We study one-signed periodic solutions of the first-order functional differential equation u′(t)=-a(t)u(t)+λb(t)f(u(t-τ(t))), t∈ℝ by using global bifurcation techniques. Where a,b∈C(ℝ,[0,∞)) are ω-periodic functions with ∫0ωa(t)dt>0, ∫0ωb(t)dt>0, τ is a continuous ω-periodic function, and λ>0 is a parameter. f∈C(ℝ,ℝ) and there exist two constants s2<0<s1 such that f(s2)=f(0)=f(s1)=0, f(s)>0 for s∈(0,s1)∪(s1,∞) and f(s)<0 for s∈(-∞,s2)∪(s2,0).
1. Introduction
In recent years, there has been considerable interest in the existence of periodic solutions of the following equation: u′(t)=-a(t)u(t)+λb(t)f(u(t-τ(t))),
where a,b∈C(ℝ,[0,∞)) are ω-periodic functions, and ∫0ωa(t)dt>0,∫0ωb(t)dt>0, τ is a continuous ω-periodic function, λ>0 is a parameter. (1.1) has been proposed as a model for a variety of physiological processes and conditions including production of blood cells, respiration, and cardiac arrhythmias; see, for example, [1–12] and the references therein. Roughly speaking, u(t) represents the number of adult (sexually mature) members in a population at time t,a(t) is the per capita death rate, and f(u(t-τ(t))) is the rate at which new members are recruited into the population at time t ( τ is the age at which members mature, and it is assumed that the birth rate at a given time depends only on the adult population size). The most famous models of this type are
the Nicholson's blowflies equation proposed in [1] to explain the oscillatory population fluctuations observed by A. J. Nicholson in 1957 in his studies of the sheep blowfly Lucilia cuprina:
u′(t)=-au(t)+p⋅u(t-h)e-γu(t-h),a,p,γ,h>0;
the model for blood cell populations proposed by Mackey and Glass in [2]
u′(t)=-au(t)+pu(t-h)1+[u(t-h)]n,a,p,γ,h>0,n>1;
the model for the survival of red blood cells in an animal proposed by Wazewska-Czyzewska and Lasota in [3]
u′(t)=-au(t)+p⋅e-γu(t-h),a,p,γ,h>0.
Recently, Cheng and Zhang [7] studied the existence of positive ω-periodic solutions of the functional equation (1.1) under the assumptions:
f∈C([0,∞),[0,∞)), and f(s)>0 for s>0;
a,b∈C(ℝ,[0,∞)) are ω-periodic functions, ∫0ωa(t)dt>0, ∫0ωb(t)dt>0, τ∈C(ℝ,ℝ) is a ω-periodic function;
there exist f0,f∞∈(0,∞) such that
f0=lim|s|→0f(s)s,f∞=lim|s|→∞f(s)s.
They proved the following.
Theorem A.
Assume (H1)–(H3)hold. Then for each λ satisfying 1σBf∞<λ<1Af0,or1σAf0<λ<1Bf∞,
equation (1.1) has a positive periodic solution, where
A=maxt∈[0,ω]∫0ωG(t,s)b(s)ds,B=mint∈[0,ω]∫0ωG(t,s)b(s)ds,σ=e∫0ωa(t)dt.
However, the condition used in [7] is not sharp, and the main results in [7] give no any information about the global structure of the set of positive periodic solutions. Moreover, f satisfied (H1) in [7], so a natural question is what would happen if f is allowed to have some zeros in ℝ? The purpose of this work is to study the global behavior of the components of one-signed solutions of (1.1) under the condition
f∈C(ℝ,ℝ); there exist two constants s2<0<s1 such that f(s2)=f(0)=f(s1)=0, f(s)>0 for s∈(0,s1)∪(s1,∞), and f(s)<0 for s∈(-∞,s2)∪(s2,0).
The rest of this paper is organized as follows. In Section 2, we give some notations and the main results. Section 3 is devoted to proving the main results.
2. Statement of the Main Results
Let Y={u∈C(ℝ,ℝ):u(t)=u(t+ω)} with the norm ‖u‖∞=maxt∈[0,ω]|u(t)|.
Then (Y,∥·∥∞) is a Banach space. Let E={u∈C1(R,R):u(t)=u(t+ω)}
be the Banach space with the norm ∥u∥=max{∥u∥∞,∥u′∥∞}.
It is well known that (1.1) is equivalent to
u(t)=λ∫tt+ωG(t,s)b(s)f(u(s-τ(s)))ds:=(Au)(t),
where
G(t,s)=e∫tsa(θ)dθe∫0ωa(θ)dθ-1,s∈[t,t+ω].
Notice that ∫0wa(t)dt>0, we have
1σ-1≤G(t,s)≤σσ-1,
where σ=e∫0ωa(t)dt, and 0<1/σ<1.
Define that K is a cone in Y by K={u∈Y:u(t)≥0,u(t)≥1σ‖u‖}.
It is not difficult to prove that A(K)⊂K and A:K→K is completely continuous.
Let us consider the spectrum of the linear eigenvalue problem u′(t)=-a(t)u(t)+λb(t)u(t-τ(t)),t∈R.
Lemma 2.1.
Assume that (H2) holds. Then the linear problem (2.7) has a unique eigenvalue λ1, which is positive and simple, and the corresponding eigenfunction φ is of one sign.
Proof.
It is a direct consequence of the Krein-Rutman Theorem [13, Theorem 19.3].
In the rest of the paper, we always assume that
‖φ‖=1,φ(t)>0,t∈R.
Define L:E→Y by setting Lu:=u′(t)+a(t)u(t),u∈E.
Then L-1:Y→E is completely continuous.
Let ζ,ξ∈C(ℝ,ℝ) be such that f(s)=f0s+ζ(s),f(s)=f∞s+ξ(s).
Clearly, lim|s|→0ζ(s)s=0,lim|s|→∞ξ(s)s=0.
Let us consider Lu(t)-λb(t)f0u(t-τ(t))=λb(t)ζ(u(t-τ(t)))
as a bifurcation problem from the trivial solution u≡0 and Lu(t)-λb(t)f∞u(t-τ(t))=λb(t)ξ(u(t-τ(t)))
as a bifurcation problem from infinity. We note that (2.12) and (2.13) are the same and each of them is equivalent to (1.1).
Let 𝔼=ℝ×E under the product topology. We add the points {(λ,∞)∣λ∈ℝ} to our space 𝔼. Let S+ denote the set of positive functions in E and S-=-S+, and S=S-∪S+. They are disjoint and open in E. Finally, let Φ±=ℝ×S± and Φ=ℝ×S.
Remark 2.2.
It is worth remaking that if u is a nontrivial solution of (1.1) and a,b, and f satisfy (H2)–(H4), then u∈Sν for some ν={+,-}. To see this, define
q(t)={f(u(t))u(t),u(t)≠0,f0,u(t)=0.
Thus (1.1) is equivalent to
u′(t)=-a(t)u(t)+λb(t)q(t-τ(t))u(t-τ(t)),t∈R.
Obviously, b(·)q(·-τ(·)) satisfies (H2). From Lemma 2.1, the nontrivial solution u∈Sν for some ν∈{+,-}.
The result of Rabinowitz [14] for (2.12) can be stated as follows: for each ν∈{+,-}, there exists a continuum 𝒞ν of solutions of (2.12) joining (λ1/f0,0) to infinity, and 𝒞ν∖{(λ1/f0,0)}⊂Φν.
The result of Rabinowitz [15] for (2.13) can be stated as follows: for each ν∈{+,-}, there exists a continuum 𝒟ν of solutions of (2.13) meeting (λ1/f∞,∞), and 𝒟ν∖{(λ1/f∞,∞)}⊂Φν.
Our main result is the following.
Theorem 2.3.
Assume (H2)–(H4) hold. Moreover, suppose that
f satisfies the Lipschitz condition in [s2,s1].
Then
for (λ,u)∈𝒞+∪𝒞-,
s2<u(t)<s1,t∈[0,ω];
for (λ,u)∈𝒟+∪𝒟-, we have that either
maxt∈[0,ω]u(t)>s1
or
mint∈[0,ω]u(t)<s2.
Corollary 2.4.
Let (H2)–(H5) hold. Then
if λ∈(λ1/f∞,λ1/f0], then (1.1) has at least two solutions u∞+ and u∞-, such that u∞+ is positive on [0,ω] and u∞- is negative on [0,ω];
if λ∈(λ1/f0,∞), then (1.1) has at least four solutions u∞+, u∞-, u0+, and u0-, such that u∞+, u0+ are positive on [0,ω] and u∞-, u0- are negative on [0,ω].
Corollary 2.5.
Let (H2)–(H5) hold. Then
if λ∈(λ1/f0,λ1/f∞], then (1.1) has at least two solutions u0+ and u0-, such that u0+ is positive on [0,ω] and u0- is negative on [0,ω];
if λ∈(λ1/f∞,∞), then (1.1) has at least four solutions u∞+, u∞-, u0+, and u0-, such that u∞+, u0+ are positive on [0,ω] and u∞-, u0- are negative on [0,ω].
3. Proof of the Main Results
To prove Theorem 2.3, we give a Proposition.
Proposition 3.1.
(i) The first-order boundary value problem
u′(t)+a(t)u(t)=h(t),u(0)=u(ω)
has a unique solution for all h∈L1[0,ω] if and only if ∫0ωa(s)ds≠0.
(ii) Assume that u is a solution of (3.1). If h≥0 and h(·)≢0 on any subinterval of [0,ω], then u(t)∫0ωa(s)ds>0 on [0,ω].
Proof.
(i) The equation u′(t)+a(t)u(t)=0 has a solution u(t)=Ce-∫0ta(s)ds, where C is a constant. If u(t) is a nontrivial solution, then by u(0)=C,u(ω)=Ce-∫0ωa(s)ds, we can get that ∫0ωa(s)ds=0.
On the other hand, from ∫0ωa(s)ds=0, we can get that u′(t)+a(t)u(t)=0 has a nontrivial solution u(t)=Ce-∫0ta(s)ds, where C∈ℝ∖{0}.
(ii) We claim that u(t)≠0,t∈[0,ω]. Suppose on the contrary that there exists t0∈[0,ω], such that u(t0)=0; it is not difficult to compute that
u′(t)+a(t)u(t)=h(t),u(t0)=u0
has a solution
u(t)=∫t0th(s)e∫tsa(τ)dτds.
Since h≥0, we have
u(0)≤u(t0)≤u(ω).
If h(t̂)>0,t̂∈[0,t0), then there exists a neighborhood U(t̂)⊂[0,t0) of t̂, such that h(t)>0 on U(t̂). Thus, u(0)=∫t00h(s)e∫0sa(τ)dτds<0; this contradicts with u(0)=u(ω).
If h(t̅)>0,t̅∈(t0,ω], then there exists a neighborhood U(t̅)⊂(t0,ω] of t̅, such that h(t)>0 on U(t̂). By using a similar way, we can prove that u(ω)>0, which also contradicts with u(0)=u(ω).
Hence u(t)≠0 on [0,ω]. Moreover, it follows that
∫0ωu′(t)u(t)dt+∫0ωa(t)dt=∫0ωh(t)u(t)dt,
that is,
(lnu(t))∣0ω+∫0ωa(t)dt=∫0ωh(t)u(t)dt.
Thus ∫0ωa(t)dt=∫0ωh(t)/u(t)dt, that is u∫0ωa(s)ds>0.
Next, we prove Theorem 2.3 and Corollaries 2.4 and 2.5.
Proof of Theorem 2.3.
Suppose on the contrary that there exists (λ,u)∈𝒞+∪𝒞-∪𝒟+∪𝒟- such that either
max{u(t)∣t∈[0,ω]}=s1
or
min{u(t)∣t∈[0,ω]}=s2.
We divide the proof into two cases.
Case 1 (max{u(t)|t∈[0,ω]}=s1).
In this case, we know that
0≤u(t)≤s1,0≤u(t-τ(t))≤s1,t∈[0,ω].
Let us consider the functional differential equation
u′(t)+a(t)u(t)=λb(t)f(u(t-τ(t))),t∈R.
By (H2), (H4) and (H5), there exists m≥0 such that b(t)f(s)+ms is strictly increasing on s for s∈[s2,s1]. Then (3.10) can be rewritten to the form
Lu+λmu(t-τ(t))=λ[b(t)f(u(t-τ(t)))+mu(t-τ(t))],
and since Ls1-a(t)s1=0=f(s1),
Ls1-a(t)s1+λms1=λ[b(t)f(s1)+ms1].
Subtracting, we get
L(s1-u)+λm(s1-u(t-τ(t)))-a(t)s1≥0.
That is,
L(s1-u)+λms1≥0,t∈[0,ω],s1-u(0)=s1-u(ω)>0.
From Proposition 3.1, we deduce that s1>u(t),t∈[0,ω], which contradicts with that max{u(t)|t∈[0,ω]}=s1. Hence,
u(t)<s1,t∈[0,ω].
Case 2 (min{u(t)|t∈[0,ω]}=s2).
In this case, we know that
s2≤u(t)≤0,s2≤u(t-τ(t))≤0,t∈[0,ω].
Let us consider (3.10); by (H2), (H4), and (H5), there exists m≥0 such that b(t)f(s)+ms is strictly increasing in s for s∈[s2,s1]. Then
Lu+λmu(t-τ(t))=λ[b(t)f(u(t-τ(t)))+mu(t-τ(t))]
and since Ls2-a(t)s2=0=f(s2),
Ls2-a(t)s2+λms2=λ[b(t)f(s2)+ms2].
Subtracting, we get
L(s2-u)+λm(s2-u(t-τ(t)))-a(t)s2≤0.
That is
L(s2-u)+λms2≤0,t∈[0,ω],s2-u(0)=s2-u(ω)<0.
From Proposition 3.1, we deduce that s2-u(t)<0,t∈[0,ω], this contradicts with that min{u(t)∣t∈[0,ω]}=s2. Therefore,
s2<u(t),t∈[0,ω].
Proof of Corollaries 2.4 and 2.5.
Since boundary value problem
u′(t)+a(t)u(t)=0,u(0)=u(ω)
has a unique solution u≡0, we get
(C+∪C-∪D+∪D-)⊂{(λ,u)∈R×E∣λ≥0}.
Take Λ∈ℝ as an interval such that Λ∩{λ1/f∞}={λ1/f∞} and ℳ as a neighborhood of (λ1/f∞,∞) whose projection on ℝ lies in Λ and whose projection on E is bounded away from 0. Then by [15, Theorem 1.6, and Corollary 1.8], we have that for each ν∈{+,-}, either
𝒟ν∖ℳ is bounded in ℝ×E in which case 𝒟ν∖ℳ meets {(λ,0)|λ∈ℝ}, or
𝒟ν∖ℳ is unbounded.
Moreover, if (1) occurs and 𝒟ν∖ℳ has a bounded projection on ℝ, then 𝒟ν∖ℳ meets (λk/f∞,∞), where λk≠λ1 is another eigenvalue of (2.7).
Obviously, Theorem 2.3 (ii) implies that (1) does not occur. So 𝒟+∖ℳ is unbounded.
Remark 2.2 guarantees that 𝒟+ is a component of solutions of (2.12) in S+ which meets (λ1/f∞,∞), and consequently Projℝ(𝒟+∖ℳ) is unbounded. Thus
ProjR(D+)⊃(λ1f∞,+∞).
Similarly, we get
ProjR(D-)⊃(λ1f∞,+∞).
By Theorem 2.3, for any (λ,u)∈(𝒞+∪𝒞-),
‖u‖∞<max{s1,|s2|}:=s*.
(3.26) and (2.12) imply that
‖u‖<max{s*,‖a‖∞s*+λ‖b‖∞max|s|≤s*|f(s)|},
which means that the sets {(λ,u)∈𝒞+∣λ∈[0,d]} and {(λ,u)∈𝒞-∣λ∈[0,d]} are bounded for any fixed d∈(0,∞). This together with the fact that 𝒞+ and 𝒞- join (λ1/f0,0) to infinity yields, respectively, that
ProjR(C+)⊃(λ1f0,+∞),ProjR(C-)⊃(λ1f0,+∞).
Combining (3.24), (3.25), and (3.28), we conclude the desired results.
Remark 3.2.
The methods used in the proof of Theorem 2.3, Corollaries 2.4, and 2.5 have been used in the study of other kinds of boundary value problems; see [16–18] and the references therein.
Remark 3.3.
The conditions in Corollaries 2.4 and 2.5 are sharp. Let us take
a(t)≡a>0,λ=a,b(t)=1,f(s)=s+h(s),τ(t)≡0.
Let
h(s)={-2ss2+1,s∈(-∞,-1)∪(1,+∞),-2s3s2+1,s∈[-1,1],
and consider problem
u′(t)=-a(t)u(t)+a[u(t)+h(u(t))],t∈[0,ω],u(0)=u(ω).
It is easy to see that λ1=a,f0=f∞=1. Since
λ1f∞=a=λ1f0,
the conditions of Corollaries 2.4 and 2.5 are not valid. In this case, (3.31) has no nontrivial solution. In fact, if u is a nontrivial solution of (3.31), then
0=∫0ωu′(t)dt=a∫0ωh(u(t))dt≠0,
which is a contradiction.
Acknowledgment
The paper is supported by the NSFC (no. 11061030), the Fundamental Research Funds for the Gansu Universities.
GurneyW. S.BlytheS. P.NisbetR. N.Nicholsons blowflies revisited1980287172110.1038/287017a0MackeyM. C.GlassL.Oscillation and chaos in physiological control systems1977197430028728910.1126/science.267326Wazewska-CzyzewskaM.LasotaA.Mathematical problems of the dynamics of a system of red blood cells1976623400682081ChowS. N.Existence of periodic solutions of autonomous functional differential equations197415350378033600310.1016/0022-0396(74)90084-9ZBL0295.34055FreedmanH. I.WuJ.Periodic solutions of single-species models with periodic delay199223368970110.1137/05230351158828ZBL0764.92016WuJ.WangZ.Positive periodic solutions of second-order nonlinear differential systems with two parameters2008561435410.1016/j.camwa.2007.07.0172427683ZBL1145.34333ChengS.ZhangG.Existence of positive periodic solutions for non-autonomous functional differential equations200159181863778ZBL1003.34059JiangD.WeiJ.Existence of positive periodic solutions for nonautonomous delay differential equations19992067157201741001YeD.FanM.WangH.Periodic solutions for scalar functional differential equations20056271157118110.1016/j.na.2005.03.0842154102ZBL1089.34056LiY.FanX.ZhaoL.Positive periodic solutions of functional differential equations with impulses and a parameter200856102556256010.1016/j.camwa.2008.05.0072460064ZBL1165.34401WangH.Positive periodic solutions of functional differential equations2004202235436610.1016/j.jde.2004.02.0182069005ZBL1064.34052JinZ.WangH.A note on positive periodic solutions of delayed differential equations201023558158410.1016/j.aml.2010.01.0152602413ZBL1194.34130DeimlingK.1985Berlin, GermanySpringer787404RabinowitzP. H.On bifurcation from infinity197314462475032870510.1016/0022-0396(73)90061-2ZBL0272.35017RabinowitzP. H.Some global results for nonlinear eigenvalue problems197174875130301587ZBL0212.16504AmbrosettiA.HessP.Positive solutions of asymptotically linear elliptic eigenvalue problems198073241142210.1016/0022-247X(80)90287-5563992ZBL0433.35026MaR.Global behavior of the components of nodal solutions of asymptotically linear eigenvalue problems200821775476010.1016/j.aml.2007.07.0292423057ZBL1152.34319AnY.MaR.Global behavior of the components for the second order m-point boundary value problems20082008102392907254593