
Citation: Filippo Gazzola, Elsa M. Marchini. The moon lander optimal control problem revisited[J]. Mathematics in Engineering, 2021, 3(5): 1-14. doi: 10.3934/mine.2021040
[1] | Asma Al-Jaser, Osama Moaaz . Second-order general Emden-Fowler differential equations of neutral type: Improved Kamenev-type oscillation criteria. Electronic Research Archive, 2024, 32(9): 5231-5248. doi: 10.3934/era.2024241 |
[2] | Osama Moaaz, Asma Al-Jaser, Amira Essam . New comparison theorems for testing the oscillation of solutions of fourth-order differential equations with a variable argument. Electronic Research Archive, 2025, 33(7): 4075-4090. doi: 10.3934/era.2025182 |
[3] | Shufen Zhao . The S-asymptotically ω-periodic solutions for stochastic fractional differential equations with piecewise constant arguments. Electronic Research Archive, 2023, 31(12): 7125-7141. doi: 10.3934/era.2023361 |
[4] | Xiaolei Dong . Well-posedness of the MHD boundary layer equations with small initial data in Sobolev space. Electronic Research Archive, 2024, 32(12): 6618-6640. doi: 10.3934/era.2024309 |
[5] | S. Bandyopadhyay, M. Chhetri, B. B. Delgado, N. Mavinga, R. Pardo . Maximal and minimal weak solutions for elliptic problems with nonlinearity on the boundary. Electronic Research Archive, 2022, 30(6): 2121-2137. doi: 10.3934/era.2022107 |
[6] | Liju Yu, Jingjun Zhang . Global solution to the complex short pulse equation. Electronic Research Archive, 2024, 32(8): 4809-4827. doi: 10.3934/era.2024220 |
[7] | Shuang Wang, FanFan Chen, Chunlian Liu . The existence of periodic solutions for nonconservative superlinear second order ODEs: a rotation number and spiral analysis approach. Electronic Research Archive, 2025, 33(1): 50-67. doi: 10.3934/era.2025003 |
[8] | Dan-Andrei Geba, Evan Witz . Revisited bilinear Schrödinger estimates with applications to generalized Boussinesq equations. Electronic Research Archive, 2020, 28(2): 627-649. doi: 10.3934/era.2020033 |
[9] | Dan-Andrei Geba . Unconditional well-posedness for the periodic Boussinesq and Kawahara equations. Electronic Research Archive, 2024, 32(2): 1067-1081. doi: 10.3934/era.2024052 |
[10] | Yong Zhou, Jia Wei He, Ahmed Alsaedi, Bashir Ahmad . The well-posedness for semilinear time fractional wave equations on RN. Electronic Research Archive, 2022, 30(8): 2981-3003. doi: 10.3934/era.2022151 |
Differential equations (DEs) are an essential tool in understanding and predicting many complex phenomena in science, engineering, and they are an important part of modeling and describing natural phenomena. It plays an important role in describing many physical phenomena, as in Newton's laws of motion, and can also be used in many real-life models, such as the propagation of epidemics, and the absorption of glucose by the body (see [1]).
There is also another type of DE called functional differential equations (FDEs), which originate from studies and reflections on dynamic phenomena that cannot be adequately modeled using traditional DEs. In the late nineteenth and early twentieth centuries, scientists began to recognize the need for models that account for both delayed and forward-time effects. In recent decades, the use of FDEs has expanded greatly in many fields, such as engineering, economics, biology, and the social sciences. These equations have become an essential tool for understanding the dynamic behavior of complex systems that contain time effects [2].
In the decade 1920s, delay differential equations (DDEs) began to appear in the scientific literature, where they were used to study biological systems and model the effects of temporal delays on population dynamics in various environments. The decade 1950–1960 witnessed major developments in the mathematical theory of functional differential equations, as rigorous mathematical foundations were laid for solving and studying these equations [3].
DDEs are a type of DE that is an equation with just one variable, typically referred to as time. It gives the function's derivative at a specific moment in terms of its values at previous times. The dependence on the past appears naturally in numerous applications in biology, electrical engineering, and physiology. A simple example in nature is reforestation, which occurs because animals must take time to digest their food before further activities and responses take place. Hence, any mathematical model of forest harvesting and regeneration has time delays built into it (see [4]).
Oscillation theory for ordinary DEs was developed at the beginning of the 1840s. For instance, Sturm's classical work [5] appeared, where oscillation theories were proven, and solutions of second-order linear homogeneous ordinary DEs were compared. In 1921, Fite [6] obtained the first oscillation results for DEs with a translated argument. In recent years, the number of investigations into the oscillation theory of functional DEs has considerably increased. Since then, much research has been done to progress the field of knowledge. To gain an overview of the most significant contributions, we recommend reading the monographs written by Agarwal et al. [7,8,9], Dosly and Rehak [10], and Gyori and Ladas [11].
Researchers have focused a lot of effort and attention on studying the oscillatory behavior of DEs. Special interest in first-order DEs was shown by Philos [12] and Santra et al. [13,14]. Even- and odd-order DEs, along with their diverse classifications, were the focus of the works in [15,16,17,18]. Examining earlier research reveals that the oscillation criteria conclusion involves three fundamental processes. All of the positive solutions to the studied equation should be categorized first. Next, a new or updated method should be developed to obtain improved relations and inequalities connecting the solution, its derivatives, and its corresponding function. Finally, the positive solutions should be excluded using the chosen technique and these improved relations.
Therefore, the purpose of this work is to establish new oscillation criteria for neutral differential equation (NDE)
(λ(μ)[w′(μ)]ℓ)′+∫baq(μ,ϱ)xℓ(φ(μ,ϱ))dϱ=0, | (1.1) |
where μ∈[μ0,∞) and w(μ):=x(μ)+p(μ)x(h(μ)). In our work, we assume the following conditions:
(A1) ℓ is a ratio of odd natural numbers, λ∈C1([μ0,∞),(0,∞)) and
π(μ)=∫∞μ0λ−1/ℓ(ϱ)dϱ=∞; |
(A2) h∈C1([μ0,∞),R), h(μ)≤μ, h′(μ)>0, and limμ→∞h(μ)=∞;
(A3) q∈C([μ0,∞)×[a,b],(0,∞)), p∈C([μ0,∞),R) and 0≤p(μ)≤p0;
(A4) φ∈C1([μ0,∞)×[a,b],(0,∞)), φ(μ,ϱ)<μ for ϱ∈[a,b], and limμ→∞φ(μ,ϱ)=∞.
We recall the definition of solutions of (1.1), using the notations introduced earlier in (A1)–(A4).
Definition 1. [19] A function x∈C([μ0,∞),R), μx≥μ0, which has the property λ(μ)[w′(μ)]ℓ∈C1([μ0,∞),R) and satisfies Eq (1.1) on [μx,∞), is called a solution of (1.1).
Our focus lies in the solution of Eq (1.1) that satisfy the requirement
sup{|x(μ)|:μ>μc}>0, |
for all μ>μc.
Definition 2 [19] A solution of (1.1) is called oscillatory if it is neither eventually positive nor eventually negative. Otherwise, it is said to be non-oscillatory. Equation (1.1) is said to be oscillatory if all of its solutions are oscillatory.
We review the most important previous works and studies that contributed to the development of the study of oscillation of DEs with neutral delay. The following discussion is a summary of some of the relationships that have been presented in the literature.
Under the canonical situation, the customary inequality
x(μ)>(1−p(μ))w(μ) | (1.2) |
is usually used. While the inequality
x(μ)>(1−p(μ)π(h(μ))/π(μ))w(μ) | (1.3) |
is used in the non-canonical situation, see [20,21].
In 2007, Xu and Weng [22] established the criteria for oscillation of
(λ(μ)ψ(x(μ))[w′(μ)]ℓ)′+∫baq(μ,ϱ)f(x(φ(μ,ϱ)))dϱ=0, |
where ψ∈C(R,[m1,m2]), m1,m2∈R+, w(μ)=x(μ)+p(μ)(μ−ho) and ho≥0. They used generalized Riccati substitution and obtained criteria of Philos-type.
In 2008, Xu and Ye [23] studied the oscillation problem of the second order quasilinear NDE
(λ(μ)ψ(x(μ))|w′(μ)|ℓ−1w′(μ))′+q(μ)f(x(φ(μ)))=0, μ≥μo>0, | (1.4) |
where 0≤p(μ)≤1 and are established under the canonical or noncanonical case. They studied the oscillation properties of Eq (1.4) by using the Riccati technique and some inequalities techniques.
In 2009, Han et al. [24] proved this using the following example:
(e2μ(x(μ)+12x(μ−2))′)′+(e2μ+12e2μ+2)x(μ)=0, μ≥μ0>0, | (1.5) |
that x(μ)=e−μ is a positive solution of Eq (1.5). Let x(μ)=e−μ and w(μ)=x(μ)+x(μ−2)/2. Then w′(μ)<0 and
(λ(μ)ψ(x(μ))(w′(μ))ℓ)′+kQ(μ)wℓ(φ(μ))=e4−48eμ>0, |
where
Q(μ)=∫baq(μ,ϱ)Pℓ(φ(μ,ϱ))dμ. | (1.6) |
Hence, the
(λ(μ)ψ(x(μ))(w′(μ))ℓ)′+kQ(μ)wℓ(φ(μ))≤0, μ≥μ0, | (1.7) |
does not hold under the case w′(μ)<0. Furthermore, we have
∫∞[Q(μ)πℓ(φ(μ))−εLkφ′(μ)π(φ(μ))λ1/ℓ(φ(μ))]dμ=∫∞[Q(μ)πℓ(φ(μ))−εLkλ(φ(μ))π(μ)(φ′(μ))ℓλ(ℓ+1)/ℓ(μ)]dμ=∫∞[Q(μ)π(φ(μ))−1Lk(ℓ+1)ℓ+1λ(φ(μ))λ1/ℓ(φ(μ))πℓ(φ(μ))]dμ=∫∞e2−28dμ=∞. |
This proves that x(μ)=e−μ is a nonoscillatory solution of Eq (1.5); although Xu and Ye [23] proved otherwise. This confirms that there is an error in their results. Which then prompted Han et al. [24] to complete and correct the results obtained by Xu and Ye [23].
In 2010, Baculikova and Dzurina [25] studied the NDE of the form
(λ(μ)[x(μ)+p(μ)x(h(μ))]′)′+q(μ)x(φ(μ))=0. |
The new oscillation criteria that were proposed for the situation where
0≤p(μ)≤po<∞ and h∘φ=φ∘h, |
and assume that the canonical case holds.
In 2014, Li et al. [26] studied the oscillatory properties of second-order nonlinear NDEs with distributed deviating arguments
(λ(μ)|w′(μ)|ℓ−1w′(μ))′+∫baq(μ,ξ)|x[φ(μ,ξ)]|ℓ−1x[φ(μ,ξ)]dσ(ξ)=0, | (1.8) |
where σ∈C([a,b],R). Some new criteria are offered based on less restrictive assumptions made regarding the neutral coefficient, where p(μ)≥0. They used the Riccati substitution technique and some inequalities to establish several new oscillation criteria for (1.8), assuming that the canonical or noncanonical case holds.
In 2015, Candan [27] studied the oscillation of more general NDE
[λ(μ)|w′(μ)|ℓ−1w′(μ)]′+∫daF(μ,x(φ(μ,ξ)))dξ=0, |
where ℓ∈R+, w(μ)=x(μ)+∫dap(μ,ϱ)x(h(μ,ϱ))dϱ and F(μ,x(φ(μ,ξ)))≥q(μ,ϱ)|x(μ)|ℓ. He used the Riccati method and assumed that
0≤∫dap(μ,ϱ)dϱ≤1, |
and considered the canonical case.
Moaaz et al. [28] obtained conditions for oscillation of
(λ(μ)[w′(μ)]ℓ)′+∫baq(μ,ϱ)xℓ(φ(μ,ϱ))dϱ=0, | (1.9) |
where w(μ)=x(μ)+∫dcp(μ,ϱ)x(h(μ,ϱ))dϱ. They obtained oscillation criteria that complement and enhance several results published in the literature by using an improved version of the Riccati transformations and comparison concepts.
Moaaz et al. [29] studied the oscillatory behavior of
(λ(μ)(w′(μ))γ)′+l∑j=1qj(μ)xβ(φj(μ))=0, | (1.10) |
As an enhancement to (1.2), they used p(t)=ρ0
x(μ)>w(μ)κ/2∑m=11ρ2m−10(1−1ρ0π(h−2m(μ))π(h−(2m−1)(μ))), for ρ0>1 and κ is even, |
and
x(μ)>w(μ)(1−ρ0)(κ−1)/2∑m=0ρ2m0π(h2m+1(μ))π(μ), for ρ0<1, and κ is odd, |
where h[h](μ)=h(h[h−1](μ)), for h=1,2,...,2m.
Hassan et al. [30] investigated the oscillatory features of
(λ(μ)(w′(μ))γ)′+q(μ)xγ(φ(μ))=0, |
they offered
x(μ)>w(μ)(κ−1)/2∑κ=0ρ2κ0(1−p(t)π(h[2κ+1](μ))π(h2κ(μ))), |
which as enhancement to (1.3), where κ is odd.
Equation
(λ(μ)(w′(μ))γ)′+q(μ)xγ(φ(μ))=0 |
has been studied by Bohner et al. [31]. They used
x(μ)>w(μ)(1−p(μ))(1+Hk(μ)), |
where
Hk(μ)={0,for k=0,k∑i=12i−1∏j=0p(hj(μ)),for h(μ)≤μ and k∈N,k∑i=1π(h2i(μ))π(μ)2i−1∏j=0p(hj(μ)),for h(μ)≥μ and k∈N, |
h[0](μ)=μ and h[j](μ)=h(hj−1(μ)) for all j∈N, which is an enhancement of (1.3).
New better relations between the solution and its related function have been derived by a number of researchers who have lately studied the oscillatory behavior of solutions to the NDE of various orders, (see, for example, [32,33,34,35]).
The following lemmas will be used in deriving our results.
Lemma 1. [36] Let F, H∈[0,∞). Then,
(F+H)ℓ≤δ(Fℓ+Hℓ), |
where
δ:={1ℓ≤1;2ℓ−1ℓ>1. |
Lemma 2. [37] Let ℓ be a ratios of two odd positive integers. Then,
Lv(ℓ+1)/ℓ−Kv≥−ℓℓ(ℓ+1)ℓ+1Kℓ+1Lℓ, L>0. | (1.11) |
In this study, we investigate the oscillatory behavior of a neutral second-order nonlinear differential equation with distributed deviating arguments, denoted as (1.1). By employing advanced analytical techniques, we establish three significant oscillation theorems that apply to all solutions of this equation. Our findings not only generalize and extend existing results in the literature but also introduce novel oscillatory criteria based on comparison principles and Riccati-type transformations. These contributions offer a more robust and comprehensive framework for understanding the oscillatory dynamics of Eq (1.1).
Moreover, we explore the implications of these theorems in the context of fractional delay differential equations, which are particularly relevant in systems characterized by memory effects. To validate the theoretical results, we present numerical simulations (Figures 1 and 2), which demonstrate strong agreement between analytical predictions and computational outcomes, thereby confirming the effectiveness and applicability of the proposed approach.
The paper is divided into four sections. In the next section, we introduce the initial lemmas, including the derivation of improved relations between the solution and its derivatives. We then derive new oscillation theorems using several approaches, and, we apply these results through a practical example to confirm their validity and applicability. In the third section, we provide a numerical analysis of the results from the previous example and schematics of the oscillatory nature of the solutions. The final section contains an analytical summary of the findings in this paper and an overview of our results in real-world applications.
In what follows, we derive some improved relationships between the dependent solution, its derivatives, and the corresponding function, and use them to derive our oscillation theorems. We also give several examples of the practical application of our results, emphasizing their easy applicability. For simplicity, we use the notations listed below:
λ(μ):=ℓλ1/ℓ(μ),Q(μ):=∫baq(μ,ϱ)Pℓ(φ(μ,ϱ))dμ,˜Q(μ)=∫∞μQ(ϱ)dϱ,P(μ;m):=m∑i=0(2i∏j=0p(hj(μ)))[1p(h2i(μ))−1]π(h2i(μ))π(μ),h0(μ):=μ,hi(μ):=h(hi−1(μ)),for i=1,2,.... |
Remark 1. All functional inequalities are assumed to hold for all sufficiently large μ in this section.
The following lemma provides a crucial inequality used in the proof of our main Theorems.
Lemma 3. Assume that x(μ) is an eventually positive solution of (1.1). Then, eventually
x(μ)>w(μ)m∑i=0(2i∏j=0p(hj(μ)))[1p(h2i(μ))−1]π(h2i(μ))π(μ), | (2.1) |
for any integer m≥0. So, Eq (1.1) became as follows:
(λ(μ)[w′(μ)]ℓ)′+∫baq(μ,ϱ)Pℓ(φ(μ,ϱ))wℓ(φ(μ,ϱ))dϱ≤0. |
Proof. The proof of this Lemma comes directly from the proof of Lemma 2.2 in reference [33]. Thus, it has been omitted.
Theorem 1. Assume that
∫∞μ0Q(ϱ)dϱ=∞. |
Then, Eq (1.1) is oscillatory.
Proof. Assume that x(μ) is a positive solution of Eq (1.1), meaning there is exists a μ0≥0 such that x(μ)>0 for μ≥μ0, and as a result, there exists a μ1≥μ0 such that x(h(μ))>0 for μ≥μ0, x(φ(μ,ϱ)) for μ≥μo. This same argument can be used to show if x(μ) is an eventually negative solution. From Eq (1.1), we have
(λ(μ)[w′(μ)]ℓ)′=−∫baq(μ,ϱ)xℓ(φ(μ,ϱ))dϱ. | (2.2) |
Hence, λ(μ)[w′(μ)]ℓ is decreasing. Consequently, we have two cases for w′(μ).
(ⅰ) w′(μ)<0 eventually.
(ⅱ) w′(μ)>0 eventually.
First, assume that w′(μ)<0 for μ≥μ1. By using decreasing nature of λ(μ)[w′(μ)]ℓ, we obtain
(λ(μ)[w′(μ)]ℓ)′≤λ(μ2)[w′(μ2)]ℓ for μ≥μ2≥μ1. | (2.3) |
Dividing both sides of (2.3) by λ(μ), integrating from μ2 to μ, and using (A1), we obtain
w(μ)≤w(μ2)−λ1/ℓ(μ2)w′(μ2)∫μμ2λ−1/ℓ(ϱ)dϱ→−∞ as μ→∞, |
which contradicts the positive nature of w(μ).
Second, assume that w′(μ)>0 for μ≥μ1. Since w(μ)≥x(μ) and w(μ) is increasing, we have
x(μ)>w(μ)P(μ;m), | (2.4) |
according to Lemma 3. Hence,
(λ(μ)[w′(μ)]ℓ)′≤−∫baq(μ,ϱ)Pℓ(φ(μ,ϱ))wℓ(φ(μ,ϱ))dϱ, | (2.5) |
or
(λ(μ)[w′(μ)]ℓ)′≤−Q(μ)wℓ(φ(μ,ϱ)). |
On the other hand, since λ(μ)[w′(μ)]ℓ is decreasing, we have
λ(μ)[w′(μ)]ℓ≤λ(φ(μ))[w′(φ(μ))]ℓ, |
or
[λ(μ)λ(φ(μ))]1/ℓ≤w′(φ(μ))w′(μ). | (2.6) |
Let
ω(μ)=λ(μ)[w′(μ)]ℓwℓ(φ(μ)), μ≥μ3, | (2.7) |
Taking the derivative of ω(μ), we have
ω′(μ)=(λ(μ)[w′(μ)]ℓ)′wℓ(φ(μ))−ℓλ(μ)φ′(μ)(w′(μ))ℓw′(φ(μ))(w(φ(μ)))ℓ+1≤−Q(μ)−λ(μ)ω1+1/ℓ(μ), |
or
ω′(μ)+Q(μ)+λ(μ)ω1+1/ℓ(μ)≤0, | (2.8) |
or
ω′(μ)+Q(μ)≤0. | (2.9) |
Integrating (2.9) from μ3 to μ and using (2.6), we obtain
ω(μ)≤ω(μ3)−∫μμ3Q(ϱ)dϱ→−∞ as μ→∞, |
contradiction, since ω(μ) is positive and decreasing. This completes the proof.
Definition 3. We define a sequence of functions {ϝn(μ)}∞n=0 as
ϝ0(μ)=˜Q(μ), μ≥μ0, | (2.10) |
and
ϝn(μ)=∫∞μλ(ϱ)ϝ1+1/ℓn−1(ϱ)dϱ+ϝ0(μ), μ≥μ0 | (2.11) |
and n=1,2,3.....We see that by induction
ϝn(μ)≤ϝn+1(μ), μ≥μ0 and n=1,2,3... | (2.12) |
Lemma 4. Assume that x(μ) is an eventually positive solution of Eq (1.1). Then, ϝn(μ)≤ω(μ), where ω(μ) and ϝn(μ) are as defined in (2.7) and (2.11), respectively. There exists a positive function ϝ(μ) on [μ,∞), such that limn→∞ϝn(μ)=ϝ(μ) for μ≥μ1≥μ0 and
ϝ(μ)=∫∞μλ(ϱ)ω1+1/ℓ(ϱ)dϱ+ϝ0(μ),mu≥μ1. | (2.13) |
Proof. By following the steps in the proof of Theorem 2.1, we arrive at (2.8). Integrating (2.8) from μ to μ∗, we get
ω(μ∗)−ω(μ)+∫μ∗μQ(ϱ)dϱ+∫μ∗μλ(ϱ)ω1+1/ℓ(ϱ)dϱ≤0. | (2.14) |
From (2.14), it is obvious that
ω(μ∗)−ω(μ)+∫μ∗μλ(ϱ)ω1+1/ℓ(ϱ)dϱ≤0. | (2.15) |
Then, from (2.15) we conclude the following relation
∫∞μλ(ϱ)ω1+1/ℓ(ϱ)dϱ<∞, | (2.16) |
for μ≥μ1 and
ω(μ∗)≤ω(μ)−∫μ∗μλ(ϱ)ω1+1/ℓ(ϱ)dϱ⟶−∞ as μ∗⟶∞, |
we get a contradiction. Since ω(μ) is positive and decreasing, we note that
limμ→∞ω(μ)=k≥0, |
We know that k=0 by virtue of (2.16). We therefore obtain
ω(μ∗)+˜Q(μ)+∫μ∗μλ(ϱ)ω1+1/ℓdϱ≤ω(μ) |
as a consequence of (2.14), i.e.,
ω(μ)≥˜Q(μ)=ϝ0. |
Furthermore, we may also observe by induction that ω(μ)≥ϝn(μ) for μ≥μ0 and n=1,2,3....Consequently, the sequence ϝn(μ) converges to ϝ(μ) since it is monotone increasing and bounded above. Using Lebesgue's monotone convergence theorem and letting n→∞ in (2.11), we arrive at (2.13). The proof is finished.
Theorem 2. Assume that
liminfμ→∞1ϝ0(μ)∫∞μϝ1+1/ℓ0(ϱ)λ(ϱ)dϱ>ℓ(ℓ+1)1+1/ℓ. | (2.17) |
Then Eq (1.1) is oscillatory.
Proof. Let x(μ) be a nonoscillatory solution of Eq (1.1). By following the steps in the Lemma 4 proof, we obtain
ω(μ)ϝ0(μ)≥1+1ϝ0(μ)∫∞μλ(ϱ)ϝ1+/ℓ0(ϱ)(ω(ϱ)ϝ0(ϱ))1+1/ℓdϱ. | (2.18) |
Let
β=infμ≥μ1ω(μ)ϝ0(μ), |
then β≥1. From (2.17) and (2.18), we obtain
β≥1+ℓ(βℓ+1)1+1/ℓ, |
or
βℓ+1≥1ℓ+1+ℓℓ+1(βℓ+1)1+1/ℓ, |
which contradicts the admissible value of ℓ and β, then the proof is finished.
Theorem 3. Let ϝn(μ) be as defined in (2.11). There exists ϝn(μ) such that
limsupμ→∞ϝn(μ)(∫μμ0λ−1/ℓ(ϱ)dϱ)ℓ>1, | (2.19) |
then Eq (1.1) is oscillatory.
Proof. Let x(μ) be a nonoscillatory solution of Eq (1.1) and defining ω(μ) as in (2.7), then we have
1ω(μ)=1λ(μ)[w(φ(μ))w′(μ)]ℓ=1λ(μ)[w(μ1)+∫μ1μλ−1/ℓ(ϱ)w′(ϱ)λ1/ℓ(ϱ)dϱw′(μ)]ℓ≥1λ(μ)[w′(μ)λ1/ℓ(μ)∫μ1μλ−1/ℓ(ϱ)dϱw′(μ)]ℓ=[∫μ1μλ−1/ℓ(ϱ)dϱ]ℓ, μ≥μ1. | (2.20) |
Thus, we conclude from (2.20) that
ω(μ)[∫μμ0λ−1/ℓ(ϱ)dϱ]ℓ≤[∫μμ0λ−1/ℓ(ϱ)dϱ∫μμ1λ−1/ℓ(ϱ)dϱ]ℓ, |
and therefore, we have
limsupμ→∞ ω(μ)[∫μμ0λ−1/ℓ(ϱ)dϱ]ℓ≤1. |
Which contradicts (2.19) and completes the proof.
We can derive the following corollary from the results in the previous theorem.
Corollary 1. Let ϝn(μ) be as defined in (2.12). If there exists ϝn(μ) such that either
∫∞μoQ(μ)exp(∫μμoλ(ϱ)ϝ1/ℓn(ϱ)dϱ)dμ=∞, | (2.21) |
or
∫∞μoλ(μ)ϝ1/ℓn(μ)ϝo(μ)exp(∫μμoλ(ϱ)ϝ1/ℓn(ϱ)dϱ)dμ=0. | (2.22) |
Then, Eq (1.1) is oscillatory.
Proof. Let x(μ) be a nonoscillatory solution of Eq (1.1), and, therefore we may assume that x(μ) is eventually positive. Then, from (2.11), we have that
ϝ′(μ)=−λ(μ)ϝ1+1/ℓ(μ)−Q(μ), μ≥μ1. | (2.23) |
Since ϝn(μ)≤ϝ(μ), from (2.23), it follows that
ϝ′(μ)≤−λ(μ)ϝ1/ℓn(μ)ϝ(μ)−Q(μ), μ≥μ1. | (2.24) |
By (2.24), we get
ϝ(μ)≤exp(−∫μμ1λ(ϱ)ϝ1/ℓn(ϱ)dϱ)(ϝ(μ1)−∫μμ1Q(ϱ)exp(∫ϱμλ(ω)ϝ1/ℓn(ω)dω)dϱ). |
The above inequality follows
∫μμ1Q(ϱ)exp(∫ϱμ1λ(ω)ϝ1/ℓn(ω)dω)dϱ≤ϝ(μ1)<∞, |
which contradicts with (2.21). Furthermore, define
v(μ)=∫∞μλ(ϱ)ϝ1+1/ℓ(ϱ)dϱ, μ≥μ1. | (2.25) |
Then, taking the derivative of (2.25) and using (2.13), we obtain
v′(μ)=−λ(μ)ϝ1+1/ℓ(μ)≤−λ(μ)ϝ1/ℓn(μ)ϝ(μ)=−λ(μ)ϝ1/ℓn(μ)[v(μ)+ϝ0(μ)]. |
Similarly to the above, we conclude that
∫∞μ1λ(μ)ϝ1/ℓnϝ0(μ)exp(∫μμ1λ(ϱ)ϝ1/ℓn(ϱ)dϱ)dμ<∞, |
which contradicts with (2.22) and completes the proof.
First, let us define
Q(μ,ϱ)=min{q(μ,ϱ),q(h(μ),ϱ)}. |
Furthermore, an additional assumption,
φ∘h=h∘φ and h′(μ)≥h0>0 |
must be added to the fundamental assumptions (A1) through (A4) in the introduction section to achieve the following conclusions.
Theorem 4. Suppose that φ(μ,a)∈C1([μ0,∞),R), φ′(μ,a)>0, and φ(μ,a)≤h(μ) for μ∈[μ0,∞). If there exists a real-valued function ρ∈C1([μ0,∞),(0,∞)) such that
limsupμ→∞∫μμ0(ρ(u)∫baQ(μ,ϱ)dϱδ−(1+pℓoho)λ(φ(u,a))(ρ′(u))ℓ+1(ℓ+1)ℓ+1(ρ(u)φ′(u,ϱ))ℓ)du=∞, | (2.26) |
then Eq (1.1) is oscillatory.
Proof. Assume that x(μ) is a nonoscillatory solution for Eq (1.1). We suppose, without losing generality, that for all μ≥μ1 and ϱ∈[a,b], there exists a μ∈[μ0,∞) such that x(μ)>0,x(h(μ))>0, and x(φ(μ,ϱ))>0. Consequently, w(μ)>0. By Lemma 1, we get
wℓ(φ(μ,ϱ))≤δ(xℓ(φ(μ,ϱ))+pℓoxℓ(φ(h(μ,ϱ)))). | (2.27) |
By (A2) and (1.1), we have
1h′(μ)(λ(h(μ))[w′(h(μ))]ℓ)′+∫baq(h(μ),ϱ)xℓ(φ(h(μ),ϱ))dϱ≤0. |
That is,
pℓoho(λ(h(μ))[w′(h(μ))]ℓ)′+pℓo∫baq(h(μ),ϱ)xℓ(φ(h(μ)),ϱ)dϱ≤0. | (2.28) |
It follows from using (1.1) and (2.28) with (2.27) that
0≥(λ(μ)[w′(μ)]ℓ)′+∫baq(μ,ϱ)xℓ(φ(μ),ϱ)dϱ+pℓoho(λ(h(μ))[w′(h(μ))]ℓ)′+pℓo∫baq(h(μ),ϱ)xℓ(φ(h(μ)),ϱ)dϱ, |
hence,
0≥(λ(μ)[w′(μ)]ℓ)′+pℓoho(λ(h(μ))[w′(h(μ))]ℓ)′+∫baQ(μ,ϱ)(xℓ(φ(μ,ϱ))+pℓoxℓ(φ(h(μ)),ϱ))dϱ. |
From that fact that w′(μ)>0 and φ(μ,ϱ)≥φ(μ,a), we have
(λ(μ)[w′(μ)]ℓ)′+pℓoho(λ(h(μ))[w′(h(μ))]ℓ)′≤−1δwℓ(φ(μ,a))∫baQ(μ,ϱ)dϱ. | (2.29) |
We define a Riccati substitution:
Ω1(μ)=ρ(μ)λ(μ)[w′(μ)]ℓwℓ(φ(μ,ϱ)), | (2.30) |
Differentiating (2.30) and from (2.6), we get
Ω′1(μ)≤ρ(μ)(λ(μ)[w′(μ)]ℓ)′wℓ(φ(μ,ϱ))+ρ′(μ)ρ(μ)Ω1(μ)−ℓφ′(μ,ϱ)ρ1/ℓ(μ)λ1/ℓ(φ(μ,ϱ))Ω(1+1/ℓ)1(μ). | (2.31) |
Parallel to that, we present an additional Riccati transformation:
Ω2(μ)=ρ(μ)λ(h(μ))[w′(h(μ))]ℓwℓ(φ(μ,ϱ)). | (2.32) |
Similarly, by differentiating (2.32) and from (2.6), we get
Ω′2(μ)≤ρ(μ)(λ(h(μ))[w′(h(μ))]ℓ)′wℓ(φ(μ,ϱ))+ρ′(μ)ρ(μ)Ω2(μ)−ℓφ′(μ,ϱ)ρ1/ℓ(μ)λ1/ℓ(φ(μ,ϱ))Ω(1+1/ℓ)2(μ). | (2.33) |
Combining (2.31) and (2.33), we get
Ω′1(μ)+pℓohoΩ′2(μ)≤ρ(μ)(λ(μ)[w′(μ)]ℓ)′+pℓoho((λ(h(μ))[w′(h(μ))]ℓ)′)wℓ(φ(μ,ϱ))+ρ′(μ)ρ(μ)Ω1(μ)−ℓφ′(μ,ϱ)ρ1/ℓ(μ)λ1/ℓ(φ(μ,ϱ))Ω(1+1/ℓ)1(μ)+pℓohoρ′(μ)ρ(μ)Ω2(μ)−pℓohoℓφ′(μ,ϱ)ρ1/ℓ(μ)λ1/ℓ(φ(μ,ϱ))Ω(1+1/ℓ)2(μ). |
From (2.29), it is evident that
Ω′1(μ)+pℓohoΩ′2(μ)≤−ρ(μ)δ∫baQ(μ,ϱ)dϱ+ρ′(μ)ρ(μ)Ω1(μ)−ℓφ′(μ,ϱ)ρ1/ℓ(μ)λ1/ℓ(φ(μ,ϱ))Ω(1+1/ℓ)1(μ)+pℓohoρ′(μ)ρ(μ)Ω2(μ)−pℓohoℓφ′(μ,ϱ)ρ1/ℓ(μ)λ1/ℓ(φ(μ,ϱ))Ω(1+1/ℓ)2(μ). |
By integrating the aforementioned inequality from μ2 to μ, we arrive at
Ω1(μ)−Ω1(μ2)+pℓohoΩ2(μ)−pℓohoΩ2(μ2)≤−∫μμ2ρ(u)δ∫baQ(μ,ϱ)dϱdu+∫μμ2[ρ′(ϱ)ρ(ϱ)Ω1(ϱ)−ℓφ′(ϱ,a)ρ1/ℓ(ϱ)λ1/ℓ(φ(ϱ,a))Ω(1+1/ℓ)1(ϱ)]dϱ+∫μμ2pℓoho[ρ′(ϱ)ρ(ϱ)Ω2(ϱ)−ℓφ′(ϱ,a)ρ1/ℓ(ϱ)λ1/ℓ(φ(ϱ,a))Ω(1+1/ℓ)2(ϱ)]dϱ. |
By using the inequality (1.11) with K=ρ′(μ)/ρ(μ) and L=ℓφ′(μ,ϱ)/ρ1/ℓ(μ)λ1/ℓ(φ(μ,ϱ)), we obtain
ρ′(μ)ρ(μ)Ω1(μ)−ℓφ′(μ,ϱ)ρ1/ℓ(μ)λ1/ℓ(φ(μ,ϱ))Ω(1+1/ℓ)1(μ)≤1(ℓ+1)ℓ+1λ(φ(μ,ϱ))(ρ′(μ))ℓ+1(ρ(μ))ℓ(φ′(μ,ϱ))ℓ. |
In the same way, we find that
ρ′(μ)ρ(μ)Ω2(μ)−ℓφ′(μ,ϱ)ρ1/ℓ(μ)λ1/ℓ(φ(μ,ϱ))Ω(1+1/ℓ)2(μ)≤1(ℓ+1)ℓ+1λ(φ(μ,ϱ))(ρ′(μ))ℓ+1(ρ(μ))ℓ(φ′(μ,ϱ))ℓ. |
Hence,
Ω1(μ)−Ω1(μ2)+pℓohoΩ2(μ)−pℓohoΩ2(μ2)≤−∫μμ2ρ(u){∫baQ(μ,ϱ)dϱδ−(1+pℓoho)λ(φ(u,a))(ρ′(u))ℓ+1(ℓ+1)ℓ+1(ρ(u)φ′(u,ϱ))ℓ}du. |
This contradicts with (2.26). This ends the proof.
Theorem 5. If the first-order delay NDE
ϖ′(μ)+hoδ(ho+pℓo)ϖ(φ(μ,a))ˆQ(μ)=0 | (2.34) |
is oscillatory, then Eq (1.1) is oscillatory. Where
ˆQ(μ)=(∫φ(μ,a)μ1λ−1/ℓ(u)du)ℓ∫baQ(μ,ϱ)dϱ. |
Proof. Continuing from (2.29), we find that
(λ(μ)[w′(μ)]ℓ+pℓohoλ(h(μ))[w′(h(μ))]ℓ)′≤−1δwℓ(φ(μ,a))∫baQ(μ,ϱ)dϱ. | (2.35) |
Equation (1.1) indicates that η(μ)=λ(μ)[w′(μ)]ℓ is decreasing, which implies that
w(μ)≥∫μμ1(λ(ϱ)(w′(ϱ))ℓ)1/ℓλ−1/ℓ(ϱ)dϱ≥η1/ℓ(μ)∫μμ1λ−1/ℓ(ϱ)dϱ. | (2.36) |
Consequently, from (2.36) in (2.35), we can observe that η(μ) is a positive solution of
(η(μ)+pℓohoη(h(μ)))′+1δη(φ(μ,a))(∫φ(μ,a)μ1λ−1/ℓ(u)du)ℓ∫baQ(μ,ϱ)dϱ≤0. | (2.37) |
Now, define a new function ϖ(μ) with
ϖ(μ)=η(μ)+pℓohoη(h(μ)). |
Then, by monotonicity of η(μ),
ϖ(μ)≤η(μ)(1+pℓoho). | (2.38) |
By replacing (2.38) with (2.37), we find that ϖ(μ) is a positive solution to the delay differential inequality
ϖ′(μ)+hoδ(ho+pℓo)ϖ(φ(μ,a))(∫φ(μ,a)μ1λ−1/ℓ(u)du)ℓ∫baQ(μ,ϱ)dϱ≤0. |
Thus, the last inequality has a positive solution. Using Theorem 1 in [12], we see that (2.34) also has a positive solution, which is a contradiction. The proof is complete.
Using Theorem 2.1.1 in [38], we get the following corollary.
Corollary 2. If φ(μ,a)<μ and
liminfμ→∞∫μφ(μ,a)ˆQ(u,a)du>δ(ho+pℓo)hoe, |
then Eq (1.1) is oscillatory.
We apply our results in the previous theories to practical examples.
Example 1. Consider
(((x(μ)+p0x(h0μ))′)ℓ)′+∫1νq0μℓ+1xℓ(μϱ)dϱ=0, | (2.39) |
where h0,ν∈(0,1). Note that π(μ)=μ,
P(μ;m):=m∑i=0(2i∏j=0p(hj(μ)))[1p(h2i(μ))−1]π(h2i(μ))π(μ)=(1−p0)m∑i=0p02ih02i, |
Q(μ):=q0μℓ+1(1−ν)((1−p0)m∑i=0p02ih02i)ℓ, |
and
˜Q(μ)=∫∞μQ(ϱ)dϱ=q0ℓμℓ(1−ν)((1−p0)m∑i=0p02ih02i)ℓ. |
From Theorem 2, Eq (2.39) is oscillatory if
q0>ℓℓ+1[(1−p0)m∑i=0p02ih02i]ℓ(1−ν)(ℓ+1)ℓ+1. | (2.40) |
Applying Theorem 5 in [28], Eq (2.39) is oscillatory if
q0>ℓℓ+1(1−p0)ℓ(1−ν)(ℓ+1)ℓ+1(ν)ℓ/(1+νℓq0ℓ(1−ν)(1−p0)ℓ). | (2.41) |
Remark 2. In the special case where p0 =1/2, h0=1/2, ν=1/3 and ℓ=1, condition (2.40) becomes q>0.703.
On the other hand, using Theorem 5 in [28], Eq (2.39) is oscillatory if q0>1.8635. As a result, our findings offer an improved oscillation criterion.
The following example illustrates the sharpness of the condition (2).
Example 2. Cosider
((x(μ)+3x(μ−π))′)′+∫3π/2π/2x(μ−ϱ)dϱ=0, | (2.42) |
where φ(μ,a)<μ such that
μ>0, λ(μ)=1, ℓ=1, p(μ)=p0=3, q(μ,ϱ)=1>0, φ(μ,ϱ)=μ−ϱ, |
φ∘h=h∘φ, h′(μ)≥h0=1>0, φ(μ,a)=μ−π2, |
and
π(μ0)=∫∞μ0dϱ=∞. |
Also,
Q(μ,ϱ)=1 |
and
ˆQ(μ)=(∫φ(μ,a)μ1λ−1/ℓ(u)du)ℓ∫baQ(μ,ϱ)dϱ=(μ−32π) |
By applying (2), we have
liminfμ→∞∫μφ(μ,a)ˆQ(u,a)du>δ(ho+pℓo)hoe. |
Then, all solutions of (2.42) are oscillatory. Moreover, x(μ)=sin(μ) is one such oscillatory solution.
In this section, we present numerical solutions to show the oscillatory behavior of specific examples. These numerical simulations can then be compared with the theoretical findings obtained in previous sections.
We solve Example 2 numerically, where Eq (1.1) is initially transformed to a system of first order. The system is then solved using Matlab 2024. Numerical solutions are shown in Figures 1 and 2, where x(t)=sin(t) is the history of the solution considered.
When comparing the numerical solution with the theoretical analysis, we observe a consistent oscillatory behavior in accordance with the conditions of the oscillation theorems. The mathematical analysis, based on the assumptions imposed on the coefficients and functions in the differential equation, confirms that all solutions must be oscillatory. This theoretical conclusion is supported by the numerical example given in Example 2, where the solution takes the explicit form sin(t), which clearly oscillates.
Numerically, the equation was solved using MATLAB 2024, where it was transformed into a system of first-order differential equations, and then numerical methods were applied to solve it. The numerical solutions also exhibited an oscillatory behavior consistent with the theoretical conclusions.
This agreement between the theoretical and numerical solutions indicates that the mathematical approach used in the theoretical analysis is accurate and effective in predicting oscillatory solutions. Consequently, the numerical results reinforce the validity of the theoretical proof and the approach used in the study, confirming that all solutions of the studied equation remain oscillatory over time.
This paper investigated the oscillatory behavior of a class of second-order neutral nonlinear differential equations with distributed deviating arguments, as given in Eq (1.1). By employing the Riccati transformation and comparison principles, we established three new oscillation theorems under suitable conditions on the equation's coefficients and deviating arguments.
These results not only generalize but also extend several earlier contributions in the literature, as discussed in the introduction. Furthermore, the validity of the theoretical results has been supported by numerical examples (see Examples 1 and 2), which exhibit behaviors consistent with the predicted oscillation criteria.
Our findings provide a comprehensive analytical framework for addressing oscillatory behavior in this class of equations. In future work, it would be of interest to investigate how these results might be extended to fractional differential equations with delay, especially those involving memory effects characterized by power-law, exponential, or Mittag-Leffler kernels (see [39,40]).
The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.
The authors gratefully acknowledge Qassim University, represented by the Deanship of Graduate Studies and Scientific Research, on the financial support for this research under the number (QU-J-PG-2-2025-53678) during the academic year 1446AH / 2024 AD.
The authors declare there is no conflict of interest.
[1] | Evans LC, An Introduction to Mathematical Optimal Control Theory. Available from: https://math.berkeley.edu/evans/control.course.pdf. |
[2] | Fattorini HO (1999) Infinite-Dimensional Optimization and Control Theory, Cambridge: Cambridge University Press. |
[3] | Fleming W, Rishel R (1975) Deterministic and Stochastic Optimal Control, Springer. |
[4] | Meditch JS (1964) On the problem of optimal thrust programming for a lunar soft kanding. IEEE T Automat Contr 9: 477–484. |
[5] | Miele A (1962) The calculus of variations in applied aerodynamics and flight mechanics, In: Optimization Techniques with Applications to Aerospace Systems, Academic Press, 99–170. |