• No results found

Stochastic inequalities and applications to dynamics analysis of a novel SIVS epidemic model with jumps

N/A
N/A
Protected

Academic year: 2020

Share "Stochastic inequalities and applications to dynamics analysis of a novel SIVS epidemic model with jumps"

Copied!
25
0
0

Loading.... (view fulltext now)

Full text

(1)

R E S E A R C H

Open Access

Stochastic inequalities and applications to

dynamics analysis of a novel SIVS epidemic

model with jumps

Xiaona Leng

1

, Tao Feng

1

and Xinzhu Meng

1,2*

*Correspondence: [email protected] 1College of Mathematics and Systems Science, Shandong University of Science and Technology, Qingdao, 266590, P.R. China

2State Key Laboratory of Mining Disaster Prevention and Control Co-founded by Shandong Province and the Ministry of Science and Technology, Shandong University of Science and Technology, Qingdao, 266590, P.R. China

Abstract

This paper proposes a new nonlinear stochastic SIVS epidemic model with double epidemic hypothesis and Lévy jumps. The main purpose of this paper is to investigate the threshold dynamics of the stochastic SIVS epidemic model. By using the

technique of a series of stochastic inequalities, we obtain sufficient conditions for the persistence in mean and extinction of the stochastic system and the threshold which governs the extinction and the spread of the epidemic diseases. Finally, this paper describes the results of numerical simulations investigating the dynamical effects of stochastic disturbance. Our results significantly improve and generalize the

corresponding results in recent literatures. The developed theoretical methods and stochastic inequalities technique can be used to investigate the high-dimensional nonlinear stochastic differential systems.

Keywords: stochastic SIVS epidemic model; Lévy jumps; persistence in mean; double epidemic diseases; Doob’s martingale inequality; Hölder’s inequality

1 Introduction

Mathematical inequalities are widely used in many fields of mathematical analysis, es-pecially differential systems [–]. Recently, the inequality technique was applied to stochastic differential systems [–], impulsive differential systems [–], and impul-sive stochastic differential systems [], thus some new results have been obtained.

As an important factor threatening the safety of human life and property, the investiga-tion of epidemic has received extensive atteninvestiga-tion from experts in various fields [–]. Generally speaking, medical researchers often use observation and experimental meth-ods to study the behavior of epidemics. Recently, however, a number of experts in the field of mathematics have also been interested in the study of epidemics. They have used mathematical methods to analyze the spread and control of epidemics [–]. Kermack and McKendrick’s pioneering work on the development of an epidemic disease is one of the typical examples. They established an SIS compartment model and proposed the fa-mous threshold theory, which has laid a solid foundation for the study of the dynamics of infectious diseases [].

(2)

The SIS model based on the deterministic ordinary differential equation is given by

⎧ ⎨ ⎩ ˙

S(t) =AβS(t)I(t) –uS(t) +rI(t), ˙

I(t) =βS(t)I(t) – (u+α+r)I(t). ()

In system (),βS(t) represents the number of people infected by a patient within a unit time att. But in reality, the number of people who can be exposed to a patient at a time is limited. To this end, some authors have introduced a saturated infection rate to study the dynamic behavior of the disease [–]. In addition, all creatures on the earth are infected by a variety of environmental noises, of course, the disease is no exception. Motivated by this, some scholars have studied the infection system with environmental noises (such as Brownion noise, Markov noise and Lévy noise) [–]. Meanwhile, populations may be affected by different kinds of infectious diseases at the same time. Therefore, it is of great significance to study the epidemic model with multiple diseases [–].

Recently, Meng et al. [] considered a novel nonlinear stochastic SIS epidemic model with double epidemic hypothesis as follows:

⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩

dS= (AuS(t) –βS(t)I(t)

a+I(t) –

βS(t)I(t)

a+I(t) +rI(t) +rI(t))dt

σS(t)I(t)

a+I(t) dB(t) –

σS(t)I(t)

a+I(t) dB(t),

dI= (βaS+(tI)I(t()t)– (u+α+r)I(t))dt+σaS(+tI)I(t()t)dB(t),

dI= (βaS+(tI)I(t()t)– (u+α+r)I(t))dt+σaS(+tI)I(t()t)dB(t).

()

They obtained the threshold of system () for the extinction and the persistence in mean of the epidemic diseases. Based on system (), recently, Zhang et al. [] proposed an SIS system with double epidemic diseases driven by Lévy jumps as follows:

⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩

dS= (AuS(t) –βS(t)I(t)

a+I(t) –

βS(t)I(t)

a+I(t) +rI(t) +rI(t))dt

+σS(t)dB(t) +

Zγ(u)S(t–)N(dt,du),

dI= (βaS+(tI)I(t()t)– (u+α+r)I(t))dt +σI(t)dB(t) +

Zγ(u)I(t–)N(dt,du),

dI= (βaS(+tI)I(t()t)– (u+α+r)I(t))dt +σI(t)dB(t) +

Zγ(u)I(t–)N(dt,du).

()

(3)

Motivated by the above works, in this paper, we propose a stochastic SIVS model with double epidemic diseases and Lévy jumps under vaccination as follows:

⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩

dS= (( –q)– (u+p)S(t) – βS(t)I(t)

α+I(t) –

βS(t)I(t)

α+I(t)

+rI(t) +rI(t) +δV(t))dt +σS dB(t) +

Zγ(u)S(t–)N(dt,du),

dI= (βαS+(tI)I(t()t)– (u+d+r)I(t))dt

+σIdB(t) +

Zγ(u)I(t–)N(dt,du),

dI= (βαS(+tI)I(t()t)– (u+d+r)I(t))dt

+σIdB(t) +

Zγ(u)I(t–)N(dt,du),

dV= [q+pS(t) – (u+δ)V(t)]dt+σV dB(t) +Zγ(u)V(t–)N(dt,du),

()

whereS(t),I(t),I(t),V(t), respectively, stand for the density of susceptible, infectiveA, infectiveBand vaccinated individuals at timet,is a constant input of new numbers into the population,qmeans a fraction of vaccinated for the newborn,βiis the infection rate

coefficient fromIi(t) (i= , ) toS(t), respectively.urepresents the natural death rate of

S(t),I(t),I(t),V(t),pis the proportional coefficient of vaccinated for the susceptible,ri,

diis the recovery rate and disease-caused death rate ofIi(t),i= , , respectively.δstands

for the rate of losing their immunity for vaccinated individuals,αandαare the so-called half-saturation constants, respectively.B(t) = (B(t),B(t),B(t),B(t)) is a standard Brow-nian motion with intensityσi>  (i= , , , ).

Throughout this paper, let (,F,{F}t≥,P) be a complete probability space with a

fil-tration{Ft}t≥ satisfying the usual conditions (i.e. it is increasing and right continuous

while F contains all P-null sets). Function Bi(t) (i= , , , ) is a Brownian motion

defined on the complete probability space , the intensity ofBi(t) is σi (i= , , , ).

N(dt,du) =N(dt,du) –λ(du)dt,N is a Poisson counting measure on (, +∞)×Z,λis the characteristic measure ofNon a measurable subsetZ,λ(Z) < +∞,γi(i= , , , ) is

bounded and continuous with respect toλand isB(Z)×Ft-measurable. For an integrable

functionX(t) on [, +∞), we defineX(t)=ttX(s)ds.

(4)

2 Main results

The main purpose of this paper is to investigate the threshold dynamics of the stochas-tic SIVS epidemic model. In this section, by using the technique of a series of stochasstochas-tic inequalities, we obtain sufficient conditions for the persistence in mean and extinction of the stochastic system and the threshold which governs the extinction and the spread of epidemic diseases.

2.1 Preliminary knowledge

For the sake of notational simplicity, we define

bi=

 σ

i +

Z

γi(u) –ln

 +γi(u)

λ(du), i= , , , ;

Ri=

βi(u+δuq)

u++upαi(u+di+ri+bi), i= , ; ˇ

γ(u) =maxγ(u),γ(u),γ(u),γ(u)

;

ˆ

γ(u) =minγ(u),γ(u),γ(u),γ(u)

;

φ=

Z  +γˇ(u)

–  –γˆ(u)v(du);

σ=max{σ,σ,σ,σ}.

Throughout this paper, suppose that the following two assumptions hold.

Assumption . The following hold:

(i)  +γi(u) > ;

(ii) Z[γi(u) –ln( +γi(u))]λ(du) <∞,i= , , , ,u∈Z.

Remark . This assumption means that the intensities of Lévy noises are not infi-nite.

Assumption . Suppose that there exists some>  such that the following inequality holds:

b=u– 

σ

φ

> .

Definition .([])

(i) The speciesX(t)is said to be extinctive iflimt→+∞X(t) = ;

(ii) The speciesX(t)is said to be persistent in mean iflimt→+∞X(t)∗> .

The following elementary inequality will be used frequently in the sequel.

Lemma .(Burkholder-Davis-Gundy inequality []) Let gL(R

+;Rd×m).For any t≥ ,define

x(t) =

t

g(s)dB(s), A(t) =

t

(5)

Then,for every p> ,there exist two positive constants cp,Cpsuch that

cpEA(t) p/

E sup

≤st|x(s)|p

CpEA(t) p/

, t≥,

where cp,Cponly depend on p.

Lemma .(Chebyshev inequality []) For any c> ,p> ,XLp, the following

in-equality holds:

Pw:X(w)≥ccpE|X|p.

Lemma .(Hölder inequality []) For any ai,biR and k≥,if p,q> andp+q= ,

the following inequality holds:

k

i=

aibi

k

i= |ai|p

/pk

i= |bi|q

/q

.

Lemma .(Doob’s martingale inequality []) Let X be a submartingale taking nonneg-ative real values,either in discrete or continuous time.That is,for all times s and t with s<t,

XsE[Xt|Fs].

Then,for any constant C> ,

P sup

≤tT

XtC

E[|XT|]

C ,

where P denotes the probability measure on the sample spaceof the stochastic process X: [,T→[, +∞)and E denotes the expected value with respect to the probability measure P.

Lemma .([, ]) Assume that X(t)∈R+is an Itô’s-Lévy process of the form

dX(t) =FXt–,tdt+GXt–,tdB(t) +

ZH

Xt–,t–,uN(dt,du),

where F:Rn×R

SRn,G:Rn×RSRnand H:Rn×RS×ZRnare

measurable functions.

Given VC,(Rn×R

S;R+),we define the operator LV by

LV(X,t) =Vt(X,t) +VX(X,t)F(X,t) +

 trace G

T(X,t)V

XX(X,t)G(X,t)

+

Z

VX+H(X,t)–V(X,t) –VX(X,t)H(X,t,u)

λ(du),

where

Vt(X,t) =

∂VX(X,t)

(6)

VX(X,t) =

∂VX(X,t)

∂X

, . . . ,∂VX(X,t)

∂Xn

,

VXX(X,t) =

VX(X,t)

∂Xi∂Xj

n×n

.

Then the generalized Itô’s formula with Lévy jumps is given by

dV(X,t) =LV(X,t)dt+VX(X,t)G(X,t)dB(t) +

Z

VX+H(X,t)–V(X,t)N(dt,du).

Lemma .([]) Let X(t)∈C(×[, +∞),R+).We have the following conclusions. (i) If there existT> ,λ> ,λ,m,nisuch that whentT,

lnX(t)≤λtλ

t

X(s)ds+mB(t) +

j

i=

ni

t

Z

ln +γi(u)(ds,du) a.s.,

then

⎧ ⎨ ⎩

X∗≤ λ

λa.s., ifλ≥;

limt→+∞X(t) =  a.s., ifλ< .

(ii) If there existT> ,λ> ,λ> ,m,nisuch that whentT,

lnX(t)≥λtλ

t

X(s)ds+mB(t) +

j

i=

ni

t

Z

ln +γi(u)(ds,du) a.s.,

thenX∗≥λλa.s.

Lemma . For any initial value(S(),I(),I(),V())∈R+,the solution(S(t),I(t),I(t),

V(t))of model()has the following property:

lim

t→∞

S(t) +I(t) +I(t) +V(t)

t =  a.s.

Moreover,

lim

t→∞

S(t)

t = , tlim→∞

I(t)

t = ,

lim

t→∞

I(t)

t = , tlim→∞

V(t)

t =  a.s.

lim

t→∞

lnS(t)

t ≤, tlim→∞

lnI(t)

t ≤,

lim

t→∞

lnI(t)

t ≤, tlim→∞

lnV(t)

t ≤ a.s.

Proof Define

(7)

Applying the generalized Itô’s formula toQ(X), we have

dQ(X)≤LQdt+X–σIdB(t) +σIdB(t) +σS dB(t) +σV dB(t)

+X

Z  +γˇ(u)

γˆN(dt,du), ()

where

LQX–(uXdI–dI) +

(– )

X

–σX+φX

X–

X

u– 

σ

φ

X

.

Choose a positive constant>  that satisfies

b=u– 

σ

φ

> .

For any constantksatisfyingk∈(,b), one has

dektQX(t)≤L ektQX(t)dt+ektX– σI(s)dB(s) +σI(s)dB(s) +σS(s)dB(s) +σV(s)dB(s)

+ektX

Z  +γˇ(u)

γˆN(dt,du).

Integrating from  totand taking expectation on both sides of (), we have

EektQX(t)≤QX()+E

t

keksQX(s)+eksLQX(s)ds

.

Easily, one has

kektQX(t)+ektLQX(t)≤kektX(t) +ektX–(t) –bX(t) +X(t) ≤ekt sup

XR+

X–

bk

X+X

+ 

:=ektH.

Therefore

EXX ()

ekt +

H kX

() +H:=M. ()

By Lemma ., applying the Burkholder-Davis-Gundy inequality, integrating equation () from  tot, and for an arbitrarily small positive constantδ, one has

E sup

t≤(k+)δ

X(t)≤EX()+Y+Y

(8)

where

Y =E

sup

t≤(k+)δ

kδtX–(s) –bX(s) +X(s)ds

cE

sup

t≤(k+)δ

kδtX(s)ds

cE (k+)δ

X(s)ds

E

sup

t≤(k+)δ

X(s)ds

, k= , , . . .

and

Y =E

sup

t≤(k+)δ

t

X–(s) σI(s)dB(s) +σI(s)dB(s) +σS(s)dB(s)

+σV(s)dB(s)

+

t

X(s)

Z  +γˇ(u)

γˆN(dt,du)

CE (k+)δ

X(–)σI+σI+σS+σVds

 

+CE (k+)δ

X

Z  +γˇ(u)

γˆ(u)v(du)ds

 

 

σ+

Z  +ˇ

γ(u)γˆ(u)v(du)

E sup

t≤(k+)δ

X

, k= , , . . . ,

wherec,C> .

So we have

E sup

t≤(k+)δ

X(t)≤EX()+E

sup

t≤(k+)δ

X(s)ds

+

 

σ+

Z  +γˇ(u)

γˆ(u)v(du)

×E sup

t≤(k+)δ

X

.

Choose a positive constantδthat satisfies

+

 

σ+

Z  + ˇ

γ(u)γˆ(u)v(du)

≤  .

Combining it with equation (), one has

E sup

t≤(k+)δ

X(t)≤EX()≤M.

Applying the arbitrariness ofκX>  and Lemma . for Chebyshev’s inequality, one

ob-tains

P sup

t≤(k+)δ

X(t) > ()+κX

≤E[supt≤(k+)δX(t)]

()+κX

≤ M

(9)

Applying the Borel-Cantelli lemma [], for almost allω, one has

sup

t≤(k+)δ

X(t)≤()+κX ()

holds for all but finitely manyk. Therefore, for any positive constantkkand almost all

ω, there isk(ω) such that equation () holds.

Thus, for almost allω, once conditionskkandt≤(k+ )δhold, then we have

lnX(t)

lnt

( +κX)ln()

ln() =  +κX. ()

Taking the limit superior on both sides of equation () and applying the arbitrariness of

κX> , one has

lim sup

t→∞

lnX(t)

lnt ≤ a.s.

Easily, for anysatisfying  <<  +(uσ–φ), one hasu>

–  σ

+φ. Therefore

lim sup

t→∞

lnX(t)

lnt

a.s.

That is to say, for any constantτ satisfying  <τ<  –, there is a constantN=N(ω), and once conditiontNholds, then we have

lnX(t)≤

+τ

lnt.

Therefore

lim

t→∞

X(t)

t =tlim→∞

S(t) +I(t) +I(t) +V(t)

t = ≤lim supt→∞

t+τ

t =  a.s.

So

lim

t→∞

S(t)

t = , tlim→∞

I(t)

t = ,

lim

t→∞

I(t)

t = , tlim→∞

V(t)

t =  a.s.

and

lim

t→∞

lnS(t)

t ≤, tlim→∞

lnI(t)

t ≤,

lim

t→∞

lnI(t)

t ≤, tlim→∞

lnV(t)

t ≤ a.s.

(10)

Lemma . For any initial value(S(),I(),I(),V())∈R+,the solution(S(t),I(t),I(t),

V(t))of model()has the following property:

lim

t→∞

t

I(s)dB(s)

t = , tlim→∞

t

Zγ(u)I(s)N(ds,du)

t =  a.s.,

lim

t→∞

t

I(s)dB(s)

t = , tlim→∞

t

Zγ(u)I(s)N(ds,du)

t =  a.s.,

lim

t→∞

t

S(s)dB(s)

t = , tlim→∞

t

Zγ(u)S(s)N(ds,du)

t =  a.s.,

lim

t→∞

t

V(s)dB(s)

t = , tlim→∞

t

Zγ(u)V(s)N(ds,du)

t =  a.s.

Proof Define

X(t) =

t

I(s)dB(s), Y(t) =

t

Zγ(u)I(s)

N(ds,du),

X(t) =

t

I(s)dB(s), Y(t) =

t

Z

γ(u)I(s)N(ds,du),

X(t) =

t

S(s)dB(s), Y(t) =

t

Zγ(u)S(s)

N(ds,du),

X(t) =

t

V(s)dB(s), Y(t) =

t

Zγ(u)V(s)

N(ds,du).

Applying Lemma . for the Burkholder-Davis-Gundy inequality and Lemma . for Hölder’s inequality, one has

Esup

≤st

X(s)

CE

t

I(θ)

CE

t

I(θ)

,

Esup

≤st

Y(s)

CE

t

ZI

(θ)γ(u)

 ≤C Zγ

(u)v(du)

 E

t

I(θ)

for  <<  +(uσ–φ). HereC= [ +

(–)–]

>  is a constant.

Applying equation (), we have

E sup

kt≤(k+) X(s)

≤MC(k+ )

 ≤+

MCk

.

For any constantκX> , applying Lemma . for Doob’s martingale inequality, one

ob-tains

Pω: sup

kt≤(k+) X(t)

>k+κX+

≤E[supkt≤(k+)|X(k+ )|]

k+κX+ ≤

+MC k

k+κX+

≤+

MC

(11)

Applying the Borel-Cantelli lemma, one has

ln|X(t)|

lnt

( +κX+

)lnk

lnk =  +κX+

. ()

Taking the limit superior on both sides of equation () and applying the arbitrariness of

κX> , one has

lim sup

t→∞

ln|X(t)|

lnt

 +

a.s.

That is to say, for any constantτ satisfying  <τ<, there is a constantN=N(ω), and oncetN,wτ holds, then we have

lnX(t)≤

 +

+τ

lnt. ()

Dividing both sides of equation () bytand taking the limit superior, we have

lim sup

t→∞ |X(t)|

t ≤lim supt→∞

t++τ

t = .

Combining it withlim inft→∞|Xt(t)|≥, one has

lim

t→∞

|X(t)|

t =tlim→∞

X(t)

t =  a.s.

Similarly, one obtains

lim

t→∞

lnX(t)

t = , tlim→∞

lnX(t)

t = ,

lim

t→∞

lnX(t)

t = , tlim→∞

lnY(t)

t = ,

lim

t→∞

lnY(t)

t = , tlim→∞

lnY(t)

t = , tlim→∞

lnY(t)

t = .

This completes the proof.

Lemma . For any initial value(S(),I(),I(t),V())∈R+,model()has a unique

pos-itive solution(S(t),I(t),I(t),V(t))∈R+on t≥with probability.

Proof The proof is similar to Refs. [, ] by definingQ(S,I,I,V) =S–  –lnS+I–  –

lnI+I–  –lnI+V–  –lnV, and hence is omitted.

2.2 Stochastic disease-free dynamics

Theorem . Suppose that conditions R< and R< hold.Then,for any initial value (S(),I(),I(),V())∈R+,the solution(S(t),I(t),I(t),V(t))of model()has the

follow-ing property:

lim

t→∞Ii(t) = , i= , , tlim→∞

S(t) =(u+δuq)

(12)

lim

t→∞

V(t) = (p+uq)

u++up.

That is to say,the two epidemic diseases go to extinct almost surely.

Proof By equation (), one has

d

S+I+I+

δ u+δV

=(u+δuq)

u+δ

u++up

u+δ S

i=

(u+di)Ii

+σS dB(t) +

Zγ(u)S

tN(dt,du)

+ 

i=

σiIidBi(t) +

Zγi(u)Ii

tN(dt,du)

+ δ

u+δ

σV dB(t) +

Zγ(u)V

tN(dt,du)

. ()

Dividing both sides of equation () bytand integrating over the time interval  totyield

S(t) = u+δ

u++up !

(u+δuq) u+δ

i= (u+di)

Ii(t) –(t)

"

, ()

where

(t) =

t

#

S(t) –S() + 

i=

Ii(t) –Ii()

+ δ

u+δ

V(t) –V()

– 

i=

t

σiIidBi(s) +

Zγi(u)Ii(s)

N(dt,du)

t

σS dB(s) +

Zγ(u)S(s)

N(dt,du)

δ

u+δ

t

σV dB(s) +

Zγ(u)V(s)

N(dt,du) $

.

Applying Lemmas . and ., we obtain that

lim

t→+∞(t) =  a.s. ()

Applying the generalized Itô’s formula in Lemma . toαlnI(t) +I(t) yields

d αlnI(t) +I(t)

= βS– (u+d+r)I–α(u+d+r) –αb

dt

+ (α+I)σdB(t) +

Z αln

 +γ(u)

(13)

Dividing both sides of equation () by t, integrating over the time interval  to tand taking the limit, one obtains that

αlnI(t) +I(t)

t =

αlnI() +I()

t +β

S(t) – (u+d+r)

I(t)

α(u+d+r) –αb+ 

t

t

α+I(s)

σdB(s)

+

t

t

Z α

ln +γ(u)

+I(s)γ(u)N(dt,du). ()

Combining equations () and (), one obtains

αlnI(t)

t =

β(u+δuq)

u++upα(u+d+r+b) –

β(u+δ)(u+d)

u++up

I(t)

β(u+δ)(u+d)

u++up + (u+d+r)

I(t) +

αlnI() +I()

t

I(t)

t

β(u+δ)

u++up(t) + 

t

t

α+I(s)

σdB(s)

+

t

t

Z α

ln +γ(u)

+I(s)γ(u)N(dt,du)

=β(u+δuq)

u++upα(u+d+r+b) –

β(u+δ)(u+d)

u++up

I(t)

β(u+δ)(u+d)

u++up + (u+d+r)

I(t) +(t), ()

where

(t) =

αlnI() +I()

tI(t)

t

β(u+δ)

u++up(t)

+

t

t

α+I(s)

σdB(s)

+

t

t

Z αln

 +γ(u)

+I(s)γ(u)N(dt,du).

Similarly, applying the generalized Itô’s formula in Lemma . toαlnI(t) +I(t) yields

αlnI(t)

t =

β(u+δuq)

u++upα(u+d+r+b) –

β(u+δ)(u+d)

u++up

I(t)

β(u+δ)(u+d)

u++up + (u+d+r)

I(t) +(t), ()

where

(t) =

αlnI() +I()

tI(t)

t

β(u+δ)

u++up(t)

+

t

t

α+I(s)

σdB(s)

+ t t  Z αln

 +γ(u)

(14)

Applying Lemmas . and ., we obtain that

lim

t→+∞i(t) = , i= ,  a.s. ()

By taking the limit superior of both sides of equation () and equation (), respectively, one has

lim sup

t→∞

αlnI(t)

t

β(u+δuq)

u++upα(u+d+r+b) =R< ,

lim sup

t→∞

αlnI(t)

t

β(u+δuq)

u++upα(u+d+r+b) =R< .

That is to say,

lim

t→∞Ii(t) = , i= ,  a.s. ()

Applying () and () into equation (), we obtain that

lim

t→∞

S(t) = u+δ

u++up !

(u+δuq) u+δ

i=

(u+di)lim t→∞

Ii(t) – lim t→∞(t)

"

=(u+δuq)

u++up. ()

By equation (), one has

d(S+I+I+V) = uSuV– (u+d)I– (u+d)I

dt

+ 

i=

σiIidBi(t) +

Z γi(u)Ii

tN(dt,du)

+σS dB(t) +

Zγ(u)S

tN(dt,du)

+σV dB(t) +

Z γ(u)V

tN(dt,du). ()

Dividing both sides of equation () byt, integrating over the time intervalt=  totand taking the limit, one obtains that

lim

t→∞

V(t) =

utlim→∞

S(t) – 

i=

u+di

u tlim→∞

Ii(t)

–lim

t→∞

S(t) –S() +%i=(Ii(t) –Ii()) +V(t) –V()

ut

+

utlim→∞

t

t

#

i=

σiIi(s)dBi(s) +

Z γ(u)I

sN(ds,du)

+σS(s)dB(s) +

Zγ(u)S

sN(ds,du)

+σV(s)dB(s) +

Z γ(u)V

sN(ds,du) $

(15)

Applying (), (), Lemmas . and ., we have

lim

t→∞

V(t) =

u

(u+δuq) u++up =

(p+uq) u++up.

This completes the proof.

2.3 Stochastic endemic dynamics

Theorem . For any initial value(S(),I(),I(),V())∈R+,the solution(S(t),I(t),

I(t),V(t))of model()has the following property:

(i) IfR> andR< ,then the epidemic diseaseI(t)is persistent in mean andI(t)

goes extinct,i.e.limt→∞I(t)=ϒR > ,limt→∞I(t) = a.s.Moreover,

lim

t→∞

S(t) =(u+δuq)

u++up

(u+δ)(u+d)

u++up

R

ϒ

a.s.,

lim

t→∞

V(t) = (p+uq)

u++up

(u+d)p

u(u+δ+p)

R

ϒ

a.s.

(ii) IfR< andR> ,then the epidemic diseaseI(t)goes extinct andI(t)is persistent

in mean,i.e.limt→∞I(t)= ,limt→∞I(t) = ϒR > a.s.Moreover,

lim

t→∞

S(t) =(u+δuq)

u++up

(u+δ)(u+d)

u++up

R

ϒ

a.s.,

lim

t→∞

V(t) = (p+uq)

u++up

(u+d)p

u(u+δ+p)

R

ϒ

a.s.

Proof Case (i): From equation () we have

αlnI(t)

t =

β(u+δuq)

u++upα(u+d+r+b)

β(u+δ)(u+d)

u++up + (u+d+r)

I(t)

β(u+δ)(u+d)

u++up

I(t) +(t)

=R–ϒ

I(t) –ϒ

I(t) +(t), ()

where

ϒ=

β(u+δ)(u+d)

u++up + (u+d+r), ϒ=

β(u+δ)(u+d)

u++up .

From Theorem ., whenR<  one has

lim

t→∞I(t) =  a.s. ()

Therefore, there exists an arbitrarily small constantε>  such that whentis large enough, we haveI(t) <ε. Applying this into equation () leads to

R–ϒ

I(t) +(t)≥

αlnI(t)

tR–ϒ

(16)

Applying Lemma . and the arbitrariness ofε, we obtain

lim

t→∞

I(t) =

R

ϒ

a.s. ()

Applying (), () and () into equation (), we obtain that

lim

t→∞

S(t) = u+δ

u++up !

(u+δuq) u+δ

i=

(u+di)lim t→∞

Ii(t) – lim t→∞(t)

"

=(u+δuq)

u++up

(u+δ)(u+d)

u++up

R

ϒ

. ()

Applying (), (), (), Lemmas . and . into equation (), we have

lim

t→∞

V(t) =

u

(u+δuq) u++up +

(u+δ)(u+d)

u++up

R

ϒ

u+d

u R

ϒ

= (p+uq)

u++up

(u+d)p

u(u+δ+p)

R

ϒ .

Case (ii): From equation () we have

αlnI(t)

t =

β(u+δuq)

u++upα(u+d+r+b) –

β(u+δ)(u+d)

u++up

I(t)

β(u+δ)(u+d)

u++up + (u+d+r)

I(t) +(t)

=R–ϒ

I(t) –ϒ

I(t) +(t), ()

where

ϒ=

β(u+δ)(u+d)

u++up , ϒ=

β(u+δ)(u+d)

u++up + (u+d+r).

From Theorem ., whenR<  one has

lim

t→∞I(t) =  a.s. ()

Therefore, there exists an arbitrarily small constantε>  such that whentis large enough, we haveI(t) <ε. Applying this into equation () leads to

R–ϒ

I(t) +(t)≥

αlnI(t)

tR–ϒ

I(t) –ϒε+(t).

Applying Lemma . and the arbitrariness ofε, we obtain

lim

t→∞

I(t) =

R

ϒ

(17)

Applying equations (), (), () into equation (), we obtain that

lim

t→∞

S(t) = u+δ

u++up !

(u+δuq) u+δ

i=

(u+di)lim t→∞

Ii(t) – lim t→∞(t)

"

=(u+δuq)

u++up

(u+δ)(u+d)

u++up

R

ϒ

. ()

Applying (), (), (), Lemmas . and . into equation (), we have

lim

t→∞

V(t) =

u

(u+δuq) u++up +

(u+δ)(u+d)

u++up

R

ϒ

u+d

u R

ϒ

= (p+uq)

u++up

(u+d)p

u(u+δ+p)

R

ϒ .

This completes the proof.

Theorem . Suppose that conditions R> and R> hold.Let(S(t),I(t),I(t),V(t))be

the solution of model()with the initial value(S(),I(),I(),V())∈R+.

(i) IfϒR<ϒR,then the epidemic diseaseI(t)is persistent in mean andI(t)goes

extinct,i.e.limt→∞I(t)=ϒR > ,limt→∞I(t) = a.s.Moreover,

lim

t→∞

S(t) =(u+δuq)

u++up

(u+δ)(u+d)

u++up

Rϒ a.s., lim t→∞

V(t) = (p+uq)

u++up

(u+d)p

u(u+δ+p)

R

ϒ

a.s.

(ii) IfϒR<ϒR,then the epidemic diseaseI(t)goes extinct andI(t)is persistent in

mean,i.e.limt→∞I(t)= ,limt→∞I(t) = ϒR > a.s.Moreover,

lim

t→∞

S(t) =(u+δuq)

u++up

(u+δ)(u+d)

u++up

Rϒ a.s., lim t→∞

V(t) = (p+uq)

u++up

(u+d)p

u(u+δ+p)

R

ϒ

a.s.

(iii) IfϒR>ϒR,ϒR>ϒR,then the epidemic diseasesIandIare persistent

in mean.Moreover,

lim

t→∞

I(t) =

ϒR–ϒR

ϒϒ–ϒϒ

, lim

t→∞

I(t) =

ϒR–ϒR

ϒϒ–ϒϒ

a.s.,

lim

t→∞

S(t) =(u+δuq)

u++up

(u+δ)(u+d)

u++up

ϒR–ϒR

ϒϒ–ϒϒ

–(u+δ)(u+d)

u++up

ϒR–ϒR

ϒϒ–ϒϒ

a.s.,

lim

t→∞

V(t) = (p+uq)

u++up

(u+d)p

u(u+δ+p)

ϒR–ϒR

ϒϒ–ϒϒ

– (u+d)

u(u+δ+p)

ϒR–ϒR

ϒϒ–ϒϒ

(18)

Proof Case (i): Note that

lim sup

t→+∞

lnI(t)

t ≤,

there exists an arbitrarily small constantε>  such that whentis large enough, we have

lnI(t)

t <ε.

From equation () and equation (), whentis large enough, one has

ϒαlnI(t)

t =ϒR–ϒR– (ϒϒ–ϒϒ)

I(t) +ϒα

lnI(t)

t

+ϒ(t) –ϒ(t)

ϒR–ϒR– (ϒϒ–ϒϒ)

I(t) +ϒαε

+ϒ(t) –ϒ(t). ()

SinceϒR<ϒRandϒϒ>ϒϒ, taking the limit superior of both sides of equa-tion (), applying equaequa-tion () and the arbitrariness ofε, we have

lim sup

t→+∞

lnI(t)

t

ϒR–ϒR

ϒα

< .

That is to say,

lim

t→∞I(t) =  a.s.

By using the method of Case (ii) in Theorem ., one obtains the persistence in mean of

I(t),S(t) andV(t), and hence is omitted.

Case (ii): The proof of Case (ii) is similar to the proof of Case (i) in this subsection and hence is omitted.

Case (iii): SinceϒR>ϒRandϒϒ>ϒϒ, using Lemma . and the arbitrari-ness ofεfor equation (), one obtains that

lim sup

t→+∞

I(t) ≤

ϒR–ϒR

ϒϒ–ϒϒ

a.s. ()

Similarly, whenϒR>ϒR, we have

lim sup

t→+∞

I(t) ≤

ϒR–ϒR

ϒϒ–ϒϒ

a.s. ()

From equation (), there exists an arbitrarily small constantε>  such that whentis large enough, we have

I(t) ≤

ϒR–ϒR

ϒϒ–ϒϒ

(19)

Applying equation () into equation (), one obtains that

αlnI(t)

t =R–ϒ

I(t) –ϒ

I(t) +(t)

R–ϒ

I(t) –ϒεϒ

ϒR–ϒR

ϒϒ–ϒϒ

+(t).

By using Lemma . and the arbitrariness ofε, we obtain that

lim inf

t→+∞

I(t) ≥

ϒR–ϒR

ϒϒ–ϒϒ

a.s. ()

Similarly, one obtains

lim inf

t→+∞

I(t) ≥

ϒR–ϒR

ϒϒ–ϒϒ

a.s. ()

Applying equations (), (), () and () leads to

lim

t→+∞

I(t) =

ϒR–ϒR

ϒϒ–ϒϒ

, lim

t→+∞

I(t) =

ϒR–ϒR

ϒϒ–ϒϒ

a.s. ()

Applying () and () into equation (), we obtain that

lim

t→∞

S(t) = u+δ

u++up !

(u+δuq) u+δ

i=

(u+di)lim t→∞

Ii(t) – lim t→∞(t)

"

=(u+δuq)

u++up

(u+δ)(u+d)

u++up

ϒR–ϒR

ϒϒ–ϒϒ

–(u+δ)(u+d)

u++up

ϒR–ϒR

ϒϒ–ϒϒ

. ()

Applying (), (), Lemmas . and . into equation (), we have

lim

t→∞

V(t) =

u

(u+δuq) u++up +

(u+δ)(u+d)

u++up

ϒR–ϒR

ϒϒ–ϒϒ

u+d

u

ϒR–ϒR

ϒϒ–ϒϒ

+(u+δ)(u+d)

u++up

ϒR–ϒR

ϒϒ–ϒϒ

u+d

u

ϒR–ϒR

ϒϒ–ϒϒ

= (p+uq)

u++up

(u+d)p

u(u+δ+p)

ϒR–ϒR

ϒϒ–ϒϒ

– (u+d)p

u(u+δ+p)

ϒR–ϒR

ϒϒ–ϒϒ .

This completes the proof.

3 Conclusions and numerical simulations

(20)

persistence in mean and extinction of the two diseases. Compared with the existing work in Refs. [] and [], the model constructed in this paper also considers the efficiency of vaccination. When all the coefficients related to the vaccination are , system () is similar to systems () and () in Refs. [] and [], in addition, our conclusion is consistent with them. That is to say, systems () and () in Refs. [] and [] are a special case of our system (). The theoretical results of this article can be used as a reference for the control of infectious diseases.

To sum up, we have the following conclusions:

I. Stochastic disease-free dynamics WhenR< andR< hold, we have

lim

t→∞Ii(t) = , i= , , tlim→∞

S(t) =(u+δuq)

u++up,

lim

t→∞

V(t) = (p+uq)

u++up.

That is to say, the two epidemic diseases go to extinct almost surely. II. Stochastic endemic dynamics

(i) If one of the following conditions holds: • R> ,R< ,

R,R> ,ϒR<ϒR, then we have

lim

t→∞

I(t) =

R

ϒ

> , lim

t→∞I(t) =  a.s.,

lim

t→∞

S(t) =(u+δuq)

u++up

(u+δ)(u+d)

u++up

R

ϒ a.s.,

lim

t→∞

V(t) = (p+uq)

u++up

(u+d)p

u(u+δ+p)

R

ϒ a.s.

That is to say, the epidemic diseaseI(t)is persistent in mean andI(t)is extinct. (ii) If one of the following conditions hold:

R< ,R> ,

R,R> ,ϒR<ϒR, then we have

lim

t→∞

I(t) = , lim

t→∞I(t) =

R

ϒ

>  a.s.,

lim

t→∞

S(t) =(u+δuq)

u++up

(u+δ)(u+d)

u++up

R

ϒ a.s.,

lim

t→∞

V(t) = (p+uq)

u++up

(u+d)p

u(u+δ+p)

R

ϒ a.s.

That is to say, the epidemic diseaseI(t)is extinct andI(t)is persistent in mean. (iii) IfϒR>ϒR,ϒR>ϒRhold, then we have

lim

t→∞

I(t) =

ϒR–ϒR

ϒϒ–ϒϒ

, lim

t→∞

I(t) =

ϒR–ϒR

Figure

Figure (a) is the time sequence diagram of system () withiure (b) is the corresponding phase diagram of =i = , i = ,,,; Fig- I(t) and I(t)
Figure 1 Time sequence diagram and phase diagram of model (4) without stochastic effects.
Figure 3 Time sequence diagram and phase diagram of model (4) for extinctions of disease 2 andpersistence of disease 1.

References

Related documents