Hostname: page-component-54dcc4c588-mz6gc Total loading time: 0 Render date: 2025-09-24T12:54:58.041Z Has data issue: false hasContentIssue false

Global dynamics of a degenerate reaction-diffusion cholera model with phage-bacteria interaction in a heterogeneous environment

Published online by Cambridge University Press:  22 September 2025

Wei Wang
Affiliation:
College of Mathematics and Systems Science, Shandong University of Science and Technology, Qingdao, Shandong 266590, China
Teng Liu
Affiliation:
College of Mathematics and Systems Science, Shandong University of Science and Technology, Qingdao, Shandong 266590, China
Hao Wang*
Affiliation:
Department of Mathematical and Statistical Sciences, University of Alberta, Edmonton, Alberta, T6G 2G1, Canada
*
Corresponding author: Hao Wang; Email: hao8@ualberta.ca
Rights & Permissions [Opens in a new window]

Abstract

To investigate multiple effects of the interaction between V. cholerae and phage on cholera transmission, we propose a degenerate reaction-diffusion model with different dispersal rates, which incorporates a short-lived hyperinfectious (HI vibrios) state of V. cholerae and lower-infectious (LI vibrios) state of V. cholerae. Our main purpose is to investigate the existence and stability analysis of multi-class boundary steady states, which is much more complicated and challenging than the case when the boundary steady state is unique. In a spatially heterogeneous case, the basic reproduction number $\mathscr{R}_{0}$ is defined as the spectral radius of the sum of two linear operators associated with HI vibrios infection and LI vibrios infection. If $\mathscr{R}_{0}\leq 1$, the disease-free steady state is globally asymptotically stable. If $\mathscr{R}_{0}\gt 1$, the uniform persistence of phage-free model, as well as the existence of the phage-free steady state, are established. In a spatially homogeneous case, when $\ \;\widetilde{\!\!\!\mathscr{R}}_{0}\gt 1$, the global asymptotic stability of phage-free steady state and the uniform persistence of the phage-present model are discussed under some additional conditions. The mathematical approach here has wide applications in degenerate Partial Differential Equations.

Information

Type
Papers
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2025. Published by Cambridge University Press

1 Introduction

Cholera, a waterborne disease caused by V. cholerae, can be found in diverse aquatic environments, such as the ocean, estuaries, rivers, and lakes [Reference Garay, Arnau and Amaro11, Reference Perez-Rosas and Hazen33, Reference Yamazaki, Yang and Wang59]. It is characterised by severe vomiting and diarrhoea, and if not promptly treated, the disease can lead to severe dehydration and death [Reference Codeço7, Reference Hove-Musekwa, Nyabadza, Chiyaka, Das, Tripathi and Mukandavire16, Reference Kierek and Watnick22]. This is attributed to the ability of V. cholerae to produce cholera toxin, which stimulates water and electrolyte secretion by intestinal endothelial cells [Reference Mari, Bertuzzo, Righetto, Casagrandi, Gatto, Rodriguez-Iturbe and Rinaldo27]. The primary symptoms of cholera include diarrhoea, dehydration, abdominal cramps, a drop in blood pressure and kidney failure [Reference Cai, Modnak and Wang5]. Besides, the dynamics of cholera epidemics involve a complex web of interactions between human hosts, pathogens and environments [Reference Wang and Wang51]. The disease is primarily transmitted to humans by ingesting water or food contaminated with toxigenic forms of V. cholerae O1 and O139 from the environment [Reference Colwell and Huq8, Reference Hartley, Morris and Smith14, Reference Kaper, Morris and Levine21]. Cholera outbreaks frequently arise in regions lacking access to antibiotics and adequate public health infrastructure, especially in developing countries with limited healthcare resources, such as the Indian subcontinent, parts of Asia, Africa and Latin America. Cholera remains a persistent health challenge [Reference Kong, Davis and Wang23, Reference Safi, Melesse and Gumel35].

Research on mathematical models of cholera can be traced back to 1973, when Capasso and Paveri-Fontana [Reference Capasso and Paveri-Fontana6] introduced an ordinary differential equation (ODE) model to study the spread of cholera in the Mediterranean region. The model described compartments for V. cholerae and infected individuals, investigating the transmission of cholera in the European Mediterranean region. Joh et al. first proposed an iSIR model describing indirectly transmitted infectious diseases with immunological threshold [Reference Joh, Wang, Weiss and Weitz20], with a follow up work [Reference Luo, Wang and Wang25] studying a seasonal forcing iSIR model with a smoothing immunological threshold. Tien and Earn [Reference Tien and Earn43] proposed a waterborne disease model incorporating both direct transmission and indirect transmission with bilinear incidence. Wu and Zou [Reference Wu and Zou58] examined a diffusive host-pathogen model that incorporates distinct dispersal rates for susceptible and infected hosts. They analysed the asymptotic profiles of the positive steady state as the dispersal rate of the susceptible or infected hosts tends to zero. The findings indicate that the infected hosts concentrate at certain points, which can be characterised as the pathogen’s most favoured sites when the mobility of the infected host is limited. Wang et al. [Reference Wang, Zhao and Wang52] proposed a new reaction-convection-diffusion model to investigate the spatiotemporal dynamics of cholera transmission. The model incorporates time-periodic parameters to describe the seasonality of the disease transmission and bacterial growth rates. Wang and Wang [Reference Wang and Wang45] proposed a reaction-diffusion cholera model incorporating the different dispersal rates of the susceptible and infected hosts in the absence of diffusion term for the cholera equation.

Recent laboratory findings in [Reference Andrews and Basu3, Reference Hartley, Morris and Smith14] suggested that the V. cholerae induces a short-lived, hyperinfectious (HI) state through the gastrointestinal tract and decays into a lower infectiousness (LI) state within hours [Reference Hartley, Morris and Smith14, Reference Nelson, Chowdhury, Flynn, Schild, Bourassa, Shao, LaRocque, Calderwood, Qadri, Camilli and Ausubel30]. Moreover, the infectivity of freshly shed V. cholerae greatly out-competes bacteria grown in vitro, exhibiting infectivity levels up to 700 times higher [Reference Nelson, Harris, Glenn Morris, Calderwood and Camilli31, Reference Wang and Wang51]. HI vibrios, being closer to human hosts than environmental vibrios, are more likely to come into contact with human susceptible individuals [Reference Hartley, Morris and Smith14]. To investigate hyperinfectious state of V. cholerae is crucial and also holds substantial practical significance. Furthermore, incorporating these hyperinfectious states and lower infectiousness states of V. cholerae into cholera disease models may lead to a better understanding of the observed cholera epidemic patterns.

Research on hyperinfectivity of V. cholerae has received increasing attention in recent years. Hartley et al. [Reference Hartley, Morris and Smith14] incorporated hyperinfectivity vibrios into the mathematical model. The results suggest that for minimising the epidemic spread of cholera, intervention measures should focus on minimising the transmission risk of short-lived, highly infectious cholera vibrios. Shuai et al. [Reference Shuai, Tien and van den Driessche39] investigated cholera dynamics with both hyperinfectivity and temporary immunity. Wang and Wang [Reference Wang and Wang51] introduced a novel modelling framework to investigate the impact of bacterial hyperinfectivity on cholera epidemics in a spatially heterogeneous environment. Specifically, this model categorised V. cholerae into HI vibrios compartment and LI vibrios compartment. Wang and Wu [Reference Wang and Wu46] extended the work in [Reference Wang and Wang45] by incorporating bacterial hyperinfectivity and saturation mechanism for indirect transmission pathway. Wang et al. [Reference Wang, Wu, Wang and Feng49] developed a reaction-advection-diffusion model with a general boundary conditions, considering HI and LI vibrio strains, convection factors, and human behaviour change, to establish the threshold-type results of cholera transmission in spatial–temporal heterogeneous environment (see also [Reference Wang, Wang and Fan53]). Wang et al. [Reference Wang, Wang and Wang48] formulated a generalised cholera model incorporating nonlocal time delays to investigate the effects of bacterial hyperinfectivity on cholera outbreaks and to derive the detailed classifications of global dynamics in a spatially heterogeneous environment.

Phages, viruses that specifically infect and destroy bacteria, have been characterised as bacterial parasites, with each phage type exhibiting a distinct ability to replicate within specific strains of host bacteria [Reference Wang and Wang55]. The interaction mechanism between bacteriophages and bacteria begins when a lytic phage inserts its genetic material into a bacterial cell, where it proliferates, leading to cell lysis and the release of new phages into the environment [Reference Botelho, Kong, Lucien, Shuai and Wang4]. This process can significantly impact the severity of cholera outbreaks. For example, the investigation of the cholera epidemic in Dhaka, Bangladesh, indicates that lytic bacteriophages may mitigate epidemic severity by eliminating bacteria in both reservoirs and infected individuals [Reference Faruque, Naser, Islam, Faruque, Ghosh, Nair, Sack and Mekalanos10, Reference Jensen, Faruque, Mekalanos and Levin19].

Phages (viruses of bacteria) play a pivotal role in shaping both the evolution and dynamics of bacterial species, especially for V. cholerae [Reference Jensen, Faruque, Mekalanos and Levin19, Reference Kong, Davis and Wang23, Reference Nelson, Chowdhury, Flynn, Schild, Bourassa, Shao, LaRocque, Calderwood, Qadri, Camilli and Ausubel30, Reference Netter, Dunham and Seed32]. Kong et al. [Reference Kong, Davis and Wang23] proposed an ODE model incorporating a Holling II response function to depict the interaction between V. cholerae and bacteriophages. Misra et al. [Reference Misra and Gupta29] investigated a reaction-diffusion system for the biological control of cholera epidemics. They focused on temporal evolution of cholera within a region and explored its control using lytic bacteriophage in aquatic reservoirs. Botelho et al. [Reference Botelho, Kong, Lucien, Shuai and Wang4] proposed an ODE model with the bacteria-phage interaction of Holling type I. This model includes human populations (SIRS), bacteria population (B) and phage population (p) to represent these interactions. In their study, they derived threshold parameters to characterise the stability of equilibria. The findings suggest that the reservoir environment might contribute to the periodicity of cholera outbreaks. Hu et al. [Reference Hu, Wang and Nie18] proposed a cholera model comprising coupled reaction-diffusion equations and ODEs to discuss the effects of spatial heterogeneity, horizontal transmission, environmental viruses and phages on the spread of V. cholerae.

The aim of the paper is to investigate multiple effects of the interaction between V. cholerae and phage on cholera transmission, thereby improving our understanding of the transmission mechanism of cholera diseases and proposing targeted disease control measures. Generally speaking, it is very challenging to discuss the threshold-type results in the case of multi-class steady states. Fortunately, in this paper, we derive the existence and stability analysis of multi-class steady states for some special cases.

The remainder of the paper is organised as follows. In Section 2, we propose a degenerate reaction-diffusion model with different dispersal rates, which incorporates short-lived hyperinfectious (HI vibrios) state of V. cholerae and lower-infectious (LI vibrios) state of V. cholerae simultaneously in a heterogeneous environment. In Section 3, we present the main results of this paper, including the well-posedness, dynamics of the disease-free steady state, dynamics of the phage-free steady state and dynamics of the phage-present steady state. In Section 4, we give the proofs of the main results. A brief discussion of this paper is given in Section 5.

2 Mathematical model

Figure 1. Schematic diagram of model (2.1). The green solid line represents the recruitment and mortality rates of human hosts and V. cholerae, and the blue solid line denotes the direct development of cholera. The purple dashed line represents the infection process and the interaction between phages and V. cholerae.

Building upon the model presented by Jensen et al. [Reference Jensen, Faruque, Mekalanos and Levin19], which integrates cholera epidemiology with bacterial and bacteriophage population dynamics, we examine the interaction between HI vibrios and LI vibrios with bacteriophages, as well as the intrinsic growth rate of V. cholerae, and divide the infected human hosts into two parts, one consists of human hosts infected only with V. cholerae, denoted as $I_{1}$ , while the other consists of human hosts that are simultaneously infected with V. cholerae and bacteriophages, indicating the parasitism of bacteriophages within the host cells (bacteria), denoted as $I_{2}$ . Based on the above considerations, we propose a degenerate reaction-diffusion cholera model:

(2.1) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial S}{\partial t}= d_{S}\Delta S +\Lambda (x) - \alpha _{1}(x) f_{1}(B_{1})S -\alpha _{2}(x) f_{2}(B_{2})S -\mu (x) S,\\[3pt] &\dfrac {\partial I_{1}}{\partial t}= d_{I}\Delta I_{1} + \alpha _{1}(x)\left (1-\dfrac {P}{L+P}\right ) f_{1}(B_{1})S +\alpha _{2}(x)\left (1-\dfrac {P}{L+P}\right ) f_{2}(B_{2})S -\mu (x) I_{1},\\[3pt] &\dfrac {\partial I_{2}}{\partial t}= d_{I}\Delta I_{2} + \alpha _{1}(x)\dfrac {P}{L+P} \, f_{1}(B_{1})S + \alpha _{2}(x)\dfrac {P}{L+P} \, f_{2}(B_{2})S-\mu (x) I_{2},\\[3pt] &\dfrac {\partial B_{1}}{\partial t}=h_{1}(x,B_{1})+\eta (x) (I_{1}+I_{2}) -b_{1}(x)B_{1}P -\delta _{1}(x) B_{1},\\[3pt] &\dfrac {\partial B_{2}}{\partial t}=h_{2}(x,B_{2}) +\delta _{1}(x) B_{1} -b_{2}(x)B_{2}P- \delta _{2}(x) B_{2},\\[3pt] &\dfrac {\partial P}{\partial t}=\alpha (x)\eta (x) I_{2}+ \chi _{1}(x) b_{1}(x)B_{1}P +\chi _{2}(x) b_{2}(x)B_{2}P-m(x)P. \end{aligned} \right . \end{align}

The population density of susceptible individuals at location $x$ and time $t$ is denoted by $S(x,t)$ . The population densities of phage-negative and phage-positive infected individuals at location $x$ and time $t$ are denoted by $I_{1}(x,t),I_{2}(x,t)$ , respectively. Let $ I(x,t)=I_{1}(x,t)+I_{2}(x,t)$ , where $ I(x,t)$ denotes all human hosts infected with V. cholerae. The concentrations of HI and LI vibrios in the water environment at location $x$ and time $t$ are denoted by $ B_{1}(x,t), B_{2}(x,t)$ , respectively. The concentration of phage in the water environment at location $x$ and time $t$ is denoted by $P(x,t)$ . The recruitment rate of susceptible human hosts is represented by $\Lambda (x)$ . The parameters $\alpha _{1}(x)$ and $\alpha _{2}(x)$ can be interpreted as the rates of HI and LI vibrios consumption. $L(x)$ represents the half-saturation concentration of phage. For simplicity, we consider only natural deaths by $\mu (x)$ and disregard deaths caused by the disease. The functions $ h_{1}(x,B_{1})$ and $ h_{2}(x,B_{2})$ are the intrinsic growth rates of HI and LI vibrios, respectively. The rate of bacterial shedding is represented by $\eta (x)$ , while $\delta _{1}(x), \delta _{2}(x)$ denote the natural death rate of HI and LI vibrios, respectively. Phage interacts with both HI and LI vibrios, resulting in bacterial death rates of $ b_{1}(x)$ and $b_{2}(x)$ , respectively. Meanwhile, the phage has a gain from two vibrios’ deaths represented by $\chi _{1}(x)$ and $ \chi _{2}(x)$ . The mean phage shed rate is denoted as $\alpha (x)$ , and $m(x)$ represents the phage decay rate. The cholera transmission process is shown in Figure 1. In model (2.1), we choose

\begin{align*} f_{i}(B_{i})=\frac {B_{i}}{B_{i}+H_{i}(x)},\ i=1,2,\ x\in \Omega , \end{align*}

where $H_{i}(x)$ denotes the half-saturation concentration of bacteria. $ d_{S}, d_{I}$ represent the dispersal rates of susceptible and infected human hosts, respectively. Here, we assume that the dispersal rate $d_{I}$ for both phage-negative and phage-positive infections are equal. We also consider an isolated habitat $\Omega$ , which is characterised by the homogeneous Neumann boundary condition.

(2.2) \begin{equation} \dfrac {\partial S}{\partial \nu }=\dfrac {\partial I_{1}}{\partial \nu }=\dfrac {\partial I_{2}}{\partial \nu }=0,\ \ x\in \partial \Omega ,\ \ t\gt 0, \end{equation}

and the initial conditions are

(2.3) \begin{equation} S(x,0)=S^{0}(x),\ I_{1}(x,0)=I_{1}^{0}(x),\ I_{2}(x,0)=I_{2}^{0}(x),\ B_{1}(x,0) = B_{1}^{0}(x),\ B_{2}(x,0) = B_{2}^{0}(x),\ P(x,0) = P^{0}(x), \end{equation}

for $ x\in \bar {\Omega }$ , and where $ S^{0}(x),I_{1}^{0}(x),I_{2}^{0}(x),B_{1}^{0}(x),B_{2}^{0}(x),P^{0}(x)$ are nonnegative continuous functions. Furthermore, if spatial heterogeneity is not considered, model (2.1) degenerates to the homogeneous model:

(2.4) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial S}{\partial t}= d_{S}\Delta S +\Lambda - \alpha _{1} f_{1}(B_{1})S -\alpha _{2} f_{2}(B_{2})S -\mu S,\quad x\in \Omega ,\ t\gt 0, \\[3pt] &\dfrac {\partial I_{1}}{\partial t}= d_{I}\Delta I_{1} + \alpha _{1}\left (1-\dfrac {P}{L+P}\right ) f_{1}(B_{1})S+ \alpha _{2}\left (1-\dfrac {P}{L+P}\right ) f_{2}(B_{2})S -\mu I_{1},\quad x\in \Omega ,\ t\gt 0, \\[3pt] &\dfrac {\partial I_{2}}{\partial t}= d_{I}\Delta I_{2} + \alpha _{1}\dfrac {P}{L+P} \, f_{1}(B_{1})S + \alpha _{2}\dfrac {P}{L+P} \, f_{2}(B_{2})S-\mu I_{2},\quad x\in \Omega ,\ t\gt 0, \\[3pt] &\dfrac {\partial B_{1}}{\partial t}=h_{1}(B_{1}) +\eta (I_{1}+I_{2}) -b_{1}B_{1}P -\delta _{1} B_{1},\quad x\in \Omega ,\ t\gt 0, \\[3pt] &\dfrac {\partial B_{2}}{\partial t}=h_{2}(B_{2}) +\delta _{1} B_{1} -b_{2}B_{2}P- \delta _{2} B_{2},\quad x\in \Omega ,\ t\gt 0, \\[3pt] &\dfrac {\partial P}{\partial t}=\alpha \eta I_{2}+ \chi _{1} b_{1}B_{1}P +\chi _{2} b_{2}B_{2}P -mP,\quad x\in \Omega ,\ t\gt 0. \end{aligned} \right . \end{align}

In the following, we make some basic assumptions:

  1. (H1 ) $ d_{S}, d_{I}$ are positive $ C^{1}$ -function on $\bar {\Omega }$ ;

  2. (H2 ) $ (I_{1}^{0},I_{2}^{0},B_{1}^{0},B_{2}^{0},P^{0})\not \equiv 0$ on $\bar {\Omega }$ ;

  3. (H3 ) $ h_{1}(x,v), h_{2}(x,v) \in C^{0,1}(\bar {\Omega }\times \mathbb{R}_{+})$ are nonnegative and strictly concave down in relation to the second variable, and $ h_{i}(x,v)=0, i= 1, 2,$ if and only if $ v=0$ , then

    (2.5) \begin{equation} \lim \limits _{v\to \infty }\dfrac {h_{i}(x,v)}{v}\lt \delta _{i}(x), \ i=1,2, \ x\in \Omega . \end{equation}

For the assumption ( $ \mathbf{H}_{\mathbf{3}}$ ), we also refer to [Reference Botelho, Kong, Lucien, Shuai and Wang4, Reference Wang and Wang51], these general incidence functions are set to be

\begin{align*} h_{i}(x,B_{i})=\theta (x)B_{i}\left (1-\dfrac {B_{i}}{H_{B_{i}}(x)}\right ),\ i=1,2,\ x\in \Omega , \end{align*}

where $\theta (x)$ is the intrinsic growth rate of bacteria, and $H_{B_{i}}(x)$ denotes the maximum capacity of the bacteria.

It should be pointed out that since multiple effects of the interaction between V. cholerae and phages on cholera transmission, the complexity of model (2.1)–(2.3) leads to several mathematical difficulties:

  1. (i) We prove the global asymptotic stability of the disease-free steady state for the critical case when $\mathscr{R}_{0}=1$ for the high-dimensional system, which is instituted by six equations.

  2. (ii) Generally speaking, it is very challenging to discuss the threshold-type results in the case of multi-class steady states. Fortunately, in this paper, we derive the existence and stability analysis of multi-class steady states for some special cases. We show the existence of phage-free steady state in a heterogeneous environment. An appropriate Lyapunov function is constructed to discuss the global stability of the phage-free steady state in a homogeneous environment.

3 Main results

In this section, we state the main results of this paper, whose proofs are given in Section 4.

3.1 Well-posedness of the model

Define $\mathit{H}=C(\bar {\Omega },\mathbb{R}^{6})$ , which is assigned the following supremum norm:

\begin{align*} \|\varphi \|_{\mathit{H}}:=\text{max}\left \{\sup _{x\in \bar {\Omega }}|\varphi _{1}(x)|, \sup _{x\in \bar {\Omega }}|\varphi _{2}(x)|, \sup _{x\in \bar {\Omega }}|\varphi _{3}(x)|,\sup _{x\in \bar {\Omega }}|\varphi _{4}(x)|,\sup _{x\in \bar {\Omega }}|\varphi _{5}(x)|,\sup _{x\in \bar {\Omega }}|\varphi _{6}(x)|\right \}, \end{align*}

with $ \varphi =(\varphi _{1},\varphi _{2},\varphi _{3},\varphi _{4},\varphi _{5},\varphi _{6})\in \mathit{H}.$ Define $H^{+}=C(\bar {\Omega },\mathbb{R}^{6}_{+})$ as the positive cone of $\mathit{H}$ . Let $L_{p}(\Omega )$ be the Banach space of function $y$ whose $p$ -th power of absolution value is integrable on $\Omega$ for $1 \leq p\leq \infty$ with

\begin{align*} \left \{ \begin{aligned} &\|y\|_{p}=\left (\int _{\Omega }\left |y\right |^{p}\right )^{\frac {1}{p}},\ 1 \leq p\lt \infty ,\\ &\|y\|_{\infty }=\text{ess}\sup \left |y(x)\right |,\ p=\infty . \end{aligned} \right . \end{align*}

Define

\begin{align*} z^{m}=\max _{x\in \bar {\Omega }}z(x),\ z_{m}=\min _{x\in \bar {\Omega }}z(x), \end{align*}

where $ z(x)\in C(\bar {\Omega },\mathbb{R})$ . Define $ \mathbb{B}_{i} : D(\mathbb{B}_{i})\to C(\bar {\Omega },\mathbb{R})$ as the linear operator with

\begin{align*} \mathbb{B}_{1}\phi :=d_{S}\Delta \phi (x),\ \ \mathbb{B}_{2}\phi :=d_{I}\Delta \phi (x), \end{align*}

where $ D(\mathbb{B}_{i}) :=\left \{\phi \in \cap _{y\ge 1}W^{2,y}(\Omega ):\ \frac {\partial \phi }{\partial \nu }=0\ \text{on}\ \partial \Omega \ \text{and}\ \mathbb{B}_{i}\phi \in C(\bar {\Omega },\mathbb{R}) \right \}$ . By [Reference Webb56], we know that the operator $ \mathbb{B}_{i}$ is the infinitesimal generator of the strongly continuous semigroup $\left \{e^{t\mathbb{B}_{i}}\right \}_{t\ge 0}, i=1,2$ . The operator $ \mathscr{B}\,:\, \mathit{H}\to \mathit{H}$ defined by

(3.1) \begin{equation} \mathscr{B}\varphi (x)\,:= \left ( \begin{array}{l} \mathbb{B}_{1}\varphi _{1}(x)\\[3pt] \mathbb{B}_{2}\varphi _{2}(x)\\[3pt] \mathbb{B}_{2}\varphi _{3}(x)\\[3pt] \quad \ 0\\[3pt] \quad \ 0\\[3pt] \quad \ 0 \end{array} \right ), \ \varphi =(\varphi _{1},\varphi _{2},\varphi _{3},\varphi _{4},\varphi _{5},\varphi _{6})\in D(\mathbb{B}_{1})\times D(\mathbb{B}_{2})\times D(\mathbb{B}_{2})\times [C(\bar {\Omega },\mathbb{R})]^{3}\subset \mathit{H} \end{equation}

is also the infinitesimal generator of the strongly continuous semigroup $\left \{e^{t\mathscr{B}}\right \}_{t\ge 0}$ in $\mathit{H}$ . Moreover, we define the nonlinear operator $ \mathscr{L}\,\,:\,\mathit{H}\to \mathit{H}$ as

(3.2) \begin{equation} \mathscr{L}\,(\varphi )(x)\,:= \left ( \begin{array}{c} \Lambda (x) - \alpha _{1}(x) f_{1}(\varphi _{4})\varphi _{1} -\alpha _{2}(x) f_{2}(\varphi _{5})\varphi _{1} -\mu (x)\varphi _{1}\\[5pt] \dfrac {\alpha _{1}(x) f_{1}(\varphi _{4})L\varphi _{1}}{L+\varphi _{6}} + \dfrac {\alpha _{2}(x) f_{2}{(\varphi _{5})}L\varphi _{1}}{L+\varphi _{6}} -\mu (x) \varphi _{2}\\[12pt] \dfrac {\alpha _{1}(x) f_{1}(\varphi _{4})\varphi _{6}\varphi _{1}}{L+\varphi _{6}} + \dfrac {\alpha _{2}(x) f_{2}(\varphi _{5})\varphi _{6}\varphi _{1}}{L+\varphi _{6}} -\mu (x) \varphi _{3}\\[10pt] h_{1}(x,\varphi _{4}) +\eta (x)(\varphi _{2}+\varphi _{3}) -b_{1}(x)\varphi _{4}\varphi _{6} -\delta _{1}(x)\varphi _{4}\\[5pt] h_{2}(x,\varphi _{5}) +\delta _{1}(x) \varphi _{4}-b_{2}(x)\varphi _{5}\varphi _{6}- \delta _{2}(x) \varphi _{5}\\[5pt] \alpha (x)\eta (x) \varphi _{3}+\chi _{1}(x) b_{1}(x)\varphi _{4}\varphi _{6} +\chi _{2}(x) b_{2}(x)\varphi _{5}\varphi _{6} -m(x)\varphi _{6}\\ \end{array} \right ). \end{equation}

Thus, model (2.1) can be expressed as

(3.3) \begin{equation} \dfrac {d}{dt}c(\cdot ,t\,;\,c^{0})=\mathscr{B}c(\cdot ,t\,;\,c^{0})+\mathscr{L}\,(c(\cdot ,t\,;\,c^{0})), \ \ c(\cdot ,0\,;\,c^{0})=c^{0}. \end{equation}

Theorem 3.1. For any $c^{0}(x)= (S^{0}(x),I_{1}^{0}(x),I_{2}^{0}(x),B_{1}^{0}(x),B_{2}^{0}(x),P^{0}(x))\in \mathit{H^{+}}$ , model (2.1)–(2.3) admits a unique global nonnegative classical solution defined on $\bar {\Omega }\times [0,\infty )$ . Moreover, model (2.1)–(2.3) has a connected global attractor in $\mathit{H}^{+}$ .

3.2 Dynamics of the disease-free steady state

A steady state of model (2.1)–(2.3) is a solution of the following model

(3.4) \begin{align} \left \{ \begin{aligned} &d_{S}\Delta S +\Lambda (x) - \alpha _{1}(x) \, f_{1}(B_{1})S -\alpha _{2}(x) \, f_{2}(B_{2})S -\mu (x) S=0,\ x\in \Omega ,\\[5pt] &d_{I}\Delta I_{1} + \alpha _{1}(x)\dfrac {L}{L+P} \, f_{1}(B_{1})S + \alpha _{2}(x)\dfrac {L}{L+P} \, f_{2}(B_{2})S -\mu (x) I_{1}=0,\ x\in \Omega ,\\[6pt] &d_{I}\Delta I_{2} + \alpha _{1}(x)\dfrac {P}{L+P} \, f_{1}(B_{1})S + \alpha _{2}(x)\dfrac {P}{L+P} \, f_{2}(B_{2})S -\mu (x) I_{2}=0, \ x\in \Omega ,\\[6pt] &h_{1}(x,B_{1}) +\eta (x) (I_{1}+I_{2}) -b_{1}(x)B_{1}P -\delta _{1}(x) B_{1}=0,\ x\in \Omega ,\\[5pt] &h_{2}(x,B_{2}) +\delta _{1}(x) B_{1}-b_{2}(x)B_{2}P- \delta _{2}(x) B_{2}=0,\ x\in \Omega ,\\[5pt] &\alpha (x)\eta (x) I_{2}+ \chi _{1}(x) b_{1}(x)B_{1}P +\chi _{2}(x) b_{2}(x)B_{2}P -m(x)P=0,\ x\in \Omega ,\\ \end{aligned} \right . \end{align}

for $ t\gt 0$ . Model (2.1)–(2.3) admits a unique disease-free steady state $F_{0}=(S^{*}(x),0,0,0,0,0)$ , where $ S^{*}(x)$ is the unique positive solution of

(3.5) \begin{align} \left \{ \begin{aligned} &d_{S}\Delta S +\Lambda (x)-\mu (x) S =0, \ x\in \Omega ,\\[3pt] &\dfrac {\partial S}{\partial \nu }=0, \ x\in \partial \Omega . \end{aligned} \right . \end{align}

Define

\begin{align*} \,\hat{\!h}_{i}(x)=\dfrac {\partial h_{i}(x,0)}{\partial B_{i}},\ i= 1,2, \ x\in \Omega . \end{align*}

Linearising model (2.1)–(2.3) at $F_{0}$ and adding the equations for $ I_{1}(x,t)$ and $I_{2}(x,t)$ , we obtain

(3.6) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial I}{\partial t}= d_{I}\Delta I + \dfrac {\alpha _{1}(x)S^{*}B_{1}}{H_{1}} + \dfrac {\alpha _{2}(x)S^{*}B_{2}}{H_{2}} -\mu (x) I,\ \ x\in \Omega ,\ \ t\gt 0, \\[3pt] &\dfrac {\partial B_{1}}{\partial t}=\,\hat{\!h}_{1}(x)B_{1} + \eta (x) I -\delta _{1}(x) B_{1},\ \ x\in \Omega ,\ \ t\gt 0, \\[3pt] &\dfrac {\partial B_{2}}{\partial t}=\,\hat{\!h}_{2}(x)B_{2} + \delta _{1}(x) B_{1}- \delta _{2}(x) B_{2},\ \ x\in \Omega ,\ \ t\gt 0,\\[3pt] &\dfrac {\partial I}{\partial \nu }=0,\ \ x\in \partial \Omega , \ \ t\gt 0. \end{aligned} \right . \end{align}

Let $J(t)$ be the solution semiflow associated with model (3.6), where $ J(t)\varphi =(I(\cdot ,t,\varphi ),B_{1}(\cdot ,t,\varphi ),B_{2}(\cdot ,t,\varphi ))$ for $\varphi \in C(\bar {\Omega },\mathbb{R}^{3})$ . Since model (3.6) is cooperative, $ J(t)$ is a $ C_{0}$ -semigroup with generator

\begin{align*} \mathscr{A}\!\,=\! \left ( \begin{array}{c@{\quad\!}c@{\quad\!}c} d_{I}\Delta -\mu & \dfrac {\alpha _{1}S^{*}}{H_{1}} & \dfrac {\alpha _{2}S^{*}}{H_{2}} \\[3pt] \eta & \,\hat{\!h}_{1}-\delta _{1} & 0 \\[3pt] 0 & \delta _{1} & \,\hat{\!h}_{2}-\delta _{2} \\ \end{array} \right )\!=\! \left ( \begin{array}{c@{\quad\!}c@{\quad\!}c} d_{I}\Delta -\mu & 0 & 0 \\[3pt] \eta & \,\hat{\!h}_{1}-\delta _{1} & 0 \\[3pt] 0 & \delta _{1} & \,\hat{\!h}_{2}-\delta _{2} \\ \end{array} \right )+ \left ( \begin{array}{c@{\quad\!\!}c@{\quad\!}c} 0 \ \ &\dfrac {\alpha _{1}S^{*}}{H_{1}} &\dfrac {\alpha _{2}S^{*}}{H_{2}} \\[3pt] 0 \ \ & 0 & 0 \\[3pt] 0 \ \ & 0 & 0 \\[3pt] \end{array} \right )\! =:\,V+F. \end{align*}

To ensure the well-definedness of $\mathscr{A}$ , we impose the following assumption on $ \,\hat{\!h}_{i}(x)$ for the remainder of this paper:

(3.7) \begin{equation} \,\hat{\!h}_{i}(x)\lt \delta _{i}(x),\ \forall x\in \Omega ,\ i=1,2. \end{equation}

Following [Reference Thieme42, Reference Wang and Zhao50], the basic reproduction number $\mathscr{R}_{0}$ of model (2.1)–(2.3) is defined as the spectral radius of $-FV^{-1}$ , which is denoted by $ r(\!-FV^{-1})$ , namely

\begin{equation*} \mathscr{R}_{0}:=r(\!-FV^{-1}). \end{equation*}

From the similar results in [Reference Thieme42, Reference Wang and Zhao50], we can assert the following statement.

Lemma 3.2. $\mathscr{R}_{0}-1$ has the same sign as $s(\mathscr{A} \, )$ , where $s(\mathscr{A} \, )$ denotes the spectrum bound of $\mathscr{A}$ with $s(\mathscr{A} \, )=\sup \{Re\lambda , \lambda \in \sigma (\mathscr{A} \, )\}$ .

To derive an equivalent formula for the basic reproduction number $\mathscr{R}_0$ , similar to the proof in [Reference Shu, Ma and Wang38, Reference Wang and Feng47], we introduce the following result involving the next generation operators $D_H$ and $D_L$ for HI vibrios and LI vibrios infections, respectively.

Lemma 3.3. Let

\begin{align*} F= \left ( \begin{array}{c@{\quad}c@{\quad}c} 0 & F_{11} & F_{12} \\ 0 & 0 & 0 \\ 0 & 0 & 0 \\ \end{array} \right ) \end{align*}

be a positive operator, and

\begin{align*} V= \left ( \begin{array}{c@{\quad}c@{\quad}c} d_{I}\Delta -V_{11} & 0 & 0 \\ V_{21} & -V_{22} & 0 \\ 0 & V_{32} & -V_{33} \\ \end{array} \right ) \end{align*}

be a resolvent-positive operator with $s(V)\lt 0$ , where $s(V)$ denotes the spectral bound of $V$ , then we obtain

\begin{align*} r(\!-FV^{-1}) = r(D_{H}+D_{L}), \end{align*}

where $D_{H}=F_{11}V_{22}^{-1}V_{21}(V_{11}-d_{I}\Delta )^{-1}$ and $D_{L}=F_{12}V_{33}^{-1}V_{32}V_{22}^{-1}V_{21}(V_{11}-d_{I}\Delta )^{-1}.$

In addition, based on Lemma 3.3, we can derive a specific form for $\mathscr{R}_{0}$ as follows:

\begin{align*} \mathscr{R}_{0}=r(D_{H}+D_{L}), \end{align*}

where

\begin{align*} D_{H}=\dfrac {\alpha _{1}(x)\Lambda (x)\eta (x)(\mu (x)-d_{I}\Delta )^{-1}}{\mu (x)H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))} \end{align*}

is the next generation operator for HI vibrios transmission to human hosts, and

\begin{align*} D_{L}=\dfrac {\alpha _{2}(x)\Lambda (x)\delta _{1}(x)\eta (x)(\mu (x)-d_{I}\Delta )^{-1}}{\mu (x)H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x))} \end{align*}

is the next generation operator for LI vibrios transmission to human hosts.

If considering exclusively the infection of human hosts by HI vibrios, the basic reproduction number $\mathscr{R}_{0}^{\,H}$ can be represented as

\begin{align*} \begin{aligned} \mathscr{R}_{0}^{\,H}&=r(D_{H})\\ &=r\left (\dfrac {\alpha _{1}(x)\Lambda (x)\eta (x)(\mu (x)-d_{I}\Delta )^{-1}}{\mu (x)H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))}\right )\\ &=\sup _{\varphi \in H^{1}(\Omega ),\ \varphi \not =0} \dfrac {\int _{\Omega }\alpha _{1}(x)\Lambda (x)\eta (x) \varphi ^{2}/\mu (x)H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))dx}{\int _{\Omega }d_{I}|\nabla \varphi |^{2}+\mu (x)\varphi ^{2}dx}. \end{aligned} \end{align*}

On the other hand, if considering exclusively the infection of human hosts by LI vibrios, the basic reproduction number $\mathscr{R}_{0}^{\,L}$ can be represented as

\begin{align*} \begin{aligned} \mathscr{R}_{0}^{\,L}&=r(D_{L})\\ &=r\left (\dfrac {\alpha _{2}(x)\Lambda (x)\delta _{1}(x)\eta (x)(\mu (x)-d_{I}\Delta )^{-1}}{\mu (x)H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x))}\right )\\ &=\sup _{\varphi \in H^{1}(\Omega ),\ \varphi \not =0} \dfrac {\int _{\Omega }\alpha _{2}(x)\Lambda (x)\delta _{1}(x)\eta (x) \varphi ^{2}/\mu (x)H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x))dx}{\int _{\Omega }d_{I}|\nabla \varphi |^{2}+\mu (x)\varphi ^{2}dx}. \end{aligned} \end{align*}

By [Reference Allen, Bolker and Lou2, Theorem 2] and the expressions of $\mathscr{R}_{0}^{\,H}$ and $\mathscr{R}_{0}^{\,L}$ , we immediately get Theorems3.4 and 3.5.

Theorem 3.4. The following statements are valid:

  1. (i) $\mathscr{R}_{0}^{\,H}$ is decreasing in $ d_{I}$ with

    \begin{align*} \lim \limits _{d_{I}\to 0}\mathscr{R}_{0}^{\,H}=\max \left \{ \dfrac {\alpha _{1}\Lambda \eta }{\mu ^{2}H_{1}(\delta _{1}-\,\hat{\!h}_{1})}\,:\,x\in \bar {\Omega }\right \} \end{align*}
    and
    \begin{align*} \lim \limits _{d_{I}\to \infty }\mathscr{R}_{0}^{\,H}=\dfrac {\int _{\Omega }\alpha _{1}\Lambda \eta /\mu H_{1}(\delta _{1}-\,\hat{\!h}_{1})dx}{\int _{\Omega } \mu dx}. \end{align*}
  2. (ii) If $\Omega$ is a favourable environment for HI vibrios in the sense that

    \begin{align*} \int _{\Omega }\dfrac {\alpha _{1}\Lambda \eta }{\mu H_{1}(\delta _{1}-\,\hat{\!h}_{1})}dx\gt \int _{\Omega }\mu dx, \end{align*}
    then $\mathscr{R}_{0}^{\,H}\gt 1$ for all $ d_{I}\gt 0$ .
  3. (iii) If $\Omega$ is a non-favourable environment for HI vibrios in the sense that

    \begin{align*} \int _{\Omega }\dfrac {\alpha _{1}\Lambda \eta }{\mu H_{1}(\delta _{1}-\,\hat{\!h}_{1})}dx\lt \int _{\Omega }\mu dx, \end{align*}
    Meanwhile, there is a favourable site $ x$ within the domain in the sense that $ \alpha _{1}(x)\Lambda (x)\eta (x) \gt \mu ^{2}(x)H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x)),$ then there exists $ \bar {d}_{I}$ such that $\mathscr{R}_{0}^{\,H}\gt 1$ when $d_{I}\lt \bar {d}_{I}$ , and $\mathscr{R}_{0}^{\,H}\lt 1$ when $d_{I}\gt \bar {d}_{I}$ .

Theorem 3.5. The following statements are valid:

  1. (i) $\mathscr{R}_{0}^{\,L}$ is decreasing in $ d_{I}$ with

    \begin{align*} \lim \limits _{d_{I}\to 0}\mathscr{R}_{0}^{\,L}=\max \left \{\dfrac {\alpha _{2}\Lambda \delta _{1}\eta }{\mu ^{2}H_{2}(\delta _{1}-\,\hat{\!h}_{1})(\delta _{2}-\,\hat{\!h}_{2})}\,:\,x\in \bar {\Omega }\right \} \end{align*}
    and
    \begin{align*} \lim \limits _{d_{I}\to \infty }\mathscr{R}_{0}^{\,L}=\dfrac {\int _{\Omega }\alpha _{2}\Lambda \delta _{1}\eta /\mu H_{2}(\delta _{1}-\,\hat{\!h}_{1})(\delta _{2}-\,\hat{\!h}_{2})dx}{\int _{\Omega } \mu dx}. \end{align*}
  2. (ii) If $\Omega$ is a favourable environment for LI vibrios in the sense that

    \begin{align*} \int _{\Omega }\dfrac {\alpha _{2}\Lambda \delta _{1}\eta }{\mu H_{2}(\delta _{1}-\,\hat{\!h}_{1})(\delta _{2}-\,\hat{\!h}_{2})}dx\gt \int _{\Omega }\mu dx, \end{align*}
    then $\mathscr{R}_{0}^{\,L}\gt 1$ for all $ d_{I}\gt 0$ .
  3. (iii) If $\Omega$ is a non-favourable environment for LI vibrios in the sense that

    \begin{align*} \int _{\Omega }\dfrac {\alpha _{2}\Lambda \delta _{1}\eta }{\mu H_{2}(\delta _{1}-\,\hat{\!h}_{1})(\delta _{2}-\,\hat{\!h}_{2})}dx\lt \int _{\Omega }\mu dx, \end{align*}
    Meanwhile, there is a favourable site $ x$ within the domain in the sense that $ \alpha _{2}(x)\Lambda (x)\delta _{1}(x)\eta (x) \gt \mu ^{2}(x)H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x)),$ then there exists $ \,\hat{\!d}_{I}$ such that $\mathscr{R}_{0}^{\,L}\gt 1$ when $d_{I}\lt \,\hat{\!d}_{I}$ , and $\mathscr{R}_{0}^{\,L}\lt1$ when $d_{I}\gt \,\hat{\!d}_{I}$ .

Remark 3.6. For model (2.4), there exists a disease-free steady state $ \tilde {F}_{0} = (S^{*},0,0,0,0,0)$ , where $ S^{*}=\frac {\Lambda }{\mu }$ . By a simple computation, the basic reproduction number of model (2.4) is

\begin{align*} \ \;\widetilde{\!\!\!\mathscr{R}}_{0}=\dfrac {\alpha _{1}\Lambda \eta }{\mu ^{2}H_{1}(\delta _{1}-\,\hat{\!h}_{1})}+\dfrac {\alpha _{2}\Lambda \eta \delta _{1}}{\mu ^{2}H_{2}(\delta _{1}-\,\hat{\!h}_{1})(\delta _{2}-\,\hat{\!h}_{2})}. \end{align*}

The extinction of the disease for model (2.1)–(2.3) in terms of $\mathscr{R}_{0}$ can be expressed as follows:

Theorem 3.7. The following two statements are valid:

  1. (i) If $\mathscr{R}_{0}\lt 1$ , the disease-free steady state $ F_0$ of model (2.1)–(2.3) is globally asymptotically stable;

  2. (ii) If $\mathscr{R}_{0}=1$ , the disease-free steady state $F_0$ of model (2.1)–(2.3) is globally asymptotically stable.

3.3 Dynamics of the phage-free steady state

In this subsection, we discuss the case that phages and phage-positive infected individuals are absent in model (2.1), and analyse the dynamics of the model under this case.

Theorem 3.8. If $\mathscr{R}_{0}\gt 1$ , model (2.1) has at least one phage-free steady state $ F_{1} =(S^{a}(x),I_{1}^{a}(x),0, B_{1}^{a}(x),B_{2}^{a}(x),0)$ .

Remark 3.9. Although we establish the existence of phage-free steady state for model (2.1), the problems of uniqueness and local/global stability are still unresolved. However, if the heterogeneous space degenerates to a homogeneous one, that is, model (2.1) degenerates to model (2.4), we can determine the uniqueness and stability of $ F_{1}$ .

For model (2.4), if $\ \;\widetilde{\!\!\!\mathscr{R}}_{0}\gt 1$ , there exists a phage-free steady state $ \tilde {F}_{1}=(\tilde {S}^{a},\tilde {I}_{1}^{a},0,\tilde {B}_{1}^{a},\tilde {B}_{2}^{a},0)$ , where

\begin{align*} \tilde {S}^{a}= \dfrac {\Lambda -\mu \tilde {I}_{1}^{a}}{\mu }\gt 0,\ \tilde {B}_{1}^{a}= \dfrac {\eta \tilde {I}_{1}^{a}}{(\delta _{1}-\,\hat{\!h}_{1})}\gt 0,\ \tilde {B}_{2}^{a}= \dfrac {\delta _{1}\eta \tilde {I}_{1}^{a}}{(\delta _{1}-\,\hat{\!h}_{1})(\delta _{2}-\,\hat{\!h}_{2})}\gt 0, \end{align*}

and $ \tilde {I}_{1}^{a}$ is the positive root of $ f(I_{1}) =\tilde {A}I_{1}^{2}+\tilde {B}I_{1}+\tilde {C}$ , where

\begin{align*} \tilde {A}= -\dfrac {\delta _{1}\eta ^{2}(\alpha _{1}+\alpha _{2}+\mu )}{(\delta _{1}-\,\hat{\!h}_{1})^{2}(\delta _{2}-\,\hat{\!h}_{2})},\ \tilde {B}= \left ( \dfrac {\Lambda \delta _{1}\eta ^{2}(\alpha _{1}+\alpha _{2})}{\mu (\delta _{1}-\,\hat{\!h}_{1})^{2}(\delta _{2}-\,\hat{\!h}_{2})}-\dfrac {\eta H_{2}(\alpha _{1}+\mu )}{(\delta _{1}-\,\hat{\!h}_{1})}-\dfrac {\delta _{1}\eta H_{1}(\alpha _{2}+\mu )}{(\delta _{1}-\,\hat{\!h}_{1})(\delta _{2}-\,\hat{\!h}_{2})}\right ), \end{align*}
\begin{align*} \tilde {C}=\left ( \dfrac {\Lambda \alpha _{1}\eta H_{2}}{\mu (\delta _{1}-\,\hat{\!h}_{1})}+\dfrac {\Lambda \alpha _{2}\delta _{1}\eta H_{1}}{\mu (\delta _{1}-\,\hat{\!h}_{1})(\delta _{2}-\,\hat{\!h}_{2})}-\mu H_{1}H_{2}\right ) = \mu H_{1}H_{2} (\, \ \;\widetilde{\!\!\!\mathscr{R}}_{0}-1). \end{align*}

If $ \, \ \;\widetilde{\!\!\!\mathscr{R}}_{0}\gt 1$ , we find that $f(0)=\tilde {C}\gt 0$ . Additionally, since $\tilde {A}\lt 0$ , it follows that $ f(I_{1})=0$ has two real roots: one positive and one negative. Hence, model (2.4) has a unique phage-free positive steady state $\tilde {F}_{1} =(\tilde {S}^{a}, \tilde {I}_{1}^{a},0,\tilde {B}_{1}^{a},\tilde {B}_{2}^{a},0)$ for $\, \ \;\widetilde{\!\!\!\mathscr{R}}_{0}\gt 1$ . Assume that $b_{2}=0$ , the following theorem presents a result regarding the global stability of the phage-free steady state $\tilde {F}_{1}$ .

Theorem 3.10. If $ \, \ \;\widetilde{\!\!\!\mathscr{R}}_{0}\gt 1$ , $ \tilde {B}_{1}^{a}\leq \tilde {B}_{1}^{b}$ and $l_{1}\leq \frac {\mu \chi _{1}}{\alpha \eta }$ hold, then the phage-free steady state $\tilde {F}_{1}$ of model (2.4) is globally asymptotically stable.

3.4 Dynamics of phage-present steady state

In this section, the existence and uniform persistence of phage-present steady state of model (2.1) are difficult to obtain due to the spatial heterogeneity and other mathematical difficulties. Therefore, we focus on proving the existence and uniform persistence of the phage-present steady state for its homogeneous case. Assuming $\alpha =b_{2}=0$ and $ I=I_{1}+I_{2}$ in this subsection. From model (2.4), we assume that there exists the phage-present positive steady state $\tilde {F}_{2}=(\tilde {S}^{b},\tilde {I}^{b},\tilde {B}_{1}^{b},\tilde {B}_{2}^{b},\tilde {P}^{b})$ . We have

(3.8) \begin{align} \left \{ \begin{aligned} &\Lambda - \alpha_{1} \, f_{1}(\tilde {B}_{1}^{b})\tilde {S}^{b} -\alpha_{2} \, f_{2}(\tilde {B}_{2}^{b})\tilde {S}^{b} -\mu \tilde {S}^{b}=0,\\[3pt] &\alpha _{1} \, f_{1}(\tilde {B}_{1}^{b})\tilde {S}^{b} +\alpha _{2} \, f_{2}(\tilde {B}_{2}^{b})\tilde {S}^{b} -\mu \tilde {I}^{b}=0,\\[3pt] &h_{1}(\tilde {B}_{1}^{b}) +\eta \tilde {I}^{b} -b_{1}\tilde {B}_{1}^{b}\tilde {P}^{b}-\delta _{1} \tilde {B}_{1}^{b}=0,\\[3pt] &h_{2}(\tilde {B}_{2}^{b}) +\delta _{1}\tilde {B}_{1}^{b} -\delta _{2} \tilde {B}_{2}^{b}=0,\\[3pt] &\chi _{1}b_{1}\tilde {B}_{1}^{b}\tilde {P}^{b}-m\tilde {P}^{b}=0. \end{aligned} \right . \end{align}

One gets

\begin{align*} \tilde {S}^{b}=\dfrac {\Lambda }{\dfrac {\alpha _{1}m}{m+H_{1}\chi _{1}b_{1}}+\dfrac {\alpha _{2}(h_{2}\chi _{1}b_{1}+\delta _{1}m)}{h_{2}\chi _{1}b_{1}+\delta _{1}m+H_{2}\delta _{2}\chi _{1}b_{1}}+\mu },\ \tilde {I}^{b}=\dfrac {\Lambda -\mu \tilde {S}^{b}}{\mu }\gt 0, \end{align*}
\begin{align*} \tilde {B}_{1}^{b}=\dfrac {m}{\chi _{1}b_{1}},\ \tilde {B}_{2}^{b}=\dfrac {(h_{2}\chi _{1}b_{1}+\delta _{1}m)}{\delta _{2}\chi _{1}b_{1}},\ \tilde {P}^{b}=\dfrac {\chi _{1}h_{1}b_{1}+\chi _{1}\eta b_{1}\tilde {I}^{b}-\delta _{1}m}{mb_{1}}. \end{align*}

We define the phage invasion reproduction number as

\begin{align*} \ \;\widetilde{\!\!\!\mathscr{R}}_{0}^{\, b}=\dfrac {\chi _{1}h_{1}b_{1}+\chi _{1}\eta b_{1}\tilde {I}^{b}}{\delta _{1}m}. \end{align*}

If $ \ \;\widetilde{\!\!\!\mathscr{R}}_{0}^{\, b}\gt 1$ , then $\tilde {P}^{b}\gt 0$ . Consequently, model (2.4) has a unique phage-present positive steady state $\tilde {F}_{2}=( \tilde {S}^{b},\tilde {I}^{b},\tilde {B}_{1}^{b},\tilde {B}_{2}^{b},\tilde {P}^{b})$ . The following theorem demonstrates the uniform persistence of the disease for model (2.4) when $\ \;\widetilde{\!\!\!\mathscr{R}}_{0}\gt 1$ .

Theorem 3.11. If $\ \;\widetilde{\!\!\!\mathscr{R}}_{0}\gt 1$ , and $\tilde {B}_{1}^{a}\gt \tilde {B}_{1}^{b}$ hold, there exists a $\tilde {\vartheta }\gt 0$ such that for the initial condition $ c^{0}(\!\cdot\!)=(S^{0},I_{1}^{0},I_{2}^{0},B_{1}^{0},B_{2}^{0},P^{0})(\!\cdot\!)\in \mathit{H^{+}}$ with $ I_{1}^{0}(x)\not \equiv 0$ or $ I_{2}^{0}(x)\not \equiv 0$ or $ B_{1}^{0}(x)\not \equiv 0$ or $ B_{2}^{0}(x)\not \equiv 0$ or $ P^{0}(x)\not \equiv 0$ , the solution $ \tilde {c}(x,t\,;\,c^{0})=(S(x,t),I_{1}(x,t),I_{2}(x,t),B_{1}(x,t),B_{2}(x,t),P(x,t))$ of model (2.4) satisfies $ \lim _{t\to \infty }\inf \tilde {c}(x,t\,;\,c^{0})\ge \tilde {\vartheta }$ uniformly for $x\in \bar {\Omega }.$

4 Proofs

4.1 Proof of Theorem3.1

In this section, we present a series of lemmas to prove Theorem3.1. The first lemma is just a consequence of applying the general results in [Reference Martin and Smith28].

Lemma 4.1. Let $\mathscr{B}$ and $\mathscr{L}\,$ be defined by (3.1)–(3.2). For any $c^{0}\in D(\mathscr{B})\subset \mathit{H^{+}}$ , there exists a $ T_{M} \gt 0$ satisfying that model (3.3) admits a unique nonnegative solution

\begin{align*} c(\cdot ,t\,;\,c^{0})=e^{t\mathscr{B}}c^{0}+\int _{0}^{t}e^{(t-s)\mathscr{B}} \mathscr{L}\,(c(\cdot ,s\,;\,c^{0}))ds,\ t\in [0,T_{M}), \end{align*}

where $T_{M}\leq +\infty$ , then we have $\lim \limits _{t\to T_{M}}\|c(\cdot ,t\,;\,c^{0})\|=\infty$ if $T_{M}=\infty$ .

Lemma 4.2. For any $c^{0}(x)= (S^{0}(x),I_{1}^{0}(x),I_{2}^{0}(x),B_{1}^{0}(x),B_{2}^{0}(x),P^{0}(x))\in \mathit{H^{+}}$ , model (2.1)–(2.3) admits a unique nonnegative global solution defined on $\bar {\Omega }\times [0,\infty )$ .

Proof. Let $c(x,t)=(S,I_{1},I_{2},B_{1},B_{2},P)$ be a solution associated with $c^{0}(x)=(S^{0}(x),I_{1}^{0}(x),I_{2}^{0}(x), B_{1}^{0}(x),B_{2}^{0}(x),P^{0}(x))$ . The first equation of model (2.1) implies that $\partial S/\partial t \leq d_{S}\Delta S+\Lambda (x)-\mu (x)S$ . By [Reference Lou and Zhao24, Lemma 1], we derive

(4.1) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial \bar {S}}{\partial t}= d_{S}\Delta \bar {S} +\Lambda (x)-\mu (x) \bar {S},\ \ x\in \Omega , \ \ t\gt 0, \\[6pt] &\dfrac {\partial \bar {S}}{\partial \nu }=0,\ \ x\in \partial \Omega , \ \ t\gt 0,\\ \end{aligned} \right . \end{align}

has a unique positive steady state $S^{*}(x)$ , which is globally asymptotically stable. According to the comparison principle, we have

(4.2) \begin{equation} \lim \limits _{t\to \infty }\sup S(x,t)\leq \lim \limits _{t\to \infty }\sup \bar {S}(x,t)=S^{*}(x). \end{equation}

Thus, there exists a $ Q_{1} \gt 0$ satisfying

(4.3) \begin{equation} \|S(x,t)\|\leq Q_{1},\ \ t\ge 0. \end{equation}

Let $ \{J_{2}(t)\}_{t\ge 0}$ be the semigroup generated by the operator $d_{I}\Delta -\mu (\!\cdot\!)$ , then from the second equation of model (2.1), we have

\begin{align*} I_{1}(x,t)=J_{2}(t)I_{1}^{0}(x)+\int _{0}^{t}J_{2}(t-s) \left (\dfrac {\alpha _{1}(x)LB_{1}(x,s)S(x,s)}{(L+P(x,s))(B_{1}(x,s)+H_{1})}+\dfrac {\alpha _{2}(x)LB_{2}(x,s)S(x,s)}{(L+P(x,s))(B_{2}(x,s)+H_{2})}\right )ds, \end{align*}

from (4.3), we derive

(4.4) \begin{align} \begin{aligned} \|I_{1}(x,t)\|&\leq e^{-\lambda _{1} t}\|I_{1}^{0}\|+Q_{1}\alpha _{1}^{m}\int _{0}^{t}e^{-\lambda _{1}(t-s)}\left (\dfrac {\|L\|\|B_{1}(x,s)\|}{(\|L\|+\|P(x,s)\|)(\|B_{1}(x,s)\|+\|H_{1}\|)}\right )ds\\ &\quad +Q_{1}\alpha _{2}^{m}\int _{0}^{t}e^{-\lambda _{1}(t-s)}\left (\dfrac {\|L\|\|B_{2}(x,s)\|}{(\|L\|+\|P(x,s)\|)(\|B_{2}(x,s)\|+\|H_{2}\|)}\right )ds\\ &\leq e^{-\lambda _{1} t}\|I_{1}^{0}\|+2Q_{1}\,\hat{\!\alpha }^{m}\int _{0}^{t}e^{-\lambda _{1}(t-s)}ds\\ &\leq \|I_{1}^{0}\|+2Q_{1}\,\hat{\!\alpha }^{m}\dfrac {1}{\lambda _{1}},\ \ t\ge 0, \end{aligned} \end{align}

where $\,\hat{\!\alpha }^{m}=\max _{x\in \bar {\Omega }}\{\alpha _{1}^{m}(x),\alpha _{2}^{m}(x)\}$ , and $\lambda _{1}\gt 0$ is the principal eigenvalue of $ -d_{I}\Delta +\mu (\!\cdot\!)$ . Similarly, from the third equation of model (2.1), we have

\begin{align*} I_{2}(x,t)=J_{2}(t)I_{2}^{0}(x)+\int _{0}^{t}J_{2}(t-s) \left (\dfrac {\alpha _{1}(x)P(x,s)B_{1}(x,s)S(x,s)}{(L+P(x,s))(B_{1}(x,s)+H_{1})}+\dfrac {\alpha _{2}(x)P(x,s)B_{2}(x,s)S(x,s)}{(L+P(x,s))(B_{2}(x,s)+H_{2})}\right )ds, \end{align*}

one gets

(4.5) \begin{align} \begin{aligned} \|I_{2}(x,t)\|&\leq e^{-\lambda _{1} t}\|I_{2}^{0}\|+Q_{1}\alpha _{1}^{m}\int _{0}^{t}e^{-\lambda _{1}(t-s)}\left (\dfrac {\|P(x,s)\|\|B_{1}(x,s)\|}{(\|L\|+\|P(x,s)\|)(\|B_{1}(x,s)\|+\|H_{1}\|)}\right )ds\\ &\quad +Q_{1}\alpha _{2}^{m}\int _{0}^{t}e^{-\lambda _{1}(t-s)}\left (\dfrac {\|P(x,s)\|\|B_{2}(x,s)\|}{(\|L\|+\|P(x,s)\|)(\|B_{2}(x,s)\|+\|H_{2}\|)}\right )ds\\ &\leq e^{-\lambda _{1} t}\|I_{2}^{0}\|+2Q_{1}\,\hat{\!\alpha }^{m}\int _{0}^{t}e^{-\lambda _{1}(t-s)}ds\\ &\leq \|I_{2}^{0}\|+2Q_{1}\,\hat{\!\alpha }^{m}\dfrac {1}{\lambda _{1}},\ \ t\ge 0. \end{aligned} \end{align}

By [Reference Shu, Ma and Wang37, proof of Theorem 2.3] and assumption ( $ \mathbf{H}_{\mathbf{3}}$ ), for $v\ge 0$ , there exist $C_{0}$ and $C_{1}$ satisfying $ h_{1}(x,v)-\delta _{1}v\leq C_{0}-C_{1}v$ , we derive

(4.6) \begin{equation} h_{1}(x,\,\hat{\!B}_{1}(x,t))-\delta _{1}\,\hat{\!B}_{1}(x,t)\leq C_{0}-C_{1}\,\hat{\!B}_{1}(x,t),\ x\in \Omega ,\ t\gt 0, \end{equation}

in which $ \,\hat{\!B}_{1}(x,t)$ satisfies

(4.7) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial \,\hat{\!B}_{1}(x,t)}{\partial t}=\eta (x)(I_{1}(x,t)+I_{2}(x,t))+C_{0}-C_{1}\,\hat{\!B}_{1}(x,t),\ x\in \Omega ,\ t\gt 0,\\[6pt] &\dfrac {\partial \,\hat{\!B}_{1}(x,t)}{\partial \nu }=0,\ x\in \partial \Omega ,\ t\gt 0,\\[6pt] &\,\hat{\!B}_{1}(x,0)=B_{1}(x,0)=B_{1}^{0}(x),\ x\in \Omega . \end{aligned} \right . \end{align}

For the first equation of model (4.7), applying the Gronwall’s inequality yields

\begin{align*} \,\hat{\!B}_{1}(x,t)=e^{-C_{1}t}B_{1}^{0}(x)+\int _{0}^{t}e^{-C_{1}(t-s)}\left (C_{0}+\eta (x)(I_{1}(x,s)+I_{2}(x,s))\right )ds, \end{align*}

based on (4.4)–(4.5), there exist $Q_{2}\gt 0$ and $Q_{3}\gt 0$ satisfying $\|I_{1}(x,t)\|\leq Q_{2}$ and $\|I_{2}(x,t)\|\leq Q_{3}$ , along with the comparison principle, we derive

(4.8) \begin{align} \begin{aligned} \|B_{1}(x,t)\|\leq \|\,\hat{\!B}_{1}(x,t)\| &\leq e^{-C_{1}t}\|B_{1}^{0}\|+\int _{0}^{t}e^{-C_{1}(t-s)}C_{0} ds+\int _{0}^{t}e^{-C_{1}(t-s)}\eta (x)\left (\|I_{1}(x,s)\|+\|I_{2}(x,s)\|\right )ds\\ &\leq \|B_{1}^{0}\|+\dfrac {C_{0}}{C_{1}}+\eta ^{m}(Q_{2}+Q_{3})\int _{0}^{t}e^{-C_{1}(t-s)}ds\\ &\leq \|B_{1}^{0}\|+\dfrac {1}{C_{1}}(C_{0}+\eta ^{m}(Q_{2}+Q_{3})),\ \ t\ge 0. \end{aligned} \end{align}

Similarly, there exist $ C_{2}\gt 0$ and $ C_{3}\gt 0$ satisfying $ h_{2}(x,v)-\delta _{2}v\leq C_{2}-C_{3}v$ , we derive

(4.9) \begin{equation} h_{2}(x,\,\hat{\!B}_{2}(x,t))-\delta _{2}\,\hat{\!B}_{2}(x,t)\leq C_{2}-C_{3}\,\hat{\!B}_{2}(x,t),\ x\in \Omega ,\ t\gt 0, \end{equation}

in which $ \,\hat{\!B}_{2}(x,t)$ satisfies

(4.10) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial \,\hat{\!B}_{2}(x,t)}{\partial t}=\delta _{1}(x)B_{1}(x,t)+C_{2}-C_{3}\,\hat{\!B}_{2}(x,t),\ x\in \Omega ,\ t\gt 0,\\[6pt] &\dfrac {\partial \,\hat{\!B}_{2}(x,t)}{\partial \nu }=0,\ x\in \partial \Omega ,\ t\gt 0,\\[6pt] &\,\hat{\!B}_{2}(x,0)=B_{2}(x,0)=B_{2}^{0}(x),\ x\in \Omega . \end{aligned} \right . \end{align}

For the first equation of model (4.10), applying the Gronwall’s inequality again yields

\begin{align*} \,\hat{\!B}_{2}(x,t)=e^{-C_{3}t}B_{2}^{0}(x)+\int _{0}^{t}e^{-C_{3}(t-s)}(C_{2}+\delta _{1}(x)B_{1}(x,s))ds, \end{align*}

based on (4.8), there is a $Q_{4}\gt 0$ satisfying $\|B_{1}(x,t)\|\leq Q_{4}$ , along with the comparison principle, we derive

(4.11) \begin{align} \begin{aligned} \|B_{2}(x,t)\|\leq \|\,\hat{\!B}_{2}(x,t)\| &\leq e^{-C_{3}t}\|B_{2}^{0}\|+\int _{0}^{t}e^{-C_{3}(t-s)}C_{2} ds+\int _{0}^{t}e^{-C_{3}(t-s)}\delta _{1}(x)\|B_{1}(x,s)\|ds\\ &\leq \|B_{2}^{0}\|+\dfrac {C_{2}}{C_{3}}+\delta _{1}^{m}Q_{4}\int _{0}^{t}e^{-C_{3}(t-s)}ds\\ &\leq \|B_{2}^{0}\|+\dfrac {1}{C_{3}}(C_{2}+\delta _{1}^{m}Q_{4}),\ \ t\ge 0. \end{aligned} \end{align}

Considering the last equation of model (2.1), we have

\begin{align*} P(x,t)= e^{\int _{0}^{t}(\chi _{1}(x)b_{1}(x)B_{1}(x,s)+\chi _{2}(x)b_{2}(x)B_{2}(x,s)-m(x))ds}P^{0}(x)+\int _{0}^{t}\alpha (x)\eta (x)I_{2}(x,s)ds,\ x\in \Omega . \end{align*}

Based on (4.11), there is a $Q_{5}\gt 0$ satisfying $\|B_{2}(x,t)\|\leq Q_{5}$ , then

(4.12) \begin{align} \begin{aligned} \|P(x,t)\|&\leq e^{\int _{0}^{t}\chi ^{m}b^{m}(\|B_{1}(x,s)\|+\|B_{2}(x,s)\|)ds}\|P^{0}\|+\alpha ^{m}\eta ^{m}\int _{0}^{t}\|I_{2}(x,s)\|ds \\ &\leq e^{\chi ^{m}b^{m}(Q_{4}+Q_{5})t}\|P^{0}\|+\alpha ^{m}\eta ^{m}Q_{3},\ \ t\ge 0, \end{aligned} \end{align}

where

\begin{align*} \chi ^{m}=\max _{x\in \bar {\Omega }}\left \{\chi _{1}(x),\chi _{2}(x)\right \},\ b^{m}=\max _{x\in \bar {\Omega }}\left \{b_{1}(x),b_{2}(x)\right \}. \end{align*}

By (4.3)–(4.5), (4.8), (4.11)–(4.12) and Lemma 4.1, the solution $c(x,t)=(S,I_{1},I_{2},B_{1},B_{2},P)$ exists globally.

Lemma 4.3. There exists a constant $ \,\hat{\!M}_{\infty }\gt 0$ independent of $c^{0}=(S^{0}(x), I_{1}^{0}(x),I_{2}^{0}(x), B_{1}^{0}(x), B_{2}^{0}(x), P^{0}(x)) \in \mathit{H^{+}}$ satisfying

\begin{align*} \lim \limits _{t\to \infty }\sup \left (S(x,t)+I_{1}(x,t)+I_{2}(x,t)+B_{1}(x,t)+B_{2}(x,t)+P(x,t)\right ) \leq \,\hat{\!M}_{\infty }. \end{align*}

Proof. From the first equation of model (2.1), we derive

\begin{align*} \dfrac {\partial S(x,t)}{\partial t}\leq d_{S}\Delta S +\Lambda ^{m}-\mu _{m}S(x,t). \end{align*}

By the comparison principle, there exists $ T_{1}\gt 0$ such that for all $ t\ge T_{1}$ and $x\in \bar {\Omega }$ ,

(4.13) \begin{equation} S(x,t)\leq M_{1},\ \text{where}\ M_{1}=\dfrac {\Lambda ^{m}}{\mu _{m}}+1. \end{equation}

The second equation $ I_{1}(x,t)$ of model (2.1) yields

\begin{align*} \dfrac {\partial I_{1}(x,t)}{\partial t}\leq d_{I}\Delta I_{1} +(\alpha _{1}^{m}+\alpha _{2}^{m})M_{1}-\mu _{m}I_{1}(x,t). \end{align*}

Applying the comparison principle, there exists $ T_{2}\gt T_{1}$ such that for all $ t\ge T_{2}$ and $x\in \bar {\Omega }$ ,

(4.14) \begin{equation} I_{1}(x,t)\leq M_{2},\ \text{where}\ M_{2}=\dfrac {(\alpha _{1}^{m}+\alpha _{2}^{m})M_{1}}{\mu _{m}}+1. \end{equation}

Similarly, one gets

(4.15) \begin{equation} I_{2}(x,t)\leq M_{2},\ \ x\in \bar {\Omega },\ \ t\ge T_{2}. \end{equation}

From the fourth equation of models (2.1) and (4.6), we derive

\begin{align*} \dfrac {\partial B_{1}(x,t)}{\partial t}\leq 2\eta ^{m}M_{2}+C_{0}-C_{1}B_{1}(x,t). \end{align*}

By the comparison principle, there exists $ T_{3}\gt T_{2}$ such that for all $ t\ge T_{3}$ and $x\in \bar {\Omega }$ ,

(4.16) \begin{equation} B_{1}(x,t)\leq M_{3}\ \text{where}\ M_{3}=\dfrac {2\eta ^{m}M_{2}+C_{0}}{C_{1}}+1. \end{equation}

Similarly, for equation $ B_{2}(x,t)$ of models (2.1) and (4.9), one gets

\begin{align*} \dfrac {\partial B_{2}(x,t)}{\partial t}\leq \delta _{1}^{m}M_{3}+C_{2}-C_{3}B_{2}(x,t). \end{align*}

The comparison principle yields $ T_{4}\gt T_{3}$ such that for all $ t\ge T_{4}$ and $x\in \bar {\Omega }$ ,

(4.17) \begin{equation} B_{2}(x,t)\leq M_{4},\text{where}\ M_{4}=\dfrac {\delta _{1}^{m}M_{3}+C_{2}}{C_{3}}+1. \end{equation}

Let $\bar {N}(x,t)=\chi _{1}(x)B_{1}(x,t)+\chi _{2}(x)B_{2}(x,t)+P(x,t)$ , we obtain

\begin{align*} \begin{aligned} \dfrac {\partial \bar {N}(x,t)}{\partial t} &\leq \chi _{1}^{m}h_{1}^{m}+\chi _{2}^{m}h_{2}^{m}+(\alpha ^{m}+2\chi _{1}^{m})\eta ^{m}M_{2}+\delta _{1}^{m}\chi _{2}^{m}M_{3}-\delta _{m}(\chi _{1}(x)B_{1}(x,t)+\chi _{2}(x)B_{2}(x,t)+P(x,t)),\\ \end{aligned} \end{align*}

where $\delta _{m}=\min _{x\in \bar {\Omega }}\{\delta _{1}(x),\delta _{2}(x), m(x)\}, h_{i}^{m}=\max _{x\in \bar {\Omega }}h_{i}(x,B_{i}),i=1,2$ . By the comparison principle, there exists a $ T_{5}\gt T_{4}$ such that for all $ t\ge T_{5}$ and $x\in \bar {\Omega }$ ,

(4.18) \begin{equation} P(x,t)\leq M_{5}, \text{where}\ M_{5}=\dfrac {\chi _{1}^{m}h_{1}^{m}+\chi _{2}^{m}h_{2}^{m}+(\alpha ^{m}+2\chi _{1}^{m})\eta ^{m}M_{2}+\delta _{1}^{m}\chi _{2}^{m}M_{3}}{\delta _{m}}+1. \end{equation}

Let $ \,\hat{\!M}_{\infty }=M_{2}+\sum _{i=1}^{5}M_{i}$ . Then by (4.13)–(4.18), for all $ t\ge T_{5}$ , the solution satisfies

\begin{align*} \lim \limits _{t\to \infty }\sup \left (S(x,t)+I_{1}(x,t)+I_{2}(x,t)+B_{1}(x,t)+B_{2}(x,t)+P(x,t)\right ) \leq \,\hat{\!M}_{\infty }. \end{align*}

This establishes Lemma 4.3.

Define $\,\,\hat{\!\!J}(t)\,:\,\mathit{H}^{+}\to \mathit{H}^{+}$ as the semiflow generated by model (2.1)–(2.3), namely $\,\,\hat{\!\!J}(t)c^{0}=c(x,t)$ for $t\gt 0$ . We observe that the last three equations in model (2.1)–(2.3) lack diffusion terms; the solution semiflow $\,\,\hat{\!\!J}(t)$ loses its compactness. We introduce the Kuratowski measure of noncompactness, denoted as $ \kappa (\!\cdot\!)$ ,

\begin{align*} \kappa =\inf \{R\,:\,\mathscr{P}\ \text{has a finite cover of diameter}\lt R\} . \end{align*}

Denote

\begin{align*} K(I_{1},I_{2},B_{1},B_{2},P)= \left ( \renewcommand \arraystretch {1.6} \begin{array}{c} h_{1}(x,B_{1})+\eta (x)(I_{1}+I_{2})-b_{1}(x)B_{1}P-\delta _{1}(x)B_{1}\\ h_{2}(x,B_{2})+\delta _{1}(x)B_{1}-b_{2}(x)B_{2}P-\delta _{2}(x)B_{2}\\ \alpha (x)\eta (x)I_{2}+\chi _{1}(x)b_{1}(x)B_{1}P+\chi _{2}(x)b_{2}(x)B_{2}P-m(x)P \end{array} \right ) \end{align*}

as the vector field associated with the last three equations of model (2.1). The Jacobian of $K$ with $(B_{1},B_{2},P)$ is defined as

\begin{align*} K_{13}=\dfrac {\partial K(I_{1},I_{2},B_{1},B_{2},P)}{\partial (B_{1},B_{2},P)}= \left ( \begin{array}{c@{\quad}c@{\quad}c@{\quad}c@{\quad}c} \dfrac {\partial h_{1}}{\partial B_{1}}-b_{1}P-\delta _{1} & 0 & -b_{1}B_{1}\\[6pt] \delta _{1} & \dfrac {\partial h_{2}}{\partial B_{2}}-b_{2}P-\delta _{2} & -b_{2}B_{2} \\[6pt] \chi _{1}b_{1}P & \chi _{2}b_{2}P & \chi _{1}b_{1}B_{1}+\chi _{2}b_{2}B_{2}-m\\ \end{array} \right ). \end{align*}

Lemma 4.4. $\,\,\hat{\!\!J}(t)$ is asymptotically smooth and $ \kappa$ -contracting if there exists a $ r^{*}\gt 0$ satisfying

(4.19) \begin{equation} u^{T}K_{13}u\leq -r^{*}u^{T}u,\ \forall u\in \mathbb{R}^{3},\ x\in \bar {\Omega },\ c\in \Gamma _{Z}. \end{equation}

Proof. By using similar proof in [Reference Sell and You36, Lemma 23.1(2)], we can derive that $\,\,\hat{\!\!J}(t)$ is asymptotically compact on any closed bounded set $\mathscr{P}$ for $\,\,\hat{\!\!J}(t)\mathscr{P} \subset \mathscr{P}$ . Thus, the omega limit $\omega (\mathscr{P})$ is nonempty, invariant and compact, and attracts $\mathscr{P}$ . This proves the asymptotic smoothness of $\,\,\hat{\!\!J}(t)$ . By [Reference Magal and Zhao26, Lemma 2.1(b)], we have

\begin{align*} \kappa (\,\hat{\!J}(t)\mathscr{P})\leq \kappa (\omega (\mathscr{P}))+\bar {d}(\,\hat{\!J}(t),\omega (\mathscr{P}))=\bar {d}(\,\hat{\!J}(t)\mathscr{P},\omega (\mathscr{P})), \end{align*}

where $ \bar {d}(\,\hat{\!J}(t)\mathscr{P},\omega (\mathscr{P}))$ denotes the distance from $\,\,\hat{\!\!J}(t)\mathscr{P}$ to $\omega (\mathscr{P})$ , which tends to zero as $t\to + \infty$ . Consequently, $\,\,\hat{\!\!J}(t)$ is $\kappa$ -contracting.

Remark 4.5. A sufficient condition for (4.19) is that

\begin{align*} \dfrac {\partial h_{1}}{\partial B_{1}}\lt 2b_{1}\,\hat{\!M}_{1},\ \dfrac {\partial h_{2}}{\partial B_{2}}+\delta _{1}\lt 2b_{1}\,\hat{\!M}_{1}+\delta _{2},\ 2\chi ^{m}b^{m}\,\hat{\!M}_{1}\lt m. \end{align*}

Proof of Theorem 3.1 The global existence and uniqueness of solution of model (2.1)–(2.3) can be obtained from Lemmas 4.1–4.2. In view of Lemma 4.3, $\,\,\hat{\!\!J}(t)$ is point dissipative. According to Lemma 4.4, $\,\,\hat{\!\!J}(t)$ is asymptotically smooth. Thus, by [Reference Hale and Waltman13, Theorem 2.1], model (2.1)–(2.3) has a connected global attractor in $\mathit{H}^{+}$ .

4.2 Proof of Lemma 3.3

Let $\psi =Fw$ and $w=-V^{-1}\varphi$ . Then,

\begin{align*} \varphi _{1}=(V_{11}-d_{I}\Delta )w_{1},\ \ \varphi _{2}=V_{22}w_{2}-V_{21}w_{1},\ \ \varphi _{3}=V_{33}w_{3}-V_{32}w_{2}, \end{align*}

we can easily derive

\begin{equation*} w_{1}=(V_{11}-d_{I}\Delta )^{-1}\varphi _{1},\\[-12pt] \end{equation*}
\begin{equation*} w_{2}=V_{22}^{-1}\left (\varphi _{2}+V_{21}(V_{11}-d_{I}\Delta )^{-1}\varphi _{1}\right ),\\[-12pt] \end{equation*}
\begin{equation*} w_{3}=V_{33}^{-1}\left (\varphi _{3}+V_{32}V_{22}^{-1}\left (\varphi _{2}+V_{21}(V_{11}-d_{I}\Delta )^{-1}\varphi _{1}\right )\right ). \end{equation*}

By a straightforward calculation, we get

\begin{align*} \begin{aligned} \psi _{1}&=F_{12}V_{33}^{-1}\left (\varphi _{3}+V_{32}V_{22}^{-1}\left (\varphi _{2}+V_{21}(V_{11}-d_{I}\Delta )^{-1}\varphi _{1}\right )\right ) +F_{11}V_{22}^{-1}\left (\varphi _{2}+V_{21}(V_{11}-d_{I}\Delta )^{-1}\varphi _{1}\right ),\\ &\,\,\,\,\quad \qquad \qquad \qquad \qquad \qquad \qquad \qquad \ \ \psi _{2}=0, \ \ \psi _{3}=0, \end{aligned} \end{align*}

and

\begin{align*} -FV^{-1} \left ( \renewcommand \arraystretch {1.3} \begin{array}{c} \varphi _{1}\\ \varphi _{2}\\ \varphi _{3} \end{array} \right ) = \left ( \renewcommand \arraystretch {1.3} \begin{array}{c} D_{1}\varphi _{1}+D_{2}\varphi _{2}+D_{3}\varphi _{3}\\ 0\\ 0 \end{array} \right ), \end{align*}

where

\begin{align*} \begin{aligned} &D_{1}=F_{11}V_{22}^{-1}V_{21}(V_{11}-d_{I}\Delta )^{-1}+F_{12}V_{33}^{-1}V_{32}V_{22}^{-1}V_{21}(V_{11}-d_{I}\Delta )^{-1},\\[6pt] &D_{2}=F_{11}V_{22}^{-1}+F_{12}V_{33}^{-1}V_{32}V_{22}^{-1},\\[6pt] &D_{3}=F_{12}V_{33}^{-1}. \end{aligned} \end{align*}

It follows by iteration that

\begin{align*} (\!-FV^{-1})^{n} \left ( \renewcommand \arraystretch {1.3} \begin{array}{c} \varphi _{1}\\ \varphi _{2}\\ \varphi _{3} \end{array} \right ) = \left ( \renewcommand \arraystretch {1.3} \begin{array}{c} D_{1}^{n}\varphi _{1}+D_{1}^{n-1}D_{2}\varphi _{2}+D_{1}^{n-1}D_{3}\varphi _{3}\\ 0\\ 0 \end{array} \right ). \end{align*}

Then, we have

\begin{align*} \|D_{1}^{n}\|\leq \|(\!-FV^{-1})^{n}\|\leq \|D_{1}^{n-1}\|(\|D_{1}\|+\|D_{2}\|+\|D_{3}\|). \end{align*}

By applying the Gelfand’s formula and the squeeze theorem, we obtain $ r(\!-FV^{-1}) = r(D_{H}+D_{L})$ . This establishes Lemma 3.3.

4.3 Proof of Theorem3.7

Before proving Theorem3.7, we first give some preliminaries.

Lemma 4.6. Define $\lambda ^{0}$ as the principal eigenvalue of the eigenvalue problem

(4.20) \begin{align} \left \{ \begin{aligned} &d_{I}\Delta \varphi -\mu (x)\varphi +\left (\dfrac {\alpha _{1}(x)S^{*}\eta (x)}{H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))}+\dfrac {\alpha _{2}(x)S^{*}\eta (x)\delta _{1}(x)}{H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x))}\right ) \varphi =\lambda \varphi ,\ x\in \Omega ,\\ &\dfrac {\partial \varphi }{\partial \nu }=0,\ x\in \partial \Omega .\\ \end{aligned} \right . \end{align}

Then $\mathscr{R}_{0}-1$ and $ s(\mathscr{A} \, )$ have the same sign as $\lambda ^{0}$ .

Proof. It is well-known that there exists a least eigenvalue $\lambda ^{0}$ associated with the eigenvalue problem (4.20), its corresponding eigenfunction $\bar {\varphi }$ can be chosen to be positive on $\Omega$ , that is,

(4.21) \begin{align} \left \{ \begin{aligned} &d_{I}\Delta \bar {\varphi }-\mu (x)\bar {\varphi }+\left (\dfrac {\alpha _{1}(x)S^{*}\eta (x)}{H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))}+\dfrac {\alpha _{2}(x)S^{*}\eta (x)\delta _{1}(x)}{H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x))}\right ) \bar {\varphi }=\lambda ^{0} \bar {\varphi },\ x\in \Omega ,\\ &\dfrac {\partial \bar {\varphi }}{\partial \nu }=0,\ x\in \partial \Omega .\\ \end{aligned} \right . \end{align}

Next, we consider the following eigenvalue problem

(4.22) \begin{align} \left \{ \begin{aligned} &d_{I}\Delta \varphi -\mu (x)\varphi +\left (\dfrac {\alpha _{1}(x)S^{*}\eta (x)}{H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))}+\dfrac {\alpha _{2}(x)S^{*}\eta (x)\delta _{1}(x)}{H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x))}\right )\dfrac {1}{\mathscr{R}_{0}} \varphi =0,\ x\in \Omega ,\\ &\dfrac {\partial \varphi }{\partial \nu }=0,\ x\in \partial \Omega .\\ \end{aligned} \right . \end{align}

By multiplying (4.21) with $\varphi$ and (4.22) with $\bar {\varphi }$ , and performing integration by parts on $\Omega$ , subtracting the resulting equation, we get

\begin{align*} \left (1-\dfrac {1}{\mathscr{R}_{0}}\right )\int _{\Omega }\left (\dfrac {\alpha _{1}(x)S^{*}\eta (x)}{H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))}+\dfrac {\alpha _{2}(x)S^{*}\eta (x)\delta _{1}(x)}{H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x))}\right )\bar {\varphi }\varphi dx=\lambda ^{0} \int _{\Omega }\bar {\varphi }\varphi dx. \end{align*}

Apparently,

\begin{align*} \int _{\Omega }\left (\frac {\alpha _{1}(x)S^{*}\eta (x)}{H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))}+\frac {\alpha _{2}(x)S^{*}\eta (x)\delta _{1}(x)}{H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x))}\right )\bar {\varphi }\varphi dx \end{align*}

and $\int _{\Omega }\bar {\varphi }\varphi dx$ are both positive, it implies that $\left (1-\frac {1}{\mathscr{R}_{0}}\right )$ and $\lambda ^{0}$ have the same sign, namely $\mathscr{R}_{0}$ and $\lambda ^{0}$ have the same sign. This establishes Lemma 4.6.

Let $I(x,t)=e^{\lambda t}\phi _{2},\ B_{1}(x,t)=e^{\lambda t}\phi _{3},\ B_{2}(x,t)=e^{\lambda t}\phi _{4}$ in model (3.6), then

(4.23) \begin{align} \left \{ \begin{aligned} &\lambda \phi _{2}= d_{I}\Delta \phi _{2} + \dfrac {\alpha _{1}(x)S^{*}}{H_{1}}\phi _{3} + \dfrac {\alpha _{2}(x)S^{*}}{H_{2}}\phi _{4} -\mu (x)\phi _{2},\ \ x\in \Omega ,\\ &\lambda \phi _{3}=\eta (x) \phi _{2} +\,\hat{\!h}_{1}(x)\phi _{3} -\delta _{1}(x) \phi _{3},\ \ x\in \Omega ,\\ &\lambda \phi _{4}=\delta _{1}(x) \phi _{3}+\,\hat{\!h}_{2}(x)\phi _{4}- \delta _{2}(x) \phi _{4},\ \ x\in \Omega ,\\ &\dfrac {\partial \phi _{2}}{\partial \nu }=0,\ \ x\in \partial \Omega . \end{aligned} \right . \end{align}

Lemma 4.7. If $\mathscr{R}_{0}\ge 1$ , then $ s(\mathscr{A} \, )$ is the principal eigenvalue of problem (4.23) with respect to a strongly positive eigenfunction.

Proof. Let

\begin{align*} T_{\lambda }=d_{I}\Delta +\dfrac {\alpha _{1}(x)S^{*}\eta (x)}{\lambda +H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))}+\dfrac {\alpha _{2}(x)S^{*}\eta (x)\delta _{1}(x)}{\lambda + H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x))}-\mu (x) \end{align*}

be a family of linear operators on $ C(\bar {\Omega })$ . Notice that $ s(T_{\lambda })$ is decreasing and continuously dependent on $\lambda$ , where $\lambda$ denotes the principal eigenvalue of $T_{\lambda }c=\lambda c$ . As a result, it has the following variational characterisation:

\begin{align*} s(T_{\lambda })=\sup _{\varphi \in H^{1}(\Omega ),\ \varphi \not =0}\left \{\dfrac {\int _{\Omega } \left (\dfrac {\alpha _{1}S^{*}\eta }{\lambda +H_{1}(\delta _{1}-\,\hat{\!h}_{1})}+\dfrac {\alpha _{2}S^{*}\eta \delta _{1}}{\lambda +H_{2}(\delta _{1}-\,\hat{\!h}_{1})(\delta _{2}-\,\hat{\!h}_{2})}\right )\varphi ^{2}-d_{I}|\nabla \varphi |^{2}-\mu \varphi ^{2}dx}{\int _{\Omega }\varphi ^{2}dx}\right \}. \end{align*}

Clearly, we have $ s(T_{\lambda })\lt 0$ if $\lambda$ is large enough. From $\mathscr{R}_{0}\ge 1$ and Lemma 4.6, the following equation holds $s(T_{0})=\lambda ^{0}\ge 0.$ Therefore, we duduce that there exists a unique $\bar {\lambda }$ satisfying $ s(T_{\bar {\lambda }})= \bar {\lambda }$ . Let $\tilde {\phi }\gt 0$ be an eigenvector with respect to $ s(T_{\bar {\lambda }})$ , we get $ T_{\bar {\lambda }}\tilde {\phi }=\bar {\lambda }\tilde {\phi }.$ Similar to the results in [Reference Wang and Zhao50, Theorem 2.3], we have $\bar {\lambda }=s(\mathscr{A} \, )$ . This establishes Lemma 4.7.

Proof of Theorem 3.7 (i) The locally asymptotically stability of $ F_{0}$ for model (2.1)–(2.3) follows from [Reference Wang and Zhao50, Theorem3.1]. Next, we only need to establish the global attractivity of $ F_{0}$ . Initially, we set $\varepsilon _{0}\gt 0$ . From (4.2), we deduce that there exists a $t_{1}\gt 0$ such that $ 0\leq S(\cdot ,t) \leq S^{*}(x)+\varepsilon _{0}$ for all $ t\gt t_{1}$ , and from the comparison principle for cooperative models, we get $(I(x,t),B_{1}(x,t),B_{2}(x,t))\leq (\bar {I}(x,t),\bar {B}_{1}(x,t),\bar {B}_{2}(x,t))$ , where $(\bar {I}(x,t),\bar {B}_{1}(x,t),\bar {B}_{2}(x,t))$ is the solution of the following model

(4.24) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial \bar {I}}{\partial t}= d_{I}\Delta \bar {I} + \dfrac {\alpha _{1}(x)(S^{*}+\varepsilon _{0})}{H_{1}}\bar {B}_{1} + \dfrac {\alpha _{2}(x)(S^{*}+\varepsilon _{0})}{H_{2}}\bar {B}_{2} -\mu (x) \bar {I},\ \ x\in \Omega ,\ \ t\gt t_{1}, \\[6pt] &\dfrac {\partial \bar {B}_{1}}{\partial t}=\eta (x) \bar {I} +\,\hat{\!h}_{1}(x)\bar {B}_{1} -\delta _{1}(x) \bar {B}_{1},\ \ x\in \Omega ,\ \ t\gt t_{1}, \\[6pt] &\dfrac {\partial \bar {B}_{2}}{\partial t}=\delta _{1}(x) \bar {B}_{1}+\,\hat{\!h}_{2}(x)\bar {B}_{2}- \delta _{2}(x) \bar {B}_{2},\ \ x\in \Omega ,\ \ t\gt t_{1},\\[6pt] &\dfrac {\partial \bar {I}}{\partial \nu }=0,\ \ x\in \partial \Omega ,\ \ t\gt t_{1}, \\[6pt] &\bar {I}(x,t_{1})=I(x,t_{1}),\ \bar {B}_{1}(x,t_{1})=B_{1}(x,t_{1}),\ \bar {B}_{2}(x,t_{1})=B_{2}(x,t_{1}), \ \ x\in \Omega .\\ \end{aligned} \right . \end{align}

Let $J_{\varepsilon _{0}}(t)$ be the linear semigroup of model (4.24) with respect to $C_{\varepsilon _{0}}$ , where $C_{\varepsilon _{0}}$ is defined as the following generator

\begin{align*} C_{\varepsilon _{0}}= \left ( \begin{array}{c@{\quad}c@{\quad}c} d_{I}\Delta -\mu & \dfrac {\alpha _{1}(S^{*}+\varepsilon _{0})}{H_{1}} & \dfrac {\alpha _{2}(S^{*}+\varepsilon _{0})}{H_{2}} \\[8pt] \eta & \,\hat{\!h}_{1}-\delta _{1} & 0 \\[2pt] 0 & \delta _{1} & \,\hat{\!h}_{2}-\delta _{2} \\ \end{array} \right ). \end{align*}

Next, we prove that $ \|J_{\varepsilon _{0}}(t)\|\leq Ne^{\gamma _{\varepsilon _{0}}t} \text{ for some } N\gt 0,$ where $\gamma _{\varepsilon _{0}}:=\lim _{t\to \infty }\frac {\ln \|J_{\varepsilon _{0}}(t)\|}{t}$ is the exponential growth bound of $J_{\varepsilon _{0}}(t)$ . Note that

\begin{align*} \gamma _{\varepsilon _{0}}=\max \{s(C_{\varepsilon _{0}}),\gamma _{ess}(J_{\varepsilon _{0}}(t))\}, \end{align*}

where $ \gamma _{ess}(J_{\varepsilon _{0}}(t))$ := $\lim _{t\to \infty }\frac {n(J_{\varepsilon _{0}}(t))}{t}$ , and $ n(\!\cdot\!)$ represents the measure of non-compactness. Similar to the arguments in [Reference Wu and Zou58, Lemma 3.5], there exists a $\delta _{*}\gt 0$ such that $\gamma _{ess}(J_{\varepsilon _{0}})\leq -\delta _{*}$ . Then we can choose $\varepsilon _{0}$ small enough such that $\gamma _{\varepsilon _{0}}\lt 0$ . To see this, $\gamma _{\varepsilon _{0}}$ has the same sign as $ s(C_{\varepsilon _{0}})$ . Moreover, $ s(C_{\varepsilon _{0}})$ has the same sign as $\lambda _{\varepsilon _{0}}$ , in which $\lambda _{\varepsilon _{0}}$ represents the principal eigenvalue of the eigenvalue problem

(4.25) \begin{align} \left \{ \begin{aligned} &d_{I}\Delta \varphi -\mu (x)\varphi +\left (\dfrac {\alpha _{1}(x)(S^{*}+\varepsilon _{0})\eta (x)}{H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))}+\dfrac {\alpha _{2}(x)(S^{*}+\varepsilon _{0})\eta (x)\delta _{1}(x)}{H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x))}\right )\varphi =\lambda \varphi ,\ x\in \Omega ,\\ &\dfrac {\partial \varphi }{\partial \nu }=0,\ x\in \partial \Omega .\\ \end{aligned} \right . \end{align}

According to Lemma 4.6, $ \mathscr{R}_{0}\lt 1$ and the continuous dependence of $\lambda _{\varepsilon _{0}}^{0}$ on $\varepsilon _{0}$ , it can be inferred that $ \lambda _{\varepsilon _{0}}^{0} \lt 0$ if $\varepsilon _{0}$ is small enough. Then we get $\gamma _{\varepsilon _{0}}\lt 0$ . It implies that $(\bar {I}(x,t),\bar {B}_{1}(x,t),\bar {B}_{2}(x,t))\to (0,0,0)$ as $ t\to \infty$ uniformly for $x\in \bar {\Omega }$ . Clearly, we derive $(I(x,t),B_{1}(x,t),B_{2}(x,t))\to (0,0,0)$ as $ t\to \infty$ uniformly for $x\in \bar {\Omega }$ . Then, we have $(I_{1}(x,t),I_{2}(x,t),B_{1}(x,t),B_{2}(x,t))\to (0,0,0,0)$ as $ t\to \infty$ uniformly for $x\in \bar {\Omega }$ . Furthermore, it follows from model (4.2) that $S(x,t)\to S^{*}(x)$ as $ t\to \infty$ uniformly for $x\in \bar {\Omega }$ . This establishes Theorem3.7 (i).

Proof of Theorem 3.7 (ii) First, we prove the local stability of $F_{0}$ . For any given $\varepsilon \gt 0$ , we assume that $\vartheta \gt 0$ and initial conditions $c^{0}=(S^{0},I_{1}^{0},I_{2}^{0},B_{1}^{0},B_{2}^{0},P^{0})$ with $ \|c^{0}-F_{0}\|\leq \vartheta$ . Let

(4.26) \begin{equation} v_{1}(x,t)=\dfrac {S(x,t)}{S^{*}(x)}-1,\ \ a(t)=\max _{x\in \bar {\Omega }}\{v_{1}(x,t),0\}. \end{equation}

From model (3.5), we have $ d_{S}\Delta S^{*} +\Lambda -\mu S^{*} =0,$ and by the first equation of model (2.1), we derive

\begin{align*} \dfrac {\partial v_{1}}{\partial t}-d_{S}\Delta v_{1}-2d_{S}\dfrac {\nabla S^{*}\cdot \nabla v_{1}}{S^{*}}+\dfrac {\Lambda }{S^{*}}v_{1}=-\dfrac {\alpha _{1}B_{1}S}{S^{*}(B_{1}+H_{1})}-\dfrac {\alpha _{2}B_{2}S}{S^{*}(B_{2}+H_{2})}. \end{align*}

Define $\,\,\hat{\!\!J}_{1}(t)$ as the positive semigroup generated by the following operator

\begin{align*} d_{S}\Delta +2d_{S}\dfrac {\nabla S^{*}\cdot \nabla }{S^{*}}-\dfrac {\Lambda }{S^{*}}. \end{align*}

By [Reference Webb57, Proposition 2.3], there exists a $z_{1}\gt 0$ satisfying $\|\,\hat{\!J}_{1}(t)\|\leq \bar {M}e^{-z_{1}t}$ for some $\bar {M}\gt 0$ . We derive

\begin{align*} v_{1}(\cdot ,t)=\,\hat{\!J}_{1}(t)v_{10}-\int _{0}^{t}\,\hat{\!J}_{1}(t-s)\left (\dfrac {\alpha _{1}B_{1}(\cdot ,s)S(\cdot ,s)}{S^{*}(B_{1}(\cdot ,s)+H_{1})}+\dfrac {\alpha _{2}B_{2}(\cdot ,s)S(\cdot ,s)}{S^{*}(B_{2}(\cdot ,s)+H_{2})}\right )ds, \end{align*}

where $v_{10}=\dfrac {S^{0}}{S^{*}}-1$ . In view of the positivity of $\,\,\hat{\!\!J}_{1}(t)$ , one gets

(4.27) \begin{align}a(t)=\max _{x\in \bar {\Omega }}\left \{v_{1}(x,t),0\right \}&=\max _{x\in \bar {\Omega }}\left \{\,\hat{\!J}_{1}(t)v_{10}-\int _{0}^{t}\,\hat{\!J}_{1}(t-s)\left (\dfrac {\alpha _{1}B_{1}(\cdot ,s)S(\cdot ,s)}{S^{*}(B_{1}(\cdot ,s)+H_{1})}+\dfrac {\alpha _{2}B_{2}(\cdot ,s)S(\cdot ,s)}{S^{*}(B_{2}(\cdot ,s)+H_{2})}\right )ds,0\right \}\nonumber\\ &\leq \max _{x\in \bar {\Omega }}\{\,\hat{\!J}_{1}(t)v_{10},0\} \leq \|\,\hat{\!J}_{1}(t)v_{10}\| \nonumber\\ &\leq \bar {M}e^{-z_{1}t}\left \| {\frac {S^{0}}{S^{*}}-1} \right \|\nonumber\\ &\leq \dfrac {\vartheta \bar {M}e^{-z_{1}t}}{S_{m}^{*}}, \end{align}

where $ S_{m}^{*}=\min _{x\in \bar {\Omega }}S^{*}(x).$ From $\mathscr{R}_{0}=1$ and [Reference Wu and Zou58, Lemma 3.6], we obtain $\|J(t)\|\leq \bar {M}$ for $t\ge 0$ and $\bar {M}\gt 0$ . Recall that $ a(t)\leq \vartheta \bar {M}e^{-z_{1}t}/S^{*}_{m}$ , we derive

\begin{align*} \left ( \renewcommand \arraystretch {1.6} \begin{array}{c@{\quad}c@{\quad}c@{\quad}c} I(\cdot ,t)\\ B_{1}(\cdot ,t)\\ B_{2}(\cdot ,t) \end{array} \right ) &\leq J(t) \left ( \renewcommand \arraystretch {1.6} \begin{array}{c@{\quad}c@{\quad}c@{\quad}c} I^{0}\\ B_{1}^{0}\\ B_{2}^{0} \end{array} \right )+\int _{0}^{t}J(t-s) \left ( \renewcommand \arraystretch {1.6} \begin{array}{c} \left (\dfrac {\alpha _{1}B_{1}(\cdot ,s)S^{*}}{H_{1}}+\dfrac {\alpha _{2}B_{2}(\cdot ,s)S^{*}}{H_{2}}\right )\left (\dfrac {S(\cdot ,s)}{S^{*}}-1\right )\\ 0\\ 0 \end{array} \right )ds\\ &\leq \bar {M}\vartheta +\int _{0}^{t}J(t-s)\,\hat{\!\alpha }^{m}S^{*}\left (B_{1}(\cdot ,s)+B_{2}(\cdot ,s)\right )\left (\dfrac {S(\cdot ,s)}{S^{*}}-1\right )ds, \end{align*}

where $I^{0}=I_{1}^{0}+I_{2}^{0}$ . Thus,

(4.28) \begin{align} \begin{aligned} \max \{\|I(x,t)\|,\|B_{1}(x,t)\|,\|B_{2}(x,t)\|\} &\leq \bar {M}\vartheta + \bar {M}\|\,\hat{\!\alpha }^{m}\|\|S^{*}\|\int _{0}^{t}a(s)(\|B_{1}(\cdot ,s)\|+\|B_{2}(\cdot ,s)\|)ds\\ &\leq \bar {M}\vartheta + \tilde {M}_{1}\vartheta \int _{0}^{t}e^{-z_{1}s}(\|B_{1}(\cdot ,s)\|+\|B_{2}(\cdot ,s)\|)ds, \end{aligned} \end{align}

where $\tilde {M}_{1}=\bar {M}^{2}\|\,\hat{\!\alpha }^{m}\|\|S^{*}\|/S^{*}_{m}$ . This yields that

\begin{align*} \|B_{1}(\cdot ,t)\|+\|B_{2}(\cdot ,t)\|\leq 2\bar {M}\vartheta + 2\tilde {M}_{1}\vartheta \int _{0}^{t}e^{-z_{1}s}(\|B_{1}(\cdot ,s)\|+\|B_{2}(\cdot ,s)\|)ds. \end{align*}

Utilising Gronwall’s inequality, we derive

(4.29) \begin{align} \begin{aligned} \|B_{1}(\cdot ,t)\|+\|B_{2}(\cdot ,t)\|&\leq 2\bar {M}\vartheta e^{\int _{0}^{t}2\tilde {M}_{1}\vartheta e^{-z_{1}s}ds} \\ &\leq 2\bar {M}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}}. \end{aligned} \end{align}

Moreover, we can also derive

(4.30) \begin{align} \begin{aligned} \|I(\cdot ,t)\|&\leq \bar {M}\vartheta + \tilde {M}_{1}\vartheta \int _{0}^{t}e^{-z_{1}s}(\|B_{1}(\cdot ,s)\|+\|B_{2}(\cdot ,s)\|)ds\\ &\leq \bar {M}\vartheta + 2\tilde {M}_{1}\bar {M}\vartheta ^{2} e^{2\tilde {M}_{1}\vartheta /z_{1}} \int _{0}^{t}e^{-z_{1}s}ds\\ &\leq \bar {M}\vartheta \left (1 + 2\tilde {M}_{1}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}}/ z_{1}\right ).\\ \end{aligned} \end{align}

The last equation in model (2.1) yields

\begin{align*} \dfrac {\partial P}{\partial t}\leq \alpha ^{m}\eta ^{m}\bar {M}\vartheta \left (1 + 2\tilde {M}_{1}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}}/ z_{1}\right )+\chi ^{m}b^{m}2\bar {M}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}}P - mP. \end{align*}

Let $ \tilde {M}_{2}=\alpha ^{m}\eta ^{m}\bar {M}\vartheta (1 + 2\tilde {M}_{1}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}}/ z_{1}),\ \tilde {M}_{3}=\chi ^{m}b^{m}2\bar {M}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}},$ we have

\begin{align*} \|P(x,t)\|\leq e^{-mt}\|P^{0}\|+\int _{0}^{t} e^{-m(t-s)} (\tilde {M}_{2}+\tilde {M}_{3}\|P(\cdot ,s)\|)ds. \end{align*}

Applying Gronwall’s inequality, we obtain

(4.31) \begin{align} \begin{aligned} \|P(x,t)\| &\leq (\vartheta e^{-mt}+\tilde {M}_{2})e^{\int _{0}^{t}\tilde {M}_{3}e^{-m(t-s)}ds}\\ &\leq (\vartheta e^{-mt}+\tilde {M}_{2})e^{\tilde {M}_{3}t}. \end{aligned} \end{align}

By the first equation of models (2.1) and (4.28), we get

\begin{align*} \dfrac {\partial S}{\partial t}-d_{S}\Delta S\gt \Lambda (x)-\mu (x)S-(2\bar {M}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}} \,\hat{\!\alpha }^{m})S. \end{align*}

Let $\,\hat{\!S}$ be the solution of the following model

(4.32) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial \,\hat{\!S}}{\partial t}=d_{S}\Delta \,\hat{\!S}+\Lambda (x)-\mu (x)\,\hat{\!S}-(2\bar {M}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}} \,\hat{\!\alpha }^{m})\,\hat{\!S}, \ \ x\in \Omega ,\ \ t\gt 0,\\[6pt] &\dfrac {\partial \,\hat{\!S}}{\partial \nu }=0,\ \ x\in \partial \Omega ,\ \ t\gt 0, \\[6pt] &\,\hat{\!S}(x,0)=S^{0}, \ \ x\in \Omega .\\ \end{aligned} \right . \end{align}

The comparison principle yields that $S(x,t)\ge \,\hat{\!S}(x,t)$ for $x\in \bar {\Omega }$ and $t\ge 0$ . Define $ S^{*}_{\vartheta }$ as the positive steady state of model (4.32) and $\,\hat{\!v}=\,\hat{\!S}-S^{*}_{\vartheta }$ , $ \,\hat{\!v}(x,t)$ satisfies

(4.33) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial \,\hat{\!v}}{\partial t}=d_{S}\Delta \,\hat{\!v}-(\mu (x)+2\bar {M}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}} \,\hat{\!\alpha }^{m})\,\hat{\!v}, \ \ x\in \Omega ,\ \ t\gt 0,\\[6pt] &\dfrac {\partial \,\hat{\!v}}{\partial \nu }=0,\ \ x\in \partial \Omega ,\ \ t\gt 0, \\[6pt] &\,\hat{\!v}(x,0)=S^{0}-S_{\vartheta }^{*}, \ \ x\in \Omega .\\ \end{aligned} \right . \end{align}

Define $ J_{1}(t)$ as the semigroup generated by $ d_{S}\Delta -\mu$ , then we have $ \|J_{1}(t)\|\leq \bar {M}e^{-\mu _{m}t}$ , where $ \bar {M}\gt 0$ and large enough. From model (4.33), we derive

\begin{align*} \,\hat{\!v}(\cdot ,t)=J_{1}(t)(S^{0}-S_{\vartheta }^{*})-\int _{0}^{t}J_{1}(t-s)2\bar {M}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}} \,\hat{\!\alpha }^{m}\,\hat{\!v}(\cdot ,s)ds. \end{align*}

Then

\begin{align*} \|\,\hat{\!v}(\cdot ,t)\|\leq \bar {M}\|S^{0}-S_{\vartheta }^{*}\|e^{-\mu _{m}t}+\int _{0}^{t}\bar {M}e^{-\mu _{m}(t-s)}2\bar {M}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}} \|\,\hat{\!\alpha }^{m}\|\|\,\hat{\!v}(\cdot ,s)\|ds. \end{align*}

Applying the Gronwall’s inequality again yields

(4.34) \begin{equation} \|\,\hat{\!S}(\cdot ,t)-S_{\vartheta }^{*}\|=\|\,\hat{\!v}(\cdot ,t)\|\leq \bar {M}\|S^{0}-S^{*}_{\vartheta }\| e^{(\tilde {M}_{4}-\mu _{m})t}, \end{equation}

where $\tilde {M}_{4}=2\bar {M}^{2}\vartheta e^{2\tilde {M}_{1}\vartheta /z_{1}}\|\,\hat{\!\alpha }^{m}\|.$

Choosing $\vartheta \gt 0$ small enough such that $ \tilde {M}_{4}\lt \mu _{m}/2$ , (4.34) yields that

(4.35) \begin{equation} \|\,\hat{\!S}(\cdot ,t)-S_{\vartheta }^{*}\|\leq \bar {M}\|S^{0}-S^{*}_{\vartheta }\| e^{-\mu _{m}t/2}. \end{equation}

Inequality (4.35) implies that

(4.36) \begin{align} \begin{aligned} S(\cdot ,t)-S^{*}&\ge \,\hat{\!S}(\cdot ,t)-S^{*}\\ &=\,\hat{\!S}(\cdot ,t)-S_{\vartheta }^{*}+S_{\vartheta }^{*}-S^{*}\\ &\ge -\bar {M}\|S^{0}-S^{*}_{\vartheta }\| e^{-\mu _{m}t/2}+S^{*}_{\vartheta }-S^{*}\\ &\ge -\bar {M}\left (\|S^{0}-S^{*}\|+\|S^{*}-S^{*}_{\vartheta }\|\right )-\|S^{*}_{\vartheta }-S^{*}\|\\ &\ge -\bar {M}\vartheta -(\bar {M}+1)\|S^{*}_{\vartheta }-S^{*}\|.\\ \end{aligned} \end{align}

Since $ a(t)\leq \vartheta \bar {M} e^{-z_{1}t}/S^{*}_{m} \leq \vartheta \bar {M}/S^{*}_{m}$ , we have

(4.37) \begin{equation} S(\cdot ,t)-S^{*}=S^{*}\left (\dfrac {S(x,t)}{S^{*}}-1\right )\leq \|S^{*}\|a(t)\leq \vartheta \bar {M}\|S^{*}\|/S^{*}_{m}. \end{equation}

From (4.36)–(4.37), we have

(4.38) \begin{equation} \|S(\cdot ,t)-S^{*}\|\leq \max \left \{\bar {M}\vartheta +(\bar {M}+1)\|S^{*}_{\vartheta }-S^{*}\|,\ \vartheta \bar {M}\|S^{*}\|/S^{*}_{m}\right \} . \end{equation}

Thus, from (4.28)–(4.31), (4.38) and $\lim _{\vartheta \to 0}S_{\vartheta }^{*}=S^{*}$ , we choose $\vartheta \gt 0$ small enough satisfying for $ t\gt 0$ ,

\begin{align*} \|S(x,t)-S^{*}\|, \|I_{1}(\cdot ,t)\|, \|I_{2}(\cdot ,t)\|, \|B_{1}(\cdot ,t)\|, \|B_{2}(\cdot ,t)\|, \|P(\cdot ,t)\| \leq \varepsilon . \end{align*}

This proves the local stability of $F_{0}=(S^{*},0,0,0,0,0)$ . Then we need to discuss the global attractivity of $F_{0}$ . Theorem3.1 implies that $\,\,\hat{\!\!J}(t)$ has a connected global attractor $\mathscr{D}$ . From Lemma 4.7, the eigenvalue problem (4.23) has a positive eigenvector $ (\phi _{2},\phi _{3},\phi _{4})$ with $ s(\mathscr{A} \, )=0$ . Let

\begin{align*} \partial Y_{1}=\left \{(\widehat{S},\widehat{I_{1}},\widehat{I_{2}},\widehat{B}_{1},\widehat{B}_{2},\widehat{P})\in \mathit{H^{+}}\,:\,\widehat{I}_{1}=\widehat{I}_{2}=\widehat{B}_{1}=\widehat{B}_{2}=\widehat{P}=0\right \}. \end{align*}

We present the following two claims.

Claim 1. For any $ c^{0}=(S^{0},I_{1}^{0},I_{2}^{0},B_{1}^{0},B_{2}^{0},P^{0}) \in \mathscr{D}$ , the omega limit set $\omega (c^{0})\subset \partial Y_{1}$ .

In view of (4.2), we have $S^{0}\leq S^{*}$ . If $ I_{1}^{0}=I_{2}^{0}=B_{1}^{0}=B_{2}^{0}=P^{0}=0$ , the claim easily follows from the fact that $\partial Y_{1}$ is invariant for $\,\,\hat{\!\!J}(t)$ . Hence, we assume that either $ I_{1}^{0}\not =0$ or $ I_{2}^{0}\not =0$ or $ B_{1}^{0}\not =0$ or $ B_{2}^{0}\not =0$ or $ P^{0}\not =0$ . From the results in Theorem3.1, we easily know that $c(x,t,c^{0})\gt 0$ for $ x\in \bar {\Omega }$ and $ t\gt 0$ . $ S(x,t)$ satisfies

\begin{align*} \left \{ \begin{aligned} &\dfrac {\partial S}{\partial t}\lt d_{S}\Delta S +\Lambda (x)-\mu (x) S, \ \ x\in \Omega , \ t\gt 0, \\[6pt] &\dfrac {\partial S}{\partial \nu }=0, \ \ x\in \partial \Omega ,\ t\gt 0,\\[6pt] &S(x,0)\leq S^{*}(x),\ \ x\in \Omega . \end{aligned} \right . \end{align*}

Applying the comparison principle, we have $ S(x,t)\lt S^{*}(x)$ for $ x\in \bar {\Omega }$ and $t\gt 0$ . Motivated by [Reference Cui, Lam and Lou9], we assume

\begin{align*} u(t\,;\,c^{0})= \inf \left \{\tilde {u}\in \mathbb{R}: I(\cdot ,t)\leq \tilde {u}\phi _{2} \text{ and } B_{1}(\cdot ,t)\leq \tilde {u}\phi _{3} \text{ and } B_{2}(\cdot ,t)\leq \tilde {u}\phi _{4}\right \}, \end{align*}

and $ u(t\,;\,c^{0})\gt 0$ holds for all $ t\gt 0$ , we also know that $ u(t\,;\,c^{0})$ is strictly decreasing. Then we choose a $ t_{2}\gt 0$ satisfying $ \widetilde{I}(\cdot ,t)=u(t_{2}\,;\,c^{0})\phi _{2}, \widetilde{B}_{1}(\cdot ,t)=u(t_{2}\,;\,c^{0})\phi _{3}, \widetilde{B}_{2}(\cdot ,t)=u(t_{2}\,;\,c^{0})\phi _{4}$ for $ t\ge t_{2}$ . Note that $ S(x,t)\lt S^{*}$ , we have

(4.39) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial \widetilde{I}}{\partial t}\gt d_{I}\Delta \widetilde{I} + \dfrac {\alpha _{1}(x)(S^{*}-\vartheta )\widetilde{B}_{1}}{H_{1}+\vartheta } + \dfrac {\alpha _{2}(x)(S^{*}-\vartheta )\widetilde{B}_{2}}{H_{2}+\vartheta } -\mu (x) \widetilde{I},\ \ x\in \Omega ,\ \ t\gt t_{2}, \\[3pt] &\dfrac {\partial \widetilde{B}_{1}}{\partial t}=\eta (x) \widetilde{I} +\dfrac {\partial h_{1}(x,\vartheta )}{\partial B_{1}}\widetilde{B}_{1}-b_{1}(x)\vartheta \widetilde{B}_{1} -\delta _{1}(x) \widetilde{B}_{1},\ \ x\in \Omega ,\ \ t\gt t_{2}, \\[3pt] &\dfrac {\partial \widetilde{B}_{2}}{\partial t}=\delta _{1}(x) \widetilde{B}_{1}+\dfrac {\partial h_{2}(x,\vartheta )}{\partial B_{2}}\widetilde{B}_{2}-b_{2}(x)\vartheta \widetilde{B}_{2} -\delta _{2}(x) \widetilde{B}_{2},\ \ x\in \Omega ,\ \ t\gt t_{2},\\[3pt] &\dfrac {\partial \widetilde{I}}{\partial \nu }=0,\ \ x\in \partial \Omega , \ \ t\gt t_{2},\\[3pt] &\widetilde{I}(x,t_{2})\ge I(x,t_{2}),\ \widetilde{B}_{1}(x,t_{2})\ge B_{1}(x,t_{2}),\ \widetilde{B}_{2}(x,t_{2})\ge B_{2}(x,t_{2}),\ \ x\in \Omega . \end{aligned} \right . \end{align}

Applying the comparison principle, we have $ (\widetilde{I}(x,t),\widetilde{B}_{1}(x,t), \widetilde{B}_{2}(x,t))\ge (I(x,t),B_{1}(x,t), B_{2}(x,t))$ for $ x\in \bar {\Omega }$ and $ t\ge t_{2}$ . From model (4.39) and the strong comparison principle, we have $u(t_{2}\,;\,c^{0})\phi _{2} =\widetilde{I}(x,t) \gt I(x,t)$ , $u(t_{2}\,;\,c^{0})\phi _{3} =\widetilde{B}_{1}(x,t) \gt {B_{1}}(x,t),\ u(t_{2}\,;\,c^{0})\phi _{4} =\widetilde{B}_{2}(x,t) \gt B_{2}(x,t)$ for $x\in \bar {\Omega }$ and $ t\gt t_{2}$ . Due to $ t_{2}\gt 0$ is arbitrary, $ u(t\,;\,c^{0})$ is strictly decreasing function. Define $ u_{*}=\lim _{t\to \infty }u(t\,;\,c^{0})$ , which implies $ u_{*}=0$ , and define $ \mathbb{T} =(\mathbb{T}_{1}, \mathbb{T}_{2},\mathbb{T}_{3},\mathbb{T}_{4},\mathbb{T}_{5},\mathbb{T}_{6})\in \omega (c^{0})$ . Then there exists a sequence $ \{t_{k}\}$ with $ t_{k}\to \infty$ satisfying $\,\,\hat{\!\!J}(t_{k})c^{0}\to \mathbb{T}$ . We easily obtain $u(t,\mathbb{T})=u_{*}$ for $ t\ge 0$ . In view of $\lim _{t_{k}\to \infty }\,\hat{\!J}(t+t_{k})c^{0}=\,\hat{\!J}(t)\lim _{t_{k}\to \infty }\,\hat{\!J}(t_{k})c^{0}=\,\hat{\!J}(t)\mathbb{T}.$ If $ \mathbb{T}_{2}\not =0$ or $ \mathbb{T}_{3}\not =0$ or $ \mathbb{T}_{4}\not =0$ or $ \mathbb{T}_{5}\not =0$ or $ \mathbb{T}_{6}\not =0$ , it follows from [Reference Wu and Zou58, Theorem 3.12] that $u(t\,;\,\mathbb{T})$ is strictly decreasing. This leads to a contradiction. Hence, we have $ \mathbb{T}_{2}=\mathbb{T}_{3}=\mathbb{T}_{4}=\mathbb{T}_{5}=\mathbb{T}_{6}=0$ . By the theory of asymptotically autonomous semiflows in [Reference Thieme41], it follows that

\begin{align*} \lim _{t\to \infty }\|(S(x,t),I_{1}(x,t),I_{2}(x,t),B_{1}(x,t),B_{2}(x,t), P(x,t))-(S^{*},0,0,0,0,0)\|=0 . \end{align*}

Claim 2. $\mathscr{D}=\{F_{0}\}$ .

Since $ \{F_{0}\}$ is globally attractive in $ \partial Y_{1}$ , $ \{F_{0}\}$ is the unique invariant subset of model (2.1)–(2.3) in $ \partial Y_{1}$ . In view of the fact that the omega limit set $\omega (c^{0})$ is compact invariant and $\omega (c^{0})\subset \partial Y_{1}$ , for any $ c^{0}\in \mathscr{D}$ , we get $\omega (c^{0})=\{F_{0}\}$ . Since the global attractor $\mathscr{D}$ is compact invariant in $H^+$ , $F_0$ is stable, and according to [Reference Wu and Zou58, Lemma 3.11], one gets $\mathscr{D}=\left \{F_{0}\right \}$ . Combining the global attractivity and local stability, we immediately obtain the globally asymptotical stability of $F_0$ , completing the proof of Theorem3.7 (ii).

4.4 Proof of Theorem3.8

If $ \mathscr{R}_{0}\gt 1$ , model (2.1) has a phage-free steady state $ F_{1} = (S^{a}(x),I_{1}^{a}(x),0,B_{1}^{a}(x),B_{2}^{a}(x),0)$ . Let $ I_{2}(x,t)=P(x,t)=0$ in model (2.1), we have

(4.40) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial S}{\partial t}= d_{S}\Delta S +\Lambda (x) - \alpha _{1}(x) f_{1}(B_{1})S -\alpha _{2}(x) f_{2}(B_{2})S -\mu (x) S,\quad x\in \Omega ,\ t\gt 0,\\[3pt] &\dfrac {\partial I_{1}}{\partial t}= d_{I}\Delta I_{1} + \alpha _{1}(x)f_{1}(B_{1})S + \alpha _{2}(x)f_{2}(B_{2})S -\mu (x) I_{1},\quad x\in \Omega ,\ t\gt 0,\\[3pt] &\dfrac {\partial B_{1}}{\partial t}=h_{1}(x,B_{1}) +\eta (x)I_{1} -\delta _{1}(x) B_{1},\quad x\in \Omega ,\ t\gt 0,\\[3pt] &\dfrac {\partial B_{2}}{\partial t}=h_{2}(x,B_{2}) +\delta _{1}(x)B_{1} - \delta _{2}(x) B_{2},\quad x\in \Omega ,\ t\gt 0,\\[3pt] &\dfrac {\partial S}{\partial \nu } = \dfrac {\partial I_{1}}{\partial \nu }=0,\quad x\in \partial \Omega ,\ t\gt 0,\\[3pt] &S(x,0)=S^{0}(x),\ I_{1}(x,0)=I_{1}^{0}(x),\ B_{1}(x,0) = B_{1}^{0}(x),\ B_{2}(x,0) = B_{2}^{0}(x),\quad x\in \Omega . \end{aligned} \right . \end{align}

By applying a similar proof as in Theorem3.1, for model (4.40), we have the following corollary:

Corollary 4.8. Let $ \mathbb{E}=C(\bar {\Omega },\mathbb{R}^{4})$ be the Banach space and its positive cone is denoted by $ \mathbb{E}^{+}$ .

  1. (i) For any $ c^{0}(x) = (S^{0}(x),I_{1}^{0}(x),B_{1}^{0}(x),B_{2}^{0}(x))\in \mathbb{E}^{+}$ , model (4.40) admits a unique global nonnegative classical solution $\bar {c}(\cdot ,t\,;\,c^{0})$ defined on $\Omega \times [0,\infty )$ .

  2. (ii) Let $\tilde {J}_{1}(t): \mathbb{E}^{+}\to \mathbb{E}^{+}$ be the semiflow generated by model (4.40), namely $ \tilde {J}_{1}(t)c^{0}=\bar {c}(x,t)$ for $ t\gt 0$ . Moreover, $ \tilde {J}_{1}(t)$ is point dissipative.

  3. (iii) The semiflow $ \tilde {J}_{1}(t)$ of model (4.40) has a connected global attractor in $ \mathbb{E}^{+}$ .

Note that model (4.40) has a disease-free steady state $ \bar {F}_{0}=(S^{*}(x),0,0,0)$ . Linearising model (4.40) at $ \bar {F}_{0}$ yields

(4.41) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial I_{1}}{\partial t}= d_{I}\Delta I_{1} + \dfrac {\alpha _{1}(x)S^{*}B_{1}}{H_{1}} + \dfrac {\alpha _{2}(x)S^{*}B_{2}}{H_{2}} -\mu (x) I_{1},\ \ x\in \Omega ,\ \ t\gt 0, \\[3pt] &\dfrac {\partial B_{1}}{\partial t}=\,\hat{\!h}_{1}(x)B_{1} + \eta (x) I_{1} -\delta _{1}(x) B_{1},\ \ x\in \Omega ,\ \ t\gt 0, \\[3pt] &\dfrac {\partial B_{2}}{\partial t}=\,\hat{\!h}_{2}(x)B_{2} + \delta _{1}(x) B_{1}- \delta _{2}(x) B_{2},\ \ x\in \Omega ,\ \ t\gt 0,\\[3pt] &\dfrac {\partial I_{1}}{\partial \nu }=0,\ \ x\in \partial \Omega , \ \ t\gt 0. \end{aligned} \right . \end{align}

We continue the discussion on the threshold dynamics of model (4.40). The following theorem presents the uniform persistence and the existence of phage-free steady state for model (4.40) when $\mathscr{R}_{0}\gt 1$ .

Theorem 4.9. If $\mathscr{R}_{0}\gt 1$ , there exists $\bar {\vartheta }\gt 0$ such that for any $ c^{0}(\!\cdot\!)=(S^{0},I_{1}^{0},B_{1}^{0},B_{2}^{0})(\!\cdot\!)\in \mathbb{E}^{+}$ with $ I_{1}^{0}(x)\not \equiv 0$ or $ B_{1}^{0}(x)\not \equiv 0$ or $ B_{2}^{0}(x)\not \equiv 0$ , the solution $ \bar {c}(x,t\,;\,c^{0})=(S(x,t),I_{1}(x,t),B_{1}(x,t),B_{2}(x,t))$ of model (4.40) satisfies $\lim _{t\to \infty }\inf \bar {c}(x,t)\ge \bar {\vartheta } \text{ uniformly for } x\in \bar {\Omega }.$ Moreover, model (4.40) has at least one positive steady state.

Before proving Theorem4.9, we first give some preliminaries. We define

\begin{align*} \mathbb{E}_{0}=\{\varphi \in \mathbb{E}^{+}\,:\,\varphi _{1}(\!\cdot\!)\gt 0,\varphi _{2}(\!\cdot\!)\not \equiv 0,\varphi _{3}(\!\cdot\!)\not \equiv 0,\varphi _{4}(\!\cdot\!)\not \equiv 0 \}, \end{align*}

and

\begin{align*} \partial \mathbb{E}_{0}\,:=\mathbb{E}^{+}\backslash \mathbb{E}_{0}=\{\varphi \in \mathbb{E}^{+}\,:\,\varphi _{2}(\!\cdot\!)\equiv 0\text{ or } \varphi _{3}(\!\cdot\!)\equiv 0 \text{ or } \varphi _{4}(\!\cdot\!)\equiv 0 \}. \end{align*}

Define

\begin{align*} F_{\partial }\,:=\{\varphi \in \partial \mathbb{E}_{0}\,:\,\tilde {J}_{1}(t)\varphi \in \partial \mathbb{E}_{0}\}, \end{align*}

for $t\ge 0$ , and $\omega (\varphi )$ be the omega limit set of the orbit $\mathit{G}^{+}\,:=\{\tilde {J}_{1}(t)\varphi\,:\,t\geq 0\}$ .

Claim 1. $\mathbb{E}_{0}$ is positively invariant regarding $\tilde {J}_{1}(t)$ , namely $\tilde {J}_{1}(t)\mathbb{E}_{0}\subseteq \mathbb{E}_{0}$ for all $t\ge 0$ .

Let $c^{0}\in \mathbb{E}_{0}$ , which implies $ I_{1}^{0}\not \equiv 0$ and $ B_{1}^{0}\not \equiv 0$ and $ B_{2}^{0}\not \equiv 0$ . We derive that $\partial I_{1}/\partial t \ge d_{I}\Delta I_{1}-\mu (x)I_{1}$ from model (4.40). Thus, $ I_{1}$ is an upper solution of the problem

\begin{align*} \left \{ \begin{aligned} &\dfrac {\partial I_{1}^{*}(x,t)}{\partial t}= d_{I}\Delta I_{1}^{*}-\mu (x)I_{1}^{*}(x,t),\ \ x\in \Omega ,\ \ t\gt 0, \\[3pt] &\dfrac {\partial I_{1}^{*}(x,t)}{\partial \nu }=0,\ \ x\in \partial \Omega ,\ \ t\gt 0, \\[3pt] &I_{1}^{*}(\cdot ,0)=I_{1}(\cdot ,0)=I_{1}^{0}(\!\cdot\!),\ \ x\in \Omega .\\ \end{aligned} \right . \end{align*}

It follows from the maximum principle and $I_{1}^{0}\not \equiv 0$ that $I_{1}^{*}(x,t)\gt 0$ for $x\in \bar {\Omega }$ and $t\gt 0$ . It holds that $I_{1}(x,t)\ge I_{1}^{*}(x,t)\gt 0$ for $x\in \bar {\Omega }$ and $ t\gt 0$ . Assume that there exist $ \tilde {t}\gt 0$ and $ \tilde {x}\in \bar {\Omega }$ such that $ B_{1}(\tilde {x},\tilde {t})=0$ , and from the third equation of model (4.40), along with $(\mathbf{H}_{\mathbf{3}})$ , we derive

\begin{align*} 0=\dfrac {\partial B_{1}(\tilde {x},\tilde {t})}{\partial t}= \eta (\tilde {x})I_{1}(\tilde {x},\tilde {t}). \end{align*}

It implies that $\eta (\tilde {x})I_{1}(\tilde {x},\tilde {t})=0$ , which leads to a contradiction. Hence, we get $B_{1}(x,t)\gt 0$ for $ x\in \bar {\Omega }$ and $ t\gt 0$ . From the fourth equation of model (4.40), one gets

\begin{align*} \dfrac {\partial B_{2}}{\partial t}\ge h_{2}(x,B_{2})+\delta _{1}(x)B_{1}-\delta _{2}(x)B_{2},\ x\in \Omega , \ t\gt 0. \end{align*}

Similar to [Reference Hsu, Jiang and Wang17, Lemma 2.1] and [Reference Wang44, Proposition 3.1], by strong maximum principle [Reference Protter and Weinberger34, Theorem 4] and Hopf boundary theorem [Reference Protter and Weinberger34, Theorem 3], we derive $ B_{2}(x,t)\gt 0$ for $x\in \bar {\Omega }$ and $ t\gt 0$ . Then $\tilde {J}_{1}(t)c^{0}\in \mathbb{E}_{0}$ .

Claim 2. If $\mathscr{R}_{0}\gt 1$ , there exists $\bar {\delta }\gt 0$ such that the semiflow $\tilde {J}_{1}(t)$ of model (4.40) satisfies $\lim _{t\to \infty }\sup \|\tilde {J}_{1}(t)\varphi -\bar {F}_{0}\|\ge \bar {\delta }$ for all $ \varphi \in \mathbb{E}_{0}$ .

By way of contradiction, assuming that there exists a $\varphi _{0}\in \mathbb{E}_{0}$ satisfying

\begin{align*} \lim \limits _{t\to +\infty } \|\tilde {J}_{1}(t)\varphi _{0}-(S^{*}(x),0,0,0)\|\lt \bar {\delta }. \end{align*}

Thus, there exists $ t^{*}\gt 0$ satisfying $ S(x,t,\varphi _{0})\ge S^{*}(x)-\bar {\delta },\ \forall t\ge t^{*}.$ Therefore, $ (I_{1}(x,t,\varphi _{0}), B_{1}(x,t,\varphi _{0}), B_{2}(x,t,\varphi _{0}))$ is an upper solution of the following linear model

(4.42) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial \tilde {I}_{1}}{\partial t}= d_{I}\Delta \tilde {I}_{1} + \dfrac {\alpha _{1}(x)(S^{*}-\bar {\delta })\tilde {B}_{1}}{H_{1}+\bar {\delta }} + \dfrac {\alpha _{2}(x)(S^{*}-\bar {\delta })\tilde {B}_{2}}{H_{2}+\bar {\delta }} -\mu (x) \tilde {I}_{1},\ \ x\in \Omega ,\ \ t\ge t^{*},\\[3pt] &\dfrac {\partial \tilde {B}_{1}}{\partial t}=\eta (x) \tilde {I_{1}} +\dfrac {\partial h_{1}(x,\bar {\delta })}{\partial B_{1}}\tilde {B}_{1}-b_{1}(x)\bar {\delta }\tilde {B}_{1} -\delta _{1}(x) \tilde {B}_{1},\ \ x\in \Omega ,\ \ t\ge t^{*},\\[3pt] &\dfrac {\partial \tilde {B}_{2}}{\partial t}=\delta _{1}(x) \tilde {B}_{1}+\dfrac {\partial h_{2}(x,\bar {\delta })}{\partial B_{2}}\tilde {B}_{2}-b_{2}(x)\bar {\delta }\tilde {B}_{2} -\delta _{2}(x) \tilde {B}_{2},\ \ x\in \Omega ,\ \ t\ge t^{*},\\[3pt] &\dfrac {\partial \tilde {I}_{1}}{\partial \nu }=0,\ \ x\in \partial \Omega , \ \ t\ge t^{*},\\[3pt] &\tilde {I}_{1}(x,t^{*})=\varphi _{2}\leq I_{1}(\cdot ,t^{*}),\ \tilde {B}_{1}(x,t^{*})=\varphi _{3}\leq B_{1}(\cdot ,t^{*}),\ \tilde {B}_{2}(x,t^{*})=\varphi _{4}\leq B_{2}(\cdot ,t^{*}).\\ \end{aligned} \right . \end{align}

Denote $\lambda _{\bar {\delta }}^{0}$ as the principal eigenvalue of the following eigenvalue problem

(4.43) \begin{align} \left \{ \begin{aligned} &d_{I}\Delta \varphi -\mu (x)\varphi +\left (\dfrac {\alpha _{1}(x)(S^{*}-\bar {\delta })\eta (x)}{H_{1}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))}+\dfrac {\alpha _{2}(x)(S^{*}-\bar {\delta })\eta (x)\delta _{1}(x)}{H_{2}(\delta _{1}(x)-\,\hat{\!h}_{1}(x))(\delta _{2}(x)-\,\hat{\!h}_{2}(x))}\right )\varphi =\lambda \varphi ,\ x\in \Omega ,\\ &\dfrac {\partial \varphi }{\partial \nu }=0,\ x\in \partial \Omega .\\ \end{aligned} \right . \end{align}

Then $\lambda _{\bar {\delta }}^{0}$ is continuous in $\bar {\delta }$ . The following eigenvalue problem

(4.44) \begin{align} \left \{ \begin{aligned} &\lambda \varphi _{2}= d_{I}\Delta \varphi _{2} + \dfrac {\alpha _{1}(x)(S^{*}-\bar {\delta })}{H_{1}}\varphi _{3} + \dfrac {\alpha _{2}(x)(S^{*}-\bar {\delta })}{H_{2}}\varphi _{4} -\mu (x)\varphi _{2},\ \ x\in \Omega ,\\[3pt] &\lambda \varphi _{3}=\eta (x) \varphi _{2} +\,\hat{\!h}_{1}(x)\varphi _{3} -\delta _{1}(x) \varphi _{3},\ \ x\in \Omega ,\\[3pt] &\lambda \varphi _{4}=\delta _{1}(x) \varphi _{3}+\,\hat{\!h}_{2}(x)\varphi _{4}- \delta _{2}(x) \varphi _{4},\ \ x\in \Omega ,\\[3pt] &\dfrac {\partial \varphi _{2}}{\partial \nu }=0,\ \ x\in \partial \Omega , \end{aligned} \right . \end{align}

has a principal eigenvalue $\tilde {\lambda }_{\bar {\delta }}^{0}$ with respect to positive eigenvector $ \left (\varphi _{2}^{\bar {\delta }}(x),\varphi _{3}^{\bar {\delta }}(x),\varphi _{4}^{\bar {\delta }}(x)\right )$ . We can choose a small enough $ \rho _{1}\gt 0$ such that

\begin{align*} (I_{1}(\cdot ,t^{*}\,;\,\varphi _{0}), B_{1}(\cdot ,t^{*}\,;\,\varphi _{0}), B_{2}(\cdot ,t^{*}\,;\,\varphi _{0}))\ge \rho _{1}\left (\varphi _{2}^{\bar {\delta }}(x),\varphi _{3}^{\bar {\delta }}(x),\varphi _{4}^{\bar {\delta }}(x)\right ). \end{align*}

Then the linear model (4.42) with initial conditions $ (\tilde {I}_{1}(x,t^{*}), \tilde {B}_{1}(x,t^{*}), \tilde {B}_{2}(x,t^{*}))=\rho _{1}\left (\varphi _{2}^{\bar {\delta }}(x),\varphi _{3}^{\bar {\delta }}(x),\varphi _{4}^{\bar {\delta }}(x)\right )$ has a unique solution

\begin{align*} \left (\tilde {I}_{1}(x,t^{*}\,;\,\varphi _{0}), \tilde {B}_{1}(x,t^{*}\,;\,\varphi _{0}), \tilde {B}_{2}(x,t^{*}\,;\,\varphi _{0})\right )=\rho _{1}e^{\tilde {\lambda }_{\bar {\delta }}^{0}(t-t^{*})}\left (\varphi _{2}^{\bar {\delta }}(x),\varphi _{3}^{\bar {\delta }}(x),\varphi _{4}^{\bar {\delta }}(x)\right ). \end{align*}

By the comparison principle of quasimonotone model, we have

\begin{align*} (I_{1}(x,t^{*}\,;\,\varphi _{0}), B_{1}(x,t^{*}\,;\,\varphi _{0}), B_{2}(x,t^{*}\,;\,\varphi _{0}))\ge \left (\tilde {I}_{1}(x,t^{*}\,;\,\varphi _{0}), \tilde {B}_{1}(x,t^{*}\,;\,\varphi _{0}), \tilde {B}_{2}(x,t^{*}\,;\,\varphi _{0})\right ), \end{align*}

on $\bar {\Omega }\,\times [t^{*},\infty )$ , which implies that $\lim _{t\to \infty }\|(I_{1}(\cdot ,t\,;\,\varphi _{0}),B_{1}(x,t\,;\,\varphi _{0}), B_{2}(x,t\,;\,\varphi _{0}))\|=\infty$ , which contradicts the boundedness of $ (I_{1}(\cdot ,t), B_{1}(\cdot ,t), B_{2}(\cdot ,t))$ by Corollary 4.8. This establishes Claim 2.

Proof of Theorem 4.9 First, we prove that $\omega (\varphi )=\{\bar {F}_{0}\}, \varphi \in F_{\partial }$ . For any $\varphi \in F_{\partial }$ , we have $\tilde {J}_{1}(t)\varphi \in F_{\partial }$ , $t\ge 0$ . Then, $I_{1}(\cdot ,t)\equiv 0$ or $B_{1}(\cdot ,t)\equiv 0$ or $B_{2}(\cdot ,t)\equiv 0$ for $ t\ge 0$ . In the case that $I_{1}(\cdot ,t)\equiv 0$ , we can derive that $B_{1}(\cdot ,t)\equiv 0$ and $B_{2}(\cdot ,t)\equiv 0$ from the second equation of model (4.40). Therefore, $ S(\cdot ,t)\to S^{*}(x)$ uniformly for $x\in \bar {\Omega }.$ In the case that $B_{1}(\cdot ,t)\equiv 0$ , we can derive that $I_{1}(\cdot ,t)\equiv 0$ from the third equation of model (4.40), then $B_{2}(\cdot ,t)\equiv 0$ and $ S(\cdot ,t)\to S^{*}(x)$ uniformly for $x\in \bar {\Omega }.$ In the case that $B_{2}(\cdot ,t)\equiv 0$ , we can derive that $B_{1}(\cdot ,t)\equiv 0$ from the fourth equation of model (4.40), then $I_{1}(\cdot ,t)\equiv 0$ and $ S(\cdot ,t)\to S^{*}(x)$ uniformly for $x\in \bar {\Omega }.$ This shows $\omega (\varphi ) =\{\bar {F}_{0}\}$ .

Define a continuous function

\begin{align*} \tau (\varphi )=\min \left \{ \min _{x\in \bar {\Omega }} \varphi _{2}(x), \min _{x\in \bar {\Omega }} \varphi _{3}(x), \min _{x\in \bar {\Omega }} \varphi _{4}(x)\right \},\ \varphi \in \mathbb{E}^{+}. \end{align*}

Apparently, $\tau ^{-1}(0,\infty )\subseteq \mathbb{E}_{0}$ . The function $\tau$ is a generalised distance function for the semiflow $\tilde {J}_{1}(t)$ . In view of Claim 1 and Claim 2, we know that the sigleton $ \{\bar {F}_{0}\}$ is an isolated invariant set for $\tilde {J}_{1}(t)$ in $ \mathbb{E}^{+}$ , then $ W^{s}(\{\bar {F}_{0}\})\cap \mathbb{E}_{0}=\emptyset$ , where $ W^{s}(\{\bar {F}_{0}\})$ represents the stable subset of $ \{\bar {F}_{0}\}$ . Besides, no subset of $ \{\bar {F}_{0}\}$ forms a cycle in $ \partial \mathbb{E}_{0}$ . By [Reference Smith and Zhao40, Theorem 3], there exists a $\vartheta _{1}\gt 0$ satisfying $\lim _{t\to \infty }\inf \tau (\tilde {J}_{1}(t)\varphi )\ge \vartheta _{1},\ \forall \varphi \in \mathbb{E}_{0}$ . Moreover, from Corollary 4.8, there exists $\tilde {t}_{1}\gt 0$ satisfying that $ B_{1}(\cdot ,t)\leq \,\hat{\!M}_{\infty }$ , $ B_{2}(\cdot ,t)\leq \,\hat{\!M}_{\infty }$ for $ x\in \Omega$ and $ t\ge \tilde {t}_{1}$ . Then, we get

\begin{align*} \dfrac {\partial S}{\partial t}\ge d_{S}\Delta S+\Lambda _{m}-((\alpha _{1}+\alpha _{2})\,\hat{\!M}_{\infty }+\mu ^{m})S, \quad x\in \Omega ,\ t\gt \tilde {t}_{1}. \end{align*}

Thus, we have $\lim _{t\to \infty }\inf S(x,t\,;\,\varphi )\gt \vartheta _{2}:=\Lambda _{m}/((\alpha _{1}+\alpha _{2})\,\hat{\!M}_{\infty }+\mu ^{m})$ . Let $ \bar {\vartheta }=\min \{\vartheta _{1},\vartheta _{2}\}$ . This establishes the uniform persistence. It follows from [Reference Magal and Zhao26, Theorem 4.7] that model (4.40) admits at least one steady state in $\mathbb{E}_{0}$ , which is positive. This establishes Theorem4.9.

Based on above theorem and [Reference Zhao60, Theorem 1.3.6]. This establishes Theorem3.8.

4.5 Proof of Theorem3.10

We choose the Lyapunov function

\begin{align*} \begin{aligned} \mathit{L}_{1}(t)=&\int _{\Omega }\tilde {S}^{a}Y\left (\dfrac {S}{\tilde {S}^{a}}\right )dx+\int _{\Omega }\tilde {I}_{1}^{a}Y\left (\dfrac {I_{1}}{\tilde {I}_{1}^{a}}\right )dx+\int _{\Omega }I_{2}dx+l_{1}\int _{\Omega }\tilde {B}_{1}^{a}Y\left (\dfrac {B_{1}}{\tilde {B}_{1}^{a}}\right )dx \\ &+ l_{2}\int _{\Omega }\tilde {B}_{2}^{a}Y\left (\dfrac {B_{2}}{\tilde {B}_{2}^{a}}\right )dx+l_{3} \int _{\Omega }Pdx, \end{aligned} \end{align*}

where the constants $ l_{1}\gt 0,l_{2}\gt 0,l_{3} \gt 0$ are to be determined, and $ Y(x)=x-1-\ln x (x\gt 0)$ . By calculating the derivative of $ \mathit{L}_{1}(t)$ , one gets

\begin{align*}\dfrac {d\mathit{L}_{1}(t)}{dt}=&\int _{\Omega }\left (1-\dfrac {\tilde {S}^{a}}{S}\right )\left (d_{S}\Delta S +\Lambda - \dfrac {\alpha _{1}B_{1}S}{B_{1}+H_{1}} -\dfrac {\alpha _{2}B_{2}S}{B_{2}+H_{2}} -\mu S\right )dx\\[3pt] &+\int _{\Omega }\left (1-\dfrac {\tilde {I}_{1}^{a}}{I_{1}}\right )\left (d_{I}\Delta I_{1} +\dfrac {\alpha _{1}LB_{1}S}{(L+P)(B_{1}+H_{1})} +\dfrac {\alpha _{2}LB_{2}S}{(L+P)(B_{2}+H_{2})} -\mu I_{1}\right )dx \\ &+\int _{\Omega }\left (d_{I}\Delta I_{2} +\dfrac {\alpha _{1}PB_{1}S}{(L+P)(B_{1}+H_{1})} +\dfrac {\alpha _{2}PB_{2}S}{(L+P)(B_{2}+H_{2})} -\mu I_{2}\right )dx\\[3pt] &+l_{1}\int _{\Omega }\left (1-\dfrac {\tilde {B}_{1}^{a}}{B_{1}}\right )\left (h_{1}(B_{1})+\eta (I_{1}+I_{2}) - b_{1}B_{1}P- \delta _{1}B_{1}\right )dx\\[3pt] &+l_{2}\int _{\Omega }\left (1-\dfrac {\tilde {B}_{2}^{a}}{B_{2}}\right )\left (h_{2}(B_{2})+\delta _{1}B_{1} - \delta _{2}B_{2}\right )dx\\[3pt] &+l_{3}\int _{\Omega }\left (\alpha \eta I_{2}+ \chi _{1}b_{1}B_{1}P-mP \right )dx.\end{align*}

Since $ \tilde {F}_{1}=(\tilde {S}^{a},\tilde {I}_{1}^{a},0,\tilde {B}_{1}^{a},\tilde {B}_{2}^{a},0)$ is the phage-free steady state of model (2.4), we derive

\begin{align*}&\dfrac {d\mathit{L}_{1}(t)}{dt}\leq -d_{S}\int _{\Omega }\tilde {S}^{a}\dfrac {|\nabla S|^{2}}{S^{2}}dx-d_{I}\int _{\Omega }\tilde {I}_{1}^{a}\dfrac {|\nabla I_{1}|^{2}}{I_{1}^{2}}dx+d_{I}\int _{\Omega } \Delta I_{2}dx-\int _{\Omega }\mu S \left (1-\dfrac {\tilde {S}^{a}}{S}\right )^{2}dx\\[3pt] &\quad+\int _{\Omega }(l_{3}\alpha \eta -\mu )I_{2}dx+\int _{\Omega }l_{1}\eta \tilde {I}_{1}^{a}\left (1-\dfrac {B_{1}}{\tilde {B}_{1}^{a}}+\dfrac {I_{1}}{\tilde {I}_{1}^{a}}-\dfrac {I_{1}\tilde {B}_{1}^{a}}{\tilde {I}_{1}^{a}B_{1}}\right )dx+\int _{\Omega }l_{2}\delta _{1}\tilde {B}_{1}^{a}\left (1-\dfrac {B_{2}}{\tilde {B}_{2}^{a}}+\dfrac {B_{1}}{\tilde {B}_{1}^{a}}-\dfrac {B_{1}\tilde {B}_{2}^{a}}{\tilde {B}_{1}^{a}B_{2}}\right )dx\\[3pt] &\quad+\int _{\Omega }l_{1}\left (B_{1}-\tilde {B}_{1}^{a}\right )\left (\dfrac {h_{1}(B_{1})}{B_{1}}-\dfrac {h_{1}(\tilde {B}_{1}^{a})}{\tilde {B}_{1}^{a}}\right )dx+\int _{\Omega }l_{2}\left (B_{2}-\tilde {B}_{2}^{a}\right )\left (\dfrac {h_{2}(B_{2})}{B_{2}}-\dfrac {h_{2}(\tilde {B}_{2}^{a})}{\tilde {B}_{2}^{a}}\right )dx \\[3pt] &\quad+\int _{\Omega }\dfrac {\alpha _{1}\tilde {S}^{a}\tilde {B}_{1}^{a}}{H_{1}+\tilde {B}_{1}^{a}}\left [ 2-\dfrac {\tilde {S}^{a}}{S}-\dfrac {I_{1}}{\tilde {I}_{1}^{a}}+\dfrac {B_{1}/(H_{1}+B_{1})}{\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}-\dfrac {S\tilde {I}_{1}^{a}B_{1}/(H_{1}+B_{1})}{\tilde {S}^{a}I_{1}\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}\right ]dx\\[3pt] &\quad+\int _{\Omega }\dfrac {\alpha _{2}\tilde {S}^{a}\tilde {B}_{2}^{a}}{H_{2}+\tilde {B}_{2}^{a}}\left [ 2-\dfrac {\tilde {S}^{a}}{S}-\dfrac {I_{1}}{\tilde {I}_{1}^{a}}+\dfrac {B_{2}/(H_{2}+B_{2})}{\tilde {B}_{2}^{a}/(H_{2}+\tilde {B}_{2}^{a})}-\dfrac {S\tilde {I}_{1}^{a}B_{2}/(H_{2}+B_{2})}{\tilde {S}^{a}I_{1}\tilde {B}_{2}^{a}/(H_{2}+\tilde {B}_{2}^{a})}\right ]dx\\[3pt] &\quad+\int _{\Omega }\left [b_{1}\tilde {B}_{1}^{a}P\left (\!-\dfrac {l_{1}B_{1}}{\tilde {B}_{1}^{a}}+l_{1}+\dfrac {l_{3}\chi _{1}B_{1}}{\tilde {B}_{1}^{a}}\right )-l_{3}mP\right ]dx. \end{align*}

After a simple calculation, we have

\begin{align*} 1-\dfrac {B_{1}}{\tilde {B}_{1}^{a}}+\dfrac {I_{1}}{\tilde {I}_{1}^{a}}-\dfrac {I_{1}\tilde {B}_{1}^{a}}{\tilde {I}_{1}^{a}B_{1}}\leq \left (\dfrac {I_{1}}{\tilde {I}_{1}^{a}}-\ln \dfrac {I_{1}}{\tilde {I}_{1}^{a}} \right )- \left (\dfrac {B_{1}}{\tilde {B}_{1}^{a}}-\ln \dfrac {B_{1}}{\tilde {B}_{1}^{a}} \right ), \end{align*}
\begin{align*} 1-\dfrac {B_{2}}{\tilde {B}_{2}^{a}}+\dfrac {B_{1}}{\tilde {B}_{1}^{a}}-\dfrac {B_{1}\tilde {B}_{2}^{a}}{\tilde {B}_{1}^{a}B_{2}}\leq \left (\dfrac {B_{1}}{\tilde {B}_{1}^{a}}-\ln \dfrac {B_{1}}{\tilde {B}_{1}^{a}} \right )- \left (\dfrac {B_{2}}{\tilde {B}_{2}^{a}}-\ln \dfrac {B_{2}}{\tilde {B}_{2}^{a}} \right ). \end{align*}

By $(\mathbf{H}_{\mathbf{3}})$ , one gets

\begin{align*} \left (B_{1}-\tilde {B}_{1}^{a}\right )\left (\dfrac {h_{1}(B_{1})}{B_{1}}-\dfrac {h_{1}(\tilde {B}_{1}^{a})}{\tilde {B}_{1}^{a}}\right )\leq 0,\ \ \left (B_{2}-\tilde {B}_{2}^{a}\right )\left (\dfrac {h_{2}(B_{2})}{B_{2}}-\dfrac {h_{2}(\tilde {B}_{2}^{a})}{\tilde {B}_{2}^{a}}\right )\leq 0. \end{align*}

Since $ 1-x\leq -\ln x$ for $ x\gt 0$ , we have

\begin{align*}&2-\dfrac {\tilde {S}^{a}}{S}-\dfrac {I_{1}}{\tilde {I}_{1}^{a}}+\dfrac {B_{1}/(H_{1}+B_{1})}{\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}-\dfrac {S\tilde {I}_{1}^{a}B_{1}/(H_{1}+B_{1})}{\tilde {S}^{a}I_{1}\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}\\[3pt] &=\left (2-\dfrac {\tilde {S}^{a}}{S}-\dfrac {I_{1}}{\tilde {I}_{1}^{a}}-\dfrac {S\tilde {I}_{1}^{a}B_{1}/(H_{1}+B_{1})}{\tilde {S}^{a}I_{1}\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}+1-\dfrac {B_{1}\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}{\tilde {B}_{1}^{a}B_{1}/(H_{1}+B_{1})} +\dfrac {B_{1}}{\tilde {B}_{1}^{a}}\right )-\dfrac {H_{1}(B_{1}-\tilde {B}_{1}^{a})^{2}}{\tilde {B}_{1}^{a}(H_{1}+\tilde {B}_{1}^{a})(H_{1}+B_{1})}\\[3pt] & \leq 3-\dfrac {\tilde {S}^{a}}{S}-\dfrac {I_{1}}{\tilde {I}_{1}^{a}}-\dfrac {S\tilde {I}_{1}^{a}B_{1}/(H_{1}+B_{1})}{\tilde {S}^{a}I_{1}\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}-\dfrac {B_{1}\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}{\tilde {B}_{1}^{a}B_{1}/(H_{1}+B_{1})}+\dfrac {B_{1}}{\tilde {B}_{1}^{a}} \\ &=\left (\dfrac {B_{1}}{\tilde {B}_{1}^{a}}-\dfrac {I_{1}}{\tilde {I}_{1}^{a}}\right )+\left (1-\dfrac {\tilde {S}^{a}}{S}\right )+\left (1-\dfrac {S\tilde {I}_{1}^{a}B_{1}/(H_{1}+B_{1})}{\tilde {S}^{a}I_{1}\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}\right )+\left (1-\dfrac {B_{1}\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}{\tilde {B}_{1}^{a}B_{1}/(H_{1}+B_{1})}\right )\\[3pt] & \leq \left (\dfrac {B_{1}}{\tilde {B}_{1}^{a}}-\dfrac {I_{1}}{\tilde {I}_{1}^{a}}\right )-\ln \dfrac {\tilde {S}^{a}}{S}-\ln \dfrac {S\tilde {I}_{1}^{a}B_{1}/(H_{1}+B_{1})}{\tilde {S}^{a}I_{1}\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}-\ln \dfrac {B_{1}\tilde {B}_{1}^{a}/(H_{1}+\tilde {B}_{1}^{a})}{\tilde {B}_{1}^{a}B_{1}/(H_{1}+B_{1})}\\[3pt] &=\left (\dfrac {B_{1}}{\tilde {B}_{1}^{a}}-\ln \dfrac {B_{1}}{\tilde {B}_{1}^{a}} \right )- \left (\dfrac {I_{1}}{\tilde {I}_{1}^{a}}-\ln \dfrac {I_{1}}{\tilde {I}_{1}^{a}} \right ). \end{align*}

Similarly, we also have

\begin{align*} 2-\dfrac {\tilde {S}^{a}}{S}-\dfrac {I_{1}}{\tilde {I}_{1}^{a}}+\dfrac {B_{2}/(H_{2}+B_{2})}{\tilde {B}_{2}^{a}/(H_{2}+\tilde {B}_{2}^{a})}-\dfrac {S\tilde {I}_{1}^{a}B_{2}/(H_{2}+B_{2})}{\tilde {S}^{a}I_{1}\tilde {B}_{2}^{a}/(H_{2}+\tilde {B}_{2}^{a})}\leq \left (\dfrac {B_{2}}{\tilde {B}_{2}^{a}}-\ln \dfrac {B_{2}}{\tilde {B}_{2}^{a}} \right )- \left (\dfrac {I_{1}}{\tilde {I}_{1}^{a}}-\ln \dfrac {I_{1}}{\tilde {I}_{1}^{a}} \right ). \end{align*}

Let $ l_{1} = \left (\dfrac {\alpha _{1}\tilde {S}^{a}\tilde {B}_{1}^{a}}{\eta \tilde {I}_{1}^{a}(H_{1}+\tilde {B}_{1}^{a})}+\dfrac {\alpha _{2}\tilde {S}^{a}\tilde {B}_{2}^{a}}{\eta \tilde {I}_{1}^{a}(H_{2}+\tilde {B}_{2}^{a})}\right ),\ l_{2}=\dfrac {\alpha _{2}\tilde {S}^{a}\tilde {B}_{2}^{a}}{\delta _{1}\tilde {B}_{1}^{a}(H_{2}+\tilde {B}_{2}^{a})}, \ l_{3}=\dfrac {l_{1}}{\chi _{1}}$ , we derive

\begin{align*} & \dfrac{d\mathit{L}_{1}(t)}{dt}\leq -d_{S}\int _{\Omega }\tilde {S}^{a}\dfrac {|\nabla S|^{2}}{S^{2}}dx-d_{I}\int _{\Omega }\tilde {I}_{1}^{a}\dfrac {|\nabla I_{1}|^{2}}{I_{1}^{2}}dx-\int _{\Omega }\mu S \left (1-\dfrac {\tilde {S}^{a}}{S}\right )^{2}dx+\int _{\Omega }(l_{3}\alpha \eta -\mu )I_{2}dx\\[3pt] &\;\;+\int _{\Omega }\dfrac {\alpha _{1}\tilde {S}^{a}\tilde {B}_{1}^{a}}{H_{1}+\tilde {B}_{1}^{a}}\left [ \left (\dfrac {I_{1}}{\tilde {I}_{1}^{a}}-\ln \dfrac {I_{1}}{\tilde {I}_{1}^{a}} \right )- \left (\dfrac {B_{1}}{\tilde {B}_{1}^{a}}-\ln \dfrac {B_{1}}{\tilde {B}_{1}^{a}} \right )\right ]dx+\int _{\Omega }\dfrac {\alpha _{2}\tilde {S}^{a}\tilde {B}_{2}^{a}}{H_{2}+\tilde {B}_{2}^{a}}\left [ \left (\dfrac {I_{1}}{\tilde {I}_{1}^{a}}-\ln \dfrac {I_{1}}{\tilde {I}_{1}^{a}} \right )- \left (\dfrac {B_{1}}{\tilde {B}_{1}^{a}}-\ln \dfrac {B_{1}}{\tilde {B}_{1}^{a}} \right )\right ]dx\\[3pt] &\;\;+\int _{\Omega }\dfrac {\alpha _{2}\tilde {S}^{a}\tilde {B}_{2}^{a}}{H_{2}+\tilde {B}_{2}^{a}}\left [ \left (\dfrac {B_{1}}{\tilde {B}_{1}^{a}}-\ln \dfrac {B_{1}}{\tilde {B}_{1}^{a}} \right )- \left (\dfrac {B_{2}}{\tilde {B}_{2}^{a}}-\ln \dfrac {B_{2}}{\tilde {B}_{2}^{a}} \right )\right ]dx+\int _{\Omega }\dfrac {\alpha _{1}\tilde {S}^{a}\tilde {B}_{1}^{a}}{H_{1}+\tilde {B}_{1}^{a}}\left [ \left (\dfrac {B_{1}}{\tilde {B}_{1}^{a}}-\ln \dfrac {B_{1}}{\tilde {B}_{1}^{a}}\right )-\left (\dfrac {I_{1}}{\tilde {I}_{1}^{a}}-\ln \dfrac {I_{1}}{\tilde {I}_{1}^{a}} \right )\right ]dx\\[3pt] &\;\;+\int _{\Omega }\dfrac {\alpha _{2}\tilde {S}^{a}\tilde {B}_{2}^{a}}{H_{2}+\tilde {B}_{2}^{a}}\left [ \left (\dfrac {B_{2}}{\tilde {B}_{2}^{a}}-\ln \dfrac {B_{2}}{\tilde {B}_{2}^{a}}\right )-\left (\dfrac {I_{1}}{\tilde {I}_{1}^{a}}-\ln \dfrac {I_{1}}{\tilde {I}_{1}^{a}} \right )\right ]dx+\int _{\Omega }l_{1}P\left (b_{1}\tilde {B}_{1}^{a}-\dfrac {m}{\chi _{1}}\right )dx\\[3pt] &\;\;\leq-\int _{\Omega }\mu S \left (1-\dfrac {\tilde {S}^{a}}{S}\right )^{2}dx+\int _{\Omega }(l_{3}\alpha \eta -\mu )I_{2}dx+\int _{\Omega }l_{1}P\left (b_{1}\tilde {B}_{1}^{a}-\dfrac {m}{\chi _{1}}\right )dx. \end{align*}

Hence, based on $l_{1}\leq \mu \chi _{1}/\alpha \eta$ and $ \tilde {B}_{1}^{a}\leq m/\chi _{1}b_{1} = \tilde {B}_{1}^{b}$ , we have $ d\mathit{L}_{1}(t)/dt\leq 0$ . Moreover, $ d\mathit{L}_{1}(t)/dt=0$ if and only if $ S=\tilde {S}^{a},I_{1}=\tilde {I}_{1}^{a},I_{2}=0, B_{1}=\tilde {B}_{1}^{a}, B_{2}=\tilde {B}_{2}^{a}, P=0$ . The largest invariant set of $ \left \{(S,I_{1},I_{2},B_{1},B_{2},P)|\frac {d\mathit{L}_{1}(t)}{dt}=0\right \}$ is the singleton $\{\tilde {F}_{1}\}$ . Then, from LaSalle invariant principle [Reference Hale and Lunel12, Reference Henry15], $\tilde {F}_{1}$ is globally asymptotically stable.

4.6 Proof of Theorem3.11

Before proving Theorem3.11, we first give some preliminaries.

We define

\begin{align*} \mathbb{H}_{0}=\{\varphi \in H^{+}\,:\,\varphi _{1}(\!\cdot\!)\gt 0,\varphi _{2}(\!\cdot\!)\not \equiv 0,\varphi _{3}(\!\cdot\!)\not \equiv 0,\varphi _{4}(\!\cdot\!)\not \equiv 0,\varphi _{5}(\!\cdot\!)\not \equiv 0,\varphi _{6}(\!\cdot\!)\not \equiv 0 \}, \end{align*}

and

\begin{align*} \partial \mathbb{H}_{0}\,:=H^{+}\backslash \mathbb{H}_{0}=\{\varphi \in H^{+}\,:\,\varphi _{2}(\!\cdot\!)\equiv 0\text{ or } \varphi _{3}(\!\cdot\!)\equiv 0 \text{ or } \varphi _{4}(\!\cdot\!)\equiv 0 \text{ or } \varphi _{5}(\!\cdot\!)\equiv 0 \text{ or } \varphi _{6}(\!\cdot\!)\equiv 0 \}. \end{align*}

Define $\widetilde{F}_{\partial}\,:=\{\varphi \in \partial \mathbb{H}_{0}\,:\,\bar {J}_{1}(t)\varphi \in \partial \mathbb{H}_{0}\},$ for $t\ge 0$ , and $\widetilde{\omega }(\varphi )$ be the omega limit set of the orbit $\widetilde{\mathit{G}}^{+}\,:=\{\bar {J}_{1}(t)\varphi\,:\,t\geq 0\}$ . Similar to the proof of Theorem4.9, $ \mathbb{H}_{0}$ is the positive invariant set for solution semiflow $\bar {J}_{1}(t)$ of model (2.4), and we have the following claim.

Claim 1. If $\ \;\widetilde{\!\!\!\mathscr{R}}_{0}\gt 1$ , there exists $\tilde {\delta }\gt 0$ such that the semiflow $\bar {J}_{1}(t)$ of model (2.4) satisfies $\lim _{t\to \infty }\sup \|\bar {J}_{1}(t)\varphi -\tilde {F}_{0}\|\ge \tilde {\delta }$ for all $ \varphi \in \mathbb{H}_{0}$ .

Claim 2. If $\ \;\widetilde{\!\!\!\mathscr{R}}_{0}\gt 1$ , and $\tilde {B}_{1}^{a}\gt \tilde {B}_{1}^{b}$ holds, there exists $\,\hat{\!\delta }\gt 0$ such that the semiflow $\bar {J}_{1}(t)$ of model (2.4) satisfies $\lim _{t\to \infty }\sup \|\bar {J}_{1}(t)\varphi -\tilde {F}_{1}\|\ge \,\hat{\!\delta }$ for all $ \varphi \in \mathbb{H}_{0}$ .

Based on $\tilde {B}_{1}^{a}\gt \tilde {B}_{1}^{b}$ , we can choose a sufficiently small $\,\hat{\!\delta }\gt 0$ such that

(4.45) \begin{equation} \chi _{1}b_{1}(\tilde {B}_{1}^{a}-\,\hat{\!\delta })-m \gt \chi _{1}b_{1}\tilde {B}_{1}^{b}-m =0. \end{equation}

By way of contradiction, assuming that there exists a $\varphi _{0}\in \mathbb{H}_{0}$ satisfying $\lim \limits _{t\to \infty }\sup \|\bar {J}_{1}(t)\varphi _{0}-\tilde {F}_{1}\|\lt \,\hat{\!\delta }.$ This implies that there exists a $ \tilde {t}_{2} \gt 0$ satisfying $\tilde {B}_{1}^{a}-\,\hat{\!\delta }\lt B_{1}(x,t\,;\,\varphi _{0})$ . Hence, we have

\begin{align*} \dfrac {\partial P}{\partial t}\ge \chi _{1} b_{1}(\tilde {B}_{1}^{a}-\,\hat{\!\delta })P -mP,\quad x\in \Omega ,\ t\gt \tilde {t}_{2}. \end{align*}

Since $\varphi _{0}\in \mathbb{H}_{0}$ , and the results mentioned above, it follows that $ P(x,t)\gt 0$ for $x\in \bar {\Omega }$ and $ t\gt 0$ . This implies that there is a constant $p_{1}\gt 0$ such that $ P(x,t\,;\,\varphi _{0}) \ge p_{1}P^{0}(x)$ . By applying the standard comparison principle, one gets

\begin{align*} P(x,t) \ge p_{1}P^{0}(x) e^{\left (\chi _{1}b_{1}(\tilde {B}_{1}^{a}-\,\hat{\!\delta })-m\right )(t-\tilde {t}_{2})},\quad x\in \Omega ,\ t\gt \tilde {t}_{2}. \end{align*}

From (4.45), we obtain $\lim _{t\to \infty }P(x,t)=\infty$ , which contradicts the boundedness of $ P(x,t)$ . This establishes Claim 2.

Proof of Theorem 3.11 First, we prove that $\widetilde{\omega }(\varphi )=\{ \tilde {F}_{0}\}\cup \{ \tilde {F}_{1}\}, \varphi \in \widetilde{F}_{\partial }$ . For any $\varphi \in \widetilde{F}_{\partial }$ , we have $\bar {J}_{1}(t)\varphi \in \widetilde{F}_{\partial }, t\ge 0$ . Thus, $ I_{1}(\cdot ,t)\equiv 0$ or $ I_{2}(\cdot ,t)\equiv 0$ or $ B_{1}(\cdot ,t)\equiv 0$ or $ B_{2}(\cdot ,t)\equiv 0$ or $ P(\cdot ,t)\equiv 0$ for $ t\ge 0$ . In the case that $ I_{1}(\cdot ,t)\equiv 0$ , we can derive that $ B_{1}(\cdot ,t)\equiv 0$ , $ B_{2}(\cdot ,t)\equiv 0$ from the second equation of model (2.4). Then we have $ I_{2}(\cdot ,t)\equiv 0$ from the fourth equation of model (2.4). Besides, we obtain $ S(\cdot ,t)\to S^{*}(x)$ uniformly for $ x\in \bar {\Omega }$ , then from the sixth equation of model (2.4), one gets $\lim _{t\to \infty }P(\cdot ,t)=0$ . This shows $\widetilde{\omega }(\varphi )=\{\tilde {F}_{0}\}$ . In the case that $ B_{1}(\cdot ,t)\equiv 0$ , we can derive that $ I_{1}(\cdot ,t)\equiv 0$ , $ I_{2}(\cdot ,t)\equiv 0$ . Similarly, we also have $ B_{2}(\cdot ,t)\equiv 0$ , $ S(\cdot ,t)\to S^{*}(x)$ uniformly for $ x\in \bar {\Omega }$ and $\lim _{t\to \infty }P(\cdot ,t)=0$ . This shows $\widetilde{\omega }(\varphi )=\{\tilde {F}_{0}\}$ . In the case that $ B_{2}(\cdot ,t)\equiv 0$ , we can derive that $ B_{1}(\cdot ,t)\equiv 0$ from the fifth equation of model (2.4). Then we have $ I_{1}(\cdot ,t)\equiv 0$ , $ I_{2}(\cdot ,t)\equiv 0$ and $ S(\cdot ,t)\to S^{*}(x)$ uniformly for $ x\in \bar {\Omega }$ and $\lim _{t\to \infty }P(\cdot ,t)=0$ . This also shows $\widetilde{\omega }(\varphi )=\{\tilde {F}_{0}\}$ . In the case that $ P(\cdot ,t)\equiv 0$ , it implies that $ I_{2}(\cdot ,t)\equiv 0$ . Conversely, if $ I_{2}(\cdot ,t)\equiv 0$ , then we also have $ P(\cdot ,t)\equiv 0$ . Thus, model (2.4) becomes

(4.46) \begin{align} \left \{ \begin{aligned} &\dfrac {\partial S}{\partial t}= d_{S}\Delta S +\Lambda - \alpha _{1} f_{1}(B_{1})S -\alpha _{2} f_{2}(B_{2})S -\mu S,\quad x\in \Omega ,\ t\gt 0,\\[3pt] &\dfrac {\partial I_{1}}{\partial t}= d_{I}\Delta I_{1} + \alpha _{1}f_{1}(B_{1})S + \alpha _{2}f_{2}(B_{2})S -\mu I_{1},\quad x\in \Omega ,\ t\gt 0,\\[3pt] &\dfrac {\partial B_{1}}{\partial t}=h_{1}(B_{1}) +\eta I_{1} -\delta _{1}B_{1},\quad x\in \Omega ,\ t\gt 0,\\[3pt] &\dfrac {\partial B_{2}}{\partial t}=h_{2}(B_{2}) +\delta _{1}B_{1} - \delta _{2} B_{2},\quad x\in \Omega ,\ t\gt 0,\\[3pt] &\dfrac {\partial S}{\partial \nu } = \dfrac {\partial I_{1}}{\partial \nu }=0,\quad x\in \partial \Omega ,\ t\gt 0,\\[3pt] &S(x,0)=S^{0}(x),\ I_{1}(x,0)=I_{1}^{0}(x),\ B_{1}(x,0) = B_{1}^{0}(x),\ B_{2}(x,0) = B_{2}^{0}(x),\quad x\in \Omega . \end{aligned} \right . \end{align}

If $ \, \ \;\widetilde{\!\!\!\mathscr{R}}_{0}\gt 1$ , model (4.46) has a positive steady state $\bar {F}_{1}=(\tilde {S}^{a},\tilde {I}_{1}^{a},\tilde {B}_{1}^{a},\tilde {B}_{2}^{a})$ . We choose the Lyapunov function

\begin{align*} \begin{aligned} \mathit{L}_{2}(t)=&\int _{\Omega }\tilde {S}^{a}Y\left (\dfrac {S}{\tilde {S}^{a}}\right )dx+\int _{\Omega }\tilde {I}_{1}^{a}Y\left (\dfrac {I_{1}}{\tilde {I}_{1}^{a}}\right )dx+\left (\dfrac {\alpha _{1}\tilde {S}^{a}\tilde {B}_{1}^{a}}{\eta \tilde {I}_{1}^{a}(H_{1}+\tilde {B}_{1}^{a})}+\dfrac {\alpha _{2}\tilde {S}^{a}\tilde {B}_{2}^{a}}{\eta \tilde {I}_{1}^{a}(H_{2}+\tilde {B}_{2}^{a})}\right )\int _{\Omega }\tilde {B}_{1}^{a}Y\left (\dfrac {B_{1}}{\tilde {B}_{1}^{a}}\right )dx \\[3pt] &+ \dfrac {\alpha _{2}\tilde {S}^{a}\tilde {B}_{2}^{a}}{\delta _{1}\tilde {B}_{1}^{a}(H_{2}+\tilde {B}_{2}^{a})}\int _{\Omega }\tilde {B}_{2}^{a}Y\left (\dfrac {B_{2}}{\tilde {B}_{2}^{a}}\right )dx. \end{aligned} \end{align*}

By applying the similar proof in Theorem3.10, we have

\begin{align*} \dfrac {d\mathit{L}_{2}(t)}{dt}\leq -d_{S}\int _{\Omega }\tilde {S}^{a}\dfrac {|\nabla S|^{2}}{S^{2}}dx-d_{I}\int _{\Omega }\tilde {I}_{1}^{a}\dfrac {|\nabla I_{1}|^{2}}{I_{1}^{2}}dx -\int _{\Omega }\mu S \left (1-\dfrac {\tilde {S}^{a}}{S}\right )^{2}dx. \end{align*}

Apparently, we can obtain $d\mathit{L}_{2}(t)/dt\leq 0$ . Moreover, $ d\mathit{L}_{2}(t)/dt=0$ if and only if $ S=\tilde {S}^{a}, I_{1}=\tilde {I}_{1}^{a}, B_{1}=\tilde {B}_{1}^{a}, B_{2}=\tilde {B}_{2}^{a}$ . The largest invariant set of $ \left \{(S,I_{1},B_{1},B_{2})|\frac {d\mathit{L}_{2}(t)}{dt}=0\right \}$ is the singleton $\{\bar {F}_{1}\}$ . Then, from LaSalle invariant principle [Reference Hale and Lunel12], $\bar {F}_{1}$ is globally asymptotically stable. Hence, we have $ \lim _{t\to \infty }(S(x,t),I_{1}(x,t),B_{1}(x,t),B_{2}(x,t))=(\tilde {S}^{a},\tilde {I}_{1}^{a},\tilde {B}_{1}^{a},\tilde {B}_{2}^{a})$ . This also shows $\widetilde{\omega }(\varphi )=\{\tilde {F}_{1}\}$ . We conclude that $\widetilde{\omega }(\varphi )=\{ \tilde {F}_{0}\}\cup \{ \tilde {F}_{1}\}$ for any $ \varphi \in \widetilde{F}_{\partial }$ . Define a continuous function

\begin{align*} \tilde {\tau }(\varphi )=\min \left \{ \min _{x\in \bar {\Omega }} \varphi _{2}(x), \min _{x\in \bar {\Omega }} \varphi _{3}(x), \min _{x\in \bar {\Omega }} \varphi _{4}(x),\min _{x\in \bar {\Omega }} \varphi _{5}(x),\min _{x\in \bar {\Omega }} \varphi _{6}(x) \right \},\ \varphi \in H^{+}. \end{align*}

Apparently, $\tilde {\tau }^{-1}(0,\infty )\subseteq \mathbb{H}_{0}$ . The function $\tilde {\tau }$ is a generalised distance function for the semiflow $\bar {J}_{1}(t)$ . By the above discussions, we know that the sigleton $\widetilde{\omega }(\varphi )=\{ \tilde {F}_{0}\}\cup \{ \tilde {F}_{1}\}$ is an isolated invariant set for $\bar {J}_{1}(t)$ in $ H^{+}$ , then $ W^{s}(\{\tilde {F}_{0}\})\cap \mathbb{H}_{0}=\emptyset$ , $ W^{s}(\{\tilde {F}_{1}\})\cap \mathbb{H}_{0}=\emptyset$ , where $ W^{s}(\{\tilde {F}_{0}\})$ and $ W^{s}(\{\tilde {F}_{1}\})$ represent the stable subset of $ \{\tilde {F}_{0}\}$ and $ \{\tilde {F}_{1}\}$ , respectively. Besides, no subset of $ \{ \tilde {F}_{0}\}\cup \{ \tilde {F}_{1}\}$ forms a cycle in $ \partial \mathbb{H}_{0}$ . By [Reference Smith and Zhao40, Theorem 3], there exists a $\tilde {\vartheta }_{1}\gt 0$ satisfying $\lim _{t\to +\infty }\inf \tilde {\tau }(\bar {J}_{1}(t)\varphi )\ge \tilde {\vartheta }_{1},\ \forall \varphi \in \mathbb{H}_{0}$ . Recall that $\vartheta _{2}$ in proof of Theorem4.9. Let $\tilde {\vartheta }=\min \{\tilde {\vartheta }_{1},\vartheta _{2}\}$ . The proof is complete.

5 Concluding remarks

Cholera is a waterborne infectious disease that can easily lead to large-scale outbreaks in areas with poor sanitation, causing persistent distress and threats. Consequently, researchers actively seek methods and measures to control cholera outbreaks. Jensen et al. discovered that under biologically plausible conditions, bacteriophages can mitigate cholera epidemics [Reference Jensen, Faruque, Mekalanos and Levin19]. This is attributed to the cholera-specific lytic bacteriophages potentially reducing cholera prevalence by eliminating bacteria present in reservoirs and infected human hosts. Additionally, recent research findings in [Reference Hartley, Morris and Smith14] indicated that V. cholerae exhibits a hyperinfectious state upon entering the gastrointestinal tract, diminishing to a lower-infectious state within hours. The different infectivity states of cholera vibrio influence the transmission dynamics of cholera outbreaks differently.

This paper incorporated the interaction between bacteriophages and HI vibrios and LI vibrios, as well as the intrinsic growth rate of V. cholerae, and proposed a degenerate reaction-diffusion cholera model. We divided the infected human hosts into two parts for study: one part consists of human hosts infected only with V. cholerae, denoted as $I_{1}$ , while the other part consists of human hosts that are simultaneously infected with V. cholerae and bacteriophages, indicating the parasitism of bacteriophages within the host cells (bacteria), denoted as $I_{2}$ . We also introduce the interaction between HI vibrios and LI vibrios and bacteriophages in this process.

In this work, we originally established the existence and uniform boundedness of the solution, and then derived the well-posedness of the solutions. In a spatially heterogeneous case, the basic reproduction number $\mathscr{R}_{0}$ is defined as the spectral radius of the sum of two linear operators associated with HI vibrios infection and LI vibrios infection. Generally speaking, it is very challenging to discuss the threshold-type results in the case of multi-class steady states. Fortunately, in this paper, we derived the existence and stability analysis of multi-class steady states for some special cases. We showed the existence of phage-free steady state in a heterogeneous environment. An appropriate Lyapunov function was constructed to discuss the global stability of the phage-free steady state in a homogeneous environment.

In considering the constraints established by our mathematical model, which mandates that disease transmission occurs through the consumption of bacteria rather than through human-to-human contact, there is potential for future research to explore the incorporation of direct interpersonal transmission pathways to enhance this approach. Furthermore, the examination of global stability of the phage-free steady state in a homogeneous environment requires two additional conditions. Future studies could seek to investigate the global stability of the phage-free steady state independently of additional conditions. The global stability of the phage-present steady state of model (2.4) also poses some challenges [Reference Wang, Wang and Fan54]. Moreover, the existence and uniform persistence of the phage-present steady state of model (2.1) are difficult to obtain due to the spatial heterogeneity and other mathematical difficulties. The situation in a heterogeneous environment presents several interesting open problems. We consider these challenges for further investigation.

Acknowledgements

We thank the anonymous referees and the handling editor for their valuable comments, which have led to a substantial improvement in the revision.

Financial support

Wei Wang gratefully acknowledges support from Shandong Provincial Natural Science Foundation of China (ZR2024QA006) and the National Natural Science Foundation of China (No. 12271308, 11901360). Hao Wang gratefully acknowledges support from the Natural Sciences and Engineering Research Council of Canada (Discovery Grant RGPIN-2020-03911 and NSERC Accelerator Grant RGPAS-2020-00090) and the Canada Research Chairs program (Tier 1 Canada Research Chair Award).

Competing interests

The authors declare no competing interests.

References

Alikakos, N. (1979An application of the invariance principle to reaction-diffusion equations. J. Differ. Equ. 33 (2), 201225.CrossRefGoogle Scholar
Allen, L., Bolker, B. & Lou, Y., et al. (2008) Asymptotic profiles of the steady states for an SIS epidemic reaction-diffusion model. Discrete Contin. Dyn. Syst. 21, 120.CrossRefGoogle Scholar
Andrews, J. & Basu, S. (2011) Transmission dynamics and control of cholera in Haiti: An epidemic model. Lancet 377 (9773), 12481255.CrossRefGoogle ScholarPubMed
Botelho, C., Kong, J. D., Lucien, M. A. B., Shuai, Z., & Wang, H. (2021) A mathematical model for Vibrio-phage interactions. Math. Biosci. Eng. 18 (3), 26882712.CrossRefGoogle ScholarPubMed
Cai, L., Modnak, C. & Wang, J. (2017) An age-structured model for cholera control with vaccination. Appl. Math. Comput. 299, 127140.Google Scholar
Capasso, V. & Paveri-Fontana, S. (1979) A mathematical model for the 1973 cholera epidemic in the European Mediterranean region. Rev. éPidéMiol. Santé Publique 27, 121132.Google ScholarPubMed
Codeço, C. (2001) Endemic and epidemic dynamics of cholera: The role of the aquatic reservoir. BMC Infect. Dis. 1, 114.CrossRefGoogle ScholarPubMed
Colwell, R. & Huq, A. (1994) Environmental reservoir of Vibrio cholerae, the causative agent of cholera. Ann. Ny. Acad. Sci. 740 (1), 4454.CrossRefGoogle ScholarPubMed
Cui, R., Lam, K. & Lou, Y. (2017) Dynamics and asymptotic profiles of steady states of an epidemic model in advective environments. J. Differ. Equ. 263 (4), 23432373.CrossRefGoogle Scholar
Faruque, S. M., Naser, I. B., Islam, M. J., Faruque, A. S. G., Ghosh, A. N., Nair, G. B., Sack, D. A., & Mekalanos, J. J. (2005) Seasonal epidemics of cholera inversely correlate with the prevalence of environmental cholera phages. Proc. Natl. Acad. Sci. USA 102 (5), 17021707.CrossRefGoogle ScholarPubMed
Garay, E., Arnau, A. & Amaro, C. (1985) Incidence of Vibrio cholerae and related vibrios in a coastal lagoon and seawater influenced by lake discharges along an annual cycle. Appl. Environ. Microbiol. 50 (2), 426430.CrossRefGoogle Scholar
Hale, J. & Lunel, S. (1993) Introduction to Functional Differential Equations, New York, Springer-Verlag.CrossRefGoogle Scholar
Hale, J. & Waltman, P. (1989) Persistence in infinite-dimensional systems. SIAM J. Math. Anal. 20 (2), 388395.CrossRefGoogle Scholar
Hartley, D., Morris, J. Jr, & Smith, D. (2006) Hyperinfectivity: A critical element in the ability of V. cholerae to cause epidemics? PLoS Med. 3 (1), e7.CrossRefGoogle ScholarPubMed
Henry, D. (1981) Geometric Theory of Semilinear Parabolic Equations, Berlin, Springer-Verlag.CrossRefGoogle Scholar
Hove-Musekwa, S. D., Nyabadza, F., Chiyaka, C., Das, P., Tripathi, A., & Mukandavire, Z. (2011) Modelling and analysis of the effects of malnutrition in the spread of cholera. Math. Comput. Model. 53 (9–10), 15831595.CrossRefGoogle Scholar
Hsu, S., Jiang, J. & Wang, F. (2010) On a system of reaction-diffusion equations arising from competition with internal storage in an unstirred chemostat. J. Differ. Equ. 248 (10), 24702496.CrossRefGoogle Scholar
Hu, Z., Wang, S. & Nie, L. (2023) Dynamics of a partially degenerate reaction-diffusion cholera model with horizontal transmission and phage-bacteria interaction. Electron. J. Differ. Equ. 2023 (8), 138.Google Scholar
Jensen, M. A., Faruque, S. M., Mekalanos, J. J., & Levin, B. R. (2006) Modeling the role of bacteriophage in the control of cholera outbreaks. Proc. Natl. Acad. Sci. USA 103 (12), 46524657.CrossRefGoogle ScholarPubMed
Joh, R. I., Wang, H., Weiss, H., & Weitz, J. S. (2009) Dynamics of indirectly transmitted infectious diseases with immunological threshold. Bull. Math. Biol. 71, 845862.CrossRefGoogle ScholarPubMed
Kaper, J. B., Morris, J. G. Jr, & Levine, M. M. (1995) Cholera. Clin. Microbiol. Rev. 8 (1), 4886.CrossRefGoogle ScholarPubMed
Kierek, K. & Watnick, P. (2003) Environmental determinants of Vibrio cholerae biofilm development. Appl. Environ. Microbiol. 69 (9), 50795088.CrossRefGoogle ScholarPubMed
Kong, J., Davis, W. & Wang, H. (2014) Dynamics of a cholera transmission model with immunological threshold and natural phage control in reservoir. Bull. Math. Biol. 76 (8), 20252051.CrossRefGoogle ScholarPubMed
Lou, Y. & Zhao, X. (2011) A reaction-diffusion malaria model with incubation period in the vector population. J. Math. Biol. 62 (4), 543568.CrossRefGoogle ScholarPubMed
Luo, J., Wang, J. & Wang, H. (2017) Seasonal forcing and expoential threshold incidence in cholera dynamics. Discrete & Continuous Dynamical Systems-Series B 22 (6), 22612290.CrossRefGoogle Scholar
Magal, P. & Zhao, X. (2005) Global attractors and steady states for uniformly persistent dynamical systems. SIAM J. Math. Anal. 37 (1), 251275.CrossRefGoogle Scholar
Mari, L., Bertuzzo, E., Righetto, L., Casagrandi, R., Gatto, M., Rodriguez-Iturbe, I., & Rinaldo, A. (2012) Modelling cholera epidemics: The role of waterways, human mobility and sanitation. J. R. Soc. Interface 9 (67), 376388.CrossRefGoogle ScholarPubMed
Martin, R. & Smith, H. (1990) Abstract functional differential equations and reaction-diffusion systems. Trans. Amer. Math. Soc. 321, 144.Google Scholar
Misra, A. & Gupta, A. (2016) A reaction-diffusion model for the control of cholera epidemic. J. Biol. Syst. 24 (04), 431456.CrossRefGoogle Scholar
Nelson, E. J., Chowdhury, A., Flynn, J., Schild, S., Bourassa, L., Shao, Y., LaRocque, R. C., Calderwood, S. B., Qadri, F., Camilli, A., Ausubel, F. M. (2008) Transmission of Vibrio cholerae is antagonized by lytic phage and entry into the aquatic environment. PLoS Pathog. 4 (10), e1000187.CrossRefGoogle ScholarPubMed
Nelson, E. J., Harris, J. B., Glenn Morris, J. Jr, Calderwood, S. B., & Camilli, A. (2009) Cholera transmission: The host, pathogen and bacteriophage dynamic. Nat. Rev. Microbiol. 7 (10), 693702.CrossRefGoogle ScholarPubMed
Netter, Z., Dunham, D. & Seed, K. (2023) Adaptation to bile and anaerobicity limits Vibrio cholerae phage adsorption. mBio. 14 (6), e0198523.CrossRefGoogle ScholarPubMed
Perez-Rosas, N. & Hazen, T. (1989) In situ survival of Vibrio cholerae and Escherichia coli in a tropical rain forest watershed. Appl. Environ. Microbiol 55 (2), 495499.CrossRefGoogle Scholar
Protter, M. & Weinberger, H. (1984) Maximum Principles in Differential Equations, New York, Springer.CrossRefGoogle Scholar
Safi, M., Melesse, D. & Gumel, A. (2013) Dynamics analysis of a multi-strain cholera model with an imperfect vaccine. Bull. Math. Biol. 75, 11041137.CrossRefGoogle ScholarPubMed
Sell, G. & You, Y. (2002) Dynamics of Evolutionary Equations, New York, Springer-Verlag.CrossRefGoogle Scholar
Shu, H., Ma, Z. & Wang, X. (2021) Threshold dynamics of a nonlocal and delayed cholera model in a spatially heterogeneous environment. J. Math. Biol. 83 (4), 41.CrossRefGoogle Scholar
Shu, H., Ma, Z. & Wang, X., et al. (2020) Viral diffusion and cell-to-cell transmission: Mathematical analysis and simulation study. J. Math. Pures Appl. 137, 290313.CrossRefGoogle Scholar
Shuai, Z., Tien, J. & van den Driessche, P. (2012) Cholera models with hyperinfectivity and temporary immunity. Bull. Math. Biol. 74 (10), 24232445.CrossRefGoogle ScholarPubMed
Smith, H. & Zhao, X. (2001) Robust persistence for semidynamical systems. Nonlinear Anal. 47 (9), 61696179.CrossRefGoogle Scholar
Thieme, H. (1992) Convergence results and Poincare-Bendixson trichotomy for asymptotically autonomous differentialequations. J. Math. Biol. 30 (7), 755763.CrossRefGoogle Scholar
Thieme, H. (2009) Spectral bound and reproduction number for intinite-dimensional population structure and time heterogeneity. SIAM J. Appl. Math. 70 (1), 188211.CrossRefGoogle Scholar
Tien, J. & Earn, D. (2010) Multiple transmission pathways and disease dynamics in a waterborne pathogen model. Bull. Math. Biol. 72, 15061533.CrossRefGoogle Scholar
Wang, F. (2010) A system of partial differential equations modeling the competition for two complementary resources in flowing habitats. J. Differ. Equ. 249 (11), 28662888.CrossRefGoogle Scholar
Wang, J. & Wang, J. (2021) Analysis of a reaction-diffusion cholera model with distinct dispersal rates in the human population. J. Dyn. Differ. Equ. 33 (1), 549575.CrossRefGoogle Scholar
Wang, J. & Wu, X. (2023) Dynamics and profiles of a diffusive cholera model with bacterial hyperinfectivity and distinct dispersal rates. J. Dyn. Differ. Equ. 35 (2), 12051241.CrossRefGoogle Scholar
Wang, W. & Feng, Z. (2023) Global dynamics of a diffusive viral infection model with spatial heterogeneity. Nonlinear Anal., Real World Appl. 72, 103763.CrossRefGoogle Scholar
Wang, W., Wang, X. & Wang, H. (2024) Spatial dynamics of a generalized cholera model with nonlocal time delay in a heterogeneous environment. J. Differ. Equ. 405, 103150.CrossRefGoogle Scholar
Wang, W., Wu, G., Wang, X., & Feng, Z. (2023) Dynamics of a reaction-advection-diffusion model for cholera transmission with human behavior change. J. Differ. Equ. 373, 176215.CrossRefGoogle Scholar
Wang, W. & Zhao, X. (2012) Basic reproduction number for reaction-diffusion epidemic models. SIAM J. Appl. Dyn. Syst. 11 (4), 16521673.CrossRefGoogle Scholar
Wang, X. & Wang, F. (2019) Impact of bacterial hyperinfectivity on cholera epidemics in a spatially heterogeneous environment. J. Math. Anal. Appl. 480 (2), 123407.CrossRefGoogle Scholar
Wang, X., Zhao, X. & Wang, J. (2018) A cholera epidemic model in a spatiotemporally heterogeneous environment. J. Math. Anal. Appl. 468 (2), 893912.CrossRefGoogle Scholar
Wang, W., Wang, X. & Fan, X. (2025) Threshold dynamics of a reaction-advection-diffusion waterborne disease model with seasonality and human behavior change. Int. J. Biomath. 18, 2350106.CrossRefGoogle Scholar
Wang, W., Wang, X. & Fan, X. (2025) On the global attractivity of a diffusive viral infection model with spatial heterogeneity. Math. Method Appl. Sci., 15. DOI: 10.1002/mma.70040.Google Scholar
Wang, Z. & Wang, H. (2024) Nontrivial traveling waves of phage-bacteria models in different media types. SIAM J. Appl. Math. 84 (2), 556580.CrossRefGoogle Scholar
Webb, G. (1981) A reaction-diffusion model for a deterministic diffusive models. J. Math. Anal. Appl. 84, 150161.CrossRefGoogle Scholar
Webb, G. (1987) An operator-theoretic formulation of asynchronous exponential growth. Trans. Amer. Math. Soc. 303 (2), 751763.CrossRefGoogle Scholar
Wu, Y. & Zou, X. (2018) Dynamics and profiles of a diffusive host-pathogen system with distinct dispersal rates. J. Differ. Equ. 264 (8), 49895024.CrossRefGoogle Scholar
Yamazaki, K., Yang, C. & Wang, J. (2021) A partially diffusive cholera model based on a general second-order differential operator. J. Math. Anal. Appl. 501 (2), 125181.CrossRefGoogle Scholar
Zhao, X. (2003) Dynamical systems in population biology. In CMS Books in Mathematics (Ouvrages De Mathématiquesde LA SMC), Vol. 16, Springer-Verlag, New York.Google Scholar
Figure 0

Figure 1. Schematic diagram of model (2.1). The green solid line represents the recruitment and mortality rates of human hosts and V. cholerae, and the blue solid line denotes the direct development of cholera. The purple dashed line represents the infection process and the interaction between phages and V. cholerae.