\phi^{-}(t-\tau,z-c\tau-y),\psi^{-}(t-\tau,z-c\tau-y)){\rm d}y \notag\\& -\int_\mathbb{R}\Gamma(t,t-\tau;y)\beta(t-\tau)\partial_2f(S_0,0)\psi^-(t-\tau,z-c\tau-y){\rm d}y.\end{matrix}$

根据$K'_{\mu_1}(t)$和$K'_{\mu_{\varepsilon_{2}}}(t)$的定义式, 计算 (2.14) 式的左边, 可得

$\begin{matrix}\ &\psi_t^{-}(t,z)-d_2\psi_{zz}^{-}(t,z)+c\psi_z^{-}(t,z)+\gamma(t)\psi^{-}(t,z) \notag\\&-\int_\mathbb{R}\Gamma(t,t-\tau;y)\beta(t-\tau)\partial_2f(S_0,0)\psi^-(t-\tau,z-c\tau-y){\rm d}y \notag\\=\ &{\rm e}^{\mu_{1}z}\Big[K_{\mu_1}^{\prime}(t)-d_2\mu_1^2K_{\mu_1}(t)+c\mu_1K_{\mu_1}(t)+\gamma(t)K_{\mu_1}(t)-\beta(t-\tau)\int_{\mathbb{R}}\Gamma(t,t-\tau;y) \notag\\&\times\partial_2f(S_0,0){\rm e}^{-\mu_1(c\tau+y)}K_{\mu_1}(t-\tau){\rm d}y\Big]-M_2{\rm e}^{\mu_{\varepsilon_2}z}\Big[K_{\mu_{\varepsilon_{2}}}^{\prime}(t)-d_{2}\mu_{\varepsilon_{2}}^{2}K_{\mu_{\varepsilon_{2}}}^{}(t)+c\mu_{\varepsilon_{2}}K_{\mu_{\varepsilon_{2}}}^{}(t) \notag\\&+\gamma(t)K_{\mu_{\varepsilon_{2}}}\big(t)-\beta(t-\tau)\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\partial_{_2}f(S_{_0},0){\rm e}^{-\mu_{\varepsilon_2}(c\tau+y)}K_{\mu_{\varepsilon_{2}}}(t-\tau){\rm d}y\Big] \notag\\=\ &{\rm e}^{\mu_{1}z}\Big[d_{2}\mu_{1}^{2}K_{\mu_{1}}(t)-c\mu_{1}K_{\mu_{1}}(t)-\gamma(t)K_{\mu_{1}}(t)+\beta(t-\tau)\int_{\mathbb{R}}\Gamma(t,t-\tau;y) \\&\times\partial_{2}f(S_{0},0){\rm e}^{-\mu_{1}(c\tau+y)} K_{\mu_1}\left(t-\tau\right){\rm d}y-d_2\mu_1^2K_{\mu_1}\left(t\right)+c\mu_1K_{\mu_1}\left(t\right)+\gamma\bigl(t\bigr)K_{\mu_1}\bigl(t\bigr) \\&-\beta\bigl(t-\tau\bigr)\int_{\mathbb{R}}\Gamma\bigl(t,t-\tau;y) \partial_2f(S_0,0){\rm e}^{-\mu_1(c\tau+y)}K_{\mu_1}\big(t-\tau\big){\rm d}y\Big] \\&-M_2{\rm e}^{\mu_{\varepsilon_2}z}\Big[d_{2}\mu_{\varepsilon_{2}}^{2}K_{\mu_{\varepsilon_{2}}}(t)-c_{\varepsilon_{2}}\mu_{\varepsilon_{2}}K_{\mu_{\varepsilon_{2}}}(t)-\gamma(t) K_{\mu_{\varepsilon_{2}}}(t)+\beta(t-\tau)\int_\mathbb{R}\Gamma\text{}(t,t-\tau;y) \\&\times \partial_2f(S_0,0){\rm e}^{-\mu_{\varepsilon_{2}}\left(c_{\varepsilon_{2}}\tau+y\right)}K_{\mu_{\varepsilon_{2}}}\left(t-\tau\right){\rm d}y-d_{2}\mu_{\varepsilon_{2}}^{2}K_{\mu_{\varepsilon_{2}}}\left(t\right)+c\mu_{\varepsilon_{2}}K_{\mu_{\varepsilon_{2}}}(t) \\&+\gamma(t)K_{\mu_{\varepsilon_{2}}}(t)-\beta(t-\tau)\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\partial_2f\left(S_0,0\right){\rm e}^{-\mu_{\varepsilon_2}\left(c\tau+y\right)}K_{\mu_{\varepsilon_2}}\left(t-\tau\right){\rm d}y\Big] \notag\\:=&-M_2{\rm e}^{\mu_{\varepsilon_2}z}\tilde{H}(t),\end{matrix}$

其中$\tilde{H}(t):=(c\!-\!c_{\varepsilon_2})\mu_{\varepsilon_2}K_{\mu_{\varepsilon_2}}(t)-\beta(t-\tau)\partial_2f\big(S_0,0)({\rm e}^{-\mu_{\varepsilon_2}c\tau}\!-\!{\rm e}^{-\mu_{\varepsilon_2}c_{\varepsilon_2}\tau})\int_{\mathbb{R}}\Gamma(t,t-\tau;y){\rm e}^{-\mu_{\varepsilon_2}y} \times K_{\mu_{\varepsilon_2}}(t-\tau){\rm d}y$ 且 $\tilde{H}(t)>0$, 另一方面, 计算 (2.14) 式的右边, 可得

$\begin{matrix}\ & \int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\phi^{-}(t-\tau,z-c\tau-y),\psi^{-}(t-\tau,z-c\tau-y)){\rm d}y \notag\\& -\int_\mathbb{R}\Gamma(t,t-\tau;y)\beta(t-\tau)\partial_2f(S_0,0)\psi^-(t-\tau,z-c\tau-y){\rm d}y \notag\\&=\beta(t-\tau)\int_{\mathbb{R}}\Gamma(t,t-\tau;y)f(\phi^{-},\psi^{-})\text{}{\rm d}y-\beta(t-\tau)\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\partial_2f(S_0,0){\rm e}^{\mu_1\left(z-c\tau-y\right)}\notag\\& \times K_{\mu_1}(t-\tau){\rm d}y+M_2\beta(t-\tau)\int_\mathbb{R}\Gamma(t,t-\tau;y)\partial_2f(S_0,0){\rm e}^{\mu_{\varepsilon_2}\left(z-c\tau-y\right)}K_{\mu_{\varepsilon_2}}\left(t-\tau\right){\rm d}y \notag\\&\geq-\beta(t-\tau)\int_\mathbb{R}\Gamma(t,t-\tau;y)\partial_2f(S_0,0){\rm e}^{\mu_1(z-c\tau-y)}K_{\mu_1}(t-\tau){\rm d}y.\end{matrix}$

因此, 由 (2.15) 和 (2.16) 式, 只需证明

$\begin{matrix}\ M_2{\rm e}^{\mu_{\varepsilon_2}z}\tilde{H}(t)\geq\beta(t-\tau)\int_\mathbb{R}\Gamma(t,t-\tau;y)\partial_2f(S_0,0){\rm e}^{\mu_1(z-c\tau-y)}K_{\mu_1}(t-\tau){\rm d}y.\end{matrix}$

因为$\beta(t),K_{_{\mu_1}}(t),\Gamma(t,t-\tau;\cdot)$是正周期函数, 取$M_{2}$充分大时, 则 (2.17) 式成立, 故 (2.13) 式成立. 证毕.

2.3 有界区间上的不动点问题

令$z^{\ast}=\max\limits_{t\in [T]}(-z_{2}(t))$, 其中$z_{2}(t)=\frac{1}{\varepsilon_{2}}\ln\frac{K_{\mu_{1}}(t)}{M_{2} K_{\mu_{\varepsilon_{2}}}(t)}(z_{2}(t)<0)$. 对任意的$l>0$, 设$C_{l}:=C(\mathbb{R}\times[-l,l],\mathbb{R}^{2})$. 取$l>z^{\ast}$, 定义集合

$\begin{eqnarray*}\Gamma_{l}:= \left\{\begin{array}{ll}(\widetilde{\phi},\widetilde{\psi}) \in C_{l}:\end{array}\left|{\begin{array}{ll}(\widetilde{\phi}(t+T,z),\widetilde{\psi}(t+T,z))=(\widetilde{\phi}(t,z),\widetilde{\psi}(t,z)),\\(\phi^{-}(t,z),\psi^{-}(t,z))\leq(\widetilde{\phi}(t,z),\widetilde{\psi}(t,z))\\\leq(\phi^{+}(t,z),\psi^{+}(t,z)), \forall (t,z)\in \mathbb{R}\times[-l,l];\\(\widetilde{\phi}(t,\pm l),\widetilde{\psi}(t,\pm l))=(\phi^{-}(t,\pm l),\psi^{-}(t,\pm l)),\forall t\in\mathbb{R}.\end{array}}\right.\right\}\end{eqnarray*}$

则对任意给定的$(\tilde{\phi}, \tilde{\psi})\in\Gamma_l,$定义两个函数

$\begin{eqnarray*}\ \chi_l[\tilde{\phi}](t,z)=\begin{cases}\tilde{\phi}(t,z),t\in\mathbb{R},|z|\leq l;\\ \phi^-(t,z),t\in\mathbb{R},|z|>l;\end{cases}\quad\chi_l[\tilde{\psi}](t,z)=\begin{cases}\tilde{\psi}(t,z),t\in\mathbb{R},|z|\leq l;\\ \psi^-(t,z),t\in\mathbb{R},|z|>l.\end{cases}\end{eqnarray*}$

和两个映射

$\begin{align} f_1[\tilde{\phi},\tilde{\psi}](t,z)=\ &\alpha_1\tilde{\phi}(t,z)-\beta(t)f(\tilde{\phi}(t,z),\tilde{\psi}(t,z)),\notag\\ f_2[\tilde{\phi},\tilde{\psi}](t,z)=\ &\alpha_2\tilde{\psi}(t,z)-\gamma(t)\tilde{\psi}(t,z)+\int_{\mathbb{R}}\Gamma(t,t-\tau;y){\beta}(t-\tau) \notag\\&\times f(\chi_l[\tilde{\phi}](t-\tau,z-c\tau-y),\chi_l[\tilde{\psi}](t-\tau,z-c\tau-y))){\rm d}y. \notag\end{align}$

这里$\alpha_{i},i=1,2$是两个正常数分别满足$\alpha_1\geq\underset{t\in[T]}{\text{max}}\beta(t)N,\alpha_2\geq\underset{t\in[T]}{\text{max}}\gamma(t)$, 其中常数$N$满足$\partial_1f(\tilde{\phi}, \tilde{\psi})\leq N$. 显然, 对任意$l>z^{\ast}$,$\chi_{l}[\tilde{\phi}](t,\cdot)$和$\chi_{l}[\tilde{\psi}](t,\cdot)$是关于$t$的$T$周期函数.

记$\mathcal{A}_i\varphi=d_i\varphi_{zz}-c\varphi_z-\alpha_i\varphi,i=1,2\text{.}$对任意给定的$(\tilde{\phi}, \tilde{\psi})\in\Gamma_l,$考虑如下线性抛物系统的初边值问题

$\begin{cases}\phi_t(t,z)-\mathcal{A}_{1}\phi(t,z)=f_{1}[\widetilde{\phi},\widetilde{\psi}](t,z),\\\psi_t(t,z)-\mathcal{A}_{2}\psi(t,z)=f_{2}[\widetilde{\phi},\widetilde{\psi}](t,z),(t,z)\in(0,\infty)\times[-l,l],\\\phi(0,z)=\phi_0(z),\psi(0,z)=\psi_0(z), z\in[-l,l],\\\phi(t,\pm l)=G_{1}(t,\pm l),\psi(t,\pm l)=G_{2}(t,\pm l),t\geq0,\end{cases}$

其中$\phi_0,\psi_0\in\left[-l,l\right],G_1\left(t,z\right)=\frac{1}{2}\phi^-\left(t,-l\right)-\frac{z}{2l}\phi^-\left(t,-l\right),G_2(t,z)\text{=}\frac{1}{2}\psi^-(t,-l)-\frac{z}{2l}\psi^-(t,-l),(t,z)\in(0,T)\times[-l,l]$. 显然$G_1(t,\pm l)=\phi^-(t,\pm l),G_2(t,\pm l)=\psi^-(t,\pm l),\forall t\in\Bbb R.$$G_i\in C^{1,2}$$(\mathbb{R}\times[-l,l]),i=1,2,$是在时间$t$上的$T$周期函数.

令$V_1(t,z)=\phi(t,z)-G_1(t,z),V_2(t,z)=\psi(t,z)-G_2(t,z),\widetilde{G}_i(t,z)=\mathcal{A}_i G_i(t,z)-\partial_{t}G_i(t,z),$$i=1,2$, 则 (2.18) 式化为关于$(V_{1},V_{2})$具有齐次边界条件的初边值问题

$\begin{cases}\partial_{t} V_i(t,z)-\mathcal{A}_{i}V_i(t,z)=f_i[\tilde{\phi},\tilde{\psi}](t,z)+\widetilde{G}_i(t,z),(t,z)\in(0,\infty)\times[-l,l],\\ V_1(0,z)=\phi_0(z)-G_1(0,z),V_2(0,z)=\psi_0(z)-G_2(0,z),z\in[-l,l],\\ V_i(t,\pm l)=0,t\ge0,i=1,2.\end{cases}$

定义$\mathcal{B}_i\varphi=\mathcal{A}_i\varphi,D\left(\mathcal{B}_i\right)=\Big\{\varphi\in\bigcap\limits_{p\geq1}W^{2,p}_{\text{loc}}:\varphi,\mathcal{A}_i\varphi\in C\left(\left[-l,l\right]\right),\varphi\left(\pm l\right)=0\Big\},i=1,2$, 则$\mathcal{B}_i:D(\mathcal{B}_i)\subset C([- l,l])\to C\big([- l,l])$生成强连续解析半群$T_i(t)_{t\geq0}$:

$\begin{equation*} T_i(t)\varphi(x)={\rm e}^{-\alpha_it}\int_{-l}^l\Gamma_i(t;x,y)\varphi(y){\rm d}y,\varphi \in C([-l,l]),(t,z)\in(0,\infty)\times[-l,l],\end{equation*}$

其中$\Gamma_{i}$是$d_{i}\partial_{xx}-c\partial_{x}$关于 (2.18) 式中 Dirichlet 边界条件的 Green 函数,$i=1,2.$将 (2.19) 式改写为

$\begin{cases} V_1(t,z)=T_1(t)(\phi_0-G_1(0))(z)+\int_0^t T_1(t-s)(f_1[\tilde{\phi},\tilde{\psi}](s)+\widetilde{G}_1(s))(z){\rm d}s,\\[3mm] V_2(t,z)=T_2(t)(\psi_0-G_2(0))(z)+\int_0^t T_2(t-s)(f_2[\tilde{\phi},\tilde{\psi}](s)+\widetilde{G}_2(s))(z){\rm d}s.\end{cases}$

由$V_{i}(t,z)\ (i=1,2)$的定义结合 (2.20) 式, 可得

$\begin{cases}\phi(t,z)=T_1(t)(\phi_0 -G_{1}(0))(z)+\int_{0}^{t}T_1(t-s)(f_{1}[\widetilde{\phi},\widetilde{\psi}](s)+\widetilde{G}_{1}(s))(z){\rm d}s+G_{1}(t,z),\\[3mm] \psi(t,z)=T_2(t)(\psi_0-G_{2}(0))(z)+\int_{0}^{t}T_2(t-s)(f_{2}[\widetilde{\phi},\widetilde{\psi}](s)+\widetilde{G}_{2}(s))(z){\rm d}s+G_{2}(t,z),\end{cases}$

从而系统 (2.21) 的任意连续解是系统 (2.19) 的一个 mild 解. 另外, 考虑到$f_i[\tilde{\phi},\tilde{\psi}]\in C(\mathbb{R}\times[-l,l]),$$i=1,2,$且$\phi_0(z),\psi_0(z)\in C([-l,l]),$根据文献[13,定理 5.1.18,定理 5.1.19] 可知, 对于满足 (2.19) 式关于时间的可微函数$\phi$和$\psi$, 对任意的$\varepsilon\in(0,2T),\theta\in(0,1)$, 有$\phi,\psi\in C\left(\left[2T\right]\times\left[-l,l\right]\right) \cap C^{\theta,2\theta}\left(\left[\varepsilon,2T\right]\times\left[-l,l\right]\right)$成立.

定义如下集合

$\begin{eqnarray*}\Gamma_{l}^{\prime}:= \left\{\begin{array}{ll}(\phi_{0},\psi_{0}) \in C([-l,l],\mathbb{R}^{2}):\end{array}\left|{\begin{array}{ll}(\phi^{-}(0,z),\psi^{-}(0,z))\leq(\phi_{0}(z),\psi_{0}(z))\\\leq(\phi^{+}(0,z),\psi^{+}(0,z)), z\in[-l,l];\\(\phi_{0}(\pm l),\psi_{0}(\pm l))=(\phi^{-}(0,\pm l),\psi^{-}(0,\pm l)),\end{array}}\right.\right\}\end{eqnarray*}$

则$\Gamma_{l}^{\prime}$是$C([-l,l],\mathbb{R}^{2})$上具有上确界范数的闭凸集.

引理 2.5 任意给定$(\phi_{0},\psi_{0})\in\Gamma_{l}^{\prime}$, 记$(\phi_{l} (t,z;\phi_{0},\psi_{0}),\psi_{l} (t,z;\phi_{0},\psi_{0}))$是系统$(2.21)$以$(\phi_{0},\psi_{0})$为初值的解, 则对所有的$t>0,z\in[-l,l]$, 有

$\phi^{-}(t,z)\leq\phi_{l} (t,z;\phi_{0},\psi_{0})\leq\phi^{+}(t,z),$
$\psi^{-}(t,z)\leq\psi_{l} (t,z;\phi_{0},\psi_{0})\leq\psi^{+}(t,z).$

对所有的$(t,z)\in(0,\infty)\times[-l,l],$有$\phi_l(t,z;\phi_0,\psi_0)\le\phi^+(t,z)$. 令

$ \overline{\phi}(t,z)=T_1(t)(\phi_0-G_1(0))(z)+\int_0^t T_1(t-s)(f_1[\tilde{\phi},\tilde{\psi}](s)+\widetilde{G}_1(s))(z){\rm d}s+G_1(t,z),$

这里$(t,z)\in(0,\infty)\times\mathbb{R}$. 因为$\alpha_1>\underset{t\in[T]}{\text{max}}\beta(t)N$,则对任意的$(t,z)\in(0,\infty)\times[-l,l]$, 有

$f_1[\tilde{\phi},\tilde{\psi}](t,z)\leq f_1[\phi^+,\psi^-](t,z),$

这是因为: 为使

$\begin{align*}&f_{1}[\phi^{+},\psi^{-}](t,z)-f_{1}[\tilde{\phi},\tilde{\psi}](t,z)\\=\ &\alpha_{1}\phi^{+}(t,z)-\beta(t)f(\phi^{+}(t,z),\psi^{-}(t,z))-\alpha_{1}\tilde{\phi}(t,z)+\beta(t)f(\tilde{\phi}(t,z),\tilde{\psi}(t,z))\\=\ &\alpha_{1}[\phi^{+}(t,z)-\tilde{\phi}(t,z)]-\beta(t)[f(\phi^{+}(t,z),\psi^{-}(t,z))-f(\tilde{\phi}(t,z),\tilde{\psi}(t,z))]\geq0,\end{align*}$

只需使

$\begin{align*}\alpha_{1}&\geq\beta(t)\partial_{1}f(\tilde{\phi}(t,z),\tilde{\psi}(t,z))\\&=\beta(t)\frac{f(\phi^{+}(t,z),\tilde{\psi}(t,z))-f(\tilde{\phi}(t,z),\tilde{\psi}(t,z))}{\phi^{+}(t,z)-\tilde{\phi}(t,z)}\\&\geq\beta(t)\frac{f(\phi^{+}(t,z),\psi^{-}(t,z))-f(\tilde{\phi}(t,z),\tilde{\psi}(t,z))}{\phi^{+}(t,z)-\tilde{\phi}(t,z)},\end{align*}$

因此当$\alpha_{1}\geq\beta(t)\partial_{1}f(\tilde{\phi}(t,z),\tilde{\psi}(t,z))$时, 由条件 (A$_{2})$中:$\partial_{1}f(S,I)>0$且对所有$S\geq0$有界可知, 存在正常数$N$使得$\partial_{1}f(\tilde{\phi}(t,z),\tilde{\psi}(t,z))\leq N$, 所以当$\alpha_1>\underset{t\in[T]}{\text{max}}\beta(t)N$时, 可以得到$f_1[\tilde{\phi},\tilde{\psi}](t,z)\leq f_1[\phi^+,\psi^-](t,z)$.那么对任意的$(t,z)\in(0,\infty)\times[-l,l]$, 可得$\phi_l(t,z;\phi_0,\psi_0)\leq\overline{\phi}(t,z;\phi_0,\psi_0)$. 又因为当$\theta\in(0,1)$时, 有$f_1[\phi^+,\psi^-]\in C^{\frac{\theta}{2},\theta}(\mathbb{R}\times[-l,l])$,$G_{1}(\cdot,\cdot)\in C^{1,2}(\mathbb{R}\times[-l,l])$且$\overline{\phi}_{0}(\cdot)\in C([-l,l])$, 从而由文献[13,定理 5.1.18] 可知, 对任意的$\delta>0$, 有$\partial_t\overline{\phi},\mathcal{A}_1\overline{\phi}\in C^{\frac{\theta}{2},0}([\delta,+\infty)\times[-l,l])$.又由文献[13,定理 5.1.19] 可知, 对任意的$\delta>0$, 有$\partial_t\overline{\phi},\mathcal{A}_1\overline{\phi}\in C^{0,\theta}([\delta,+\infty)\times[-l,l])$, 所以$\partial_t\overline{\phi},\mathcal{A}_1\overline{\phi}\in C^{\frac{\theta}{2},\theta}([\delta,+\infty)\times[-l,l])$. 此外, 由文献[13,定理 5.1.18 和定理 5.1.19] 可知,$\overline{\phi}$在$(0,\infty)\times[-l,l]$上关于$t$是可微的,且对任意的$p\geq1$, 有$\overline{\phi}(t,\cdot)\in W_{loc}^{2,p}((-l,l))$, 其中$W_{loc}^{2,p}((-l,l))=\{u|u\in L_{loc}^{p}((-l,l)), D^{r}u\in L_{loc}^{p}((-l,l)), \forall|r|\leq2\}$.因此, 由空间$W_{loc}^{2,p}((-l,l))$的定义并结合$\overline{\phi}_{0}(\cdot)\in C([-l,l])$, 以及$\overline{\phi}$在$(0,\infty)\times[-l,l]$上关于$t$是可微的, 从而有$\overline{\phi}\in C([l])\cap C^{1,2}((0,+\infty)\times[-l,l])$并满足

$\begin{cases}\ \overline{\phi}_{t}(t,z)-\mathcal{A}_1\overline{\phi}(t,z)=f_1[\phi^+,\psi^-](t,z),(t,z)\in(0,\infty)\times[-l,l],\\ \overline{\phi}(0,z)=\phi_0(z),z\in[-l,l],\\ \overline{\phi}(t,\pm l)=G_1(t,\pm l)=\phi^-(t,\pm l),t\in(0,\infty).\end{cases}$

另一方面, 由引理 2.1 可知,$\phi^{+}$满足

$\begin{cases}\ \phi^{+}_{t}\left(t,z\right)-\mathcal{A}_{1}\phi^{+}\left(t,z\right)\geq f_{1}\left[\phi^{+},\psi^{-}\right]\left(t,z\right),\left(t,z\right)\in\left(0,\infty\right)\times\left[- l,l\right],\\ \phi^{+}\left(0,z\right)=S_{0}\geq\phi_{0}\left(z\right),z\in\left[- l,l\right],\\ \phi^{+}\left(t,\pm l\right)=S_{0}\geq\phi^{-}\left(t,\pm l\right),t\in\left(0,\infty\right).\end{cases}$

由比较原理可知,$\overline{\phi}(t,z)\leq\phi^+(t,z),\forall(t,z)\in(0,\infty)\times[-l,l].$因此, 对任意的$(t,z)\in(0,\infty)\times[-l,l],$有$\phi_l(t,z;\phi_0,\psi_0)\leq\overline{\phi}(t,z;\phi_0,\psi_0)\leq\phi^+(t,z)\text{.}$

下证对任意的$(t,z)\in(0,\infty)\times[-l,l]$, 有$\phi^-\left(t,z\right)\leq\phi_l\left(t,z;\phi_0,\psi_0\right)$. 令

$\begin{eqnarray*}\ \underline{\phi}(t,z)=T_1(t)(\phi_0-G_1(0))(z)+\int_0^t T_1(t-s)(f_1[\tilde{\phi},\tilde{\psi}](s)+\widetilde{G}_1(s))(z){\rm d}s+G_1(t,z),\end{eqnarray*}$

这里$(t,z)\in(0,\infty)\times\mathbb{R}$. 因为$\alpha_1>\underset{t\in[T]}{\text{max}}\beta(t)N,$则对任意的$(t,z)\in(0,\infty)\times[-l,l]$, 有$f_1[\phi^-,\psi^+](t,z)\leq f_1[\tilde{\phi},\tilde{\psi}](t,z),$即对任意的$(t,z)\in(0,\infty)\times[-l,l],$有$\underline{\phi}(t,z;$$\phi_0,\psi_0)\leq\phi_l(t,z;\phi_0,\psi_0)$. 类似于$\overline{\phi}$的讨论, 有

$\begin{cases}\underline{\phi}_{t}(t,z)-\mathcal{A}_{1}\underline{\phi}(t,z)=f_1[\phi^-,\psi^+](t,z),(t,z)\in(0,\infty)\times[-l,l],\\ \underline{\phi}(0,z)=\phi_0(z),z\in[-l,l],\\ \underline{\phi}(t,\pm l)=G_1(t,\pm l)=\phi^-(t,\pm l),t\geq0.\end{cases}$

令$\underline{\phi}^*\equiv0,$则$\underline{\phi}^*$满足$\underline{\phi}_t^*(t,z)-\mathcal{A}_{1}\underline{\phi}^*(t,z)\leq f_1[\phi^-,\psi^+](t,z),(t,z)\in\big(0,\infty\big)\times\big[-l,l],$由抛物比较原理可得, 对任意的$(t,z)\in(0,\infty)\times[-l,l]$,有$\underline{\phi}(t,z)\geq0$, 然后根据引理 2.2 得

$\begin{cases} \phi^{-}_{t}\left(t,z\right)-\mathcal{A}_{1}\phi^{-}\left(t,z\right)\leq f_{1}\left[\phi^{-},\psi^{+}\right](t,z),\left(t,z\right)\in\left(0,\infty\right)\times\left[-l,z_{1}\right],\\ \phi^{-}\left(0,z\right)\leq\phi_{0}\left(z\right)=\underline{\phi}\left(0,z\right),z\in\left[-l,z_{1}\right],\\ \phi^{-}\left(t,-l\right)=G_{1}\left(t,-l\right)=\underline{\phi}\left(t,-l\right),\phi^{-}\left(t,z_{1}\right)=0\leq\underline{\phi}\left(t,z_{1}\right),t\geq0,\end{cases}$

其中$z_{1}$已在引理 2.2 中定义, 再次利用抛物比较原理得$\phi^-(t,z)\leq\underline{\phi}(t,z),\forall(t,z)\in(0,\infty)\times[-l,z_1],$所以$\phi^-(t,z)\leq\underline{\phi}(t,z;\phi_0,\psi_0)\leq\phi_l(t,z;\phi_0,\psi_0),$结合$\phi_l(t,z;\phi_0,\psi_0)\leq\overline{\phi}(t,z;\phi_0,\psi_0)\leq\phi^+(t,z)$, 可得$\phi^-\left(t,z\right)\leq\phi_l\left(t,z;\phi_0,\psi_0\right)\leq\phi^+\left(t,z\right).$同理可得, 对任意的$(t,z)\in(0,\infty)\times[-l,l]$, 有$\psi^-\left(t,z\right)\leq\psi_l\left(t,z;\phi_0,\psi_0\right)\leq\psi^+\left(t,z\right).$证毕.

对任意给定的$(\tilde{\phi}, \tilde{\psi})\in\Gamma_{l},$定义$\Gamma'_{l}\to C([-l,l],\mathbb{R})$的映射$F_{(\tilde{\phi},\tilde{\psi})}$,

$\begin{eqnarray*}\ F_{(\tilde{\phi},\tilde{\psi})}[\phi_0,\psi_0](\cdot)=(\phi_l(T,\cdot;\phi_0,\psi_0),\psi_l(T,\cdot;\phi_0,\psi_0)).\end{eqnarray*}$

这里$(\phi_l(t,\cdot;\phi_0,\psi_0),\psi_l(t,\cdot;\phi_0,\psi_0))$是 (2.21) 式以$\text{(}\phi_0,\psi_0\text{)}\in\Gamma'_l$为初值的解. 由引理 2.5 可知,$F_{(\tilde{\phi},\tilde{\psi})}(\Gamma'_l)\subset\Gamma'_l.$此外$\Gamma'_l$是完备的距离空间. 由解析半群$T_{i}(t)$和 (2.21) 式可知, 对任意的$\left(\phi_0^i,\psi_0^i\right)\text{}\in\Gamma_l',i=1,2,$有

$\left\|\phi_l\left(T,\cdot;\phi_0^1,\psi_0^1\right)-\phi_l\left(T,\cdot;\phi_0^2,\psi_0^2\right)\right\|_{C[-l,l]}\leq {\rm e}^{-\alpha_1 T}\left\|\phi_0^1-\phi_0^2\right\|_{C[-l,l]},$
$\left\|\psi_l\left(T,\cdot;\phi_0^1,\psi_0^1\right)-\psi_l\left(T,\cdot;\phi_0^2,\psi_0^2\right)\right\|_{{C}[-l,l]}\leq {\rm e}^{-\alpha_2T}\left\|\psi_0^1-\psi_0^2\right\|_{{C}[-l,l]}.$

由于${\rm e}^{-\alpha_{i}T}<1,\text{}i=1,2,$故$F_{(\tilde{\phi},\tilde{\psi})}:\Gamma'_l\rightarrow\Gamma'_l$是压缩映射. 则由 Banach 不动点定理, 可知$F_{(\tilde{\phi},\tilde{\psi})}$在$\Gamma'_l$上存在唯一的不动点$\left(\phi_0^*,\psi_0^*\right)$. 令$\left(\phi_l^*\left(t,z\right),\psi_l^*\left(t,z\right)\right):=\left(\phi_l^*\left(t,z;\phi_0^*,\psi_0^*\right),\psi_l^*\left(t,z;\phi_0^*,\psi_0^*\right)\right)$为 (2.21) 式以$\left(\phi_0^*,\psi_0^*\right)$为初值的解, 这里$(t,z)\in(0,\infty)\times[-l,l]$, 那么, 由解的唯一性可得,$(\phi_l^*(t+T,z),\psi_l^*(t+T,z))=(\phi_l^*(t,z),\psi_l^*(t,z)).$进一步, 通过定义

$(\phi_l^*(t,z),\psi_l^*(t,z))=(\phi_l(t-kT,z;\phi_0^*,\psi_0^*),\psi_l(t-kT,z;\phi_0^*,\psi_0^*)),(t,z)\in\mathbb{R}\times[-l,l], $

其中$k\in\mathbb{Z}$满足$kT<t<(k+1)T$, 则$\text{(}\overset{\text{.}}{\phi_l^{\ast},\psi_l^{\ast}}\text{)}(t,\cdot)$的定义区间可以延拓至整个$\mathbb{R}$, 同时保证对所有$z\in\left[-l,l\right],\left(\phi_l^*,\psi_l^*\right)$关于$t\in\mathbb{R}$是$T$周期的. 此外, 由引理 2.5 知$\left(\phi_l^*(t,z), \psi_l^*(t,z)\right)\in\Gamma_l.$于是, 对所有的$(t,z)\in(s,\infty)\times[-l,l]$, 可知$\left(\phi_l^*,\psi_l^*\right)$满足

$\begin{cases} \phi_l^*(t,z)=T_1(t-s)(\phi_l^*(s)-G_1(s))(z)+\int_s^t(f_1[\tilde{\phi},\tilde{\psi}](\theta)+\widetilde{G}_1(\theta))(z)d\theta+G_1(t,z),\\[3mm] \psi_l^*(t,z)=T_2(t-s)(\psi_l^*(s)-G_2(s))(z)+\int_s^t(f_2[\tilde{\phi},\tilde{\psi}](\theta)+\widetilde{G}_2(\theta))(z)d\theta+G_2(t,z).\end{cases}$

由上述讨论可知, 对任意给定的$(\tilde{\phi}, \tilde{\psi})\in\Gamma_l,$存在唯一的$\left(\phi_l^*,\psi_l^*\right) \in\Gamma_l $使得 (2.24) 式成立. 因此, 由此定义的$\Gamma_{l}\rightarrow\Gamma_{l}$上的算子$\mathcal{F}$, 满足$\mathcal{F}(\tilde{\phi}, \tilde{\psi})=(\phi_l^*,\psi_l^*).$

引理 2.6$\mathcal{F}:\Gamma_{l}\rightarrow\Gamma_{l}$是一个全连续算子.

证明类似文献[7], 故此省略.

由引理 2.6, 结合 Schauder 不动点定理可知$\mathcal{F}$存在不动点$\left(\phi_l^*,\psi_l^*\right) \in\Gamma_l$. 特别地, 对任意的$t\in\mathbb{R}$可得$\left(\phi_l^*(t+T,\cdot),\psi_l^*(t+T,\cdot)\right)\text{=}(\left(\phi_l^*(t,\cdot),\psi_l^*(t,\cdot)\right).$进一步, 存在常数$\theta\in(0,1)$使得$\phi^*_l, \psi^*_l\in C^{\frac{\theta}{2},\theta}(\mathbb{R}\times[-l,l])$, 再利用文献[13,定理 5.1.18 和定理 5.1.19] 可知,$\phi_{l}^{*}, \psi_{l}^{*}$满足

$\begin{cases}\partial_{t}\phi_l^*(t,z)=d_{1}\partial_{zz}\phi_l^*(t,z)-c\partial_{z}\phi_l^*(t,z)-\beta(t)f(\phi_l^*(t,z),\psi_l^*(t,z)),\\ \partial_{t}\psi_l^*(t,z)=d_{2}\partial_{zz}\psi_l^*(t,z)-c\partial_{z}\psi_l^*(t,z)-\gamma(t)\psi_l^*(t,z)+\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\\ \times\beta(t-\tau)f(\chi_{l}(\phi_l^*)(t-\tau,z-c\tau-y),\chi_{l}(\psi_l^*)(t-\tau,z-c\tau-y)){\rm d}y,\\\phi_l^*(t,\pm l)=\phi^{-}(t,\pm l),\psi_l^*(t,\pm l)=\psi^{-}(t,\pm l),t\in\mathbb{R}.\end{cases}$

下面定理给出$\phi_l^*,\psi_l^*$的局部一致估计.

引理 2.7 令$p\geq2$. 对任意给定的$Z>0$, 存在常数$B(p,Z)>0$, 使得对充分大的$l>\max\{Z,z^{\ast}\}$, 有

$\begin{eqnarray*}\|\phi_{l}^{*}\|_{W_{p}^{1,2} ((0,T)\times[-Z,Z])}\leq B(p,Z), \|\psi_{l}^{*}\|_{W_{p}^{1,2} ((0,T)\times[-Z,Z])}\leq B(p,Z).\end{eqnarray*}$

进一步, 存在常数$B^{\prime}(Z)>0$, 使得对任意的$z_{0} \in\mathbb{R}$和充分大的$l>\max\{Z+|z_{0}|,z^{\ast}\}$, 均有

$\begin{eqnarray*}\|\phi_{l}^{*}\|_{C^{\frac{1+\theta}{2},1+\theta}((0,T)\times[z_{0}-Z,z_{0}+Z])}\leq B^{\prime}(Z), \|\psi_{l}^{*}\|_{C^{\frac{1+\theta}{2},1+\theta}((0,T)\times[z_{0}-Z,z_{0}+Z])}\leq B^{\prime}(Z),\end{eqnarray*}$

其中$\theta\in (0,1)$是一个常数.

固定$Z>0$和$z_{0}\in\mathbb{R}$, 令$l>\max\{Z+|z_{0}|,z^{\ast}\}$. 因为$\left(\phi_l^*,\psi_l^*\right) \in\Gamma_l $且满足 (2.25) 式, 则存在与$l$无关的正常数$M$, 使得

$\sup\limits_{\left(t,z\right)\in\left[T\right]\times\left[-l,l\right]}\phi_l^*\left(t,z\right)<M,\sup\limits_{\left(t,z\right)\in\left[T\right]\times\left[-l,l\right]}\psi_l^*\left(t,z\right)<M\text{.}$

令$\hat{\phi_l}\left(t,z\right)\text{:}={\rm e}^{-\frac{c\left(z-z_0\right)}{2d_1}}\phi_l^*\left(t,z\right),\hat{\psi_l}\left(t,z\right)\text{:}={\rm e}^{-\frac{c\left(z-z_0\right)}{2d_2}}\psi_l^*\left(t,z\right),\forall t\in\mathbb{R},\text{}z\in[-l,l],$则对任意的$t\in\mathbb{R},\text{}z\in[-l,l],\hat{\phi_l},\hat{\psi_l}$满足

$\begin{cases} \partial_t\hat{\phi_l}(t,z)=d_1\partial_{zz}\hat{\phi_l}(t,z)-\frac{c^2}{4d_1}{\rm e}^{-\frac{c\left(z-z_0\right)}{2d_1}}\phi_l^*\left(t,z\right)-{\rm e}^{-\frac{c\left(z-z_0\right)}{2d_1}}\beta(t)f(\phi_l^*(t,z),\psi_l^*(t,z)),\\[3mm] \partial_t\hat{\psi}_l(t,z)=d_2\partial_{zz}\hat{\psi}_l(t,z)+{\rm e}^{-\frac{c(z-z_0)}{2d_2}}\Big[\frac{c^2}{4d_2}\psi_l^*\left(t,z\right)-\gamma(t)\psi_l^*\left(t,z\right)+\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\\[3mm] \times\beta(t-\tau)f(\chi_l[\phi^*_l](t-\tau,z-c\tau-y),\chi_l[\psi^*_l](t-\tau,z-c\tau-y)){\rm d}y\Big],\end{cases}$

对任意的$r>0$和$t',z'\in\mathbb{R}$, 令$\mathcal{Q}((t',z'),r):=\{(t,z)\in\mathbb{R}^2:|t-t'|<r,|z-z'|<r,t<t'\}$, 取$R=\max\big\{2Z,\sqrt{3T}\big\}$, 定义

$\begin{eqnarray*}\ h_l^\phi\left(t,z\right)&=&{\rm e}^{-\frac{c\left(z-z_0\right)}{2d_1}}\Big[-\frac{c^2}{4d_1}\phi_l^*\left(t,z\right)-\beta(t)f\left(\phi_l^*\left(t,z\right),\psi_l^*\left(t,z\right)\right)\Big],\\\ h_l^\psi(t,z)&=&{\rm e}^{-\frac{c(z-z_0)}{2d_2}}\Big[-\frac{c^2}{4d_2}\psi_l^\ast(t,z)-\gamma(t)\psi_l^\ast(t,z)+\int_{\mathbb{R}}\beta(t-\tau)\Gamma(t,t-\tau;y)\\&&\times f(\chi_l[\phi_l^\ast](t-\tau,z-c\tau-y),\chi_l[\psi_l^\ast](t-\tau,z-c\tau-y)){\rm d}y\Big].\end{eqnarray*}$

由文献[14,性质 7.14], 对$l>72R+\left|z_0\right|$存在不依赖于$l$的常数$B_1(p,R)>0$使得

$ \left\|\partial_z\hat{\phi_l}\right\|_{L^p\left(\mathcal{Q}((2T,z_0),2R)\right)}\leq B_1\Big(\left\|\hat{\phi_l}\right\|_{L^p\left(\mathcal{Q}((2T,z_0),72R)\right)}+\left\|h_l^{\phi}\right\|_{L^p\left(\mathcal{Q}((2T,z_0),72R)\right)}\Big),$
$\left\|\partial_z\hat{\psi_l}\right\|_{L^p\left(\mathcal{Q}((2T,z_0),2R)\right)}\leq B_1\Big(\left\|\hat{\psi_l}\right\|_{L^p\left(\mathcal{Q}((2T,z_0),72R)\right)}+\left\|h_l^{\psi}\right\|_{L^p\left(\mathcal{Q}((2T,z_0),72R)\right)}\Big),$

结合$\hat{\phi_l}$和$\hat{\psi_l}$的定义可知, 存在不依赖于$l$的常数$B_2(p,R)$使得

$\begin{eqnarray*}\ \left\|\partial_z\phi_l^*\right\|_{L^p(\mathcal{Q}((2T,z_0),2R))}\leq B_2,\quad\left\|\partial_z\psi_l^*\right\|_{L^p(\mathcal{Q}((2T,z_0),2R))}\leq B_2,\end{eqnarray*}$

另外, 由文献[14,性质 7.18] 可知, 存在不依赖于$l$的常数$B_3(p,R)$使得

$\left\|\partial_t\phi_l^*\right\|_{L^p(\mathcal{Q}((2T,z_0),2R))}+\left\|\partial_{zz}\phi_l^*\right\|_{L^p(\mathcal{Q}((2T,z_0),2R))}\leq B_3,$
$\left\|\partial_t\psi_l^*\right\|_{L^p(\mathcal{Q}((2T,z_0),2R))}+\left\|\partial_{zz}\psi_l^*\right\|_{L^p(\mathcal{Q}((2T,z_0),2R))}\leq B_3.$

因此, 存在不依赖于$l$的常数$B(p,R)$, 使得

$\left\|\phi_l^*\right\|_{W_p^{1,2}(\mathcal{Q}((2T,z_0),2R))}\leq B, \left\|\psi_l^*\right\|_{W_p^{1,2}(\mathcal{Q}((2T,z_0),2R))}\leq B,$

注意到$[T]\times[- Z,Z]\subset\mathcal{Q}((2T,z_0), 2R)$且$R$只依赖于$Z$. 那么,$\left\|\phi_l^*\right\|_{W_p^{1,2}([T]\times[-Z,Z])}\leq B,$$\left\|\psi_l^*\right\|_{W_p^{1,2}([T]\times[-Z,Z])}\leq B$.取$p>3$, 由空间$W_p^{1,2}([T]\times[-Z,Z])=\{u|u\in L_{p}([T]\times[-Z,Z]), D_{t}^{r}D_{x}^{s}u\in L_{p}([T]\times[-Z,Z]), \forall 2r+|s|\leq2\}$的定义以及文献[15,定理 5.5.5] (嵌入定理) 可知, 当$p>3,n=1,u\in W_{p}^{1,2}([T]\times[-Z,Z]), \partial([-Z,Z])\in C^{2}(\mathbb{R})$时, 则满足定理 5.5.5 中的$p>\frac{3}{2}$且$\alpha=2-\frac{3}{p}$不是整数$(1<\alpha<2)$, 从而有$u\in C^{\frac{\alpha}{2},\alpha}([T]\times[-Z,Z])$, 并且$|u|_{C^{\frac{\alpha}{2},\alpha}([T]\times[-Z,Z])}\leq C\|u\|_{W_p^{1,2}([T]\times[-Z,Z])} (C$为常数). 因此, 对某个$\theta\in(0,1)$和$z_{0}\in\mathbb{R}$, 有$W_{p}^{1,2}([T]\times[-Z,Z])\hookrightarrow C^{\frac{1+\theta}{2},1+\theta}([T]\times[z_0-Z,z_0+Z])$, 故$\phi_l^*,\psi_l^*\in C^{\frac{1+\theta}{2},1+\theta}([T]\times[z_0-Z,z_0+Z])$, 进而存在不依赖于$l$的常数$B'(p,R)$使得

$\begin{matrix}\left\|\phi_l^*\right\|_{C^{\frac{1+\theta}{2},1+\theta}([T]\times[z_0-Z,z_0+Z])}\leq B',\left\|\psi_l^*\right\|_{C^{\frac{1+\theta}{2},1+\theta}([T]\times[z_0-Z,z_0+Z])}\leq B'.\end{matrix}$

证毕.

2.4 周期行波解的存在性

定理 2.1 若$\Re_0>1$且$c>c^*$时$,$则系统$(1.4)$存在周期行波解$(\phi^{*}(t,z),\psi^{*}(t,z))$.

对$\forall m\in\mathbb{N}$, 令$l_{m}$是递增序列满足$l_{m}>z^{\ast}$且$\lim\limits_{m\to+\infty}l_m=+\infty$, 由前面的讨论可知,$\left(\phi_{l_m}^*, \psi_{l_m}^*\right)\in\Gamma_{l_{m}}$满足 (2.25) 式和引理 2.7. 由引理 2.7 以及$(\phi_{l_m}^*, \psi_{l_m}^*)$关于$t\in\mathbb{R}$的周期性, 可知存在$(\phi_{l_m}^*, \psi_{l_m}^*)$的子列, 不妨仍记为$(\phi_{l_m}^*, \psi_{l_m}^*)$以及$\left(\phi^*,\psi^*\right)\text{}\in C\left(\mathbb{R}^2\right)$使得在$C^{\frac{1+\beta}{2},1+\beta}_{loc}(\mathbb{R}^2),H^1_{loc}(\mathbb{R}^2)$以及$L_{loc}^2\left(\mathbb{R},H_{loc}^2\left(\mathbb{R}\right)\right)$中有

$\begin{matrix}\left(\phi_{l_m}^*,\psi_{l_m}^*\right)\stackrel{\text{弱}}{\rightarrow}\left(\phi^*,\psi^*\right),\end{matrix}$

其中$\beta\in\left(0,\theta\right),\theta\in\left(0,1\right)$. 显然

$\begin{eqnarray*}\left(\phi^*,\psi^*\right)\in C^{\frac{1+\beta}{2},1+\beta}_{loc}(\mathbb{R}^2)\cap H^1_{loc}(\mathbb{R}^2)\cap L^2_{loc}(\mathbb{R},H^2_{loc}(\mathbb{R})).\end{eqnarray*}$

因为$(\phi_{l_m}^*, \psi_{l_m}^*)$是$T$-周期函数, 所以对$(t,z)\in\mathbb{R}^{2}$, 有

$\begin{eqnarray*}\ (\phi^*(t+T,z),\psi^*(t+T,z))=(\phi^*(t,z),\psi^*(t,z)).\end{eqnarray*}$

进一步, 由 (2.27) 式可知, 对所有的$l>0$, 存在与$l$无关的常数$B_{4}>0$使得

$\begin{matrix}\left\|\phi^*\right\|_{C_{[T]\times[-l,l]}^{\frac{1+\beta}{2},1+\beta}(\mathbb{R}^{2})}+\left\|\psi^*\right\|_{C_{[T]\times[-l,l]}^{\frac{1+\beta}{2},1+\beta}(\mathbb{R}^{2})}\leq B_{4}.\end{matrix}$

对给定的$u,\nu \in C^\infty_0\left(\mathbb{R}^2\right)$, 对于任意足够大的$m$, 有$\text{supp}(u)\times\text{supp}(\nu)\subset\mathbb{R}\times(-l_m,l_m)$. 那么,$(\phi_{l_m}^*, \psi_{l_m}^*)$满足

$\begin{matrix}&\iint_{\mathbb{R}^2}\partial_t u\left(t,z\right)\phi_{l_m}^*\left(t,z\right){\rm d}t{\rm d}z-d_1\iint_{\mathbb{R}^2}\partial_z u\left(t,z\right)\partial_z\phi_{l_m}^*\left(t,z\right){\rm d}t{\rm d}z \notag\\=\ &c\iint_{\mathbb{R}^2}u(t,z)\partial_z\phi_{l_m}^*(t,z){\rm d}t{\rm d}z+\iint_{\mathbb{R}^2}\beta(t)u(t,z)f(\phi_{l_m}^*(t,z),\psi_{l_m}^*(t,z)){\rm d}t{\rm d}z,\end{matrix}$

$\begin{matrix}&\iint_{\mathbb{R}^2}\partial_t \nu\left(t,z\right)\psi_{l_m}^*\left(t,z\right){\rm d}t{\rm d}z-d_2\iint_{\mathbb{R}^2}\partial_z \nu\left(t,z\right)\partial_z\psi_{l_m}^*\left(t,z\right){\rm d}t{\rm d}z \notag\\=\ &c\iint_{\mathbb{R}^2}\nu(t,z)\partial_z\psi_{l_m}^*(t,z){\rm d}t{\rm d}z+\iint_{\mathbb{R}^2}\nu(t,z)\gamma(t)\psi_{l_m}^*(t,z)\text{}{\rm d}t{\rm d}z-\iint_{\mathbb{R}^2}\nu(t,z)\int_{\mathbb{R}}\Gamma(t,t-\tau;y) \notag\\&\times \beta(t-\tau)f(\chi_{l_m}[\phi_{l_m}^*](t-\tau,z-c\tau-y),\chi_{l_m}[\psi_{l_m}^*](t-\tau,z-c\tau-y)){\rm d}y{\rm d}t{\rm d}z.\end{matrix}$

由勒贝格控制收敛定理可知, 当$m\rightarrow\infty$时, 对任意的$(t,z)\in\mathbb{R}^{2}$都有

$\begin{matrix}&\int_\mathbb{R}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\chi_{l_m}[\phi_{l_m}^*](t-\tau,z-c\tau-y),\chi_{l_m}[\psi_{l_m}^*](t-\tau,z-c\tau-y)){\rm d}y \notag\\\rightarrow &\int_\mathbb{R}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\phi^*(t-\tau,z-c\tau-y),\psi^*(t-\tau,z-c\tau-y)){\rm d}y,\end{matrix}$

结合 (2.27) 和 (2.29)-(2.31) 式可知,$\phi^*,\psi^*$满足

$\begin{align}&\iint_{\mathbb{R}^2}\partial_t u\left(t,z\right)\phi^*\left(t,z\right){\rm d}t{\rm d}z-d_1\iint_{\mathbb{R}^2}\partial_z u\left(t,z\right)\partial_z\phi^*\left(t,z\right){\rm d}t{\rm d}z \notag\\=\ &c\iint_{\mathbb{R}^2}u(t,z)\partial_z\phi^*(t,z){\rm d}t{\rm d}z+\iint_{\mathbb{R}^2}\beta(t)u(t,z)f(\phi^*(t,z),\psi^*(t,z)){\rm d}t{\rm d}z, \notag\end{align}$

$\begin{align*}&\iint_{\mathbb{R}^2}\partial_t \nu\left(t,z\right)\psi^*\left(t,z\right){\rm d}t{\rm d}z-d_2\iint_{\mathbb{R}^2}\partial_z \nu\left(t,z\right)\partial_z\psi^*\left(t,z\right){\rm d}t{\rm d}z \notag\\=\ &c\iint_{\mathbb{R}^2}\nu(t,z)\partial_z\psi^*(t,z){\rm d}t{\rm d}z+\iint_{\mathbb{R}^2}\nu(t,z)\gamma(t)\psi^*(t,z)\text{}{\rm d}t{\rm d}z\\&-\iint_{\mathbb{R}^2}\nu(t,z)\int_{\mathbb{R}}\Gamma(t,t-\tau;y) \beta(t-\tau)f(\phi^*(t-\tau,z-c\tau-y),\psi^*(t-\tau,z-c\tau-y)){\rm d}y{\rm d}t{\rm d}z. \notag\end{align*}$

由$u,\nu\in C^\infty_0\left(\mathbb{R}^2\right)$的任意性, 可知$\left(\phi^{*}, \psi^{*}\right)$在$(t,z)\in\mathbb{R}$几乎处处满足

$\begin{cases}\partial_{t}\phi^{*}(t,z)=d_{1}\partial_{zz}\phi^{*}(t,z)-c\partial_{z}\phi^{*}(t,z)-\beta(t)f(\phi^{*}(t,z),\psi^{*}(t,z)),\\\partial_t\psi^*(t,z)=d_2\partial_{zz}\psi^*(t,z)-c\partial_z\psi^*(t,z)-\gamma(t)\psi^*(t,z)\\ +\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\phi^*(t-\tau,z-c\tau-y),\psi^*(t-\tau,z-c\tau-y)){\rm d}y.\end{cases}$

考虑如下柯西问题

$\begin{cases}\partial_{t}\upsilon_{1}(t,z)=d_{1}\partial_{zz}\upsilon_{1}(t,z)-c\partial_{z}\upsilon_{1}(t,z)-\beta(t)f(\upsilon_{1}(t,z),\upsilon_{2}(t,z)),\\\partial_t\upsilon_{2}(t,z)=d_2\partial_{zz}\upsilon_{2}(t,z)-c\partial_z\upsilon_{2}(t,z)-\gamma(t)\upsilon_{2}(t,z)\\ +\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\upsilon_{1}(t-\tau,z-c\tau-y),\upsilon_{2}(t-\tau,z-c\tau-y)){\rm d}y,\\\upsilon_1(0,z)=\phi^*(0,z),\upsilon_2(0,z)=\psi^*(0,z),\text{}z\in\mathbb{R}.\end{cases}$

由文献[13,定理 5.1.3、定理 5.1.4] 可知$\left(\phi^{*}, \psi^{*}\right)$是 (2.32) 式的唯一解, 进而, 对于某个$\theta\in(0,1),$$\phi^*,\psi^* \in C^{1+\frac{\theta}{2},2+\theta}(\mathbb{R}\times\mathbb{R},\mathbb{R})$满足行波方程 (1.8), 即

$\begin{cases}\partial_{t}\phi^{*}(t,z)=d_{1}\partial_{zz}\phi^{*}(t,z)-c\partial_{z}\phi^{*}(t,z)-\beta(t)f(\phi^{*}(t,z),\psi^{*}(t,z)),\\\partial_t\psi^*(t,z)=d_2\partial_{zz}\psi^*(t,z)-c\partial_z\psi^*(t,z)-\gamma(t)\psi^*(t,z)\\+\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\phi^*(t-\tau,z-c\tau-y),\psi^*(t-\tau,z-c\tau-y)){\rm d}y,\end{cases}$

且有

$\begin{matrix}\left\|\phi^*\right\|_{C^{1+\frac{\theta}{2},{2+\theta}}(\mathbb{R}\times\mathbb{R},\mathbb{R})}+\left\|\psi^*\right\|_{C^{1+\frac{\theta}{2},{2+\theta}}(\mathbb{R}\times\mathbb{R},\mathbb{R})}<\infty.\end{matrix}$

证毕.

定理 2.2 假设$\mathfrak{R}_0>1,c>c^*\text{},$则系统$(1.8)$的周期解满足

$\begin{eqnarray*}\lim_{z\to-\infty}\phi^*(t,z)=S_0,\lim_{z\to-\infty}\psi^*(t,z)=0,\lim_{z\to+\infty}\phi^*\left(t,z\right)=S^\infty,\text{}\lim_{z\to+\infty}\psi^*\left(t,z\right)=0.\end{eqnarray*}$

由定理 2.1 可知

$\begin{eqnarray*}\phi^{-}(t,z)\le\phi^{*}(t,z)\le S_{0},\psi^{-}(t,z)\le\psi^{*}(t,z)\le\psi^{+}(t,z),(t,z)\in\mathbb{R}^{2}.\end{eqnarray*}$

因此, 对任意的$t\in\mathbb{R}$, 有$\lim\limits_{z\to-\infty}\phi^*(t,z)=S_0,\lim\limits_{z\to-\infty}\psi^*(t,z)=0$一致成立. 由 (2.34) 式和文献[16]中的 Landau 型不等式可知, 当$\phi^*(t,z):[T]\times(-\infty,N]\rightarrow\mathbb{R}$关于$z$二阶可微且$\phi^*(t,z),\partial_{zz}\phi^*(t,z)\in L^{p}([T]\times(-\infty,N]),p\in[\infty]$时, 有$\partial_{z}\phi^*(t,z)\in L^{p}([T]\times(-\infty,N])$, 其中$N\in\mathbb{R}_{+}$是正常数 (且可以充分大). 并且, 当$p=\infty$, 常数$C_{\infty}(\mathbb{R}_{+})=2$时, 有

$\begin{eqnarray*}\left\|\partial_z\phi^*(t,z)\right\|_{L^\infty([T]\times(-\infty,N])}\leq2\left\|\phi^*(t,z)\!-\!S_0\right\|^{\frac{1}{2}}_{L^\infty([T]\times(-\infty,N])}\left\|\partial_{z z}\phi^*(t,z)\right\|_{L^\infty([T]\times(-\infty,N])}^{\frac{1}{2}}.\end{eqnarray*}$

同理可得

$\begin{eqnarray*}\left\|\partial_z\psi^*(t,z)\right\|_{L^\infty([T]\times(-\infty,N])}\leq2\left\|\psi^*(t,z)\right\|^{\frac{1}{2}}_{L^\infty([T]\times(-\infty,N])}\left\|\partial_{z z}\psi^*(t,z)\right\|_{L^\infty([T]\times(-\infty,N])}^{\frac{1}{2}}.\end{eqnarray*}$

因此, 对$t\in\mathbb{R}$, 有$\lim\limits_{z\to-\infty}(\partial_z\phi^*,\partial_z\psi^*)=(0,0)$一致成立. 进一步, 对 (2.33) 式两边关于$z$微分, 可得

$\begin{cases}\partial_t\left(\partial_z\phi^*\right)=d_1\partial_{zz}\left(\partial_z\phi^*\right)-c\partial_z\left(\partial_z\phi^*\right)-\beta(t)[\partial_1f(\phi^*,\psi^*)\partial_z\phi^*+\partial_2f(\phi^*,\psi^*)\partial_z\psi^*],\\ \partial_t\left(\partial_z\psi^*\right)=d_2\partial_{zz}\left(\partial_z\psi^*\right)-c\partial_z\left(\partial_z\psi^*\right)-\gamma(t)\left(\partial_z\psi^*\right)+\int_\mathbb{R}\Gamma\left(t,t-\tau;y\right)\beta\left(t-\tau\right)\\ \times[\partial_1f(\phi^*(t-\tau,z-c\tau-y),\psi^*(t-\tau,z-c\tau-y))\partial_z\phi^*(t-\tau,z-c\tau-y)\\+\partial_2f(\phi^*(t-\tau,z-c\tau-y),\psi^*(t-\tau,z-c\tau-y))\partial_z\psi^*(t-\tau,z-c\tau-y)]{\rm d}y.\\ \end{cases}$

类似 (2.34) 式的证明, 由$\phi^*,\psi^*$的周期性及文献[13,定理 5.1.3、定理 5.1.4] 可知, 对任意的$\theta\in(0,1)$, 有

$\begin{eqnarray*}\left\|\partial_z\phi^*\right\|_{C^{1+\frac{\theta}{2},2+\theta}(\mathbb R\times\mathbb R,\mathbb R)}<+\infty, \left\|\partial_z\psi^*\right\|_{C^{1+\frac{\theta}{2},2+\theta}(\mathbb R\times\mathbb R,\mathbb R)}<+\infty.\end{eqnarray*}$

同理, 由 Landau 型不等式可知$\lim\limits_{z\to-\infty}(\partial_{zz}\phi^*,\partial_{zz}\psi^*)=(0,0)$关于$t\in\mathbb{R}$一致成立.

令$\Phi(z)=\frac{1}{T}\int_0^T\phi^*(t,z)\text{}{\rm d}t$, 由 (2.33) 式的第一个方程可知,

$\begin{matrix}c\partial_z\Phi=d_1\partial_{zz}\Phi-\frac{1}{T}\int_0^T\beta(t)f(\phi^*(t,z),\psi^*(t,z)){\rm d}t.\end{matrix}$

因为$\Phi(-\infty)=S_{0}$且$\lim\limits_{z\to-\infty}\partial_z\Phi(z)=0$, 对 (2.35) 式从$-\infty$到$z$关于$z$积分, 可得

$\begin{matrix}d_1\partial_z\Phi(z)=c[\Phi(z)-S_0]+\frac{1}{T}\int_{-\infty}^z\int_0^T\beta(t)f(\phi^*(t,y),\psi^*(t,y)){\rm d}t{\rm d}y.\end{matrix}$

由于$\phi^*(t,z),\partial_z\phi^*(t,z)$一致有界, 则$\Phi(z)=\frac{1}{T}\int_0^T\phi^*(t,z){\rm d}t, \partial_{z}\Phi(z)=\frac{1}{T}\int_0^T\partial_{z}\phi^*(t,z){\rm d}t$一致有界, 从而$\int_0^T\beta(t)f(\phi^*(t,z), \psi^*(t,z)){\rm d}t$在$\mathbb{R}$上可积. 对 (2.35) 式两边同乘$\frac{1}{d_1}{\rm e}^{-\frac{cz}{d_1}}$并从$z$到$+\infty$关于$z$积分, 可得

$\begin{eqnarray*}{\rm e}^{-\frac{cz}{d_1}}\partial_z\Phi(z)=-\frac{1}{d_1T}\int_z^{\infty}{\rm e}^{-\frac{cy}{d_1}}\int_0^T\beta(t)f(\phi^*(t,y),\psi^*(t,y)){\rm d}t{\rm d}y.\end{eqnarray*}$

因此, 对任意的$z\in\mathbb{R}$, 有$\partial_z\Phi(z)<0$. 又因为$\Phi(z)$是一致有界的, 所以$\Phi(+\infty)$存在且$\Phi(+\infty)<\Phi(-\infty)=S_{0}$.由 Barbalat's 引理[17]可知, 当$z\rightarrow+\infty$时,$\partial_z\Phi(z)\rightarrow0$. 在 (2.36) 式中令$z\rightarrow+\infty$, 则

$\begin{matrix}\frac{1}{T}\int_{-\infty}^{+\infty}\int_0^T\beta(t)f(\phi^*(t,z),\psi^*(t,z)){\rm d}t{\rm d}z=c[S_0-\Phi(+\infty)]=c[S_0-S^\infty],\end{matrix}$

其中$S^\infty:=\Phi(+\infty)<S_0$.

下证对任意的$t\in\mathbb{R},$有$\lim\limits_{z\to+\infty}\psi^{*}(t,z)=0$一致成立. 令$\Psi(z)=\frac{1}{T}\int_0^T{\psi^{*}(t,z){\rm d}t},$则$\Psi(z)$满足

$\begin{matrix} &-d_2\partial_{zz}\Psi(z)+c\partial_{z}\Psi(z)+\hat\gamma\Psi(z)\notag\\=\ &\frac{1}{T}\int_0^T{\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\phi^{*}(t-\tau,z-c\tau-y),\psi^{*}(t-\tau,z-c\tau-y)){\rm d}y{\rm d}t}\notag\\&-\frac{1}{T}\int_0^T(\gamma(t)-\hat\gamma)\psi^{*}(t,z){\rm d}t,\end{matrix}$

其中$\hat\gamma:=\mathop{\min}\limits_{t\in[T]}\gamma(t).$令$\hat\lambda^{\pm}:=\frac{c\pm\sqrt{c^2+4d_2\hat\gamma}}{2d_2}$是特征方程$-d_2\lambda^2+c\lambda+\hat\gamma=0$的两相异实根. 定义$\hat\rho:=d_2(\hat\lambda^{+}-\hat\lambda^{-})=\sqrt{c^2+4d_2\hat\gamma}.$显然,$\hat\lambda^{-}<0<\hat\lambda^{+}.$则由 (2.38) 式, 可得

$\begin{matrix} \Psi(z)=\ &\frac{1}{\hat\rho T}\int_{-\infty}^{z}{\rm e}^{\hat\lambda^{-}(z-\eta)}[-\int_{0}^{T}(\gamma(t)-\hat\gamma)\psi^{*}(t,\eta){\rm d}t \notag\\&+\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\phi^{*}(t-\tau,\eta-c\tau-y),\psi^{*}(t-\tau,\eta-c\tau-y)){\rm d}y{\rm d}t]{\rm d}\eta \notag\\&+\frac{1}{\hat\rho T}\int_{z}^{\infty}{\rm e}^{\hat\lambda^{+}(z-\eta)}[-\int_{0}^{T}(\gamma(t)-\hat\gamma)\psi^{*}(t,\eta){\rm d}t \notag\\&+\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\phi^{*}(t-\tau,\eta-c\tau-y),\psi^{*}(t-\tau,\eta-c\tau-y)){\rm d}y{\rm d}t]{\rm d}\eta \notag\\\le\ &\frac{1}{\hat\rho T}\int_{-\infty}^{z}{\rm e}^{\hat\lambda^{-}(z-\eta)}\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau) \notag\\&\times f(\phi^{*}(t-\tau,\eta-c\tau-y),\psi^{*}(t-\tau,\eta-c\tau-y)){\rm d}y{\rm d}t{\rm d}\eta \notag\\&+\frac{1}{\hat\rho T}\int_{z}^{\infty}{\rm e}^{\hat\lambda^{+}(z-\eta)}\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)\notag\\&\times f(\phi^{*}(t-\tau,z-c\tau-y),\psi^{*}(t-\tau,z-c\tau-y)){\rm d}y{\rm d}t{\rm d}\eta \notag\\=\ &\frac{1}{\hat\rho T}\int_{0}^{+\infty}{\rm e}^{\hat\lambda^{-}\eta}\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau) \notag\\&\times f(\phi^{*}(t-\tau,z-\eta-c\tau-y),\psi^{*}(t-\tau,z-\eta-c\tau-y)){\rm d}y{\rm d}t{\rm d}\eta \notag\\&+\frac{1}{\hat\rho T}\int_{-\infty}^{0}{\rm e}^{\hat\lambda^{+}\eta}\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau) \notag\\&\times f(\phi^{*}(t-\tau,z-\eta-c\tau-y),\psi^{*}(t-\tau,z-\eta-c\tau-y)){\rm d}y{\rm d}t{\rm d}\eta,\end{matrix}$

将 (2.39) 式从$\zeta$到$\varsigma,$可得

$\begin{align*}\int_{\zeta}^{\varsigma}\Psi(z){\rm d}z\le\ &\frac{1}{\hat\rho T}\int_{0}^{+\infty}{\rm e}^{\hat\lambda^{-}\eta}\int_{\zeta}^{\varsigma}\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)\notag\\&\times f(\phi^{*}(t-\tau,z-\eta-c\tau-y),\psi^{*}(t-\tau,z-\eta-c\tau-y)){\rm d}y{\rm d}t{\rm d}z{\rm d}\eta\notag\\&+\frac{1}{\hat\rho T}\int_{-\infty}^{0}{\rm e}^{\hat\lambda^{+}\eta}\int_{\zeta}^{\varsigma}\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)\notag\\&\times f(\phi^{*}(t-\tau,z-\eta-c\tau-y),\psi^{*}(t-\tau,z-\eta-c\tau-y)){\rm d}y{\rm d}t{\rm d}z{\rm d}\eta,\notag\end{align*}$

因为$\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)f(\phi^{*}(t-\tau,z-c\tau-y),\psi^{*}(t-\tau,z-c\tau-y)){\rm d}y{\rm d}t$在$\mathbb{R}$上可积, 根据 Fubini 定理[18]可知,$\Psi(z)$在$\mathbb{R}$上可积, 并有

$\begin{align}\int_{\zeta}^{\varsigma}\Psi(z){\rm d}z\le\ &\frac{1}{\hat\rho T}(\frac{1}{\hat\lambda^{+}}-\frac{1}{\hat\lambda^{-}})\int_{\zeta}^{\varsigma}\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)\notag\\&\times f(\phi^{*}(t-\tau,z-c\tau-y),\psi^{*}(t-\tau,z-c\tau-y)){\rm d}y{\rm d}t{\rm d}z,\notag\end{align}$

$\begin{align}\int_{-\infty}^{+\infty}\Psi(z){\rm d}z\le\ &\frac{1}{\hat\rho T}(\frac{1}{\hat\lambda^{+}}-\frac{1}{\hat\lambda^{-}})\int_{-\infty}^{+\infty}\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)\notag\\&\times f(\phi^{*}(t-\tau,z-c\tau-y),\psi^{*}(t-\tau,z-c\tau-y)){\rm d}y{\rm d}t{\rm d}z\notag\\=\ &\frac{1}{\hat\gamma T}\int_{-\infty}^{+\infty}\int_{0}^{T}\int_{\mathbb{R}}\Gamma(t,t-\tau;y)\beta(t-\tau)\notag\\&\times f(\phi^{*}(t-\tau,z-c\tau-y),\psi^{*}(t-\tau,z-c\tau-y)){\rm d}y{\rm d}t{\rm d}z,\notag\end{align}$

因为$\partial_z\psi^{*}(t,z)$一致有界, 所以$\partial_z\Psi(z)$在$\mathbb{R}$上也有界. 因此由 Barbalat's 引理可知, 当$z\to+\infty$时,$\Psi(z)\to0.$另外, 由 (2.34) 式和$\psi^{\ast}\in C^{1+\frac{\theta}{2},2+\theta}$, 可知$\partial_t\psi^*(t,z)$在$[T]\times\mathbb{R}$上一致有界, 进而由$\psi^*(t,z)$的非负性和周期性可得, 当$z\to+\infty$时, 对任意的$t\in\mathbb{R},$有$\psi^*(t,z)\to0$一致成立.

进一步, 可以证明对任意的$t\in\mathbb{R},$有$\lim\limits_{z\to+\infty}\phi^{*}(t,z)=S^{\infty}$一致成立. 由于$\phi^{*}$有界则上下极限必存在, 故令

$\begin{matrix}\limsup\limits_{z\to+\infty}\mathop {\max }\limits_{t \in [T]} \phi^{*}(t,z)=:S_{+}^{\infty}, \liminf\limits_{z\to+\infty}\mathop {\max }\limits_{t \in [T]} \phi^{*}(t,z)=:S_{-}^{\infty}\end{matrix}$

结合$\phi^*$的$T$周期性, 仅需证明$S_{+}^{\infty}=S^{\infty}=S_{-}^{\infty}$即可. 由 (2.40) 式可知, 存在序列$\{t_n\}$和$\{z_n\},$当$n\to+\infty$时有$\{t_n\}\subset[T]$且$z_n\to+\infty,$满足$\lim\limits_{n\to+\infty}\phi^{*}(t_n,z_n)=S_+^\infty.$令

$\phi_n(t,z):=\phi^*(t+t_n,z+z_n), \psi_n(t,z):=\psi^*(t+t_n,z+z_n), n\in\mathbb{N}_+, t\in\mathbb{R},z\in\mathbb{R}.$

由 (2.34) 式可知, 存在函数列$(\phi_n(t,z),\psi_n(t,z))$的子列 (仍记为$(\phi_n(t,z),\psi_n(t,z))$), 使得当$n\to+\infty$时,$(\phi_n(t,z),\psi_n(t,z))\to(\phi_*(t,z),0)\in C_{loc}^{1+\frac{\theta}{2},2+\theta}(\mathbb{R}\times\mathbb{R}).$

注意到$\phi_*(0,0)=S_+^\infty,$且$\phi_*(t,z)$是$T$-周期函数, 并对任意的$(t,z)\in\ \mathbb{R}\times\mathbb{R},$有$\phi_*(t,z)\le S_+^\infty.$由于$\{t_n\}\subset[T],$故存在$t^*\subset[T]$使得当$n\to+\infty$时$t_n\to t^*.$令$\phi_*^+(t,z):=\phi_*(t-t^*,z),$由 (2.21) 式可知,$\phi_*^+(t,z)$满足

$\begin{align}\phi_*^+(t)&=T_1(t)\phi_*^+(0)+\int_0^{t}T_1(t-s)f_1[\phi_*^+,0](s){\rm d}s\notag\\&=T_1(t)\phi_*^+(0)+\int_0^{t}T_1(t-s)\alpha_1\phi_*^+ (s){\rm d}s,\notag\end{align}$

且$\partial_t\phi_*^+=d_1\partial_{zz}\phi_*^+ -c\partial_z\phi_*^+.$由于$\phi_*^+(t^*,0)=S_+^\infty$和$\phi_*^+(t^*,z)\le S_+^\infty,$则由最大值原理可知, 当$t<t^*$时, 有$\phi_*^+(t,z)=S_+^\infty.$即对任意的$t\in\mathbb{R},$有$\frac{1}{T}\int_0^{T}\phi_*^+(t,z){\rm d}t=S_+^\infty.$下面, 记$\Phi_*^+(z):=\frac{1}{T}\int_0^{T}\phi_*^+(t,z){\rm d}t,$则$\Phi_*^+(z)=S_+^\infty.$同时, 有

$\begin{align}\Phi_*^+(z)&=\frac{1}{T}\int_0^{T}\phi_*^+(t,z){\rm d}t=\frac{1}{T}\int_0^{T}\phi_*(t-t^*,z){\rm d}t=\lim\limits_{n\to+\infty}\frac{1}{T}\int_0^{T}\phi_n(t-t^*,z){\rm d}t\notag\\&=\lim\limits_{n\to+\infty}\frac{1}{T}\int_0^{T}\phi^*(t-t^*+t_n,z+z_n){\rm d}t=S_+^\infty,\notag\end{align}$

又因为$\Phi(+\infty)=S^\infty,$从而$S_+^\infty=S^\infty.$同理可证,$S_-^\infty=S^\infty.$因此$\lim\limits_{z\to+\infty}\phi^*(t,z)=S^\infty$关于$t\in\mathbb{R}$一致成立. 证毕.

3 周期行波解的不存在性

本节建立$\Re_{0}<1$时, 系统 (1.4) 周期行波解的不存在性.

定理 3.1 若$\Re_{0}<1,$则对任意的$c\geq0,$系统$(1.4)$不存在满足渐近边界条件$(1.9)$的周期行波解.

利用反证法, 则系统 (1.4) 存在满足渐近边界条件 (1.9) 的周期行波解$(\phi^*,\psi^*)$. 显然, 对任意的$t\in \mathbb{R}$, 总有$\phi(t,z)\leq S_0$, 因此, 当$t>0$,$z\in \mathbb{R}$时, 有

$\begin{matrix}\psi_{t}(t, z)=\ &d_{2} \psi_{z z}(t, z)-c \psi_{z}(t, z)-\gamma(t) \psi(t, z) \notag\\&+\int_{\mathbb{R}} \Gamma(t, t-\tau ; y) \beta(t-\tau) f(\phi(t-\tau, z-c \tau-y), \psi(t-\tau, z-c \tau-y)){\rm d}y \notag\\\leq\ &d_{2} \psi_{z z}(t, z)-c \psi_{z}(t, z)-\gamma(t) \psi(t, z) \notag\\&+\int_{\mathbb{R}} \Gamma(t, t-\tau ; y) \beta(t-\tau) \partial_{2} f\left(S_{0}, 0\right) \psi(t-\tau, z-c \tau-y){\rm d}y,\end{matrix}$

记$\xi:=\sup _{z \in \mathbb{R}} \psi(0, z)<\infty$, 则$\psi(0,z)\leq \xi$, 由比较原理可知, 对任意的$z\in \mathbb{R}$有

$\begin{matrix}\psi(t,z)\leq \omega(t;\xi),\end{matrix}$

这里$\omega(t;\xi)$是如下常微分初值问题

$\begin{cases}\frac{{\rm d}\omega(t)}{{\rm d}t}=-\gamma(t)\omega(t)+{\rm e}^{\int_{t-\tau}^{t}\gamma_L(s){\rm d}s}\beta(t-\tau)\partial_2f(S_0,0)\omega(t-\tau),&t>0,\\[2mm]\omega(t)=\xi,&t\in[-\tau,0]\end{cases}$

的解. 记$\rho_{0}$为方程 (3.3) Poincaré映射的谱半径, 则由$\Re_{0}<1$和文献[10,定理 2.1] 可知$\Re_{0}-1$与$\rho_{0}-1$符号相同, 即$\rho(0)=\rho_{0}<1$. 由文献[11,命题 2.1] 可知, 存在以$T$为周期的正函数$H_{\mu}(t)$使得$\omega(t;\xi)={\rm e}^{\lambda(0)t}H_{\mu}(t)$满足 (3.3), 其中$\lambda(0)=\frac{\ln \rho(0)}{T}<0$, 从而$t\rightarrow\infty$时$\omega(t;\xi)$收敛于$0$. 于是$\lim\limits_{t\rightarrow\infty}\psi(t,z)=0$关于$z\in \mathbb{R}$一致成立, 这与$\psi(t,\cdot)$不恒等于$0$产生了矛盾. 证毕.

参考文献

Alikakos N, Bates P W, Chen X.

Periodic traveling waves and locating oscillating patterns in multidimensional domains

Transactions of the American Mathematical Society, 1999, 351(7): 2777-2805

DOI:10.1090/tran/1999-351-07      URL     [本文引用: 1]

Bates P W, Chen F.

Periodic traveling waves for a nonlocal integro-differential model

Electronic Journal of Differential Equations, 1999, 26: 1-19

[本文引用: 1]

杨瑜.

一类非局部扩散的SIR模型的行波解

数学物理学报, 2022, 42A(5): 1409-1415

[本文引用: 1]

Yang Y.

Traveling wave solutions of a class of SIR model with nonlocal diffusion

Acta Math Sci, 2022, 42A(5): 1409-1415

[本文引用: 1]

王凯, 赵洪涌.

一类具有时滞的反应扩散登革热传染病模型的行波解

数学物理学报, 2022, 42A(4): 1209-1226

[本文引用: 2]

Wang K, Zhao H Y.

Traveling wave of a reaction-diffusion dengue epidemic model with time delays

Acta Math Sci, 2022, 42A(4): 1209-1226

[本文引用: 2]

Shen W X.

Traveling waves in time periodic lattice differential equations

Nonlinear Analysis, 2003, 54(2): 319-339

DOI:10.1016/S0362-546X(03)00065-8      URL     [本文引用: 2]

Zhang L, Wang Z C, Zhao X Q.

Time periodic traveling wave solutions for a Kermack-McKendrick epidemic model with diffusion and seasonality

Journal of Evolution Equations, 2020, 20(3): 1029-1059

DOI:10.1007/s00028-019-00544-2      [本文引用: 6]

Wang S M, Feng Z, Wang Z C, et al.

Periodic traveling wave of a time periodic and diffusive epidemic model with nonlocal delayed transmission

Nonlinear Analysis: Real World Applications, 2020, 55: 103117

DOI:10.1016/j.nonrwa.2020.103117      URL     [本文引用: 8]

Wu W X, Teng Z D.

The periodic traveling waves in a diffusive periodic SIR epidemic model with nonlinear incidence

Chaos, Solitons and Fractals, 2021, 144: 1-18

[本文引用: 4]

宋雪, 杨赟瑞, 杨璐.

带有外部输入项的时间周期SIR传染病模型的周期行波解

应用数学和力学, 2022, 43(10): 1164-1176

[本文引用: 1]

Song X, Yang Y R, Yang L.

Periodic traveling wave solutions of a time-period SIR epidemic model with external supplies

Applied Mathematics and Mechanics, 2022, 43(10): 1164-1176

[本文引用: 1]

Zhao X Q.

Basic reproduction ratios for periodic compartmental models with time delay

Journal of Dynamics and Differential Equations, 2017, 29(1): 67-82

DOI:10.1007/s10884-015-9425-2      URL     [本文引用: 5]

Xu D S, Zhao X Q.

Dynamics in a periodic competitive model with stage structure

Math Anal Appl, 2005, 311(2): 417-438

DOI:10.1016/j.jmaa.2005.02.062      URL     [本文引用: 2]

Liang X, Zhao X Q.

Asymptotic speeds of spread and traveling waves for monotone semiflows with applications. Comm.Pure Appl

Math., 2007, 60(1): 1-40

[本文引用: 1]

Lunardi A. Analytic Semigroups and Optimal Regularity in Parabolic Problems. Berlin: Springer, 2012

[本文引用: 7]

Lieberman G M.

Second Order Parabolic Differential Equations

Singapore: World Scientific, 1996

[本文引用: 2]

王明新. 索伯列夫空间. 北京: 高等教育出版社, 2013

[本文引用: 1]

Wang M X. Sobolev Space. Beijing: Higher Education Press, 2013

[本文引用: 1]

Barnett N S, Dragomir S S.

Some Landau type inequalities for functions whose derivatives are of locally bounded variation

Research report collection, 2005, 8(1). https://vuir.vu.edu.au/17413/ [2023-01-06]

URL     [本文引用: 1]

Hou M, Duan G, Guo M.

New versions of Barbalat's lemma with applications

Journal of Control Theory and Applications, 2010, 8(4): 545-547

DOI:10.1007/s11768-010-8178-z      URL     [本文引用: 1]

Dosla Z, Vrkoc I.

On an extension of the Fubini theorem and its applications in ODEs. Nonlinear Analysis: Theory

Methods Applications, 2014, 57(4): 531-548

[本文引用: 1]

/