Loading [MathJax]/jax/output/SVG/jax.js
Research article Special Issues

Some dynamic Hardy-type inequalities with negative parameters on time scales nabla calculus

  • Received: 13 December 2023 Revised: 08 January 2024 Accepted: 19 January 2024 Published: 25 January 2024
  • MSC : 26D10, 26D15, 34N05, 47B38, 39A12

  • In this paper, we establish some new dynamic Hardy-type inequalities with negative parameters on time scales nabla calculus by applying the reverse H ölder's inequality, integration by parts, and chain rule on time scales nabla calculus. As special cases of our results (when T=R), we get the continuous analouges of inequalities proven by Benaissa and Sarikaya, and when T=N0, the results to the best of the authors' knowledge are essentially new.

    Citation: Elkhateeb S. Aly, Y. A. Madani, F. Gassem, A. I. Saied, H. M. Rezk, Wael W. Mohammed. Some dynamic Hardy-type inequalities with negative parameters on time scales nabla calculus[J]. AIMS Mathematics, 2024, 9(2): 5147-5170. doi: 10.3934/math.2024250

    Related Papers:

    [1] Adel M. Al-Mahdi, Mohammad M. Al-Gharabli, Nasser-Eddine Tatar . On a nonlinear system of plate equations with variable exponent nonlinearity and logarithmic source terms: Existence and stability results. AIMS Mathematics, 2023, 8(9): 19971-19992. doi: 10.3934/math.20231018
    [2] Adel M. Al-Mahdi . The coupling system of Kirchhoff and Euler-Bernoulli plates with logarithmic source terms: Strong damping versus weak damping of variable-exponent type. AIMS Mathematics, 2023, 8(11): 27439-27459. doi: 10.3934/math.20231404
    [3] Mohammad M. Al-Gharabli, Adel M. Al-Mahdi, Mohammad Kafini . Global existence and new decay results of a viscoelastic wave equation with variable exponent and logarithmic nonlinearities. AIMS Mathematics, 2021, 6(9): 10105-10129. doi: 10.3934/math.2021587
    [4] Sarra Toualbia, Abderrahmane Zaraï, Salah Boulaaras . Decay estimate and non-extinction of solutions of p-Laplacian nonlocal heat equations. AIMS Mathematics, 2020, 5(3): 1663-1679. doi: 10.3934/math.2020112
    [5] Salim A. Messaoudi, Mohammad M. Al-Gharabli, Adel M. Al-Mahdi, Mohammed A. Al-Osta . A coupled system of Laplacian and bi-Laplacian equations with nonlinear dampings and source terms of variable-exponents nonlinearities: Existence, uniqueness, blow-up and a large-time asymptotic behavior. AIMS Mathematics, 2023, 8(4): 7933-7966. doi: 10.3934/math.2023400
    [6] Adel M. Al-Mahdi, Mohammad M. Al-Gharabli, Maher Nour, Mostafa Zahri . Stabilization of a viscoelastic wave equation with boundary damping and variable exponents: Theoretical and numerical study. AIMS Mathematics, 2022, 7(8): 15370-15401. doi: 10.3934/math.2022842
    [7] Abdelbaki Choucha, Salah Boulaaras, Asma Alharbi . Global existence and asymptotic behavior for a viscoelastic Kirchhoff equation with a logarithmic nonlinearity, distributed delay and Balakrishnan-Taylor damping terms. AIMS Mathematics, 2022, 7(3): 4517-4539. doi: 10.3934/math.2022252
    [8] Adel M. Al-Mahdi, Mohammad M. Al-Gharabli, Mohamed Alahyane . Theoretical and numerical stability results for a viscoelastic swelling porous-elastic system with past history. AIMS Mathematics, 2021, 6(11): 11921-11949. doi: 10.3934/math.2021692
    [9] Shuai Li, Tianqing An, Weichun Bu . Existence results for Schrödinger type double phase variable exponent problems with convection term in RN. AIMS Mathematics, 2024, 9(4): 8610-8629. doi: 10.3934/math.2024417
    [10] Yousif Altayeb . New scenario of decay rate for system of three nonlinear wave equations with visco-elasticities. AIMS Mathematics, 2021, 6(7): 7251-7265. doi: 10.3934/math.2021425
  • In this paper, we establish some new dynamic Hardy-type inequalities with negative parameters on time scales nabla calculus by applying the reverse H ölder's inequality, integration by parts, and chain rule on time scales nabla calculus. As special cases of our results (when T=R), we get the continuous analouges of inequalities proven by Benaissa and Sarikaya, and when T=N0, the results to the best of the authors' knowledge are essentially new.



    Malaria is a lethal infectious disease caused by Plasmodium parasites, which infects humans through the bite of infected female geneus Anopheles mosquitoes. It is endemic in about 100 countries and an estimated 3.3 billion people are at risk of being infected in the word [1]. Therefore, it is necessary to study the spread of malaria in the population. The pioneering work was given by Ross [2] which showed that malaria is transmitted by mosquitoes and proposed the first mathematical model of this disease. Subsequently, Macdonald promoted this [3], which plays a particularly important role in understanding the disease. However, the Ross-Macdonald model is extremely simplified, ignoring many critical factors of real-world ecology and epidemiology [4]. One omission is spatial heterogeneity, for example, the distribution of urban and rural areas and the spatial movement of mosquitoes and people, which increases the spatial transmission of malaria. Hence, in the malaria epidemiological model, spatial heterogeneity must be considered [5,6,7]. In 1951, Skellam specially examined the reaction-diffusion model of species population density in a bounded habitat [8]. One advantage of reaction-diffusion models is that they can easily incorporate simple rules about edge effects. Use these features of reaction-diffusion model to study how environmental heterogeneity affects malaria [9,10,11]. However, in these studies, it seems that the following important biological aspects are rarely paid, simultaneously.

    First, "vector-bias" is a typical feature of malaria transmission, which describes the different probabilities of a mosquito bitting humans. Lacroix et al. [12] proved that mosquitoes are generally more inclined to bite infected individuals. To describe this preference, Chamchod and Britton incorporated the vector-bias effect into malaria model [13]. Subsequently, a wealth of papers take into account the effect of vector-bias when studying malaria transmission [14,15,16]. Due to the different densities of mosquitoes in various regions, the infectivity of malaria is also different. The analysis of vector-bias effects is of great significance for understanding the dynamics of malaria transmission.

    Secondly, in recent years, climate has become more fluctuating [17], and the rise in temperature is related to global warming. Climate warming is directly connected with the number of Anopheles, the main vector of malaria. Because temperature affects the immature development rate, immature stage survival rate, adult size, adult lifespan, blood feeding, and reproduction ability of mosquitoes [18]. It is widely accepted that both the life cycle of mosquitoes and the development of parasites are extremely sensitive to temperature. Empirical evidence indicates that the life cycle of mosquitoes experiences four different stages: eggs, larvae, pupae and adults. The first three stages are also called aquatic or immature stages [19]. In addition, a very important process in the development of parasites is the extrinsic incubation period (EIP), i.e., mosquitoes may not be able to transmit the disease to humans for a period of time after taking infected blood meal. The lifespan of female adult mosquitoes is generally 3 to 100 days, and EIP can reach 30 days [20]. These infected mosquitoes that live pass EIP will keep infectious until they die. There is considerable evidence that temperature rise will shorten the length of each immature stage [21] and EIP is extremely sensitive to seasonal changes in temperature [22,23]. In view of climate change, exploring the effect of temperature in the spread of malaria is particularly essential. Therefore, these seasonal forced EIP and mosquito life cycles need to be considered in the malaria model.

    Note that mathematical models of malaria transmission rarely consider key features such as vector-bias, climate/seasonality, temperature-dependent mosquito life cycle and EIP in a spatially heterogeneous environment, simultaneously. A reaction-diffusion model with periodic delays is proposed to study the influence of the above factors on the transmission of malaria. This work is inspired by the following biological questions: (I) What is the impact of vector-bias on malaria transmission? (II) What effect does the maturity and incubation period of seasonal temperature changes have on the spread of malaria? (III) Does the movement of mosquitoes and humans have different effects for the transmission of malaria in different areas? These issues that are closely related to the spread and control of malaria need to be explored in more depth. How to add the temperature-dependent maturity period to the model is the first issue we consider, and the detailed model derivation is also the difficulty of this paper. The existence of temperature-dependent periodic delays makes it difficult to understand the asymptotic behavior of the model. We consider the stage structure, because the mosquitoes in the aquatic stage cannot fly, which makes it impossible for us to directly use the theoretical results in references [22,24]. Our theoretical results indicate the model has a threshold parameter, the basic reproduction number R0, which determines wether the malaria can persist or become extinct in susceptible populations.

    The structure of the rest of this paper is shown below. Section 2 establishes human and mosquito populations models and malaria transmission model. This section also gives the biological significance of these parameters in models. Section 3 is devoted to well-posedness and some properties of the model. In Section 4, we first define the basic reproduction number R0, and then analyze the threshold dynamics of models in terms of R0. Section 5 gives the basic framework of the system threshold dynamics through numerical simulation, and discusses the influence of parameters on basic reproduction number, mosquitoes and humans in the process of malaria transmission. A brief discussion section concludes the paper.

    Motivated by models in references [19,25,26], we introduce a seasonal stage-structured model of malaria to explain the cross-infection between mosquitoes and individuals. In this subsection, we first consider the mosquito population. According to the biological cycle, mosquitoes can go through several stages and have different habits at various stages. We divide them into immature aquatic stage M(t,x) and mature stage V(t,x) at time t0 and location x in a spatial bounded domain ΩRn with smooth boundary Ω. Let η(t,a,x) denote the density of the mosquito population at time t, chronological age a0, and location x. Metz and Diekmann [27] gave

    {(t+a)η(t,a,x)=D(a)Δη(t,a,x)d(t,a,x)η(t,a,x),η(0,a,x)=ˆϕ(a,x),a0,xˉΩ,η(t,0,x)=Λv(t,x),t0,xˉΩ, (2.1)

    through the standard argument for structured populations and spatial diffusion. Here, Δ denotes the usual Laplacian operator. ˆϕ(,x) represents the non-negative initial age and location distribution of mosquitoes, the term Λv(t,x) is the recruitment rate of the immature stage, which is Hölder continuous and nonnegative function on R×ˉΩ, D(a) is the age-related diffusion coefficient, and d(t,a,x) indicates the death rate of mosquitoes with time t, age a and location x, which is Hölder continuous and positive on R×ˉΩ. Λv(t,) and d(t,,) are periodic in t and have the same period ω.

    Let us first assume some age thresholds so that the growth of mosquitoes can be divided into discrete age stages. When modeling species dynamics with a stage structure, we divide mosquitoes into immature stage (also known as aquatic stage) and mature stage, the time delay is the mature period. The temperature T(t) is an ω-periodic function of time t. Generally speaking, the maturity age is determined by seasonal changes in temperature. We will use a time-dependent positive function τ1(t) to describe the duration from newborn to adult. That is, an individual at time t becomes mature only if its age exceeds τ1(t), in other words, the adult mosquitoes that develop at time t enter the aquatic stage at tτ1(t). We assume that the maximum chronological age of aquatic mosquitoes at time t is A(t). Here and in the following, τ1(t) and A(t) are C1 (a section of continuous differentiable functions) periodic functions of time t with the same periodic ω. The relationship between time-dependent threshold age and time-dependent delay can be indicated as follows:

    A(t)=τ1(t).
    Figure 1.  At time t the mosquitoes reach the threshold age A(t) and develop into the mature stage.

    Here we introduce another quantity f to describe the maturity level of mosquitoes. Let f=fM=0, that is, the maturity level of mosquitoes that have just entered the aquatic stage is 0. At time t, the mosquitoes reach the maximum chronological age A(t) of the aquatic stage. At this time, the maturity level is also the maximum, fV. Then one has

    fVfM=ttτ1(t)ˆρ(r1)dr1

    where ˆρ(r1) is the proportion of maturity at time r1. Taking the derivative of the above formula with respect to t, we can get

    1τ1(t)=ˆρ(t)ˆρ(tτ1(t)),

    and hence 1τ1(t)>0. Recall that τ1(t) represents the time it takes for a mosquito to go from the aquatic stage to the adult stage at time t. For the case where δ>0 is small enough, if the mosquito is an adult mosquito at time t, then it is still an adult mosquito at time t+δ. Thus τ1(t+δ)<τ1(t)+δ, and hence, τ1(t)=limδ0τ1(t+δ)τ1(t)δ1. This condition also makes sense biologically. 1τ1(t)>0, is called maturation rate [26], which excludes the possibility of mosquitoes returning from the mature stage to the immature stage unless they are born. Mathematically speaking, tτ1(t) is strictly increased with respect to t, that is, since the developmental rate depends only on time, it is impossible for juveniles to reach maturity before they are born.

    M(t,x) and V(t,x) denote the density of mosquitoes in the aquatic and mature stages at time t and location x, respectively. The accumulative density between the two age thresholds can be recorded as the population size at each stage. In particular, for the aquatic stage M and the mature stage V (for the sake of simplicity, we use M and V instead of M(t,x) and V(t,x), respectively), we have the following mathematical expressions:

    M=A(t)0η(t,a,x)da,V=A(t)η(t,a,x)da. (2.2)

    Next, we put forward natural biological assumptions about birth rate, death rate and diffusion coefficient in Eq (2.1). Since immature mosquitoes cannot fly, we assume that the diffusion coefficient satisfies

    D(a)={0,0a<A(t),Dv,aA(t).

    At each age stage, all mosquitoes undergo the same birth rate and death rate which are age-independent, then the natural death rates in the aquatic and mature stages are μm(t,x) and μv(t,x), respectively, i.e.,

    d(t,a,x)={μm(t,x),0a<A(t),μv(t,x),aA(t),

    where μm(t,x) and μv(t,x) are non-negative non-trivial and Hölder continuous on R×ˉΩ and ω-periodic in t. Differentiating the equations in system (2.2) with respect to time t on both sides yields

    Mt=A(t)0η(t,a,x)tda+η(t,A(t),x)A(t)=μm(t,x)Mη(t,A(t),x)+η(t,0,x)+η(t,A(t),x)A(t),

    and

    Vt=A(t)η(t,a,x)tdaη(t,A(t),x)A(t)=DvΔVμv(t,x)Vη(t,,x)+(1A(t))η(t,A(t),x).

    In a biological sense, because no mosquito can live forever, η(t,,x)=0. η(t,0,x) represents the inflow rate of mosquitoes in the aquatic stage at time t, then η(t,0,x)=Λv(t,x) is a function of t.

    {Mt=Λv(t,x)μm(t,x)M(1A(t))η(t,A(t),x),Vt=DvΔVμv(t,x)V+(1A(t))η(t,A(t),x).

    Mathematically, assume that the delay τ1(t) is continuously differentiable in [0,), bounded and far away from zero and infinity. The expression of η(t,A(t),x) in the above system will be obtained by integrating along characteristics. For tA(t) and t<A(t), η(t,A(t),x) has different expressions. Without loss of generality, we study tA(t) (see, [28]), which is practicable due to the boundedness of A(t), because we are concerned with long-term behavior of population dynamics. Here, η(tτ1(t),0,)=Λv(tτ1(t),). Let t=t0+h, a=a0+h, P(h,x)=η(t0+h,a0+h,x) and ˆd(h,x)=d(t0+h,a0+h,x), then

    P(h,x)h=η(t0+h,a0+h,x)hdhdt+η(t0+h,a0+h,x)hdhda=(t+a)η(t,a,x)=DvΔP(h,x)ˆd(h,x)P(h,x). (2.3)

    Integrating Eq (2.3) from h1 to h2 yields

    P(h2,x)=P(h1,x)eh2h1ˆd(r2,)dr2.

    For tA(t), setting t0=tA(t), h=A(t) and a0=0, we have

    η(t,A(t),x)=η(tA(t),0,x)ettA(t)μm(r2,x)dr2=Λv(tτ1(t),x)ettτ1(t)μm(r2,x)dr2.

    Since we are concerned about the long-term behavior of population dynamics, the model for tA(t) is as follows

    {Mt=Λv(t,x)μm(t,x)M(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r2,x)dr2,xΩ,Vt=DvΔVμv(t,x)V+(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r2,x)dr2,xΩ,Vν=0,xΩ, (2.4)

    where ν is the outward unit normal vector on Ω and ν indicates the normal derivative along ν on Ω, ettτ1(t)μm(r2,x)dr2 represents probability surviving through natural death during aquatic stage. It is easy to see that the equation V(t,x) is decoupled, so we can get the following system

    {Vt=DvΔVμv(t,x)V+(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r2,x)dr2,xΩ,Vν=0,xΩ. (2.5)

    for any tA(t). Alternatively, according to biological significance, M and V can be written as integral form. Note that τ1(t) is the maturation time of aquatic mosquitoes at time t. Thence, aquatic mosquitoes at time t contain all newly born mosquitoes at a previous time ξ1 with ξ1(tτ1(t),t) and survive to time t. Hence, one has

    M=ttτ1(t)Λv(ξ1,x)etξ1μm(r2,x)dr2dξ1,xˉΩ. (2.6)

    Similarly, for any xˉΩ,

    V=ΩΓ(t,0,x,y)V(0,y)dy+t0ΩΓ(t,ξ1,x,y)Λv(ξ1τ1(ξ1),y)(1τ1(ξ1))eξ1ξ1τ1(ξ1)μm(r2,)dr2dydξ1, (2.7)

    where V(0,) is the initial data of V(t,).

    Figure 2.  Experience of mature mosquitoes at time t, age a.

    The first term in Eq (2.7) represents those mosquitoes that were adults at time 0 at position y and are still in V class at time t at position x by diffusion. The second term in Eq (2.7) denotes the total density of mosquitoes at time t. These mosquitoes were newly developed adults at time ξ1 and survived until time t. These new adults developed from the aquatic stage at previous time ξ1τ1(ξ1), which grown through τ1(ξ1) periodic of time and developed into adult stage at time ξ1 with "maturation rate" 1τ1(ξ1). Γ(t,t0,x,y) with tt00 is the Green function related to Vt=DvΔVμv(t,x)V that meets the Neumann boundary condition.

    Since the incubation period of the pathogen in the human is 1–12 days, which is short compared with the human life span, we do not consider the incubation period of the pathogen in the human body. In this subsection, inspired by [13,19,20,24,29], we give a model of malaria transmission, including the mobility of mosquitoes and individuals, vector-bias, temperature-dependent maturation delay and EIP. Because mosquitoes are the vectors of malaria transmission from person to person, the modeling takes into account the dynamics of human and mosquito populations. One supposes that individuals will become susceptible again after recovery. Thus, humans include two epidemiological types: susceptible (Sh) and infected (Ih) compartments. Assume that all populations walk unbiased and random. Since climate change has almost no impact on human birth and death, according to [29], the following reaction-diffusion equation is suitable:

    Nh(t,x)t=DhΔNh(t,x)+Λh(x)μh(x)Nh(t,x),t>0,xˉΩ. (2.8)

    Dh>0 is the human diffusion coefficient. Suppose the dispersal pattern is random diffusion, that is, individual walkers use a fixed step to walk randomly on the solid line [30,31]. Λh(x) and μh(x) are the recruitment rate and natural death rate of human at location x, respectively. In particular, we should note that the disease dose not have a significant impact on the mobility of individuals and mosquitoes. It is mathematically assumed that the diffusion coefficient Dh>0 of all individuals is the same, and the diffusion coefficient Dv>0 of all mosquitoes is recognized as the same. Assuming that no population flow crosses the boundary Ω, therefore, the Neumann boundary condition is imposed:

    Nh(t,x)ν=0,t>0,xΩ, (2.9)

    here, the definition of ν is same as Section 2.1. Following Theorems 3.1.5 and 3.1.6 in reference [32] and Section 2 in reference [29], we realize that the positive steady state N(x)C(ˉΩ,R+){0} of systems (2.8) and (2.9) is globally attractive, where C(ˉΩ,R+) is Banach space of continuous functions from ˉΩ to R+. According to Appendix 8.1 in reference [33], it is assumed that the total human density of at time t and location x is stable at N(x), i.e., Nh(t,x)N(x), for t0 and xΩ. Then we know that Sh=N(x)Ih. Here and later, for the sake of simplicity, we use Sh and Ih instead of Sh(t,x) and Ih(t,x), respectively.

    From a biological and epidemiological point of view, immature mosquitoes do not take part in infection, and are primarily in a waiting period, because they cannot fly and bite humans. Since adult male mosquitoes do not eat blood meal, they cannot transmit and acquire the virus, so only adult female mosquitoes are modeled. Mature female mosquitoes are divided into susceptible (Sv(t,x)), exposed (Ev(t,x)) and infectious (Ev(t,x)), t is time and x denotes location. The exposed mosquitoes are already infected, but not infectious. From Section 2.1, we know that V=Sv+Ev+Iv is the total density of female adult mosquitoes and satisfies reaction-diffusion system (2.5). Here and later, for the sake of simplicity, we use V, Sv, Ev and Iv instead of V(t,x), Sv(t,x), Ev(t,x) and Iv(t,x), respectively.

    Vector-bias effect is a situation in which mosquitoes display the weight of preference when choosing a host. In order to add a vector-bias term to the model, parameters p and l are introduced, which are the probability of mosquitoes arriving at humans randomly, and select infectious and susceptible humans, respectively [13]. Since mosquitoes are more attracted to infected humans, we assume that the vector-bias parameter χ=pl1. To comprehend the influences of spatial structure and climate change, we allow a space-dependent bite rate β(t,x) of female mosquitoes to be Hölder continuous non-negative non-trivial on R×ˉΩ with β(t,x)0, representing the number of mosquitoes biting per unit time at location x, and β(t,) is a ω-period function of time t. Let B represent the mosquitoes bite incident, and A be the event of mosquitoes reaching humans infected with malaria. Then P(A)=IhN(x), P(Ac)=1IhN(x), P(B|A)=p and P(B|Ac)=l. So the probability that an infectious human is chosen as P(A|B),

    then one has

    P(Ac|B)=l[N(x)Ih]pIh+l[N(x)Ih]andP(A|B)=pIhpIh+l[N(x)Ih],

    which are the probability of infected individuals and susceptible individuals are bitten. Then at time t and location x, the numbers of newly infected individuals and mosquitoes per unit time are

    cβ(t,x)l[N(x)Ih]IvpIh+l[N(x)Ih]andαβ(t,x)pIhSvpIh+l[N(x)Ih],

    respectively, where c is the probability that an infected mosquito will spread to susceptible individuals per bite, and α is probability that an infected individual will transmission to susceptible mosquitoes per bite.

    During the malaria transmission cycle, the EIP of parasite in mosquitoes is one of the most pivotal parameters. Mosquitoes can fly around during EIP. The malaria parasites undergo different growth stages before migrating to the salivary glands where can spread to humans. The degree of this development is related to environmental factors, such as temperature. EIP is exceedingly sensitive to temperature. Therefore, it is necessary to incorporate this seasonal forced incubation period when describing malaria transmission. We will use a time-dependent positive function τ2(t) to describe the duration from new infection to infectiousness. That is, a mosquito is infectious at time t only if its age of infection exceeds τ2(t), where τ2(t) is a C1 periodic function in [0,) with the periodic ω, and tτ2(t) is strictly increasing in t.

    Based on the above discussions, we will pay attention to the dynamics of the following malaria transmission reaction-diffusion model

    {Mt=Λv(t,x)(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r2,x)dr2μm(t,x)M,xΩ,Svt=DvΔSv+(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r2,x)dr2αβ(t,x)pIhSvpIh+l[N(x)Ih]μv(t,x)Sv,xΩ,Evt=DvΔEvEm(t,x)μv(t,x)Ev+αβ(t,x)pIhSvpIh+l[N(x)Ih],xΩ,Ivt=DvΔIv+Em(t,x)μv(t,x)Iv,xΩ,Iht=DhΔIh(μh(x)+ϱ(x))Ih+cβ(t,x)l[N(x)Ih]IvpIh+l[N(x)Ih],xΩ,Svν=Evν=Ivν=Ihν=0,xΩ, (2.10)

    for t>0. Similar to the derivation in Section 2.1, we get

    Em(t,x)=(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)pIh(tτ2(t),y)Sv(tτ2(t),y)pIh(tτ2(t),y)+l[N(y)Ih(tτ2(t),y)]dy.

    Substituting Em(t,x) into system (2.10), one has

    {Mt=Λv(t,x)(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r,x)drμm(t,x)M,xΩ,Svt=DvΔSv+(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r,x)drαβ(t,x)pIhSvpIh+l[N(x)Ih]μv(t,x)Sv,xΩ,Evt=DvΔEv+αβ(t,x)pIhSvpIh+l[N(x)Ih]μv(t,x)Ev(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)pIh(tτ2(t),y)Sv(tτ2(t),y)pIh(tτ2(t),y)+l[N(y)Ih(tτ2(t),y)]dy,xΩ,Ivt=DvΔIvμv(t,x)Iv+(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)pIh(tτ2(t),y)Sv(tτ2(t),y)pIh(tτ2(t),y)+l[N(y)Ih(tτ2(t),y)]dy,xΩ,Iht=DhΔIh(μh(x)+ϱ(x))Ih+cβ(t,x)l[N(x)Ih]IvpIh+l[N(x)Ih],xΩ,Svν=Evν=Ivν=Ihν=0,xΩ, (2.11)

    for t>0, where α, p, l and c are positive constants, and ϱ(x) denotes the recovery rate of humans in position x, which is Hölder continuous and positive on Ω. Neumann boundary condition indicates that all individuals and mosquitoes still remain in this domain Ω.

    Since M(t,x) and Ev(t,x) of the system (2.11) are decoupled from the other equations, it is sufficient to explore the system

    {u1t=DvΔu1μv(t,x)u1+(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r,x)drαβ(t,x)pu3u1pu3+l[N(x)u3],xΩ,u2t=DvΔu2μv(t,x)u2+(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)pu3(tτ2(t),y)u1(tτ2(t),y)pu3(tτ2(t),y)+l[N(y)u3(tτ2(t),y)]dy,xΩ,u3t=DhΔu3+cβ(t,x)l[N(x)u3]u2pu3+l[N(x)u3](μh(x)+ϱ(x))u3,xΩ,u1t=u2t=u3t=0,xΩ, (2.12)

    for t>0, where (u1,u2,u3)=(Sv,Iv,Ih).

    In this section, the well-posedness of the model is proved. Let X1:=C(ˉΩ,R3) be the Banach space of continuous functions from ˉΩ to R3 the supremum norm be X, and X+1:=C(ˉΩ,R3+). Let ˆτ1:=maxt[0,ω]τ1(t), ˆτ2:=maxt[0,ω]τ2(t), and ˆτ:=max{ˆτ1,ˆτ2}. Define X1:=C([ˆτ,0],X1) to be a Banach space, with the norm ϕ=maxθ[ˆτ,0]ϕ(θ)X1, ϕX1, and X+1:=C([ˆτ,0],X+1). Then (X1,X+1) and (X1,X+1) are order Banach spaces. A function z:[ˆτ,σ)X1 for σ>0 is given and ztX1 is defined by

    zt(θ)=(z1(t+θ),z2(t+θ),z3(t+θ)),θ[ˆτ,0],

    for any t[0,σ).

    Denote Y:=C(ˉΩ,R), Y+:=C(ˉΩ,R+), Yh:={φY+:0φ(x)N(x),xˉΩ}. Suppose that T1(t,s),T2(t,s):YY, are the evolution operators associated

    u1t=DvΔu1μv(t,x)u1(t,x):=A1(t)u1,

    and

    u3t=DhΔu3(μh(x)+ϱ(x))u3:=A2u3,

    obey the Neumann boundary condition, respectively. Since Ti(t,s)=Ti(ts) (i=1,2) and μv(t,) is ω-periodic in t, = [34,Lemma 6.1] indicates that Ti(t+ω,s+ω)=Ti(t,s) for (t,s)R2 with ts. Furthermore, Ti(t,s) is compact and strongly positive [35,Section 7.1 and Corollary 7.2.3] for (t,s)R2 with t>s (i=1,2). Let T(t,s)=diag{T1(t,s),T1(t,s),T2(t,s)}:X1X1, ts, is a strongly continuous semigroup and A(t)=diag{A1(t),A1(t),A2}, defined on D(A(t))=D(A1(t))×D(A1(t))×D(A2). A1(t) is defined by

    D(A1(t)):={φC2(ˉΩ):φν=0onΩ},A1(t)φ=DvΔφμv(t,x)φ,φD(A1(t)),

    and A2 is defined by

    D(A2):={φC2(ˉΩ):φν=0onΩ},A2φ=DhΔφ(μh(x)+ϱ(x))φ,φD(A2).

    Set

    Wh=C([ˆτ,0],Y+×Y+×Yh).

    Write F=(F1,F2,F3):[0,)×WhX1 as

    {F1(t,ϕ)=(1τ1(t))Λv(tτ1(t),)ettτ1(t)μm(r2,)dr2αβ(t,)pϕ3(0,)ϕ1(0,)pϕ3(0,)+l[N()ϕ3(0,)],F2(t,ϕ)=(1τ2(t))ΩΓ(t,tτ2(t),,y)αβ(tτ2(t),y)pϕ3(τ2(t),y)ϕ1(τ2(t),y)pϕ3(τ2(t),y)+l[N(y)ϕ3(τ2(t),y)]dy,F3(t,ϕ)=cβ(t,)l[N()ϕ3(0,)]ϕ2(0,)pϕ3(0,)+l[N()ϕ3(0,)], (3.1)

    for t0, xˉΩ and ϕ=(ϕ1,ϕ2,ϕ3)Wh. Rewrite system (2.12) as

    {dut=A(t)u+F(t,ut),t>0,u0=ϕ, (3.2)

    it then follows from references [36,Corollary 4 and Theorem 1] and [37,Corollary 8.1.3] that for any ϕWh, a mild solution can be get as a continuous solution of the integral equation:

    {u(t,ϕ)=T(t,0)ϕ(0)+t0T(t,s)F(s,us)ds,t>0,u0=ϕ. (3.3)

    Lemma 3.1. For any ϕWh, system (2.12) has the unique solution, denoted by z(t,,ϕ) on its maximal existence interval [0,tϕ) with z0=ϕ, where tϕ.Moreover, z(t,,ϕ)Y+×Y+×Yh for any t[0,tϕ), andthis solution is a classical solution of system (2.12) for all t>ˆτ.

    Before studying the global existence, go back to system (2.11) to make more information. We impose the compatibility conditions based on the biological meaning of τ1(t) and τ2(t):

    M(0,)=0τ1(0)Λv(ξ1,)e0ξ1μm(r2,)dr2dξ1, (3.4)

    and

    Ev(0,)=0τ2(0)T1(0,ξ1)αβ(ξ1,)pIh(ξ1,)Sv(ξ1,)pIh(ξ1,)+l[N()Ih(ξ1,)]dξ1. (3.5)

    Now we define

    D:={ψC([ˆτ,0],C(ˉΩ,R5+)):ψ5(s,)N(),s[ˆτ,0],ψ1(0,)=0τ1(0)Λv(ξ1,)e0ξ1μm(r2,)dr2dξ1,ψ3(0,)=0τ2(0)T1(0,ξ1)αβ(ξ1,)pIh(ξ1,)Sv(ξ1,)pIh(ξ1,)+l[N()Ih(ξ1,)]dξ1}.

    Next, for any ψD, system (2.11) has the unique solution U(t,)=(M(t,),Sv(t,),Ev(t,),Iv(t,),Ih(t,)) satisfying U0=ψ. Corollary 4 in reference [36] implies that Sv(t,x)0, Iv(t,x)0 and Ih(t,x)0 on maximal existence interval. According to the uniqueness of the solutions and the compatibility conditions (3.4) and (3.5), one has M(t,) (see(2.6)) and

    Ev(t,)=ttτ2(t)T1(t,ξ1)αβ(ξ1,)pIh(ξ1,)Sv(ξ1,)pIh(ξ1,)+l[N()Ih(ξ1,)]dξ1. (3.6)

    Hence, M(t,x)0 and Ev(t,x)0 on the maximal interval of existence.

    Lemma 3.2. For any ϕWh, system (2.12)has a unique solution u(t,,ϕ) on [0,) with u0=ϕ.Moreover, system (2.12) generates an ω-periodic semiflowQt:WhWh, defined by Qt:=ut(), t0, i.e., Qt(ϕ)(s,x)=u(t+s,x,ϕ), ϕWh, t0, s[ˆτ,0], xΩ, and Q:=Qω has a strong global attractor in Wh.

    The proof of Lemma 3.2 is given in Appendix A.

    Here, the basic reproduction number are introduced by the theory in reference [41], and then investigate the dynamics of system (2.12).

    The current main work is to explore the extinction and persistence of malaria in a spatially heterogeneous environment. The basic reproduction number R0 is one of the most necessary concepts in epidemiology. It is usually defined as the mean number of secondary infections during the entire period when a typical infected person is introduced into a completely susceptible population. From a biological point of view, the basic reproduction number is a characteristic of the risk of infection. More meaningfully, it describes the threshold behavior of many infectious disease models.

    Let u2=u3=0, we have the equation of mosquitoes compartment

    u1t=DvΔu1+(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r2,x)dr2μv(t,x)u1, (4.1)

    obeys the Neumann boundary condition. Based on reference [38,Lemma 2.1], it is easy to know that system (4.1) has a unique globally attractive positive ω-periodic solution u1(t,) in Y+{0}. Linearize system (2.12) at (u1(t,),0,0) and consider the equations of infectious compartments,

    {w1t=DvΔw1μv(t,x)w1+(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)pw2(tτ2(t),y)u1(tτ2(t),y)lN(y)dy,xΩ,w2t=DhΔw2+cβ(t,x)w1(μh(x)+ϱ(x))w2,xΩ,w1ν=w2ν=0,xΩ, (4.2)

    for t>0. Denote w:=(w1,w2)=(u2,u3).

    Set X2:=C(ˉΩ,R2), X+2:=C(ˉΩ,R2+), and Cω(R,X2) to be the Banach space consisting of all ω-periodic and continuous functions from R to X2, where ψCω(R,X2):=maxθ[0,ω]ψ(θ)X2 for any ψCω(R,X2). Let X2:=C([ˆτ2,0],X2) and X+2:=C([ˆτ2,0],X+2). Write F(t):X2X2 as

    F(t)(˜ψ1˜ψ2)=((1τ2(t))ΩΓ(t,tτ2(t),,y)αβ(tτ2(t),y)p˜ψ2(τ2(t),y)u1(tτ2(t),y)lN(y)dycβ(t,)˜ψ1(0,))

    for any tR, ˜ψ=(˜ψ1,˜ψ2)X2 and V1(t)w=ˉDΔwW(t)w, where ˉD=diag(Dv,Dh) and

    [W(t)](x)=(μv(t,x)00(μh(x)+ϱ(x))),xˉΩ.

    Denote that Φ(t,s)=diag(T1(t,s),T2(t,s)) is an evolution operator related to the following system

    dwdt=V1(t)w, (4.3)

    ts, with the Neumann boundary condition. Obviously, Φ(t,s) is a positive operator, that is, Φ(t,s)X+2X+2 for all ts. Then V1(t) is resolvent positive, based on reference [39,Theorem 3.12].

    The exponential growth bound of Φ(t,s) can be written as

    ˉω=inf{˜ω:L1suththatΦ(t+s,s)Le˜ωt,sR,t0}.

    By [39,Proposition A.2], one has

    ˉω(Φ)=lnr(Φ(ω+ˉs,ˉs))ω,ˉs[0,ω].

    According to Krein-Rutman theorem and [40,Lemma 14.2], one has

    0<r(Φ(ω,0))=max{r(T1(ω,0)),r(T2(ω,0))}<1,

    here, r(Φ(ω,0)) denotes the spectral radius of Φ(ω,0). By [39,Proposition 5.6] with s=0, we obtain ˉω(Φ)<0. Obviously, F(t) and W(t) meet the assumptions: (1) F(t):X2X2 is positive, i.e., F(t)X+2X+2, for any t0. (2) W(t) is cooperative, and ˉω(Φ)<0.

    To introduce R0 of system (2.12) based on references [41] and [42], it is assumed that both human and mosquito populations are close to (u1(t,),0,0). Suppose wCω(R,X2), and w(t)=w(t) is the initial distribution of infected mosquitoes and humans at tR. For any fixed s0, the distribution of newly infected mosquitoes and individuals at time ts (ts) is F(ts)wts, which is engendered by the infectious mosquitoes and humans introduced in [tsˆτ2,ts]. Thus, the distribution of infectious who were newly infected at time ts and keep survive at time t, is Φ(t,ts)F(ts)wts, ts. Then, at time t, the distribution of the accumulative infectious humans and mosquitoes produced by all people introduced to t at all previous times is

    +0Φ(t,ts)F(ts)wtsds=+0Φ(t,ts)F(ts)w(ts+)ds.

    Define the next generation operator L associated with system (2.12) as

    [Lw](t):=+0Φ(t,ts)F(ts)w(ts+)ds,tR,wCω(R,X2).

    Inspired by the definition of next generation operators in references [39,42], the basic reproduction number of system (2.12) is defined by the spectral radius L, i.e.,

    R0:=r(L). (4.4)

    For any ˜ψX2, set ˉQt to be solution map of system (4.2) on X2, i.e., ˉQt(˜ψ)=wt(˜ψ), t0, where wt(˜ψ)(θ)(x)=w(t+θ,x,˜ψ)=(w1(t+θ,x,˜ψ),w2(t+θ,x,˜ψ)), and w(t,x,˜ψ) is the unique solution of system (4.2) with w(θ,x)=˜ψ(θ,x) for all θ[ˆτ2,0], xˉΩ. Thus ˉQ:=ˉQω is the Poincaré map related to system (4.2). r(ˉQ) represents the spectral radius of ˉQ. Based on reference [41,Theorem 3.7], the following result can be obtained.

    Lemma 4.1. R01 has the same sign as r(ˉQ)1.

    To explore the global dynamic behaviors of system (2.12) according to R0, we demonstrate that system (4.2) engenders an eventually strong monotone periodic semiflow on the phase space:

    E1:=Y×C([τ2(0),0],Y),andE+1:=Y×C([τ2(0),0],Y+).

    Thus (E1,E+1) is an ordered Banach space. For every ˆψE+1, ˜w(t,x,ˆψ)=(˜w1(t,x,ˆψ),˜w2(t,x,ˆψ)) denotes the unique solution of (4.2) with ˜w0(ˆψ)(θ,x) for xˉΩ, θ[τ2(0),0], where

    ˜wt(ˆψ)(θ,x)=˜w(t+θ,x,ˆψ)=(˜w1(t,x),˜w2(t+θ2,x)),θ[τ2(0),0],t0.

    Lemma 4.2. For any ˆψE+1, system (4.2)has a unique nonnegative solution ˜w(t,,ˆψ) on [0,)with ˜w0=ˆψ.

    The proof of Lemma 4.2 is given in Appendix B.

    Remark 1. Through the uniqueness of solutions in Lemmas 3.2 and 4.2, then for any ˜ψX+2 and ˆψE+1 with ˜ψ1()=ˆψ1() and ˜ψ2(θ,)=ˆψ2(θ,), θ[τ2(0),0], one has w(t,,˜ψ)=˜w(t,,ˆψ), t0, where w(t,,˜ψ) and ˜w(t,,ˆψ) are solutions of system (4.2) meeting w0=˜ψ and ˜w0=ˆψ, respectively.

    For every fixed t0, let ˜Qt be the solution map of system (4.2) on the space E1, i.e., ˜Qt(ˆψ)=˜wt(ˆψ), for t0 and ˆψE1. Thus, ˜Q:=˜Qω is the Poincaré map related to system (4.2) and r(˜Q) is the spectral radius of ˜Q. The next lemma indicates that the periodic semiflow ˜Qt is eventually strongly positive.

    Lemma 4.3. For each ξ in E+1, with ξ0, the solution ˜w(t,,ξ) of (4.2) with ˜w0=ξ, meets ˜wi(t,)>0 (i=1,2) for all t>2ˆτ, therefore, ˜Qtξ0 for every t>3ˆτ.

    The proof of Lemma 4.3 is given in Appendix C.

    Fix an integer n0 satisfying n0ω>3ˆτ. Obviously, ˜Qn0=˜Qn0ω is strongly monotone, according to proof of Lemma 4.3. In addition, through similar arguments in reference [43,Lemma 2.6], one can get that ˜Qn0 is compact. Apply the Krein-Rutman theorem to the linear operator ˜Qn0, and because r(˜Qn0)=(r(˜Q))n0, one has ˉλ=r(˜Q)>0, among them ˉλ is the simple eigenvalue of ˜Q and there is a eigenvector ϑInt(E+1) that is strongly positive. Based on reference [22,Lemma 3.8], one has the following lemma.

    Lemma 4.4. The spectral radii of the two Poincaré maps ˉQ:X2X2and ˜Q:E1E1 are the samei.e., r(ˉQ)=r(˜Q).Furthermore, R01 has the same sign as r(˜Q)1.

    The fact that system has a special solution is given by the following lemma, which can be used to study long-term dynamics. The following argument is inspired by references [29,Lemma 5] and [44,Proposition 1.1].

    Lemma 4.5. There is a positive ω-periodic function w(t,x), such that e˜μtw(t,x) is a solution of (4.2), where ˜μ=lnr(˜Q)ω.

    The proof of Lemma 4.5 is given in Appendix D.

    Next, we use R0 to give the threshold results of the global dynamics of system (2.12).

    Denote

    E2:=C([τ2(0),0],Y+)×Y+×C([τ2(0),0],Yh). (4.5)

    Through similar arguments in Lemma 4.2, it can be concluded that for any ϕWh and ˉφE2 with ϕ1(θ,)=ˉφ1(θ,), ϕ2(0,)=ˉφ2(), and ϕ3(θ,)=ˉφ3(θ,), θ[τ2(0),0], z(t,,ϕ)=ˉz(t,,ˉφ) is established, t0, where z(t,,ϕ) and ˉz(t,,ˉφ) are solutions of system (2.12) meeting z0=ϕ and ˉz0=ˉφ, respectively. Then we get that solutions of system (2.12) on E2 exist globally on [0,+) and ultimately bounded. In addition, it is easy to get the next result based on the discussion on references [22,Lemma 3.5], [43] and [45,Theorem 2.9].

    Lemma 4.6. Set ˆQt to be the solution map of system (2.12) on the spaceE2, i.e., ˆQt(ˉφ)=ut(ˉφ), for t0 and ˉφE2. Thus, ˆQt is an ω-periodic semiflow on E2, in this sense, one has

    (i) ˆQ0=I;

    (ii) ˆQt+ω=ˆQtˆQω, t0;

    (iii) ˆQt(ˉφ) is continuous in (t,ˉφ)[0,)×E2.

    Furthermore, ˆQ:=ˆQω admits a strong global attractor in E2.

    Next, we prove that the solution of system (2.12) is strictly positive.

    Lemma 4.7. Use u(t,,ˉφ)=(u1(t,,ˉφ),u2(t,,ˉφ),u3(t,,ˉφ)) to be the solution of (2.12) with the initial data u0=ˉφE2, if there is t10 such that u2(t1,,ˉφ)0, andu3(t1,,ˉφ)0, in that way, the solution of system (2.12) meets

    u2(t,x,ˉφ)>0,u3(t,x,ˉφ)>0,t>t1,xˉΩ. (4.6)

    Besides, for t>0, xˉΩ and any initial data ˉφE2, one has u1(t,x,ˉφ)>0, and

    lim inftu1(t,x,ˉφ)ε,unifromlyforxˉΩ, (4.7)

    where ε>0 is a ˉφ-independent positive constant.

    Proof. According to the comparison principle for cooperative systems, we have u2(t,x)0 and u3(t,x)0 for t>0 and xˉΩ. Moreover, for a fixed ˉφE2, it is easy to know that u2(t,x,ˉφ) and u3(t,x,ˉφ) satisfy

    {u2tDvΔu2ˆμvu2,u3tDhΔu3(ˆμh+ˆϱ)u3,u2ν=u3ν=0,xΩ,

    where ˆμv=maxt[0,ω],xˉΩμ(t,x), ˆμh=maxxˉΩμh(x) and ˆϱ=maxxˉΩϱ(x). If there exists t10, such that u2(t1,,ˉφ)0 and u3(t1,,ˉφ)0, one has u2(t,,ˉφ)>0 and u3(t,,ˉφ)>0, t>t1 based on the strong maximum principle [40,Proposition 13.1]. For t>0, let ˉn(t,x,ˉφ2) be the solution of

    {ˉn(t,x)t=DvΔˉn(t,x)+(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r,x)dr(αβ(t,x)pl+μv(t,x))ˉn(t,x),xΩ,ˉn(t,x)ν=0,xΩ. (4.8)

    Clearly, Λv(t,x)0 is non-negative non-trivial and Hölder continuous on R×ˉΩ. Using the comparison principle, one has

    u1(t,x)ˉn(t,x)>0,t>0,xˉΩ.

    In addition, from reference [38,Lemma 2.1], system (4.8) allows a unique globally attractive periodic solution ˉn(t,). Then

    lim inftu1(t,x,ˉφ1)ε:=mint[0,ω],xˉΩˉn(t,x),uniformlyforxˉΩ.

    This proof is complete.

    For every fixed ˉφE2, u(t,x,ˉφ) is the unique solution of system (2.12) with u0=ˉφ. The following conclusion indicates that R0 is a threshold for the transmission of disease.

    Theorem 4.8. Let R0 be defined in (4.4).We have the following statements:

    (i) When R0<1, the disease-free ω-periodic solution (u1(t,x),0,0) of system (2.12) is globally attractive.

    (ii) When R0>1, system (2.12) has at least one positive ω-periodic solution (ˉz1(t,x),ˉz2(t,x),ˉz3(t,x)), and there exists a constant ˉσ>0such that for every ˉφE2 withˉφ2()0, ˉφ3(0,)0, on has

    lim inftminxˉΩui(t,x,ˉφ)ˉσ,i=1,2,3. (4.9)

    Proof. Since ˆQt is the solution map of system (2.12) on E2, based on reference [22,Lemma 3.5], ˆQt is an ω-periodic semiflow of system (2.12). Thence, ˆQ:=ˆQω is the Poincaré map, and {ˆQn}n0 is a discrete-time system on E2. For every fixed ˉφE2, the omega limit set of the orbit {ˆQn(ˉφ)}n0 is ω(ˉφ). Then ω(ˉφ) is an internally chain transitive set for {ˆQn} on E2.

    (i) As R0<1, according to Lemma 4.4, one has r(˜Q)<1, and therefore ˜μ=lnr(˜Q)ω<0. For t>0, discuss the following system with parameter ϵ>0,

    {wϵ1t=DvΔwϵ1μv(t,x)wϵ1+(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)p[u1(tτ2(t),y)+ϵ]wϵ2(tτ2(t),y)lN(y)dy,xΩ,wϵ2t=DhΔwϵ2+cβ(t,x)wϵ1(μh(x)+ϱ(x))wϵ2,xΩ,wϵ1ν=wϵ2ν=0,xΩ. (4.10)

    Denote wϵ(t,x,ˆφ)=(wϵ1(t,x,ˆφ),wϵ2(t,x,ˆφ)) to be the unique solution of system (4.10) for any ˆφE1, with wϵ0(ˆφ)(θ,x)=(ˆφ(0,x),ˆφ(θ,x)) for all θ[τ2(0),0], xˉΩ, where

    wϵt(ˆφ)(θ,x)=wϵ(t+θ,x,ˆφ)=(wϵ1(t,x,ˆφ),wϵ2(t+θ,x,ˆφ)),θ[τ2(0),0],t0,xˉΩ.

    Let ˜Qϵt:=E1E1 be the solution map of system (4.10) and ˜Qϵ:=˜Qϵω be the corresponding Poincaré map, i.e., ˜Qϵ(ˆφ)=wϵω(ˆφ), for ˆφE1, and make r(˜Qϵ) be the spectral radius of ˜Qϵ. Then the eventual strong monotonicity and compactness of ˜Qϵ on E1 are easy to prove, for t3ˆτ2. From the continuity of the principle eigenvalue to parameters, one has limϵ0r(˜Qϵ)=r(˜Q). Allow ϵ>0 to be small enough, one has r(˜Qϵ)<1. From Lemma 4.5, a positive ω-periodic function wϵ(t,x) exists such that wϵ(t,x)=e˜μϵtwϵ(t,x) is a solution of system (4.10), where ˜μϵ=lnr(˜Qϵ)ω<0. By system (4.1), u1(t,) is globally attractive and the comparison principle, there must be a large enough integer n1>0 such that n1ω>ˆτ2 and

    u1(t,x)u1(t,x)+ϵ,tn1ωˆτ2,xˉΩ.

    Thus, as tn1ω, one has

    {u2tDvΔu2μv(t,x)u2+(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)p[u1(tτ2(t),y)+ϵ]u3(tτ2(t),y)lN(y)dy,xΩ,u3tDhΔu3+cβ(t,x)u2(μh(x)+ϱ(x))u3,xΩ,u2ν=u3ν=0,xΩ, (4.11)

    there exists some C1>0, for any given ˉφE2, such that

    (u2(t,x,ˉφ),u3(t,x,ˉφ))C1(wϵ1(t,x),wϵ2(t,x)),t[n1ωˆτ2,n1ω],xˉΩ.

    Using systems (4.10) and (4.11), and the comparison theorem [36,Proposition 1], we have

    (u2(t,x,ˉφ),u3(t,x,ˉφ))C1e˜μϵtwϵ(t,x),tn1ω,xˉΩ.

    Accordingly, limt(u2(t,x,ˉφ),u3(t,x,ˉφ))=(0,0) uniformly for xˉΩ. Using the theory of internally chain transitive sets [32,Section 1.2], we can prove that limt(u1(t,x,ˉφ)u1(t,x))=0 uniformly for xˉΩ, where u1(t,x) is a globally attractive solution of system (4.1). Based on the above argument, u1(t,,ˉφ) meets a non-autonomous system, and this system tends to periodic system (4.1) asymptotically. It is known that system (4.1) can produce a solution semiflow ˆPt, t0 on C([τ2(0),0],Y+). Then ˆP:=ˆPω is a Poincaré map related to system (4.1). Obviously, ˆP has a global attractor in C([τ2(0),0],Y+).

    Let J=ω(ˉφ) be the omega limit set of ˉφ=(ˉφ1,ˉφ2,ˉφ3)E2 for ˆQt. Owing to limtu2(t,x,ˉφ)=0 and limtu3(t,x,ˉφ)=0 uniformly for xˉΩ, one has J=J×{0}×{ˆ0}. Here, ˆ0 represents function which is identical to 0, i.e., ˆ0(θ,)=0, θ[τ2(0),0]. Following Lemma 4.7, we have ˆ0J. Based on [32,Lemma 1.2.1], we know that J is an internally chain transitive set for ˆQ, then J is an internally transitive chain set for ˆP. Define u0C([τ2(0),0],Y+) by u0(θ,)=u1(θ,) for θ[τ2(0),0]. Since J{ˆ0} and u0 is globally attractive in C([τ2(0),0],Y+){ˆ0}, one has JWs(u0), where Ws(u0) is the stable set of u0. According to [32,Lemma 1.2.1], one has J={u0}. J={u0,0,ˆ0} is proved, thence

    limt(u1(t,x,ˉφ),u2(t,x,ˉφ),u3(t,x,ˉφ))(u1(t,),0,0)=0.

    (ii) When R0>1, Lemmas 4.1, 4.4 and 4.5 indicate r(˜Q)>1, consequently, ˜μ=lnr(˜Q)ω>0. Let

    P0={ˉφE2:ˉφ2()0andˉφ3(0,)0},

    and

    P0=E2P0={ˉφE2:ˉφ2()0orˉφ3(0,)0}.

    For every ˉφP0, Lemma 4.7 reveals that u2(t,x,ˉφ)>0 and u3(t,x,ˉφ)>0, t>0, xˉΩ. Thus, ˆQn(P0)P0, nN. Lemma 4.6 evidences that ˆQ has a strong global attractor in E2.

    Set

    M:={ˉφP0:ˆQn(ˉφ)P0,nN},

    and the omega limit set of the orbit γ+:={ˆQn(ˉφ):nN} is ω(ˉφ). Denote H=(u1(t,x),0,ˆ0). Next, we prove that H cannot form a cycle for ˆQ in P0.

    Claim 1. For every ˜ϕM, the omega limit set ω(˜ϕ)=H.

    For every fixed ˜ϕM, ˆQn(˜ϕ)P0, nN. Then either u2(nω,,˜ϕ)0 or u3(nω,,˜ϕ)0, for each nN. Furthermore, by contradiction and Lemma 4.7, it is obvious that for each t0, either u2(t,,˜ϕ)0 or u3(t,,˜ϕ)0. If u2(t,,˜ϕ)0 for t0, then limt(u1(t,x,˜ϕ)u1(t,x))=0 uniformly for xˉΩ. Thereby, the u3 equation in system (2.12) meets

    u3tDhΔu3(ˉμh+ˉϱ)u3,

    where ˉϱ=minxˉΩϱ(x) and ˉμh=minxˉΩμh(x). According to the comparison principle, one has limtu3(t,x,˜ϕ)=0 uniformly for xˉΩ. Suppose that u2(t2,,˜ϕ)0 for some t20, in the light of Lemma 4.7, we obtain u2(t,,˜ϕ)>0, tt2. Accordingly, u3(t,,˜ϕ)0. From the u2 equation in system (2.12), one has limtu2(t,x,˜ϕ)=0 uniformly for xˉΩ. Consequently, the u1 equation in system (2.12) abides by a nonautonomous system, which is asymptomatic to the periodic system (4.1). The theory of asymptomatically periodic system [32,Section 3.2] implies that limt(u1(t,x,˜ϕ)u1(t,x))=0 uniformly for xˉΩ. Therefore, ω(˜ϕ)=H for any ˜ϕM, and H can not form a cycle for ˆQ in P0.

    For t>0, study the time-periodic parabolic system with parameter ˉδ>0

    {wˉδ1t=DvΔwˉδ1μv(t,x)wˉδ1+(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)p[u1(tτ2(t),y)ˉδ]wˉδ2(tτ2(t),y)pˉδ+lN(y)dy,xΩ,wˉδ2t=DhΔwˉδ2+cβ(t,x)l[N(x)ˉδ]wˉδ1pˉδ+lN(x)(μh(x)+ϱ(x))wˉδ2,xΩ,wˉδ1ν=wˉδ2ν=0,xΩ. (4.12)

    For every ˆφE1, set wˉδ(t,x,ˆφ)=(wˉδ1(t,x,ˆφ),wˉδ2(t,x,ˆφ)) to be the unique solution of system (4.12) with wˉδ0(ˆφ)(θ,x)=(ˆφ1(0,x),ˆφ2(θ,x)) for all θ[τ2(0),0], xˉΩ, where

    wδt(ˆφ)(θ,x)=wˉδ(t+θ,x,ˆφ)=(wˉδ1(t,x,ˆφ),wˉδ2(t+θ,x,ˆφ)),

    for any t0 and xˉΩ. Let ˜Qˉδt:E1E1 be the solution map of system (4.12), i.e., ˜Qˉδt(ˆφ)=wˉδt(ˆφ), and ˜Qˉδ:=˜Qˉδω be the corresponding Poincaré map, for ˆφE1. Let r(˜Qˉδ) be the spectral radius of ˜Qˉδ. Because limˉδ0r(˜Qˉδ)=r(˜Q)>1, a fully small ˉδ>0 can be selected, such that

    ˉδ<min{mint[0,ω],xˉΩu1(t,x),minxˉΩN(x)},andr(˜Qˉδ)>1.

    For the certain ˉδ>0, according to the continuous dependence of the solution on the initial data, there must be a constant δ>0 such that for all ˉφ with ˉφH<δ, which leads to ˆQt(ˉφ)ˆQt(H)<ˉδ for all t[0,ω].

    Claim 2. For any ˉφP0, then lim supnˆQn(φ)Hδ.

    Assuming for some ˉφ0P0, one has lim supnˆQn(ˉφ0)H<δ, by contradiction. So there is n21 such that ˆQn(ˉφ0)H<δ for all nn2. For each tn2ω, set t=nω+t with n=[t/ω] and t[0,ω), one has

    ˆQt(ˉφ0)ˆQt(H)=ˆQt(ˆQn(ˉφ0))ˆQt(H)<ˉδ. (4.13)

    Following system (4.13) and Lemma 4.7, we can receive

    u1(t,x)ˉδ<u1(t,x,ˉφ0)<u1(t,x)+ˉδ,0<u2(t,x,ˉφ0)<ˉδ,0<u3(t,x,ˉφ0)<ˉδ,

    for each t>n2ωˆτ2, and xˉΩ. As tn2ω, u2(t,x,ˉφ0) and u3(t,x,¯φ0) meet

    {u2tDvΔu2μv(t,x)u2+(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)p[u1(tτ2(t),y)ˉδ]u3(tτ2(t),y)pˉδ+lN(y)dy,xΩ,u3tDhΔu3+cβ(t,x)l[N(x)ˉδ]u2pˉδ+lN(x)(μh(x)+ϱ(x))u3,xΩ,u2ν=u3ν=0,xΩ. (4.14)

    Based on u(t,x,ˉφ0)0 for every t0 and xˉΩ, there must be a constant C2>0, such that

    (u2(t,x,ˉφ0),u3(t,x,ˉφ0))C2e˜μˉδtuˉδ(t,x),t[n2ωˆτ2,n2ω],xˉΩ,

    where uˉδ(t,x) is a positive ω-periodic function and e˜μˉδtuˉδ(t,x) is a solution of system (4.12), then ˜μˉδ=lnr(˜Qˉδ)ω>1. By system (4.14) and the comparison theorem

    (u2(t,x,ˉφ0),u3(t,x,ˉφ0))C2e˜μˉδtuˉδ(t,x),tn2ω,xˉΩ.

    Obviously, ˜μˉδ>0, then u2(t,,ˉφ0)+ and u3(t,,ˉφ0)+ as t+, which leads to a contradiction.

    With the above claim, it is easy to obtain that H is an isolated invariant set for ˆQ in E2, and Ws(H)P0=, where Ws(H) is the stable set of H for ˆQ. For every integer n and nω>ˆτ2, ˆQn:=ˆQnω is compact, then ˆQ is compact, and ˆQ is asymptotically smooth on E2. Moreover, similar to Lemma 4.6, we get that ˆQ has a global attractor on E2. According to reference [45,Theorem 3.7], ˆQ has a global attractor A0 in P0. Based on the acyclicity theorem on uniform persistence for map [32,Theorem 1.3.1 and Remark 1.3.1], we obtain that ˆQ is uniformly persistent with respect to (P0,P0), in this sense, there exists an γ1>0, such that

    lim infnd(ˆQn(ˉφ),P0)γ1,ˉφP0. (4.15)

    Now, the practical persistence are derived. Following A0=ˆQ(A0), we obtain that ˉφ2()>0 and ˉφ3(0,)>0 for any ˉφA0. Set B:=t[0,ω]ˆQt(A0). Obviously, we have BP0 and limtd(ˆQt(ˉφ),B)=0, ˉφP0. The continuous function q:E2R+ is defined by

    q(ˉφ):=min{minxˉΩˉφ2(0,x),minxˉΩˉφ3(0,x)},ˉφ(ˉφ1,ˉφ2,ˉφ3)E2.

    Because B is a compact subset of P0, we obtain that infˉφBq(ˉφ)=minˉφBq(ˉφ)>0. Consequently, there is an γ2>0 such that

    lim inftq(ˆQt(ˉφ))=lim inftmin(minxˉΩu2(t,x,ˉφ),minxˉΩu3(t,x,ˉφ))γ2,ˉφP0.

    Furthermore, according to Lemma 4.7, there must be an γ3(0,γ2) such that

    lim inftminxˉΩuj(t,x,ˉφ)γ3,ˉφP0(j=1,2,3).

    By references [29,Lemma 8] and [32,Theorem 3.5.1], for every t>0, the solution map Qt:=WhWh of system (2.12), defined in Lemma 3.2, is a κ-contraction about an equivalent norm on Wh, where κ is Kuratowski measure of noncompactness. Define

    W0={ϕWh:ϕ2(0,)0andϕ3(0,)0},

    and

    W0=Wh/W0={ϕWh:ϕ2(0,)=0orϕ3(0,)=0}.

    For any ϕW0, Q is ρ-uniformly persistent with ρ(ϕ)=d(ϕ,W0) is easily attainable. It then follows from [45,Theorem 4.5], applicable to Q, that system (2.12) has an ω-periodic solution (z1(t,),z2(t,),z3(t,)) with (z1t,z2t,z3t)W0. Let ˉz1(θ,)=z1(θ,), ˉz2(0,)=ˉz2(0,) and ˉz3(θ,)=z3(θ,) where θ[τ2(0),0]. Combing the uniqueness of solutions and Lemma 4.7, we have that (ˉz1(t,),ˉz2(t,),ˉz3(t,)) is periodic solution system (2.12) and it is also strictly positive.

    Next, we prove the asymptomatic behaviour of M and Ev in system (2.11). By Eq (2.6), and Λv(t,x)0 is nonnegative and Hölder continuous function for t>0 and xˉΩ. Then there is a unique positive ω-periodic solution M(t,x) which is globally attractive. From Eq (3.6) and Theorem 4.8(i), when R0<1, one has limtEv(t,)=0. With the arguments similar to those above, we can prove that if R0<1, then limt(Ev(t,x),Iv(t,x),Ih(t,x))=(0,0,0) and limt((M(t,),Sv(t,x))(M(t,x),u1(t,x)))=0, uniformly for xˉΩ. By Eq (2.12), Theorem (4.8)(ii) and Eqs (2.6) and (3.6), then if R0>1, for any ˉφE2 with ˉφ2()0 and ˉφ3(0,)0, one has

    lim inftSv(t,x)γ3,lim inftIv(t,x)γ3,lim inftIh(t,x)γ3,

    uniformly for xˉΩ. Since Λv(t,x) is Hölder continuous and non-negative function for t>0 and xΩ from Eq (2.6), then there exists γ4>0 such that

    lim inftM(t,x)γ4,

    uniformly for xˉΩ with M(0,x) satisfies compatibility condition (3.4). By the integral form Eq (3.6), then exists γ5>0 such that

    lim inftminxˉΩEv(t,x)γ5,

    with Ev(0,x) satisfies compatibility condition (3.5). In addition, if (Sv(t,),Iv(t,),Ih(t,)) is ω-periodic in t, then Ev(t,) is ω-periodic is easy to be checked. Therefore, the following result is obtained.

    Theorem 4.9. Let R0 be defined as in (4.4).Set U(t,x,ψ) to be the solution of system (2.11) with U0=ψ, where(U1(t,x),U2(t,x),U3(t,x),U4(t,x),U5(t,x))=(M(t,x),Sv(t,x),Ev(t,x),Iv(t,x),Ih(t,x)).One has:

    (i) When R0<1, the disease-free ω-periodic solution (M(t,),u1(t,),0,0,0) of system (2.11) is globally attractive.

    (ii) When R0>1, system (2.11) has at least one positive ω-periodic solution ˉU(t,x), where ˉU2(t,x)=ˉz1(t,x), ˉU4(t,x)=ˉz2(t,x), ˉU5(t,x)=ˉz3(t,x) and there is a constantγ6=max{γ3,γ4,γ5}>0that makes for every ψD withψ4()0, ψ5(0,)0, we have

    lim inftminxˉΩˉUi(t,x,ψ)γ6,i=1,2,3,4,5,6. (4.16)

    In this part, we use numerical simulations to clarify the impact of spatial heterogeneity, periodic delays, seasonality, and vector-bias on the spread of malaria, to show how to derive some understanding of epidemiology from our analysis results.

    Here, some simulations to support the analysis results built in Section 4 are provided. More precisely, the potential dynamics outcomes of system (2.11) are given. For convenience, we will focus our attention on the domain Ω, which is one-dimensional, denoted by [0,π] to simulate the long-time behavior of system (2.11). The basic reproduction number R0 often (but not always) controls the extinction and persistence of the disease. Specifically, system (2.11) applies to the transmission of malaria in Maputo Province, Mozambique, a sub-Saharan African country where malaria is at risk of spreading in the country. The WHO report shows that the number of confirmed malaria cases has been increasing in recent years, and the condition deteriorated severely in 2014.

    According to reference [24], suppose the total population density N(x)=ˉN=53(km2)1. The unit of time is month. Set the periodic ω=1year=12 months. We set ϱ=a1(2.05cos(2x))Month1, where a1=0.0187 means that people living in urban areas (near the center area) can get better medical services (number of doctors and hospitals, advanced medical equipment, supply of medicines) than people in rural areas. This results in a higher recovery rate near the center. See Appendix E for the values and definitions of other parameters.

    To compute R0, we use the numerical method proposed in reference [41,Lemma 2.5 and Remark 3.2]. Using the above parameters set, the basic reproduction number can be numerically calculate as R0=1.2233>1, which implies that in mosquito and human populations, the disease persist. Figure 3 shows numerical plots of humans and mosquitoes compartments with initial data

    ˆu(θ,x)=(5005cos2x3005cos2x325cos2x405cos2x52cos2x),θ[ˆτ,0],x[0,π],

    where ˆu(t,x):=(M,Sv,Ev,Iv,Ih). This conforms to the conclusion of Theorem 4.9(ii). The time interval can be truncated [20,30] to prove that the existence of the positive periodic solution.

    Figure 3.  The long-term behavior of the solutions of system (2.11) when R0>1.

    Reduce the bite rate to 0.5β(t), and increase the death rate of mosquito to 1.5μv(t), which can be achieved by using insecticide-treated nets, then R0=0.2983<1. Thus, immature mosquitoes and susceptible adult mosquitoes show periodic fluctuations. The exposed adult mosquitoes, infectious adult mosquitoes and infectious humans converge to 0 (Figure 4). This is coincident with the conclusion of Theorem 4.9(i).

    Figure 4.  The long-term behavior of system (2.11) solution, when R0<1.

    We are interested in how system parameters affect the disease risk R0. In this section, we choose the parameters that are consistent with those in Section 5.1.

    In this subsection, we explore the impact of mature period affected by seasonal climate change on malaria transmission. The results are given in Figure 5, the green one represents the time-averaged value of maturation delay [τ1] in system (2.11), where [τ1]=1ωω0τ1(t)dt=0.5605Month, and the red one indicates system (2.11) is under a time-periodic maturation delay. An intuitive result is that in these four cases, the green line is always below the red line, which fully indicates that the time-average maturity delay will underestimate the risk of malaria transmission, so the impact of climate change cannot be ignored. In the following subsections, we analyze in detail the sensitivity of R0 to these four parameters.

    Figure 5.  The relationship between R0 and parameters. There are two curves. The green one represents the time-average maturity delay [τ1], and the red one represents τ1(t) affected by the periodic temperature.

    Looking at Figure 5, we let l/p vary in [0,1], and other parameters are same as those in Figure 3 and explore the impact of the vector-bias effect, l/p are used to represent the relative attractiveness of susceptible to infected humans. From Figure 5(a), it can be easily seen that R0 is an increasing function of l/p. Therefore, ignoring the vector-bias effect will ultimately underestimate the risk of disease transmission. And this conclusion is valid regardless of whether the periodic or the time-average maturation delay is considered.

    In this subsection, we examine the impact of interventions on R0. Note that in Figure 3(e), the population density of infected individuals in urban (central) areas is less than in rural (border) areas, which prompts us to explore the relationship between human recovery rate ϱ and disease risk R0 in a heterogeneous space. Since medical resources (advanced medical equipment, the number of hospitals and doctors, medicine supply) are abundant in urban areas than in rural areas, humans can get better medical services in urban areas with a higher recovery rate.

    In this subsection, we analyze the impact of medical resources on the spread of malaria, including two aspects: (i) the impact of increasing medical resources; (ii) for fixed medical resources, the impact of different allocation measures on disease transmission. In Figure 5(b), we let a1 in ϱ vary in [0.02,0.5], and the other parameters are same as those in Figure 3. It proposes a strategy, i.e., in the current allocation of the medical resources, the recovery rate can be improved through efforts to develop new drugs. Figure 5(b) shows that increasing medical resources can reduce R0 to less than 1, which means that the disease can be controlled. In this paper, we use the recovery rate π0ϱ(x)dx of the entire region to represent medical resources.

    In the case of fixed medical resources, we explore the impact of different allocation on the spread of disease. Choose a1=0.058 and a new parameter δ1 is added to ϱ, i.e., ϱ(x)=0.058×(1(1δ1)cos(2x)), where δ1[0,1], which measures the difference in the allocation of urban and rural medical resources. Other parameters are the same as those in Figure 3. When δ1=0, the gap between rural and urban medical resources is the largest, and the recovery rate in urban area is the highest (near the center of space area, i.e., x=π2). As δ1 changes from 0 to 1, medical resources are allocated to nearby rural area (near x=0 and x=π), and the gap in medical conditions between rural and urban areas gradually narrows, and is eventually evenly distributed in space. Through calculation, it can be found that if δ1=0, in order to make R0 less than 1, a10.0637 is needed, but if δ1=1, the medical resources only need 0.0576, which can make R0<1. From Figure 5(c), we come to an interesting conclusion that in the case of limited medical resources, the spatial heterogeneity of the distribution of medical resources increases the risk of disease transmission. This result indicates that the distribution of medical resources dose play an important role in the design of malaria control programs.

    We explore the impact of human mobility on the transmission of malaria in a heterogeneous space. Let Dh vary in [0.02,0.5] and other parameters are consistent with those in Figure 3. The result is shown in Figure 5(d), which describes the relationship between R0 and Dh. It can be directly observed from Figure 5(d) that R0 decreases with respect to Dh. When Dh is very small, R0 will drop sharply and as Dh continues to increase, the rate of decline of R0 will slow down. This seems to indicate that malaria cannot be controlled if no protective measures are taken and only by improving a high diffusion to avoid mosquito bites without taking any protective measures [49] systematically analyzes this problem.

    In Section 5.2.4, we explore the impact of diffusion on R0, and study the impact of diffusion on the final size of disease transmission. Set ϱ(x)=0.04(2.05cos(2x)) and other parameters are consistent with those in Figure 3. When Dh=0.1 and Dv=0.0125, calculate R0=0.8448, the disease is extinct. However, without considering the movement of mosquitoes and humans, that is, when the diffusion coefficients are Dh=0 and Dv=0, it is found that R0=1.1635 is greater than 1, and the disease is persistent. Then, we will get the result shown in Figure 6, which shows some interesting phenomena. Under this set of parameters, if the diffusion of mosquitoes and humans is not considered, then the disease is extinct in urban areas, and is persistent in rural areas. The reason for this result is that we study spatial heterogeneity. Due to urban environment is not suitable for mosquitoes to survive, the number of mosquitoes is relatively small, and the medical resources are relatively large, so the urban disease control is better. Compared with the right column of Figure 6, the movement of mosquitoes and humans will reduce the risk of disease transmission to a certain extent, which is consistent with the conclusions in Figure 5(d) and reference [24]. There are two explanations: (i) The existence of diffusion indicates that the infected humans in rural areas can be treated in cities with better medical conditions. (ii) As mentioned in reference [24], to a certain extent, it is difficult for mosquitoes to take a blood meal among moving humans.

    Figure 6.  The evolution of infection compartments of humans and mosquitoes with/without diffusion.

    As we all know, in the study of malaria transmission, the combined effects of temperature-dependent EIP and maturation delay, vector-bias, spatial structure and seasonal variation are worth studying. Here, we establish and analyze a reaction-diffusion model of malaria, including periodic delays and vector-bias in a heterogeneous environment.

    Using the theory proposed by Liang et al. [41] and Zhao [42], the basic reproduction number R0 can be introduced. We first prove that in a defined phase space, linear system (4.2) generates a periodic semiflow which is eventually strongly monotonic. Then, we describe R0 as a threshold for the extinction and persistence of malaria. It is proved that if R0<1, then malaria will be eliminated. When R0>1, there admits at least one positive ω-periodic solution.

    In the numerical part, data for exploring the transmission of malaria in Maputo Provence, Mozambique have been published. The numerical simulation is divided into three parts. First, we give some simulation conclusions to support the theoretical analysis. Second, we discuss the sensitivity of R0 to the parameters in the model. (I) Our numerical results show that if we only consider the time-average maturation delay, we can underestimate the risk of malaria transmission assessed by R0. (II) It is easy to know that R0 is decreasing with respect to l/p. Thence, if the impact of vector-bias effect on malaria transmission is ignored, the risk of transmission will be underestimated. (III) We analyze the impact of intervention measures on R0 from two aspects: (i) the impact of increasing medical resources. This case is mainly due to the uneven distribution of medical resources in rural and urban areas, which increases the impact of medical resources on R0; (ii) for fixed medical resources, the impact of different allocation measures on disease transmission. Through the analysis of this case, we find that in the case of limited medical resources, narrowing the gap in the allocation of medical resources between rural and urban areas can better reduce the risk of malaria transmission. Third, we explore the impact of diffusion on R0 in a spatially heterogeneous environment, assuming that the distribution of medical resources and the initial data of mosquitoes and humans depend on space. Biologically speaking, we understand this spatial heterogeneity as the difference between rural and urban areas. In addition, we also study the impact of diffusion on the final size of malaria. We find that under the set parameters, if the diffusion of mosquitoes and humans is not considered, the disease will become extinct in urban areas, and persist in rural areas. But under this set of parameters, the disease will become extinct if the diffusion is not considered.

    We are very grateful to anonymous referees and editor for careful reading and helpful suggestions which led to an improvement of our original manuscript. The authors are grateful for the support and useful comments provided by Professor Xiaoqiang Zhao (Memorial University of Newfoundland). This research is supposed by National Natural Science Foundation of China (grant number: 11971013) and Postgraduate Research & Practice Innovation Program of Jiangsu Province (KYCX21_0176) and Nanjing University of Aeronautics and Astronautics PhD short-term visiting scholar project (ZDGB2021027).

    The authors declare there is no conflict of interest.

    Proof. Firstly, the local existence of the unique mild solution is proved. Clearly, F is locally Lipschitz continuous. It is necessary to show

    limˆθ0dist(ϕ(0,)+ˆθF(t,ϕ),X+1)=0,(t,ϕ)[0,+)×Wh. (A.1)

    For all (t,ϕ)[0,+)×Wh and ˆθ0, because pl>0, one has

    ϕ(0,x)+ˆθF(t,ϕ)(x)=(ϕ1(0,x)+ˆθ((1τ1(t))Λv(tτ1(t))ettτ1(t)μm(r,x)drαβ(t,x)pϕ3(0,x)ϕ1(0,x)pϕ3(0,x)+l[N(x)ϕ3(0,x)])ϕ2(0,x)+ˆθ(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)pϕ3(τ2(t),y)ϕ1(τ2(t),y)pϕ3(τ2(t),y)+l[N(y)ϕ3(τ2(t),y)]dyϕ3(0,x)+ˆθcβ(t,x)l[N(x)ϕ3(0,x)]ϕ2(0,x)pϕ3(0,x)+l[N(x)ϕ3(0,x)])(ϕ1(0,x)+ˆθ((1τ1(t))Λv(tτ1(t))ettτ1(t)μm(r,x)drαβ(t,x)pϕ3(0,x)ϕ1(0,x)lN(x))ϕ2(0,x)+ˆθ(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)pϕ3(τ2(t),y)ϕ1(τ2(t),y)lN(y)dyϕ3(0,x)+ˆθcβ(t,x)l[N(x)ϕ3(0,x)]ϕ2(0,x)pN(x))(ϕ1(0,x)(1ˆθαˆβpϕ3(0,x)l˜N)ϕ2(0,x)ϕ3(0,x)(1ˆθcˆβlϕ2(0,x)p˜N)),

    for t0 and xˉΩ, where ˆβ=maxt[0,ω],xˉΩβ(t,x) and ˜N=minxˉΩN(x), and

    N(x)[ϕ3(0,x)+ˆθF3(t,ϕ)(x)]=[N(x)ϕ3(0,x)][1ˆθcβ(t,x)lϕ2(0,x)pϕ3(0,x)+l[N(x)ϕ3(0,x)]][N(x)ϕ3(0,x)][1ˆθcβ(t,x)ϕ2(0,x)N(x)].

    This implies that

    limˆθ0dist(ϕ(0,)+ˆθF(t,ϕ),Y+×Y+×Yh)=0,(t,ϕ)[0,+)×Wh.

    Consequently, by reference [36,Corollary 4] with K=Y+×Y+×Yh and S(t,s)=T(t,s), system (2.12) has a unique mild solution u(t,,ϕ) on its maximum existence interval t[0,tϕ) and u0=ϕ where tϕ, u(t,,ϕ)Y+×Y+×Yh. In addition, when t>ˆτ, u(t,,ϕ) is a classical solution, based on the analyticity of T(t,s), s,tR and t>s.

    Analyze the following time-periodic reaction-diffusion equation

    {ˆw(t,x)t=DvΔˆw(t,x)μv(t,x)ˆw(t,x)+(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r,x)dr,xΩ,ˆw(t,x)ν=0,xΩ. (A.2)

    By reference [38,Lemma 2.1], system (A.2) has a unique globally attractive positive ω-periodic solution u1(t,x) in Y+. Since system (A.2) can dominate the first equation of (2.12), there is a B1>0 such that for any ϕWh, a positive integer l1=l1(ϕ)>0 exists and meets u1(t,x,ϕ)B1 for all tl1ω and xˉΩ.

    Next, similar to reference [46,Theorem 2.1], the ultimate boundedness of solutions can be proved. Set

    ˜Λv=maxt[0,ω]ΩΛv(t,x)dx,

    for any ϕWh, let (u1(t,x),u2(t,x),u3(t,x)):=(u1(t,ϕ)(x),u2(t,ϕ)(x),u3(t,ϕ)(x)), t0, xˉΩ, and ˉui(t)=Ωui(t,x)dx where i=1,2,3. Using the Green's formula and integrating the first equation of system (2.12), one has

    dˉu1(t)dt=Ω(1τ1(t))Λv(tτ1(t),x)ettτ1(t)μm(r,x)drdxΩαβ(t,x)pu3u1pu3+l[N(x)u3]dxΩμv(t,x)u1dx˜Λvˉμvˉu1(t)Ωαβ(t,x)pu3u1pu3+l[N(x)u3]dx,

    that is,

    Ωαβ(t,x)pu3u1pu3+l[N(x)u3]dx˜Λvˉμvˉu1(t)dˉu1(t)dt,t>0. (A.3)

    By system (A.3), the property of the fundamental solutions [47] and Green's formula, the second equation of (2.12) are integrated to get

    dˉu2(t)dt=Ωμv(t,x)u2dx+(1τ2(t))ΩΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)pu3(tτ2(t),y)u1(tτ2(t),y)pu3(tτ2(t),y)+l[N(y)u3(tτ2(t),y)]dydxˉμvˉu2(t)k1ˉu2(tτ2(t))k2dˉu1(tτ2(t))dt+k3,tl1ω+ˆτ, (A.4)

    where k1, k2 and k3 are positive numbers independent of ϕ. Select k1ˉμvk2 in system (15.4), so that

    ddt[ˉu2(t)+k2ˉu1(tτ2(t))]ˉμvˉu2(t)k1ˉu1(tτ2(t))+k3k1k2ˉu2(t)k1ˉu1(tτ2(t))+k3k1k2[ˉu2(t)+k2ˉu1(tτ2(t))]+k3,tl1ω+ˆτ,

    which yields ˉu2(t)+k2ˉu1(tτ2(t))k2k3k1+1 for tl1ω+ˆτ, where l1>l1 is some integer. Hence,

    ˉu2(t)=u2(t,)L1(Ω)k2k3k1+1,tl1ω.

    Integrating the third equation on model (2.12), and applying Green's formula, we have

    dˉu3(t)dt=Ωcβ(t,x)l[N(x)u3(t,x)]u2(t,x)pu3(t,x)+l[N(x)u3(t,x)]dxΩ(μh(x)+ϱ(x))u3(t,x)dxcˆβˉu2(t)(ˉμh+ˉϱ)ˉu3(t).

    Thus,

    ˉu3(t)(k2k3k1+1)cˆβˉμh+ˉϱ,

    for tl2ω+ˆτ, (l2l1) and xˉΩ. Therefore, we have tϕ= for each ϕWh.

    Let {Qt}t0 be a family of operators on Wh and Qt(ϕ)(s,x)=ut(s,x,ϕ)=u(t+s,x,ϕ) for t0, s[ˆτ,0], xΩ and ϕWh. Similar to the proof of Lemma 2.1 in reference [38], we can show that {Qt}t0 is an ω-periodic semiflow on Wh. From the above proofs, we conclude that Qt:WhWh is point dissipative. Moreover Q:=Qω is κ-contraction, where κ is Kuratowski measure of noncompactness, and hence, Q is asymptotically smooth. By references [48,Theorem 2.4.7] and [32,Theorem 1.1.2], it follows that Q has a global compact attractor.

    Proof. Let ˉτ2=mint[0,ω]τ2(t). For all t[0,ˉτ2], since tτ2(t) strictly increase with respect to t, one has

    τ2(0)=0τ2(0)tτ2(t)ˉτ2τ2(ˉτ2)ˉτ2ˉτ2=0,

    and therefore, ˜w2(tτ2(t),)=ˆψ2(tτ2(t),). Thus, for any t[0,ˉτ2], there holds

    {˜w1t=DvΔ˜w1μv(t,x)˜w1+(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)pˆψ2(tτ2(t),y)u1(tτ2(t),y)lN(y)dy,xΩ,˜w2t=DhΔ˜w2+cβ(t,x)˜w1(μh(x)+ϱ(x))˜w2,xΩ,˜w1ν=˜w2ν=0,xΩ.

    Fix ˆψE+1, for t[0,ˉτ2], the solution (˜w1(t,),˜w1(t,)) of the above linear system exists uniquely. To put this another way, one has z1(θ3,)=˜w1(θ3,), where θ3[0,ˉτ2] and z2(θ4,)=˜w2(θ4,), θ4[τ2(0),ˉτ2].

    By repeating the above step to [nˉτ2,(n+1)ˉτ2], we see that the solution with initial date ˆψE+1 exists uniquely for all t0.

    Proof. Similar to the proof of Lemma 4.2, on each interval [nˉτ2,(n+1)ˉτ2], nN, one easily obtain that ˜wi(t,)0, (i=1,2) for all t0. Select a K>max{ˆμv,ˆμh+ˆϱ}, such that for every tR, g1(t,,˜w1):=μv(t,)˜w1+K˜w1 is strictly increasing with respect to ˜w1, and g2(t,,˜w2):=(μh()+ϱ())˜w2+K˜w2 is increasing in ˜w2. Then, for t>0, both ˜w1 and ˜w2 meet the following system

    {˜w1t=DvΔ˜w1K˜w1+g1(t,,˜w1)+(1τ2(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)p˜w2(tτ2(t),y)u1(tτ2(t),y)lN(y)dy,xΩ,˜w2t=DhΔ˜w2K˜w2+g2(t,,˜w2)+cβ(t,x)˜w1,xΩ,˜w1ν=˜w2ν=0,xΩ.

    Therefore, for a fixed ˉξE+1, one has

    {˜w1(t,,ˉξ)=˜T1(t,0)ˉξ1(0)+t0˜T1(t,s)g1(s,,˜w1(s,))ds+t0˜T1(t,s)(1τ2(t))ΩΓ(s,sτ2(s),,y)αβ(sτ2(s),y)p˜w2(sτ2(s),y)u1(sτ2(s),y)lN(y)dyds,˜w2(t,,ˉξ)=˜T2(t,0)ˉξ2(0)+t0˜T2(t,s)g2(s,,˜w2(s,))ds+t0˜T2(t,s)cβ(s,)˜w1(s,)ds, (C.1)

    where ˜T1(t,s),˜T2(t,s):=YY are evolution operators related to ˜w1t=DvΔ˜w1K˜w1 and ˜w2t=DhΔ˜w2K˜w2 with Nuemann boundary condition. Since m(t):=tτ2(t) are increasing in tR, it easily follows that [τ2(0),0]m([0,ˆτ2]). Generally, we assume that ξ2>0. Then there exists an (θ,x0)[τ2(0),0]×Ω such that ˜w2(θ,x0)>0. According to the equation of (17.1), we have ˜w1(t,,ξ)>0 for all t>ˆτ2. If s>2ˆτ2, then sτ2(s)>2ˆτ2ˆτ2=ˆτ2. It is easy to know that t0, ˜w2(t,,ξ)>0 for all t>2ˆτ2 according to the second equation of (17.1). This means that there is ˜wi(t,)0 for all t>2ˆτ2, i=1,2, so the solution map ˜Qt is strongly positive for all t>3ˆτ2.

    Proof. For every ˆψE+1, let ˜w(t,x,ˆψ)=(˜w1(t,x,ˆψ),˜w2(t,x,ˆψ)) be the solution of (4.2) with ˜w0(ˆψ)=ˆψ. Since ˆψ0, we get ˜wt(ˆψ)0 for all t0 will be a piece of cake. Denote

    w1(t,x)=e˜μt˜w1(t,x,ˆψ),t0,xˉΩ,w2(t,x)=e˜μt˜w2(t,x,ˆψ),tτ2(0),xˉΩ.

    Then w(t,x)=(w1(t,x),w2(t,x))0 for t0, xˉΩ, and w(t,x) meets the linear-periodic system with ˜μ:

    {w1(t,x)t=DvΔw1(t,x)(μv(t,x)+˜μ)w1(t,x)+(1τ1(t))ΩΓ(t,tτ2(t),x,y)αβ(tτ2(t),y)pu1(tτ2(t),y)e˜μτ2(t)w2(tτ2(t),y)lN(y)dy,xΩ,w2(t,x)t=DhΔw2(t,x)+cβ(t,x)w1(t,x)(μh(x)+ϱ(x)+˜μ)w2(t,x),xΩ,w1(t,x)ν=w2(t,x)ν=0,xΩ, (D.1)

    for t>0. Hence, for all θ[τ2(0),0], xˉΩ, w(t,x) is the solution of the ω-periodic system (D.1) with w0(θ,x)=(w1(0,x),w2(θ,x))=(ˆψ1(x),e˜μθˆψ2(θ,x)), where wt(,)=(w1(t,),w2t(,)) for each t0 with

    w1(t,x)=e˜μt˜w1(t,x,ˆψ),xˉΩ,w2t(θ,x)=˜w2(t+θ,x)=e˜μ(t+θ)˜w2(t+θ,x,ˆψ),(θ,x)[τ2(0),0]×ˉΩ.

    For every θ[τ2(0),0], xˉΩ, one has

    w1(ω,x)=e˜μω(˜Q(ˆψ))1(θ,x)=e˜μωr(˜Q)ˆψ1(x)=ˆψ1(x)=w1(0,x),w2(ω+θ,x)=e˜μ(ω+θ)(˜Q(ˆψ))2(θ,x)=e˜μ(ω+θ)r(˜Q)ˆψ2(θ,x)=e˜μθˆψ2(θ,x)=w2(θ,x).

    Consequently, w0(θ,)=wω(θ,) for all θ[τ2(0),0], and by the existence and uniqueness of system (D.1), one has

    w1(t,x)=w1(t+ω,x),t0,xˉΩ,w2(t,x)=w2(t+ω,x),tτ2(0),xˉΩ.

    Thus, w(t,x) is an ω-periodic solution of (D.1) and e˜μtw(t,x) is a solution of (4.2).

    The authors in reference [20] discussed the impact of seasonality on the spread of malaria including assessing the seasonal related bite rate β(t), periodic maturity delay τ1(t), periodic EIP τ2(t), periodic death μv(t) and recruitment rate Λv(t) of mosquitoes, where

    β(t)=6.9830.0709sin(πt/2)1.993cos(πt/6)0.128cos(πt/2)0.007642sin(πt/3)0.4247cos(πt/3)+0.05452sin(2πt/3)1.459sin(πt/6)0.04095cos(2πt/3)+0.0005486cos(5πt/6)0.06235sin(5πt/3)Month1, (E.1)
    τ1(t)=0.56050.03366cos(πt/6)+0.01178sin(πt/6)0.0008135cos(πt/2)0.00266cos(2πt/3)0.0001281cos(5πt/6)+0.005958cos(πt/3)+0.01178sin(πt/6)0.001072sin(2πt/3)0.001936sin(πt/2)0.0001059sin(πt/3)+0.005026sin(5πt/6)Month, (E.2)
    τ2(t)=1/30.4(17.25+4.806sin(πt/6)0.3257cos(πt)+8.369cos(πt/6)+2.857sin(πt/3)+1.197cos(πt/2)+0.03578cos(2πt/3)+0.6354sin(5πt/6)+1.035sin(2πt/3)0.3505cos(5πt/6)+1.963sin(πt/2)+3.27cos(πt/3)+0sin(πt))Month, (E.3)
    μv(t)=3.086+0.01942cos(πt/3)+0.0007665cos(2πt/3)+0.007133cos(πt/2)+0.04788cos(πt/6)0.001459cos(5πt/6)+0.005687sin(2πt/3)+0.01819sin(πt/3)+0.01135sin(πt/2)+0.02655sin(πt/6)+0.003198sin(5πt/3)Month1, (E.4)
    Λv(t)=ˆk×β(t)(km2Month)1,whereˆk=53.13×5. (E.5)
    μm(t)=9.0288(1+0.08cos(π6t))Month1. (E.6)

    Other parameter values are defined as

    Dv=0.0125km2Month1,Dh=0.1km2Month1,α=0.64,c=0.21,μh=0.001574,p=0.8,l=0.6.
    Table A1.  Parameter definitions and values.
    Parameter Definition Value(range) References
    Nh Total human population number 53 (km2)1 [24]
    μh Natural mortality rate of humans 0.00157 Month1 [20]
    ϱ Constant recovery rate of humans (0.04256, 0.5168) Month1 [24]
    l/p Vector-bias parameter (0,1) [13]
    c Transmission probability (0.01, 0.27) [50]
    from infectious mosquitoes to humans
    Dh Diffusion rate of humans 0.1 km2Month1 [24]
    Dv Diffusion rate of mosquitoes 0.0125 km2Month1 [24]
    α Transmission probability from (0.072, 0.64) [50]
    infectious humans to mosquitoes per bite
    β Biting rate of mosquitoes (E.1) [20]
    μv Death rate of mosquitoes (E.4) [20]
    τ1 Maturation delay of (E.2) [51]
    mosquitoes in the aquatic stage
    τ2 EIP of mosquitoes (E.3) [24]
    Λv Recruitment rate of mosquitoes (E.5) [24]
    μm Natural death rate for the aquatic stage (E.6) [52]

     | Show Table
    DownLoad: CSV


    [1] G. H. Hardy, Notes on a theorem of Hilbert, Math. Z., 6 (1920), 314–317. https://doi.org/10.1007/BF01199965 doi: 10.1007/BF01199965
    [2] G. H. Hardy, Notes on some points in the integral calculus (LX), Mess. Math., 54 (1925), 150–156.
    [3] G. H. Hardy, J. E. Littlewood, Elementary theorems concerning power series with positive coefficents and moment constants of positive functions, J. Reine Angew. Math., 157 (1927), 141–158. https://doi.org/10.1515/crll.1927.157.141 doi: 10.1515/crll.1927.157.141
    [4] Y. Bicheng, On a new Hardy type integral inequalities, Int. Math. Forum., 2 (2007), 3317–3322.
    [5] B. Benaissa, M. Z. Sarikaya, Generalization of some Hardy-type integral inequality with negative parameter, Bull. Transilv. Univ. Bras. III, 13 (2020), 69–76. https://doi.org/10.31926/but.mif.2020.13.62.1.6 doi: 10.31926/but.mif.2020.13.62.1.6
    [6] G. Al Nemer, A. I. Saied, A. M. Hassan, C. Cesarano, H. M. Rezk, M. Zakarya, On some new dynamic inequalities involving C-Monotonic functions on time scales, Axioms, 11 (2022), 644. https://doi.org/10.3390/axioms11110644 doi: 10.3390/axioms11110644
    [7] E. Awwad, A. I. Saied, Some new multidimensional Hardy-type inequalities with general kernels on time scales, J. Math. Inequal., 16 (2022), 393–412. https://doi.org/10.7153/jmi-2022-16-29 doi: 10.7153/jmi-2022-16-29
    [8] R. Bibi, M. Bohner, J. Pečarić, S. Varošanec, Minkowski and Beckenbach-Dresher inequalities and functionals on time scales, J. Math. Inequal., 7 (2013), 299–312. https://doi.org/10.7153/jmi-07-28 doi: 10.7153/jmi-07-28
    [9] J. A. Oguntuase, L. E. Persson, Time scales Hardy-type inequalities via superquadracity, Ann. Funct. Anal., 5 (2014), 61–73. https://doi.org/10.15352/afa/1396833503 doi: 10.15352/afa/1396833503
    [10] P. Řehak, Hardy inequality on time scales and its application to half-linear dynamic equations, J. Inequal. Appl., 2005 (2005), 495–507. https://doi.org/10.1155/JIA.2005.495 doi: 10.1155/JIA.2005.495
    [11] H. M. Rezk, A. I. Saied, G. AlNemer, M. Zakarya, On Hardy-Knopp type inequalities with Kernels via time scale calculus, J. Math., 2022 (2022), 7997299. https://doi.org/10.1155/2022/7997299 doi: 10.1155/2022/7997299
    [12] S. H. Saker, E. Awwad, A. I. Saied, Some new dynamic inequalities involving monotonic functions on time scales, J. Funct. Space., 2019 (2019), 7584836. https://doi.org/10.1155/2019/7584836 doi: 10.1155/2019/7584836
    [13] S. H. Saker, A. I. Saied, M. Krnić, Some new dynamic Hardy-type inequalities with kernels involving monotone functions, Racsam Rev. R. Acad. A, 114 (2020), 1–16. https://doi.org/10.1007/s13398-020-00876-6 doi: 10.1007/s13398-020-00876-6
    [14] S. H. Saker, A. I. Saied, M. Krnić, Some new weighted dynamic inequalities for Monotone functions involving Kernels, Mediterr. J. Math., 17 (2020), 1–18. https://doi.org/10.1007/s00009-020-1473-0 doi: 10.1007/s00009-020-1473-0
    [15] S. H. Saker, J. Alzabut, A. I. Saied, D. O'Regan, New characterizations of weights on dynamic inequalities involving a Hardy operator, J. Inequal. Appl., 2021 (2021), 1–24. https://doi.org/10.1186/s13660-021-02606-x doi: 10.1186/s13660-021-02606-x
    [16] S. H. Saker, A. I. Saied, D. R. Anderson, Some new characterizations of weights in dynamic inequalities involving monotonic functions, Qual. Theory Dyn. Syst., 20 (2021), 1–22. https://doi.org/10.1007/s12346-021-00489-3 doi: 10.1007/s12346-021-00489-3
    [17] M. Bohner, T. Li, Kamenev-type criteria for nonlinear damped dynamic equations, Sci. China Math., 58 (2015), 1445–1452. https://doi.org/10.1007/s11425-015-4974-8 doi: 10.1007/s11425-015-4974-8
    [18] M. J. Huntul, Identifying an unknown heat source term in the third-order pseudoparabolic equation from nonlocal integral observation, Int. Commun. Heat Mass, 128 (2021), 105550. https://doi.org/10.1016/j.icheatmasstransfer.2021.105550 doi: 10.1016/j.icheatmasstransfer.2021.105550
    [19] M. J. Huntul, I. Tekin, On an inverse problem for a nonlinear third order in time partial differential equation, Results Appl. Math., 15 (2022), 100314. https://doi.org/10.1016/j.rinam.2022.100314 doi: 10.1016/j.rinam.2022.100314
    [20] M. J. Huntul, M. Abbas, An inverse problem of fourth-order partial differential equation with nonlocal integral condition, Adv. Contin. Discrete Models, 2022 (2022), 1–27. https://doi.org/10.1186/s13662-022-03727-3 doi: 10.1186/s13662-022-03727-3
    [21] W. W. Mohammed, F. M. Al-Askar, C. Cesarano, On the dynamical behavior of solitary waves for coupled stochastic Korteweg-De Vries equations, Mathematics, 11 (2023), 3506. https://doi.org/10.3390/math11163506 doi: 10.3390/math11163506
    [22] W. W. Mohammed, F. M. Al-Askar, C. Cesarano, E. S. Aly, The soliton solutions of the stochastic shallow water wave equations in the sense of Beta-derivative, Mathematics, 11 (2023), 1338. https://doi.org/10.3390/math11061338 doi: 10.3390/math11061338
    [23] M. Alshammari, N. Iqbal, W. W. Mohammed, T. Botmart, The solution of fractional-order system of KdV equations with exponential-decay kernel, Results Phys., 38 (2022), 105615. https://doi.org/10.1016/j.rinp.2022.105615 doi: 10.1016/j.rinp.2022.105615
    [24] C. Zhang, T. Li, S. H. Saker, Oscillation of fourth-order delay differential equations, J. Math. Sci., 201 (2014), 296–309. https://doi.org/10.1007/s10958-014-1990-0 doi: 10.1007/s10958-014-1990-0
    [25] Y. Tian, Y. Yang, X. Ma, X. Su, Stability of discrete-time delayed systems via convex function-based summation inequality, Appl. Math. Lett., 2023 (2023), 108764. https://doi.org/10.1016/j.aml.2023.108764 doi: 10.1016/j.aml.2023.108764
    [26] Y. Tian, X. Su, C. Shen, X. Ma, Exponentially extended Dissipativity-based filtering of switched neural networks, Automatica, 161 (2024), 111465. https://doi.org/10.1016/j.automatica.2023.111465 doi: 10.1016/j.automatica.2023.111465
    [27] M. Bohner, A. Peterson, Dynamic equations on time scales: An introduction with applications, Boston: Birkhäuser, 2001.
    [28] D. Anderson, J. Bullock, L. Erbe, A. Peterson, H. Tran, Nabla dynamic equations on time scales, Panam. Math. J., 13 (2003), 1–47.
    [29] A. F. Güvenilir, B. Kaymakçalan, N. N. Pelen, Constantin's inequality for nabla and diamond-alpha derivative, J. Inequal. Appl., 2015 (2015), 1–17. https://doi.org/10.1186/s13660-015-0681-9 doi: 10.1186/s13660-015-0681-9
    [30] R. P. Agarwal, D. O'Regan, S. H. Saker, Dynamic inequalities on time scales, New York: Springer Cham, 2014. https://doi.org/10.1007/978-3-319-11002-8
  • This article has been cited by:

    1. Hocine Makheloufi, Stability Analysis of Viscoelastic Swelling Porous Elastic Soils With Nonlinear Kelvin–Voigt Dampings, 2025, 0170-4214, 10.1002/mma.10889
  • Reader Comments
  • © 2024 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(1203) PDF downloads(93) Cited by(4)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog