Volume 2018, Issue 1 5321314
Research Article
Open Access

On Solvability Theorems of Second-Order Ordinary Differential Equations with Delay

Nai-Sher Yeh

Corresponding Author

Nai-Sher Yeh

Department of Mathematics, Fu Jen Catholic University, Xinzhuang District, New Taipei City 24205, Taiwan fju.edu.tw

Search for more papers by this author
First published: 01 April 2018
Academic Editor: Chun-Lei Tang

Abstract

For each x0 ∈ [0,2π) and kN, we obtain some existence theorems of periodic solutions to the two-point boundary value problem u′′(x) + k2u(xx0) + g(x, u(xx0)) = h(x) in (0, 2π) with u(0) − u(2π) = u(0) − u(2π) = 0 when g : (0,2π) × RR is a Caratheodory function which grows linearly in u as |u| → ∞, and hL1(0,2π) may satisfy a generalized Landesman-Lazer condition for all vN(L)\{0}. Here N(L) denotes the subspace of L1(0,2π) spanned by sin⁡kx and cos⁡kx, −1 < β ≤ 0, , and .

1. Introduction

Let x0 ∈ [0,2π) and kN be fixed. We consider the following two-point boundary value problems:
()
()
where hL1(0,2π) is given and g : (0,2π) × RR is a Caratheodory function; that is, g(x, u) is continuous in uR, for a.e. x ∈ (0,2π), is measurable in x ∈ (0,2π) for all uR, and satisfies, for each r > 0, the fact that there exists an arL1(0,2π) such that
()
for a.e. x ∈ (0,2π) and all |u| ≤ r. Concerning the growth condition of the nonlinear term g to (1) and (2), we assume that
  • (H)

    there exist constants −1 < β ≤ 0, r0 > 0, and a, b, c, dL1(0,2π),   a, b ≥ 0 and a(x) ≤ 2k + 1 for a.e. x ∈ (0,2π) with strict inequality on a positive measurable subset of (0,2π), such that for a.e. x ∈ (0,2π) and all ur0

    ()

  • and for a.e. x ∈ (0,2π) and all u ≤ −r0

    ()

  • (G)

    there exist constants −1 < β ≤ 0, r0 > 0, and a, b, c, dL1(0,2π),   a, b ≥ 0 and a(x) ≤ 2k − 1 for a.e. x ∈ (0,2π) with strict inequality on a positive measurable subset of (0,2π), such that for a.e. x ∈ (0,2π) and all ur0

    ()

  • and for a.e. x ∈ (0,2π) and all u ≤ −r0

    ()

respectively, and a generalized Landesman-Lazer condition
()
for all vN(L)\{0}, may be satisfied. Here N(L) denotes the subspace of L1(0,2π) spanned by sin⁡kx and cos⁡kx, βR, , and . Under assumptions and either with or without the Landesman-Lazer condition
()
for all vN(L)\{0}, the solvability of the problem (1) has been extensively studied if the nonlinearity g(x, u) has at most linear growth in u as |u| → (see [113] for the case x0 = 0 and [1416] for the general case) or grows superlinearly in u in one of directions u and u → − and may be bounded in the other (see [8, 17] for the case x0 = 0 and [14] for the general case when k = 0). Based on the well-known Leray-Schauder continuation method (see [18, 19]), we obtain solvability theorems to (1) (resp., (2)) when g(x, u) satisfies (H) (resp., (G)) and either (8) with −1 < β < 0 or (9) with β = 0 is satisfied, which extends the results of [15] for the nonresonance case, and has been established in [9] for the case x0 = 0 and g(x, u) grows sublinearly in u as |u| → with −1 < β ≤ 1. Unfortunately, it is still unknown when kN,  g(x, u) grows linearly in u as |u| → and the assumption of (8) is replaced by
()
for all vN(L)\{0} with β > 0. In the following we will make use of real Banach spaces Lp(0,2π),   C[0,2π] and Sobolev spaces W2,1(0,2π) and H1(0,2π). The norms of Lp(0,2π),   C[0,2π] and H1(0,2π) are denoted by and , respectively. By a solution of (1), we mean a periodic function u : RR of period 2π which belongs to W2,1(0,2π) and satisfies the differential equation in (1) a.e. x ∈ (0,2π).

2. Existence Theorems

For each vW2,1(0,2π) with v(0) − v(2π) = v(0) − v(2π) = 0 and kN, we write ,  , and v = ∑0≤jkPjv. Here Pjv denotes the projection of v on the eigenspace of d2/dx2 spanned by sin⁡jx and cos⁡jx for jN{0}. Just as an application of [11, Lemma 2] or [1, Lemma 2.2], we can modify slightly the proof of [15, Lemma 1] to obtain the next lemma.

Lemma 1. Let kN ∪ {0} and Γ be a nonnegative L1(0,2π)-function such that for a.e. x ∈ (0,2π),   Γ(x) ≤ 2k + 1 with strict inequality on a positive measurable subset of (0,2π). Then there exists a constant K1 > 0 such that

()
whenever pL1(0,2π) with 0 ≤ p(x) ≤ Γ(x) for a.e. x ∈ (0,2π) and uW2,1(0,2π) is a periodic function of period 2π with u(0) − u(2π) = u(0) − u(2π) = 0.

Proof. Just as in [20, Lemma 1], we can modify slightly the proof of [11, Lemma 2] or [1, Lemma 2.2] to obtain the fact that there exists a constant K1 > 0 such that

()
whenever pL1(0,2π) with 0 ≤ p(x) ≤ Γ(x) for a.e. x ∈ (0,2π) and uW2,1(0,2π) with u(0) − u(2π) = u(0) − u(2π) = 0. Let us extend u(x) and p(x)  2π periodically in x to all of R and then use the same notations for the periodic extensions as for the original functions. In this case, we have and
()
Since ,  p(x) ≥ 0 for a.e. x ∈ (0,2π), and for all v, wW2,1(0,2π) with v(0) − v(2π) = v(0) − v(2π) = 0 and w(0) − w(2π) = w(0) − w(2π) = 0, we have and
()
Combining (12) with (13), we have
()

Lemma 2. Let kN and Γ be a nonnegative L1(0,2π)-function such that for a.e. x ∈ (0,2π),   Γ(x) ≤ 2k − 1 with strict inequality on a positive measurable subset of (0,2π). Then there exists a constant K2 > 0 such that

()
whenever pL1(0,2π) with 0 ≤ p(x) ≤ Γ(x) for a.e. x ∈ (0,2π) and uW2,1(0,2π) is a periodic function of period 2π with u(0) − u(2π) = u(0) − u(2π) = 0. Here and for each vW2,1(0,2π) with v(0) − v(2π) = v(0) − v(2π) = 0.

Theorem 3. Let kN ∪ {0} and g : (0,2π) × RR be a Caratheodory function satisfying (H). Then for each hL1(0,2π) problem (1) has a solution u, provided that either (8) with −1 < β < 0 or (9) with β = 0 holds.

Proof. Let αR be fixed and 0 < α < 2k + 1. We consider the boundary value problems

()
for 0 ≤ t ≤ 1, which becomes the original problem when t = 1. Since 0 < α < 2k + 1, we observe from Lemma 1 that (17) has only a trivial solution when t = 0. To apply the Leray-Schauder continuation method, it suffices to show that solutions to (17) for 0 < t < 1 have an a priori bound in H1(0,2π). To this end, let θ : RR be a continuous function such that 0 ≤ θ ≤ 1,  θ(u) = 0 for |u| ≤ r0, and θ(u) = 1 for |u| ≥ 2r0. We define ,
()
and g2(x, u) = g(x, u)g1(x, u). Then g1, g2 : (0,2π) × RR are Caratheodory functions, such that for a.e. x ∈ (0,2π) and uR,  u ≠ 0
()
()
If u is a possible solution to (17) for some 0 < t < 1, then using (19), (20), and Lemma 1, we have
()
which implies that
()
for some constants C1, C2 > 0 independent of u. It remains to show that solutions to (17) for 0 < t < 1 have an a priori bound in H1(0,2π). We argue by contradiction and suppose that there exists a sequence {un} of periodic functions with period 2π and a corresponding sequence {tn} in (0,1) such that un is a solution to (17) with t = tn and for all n. Let ; then for all nN, and by (22) we have as n. Since and for all nN, we have a bounded sequence {Pkvn} in H1(0,2π). For simplicity, we may assume that vn converges to v in H1(0,2π) for some vN(L) with . In particular, vnv in C[0,2π]. Clearly, v(·−x0) ∈ N(L) and . It follows that un(x) → for each xR with v(x) > 0, and un(x)→− for each xR with v(x) < 0. Since and , we have
()
Multiplying each side of (17) by Pkvn(xx0), and then integrating them over [0,2π] when u = un and t = tn, we get
()
By (19) and the assumption of −1 < β ≤ 0, we have
()
for a.e. x ∈ (0,2π). Combining (22) with (25), we get that is bounded from below by an L1(0,2π)-function independent of n. By (20) and the assumption of −1 < β ≤ 0, we have
()
for a.e. x ∈ (0,2π), In particular, is bounded from below by an L1(0,2π)-function independent of n, which implies that is also so, , and
()
for all nN with un(xx0) ≠ 0. Here sign(w) = 1 if w > 0, sign(w) = 0 if w = 0, and sign(w) = −1 if w < 0. Applying Fatou’s lemma to the integral , we have
()
which is a contradiction when either (8) with −1 < β < 0 or (9) with β = 0 is satisfied. Hence, the proof of this theorem is complete.

By slightly modifying the proof of Theorem 3, we can apply Lemma 2 to obtain an existence theorem to (2) when condition (H) is replaced by (G) and either (8) with −1 < β < 0 or (9) with β = 0 is satisfied, which has been established in [20] for the case x0 = 0 when (9) with β = 0 is satisfied and in [9] for the case x0 = 0 when (8) with β = −1 is satisfied.

Theorem 4. Let kN and g : (0,2π) × RR be a Caratheodory function satisfying (G). Then for each hL1(0,2π) problem (2) has a solution u, provided that either (8) with −1 < β < 0 or (9) with β = 0 holds.

Conflicts of Interest

There are no conflicts of interest involved.

    The full text of this article hosted at iucr.org is unavailable due to technical difficulties.