
Citation: Hilla Behar, Alexandra Agranovich, Yoram Louzoun. Diffusion rate determines balance between extinction and proliferationin birth-death processes[J]. Mathematical Biosciences and Engineering, 2013, 10(3): 523-550. doi: 10.3934/mbe.2013.10.523
[1] | Zhisheng Shuai, P. van den Driessche . Impact of heterogeneity on the dynamics of an SEIR epidemic model. Mathematical Biosciences and Engineering, 2012, 9(2): 393-411. doi: 10.3934/mbe.2012.9.393 |
[2] | Junyuan Yang, Rui Xu, Xiaofeng Luo . Dynamical analysis of an age-structured multi-group SIVS epidemic model. Mathematical Biosciences and Engineering, 2019, 16(2): 636-666. doi: 10.3934/mbe.2019031 |
[3] | Cheng-Cheng Zhu, Jiang Zhu . Spread trend of COVID-19 epidemic outbreak in China: using exponential attractor method in a spatial heterogeneous SEIQR model. Mathematical Biosciences and Engineering, 2020, 17(4): 3062-3087. doi: 10.3934/mbe.2020174 |
[4] | Jinhu Xu, Yicang Zhou . Global stability of a multi-group model with vaccination age, distributed delay and random perturbation. Mathematical Biosciences and Engineering, 2015, 12(5): 1083-1106. doi: 10.3934/mbe.2015.12.1083 |
[5] | Qiuyi Su, Jianhong Wu . Impact of variability of reproductive ageing and rate on childhood infectious disease prevention and control: insights from stage-structured population models. Mathematical Biosciences and Engineering, 2020, 17(6): 7671-7691. doi: 10.3934/mbe.2020390 |
[6] | Cheng-Cheng Zhu, Jiang Zhu, Xiao-Lan Liu . Influence of spatial heterogeneous environment on long-term dynamics of a reaction-diffusion SVIR epidemic model with relaps. Mathematical Biosciences and Engineering, 2019, 16(5): 5897-5922. doi: 10.3934/mbe.2019295 |
[7] | Abdennasser Chekroun, Mohammed Nor Frioui, Toshikazu Kuniya, Tarik Mohammed Touaoula . Global stability of an age-structured epidemic model with general Lyapunov functional. Mathematical Biosciences and Engineering, 2019, 16(3): 1525-1553. doi: 10.3934/mbe.2019073 |
[8] | Gang Huang, Edoardo Beretta, Yasuhiro Takeuchi . Global stability for epidemic model with constant latency and infectious periods. Mathematical Biosciences and Engineering, 2012, 9(2): 297-312. doi: 10.3934/mbe.2012.9.297 |
[9] | Xichao Duan, Sanling Yuan, Kaifa Wang . Dynamics of a diffusive age-structured HBV model with saturating incidence. Mathematical Biosciences and Engineering, 2016, 13(5): 935-968. doi: 10.3934/mbe.2016024 |
[10] | Jinliang Wang, Hongying Shu . Global analysis on a class of multi-group SEIR model with latency and relapse. Mathematical Biosciences and Engineering, 2016, 13(1): 209-225. doi: 10.3934/mbe.2016.13.209 |
Since the pioneering work of Kermack and McKendrick [1], many mathematical models have been proposed attempting to gain a better understanding of disease transmission, especially for the control strategy and dynamical behavior of infectious diseases [2,3,4,5,6,7,8]. Simple models with assumption that individuals are well mixed, which implies each individual has the same probability to be infected, are beneficial in that one can obtain analytical results easily but may be lack of realism. Epidemic models with population structures, like age, sex and patch (such as communities, cities, or counties), may be a more realistic way to describe complex disease dynamics. As a matter of fact, the total population should be classified into different groups and the vital epidemic parameters should vary among different population groups. In addition, at different age stages, the effects of infectious transmission are various, which is another important and key factor that needs necessarily to be included in modeling this infectious transmission process. Thus, considering multi-group and age structure in epidemic models is very necessary and reasonable. Some recent developments on the transmission dynamics of multi-group and age structured epidemic models have been discussed in [8,9,10,11,12,13].
Since that the population distribute heterogeneously in different spatial location in the real life and they will move or diffuse for many reasons, in epidemiology, there is increasing evidence that environmental heterogeneity and individual motility have significant impact on the spread of infectious diseases [14,15]. In recent years, global behavior of spatial diffusion systems, which are suitable for diseases such as the rabies and the Black Death, has been attracted extensive attention of researchers and has been one of the hot topics [16,17,18,19,20,21,22,23,24,25]. Among these works, few take age structure or multi-group into consideration. Yang et al. [24] proposed a novel model incorporated with both age-since-infection and spacial diffusion of brucellosis infection, and the basic reproduction number and global behaviors of this system were completely investigated. Fitzgibbon et al. [19] considered a diffusive epidemic model with age structure where the disease spreads between vector and host populations. Then, the existence of solutions of the model was studied based on semigroup theory and the asymptotic behavior of the solution was analyzed. Luo et al. [21] incorporated spatial heterogeneity in $ n $-group reaction-diffusion SIR model with nonlinear incidence rate to investigate the global dynamics of the disease-free and endemic steady states for this model. Zhao et al. [25] modeled host heterogeneity by introducing multi-group structure in a time delay SIR epidemic model and showed that basic reproduction number determines the existence of traveling waves of this system. To determine how age structure, multi-group population and diffusion of individuals affect the consequences of epidemiological processes, Ducrot et al. [18] formulated a multi-group age-structured epidemic model with the classical Fickian diffusion and studied the existence of travelling wave solutions for this model.
To the best of our knowledge, epidemic models established by researchers except for [18] only include one or two characteristics of multi-group, age structure and spatial diffusion. All the three characteristics are incorporated into epidemic model in [18], however, this model does not include the class of latent individuals. For some epidemic diseases like malaria, HIV/AIDS and West Nile virus, latent individuals may take days, months, or even years to become infectious. Moreover, the travel of latent individuals showing no symptoms can spread the disease geographically which makes disease harder to control. Motivated by the above discussion, in this paper, we investigate a diffusive version of multi-group epidemic system with age structure which is generalization of the model studied in [26] for the first time to allow for individuals moving around on the spatial habitat $ x\in\Omega\subset\mathbb{R}^n $ with smooth boundary $ \partial\Omega $. The organization of this paper is as follows. Firstly, we present our model in the next section. In section 3, some preliminaries including the positivity, boundedness, existence and uniqueness of solution, and the existence of compact global attractor of the associated solution semiflow, are established. In section 4, the sufficient conditions on the existence and global stability of disease-free and endemic steady states are stated and proved. In section 5, we conduct numerical simulations to illustrate the validity of our theoretical results. In section 6, a brief conclusion is given.
In 2015, Liu et al. [26] introduced age-of-latent and age-of-relapse into epidemic model which is appropriate for diseases such as tuberculosis and herpes virus infection. For these diseases, latent individuals may take days, months, or even years to become infectious and the treatment efficacy may decline with time for recovered individuals then cause recurrence of disease. In order to study the global dynamics o for these diseases, they formulated the following SEIR epidemic system with continuous age-dependent latency and relapse
$ {dS(t)dt=Λ−μS(t)−βS(t)I(t),∂e(t,a)∂t+∂e(t,a)∂a=−σ(a)e(t,a)−(μ+δ1)e(t,a),dI(t)dt=∫∞0σ(a)e(t,a)da−(μ+δ2+c)I(t)+∫∞0γ(b)r(t,b)db,∂r(t,b)∂t+∂r(t,b)∂b=−γ(b)r(t,b)−μr(t,b),e(t,0)=βS(t)I(t),r(t,0)=cI(t), $
|
(2.1) |
for $ t\geq0 $ and with initial conditions
$ {S(0)=S0>0,e(0,a)=e0(a)∈L1+(0,+∞),I(0)=I0>0,r(0,b)=r0(b)∈L1+(0,+∞), $
|
for $ a, b\geq0 $, where $ L_+^1(0, +\infty) $ is the space of functions on $ (0, +\infty) $ that are nonnegative and Lebesgue integrable. At time $ t $, the densities of susceptible individuals, latent individuals with latent age $ a $, infectious individuals and removed individuals with relapse age $ b $ are denoted by $ S(t) $, $ e(t, a) $, $ I(t) $, $ r(t, b) $, respectively. $ \sigma(a) $ and $ \gamma(b) $ denote the conversional rate from the latent class and the relapse rate in the removed class, which depend on age $ a $ and age $ b $, respectively. Furthermore, $ \beta $ is the transmission rate of the disease between susceptible and infectious individuals, $ \Lambda $ is the density of the recruitment into the susceptible class (including the births and immigration), $ \mu $ is the natural death rate of all individuals, $ \delta_1 $ and $ \delta_2 $ are the additional death rate induced by the infectious diseases, and $ c $ is the recovery rate from the infectious class. All parameters are assumed to be positive.
It is clear that the variations of different epidemic parameters between or within different groups can be well realized according to the description of multi-group epidemic models. Hence, Liu and Feng [27] extended model (2.1) to the situation in which the population is divided into $ n $ groups according to different contact patterns and derived the following multi-group SEIR epidemic model
$ {dSk(t)dt=Λk−μkSk(t)−n∑j=1βkjSk(t)Ij(t),∂ek(t,a)∂t+∂ek(t,a)∂a=−σk(a)ek(t,a)−(μk+δ1k)ek(t,a),dIk(t)dt=∫∞0σk(a)ek(t,a)da−(μk+δ2k+ck)Ik(t)+∫∞0γk(b)rk(t,b)db,∂rk(t,b)∂t+∂rk(t,b)∂b=−γk(b)rk(t,b)−μkrk(t,b),ek(t,0)=n∑j=1βkjSk(t)Ij(t),rk(t,0)=ckIk(t), $
|
(2.2) |
for $ t\geq0 $ and with initial conditions
$ {Sk(0)=S0k>0,ek(0,a)=e0k(a)∈L1+(0,+∞),Ik(0)=I0k>0,rk(0,b)=r0k(b)∈L1+(0,+∞), $
|
for $ a, b > 0 $. $ \Lambda_k $, $ \mu_k $ and $ c_k $ denote the recruitment rate of the susceptible class, the per-capita natural death rate and the recovery rate from the infectious class in group $ k $, respectively. $ \beta_{kj} $ denotes the transmission rate of the disease between susceptible individuals in group $ k $ and infectious individuals in group $ j $. $ \delta_{1k} $ and $ \delta_{2k} $ denote the additional death rates of exposed and infectious individuals induced by the infectious diseases in group $ k $, respectively. $ \sigma_k(a) $ denotes the conversional rate from the latent class in group $ k $, which depends on age $ a $ and $ \gamma_k(b) $ denotes the relapse rate from the removed class into the infectious class in group $ k $, which depends on age $ b $.
Spatial diffusion is an intrinsic characteristic for investigating the roles of spatial heterogeneity on diseases mechanisms and transmission routes and can lead to rich dynamics. Based on this fact, we generalize (2.2) by taking account of the case that individuals move or diffuse around on the spatial habitat $ x\in\Omega\subset\mathbb{R}^n $ with smooth boundary $ \partial\Omega $. Let $ S_k(t, x) $ and $ I_k(t, x) $ be the densities of susceptible individuals and infectious individuals at time $ t $ and location $ x\in\Omega $ in group $ k $, respectively, where the habitat $ \Omega $ is bounded and connected. And let $ e_k(t, a, x) $ and $ r_k(t, b, x) $ denote the densities of individuals in the latent class with age $ a $ and the removed class with age $ b $ at time $ t $ and location $ x $ in group $ k $, respectively. Hence, the $ n $-group diffusive SEIR epidemic model with age-dependent latent and relapse has the following form
$ {∂Sk(t,x)∂t=d1k△Sk(t,x)+Λk−μkSk(t,x)−n∑j=1βkjSk(t,x)Ij(t,x),∂ek(t,a,x)∂t+∂ek(t,a,x)∂a=d2k△ek(t,a,x)−σk(a)ek(t,a,x)−(μk+δ1k)×ek(t,a,x),∂Ik(t,x)∂t=d3k△Ik(t,x)+∫∞0σk(a)ek(t,a,x)da−(μk+δ2k+ck)Ik(t,x)+∫∞0γk(b)rk(t,b,x)db,∂rk(t,b,x)∂t+∂rk(t,b,x)∂b=d4k△rk(t,b,x)−γk(b)rk(t,b,x)−μkrk(t,b,x),ek(t,0,x)=n∑j=1βkjSk(t,x)Ij(t,x),rk(t,0,x)=ckIk(t,x), $
|
(2.3) |
for $ x\in\Omega $, $ a, b\in \mathbb{R}_+ = (0, +\infty) $, with the homogeneous Neumann boundary conditions
$ \frac{\partial S_k(t,x)}{\partial\nu} = \frac{\partial e_k(t,a,x)}{\partial\nu} = \frac{\partial I_k(t,x)}{\partial\nu} = \frac{\partial r_k(t,b,x)}{\partial\nu} = 0,\;x\in\partial\Omega, $ |
and initial functions
$ S_k(0,x) = S_k^0(x),\;e_k(0,a,x) = e_k^0(a,x),\;I_k(0,x) = I_k^0(x),\;r_k(0,b,x) = r_k^0(b,x). $ |
$ d_{1k} $, $ d_{2k} $, $ d_{3k} $, $ d_{4k} $ denote the diffusion coefficients of susceptible individuals, exposed individuals, infectious individuals and removed individuals in group $ k $, respectively. And the other parameters have the same biological meanings as in (2.2). The homogeneous Neumann boundary conditions imply that there is no population flux across the boundary $ \partial\Omega $.
We define the functional spaces $ X = C(\bar{\Omega}, \mathbb{R}) $ and $ Y = L^1(\mathbb{R}_+, X) $ for model (2.3) equipped, respectively, with the norms
$ |\phi|_X = \sup\limits_{x\in\bar{\Omega}}|\phi(x)|,\quad\|\varphi\|_Y = \int_0^\infty|\varphi(a,\cdot)|_Xda, $ |
for $ \phi\in X $, $ \varphi\in Y $. The positive cones are denoted by $ X_+ $ and $ Y_+ $. In addition, we define a vector space $ Z = (C([0, T], X))^{2n} $ with the norm
$ \|\psi\|_{Z} = \max\limits_i\sup\limits_{0\leq t\leq T}|\psi_i(t,\cdot)|_X,\;\psi = (\psi_1,\psi_2,...,\psi_{2n})\in Z. $ |
Throughout this paper, for convenience, we always denote $ S = (S_1, S_2, ..., S_n) $, $ e = (e_1, e_2, ..., e_n) $, $ I = (I_1, I_2, ..., I_n) $, $ r = (r_1, r_2, ..., r_n) $, and $ S^0 = (S^0_1, S^0_2, ..., S^0_n) $, $ e^0 = (e^0_1, e^0_2, ..., e^0_n) $, $ I^0 = (I^0_1, I^0_2, ..., I^0_n) $, $ r^0 = (r^0_1, r^0_2, ..., r^0_n) $. We also denote $ (y_1, y_2, ..., y_n)^T > (z_1, z_2, ..., z_n)^T $ as $ y_i > z_i $ for all $ i = 1, 2, ..., n $. For each $ i = 1, 2, 3, 4 $, we suppose that $ T_{ik}:C(\bar{\Omega}, \mathbb{R})\rightarrow C(\bar{\Omega}, \mathbb{R}) $ is the $ C_0 $ semigroup generated by $ d_{ik}\Delta $ subjects to the Neumann boundary condition in group $ k $. From subsection 2.1 in [28], we have
$ (T_{ik}(t)[\phi])(x) = \int_\Omega\Gamma_{ik}(t,x,y)\phi(y)dy, $ |
for all $ t > 0 $ and $ \phi\in C(\bar{\Omega}, \mathbb{R}) $, where $ \Gamma_{ik}(t, x, y) $ is the Green function. We have that $ T_{ik} $, $ i = 1, 2, 3, 4 $, $ k = 1, 2, ..., n $ are compact and strongly positive for each $ t > 0 $ by the Corollary 7.2.3 in [29]. Integrating the second equation in model (2.3) along the characteristic line $ t-a = c $, where $ c $ is a constant, we obtain
$ ek(t,a,x)={∫ΩΓ2k(a,x,y)ek(t−a,0,y)dyπ1k(a),t≥a,∫ΩΓ2k(a,x,y)e0k(a−t,y)dyπ1k(a)π1k(a−t),t<a, $
|
(2.4) |
where $ \pi_{1k}(a) = e^{-\int_0^a[\mu_k+\delta_{1k}+\sigma_k(s)]ds} $. Similarly,
$ rk(t,b,x)={∫ΩΓ4k(b,x,y)rk(t−b,0,y)dyπ2k(b),t≥b,∫ΩΓ4k(b,x,y)r0k(b−t,y)dyπ2k(b)π2k(b−t),t<b, $
|
(2.5) |
where $ \pi_{2k}(b) = e^{-\int_0^b[\mu_k+\gamma_k(s)]ds} $. To study the asymptotic behaviors of the dynamics of model (2.3), we require the following assumptions on the model parameters.
Assumption 2.1. For each $ k, j = 1, 2, ..., n $,
$ (H_1) \; d_{1k}, d_{2k}, d_{3k}, d_{4k}, \Lambda_k, \mu_k, \delta_{1k}, \delta_{2k}, c_k > 0 $.
$ (H_2) \; \beta_{kj}\in\mathbb{R}_+ $, and the $ n $-dimensional square matrix $ (\beta_{kj})_{n\times n} $ is irreducible.
$ (H_3) \; \sigma_k(\cdot), \gamma_k(\cdot)\in L^{\infty}(\mathbb{R}_+, \mathbb{R}_+) $, $ \bar\sigma_k > 0 $ and $ \bar\gamma_k > 0 $, where $ \bar{\sigma}_k: = \mathop{\rm{ess\; sup}}\limits_{a\in\mathbb{R}_+}\sigma_k(a) $, $ \bar{\gamma}_k: = \mathop{\rm{ess\; sup}}\limits_{b\in\mathbb{R}_+}\gamma_k(b). $
We define $ A_k(t, x) = e_k(t, 0, x) $, $ B_k(t, x) = r_k(t, 0, x) $ for $ (t, x)\in\mathbb{R}_+\times\Omega $, and let $ A = (A_1, A_2, ..., A_n) $, $ B = (B_1, B_2, ..., B_n) $ and $ C = (A, B) $. Positiveness of the solutions of model (2.3) is given below.
Theorem 3.1. Suppose that there exists a solution $ (S(t, \cdot), e(t, \cdot, \cdot), I(t, \cdot), r(t, \cdot, \cdot))\in X^n\times Y^n\times X^n\times Y^n $ of (2.3) corresponding to $ (S^0, e^0, I^0, r^0)\in X_+^n\times Y_+^n\times X_+^n\times Y_+^n $ with an interval of existence $ [0, T] $, $ T > 0 $. Then
$ (S(t,\cdot),e(t,\cdot,\cdot),I(t,\cdot),r(t,\cdot,\cdot))\in X_+^n\times Y_+^n\times X_+^n\times Y_+^n, $ |
for all $ t\in[0, T] $.
Proof. From the first equation of (2.3), we have
$ Sk(t,x)=ˆFSk(t,x)+∫t0e−∫ta[μk+n∑j=1βkjIj(τ,x)]dτΛk∫ΩΓ1k(t−a,x,y)dyda, $
|
where $ \hat{F}_{S_k}(t, x) = e^{-\int_{0}^t\big[\mu_k+{\sum\limits_{j = 1}^n}{\beta_{kj}I_j(\tau, x)}\big]d\tau}\int_\Omega\Gamma_{1k}(t, x, y)S_k^0(y)dy $. The positivity of $ \Lambda_k $ and $ S_k^0 $ ensures $ S_k(t, x) > 0 $ for each $ (t, x)\in[0, T]\times\Omega $. The positivity of $ C $ which means the positivity for $ A_k $ and $ B_k $, $ k = 1, 2, ..., n $ is established by constructing Picard sequences as follows.
Solving equation $ I_k $ for system (2.3), we have
$ Ik(t,x)=FIk(t,x)+∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)×[∫∞0σk(b)ek(a,b,y)db+∫∞0γk(b)rk(a,b,y)db]dyda, $
|
(3.1) |
where $ F_{I_k}(t, x) = e^{-(\mu_k+\delta_{2k}+c_k)t}\int_\Omega\Gamma_{3k}(t, x, y)I_k^0(y)dy. $ For $ (t, x)\in[0, T]\times\Omega $, by (2.4) and (2.5), we obtain
$ ∫∞0σk(b)ek(t,b,y)db=FAk(t,y)+∫t0σk(b)π1k(b)∫ΩΓ2k(b,y,z)Ak(t−b,z)dzdb, $
|
(3.2) |
where $ F_{A_k}(t, y) = \int_0^\infty\sigma_k(b+t)\frac{\pi_{1k}(b+t)}{\pi_{1k}(b)}\int_\Omega\Gamma_{2k}(b+t, y, z)e_k^0(b, z)dzdb $, and
$ ∫∞0γk(b)rk(t,b,y)db=FBk(t,y)+∫t0γk(b)π2k(b)∫ΩΓ4k(b,y,z)Bk(t−b,z)dzdb, $
|
(3.3) |
where $ F_{B_k}(t, y) = \int_0^\infty\gamma_k(b+t)\frac{\pi_{2k}(b+t)}{\pi_{2k}(b)}\int_\Omega\Gamma_{4k}(b+t, y, z)r_k^0(b, z)dzdb $. From (3.1)–(3.3) and the definitions of $ A_k $ and $ B_k $, we have
$ Ak(t,x)=n∑j=1βkjSk(t,x){FIj(t,x)+∫t0e−(μj+δ2j+cj)(t−a)∫ΩΓ3j(t−a,x,y)[FAj(a,y)+∫a0σj(b)π1j(b)∫ΩΓ2j(b,y,z)Aj(a−b,z)dzdb+FBj(a,y)+∫a0γj(b)π2j(b)×∫ΩΓ4j(b,y,z)Bj(a−b,z)dzdb]dyda}, $
|
and
$ Bk(t,x)=ck{FIk(t,x)+∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)[FAk(a,y)+∫a0σk(b)×π1k(b)∫ΩΓ2k(b,y,z)Ak(a−b,z)dzdb+FBk(a,y)+∫a0γk(b)π2k(b)×∫ΩΓ4k(b,y,z)Bk(a−b,z)dzdb]dyda}. $
|
Let
$ A(0)k(t,x)=n∑j=1βkjSk(t,x){FIj(t,x)+∫t0e−(μj+δ2j+cj)(t−a)∫ΩΓ3j(t−a,x,y)×[FAj(a,y)+FBj(a,y)]dyda}, $
|
and
$ B(0)k(t,x)=ck{FIk(t,x)+∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)×[FAk(a,y)+FBk(a,y)]dyda}. $
|
Then it is obvious that $ A_k^{(0)}(t, x) > 0 $, $ B_k^{(0)}(t, x) > 0 $. Now we assume that $ A_k^{(m)}(t, x) > 0 $, $ B_k^{(m)}(t, x) > 0 $ ($ m\in\mathbb{N} $) for $ e^0_k > 0 $, $ r^0_k > 0 $ and $ (t, x)\in[0, T]\times\Omega $. Then
$ A(m+1)k(t,x)=A(0)k(t,x)+n∑j=1βkjSk(t,x){∫t0e−(μj+δ2j+cj)(t−a)∫ΩΓ3j(t−a,x,y)[∫a0σj(b)×π1j(b)∫ΩΓ2j(b,y,z)A(m)j(a−b,z)dzdb+∫a0γj(b)π2j(b)∫ΩΓ4j(b,y,z)×B(m)j(a−b,z)dzdb]dyda}, $
|
and
$ B(m+1)k(t,x)=B(0)k(t,x)+ck{∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)[∫a0σk(b)π1k(b)×∫ΩΓ2k(b,y,z)A(m)k(a−b,z)dzdb+∫a0γk(b)π2k(b)∫ΩΓ4k(b,y,z)×B(m)k(a−b,z)dzdb]dyda}. $
|
From the positivity of $ \beta_{kj} $, $ \sigma_k $ and $ \gamma_k $, together with the positivity of $ \Gamma_{2k} $, $ \Gamma_{3k} $ and $ \Gamma_{4k} $, it follows that
$ A(1)k(t,x)−A(0)k(t,x)=n∑j=1βkjSk(t,x){∫t0e−(μj+δ2j+cj)(t−a)∫ΩΓ3j(t−a,x,y)[∫a0σj(b)×π1j(b)∫ΩΓ2j(b,y,z)A(0)j(a−b,z)dzdb+∫a0γj(b)π2j(b)∫ΩΓ4j(b,y,z)×B(0)j(a−b,z)dzdb]dyda}>0, $
|
and
$ B(1)k(t,x)−B(0)k(t,x)=ck{∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)[∫a0σk(b)π1k(b)∫ΩΓ2k(b,y,z)×A(0)k(a−b,z)dzdb+∫a0γk(b)π2k(b)∫ΩΓ4k(b,y,z)B(0)k(a−b,z)dzdb]dyda}>0, $
|
which lead to $ C^{(1)}(t, x)-C^{(0)}(t, x) > 0 $ for $ (t, x)\in[0, T]\times\Omega $. We assume that $ C^{(m)}(t, x)-C^{(m-1)}(t, x) > 0 $ for all $ m\geq2 $, that is, $ A_k^{(m)}(t, x)-A_k^{(m-1)}(t, x) > 0 $ and $ B_k^{(m)}(t, x)-B_k^{(m-1)}(t, x) > 0 $, $ k = 1, 2, ..., n $. Then,
$ A(m+1)k(t,x)−A(m)k(t,x)=n∑j=1βkjSk(t,x){∫t0e−(μj+δ2j+cj)(t−a)∫ΩΓ3j(t−a,x,y)(∫a0σj(b)×π1j(b)∫ΩΓ2j(b,y,z)[A(m)j(a−b,z)−A(m−1)j(a−b,z)]dzdb+∫a0γj(b)×π2j(b)∫ΩΓ4j(b,y,z)[B(m)j(a−b,z)−B(m−1)j(a−b,z)]dzdb)dyda}>0, $
|
and
$ B(m+1)k(t,x)−B(m)k(t,x)=ck{∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)(∫a0σk(b)π1k(b)∫ΩΓ2k(b,y,z)×[A(m)k(a−b,z)−A(m−1)k(a−b,z)]dzdb+∫a0γk(b)π2k(b)∫ΩΓ4k(b,y,z)×[B(m)k(a−b,z)−B(m−1)k(a−b,z)]dzdb)dyda}>0. $
|
Hence applying mathematical induction, we show that the sequence $ \{C^{(m)}\}_0^\infty $ is monotonically increasing.
Next, applying the contraction mapping principle, we show the sequence $ \{C^{(m)}\}_0^\infty $ converges to $ C(t, x) $ for any $ (t, x)\in[0, T]\times\Omega $ as $ m $ approaches infinity. To this end, we define a variable
$ \hat{C}^{(m)}(t,x) = e^{-\lambda t}C^{(m)}(t,x),\;\text{for some}\;\lambda\in\mathbb{R}_+. $ |
By the definitions of $ A_k^{(m)} $ and $ B_k^{(m)} $, we have
$ ˆA(m+1)k(t,x)=e−λtA(0)k(t,x)+n∑j=1βkjSk(t,x){∫t0e−(μj+δ2j+cj)a∫ΩΓ3j(a,x,y)[∫t−a0σj(b)×π1j(b)∫ΩΓ2j(b,y,z)e−λ(a+b)ˆA(m)j(t−a−b,z)dzdb+∫t−a0γj(b)π2j(b)×∫ΩΓ4j(b,y,z)e−λ(a+b)ˆB(m)j(t−a−b,z)dzdb]dyda}, $
|
and
$ ˆB(m+1)k(t,x)=e−λtB(0)k(t,x)+ck{∫t0e−(μk+δ2k+ck)a∫ΩΓ3k(a,x,y)[∫t−a0σk(b)π1k(b)×∫ΩΓ2k(b,y,z)e−λ(a+b)ˆA(m)k(t−a−b,z)dzdb+∫t−a0γk(b)π2k(b)×∫ΩΓ4k(b,y,z)e−λ(a+b)ˆB(m)k(t−a−b,z)dzdb]dyda}. $
|
For any $ m\in\mathbb{N} $,
$ ˆA(m+1)k−ˆA(m)k≤n∑j=1βkjˆSk{∫t0e−(μj+δ2j+cj)(a)∫ΩΓ3j(a,x,y)[∫t−a0σj(b)π1j(b)∫ΩΓ2j(b,y,z)×e−λ(a+b)dzdb+∫t−a0γj(b)π2j(b)∫ΩΓ4j(b,y,z)e−λ(a+b)dzdb]dyda}×‖ˆC(m)−ˆC(m−1)‖Z≤n∑j=1βkjˆSk(ˉσj+ˉγj)λ2‖ˆC(m)−ˆC(m−1)‖Z, $
|
where $ \hat{S}_k = \max\limits_{t\in[0, T]}|S_k(t, \cdot)|_X $, and
$ ˆB(m+1)k−ˆB(m)k≤ck{∫t0e−(μk+δ2k+ck)a∫ΩΓ3k(a,x,y)[∫t−a0σk(b)π1k(b)∫ΩΓ2k(b,y,z)×e−λ(a+b)dzdb+∫t−a0γk(b)π2k(b)∫ΩΓ4k(b,y,z)e−λ(a+b)dzdb]dyda}×‖ˆC(m)−ˆC(m−1)‖Z≤ck(ˉσk+ˉγk)λ2‖ˆC(m)−ˆC(m−1)‖Z. $
|
Hence,
$ ‖ˆC(m+1)−ˆC(m)‖Z≤Kλ‖ˆC(m)−ˆC(m−1)‖Z≤Kmλ‖ˆC(1)−ˆC(0)‖Z, $
|
where $ K_\lambda = \max\left\{M_\lambda, N_\lambda\right\} $, $ M_\lambda = \max\limits_k\bigg\{{\sum\limits_{j = 1}^n}\frac{{\beta_{kj}}\hat{S}_k(\bar{\sigma}_j+\bar{\gamma}_j)}{\lambda^2}\bigg\} $, $ N_\lambda = \max\limits_k\left\{\frac{c_k(\bar{\sigma}_k+\bar{\gamma}_k)}{\lambda^2}\right\} $. Therefore, for any $ m_1 > m_2 $, $ m_1, m_2\in\mathbb{N} $,
$ \|\hat{C}^{(m_1)}-\hat{C}^{(m_2)}\|_Z\leq\frac{K^{m_1}_\lambda}{1-K_\lambda}\|\hat{C}^{(1)}-\hat{C}^{(0)}\|_Z. $ |
We choose $ \lambda $ sufficiently large such that $ {\sum\limits_{j = 1}^n}\frac{{\beta_{kj}}\hat{S}_k(\bar{\sigma}_j+\bar{\gamma}_j)}{\lambda^2} < 1 $ and $ \frac{c_k(\bar{\sigma}_k+\bar{\gamma}_k)}{\lambda^2} < 1 $ for all $ k = 1, 2, ..., n $, then $ K_\lambda < 1 $. Hence, $ \|\hat{C}^{(m_1)}-\hat{C}^{(m_2)}\|_Z\rightarrow0 $ as $ m_2\rightarrow\infty $ which implies that $ \hat{C}^{(m)}\rightarrow\hat{C} $ and thus $ C^{(m)}\rightarrow C $ as $ m\rightarrow\infty $. Furthermore, we have $ A_k^{(m)}\rightarrow A_k $ and $ B_k^{(m)}\rightarrow B_k $ for $ k = 1, 2, ..., n $ as $ m\rightarrow\infty $. Since sequence $ \{C^{(m)}\}_0^\infty $ is monotonically increasing, we obtain $ A_k $ and $ B_k $ are positive for $ k = 1, 2, ..., n $.
By (2.4) and (2.5), together with the positivity of $ e_k^0 $, $ r_k^0 $, $ A_k $ and $ B_k $, we conclude that $ e_k(t, a, x) $ and $ r_k(t, a, x) $ are positive. For the positivity of $ I_k $, we prove this by contradiction. Suppose that there exist $ x_0\in\Omega $ and $ t_0 = \inf\{t\in\mathbb{R}_+|I_k(t, x_0) = 0\} $ such that
$ I_k(t_0,x_0) = 0,\;I_k(t,x_0) \gt 0,\;\frac{\partial I_k(t_0,x_0)}{\partial t}\leq0,\;t\in[0,t_0). $ |
By the third equation of system (2.3), we can easily obtain
$ ∂Ik(t0,x0)∂t=FAk(t0,x0)+FBk(t0,x0)+∫t00σk(a)π1k(a)∫ΩΓ2k(a,x0,y)Ak(t0−a,y)dyda+∫t00γk(a)π2k(a)∫ΩΓ4k(a,x0,y)Bk(t0−a,y)dyda>0. $
|
This leads to a contradiction. Hence, for any $ t\in[0, T] $, we have $ (S(t, \cdot), e(t, \cdot, \cdot), I(t, \cdot), r(t, \cdot, \cdot))\in X_+^n\times Y_+^n\times X_+^n\times Y_+^n $.
Let $ N_k(t) = \int_\Omega\left[S_k(t, x)+\int_0^\infty e_k(t, a, x)da+I_k(t, x)+\int_0^\infty r_k(t, b, x)db\right]dx $ denotes the total population at time $ t $ in group $ k $ and region $ \Omega $.
Theorem 3.2. If
$ lima→+∞ek(t,a,x)=0,limb→+∞rk(t,b,x)=0, $
|
(3.4) |
for all $ t > 0 $, $ x\in\Omega $, the region $ \Pi $ defined by
$ \Pi = \bigg\{\left(S_{k}, e_{k}, I_{k}, r_{k}\right)|\;N_k\leq\frac{\Lambda_k}{\mu_k}|\Omega|\bigg\}, $ |
is positively invariant for system (2.3).
Proof. Following condition (3.4) and the equations of system (2.3), we have
$ ∂Sk(t,x)∂t+∫∞0∂ek(t,a,x)∂tda+∂Ik(t,x)∂t+∫∞0∂rk(t,b,x)∂tdb=d1k△Sk(t,x)+d2k∫∞0△ek(t,a,x)da+d3k△Ik(t,x)+d4k∫∞0△rk(t,b,x)da+Λk−μkSk(t,x)−n∑j=1βkjSk(t,x)Ij(t,x)−∫∞0∂ek(t,a,x)∂ada−∫∞0σk(a)×ek(t,a,x)da−∫∞0(μk+δ1k)ek(t,a,x)da+∫∞0σk(a)ek(t,a,x)da−(μk+δ2k+ck)Ik(t,x)+∫∞0γk(b)rk(t,b,x)db−∫∞0∂rk(t,b,x)∂bdb−∫∞0γk(b)rk(t,b,x)db−∫∞0μkrk(t,b,x)db=d1k△Sk(t,x)+d2k∫∞0△ek(t,a,x)da+d3k△Ik(t,x)+d4k∫∞0△rk(t,b,x)da+Λk−μkSk(t,x)−∫∞0(μk+δ1k)ek(t,a,x)da−(μk+δ2k)Ik(t,x)−∫∞0μkrk(t,b,x)db<d1k△Sk(t,x)+d2k∫∞0△ek(t,a,x)da+d3k△Ik(t,x)+d4k∫∞0△rk(t,b,x)da+Λk−μkSk(t,x)−μk∫∞0ek(t,a,x)da−μkIk(t,x)−μk∫∞0rk(t,b,x)db. $
|
Noting the Neumann boundary conditions of system (2.3) and using the Gauss formula, we derive
$ ∫Ωd1k△Sk(t,x)dx=∫Ω∫∞0d2k△ek(t,a,x)dadx=∫Ωd3k△Ik(t,x)dx=∫Ω∫∞0d4k△rk(t,b,x)dadx=0. $
|
It follows that
$ dNk(t)dt=∫Ω[∂Sk(t,x)∂t+∫∞0∂ek(t,a,x)∂tda+∂Ik(t,x)∂t+∫∞0∂rk(t,b,x)∂tdb]dx<∫Ω{Λk−μk[Sk(t,x)+∫∞0ek(t,a,x)da+Ik(t,x)+∫∞0rk(t,b,x)db]}dx=Λk|Ω|−μkNk(t). $
|
Thus if $ N_k(t) > \frac{\Lambda_k}{\mu_k}|\Omega| $, then $ \frac{dN_k(t)}{dt} < 0 $. Moreover, we observe the ordinary differential equation
$ \frac{dN_k(t)}{dt} = \Lambda_k|\Omega|-\mu_kN_k(t), $ |
with general solution
$ N_k(t) = \frac{\Lambda_k}{\mu_k}|\Omega|+\left[N_k(0)-\frac{\Lambda_k}{\mu_k}|\Omega|\right]e^{-{\mu_k}t}, $ |
where $ N_k(0) $ means the initial value of total population in group $ k $ and region $ \Omega $. By applying the standard comparison theorem, we have for all $ t\geq0 $,
$ N_k(t)\leq\frac{\Lambda_k}{\mu_k}|\Omega|,\quad \text{if}\;N_k(0)\leq\frac{\Lambda_k}{\mu_k}|\Omega|. $ |
Hence, $ \Pi $ is positive invariant for system (2.3).
The existence and uniqueness of the solution of model (2.3) follow from Banach-Picard fixed point theorem.
Theorem 3.3. Let initial functions satisfy $ (S^0, e^0, I^0, r^0)\in X_+^n\times Y_+^n\times X_+^n\times Y_+^n $. Then the system (2.3) has a unique solution $ (S(t, \cdot), e(t, \cdot, \cdot), I(t, \cdot), r(t, \cdot, \cdot))\in X_+^n\times Y_+^n\times X_+^n\times Y_+^n $ for $ t\in[0, T] $.
Proof. Solving equation $ S_k $ for system (2.3), we have
$ Sk(t,x)=FSk(t,x)+∫t0e−μk(t−a)∫ΩΓ1k(t−a,x,y)[Λk−Ak(a,y)]dyda, $
|
(3.5) |
for $ (t, x)\in[0, T]\times\Omega $, where $ F_{S_k}(t, x) = e^{-\mu_kt}\int_\Omega\Gamma_{1k}(t, x, y)S_k^0(y)dy. $ From (3.1)–(3.5) and the definitions of $ A_k $ and $ B_k $, we have
$ Ak(t,x)=n∑j=1βkj{FSk(t,x)+∫t0e−μk(t−a)∫ΩΓ1k(t−a,x,y)[Λk−Ak(a,y)]dyda}×{FIj(t,x)+∫t0e−(μj+δ2j+cj)(t−a)∫ΩΓ3j(t−a,x,y)[FAj(a,y)+∫a0σj(b)×π1j(b)∫ΩΓ2j(b,y,z)Aj(a−b,z)dzdb+FBj(a,y)+∫a0γj(b)π2j(b)×∫ΩΓ4j(b,y,z)Bj(a−b,z)dzdb]dyda}:=F1k[C](t,x), $
|
(3.6) |
and
$ Bk(t,x)=ck{FIk(t,x)+∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)[FAk(a,y)+∫a0σk(b)×π1k(b)∫ΩΓ2k(b,y,z)Ak(a−b,z)dzdb+FBk(a,y)+∫a0γk(b)π2k(b)×∫ΩΓ4k(b,y,z)Bk(a−b,z)dzdb]dyda}:=F2k[C](t,x), $
|
(3.7) |
where $ \mathcal{F}_{1k} $, $ \mathcal{F}_{2k}:Z\rightarrow C([0, T], X) $ are nonlinear operators for each $ k = 1, 2, ..., n $. For the sake of convenience, we define for each $ (t, x)\in[0, T]\times\Omega $,
$ FCk(t,x)=FSk+∫t0e−μk(t−a)∫ΩΓ1k(t−a,x,y)Λkdyda,FDk(t,x)=FIk(t,x)+∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)[FAk(a,y)+FBk(a,y)]dyda,Θ1(Ak)=∫t0e−μk(t−a)∫ΩΓ1k(t−a,x,y)Ak(a,y)dyda,Θ2(Ak)=∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)∫a0σk(b)π1k(b)×∫ΩΓ2k(b,y,z)Ak(a−b,z)dzdbdyda,Θ3(Bk)=∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)∫a0γk(b)π2k(b)×∫ΩΓ4k(b,y,z)Bk(a−b,z)dzdbdyda. $
|
Then
$ F1k[C]=n∑j=1βkj[FCk−Θ1(Ak)][FDj+Θ2(Aj)+Θ3(Bj)],F2k[C]=ck[FDk+Θ2(Ak)+Θ3(Bk)]. $
|
For any $ C $, $ \bar{C}\in Z $, we set $ \tilde{C} = C-\bar{C} $. Then, from the positivity of $ A_k $ and $ B_k $ proved in Theorem 3.1, we have
$ F1k[C]−F1k[ˉC]=n∑j=1βkj{FCk[Θ2(˜Aj)+Θ3(˜Bj)]−FDjΘ1(˜Ak)−Θ1(˜Ak)×[Θ2(Aj)+Θ3(Bj)]−Θ1(ˉAk)[Θ2(˜Aj)+Θ3(˜Bj)]}≤n∑j=1βkjFCk[Θ2(˜Aj)+Θ3(˜Bj)]≤n∑j=1βkj|FCk(ˆΘ2+ˆΘ3)|‖˜C‖Z, $
|
and
$ F2k[C]−F2k[C]=ck[Θ2(˜Ak)+Θ3(˜Bk)]≤ck|ˆΘ2+ˆΘ3|‖˜C‖Z, $
|
where
$ ˆΘ2=∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)∫a0σk(b)π1k(b)×∫ΩΓ2k(b,y,z)dzdbdyda,ˆΘ3=∫t0e−(μk+δ2k+ck)(t−a)∫ΩΓ3k(t−a,x,y)∫a0γk(b)π2k(b)×∫ΩΓ4k(b,y,z)Bk(a−b,z)dzdbdyda. $
|
Denote
$ m1k(T)=n∑j=1βkj|FCk(T,⋅)[ˆΘ2(T,⋅)+ˆΘ3(T,⋅)]|X,m2k(T)=ck|ˆΘ2(T,⋅)+ˆΘ3(T,⋅)|X,m(T)=max{m11(T),m12(T),...,m1n(T),m21(T),m22(T),...,m2n(T)}, $
|
and
$ \mathcal{F}[C] = (\mathcal{F}_{11},\mathcal{F}_{12},...,\mathcal{F}_{1n},\mathcal{F}_{21},\mathcal{F}_{22},...,\mathcal{F}_{2n})[C]:Z\rightarrow Z. $ |
Clearly, we can choose $ T $ small enough such that $ m_{1k}(T) < 1 $ and $ m_{2k}(T) < 1 $ for all $ k = 1, 2, ..., n $. Consequently, we have $ m(T) < 1 $. Then
$ \|\mathcal{F}C-\mathcal{F}\bar{C}\|_Z\leq m(T)\|C-\bar{C}\|_Z. $ |
Hence, applying contraction operator theorem [30], we conclude that $ \mathcal{F} $ has a unique fixed point $ C = (A_1, A_2, ..., A_n, B_1, B_2, ..., B_n) $. From Theorem 3.1, (2.4) and (2.5), together with $ A_k(t, x) = {\sum\limits_{j = 1}^n}{\beta_{kj}S_k(t, x)I_j(t, x)} $ and $ B_k(t, x) = c_kI_k(t, x) $, we derive the existence and uniqueness of the solution $ (S(t, \cdot), e(t, \cdot, \cdot), I(t, \cdot), r(t, \cdot, \cdot))\in X_+^n\times Y_+^n\times X_+^n\times Y_+^n $ for system (2.3).
To further establish the global existence of the solution of system (2.3), we need the following lemma.
Lemma 3.1. [31]. The following problem
$ {∂ω(t,x)∂t=dω△ω(t,x)+Λ−μω(t,x),x∈Ω,∂ω(t,x)∂ν=0,x∈∂Ω, $
|
admits a unique positive steady state $ \omega^* = \frac{\Lambda}{\mu} $, which is globally attractive in $ X $.
Theorem 3.4. Let initial functions satisfy $ (S^0, e^0, I^0, r^0)\in X_+^n\times Y_+^n\times X_+^n\times Y_+^n $. Then the system (2.3) has a unique solution $ (S(t, \cdot), e(t, \cdot, \cdot), I(t, \cdot), r(t, \cdot, \cdot))\in X_+^n\times Y_+^n\times X_+^n\times Y_+^n $ for $ t\in\mathbb{R}_+ $.
Proof. To extend the domain of existence from $ t\in[0, T] $ to $ t\in\mathbb{R}_+ $, it suffices to show that the solution does not blow up in finite time. In fact, by Theorem 3.1, we have that
$ \frac{\partial S_k(t,x)}{\partial t}\leq d_{1k}\triangle S_k(t,x)+\Lambda_k-\mu_k S_k(t,x), $ |
for all $ t > 0 $ and $ x\in\Omega $. From Lemma 3.1 and the comparison principle, we get that $ S_k(t, x) $ is bounded above by the upper solution $ \frac{\Lambda_k}{\mu_k} $.
We now claim that $ e_k(t, a, x) < +\infty $ for all $ t > 0 $, $ a > 0 $, $ x\in\Omega $ and $ k = 1, 2, ..., n $. From (2.4), it is sufficient to show that $ e_k(t, 0, x) < +\infty $ for all $ t > 0 $ and $ x\in\Omega $. Suppose on the contrary that there exist $ t_e > 0 $ and $ x_e\in\Omega $ such that
$ \lim\limits_{t\rightarrow t_e-0}e_k(t,0,x_e) = +\infty. $ |
We then have from (3.5) that
$ \lim\limits_{t\rightarrow t_e-0}S_k(t,x_e) = -\infty, $ |
which implies that $ S_k(t, x_e) $ is negative in the neighborhood of $ t_e $. This contradicts to the positivity of $ S_k $, which has been proved in Theorem 3.1. Furthermore, from $ e_k(t, 0, x) = {\sum\limits_{j = 1}^n}{\beta_{kj}S_k(t, x)I_j(t, x)} $ and $ r_k(t, 0, x) = c_kI_k(t, x) $ in (2.4), we obtain $ I_k(t, x) < +\infty $ and $ r_k(t, 0, x) < +\infty $. And from (2.5), we get $ r_k(t, a, x) < +\infty $ for all $ t > 0 $, $ a > 0 $, $ x\in\Omega $ and $ k = 1, 2, ..., n $. Thus, blow up never occurs, and we obtain a solution $ (S(t, \cdot), e(t, \cdot, \cdot), I(t, \cdot), r(t, \cdot, \cdot))\in X_+^n\times Y_+^n\times X_+^n\times Y_+^n $ for $ t\in\mathbb{R}_+ $.
Theorem 3.5. The solution semiflow $ \Phi(t) = \left(S(t, \cdot), e(t, \cdot, \cdot), I(t, \cdot), r(t, \cdot, \cdot)\right):X^n_+\times Y^n_+\times X^n_+\times Y^n_+\backslash(\bf{0}, \bf{0}, \bf{0}, \bf{0})\rightarrow X^n_+\times Y^n_+\times X^n_+\times Y^n_+\backslash(\bf{0}, \bf{0}, \bf{0}, \bf{0}) $ of model (2.3) has a compact and global attractor as condition (3.4) holds.
Proof. According to Theorem 3.2, we know that system (2.3) is ultimately bounded, which implies that solution semiflow $ \Phi(t) $ is point dissipative on $ X^n_+\times Y^n_+\times X^n_+\times Y^n_+\backslash(\bf{0}, \bf{0}, \bf{0}, \bf{0}) $. By Theorem 2.6 in [15], we can get that $ \Phi(t) $ is compact for any $ t > 0 $. Thus, from Theorem 3.4.8 in [32], we further obtain that $ \Phi(t) $ has a compact and global attractor in $ X^n_+\times Y^n_+\times X^n_+\times Y^n_+\backslash(\bf{0}, \bf{0}, \bf{0}, \bf{0}) $.
It is easy to see that model (2.3) has a unique disease-free steady state $ \bar{E} = (\bar{S}, \bf{0}, \bf{0}, \bf{0})^T $ where $ \bar{S} = (\bar{S}_1, \bar{S}_2, ..., \bar{S}_n)^T $ and $ \bar{S}_k = \frac{\Lambda_k}{\mu_k} $, $ k = 1, 2, ..., n. $
We denote $ E^* = (S^*, e^*(\cdot), I^*, r^*(\cdot))^T $ as the space-independent endemic steady state of (2.3), where $ S^* = (S_1^*, S_2^*, ..., S_n^*)^T $, $ e^*(\cdot) = (e_1^*(\cdot), e_2^*(\cdot), ..., e_n^*(\cdot))^T $, $ I^* = (I_1^*, I_2^*, ..., I_n^*)^T $ and $ r^*(\cdot) = (r_1^*(\cdot), r_2^*(\cdot), ..., r_n^*(\cdot))^T $. Then, $ E^* $ satisfies
$ {Λk−μkS∗k−n∑j=1βkjS∗kI∗j=0,de∗k(a)dt=−[σk(a)+μk+δ1k]e∗k(a),∫∞0σk(a)e∗k(a)da−(μk+δ2k+ck)I∗k+∫∞0γk(b)r∗k(b)db=0,dr∗k(b)dt=−[γk(b)+μk]r∗k(b),e∗k(0)=n∑j=1βkjS∗kI∗j,r∗k(0)=ckI∗k. $
|
(4.1) |
We denote $ f_k(I^*) = {\sum\limits_{j = 1}^n}{\beta_{kj}I_j^*}:\mathbb{R}_+^n\rightarrow\mathbb{R}_+ $, $ h_k(I^*) = \frac{\Lambda_k}{\mu_k+f_k(I^*)}:\mathbb{R}_+^n\rightarrow\mathbb{R}_+ $, $ k = 1, 2, ..., n $ for brevity. By solving the Eq (4.1), we get
$ S∗k=hk(I∗),e∗k(a)=S∗kπ1k(a)fk(I∗),r∗k(b)=ckπ2k(b)I∗k, $
|
(4.2) |
where $ \pi_{1k}(\cdot) $ and $ \pi_{2k}(\cdot) $ are given in (2.4) and (2.5), respectively. Substituting (4.2) into the third equation in (4.1) and rearranging it, we have
$ I∗k=Lkhk(I∗)fk(I∗)(μk+δ2k+ck)−ckPk, $
|
(4.3) |
where $ L_k = {\int_0^\infty}{\sigma_k(a)\pi_{1k}(a)da} $ and $ P_k = {\int_0^\infty}{\gamma_k(b)\pi_{2k}(b)db} $. Let us define a matrix-valued function $ M(x) $ on $ \mathbb{R}^n $ to $ \mathbb{R}^{n\times n} $, where $ M(x)_{ij} = \frac{L_i\beta_{ij}h_i(x)}{(\mu_i+\delta_{2i}+c_i)-c_iP_i} $, $ x = (x_1, x_2, ..., x_n)^T\in\mathbb{R}^n $. Then, (4.3) is equivalent to
$ I∗=M(I∗)I∗. $
|
(4.4) |
On the existence of the endemic equilibrium $ E^* $ of system (2.3), we prove the following theorem.
Theorem 4.1. Let $ M^0 = \left(\frac{{\bar{S}_iL_i\beta_{ij}}}{(\mu_i+\delta_{2i}+c_i)-c_iP_i}\right)_{n\times n} $. If $ \rho(M^0) > 1 $, where $ \rho(M^0) $ represents the spectral radius of $ M^0 $, then system (2.3) has a space-independent steady state $ E^* $.
Proof. From (4.4), we only need to show that the nonlinear operator $ \tilde{M}(x): = M(x)x $, $ x\in\mathbb{R}_+^n $, has at least one positive fixed point $ x^*\in\mathbb{R}_+^n $. We define $ \|\tilde{M}(x)\|: = \max\limits_{1\leq k\leq n}|\tilde{M}(x)_k| $, where $ \tilde{M}(x)_k $ denotes the $ k $-th entry of vector $ \tilde{M}(x) $. Then, $ \tilde{M}(x) $ is monotone increasing with respect to $ x\in\mathbb{R}_+^n $ and uniformly bounded above by $ \max\limits_{1\leq k\leq n}\left|\frac{{\Lambda_kL_k}}{(\mu_k+\delta_{2k}+c_k)-c_kP_k}\right| $.
It is obvious that $ \tilde{M}(\bf{0}) = \bf{0} $ and $ {M}(\bf{0}) $ is the strong Fréchet derivative of $ \tilde{M}(\cdot) $ at the origin. Since $ M(\bf{0}) = M^0 $, we have $ \rho(M(\bf{0})) > 1 $. Thus, it follows from the Perron-Frobenius theorem (see [33]) that $ \rho(M(\bf{0})) $ is a simple eigenvalue of $ M(\bf{0}) $ corresponding to a strictly positive eigenvector, and there exists no nonnegative eigenvector of $ M(\bf{0}) $ corresponding to eigenvalue 1. Hence, we apply Theorem 4.11 of [34], to conclude that operator $ \tilde{M}(\cdot) $ has at least one positive fixed point $ x^*\in\mathbb{R}_+^n $.
Theorem 4.2. For all $ k = 1, 2, ..., n $, if
$ ∫∞0αkσk(θ)π1k(θ)dθ<1,∫∞0γk(θ)π2k(θ)dθ<1, $
|
(4.5) |
where $ \alpha_k = \frac{{\sum\limits_{j = 1}^n}\bar{S}_j\beta_{jk}}{\mu_k+\delta_{2k}} $, $ \pi_{1k}(\cdot) $ and $ \pi_{2k}(\cdot) $ are given in (2.4) and (2.5), respectively, then the disease-free steady state $ \bar{E} $ is globally asymptotically stable.
Proof. We construct a Lyapunov function
$ V(t) = {\sum\limits_{k = 1}^n}{V_k(t)}, $ |
where
$ V_k(t) = \int_\Omega\left[V_{1k}(t,x)+V_{2k}(t,x)+V_{3k}(t,x)+V_{4k}(t,x)\right]dx, $ |
$ V1k(t,x)=ˉSk[Sk(t,x)ˉSk−1−lnSk(t,x)ˉSk],V2k(t,x)=∫∞0χk(a)ek(t,a,x)da,V3k(t,x)=αkIk(t,x),V4k(t,x)=∫∞0ψk(b)rk(t,b,x)db, $
|
and
$ χk(a)=∫∞aαkσk(θ)π1k(θ)π1k(a)dθ,ψk(b)=∫∞bαkγk(θ)π2k(θ)π2k(b)dθ. $
|
(4.6) |
Taking the derivation of $ V_{1k}(t, x) $ along the trajectory of (2.3) with respect to $ t $, we have
$ ∂V1k∂t=Sk(t,x)−ˉSkSk(t,x)∂Sk(t,x)∂t=d1k[Sk(t,x)−ˉSk]△Sk(t,x)Sk(t,x)−μk[Sk(t,x)−ˉSk]2Sk(t,x)+ˉSkn∑j=1βkjIj(t,x)−Sk(t,x)n∑j=1βkjIj(t,x). $
|
(4.7) |
Recalling (2.4), we can rewrite $ V_{2k}(t, x) $ as follows
$ V2k(t,x)=∫t0χk(t−a)∫ΩΓ2k(t−a,x,y)Ak(a,y)dyπ1k(t−a)da+∫∞0χk(a+t)∫ΩΓ2k(a+t,x,y)e0k(a,y)dyπ1k(a+t)π1k(a)da. $
|
Thus, we calculate $ \frac{\partial V_{2k}}{\partial t} $ along the solution of system (2.3) and get
$ ∂V2k∂t=χk(0)Ak(t,x)+∫∞0{χ′k(a)−[μk+δ1k+σk(a)−d2k△]χk(a)}×ek(t,a,x)da. $
|
(4.8) |
Similarly,
$ ∂V3k∂t=αkd3k△Ik(t,x)+αk∫∞0σk(a)ek(t,a,x)da+αk∫∞0γk(b)rk(t,b,x)db−n∑j=1ˉSjβjkIk(t,x)−αkckIk(t,x). $
|
(4.9) |
From (2.5), we get
$ V4k(t,x)=∫t0ψk(t−b)∫ΩΓ4k(t−b,x,y)Bk(b,y)dyπ2k(t−b)db+∫∞0ψk(b+t)∫ΩΓ4k(b+t,x,y)r0k(b,y)dyπ2k(b+t)π2k(b)db. $
|
Thus,
$ ∂V4k∂t=ψk(0)Bk(t,x)+∫∞0{ψ′k(b)−[μk+γk(b)−d4k△]ψk(b)}rk(t,b,x)db. $
|
(4.10) |
Hence, combining (4.7)–(4.10), we calculate the derivative of $ V_{k}(t) $ along the solution trajectory of (2.3) as
$ dVkdt=−d1kˉSk∫Ω|△Sk(t,x)|2S2k(t,x)dx−∫Ωμk[Sk(t,x)−ˉSk]2Sk(t,x)dx+∫Ω[χk(0)−1]×Ak(t,x)dx+∫Ω[ψk(0)−αk]Bk(t,x)dx+∫Ω∫∞0{αkσk(a)−[μk+δ1k+σk(a)−d2k△]χk(a)+χ′k(a)}ek(t,a,x)da+∫Ω∫∞0{αkγk(b)−[μk+γk(b)−d4k△]ψk(b)+ψ′k(b)}×rk(t,b,x)db+∫ΩˉSkn∑j=1βkjIj(t,x)dx−∫Ωn∑j=1ˉSjβjkIk(t,x)dx. $
|
Using (4.6), we yield that
$ dVdt=−n∑k=1d1kˉSk∫Ω|△Sk(t,x)|2S2kdx−n∑k=1∫Ωμk[Sk(t,x)−ˉSk]2Sk(t,x)dx+n∑k=1∫Ω[χk(0)−1]Ak(t,x)dx+n∑k=1∫Ω[ψk(0)−αk]Bk(t,x)dx. $
|
Thus, from (4.5), we have
$ \chi_k(0) \lt 1,\quad\psi_k(0) \lt \alpha_k , $ |
which implies the global asymptotic stability of disease-free steady state $ \bar{E} $.
Theorem 4.3. If $ \rho(M^0) > 1 $, $ \lim\limits_{a\rightarrow+\infty}e_k(t, a, x) = 0 $ and $ \lim\limits_{b\rightarrow+\infty}r_k(t, b, x) = 0 $ for all $ t > 0 $, $ x\in\Omega $, then the space-independent steady state $ E^* $ is globally asymptotically stable.
Proof. We define $ g(x) = x-1-\ln x $, clearly, $ g(x) $ is always positive for $ x > 0 $ and $ g'(x) = 1-\frac{1}{x} $. Consider a matrix $ D = (\bar{\beta}_{kj})_{n\times n} $ with entry $ \bar{\beta}_{kj} = \beta_{kj}L_kS_k^*I_j^* $ and a digraph $ \mathcal{G} = (U, H) $ which contains a set $ U = \{1, 2, ..., n\} $ of vertices and a set $ H $ of arcs $ (k, j) $ leading from initial vertex $ k $ to terminal vertex $ j $, then, we denote a weighted digraph as $ (\mathcal{G}, D) $ for which each arc $ (j, k) $ is assigned a positive weight $ \bar{\beta}_{kj} $. Furthermore, we denote $ \bar{D} $ as the Laplacian matrix of matrix $ (\mathcal{G}, D) $. Then, the irreducibility of matrix $ (\beta_{kj})_{n\times n} $ implies that $ \bar{D} $ is also irreducible. Let $ q_k $ denote the cofactor of $ k $-th diagonal element of $ \bar{D} $. And we construct a Lyapunov function as the following form
$ W(t) = {\sum\limits_{k = 1}^n}{q_kW_k(t)}, $ |
where
$ W_k(t) = \int_\Omega[W_{1k}(t,x)+W_{2k}(t,x)+W_{3k}(t,x)+W_{4k}(t,x)]dx, $ |
$ W1k(t,x)=LkS∗kg[Sk(t,x)S∗k],W2k(t,x)=∫∞0Ψ1k(a)e∗k(a)g[ek(t,a,x)e∗k(a)]da,W3k(t,x)=I∗kg[Ik(t,x)I∗k],W4k(t,x)=∫∞0Ψ2k(b)r∗k(b)g[rk(t,b,x)r∗k(b)]db, $
|
and
$ Ψ1k(a)=∫∞aσk(s)π1k(s)π1k(a)ds,Ψ2k(b)=∫∞bγk(s)π2k(s)π2k(b)ds. $
|
(4.11) |
For convenience, we denote $ J_k(t, x) = {\sum\limits_{j = 1}^n}\beta_{kj}I_j(t, x) $ and $ J_k^* = {\sum\limits_{j = 1}^n}\beta_{kj}I_j^* $. Taking the derivation of $ W_{1k} $ along the trajectory of (2.3) with respect to $ t $, we have
$ ∂W1k∂t=LkS∗k[1S∗k−1Sk(t,x)]∂Sk(t,x)∂t=Lk[1−S∗kSk(t,x)]d1k△Sk(t,x)−LkμkSk(t,x)[Sk(t,x)−S∗k]2+LkS∗kJ∗k×[1+Jk(t,x)J∗k−S∗kSk(t,x)−Sk(t,x)Jk(t,x)S∗kJ∗k]. $
|
(4.12) |
Calculating the derivative of $ W_{2k} $ along the solution of system (2.3) yields
$ ∂W2k∂t=∫∞0Ψ1k(a)e∗k(a)∂∂tg[ek(t,a,x)e∗k(a)]da=∫∞0Ψ1k(a)[1−e∗k(a)ek(t,a,x)]{d2kΔek(t,a,x)−∂∂aek(t,a,x)−[μk+δ1k+σk(a)]ek(t,a,x)}da=∫∞0Ψ1k(a)[1−e∗k(a)ek(t,a,x)]d2kΔek(t,a,x)da−∫∞0Ψ1k(a)e∗k(a)×[ek(t,a,x)e∗k(a)−1][1ek(t,a,x)∂∂aek(t,a,x)+μk+δ1k+σk(a)]da=∫∞0Ψ1k(a)[1−e∗k(a)ek(t,a,x)]d2kΔek(t,a,x)da−∫∞0Ψ1k(a)e∗k(a)×∂∂ag[ek(t,a,x)e∗k(a)]da=∫∞0Ψ1k(a)[1−e∗k(a)ek(t,a,x)]d2kΔek(t,a,x)da+Ψ1k(0)e∗k(0)×g[ek(t,0,x)e∗k(0)]+∫∞0g[ek(t,a,x)e∗k(a)]dda[Ψ1k(a)e∗k(a)]da=∫∞0Ψ1k(a)[1−e∗k(a)ek(t,a,x)]d2kΔek(t,a,x)da+LkS∗kJ∗kg[Sk(t,x)Jk(t,x)S∗kJ∗k]+∫∞0σk(a)e∗k(a)[1−ek(t,a,x)e∗k(a)+lnek(t,a,x)e∗k(a)]da. $
|
(4.13) |
Similarly,
$ ∂W3k∂t=[1−I∗kIk(t,x)][d3k△Ik(t,x)+∫∞0σk(a)ek(t,a,x)da−(μk+δ2k+ck)×Ik(t,x)+∫∞0γk(b)rk(t,b,x)db]=[1−I∗kIk(t,x)]{d3k△Ik(t,x)+∫∞0σk(a)e∗k(a)[ek(t,a,x)e∗k(a)−Ik(t,x)I∗k]da+∫∞0γk(b)r∗k(b)[rk(t,b,x)r∗k(b)−Ik(t,x)I∗k]db}=[1−I∗kIk(t,x)]d3k△Ik(t,x)+∫∞0σk(a)e∗k(a)[1−Ik(t,x)I∗k+ek(t,a,x)e∗k(a)−I∗kek(t,a,x)Ik(t,x)e∗k(a)]da+∫∞0γk(b)r∗k(b)[1−Ik(t,x)I∗k+rk(t,b,x)r∗k(b)−I∗krk(t,b,x)Ik(t,x)r∗k(b)]db, $
|
(4.14) |
and
$ ∂W4k∂t=∫∞0Ψ2k(b)[1−r∗k(b)rk(t,b,x)]d4kΔrk(t,b,x)db+PkckI∗kg[Ik(t,x)I∗k]+∫∞0γk(b)r∗k(b)[1−rk(t,b,x)r∗k(b)+lnrk(t,b,x)r∗k(b)]db. $
|
(4.15) |
Hence, combining (4.12)–(4.15), we calculate the derivative of $ W_{k}(t) $ along the solution trajectory of (2.3) as
$ dWkdt=−Lkd1kS∗k∫Ω|∇Sk(t,x)|2S2k(t,x)dx−∫Ω∫∞0d2ke∗k(a)Ψ1k(a)|∇ek(t,a,x)|2e2k(t,a,x)dadx−d3kI∗k∫Ω|∇Ik(t,x)|2I2k(t,x)dx−∫Ω∫∞0d4kr∗k(b)Ψ2k(b)|∇rk(t,b,x)|2r2k(t,b,x)dbdx−Lkμk∫Ω[Sk(t,x)−S∗k]2Sk(t,x)dx+PkckI∗k∫Ωg[Ik(t,x)I∗k]dx+LkS∗kn∑j=1βkj×I∗j∫Ω[Jk(t,x)J∗k−S∗kSk(t,x)−lnSk(t,x)Jk(t,x)S∗kJ∗k]dx+∫Ω∫∞0σk(a)e∗k(a)×[2−Ik(t,x)I∗k−I∗kek(t,a,x)Ik(t,x)e∗k(a)+lnek(t,a,x)e∗k(a)]dadx+∫Ω∫∞0γk(b)r∗k(b)×[2−Ik(t,x)I∗k−I∗krk(t,b,x)Ik(t,x)r∗k(b)+lnrk(t,b,x)r∗k(b)]dbdx=−Lkd1kS∗k∫Ω|∇Sk(t,x)|2S2k(t,x)dx−∫Ω∫∞0d2ke∗k(a)Ψ1k(a)|∇ek(t,a,x)|2e2k(t,a,x)dadx−d3kI∗k∫Ω|∇Ik(t,x)|2I2k(t,x)dx−∫Ω∫∞0d4kr∗k(b)Ψ2k(b)|∇rk(t,b,x)|2r2k(t,b,x)dbdx−Lkμk∫Ω[Sk(t,x)−S∗k]2Sk(t,x)dx−LkS∗kJ∗k∫Ωg[S∗kSk(t,x)]dx−∫Ω∫∞0σk(a)×e∗k(a)g[I∗kek(t,a,x)Ik(t,x)e∗k(a)]dadx−∫Ω∫∞0γk(b)r∗k(b)g[I∗krk(t,b,x)Ik(t,x)r∗k(b)]dbdx−LkS∗kJ∗k∫Ω{g[Ik(t,x)I∗k]−g[Jk(t,x)J∗k]}dx. $
|
By Theorem 2.3 in [35], the following identity holds
$ n∑k=1qk{LkS∗kJ∗k∫Ω[g(Ik(t,x)I∗k)−g(Jk(t,x)J∗k)]dx}=0. $
|
Hence, together with the property of $ g $, we have $ \frac{dW}{dt}\leq0 $. Furthermore, the strict equality holds only if $ S_k(t, x) = S_k^* $, $ e_k(t, a, x) = e_k^*(a) $, $ I_k(t, x) = I_k^* $, and $ r_k(t, b, x) = r_k^*(b) $. Thus, $ T^* = \left\{E^*\right\}\subset\Omega $ is the largest invariant subset of $ \left\{(S, e, I, r):\frac{dW}{dt} = 0\right\} $, and the Lyapunov-LaSalle invariance principle implies that the endemic equilibrium $ E^* $ is globally asymptotically stable when $ \rho(M^0) > 1 $.
In this section, we present numerical examples to demonstrate the validity and applicability of our main results, Theorems 4.2 and 4.3. For simplicity, we consider the case of two groups, a normalized 1-dimensional space ($ \Omega $ = [0, 3]) and a normalized maximum age ($ a $ = 2). Firstly, we fix some parameter values as follows,
$ d11=0.2,d21=0.5,d31=0.4,d41=0.1,μ1=0.6,c1=0.006,d12=0.4,d22=0.1,d32=0.3,d42=0.2,μ2=0.7,c2=0.006. $
|
Then, we take the initial functions as
$ S01(x)=0.8×(1+0.3×sinπx),e01(a,x)=1.2×e−a,I01(x)=0.5×(1.5+0.2×sinπx),r01(a,x)=1.5×e−a,S02(x)=0.7×(1+0.2×sinπx),e02(a,x)=1.3×e−a,I02(x)=0.6×(1.5+0.3×sinπx),r02(a,x)=1.4×e−a. $
|
Example 5.1. If we choose other parameters as
$ Λ1=3,β11=0.04,β12=0.07,δ11=0.2,δ21=0.9,Λ2=2,β21=0.06,β22=0.05,δ12=0.5,δ22=0.8,σ1(a)=0.003×(1+sin(a−5)π2),γ1(a)=0.1×(1+sin(a−5)π2),σ2(a)=0.002×(1+sin(a−5)π2),γ2(a)=0.2×(1+sin(a−5)π2), $
|
(5.1) |
then we have
$ ∫∞0α1σ1(θ)π11(θ)dθ≈0.00051<1,∫∞0γ1(θ)π21(θ)dθ≈0.0817<1,∫∞0α2σ2(θ)π12(θ)dθ≈0.00027<1,∫∞0γ2(θ)π22(θ)dθ≈0.13319<1. $
|
In this case, from Theorem 4.2, we expect the disease-free steady state $ \bar{E} $ to be globally asymptotically stable. In fact, in Figure 1, the infective population $ I_1(t, x) $ and $ I_2(t, x) $ converge to zero over time.
Example 5.2. If we choose other parameters as
$ Λ1=5,β11=0.99,β12=0.97,δ11=0.002,δ12=0.005,Λ2=7,β21=0.96,β22=0.98,δ21=0.009,δ22=0.008,σ1(a)=35×(1+sin(a−5)π2),γ1(a)=0.1×(1+sin(a−5)π2),σ2(a)=30×(1+sin(a−5)π2),γ2(a)=0.2×(1+sin(a−5)π2), $
|
(5.2) |
then we have $ \rho(M^0)\approx3.89217 > 1 $. In this case, from Theorem 4.3, we expect the space-independent steady state $ E^* $ to be globally asymptotically stable. In fact, in Figure 2, the infective population $ I_1(t, x) $ and $ I_2(t, x) $ converge to the positive distribution over time.
In this paper, as an additional structure of the system, we focus on the spatial diffusion of the population. Models with spatial diffusion allow individuals to move to adjacent positions through a random walk process, this is a key factor in considering the geographical spread of infectious diseases. Firstly, we propose the $ n $-group diffusive SEIR epidemic model with age-dependent latent and relapse, it is a generalization of the model in [27] to a spatially diffusive system. Then, we investigate the positivity, boundedness, existence and uniqueness of solution and the existence of compact global attractor of the associated solution semiflow for this system. For these results, we use the method of constructing Picard sequences, Banach-Picard fixed point theorem and theories of partial functional differential equations. Thereafter, we establish the existence of disease-free and endemic steady states based on Perron-Frobenius theorem. we utilize appropriate Lyapunov functionals, graph-theoretical results and the LaSalle's invariance principle to prove the global stability of disease-free and endemic steady states. Thus, we presented the results of numerical simulations to verify the validity of our main theorems. This is important from the viewpoint of applications.
In this epidemic model, we are concerned with two kinds of spatial heterogeneity: the patch structure and spatial diffusion. Furthermore, age-of-latent and age-of-relapse are included into the epidemic model which is appropriate for diseases such as tuberculosis and herpes virus infection. Dynamical results obtained in this paper provide theoretical foundation for seeking effective measures to prevent, control and study disease transmission.
The expressions of basic reproduction number and endemic steady state depends on space are not analyzed in this paper owing to the complexity of model. In addition, how to improve the sufficient conditions that ensure the stabilities of steady states and make them be depended on basic reproduction number is also need to investigate. We leave these issues for future research.
The authors are very grateful to the editor and the anonymous referees for their valuable comments and suggestions, which helped us to improve the presentation of this work significantly.
This work was supported by the National Natural Science Foundation of China (Grant Number: 11971329).
The authors declare there is no conflict of interest.
[1] | Saunders, Philadelphia, (2003), 243-274. |
[2] | Physical Review E, 85 (2012), p. 031911. |
[3] | Journal of Theoretical Biology, 241 (2006), 307-320. |
[4] | Physical Review, 109 (1958), p. 1492. |
[5] | Journal of Theoretical Biology, 139 (1989), 311-326. |
[6] | John Wiley & Sons, 2008. |
[7] | The Journal of Experimental Medicine, 183 (1996), 2259-2269. |
[8] | Physical Review E, 86 (2012), p. 031146. |
[9] | Ecology, (1992), 1530-1535. |
[10] | Theoretical Population Biology, 65 (2004), 299-315. |
[11] | in "Proceedings of the Cambridge Philosophical Society," 53 1957, 629-641. |
[12] | Immunology Letters, 65 (1999), 93-98. |
[13] | Ecological Complexity, 4 (2007), 242-249. |
[14] | Theoretical Population Biology, 53 (1998), 108-130. |
[15] | Journal of Mathematical Biology, 34 (1996), 579-612. |
[16] | Journal of Theoretical Biology, 176 (1995), 91-102. |
[17] | Physical Review Letters, 101 (2008), p. 258102. |
[18] | Journal of Theoretical Biology, 227 (2004), 535-545. |
[19] | Journal of Differential Equations, 203 (2004), 331-364. |
[20] | Annals of Human Genetics, 7 (1937), 355-369. |
[21] | Applicable Analysis, 31 (1989), 247-266. |
[22] | Bulletin of Mathematical Biology, 48 (1986), 493-508. |
[23] | 2002. |
[24] | Science, 79 (1934), p. 340. |
[25] | Zeitschrift fur Physik B Condensed Matter, 47 (1982), 365-374. |
[26] | Journal of Physics A: Mathematical and General, 22 (1989), 3673-3679. |
[27] | Physical Review E, 55 (1997), p. 2488. |
[28] | Journal of Physics A: Mathematical and General, 17 (1999), p. L105. |
[29] | Journal of Mathematical Biology, 5 (1977), 399-403. |
[30] | Physica A: Statistical Mechanics and its Applications, 289 (2001), 178-190. |
[31] | Advances In Physics, 49 (2000), 815-958. |
[32] | Ecology Letters, 8 (2004), 102-116. |
[33] | Garland Publ., Inc., New York, NY, 1997. |
[34] | Zeitschrift für Physik B Condensed Matter, 42 (1981), 151-154. |
[35] | Physica A: Statistical Mechanics and its Applications, 203 (1994), 175-188. |
[36] | Nonlinear Analysis: Theory, Methods & Applications, 28 (1997), 145-164. |
[37] | International Journal of Modern Physics C, 17 (2006), 1647-1662. |
[38] | Open ISBN, (2007). |
[39] | Bulletin of Mathematical Biology, 53 (1991), 33-55. |
[40] | Electron. J. Probab., 8 (2003), 1-51. |
[41] | Mosc. Univ. Bull. Math, 1 (1937), 1-25. |
[42] | Cambridge University Press, 2001. |
[43] | Ecology, 90 (2009), 802-811. |
[44] | Biometrika, (1960), 219-234. |
[45] | Biophysical Journal, 87 (2004), 75-80. |
[46] | Journal of the American Chemical Society, 42 (1920), 1595-1599. |
[47] | Proceedings of the National Academy of Sciences of the United States of America, 8 (1922), p. 147. |
[48] | Williams & Wilkins Baltimore, 1925. |
[49] | Physica A, 297 (2001), 242-252. |
[50] | preprint, arXiv:nlin/0006043, (2000). |
[51] | Physica A: Statistical Mechanics and its Applications, 297 (2001), 242-252. |
[52] | Bulletin of Mathematical Biology, 65 (2003), 375-396. |
[53] | Artificial Life, 9 (2003), 357-370. |
[54] | New York: Penguin, (1798). |
[55] | Physical Review E, 73 (2006), p. 040903. |
[56] | Stochastic models in biological sciences, 253-257, Banach Center Publ., 80, Polish Acad. Sci. Inst. Math., Warsaw, 2008. |
[57] | 1989. |
[58] | Physical Review E, 58 (1998), 1383-1403. |
[59] | Mathematical Biosciences, 110 (1992), 45-66. |
[60] | Theoretical Population Biology, 48 (1995), 7-43. |
[61] | in "Proceedings of the Zoological Society of London," 105, Wiley Online Library, (1935), 551-598. |
[62] | Springer, New York, 1980, xiii+254. |
[63] | Veterinary Immunology and Immunopathology, 96 (2003), 193-205. |
[64] | Wiley, New York, 1998. |
[65] | Physical Review E, 68 (2003), p. 046121. |
[66] | Science, 171 (1971), 385-387. |
[67] | Physics Letters A, 280 (2001), 45-52. |
[68] | Mathematical Biosciences, 206 (2007), 108-119. |
[69] | The Journal of Immunology, 182 (2009), 6379-6393. |
[70] | Physica D: Nonlinear Phenomena, 117 (1998), 145-166. |
[71] | Journal of the Royal Society Interface, 5 (2008), 483-505. |
[72] | Physical Review E, 63 (2001), p. 021103. |
[73] | Proceedings of the National Academy of Sciences, 97 (2000), 10322-10324. |
[74] | Biometrika, (1951), 196-218. |
[75] | Physical review letters, 99 (2007), 234503. |
[76] | Memoires de lAcademie Royale des Sciences et des Belles-Lettres de Bruxelles, 18 (1845), 1-45. |
[77] | Mem. Acad, Lincei 22, 31, 113 (1926). |
[78] | Journal of Physics: Condensed Matter, 19 (2007), 065139. |
[79] | Physica D: Nonlinear Phenomena, 237 (2008), 2553-2562. |
1. | Yi Li, Mandy Chen, Joshy George, Edison T. Liu, R. Krishna Murthy Karuturi, Adaptive Sentinel Testing in Workplace for COVID-19 Pandemic, 2022, 1557-8666, 10.1089/cmb.2022.0291 | |
2. | Peng Wu, Zhaosheng Feng, Xuebing Zhang, Global dynamics of an age–space structured HIV/AIDS model with viral load-dependent infection and conversion rates, 2022, 412, 03770427, 114309, 10.1016/j.cam.2022.114309 | |
3. | Yifan Xing, Hong-Xu Li, Almost periodic solutions for a SVIR epidemic model with relapse, 2021, 18, 1551-0018, 7191, 10.3934/mbe.2021356 | |
4. | Yanlin Ding, Jianjun Jiao, Global analysis and control for a vector-borne epidemic model with multi-edge infection on complex networks, 2023, 21, 2391-5455, 10.1515/math-2022-0580 | |
5. | Victoria Chebotaeva, Paula A. Vasquez, Erlang-Distributed SEIR Epidemic Models with Cross-Diffusion, 2023, 11, 2227-7390, 2167, 10.3390/math11092167 | |
6. | Peng Wu, Ran Zhang, Anwarud Din, Mathematical analysis of an age-since infection and diffusion HIV/AIDS model with treatment adherence and Dirichlet boundary condition, 2023, 214, 03784754, 1, 10.1016/j.matcom.2023.06.018 | |
7. | Zeyan Zhang, Chunmei Zhang, 2024, Exponential stability of stochastic multi-group models with regime-switching diffusion under the local conditions, 978-9-8875-8158-1, 1171, 10.23919/CCC63176.2024.10662306 |