In this work, we formulate the Beverton–Holt model on isolated time scales and extend existing results known in the discrete and quantum calculus cases. Applying a recently introduced definition of periodicity for arbitrary isolated time scales, we discuss the effects of periodicity onto a population modeled by a dynamic version of the Beverton–Holt equation. The first main theorem provides conditions for the existence of a unique ω -periodic solution that is globally asymptotically stable, which addresses the first Cushing–Henson conjecture on isolated time scales. The second main theorem concerns the generalization of the second Cushing–Henson conjecture. It investigates the effects of periodicity by deriving an upper bound for the average of the unique periodic solution. The obtained upper bound reveals a dependence on the underlying time structure, not apparent in the classical case. This work also extends existing results for the Beverton–Holt model in the discrete and quantum cases, and it complements existing conclusions on periodic time scales. This work can furthermore guide other applications of the recently introduced periodicity on isolated time scales.
Citation: Martin Bohner, Jaqueline Mesquita, Sabrina Streipert. The Beverton–Hold model on isolated time scales[J]. Mathematical Biosciences and Engineering, 2022, 19(11): 11693-11716. doi: 10.3934/mbe.2022544
[1] | Martin Bohner, Sabrina Streipert . Optimal harvesting policy for the Beverton--Holt model. Mathematical Biosciences and Engineering, 2016, 13(4): 673-695. doi: 10.3934/mbe.2016014 |
[2] | Qianhong Zhang, Fubiao Lin, Xiaoying Zhong . On discrete time Beverton-Holt population model with fuzzy environment. Mathematical Biosciences and Engineering, 2019, 16(3): 1471-1488. doi: 10.3934/mbe.2019071 |
[3] | John E. Franke, Abdul-Aziz Yakubu . Periodically forced discrete-time SIS epidemic model with disease induced mortality. Mathematical Biosciences and Engineering, 2011, 8(2): 385-408. doi: 10.3934/mbe.2011.8.385 |
[4] | Yang Li, Jia Li . Stage-structured discrete-time models for interacting wild and sterile mosquitoes with beverton-holt survivability. Mathematical Biosciences and Engineering, 2019, 16(2): 572-602. doi: 10.3934/mbe.2019028 |
[5] | Shishi Wang, Yuting Ding, Hongfan Lu, Silin Gong . Stability and bifurcation analysis of $ SIQR $ for the COVID-19 epidemic model with time delay. Mathematical Biosciences and Engineering, 2021, 18(5): 5505-5524. doi: 10.3934/mbe.2021278 |
[6] | Jaqueline G. Mesquita, Urszula Ostaszewska, Ewa Schmeidel, Małgorzata Zdanowicz . Global attractors, extremal stability and periodicity for a delayed population model with survival rate on time scales. Mathematical Biosciences and Engineering, 2021, 18(5): 6819-6840. doi: 10.3934/mbe.2021339 |
[7] | Guilherme M Lopes, José F Fontanari . Influence of technological progress and renewability on the sustainability of ecosystem engineers populations. Mathematical Biosciences and Engineering, 2019, 16(5): 3450-3464. doi: 10.3934/mbe.2019173 |
[8] | Jordi Ripoll, Jordi Font . Numerical approach to an age-structured Lotka-Volterra model. Mathematical Biosciences and Engineering, 2023, 20(9): 15603-15622. doi: 10.3934/mbe.2023696 |
[9] | Michael Leguèbe . Cell scale modeling of electropermeabilization by periodic pulses. Mathematical Biosciences and Engineering, 2015, 12(3): 537-554. doi: 10.3934/mbe.2015.12.537 |
[10] | Islam A. Moneim, David Greenhalgh . Use Of A Periodic Vaccination Strategy To Control The Spread Of Epidemics With Seasonally Varying Contact Rate. Mathematical Biosciences and Engineering, 2005, 2(3): 591-611. doi: 10.3934/mbe.2005.2.591 |
In this work, we formulate the Beverton–Holt model on isolated time scales and extend existing results known in the discrete and quantum calculus cases. Applying a recently introduced definition of periodicity for arbitrary isolated time scales, we discuss the effects of periodicity onto a population modeled by a dynamic version of the Beverton–Holt equation. The first main theorem provides conditions for the existence of a unique ω -periodic solution that is globally asymptotically stable, which addresses the first Cushing–Henson conjecture on isolated time scales. The second main theorem concerns the generalization of the second Cushing–Henson conjecture. It investigates the effects of periodicity by deriving an upper bound for the average of the unique periodic solution. The obtained upper bound reveals a dependence on the underlying time structure, not apparent in the classical case. This work also extends existing results for the Beverton–Holt model in the discrete and quantum cases, and it complements existing conclusions on periodic time scales. This work can furthermore guide other applications of the recently introduced periodicity on isolated time scales.
The Beverton–Holt recurrence
xt+1=ηKxtK+(η−1)xt, | (1.1) |
where η is the proliferation rate and K is the carrying capacity, was derived in [1] in the context of fisheries. The solution of the logistic differential equation evaluated at time T+t0 is used to describe the adult fish population (new generation), and the solution at time t0 represents the juveniles (old generation). The derivation led to a proliferation rate η=erT>1 , where r>0 is the growth rate of the underlying continuous model and T is the time span until adulthood. Equation (1.1) for more general parameters is also known as the Pielou equation [2]. For related work on the Beverton–Holt equation, see [3,4,5,6,7,8,9,10,11,12], and for other related work, see [13,14,15,16].
In [17], Cushing and Henson investigated the effects of a periodically enforced carrying capacity onto flour beetles, which were modeled by the Beverton–Holt recurrence (1.1). Based on their observations, the authors conjectured that the introduction of a periodic environment on populations, modeled by (1.1), results in the existence of a unique periodic solution. Further, the authors predicted that a periodic environment is deleterious to the population, as the average of the unique periodic solution is bounded above by the mean of the periodic environment. These conclusions were formulated as the first and second Cushing–Henson conjectures.
In the case of a two-periodic Kt and constant η , the conjectures have been analytically verified in [18]. For higher-order periodic carrying capacities Kt , the conjectures have been the focus of the works [19,20]. A discussion assuming additionally time-dependent growth rates can be found in [21]. The extension of the conjectures to periodic time scales was addressed in [22]. Periodic time scales are time domains such that if t is in the time scale, then so is t+ω , hence requiring an additive time structure. The discrete time setting obeys this additive property and is a special case of a periodic time scale, in contrast to the quantum time setting qN0 , which is not periodic. Periodic time scales are a subset of arbitrary time scales, a theory developed by Stefan Hilger in 1988, that unifies the discrete and continuous theories. Studying the Beverton–Holt model on time scales provides a platform to consider time-dependent time spans until adulthood instead of a constant time span T , as assumed in the derivation of the Beverton–Holt model. Due to a lack of existing periodicity definitions for general time domains, the conjectures remained unsolved for this and other examples of nonperiodic time scales. In [23], in order to extend the study to the quantum time scale, the authors defined periodicity in the quantum setting and discussed the Cushing–Henson conjectures. In the quantum time setting, the Beverton–Holt model reads as
x(qt)=ηK(t)x(t)K(t)+(η−1)x(t) |
with the carrying capacity K:qN0→R+:=(0,∞) ( q>1 ) and proliferation rate η>1 . In [23], the authors proved the existence and global stability of a unique periodic solution for periodic carrying capacities, i.e., for K such that qωK(qωt)=K(t) , confirming the first Cushing–Henson conjecture for the q -Beverton–Holt model. In [24], the authors were able to extend this result to time-dependent proliferation rates. The second Cushing–Henson conjecture, however, only remained true in the quantum time scale with a slight modification as follows.
Theorem 1.1 (See [23, Theorem 5.6]). The average of the ω -periodic solution ˉx of the q -Beverton–Holt model is strictly lessthan the average of the ω -periodic carrying capacitytimes the constant q−λ1−λ , i.e.,
∫qω1ˉx(t)Δt≤q−λ1−λ∫qω1K(t)Δt, |
where λ=1−(q−1)A , η=11−(q−1)A , and A>0 .
The multiplicative constant in Theorem 1.1 can be expressed by
q−λ1−λ=(q−1)(1+A)(q−1)A=(q−1)η(q−1)+η−1η(q−1)η−1η=qη−1η−1. | (1.2) |
This indicates that the second conjecture does not necessarily hold, and it reveals that the time scale determines the proportionality constant linearly. Given this reformulation of the classical second Cushing–Henson conjecture in the special case of a quantum time scale, we aim to find a general formulation of this conjecture on arbitrary isolated time scales. On any isolated time scale T , we therefore consider the Beverton–Holt model
xσ=ηKxK+(η−1)x, | (1.3) |
where η,K:T→R+ , η>1 , and where σ(t) is the next time step following t . We can express the recurrence (1.3) as a dynamic equation
xσ−xμ=:xΔ=αxσ(1−xK)withα=η−1μη, | (1.4) |
where μ(t) is the distance to the time point following t , formally introduced in Section 2. Equation (1.4) is known as the logistic dynamic equation [25], and it can be transformed equivalently into a linear dynamic equation using the variable substitution u=1/x for x≠0 , namely
uΔ=−αu+αK. | (1.5) |
To study the effects of periodicity on (1.3) on arbitrary time scales, we utilize the recently introduced definition of periodicity in [26]. This new concept was already successfully applied in [27]. This and other useful definitions are stated in Section 2. Section 3 concerns the generalization of the first Cushing–Henson conjecture, discussing the existence of a unique globally asymptotically stable periodic solution. The generalization of the second Cushing–Henson conjecture is addressed in Section 4. The paper is completed in Section 5 with some concluding remarks.
In this section, we introduce some necessary time scales fundamentals. A time scale T is a closed nonempty subset of R .
Definition 2.1 (See [28, Definition 1.1]). For t∈T , the forward jump operator σ:T→T is defined by
σ(t)=inf{s∈T:s>t}. |
We adopt the convention that inf∅=supT . If σ(t)>t , then we say that t is right-scattered. If σ(t)=t , then we say that t is right-dense. Similarly, left-scattered and left-dense points are defined. The graininess μ:T→R+0 is defined by μ(t):=σ(t)−t . We define fσ:T→R by fσ:=f∘σ . If T has a left-scattered maximum M , then we define Tκ=T∖{M} ; otherwise, Tκ=T .
In this work, we focus on isolated time scales, i.e., all points are left-scattered and right-scattered. Hence, in what follows, throughout, T refers to an isolated time scale, and the following definitions and results are taking this special time structure already into account, as well as the entire remainder of this paper.
Definition 2.2 (See [25, Definition 2.25]). A function p:T→R is called regressive, denoted by R , provided
1+μ(t)p(t)≠0forallt∈T. |
Moreover, p is called positively regressive, denoted by R+ , provided
1+μ(t)p(t)>0forallt∈T. |
Remark 2.3. Assume f:T→R and t∈Tκ . Then the delta-derivative of f , denoted by fΔ (see [25, Definition 1.10]), is
fΔ(t)=f(σ(t))−f(t)μ(t). |
For f,g:T→R , the product-rule for t∈Tκ (see [25]) reads as
(fg)Δ=fΔgσ+fgΔ=fΔg+fσgΔ, | (2.1) |
and, if g≠0 , the quotient-rule (see [25]) reads as
(fg)Δ=fΔg−fgΔggσ. |
The delta integral is defined for a,b∈T with a<b by
∫baf(τ)Δτ=∑τ∈[a,b)∩Tμ(τ)f(τ), |
and consequently,
FΔ=f,ifF(t)=∫taf(τ)Δτ. |
Using the product rule, we get the integration by parts formula (see [25])
∫ba(fΔg)(t)Δt=(fg)(b)−(fg)(a)−∫ba(fσgΔ)(t)Δt. | (2.2) |
The following circle-plus addition turns (R,⊕) into an Abelian group.
Definition 2.4 (See [28, p. 13]). Define circle plus and circle minus by
p⊕q=p+q+μpq,p⊖q=p−q1+μqforq∈R. |
Theorem 2.5 (See [25, Theorem 2.33). Let p∈R and t0∈T . Then the initial value problem
yΔ=p(t)y,y(t0)=1 |
possesses a unique solution.
The unique solution from Theorem 2.5 is called the dynamic exponential function and is denoted by ep(⋅,t0) . On an isolated time scale, the dynamic exponential function for p∈R is
ep(t,t0)=∏s∈[t0,t)∩T(1+μ(s)p(s)),t>t0. |
Useful properties of the dynamic exponential function follow. Part 6 can easily be shown using parts 4 and 5. Part 8 is is the content of [25, Theorem 2.39]. Part 9 is in [25, Theorem 2.48(i)], and the remaining parts are from [25, Theorem 2.36].
Theorem 2.6. If p,q∈R and t,s,r∈T , then
1. e0(t,s)=1 and ep(t,t)=1 ,
2. ep⊕q(t,s)=ep(t,s)eq(t,s) ,
3. ep⊖q(t,s)=ep(t,s)eq(t,s) ,
4. e⊖p(t,s)=ep(s,t)=1ep(t,s) ,
5. ep(σ(t),s)=(1+μ(t)p(t))ep(t,s) ,
6. ep(t,σ(s))=ep(t,s)1+μ(s)p(s) ,
7. eΔp(⋅,s)=pep(⋅,s) ,
8. eΔp(s,⋅)=−peσp(s,⋅) ,
9. p∈R+ implies ep(t,s)>0 , and
10. the semigroup property holds: ep(t,r)ep(r,s)=ep(t,s) .
The following result was used in [26, Proof of Theorem 6.2]. A variant of it was also included in [29, Theorem 2.1], see also [30, First formula in the line two lines after (3)]). Here we state it explicitly and include its short proof.
Lemma 2.7. Let f:T→R∖{0} be delta-differentiable and s,t∈T . Then
efΔf(s,t)=f(s)f(t). | (2.3) |
Proof. For fixed t∈T , define w(s):=f(s)f(t) . Since f≠0 , fΔf∈R and
wΔ(s)=fΔ(s)f(t)=f(s)f(t)fΔ(s)f(s)=w(s)fΔ(s)f(s),w(t)=1. |
By Theorem 2.5, the claim follows.
Theorem 2.8 (See [31]). If f is nonnegative with −f∈R+ , then
1−∫tsf(τ)Δτ≤e−f(t,s)≤exp(−∫tsf(τ)Δτ). | (2.4) |
Theorem 2.9. (Variation of Constants, see [28, Theorem 2.77]). If p∈R , f:T→R , t0∈T , and y0∈R , then the unique solution of the IVP
yΔ=p(t)y+f(t),y(t0)=y0 |
is given by
y(t)=ep(t,t0)y0+∫tt0ep(t,σ(s))f(s)Δs. | (2.5) |
As mentioned in the introduction, the definition of periodicity is crucial in the discussion of effects of periodicity. We refer to our recent work [26], where we introduced periodicity on isolated time sales as follows.
Definition 2.10 (See [26, Definition 4.1]). Let ω∈N . A function f:T→R is called ω -periodic, denoted by f∈Pω , provided
νΔfν=f,whereν=σωandfν=f∘ν. | (2.6) |
Example 2.11. If T=Z , then ν(t)=t+ω , νΔ(t)=1 , and (2.6) reduces to the known definition of periodicity, f(t+ω)=f(t) .
Example 2.12. If T=qN0 , then ν(t)=qωt , νΔ(t)=qω , and (2.6) reduces to the known definition of periodicity, qωf(qωt)=f(t) , which was introduced in [23].
Now, from our recent paper [26], we collect some results, supplemented by some new tools (substitution rule and change of order of integration formula), that will be used in the remainder of this study.
Lemma 2.13. (See [26, Theorem 5.6 and Corollary 5.8]). If f,g∈Pω , then
f+g,f−g,⊖f,f⊕g∈Pω. |
Lemma 2.14. (See [26, Theorem 5.1]). f∈P1 iff μf is constant.
Lemma 2.15. (See [26, Lemma 4.6]). We have P1⊂Pω for all ω∈N .
Lemma 2.16. (See [26, Lemma 3.1]). We have the formula
μνΔ=μν. | (2.7) |
Theorem 2.17 (Chain Rule, Substitution Rule). For f:T→R , we have
FΔν=νΔfν−f,ifFν(t)=∫ν(t)tf(τ)Δτ. | (2.8) |
Moreover, if s,t∈T , then
∫ν(t)ν(s)f(τ)Δτ=∫tsνΔ(τ)f(ν(τ))Δτ. | (2.9) |
Proof. Equation (2.8) is the content of [26, Lemma 3.8]. Using (2.8), we get
∫ν(t)ν(s)f(τ)Δτ=Fν(t)−Fν(s)+∫tsf(τ)Δτ=∫ts(FΔν(τ)+f(τ))Δτ=∫tsνΔ(τ)fν(τ)Δτ, |
i.e., (2.9) holds.
Theorem 2.18. (See [26, Theorem 4.9]). If p∈Pω∩R and t,s∈T , then
ep(ν(t),t)=ep(ν(s),s)andep(ν(t),ν(s))=ep(t,s). | (2.10) |
To conclude this section, we include a result on how to change the order of integration in a double integral. The final formula in the following theorem will be needed in our proof of the second Cushing–Henson conjecture in Section 4, while the other formulas are included for future reference.
Theorem 2.19. Let f,g:T→R and a,b,c∈T . For Φ(t,s):=f(t)g(s) , we have
∫ba∫tcΦ(t,s)ΔsΔt=∫ba∫acΦ(t,s)ΔtΔs+∫ba∫bσ(s)Φ(t,s)ΔtΔs, | (2.11) |
∫ba∫ν(t)cΦ(t,s)ΔsΔt=∫ba∫ν(a)cΦ(t,s)ΔtΔs+∫ν(b)ν(a)∫bν−1(σ(s))Φ(t,s)ΔtΔs, | (2.12) |
∫ba∫ν(t)tΦ(t,s)ΔsΔt=∫ba∫ν(a)aΦ(t,s)ΔtΔs−∫ba∫bσ(s)Φ(t,s)ΔtΔs+∫ν(b)ν(a)∫bν−1(σ(s))Φ(t,s)ΔtΔs, | (2.13) |
∫ν(a)a∫ν(t)tΦ(t,s)ΔsΔt=∫ν(a)a∫σ(s)aΦ(t,s)ΔtΔs+∫ν(ν(a))ν(a)∫ν(a)ν−1(σ(s))Φ(t,s)ΔtΔs. | (2.14) |
Proof. In what follows, we use the notation
F(t):=−∫btf(s)Δs,G(t):=∫tcg(s)Δs,Gν(t):=∫ν(t)tg(s)Δs,ψ(t):=g(t)Fν−1(σ(t)), |
which imply
F(b)=0,FΔ=f,GΔ=g,GΔν(2.8)=νΔgν−g. |
First,
∫ba∫tcΦ(t,s)ΔsΔt(2.2)=∫baFΔ(t)G(t)Δt(2.2)=F(b)G(b)−F(a)G(a)−∫baF(σ(s))GΔ(s)Δs, |
so (2.11) holds. Next,
∫ba∫ν(t)tΦ(t,s)ΔsΔt(2.2)=∫baFΔ(t)Gν(t)Δt(2.2)=F(b)Gν(b)−F(a)Gν(a)−∫baF(σ(s))GΔν(s)Δs(2.2)=−F(a)Gν(a)−∫baνΔ(s)ψν(s)Δs−∫baF(σ(s))g(s)Δs(2.9)=−F(a)Gν(a)−∫ν(b)ν(a)ψ(s)Δs−∫baF(σ(s))g(s)Δs |
shows (2.13). Finally, (2.12) follows by adding (2.11) and (2.13), while (2.14) is the same as (2.13) with b=ν(a) .
Recall that throughout, ω∈N and ν=σω . Recall also that η>1 and
α:=η−1μη=−(⊖η−1μ),i.e.,η=1+μ(⊖(−α))=11−μα. | (3.1) |
Throughout the remainder of this paper, we assume
η:T→(1,∞),α,K:T→(0,∞),−α∈R+. |
We formulate some assumptions.
( A 1 ) (σΔη)ν=σΔη ,
( A 2 ) Kη−1∈Pω .
Theorem 3.1 (First Cushing–Henson Conjecture). Let t0∈T . Assume ( A 1 ) and ( A 2 ) . Define
λ:=νΔ(t0)e⊖(−α)(ν(t0),t0)−1. | (3.2) |
If λ≠0 , then (1.4) has a unique ω -periodic solution ˉx , given by
ˉx(t)=λ∫ν(t)te⊖(−α)(σ(s),t)α(s)K(s)Δs. | (3.3) |
If additionally, T is unbounded above, ∫∞t0α(s)Δs=∞ , and ˉx and K are bounded above, then ˉx is globally asymptotically stablefor solutions with positive initial conditions.
Given the structure of (1.3), we immediately obtain that solutions remain positive for positive initial conditions, i.e., for x0>0 , the solution x satisfies x(t)>0 for all t∈T , t≥t0 .
Before proving Theorem 3.1, we give a series of auxiliary results.
Lemma 3.2. Consider
( A 3 ) −α+1μσΔ∈Pω ,
( A 4 ) (⊖(−α)K)ν=⊖(−α)K .
( A 5 ) φ:=(−α)⊖μΔμ∈Pω .
Then ( A 1 ) holds iff ( A 3 ) holds iff ( A 5 ) holds, and ( A 2 ) holds iff ( A 4 ) holds.
Proof. The three calculations
μ{νΔ(−α+1μσΔ)ν−−α+1μσΔ}(2.7)=μ{νΔ(1−μαμσΔ)ν−1−μαμσΔ}(2.7)=(1−μασΔ)ν−1−μασΔ(3.1)=1(σΔη)ν−1σΔη, |
μ{νΔ(Kη−1)ν−Kη−1}(3.1)=μ{νΔ(Kμ(⊖(−α)))ν−Kμ(⊖(−α))}(2.7)=(K⊖(−α))ν−K⊖(−α), |
and (using 1+μΔ=σΔ and Lemmas 2.13, 2.14, and 2.15)
−α+1μσΔ−φ=1μ∈P1⊂Pω | (3.4) |
complete the proof.
Lemma 3.3. Let t,s∈T . For φ∈R defined in ( A 5 ) ,
eφ(t,s)=e−α(t,s)μ(s)μ(t) | (3.5) |
and
e⊖φ(ν(t),t)=νΔ(t)e⊖(−α)(ν(t),t). | (3.6) |
If ( A 5 ) holds, then
νΔ(t)e⊖(−α)(ν(t),t)=νΔ(s)e⊖(−α)(ν(s),s) | (3.7) |
and
νΔ(t)e⊖(−α)(ν(t),ν(s))=νΔ(s)e⊖(−α)(t,s). | (3.8) |
Proof. By Theorem 2.6 (part 3) and (2.3), we get (3.5). Using (3.5) with (2.7) yields (3.6). If φ∈Pω , then ⊖φ∈Pω by Lemma 2.13. Employing (3.5) and (3.6) together with (2.10) (applied to p=⊖φ ) implies (3.7). Theorem 2.6 and (3.7) result in (3.8).
Lemma 3.4. Assume ( A 3 ) and ( A 4 ) . Let t0∈T and define
Hν(t):=∫ν(t)th(s)Δswithh(t):=e⊖(−α)(σ(t),t0)α(t)K(t). | (3.9) |
Then, for λ defined in (3.2), we have
HΔν(t)=λh(t) | (3.10) |
and
Hνν(t)=(λ+1)Hν(t). | (3.11) |
Proof. First, by Theorem 2.6 (part 5), we have
h(t)=e⊖(−α)(t,t0)α(t)(1−μ(t)α(t))K(t)=(⊖(−α)K)(t)e⊖(−α)(t,t0), |
and hence
hν(t)(A4)=(⊖(−α)K)ν(t)e⊖(−α)(ν(t),t0)(A4)=(⊖(−α)K)(t)e⊖(−α)(ν(t),t0)=e⊖(−α)(ν(t),t)h(t). |
Thus,
HΔν(t)(2.8)=νΔ(t)hν(t)−h(t)(3.7)=(λ+1−1)h(t)=λh(t), |
which shows (3.10). Next,
Hνν(t)(3.10)=Hν(ν(t))=Hν(t)+∫ν(t)tHΔν(s)Δs(3.10)=Hν(t)+λ∫ν(t)th(s)Δs(3.10)=Hν(t)+λHν(t)=(λ+1)Hν(t) |
proves (3.11).
With Lemmas 3.3 and 3.4, we now have sufficient machinery to prove Theorem 3.1.
Proof of Theorem 3.1. In a first step, we show that ˉx given by (3.3) is an ω -periodic solution of (1.4). Note that
ˉx(t)(3.9)=λe−α(t,t0)Hν(t), |
and thus,
νΔ(t)ˉxν(t)(3.11)=λνΔ(t)e−α(ν(t),t0)Hνν(t)(3.11)=λνΔ(t)e−α(ν(t),t0)(λ+1)Hν(t)(3.11)=νΔ(t)e−α(ν(t),t)(λ+1)ˉx(t)(3.7)=ˉx(t) |
(use also (3.2) in the last equality), so ˉx is ω -periodic. With ˉu=1/ˉx , we get
ˉu(t)=1λe−α(t,t0)Hν(t), |
and thus,
ˉuΔ(t)(2.1)=1λ{e−α(σ(t),t0)HΔν(t)−α(t)e−α(t,t0)Hν(t)}(3.10)=e−α(σ(t),t0)h(t)−α(t)ˉu(t)(3.9)=α(t)K(t)−α(t)ˉu(t), |
so ˉu solves (1.5), and thus ˉx=1/ˉu solves (1.4). Altogether, ˉx is an ω -periodic solution of (1.4).
Conversely, we assume that ˜x is any ω -periodic solution of (1.4). Then ˜u=1/˜x satisfies (1.5), i.e., ˜uΔ(t)=−α(t)˜u(t)+α(t)K(t) . Hence,
νΔ(t)˜u(t)=νΔ(t)˜x(t)=1˜xν(t)=˜uν(t)=˜u(ν(t))(2.5)=e−α(ν(t),t)˜u(t)+∫ν(t)te−α(ν(t),σ(s))α(s)K(s)Δs=e−α(ν(t),t){˜u(t)+e−α(t,t0)∫ν(t)te⊖(−α)(σ(s),t0)α(s)K(s)Δs}=e−α(ν(t),t){˜u(t)+e−α(t,t0)Hν(t)} |
(note that (2.5) was applied with t0 replaced by t and t replaced by ν(t) ), so
(1+λ)˜u(t)(3.7)=νΔ(t)e⊖(−α)(ν(t),t)˜u(t)(3.7)=˜u(t)+e−α(t,t0)Hν(t), |
which, upon solving for ˜u(t) , results in
˜u(t)=e−α(t,t0)Hν(t)λ, |
i.e.,
˜x(t)=1˜u(t)=λe−α(t,t0)Hν(t)=ˉx(t). |
To prove the global asymptotic stability of ˉx , let x be the unique solution of (1.4) with initial condition x0>0 , and let ˉx0:=ˉx(t0) . Since 1/x solves (1.5), using (2.5), we get
x(t)=x0e−α(t,t0)(1+x0∫tt0h(s)Δs). |
Thus, we obtain
x(t)−ˉx(t)=x0e−α(t,t0)(1+x0∫tt0h(s)Δs)−ˉx0e−α(t,t0)(1+ˉx0∫tt0h(s)Δs)=x0−ˉx0e−α(t,t0)(1+x0∫tt0h(s)Δs)(1+ˉx0∫tt0h(s)Δs)=(x0−ˉx0)ˉx(t)ˉx0(1+x0∫tt0h(s)Δs), |
which tends to zero as t→∞ because α>0 and −α∈R+ so that
1+x0∫tt0h(s)Δs(2.4)≥1+x0‖ |
completing the proof.
Example 3.5. If \mathbb{T} = \mathbb{Z} , then
\begin{gather*} \sigma(t) = t+1, \quad \sigma ^{\Delta}(t) = 1, \quad \nu(t) = t+\omega, \quad \nu ^{\Delta}(t) = 1, \\ \mu(t) = 1, \quad \mu ^{\Delta}(t) = 0 \end{gather*} |
and, as noted in Example 2.11, periodicity defined in (2.6) is consistent with the classical periodicity definition, i.e., f is \omega -periodic if f(t+\omega) = f(t) for all t\in \mathbb{Z} . In this case, \rm ( A _{{1}} \rm ) states that
\eta(t+\omega) = \eta(t), |
and \rm ( A _{{2}} \rm ) says that
\frac{K(t+\omega)}{ \eta(t+\omega)-1} = \frac{K(t)}{ \eta(t)-1}. |
Together, \rm ( A _{{1}} \rm ) and \rm ( A _{{2}} \rm ) are equivalent to
\eta(t+\omega) = \eta(t) \quad{\rm{and}}\quad K(t+\omega) = K(t), |
i.e., both \eta and K are \omega -periodic. Next, \rm ( A _{{3}} \rm ) says that
- \alpha(t+\omega)+1 = - \alpha(t)+1, |
and \rm ( A _{{4}} \rm ) states that
\frac{ \alpha(t+\omega)}{1- \alpha(t+\omega)}\cdot\frac{1}{K(t+\omega)} = \frac{ \alpha(t)}{1- \alpha(t)}\cdot\frac{1}{K(t)}. |
Together, \rm ( A _{{3}} \rm ) and \rm ( A _{{4}} \rm ) are equivalent to
\alpha(t+\omega) = \alpha(t) \quad{\rm{and}}\quad K(t+\omega) = K(t), |
i.e., both \alpha and K are \omega -periodic. We also note that \varphi = (- \alpha)\ominus 0 = - \alpha , and so \varphi is \omega -periodic if and only if \alpha is \omega -periodic. If \alpha > 1 is constant and K is \omega -periodic, then \rm ( A _{{3}} \rm ) and \rm ( A _{{4}} \rm ) are satisfied, and \bar{x} from (3.3) is consistent with the unique \omega -periodic solution derived in [22]. In that case, \bar{x} and K are bounded, as any periodic function on \mathbb{Z} is bounded, and
\int_{t_0}^t \alpha\Delta s = \alpha(t-t_0)\to\infty \quad{\rm{as}}\quad t\to\infty. |
Hence, the \omega -periodic solution is globally asymptotically stable for solutions with positive initial conditions. The classical first Cushing–Henson conjecture is therefore a special case of Theorem 3.1. Further, Theorem 3.1 for \mathbb{T} = \mathbb{Z} also contains an extension of the classical first Cushing–Henson conjecture as presented in [21], where the authors considered both K and \alpha to be \omega -periodic. Again, in this case, all assumptions of Theorem 3.1 are satisfied, and the unique \omega -periodic solution is globally asymptotically stable.
Example 3.6. If \mathbb{T} = q^{ \mathbb{N}_0} , then
\begin{gather*} \sigma(t) = qt, \quad \sigma ^{\Delta}(t) = q, \quad \nu(t) = q^\omega t, \quad\nu ^{\Delta}(t) = q^\omega, \\ \mu(t) = (q-1)t, \quad \mu ^{\Delta}(t) = q-1 \end{gather*} |
and, as noted in Example 2.12, periodicity defined in (2.6) is consistent with the periodicity definition from [23], i.e., f is \omega -periodic if q^\omega f(q^\omega t) = f(t) for all t\in q^{ \mathbb{N}_0} . In this case, \rm ( A _{{1}} \rm ) states that
q \eta(q^\omega t) = q \eta(t), |
and \rm ( A _{{2}} \rm ) says that
q^\omega\frac{K(q^\omega t)}{ \eta(q^\omega t)-1} = \frac{K(t)}{ \eta(t)-1}. |
Together, \rm ( A _{{1}} \rm ) and \rm ( A _{{2}} \rm ) are equivalent to
\eta(q^\omega t) = \eta(t) \quad{\rm{and}}\quad q^\omega K(q^\omega t) = K(t), |
i.e., both \eta/\mu and K are \omega -periodic. Next, \rm ( A _{{3}} \rm ) says that
q^\omega\frac{- \alpha(q^\omega t)+\frac{1}{(q-1)q^\omega t}}{q} = \frac{- \alpha(t)+\frac{1}{(q-1)t}}{q}, |
and \rm ( A _{{4}} \rm ) states that
\frac{ \alpha(q^\omega t)}{1-(q-1)q^\omega t \alpha(q^\omega t)} \cdot\frac{1}{K(q^\omega t)} = \frac{ \alpha(t)}{1-(q-1)t \alpha(t)}\cdot\frac{1}{K(t)}. |
Together, \rm ( A _{{3}} \rm ) and \rm ( A _{{4}} \rm ) are equivalent to
q^\omega \alpha(q^\omega t) = \alpha(t) \quad{\rm{and}}\quad q^\omega K(q^\omega) = K(t), |
i.e., both \alpha and K are \omega -periodic. We also note that
\varphi(t) = \frac{- \alpha(t)-\frac{1}{t}}{1+\frac{(q-1)t}{t}} = -\frac{ \alpha(t)+\frac{1}{t}}{q}, |
and so \varphi is \omega -periodic if and only if \alpha is \omega -periodic. Since these assumptions coincide with the assumptions in [24], [24, Conjecture 1] is the same as Theorem 3.1 if \mathbb{T} = q^{ \mathbb{N}_0} . We would like to remind the reader that Theorem 3.1 is therefore also a generalization of [23, Conjecture 1] that assumes \alpha to be 1 -periodic.
Examples 3.5 and 3.6 show that assumptions \rm ( A _{{4}} \rm ) and \rm ( A _{{5}} \rm ) are equivalent to \alpha and K being \omega -periodic, in the sense of (2.6), if \mathbb{T} = \mathbb{Z} or \mathbb{T} = q^{ \mathbb{N}_0} . One might wonder for which other time scales this observation is true.
Theorem 3.7. Assume \mathbb{T} is such that
\begin{equation} \frac{\mu ^{\Delta}}{\mu}\in \mathcal{P}_\omega, \quad\mathit{{\rm{i.e., }}}\quad \mu^{\Delta\nu} = \mu ^{\Delta}. \end{equation} | (3.12) |
Then \rm ( A _{{4}} \rm ) and \rm ( A _{{5}} \rm ) hold if and only ifboth \alpha and K are \omega -periodic.
Proof. Assume (3.12). First, assuming \rm ( A _{{4}} \rm ) and \rm ( A _{{5}} \rm ) hold, we get from \rm ( A _{{5}} \rm ) that
- \alpha = \varphi\oplus\frac{\mu ^{\Delta}}{\mu}\in \mathcal{P}_\omega |
due to (3.12) and Lemma 2.13. Hence, using again Lemma 2.13, we obtain \alpha\in \mathcal{P}_\omega . Then, by \rm ( A _{{4}} \rm ) ,
\frac{\ominus(- \alpha)}{K} = \left({\frac{\ominus(- \alpha)}{K}}\right)^\nu = \frac{\nu ^{\Delta}}{\nu ^{\Delta}}\left({\frac{\ominus(- \alpha)}{K}}\right)^\nu = \frac{\nu ^{\Delta}(\ominus(- \alpha))^\nu}{\nu ^{\Delta} K^\nu} = \frac{\ominus(- \alpha)}{\nu ^{\Delta} K^\nu}, |
so K is \omega -periodic. Conversely, assuming both \alpha and K are \omega -periodic, we get by Lemma 2.13 that - \alpha\in \mathcal{P}_\omega , and hence
\varphi = (- \alpha)\ominus\frac{\mu ^{\Delta}}{\mu}\in \mathcal{P}_\omega |
due to (3.12) and Lemma 2.13. Hence, \rm ( A _{{5}} \rm ) holds. Moreover,
\left({\frac{\ominus(- \alpha)}{K}}\right)^\nu = \frac{\nu ^{\Delta}}{\nu ^{\Delta}}\left({\frac{\ominus(- \alpha)}{K}}\right)^\nu = \frac{\nu ^{\Delta}(\ominus(- \alpha))^\nu}{\nu ^{\Delta} K^\nu} = \frac{\ominus(- \alpha)}{K}, |
showing \rm ( A _{{4}} \rm ) .
However, if (3.12) does not hold, then, assuming \alpha and K are \omega -periodic instead of \rm ( A _{{4}} \rm ) and \rm ( A _{{5}} \rm ) , there does not even exist an \omega -periodic solution of (1.4) in general. This can be verified easily with \omega = 1 according to the following example.
Example 3.8. Let \omega = 1 and assume \alpha and K are \omega -periodic. Let \tilde{x} be an \omega -periodic solution of (1.4). Let \tilde{u} = 1/\tilde{x} . By Lemma 2.14, \tilde{c}: = \alpha/K is constant. Moreover, since \tilde{x} = 1/\tilde{u} satisfies (2.6), we get
\begin{equation} \sigma ^{\Delta}\tilde{u} = \tilde{u} ^{\sigma}. \end{equation} | (3.13) |
Thus,
\tilde{u} ^{\Delta} = \frac{\tilde{u} ^{\sigma}-\tilde{u}}{\mu} \stackrel{(3.13)}{ = } \frac{\sigma ^{\Delta}\tilde{u}-\tilde{u}}{\mu} = \frac{\mu ^{\Delta}}{\mu}\tilde{u}, |
and hence, due to
0\stackrel{(1.5)}{ = } \tilde{u} ^{\Delta}+ \alpha\tilde{u}-\tilde{c} = \left({\frac{\mu ^{\Delta}}{\mu}+ \alpha}\right)\tilde{u}-\tilde{c}, |
we obtain
\begin{equation} \tilde{u} = \frac{\mu\tilde{c}}{\mu ^{\Delta}+\mu \alpha}. \end{equation} | (3.14) |
Hence,
\tilde{u} ^{\sigma} \stackrel{(3.14)}{ = } \frac{\mu ^{\sigma}\tilde{c}}{\mu ^{\Delta\sigma}+\mu ^{\sigma} \alpha ^{\sigma}} \stackrel{(2.7)}{ = } \frac{\mu\sigma ^{\Delta}\tilde{c}}{\mu ^{\Delta\sigma}+\mu\sigma ^{\Delta} \alpha ^{\sigma}} \stackrel{(2.6)}{ = } \frac{\mu\sigma ^{\Delta}\tilde{c}}{\mu ^{\Delta\sigma}+\mu \alpha} |
and
\sigma ^{\Delta}\tilde{u} \stackrel{(3.14)}{ = } \frac{\mu\sigma ^{\Delta}\tilde{c}}{\mu ^{\Delta}+\mu \alpha}. |
Thus, with (3.13), we obtain (3.12).
Example 3.9. Let \omega = 4 . For q > 0 , consider
\mathbb{T} = \left\{{t_m:\;m\in \mathbb{N}_0}\right\}, \quad{\rm{where}}\quad t_m = \sum\limits_{i = 0}^{m-1}q^{(-1)^i} \quad{\rm{for}}\quad m\in \mathbb{N}_0, |
where the "empty sum" is by convention zero, i.e., t_0 = 0 . Then
\begin{gather*} \sigma(t_m) = t_{m+1}, \quad \mu(t_m) = t_{m+1}-t_m = q^{(-1)^m}, \\ \sigma ^{\Delta}(t_m) = \frac{\sigma(t_{m+1})-\sigma(t_m)}{\mu(t_m)} = \frac{\mu(t_{m+1})}{\mu(t_m)} = q^{2(-1)^{m+1}}, \\ \mu ^{\Delta}(t_m) = \sigma ^{\Delta}(t_m)-1 = q^{2(-1)^{m+1}}-1. \end{gather*} |
Hence (3.12) holds. Let K_0, K_1, K_2, K_3 > 0 , \bar{K}_i = K_{i\rm{mod}4} , a\neq 1 and define
\eta(t_m) = aq^{2(-1)^{m}}, \quad K(t_m) = q^{(-1)^{m+1}}\bar{K}_m. |
Clearly, K\in \mathcal{P}_\omega by design, and since \eta^\nu = \eta , we have \alpha\in \mathcal{P}_\omega . By Theorem 3.7, \rm ( A _{{3}} \rm ) and \rm ( A _{{5}} \rm ) hold, so that by Theorem 3.1, the unique 4 -periodic solution is given by
\bar{x}(t_m) = \frac{\lambda}{( \eta(t_m)-1) \frac{\bar{K}_m+\bar{K}_{m+2}}{\bar{K}_m\bar{K}_{m+2}} + \eta(t_m)( \eta(\sigma(t_m))-1) \frac{\bar{K}_{m+1}\;+\bar{K}_{m+3}}{\bar{K}_{m+1}\;\bar{K}_{m+3}}}, |
where \lambda = a^4-1\neq 0 .
We now bring our attention to the second Cushing–Henson conjecture, which reads for the Beverton–Holt difference equation as follows. If \eta > 1 is constant, K: \mathbb{Z}\to \mathbb{R}^+ is \omega -periodic, i.e., K(t+\omega) = K(t) for all t\in \mathbb{Z} , then the average of the unique periodic solution \bar{x} of (1.1) is less than or equal (equal iff K is constant) the average of the periodic carrying capacity, i.e.,
\frac{1}{\omega}\sum\limits_{t = 0}^{\omega-1}\bar{x}(t) \leq\frac{1}{\omega}\sum\limits_{t = 0}^{\omega-1}K(t). |
Biologically, this inequality is interpreted as deleterious effect of a periodic environment to the population. In order to extend this result to isolated time scales, we aim to find an upper bound for the average of the unique periodic solution. Similar to the discrete case, where the second Cushing–Henson conjecture assumed a constant proliferation rate, we adjust \rm ( A _{{1}} \rm ) accordingly for \omega = 1 . More specifically, we consider the assumptions
\rm ( A _{{6}} \rm ) \left({\sigma ^{\Delta} \eta}\right)^{\sigma} = \sigma ^{\Delta} \eta , \rm ( A _{{7}} \rm ) \frac{- \alpha+\frac{1}{\mu}}{\sigma ^{\Delta}}\in \mathcal{P}_1 , \rm ( A _{{8}} \rm ) \varphi = (- \alpha)\ominus\frac{\mu ^{\Delta}}{\mu}\in \mathcal{P}_1 .
Remark 4.1. According to Lemma 3.2, we have
\left(\mathrm{A}_{6}\right) \text { holds iff }\left(\mathrm{A}_{7}\right) \text { holds iff }\left(\mathrm{A}_{8}\right) \text { holds, } |
and, by Lemma 2.15,
\text { any of }\left(\mathrm{A}_{6}\right),\left(\mathrm{A}_{7}\right),\left(\mathrm{A}_{8}\right) \text { implies any of }\left(\mathrm{A}_{1}\right),\left(\mathrm{A}_{3}\right),\left(\mathrm{A}_{5}\right) \text {. } |
Assume now any of the conditions \rm ( A _{{6}} \rm ) , \rm ( A _{{7}} \rm ) , and \rm ( A _{{8}} \rm ) . By Remark 2.14,
\begin{equation} C: = \mu\frac{- \alpha+\frac{1}{\mu}}{\sigma ^{\Delta}} = \frac{1-\mu \alpha}{\sigma ^{\Delta}} \quad\rm{{is \; constant.}}\quad \end{equation} | (4.1) |
Moreover, due to (3.4), we have
\begin{equation} \mu \varphi = C-1 \quad{\rm{{and\; thus}}}\quad \mu(\ominus \varphi) = \frac{-\mu \varphi}{1+\mu \varphi} = \frac{1-C}{C} = :D. \end{equation} | (4.2) |
Because of
e_{\ominus \varphi}(\nu(t_0), t_0) = \prod\limits_{\tau\in[t_0, \nu(t_0))\cap \mathbb{T}}\left({1+\mu(\tau)(\ominus \varphi)(\tau)}\right) = \left({1+\frac{1-C}{C}}\right)^\omega = \frac{1}{C^\omega}, |
and thus, we get that
\lambda\stackrel{(3.2)}{ = }\nu ^{\Delta}(t_0)e_ {\ominus(-\alpha)}(\nu(t_0), t_0)-1 \stackrel{(3.6)}{ = }e_{\ominus \varphi}(\nu(t_0), t_0)-1 = \frac{1}{C^\omega}-1. |
Theorem 4.2 (Second Cushing–Henson Conjecture). Assume \rm ( A _{{4}} \rm ) , \rm ( A _{{8}} \rm ) , and C<1 , where C is defined in (4.1). Then the average of the unique \omega -periodic solution \bar{x} of (1.4) is bounded above by
\begin{equation} \frac{1}{\omega}\int_{t_0}^{\nu(t_0)}\bar{x}(t)\Delta t \leq\frac{1}{\omega}\int_{t_0}^{\nu(t_0)} \frac{1-C}{\frac{1}{\sigma ^{\Delta}(t)}-C}K(t)\Delta t, \end{equation} | (4.3) |
and equality holds iff \frac{K}{ {\ominus(-\alpha)}} is constant.
The central tool in the proof of Theorem 4.2 is the following generalized Jensen inequality from [32, Theorem 2.2], which reads for the strictly convex function 1/z as follows:
\begin{equation} \frac{ \int_a^bw(s)\Delta s} { \int_a^bw(s)v(s)\Delta s} \leq\frac{ \int_a^b\frac{w(s)}{v(s)}\Delta s} { \int_a^bw(s)\Delta s} \quad{\rm{with}}\quad w > 0. \end{equation} | (4.4) |
We apply (4.4) with
w_t(s): = - \varphi(s)e_ \varphi(t, \sigma(s)) > 0 \quad{\rm{and}}\quad v_t(s): = \mu(t)\frac{\tilde{\beta}(s)}{(\ominus \varphi)(s)}, |
where we also put
\tilde{\beta}: = \frac{ {\ominus(-\alpha)}}{\mu K} \quad{\rm{and}}\quad \beta: = \frac{1}{\mu^2\tilde{\beta}} = \frac{K}{\mu( {\ominus(-\alpha)})}. |
Note that \rm ( A _{{4}} \rm ) implies (use (2.7))
\begin{equation} \beta, \tilde{\beta}\in \mathcal{P}_\omega. \end{equation} | (4.5) |
Before proving Theorem 4.2, we offer the following auxiliary result.
Lemma 4.3. Assume \rm ( A _{{4}} \rm ) and \rm ( A _{{5}} \rm ) . Define \lambda by (3.2). We have
\begin{equation} \int_t^{\nu(t)}w_t(s)\Delta s = \lambda \end{equation} | (4.6) |
and
\begin{equation} w_t(s)v_t(s) = e_{ {\ominus(-\alpha)}}(\sigma(s), t)\frac{ \alpha(s)}{K(s)}. \end{equation} | (4.7) |
Moreover, if \rm ( A _{{8}} \rm ) holds, then
\begin{equation} \frac{w_t(s)}{v_t(s)} = -D\beta(s) \varphi(t)e_ \varphi(t, \sigma(s)), \end{equation} | (4.8) |
where D is defined in (4.2).
Proof. First, we use Theorem 2.6 (part 8) to integrate
\int_t^{\nu(t)}w_t(s)\Delta s = -\int_t^{\nu(t)} \varphi(s)e_ \varphi(t, \sigma(s))\Delta s = e_ \varphi(t, \nu(t))-1 = \lambda, |
where we also used (3.6), (3.7), and (3.2). This proves (4.6). Next, using (3.5), we get
\begin{align*} w_t(s)v_t(s) & = - \varphi(s)e_ \varphi(t, \sigma(s)) \mu(t)\frac{\tilde{\beta}(s)}{(\ominus \varphi)(s)}\\ & = e_ \varphi(t, s)\mu(t)\tilde{\beta}(s) = e_{- \alpha}(t, s)\mu(s)\tilde{\beta}(s)\\ & = e_{- \alpha}(t, s)\frac{( {\ominus(-\alpha)})(s)}{K(s)} = e_{ {\ominus(-\alpha)}}(\sigma(s), t)\frac{ \alpha(s)}{K(s)}, \end{align*} |
which shows (4.7). Finally, assuming \rm ( A _{{8}} \rm ) , we have (4.1) and (4.2). Then
\begin{align*} \frac{w_t(s)}{v_t(s)} & = \frac{- \varphi(s)e_ \varphi(t, \sigma(s))} {\mu(t)\frac{\tilde{\beta}(s)}{(\ominus \varphi)(s)}} = \frac{- \varphi(s)(\ominus \varphi)(s)}{\mu(t)}\mu^2(s)\beta(s)e_ \varphi(t, \sigma(s))\\ & = -\mu(s) \varphi(s)\mu(s)(\ominus \varphi)(s)\beta(s) \frac{e_ \varphi(t, \sigma(s))}{\mu(t)}\\ & = -\mu(t) \varphi(t)\mu(s)(\ominus \varphi)(s)\beta(s) \frac{e_ \varphi(t, \sigma(s))}{\mu(t)}\\ & = -\mu(s)(\ominus \varphi)(s)\beta(s) \varphi(t)e_ \varphi(t, \sigma(s))\\ & = -D\beta(s) \varphi(t)e_ \varphi(t, \sigma(s)) \end{align*} |
shows (4.8).
We can now bring our attention to the proof of the second Cushing–Henson conjecture on isolated time scales.
Proof of Theorem 4.2. We apply the generalized Jensen inequality (4.4) on time scales in the single forthcoming calculation to estimate
\begin{array}{l} {\int_{t_0}^{\nu(t_0)}\bar{x}(t)\Delta t \stackrel{(3.3)}{ = } \int_{t_0}^{\nu(t_0)} \frac{\lambda} { \int_t^{\nu(t)}e_{ {\ominus(-\alpha)}}(\sigma(s), t) \frac{ \alpha(s)}{K(s)}\Delta s}\Delta t}\\ \stackrel{(4.6)}{ = } \int_{t_0}^{\nu(t_0)} \frac{ \int_t^{\nu(t)}w_t(s)\Delta s} { \int_t^{\nu(t)}e_{ {\ominus(-\alpha)}}(\sigma(s), t) \frac{ \alpha(s)}{K(s)}\Delta s}\Delta t \stackrel{(4.7)}{ = } \int_{t_0}^{\nu(t_0)} \frac{ \int_t^{\nu(t)}w_t(s)\Delta s} { \int_t^{\nu(t)}w_t(s)v_t(s)\Delta s}\Delta t\\ \stackrel{(4.4)}{\leq} \int_{t_0}^{\nu(t_0)} \frac{ \int_t^{\nu(t)}\frac{w_t(s)}{v_t(s)}\Delta s} { \int_t^{\nu(t)}w_t(s)\Delta s}\Delta t \stackrel{(4.6)}{ = } \frac{1}{\lambda} \int_{t_0}^{\nu(t_0)}\int_t^{\nu(t)}\frac{w_t(s)}{v_t(s)}\Delta s\Delta t\\ \stackrel{(2.14)}{ = } \frac{1}{\lambda} \left\{{\int_{t_0}^{\nu(t_0)}\int_{t_0}^{\sigma(s)} \frac{w_t(s)}{v_t(s)}\Delta t\Delta s +\int_{\nu(t_0)}^{\nu(\nu(t_0))}\int_{\nu^{-1}(\sigma(s))}^{\nu(t_0)} \frac{w_t(s)}{v_t(s)}\Delta t\Delta s}\right\}\\ \stackrel{(4.8)}{ = } \frac{-D}{\lambda}\left\{ \int_{t_0}^{\nu(t_0)}\beta(s)\int_{t_0}^{\sigma(s)} \varphi(t)e_ \varphi(t, \sigma(s))\Delta t\Delta s\right.\\ \left.+\int_{\nu(t_0)}^{\nu(\nu(t_0))} \beta(s)\int_{\nu^{-1}(\sigma(s))}^{\nu(t_0)} \varphi(t)e_ \varphi(t, \sigma(s))\Delta t\Delta s\right\}\\ \stackrel{\phantom{(4.8)}}{ = } \frac{D}{\lambda}\left\{ \int_{t_0}^{\nu(t_0)} \beta(s)\left({e_ \varphi(t_0, \sigma(s))-1}\right)\Delta s\right.\\ \left.+\int_{\nu(t_0)}^{\nu(\nu(t_0))} \beta(s)\left({e_ \varphi(\nu^{-1}(\sigma(s)), \sigma(s)) -e_ \varphi(\nu(t_0), \sigma(s))}\right)\Delta s\right\}\\ \stackrel{(2.9)}{ = } \frac{D}{\lambda}\left\{ \int_{t_0}^{\nu(t_0)} \beta(s)\left({e_ \varphi(t_0, \sigma(s))-1}\right)\Delta s\right.\\ \left.+\int_{t_0}^{\nu(t_0)}\nu ^{\Delta}(s)\beta^\nu(s) \left({e_ \varphi(\sigma(s), \sigma(\nu(s))) -e_ \varphi(\nu(t_0), \sigma(\nu(s)))}\right)\Delta s\right\}\\ \stackrel{(2.10)}{ = } \frac{D}{\lambda}\left\{ \int_{t_0}^{\nu(t_0)} \beta(s)\left({e_ \varphi(t_0, \sigma(s))-1}\right)\Delta s\right.\\ \left.+\int_{t_0}^{\nu(t_0)}\nu ^{\Delta}(s)\beta^\nu(s) \left({e_ \varphi(t_0, \nu(t_0)) -e_ \varphi(t_0, \sigma(s))}\right)\Delta s\right\}\\ \stackrel{(4.5)}{ = } \frac{D}{\lambda}\left\{ \int_{t_0}^{\nu(t_0)}\beta(s) \left({e_ \varphi(t_0, \sigma(s))-1}\right)\Delta s\right.\\ \left.+\int_{t_0}^{\nu(t_0)}\beta(s) \left({e_ \varphi(t_0, \nu(t_0)) -e_ \varphi(t_0, \sigma(s))}\right)\Delta s\right\}\\ \stackrel{\phantom{(4.8)}}{ = } \frac{D}{\lambda} \int_{t_0}^{\nu(t_0)}\beta(s) \left({e_ \varphi(t_0, \nu(t_0))-1}\right)\Delta s\\ \stackrel{(3.6)}{ = } D\int_{t_0}^{\nu(t_0)}\beta(s)\Delta s = \int_{t_0}^{\nu(t_0)} \frac{1-C}{\frac{1}{\sigma ^{\Delta}(s)}-C}K(s)\Delta s, \end{array} |
where the last equality holds because
D\beta(s) \stackrel{(4.2)}{ = } \frac{(1-C)(1-\mu(s) \alpha(s))K(s)}{C\mu(s) \alpha(s)} \stackrel{(4.1)}{ = } \frac{(1-C)\sigma ^{\Delta} C}{C(1-\sigma ^{\Delta} C)}. |
Note also, that due to the strict convexity of 1/z (see [32, Theorem 2.2]), equality holds in (4.4) if and only if v_t(s) is independent of s , which means \beta\in \mathcal{P}_1 .
Remark 4.4. If \rm ( A _{{4}} \rm ) and \beta\in \mathcal{P}_1 hold, then the unique periodic solution of (1.4) is D\beta , where D is given by (4.2). In detail, the unique periodic solution is
\begin{equation} \frac{cD}{\mu(t)}, \quad{\rm{where}}\quad c: = \left({\frac{K}{ {\ominus(-\alpha)}}}\right)(t) \quad{\rm{{is\; constant.}}}\quad \end{equation} | (4.9) |
We can check (4.9) in two simple ways, namely by calculating it from (3.3), i.e.,
\begin{align*} \bar{x}(t) \stackrel{(3.3)}{ = }& \frac{\lambda} { \int_t^{\nu(t)}e_ {\ominus(-\alpha)}(\sigma(s), t) \frac{ \alpha(s)}{K(s)}\Delta s}\\ \stackrel{\phantom{(3.3)}}{ = }& \frac{\lambda} { \int_t^{\nu(t)}e_{- \alpha}(t, s) \left({\frac{ {\ominus(-\alpha)}}{K}}\right)(s)\Delta s} = \frac{\lambda c} { \int_t^{\nu(t)}e_{- \alpha}(t, s)\Delta s}\\ \stackrel{(3.5)}{ = }& \frac{\lambda c} { \int_t^{\nu(t)}e_ \varphi(t, s)\frac{\mu(t)}{\mu(s)}\Delta s} \stackrel{\phantom{(3.5)}}{ = } \frac{\lambda c} { \int_t^{\nu(t)}e_ \varphi(t, s) \frac{\mu(t)(\ominus \varphi)(s)}{\mu(s)(\ominus \varphi)(s)}\Delta s}\\ \stackrel{(4.2)}{ = }& \frac{\lambda cD} {\mu(t) \int_t^{\nu(t)}e_{\ominus \varphi}(s, t)(\ominus \varphi)(s)\Delta s} \stackrel{(3.6)}{ = } \frac{cD}{\mu(t)}, \end{align*} |
or by directly checking that it is 1 -periodic (this is clear from Lemma 2.14) and verifying that it solves (1.4), i.e.,
\begin{align*} & \alpha\left({\frac{cD}{\mu}}\right) ^{\sigma}\left({1-\frac{cD}{\mu K}}\right) -\left({\frac{cD}{\mu}}\right) ^{\Delta}\\ & = \frac{ \alpha cD}{\mu ^{\sigma}}\left({1-\frac{D(1-\mu \alpha)}{\mu \alpha}}\right) +\frac{cD\mu ^{\Delta}}{\mu\mu ^{\sigma}}\\ & = \frac{cD}{\mu\mu ^{\sigma}}\left({\mu \alpha-D(1-\mu \alpha)+\sigma ^{\Delta}-1}\right)\\ & = \frac{cD}{\mu\mu ^{\sigma}}\left({(\mu \alpha-1)(1+D)+\sigma ^{\Delta}}\right)\\ & = \frac{cD}{\mu\mu ^{\sigma}}\left({-C(1+D)+1}\right)\sigma ^{\Delta} \stackrel{(4.2)}{ = }0. \end{align*} |
It can also be verified easily that for (4.9), the inequality (4.3) becomes an equality with cD on both sides:
\frac{1}{\omega}\int_{t_0}^{\nu(t_0)}\frac{cD}{\mu(t)}\Delta t = cD |
and
\begin{align*} &\frac{1}{\omega}\int_{t_0}^{\nu(t_0)} \frac{1-C}{\frac{1}{\sigma ^{\Delta}(t)}-C}K(t)\Delta t\\ & = \frac{1}{\omega}\int_{t_0}^{\nu(t_0)} \frac{(1-C)c \alpha(t)} {\left({\frac{1}{\sigma ^{\Delta}(t)}-C}\right)(1-\mu(t) \alpha(t))}\Delta t\\ & = \frac{1}{\omega}\int_{t_0}^{\nu(t_0)} \frac{(1-C)c \alpha(t)} {\left({\frac{1}{\sigma ^{\Delta}(t)}-C}\right)C\sigma ^{\Delta}}\Delta t = \frac{1}{\omega}\int_{t_0}^{\nu(t_0)} \frac{(1-C)c \alpha(t)}{C-C^2\sigma ^{\Delta}(t)}\Delta t\\ & = \frac{1}{\omega}\int_{t_0}^{\nu(t_0)} \frac{(1-C)c \alpha(t)}{C\mu(t) \alpha(t)}\Delta t = \frac{cD}{\omega}\int_{t_0}^{\nu(t_0)}\frac{\Delta t}{\mu(t)} = cD. \end{align*} |
Remark 4.5. For all isolated time scales such that \sigma ^{\Delta} = c is constant for some c\in \mathbb{R} , Theorem 4.2 implies that the average of the unique periodic solution is less than or equal to the average of the carrying capacity multiplied by the constant \frac{ \eta c-1}{ \eta-1} . If \mathbb{T} = \mathbb{Z} , then c = 1 , and the classical second Cushing–Henson conjecture is retrieved. If \mathbb{T} = q^{ \mathbb{N}_0} , then c = q , and Theorem 4.2 is consistent with the second Cushing–Henson conjecture formulated in [23], see also (1.2). The inequality (4.3) reveals that the upper bound is increasing in \sigma ^{\Delta} . Since
\frac{1-C}{\frac{1}{\sigma ^{\Delta}}-C} = 1+\frac{1-\frac{1}{\sigma ^{\Delta}}}{\frac{1}{\sigma ^{\Delta}}-C} = 1+\frac{\sigma ^{\Delta}-1}{1-C\sigma ^{\Delta}} \stackrel{(4.1)}{ = } 1+\frac{\mu ^{\Delta}}{\mu \alpha}, |
we can write (4.3) also as
\int_{t_0}^{\nu(t_0)}\bar{x}(t)\Delta t \leq\int_{t_0}^{\nu(t_0)} \left({1+\frac{\mu ^{\Delta}(t)}{\mu(t) \alpha(s)}}\right)K(t)\Delta t. |
Hence if \mu ^{\Delta}(s) < 0 for all s\in[t_0, \nu(t_0)) , then the upper bound for the average periodic solution is smaller than the average of the carrying capacities, suggesting an even stronger negative effect of periodicity onto the population compared to the classical case. If \mu ^{\Delta}(s) > 0 for all s\in[t_0, \nu(t_0)) , then the second Cushing–Henson conjecture does not necessarily hold as the multiplicative factor exceeds one. The inequality (4.3) exposes the effects of the time structure onto the upper bound of the mean periodic population.
In this work, we studied the Beverton–Holt model on arbitrary isolated time scales with time-dependent coefficients. Using the recently formulated periodicity concept for isolated time scales allows to address the Cushing–Henson conjectures for nonperiodic time scales. After an introduction in Section 1 and some preliminaries in Section 2, in Section 3, we provided conditions for the existence and uniqueness of a globally asymptotically stable periodic solution. This generalizes the first Cushing–Henson conjecture to an arbitrary isolated time scale. The provided theorem, when applied to the special case of the discrete time domain \mathbb{Z} , coincides with results in existing literature. The presented conditions for existence and uniqueness of the periodic solution and its global asymptotic stability are equivalent to the conditions in the first conjecture presented in [21]. It therefore generalizes the classical formulation of the first Cushing–Henson conjecture. We also showed that our result is consistent [24, Conjecture 1] in the special case of a quantum time scale. A special subcase in this time scale was discussed in [23]. As we outlined, Theorem 3.1 contains these works as special cases. In Section 4, we focused on the discussion of the second Cushing–Henson conjecture on arbitrary isolated time scales. In the classical case, when \mathbb{T} = \mathbb{Z} , the conjecture concerns the effects of a periodic environment under constant proliferation rate, mathematically formulated by an upper bound of the average periodic solution. The derived upper bound in Theorem 4.2 is, in contrast to the classical case, a weighted average dependent on changes of the time scale. This highlights that the second Cushing–Henson conjecture does not necessarily hold in general, and its statement depends on the change of the time scale. If the time scale changes with a constant rate, that is, \sigma ^{\Delta} is constant, then the average of the periodic solution is bounded by a factor times the average of the carrying capacity. Examples of this special case contain the discrete and the quantum time scale. For both of these time scales, Theorem 4.2 is consistent with existing formulations of the second Cushing–Henson conjecture in [18,21,23]. Our results complement work in [22], where the authors consider the Cushing–Henson conjectures for the Beverton–Holt model on periodic time scales. In contrast to isolated time scales, \omega -periodic time scales assume t+\omega\in \mathbb{T} for all t\in \mathbb{T} . Since periodic and isolated time scales intersect but neither is a subset of the other, modifications of the conjectures remain unknown on arbitrary time scales. We highlight that this work is an application of the new definition of periodicity on isolated time scales, defined in [26]. The introduced method of the application of periodicity on isolated time scales can now be extended to other models, such as delay Beverton–Holt models. In fact, in [33,34,35], the Cushing–Henson conjectures for different delay Beverton–Holt models in the discrete case are discussed. The tools used in this paper can be used to establish these results on an arbitrary isolated time scale.
The authors would like to thank the three anonymous referees and the handling editor for many useful comments and suggestions, leading to a substantial improvement of the presentation of this article. The second author acknowledges partial support by the project UnB DPI/DPG - 03/2020 and CNPq grant 307582/2018-3.
The authors declare there is no conflict of interest.
[1] | R. J. H. Beverton, S. J. Holt, On the dynamics of exploited fish populations, volume 19 of Fishery investigations (Great Britain, Ministry of Agriculture, Fisheries, and Food), H. M. Stationery Off., London, 1957. |
[2] | E. C. Pielou, An Introduction to Mathematical Ecology, Wiley-Interscience, 1969. |
[3] |
A. S. Al-Ghassani, Z. AlSharawi, The effect of maps permutation on the global attractor of a periodic Beverton–Holt model, Appl. Math. Comput., 370 (2018), 124905. https://doi.org/10.1016/j.amc.2019.124905 doi: 10.1016/j.amc.2019.124905
![]() |
[4] | E. J. Bertrand, M. R. S. Kulenović, Global dynamics of higher-order transcendental-type generalized Beverton–Holt equations, Int. J. Difference Equ., 13 (2018), 71–84. |
[5] |
P. H. Bezandry, Almost periodic stochastic Beverton–Holt difference equation with higher delays and with competition between overlapping generations, Nonauton. Dyn. Syst., 7 (2020), 118–125. https://doi.org/10.1515/msds-2020-0105 doi: 10.1515/msds-2020-0105
![]() |
[6] |
J. A. Filar, Z. H. Qiao, S. Streipert, Risk sensitivity in Beverton–Holt fishery with multiplicative harvest, Nat. Resour. Model., 33 (2020), e12257. https://doi.org/10.1111/nrm.12257 doi: 10.1111/nrm.12257
![]() |
[7] | T. Khyat, M. R. S. Kulenović, E, Pilav, The invariant curve caused by Neimark–Sacker bifurcation of a perturbed Beverton–Holt difference equation, Int. J. Difference Equ., 124 (2017), 267–280. |
[8] | T. Khyat, M. R. S. Kulenović, Global dynamics of delayed sigmoid Beverton–Holt equation, Discrete Dyn. Nat. Soc., (2020), 1–15, Art. ID 1364282. https://doi.org/10.1155/2020/1364282 |
[9] |
M. R. S. Kulenović, S. Moranjkić, Z. Nurkanović. Global dynamics and bifurcation of a perturbed sigmoid Beverton–Holt difference equation, Math. Methods Appl. Sci., 39 (2016), 2696–2715. https://doi.org/10.1002/mma.3722 doi: 10.1002/mma.3722
![]() |
[10] |
Y. Li, J. Li, Discrete-time models for releases of sterile mosquitoes with Beverton–Holt-type of survivability, Ric. Mat., 67 (2018), 141–162. https://doi.org/10.1007/s11587-018-0361-4 doi: 10.1007/s11587-018-0361-4
![]() |
[11] |
Y. Li, J. Li, Stage-structured discrete-time models for interacting wild and sterile mosquitoes with Beverton–Holt survivability. Math. Biosci. Eng., 16 (2019), 572–602. https://doi.org/10.3934/mbe.2019028 doi: 10.3934/mbe.2019028
![]() |
[12] |
T. Diagana, Almost automorphic solutions to a Beverton–Holt dynamic equation with survival rate, Appl. Math. Lett., 36 (2014), 19–24. https://doi.org/10.1016/j.aml.2014.04.011 doi: 10.1016/j.aml.2014.04.011
![]() |
[13] | S. Frassu, T. X. Li, G. Viglialoro, Improvements and generalizations of results concerning attraction-repulsion chemotaxis models, Math. Methods Appl. Sci., Published online 1 June 2022. https://doi.org/10.1002/mma.8437 |
[14] |
S. Frassu, G, Viglialoro, Boundedness criteria for a class of indirect (and direct) chemotaxis-consumption models in high dimensions, Appl. Math. Lett., 132 (2022), 108108. https://doi.org/10.1016/j.aml.2022.108108 doi: 10.1016/j.aml.2022.108108
![]() |
[15] |
T. X. Li, N. Pintus, G. Viglialoro, Properties of solutions to porous medium problems with different sources and boundary conditions, Z. Angew. Math. Phys., 70 (2019), 1–18. https://doi.org/10.1007/s00033-018-1046-2 doi: 10.1007/s00033-018-1046-2
![]() |
[16] | T. X. Li, G. Viglialoro, Boundedness for a nonlocal reaction chemotaxis model even in the attraction-dominated regime, Differ. Integral Equat., 34 (2021), 315–336. |
[17] |
J. M. Cushing, S. M. Henson, Global dynamics of some periodically forced, monotone difference equations, J. Differ. Equ. Appl., 7 (2001), 859–872. On the occasion of the 60th birthday of Calvin Ahlbrandt. https://doi.org/10.1080/10236190108808308 doi: 10.1080/10236190108808308
![]() |
[18] |
J. M. Cushing, S. M. Henson, A periodically forced Beverton–Holt equation, J. Difference Equ. Appl., 8 (2002), 1119–1120. https://doi.org/10.1080/1023619021000053980 doi: 10.1080/1023619021000053980
![]() |
[19] | S. Stević, A short proof of the Cushing–Henson conjecture, Discrete Dyn. Nat. Soc., (2006), pages 1–5, Art. ID 37264. https://doi.org/10.1155/DDNS/2006/37264 |
[20] |
S. Elaydi, R. J. Sacker, Periodic difference equations, population biology and the Cushing-Henson conjectures, Math. Biosci., 201 (2006), 195–207. https://doi.org/10.1016/j.mbs.2005.12.021 doi: 10.1016/j.mbs.2005.12.021
![]() |
[21] |
M. Bohner, S. Streipert, The Beverton-Holt equation with periodic growth rate, Int. J. Math. Comput., 26 (2015), 1–10. https://doi.org/10.1007/978-3-319-24747-2_1 doi: 10.1007/978-3-319-24747-2_1
![]() |
[22] |
M. Bohner, H. Warth, The Beverton–Holt dynamic equation, Appl. Anal., 86 (2007), 1007–1015. https://doi.org/10.1080/00036810701474140 doi: 10.1080/00036810701474140
![]() |
[23] |
M. Bohner, R. Chieochan, The Beverton–Holt q -difference equation, J. Biol. Dyn., 7 (2013), 86–95. https://doi.org/10.1080/17513758.2013.804599 doi: 10.1080/17513758.2013.804599
![]() |
[24] | M. Bohner, S. Streipert, The Beverton–Holt q -difference equation with periodic growth rate, In Difference equations, discrete dynamical systems, and applications, Springer-Verlag, Berlin-Heidelberg-New York, 2015, 3–14. https://doi.org/10.1007/978-3-319-24747-2_1 |
[25] | M. Bohner, A. Peterson, Dynamic equations on time scales. Birkhäuser Boston, Inc., Boston, MA, 2001, An introduction with applications. https://doi.org/10.1007/978-1-4612-0201-1 |
[26] |
M. Bohner, J. G. Mesquita, S, Streipert, Periodicity on isolated time scales, Math. Nachr., 295 (2022), 259–280. https://doi.org/10.1002/mana.201900360 doi: 10.1002/mana.201900360
![]() |
[27] |
M. Bohner, T. Cuchta, S. Streipert, Delay dynamic equations on isolated time scales and the relevance of one-periodic coefficients, Math. Methods Appl. Sci., 45 (2022), 5821–5838. https://doi.org/10.1002/mma.8141 doi: 10.1002/mma.8141
![]() |
[28] | M. Bohner, A. Peterson, Advances in dynamic equations on time scales, Birkhäuser Boston, Inc., Boston, MA, 2003. https://doi.org/10.1007/978-0-8176-8230-9 |
[29] |
B. Jackson, The time scale logarithm, Appl. Math. Lett., 21 (2008), 215–221. https://doi.org/10.1016/j.aml.2007.02.020 doi: 10.1016/j.aml.2007.02.020
![]() |
[30] |
M. Bohner, The logarithm on time scales, J. Difference Equ. Appl., 11 (2005), 1305–1306. https://doi.org/10.1080/10236190500376284 doi: 10.1080/10236190500376284
![]() |
[31] | M. Bohner, Some oscillation criteria for first order delay dynamic equations, Far East J. Appl. Math., 18 (2005), 289–304. |
[32] | F. H. Wong, C. C. Yeh, W. C. Lian, An extension of Jensen's inequality on time scales, Adv. Dyn. Syst. Appl., 1 (2006), 113–120. |
[33] |
M. Bohner, F. M. Dannan, S. Streipert, A nonautonomous Beverton–Holt equation of higher order, J. Math. Anal. Appl., 457 (2018), 114–133. https://doi.org/10.1016/j.jmaa.2017.07.051 doi: 10.1016/j.jmaa.2017.07.051
![]() |
[34] |
V. L. Kocic, A note on the nonautonomous delay Beverton–Holt model, J. Biol. Dyn., 4 (2010), 131–139. https://doi.org/10.1080/17513750902803588 doi: 10.1080/17513750902803588
![]() |
[35] |
E. Braverman, S. H. Saker, On the Cushing–Henson conjecture, delay difference equations and attenuant cycles, J. Difference Equ. Appl., 14 (2008), 275–286. https://doi.org/10.1080/10236190701565511 doi: 10.1080/10236190701565511
![]() |
1. | Martin Bohner, Jaqueline Mesquita, Sabrina Streipert, Generalized periodicity and applications to logistic growth, 2024, 186, 09600779, 115139, 10.1016/j.chaos.2024.115139 | |
2. | Manuel De la Sen, Santiago Alonso-Quesada, Asier Ibeas, Aitor J. Garrido, Ewa Pawluszewicz, On an Extended Time-Varying Beverton–Holt Equation Subject to Harvesting Monitoring and Population Excess Penalty, 2023, 2023, 1607-887X, 1, 10.1155/2023/5052799 | |
3. | Manuel De la Sen, Santiago Alonso-Quesada, Asier Ibeas, Aitor J. Garrido, A. E. Matouk, On a Coupled Time-Varying Beverton–Holt Model with Two Habitats Subject to Harvesting, Repopulation, and Mixed Migratory Flows of Populations, 2023, 2023, 1607-887X, 1, 10.1155/2023/6050789 | |
4. | Mariem Mohamed Abdelahi, Mohamed Ahmed Sambe, Elkhomeini Moulay Ely, Existence results for some generalized Sigmoid Beverton-Holt models in time scales, 2023, 10, 2353-0626, 10.1515/msds-2022-0166 | |
5. | Tri Truong, Martin Bohner, Ewa Girejko, Agnieszka B. Malinowska, Ngo Van Hoa, Granular fuzzy calculus on time scales and its applications to fuzzy dynamic equations, 2025, 690, 00200255, 121547, 10.1016/j.ins.2024.121547 |