Theoretical Mathematics & Applications, vol.4, no.2, 2014, 31-44 ISSN: 1792-9687 (print), 1792-9709 (online) Scienpress Ltd, 2014 Multiple Positive solutions of Sturm-Liouville problems for second order singular and impulsive differential equations Ying He1 Abstract In this paper,we study the existence of multiple positive solutions of nonlinear singular two-point boundary value problems for second-order impulsive differential equations.The proof is based on the theory of fixed point index in cones. Mathematics Subject Classification: 34B15 Keywords: Multiple positive solutions; Singular two-point boundary value problem; Second-order impulsive differential equations; Fixed point index in cones. 1 School of Mathematics and Statistics, Northeast Petroleum University, Daqing 163318, P.R.China. Project supported by Heilongjiang province education department natural science research item, China. (12541076). E-mail: heying65338406@163.com Article Info: Received : April 20, 2014. Revised : June 4, 2014. Published online : June 30, 2014. 32 1 Multiple Positive solutions of Sturm-Liouville problems... Introduction Impulsive differential equations play a very important role in modern applied mathematics due to their deep physical background and broad application. In this paper,we consider the existence of multiple positive solutions of two-point boundary value problems for nonlinear second-order singular and impulsive differential equations: −Lu = h(x)g(x, u), x ∈ I 0, 0 −∆(pu )|x=xk = Ik (u(xk )), k = 1, 2, · · · , m, (1.1) ∆(pu)|x=xk = I k (u(xk )), k = 1, 2, · · · , m, 0 R1 (u) = αu(0) − βu (0) = 0, R (u) = γu(1) + δu0 (1) = 0, 2 here Lu = (p(x)u0 )0 + q(x)u is sturm-liouville operator, I = [0, 1], I 0 = I \ {x1 , x2 , · · · , xm } and 0 < x1 < x2 < · · · < xm < 1 are given, R+ = [0, +∞), g ∈ C(I ×R+ , R+ ), Ik , I k ∈ C(R+ , R+ ), ∆(pu0 )|x=xk = p(xk )u0 (x+ k )− + − + 0 − 0 + 0 − p(xk )u (xk ), ∆(pu)|x=xk = p(xk )u(xk )−p(xk )u(xk ) u (xk ), u(xk ) ( u (xk ), u(x− k )) 0 denote the right limit ( left limit) of u (x) and u(x) at x = xk respectively,h(x) ∈ C(I, R+ ) and may be singular at x = 0 or x = 1. Throughout this paper, we always suppose that (S1 ) p(x) ∈ C 1 ([0, 1], R), p(x) > 0, q(x) ∈ C([0, 1], R), q(x) ≤ 0, α, β, γ, δ ≥ 0, ρ = βγ + αγ + αδ > 0. It is well known that there are abundant results about the existence of positive solutions of boundary value problems for second order impulsive differential equations. Some works can be found in [1, 3, 6 − 9] and references therein.They,mainly investigated the case p(x) = 1 and q(x) = 0. In this paper ,we will consider the case p(x) 6= 1,and q(x) 6= 0. Here we also mention that second order dynamic inclusions on time scales with impulses has been studied in [2] .We obtain the existence results of positive solutions,by means of the fixed point index theorem in cones under some conditions on g(x, u) concerning the first eigenvalue corresponding to the relevant linear operator. To conclude the introduction,we introduce the following notation: Ik (u) I k (u) g(x, u) , I0 (k) = lim inf , I 0 (k) = lim inf + + x∈[a,b] u→0 u→0 u u u g0 = lim inf min + u→0 Ik (u) g(x, u) I k (u) , I∞ (k) = lim inf , I ∞ (k) = lim inf ; u→+∞ u→+∞ x∈[a,b] u u u g∞ = lim inf min u→+∞ 33 Ying He Moreover,for the simplicity in the following discussion,we introduce the following hypotheses. (H1 ) : m P σ (I0 (k)φ1 (xk ) + I 0 (k)φ01 (xk )) g0 + k=1 > λ1 , Rb φ (x)h(x)dx a 1 m P σ (I∞ (k)φ1 (xk ) + I ∞ (k)φ01 (xk )) g∞ + k=1 > λ1 . Rb φ (x)h(x)dx 1 a here σ = min{ m(a) , n(b) } (see section 2), and φ1 (x) is the eigenfunction related m(1) n(0) to the smallest eigenvalue λ1 of the eigenvalue problem −Lφ = λφh, R1 (φ) = R2 (φ) = 0. (H2 ) : There is a p > 0 such that 0 ≤ u ≤ p and 0 ≤ x ≤ 1 implies g(x, u) ≤ ηp, Ik (u) ≤ ηk p, I¯k (u) ≤ η̄k p here η, ηk , η̄k ≥ 0, η + m P (ηk + η̄k ) > 0, η k=1 R1 0 G(y, y)h(y)dy + m P G(xk , xk )(ηk + k=1 η̄k ) < 1 and G(x, y) is the Green’s function of boundary value problem −Lu = 0, R1 (u) = R2 (u) = 0 (see section 2). R0 (H3 ) : 0 < 1 G(y, y)h(y)dy < +∞ 2 Preliminary Notes In this paper, we shall consider the following space + P C(I, R) = {u ∈ C(I, R); u|(xk ,xk+1 ) ∈ C(xk , xk+1 ), u(x− k ) = u(xk ), ∃ u(xk ), k = 1, 2, · · · , m} P C 0 (I, R) = {u ∈ C(I, R); u0 |(xk ,xk+1 ) ∈ C(xk , xk+1 ), u0 (x− k) = 0 0 + u (xk ), ∃ u (xk ), k = 1, 2, · · · , m} with the norm kukP C = sup |u(x)|, x∈[0,1] kukP C 0 = max{kukP C , ku0 kP C }, Then P C(I, R),P C 0 (I, R) are Banach spaces. Definition 2.1. A function u ∈ P C 0 (I, R) ∩ C 2 (I 0 , R) is a solution of (1.1), if it satisfies the differential equation Lu + h(x)g(x, u) = 0, x ∈ I0 and the function u satisfies conditions ∆(pu0 )|x=xk = −Ik (u(xk )), ∆(pu)|x=xk = I k (u(xk )) and R1 (u) = R2 (u) = 0. 34 Multiple Positive solutions of Sturm-Liouville problems... Let Q = I × I and Q1 = {(x, y) ∈ Q|0 ≤ x ≤ y ≤ 1}, Q2 = {(x, y) ∈ Q|0 ≤ y ≤ x ≤ 1}. Let G(x, y) is the Green’s function of the boundary value problem −Lu = 0, R1 (u) = R2 (u) = 0. Following from [4], G(x, y) can be written by ( m(x)n(y) , (x, y) ∈ Q1 , ω G(x, y) := m(y)n(x) , (x, y) ∈ Q2 . ω (2.1) Lemma 2.2. Suppose that (S1 ) holds, then the Green’s function G(x, y), defined by (2.1), possesses the following properties: (i): m(x) ∈ C 2 (I, R) is increasing and m(x) > 0, x ∈ (0, 1]. (ii): n(x) ∈ C 2 (I, R) is decreasing and n(x) > 0, x ∈ [0, 1). (iii): (Lm)(x) ≡ 0, m(0) = β, m0 (0) = α. (iv): (Ln)(x) ≡ 0, n(1) = δ, n0 (1) = −γ. (v): ω is a positive constant. Moreover, p(x)(m0 (x)n(x) − m(x)n0 (x)) ≡ ω. (vi): G(x, y) is continuous and symmetrical over Q. (vii): G(x, y) has continuously partial derivative over Q1 , Q2 . (viii): For each fixed y ∈ I, G(x, y) satisfies LG(x, y) = 0 for x 6= y, x ∈ I. Moreover, R1 (G) = R2 (G) = 0 for y ∈ (0, 1). 0 (viiii): Gx has discontinuous point of the first kind at x = y and G0x (y + 0, y) − G0x (y − 0, y) = − 1 , y ∈ (0, 1). p(y) Following from Lemma2.2, it is easy to see that: G(x, y) ≤ G(y, y) = m(y)n(y) , x, y ∈ [0, 1] ω G(x, y) ≥ σG(y, y), x ∈ [a, b], y ∈ [0, 1], (2.2) where a ∈ (0, t1 ], b ∈ [tm , 1), 0 < σ = min{ m(a) , n(b) } < 1 m(1) n(0) Consider the linear Sturm-Liouvile problem −(Lu)(x) = λu(x)h(x), R1 (u) = R2 (u) = 0. By the Sturm-Liouvile theory of ordinary differential equations, we know that there exists an eigenfunction φ1 (x) with respect to the first eigenvalue λ1 > 0 such that φ1 (x) > 0 for x ∈ (0, 1). 35 Ying He Let E be a Banach space and K ⊂ E be a closed convex cone in E. For r > 0, let Kr = {u ∈ K : ||u|| < r} and ∂Kr = {u ∈ K : ||u|| = r}. The following two Lemmas are needed in our argument. Lemma 2.3. Let Φ : K → K be a continuous and completely continuous mapping and Φu 6= u for u ∈ ∂Kr . Thus the following conclusions hold: (i) If ||u|| ≤ kΦuk for u ∈ ∂Kr , then i(Φ, Kr , K) = 0; (ii) If kuk ≥ kΦuk for u ∈ ∂Kr , then i(Φ, Kr , K) = 1. Lemma 2.4. Let Φ : K → K be a continuous and completely continuous mapping. Suppose that the following two conditions are satisfied: (i) inf ||Φu|| > 0; u∈∂Kr (ii) µΦu 6= u for every u ∈ ∂Kr and µ ≥ 1. Then, i(Φ, Kr , K) = 0. In applications below, we take E = C(I, R) and define K = {u ∈ C(I, R) : u(x) ≥ σkuk, x ∈ [a, b]}. One may readily verify that K is a cone in E. Define an operator Φ : K → K by Z 1 (Φu)(x) = G(x, y)h(y)g(y, u(y))dy+ 0 X G(x, xk )(Ik (u(xk ))+I k (u(xk ))), x ∈ I. 0<xk <x It follows form (H3 ) that φ is well defined. Lemma 2.5. If (H3 ) is satisfied,then Φ : K → K is continuous and completely continuous, Moreover, Φ(K) ⊂ K. Proof By the property of continuous of g(x, u), Ik (x), I k (x), it is easy to see that Φ : K → K is continuous and completely continuous. Thus we only need to show Φ(K) ⊂ K. In fact, for u ∈ K, by using inequalities (2.2) and (H3 ), we have that Z 1 kΦuk ≤ G(y, y)h(y)g(y, u(y))dy+ 0 X 0<xk <x G(xk , xk )(Ik (u(xk ))+I¯k (u(xk ))) < +∞ 36 Multiple Positive solutions of Sturm-Liouville problems... On the other hand ,for any x ∈ [a, b],by (2.2) we obtain Z 1 X (Φu)(x) = G(x, y)h(y)g(y, u(y))dy + G(x, xk )(Ik (u(xk )) + I k (u(xk ))) Z 0 0<xk <x X a ≥ G(x, y)h(y)g(y, u(y))dy + b G(x, xk )(Ik (u(xk )) + I k (u(xk ))) 0<xk <x ÃZ 0 ≥ σ G(y, y)h(y)g(y, u(y))dy + 1 X ! G(xk , xk )(Ik (u(xk )) + I k (u(xk ))) 0<xk <x ≥ σkΦuk Thus, Φ(K) ⊂ K. Lemma 2.6. If u is a fixed point of the operator Φ, then u is a solution of problem (1.1). 3 Main Results Lemma 3.1. If (H2 ) is satisfied, then i(Φ, Kp , K) = 1. Proof Let u ∈ K with kuk = p. It follows from (H2 ) that Z 1 kΦuk ≤ G(y, y)h(y)g(y, u(y))dy + 0 m X G(xk , xk )(Ik (u(xk )) + I k (u(xk ))) k=1 Z 1 ≤ p[η G(y, y)h(y)dy + 0 m X G(xk , xk )(ηk + η k )] < p = kuk. k=1 Thus kΦuk < kuk, ∀ u ∈ ∂Kp . It is obvious that Φu 6= u for u ∈ ∂Kp . Therefore, i(Φ, Kp , K) = 1, here we use Lemma2.3. Theorem 3.2. Assume that (H1 ) − (H3 ) are satisfied. Then problem (1.1) has at least two positive solutions u1 and u2 with 0 < ||u1 || < p < ||u2 ||. 37 Ying He Proof According to Lemma 3.1, we have that i(Φ, Kp , K) = 1. (3.1) Since (H1 ) holds, then there exists 0 < ε < 1 such that m P σ (I0 (k)φ1 (xk ) + I 0 (k)φ01 (xk )) ] > λ1 , Rb φ (x)h(x)dx 1 a k=1 (1 − ε)[g0 + σ m P (I∞ (k)φ1 (xk ) + I ∞ (k)φ01 (xk )) ] > λ1 . Rb φ (x)h(x)dx a 1 k=1 (1 − ε)[g∞ + (3.2) By the definitions of g0 , I0 , one can find 0 < r0 < p such that g(x, u) ≥ g0 (1−ε)u, Ik (u) ≥ I0 (k)(1−ε)u, I k (u) ≥ I 0 (k)(1−ε)u ∀ x ∈ [a, b], 0 < u < r0 . Let r ∈ (0, r0 ), then for u ∈ ∂Kr , we have u(x) ≥ σkuk = σr > 0. x ∈ [a, b] Thus 1 (Φu)( ) = 2 Z 1 0 Z 0<xk < 2 b ≥ a X 1 1 G( , y)h(y)g(y, u(y))dy + G( , xk )(Ik (u(xk )) + I k (u(xk ))) 2 2 1 X 1 1 G( , xk )(Ik (u(xk )) + I k (u(xk ))) G( , y)h(y)g(y, u(y))dy + 2 2 1 0<xk < 2 Z b ≥ g0 (1 − ε) a + (1 − ε) X X 1 1 G( , y)h(y)u(y)dy + (1 − ε) G( , xk )I0 (k)u(xk ) 2 2 1 0<xk < 2 1 G( , xk )I 0 (k)u(xk ) 2 1 0<xk < 2 Z b ≥ (1 − ε)σr g0 a X 1 1 G( , y)h(x)dy + G( , xk )(I0 (k) + I 0 (k)) > 0 2 2 1 0<xk < 2 from which we see that inf ||Φu|| > 0, namely, hypothesis (i) of Lemma 2.4 u∈∂Kr holds. Next we show that µΦu 6= u for any u ∈ ∂Kr and µ ≥ 1. 38 Multiple Positive solutions of Sturm-Liouville problems... If this is not true, then there exist u0 ∈ ∂Kr and µ0 ≥ 1 such that µ0 Φu0 = u0 . Note that u0 (x) satisfies Lu0 + µ0 h(x)g(x, u0 (x)) = 0, x ∈ I 0, 0 −∆(pu0 )|x=xk = µ0 Ik (u0 (xk )), k = 1, 2, · · · , m, (3.3) ∆(pu0 )|x=xk = µ0 I k (u0 (xk )), k = 1, 2, · · · , m, αu0 (0) − βu00 (0) = 0, γu (1) + δu0 (1) = 0. 0 0 Multiply equation (3.3) by φ1 (x) and integrate from a to b, note that Z b Z x1 0 0 φ1 (x)[(p(x)u0 (x)) + q(x)u0 (x)]dx = φ1 (x)[(p(x)u00 (x))0 + q(x)u0 (x)]dx + a m−1 X Z xk+1 xk k=1 b Z + = a φ1 (x)[(p(x)u00 (x))0 + q(x)u0 (x)]dx φ1 (x)[(p(x)u00 (x))0 + q(x)u0 (x)]dx xm Z x1 Z 0 0 0 φ1 (x1 )p(x1 )u0 (x1 − 0) − p(x)u0 (x)φ1 (x)dx + a + [φ1 (xk+1 )p(xk+1 )u00 (xk+1 − 0) − φ1 (xk )p(xk )u00 (xk + 0) − xk = Z p(x)u00 (x)φ01 (x)dx φ1 (xm )p(xm )u00 (xm − m X xk+1 + q(x)u0 (x)φ1 (x)dx] xk Z b + 0) − xm ∆(p(xk )u00 (xk ))φ1 (xk ) Z p(x)u00 (x)φ01 (x)dx Z . Also note that Z b Z 0 0 p(x)φ1 (x)u0 (x)dx = a Z + = b a x1 q(x)u0 (x)φ1 (x)dx xm Z p(x)φ01 (x)u00 (x)dx p(x)φ01 (x)du0 (x) a b + + b + q(x)φ1 (x)u0 (x)dx a m−1 X Z xk+1 k=1 xk p(x)φ01 (x)du0 (x) p(x)φ01 (x)du0 (x) xm m X − − b − k=1 = q(x)u0 (x)φ1 (x)dx a m−1 X Zk=1xk+1 − x1 k=1 m X k=1 Z ∆(p(xk )u0 (xk ))φ01 (xk ) b − a Z ∆(p((xk )u0 (xk ))φ01 (xk ) u0 (x)(p(x)φ01 (x))0 dx Z b + b u0 (x)q(x)φ1 (x)dx + λ1 a h(x)φ1 (x)u0 (x)dx a 39 Ying He Z b a + m X φ1 (x)[(p(x)u00 (x))0 + q(x)u0 (x)]dx = − = b h(x)φ1 (x)u0 (x)dx a k=1 m X ∆(p(xk )u00 (xk ))φ1 (xk ) k=1 Z ∆(p(xk )u0 (xk ))φ01 (xk ) − λ1 m X Z I k (u0 (xk ))φ01 (xk )) µ0 (Ik (u0 (xk ))φ1 (xk ) + = a m X u0 (x)h(x)φ1 (x)dx a k=1 So we obtain Z b λ1 u0 (x)h(x)φ1 (x)dx b − λ1 µ0 (Ik (u0 (xk ))φ1 (xk ) + I k (u0 (xk ))φ01 (xk )) k=1 Z b + µ0 φ1 (x)h(x)g(x, u0 (x))dx a ≥ (1 − ε) m X u0 (xk )(I0 (k)φ1 (xk ) + I 0 (k)φ01 (xk )) k=1 Z b φ1 (x)u0 (x)h(x)dx + (1 − ε)g0 a Since u0 (x) ≥ σ||u0 || = σr, we have Rb a φ1 (x)u0 (x)h(x)dx > 0 and m P u0 (xk )(I0 (k)φ1 (xk )+ k=1 I 0 (k)φ01 (xk )) > 0. So from the above inequality we see that λ1 > (1 − ε)g0 . Thus Z b m X [λ1 − (1 − ε)g0 ] u0 (x)h(x)φ1 (x)dx ≥ (1 − ε) (I0 (k)φ1 (xk ) + I 0 (k)φ01 (xk ))u0 (xk ) a k=1 ≥ (1 − ε)σr m X (I0 (k)φ1 (xk ) + I 0 (k)φ01 (xk )). k=1 Rb Rb Since a u0 (x)h(x)φ1 (x)dx ≤ r a φ1 (x)h(x)dx, we have Z b [λ1 − (1 − ε)g0 ] h(x)φ1 (x)dx ≥ (1 − ε)σ a m X (I0 (k)φ1 (xk ) + I 0 (k)φ01 (xk )), k=1 which contradicts (3.2) again. Hence Φ satisfies the hypotheses of Lemma 2.4 in Kr . Thus i(Φ, Kr , K) = 0. (3.4) 40 Multiple Positive solutions of Sturm-Liouville problems... On the other hand, from (H1 ), there exists H > p such that for any x ∈ [a, b], u ≥ H. g(x, u) ≥ g∞ (1 − ε)u, Ik (u) ≥ I∞ (k)(1 − ε)u, I k (u) ≥ I ∞ (k)(1 − ε)u, (3.5) Let m P max |Ik (u) − I∞ (k)(1 − ε)u| + C = max max |g(x, u) − g∞ (1 − ε)u| + k=1 0≤u≤H 0≤u≤H a≤x≤b m P max |I k (u) − I ∞ (k)(1 − ε)u|. It is clear that for any x ∈ [a, b], u ≥ 0. k=1 0≤u≤H g(x, u) ≥ g∞ (1−ε)u−C, Ik (u) ≥ I∞ (k)(1−ε)u−C, I k (u) ≥ I ∞ (k)(1−ε)u−C, (3.6) H Choose R > R0 := max{ σ , p} and let u ∈ ∂KR . Since u(x) ≥ σ||u|| = σR > H for x ∈ [a, b], from (3.5) we see that g(x, u(x)) ≥ g∞ (1 − ε)u(x) ≥ σg∞ (1 − ε)R, ∀ x ∈ [a, b]. Ik (u(xk ) ≥ σI∞ (k)(1 − ε)R, I k (u(xk ) ≥ σI ∞ (k)(1 − ε)R. Essentially the same reasoning as above yields inf ||Φu|| > 0. Next we show u∈∂KR that if R is large enough, then µΦu 6= u for any u ∈ ∂KR and µ ≥ 1. In fact, if there exist u0 ∈ ∂KR and µ0 ≥ 1 such that µ0 Φu0 = u0 , then u0 (x) satisfies equation (3.3) . Multiply equation (3.3) by φ1 (x) and integrate from a to b, using integration by parts in the left side to obtain Z b m X µ0 (Ik (u0 (xk ))φ1 (xk ) + I k (u0 (xk ))φ01 (xk )) λ1 u0 (x)h(x)φ1 (x)dx = a Z k=1 b + µ0 φ1 (x)h(x)g(x, u0 (x))dx a ≥ Z b m X 0 (1 − ε) (I∞ (k)φ1 (xk ) + I ∞ (k)φ1 (xk ))u0 (xk ) + (1 − ε)g∞ u0 (x)φ1 (x)h(x)dx à m k=1 ! Z b X 0 φ1 (x)h(x)dx . − C (φ1 (xk ) + φ1 (xk )) + a a k=1 If g∞ ≤ λ1 , then we have à m ! Z b Z b X [λ1 − (1 − ε)g∞ ] u0 (x)h(x)φ1 (x)dx + C φ1 (x)h(x)dx (φ1 (xk ) + φ01 (xk )) + a ≥ (1 − ε) m X k=1 k=1 (I∞ (k)φ1 (xk ) + I ∞ (k)φ01 (xk ))u0 (xk ). a 41 Ying He thus Z à m X b ku0 k[λ1 − (1 − ε)g∞ ] h(x)φ1 (x)dx + C a ≥ (1 − ε)σku0 k Z (φ1 (xk ) + φ01 (xk )) + ! b φ1 (x)h(x)dx a k=1 m X (I∞ (k)φ1 (xk ) + I ∞ (k)φ01 (xk )). k=1 and µ C ku0 k ≤ m P (1 − ε)σ m P (φ1 (xk ) + k=1 φ01 (xk )) + Rb a ¶ φ1 (x)h(x)dx (I∞ (k)φ1 (xk ) + I ∞ (k)φ01 (xk )) − [λ1 − (1 − ε)g∞ ] k=1 Rb a =: R̄. φ1 (x)h(x)dx (3.7a ) If g∞ > λ1 , we can choose ε > 0 such that (1 − ε)g∞ > λ1 , then we have ! à m Z b Z b X 0 (φ1 (xk ) + φ1 (xk )) + φ1 (x)h(x)dx ≥ [(1 − ε)g∞ − λ1 ] φ1 (x)u0 (x)h(x)dx C a k=1 a Z b φ1 (x)h(x)dx. ≥ [(1 − ε)g∞ − λ1 ]σku0 k a Thus µ C ku0 k ≤ m P (φ1 (xk ) + k=1 φ01 (xk )) [(1 − ε)g∞ − λ1 ]σ + Rb a Rb a ¶ φ1 (x)h(x)dx φ1 (x)h(x)dx =: R̄. (3.7b ) Let R > max{p, R̄}, then for any u ∈ ∂KR and µ ≥ 1, we have µΦu 6= u. Hence hypothesis (ii) of Lemma 2.4 is satisfied and i(Φ, KR , K) = 0. (3.8) In view of (3.1), (3.4) and (3.8), we obtain i(Φ, KR \ K̄p , K) = −1, i(Φ, Kp \ K̄r , K) = 1. Then Φ has fixed points u1 and u2 in Kp \ K̄r and KR \ K̄p , respectively, which means u1 (x) and u2 (x) are positive solution of the problem (1.1) and 0 < ku1 k < p < ku2 k. 42 Multiple Positive solutions of Sturm-Liouville problems... Corollary 3.3. The conclusion of Theorem 3.2 is valid if (H1 ) is replaced by (H1∗ ) g0 = ∞ or m X I0 (k)φ1 (xk ) = ∞ or m X k=1 I 0 (k)φ01 (xk ) = ∞; k=1 and g∞ = ∞ or m X I∞ (k)φ1 (xk ) = ∞ or k=1 m X I ∞ (k)φ01 (xk ) = ∞. k=1 Example 3.4. 1 Lu + 21 (u 3 + u3 ) = 0, x ∈ I 0 , 0 −∆(pu )|x=xk = ck u(xk ), ck ≥ 0, ∆(pu)|x=xk = dk u(xk ), dk ≥ 0, R1 (u) = αu(0) − βu0 (0) = 0, R (u) = γu(1) + δu0 (1) = 0, 2 (3.9) here Lu = (p(x)u0 )0 + q(x)u. Assume that (S1 ) is satisfied. Then problem (3.9) has at least two positive solutions u1 and u2 with 0 < ||u1 || < 1 < ||u2 || provided 1 1< d à 1− m X ! Z 1 G(xk , xk )(ck + dk ) , d = G(y, y)dy. 0 k=1 Proof To see this we will apply Theorem 3.2 (or Corollary 3.3). By (3.10), η > 0 is chosen such that m X 1 G(xk , xk )(ck + dk )). 1 < η < (1 − d k=1 Set 1 1 g(x, u) = (u 3 + u3 ). 2 Note g0 = ∞, so (H1 ) (or (H1∗ )) holds. g∞ = ∞, (3.10) 43 Ying He Let ηk = ck , η̄k = dk , then η, ηk , η̄k satisfy Z 1 η G(y, y)dy + 0 m X G(xk , xk )(ηk + η k ) < 1. k=1 Let p = 1, then for 0 ≤ u ≤ p, we have 1 1 1 1 g(x, u) = (u 3 + u3 ) ≤ + < ηp = η, 2 2 2 and Ik (u) = ck u = ηk u ≤ ηk p, I k (u) = dk u = η k u ≤ η k p thus (H2 ) holds. The result now follows from Theorem 3.2 (or Corollary3.3). References [1] R.P.Agarwal , D. O’Regan, Multiple nonnegative solutions for second order impulsive differential equations. Appl. Math. Comput., 114, (2000), 51-59. [2] W. Ding , M. Han, Periodic boundary value problem for the second order impulsive functional differential equations. Appl. Math. Comput, 155 A, (2004), 709-726. [3] D. Guo and V. Lakshmikantham, Nonlinear Problems in Abstract Cones, Academic Press, Boston, MA, 1988. [4] X. Liu, D. Q. Jiang, Z. Wei, Periodic boundary value problems for second order impulsive integrodifferential equations of mixed type in Banach spaces.J. Math. Anal. Appl., 195, (1995), 214-229. [5] S. G. Hristova , D. D. Bainov, Monotone-iterative techniques of V. Lakshmikantham for a boundary value problem for systems of impulsive differential-difference equations. J. Math. Anal. Appl., 1997, (1996), 1-13. [6] V. Lakshmik ntham, D.D. Bainov and P.S. Simeonov, Theory of Impulsive Differential Equations. World Scientific, Singapore, 1989. 44 Multiple Positive solutions of Sturm-Liouville problems... [7] E. Lee, Y. Lee, Multiple positive solutions of singular two point boundary value problems for second order impulsive differential equation, Appl. Math. Comput., 158, ( 2004), 745-759 [8] X. Liu, D. Q. Jiang, Multiple positive solutions of Dirichlet boundary value problems for second order impulsive differential equations. J. Math. Anal. Appl., 321, (2006), 501-514. [9] X. Liu, D. Q. Jiang, Z. Wei, Periodic boundary value problems for second order impulsive integrodifferential equations of mixed type in Banach spaces.J. Math. Anal. Appl., 195, (1995), 214-229. [10] I. Rachunkova, J. Tomecek, Impulsive BVPs with nonlinear boundary conditions for the second order differential equations without growth restrictions. J. Math. Anal. Appl., 292, (2004), 525-539.