Electronic Journal of Differential Equations, Vol. 2007(2007), No. 123, pp. 1–9. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu ftp ejde.math.txstate.edu (login: ftp) SOLVABILITY OF A FOUR-POINT BOUNDARY-VALUE PROBLEM FOR FOURTH-ORDER ORDINARY DIFFERENTIAL EQUATIONS JING GE, CHUANZHI BAI Abstract. In this paper we investigate the existence of solutions of a class of four-point boundary-value problems for fourth-order ordinary differential equations. Our analysis relies on a fixed point theorem due to Krasnoselskii and Zabreiko. 1. Introduction In recent years, boundary-value problems for second and higher order differential equations have been extensively studied. The monographs of Agarwal [1] and Agarwal, O’Regan, and Wong [2] contain excellent surveys of known results. Recently an increasing interest in studying the existence of solutions and positive solutions to boundary-value problems for higher order differential equations is observed; see for example [3, 4, 5, 6, 7, 8]. Very recently, Zhang, Chen and Lü [10] by using the upper and lower solution method investigated the fourth order nonlinear ordinary differential equation u(4) (t) = f (t, u(t), u00 (t)), 0 < t < 1, with the four-point boundary conditions u(0) = u(1) = 0, au00 (ξ1 ) − bu000 (ξ1 ) = 0, cu00 (ξ2 ) + du000 (ξ2 ) = 0, (1.1) (1.2) where a, b, c, d are nonnegative constants satisfying ad + bc + ac(ξ2 − ξ1 ) > 0, 0 ≤ ξ1 < ξ2 ≤ 1 and f ∈ C([0, 1] × R × R). They proved the following Lemma (a key lemma): Lemma 1.1 ([10, Lemma 2.2]). Suppose a, b, c, d, ξ1 , ξ2 are nonnegative constants satisfying 0 ≤ ξ1 < ξ2 ≤ 1, b−aξ1 ≥ 0, d−c+cξ2 ≥ 0 and δ = ad+bc+ac(ξ2 −ξ1 ) 6= 0. If u(t) ∈ C 4 [0, 1] satisfies • u(4) (t) ≥ 0, t ∈ (0, 1), • u(0) ≥ 0, u(1) ≥ 0, 2000 Mathematics Subject Classification. 34B10, 34B15. Key words and phrases. Four-point boundary-value problem; existence of solutions; fixed point theorem. c 2007 Texas State University - San Marcos. Submitted May 10, 2007. Published September 21, 2007. Supported by grants NNSFC-10771212 and 06KJB110010. 1 2 J. GE, C. BAI EJDE-2007/123 • au00 (ξ1 ) − bu000 (ξ1 ) ≤ 0, cu00 (ξ2 ) + du000 (ξ2 ) ≤ 0, then u(t) ≥ 0 and u00 (t) ≤ 0 for t ∈ [0, 1]. Unfortunately this Lemma is wrong as shown below. Counterexample to [10, Lemma 2.2]. Let u(t) = 31 t4 + 41 t3 − 43 t2 + 43 t which 1 belongs to C 4 [0, 1], ξ1 = 10 , ξ2 = 18 , a, b, c, d be nonnegative constants satisfying 15 7 1 1 ac 6= 0. Then we have b ≥ 10 a = aξ1 , d = 16 c > 8 c = (1 − ξ2 )c and δ = ad + bc + 40 u(4) (t) = 8 > 0, t ∈ (0, 1), u(0) = 0, u(1) = 0, h h 3i 8i 3 − b 8t + au00 (ξ1 ) − bu000 (ξ1 ) = a 4t2 + t − 2 3 t=1/10 2 t=1/10 3 143 a − 2 b ≤ 0, = −2 300 10 and h h 3 8i 3i cu00 (ξ2 ) + du000 (ξ2 ) = c 4t2 + t − + d 8t + 2 3 t=1/8 2 t=1/8 5 29 5 15 7 29 = − c + d = − c + · c = − c ≤ 0. 12 2 12 2 16 96 But 8 = −0.0031 < 0; 9 that is, [10, Lemma 2.2] is incorrect. So the conclusions of [10] should be reconsidered. The aim of this paper is to investigate the existence of solutions of the BVP (1.1)-(1.2) by using a fixed point theorem due to Krasnoselskii and Zabreiko in [9]. u 2. Main result First, we give some lemmas which are needed in our discussion of the main results. Lemma 2.1. Suppose a, b, c, d, ξ1 , ξ2 are nonnegative constants satisfying 0 ≤ ξ1 < ξ2 ≤ 1 and δ = ad + bc + ac(ξ2 − ξ1 ) 6= 0. If h ∈ C[0, 1], then the boundary-value problem v 00 (t) = h(t), 0 av(ξ1 ) − bv (ξ1 ) = 0, t ∈ [0, 1], (2.1) 0 cv(ξ2 ) + dv (ξ2 ) = 0, has a unique solution Z Z t 1 ξ2 v(t) = (t − s)h(s)ds + (a(ξ1 − t) − b)(c(ξ2 − s) + d)h(s)ds. δ ξ1 ξ1 (2.2) (2.3) Proof. By (2.1), it is easy to know that Z t (t − s)h(s)ds, v(t) = C1 + C2 t + 0 (2.4) EJDE-2007/123 FOUR-POINT BOUNDARY-VALUE PROBLEMS 3 where C1 , C2 are any two constants. Substituting (2.4) into boundary conditions (2.2), by a routine calculation, we get Z ξ1 Z ξ2 1 sh(s)ds + (aξ1 − b) C1 = (c(ξ2 − s) + d)h(s)ds, (2.5) δ 0 ξ1 Z ξ1 Z a ξ2 C2 = − h(s)ds − (c(ξ2 − s) + d)h(s)ds. (2.6) δ ξ1 0 Substituting (2.5) and (2.6) into (2.4), we obtain (2.3) which implies lemma. Remark 2.2. Let ξ1 = 0, ξ2 = 1, then (2.3) reduces to Z 1 v(t) = − G(t, s)h(s)ds, 0 where 1 G(t, s) = δ ( (as + b)(d + c(1 − t)), 0 ≤ s ≤ t ≤ 1, (at + b)(d + c(1 − s)), 0 ≤ t < s ≤ 1. Remark 2.3. Let 1 ((a(t − ξ1 ) + b)x3 + (c(ξ2 − t) + d)x2 ), δ ( s(1 − t), 0 ≤ s ≤ t ≤ 1, G1 (t, s) = t(1 − s), 0 ≤ t < s ≤ 1, ( 1 (a(s − ξ1 ) + b)(d + c(ξ2 − t)), ξ1 ≤ s ≤ t ≤ ξ2 , G2 (t, s) = δ (a(t − ξ1 ) + b)(d + c(ξ2 − s)), ξ1 ≤ t < s ≤ ξ2 . R(t) = In [10, Lemma 2.2] it is claimed that Z 1 Z u(t) = tx1 +(1−t)x0 − G1 (t, ξ)R(ξ)dξ+ 0 1 Z ξ2 G1 (t, ξ) 0 (2.7) G2 (ξ, s)h(s)dsdξ, (2.8) ξ1 is the solution of the boundary-value problem u(4) (t) = h(t), u(0) = x0 , 00 000 au (ξ1 ) − bu (ξ1 ) = x2 , 0 < t < 1, u(1) = x1 , cu00 (ξ2 ) + du000 (ξ2 ) = x3 . However (2.8) is wrong. Indeed, by Lemma 2.1, (2.8) should be replaced by Z 1 Z 1 u(t) = tx1 + (1 − t)x0 − G1 (t, ξ)R(ξ)dξ − G1 (t, η)v(η)dη, 0 0 where Z η (η − s)h(s)ds + v(η) = ξ1 1 δ Z ξ2 (a(ξ1 − η) − b)(c(ξ2 − s) + d)h(s)ds. ξ1 Remark 2.4. In [10, Theorem 3.1], the operator T : C[0, 1] → C[0, 1] is defined as Z 1 Z ξ2 T u(t) = G1 (t, η) G2 (η, s)f (s, u(s), u00 (s))dsdη, 0 ξ1 4 J. GE, C. BAI EJDE-2007/123 where G1 (t, s) and G2 (t, s) are as in Remark 2.2. By 2.1 and Remark 2.2, the definition of T is incorrect. In fact, the operator T should be defined as Z 1 Z η T u(t) = G1 (t, η) (s − η)f (s, u(s), u00 (s))dsdη 0 + ξ1 1 δ 1 Z Z ξ2 G1 (t, η) (b − a(ξ1 − η))(c(ξ2 − s) + d)f (s, u(s), u00 (s))dsdη. ξ1 0 The following well-known fixed point theorem [9] will play an important role in the proof of our theorem. Lemma 2.5. Let X be a Banach space, and F : X → X be completely continuous. Assume that A : X → X is a bounded linear operator such that 1 is not an eigenvalue of A and kF (x) − A(x)k lim = 0. kxk kxk→∞ Then F has a fixed point in X. Let X = C 2 [0, 1] be endowed with the norm by kuk0 = max{kuk, ku00 k}, where kuk = max0≤t≤1 |u(t)|. We are now in a position to present and prove our main result. Let (H1) a, b, c, d, ξ1 , ξ2 are nonnegative constants satisfying 0 ≤ ξ1 < ξ2 ≤ 1, b − aξ1 ≥ 0 and δ = ad + bc + ac(ξ2 − ξ1 ) 6= 0, (H2) f (t, u, v) = p(t)g(u) + q(t)h(v), where g, h : R → R are continuous with g(u) h(v) = λ, lim = µ, u→∞ u v→∞ v where p, q ∈ C[0, 1]. Moreover, there exists some t0 ∈ [0, 1] such that p(t0 )g(0)+q(t0 )h(0) 6= 0, and there exists a continuous nonnegative function w : [0, 1] → R+ such that |p(s)| + |q(s)| ≤ w(s) for each s ∈ [0, 1]. Theorem 2.6. Assume (H1)–(H2). If max{|λ|, |µ|} < min L11 , L12 , where Z Z 1 1 h ξ1 3 L1 = τ (2 − τ )w(τ )dτ + (1 − τ )3 (1 + τ )w(τ )dτ 12 0 ξ1 Z i 2(b − aξ1 ) + a ξ2 + (c(ξ2 − τ ) + d)w(τ )dτ , δ ξ1 lim and Z 1 (1 − τ )w(τ )dτ + L2 = ξ1 1 δ Z ξ2 (b + a(1 − ξ1 ))(c(ξ2 − τ ) + d)w(τ )dτ, ξ1 then BVP (1.1) and (1.2) has at least one nontrivial solution u ∈ C 2 [0, 1]. Proof. Define an operator F : C 2 [0, 1] → C 2 [0, 1] by Z 1 Z s F u(t) := G1 (t, s) (τ − s)[p(τ )g(u(τ )) + q(τ )h(u00 (τ )]dτ ds 0 ξ1 Z ξ2 1 G1 (t, s) (b − a(ξ1 − s))(c(ξ2 − τ ) + d) δ 0 ξ1 × p(τ )g(u(τ )) + q(τ )h(u00 (τ ) dτ ds, Z + 1 (2.9) EJDE-2007/123 FOUR-POINT BOUNDARY-VALUE PROBLEMS 5 where G1 (t, s) is as in (2.7). Then by Lemma 2.1 and Remark 2.4, we easily know that the fixed points of F are the solutions to the boundary-value problem (1.1) and (1.2). It is well known that the operator F is a completely continuous operator. Now, we consider the following boundary-value problem u(4) (t) = λp(t)u(t) + µq(t)u00 (t), 0<t<1 u(0) = u(1) = 0, 00 000 au (ξ1 ) − bu (ξ1 ) = 0, 00 (2.10) 000 cu (ξ2 ) + du (ξ2 ) = 0. Define 1 Z Au(t) := s Z (τ − s)[λp(τ )u(τ ) + µq(τ )u00 (τ )dηds G1 (t, s) 0 1 + δ ξ1 Z 1 Z ξ2 (b − a(ξ1 − s))(c(ξ2 − τ ) + d) G1 (t, s) (2.11) ξ1 0 [λp(τ )u(τ ) + µq(τ )u00 (τ )]dηds. Obviously, A is a bounded linear operator. Furthermore, the fixed point of A is a solution of the BVP (2.10) and conversely. We now assert that 1 is not an eigenvalue of A. In fact, if λ = 0 and µ = 0, then the BVP (2.10) has no nontrivial solution. If λ 6= 0 or µ 6= 0, suppose the BVP (2.10) has a nontrivial solution u and kuk0 > 0, then Z 1 Z s |Au(t)| ≤ G1 (t, s) |(τ − s)[λp(τ )u(τ ) + µq(τ )u00 (τ )]|dτ ds 0 ξ1 1 + δ Z 1 ξ2 Z (b − a(ξ1 − s))(c(ξ2 − τ ) + d) G1 (t, s) 0 ξ1 × [λp(τ )u(τ ) + µq(τ )u00 (τ )]dτ ds Z 1 Z s ≤ s(1 − s) (s − τ )[|λ||p(τ )||u(τ )| + |µ||q(τ )||u00 (τ )|]dτ ds 0 + ξ1 1 δ Z 1 Z ξ2 s(1 − s) 0 (b − a(ξ1 − s))(c(ξ2 − τ ) + d) ξ1 × [|λ||p(τ )||u(τ )| + |µ||q(τ )||u00 (τ )|]dτ ds Z Z s hZ 1 1 1 ≤ s(1 − s) (s − τ )[|λ||p(τ )| + |µ||q(τ )|]dτ ds + s(1 − s) δ 0 0 ξ1 Z ξ2 i × (b − a(ξ1 − s))(c(ξ2 − τ ) + d)[|λ||p(τ )| + |µ||q(τ )|]dτ ds kuk0 ξ1 Z 1 h ξ1 3 τ (2 − τ )(|λ||p(τ )| + |µ||q(τ )|)dτ = 12 0 Z 1 + (1 − τ )3 (1 + τ )(|λ||p(τ )| + |µ||q(τ )|)dτ ξ1 Z i 2(b − aξ1 ) + a ξ2 + (c(ξ2 − τ ) + d)(|λ||p(τ )| + |µ||q(τ )|)dτ kuk0 δ ξ1 Z 1 Z ξ1 h 1 τ 3 (2 − τ )w(τ )dτ + (1 − τ )3 (1 + τ )w(τ )dτ ≤ max{|λ|, |µ|} 12 0 ξ1 6 J. GE, C. BAI + 2(b − aξ1 ) + a δ Z ξ2 EJDE-2007/123 i (c(ξ2 − τ ) + d)w(τ )dτ kuk0 , t ∈ [0, 1], ξ1 which implies that |Au(t)| ≤ max{|λ|, |µ|}L1 kuk0 < 1 L1 kuk0 = kuk0 . L1 On the other hand, we have Z t |(Au)00 (t)| = (s − t)[λp(s)u(s) + µq(s)u00 (s)]ds ξ1 Z 1 ξ2 (b − a(ξ1 − t))(c(ξ2 − s) + d)[λp(s)u(s) + µq(s)u00 (s)]ds + δ ξ1 hZ 1 ≤ (1 − s)(|λ||p(s)| + |µ||q(s)|)ds ξ1 1 + δ Z ξ2 i (b + a(1 − ξ1 ))(c(ξ2 − s) + d)(|λ||p(s)| + |µ||q(s)|)ds kuk0 ξ1 ≤ max{|λ|, |µ|}L2 kuk0 < 1 L2 kuk0 = kuk0 , L2 t ∈ [0, 1]. Then kAuk0 < kuk0 . This contradiction means that BVP (2.10) has no nontrivial solution. Hence, 1 is not an eigenvalue of A. Finally, we prove that lim kuk0 →∞ According to limu→∞ R > 0 such that g(u) u kF u − Auk0 = 0. kuk0 = λ and limv→∞ |g(u) − λu| < ε|u|, h(v) v = µ, for any ε > 0, there must be |h(v) − µv| < ε|v|, |u|, |v| > R. Set R∗ = max{max|u|≤R |g(u)|, max|v|≤R |h(v)|} and select M > 0 such that R∗ + max{|λ|, |µ|} < εM . Denote E1 = {t ∈ [0, 1] : |u(t)| ≤ R, |v(t)| > R}, E2 = {t ∈ [0, 1] : |u(t)| > R, |v(t)| ≤ R}, E3 = {t ∈ [0, 1] : max{|u(t)|, |v(t)|} ≤ R}, E4 = {t ∈ [0, 1] : min{|u(t)|, |v(t)|} > R}. Thus for any u ∈ C 2 [0, 1] with kuk0 > M , when t ∈ E1 , we have |g(u(t)) − λu(t)| ≤ |g((u(t))| + |λ||u(t)| ≤ R∗ + |λ|R < εM < εkuk0 , and |h(v(t)) − µv(t)| < ε|v(t)| ≤ εkvk0 . Similarly, we conclude that for any u ∈ C 2 [0, 1] with kuk0 > M , when t ∈ Ei (i = 2, 3, 4), we also have that |g(u(t)) − λu(t)| < εkuk0 , |h(v(t)) − µv(t)| < εkvk0 . EJDE-2007/123 FOUR-POINT BOUNDARY-VALUE PROBLEMS 7 Hence, we get |F u(t) − Au(t)| Z s Z 1 (τ − s)(p(τ )[g(u(τ )) − λu(τ )] + q(τ )[h(u00 (τ )) − µu00 (τ )])dτ ds G1 (t, s) = ξ1 0 + 1 δ 1 Z Z ξ2 (b − a(ξ1 − s))(c(ξ2 − τ ) + d) G1 (t, s) 0 ξ1 × (p(τ )[g(u(τ )) − λu(τ )] + q(τ )[h(u00 (τ )) − µu00 (τ )])dτ ds Z s hZ 1 ≤ G1 (s, s) (s − τ )(|p(τ )| + |q(τ )|)dτ ds 0 ξ1 Z ξ2 i 1 G1 (s, s) (b − a(ξ1 − s))(c(ξ2 − τ ) + d)(|p(τ )| + |q(τ )|)dτ ds εkuk0 + δ 0 ξ1 Z ξ1 Z 1 h 1 ≤ τ 3 (2 − τ )w(τ )dτ + (1 − τ )3 (1 + τ )w(τ )dτ 12 0 ξ1 Z i 2(b − aξ1 ) + a ξ2 (c(ξ2 − τ ) + d)w(τ )dτ εkuk0 . + δ ξ1 Z 1 = εL1 kuk0 . (2.12) On the other hand, we have |(F u − Au)00 (t)| Z t (s − t)(p(s)[g(u(s)) − λu(s)] + q(s)[h(u00 (s)) − µu00 (s)])ds = ξ1 + 1 δ Z ξ2 (b − a(ξ1 − t))(c(ξ2 − s) + d) ξ1 × (p(s)[g(u(s)) − λu(s)] + q(s)[h(u00 (s)) − µu00 (s)])ds Z hZ 1 i 1 ξ2 ≤ (1 − s)w(s)ds + (b + a(1 − ξ1 ))(c(ξ2 − s) + d)w(s)ds εkuk0 δ ξ1 ξ1 = εL2 kuk0 . Combining the above inequality with (2.12), we have lim kuk0 →∞ kF u − Auk0 = 0. kuk0 Lemma 2.5 now guarantees that BVP (1.1) and (1.2) has a solution u∗ ∈ C 2 [0, 1]. Obviously, u∗ 6= 0 when p(t0 )g(0) + q(t0 )h(0) 6= 0 for some t0 ∈ [0, 1]. In fact, if u∗ = 0, then (0)(4) = p(t0 )g(0) + q(t0 )h(0) 6= 0 will lead to a contradiction. This completes the proof. Example 2.7. Consider the fourth-order four-point boundary-value problem u(4) (t) = t sin 2πt 1 u(t) − tecos t cos u00 (t), 2 t +1 2 u(0) = u(1) = 0, u00 (1/3) − u000 (1/3) = 0, 0 < t < 1, u00 (2/3) + u000 (2/3) = 0. (2.13) 8 J. GE, C. BAI EJDE-2007/123 To show (2.13) has at least one nontrivial solution we apply Theorem 2.6 with 2πt 1 cos t p(t) = t sin , g(u) = u, h(u) = cos u, a = b = c = d = 1, ξ1 = 1/3 t2 +1 , q(t) = 2 te and ξ2 = 2/3. Clearly (H1) is satisfied. Obviously, p(t0 )g(0) + q(t0 )h(0) = 1 cos t0 t0 e 6= 0, 2 t0 ∈ (0, 1]. Since |p(s)| + |q(s)| ≤ ( 2e + 1)s := w(s) for each s ∈ [0, 1], we have Z 1 e h Z 1/3 4 2 +1 L1 = τ (2 − τ )dτ + (e + 1) (1 − τ )3 (1 + τ )τ dτ 12 0 1/3 Z 2/3 i 5 + ( − τ )τ dτ , 1/3 3 L2 = h e +1 2 Z 1 τ (1 − τ )dτ + 1/3 5 7 Z 2/3 1/3 i 5 − τ τ dτ . 3 By simple calculation we easily know that L1 < L2 < 1 e + 1 < 1. 3 2 Notice g(u) = 1, u→∞ u λ = lim h(u) = 0, u→∞ u µ = lim we have max{λ, µ} < 1 < min{ 1 1 , }. L1 L2 So (H2) is satisfied. Thus, Theorem 2.6 now guarantees that BVP (2.13) has at least one nontrivial solution u ∈ C 2 [0, 1]. References [1] R. P. Agarwal, Focal Boundary Value Problems for Differential and Difference Equations, Kluwer Academic, Dordrecht, 1998. [2] R. P. Agarwal, D. O’Regan, P. J.Y . Wong; Positive Solutions of Differential, Difference, and Integral Equations, Kluwer Academic, Dordrecht, 1998. [3] D. R. Anderson, J. M. Davis; Multiple solutions and eigenvalues for third-order right focal boundary-value problem, J. Math. Anal. Appl. 267 (2002), 135C157. [4] Z. Bai, H. Wang; On positive solutions of some nonlinear fourth-ordr beam equations, J. Math. Anal. Appl. 270 (2002) 357-368. [5] J. R. Graef, B. Yang; On a nonlinear boundary-value problem for fourth order equations, Appl. Anal. 72 (1999) 439-448. [6] J. R. Graef, B. Yang; Positive solutions to a multi-point higher order boundary-value problem, J. Math. Anal. Appl. 316 (2006), 409-421. [7] J. R. Graef, J. Henderson, B. Yang; Positive solutions of a nonlinear higher order boundaryvalue problem, Electron. J. Differential Equations, Vol. 2007(2007), No. 45, pp. 1-10. [8] Z. Hao, L. Liu, L. Debnath; A necessary and sufficiently condition for the existence of positive solution of fourth-order singular boundary-value problems, Appl. Math. Lett. 16 (2003) 279285. [9] M. A. Krasnoselskii, P. P. Zabreiko; Geometrical Methods of Nonlinear Analysis, SpringerVerlag, New York, 1984. [10] Q. Zhang, S. Chen, J. Lü; Upper and lower solution method for fourth-order four-point boundary-value problems, J. Comput. Appl. Math. 196 (2006) 387-393. EJDE-2007/123 FOUR-POINT BOUNDARY-VALUE PROBLEMS 9 Jing Ge Department of Mathematics, Huaiyin Teachers College, Huaian, Jiangsi 223300, China E-mail address: gejing0512@163.com Chuanzhi Bai Department of Mathematics, Huaiyin Teachers College, Huaian, Jiangsi 223300, China E-mail address: czbai8@sohu.com