A class of three-dimensional Gause-type predator-prey model with delay is
considered. Firstly, a group of sufficient conditions for the existence of Hopf bifurcation is obtained via employing the polynomial theorem by analyzing the distribution of the roots of the associated characteristic equation. Secondly, the direction of the Hopf bifurcation and the stability of the bifurcated periodic solutions are determined by applying the normal form method and the center manifold theorem. Finally, some numerical simulations are carried out to illustrate the obtained results.
1. Introduction
Multispecies predator-prey models arise frequently on various ecosystems [1–4]. They usually exhibit more rich and complex dynamics as the number of species increases. Take three-species models, for example. There might be a food chain, or two predators or preys, whose relations might be cooperative or competitive. It has attracted extensive studies on the dynamics of various multispecies predator-prey models. Cheng et al. [5] derived some results to ensure the global stability of a predator-prey system with Holling’s type III functional response. klebanoff and Hastings [6] investigated a three-species food chain model. It showed that the system exhibits rich complexity features such as stable, periodic, and chaotic dynamics. Hsu and Huang [7] deals with the question of global stability of the positive locally asymptotically stable equilibrium in a class of predator-prey systems by using the Dulac’s criterion and constructing Lyapunov functions. See [8–14].
In 1977, Freedman and Waltman [12] investigated the following general Gause-type predator-prey model:dx(t)dt=xg(x)-yp(x),dy(t)dt=y[-h+ep(x)]-zq(y),dz(t)dt=z[-s+mq(y)],
where x(t), y(t), and z(t) are the population densities of prey, predator, and top predator at time t, respectively. g(x) is the intrinsic growth rate of prey; p(x) and q(y) are the specific growth rates of predator and top predator; h, s>0 are the death rates of y(t) and z(t); e, m>0 are the conversion rates for prey and predator. They established the stability criteria and argued that the unique interior equilibrium exists and is locally asymptotically stable. Ginoux et al. [13] highlighted that this model has several Hopf bifurcations and a period-doubling cascade generating a snail shell-shaped chaotic attractor. They compared the quantity and property of the equilibria, bifurcation structure, and shape of attractors among this model and the so-called Rosenzweig-MacArthur and Hastings-Powell models and also gave the bifurcation analysis for each model. Hastings and Powell [14] discussed a continuous time model of a food chain incorporating nonlinear functional responses, and the model exhibits chaotic in long-term behavior when appropriate biologically reasonable parameter values are chosen. They found that, for different values of the key parameter, the system exhibits several types of asymptotic motions, namely, stable equilibrium point, limit cycles, change in periodicity of these cycles, and the so called teacup chaos.
We know that the population outbreak may happen for the species with periodic fluctuation. The outbreaks of pests and mice are famous. For example, Finerty discovered in Canada the populations of polar rabbits, lynxs and Ondatrazibethicas have 10-year cycle fluctuations, and the populations of lemmings and some murine experience 4-year cycle fluctuations. When the peak arrived, the vegetation would be severely damaged. Some cyclical fluctuations are consistent with the periodicity of intensity of certain infectious diseases. Thus, it is of great significance to study periodic solutions of biological systems for controlling insect. It is known that the delay differential equations (DDEs) will exhibit much more complicated dynamics than ordinary differential equations (ODEs) such as the existence of Bogdanov-Takens bifurcation and even chaos. See [15–17]. Many authors have researched the different delays Gause-type predator-prey systems [18–22].
For two-dimensional Gause-type predator-prey model,dx(t)dt=x(t)g(x(t))-y(t)p(x(t)),dy(t)dt=y(t)[-d+μp(x(t))],
the time delay can be incorporated into (1.2) at three different terms [4], including the prey specific growth term g(x(t)), the predator response term p(x(t)), or the interaction term y(t)p(x(t)). The detailed analysis on stability and bifurcation can be found in [23]. Because the prey usually has a gestation or reaction time of predator, introducing time delay is necessary and hence changes the model into a DDE. It is well known that delay could cause a stable equilibrium to be unstable and induce bifurcations as well as periodic oscillations. Under the hypothesis that prey x(t) has a gestation in (1.1), we modify it to be the following one:dx(t)dt=xg(x)-yp(x-τ),dy(t)dt=y[-h+ep(x)]-zq(y),dz(t)dt=z[-s+mq(y)],
where τ is the time of gestation.
The purpose of current work is to analyze the effect of delay on the dynamics for (1.3), so we plan to employ bifurcation analysis approach with delay τ being the parameter. In particular, we choose g(x)=α(1-(x/K)), p(x)=βx/(1+px), and q(y)=ry. So the following delay model will be obtained:dx(t)dt=αx(1-xk)-βyx(t-τ)1+px(t-τ),dy(t)dt=y[-h+eβx1+px]-rzy,dz(t)dt=z[-s+mry],
where α,β,k,p,h,e,r,s,m are all positive parameters.
Our results reveal that Hopf bifurcation can occur as the delay crosses some critical values which leads to the existence of periodic solution that may conform to certain phenomena in ecosystem system.
The rest of the paper is organized as follows. In Section 2, we first investigate the stability of coexisting equilibrium and the existence of the Hopf bifurcation of (1.4) by analyzing the characteristic equation of the associated linearized system. In Section 3, we derive an explicit formula for determining the stability and the direction of bifurcating periodic solutions by the normal form method and the center manifold theory. In Section 4, we carry out some numerical simulations to illustrate the results obtained and forecast the change of population quantity.
2. Stability and Hopf Bifurcation of Coexisting Equilibrium
For the sake of convenience, we nondimensionalized (1.4) with the following: scalingx→xk,t→αt,y→y,z→z,
then (1.4) takes the formdx(t)dt=x(1-x)-ayx(t-τ)x(t-τ)+b,dy(t)dt=y(-l+cxx+b-rz),dz(t)dt=z(-s+dy),
wherea=βpkα,b=1pk,r=hα,c=eβpα,d=mr.
Obviously, the delay cannot change the number of equilibria and non-dimensionalizations cannot change the properties of system. Through simple analysis, we know (2.2) has four equilibria: E1(0,0,0), E2(1,0,0), E3(x0,y0,0), and E¯(x*,y*,z*) withx0=blc-l,y0=(1-x0)(x0+b)a,x*=(1-b)+(1-b)2+4b-(4as/d)2,y*=sd,z*=-lr+cx*r(x*+b).
There is no obvious biological significance for E1(0,0,0) and E2(1,0,0). In this paper, we mainly study the change of stability of coexisting equilibrium E¯(x*,y*,z*) with the variation of time delay. If as/d≤b<1, then E¯(x*,y*,z*) is the uniqueness equilibrium of (2.2). We consider the linearized system of (2.2) at E¯. The characteristic equation at E¯ is given by|λ-m11-n11e-λτ-m120-m21λ-m230-m32λ|=0,
wherem11=1-2x*,m12=-ax*x*+b<0,m21=bcy*(x*+b)2>0,m23=-ry*<0,m32=dz*>0,n11=-aby*(x*+b)2<0.
The eigenvalue λ satisfies the characteristic equationD(λ,τ)=λ3+a2λ2+a1λ+a0+(b2λ2+b0)e-λτ=0,
where a2=-m11, a1=-m12m21-m32m23>0, a0=m32m23m11, b2=-n11>0, and b0=m32m23n11>0.
If m11<0, then all the eigenvalues of (2.7) have negative real parts when τ=0 by the Routh-Hurwitz criterion. So we have the Lemma 2.1.
Lemma 2.1.
If as/d≤b<1 and m11<0, then the coexisting equilibrium E¯(x*,y*,z*) of (2.2) is locally asymptotically stable.
It is known that E¯(x*,y*,z*) is asymptotically stable if all roots of the corresponding characteristic equation (2.7) have negative real parts. We shall study the distribution of the roots of the (2.7) when τ≠0. We assume that the equilibrium E¯(x*,y*,z*) of the ODE model is stable, then we derive some conditions to ensure that the steady state of the delay model is still stable.
We compute the eigenvalues of the Jacobian matrix at the coexisting equilibrium E¯(x*,y*,z*). Substituting λ=iω into (2.7) yields,
when ω=0, D(0,τ)=a0+b0=m23m32(n11+m11)≠0,
when ω≠0, D(iω,τ)=(iω)3+a2(iω)2+a1iω+a0+(b0-b2ω2)e-iωτ=0. Separating the real and imaginary parts, we get
-a2ω2+a0+(b0-b2ω2)cosωτ=0,-ω3+a1ω-(b0-b2ω2)sinωτ=0.
Consequently, we getω6+(a22-2a1-b22)ω4+(a12-2a0a2+2b0b2)ω2+a02-b02=0.
Let ω2=l, A=a22-2a1-b22, B=a12-2a0a2+2b0b2 and C=a02-b02, then (2.9) becomesl3+Al2+Bl+C=0.
From Ruan and Wei [24] or Li and Wei [25], we have the following results on the distribution of roots of (2.10).
IfC≥0, then (2.10) has at least one positive root.
IfC≥0 and Δ<0, then (2.10) has no positive root.
IfC≥0, then (2.10) has positive roots ⇔ if l1>0 and h′(l1)≤0.
Suppose (2.10) has root with positive real part. Without loss of generality, we assume that it has three positive roots, denoted by l1,l2,l3, respectively. Then, (2.10) has three positive roots ωi=li(i=1,2,3). Let τk(0) be the unique root of (2.8) such that τk(0)ωk∈[0,2π). Also denote byτk(j)=τk(0)+2jπωk,
for k=1,2,3.
So (±ωk,τk(j)) is the solution of (2.7). Clearly,limτk(j)=∞,k=1,2,3.
We can defineτ0=τk0=minτk(0),ω0=ωk0,
that is, ±iω0 is the purely imaginary roots of (2.7) for τ=τ0. So we have the following.
Lemma 2.3.
τ0 is defined by (2.14), then
(1) if one of the following holds (i)C<0; (ii)C≥0, Δ≥0; (iii) C≥0,l1>0; h′(l1)>0, then all roots of (2.7) have negative real parts for τ∈[0,τ0);
(2) if none of the conditions (i)–(iii) are satisfied, then all toots of (2.7) have negative real parts for all τ≥0.
So we have the following theorem.
Theorem 2.4.
Supposed that
as/d≤b<1 and m11<0,
eitherC≥0,A2-3B<0, orC≥0,B>0,
then the equilibrium E¯(x*,y*,z*) of the delay model (2.2) is absolutely stable; that is, E¯(x*,y*,z*) is asymptotically stable for all τ≥0.
Denoting λ(τ)=α(τ)+β(τ) to be the root of (2.7) satisfying α(τ0)=0,ω(τ0)=ω0, then we have the following Lemma 2.5.
Lemma 2.5.
Suppose A2-3B<0,
±iω0 is a pair of simple purely imaginary roots of (2.7) when τ=τ0,
[dReλ(τ)/dτ]|τ=τ0>0.
Proof.
If iω0 is not simple, then ω0 must satisfy
ddτ[λ3+a2λ2+a1λ+a0+(b2λ2+b0)e-λτ]|λ=iω0=0,
that is, ω0 and τ0 must satisfy
τ(b2ω2-b0)cosωτ+2b2ωsinωτ=3ω2-a1,τ(b2ω2-b0)sinωτ-2b2ωcosωτ=2a2ω.
Eliminating τ, we have
(a1-3ω2)sinωτ+2a2ωcosωτ=-2b2ω,
substituting sinωτ, cosωτ from (2.8) into (2.17). We have
3ω4+2(a22-2a1-b22)ω2+(a12-2a2a0+2b2b0)=0.
Recall ω02=l0, which implies
3l02+2Al0+B=0.
However if A2-3B<0, we have
h′(l0)=3l02+2Al0+B>0,
a contradiction. So ±iω0 is a pair of simple purely imaginary roots of (2.7).
Note that
M=(a1-3ω02)cosω0τ0-2a2ω0sinω0τ0+(b2ω02-b0)τ0,N=(a1-3ω02)sinω0τ0+2a2ω0cosω0τ0+2b2ω0,Q=b0ω0-b2ω03.
So
dλ(τ)dτ=iQM+iN=NQ+iMQM2+N2.
Then,
dReλ(τ0)dτ=NQM2+N2
has the same sign as that of NQ.
Now,
NQ=[(a1-3ω02)sinω0τ0+2a2ω0cosω0τ0+2b2ω0](b0ω-b2ω3)=ω02[3ω04+2(a22-2a1-b22)ω02+(a12-2a2a0+2b2b0)]=ω02[3ω04+2Aω02+B].
When A2-3B<0, NQ is greater than zero, that is [dReλ(τ)/dτ]|τ=τ0>0.
By Lemma 2.5, we have the Theorem 2.6.
Theorem 2.6.
Supposed that Lemma 2.3 is right, then, when A2-3B<0, the equilibrium E¯(x*,y*,z*) of the delay model (2.2) is asymptotically stable when τ<τ0 and unstable when τ>τ0, where τ0 is defined by (2.14). When τ=τ0, a Hopf bifurcation occurs.
3. Direction and Stability of Hopf Bifurcation
Let x1(t)=x(t)-x*, x2(t)=y(t)-y*, x3(t)=z(t)-z*, Xi(t)=xi(τt) (i=1,2,3), τ=τ0+μ, μ∈ℝ, andB=(m11m120m210m230m320),C=(n1100000000).
The system (1.4) is transformed into a functional differential equation (FDE) in C=C([-1,0],ℝ3), definingLμ(ϕ)=(τ0+μ)Bϕ(0)+(τ0+μ)Cϕ(-1),
where ϕ=(ϕ1,ϕ2,ϕ3)T∈C([-1,0],ℝ3), and the nonlinear term ish(μ,ϕ)=(τ0+μ)(-2ϕ12(0)-2ab(x*+b)2ϕ2(0)ϕ1(-1)-2bcy*(b+x*)3ϕ12(0)+2bc(b+x*)2ϕ1(0)ϕ2(0)-2rϕ2(0)ϕ3(0)2drϕ2(0)ϕ3(0)).
So μ=0 is Hopf bifurcation point.
By the Riesz representation theorem, there exists a 3×3 matrix η(θ,μ) (-1≤θ≤0), whose elements are of bounded variation functions such thatLμ(ϕ)=∫-10[dη(θ,μ)]ϕ(θ),forϕ∈C([-1,0],R3).
In fact, we can chooseη(θ,μ)={(τ0+μ)B,θ=0,0,θ∈(-1,0),(τ0+μ)C,θ=-1.
Then, (3.4) is satisfied. For ϕ∈C1([-1,0],ℝ3), we defineA(μ)ϕ(θ)={dϕ(θ)dθ,θ∈[-1,0),∫-10[dη(ξ,μ)]ϕ(ξ),θ=0,R(μ)ϕ(θ)={0,θ∈[-1,0),h(μ,ϕ),θ=0.
So (3.4) is equivalent to the following abstract equation:ẋt=A(μ)xt+R(μ)xt,
where x=(x1,x2,x3)T, xt=x(t+θ), for θ∈[-1,0].
For ψ∈C1([0,1],ℝ3*), we defineA*ψ(s)={-dψ(s)ds,s∈(0,1],∫-10ψ(-ξ)dη(ξ,0),s=0,
and a bilinear form〈ψ(s),ϕ(θ)〉=ψ¯(0)ϕ(0)-∫-10∫ξ=0θψ¯(ξ-θ)dη(θ)ϕ(ξ)dξ,
where η(θ)=η(θ,0). Then, A(0) and A* are adjoint operators. We know that ±iω0τ0 are eigenvalues of A(0), and, therefore, they are also eigenvalues of A*(0). The vector q(θ)=(q1,1,q3)Teiω0τ0θ(θ∈[-1,0]) and q*(s)=D(q1*,1,q3*)eiω0τ0s(s∈[0,1]) are the eigenvectors of A(0) and A* corresponding to the eigenvalue iω0τ0 and -iω0τ0, respectively, satisfying 〈q*(s),q(θ)〉=1,〈q*(s),q¯(θ)〉=0 with q1=m12/(iω0-n11e-iω0-m11),q3=m32/iω0, q1*=m21/(-iω0-n11eiω0-m11),q3*=m23/iω0, and D=1/(q1*q1¯+1+q3*q3¯+q1¯q1*n11e-iω0τ0). Following the same algorithms as Hassard et al. [26], we can obtain the coefficients which will be used to determine the important quantities:g02=2D¯τ0[-q1¯2q1*¯-ab(b+x*)2q1¯q1*¯eiω0τ0-bcy*(b+x*)2q1¯-rq3¯+dq3¯q3*¯],g20=2D¯τ0[-q12q1*¯-abq1q1*¯(b+x*)2e-iω0τ0-bcy*(b+x*)3q12+bc(b+x*)2q1-rq3+dq3q3*¯],g11=2D¯τ0[-2q1q1¯q1*¯-ab(b+x*)2(2Re{q1e-iω0τ0})-2bcy*(b+x*)3q1q1¯+2bc(b+x*)2Re{q1}-2rRe{q3}+2dRe{q3q3*¯}2bc(b+x*)2],g21=2D¯τ0[(-4q1q1*¯-4bc(b+x*)3q1+2bc(b+x*)2)W111(0)+(2bc(b+x*)2q1-2ab(b+x*)2q1q1*¯e-iω0τ0+2dq3q3*¯-2rq3)W112(0)+(2dq3*¯-2r)W113(0)-2abq1*¯(b+x*)2W111(-1)+(-4q1*¯q1¯-2bc(b+x*)3q1¯+bc(b+x*)2)W201(0)+(-abq1*¯(b+x*)2q1¯eiω0τ0+bc(b+x*)2q1¯+dq3*¯q3¯-rq3¯)W202(0)+(dq3*¯-r)W203(0)-abq1*¯(b+x*)2W201(-1)].
Since there are W20(θ) and W11(θ) in g21, we still need to compute them. From [25], we haveW20(θ)=ig20ω0τ0q(θ)+ig¯023ω0τ0q¯(θ)+E1e2iω0τ0θ.
According to[2iω0τ0I-∫-10dη(θ)e2iω0θ]E1=hz2,
wherehz2=((-2+2aby*(b+x*)3)q12-2ab(b+x*)2q1-2rq3+bc(b+x*)2q1e-iω0τ0-4bcy*(b+x*)3q12e-2iω0τ02dq3e-iω0τ0),
we have E1=(E1(1),E1(2),E1(3))T, where E1(i)=2Δ1(i)/Δ1(i=1,2,3), withΔ1=(-4ω02τ02-m23m32-m12m21)(2iω0τ0)+(4ω02τ02+m23m32)(n11e-2iω0τ0+m11),Δ1(1)=(4ω02τ02+m23m32)(2q12+2ab(b+x*)2q1e-iω0τ0)+2m12iω0τ0[-2bcy*(b+x*)3q12+2bc(b+x*)2q1-2rq3]+2m12m23dq3,Δ1(2)=2m21iω0τ0[-2q12-2ab(b+x*)2q1e-iω0τ0]+2iω0τ0(2iω0τ0-n11e-iω0τ0-m11)[-2bcy*(b+x*)3q12+2bc(b+x*)2q1-2rq3]+2m23(2iω0τ0-n11e-2iω0τ0-m11)dq3,Δ1(3)=m21m32[-2q12-2ab(b+x*)2q1e-iω0τ0]+m32(2iω0τ0-n11e-iω0τ0-m11)[-2bcy*(b+x*)3q12+2bc(b+x*)2q1-2rq3]+2[2iω0τ0(2iω0τ0-n11e-iω0τ0-m11)-m12m21]dq3.
And, similarly,W11(θ)=-ig11ω0τ0q(θ)+ig¯11ω0τ0q¯(θ)+E2.
According to(∫-10dη(θ))E2=-hzz¯,
wherehzz¯=((-2+2aby*(b+x*)3)q1q1¯-2ab(b+x*)2(q1+q1¯)-r(q3+q3¯)+bc(b+x*)2(q1¯eiω0τ0+q1e-iω0τ0)-4bcy*(b+x*)3q1q1¯d(q3q3*¯eiω0τ0+q3q3*¯e-iω0τ0)),
we have E2=(E2(1),E2(2),E2(3))T, where E2(i)=2Δ2(i)/Δ2(i=1,2,3), withΔ2=m23m32(m11+n11),Δ2(1)=-m23m32(-4q1q1¯-2ab(b+x*)2(2Re{q1e-iω0τ0}))+2m12m23d(2Re{q3}),Δ2(2)=-2m23(m11+n11)d(2Re{q3}),Δ2(3)=m21m32(-4q1q1¯-2ab(b+x*)2(2Re{q1e-iω0τ0}))-m32(m11+n11)[-4bcy*(b+x*)3q1q1¯+2bc(b+x*)2(2Re{q1})]-2m12m21d(2Re{q3}).
Consequently, gij can be expressed explicitly by the parameters and delay. Thus, we can compute the following values:c1(0)=i2ω0τ0(g11g20-2|g11|2-|g02|23)+g212,μ2=-Re(c1(0))Re(λ′(τ0)),T2=-Imc1(0)+μ2Imλ′(τ0)ω0τ0,β2=2Re(c1(0)),
which determine the properties of bifurcating periodic solutions at the critical value τ0. That is, μ2 determines the direction of Hopf bifurcation: if μ2>0 (μ2<0), then Hopf bifurcation at τ0 is forward (or backward); β2 determines the stability of bifurcating periodic solutions: the bifurcating periodic solutions on the center manifold are stable (unstable) if β2<0 (β2>0); T2 determines the period of the bifurcating periodic solutions: the period increases (decreases) if T2>0 (T2<0).
4. Numerical Simulations and Discussions
In this part, we perform some numerical simulations. The results not only support our previous parts but also predict the existence of global Hopf bifurcations periodic solutions. However, to prove this observation theoretically is of great challenge.
Thus, all the conditions in Theorem 2.4 are satisfied. E¯(0.855, 0.7519, 1.1272) with initial value (0.812, 0.702, 0.925) for (1) and E¯(0.7915, 0.7078, 1.3339) with initial value (0.607, 0.602, 0.895) for (2) are asymptotically stable for all τ>0 (see Figure 1). From a biological sense, the prey, predator, and top predator will have a short-term fluctuation in the initial stage as the effect of τ. But the population would tend to a steady level after a long period of time.
E¯ is asymptotically stable for all τ>0.
We choose another set of parameters which satisfy the assumptions in Theorem 2.6:
Through (2.14), we have τ0=12.6774, ω0=14.7193, and c1(0)=-0.4617+i0.6024. Utilizing Theorem 2.6, we know that the equilibrium of the delay model (2.2) is asymptotically stable when τ<τ0 (see Figure 2).
E¯(0.9977, 0.6954, 0.4432) with initial value (0.985, 0.135, 0.4) is asymptotically stable when τ=8.6017<τ0=12.6774.
Hopf bifurcation occurs when τ=τ0, and the bifurcating periodic solution is orbitally asymptotically for τ>τ0 (see Figure 3).
A stable periodic orbit of system (2.2) when initial value is (0.985, 0.135, 0.4) and τ=20.7288>τ0=12.6774.
In addition, the periodic solution of system (2.2) still exists when τ is large and its amplitude is larger compared with the solution in Figure 4. The numerical results of Figure 4 show the global existence of periodic solutions generated by the Hopf bifurcation. How to explain the phenomenon theoretically needs further researches.
The stable periodic orbits of system (2.2) when τ=200.7288 and τ=400.8768 with parameters given by (3).
5. Conclusion
In this paper, we analyze the dynamics of the equilibria coexistence for a class of three-dimensional Gause-type predator-prey model. We obtain the stability of this equilibrium and also claim that the introduced delay changes its stability while a Hopf bifurcation occurs. The existence of the bifurcation periodic solutions for sufficiently large delay has been shown by numerical simulations.
Acknowledgments
The work is supported in part by the National Natural Science Foundation of China and by the Heilongjiang Provincial Natural Science Foundation (no. A200806, A200813).
AndrewsJ. F.A mathematical model for the continuous culture of microorganisms utilizing inhibitory substrates196810670772310.1002/bit.260100602BartlettM. S.On theoretical models for competitive and predatory biological systems19574427420086727ZBL0080.36301BrauerF.SoudackA. C.Stability regions and transition phenomena for harvested predator-prey systems19797431933710.1007/BF00275152648855ZBL0397.92019I. FreedmanH.1987Edmonton, CanadaHIFR ConsultingChengK. S.HsuS. B.LinS. S.Some results on global stability of a predator-prey system198112111512610.1007/BF00275207631003ZBL0464.92021KlebanoffA.HastingsA.Chaos in three-species food chains199432542745110.1007/BF001601671284166ZBL0823.92030HsuS. B.HuangT. W.Global stability for a class of predator-prey systems199555376378310.1137/S00361399932532011331585ZBL0832.34035VarrialeM. C.cris@mat.ufrgs.brGomesA. A.A study of a three species food chain1998110211913310.1016/S0304-3800(98)00062-3ZhangS. W.ChenL. S.Chaos in three species food chain system with impulsive perturbations2005241738310.1016/j.chaos.2004.07.0142110025ZBL1066.92060McCannK.YodzisP.Bifurcation structure of a 3-species food chain model19954829312510.1006/tpbi.1995.1023HsuS. B.HwangT. W.Hopf bifurcation analysis for a predator-prey system of Holling and Leslie type19993135531676021ZBL0935.34035FreedmanH. I.WaltmanP.Mathematical analysis of some three-species food-chain models1977333-4257276068226210.1016/0025-5564(77)90142-0ZBL0363.92022GinouxJ. M.RossettoB.JametJ. L.Chaos in a three-dimensional Volterra-Gause model of predator-prey type20051551689170810.1142/S02181274050129342154962ZBL1092.37541HastingsA.PowellT.Chaos in three-species food chain199172389690310.2307/1940591JiangW.WeiJ.Bifurcation analysis in a limit cycle oscillator with delayed feedback200523381783110.1016/j.chaos.2004.05.0282093772ZBL1080.34054WangH. B.JiangW. H.Hopf-pitchfork bifurcation in van der Pol's oscillator with nonlinear delayed feedback2010368191810.1016/j.jmaa.2010.03.0122609254WeiJ. J.JiangW. H.Bifurcation analysis in van der Pol's oscillator with delayed feedback2008213260461510.1016/j.cam.2007.01.0412392579XuR.ChaplainM. A. J.Persistence and global stability in a delayed Gause-type predator-prey system without dominating instantaneous negative feedbacks2002265114816210.1006/jmaa.2001.77011874262ZBL1013.34074LiuG.YanW.YanJ.Positive periodic solutions for a class of neutral delay Gause-type predator-prey system200971104438444710.1016/j.na.2009.03.0022548674ZBL1181.34089K. LiY.KuangY.Periodic solutions in periodic delayed Gause-type predator-prey systems20013375382ZhaoT.KuangY.SmithH. L.Global existence of periodic solutions in a class of delayed Gause-type predator-prey systems19972881373139410.1016/0362-546X(95)00230-S1428657ZBL0872.34047DingX.JiangJ.Positive periodic solutions in delayed Gause-type predator-prey systems200833921220123010.1016/j.jmaa.2007.07.0792377079ZBL1137.34033RuanS.Absolute stability, conditional stability and bifurcation in Kolmogorov-type predator-prey systems with discrete delays20015911591731811101ZBL1035.34084G. RuanS.J. WeiJ.On the zeros of a third degree exponential polynomial with applications to a delayed model for the control of testosterone secretion20011814152LiX.WeiJ.On the zeros of a fourth degree exponential polynomial with applications to a neural network model with delays200526251952610.1016/j.chaos.2005.01.0192143965ZBL1098.37070HassardB. D.KazarinoffN. D.WanY. H.198141Cambridgem, UKCambridge University Pressv+311 pp. (microfiche insert)London Mathematical Society Lecture Note Series603442