Home Global bifurcation of coexistence states for a prey-predator model with prey-taxis/predator-taxis
Article Open Access

Global bifurcation of coexistence states for a prey-predator model with prey-taxis/predator-taxis

  • Shanbing Li EMAIL logo and Jianhua Wu
Published/Copyright: April 27, 2023

Abstract

This article is concerned with the stationary problem for a prey-predator model with prey-taxis/predator-taxis under homogeneous Dirichlet boundary conditions, where the interaction is governed by a Beddington-DeAngelis functional response. We make a detailed description of the global bifurcation structure of coexistence states and find the ranges of parameters for which there exist coexistence states. At the same time, some sufficient conditions for the nonexistence of coexistence states are also established. Our method of analysis uses the idea developed by Cintra et al. (Unilateral global bifurcation for a class of quasilinear elliptic systems and applications, J. Differential Equations 267 (2019), 619–657). Our results indicate that the presence of prey-taxis/predator-taxis makes mathematical analysis more difficult, and the Beddington-DeAngelis functional response leads to some different phenomena.

MSC 2010: 35J57; 35J60; 92D25

1 Introduction

The present article is concerned with the following Dirichlet problem of quasilinear elliptic equations:

(1.1) div ( d N N + β N N P ) = λ N N 2 N P 1 + m N + k P , x Ω , div ( d P P β P P N ) = μ P P 2 + γ N P 1 + m N + k P , x Ω , N = P = 0 , x Ω ,

where Ω is a bounded domain in R n with smooth boundary Ω ; is the gradient operator; div is the divergence operator. The coefficients d N , d P , γ , m , and k are positive constants, β N and β P are nonnegative constants, and λ and μ may change sign. System (1.1) is the stationary problem of a prey-predator model in which unknown functions N = N ( x ) and P = P ( x ) denote the stationary population densities of the prey and the predator in the habitat Ω , respectively. In the reaction terms, λ and μ are the growth rates of respective species; γ accounts for the intrinsic predation rate; the function N / ( 1 + m N + k P ) represents the functional response of the predator, which is the so-called “Beddington-DeAngelis response” introduced by Beddington [1] and DeAngelis et al. [9]. In the diffusion terms, d N div ( N ) and d P div ( P ) denote the linear diffusion of respective species, and d N and d P are the random-diffusion rates of respective species; β N div ( N P ) describes an ecological tendency such that the prey diffuses from the high-density area of the predator toward the low-density area of the predator, and β N is the intrinsic predator-taxis rate; β P div ( P N ) describes the tendency of the predator to move from the low-density area of the prey toward the high-density area of the prey, and β P is the intrinsic prey-taxis rate.

Such a prey-predator system with predator-taxis and prey-taxis was proposed by Tsyganov et al. [35]. It should be noted that the relevance of attractive prey-taxis (i.e., “predators move towards their prey”) was first biologically verified by Kareiva and Odell in their study of heterogeneous aggregative patterns [20], and the repulsive predator-taxis (i.e., “prey moves away from their predators”) has been detected for crayfish seeking shelter [14]. As far as we know, among prey-predator systems with predator-taxis or prey-taxis, the corresponding Neumann problem has been studied most extensively (see, for instance, [2,13,18,19,22,31,34,3639] and references therein). While less extensively studied than those with Neumann boundary conditions, those with Dirichlet conditions have been mathematically examined by Cintra et al. [46].

When there are no prey-taxis and predator-taxis effects (i.e., β N = β P = 0 ), (1.1) is reduced to the classical Beddington-DeAngeli prey-predator model, which has been extensively studied by Guo and Wu in [16,17]. They gave a good understanding of the existence, nonexistence, stability, and number of positive solutions for large m or k . However, to the best of our knowledge, there are few works in the field of reaction-diffusion systems, which specialize in problem (1.1) with prey-taxis/predator-taxis (i.e., β N > 0 or β P > 0 ). From a mathematical point of view, system (1.1) is much more challenging than those “only” containing random-diffusion (i.e., β N = β P = 0 ). For instance, for those “only” containing random-diffusion, the W 2 , p ( Ω ) -estimate of solutions “automatically” follows, provided the L ( Ω ) -estimate of solutions is obtained. However, the presence of prey-taxis/predator-taxis results in the fact that the L ( Ω ) -estimate of solutions does not “automatically” implies the W 2 , p ( Ω ) -estimate of solutions. As another example, the nonexistence of positive solutions to those “only” containing random-diffusion is usually easy to obtain by the monotonicity properties of principal eigenvalue. However, for system (1.1), the nonexistence of positive solutions is not trivial. Therefore, it is not too surprising that the analysis of system (1.1) is much less developed than those “only” containing random-diffusion.

Main results. To present our main results, we introduce some notations and basic facts. For any given p ( x ) C 1 , α ( Ω ¯ ) and q ( x ) , m ( x ) C α ( Ω ¯ ) , where α ( 0 , 1 ) , m ( x ) > 0 , and p ( x ) p 0 > 0 for x Ω , it is well known that the eigenvalue problem

div ( p ( x ) ϕ ) + q ( x ) ϕ = σ m ( x ) ϕ , x Ω , ϕ = 0 , x Ω

has an infinite sequence of eigenvalues that are bound from below. We denote the i -th eigenvalue by σ i [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] , where the first eigenvalue σ 1 [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] is a simple eigenvalue and the corresponding eigenfunction does not change sign in Ω . Particularly, if p ( x ) p 0 is a positive constant, then div ( p 0 ) = p 0 Δ , and hence, we will denote σ 1 [ div ( p 0 ) + q ( x ) ; m ( x ) ] simply by σ 1 [ p 0 Δ + q ( x ) ; m ( x ) ] . In addition, for any given p ( x ) C 1 , α ( Ω ¯ ) and b ( x ) C α ( Ω ¯ ) , where b ( x ) b 0 > 0 and p ( x ) p 0 > 0 for x Ω , the following logistic equation

div ( p ( x ) ϕ ) = a ϕ b ( x ) ϕ 2 , x Ω , ϕ = 0 , x Ω

admits a unique positive solution if and only if a > σ 1 [ div ( p ( x ) ) ; 1 ] , and we denote the unique positive solution by θ p , a , b . Particularly, if b ( x ) 1 in Ω , then we will denote θ p , a , 1 simply by θ p , a . Some further properties concerning σ 1 [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] and θ p , a , b will be presented in Section 2.

The first purpose of this article is to study the positive solutions of (1.1) in the case β N = 0 and β P > 0 . That is, we focus on the solutions to the following Dirichlet problem of quasilinear elliptic equations:

(1.2) d N Δ N = λ N N 2 N P 1 + m N + k P , x Ω , div ( d P P β P P N ) = μ P P 2 + γ N P 1 + m N + k P , x Ω , N = P = 0 , x Ω .

By regarding λ and μ as main parameters, the existence and nonexistence of positive solutions to (1.2) are summarized as follows.

Theorem 1.1

For any given d N , d P , β P , γ , m , and k , the following statements hold true.

  1. If λ d N σ 1 [ Δ ; 1 ] is fixed, then (1.2) has no positive solution.

  2. If d N σ 1 [ Δ ; 1 ] < λ < d N σ 1 [ Δ ; 1 ] + 1 / k is fixed, then

    1. there exists a positive number M = M ( d N , d P , β P , λ , γ , m , k ) large enough such that (1.2) has no positive solution if μ M ;

    2. there exists a continuum C 1 of positive solutions to (1.2) such that it bifurcates from the semi-trivial solution set { ( μ , θ d N , λ , 0 ) : μ R } at ( μ λ , θ d N , λ , 0 ) , and it is a smooth curve near the bifurcation point ( μ λ , θ d N , λ , 0 ) , where

      μ λ σ 1 div ( d P e ( β P / d P ) θ d N , λ ) γ θ d N , λ 1 + m θ d N , λ e ( β P / d P ) θ d N , λ ; e ( β P / d P ) θ d N , λ ;

    3. the continuum C 1 can be extended to a bounded global continuum of positive solutions to (1.2), which meets the other semi-trivial solution set { ( μ , 0 , θ d P , μ ) : μ > d P σ 1 [ Δ ; 1 ] } at ( μ , 0 , θ d P , μ ) , where μ is uniquely determined by λ = λ μ and λ μ is given as follows:

      λ μ σ 1 d N Δ + θ d P , μ 1 + k θ d P , μ ; 1 ,

      therefore, (1.2) admits at least one positive solution for μ ( min { μ λ , μ } , max { μ λ , μ } ) .

  3. If λ d N σ 1 [ Δ ; 1 ] + 1 / k is fixed, then

    1. there exists a positive number M = M ( d N , d P , β P , λ , γ , m , k ) large enough such that (1.2) has no positive solution if μ M ;

    2. the semi-trivial solution ( 0 , θ d P , μ ) is unstable for any μ > d P σ 1 [ Δ ; 1 ] , and there is no bifurcation of positive solutions occurring from the semi-trivial solution set { ( μ , 0 , θ d P , μ ) : μ > d P σ 1 [ Δ ; 1 ] } ;

    3. there exists an unbounded continuum C 2 of positive solutions to (1.2) such that it bifurcates from the semi-trivial solution set { ( μ , θ d N , λ , 0 ) : λ R } at ( μ λ , θ d N , λ , 0 ) and goes to as μ , and therefore, (1.2) admits at least one positive solution for μ ( μ λ , ) .

In Figure 1, an indication of the behavior of the global bifurcation branches of the positive solutions to (1.2) is shown, where drastic changes can be observed between the cases d N σ 1 [ Δ ; 1 ] < λ < d N σ 1 [ Δ ; 1 ] + 1 / k and λ d N σ 1 [ Δ ; 1 ] + 1 / k .

Figure 1 
               Possible bifurcation diagram of positive solutions to (1.2). (a) 
                     
                        
                        
                           
                              
                                 d
                              
                              
                                 N
                              
                           
                           
                              
                                 σ
                              
                              
                                 1
                              
                           
                           
                              [
                              
                                 −
                                 Δ
                                 ;
                                 
                                 1
                              
                              ]
                           
                           <
                           λ
                           <
                           
                              
                                 d
                              
                              
                                 N
                              
                           
                           
                              
                                 σ
                              
                              
                                 1
                              
                           
                           
                              [
                              
                                 −
                                 Δ
                                 ;
                                 
                                 1
                              
                              ]
                           
                           +
                           1
                           
                           /
                           
                           k
                        
                        {d}_{N}{\sigma }_{1}\left[-\Delta ;\hspace{0.33em}1]\lt \lambda \lt {d}_{N}{\sigma }_{1}\left[-\Delta ;\hspace{0.33em}1]+1\hspace{0.1em}\text{/}\hspace{0.1em}k
                     
                   (b) 
                     
                        
                        
                           λ
                           ≥
                           
                              
                                 d
                              
                              
                                 N
                              
                           
                           
                              
                                 σ
                              
                              
                                 1
                              
                           
                           
                              [
                              
                                 −
                                 Δ
                                 ;
                                 
                                 1
                              
                              ]
                           
                           +
                           1
                           
                           /
                           
                           k
                        
                        \lambda \ge {d}_{N}{\sigma }_{1}\left[-\Delta ;\hspace{0.33em}1]+1\hspace{0.1em}\text{/}\hspace{0.1em}k
                     
                  .
Figure 1

Possible bifurcation diagram of positive solutions to (1.2). (a) d N σ 1 [ Δ ; 1 ] < λ < d N σ 1 [ Δ ; 1 ] + 1 / k (b) λ d N σ 1 [ Δ ; 1 ] + 1 / k .

The second purpose of this article is to study the positive solutions of (1.1) in the case β N > 0 and β P = 0 . That is, we focus on the solutions to the following Dirichlet problem of quasilinear elliptic equations:

(1.3) div ( d N N + β N N P ) = λ N N 2 N P 1 + m N + k P , x Ω , d P Δ P = μ P P 2 + γ N P 1 + m N + k P , x Ω , N = P = 0 , x Ω .

By regarding λ and μ as main parameters, the existence and nonexistence of positive solutions to (1.3) are summarized as follows.

Theorem 1.2

For any given d N , d P , β N , γ , m , and k , the following statements hold true.

  1. If μ d P σ 1 [ Δ ; 1 ] γ / m is fixed, then (1.3) has no positive solution for all λ R .

  2. If d P σ 1 [ Δ ; 1 ] γ / m < μ < d P σ 1 [ Δ ; 1 ] is fixed, then

    1. (1.3) has no positive solution for λ 0 ;

    2. there exists a continuum C 3 of positive solutions to (1.3) such that it bifurcates from the semi-trivial solution set { ( λ , θ d N , λ , 0 ) : λ > d N σ 1 [ Δ ; 1 ] } at ( λ , θ d N , λ , 0 ) and it is a smooth curve near the bifurcation point ( λ , θ d N , λ , 0 ) , where λ is uniquely determined by μ = μ λ and μ λ is given by

      μ λ σ 1 d P Δ γ θ d N , λ 1 + m θ d N , λ ; 1 ;

    3. the continuum C 3 can be extended to an unbounded global continuum of positive solutions to (1.3)and it goes to as λ , therefore, (1.3) admits at least one positive solution for λ ( λ , ) .

  3. If μ > d P σ 1 [ Δ ; 1 ] is fixed, then

    1. (1.3) has no positive solution for λ 0 ;

    2. there exists a continuum C 4 of positive solutions to (1.3) such that it bifurcates from the semi-trivial solution set { ( λ , 0 , θ d P , μ ) : λ R } at ( λ μ , 0 , θ d P , μ ) and it is a smooth curve near the bifurcation point ( λ μ , 0 , θ d P , μ ) , where

      λ μ σ 1 div ( d N e ( β N / d N ) θ d P , μ ) + θ d P , μ 1 + k θ d P , μ e ( β N / d N ) θ d P , μ ; e ( β N / d N ) θ d P , μ ;

    3. the continuum C 4 can be extended to an unbounded global continuum of positive solutions to (1.3)and it goes to as λ , and therefore, (1.3) admits at least one positive solution for λ ( λ μ , ) .

In Figure 2, an indication of the behavior of the global bifurcation branches of the positive solutions to (1.3) is shown, where the differences can be observed between the cases d P σ 1 [ Δ ; 1 ] γ / m < μ < d P σ 1 [ Δ ; 1 ] and μ > d P σ 1 [ Δ ; 1 ] .

Figure 2 
               Possible bifurcation diagram of positive solutions to (1.2). (a) 
                     
                        
                        
                           
                              
                                 d
                              
                              
                                 P
                              
                           
                           
                              
                                 σ
                              
                              
                                 1
                              
                           
                           
                              [
                              
                                 −
                                 Δ
                                 ;
                                 
                                 1
                              
                              ]
                           
                           −
                           γ
                           
                           /
                           
                           m
                           <
                           μ
                           <
                           
                              
                                 d
                              
                              
                                 P
                              
                           
                           
                              
                                 σ
                              
                              
                                 1
                              
                           
                           
                              [
                              
                                 −
                                 Δ
                                 ;
                                 
                                 1
                              
                              ]
                           
                        
                        {d}_{P}{\sigma }_{1}\left[-\Delta ;\hspace{0.33em}1]-\gamma \hspace{0.1em}\text{/}\hspace{0.1em}m\lt \mu \lt {d}_{P}{\sigma }_{1}\left[-\Delta ;\hspace{0.33em}1]
                     
                   (b) 
                     
                        
                        
                           μ
                           >
                           
                              
                                 d
                              
                              
                                 P
                              
                           
                           
                              
                                 σ
                              
                              
                                 1
                              
                           
                           
                              [
                              
                                 −
                                 Δ
                                 ;
                                 
                                 1
                              
                              ]
                           
                        
                        \mu \gt {d}_{P}{\sigma }_{1}\left[-\Delta ;\hspace{0.33em}1]
                     
                  .
Figure 2

Possible bifurcation diagram of positive solutions to (1.2). (a) d P σ 1 [ Δ ; 1 ] γ / m < μ < d P σ 1 [ Δ ; 1 ] (b) μ > d P σ 1 [ Δ ; 1 ] .

Remark 1.1

The unilateral bifurcation result used in this article is developed by Cintra et al. [5] (see Theorems 1.1 and 1.2 in [5], which is based on the unilateral bifurcation result of [23]). Precisely, Cintra et al. have extended the unilateral bifurcation result for semilinear elliptic systems to quasilinear elliptic systems. It is worth noting that another interesting bifurcation result for quasilinear elliptic systems developed by Shi and Wang has been widely used in recent years (see Theorem 4.4 in [32], which is based on Degree theory for C 1 Fredholm mappings of index 0 of [29] and the unilateral bifurcation result of [23]). For more details on the unilateral bifurcation result, one can refer to the results established by Rabinowitz [30], López-Gómez [25], López-Gómez and Mora-Corral [26], and Pejsachowicz and Rabier [29].

The contents of this article are as follows: in Section 2, we state some preliminary results that will be used repeatedly in later discussions. In Section 3, we study (1.2), where for both cases d N σ 1 [ Δ ; 1 ] < λ < d N σ 1 [ Δ ; 1 ] + 1 / k and λ d N σ 1 [ Δ ; 1 ] + 1 / k , the structure of positive solutions are investigated. Our analysis is based on the a priori estimate results (Propositions 3.1 and 3.2), the nonexistence results (Propositions 3.3 and 3.4), and a global bifurcation method adapted from [5]. In Section 4, we carry out a similar analysis for (1.3), but the phenomena revealed there are quite different from those in Section 3.

2 Preliminaries

In this section, we will introduce two important lemmas that will be used repeatedly throughout this article. The first lemma provides some important properties of the principal eigenvalue σ 1 [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] (see, for instance, Lemma 2.2 in [4]).

Lemma 2.1

For any fixed p ( x ) C 1 , α ( Ω ¯ ) and q ( x ) , m ( x ) C α ( Ω ¯ ) , where α ( 0 , 1 ) , m ( x ) > 0 , and p ( x ) p 0 > 0 for x Ω , the linear eigenvalue problem

div ( p ( x ) ϕ ) + q ( x ) ϕ = σ m ( x ) ϕ , x Ω , ϕ = 0 , x Ω

admits a principal eigenvalue, which is denoted by σ 1 [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] . Moreover, the principal eigenvalue satisfies

σ 1 [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] = inf ψ H 0 1 ( Ω ) , ψ 0 Ω p ( x ) ψ 2 d x + Ω q ( x ) ψ 2 d x Ω m ( x ) ψ 2 d x .

Furthermore, the following monotonicity properties hold:

  1. σ 1 [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] is increasing with respect to p ( x ) ;

  2. σ 1 [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] is increasing with respect to q ( x ) ;

  3. the monotonicity of σ 1 [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] with respect to m ( x ) depends on the sign of σ 1 [ div ( p ( x ) ) + q ( x ) ; 1 ] , and

    1. if σ 1 [ div ( p ( x ) ) + q ( x ) ; 1 ] > 0 , then σ 1 [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] is positive and decreasing with respect to m ( x ) ;

    2. if σ 1 [ div ( p ( x ) ) + q ( x ) ; 1 ] = 0 , then σ 1 [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] = 0 for every m ( x ) ;

    3. if σ 1 [ div ( p ( x ) ) + q ( x ) ; 1 ] < 0 , then σ 1 [ div ( p ( x ) ) + q ( x ) ; m ( x ) ] is negative and increasing with respect to m ( x ) .

The second lemma provides some information on the diffusive logistic equation (see, for instance, Lemma 2.3 in [4]).

Lemma 2.2

For any fixed p ( x ) C 1 , α ( Ω ¯ ) and b ( x ) C α ( Ω ¯ ) , where α ( 0 , 1 ) , b ( x ) b 0 > 0 , and p ( x ) p 0 > 0 for x Ω , the diffusive logistic equation

div ( p ( x ) ϕ ) = ( a b ( x ) ϕ ) ϕ , x Ω , ϕ = 0 , x Ω

admits a unique positive solution, denoted by θ p , a , b , if and only if a > σ 1 [ div ( p ( x ) ) ; 1 ] . Moreover, the map a θ p , a , b is continuous and increasing from ( σ 1 [ div ( p ( x ) ) ; 1 ] , ) to C 2 , α ( Ω ¯ ) , and it satisfies

a σ 1 [ div ( p ( x ) ) ; 1 ] b ( x ) C ( Ω ¯ ) ϕ a C ( Ω ¯ ) ϕ a θ p , a , b a b 0 in Ω ,

where ϕ a is a principal eigenfunction associated to σ 1 [ div ( p ( x ) ) ; 1 ] .

If P = 0 in Ω , then N satisfies

(2.1) d N Δ N = λ N N 2 , x Ω , N = 0 , x Ω .

By virtue of Lemma 2.2, it is clear that (2.1) admits a unique positive solution θ d N , λ if and only if λ > d N σ 1 [ Δ ; 1 ] . Likewise, if N = 0 in Ω , then P satisfies

(2.2) d P Δ P = μ P P 2 , x Ω , P = 0 , x Ω ,

and (2.2) admits a unique positive solution θ d P , μ if and only if μ > d P σ 1 [ Δ ; 1 ] .

3 Structure of solutions for β N = 0 and β P > 0

This section is devoted to the understanding of the global bifurcation structure of the set of positive solutions to (1.2) by treating μ as the main bifurcation parameter.

3.1 A priori estimates

The main purpose of this subsection is to prove some a priori estimates of positive solutions. The first proposition of this subsection gives the L ( Ω ) -estimate of any positive solution.

Proposition 3.1

If λ d N σ 1 [ Δ ; 1 ] is fixed, then (1.2) has no positive solution. If λ > d N σ 1 [ Δ ; 1 ] is fixed, then any positive solution ( N , P ) of (1.2) satisfies

N ( x ) θ d N , λ λ and P ( x ) e ( β P / d P ) λ ( μ + γ / m )

for all x Ω .

Proof

Observe that N is a subsolution of the equation (2.1), and then N θ d N , λ λ in Ω . Since θ d N , λ = 0 in Ω if λ d N σ 1 [ Δ ; 1 ] , this implies that N = 0 in Ω if λ d N σ 1 [ Δ ; 1 ] . Consequently, when λ d N σ 1 [ Δ ; 1 ] , (1.2) has no positive solution.

When λ > d N σ 1 [ Δ ; 1 ] is fixed, we assume x 1 Ω ¯ is a maximum point of N , i.e., N ( x 1 ) = max Ω ¯ N ( x ) > 0 . It is clear that x 1 Ω , and hence,

0 d N Δ N ( x 1 ) = λ N ( x 1 ) P ( x 1 ) 1 + m N ( x 1 ) + k P ( x 1 ) N ( x 1 ) .

Thus, we have

N ( x 1 ) λ P ( x 1 ) 1 + m N ( x 1 ) + k P ( x 1 ) λ .

This implies that N ( x ) λ for all x Ω if λ > d N σ 1 [ Δ ; 1 ] is fixed.

Let us denote

W e ( β P / d P ) N P .

The second equation with the boundary condition of (1.2) can be rewritten as follows:

(3.1) div ( d P e ( β P / d P ) N W ) = e ( β P / d P ) N W μ P + γ N 1 + m N + k P , x Ω , W = 0 , x Ω .

Suppose that x 2 Ω ¯ is a maximum point of W , i.e., W ( x 2 ) = max Ω ¯ W ( x ) > 0 . Then, x 2 Ω , and hence, W ( x 2 ) = 0 and Δ W ( x 2 ) 0 . A simple calculation provides

div ( d P e ( β P / d P ) N W ) x = x 2 = β P e ( β P / d P ) N N W x = x 2 + d P e ( β P / d P ) N Δ W x = x 2 0 .

By virtue of (3.1), we have

P ( x 2 ) μ + γ N ( x 2 ) 1 + m N ( x 2 ) + k P ( x 2 ) μ + γ / m ,

and so,

W ( x 2 ) = e ( β P / d P ) N ( x 2 ) P ( x 2 ) P ( x 2 ) μ + γ / m .

Thus, W ( x ) μ + γ / m for all x Ω . Since P = e ( β P / d P ) N W in Ω , we obtain

P ( x ) e ( β P / d P ) max Ω ¯ N ( x ) max Ω ¯ W ( x ) e ( β P / d P ) λ ( μ + γ / m )

for all x Ω .□

The next proposition of this subsection gives the W 2 , p ( Ω ) -estimate of any positive solution.

Proposition 3.2

Let λ > d N σ 1 [ Δ ; 1 ] be fixed. Assume that ( N , P ) is any positive solution of (1.2). Then, for any p ( 1 , ) , there exists a positive constant M, depending on the parameters of system (1.2), such that

N W 2 , p ( Ω ) M and P W 2 , p ( Ω ) M .

Proof

For simplicity, we denote the positive constants by M i depending on the parameters of system (1.2). By virtue of Proposition 3.1, there exists a positive constant M 1 such that

1 d N λ N N 2 N P 1 + m N + k P L p ( Ω ) M 1

for any p ( 1 , ) . It follows from L p -estimate for elliptic equations [15] that N W 2 , p ( Ω ) is bound for any p ( 1 , ) . That is, there exists a positive constant M 2 such that N W 2 , p ( Ω ) M 2 . Thus, the Sobolev embedding theorem ensures that there exists a positive constant M 3 such that N C 1 ( Ω ¯ ) M 3 .

As above, we apply the elliptic regularity [15] to (3.1) to conclude that there exists a positive constant M 4 such that

e ( β P / d P ) N P C 1 ( Ω ¯ ) = W C 1 ( Ω ¯ ) M 4 .

Since

( e ( β P / d P ) N P ) = e ( β P / d P ) N P ( β P / d P ) P e ( β P / d P ) N N ,

the triangular inequality and Proposition 3.1 yield

e ( β P / d P ) λ P e ( β P / d P ) N P ( e ( β P / d P ) N P ) + ( β P / d P ) P e ( β P / d P ) N N .

Thus, there exists a positive constant M 5 such that P M 5 . In view of Proposition 3.1, there exists a positive constant M 6 such that P C 1 ( Ω ¯ ) M 6 .

Note that the second equation of (1.2) can be expressed as follows:

Δ P = 1 d P β P P N + β P d N λ N N 2 N P 1 + m N + k P P + μ P P 2 + γ N P 1 + m N + k P , x Ω , P = 0 , x Ω .

Since N C 1 ( Ω ¯ ) M 3 and P C 1 ( Ω ¯ ) M 6 , there exists a positive constant M 7 such that

1 d P β P P N + β P d N λ N N 2 N P 1 + m N + k P P + μ P P 2 + γ N P 1 + m N + k P L p ( Ω ) M 7

for any p ( 1 , ) . Hence, it follows from L p -estimates for elliptic equations that P W 2 , p ( Ω ) is bounded for any p ( 1 , ) . The desired estimate is derived.□

3.2 Nonexistence of positive solutions

The main purpose of this subsection is to study the nonexistence of positive solutions. By virtue of Proposition 3.1, one has known that (1.2) has no positive solution for λ d N σ 1 [ Δ ; 1 ] . The following proposition asserts that for any fixed λ > d N σ 1 [ Δ ; 1 ] , (1.2) has no positive solution if μ is too small.

Proposition 3.3

If λ > d N σ 1 [ Δ ; 1 ] is fixed, then there exists a constant M ̲ = M ̲ ( d P , β P , λ , γ ) such that (1.2) has no positive solution for μ M ̲ .

Proof

Assume that ( N , P ) is a positive solution of (1.2). It follows from (3.1) that W satisfies

div ( d P e ( β P / d P ) N W ) + P γ N 1 + m N + k P e ( β P / d P ) N W = μ e ( β P / d P ) N W , x Ω , W = 0 , x Ω ,

where W = e ( β P / d P ) N P . Since ( N , P ) is a positive solution of (1.2), it is clear that W > 0 in Ω . Thus, the Krein-Rutman theorem implies that

μ = σ 1 div ( d P e ( β P / d P ) N ) + P γ N 1 + m N + k P e ( β P / d P ) N ; e ( β P / d P ) N .

By virtue of Lemma 2.1 and Proposition 3.1, we have

μ > σ 1 [ d P Δ γ λ e ( β P / d P ) λ ; e ( β P / d P ) N ] .

Moreover, Lemma 2.1 also shows that the monotonicity of σ 1 [ d P Δ γ λ e ( β P / d P ) λ ; e ( β P / d P ) N ] with respect to e ( β P / d P ) N is determined by the sign σ 1 [ d P Δ γ λ e ( β P / d P ) λ ; 1 ] . Thus,

μ > σ 1 [ d P Δ γ λ e ( β P / d P ) λ ; e ( β P / d P ) λ ] , if σ 1 [ d P Δ γ λ e ( β P / d P ) λ ; 1 ] > 0 , 0 , if σ 1 [ d P Δ γ λ e ( β P / d P ) λ ; 1 ] = 0 , σ 1 [ d P Δ γ λ e ( β P / d P ) λ ; 1 ] , if σ 1 [ d P Δ γ λ e ( β P / d P ) λ ; 1 ] < 0 .

Consequently, there exists a constant M ̲ = M ̲ ( d P , β P , λ , γ ) such that if (1.2) has a positive solution ( N , P ) , then μ > M ̲ . In other words, (1.2) has no positive solution for any μ M ̲ .□

For any fixed λ ( d N σ 1 [ Δ ; 1 ] , d N σ 1 [ Δ ; 1 ] + 1 / k ) , the next proposition asserts that (1.2) has no positive solution if μ is too large.

Proposition 3.4

If λ ( d N σ 1 [ Δ ; 1 ] , d N σ 1 [ Δ ; 1 ] + 1 / k ) is fixed, then there exists a positive constant M ¯ = M ¯ ( d N , d P , β P , λ , γ , m , k ) such that (1.2) has no positive solution for μ M ¯ .

Proof

To achieve the proof, we will adapt the proof of Lemma 5.5 of [5], which in turn came from Proposition 6.5 of [10]. Suppose the conclusion is false. Then (1.2) admits at least one positive solution ( N , P ) for all μ > 0 large. To make the proof more clear, we divide the proof into the following steps.

Step 1: It follows from Proposition 3.1 that

1 e ( β P / d P ) N e ( β P / d P ) λ in Ω .

In view of (3.1), we deduce that

div ( d P e ( β P / d P ) N W ) = e ( β P / d P ) N W μ P + γ N 1 + m N + k P = μ e ( β P / d P ) N W e 2 ( β P / d P ) N W 2 + γ N e ( β P / d P ) N W 1 + m N + k e ( β P / d P ) N W μ W e 2 ( β P / d P ) λ W 2 in Ω .

Thus, W is a super-solution of

(3.2) div ( d P e ( β P / d P ) N ϕ ) = ( μ e 2 ( β P / d P ) λ ϕ ) ϕ , x Ω , ϕ = 0 , x Ω .

By virtue of Lemma 2.2, (3.2) has a unique positive solution if μ > σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] , and it is θ d P e ( β P / d P ) N , μ , e 2 ( β P / d P ) λ . Moreover, Lemma 2.2 also shows that

μ σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] e 2 ( β P / d P ) λ ϕ μ C ( Ω ¯ ) ϕ μ θ d P e ( β P / d P ) N , μ , e 2 ( β P / d P ) λ in Ω ,

where ϕ μ with ϕ μ L 2 ( Ω ) = 1 is the corresponding eigenfunction associated to σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] . It is worth noting that ϕ μ depends on μ since N depends on μ . In view of the proof of Lemma 2.3 in [4], one sees that

μ σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] e 2 ( β P / d P ) λ ϕ μ C ( Ω ¯ ) ϕ μ

is a sub-solution of (3.2). Thus, the sub-supersolution method and the uniqueness of the positive solution of (3.2) show that

μ σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] e 2 ( β P / d P ) λ ϕ μ C ( Ω ¯ ) ϕ μ θ d P e ( β P / d P ) N , μ , e 2 ( β P / d P ) λ W in Ω .

Furthermore, it follows from Lemma 2.1 and Proposition 3.1 that

σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] σ 1 [ div ( d P e ( β P / d P ) λ ) ; 1 ] = d P e ( β P / d P ) λ σ 1 [ Δ ; 1 ] .

This, together with W = e ( β P / d P ) N P , implies that

μ d P e ( β P / d P ) λ σ 1 [ Δ ; 1 ] e 2 ( β P / d P ) λ ϕ μ C ( Ω ¯ ) ϕ μ θ d P e ( β P / d P ) N , μ , e 2 ( β P / d P ) λ W P in Ω .

That is, if μ > σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] , then

P μ d P e ( β P / d P ) λ σ 1 [ Δ ; 1 ] e 2 ( β P / d P ) λ ϕ μ C ( Ω ¯ ) ϕ μ in Ω .

Consequently, we prove that P has a positive lower bound.

Step 2: For writing convenience, we denote

ρ ( μ ) μ d P e ( β P / d P ) λ σ 1 [ Δ ; 1 ] e 2 ( β P / d P ) λ ϕ μ C ( Ω ¯ ) .

According to Theorem 4.1 in [33], it is well known that ϕ μ C ( Ω ¯ ) is uniformly bound with respect to μ . That is, there exists a positive constant M , which is independent of μ , such that ϕ μ C ( Ω ¯ ) M . This ensures we can obtain

ρ ( μ ) μ d P e ( β P / d P ) λ σ 1 [ Δ ; 1 ] e 2 ( β P / d P ) λ M as μ .

By virtue of the first equation of (1.2), we obtain from Lemma 2.1 that

λ = σ 1 d N Δ + N + P 1 + m N + k P ; 1 > σ 1 d N Δ + ρ ( μ ) ϕ μ 1 + m λ + k ρ ( μ ) ϕ μ ; 1 .

We further denote

t ( μ ) σ 1 d N Δ + ρ ( μ ) ϕ μ 1 + m λ + k ρ ( μ ) ϕ μ ; 1 .

Consequently, we prove that λ > t ( μ ) for all large μ .

Step 3: We next want to show

t ( μ ) d N σ 1 [ Δ ; 1 ] + 1 / k as μ .

If so, then we obtain a contradiction since λ ( d N σ 1 [ Δ ; 1 ] , d N σ 1 [ Δ ; 1 ] + 1 / k ) is fixed, and hence, the proof is complete by contradiction.

By virtue of the variational characterization of the principal eigenvalue (see Lemma 2.1), it follows that

t ( μ ) = inf ψ H 0 1 ( Ω ) , ψ 0 d N Ω ψ 2 d x + Ω ρ ( μ ) ϕ μ 1 + m λ + k ρ ( μ ) ϕ μ ψ 2 d x Ω ψ 2 d x .

Thereby, for all ψ H 0 1 ( Ω ) and ψ 0 , we have

t ( μ ) d N Ω ψ 2 d x + Ω ρ ( μ ) ϕ μ 1 + m λ + k ρ ( μ ) ϕ μ ψ 2 d x Ω ψ 2 d x d N Ω ψ 2 d x + Ω 1 k ψ 2 d x Ω ψ 2 d x .

Taking the infimum for all ψ H 0 1 ( Ω ) and ψ 0 , we find that t ( μ ) d N σ 1 [ Δ ; 1 ] + 1 / k , and hence,

limsup μ t ( μ ) d N σ 1 [ Δ ; 1 ] + 1 / k .

On the other hand,

t ( μ ) inf ψ H 0 1 ( Ω ) , ψ 0 d N Ω ψ 2 d x Ω ψ 2 d x = d N σ 1 [ Δ ; 1 ] .

This implies that t ( μ ) is bounded for all μ > σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] , and hence, there exists a sequence ψ μ H 0 1 ( Ω ) with ψ μ L 2 ( Ω ) = 1 such that

(3.3) t ( μ ) = d N Ω ψ μ 2 d x + Ω ρ ( μ ) ϕ μ 1 + m λ + k ρ ( μ ) ϕ μ ψ μ 2 d x .

Since t ( μ ) is bounded, (3.3) implies that ψ μ is bounded in H 0 1 ( Ω ) . Consequently, by choosing a subsequence if necessary, there exists some nonnegative function ψ with ψ L 2 ( Ω ) = 1 such that

ψ μ ψ weakly in H 0 1 ( Ω ) and ψ μ ψ in L 2 ( Ω )

as μ .

In order to take the limit in (3.3) with respect to μ , we need to know the limit of ϕ μ as μ . Recall that ϕ μ with ϕ μ L 2 ( Ω ) = 1 is a principal eigenfunction associated with σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] . Then, ϕ μ satisfies

(3.4) div ( d P e ( β P / d P ) N ϕ μ ) = σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] ϕ μ , x Ω , ϕ μ = 0 , x Ω .

By virtue of Lemma 2.1 and the inequality 1 e ( β P / d P ) N e ( β P / d P ) λ , we have

d P σ 1 [ Δ ; 1 ] σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] d P e ( β P / d P ) λ σ 1 [ Δ ; 1 ] .

This means that the principal eigenvalue σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] is bounded, and hence, there exists some number σ ˜ [ d P σ 1 [ Δ ; 1 ] , d P e ( β P / d P ) λ σ 1 [ Δ ; 1 ] ] such that

σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] σ ˜ as μ ,

by choosing a subsequence if necessary. On the other hand, we multiply both sides of (3.4) by ϕ μ and integrate the resulting expression over Ω to obtain

d P Ω ϕ μ 2 d x Ω d P e ( β P / d P ) N ϕ μ 2 d x = σ 1 [ div ( d P e ( β P / d P ) N ) ; 1 ] Ω ϕ μ 2 d x d P e ( β P / d P ) λ σ 1 [ Δ ; 1 ] Ω ϕ μ 2 d x .

Since ϕ μ L 2 ( Ω ) = 1 and d P > 0 , it is clear that ϕ μ is bounded in H 0 1 ( Ω ) . Consequently, by choosing a subsequence if necessary, there exists some nonnegative function ϕ with ϕ L 2 ( Ω ) = 1 such that

ϕ μ ϕ weakly in H 0 1 ( Ω ) and ϕ μ ϕ in L 2 ( Ω )

as μ . It is noted that (3.4) is verified in H 1 ( Ω ) . Thus, the homogenization technique (see, for instance, Theorem 2.1 in [21]) shows that the following equation is verified in H 1 ( Ω ) :

div ( A ϕ ) = σ ˜ ϕ , x Ω , ϕ = 0 , x Ω ,

where A ( L ( Ω ) ) N × N is a uniformly elliptic symmetric matrix. Because σ ˜ ϕ 0 ( 0 ) , we apply the strong maximum principle (see, for instance, [24]) to derive ϕ > 0 in Ω .

We now begin to take the limit in (3.3) with respect to μ . Note that

Ω ρ ( μ ) ϕ μ 1 + m λ + k ρ ( μ ) ϕ μ ψ μ 2 d x Ω 1 k ψ 2 d x Ω ρ ( μ ) ϕ μ 1 + m λ + k ρ ( μ ) ϕ μ ( ψ μ 2 ψ 2 ) d x + Ω ρ ( μ ) ϕ μ 1 + m λ + k ρ ( μ ) ϕ μ 1 k ψ 2 d x 1 k ψ μ + ψ L 2 ( Ω ) ψ μ ψ L 2 ( Ω ) + Ω ρ ( μ ) ϕ μ 1 + m λ + k ρ ( μ ) ϕ μ 1 k ψ 2 d x .

In addition, Lebesgue’s dominated convergence theorem ensures that

lim μ Ω ρ ( μ ) ϕ μ 1 + m λ + k ρ ( μ ) ϕ μ ψ 2 d x = Ω 1 k ψ 2 d x .

Thus, we derive

lim μ Ω ρ ( μ ) ϕ μ 1 + m λ + k ρ ( μ ) ϕ μ ψ μ 2 d x = Ω 1 k ψ 2 d x .

In view of (3.3), it follows from Poincaré inequality ψ L 2 ( Ω ) 2 σ 1 [ Δ ; 1 ] ψ L 2 ( Ω ) 2 and classic inequality liminf μ ψ μ L 2 ( Ω ) ψ L 2 ( Ω ) that

liminf μ t ( μ ) d N Ω ψ 2 d x + Ω 1 k ψ 2 d x d N σ 1 [ Δ ; 1 ] + 1 / k .

By summarizing the above analysis, we obtain

lim μ t ( μ ) = d N σ 1 ( Δ ; 1 ) + 1 / k .

This completes the whole proof of this proposition.□

3.3 Bifurcation structure of positive solutions

The main purpose of this subsection is to investigate the bifurcation structure of positive solutions to (1.2) by regarding μ as a bifurcation parameter and fixing all other constants.

For any μ R , (1.2) has a trivial solution branch Γ 0 { ( μ , 0 , 0 ) : μ R } . As μ increases across d P σ 1 [ Δ ; 1 ] , there is a semi-trivial solution branch

Γ P { ( μ , 0 , θ d P , μ ) : μ > d P σ 1 [ Δ ; 1 ] } ,

which bifurcates from Γ 0 . By virtue of Proposition 3.1, one has known that if λ d N σ 1 [ Δ ; 1 ] is fixed, then (1.2) has no positive solution. Thus, all nonnegative solutions of (1.2) lie on either Γ 0 or Γ P . In what follows, we always assume that λ > d N σ 1 [ Δ ; 1 ] is fixed so that (1.2) has a semi-trivial solution branch

Γ N { ( μ , θ d N , λ , 0 ) : μ R } .

We now apply the result of Crandall and Rabinowitz [7] on bifurcation from a simple eigenvalue to obtain a local result on bifurcation from Γ N . Let X = W 2 , p ( Ω ) W 0 1 , p ( Ω ) with p > n . We define an operator

L : R × X × X X × X

given by

L ( μ , N , P ) = d N Δ N λ N + N 2 + N P 1 + m N + k P div ( d P P β P P N ) μ P + P 2 γ N P 1 + m N + k P .

Clearly, ( N , P ) X × X is a nonnegative solution of (1.2) if and only if L ( μ , N , P ) = 0 .

In order to find a bifurcation point on the semi-trivial solution branch Γ N from which positive solutions of (1.2) bifurcate, the necessary condition for bifurcation is that L ( N , P ) ( μ , θ d N , λ , 0 ) is degenerate, where L ( N , P ) ( μ , θ d N , λ , 0 ) is the linearization of L ( μ , N , P ) with respect to ( N , P ) at ( θ d N , λ , 0 ) and is given by

L ( N , P ) ( μ , θ d N , λ , 0 ) = d N Δ λ + 2 θ d N , λ θ d N , λ 1 + m θ d N , λ 0 div ( d P β P θ d N , λ ) μ γ θ d N , λ 1 + m θ d N , λ .

By setting L ( N , P ) ( μ , θ d N , λ , 0 ) ( ϕ , ψ ) = 0 , we have

d N Δ ϕ + ( 2 θ d N , λ λ ) ϕ = θ d N , λ 1 + m θ d N , λ ψ , x Ω , div ( d P ψ β P θ d N , λ ψ ) γ θ d N , λ 1 + m θ d N , λ ψ = μ ψ , x Ω , ϕ = ψ = 0 , x Ω .

Let Ψ = ψ e ( β P / d P ) θ d N , λ . Then this system can be rewritten as follows:

(3.5) d N Δ ϕ + ( 2 θ d N , λ λ ) ϕ = θ d N , λ 1 + m θ d N , λ e ( β P / d P ) θ d N , λ Ψ , x Ω , div ( d P e ( β P / d P ) θ d N , λ Ψ ) γ θ d N , λ 1 + m θ d N , λ e ( β P / d P ) θ d N , λ Ψ = μ e ( β P / d P ) θ d N , λ Ψ , x Ω , ϕ = Ψ = 0 , x Ω .

Since θ d N , λ is the unique positive solution of (2.1), it is clear that σ 1 [ d N Δ + θ d N , λ λ ; 1 ] = 0 . Moreover, Lemma 2.1 shows that

σ 1 [ d N Δ + 2 θ d N , λ λ ; 1 ] > σ 1 [ d N Δ + θ d N , λ λ ; 1 ] = 0 .

This ensures the invertibility of the operator d N Δ + 2 θ d N , λ λ : X X . Consequently, (3.5) is solvable, provided the second equation of (3.5) has a solution. Since we hope to obtain positive solutions, the bifurcation should occur at the principal eigenvalue, which ensures that the eigenfunction is positive. In view of the Krein-Rutman theorem, the second equation of (3.5) has positive solutions if and only if μ = μ λ . Let Ψ μ λ be the positive eigenfunction associated to μ λ . Then,

ker [ L ( N , P ) ( μ λ , θ d N , λ , 0 ) ] = span ( ϕ μ λ , ψ μ λ ) ,

where

ψ μ λ = e ( β P / d P ) θ d N , λ Ψ μ λ , ϕ μ λ = ( d N Δ + 2 θ d N , λ λ ) 1 θ d N , λ 1 + m θ d N , λ e ( β P / d P ) θ d N , λ Ψ μ λ .

This shows that ker [ L ( N , P ) ( μ λ , θ d N , λ , 0 ) ] is one-dimensional.

We now show that codim Range [ L ( N , P ) ( μ λ , θ d N , λ , 0 ) ] = 1 . Suppose ( h , k ) Range [ L ( N , P ) ( μ λ , θ d N , λ , 0 ) ] . Then, there exists ( ϕ , ψ ) X × X such that

d N Δ ϕ + ( 2 θ d N , λ λ ) ϕ + θ d N , λ 1 + m θ d N , λ ψ = h , x Ω , div ( d P ψ β P θ d N , λ ψ ) μ λ + γ θ d N , λ 1 + m θ d N , λ ψ = k , x Ω , ϕ = ψ = 0 , x Ω .

As above, we set Ψ = ψ e ( β P / d P ) θ d N , λ . Then, Ψ satisfies

(3.6) div ( d P e ( β P / d P ) θ d N , λ Ψ ) μ λ + γ θ d N , λ 1 + m θ d N , λ e ( β P / d P ) θ d N , λ Ψ = k , x Ω , Ψ = 0 , x Ω .

Since the operator

div ( d P e ( β P / d P ) θ d N , λ ) μ λ + γ θ d N , λ 1 + m θ d N , λ e ( β P / d P ) θ d N , λ : X X

is self-adjoint, it follows from the Fredholm alternative theorem that (3.6) has a solution Ψ if and only if

Ω k Ψ μ λ d x = 0 .

Thus, the invertibility of the operator d N Δ + 2 θ d N , λ λ : X X ensures that

ϕ = ( d N Δ + 2 θ d N , λ λ ) 1 h θ d N , λ 1 + m θ d N , λ e ( β P / d P ) θ d N , λ Ψ .

Therefore,

Range [ L ( N , P ) ( μ λ , θ d N , λ , 0 ) ] = { span ( 0 , Ψ μ λ ) } .

The desired result is obtained.

We further check the transversality condition:

(3.7) L ( N , P ) , μ ( μ λ , θ d N , λ , 0 ) ϕ μ λ ψ μ λ Range [ L ( N , P ) ( μ λ , θ d N , λ , 0 ) ] .

By a simple calculation, we have

L ( N , P ) , μ ( μ λ , θ d N , λ , 0 ) ϕ μ λ ψ μ λ = 0 ψ μ λ .

Suppose that (3.7) is not true. Then, there exists ( ϕ , ψ ) X × X such that

d N Δ ϕ + ( 2 θ d N , λ λ ) ϕ + θ d N , λ 1 + m θ d N , λ ψ = 0 , x Ω , div ( d P ψ β P θ d N , λ ψ ) μ λ + γ θ d N , λ 1 + m θ d N , λ ψ = ψ μ λ , x Ω , ϕ = ψ = 0 , x Ω .

As above, if this system admits a solution ( ϕ , ψ ) , then the following identity should be true:

Ω ψ μ λ Ψ μ λ d x = Ω e ( β P / d P ) θ d N , λ Ψ μ λ 2 d x = 0 .

This is a contradiction as Ψ μ λ > 0 in Ω .

According to the local bifurcation theorem of Crandall and Rabinowitz (see Theorem 1.7 in [7]), positive solutions of (1.2) in a neighborhood of ( μ λ , θ d N , λ , 0 ) are expressed as follows:

( μ ( s ) , N ( s ) , P ( s ) ) = ( μ λ + μ λ ( s ) , θ d N , λ + s ( ϕ μ λ + ϕ ( s ) ) , s ( ψ μ λ + ψ ( s ) ) )

for s ( 0 , ε ) with some ε > 0 , where ( μ λ ( s ) , ϕ ( s ) , ψ ( s ) ) R × X × X is continuously differentiable for s ( 0 , ε ) and satisfies ( μ λ ( 0 ) , ϕ ( 0 ) , ψ ( 0 ) ) = ( 0 , 0 , 0 ) and Ω ψ μ λ ψ ( s ) d x = 0 for s ( 0 , ε ) .

We next calculate the signal of μ λ ( 0 ) , which determines the bifurcation direction of positive solutions near the bifurcation point ( μ λ , θ d N , λ , 0 ) . Since ( μ ( s ) , N ( s ) , P ( s ) ) is a positive solution of (1.2), it follows from (3.1) that

W ( s ) e ( β P / d P ) N ( s ) P ( s )

satisfies

div ( d P e ( β P / d P ) N ( s ) W ( s ) ) = e ( β P / d P ) N ( s ) W ( s ) μ ( s ) P ( s ) + γ N ( s ) 1 + m N ( s ) + k P ( s ) , x Ω , W ( s ) = 0 , x Ω .

Let us multiply the above equation by Ψ μ λ and integrate the resulting expression over Ω to obtain

(3.8) Ω d P e ( β P / d P ) N ( s ) W ( s ) Ψ μ λ d x = Ω e ( β P / d P ) N ( s ) W ( s ) μ ( s ) P ( s ) + γ N ( s ) 1 + m N ( s ) + k P ( s ) Ψ μ λ d x .

For the sake of convenience in writing, we denote

A ( s ) e ( β P / d P ) N ( s ) = e ( β P / d P ) θ d N , λ + A ˜ ( s ) , B ( s ) γ N ( s ) 1 + m N ( s ) + k P ( s ) = γ θ d N , λ 1 + m θ d N , λ + B ˜ ( s ) ,

where

lim s 0 A ˜ ( s ) / s = A ( 0 ) , lim s 0 B ˜ ( s ) / s = B ( 0 ) .

Then, (3.8) can be rewritten as follows:

(3.9) Ω d P ( e ( β P / d P ) θ d N , λ + A ˜ ( s ) ) W ( s ) Ψ μ λ d x = Ω ( e ( β P / d P ) θ d N , λ + A ˜ ( s ) ) W ( s ) ( μ λ + μ λ ( s ) ) Ψ μ λ d x + Ω ( e ( β P / d P ) θ d N , λ + A ˜ ( s ) ) W ( s ) P ( s ) + γ θ d N , λ 1 + m θ d N , λ + B ˜ ( s ) Ψ μ λ d x .

Recall that Ψ μ λ is the positive eigenfunction associated to μ λ . Then,

div ( d P e ( β P / d P ) θ d N , λ Ψ μ λ ) γ θ d N , λ 1 + m θ d N , λ e ( β P / d P ) θ d N , λ Ψ μ λ = μ λ e ( β P / d P ) θ d N , λ Ψ μ λ , x Ω , Ψ μ λ = 0 , x Ω .

Let us multiply the above equation by W ( s ) and integrate the resulting expression over Ω to obtain

(3.10) Ω d P e ( β P / d P ) θ d N , λ Ψ μ λ W ( s ) d x = Ω γ θ d N , λ 1 + m θ d N , λ + μ λ e ( β P / d P ) θ d N , λ Ψ μ λ W ( s ) d x .

By (3.9) and (3.10), we have

Ω d P A ˜ ( s ) Ψ μ λ W ( s ) d x = Ω e ( β P / d P ) θ d N , λ W ( s ) ( μ λ ( s ) P ( s ) + B ˜ ( s ) ) Ψ μ λ d x + Ω A ˜ ( s ) W ( s ) μ λ + μ λ ( s ) P ( s ) + γ θ d N , λ 1 + m θ d N , λ + B ˜ ( s ) Ψ μ λ d x .

Let us divide both sides of this equation by s 2 . Then,

Ω d P A ˜ ( s ) s W ( s ) s Ψ μ λ d x = Ω e ( β P / d P ) θ d N , λ W ( s ) s μ λ ( s ) P ( s ) + B ˜ ( s ) s Ψ μ λ d x + Ω A ˜ ( s ) s W ( s ) s μ λ + μ λ ( s ) P ( s ) + γ θ d N , λ 1 + m θ d N , λ + B ˜ ( s ) Ψ μ λ d x .

Setting s 0 + , we have

(3.11) Ω d P A ( 0 ) Ψ μ λ 2 d x = Ω e ( β P / d P ) θ d N , λ ( μ λ ( 0 ) e ( β P / d P ) θ d N , λ Ψ μ λ + B ( 0 ) ) Ψ μ λ 2 d x + Ω A ( 0 ) μ λ + γ θ d N , λ 1 + m θ d N , λ Ψ μ λ 2 d x .

By virtue of the expressions of A ( s ) and B ( s ) , a direct calculation yields

A ( 0 ) = ( β P / d P ) e ( β P / d P ) θ d N , λ ϕ μ λ and B ( 0 ) = γ ( ϕ μ λ k θ d N , λ ψ μ λ ) ( 1 + m θ d N , λ ) 2 .

Thereby, (3.11) becomes

(3.12) μ λ ( 0 ) Ω e ( β P / d P ) θ d N , λ Ψ μ λ 2 d x = Ω β P e ( β P / d P ) θ d N , λ ϕ μ λ Ψ μ λ 2 d x + Ω e 2 ( β P / d P ) θ d N , λ Ψ μ λ 3 d x Ω γ ( ϕ μ λ k θ d N , λ ψ μ λ ) ( 1 + m θ d N , λ ) 2 e ( β P / d P ) θ d N , λ Ψ μ λ 2 d x Ω ( β P / d P ) μ λ + γ θ d N , λ 1 + m θ d N , λ e ( β P / d P ) θ d N , λ ϕ μ λ Ψ μ λ 2 d x .

Therefore, the bifurcation is supercritical if μ λ ( 0 ) > 0 , and subcritical if μ λ ( 0 ) < 0 .

Summarizing, we have the following result.

Theorem 3.1

Assume that λ > d N σ 1 [ Δ ; 1 ] is fixed. Then, positive solutions of (1.2) bifurcate from the semi-trivial solution branch Γ N = { ( μ , θ d N , λ , 0 ) : μ R } if and only if μ = μ λ . To be precise, there exists a neighborhood N 1 of ( μ , N , P ) = ( μ λ , θ d N , λ , 0 ) in R × X × X such that L 1 ( 0 ) N 1 consists of the union of Γ N N 1 and the local curve

( μ ( s ) , N ( s ) , P ( s ) ) = ( μ λ + μ λ ( s ) , θ d N , λ + s ( ϕ μ λ + ϕ ( s ) ) , s ( ψ μ λ + ψ ( s ) ) )

for s ( ε , ε ) with some ε > 0 , where ( μ λ ( s ) , ϕ ( s ) , ψ ( s ) ) R × X × X is continuously differentiable for s ( ε , ε ) and satisfies ( μ λ ( 0 ) , ϕ ( 0 ) , ψ ( 0 ) ) = ( 0 , 0 , 0 ) and μ λ ( 0 ) is given by (3.12). Therefore, positive solutions contained in L 1 ( 0 ) N 1 can be expressed as follows:

S 1 { ( μ λ + μ λ ( s ) , θ d N , λ + s ( ϕ μ λ + ϕ ( s ) ) , s ( ψ μ λ + ψ ( s ) ) ) : s ( 0 , ε ) } .

Although the above analysis provides some information on positive solutions of (1.2) in the neighborhood of ( μ λ , θ d N , λ , 0 ) , there is no information on the bifurcating curve S 1 far from the bifurcation point ( μ λ , θ d N , λ , 0 ) . Therefore, a further study is necessary to understand its global structure in the ( μ , N , P ) plane, i.e., R × X × X . For this, we first prove the following proposition.

Proposition 3.5

Let

λ μ σ 1 d N Δ + θ d P , μ 1 + k θ d P , μ ; 1 .

Then, λ μ is a continuous and increasing function with respect to μ and satisfies

lim μ d P σ 1 [ Δ ; 1 ] λ μ = d N σ 1 [ Δ ; 1 ] and lim μ λ μ = d N σ 1 [ Δ ; 1 ] + 1 / k .

Proof

In view of the positivity of θ d P , μ in Ω and the monotonicity properties of principal eigenvalue (see Lemma 2.1), it is obvious that

(3.13) d N σ 1 [ Δ ; 1 ] λ μ d N σ 1 [ Δ ; 1 ] + 1 / k

for all μ ( d P σ 1 [ Δ ; 1 ] , ) . Since θ d P , μ is a continuous and increasing function with respect to μ , Lemma 2.1 ensures that λ μ is also a continuous and increasing function with respect to μ . Moreover, since lim μ d P σ 1 [ Δ ; 1 ] θ d P , μ = 0 uniformly in Ω , we deduce from Corollary 8.1 in [24] that lim μ d P σ 1 [ Δ ; 1 ] λ μ = d N σ 1 [ Δ ; 1 ] .

We next prove lim μ λ μ = d N σ 1 [ Δ ; 1 ] + 1 / k . As λ μ is bounded for all μ ( d P σ 1 [ Δ ; 1 ] , ) , there exists a sequence { ψ μ } H 0 1 ( Ω ) with ψ μ L 2 ( Ω ) = 1 such that

λ μ = d N Ω ψ μ 2 d x + Ω θ d P , μ 1 + k θ d P , μ ψ μ 2 d x .

The boundedness of λ μ implies that { ψ μ } is bounded in H 0 1 ( Ω ) . Hence, by choosing a subsequence if necessary, there exists some nonnegative function ψ with ψ L 2 ( Ω ) = 1 such that

ψ μ ψ weakly in H 0 1 ( Ω ) and ψ μ ψ in L 2 ( Ω )

as μ . Note that

Ω θ d P , μ 1 + k θ d P , μ ψ μ 2 d x Ω 1 k ψ 2 d x Ω θ d P , μ 1 + k θ d P , μ ( ψ μ 2 ψ 2 ) d x + Ω θ d P , μ 1 + k θ d P , μ 1 k ψ 2 d x 1 k ψ μ + ψ 2 ψ μ ψ 2 + Ω θ d P , μ 1 + k θ d P , μ 1 k ψ 2 d x .

In addition, Lebesgue’s dominated convergence theorem ensures that

lim μ Ω θ d P , μ 1 + k θ d P , μ ψ 2 d x = Ω 1 k ψ 2 d x .

Thus, we derive

lim μ Ω θ d P , μ 1 + k θ d P , μ ψ μ 2 d x = Ω 1 k ψ 2 d x .

It follows from Poincaré inequality ψ L 2 ( Ω ) 2 σ 1 [ Δ ; 1 ] ψ L 2 ( Ω ) 2 and the classic inequality liminf μ ψ μ L 2 ( Ω ) ψ L 2 ( Ω ) that

liminf μ λ μ d N Ω ψ 2 d x + Ω 1 k ψ 2 d x d N σ 1 [ Δ ; 1 ] + 1 k .

On the other hand, by (3.13), we have limsup μ λ μ d N σ 1 [ Δ ; 1 ] + 1 / k . Thus,

lim μ λ μ = d N σ 1 [ Δ ; 1 ] + 1 / k .

This completes the proof of Proposition 3.5.□

Proposition 3.6

Assume that λ d N σ 1 [ Δ ; 1 ] + 1 / k . Then, the semi-trivial solution ( 0 , θ d P , μ ) is unstable for any μ > d P σ 1 [ Δ ; 1 ] . Moreover, there is no bifurcation of positive solutions occurring from ( 0 , θ d P , μ ) .

Proof

The linearization of L ( N , P ) with respect to ( N , P ) at ( 0 , θ d P , μ ) is given as follows:

L ( N , P ) ( 0 , θ d P , μ ) = d N Δ λ + θ d P , μ 1 + k θ d P , μ 0 β P div ( θ d P , μ ) γ θ d P , μ 1 + k θ d P , μ d P Δ μ + 2 θ d P , μ .

By the Riesz-Schauder theory, it is well known that the spectrum of L ( N , P ) ( 0 , θ d P , μ ) , denoted by ρ ( L ( N , P ) ( 0 , θ d P , μ ) ) , consists of real eigenvalues and

ρ ( L ( N , P ) ( 0 , θ d P , μ ) ) = ρ d N Δ λ + θ d P , μ 1 + k θ d P , μ ρ ( d P Δ μ + 2 θ d P , μ ) .

Since θ d P , μ is the unique positive solution of (2.2), it is clear that σ 1 [ d P Δ + θ d P , μ μ ; 1 ] = 0 . Moreover, Lemma 2.1 shows that

σ 1 [ d P Δ + 2 θ d P , μ μ ; 1 ] > σ 1 [ d P Δ + θ d P , μ μ ; 1 ] = 0 .

Hence, ρ ( d P Δ μ + 2 θ d P , μ ) lies on the positive real axis. In addition, ρ ( d N Δ λ + θ d P , μ / ( 1 + k θ d P , μ ) ) lies on the real axis and the least eigenvalue is λ μ λ . According to Proposition 3.5, one sees that if λ d N σ 1 [ Δ ; 1 ] + 1 / k , then λ μ < λ . Therefore, the semi-trivial solution ( 0 , θ d P , μ ) is unstable for any μ > d P σ 1 [ Δ ; 1 ] , provided λ d N σ 1 [ Δ ; 1 ] + 1 / k .

According to the local bifurcation theorem of Crandall and Rabinowitz [7], the necessary condition for bifurcation from ( 0 , θ d P , μ ) is that L ( N , P ) ( 0 , θ d P , μ ) is degenerate. Thus, there exists a pair of functions ( ϕ , ψ ) C 0 1 ( Ω ¯ ) × C 0 1 ( Ω ¯ ) with ( ϕ , ψ ) ( 0 , 0 ) such that L ( N , P ) ( 0 , θ d P , μ ) ( ϕ , ψ ) = 0 . That is,

d N Δ ϕ λ ϕ + θ d P , μ 1 + k θ d P , μ ϕ = 0 , x Ω , d P Δ ψ μ ψ + 2 θ d P , μ ψ + β P div ( θ d P , μ ) γ θ d P , μ 1 + k θ d P , μ ϕ = 0 , x Ω , ϕ = ψ = 0 , x Ω .

Since we hope to obtain positive solutions by bifurcating from ( 0 , θ d P , μ ) , the bifurcation should occur at the principal eigenvalue, which ensures that the eigenfunction is positive. In view of the Krein-Rutman theorem, the first equation of this system has positive solutions if and only if λ = λ μ . Proposition 3.5 shows that this is impossible for λ d N σ 1 [ Δ ; 1 ] + 1 / k . Consequently, there is no bifurcation of positive solutions occurring from the semi-trivial solution ( 0 , θ d P , μ ) .□

The following theorem gives the global structure of the local curve S 1 in the ( μ , N , P ) plane.

Theorem 3.2

For any given d N , d P , β P , γ , m , and k , the following statements hold true.

  1. If d N σ 1 [ Δ ; 1 ] < λ < d N σ 1 [ Δ ; 1 ] + 1 / k is fixed, then the local curve S 1 can be extended to a bounded global continuum of positive solutions to (1.2), which meets the other semi-trivial solution ( 0 , θ d P , μ ) at μ = μ , where μ is uniquely determined by λ = λ μ .

  2. If λ d N σ 1 [ Δ ; 1 ] + 1 / k is fixed, then the local curve S 1 can be extended to an unbounded global continuum of positive solutions to (1.2) along the positive values of μ .

Proof

(1) Let P denote the positive cone in C 0 1 ( Ω ¯ ) and the interior of the positive cone P , denoted by int ( P ) , is nonempty. By Definition 1.1 in [5], it is easy to check that θ d N , λ is a non-degenerate solution of (2.1). Moreover, for (1.2), the hypotheses ( H P Q R S ) , ( H a b ) , ( H f g ) , and ( H F G ) given in [5] are satisfied. Therefore, we apply Theorem 1.1 in [5] to conclude that there exists a continuum

C R × int ( P ) × int ( P )

of positive solutions to (1.2) such that S 1 C and C satisfies one of the following statements:

  1. C is unbounded in R × C 0 1 ( Ω ¯ ) × C 0 1 ( Ω ¯ ) ;

  2. there exists a positive solution θ d P , μ of (2.2) such that ( μ , 0 , θ d P , μ ) C ¯ , where μ is determined by

    λ = σ 1 d N Δ + θ d P , μ 1 + k θ d P , μ ; 1 ;

  3. there exists another positive solution of (2.1), denoted by ϕ d N , λ with ϕ d N , λ θ d N , λ , such that

    σ 1 div ( d P e ( β P / d P ) ϕ d N , λ ) γ ϕ d N , λ 1 + m ϕ d N , λ e ( β P / d P ) ϕ d N , λ ; e ( β P / d P ) ϕ d N , λ , ϕ d N , λ , 0 C ¯ ;

  4. λ = d N σ 1 [ Δ ; 1 ] and ( d N σ 1 [ Δ ; 1 ] , 0 , 0 ) C ¯ .

We next claim that alternatives (a), (c), and (d) cannot occur. Since d N σ 1 [ Δ ; 1 ] < λ < d N σ 1 [ Δ ; 1 ] + 1 / k , it is clear that alternative (d) cannot be true. By virtue of Lemma 2.2, (2.1)  has a unique positive solution for λ > d N σ 1 [ Δ ; 1 ] , this means that alternative (c) cannot occur as well. From Propositions 3.3 and 3.4, we find that (1.2) has no positive solution if μ is too small or large. Moreover, Proposition 3.2 shows that any positive solution is bounded in W 2 , p ( Ω ) × W 2 , p ( Ω ) , provided μ is bounded. Thus, for any bounded μ , it follows from the Sobolev embedding theorem that any positive solution is bounded in C 0 1 ( Ω ¯ ) × C 0 1 ( Ω ¯ ) , and hence, alternative (a) cannot be satisfied either. Consequently, the continuum C of positive solutions to (1.2) must satisfy alternative (b); that is, there exists a number μ such that ( μ , 0 , θ d P , μ ) C ¯ . Moreover, Proposition 3.5 ensures that μ is unique for any given λ ( d N σ 1 [ Δ ; 1 ] , d N σ 1 [ Δ ; 1 ] + 1 / k ) . This completes the proof of part (1).

(2) For any given λ [ d N σ 1 [ Δ ; 1 ] + 1 / k , ) , as above, there exists a continuum C R × int ( P ) × int ( P ) of positive solutions to (1.2) such that S 1 C and C satisfies one of the alternatives (a)-(d). As above, alternatives (c) and (d) are unlikely to be true. By virtue of Proposition 3.6, alternative (b) cannot occur as well. The remaining possibility is that C is unbounded in R × C 0 1 ( Ω ¯ ) × C 0 1 ( Ω ¯ ) . According to Proposition 3.2 and the Sobolev embedding theorem, any positive solutions of (1.2) is bounded in C 0 1 ( Ω ¯ ) × C 0 1 ( Ω ¯ ) , provided μ is bounded in R . This means that Proj μ C is unbounded. Furthermore, it follows from Proposition 3.3 that C extends to infinity in positive values of μ .□

4 Structure of solutions for β N > 0 and β P = 0

In this section, we study (1.3) and investigate the global bifurcation structure of the set of positive solutions by treating λ as the main bifurcation parameter.

4.1 A priori estimates

Analogous to the case of (1.2), we first establish some a priori estimates of positive solutions to (1.3) so that we could make a detailed description for the global bifurcation structure of the set of positive solutions. Although the proofs of the next two propositions are similar to those of Propositions 3.1 and 3.2, we present them here for the reader’s convenience.

Proposition 4.1

If λ 0 is fixed, then (1.3) has no positive solution. If λ > 0 is fixed, then any positive solution ( N , P ) of (1.3) satisfies

N ( x ) λ e ( β N / d N ) ( μ + γ / m ) and P ( x ) μ + γ / m

for all x Ω .

Proof

Suppose (1.3) has at least one positive solution ( N , P ) for λ 0 . Then, it follows from (4.1) that

div ( d N e ( β N / d N ) P V ) 0 , x Ω , V = 0 , x Ω .

By virtue of the maximum principle (see, for instance, [24]), one sees that V 0 in Ω , and so N 0 in Ω . This is impossible since ( N , P ) is a positive solution.

When λ > 0 is fixed, we assume x 1 Ω ¯ is a maximum point of P , i.e., P ( x 1 ) = max Ω ¯ P ( x ) > 0 . Then, x 1 Ω , and hence,

0 d P Δ P ( x 1 ) = μ P ( x 1 ) + γ N ( x 1 ) 1 + m N ( x 1 ) + k P ( x 1 ) P ( x 1 ) .

Since P ( x 1 ) > 0 , we have

P ( x 1 ) μ + γ N ( x 1 ) 1 + m N ( x 1 ) + k P ( x 1 ) μ + γ / m .

Thus, P ( x ) μ + γ / m for all x Ω .

We now consider the upper bound of N in Ω . Let

V e ( β N / d N ) P N .

Then, it follows from the first equation of (1.3) that V satisfies

(4.1) div ( d N e ( β N / d N ) P V ) = e ( β N / d N ) P V λ e ( β N / d N ) P V P 1 + m N + k P , x Ω , V = 0 , x Ω .

Suppose that x 2 Ω ¯ is a maximum point of V , i.e., V ( x 2 ) = max Ω ¯ V ( x ) > 0 . Then, x 2 Ω , and hence, V ( x 2 ) = 0 and Δ V ( x 2 ) 0 . A simple calculation provides

div ( d N e ( β N / d N ) P V ) x = x 2 = β N e ( β N / d N ) P P V x = x 2 + d N e ( β N / d N ) P Δ V x = x 2 0 .

By virtue of (4.1), we have

V ( x 2 ) e ( β N / d N ) P ( x 2 ) λ P ( x 2 ) 1 + m N ( x 2 ) + k P ( x 2 ) λ e ( β N / d N ) ( μ + γ / m ) .

Since N = e ( β N / d N ) P V V in Ω , the desired estimate is derived.□

With the help of Proposition 4.1, we establish the W 2 , p -estimate for any positive solution of (1.3) in the following proposition.

Proposition 4.2

Let λ > 0 be fixed. Assume that ( N , P ) is any positive solution of (1.3). Then, for any p ( 1 , ) , there exists a positive constant M, depending on the parameters of system (1.3), such that ( N , P ) satisfies

N W 2 , p ( Ω ) M and P W 2 , p ( Ω ) M .

Proof

For simplicity, we denote the positive constants by M i depending on the parameters of system (1.3). Proposition 4.1 ensures that there exists a positive constant M 1 such that

1 d P μ P P 2 + γ N P 1 + m N + k P L p ( Ω ) M 1

for all p > 1 . We apply the L p -estimate for elliptic equations [15] to conclude that P W 2 , p ( Ω ) is bounded for all p > 1 . Thus, there exists a positive constant M 2 such that P W 2 , p ( Ω ) M 2 . It follows from the Sobolev embedding theorem that there exists a positive constant M 3 such that P C 1 ( Ω ¯ ) M 3 . Similarly, we apply the elliptic regularity [15] to (4.1) to conclude that there exists a positive constant M 4 such that

e ( β N / d N ) P N C 1 ( Ω ¯ ) = V C 1 ( Ω ¯ ) M 4 .

Note that

( e ( β N / d N ) P N ) = e ( β N / d N ) P N + ( β N / d N ) N e ( β N / d N ) P P .

By virtue of the triangular inequality, we have

N e ( β N / d N ) P N ( e ( β N / d N ) P N ) + ( β N / d N ) N e ( β N / d N ) P P .

Consequently, there exists a positive constant M 5 such that N M 5 , and hence, it follows from Proposition 4.1 that there exists a positive constant M 6 such that N C 1 ( Ω ¯ ) M 6 .

Observe that the first equation of (1.3) can be expressed as follows:

Δ N = 1 d N β N P N + β N N Δ P + λ N N 2 N P 1 + m N + k P , x Ω , N = 0 , x Ω .

Since P C 1 ( Ω ¯ ) M 3 and N C 1 ( Ω ¯ ) M 6 , there exists a positive constant M 7 such that

1 d N β N P N + β N N Δ P + λ N N 2 N P 1 + m N + k P L p ( Ω ) M 7

for all p > 1 . This ensures us to apply L p -estimates for elliptic equations to conclude that N W 2 , p ( Ω ) is bounded for all p > 1 . The desired estimate is derived.□

4.2 Nonexistence of positive solutions

The main purpose of this subsection is to give an appropriate nonexistence result of positive solutions to (1.3).

Proposition 4.3

If μ d P σ 1 [ Δ ; 1 ] γ / m , then (1.3) has no positive solution.

Proof

If ( N , P ) is a positive solution of (1.3), then we multiply both sides of the second equation of (1.3) by P and integrate the resulting expression over Ω to obtain

d P Ω P 2 d x = Ω μ P + γ N 1 + m N + k P P 2 d x < ( μ + γ / m ) Ω P 2 d x .

In view of Poincaré’s inequality, one sees that

d P σ 1 [ Δ ; 1 ] Ω P 2 d x d P Ω P 2 d x < ( μ + γ / m ) Ω P 2 d x .

This implies that μ + γ / m > d P σ 1 [ Δ ; 1 ] since P > 0 in Ω ; in other words, (1.3) has no positive solution for any fixed μ d P σ 1 [ Δ ; 1 ] γ / m .□

4.3 Bifurcation structure of positive solutions

The main purpose of this subsection is to investigate the bifurcation structure of positive solutions to (1.3) by regarding λ as a bifurcation parameter and fixing all other constants.

Suppose μ > d P σ 1 [ Δ ; 1 ] . Then, for any λ R , (1.3) has a semi-trivial solution branch

Π P { ( λ , 0 , θ d P , μ ) : λ R } .

We now use bifurcation techniques to make a direct investigation of this semi-trivial solution branch Π P for (1.3). Our argument below is very similar to those of the preceding section (see Section 3.3), and hence we will only sketch it here.

Define

T ( λ , N , P ) = div ( d N N + β N N P ) λ N + N 2 + N P 1 + m N + k P d P Δ P μ P + P 2 γ N P 1 + m N + k P .

Clearly, ( N , P ) X × X is a nonnegative solution of (1.3) if and only if T ( λ , N , P ) = 0 . The linearization of T ( λ , N , P ) with respect to ( N , P ) at ( 0 , θ d P , μ ) is given as follows:

T ( N , P ) ( λ , 0 , θ d P , μ ) = div ( d N + β N θ d P , μ ) λ + θ d P , μ 1 + k θ d P , μ 0 γ θ d P , μ 1 + k θ d P , μ d P Δ μ + 2 θ d P , μ .

By setting T ( N , P ) ( λ , 0 , θ d P , μ ) ( ϕ , ψ ) = 0 , we obtain

ker [ T ( N , P ) ( λ μ , 0 , θ d P , μ ) ] = span ( ϕ λ μ , ψ λ μ ) ,

where

ϕ λ μ = e ( β N / d N ) θ d P , μ Φ λ μ , ψ λ μ = ( d P Δ + 2 θ d P , μ μ ) 1 γ θ d P , μ 1 + k θ d P , μ e ( β N / d N ) θ d P , μ Φ λ μ ,

and Φ λ μ is the positive eigenfunction associated to λ μ . This shows that ker [ T ( N , P ) ( λ μ , 0 , θ d P , μ ) ] is one-dimensional. Moreover, by virtue of the Fredholm alternative theorem, it is not difficult to show codim Range [ T ( N , P ) ( λ μ , 0 , θ d P , μ ) ] = 1 . In addition, a simple calculation yields

T ( N , P ) λ ( λ μ , 0 , θ d P , μ ) ϕ λ μ ψ λ μ = ϕ λ μ 0 .

Since Ω ϕ λ μ Φ λ μ d x > 0 , we have

T ( N , P ) λ ( λ μ , 0 , θ d P , μ ) ϕ λ μ ψ λ μ Range [ T ( N , P ) ( λ μ , 0 , θ d P , μ ) ] .

Therefore, we apply the local bifurcation theorem of Crandall and Rabinowitz (see Theorem 1.7 in [7]) to conclude that positive solutions of (1.3) in the neighborhood of ( λ μ , 0 , θ d P , μ ) are expressed as follows:

( λ ( s ) , N ( s ) , P ( s ) ) = ( λ μ + λ μ ( s ) , s ( ϕ λ μ + ϕ ( s ) ) , θ d P , μ + s ( ψ λ μ + ψ ( s ) ) )

for s ( 0 , ε ) with some ε > 0 , where ( λ μ ( s ) , ϕ ( s ) , ψ ( s ) ) R × X × X is continuously differentiable for s ( 0 , ε ) and satisfies ( λ μ ( 0 ) , ϕ ( 0 ) , ψ ( 0 ) ) = ( 0 , 0 , 0 ) and Ω ϕ λ μ ϕ ( s ) d x = 0 for s ( 0 , ε ) . In addition, since ( λ ( s ) , N ( s ) , P ( s ) ) is a positive solution of (1.3) and V ( s ) e ( β N / d N ) P ( s ) N ( s ) , we use a standard but cumbersome calculation as before to derive

(4.2) λ μ ( 0 ) Ω e ( β N / d N ) θ d P , μ Φ λ μ 2 d x = Ω β N e ( β N / d N ) θ d P , μ ψ λ μ Φ λ μ 2 d x + Ω e 2 ( β N / d N ) θ d P , μ Φ λ μ 3 d x + Ω ψ λ μ m θ d P , μ ϕ λ μ ( 1 + k θ d P , μ ) 2 e ( β N / d N ) θ d P , μ Φ λ μ 2 d x + Ω ( β N / d N ) λ μ θ d P , μ 1 + k θ d P , μ e ( β N / d N ) θ d P , μ ψ λ μ Φ λ μ 2 d x .

Consequently, we have the following result.

Theorem 4.1

Assume that μ > d P σ 1 [ Δ ; 1 ] is fixed. Then, positive solutions of (1.3) bifurcate from the semi-trivial solution branch Π P = { ( λ , 0 , θ d P , μ ) : λ R } if and only if λ = λ μ . To be precise, there exists a neighborhood N 2 of ( λ , N , P ) = ( λ μ , 0 , θ d P , μ ) in R × X × X such that T 1 ( 0 ) N 2 consists of the union of Π P N 2 and the local curve

( λ ( s ) , N ( s ) , P ( s ) ) = ( λ μ + λ μ ( s ) , s ( ϕ λ μ + ϕ ( s ) ) , θ d P , μ + s ( ψ λ μ + ψ ( s ) ) )

for s ( ε , ε ) with some ε > 0 , where ( λ μ ( s ) , ϕ ( s ) , ψ ( s ) ) R × X × X is continuously differentiable for s ( ε , ε ) and satisfies ( λ μ ( 0 ) , ϕ ( 0 ) , ψ ( 0 ) ) = ( 0 , 0 , 0 ) , and λ μ ( 0 ) is given by (4.2). Therefore, positive solutions contained in T 1 ( 0 ) N 2 can be expressed as follows:

S 2 { ( λ μ + λ μ ( s ) , s ( ϕ λ μ + ϕ ( s ) ) , θ d P , μ + s ( ψ λ μ + ψ ( s ) ) ) : s ( 0 , ε ) } .

The following theorem gives the global structure of the local curve S 2 in the ( λ , N , P ) plane.

Theorem 4.2

If μ > d P σ 1 [ Δ ; 1 ] is fixed, then the local curve S 2 can be extended to an unbounded global continuum of positive solutions to (1.3) along the positive values of λ .

Proof

By Definition 1.1 in [5], θ d P , μ is a non-degenerate solution of (2.2). Moreover, for (1.3), the hypotheses ( H P Q R S ) , ( H a b ) , ( H f g ) , and ( H F G ) given in [5] are satisfied. Therefore, we apply Theorem 1.2 in [5] to conclude that there exists a continuum

C R × int ( P ) × int ( P )

of positive solutions to (1.3), where P is given in Theorem 3.2, such that S 2 C and C satisfies one of the following statements:

  1. C is unbounded in R × C 0 1 ( Ω ¯ ) × C 0 1 ( Ω ¯ ) ;

  2. there exists another positive solution of (2.2), denoted by ψ d P , μ with ψ d P , μ θ d P , μ , such that

    σ 1 div ( d N e ( β N / d N ) ψ d P , μ ) + ψ d P , μ 1 + k ψ d P , μ e ( β N / d N ) ψ d P , μ ; e ( β N / d N ) ψ d P , μ , 0 , ψ d P , μ C ¯ ;

  3. there exists a positive solution θ d N , λ ˜ of (2.1) such that ( λ ˜ , θ d N , λ ˜ , 0 ) C ¯ , where λ ˜ is determined by

    μ = σ 1 d P Δ γ θ d N , λ ˜ 1 + m θ d N , λ ˜ ; 1 ;

  4. μ = d P σ 1 [ Δ ; 1 ] and ( d P σ 1 [ Δ ; 1 ] , 0 , 0 ) C ¯ .

We next claim that alternatives (b), (c), and (d) cannot occur. Since μ > d P σ 1 [ Δ ; 1 ] , it is clear that alternative (d) cannot be true. By virtue of the uniqueness of positive solution to (2.2), alternative (b) cannot occur as well. In addition, it follows from the monotonicity of principal eigenvalue with respect to potential functions (see Lemma 2.1) that

σ 1 d P Δ γ θ d N , λ ˜ 1 + m θ d N , λ ˜ ; 1 σ 1 [ d P Δ ; 1 ] = d P σ 1 [ Δ ; 1 ] < μ .

This means that for any given μ > d P σ 1 [ Δ ; 1 ] , there must be no such value λ ˜ , and hence, alternative (c) cannot be satisfied either. Consequently, the continuum C of positive solutions of (1.3) must satisfy alternative (a); that is, C is unbounded in R × C 0 1 ( Ω ¯ ) × C 0 1 ( Ω ¯ ) . According to Proposition 4.2, we assert that any positive solution of (1.3) is bounded in C 0 1 ( Ω ¯ ) × C 0 1 ( Ω ¯ ) , provided λ is bounded in R . This means that Proj λ C is unbounded. Furthermore, by virtue of Proposition 4.1, C extends to infinity in positive values of λ .□

According to Proposition 4.3, one has known that if μ d P σ 1 [ Δ ; 1 ] γ / m , then (1.3) does not have positive solutions. Hence, we suppose now that d P σ 1 [ Δ ; 1 ] γ / m < μ < d P σ 1 [ Δ ; 1 ] . In this case, (1.3) has a trivial solution branch Π 0 { ( λ , 0 , 0 ) : λ R } and a semi-trivial solution branch

Π N { ( λ , θ d N , λ , 0 ) : λ > d N σ 1 [ Δ ; 1 ] } ,

which bifurcates from Π 0 as λ increases across d N σ 1 [ Δ ; 1 ] .

Since bifurcation from Π N seems likely to occur at values of λ such that

μ = σ 1 d P Δ γ θ d N , λ 1 + m θ d N , λ ; 1 ,

we first discuss the properties of this principal eigenvalue.

Proposition 4.4

Let

μ ( λ ) σ 1 d P Δ γ θ d N , λ 1 + m θ d N , λ ; 1 .

Then, μ ( λ ) is a continuous and decreasing function with respect to λ and satisfies

lim λ d N σ 1 [ Δ ; 1 ] μ ( λ ) = d P σ 1 [ Δ ; 1 ] and lim λ μ ( λ ) = d P σ 1 [ Δ ; 1 ] γ / m .

Proof

The proof can be completed by the similar arguments to those of Proposition 3.5.□

It can be proved as before that there is the local curve S 3 of positive solutions bifurcating from Π N at ( λ , θ d N , λ , 0 ) , and the local curve S 3 can be extended as a global continuum and it goes to as λ . More precisely, we have the following theorem.

Theorem 4.3

Assume that d P σ 1 [ Δ ; 1 ] γ / m < μ < d P σ 1 [ Δ ; 1 ] is fixed. Then, positive solutions of (1.3) bifurcate from the semi-trivial solution branch Π N = { ( λ , θ d N , λ , 0 ) : λ > d N σ 1 [ Δ ; 1 ] } if and only if λ = λ . Moreover, the following statements hold true.

  1. There exists a neighborhood N 3 of ( λ , N , P ) = ( λ , θ d N , λ , 0 ) in R × X × X such that T 1 ( 0 ) N 3 consists of the union of Π N N 3 and the local curve

    ( λ ( s ) , N ( s ) , P ( s ) ) = ( λ + λ ( s ) , θ d N , λ + s ( ϕ λ + ϕ ( s ) ) , s ( ψ λ + ψ ( s ) ) )

    for s ( ε , ε ) with some ε > 0 , where

    1. ( λ ( s ) , ϕ ( s ) , ψ ( s ) ) R × X × X is continuously differentiable for s ( ε , ε ) and satisfies ( λ ( 0 ) , ϕ ( 0 ) , ψ ( 0 ) ) = ( 0 , 0 , 0 ) ,

    2. ψ λ is the positive eigenfunction associated to σ 1 [ d P Δ γ θ d N , λ / ( 1 + m θ d N , λ ) ; 1 ] and

      ϕ λ = ( d N Δ λ + 2 θ d N , λ ) 1 div ( β N θ d N , λ ψ λ ) θ d N , λ 1 + m θ d N , λ ψ λ .

  2. Positive solutions contained in T 1 ( 0 ) N 3 can be expressed as follows:

    S 3 { ( λ + λ ( s ) , θ d N , λ + s ( ϕ λ + ϕ ( s ) ) , s ( ψ λ + ψ ( s ) ) ) : s ( 0 , ε ) } .

  3. The local curve S 3 can be extended to an unbounded global continuum of positive solutions to (1.3) along the positive values of λ .

5 Summary and discussion

In this article, we study the stationary problem for a prey-predator model with prey-taxis/predator-taxis under the homogeneous Dirichlet boundary conditions, where the interaction is governed by a Beddington-DeAngelis functional response. By applying the local and global bifurcation theory, eigenvalue theory of the second-order linear elliptic operators, and various elliptic estimates, we establish the sufficient conditions for the existence/nonexistence of coexistence states. These results provide an easy way to predict the coexistence of two species and manage to explain the occurrence of stationary patterns. In the following, we compare the findings of this study with those of the previous articles as follows:

  • When the Lotka-Volterra type functional response is adopted (i.e., m = k = 0 ), Cintra et al. [5] proved that the continuum of positive solutions is bounded for any given λ ( d N σ 1 [ Δ ; 1 ] , ) , and there is no positive solution for all large μ > 0 (see Lemma 5.5 and Theorem 5.2 in [5]). However, our results show that there exists a critical value for the prey’s growth rate λ such that, above this value, the continuum of positive solutions is unbounded (see Theorem 1.1(3)) and, below this critical value, the continuum of positive solutions is bounded (see Theorem 1.1(2)). Moreover, Theorem 1.1(3) also shows that above this value, (1.2) has at least one positive solution even though the predator’s growth rate μ is large. These differences suggest that the mutual interference by predators (i.e., k > 0 ) affects the behavior of positive solutions.

  • When the prey-taxis is ignored (i.e., β P = 0 ), the signal of μ λ ( 0 ) , which determines the bifurcation direction of positive solutions near the bifurcation point ( μ λ , θ d N , λ , 0 ) , is given as follows:

    μ λ ( 0 ) = Ω ψ μ λ 3 d x Ω γ ( ϕ μ λ k θ d N , λ ψ μ λ ) ( 1 + m θ d N , λ ) 2 ψ μ λ 2 d x Ω ψ μ λ 2 d x .

    This means that μ λ ( 0 ) > 0 since ψ μ λ > 0 and ϕ μ λ < 0 in Ω . Thus, the bifurcation direction is supercritical. However, when β P > 0 , μ λ ( 0 ) is given by (3.12) and the signal is difficult to determine. In other words, μ λ ( 0 ) may be positive or negative if the appropriate values of parameters in (3.12) are selected. This indicates that the bifurcation direction may be supercritical or subcritical. Moreover, this change will most likely lead to the multiplicity of positive solutions near the bifurcation point ( μ λ , θ d N , λ , 0 ) . Consequently, the introduction of prey-taxis (i.e., β P > 0 ) also changes the behavior of positive solutions.

  • Based on a priori estimate of solutions and standard elliptic regularity theory, it is not difficult to show that as β N , β P , and k tend to zero, any positive solution of (1.2) or (1.3) converges to a solution of the following Holling-Tanner prey-predator elliptic system:

    (5.1) d N Δ N = λ N N 2 N P 1 + m N , x Ω , d P Δ P = μ P P 2 + γ N P 1 + m N , x Ω , N = P = 0 , x Ω .

    Hence, it is reasonable to think that the existing results of this article have been perturbed from the existing results of [3] (see also [27]) as β N , β P , and k perturb from zero. In addition, our existing results are not limited to small β N , β P , and k but hold for any given β N , β P , and k . It is worth noting that the presence of prey-taxis (i.e., β P > 0 ) or predator-taxis (i.e., β N > 0 ) makes mathematical analysis more difficult (see, e.g., Proposition 3.4), and hence, the existing results of this article is not just a simple extension of the existing results of [3].

There are various interesting questions that deserve further exploration. For the semilinear elliptic system (5.1), Casal et al. analyzed theoretically the multiplicity of positive solutions and found numerically the existence of a Hopf bifurcation in [3]. Some time later, some of these pioneering findings were sharpened by Du and Lou in [11,12]. However, it is unclear whether or not these results hold for the quasilinear elliptic systems (1.2) and (1.3). Moreover, when the domain of habitation is one-dimensional, the uniqueness of positive solutions of (5.1) has been established in [3] (see also the result of Dancer et al. [8] and the result of López-Gómez and Pardo [28]). Whether the method developed in [3] can successfully solve the uniqueness of positive solutions of (1.2) and (1.3) is unknown. In addition, a more interesting question is how to study the positive solutions of the complete system (1.1). All these questions are very interesting and worthwhile to pursue in the future.

Acknowledgements

The authors would like to thank three reviewers for their very careful readings and helpful suggestions, which significantly improved the initial draft.

  1. Funding information: The work was partially supported by the National Science Foundation of China (11901446, 12171296), the Postdoctoral Science Foundation of China (2021T140530), the National Science Foundation in Shaanxi Province of China (2023-JC-YB-038), and the Fundamental Research Funds for the Central Universities (ZYTS23051).

  2. Conflict of interest: The authors state that there is no conflict of interest.

References

[1] J. R. Beddington, Mutual interference between parasites or predators and its effect on searching efficiency, J. Animal Ecol. 44 (1975), 331–340. 10.2307/3866Search in Google Scholar

[2] Y. L. Cai, Q. Cao, and Z. A. Wang, Asymptotic dynamics and spatial patterns of a ratio-dependent predator-prey system with prey-taxis, Appl. Anal. 101 (2022), 81–99. 10.1080/00036811.2020.1728259Search in Google Scholar

[3] A. Casal, J. C. Eilbeck, and J. López-Gómez, Existence and uniqueness of coexistence states for a predator-prey model with diffusion, Diff. Int. Eqns. 7 (1994)411–439. 10.57262/die/1369330437Search in Google Scholar

[4] W. Cintra, C. Morales-Rodrigo, and A. Suárez, Coexistence states in a cross-diffusion system of a predator-prey model with predator satiation term, Math. Models Methods Appl. Sci. 28 (2018), 2131–2159. 10.1142/S0218202518400109Search in Google Scholar

[5] W. Cintra, C. Morales-Rodrigo, and A. Suárez, Unilateral global bifurcation for a class of quasilinear elliptic systems and applications, J. Differential Equations 267 (2019), 619–657. 10.1016/j.jde.2019.01.021Search in Google Scholar

[6] W. Cintra, C. A. dos Santos, and J. Z. Zhou, Coexistence states of a Holling type II predator-prey system with self and cross-diffusion terms, Discret. Contin. Dyn. Syst. Ser. B 27 (2022), 3913–3931. 10.3934/dcdsb.2021211Search in Google Scholar

[7] M. G. Crandall and P. H. Rabinowitz, Bifurcation from simple eigenvalues, J. Funct. Anal. 8 (1971), 321–340. 10.1016/0022-1236(71)90015-2Search in Google Scholar

[8] E. N. Dancer, J. López-Gómez, and R. Ortega, Ortega, On the spectrum of some linear noncooperative elliptic systems with radial symmetry, Diff. Int. Eqns. 8 (1995), 515–523. 10.57262/die/1369316502Search in Google Scholar

[9] D. L. DeAngelis, R. A. Goldstein, and R. V. O’Neill, A model for tropic interaction, Ecology 56 (1975), 881–892. 10.2307/1936298Search in Google Scholar

[10] M. Delgado and A. Suárez, Study of an elliptic system arising from angiogenesis with chemotaxis and flux at the boundary, J. Differential Equations 244 (2008), 3119–3150. 10.1016/j.jde.2007.12.007Search in Google Scholar

[11] Y. H. Du and Y. Lou, Some uniqueness and exact multiplicity results for a predator-prey model, Trans. Amer. Math. Soc. 349 (1997), 2443–2475. 10.1090/S0002-9947-97-01842-4Search in Google Scholar

[12] Y. H. Du and Y. Lou, S-shaped global bifurcation curve and Hopf bifurcation of positive solutions to a predator-prey model, J. Differential Equations 144 (1998), no. 2, 390–440. 10.1006/jdeq.1997.3394Search in Google Scholar

[13] M. Fuest, Global solutions near homogeneous steady states in a multidimensional population model with both predator- and prey-taxis, SIAM J. Math. Anal. 52 (2020), 5865–5891. 10.1137/20M1344536Search in Google Scholar

[14] J. E. Garvey, R. A. Stein, and H. M. Thomas, Assessing how fish predation and interspecific prey competition influence a crayfish assemblage, Ecology 75 (1994), 532–547. 10.2307/1939556Search in Google Scholar

[15] D. Gilbarg and N. S. Trudinger, Elliptic Partial Differential Equations of Second Order, Second edition, Springer-Verlag, Berlin, 1983. Search in Google Scholar

[16] G. H. Guo and J. H. Wu, Multiplicity and uniqueness of positive solutions for a predator-prey model with B-D functional response, Nonlinear Anal. 72 (2010), 1632–1646. 10.1016/j.na.2009.09.003Search in Google Scholar

[17] G. H. Guo and J. H. Wu, The effect of mutual interference between predators on a predator-prey model with diffusion, J. Math. Anal. Appl. 389 (2012), 179–194. 10.1016/j.jmaa.2011.11.044Search in Google Scholar

[18] H. Y. Jin and Z. A. Wang, Global stability of prey-taxis systems, J. Differential Equations 262 (2017), 1257–1290. 10.1016/j.jde.2016.10.010Search in Google Scholar

[19] H. Y. Jin and Z. A. Wang, Global dynamics and spatio-temporal patterns of predator-prey systems with density-dependent motion, European J. Appl. Math. 32 (2021), 652–682. 10.1017/S0956792520000248Search in Google Scholar

[20] P. Kareiva and G. Odell, Swarms of predators exhibit “preytaxis” if individual predators use area-restricted search, Am. Nat. 130 (1987), 233–270. 10.1086/284707Search in Google Scholar

[21] S. Kesavan, Homogenization of elliptic eigenvalue problems. I, Appl. Math. Optim. 5 (1979), 153–167. 10.1007/BF01442551Search in Google Scholar

[22] J. M. Lee, T. Hillen, and M. A. Lewis, Pattern formation in prey-taxis systems, J. Biol. Dyn. 3 (2009), 551–573. 10.1080/17513750802716112Search in Google Scholar PubMed

[23] J. López-Gómez, Spectral Theory and Nonlinear Function Analysis, Chapman & Hall/CRC, New York, 2001. Search in Google Scholar

[24] J. López-Gómez, Linear Second Order Elliptic Operators, World Scientific Publishing Company, Singapore, 2013. 10.1142/8664Search in Google Scholar

[25] J. López-Gómez, Global bifurcation for Fredholm operators, Rend. Istit. Mat. Univ. Trieste, 48 (2016), 539–564. Search in Google Scholar

[26] J. López-Gómez and C. Mora-Corral, Counting zeros of C1 Fredholm maps of index 1, Bull. Lond. Math. Soc. 37 (2005), 778–792. 10.1112/S0024609305004716Search in Google Scholar

[27] J. López-Gómez and E. Munnnoz-Hernández, A spatially heterogeneous predator-prey model, Discret. Contin. Dyn. Syst. Ser. B 26 (2021), 2085–2113. 10.3934/dcdsb.2020081Search in Google Scholar

[28] J. López-Gómez and R. Pardo, Existence and uniqueness of coexistence states for the predator-prey model with diffusion: The scalar case, Diff. Int. Eqns. 6 (1993), 1025–1031. 10.57262/die/1370021908Search in Google Scholar

[29] J. Pejsachowicz and P. J. Rabier, Degree theory for C1 Fredholm mappings of index 0, J. Anal. Math. 76 (1998), 289–319. 10.1007/BF02786939Search in Google Scholar

[30] P. H. Rabinowitz, Some global results for nonlinear eigenvalue problems, J. Funct. Anal. 7 (1971), 487–513. 10.1016/0022-1236(71)90030-9Search in Google Scholar

[31] G. Q. Ren and B. Liu, Global existence and convergence to steady states for a predator-prey model with both predator- and prey-taxis, Discret. Contin. Dyn. Syst. 42 (2022), 759–779. 10.3934/dcds.2021136Search in Google Scholar

[32] J. P. Shi and X. F. Wang, On global bifurcation for quasilinear elliptic systems on bounded domains, J. Differential Equations 246 (2009), 2788–2812. 10.1016/j.jde.2008.09.009Search in Google Scholar

[33] G. Stampacchia, Le problème de Dirichlet pour les équations elliptiques du second ordre coefficients discontinus, Ann. Inst. Fourier (Grenoble) 15 (1965), 189–257. 10.5802/aif.204Search in Google Scholar

[34] Y. S. Tao and M. Winkler, A fully cross-diffusive two-component evolution system: existence and qualitative analysis via entropy-consistent thin-film-type approximation, J. Funct. Anal. 281 (2021), 109069. 10.1016/j.jfa.2021.109069Search in Google Scholar

[35] M. A. Tsyganov, J. Brindley, A. V. Holden, and V. N. Biktashev, Quasisoliton interaction of pursuit-evasion waves in a predator-prey system, Phys. Rev. Lett. 91 (2003), 218102. 10.1103/PhysRevLett.91.218102Search in Google Scholar PubMed

[36] M. Winkler, Asymptotic homogenization in a three-dimensional nutrient taxis system involving food-supported proliferation, J. Differential Equations 263 (2017), 4826–4869. 10.1016/j.jde.2017.06.002Search in Google Scholar

[37] S. N. Wu, J. P. Shi, and B. Y. Wu, Global existence of solutions and uniform persistence of a diffusive predator-prey model with prey-taxis, J. Differential Equations 260 (2016), 5847–5874. 10.1016/j.jde.2015.12.024Search in Google Scholar

[38] S. N. Wu, J. F. Wang, and J. P. Shi, Dynamics and pattern formation of a diffusive predator-prey model with predator-taxis, Math. Models Methods Appl. Sci. 28 (2018), 2275–2312. 10.1142/S0218202518400158Search in Google Scholar

[39] T. Xiang, Global dynamics for a diffusive predator-prey model with prey-taxis and classical Lotka-Volterra kinetics, Nonlinear Anal. Real. World Appl. 39 (2018), 278–299. 10.1016/j.nonrwa.2017.07.001Search in Google Scholar

Received: 2022-08-29
Revised: 2023-02-06
Accepted: 2023-03-14
Published Online: 2023-04-27

© 2023 the author(s), published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Articles in the same Issue

  1. Research Articles
  2. Asymptotic properties of critical points for subcritical Trudinger-Moser functional
  3. The existence of positive solution for an elliptic problem with critical growth and logarithmic perturbation
  4. On some dense sets in the space of dynamical systems
  5. Sharp profiles for diffusive logistic equation with spatial heterogeneity
  6. Generic properties of the Rabinowitz unbounded continuum
  7. Global bifurcation of coexistence states for a prey-predator model with prey-taxis/predator-taxis
  8. Multiple solutions of p-fractional Schrödinger-Choquard-Kirchhoff equations with Hardy-Littlewood-Sobolev critical exponents
  9. Improved fractional Trudinger-Moser inequalities on bounded intervals and the existence of their extremals
  10. The existence of infinitely many boundary blow-up solutions to the p-k-Hessian equation
  11. A priori bounds, existence, and uniqueness of smooth solutions to an anisotropic Lp Minkowski problem for log-concave measure
  12. Existence of nonminimal solutions to an inhomogeneous elliptic equation with supercritical nonlinearity
  13. Non-degeneracy of multi-peak solutions for the Schrödinger-Poisson problem
  14. Gagliardo-Nirenberg-type inequalities using fractional Sobolev spaces and Besov spaces
  15. Ground states of Schrödinger systems with the Chern-Simons gauge fields
  16. Quasilinear problems with nonlinear boundary conditions in higher-dimensional thin domains with corrugated boundaries
  17. A system of equations involving the fractional p-Laplacian and doubly critical nonlinearities
  18. A modified Picone-type identity and the uniqueness of positive symmetric solutions for a prescribed mean curvature problem
  19. On a version of hybrid existence result for a system of nonlinear equations
  20. Special Issue: Geometric PDEs and applications
  21. Preface for the special issue on “Geometric Partial Differential Equations and Applications”
  22. Convex hypersurfaces with prescribed Musielak-Orlicz-Gauss image measure
  23. Total mean curvatures of Riemannian hypersurfaces
  24. On degenerate case of prescribed curvature measure problems
  25. A curvature flow to the Lp Minkowski-type problem of q-capacity
  26. Aleksandrov reflection for extrinsic geometric flows of Euclidean hypersurfaces
  27. A note on second derivative estimates for Monge-Ampère-type equations
  28. The Lp chord Minkowski problem
  29. Widths of balls and free boundary minimal submanifolds
  30. Smooth approximation of twisted Kähler-Einstein metrics
  31. The exterior Dirichlet problem for the homogeneous complex k-Hessian equation
  32. A Carleman inequality on product manifolds and applications to rigidity problems
  33. Asymptotic behavior of solutions to the Monge-Ampère equations with slow convergence rate at infinity
  34. Pinched hypersurfaces are compact
  35. The spinorial energy for asymptotically Euclidean Ricci flow
  36. Geometry of CMC surfaces of finite index
  37. Capillary Schwarz symmetrization in the half-space
  38. Regularity of optimal mapping between hypercubes
  39. Special Issue: In honor of David Jerison
  40. Preface for the special issue in honor of David Jerison
  41. Homogenization of oblique boundary value problems
  42. A proof of a trace formula by Richard Melrose
  43. Compactness estimates for minimizers of the Alt-Phillips functional of negative exponents
  44. Regularity properties of monotone measure-preserving maps
  45. Examples of non-Dini domains with large singular sets
  46. Sharp inequalities for coherent states and their optimizers
  47. Gradient estimates and the fundamental solution for higher-order elliptic systems with lower-order terms
  48. Propagation of symmetries for Ricci shrinkers
  49. Linear extension operators for Sobolev spaces on radially symmetric binary trees
  50. The Neumann problem on the domain in 𝕊3 bounded by the Clifford torus
  51. On an effective equation of the reduced Hartree-Fock theory
  52. Polynomial sequences in discrete nilpotent groups of step 2
  53. Integral inequalities with an extended Poisson kernel and the existence of the extremals
  54. On singular solutions of Lane-Emden equation on the Heisenberg group
Downloaded on 1.11.2025 from https://www.degruyterbrill.com/document/doi/10.1515/ans-2022-0060/html
Scroll to top button