2026 Volume 35 Issue 8
Article Contents

Wenjie Yang, Jingxin Hu, Qianqian Zheng, Jianwei Shen, Yiping Lin. Stability switching and hopf bifurcation in a two-delay predation system with diffusion and network coupling[J]. Chinese Physics B, 2026, 35(8): 080204. doi: 10.1088/1674-1056/ae5c8a
Citation: Wenjie Yang, Jingxin Hu, Qianqian Zheng, Jianwei Shen, Yiping Lin. Stability switching and hopf bifurcation in a two-delay predation system with diffusion and network coupling[J]. Chinese Physics B, 2026, 35(8): 080204. doi: 10.1088/1674-1056/ae5c8a

Stability switching and hopf bifurcation in a two-delay predation system with diffusion and network coupling

  • Received Date: 18/01/2026
    Accepted Date: 02/04/2026
    Available Online: 01/08/2026
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

Figures(4)  /  Tables(2)

Article Metrics

Article views(87) PDF downloads(0) Cited by(0)

Access History

Stability switching and hopf bifurcation in a two-delay predation system with diffusion and network coupling

Abstract: This paper studies stability switching and Hopf bifurcation in a two-delay predation system with diffusion and network coupling. By using the Kronecker-product formulation and Laplacian-mode decomposition, the high-dimensional characteristic equation of the coupled system is reduced to a family of mode-dependent characteristic equations. Based on this reduction, explicit stability-switching curves are derived in the (τ1,τ2)-plane, and the corresponding stability region of the positive equilibrium is determined. Furthermore, by applying the Hassard method and center-manifold reduction, the direction of the Hopf bifurcation, the stability of the bifurcating periodic solutions, and the variation of their periods are obtained. Numerical examples are presented to verify the theoretical results and illustrate the temporal and spatiotemporal dynamics generated by the two delays under network coupling. These results provide an effective analytical framework for studying delay-induced stability switching and oscillatory behavior in networked predator-prey systems.

1.   Introduction
  • In ecology, predation models are classical frameworks for describing species interactions and have long been used to study local population dynamics. Predation systems have been extensively studied in recent years.[1,2] Since the original models were established,[3,4] mathematicians and biologists have progressively refined their descriptions of dynamical behaviors in predation systems. Under standard simplifying assumptions, a commonly used predator-prey model is given by the following system[5]

    However, spatial migration and time-delay effects are ubiquitous in real ecosystems, ranging from population diffusion caused by habitat fragmentation to response delays induced by predator breeding cycles. These factors collectively shape the complex dynamics of ecosystems. In recent years, with the interdisciplinary integration of network science and dynamical systems theory, predation models on undirected networks have attracted increasing attention.[6–9] Such models combine local node dynamics with edge-based coupling, thereby providing a more realistic framework for analyzing ecological processes in discrete spaces.

    Diffusion, as the migration process of species in space, quantifies their movement rate through diffusion coefficients. It not only reflects species’ responses to resource distribution and environmental pressure but also reshapes population spatial distribution through material exchange between nodes. For example, Miao and Zhu[10] found that in a two-dimensional reaction-diffusion model, the coupling of time delay and diffusion can induce Turing instability, leading to the emergence of spatial patterns such as stripes and spots. This pattern-formation mechanism is particularly important in fragmented habitats. When diffusion interacts with predation and reproduction, species may avoid local extinction through spatially heterogeneous distributions.[11,12]

    Time delay originates from inherent delays in ecological processes, such as prey maturation cycles, predator gestation delays, or response intervals to environmental changes. Suriguga et al.[13] showed that time delay can lead to bifurcation in a predation model, triggering periodic oscillations in population sizes; Wang et al.[14] further discovered that the superposition of prey spatial memory and a delay in predator gestation can exert a regulatory effect on the stability of the positive equilibrium state of the system, even triggering persistent oscillations under diffusion-free conditions. The dynamical complexity induced by such time delays makes ecosystems exhibit nonlinear responses to parameter changes, increasing the difficulty of predicting population dynamics.[15,16]

    Undirected networks, characterized by symmetric connections, provide a natural framework for describing mutualistic or competitive interactions among species. In predation systems, undirected edges usually denote the bidirectional migration of species between adjacent habitats, and topological features such as node degree and clustering coefficient directly influence the diffusion efficiency and stability of the system. For example, Gou et al.[17] pointed out that the eigenvalue distribution of the undirected network Laplacian determines the critical bifurcation parameters: fully connected graphs suppress spatial heterogeneous oscillations due to strong coupling effects, while sparse networks may promote non-uniform population distributions.

    Embedding diffusion and time delays into undirected network models enables the construction of more realistic ecological scenarios. Taking the Leslie-Gower predation system as an example, when considering time-delay reaction-diffusion under uniform Neumann boundary conditions, Jiao et al.[18] found that the system can exhibit spatial Hopf bifurcation, with oscillation patterns closely related to network connection density. Zhou et al.[19] further introduced non-local intraspecific competition and spatial memory, confirming that memory-based diffusion coefficients significantly alter the system’s bifurcation thresholds and oscillation periods. In addition, related studies published in domestic journals have also shown that bifurcation-oriented analysis is important for understanding complex oscillatory behaviors in nonlinear dynamical systems.[20] These studies indicate that the topological structure of undirected networks synergizes with diffusion-delay effects to jointly regulate ecological system dynamics.

    Despite substantial progress in the study of predation systems on undirected networks, their high dimensionality and nonlinearity still pose significant theoretical challenges. In existing research, linear stability analyses, such as Turing instability analysis and Hopf bifurcation analysis, have been well developed, but the characterization of complex dynamics after bifurcation (such as the stability of periodic solutions and multi-frequency oscillation mechanisms) still relies on normal form theories. Gou et al.[17,21] first proposed a Hopf normal form derivation framework for undirected network reaction-diffusion systems, successfully extending the center manifold reduction method to discrete space models by decomposing the eigenstructure of high-dimensional Jacobian matrices using Kronecker products. However, this framework mainly targets delay-free systems. When introducing multiple time-delays (such as predator response delays and prey migration delays), the analytical calculation of normal form coefficients and the determination of bifurcation curves still face technical bottlenecks. This difficulty involves not only eigenvalue tracking in multi-parameter spaces but also the treatment of delay-induced nonlocal nonlinear couplings.[22]

    Additionally, the interaction of multiple ecological factors has motivated the development of more complex systems. For instance, Anand et al.[23] incorporated prey fear effects and predator group hunting behavior into undirected network models, revealing that time-delay anti-predation responses can enhance system oscillation amplitudes; Hou and Ding[24] constructed a vector-borne disease-predation system with non-local effects and memory diffusion for pine wilt disease transmission, uncovering the regulatory mechanism of network topology on disease transmission thresholds. These studies indicate that the development of undirected network predation systems should focus more on the integration of multi-factor coupling and advanced mathematical methods.[25,26]

    This paper studies a two-delay predation system with diffusion and undirected network coupling. The delay τ1 characterizes the delayed growth effect of the prey, while τ2 represents the delayed predation effect. Accordingly, we consider the following two-delay predation system:

    where x(t) and y(t) denote the prey and predator densities at time t, respectively. The parameter r > 0 is the intrinsic growth rate of the prey, and k=rK>0 is the intraspecific competition coefficient, where K is the carrying capacity of the prey. The term −kx2 describes density-dependent self-limitation, and mxa+x is the Holling type-II functional response. Moreover, m > 0 denotes the capture rate, a > 0 is the half-saturation constant, c > 0 is the conversion efficiency, and p > 0 is the predator mortality rate.

    The main purpose of this paper is to analyze how the two delays affect the local stability and Hopf bifurcation of the positive equilibrium in the networked predation system. By exploiting the Kronecker-product structure and Laplacian-mode decomposition, we reduce the high-dimensional characteristic equation to a family of mode-dependent characteristic equations. This allows us to derive explicit stability-switching curves in the (τ1,τ2)-plane and characterize the corresponding Hopf bifurcation properties by means of the Hassard method and center-manifold reduction.

    The rest of this paper is organized as follows. Section 2 is devoted to the stability-switching curves of the two-delay predation system with diffusion and network coupling. Section 3 studies the stability and Hopf bifurcation of the system. Section 4 presents numerical simulations to verify the theoretical results and illustrate the corresponding dynamical behaviors.

2.   Stability-switching curves of a two-delay predation system with diffusion and network coupling
  • The basic predation system (2) has the following equilibrium points:

    The equilibrium E* is positive if and only if the conditions p < cm and (ak + r) p < cmr are satisfied.

    To incorporate spatial diffusion and network coupling, we introduce the Laplacian matrix and consider the following coupled two-delay predation system:

    For an undirected network, the Laplacian matrix L = {Lij} is symmetric and hence has real eigenvalues together with an orthonormal basis of eigenvectors. Let Λl be an eigenvalue of L with corresponding eigenvector ψl, l = 1,2,…,N. Here, k1 and k2 denote the diffusion coefficients of the prey and predator populations, respectively.[27] Let ui(t)=xi(t)−x* and vi(t)=yi(t)−y* denote the deviations from the equilibrium E*. Linearizing system (3) around the equilibrium E*, we obtain

    where

    In the above matrices, the coefficients are given by

    Let U := (u1(t), v1(t), …, uN(t), vN(t))T, Uτ1 := (u1(t – τ1), v1(t – τ1), …, uN(t – τ1), vN(t – τ1))T, Uτ2 := (u1(t – τ2), v1(t – τ2), …, uN(t – τ2), vN(t – τ2))T. Then above linearizing system can be written in matrix form as

    where IN is the N × N identity matrix and ⊗ denotes the Kronecker product.

    Assume that ψl=(ψl1,ψl2,…,ψlN)T is an eigenvector of the Laplacian matrix L corresponding to the eigenvalue Λl, i.e., Lψl = Λlψl. Specifically, for i,l = 1,2,…,N, ∑j=1NLijψlj=Λlψli.

    Substituting U(t) = ψl⊗Ceλt, where C=(C1,C2)T≠0, into system (4), we obtain the characteristic equation associated with the mode Λl

    Accordingly, the stability of the linearized system is determined by the roots of the corresponding characteristic equation. For convenience, we define

    We now state the following result.

    Theorem 1 For the mode matrices

    if Bl has eigenvalue λ (with corresponding eigenvector Cl = (Cl1,Cl2)T satisfying BlCl = λCl), then λ is also an eigenvalue of B. The corresponding eigenvector of B is ψl⊗Cl.

    Proof By direct calculation (for each l = 1,2,…,N),

    Thus, B(ψl⊗Cl) = λ(ψl⊗Cl).

    Then, expanding det(Bl – λI2) = 0, we obtain the characteristic equation for each mode l

    where αl1=−(k1Λl+k2Λl+a11+a22),αl2=a11a22+(a11k2+a22k1)Λl+k1k2Λl2,αl3=−b11,αl4=b11k2Λl+a22b11,αl5=−c12a21.

    Since our interest lies in the coexistence state, we focus on the positive equilibrium E*. We next analyze system (3) in the following two cases.

    Case 1 Delay-free (τ1 = τ2 = 0).

    When τ1 = τ2 = 0, Eq. (5) reduces to

    By the Routh-Hurwitz criterion, the positive equilibrium E* is locally asymptotically stable if and only if αl1+αl3>0 and (αl1+αl3)(αl2+αl4+αl5)>0 for all l = 1,2,…,N. This condition gives the local stability criterion for the positive equilibrium in the delay-free case.

    Case 2 Delay-included (τ1 and τ2 are not both zero).

    We rewrite Eq. (5) in the form

    where Pl0(λ)=λ2+αl1λ+αl2, Pl1(λ)=αl3λ+αl4, Pl2(λ) = αl5. The following result is based on Ref. [28].

    From the above conditions, we obtain the following lemma.

    Lemma 1 For each ω>0, the following statements hold.

    (1) If Pl0(iω)≠0 and (τ1,τ2)∈R+2, then λ = iω can be a solution of Pl(λ,τ1,τ2)=0 if and only if

    (2) If Pl0(iω)=0 and (τ1,τ2)∈R+2, then λ = iω can be a solution of Pl(λ,τ1,τ2)=0 if and only if

    For convenience, conditions (8) and (9) can be rewritten as

    Therefore, λ = iω is a solution of Pl(λ,τ1,τ2) = 0 if and only if the inequalities Fl1≤0,Fl2≤0,Fl3≤0 hold. We denote by Ωl the set of all positive ω satisfying Eqs. (8) and (9). For any given ω0 ∈ Ωl with Plk(iω)≠0 for k = 0,1,2, it follows from Eq. (10) that the pair (τ1,τ2) satisfies

    where θl1, θl2 ∈ [0,π] are

    and μ0+, μ0−, ν0+, ν0− represent the smallest feasible integers (which might be negative and potentially depend on ω0) such that the corresponding τl1μ0++, τl1μ0−−, τl2ν0++, τl2ν0−− values computed are non-negative. Hence, for each ω0 ∈ Ωl, formulas (11) and (12) determine the corresponding stability-switching curves in the τ1 – τ2 plane. Differentiating the characteristic Eq. (7) with respect to τ1 and τ2, we obtain the transversal conditions

    where

    Let

    For each mode l, let Tlj denote the j-th continuous curve segment parameterized by ω0, and let Tl0 denote the interior region enclosed by the coordinate axes and these curves. Moreover, let Tl=∪j=1mTlj, which constitutes the stability-switching curves.

    Based on the above analysis, we obtain the following theorem on the stability switching of the positive equilibrium E*.

    Theorem 2 Assume that conditions (13) and (14) hold. Let Tlj=(τl10j(ω0),τl20j(ω0)) denote the curve segments on Ωl, and let Tl0 denote the corresponding interior region. For E* of Eq. (3):

    (I) If (τ1,τ2)∈Tl0, then E* is asymptotically stable;

    (II) If (τ1,τ2)∈Tlj, then Eq. (5) has a pair of purely imaginary roots, and a Hopf bifurcation occurs at E*.

    To implement Theorem 2 numerically, we summarize the procedure for computing the stability-switching curves in the (τ1,τ2)-plane as follows.

    Algorithm 1 Numerical computation of the stability-switching curves.
    1: Step 1. Fix a Laplacian eigenvalue Λl and compute the functions Fl1(ω), Fl2(ω), and Fl3(ω) for ω > 0.
    2: Step 2. Determine the admissible frequency set Ωl={ω>0:Fl1(ω)≤0,Fl2(ω)≤0,Fl3(ω)≤0}.
    3: Step 3. For each ω ∈ Ωl, compute θl1(ω) and θl2(ω), and then evaluate τ1(ω) and τ2(ω) from Eqs. (11)–(12).
    4: Step 4. Choose the smallest integers μ and ν such that τ1(ω) ≥ 0 and τ2(ω) ≥ 0.
    5: Step 5. Use the signs of Re(dλ/dτ1) and Re(dλ/dτ2) to determine the crossing direction.
    6: Step 6. Plot the curves in the (τ1, τ2)-plane and identify the stable region according to Theorem 2.

3.   The stability associated with the Hopf bifurcation
  • By employing the Hassard method and center manifold theorem,[29] we investigate the stability of the Hopf bifurcation and derive the conditions for the bifurcation direction and the stability of bifurcating periodic solutions at the critical value (τl10,τl20). We consider two scenarios: 0 < τl20 ≤ τl10 and 0 < τl10 < τl20.

    (I) 0 < τl20 ≤ τl10

    Let τ1 = τl10 + μ, μ ∈ R, t = tτ1, xl(tτ1)=x^l(t), yl(tτ1)=y^l(t). Note that xl=x^l and yl=y^l, and denote the solution of Eq. (3) by x(t)=(x1(t),y1(t),…,xN(t),yN(t))T. Set xt(θ)=x(t+θ), θ ∈ [−1,0], and xt ∈ C([−1,0], R2N) ≐C1. It can be ascertained that at the critical value τ1 = τl10, Eq. (3) experiences a Hopf bifurcation, with the corresponding eigenvalues being ±iω0. Thereafter, we represent Eq. (3) in the vicinity of E* as follows:

    where Lμ,gμ:C1→R2N are linear and nonlinear operators, respectively. The linear operator Lμ is defined as

    and the nonlinear operator gμ is given by

    where A10,A11,A12,L,K are defined in Eq. (4)

    According to the Riesz representation theorem, there exists a bounded variational matrix function η(θ,μ)2N×2N for θ ∈ [−1,0], such that Lμυ satisfies

    In fact, we can take

    For ϕ ∈ C1, define

    When θ = 0, system (15) is equivalent to the following equation:

    where xt(θ)=x(t+θ), θ∈[−1,0]. Denote A = A(0) and η(θ)=η(θ,0).

    For υ∈C1([0,1],(ℝ2N)*), define

    and the bilinear inner product

    Then A and A* are adjoint operators. It is known that ±iω0τl10 are eigenvalues of A. Therefore, they are also eigenvalues of A*.

    Suppose that the linear operator A has a pair of purely imaginary eigenvalues ±iω0τl10 corresponding to the eigenvectors q(θ) and q¯(θ). Then Aq(θ)=iω0τl10q(θ). By calculation, we have

    where

    Similarly, we can obtain q*(s) as an eigenvector of A* corresponding to the eigenvalue −iω0τl10, where

    where the component is given by

    ψl is the real eigenvector of the network Laplacian matrix L with corresponding eigenvalue Λl, and ψl* is the real eigenvector of the matrix LT corresponding to the eigenvalue Λl. Now, we calculate the inner product 〈q*(s),q(θ)〉 as follows:

    Then we note D¯=[∑j=1Nψljψlj*¯(1+ql1ql1*¯+τl10b11e−iω0τl10+ql1τl20c12e−iω0τl20)]−1, where the normalization conditions 〈q*(t),q(θ)〉=1 and 〈q*(t),q¯(θ)〉=0 are satisfied.

    Let xt be the solution of Eq. (16) when μ = μ0. Then we decompose ℝ2N=Tc⊕Tsu, where Tc:={zq(θ)+zq¯(θ)|z∈ℂ} and Tsu:={xt|〈q*,xt〉=0}. For any

    there exist z∈ℂ and

    such that

    or, equivalently, in component form

    for j, l = 1,2,…,N. On the center manifold Cμ0, we have

    where z and z¯ represent the local coordinates of the center manifold Cμ0 in the directions of q and q¯, respectively. Note that if xt is real, then W is also real. We only consider real solutions. For the solution xt ∈ Cμ0 of Eq. (16), since μ = μ0, we have

    Denote

    Then

    To discuss the properties of the periodic solutions of the equation near the origin, it is only necessary to calculate g02, g11, g20, and g21. Below, we give an algorithm for calculating gij.

    Since f(μ,ϕ) is at least quadratic in ϕ, g02, g11, and g20 can be directly obtained from Eq. (18), while g21 is generally related to the coefficients W11 and W20 in W(z,z¯,θ). Next, we solve for W11 and W20.

    From Eqs. (16)–(18), we get

    Thus,

    Equation (16) is transformed into

    Denote

    On the other hand, in a neighborhood of the origin, we have

    Therefore, substituting Eqs. (18) and (20) into Eq. (19) and comparing the coefficients of terms with the same degree in z, we obtain

    Note that Hij is independent of Wij(θ) when i + j = 2. Also, since 2iω0 and 0 are not eigenvalues of A, W20, W11, and W02=W¯20 can be solved from Eq. (21). That is, from Eq. (21), when θ ∈ [−1,0), we have

    Comparing the coefficients in Eq. (20), we get

    From Eqs. (21) and (22) and the definition of matrix A, we obtain

    Note that q(θ)=q(0)eiω0τl10θ. Then,

    where both E1 and E2 are 2N-dimensional constant vectors.

    Also, when θ = 0, we have

    Substitute Eqs. (25) into Eqs (23) and (24) respectively, and note that

    Then we can calculate that

    where

    By solving the linear system corresponding to the complement space, we obtain the coefficients gij as following:

    and further calculate the key bifurcation parameters.

    (i) The direction of the Hopf bifurcationμ2=−Re{c1(0)}Re{λ′(τl10)}. Precisely, when Re{λ′(τl10)}>0, if μ2>0 (alternatively, μ2<0), the Hopf bifurcation is supercritical (subcritical, correspondingly), and the periodic solution comes into existence for τ1 > τl10 (respectively, τ1 < τl10). However, if Re{λ′(τl10)}<0, the bifurcating periodic solutions occur in the opposite direction.

    (ii) The stability of the bifurcated periodic solutionsβ2=2Re(c1(0)). More precisely, the bifurcated periodic solutions are stable when β2<0; conversely, they are unstable when β2>0.

    (iii) The period of the bifurcated periodic solutionsT2=−Im{c1(0)}+μ2Im{λ′(τl10)}ω0τl10. More precisely, the period increases when T2>0; in contrast, it decreases when T2<0.

    Where c1(0)=i2ω0τl10(g20g11−2|g11|2−13|g02|2)+g212.

    (II)0<τl10<τl20

    Let τ2 = τl20 + μ2. Subsequently, we recast Eq. (3) in the vicinity of E* as a functional differential equation within the space C1

    where

    Using Hassard’s method, we calculate the following values:

    and

    further calculate the key bifurcation parameters.

    (i) The direction of the Hopf bifurcationμ2=−Re{c1(0)}Re{λ′(τl20)}. Precisely, in the case where Re{λ′(τl20)}>0, if μ2>0 (alternatively, μ2<0), the Hopf bifurcation is supercritical (subcritical correspondingly), and the periodic solution comes into existence for τ2 > τl20 (respectively τ2 < τl20). If Re{λ′(τl20)}<0, however, the bifurcating periodic solutions will be in the opposite direction.

    (ii) The stability of the bifurcated periodic solutionsβ2=2Re{c1(0)}. To be exact, the bifurcation periodic solutions possess stability when β2<0; conversely, they lack stability when β2>0.

    (iii) The period of the bifurcated periodic solutionsT2=−Im{c1(0)}+μ2Im{λ′(τl20)}ω0τl20. More precisely, the period experiences an increment when T2>0; in contrast, it undergoes a decrement when T2<0, where

4.   Simulation
  • To verify the theoretical results, we consider the following specific form of system (3):

    The positive equilibrium of system (27) is E* = (0.55,0.27).

    For numerical illustration, we focus on the Laplacian mode Λ1 = 0. Here, Λ1 = 0 is considered purely as a Laplacian mode in the modal stability analysis, rather than as a criterion for network connectivity. The purpose of this section is twofold: first, to verify the stability-switching and Hopf bifurcation results obtained in Sections 2 and 3; second, to illustrate representative dynamical behaviors of the networked system for different values of the link probability p.

    For the mode Λ1 = 0, Eq. (5) reduces to

    When τ1 = τ2 = 0, the roots of Eq. (28) are −0.2112 ± 0.5367i, which implies that the positive equilibrium E* is asymptotically stable.

    When at least one of τ1 and τ2 is nonzero, Lemma 1 shows that λ=iω is a solution of Eq. (28) if and only if ω ∈ (0.1175,0.6198). Accordingly, the admissible frequency interval is obtained from the functions F11(ω), F12(ω), and F13(ω), as shown in Fig. 1(a). Based on Theorem 2, the corresponding stability-switching curve T1 in the (τ1,τ2)-plane is plotted in Fig. 1(b), and the region enclosed by the coordinate axes and T1 gives the asymptotically stable region of the positive equilibrium E*.

    To further validate the theoretical prediction, we directly compute the characteristic roots of Eq. (28) at several representative points in the (τ1,τ2)-plane. We select one point inside the stable region, one point near the switching curve, and one point in the unstable region. The corresponding rightmost roots confirm that the maximum real part is negative in the stable region, close to zero near the switching curve, and positive in the unstable region. Hence, the direct root computation is consistent with the prediction of Theorem 2, as shown in Table 1.

    Figure 2 displays three representative dynamical regimes corresponding to the above parameter choices. For (τ1,τ2)=(9,5), the solution converges to the positive equilibrium E*, which agrees with the stability prediction. For (τ1,τ2) = (17.6536,6.38558), the trajectory exhibits periodic oscillations near the stability-switching boundary, indicating the onset of Hopf bifurcation. For (τ1,τ2) = (29,10), the equilibrium loses stability, and the solution exhibits more complex oscillatory behavior.

    Next, we illustrate how different values of the link probability p affect the observed dynamical behavior of the networked system. Here, p is used only as a numerical parameter in the random network construction, and the following simulations are intended to provide qualitative illustrations rather than additional theoretical conclusions.

    We consider three representative cases: p = 0, p = 0.03, and p = 0.07.

    When p = 0, the nodes are effectively isolated from one another, and the dynamics at different nodes evolve independently. In this case, the network exhibits weak coherence, and the temporal behaviors are mainly governed by the local node dynamics.

    When p = 0.03, the coupling between nodes is present but remains relatively weak. As a result, interactions among neighboring nodes become visible, and the solutions exhibit partial coherence together with noticeable spatial heterogeneity.

    When p = 0.07, the coupling effect becomes stronger, and the collective dynamics become more coherent across the network. The simulations indicate that stronger coupling tends to suppress disorder and promote more coherent collective behavior.

    Figures 3 and 4 present the corresponding numerical solutions for these three values of p. These results show that, under the same local parameters, varying the network link probability may significantly influence the observed temporal and spatiotemporal patterns of the coupled predation system.

5.   Conclusion
  • In this paper, we investigated a two-delay predation system with diffusion and network coupling. By combining the Laplacian-mode decomposition with the Kronecker-product formulation, we reduced the stability analysis of the high-dimensional networked system to a family of mode-dependent characteristic equations. On this basis, we derived the corresponding stability-switching curves in the (τ1,τ2)-plane and obtained sufficient criteria for the stability of the positive equilibrium.

    Furthermore, by employing the Hassard method and center-manifold reduction, we analyzed the Hopf bifurcation generated at the critical delay values and characterized the direction of bifurcation, the stability of the bifurcating periodic solutions, and the variation of their periods. These results clarify how the interplay between the two delays influences the local dynamics near the positive equilibrium.

    The numerical examples support the theoretical analysis. In particular, the stability-switching curves predicted by the characteristic equation are consistent with the simulated dynamical behaviors observed at representative parameter points. The simulations further illustrate typical temporal and spatiotemporal patterns under network coupling.

    Overall, the present study establishes a tractable analytical approach for studying delay-induced stability switching and Hopf bifurcation in predation systems with diffusion and network coupling. The framework developed here may also be useful for analyzing other delayed networked biological models with similar coupling mechanisms.

    Future work may consider extending the present setting to more general ecological systems, such as models with multiple interacting species, heterogeneous delays, or stochastic perturbations, in order to further understand the influence of complex ecological factors on the stability and oscillatory behavior of networked population systems.

Figure (4)  Table (2) Reference (29)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return