An SIR epidemic model with incubation time and
saturated incidence rate is formulated, where the susceptibles are
assumed to satisfy the logistic equation and the incidence term is
of saturated form with the susceptible. The threshold value
ℜ0 determining whether the disease dies out is found. The
results obtained show that the global dynamics are completely
determined by the values of the threshold value ℜ0 and time
delay (i.e., incubation time length). If ℜ0 is less than one,
the disease-free equilibrium is globally asymptotically stable and
the disease always dies out, while if it exceeds one there will be
an endemic. By using the time delay as a bifurcation parameter, the
local stability for the endemic equilibrium is investigated, and the
conditions with respect to the system to be absolutely stable and
conditionally stable are derived. Numerical results demonstrate that
the system with time delay exhibits rich complex dynamics, such as
quasiperiodic and chaotic patterns.
1. Introduction
Epidemic models with
nonlinear incidence have been studied by many authors, and related literature
of SIR disease transmission model is quite large, where S denotes the number of
individuals that are susceptible to infection, I denotes the number of individuals that are
infectious, and R denotes the number of individuals that
have been removed with immunity. For example, a detailed dynamical analysis of the nonlinear
incidence rate βIpSq (where β is the average number of contacts per
infective per unit time) is given by Liu et al. [1, 2], Hethcote and van den Driessche [3], Moghadas and Alexander [4],
Korobeinikov and Maini [5], and others. After a study of the cholera epidemic
spread in Bari in 1973, Capasso and Serio [6] introduced a saturated incidence rate g(I)S into epidemic models, the incidence rate seems
more reasonable than βIpSq because the number of effective contacts
between infective individuals and susceptible individuals may saturate at high
infective levels due to crowding of infective individuals or due to the
protection measures by the susceptible individuals [7], such as the incidence rate β(IpS/(1+αIq)) (see [1, 8] and references therein).
Zhang and Chen [9] investigated a class of SIR epidemiological models
under assumption that the susceptible satisfies the logistic equation and the
incidence rate is of the form βISq.
More recently, Zhang et al. [10], serve as an extended version to [9], have carried out a
long-term qualitative analysis incorporating incubation time delay into incidence
rate in the case of q=1,
that is, with the force of infection βS(t)I(t−τ),
which was proposed by Cooke [11]. The incubation period τ(τ>0) is a time, during which the infectious agents
develop in the vector, and only after that time the infected vector becomes
itself infectious. The detailed biological meanings and transmission mechanisms
were given in [11].
The results obtained in [10]
represent that, if the epidemic is persistent, introducing time delay
changes the dynamical behaviors of the epidemic state. If the threshold value
determining whether the disease dies out is larger than one and less than
three, the endemic equilibrium is absolutely stable (in the sense that it is
asymptotically stable for all values of the delays [12]); when it exceeds three,
the endemic equilibrium is conditionally stable (i.e., it is asymptotically
stable for the delays in some intervals), and limit cycles arise by Hopf-type
bifurcation with increasing time delay.
In 1978, May and Anderson [13] proposed the saturated
incidence rate of the form β(SI/(1+αS)),
and used by some authors [14–17], recently. The effect of saturation factor (refer to α) stems from epidemical control. In the
absence of effective therapeutic treatment and vaccine, the epidemical control
strategies are based on taking appropriate preventive measures. For example, if
transmission vector is mosquito, these measures include mosquito reduction
mechanisms and personal protection against exposure to mosquitos. Mosquito
reduction mechanisms entail the elimination of mosquito breeding sites (such as
clearing culverts, roadside ditches, eliminating standing water, etc.),
larvaciding (killing of larvae before they become adults) and adulticiding
(killing of adult mosquitoes by spraying). On the other hand, personal
protection is based on preventing vector mosquitoes from biting humans (by
using mosquito repellents, avoiding locations where mosquitoes are biting, and
using barrier methods such as window screens and long-sleeved clothing)
[18–20].
From a practical point of view, instead of the
bilinear incidence rate in [10], we consider saturation incidence rate in this paper
and assume the force of infection is in this version β(SI(t−τ)/(1+αS)) which is saturated with the susceptible. The
susceptible host population is also assumed to have the logistic growth with
carrying capacity K, with a specific growth rate constant r.
We can get a generalized SIR epidemiological model as follows:S˙(t)=r(1−S(t)K)S(t)−βS(t)1+αS(t)I(t−τ),I˙(t)=βS(t)1+αS(t)I(t−τ)−μ1I(t)−γI(t),R˙(t)=γI(t)−μ2R(t),where K,r,α,γ,μ1, and μ2 are positive constants. α is the parameter that measures the inhibitory
effect, γ is the natural recovery rate of the infective
individuals, μ1 and μ2 represent the per capita death rates of
infectives and recovered, respectively. Notice that when α=0,
the system (1.1) becomes the system of bilinear incidence rate in [10], throughout this paper, we
assume α≠0.
By mathematical analysis, we derive a threshold value ℜ0 and prove that the values of ℜ0 and incubation time length completely
determine the global dynamics of system (1.1), that is, this two factors
determine whether the disease approaches an endemic value or whether solutions
oscillate. If ℜ0≤1,
the disease-free equilibrium is globally asymptotically stable and the disease
always dies out, whereas if ℜ0>1,
the disease persists if it is initially present. By taking the incubation time
delay as a bifurcation parameter, the local stability for the endemic
equilibrium is investigated, and the conditions with respect to the system to
be absolutely stable and conditionally stable are derived. Numerical
simulations show that the system with time delay admits rich complex dynamic,
and a sequence of periodic solutions will emanate with increasing time delay,
which exhibits quite complex periodic and chaotic patterns.
We arrange our paper as follows. In Section 2, results
on positivity and boundedness of solutions are presented. In addition, we also
consider the equilibria of system (1.1) and give the threshold for the
existence of endemic equilibrium. In Section 3, we consider the global
stability of the disease-free equilibrium and obtain the necessary and sufficient
conditions for the permanence of endemic equilibrium. The local stability
analysis of system (1.1) is considered in Section 4. Some numerical results
will be given as applications in Section 5.
2. Preliminary Results
The initial conditions ϕ=(ϕ1,ϕ2,ϕ3) of (1.1) are
defined in the Banach spaceC+={ϕ∈C([−τ,0],R+3):ϕ1(θ)=S(θ),ϕ2(θ)=I(θ),ϕ3(θ)=R(θ)},where R+3={(S,I,R)∈R3:S≥0,I≥0,R≥0}.
By a biological meaning, we assume that ϕi(0)>0(i=1,2,3).
It can be verified that the positive cone R+3 is positively invariant with respect to (1.1)
from [21, Lemma 2.1].
Lemma 2.1.
All
feasible solutions of the system (1.1) are bounded and enter the regionΩε={(S,I,R)∈R+3:S+I+R≤(r+1)μmM+ε,∀ε>0},where μm=min{1,μ1,μ2},limsupt→∞S(t)≤M:=max{S(0),K}.
Proof.
From the first
equation of (1.1), we getS˙(t)≤r(1−S(t)K)S(t),by comparison, we have limsupt→∞S(t)≤M. The total host population size N(t) can be determined by N(t)=S(t)+I(t)+R(t),
andN˙(t)=r(1−S(t)K)S(t)−μ1I(t)−μ2R(t)≤(r+1)S(t)−S(t)−μ1I(t)−μ2R(t)≤(r+1)M−μmN.Thus, we have 0≤N≤((r+1)/μm)M, as t→∞.
Therefore, all feasible solutions of the system (1.1) are bounded and enter the
region Ωε.
This completes the proof of Lemma 2.1.
Lemma 2.1 shows that the solutions of system (1.1) are
bounded and, hence, lie in a compact set and are continuable for all positive
time.
Let ℜ0=K[β−α(μ1+γ)]/(μ1+γ).
For system (1.1), there always exists the equilibria E0=(0,0,0), E1=(K,0,0),
if ℜ0>1,
there also exists an endemic equilibrium E+=(S*,I*,R*),
whereS*=μ1+γβ−α(μ1+γ),I*=rS*2K(μ1+γ)(ℜ0−1),R*=γμ2I*.
3. Permanence
Before starting our theorem, we give the following
lemma.
Lemma 3.1 (see[22]).
Consider
the following equation:u˙(t)=au(t−τ)−bu(t),where a,b,τ>0;
and u(t)>0 for −τ≤t≤0.
One has
if a<b,
then limt→∞u(t)=0;
if a>b,
then limt→∞u(t)=+∞.
Theorem 3.2.
If ℜ0≤1,
then the solutions of (1.1), with respect to Ωε for any ε,
satisfy (S(t),I(t),R(t))→(K,0,0) as t→∞.
Proof.
We consider
first the case when ℜ0<1.
From the first equation of (1.1), then there exists a ε>0 such that S(t)<K+ε for some T1>0 when t≥T1. Since S/(1+αS) is increasing function with respect to S,
then from the second equation of (1.1), we haveI˙(t)≤βK+ε1+α(K+ε)I(t−τ)−(μ1+γ)I(t).
Since ℜ0<1, we haveβK+ε1+α(K+ε)<μ1+γ.By Lemma 3.1, we have limsupt→∞I(t)=0.limsupt→∞S(t)=K in terms of S(t)=K is the global attractivor of S(t)=r(1−S(t)/K)S(t).
Next, we shall consider the case when ℜ0=1. Noticing that ℜ0=1 is equal to β(K/(1+αK))=μ1+γ. Since S′(t)≤r(K−S(t))S(t), S(t) is always decreasing when above K.
If S(t) should ever get below K then S(t) must stay strictly below K for all subsequent time. This implies there
are two possible cases, either
S(t)→K from above as t→∞,
or
there exists T such that S(t)<K for all t>T.
In the first of these cases, we have only to show that I(t)→0. Integrating the first equation for S from τ to t+τ in (1.1), we getS(t+τ)−S(τ)=∫τt+τrS(u)(1−S(u)K)du−∫τt+τβS(u)1+αS(u)I(u−τ)du,≤∫τt+τrS(u)(1−S(u)K)du−∫τt+τβK1+αKI(u−τ)du,=∫τt+τrS(u)(1−S(u)K)du−∫τt+τ(μ1+γ)I(u−τ)du.Then,(μ1+γ)∫0tI(u)du≤∫τt+τrS(u)(1−S(u)K)︸<0du−S(t+τ)+S(τ)≤S(τ)≤S(0).Letting t→∞,
we conclude that I(t)∈L1(0,∞) and, therefore, I(t)→0.
In the second of these cases, consider the
functionalV=I(t)+(μ1+γ)∫t−τtI(u)du.Then, for all t>T+τ,V˙(t)|(1.1)=I˙(t)+(μ1+γ)[I(t)−I(t−τ)]=βS(t)1+αS(t)I(t−τ)−(μ1+γ)I(t−τ)=β[S(t)1+αS(t)−K1+αK]I(u−τ)<0.
A direct application of Liapunov-LaSalle type theorem
[22] shows that limt→∞I(t)=0.
By the third equation of (1.1), we get that limt→∞I(t)=0 implies limt→∞R(t)=0. This proves ℜ0≤1 is the sufficient condition for limt→∞(S(t),I(t),R(t))=(K,0,0).
According to (1.1) and the definitions for permanence
in [23], we have the
following lemma.
Lemma 3.3.
Permanence
of S(t),I(t) in system (1.1) implies that of R(t).
Next, we represent our main results in this section.
Theorem 3.4.
System (1.1) is permanent if it satisfies ℜ0>1.
In order to prove Theorem 3.4, we present uniform
persistence theory for infinite dimensional systems from [24]. Let X be a complete metric
space. Suppose that X0 is open and dense in X and X0∪X0=X,X0∩X0=∅.
Assume that T(x) is a C0 semigroup on X satisfyingT(t):X0⟶X0,T(t):X0⟶X0.Let Tb(t)=T(t)|X0 and let Ab be the global attractor for Tb(t).
Lemma 3.5 (see[24]).
Suppose
that T(t) satisfies (3.8) and one has the following:
there
is a t0≥0 such that T(t) is compact for t>t0;
T(t) is point dissipative in X;
Ab=⋃x∈Abω(x) is isolated and thus has an acyclic covering M^, where M^={M1,M2,…,Mn};
Ws(Mi)∩X0=∅ for i=1,2,…,n.
Then X0 is a uniform repeller with respect to X0,
that is, there is an ε>0 such that for any x∈X0, liminft→+∞d(T(t)x,X0)≥ε,
where d is the distance of T(t)x from X0.
Proof.
We first prove that ℜ0>1 leads to the permanence of system (1.1).
By Lemma 3.3, we only need to consider the following
subsystem of (1.1) and prove that (S(t),I(t)) in system (3.10) are permanent if and only if ℜ0>1 holdsS˙(t)=r(1−S(t)K)S(t)−βS(t)1+αS(t)I(t−τ),I˙(t)=βS(t)1+αS(t)I(t−τ)−μ1I(t)−γI(t),where S(θ),I(θ)≥0 are continuous on −τ≤θ≤0,
and S(0),I(0)>0.
We begin by showing that the boundary planes of ℝ+2={(x,y):x≥0,y≥0} repel the positive solutions to system (3.10)
uniformly. Let C+([−τ,0],ℝ+2) denote the space of continuous functions
mapping [−τ,0] into ℝ+2.
We chooseC1={(φ1,φ2)∈C+([−τ,0],ℝ+2):φ1(θ)≡0,θ∈[−τ,0]},C2={(φ1,φ2)∈C+([−τ,0],ℝ+2):φ1(θ)>0,φ2(θ)≡0,θ∈[−τ,0]}.
Denote C0=C1∪C2,X=C+([−τ,0],ℝ+2) and C0= Int C+([−τ,0],ℝ+2).
Next, we verify that the conditions of Lemma 3.5 are
satisfied. By the definition of C0 and C0 and system (3.10), it is easy to see that C0 and C0 are positively invariant. Moreover, conditions
(i) and (ii) of Lemma 3.5 are clearly satisfied. Thus, we only need to verify
conditions (iii) and (iv). Since system (3.10)
possesses two constant solutions in C0:E0∈C1,E1∈C2 withE0={(φ1,φ2)∈C+([−τ,0],ℝ+2):φ1(θ)≡φ2(θ)≡0,θ∈[−τ,0]},E1={(φ1,φ2)∈C+([−τ,0],ℝ+2):φ1(θ)≡K,φ2(θ)≡0,θ∈[−τ,0]},and we have S˙(t)|(φ1,φ2)∈C1≡0,
then we get S(t)|(φ1,φ2)∈C1≡0 for all t≥0,
according to the second equation of (3.10), we have I˙(t)|(φ1,φ2)∈C1=−(μ1+γ)I(t)≤0,
hence all points in C1 approach E0,
C1=Ws(E0).
Similarly, we have all points in C2 approach E1,
that is, C2=Ws(E1).
This shows that invariant sets E0 and E1 are isolated invariant, then {E0,E1} is isolated and is an acyclic covering,
satisfying condition (iii) of Lemma 3.5.
Now, we show that Ws(Ei)∩C0=∅,i=0,1. We only need to prove Ws(E1)∩C0=∅,
since the proof for Ws(E0)∩C0=∅ is simple.
Assume the contrary, that is, Ws(E1)∩C0≠∅,
then there exists a positive solution (S(t),I(t)) to system (3.10) with limt→∞(S(t),I(t))=(K,0). Since ℜ0>1,
then for a sufficiently small ε>0 with μ1+γ<β((K−ε)/(1+α(K−ε))),
there exists a positive constant T=T(ε) such that S(t)>K−ε>0,0<I(t)<ε∀t≥T.
By the second
equation of (3.10) and noting that β(S/(1+αS)) is an increasing function with respect to S,
then we haveI˙(t)≥βK−ε1+α(K−ε)I(t−τ)−(μ1+γ)I(t),t≥T+τ.According to the comparison
principle, limt→∞I(t)=∞ when ℜ0>1,
contradicting I(t)<ε.
Then we have Ws(E1)∩C0=∅.
At this time, we are able to conclude from Lemma 3.5 that C0 repels the positive solutions of (3.10)
uniformly. Incorporating the above results into Lemmas 3.3 and 3.5, we know
that system (1.1) is permanent.
Next, we verify that permanence of system (1.1)
indicates ℜ0>1.
Assume that the contrary holds, that is, ℜ0≤1,
then by Theorem 3.2, (S(t),I(t),R(t))→(K,0,0), as t→∞, contradicting permanence of system (1.1). This
proves Theorem 3.4.
4. Linearized Analysis
The dynamics of model (1.1) are determined by the
first two equations. Therefore, throughout the remainder of this paper, we
consider the subsystem (3.10), and rewrite it as follows:S˙(t)=r(1−S(t)K)S(t)−βS(t)1+αS(t)I(t−τ),I˙(t)=βS(t)1+αS(t)I(t−τ)−μ1I(t)−γI(t).
Let E^=(S^,I^) be any equilibrium of (4.1), linearized system
of (4.1) at E^=(S^,I^),
we getx˙(t)=[r−2rKS^−βI^(1+αS^)2]x(t)−βS^1+αS^y(t−τ),y˙(t)=βI^(1+αS^)2x(t)+βS^1+αS^y(t−τ)−(μ1+γ)y(t).Then the characteristic equation
of (4.1) at E^ is given bydet[r−2rKS^−βI^(1+αS^)2−λ−βS^1+αS^e−λτβI^(1+αS^)2βS^1+αS^e−λτ−(μ1+γ)−λ]=0.At the equilibrium E^0=(0,0),
characteristic equation (4.3) reduces to(λ−r)(λ+μ1+γ)=0.
Obviously,
(4.4) has a positive root λ=r independent of any parameters. Hence, E^0 is always a unstable saddle point.
Theorem 4.1.
For the
system (4.1), the equilibrium E^1=(K,0) is
asymptotic stable if ℜ0<1;
linearly
neutrally stable if ℜ0=1;
unstable
if ℜ0>1.
Proof.
The characteristic equation at E^1 is(λ+r)[λ+(μ1+γ)(1−ℜ0+αK1+αKe−λτ)]=0.
Equation (4.5) has a negative real part characteristic
root λ=−r and roots ofF(λ)=λ+(μ1+γ)(1−ℜ0+αK1+αKe−λτ)=0.
(i) Assume that ℜ0<1,
(4.6) has characteristic root λ=(μ1+γ)((ℜ0+αK)/(1+αK)−1)<0 when τ=0.
If λ=iω is a root of (4.6), it must
satisfyω2=(μ1+γ)2[(ℜ0+αK1+αK)2−1].When ℜ0<1,
there are no positive real roots ω.
This shows that all roots of F(λ)=0 must have negative real parts, therefore, E^1 is an asymptotically stable equilibrium.
(ii) Assume that ℜ0=1,
then λ=0 is a root of (4.6). It is easy to verify that λ=0 is a simple characteristic root. If the other
roots are λ=α+iω,
then they must satisfy[α+(μ1+γ)]2+ω2=(μ1+γ)2e−2ατ,and we must have α≤0.
Therefore E^1 is linearly neutrally stable.
(iii) Assume that ℜ0>1,
then F(0)<0,
and F(+∞)=+∞.
Hence, F(λ) has at least one positive root and E^1 is unstable.
By the arguments to Theorems 3.2 and 4.1, we
directly have the following corollary.
Corollary 4.2.
The equilibrium E1=(K,0,0) of system (1.1) is global asymptotically
stable if ℜ0≤1 holds true in the feasible region Ωε for any ε>0.
In the following, we will study the linear stability
of the positive equilibrium E^+=(S*,I*) of (4.1). We can see that the characteristic
roots of (4.3) at positive equilibrium E^+ are the roots ofdet[r(1−2ℜ0)−r1+αS^(1−1ℜ0)−λ−(μ1+γ)e−λτr1+αS^(1−1ℜ0)(μ1+γ)e−λτ−(μ1+γ)−λ]=0.
Since β(I*/(1+αS*))=r(1−1/ℜ0) and β(S*/(1+αS*))=μ1+γ at (S*,I*),
we haveP(λ,τ)+Q(λ,τ)e−λτ=0,whereP(λ,τ)=λ2+λ[−r(1−2ℜ0)+(μ1+γ)+r1+αS*(1−1ℜ0)]+(μ1+γ)[−r(1−2ℜ0)+r1+αS*(1−1ℜ0)],Q(λ,τ)=−λ(μ1+γ)+r(μ1+γ)(1−2ℜ0).
When τ=0,
the DDE (4.1) becomes ODE which has the same equilibria E^ as follows:S˙(t)=r(1−S(t)K)S(t)−βS(t)1+αS(t)I(t),I˙(t)=βS(t)1+αS(t)I(t)−μ1I(t)−γI(t),and (4.10)
becomesλ2+λ[−r(1−2ℜ0)+r1+αS*(1−1ℜ0)]+(μ1+γ)r1+αS*(1−1ℜ0)=0.
Define ℜcc=2+1/αS*.
If −r(1−2/ℜ0)+(r/(1+αS*))(1−1/ℜ0)>0,
that is, 1<ℜ0<ℜcc=2+1/αS*,
the system (4.12) is locally asymptotically stable. If ℜ0>ℜcc,
the unique positive equilibrium of (4.12) is unstable, system (4.12) becomes
oscillatory in a stable limit cycle [25], and this limit cycle is unique [26, 27].
If λ=iω(ω>0) is a root of (4.10), then by separating the
real and imaginary parts, we get−ω2+(μ1+γ)[−r(1−2ℜ0)+r1+αS*(1−1ℜ0)]=−r(1−2ℜ0)(μ1+γ)cosωτ+ω(μ1+γ)sinωτ,ω[−r(1−2ℜ0)+(μ1+γ)+r1+αS*(1−1ℜ0)]=ω(μ1+γ)cosωτ+r(1−2ℜ0)(μ1+γ)sinωτ.
Squaring and
adding both equations, then we have ω4+ω2[−r(1−2ℜ0)+r1+αS*(1−1ℜ0)]2+r1+αS*(1−1ℜ0)(μ1+γ)2[−2r(1−2ℜ0)+r1+αS*(1−1ℜ0)]=0.
Define ℜc=2+1/(1+2αS*).
If −2r(1−2/ℜ0)+(r/(1+αS*))(1−1/ℜ0)≥0,
that is, 1<ℜ0≤ℜc,
there is no positive real ω satisfying (4.15), thus eigenvalues of (4.10)
do not approach the imaginary axis for any τ>0.
This shows that E^+ is absolutely stable when 1<ℜ0≤ℜc.
If ℜ0>ℜc,
there is a unique positive ω0 satisfying (4.15). That is, (4.10) has a unique
pair of purely imaginary roots ±iω0.
From (4.14), τn corresponding to ω0 can be obtained as follows:τn=1ω0arccos{ω02ℤ−ℂ(μ1+γ)r(1−2/ℜ0)ω02(μ1+γ)+[r(1−2/ℜ0)]2(μ1+γ)}+2nπω0,n=0,1,2,…,
where ℤ denotes [(r/(1+αS*))(1−1/ℜ0)+(μ1+γ)] and ℂ denotes [−r(1−2/ℜ0)+(r/(1+αS*))(1−1/ℜ0)].
Further,
dRe(λ)dτ|λ=iω0=Re(dλdτ)−1|λ=iω0=Re{λ2−(μ1+γ)[−r(1−2/ℜ0)+(r/(1+αS*))(1−1/ℜ0)]−λ2P(λ,τ)}λ=iω0+Re[−r(1−2/ℜ0)(μ1+γ)λ2Q(λ,τ)]λ=iω0=ω04−(μ1+γ)2[−r(1−2/ℜ0)+(r/(1+αS*))(1−1/ℜ0)]2+[r(1−2/ℜ0)(μ1+γ)]2(μ1+γ)2+[r(1−2/ℜ0)(μ1+γ)]2=ω04−(μ1+γ)2(r/(1+αS*))(1−1/ℜ0)[−2r(1−2/ℜ0)+(r/(1+αS*))(1−1/ℜ0)](μ1+γ)2+[r(1−2/ℜ0)(μ1+γ)]2.
Under the
condition ℜ0>ℜc,
that is, −2r(1−2/ℜ0)+(r/(1+αS*))(1−1/ℜ0)<0,
then we have
dRe(λ(τ))/dτ∣λ=iω0>0.
If ℜc<ℜ0<ℜcc,
there exists a critical value τ0, when τ<τ0,E^+ is stable; when τ>τ0,E^+ is unstable.
Summarizing the discussion above, we have the
following conclusion.
Theorem 4.3.
For system (4.1), one has
if 1<ℜ0≤ℜc holds true, then E^+ is absolutely stable;
if ℜc<ℜ0<ℜcc holds true, then E^+ is conditionally stable, that is, there is a
critical delay value τ0 such that E^+ is asymptotically stable when τ∈[0,τ0) and unstable when τ>τ0.
Furthermore, system (4.1) undergoes Hopf bifurcation at E^+ when τ=τn,n=0,1,2,…;
if ℜ0>ℜcc,
then there is also a critical delay value τ0 such that the periodic solution is still
stable when τ∈[0,τ0),
however, there are a sequence of periodic solutions emanate when τ=τn,n=0,1,2,….
These results are summarized in Table 1.
Compare DDE (4.1) with ODE (4.12).
Case
0<ℜ0≤1
1<ℜ0≤ℜc
ℜc<ℜ0<ℜcc
ℜ0>ℜcc
ODE
E^1 GAS
E^+ LAS
E^+ LAS
stable periodic solution
DDE
E^1 GAS
E^+ ALASa
E^+ CLASb
complex dynamic phenomena
aabsolutely stable
bconditionally stable
Remark 4.4.
In fact, in the case of ℜ0>ℜcc,
we have known that ODE (4.12) has a stable periodic solution [25]. If we consider the impact
of the incubation time on ODE (4.12), that is, DDE (4.1), from above discussion,
we can see that there are a sequence of periodic solutions bifurcate from the
positive equilibrium E^+ when the time delay takes the critical delay τn such that previous stable periodic solution
losses stability, which will lead to complex dynamic phenomena. This can be
seen from the simulation results Section 5.
In addition, we want to mention that Theorem 4.3(ii)
cannot determine the direction and stability of bifurcation periodic solutions,
this can be done by analyzing the high-order terms in terms of [28]. The method is very complex
and trivial, here we omit it.
From the point of biology, in comparison with the
results of [10], we
see that if there is the inhibition effect from the behavioral change of the
susceptible individuals when the infective increases (i.e., we take use of the
saturation incidence rate), then the threshold ℜc decline and less than 3. It is crucial for the
government to take the corresponding control measures and policies against the
disease when the epidemic outbreaks.
5. Numerical Results
The main goal of the previous section was to
qualitatively characterize the dynamic behaviors of system (1.1) at long term.
In this section, we confirm our previous theoretical analysis in Section 4 and
demonstrate that the local behaviors in the regions of the parameters space
correspond to complex population dynamics to system (1.1). The objective is to
explore the possibility of chaotic behavior in system (1.1). It is difficult to
test whether there exists chaos in a time-delayed system, but numerical
simulation analysis is a valid method for such a system. Extensive numerical
simulations are carried out for different values of saturation parameter α and recover rate γ.
The quality results are as follows.
First, we consider the property of system (1.1) in the
regions of the parameter space corresponds to complex population dynamics in
the case of ℜ0>ℜcc.
To illustrate the transition from the periodic pattern to chaotic pattern, we
concentrate on the regions of small and large time delay as an example. We
consider the set of parameter values as r=0.1, K=80, β=0.1, μ1=0.5, μ2=0.1, α=0.05, and γ=0.5.
By calculating, there exists the relation of ℜ0>ℜcc (ℜ0=4.00, ℜcc=3.00) for system (1.1). Then, from Theorem
4.3(iii), the system (1.1) has a stable period solution if the time delay is
less than the critical delay τ0≐0.23 (see Figure 1), and the periodic solution will
lost stability when the time delay is greater than the critical delay τ0≐0.23,
and then a typical chaos was observed with increasing the time delay (see
Figure 2). This phenomenon has been verified by the bifurcation diagram via
delay τ,
as shown in Figure 3.
Temporal behavior of the
infected and corresponding three-dimensional phase are plotted for the system
(1.1) subject to ℜ0>ℜcc (ℜ0=4.00, ℜcc=3.00). The parameters are r=0.1, K=80, β=0.1, μ1=0.5, μ2=0.1, α=0.05, γ=0.5, and τ=0.001 with initial value (28, 3, 7).
The parameters like in Figure 1 but τ=30.
Bifurcation diagram of system (1.1):
successive maxima of the infected are plotted for increasing values of the time
delay τ,
with parameters r=0.1, K=80, β=0.1, μ1=0.5, μ2=0.1, α=0.05, and γ=0.5.
We increase the recovery rate γ,
let γ=0.75,
and fix the other parameters as above. By calculating, the system (1.1)
satisfies ℜc<ℜ0<ℜcc (ℜc=2.23, ℜ0=2.40, ℜcc=2.6). In this context, by Theorem 4.3(ii), system
(1.1) is conditionally stable at unique positive equilibrium E+=(33.33,1.56,11.67),
that is, there exists a critical delay τ0≐0.31 such that E+ is stable if τ<τ0 (see Figure 4), E+ will lose stability by an Hopf bifurcation if τ>τ0,
as shown in Figure 5. We find that the periodic solution, quasiperiod, and
chaos patterns emerge with increasing time delay. This may be clear from the
bifurcation diagram (see Figure 7).
Temporal behavior of
the infected and corresponding three-dimensional phase are plotted for the
system (1.1) subject to 1<ℜc<ℜ0<ℜcc (ℜc=2.23, ℜ0=2.40, ℜcc=2.6). The parameters are r=0.1, K=80, β=0.1, μ1=0.5, μ2=0.1, α=0.05, γ=0.75, and τ=0.001 with initial value (35, 10, 15).
Temporal behavior of the
infected and corresponding three-dimensional phase are plotted for the system
(1.1) subject to 1<ℜc<ℜ0<ℜcc (ℜc=2.23, ℜ0=2.40, ℜcc=2.6). The parameters are r=0.1, K=80, β=0.1, μ1=0.5, μ2=0.1, α=0.05, γ=0.75,
and τ=0.5 with initial value (40, 3, 7).
The parameters like in
Figure 5 but τ=27.
Bifurcation diagram of system (1.1): successive maxima of the infected
are plotted for increasing time delay τ,
with parameters r=0.1, K=80, β=0.1, μ1=0.5, μ2=0.1, α=0.05m, and γ=0.75.
Now, we fix the parameter γ=0.75,
change α, and let α=0.055 (i.e., we take some measures to protect on
susceptibles), and the other parameters are also as above. By calculating,
system (1.1) always exists a relationship 1<ℜ0<ℜc (ℜ0=2.00, ℜc=2.19). By Theorem 4.3(i), we know that system
(1.1) is absolutely stable at unique positive equilibrium E+=(40,1.60,12) for any value of the time delay, as shown in
Figure 8. Our numerical simulations have demonstrated the validity of our
theoretical analysis, that is, the values of threshold value ℜ0 and incubation time τ length completely determine the dynamics of
system (1.1).
Temporal
behavior of the infected and corresponding three-dimensional phase are plotted
for the system (1.1) subject to 1<ℜ0<ℜc.
The parameters like in Figure 5 but α=0.055, γ=0.75,
and τ=1.5 with initial value (25, 4, 20).
It is necessary to indicate that the system (1.1) is
realistic at the initial phase of the disease emergence since the number of the
infected is rare. However, when the number of the infected is large, it is more
reasonable that one should replace the term r(1−S/K)S in the first equation of (1.1) by r(1−(S+I+R)/K)S.
Hence, a profound understanding for this case is still desirable and could
motivate further investigations.
Acknowledgments
The authors are grateful to Associate Editor Manuel de
la Sen and anonymous referee for their valuable comments and suggestions that
greatly improved the original version of this paper. This work is supported by
the National Sciences Foundation of China (60771026), Science Foundations of
Shanxi Province (2007011019), the Special Scientific Research Foundation
for the Subjects of Doctors in University
(20060110005), and supported by Program for New Century Excellent Talents in
University (NCET050271).
LiuW. M.LevinS. A.IwasaY.Influence of nonlinear incidence rates upon the behavior of SIRS epidemiological models1986232187204MR82913210.1007/BF00276956ZBL0582.92023LiuW. M.HethcoteH.W.LevinS. A.Dynamical behavior of epidemiological models with nonlinear incidence rates1987254359380MR90837910.1007/BF00277162ZBL0621.92014HethcoteH. W.van den DriesscheP.Some epidemiological models with nonlinear incidence1991293271287MR108978610.1007/BF00160539ZBL0722.92015MoghadasS. M.AlexanderM. E.Bifurcations of an epidemic model with nonlinear incidence and infection-dependent removal rate200623323125410.1093/imammb/dql011ZBL1098.92058KorobeinikovA.MainiP. K.Nonlinear incidence and stability of infectious disease models200522211312810.1093/imammb/dqi001ZBL1076.92048CapassoV.SerioG.A generalization of the Kermack-McKendrick deterministic epidemic model1978421-24361MR52909710.1016/0025-5564(78)90006-8ZBL0398.92026RuanS.WangW.Dynamical behavior of an epidemic model with a nonlinear incidence rate20031881135163MR195451110.1016/S0022-0396(02)00089-XZBL1028.34046XiaoD.RuanS.Global analysis of an epidemic model with nonmonotone incidence rate20072082419429MR234812310.1016/j.mbs.2006.09.025ZBL1119.92042ZhangX.-A.ChenL.The periodic solution of a class of epidemic models1999383-46171MR1703408ZBL0939.9203110.1016/S0898-1221(99)00206-0ZhangJ.JinZ.WangJ.Analysis of an SIR model with bilinear incidence ratesubmittedCookeK. L.Stability analysis for a vector disease model1979913142MR517971ZBL0423.92029RuanS.Absolute stability, conditional stability and bifurcation in Kolmogorov-type predator-prey systems with discrete delays2001591159173MR1811101ZBL1035.34084AndersonR. M.MayR. M.Regulation and stability of host-parasite population interactions: I. Regulatory processes197847121926710.2307/3933GaoS.ChenL.NietoJ. J.TorresA.Analysis of a delayed epidemic model with pulse vaccination and saturation incidence20062435-366037604510.1016/j.vaccine.2006.05.018ZhangT.TengZ.Pulse vaccination delayed SEIRS epidemic model with saturation incidence20083271403141610.1016/j.apm.2007.06.005MR2409386MengX.ChenL.ChengH.Two profitless delays for the SEIRS epidemic disease model with nonlinear incidence and pulse vaccination20071861516529MR231451110.1016/j.amc.2006.07.124ZBL1111.92049WeiC.ChenL.A delayed epidemic model with pulse vaccination2008200812746951MR240319710.1155/2008/746951BowmanC.GumelA. B.van den DriesscheP.WuJ.ZhuH.A mathematical model for assessing control strategies against West Nile virus200567511071133MR221689410.1016/j.bulm.2005.01.002NosalB.PellizzariR.West Nile virus20031681114431444PetersenL. R.MarfinA. A.GublerD. J.West Nile virus2003290452452810.1001/jama.290.4.524YangX.ChenL.ChenJ.Permanence and positive periodic solution for the single-species nonautonomous delay diffusive models1996324109116MR1403462ZBL0873.3406110.1016/0898-1221(96)00129-0KuangY.1993191Boston, Mass, USAAcademic Pressxii+398Mathematics in Science and EngineeringMR1218880ZBL0777.34002XiaoY.ChenL.Modeling and analysis of a predator-prey model with disease in the prey200117115982MR183920910.1016/S0025-5564(01)00049-9ZBL0978.92031HaleJ. K.WaltmanP.Persistence in infinite-dimensional systems1989202388395MR98266610.1137/0520025ZBL0692.34053WaltmanP.198345Philadelphia, Pa, USASIAMv+77CBMS-NSF Regional Conference Series in Applied MathematicsMR778562ZBL0572.92019ChengK. S.Uniqueness of a limit cycle for a predator-prey system1981124541548MR61771310.1137/0512047ZBL0471.92021KuangY.FreedmanH. I.Uniqueness of limit cycles in Gause-type models of predator-prey systems19888816784MR93000310.1016/0025-5564(88)90049-1ZBL0642.92016HassardB. D.KazarinoffN. D.WanY. H.198141Cambridge, UKCambridge University Pressv+311 ppLondon Mathematical Society Lecture Note SeriesMR603442ZBL0474.34002