\documentclass[reqno]{amsart} \AtBeginDocument{{\noindent\small {\em Electronic Journal of Differential Equations}, Vol. 2004(2004), No. 89, pp. 1--14.\newline ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu \newline ftp ejde.math.txstate.edu (login: ftp)} \thanks{\copyright 2004 Texas State University - San Marcos.} \vspace{9mm}} \begin{document} \title[\hfilneg EJDE-2004/89\hfil Multiple sign-changing solutions] {Multiple sign-changing solutions for some m-point boundary-value problems} \author[X. Xu\hfil EJDE-2004/89\hfilneg] {Xian Xu} \address{Xian Xu \hfill\break Department of Mathematics, Xuzhou Normal University, Xuzhou, Jiangsu, 221116, China} \email{xuxian68@163.com} \date{} \thanks{Submitted April 26, 2004. Published July 5, 2004.} \subjclass[2000]{34B15, 34B25} \keywords{m-point boundary-value problem; sign-changing solution; \hfill\break\indent fixed point index} \begin{abstract} In this paper, we show existence results for multiple sign-changing solutions for m-point boundary-value problems. We use fixed point index and Leray-Schauder degree methods. \end{abstract} \maketitle \numberwithin{equation}{section} \newtheorem{theorem}{Theorem}[section] \newtheorem{lemma}[theorem]{Lemma} \newtheorem{corollary}[theorem]{Corollary} \section{Introduction} In this paper, we consider the second-order multi-point boundary-value problem \begin{equation} \begin{gathered} y''(t)+f(y)=0, \quad 0\leq t\leq 1,\\ y(0)=0,\quad y(1)=\sum_{i=1}^{m-2}\alpha_i y(\eta_i), \end{gathered} \label{e1.1} \end{equation} where $0<\alpha_i$, $i=1,2,\dots, m-2$, $0<\eta_1<\eta_2<\dots<\eta_{m-2}<1$, $f\in C(\mathbb{R},\mathbb{R})$. The multi-point boundary-value problems for ordinary differential equations arise in different areas of applied mathematics and physics. For examples, the vibrations of a guy wire of uniform cross-section and composed of $N$ parts of different densities can be set up as a multi-point boundary-value problem (see \cite{m4}), many problems in the theory of elastic stability can be handled as multi-point problems (see \cite{t1}). Recently, there is much attention focused on the existence of nontrivial or positive solutions of the nonlinear multi-point boundary-value problems(see \cite{f1,g1,g2,l1,m1, m3,r1,w1,x1,x2,y1} and the references therein). For example, Ruyun Ma \cite{m1} considered the m-point boundary-value problem \begin{equation} \begin{gathered} u''(t)+a(t)f(u)=0,\quad t\in (0,1),\\ u'(0)=\sum_{i=1}^{m-2}b_iu'(\xi_i),\quad u(1)=\sum_{i=1}^{m-2}a_iu(\xi_i), \end{gathered} \label{e1.2} \end{equation} where $f\in C(\mathbb{R}^+,\mathbb{R}^+)$, $\xi_i\in (0,1)$ with $0<\xi_1<\xi_2<\dots<\xi_{m-2}<1$, $a_i,\ b_i\in \mathbb{R}^+$ with $0<\sum_{i=1}^{m-2}a_i<1$, and $0<\sum_{i=1}^{m-2}b_i<1$. Set $$ f_0=\lim_{u\to 0^+}\frac{f(u)}{u},\quad f_{\infty}=\lim_{u\to+\infty}\frac{f(u)}{u}. $$ Then $f_0=0$ and $f_{\infty}=\infty$ correspond to the super-linear case, and $f_0=\infty$ and $f_{\infty}=0$ correspond to the sub-linear case. By applying the fixed point theorem in cones, Ruyun Ma \cite{m1} showed that the m-point boundary value problem \eqref{e1.2} has at least one positive solution if $f$ is either super-linear or sub-linear. In this paper, we shall study the cases $f_0$, $f_\infty\not\in\{0,+\infty\}$. In these cases, the m-point boundary-value problem \eqref{e1.1} may have sign-changing solutions. Quite recently, the existence and qualitative properties of sign-changing solutions for elliptic boundary-value problems have been extensively studied. To the author's knowledge, however, there were fewer papers considered the sign-changing solutions for multi-point boundary value problems. The purpose of this paper is to give some existence results for multiple sign-changing solution for m-point boundary value problem \eqref{e1.1}. We shall follow the idea employed in \cite{l2} by Liu. To show the main result in this paper we need to study the the spectrum properties of the linear operator related the m-point boundary-value problem \eqref{e1.1}. Gupta and Sergej Trofimchuk \cite{g1} studied the problem of existence of solutions for the three-point boundary-value problem \begin{equation} \begin{gathered} x''(t)=f(t, x(t), x'(t)),\quad t\in (0,1),\\ x(0)=0,\quad x(1)=\alpha x(\eta), \end{gathered}\label{e1.3} \end{equation} where $\alpha\in \mathbb{R}$, $\alpha\leq 1$ and $\eta\in(0,1)$ are given. Using the spectrum radius of some related linear operators, the authors proved some existence results for nontrivial solutions of the three-point boundary-value problem \eqref{e1.3}. We shall organize this paper as follows. In $\S2$ some preliminary results are given including the study of the eigenvalues of the linear operator $A'(\theta)$ and $A'(\infty)$. In $\S3$ by using the fixed point index and Leray-Shauder degree method, we will prove the main result. \section{Preliminary Lemmas} From \cite[Theorem 2.3.1]{d1}, we have the following definition. Let $X$ be a retract of real Banach space $E$, $U$ be a relatively bounded open subset of $X$, $A:D\mapsto X$ be completely continuous operator. The integer $i(A, U, X)$ be defined by $$ i(A, U, X)=\deg(I-A\cdot r, B(\theta, R)\cap r^{-1}(U),\theta), $$ where $r:E\mapsto X$ is an arbitrary retraction and $R>0$ such that $B(\theta, R)\supset U$. Then the integer $i(A, U,X)$ be called the fixed point index of $A$ on $U$ with respect to $X$. Set $$ \beta_0=\lim_{x\to 0}\frac{f(x)}{x},\quad \beta_1=\lim_{|x|\to \infty}\frac{f(x)}{x}. $$ Let us list some conditions to be used in this paper. \begin{itemize} \item[(H0)] Assume that the sequence of positive solutions of the equation $$\sin \sqrt{x}=\sum_{i=1}^{m-2}\alpha_i \sin\eta_i\sqrt{x}$$ is $\lambda_1<\lambda_2<\dots<\lambda_n<\lambda_{n+1}<\dots$. \item[(H1)] $0<\sum_{i=1}^{m-2}\alpha_i<1$, $f\in C(\mathbb{R}, \mathbb{R})$, $f(0)=0$, $xf(x)> 0$ for all $x\in \mathbb{R}\backslash\{0\} $. \item[(H2)] There exist positive integers $n_0$ and $n_1$ such that $$ \lambda_{2n_0}<\beta_0<\lambda_{2n_0+1},\quad \lambda_{2n_1}<\beta_1<\lambda_{2n_1+1}. $$ \item[(H3)] There exists $C_0>0$ such that $$ |f(x)|<\frac{2(1-\sum_{i=1}^{m-2}\alpha_i\eta_i)} {5-\sum_{i=1}^{m-2}\alpha_i\eta_i}C_0, $$ for all $x$ with $|x|\leq C_0$. \end{itemize} The main result of this paper is the following. \begin{theorem} \label{thm2.1} Suppose that (H0)--(H3) hold. Then the m-point boundary-value problem \eqref{e1.1} has at least two sign-changing solutions. Moreover, the m-point boundary-value problem \eqref{e1.1} also has at least two positive solutions and two negative solutions. \end{theorem} Before giving the proof of Theorem \ref{thm2.1}, we list some preliminary lemmas. Let \begin{gather*} E=\{x\in C^1[0,1] : x(0)=0,\; x(1)=\sum_{i=1}^{m-2}\alpha_i x(\eta_i)\}\\ P=\{x\in E: x(t)\geq 0 \mbox{ for } t\in [0,1]\}. \end{gather*} For $x\in E$, let $\|x\|=\|x\|_0+\|x'\|_0$, where $\|x\|_0=\max_{t\in[0,1]}|x(t)|$ and $\|x'\|_0=\max_{t\in[0,1]}|x'(t)|$. It is easy to show that $E$ is a Banach space with the norm $\|\cdot\|$ and $P$ is a cone of $E$. Let the operators $K$, $F$ and $A$ be defined by \begin{equation} \begin{aligned} (Kx)(t)&= \frac{t}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}\int^1_0(1-s)x(s)ds -\int^t_0(t-s)x(s)ds\\ &\quad -\frac{t}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i} \sum_{i=1}^{m-2}\alpha_i\int^{\eta_i}_0(\eta_i-s)x(s)ds,\quad t\in [0,1],\;x\in E, \end{aligned} \label{e2.1)} \end{equation} $(Fx)(t)=f(x(t))$ for $t\in [0,1]$, $x\in E$ and $A=KF$. From \cite[Lemma 2.3.1]{d1}, we get the following Lemma. \begin{lemma} \label{lm1} Let $\theta\in \Omega$ and $A:P\cap\bar{\Omega}\mapsto P$ be condensing. Suppose that $$ Ax\neq \mu x, \quad \forall x\in P\cap\partial\Omega,\; \mu\geq 1. $$ Then $i(A, P\cap\Omega, P)=1$. \end{lemma} From \cite[Corollary 2, p.p.146]{d2}, we have the following Lemma. \begin{lemma} \label{lm2} Let $\Omega$ be a open set in $E$ and $\theta\in \Omega$, $A:\bar{\Omega}\mapsto E$ be completely continuous. Suppose that $$ \|Ax\|\leq\|x\|, \quad Ax\neq x, \; \forall x\in \partial \Omega. $$ Then $\deg(I-A, \Omega, \theta)=1$. \end{lemma} \noindent{\bf Remark}\; Obviously, Lemma \ref{lm2} can also be directly obtained by the normality and homotopic invariance property of Leray-Schauder degree. The following Lemma can be easily obtained. \begin{lemma} \label{lm3} Suppose that $\sum_{i=1}^{m-2} \alpha_i\eta_i<1$. If $u\in C[0,1]$, then $y\in C^2[0,1]$ is a solution the m-point boundary-value problem \begin{gather*} y''(t)+u(t)=0,\quad 0\leq t\leq 1,\\ y(0)=0,\quad y(1)=\sum_{i=1}^{m-2} \alpha_i y(\eta_i) \end{gather*} if and only if $y\in C[0,1]$ is a solution of the integral equation $y(t)=(Ku)(t), t\in [0,1]$. \end{lemma} \noindent{\bf Remark} %2 By Lemma \ref{lm3} we can easily show that $A:E\mapsto E$ is a completely continuous operator. \begin{lemma} \label{lm4} Suppose that (H1) and (H2) hold. Then the operator $A$ is Fr\'echet differentiable at $\theta$ and $\infty$. Moreover, $A'(\theta)=\beta_0 K$, and $A'(\infty)=\beta_1 K.$ \end{lemma} \begin{proof} For any $\varepsilon>0$, by (H2) there exists $\delta>0$ such that for any $0<|x|<\delta$, $$ \big|\frac{f(x)}{x}-\beta_0\big|<\varepsilon, $$ that is $|f(x)-\beta_0x|<\varepsilon |x|$, for all $0\leq|x|<\delta$. Then, for any $x\in E$ with $\|x\|<\delta$, we have \begin{align*} &|(Ax-A\theta-\beta_0 Kx)(t)|\\ &=|(K(Fx-\beta_0x))(t)|\\ &\leq\frac{1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}\int^1_0(1-s)\max_{s\in [0,1]}|f(x(s))-\beta_0x(s)|ds\\ &\quad +\int^1_0(1-s)\max_{s\in [0,1]}|f(x(s))-\beta_0x(s)|ds\\ &\quad +\frac{1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}\sum_{i=1}^{m-2} \alpha_i\int^{\eta_i}_0(\eta_i-s)\max_{s\in [0,1]}|f(x(s))-\beta_0x(s)|ds\\ &\leq \Big [\frac{1}{2(1-\sum_{i=1}^{m-2}\alpha_i\eta_i)}+ \frac{1}{2}+\frac{\sum_{i=1}^{m-2}\alpha_i\eta_i^2}{2(1- \sum_{i=1}^{m-2}\alpha_i\eta_i)}\Big ]\|x\|_0\varepsilon\\ &\leq \frac{1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}\|x\|\varepsilon,t\in [0,1]. \end{align*} This implies \begin{equation} \|Ax-A\theta-\beta_0 Kx\|_0\leq \frac{1}{1-\sum_{i=1}^{m-2} \alpha_i\eta_i}\|x\|\varepsilon,\quad x\in E, \; \|x\|<\delta.\label{e2.1} \end{equation} Similarly, we can show that for any $x\in E$, $\|x\|<\delta$, $$ |(Ax-A\theta-\beta_0 Kx)'(t)|\leq \frac{3-\sum_{i=1}^{m-2}\alpha_i\eta_i}{2(1-\sum_{i=1}^{m-2}\alpha_i\eta_i)}\|x\| \varepsilon, \quad t\in [0,1] $$ and so \begin{equation} \|(Ax-A\theta-\beta_0 Kx)'\|_0\leq \frac{3-\sum_{i=1}^{m-2}\alpha_i\eta_i}{2(1-\sum_{i=1}^{m-2}\alpha_i\eta_i)} \|x\|\varepsilon,\ x\in E, \ \|x\|<\delta.\label{e2.2} \end{equation} By \eqref{e2.1} and \eqref{e2.2}, we have \begin{align*} \|Ax-A\theta-\beta_0 Kx\| &=\|Ax-A\theta-\beta_0 Kx\|_0+\|(Ax-A\theta-\beta_0 Kx)'\|_0\\ &\leq\frac{5-\sum_{i=1}^{m-2}\alpha_i\eta_i}{2(1 -\sum_{i=1}^{m-2}\alpha_i\eta_i)}\|x\|\varepsilon \end{align*} Consequently, $$ \lim_{\|x\|\to 0}\frac{\|Ax-A\theta-\beta_0 Kx\|}{\|x\|}=0. $$ This means that $A$ is Fr\'echet differentiable at $\theta$, and $A'(\theta)=\beta_0K$. For each $\varepsilon>0$, by (H2), there exists $R>0$ such that $$ |f(x)-\beta_1x|<\varepsilon |x|$$ for $|x|>R$. Let $b=\max_{|x|\leq R}|f(x)-\beta_1 x|$. Then we have for any $x\in \mathbb{R}$, $$ |f(x)-\beta_1 x|\leq \varepsilon |x|+b. $$ Consequently, \begin{align*} &|(Ax-\beta_1 Kx)(t)|\\ &=|(K(Fx-\beta_1x))(t)|\\ &\leq\frac{1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}\int^1_0(1-s)\max_{s\in [0,1]}|f(x(s))-\beta_1x(s)|ds\\ &\quad +\int^1_0(1-s)\max_{s\in [0,1]}|f(x(s))-\beta_1x(s)|ds\\ &\quad +\frac{1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i} \sum_{i=1}^{m-2}\alpha_i\int^{\eta_i}_0(\eta_i-s)\max_{s\in [0,1]}|f(x(s))-\beta_1x(s)|ds\\ &\leq \Big[\frac{1}{2(1-\sum_{i=1}^{m-2}\alpha_i\eta_i)}+ \frac{1}{2}+\frac{\sum_{i=1}^{m-2}\alpha_i\eta_i^2}{2(1-\sum_{i=1}^{m-2} \alpha_i\eta_i)}\Big](\varepsilon\|x\|_0+b)\\ &\leq \frac{1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}(\varepsilon\|x\|+b),\quad t\in [0,1]. \end{align*} This implies \begin{equation} \|Ax-\beta_1 Kx\|_0\leq \frac{1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}(\varepsilon\|x\|+b),\quad x\in E.\label{e2.3} \end{equation} Similarly, we can show that \begin{equation} \|(Ax-\beta_1 Kx)'\|_0\leq \frac{3-\sum_{i=1}^{m-2}\alpha_i\eta_i}{2(1-\sum_{i=1}^{m-2}\alpha_i\eta_i)} (\varepsilon\|x\|+b),\quad x\in E.\label{e2.4} \end{equation} By \eqref{e2.3} and \eqref{e2.4}, we have \begin{align*} \|Ax-\beta_1 Kx\|&=\|Ax-\beta_1 Kx\|_0+\|(Ax-\beta_1 Kx)'\|_0\\ &\leq\frac{5-\sum_{i=1}^{m-2}\alpha_i\eta_i}{2(1-\sum_{i=1}^{m-2} \alpha_i\eta_i)}(\varepsilon\|x\|+b). \end{align*} Consequently, $$ \lim_{\|x\|\to \infty}\frac{\|Ax-\beta_1 Kx\|}{\|x\|}=0\,. $$ This means that $A$ is Fr\'echet differentiable at $\infty$, and $A'(\infty)=\beta_1K$. The proof is complete. \end{proof} \begin{lemma} \label{lm5} Suppose that (H0) and (H1) hold. Let $\beta$ be a positive number. Then the sequence of positive eigenvalues of the operator $\beta K$ is $$ \frac{\beta}{\lambda_1}>\frac{\beta}{\lambda_2}>\dots>\frac{\beta}{\lambda_n}\dots. $$ Moreover, the positive eigenvalues $\frac{\beta}{\lambda_n}$ $(n=1,2,\dots)$ have algebraic multiplicity one. \end{lemma} \begin{proof} Let $\bar{\lambda}$ be a positive eigenvalue of the linear operator $\beta K$, and $y\in E\backslash\{\theta\}$ be an eigenfunction corresponding to the eigenvalue $\bar{\lambda}$. By Lemma \ref{lm3}, we have \begin{equation} \begin{gathered} y''(t)+\frac{\beta}{\bar{\lambda}}y(t)=0,0\leq t\leq 1,\\ y(0)=0,\quad y(1)=\sum_{i=1}^{m-2}\alpha_i y(\eta_i). \end{gathered} \label{e2.5} \end{equation} The auxiliary equation of the differential equation \eqref{e2.5} has roots $\pm \sqrt{\frac{\beta}{\bar{\lambda}}}i$. Thus the general solution of \eqref{e2.5} is of the form $$ y(t)=C_1 \cos t\sqrt{\frac{\beta}{\bar{\lambda}}}+C_2 \sin t\sqrt{\frac{\beta}{\bar{\lambda}}},\quad t\in [0,1]. $$ Applying the condition $y(0)=0$, we obtain that $C_1=0$, and so the general solution can be reduce to $$ y(t)=C_2 \sin t\sqrt{\frac{\beta}{\bar{\lambda}}},\quad t\in [0,1]. $$ Applying the second condition $y(1)=\sum_{i=1}^{m-2}\alpha_i y(\eta_i)$, we obtain that $$ \sin\sqrt{\frac{\beta}{\bar{\lambda}}}= \sum_{i=1}^{m-2}\alpha_i\sin\eta_i\sqrt{\frac{\beta}{\bar{\lambda}}}. $$ Since the positive solutions of the equation $\sin \sqrt{x}=\sum_{i=1}^{m-2}\alpha_i\sin \eta_i\sqrt{x}$ are $0<\lambda_1<\lambda_2<\dots$, then $\bar{\lambda}$ is one of the values $$ \frac{\beta}{\lambda_1}>\frac{\beta}{\lambda_2}>\dots>\frac{\beta}{\lambda_n}\dots $$ and the eigenfunction corresponding to the eigenvalue $\frac{\beta}{\lambda_n}$ is $$ y_n(t)=C \sin t\sqrt{\lambda_n}, \quad t\in [0,1], $$ where $C$ is a nonzero constant. By ordinary method, we can show that any two eigenfunctions corresponding to the same eigenvalue $\frac{\beta}{\lambda_n}$ are merely nonzero constant multiples of each other. Consequently, \begin{equation} \dim \ker(\frac{\beta}{\lambda_n}I- \beta K)=\dim \ker(I-\lambda_n K)=1. \label{e2.6} \end{equation} Now we show that \begin{equation} \ker(I-\lambda_nK)=\ker (I-\lambda_nK)^2.\label{e2.7} \end{equation} Obviously, we need to show only that $$ \ker (I-\lambda_nK)^2\subset \ker(I-\lambda_nK). $$ For any $y\in \ker (I-\lambda_nK)^2$, $(I-\lambda_nK)y$ is an eigenfunction of linear operator $\beta K$ corresponding to the eigenvalue $\frac{\beta}{\lambda_n}$ if $(I-\lambda_nK)y\not =\theta$. Then there exists nonzero constant $\gamma$ such that $$ (I-\lambda_nK)y=\gamma \sin t\sqrt{\lambda_n}, \quad t\in [0,1]. $$ By direct computation, we have \begin{equation} \begin{gathered} y''(t)+\lambda_ny =-\lambda_n \gamma \sin t\sqrt{\lambda_n}, \quad t\in [0,1],\\ y(0)=0,\quad y(1)=\sum_{i=1}^{m-2}\alpha_i y(\eta_i). \end{gathered} \label{e2.8} \end{equation} It is easy to see that the general solutions of \eqref{e2.8} is of the form \begin{align*} y(t)&=C_1\cos t\sqrt{\lambda_n}+C_2\sin t\sqrt{\lambda_n}+ (\frac{\gamma t\sqrt{\lambda_n}}{2}-\frac{\gamma}{4}\sin 2t\sqrt{\lambda_n})\cos t \sqrt{\lambda_n}\\ &\quad +\frac{\gamma}{4}\cos 2t\sqrt{\lambda_n}\cdot \sin t\sqrt{\lambda_n},\quad t\in [0,1], \end {align*} where $C_1$, $C_2$ are two nonzero constants. Applying the condition $y(0)=0$, we obtain that $C_1=0$. Since $\sin\sqrt{\lambda_n}=\sum_{i=1}^{m-2}\alpha_i\sin \eta_i\sqrt{\lambda_n}$, then we have \begin{equation} \begin{aligned} y(1)&=C_2\sin \sqrt{\lambda_n}+ (\frac{\gamma \sqrt{\lambda_n}}{2}-\frac{\gamma}{4}\sin 2\sqrt{\lambda_n})\cos \sqrt{\lambda_n}+\frac{\gamma}{4}\cos 2\sqrt{\lambda_n}\cdot \sin \sqrt{\lambda_n}\\ &=\sum_{i=1}^{m-2}\alpha_iC_2 \sin \eta_i\sqrt{\lambda_n}+ \frac{\gamma\sqrt{\lambda_n}}{2}\cos\sqrt{\lambda_n}- \frac{\gamma}{2}\sum_{i=1}^{m-2} \alpha_i\sin\eta_i\sqrt{\lambda_n}\cos^2\sqrt{\lambda_n}\\ &\quad + \frac{\gamma}{4}\sum_{i=1}^{m-2}\alpha_i\cos2 \sqrt{\lambda_n}\sin\eta_i\sqrt{\lambda_n}, \end{aligned} \label{e2.9} \end{equation} and \begin{equation} \begin{aligned} \sum_{i=1}^{m-2}\alpha_i y(\eta_i) &= \sum_{i=1}^{m-2}\alpha_i C_2\sin \eta_i\sqrt{\lambda_n}+ \sum_{i=1}^{m-2}(\frac{\gamma \alpha_i\eta_i\sqrt{\lambda_n}}{2} -\frac{\gamma\alpha_i}{4}\sin 2\eta_i\sqrt{\lambda_n})\cos \eta_i\sqrt{\lambda_n}\\ &\quad +\sum_{i=1}^{m-2}\frac{\gamma\alpha_i}{4}\cos 2\eta_i\sqrt{\lambda_n}\cdot \sin \eta_i\sqrt{\lambda_n}.\ \end{aligned} \label{e2.10} \end{equation} Since $y(1)=\sum_{i=1}^{m-2}\alpha_i y(\eta_i)$, by \eqref{e2.9} and \eqref{e2.10}, we have $$ \cos\sqrt{\lambda_n}=\sum_{i=1}^{m-2}\alpha_i\eta_i\cos\eta_i\sqrt{\lambda_n}. $$ By the Schwarz inequality, we obtain \begin{align*} 1-\sin ^2\sqrt{\lambda_n} &=(\sum_{i=1}^{m-2}\alpha_i\eta_i\cos\eta_i\sqrt{\lambda_n})^2\\ &\leq (\sum_{i=1}^{m-2}\eta_i^2)(\sum_{i=1}^{m-2}\alpha_i^2\cos^2\eta_i\sqrt{\lambda_n})\\ &=(\sum_{i=1}^{m-2}\eta_i^2)(\sum_{i=1}^{m-2}\alpha_i^2)-(\sum_{i=1}^{m-2} \eta_i^2)(\sum_{i=1}^{m-2}\alpha_i^2\sin^2\eta_i\sqrt{\lambda_n}). \end{align*} Applying the condition $\sin \sqrt{\lambda_n}=\sum_{i=1}^{m-2}\alpha_i\sin\eta_i\sqrt{\lambda_n}$, we obtain \begin{align*} 1&\leq (\sum_{i=1}^{m-2}\eta_i^2)(\sum_{i=1}^{m-2}\alpha_i^2)+(\sum_{i=1}^{m-2}\alpha_i\sin\eta_i\sqrt{\lambda_n})^2-(\sum_{i=1}^{m-2}\eta_i^2) (\sum_{i=1}^{m-2}\alpha_i^2\sin^2\eta_i\sqrt{\lambda_n})\\ &=(\sum_{i=1}^{m-2}\eta_i^2)(\sum_{i=1}^{m-2}\alpha_i^2) +(1-(\sum_{i=1}^{m-2}\eta_i^2))( \sum_{i=1}^{m-2}\alpha_i^2\sin^2\eta_i\sqrt{\lambda_n}) \\ &+\sum_{i\not =j}\alpha_i\alpha_j\sin\eta_i\sqrt{\lambda_n}\sin\eta_j\sqrt{\lambda_n}\\ &\leq(\sum_{i=1}^{m-2}\eta_i^2)(\sum_{i=1}^{m-2}\alpha_i^2)+ (1-(\sum_{i=1}^{m-2}\eta_i^2))(\sum_{i=1}^{m-2}\alpha_i^2) +\sum_{i\neq j}\alpha_i\alpha_j\\ &=(\sum_{i=1}^{m-2}\alpha_i)^2, \end{align*} which is a contradiction of $\sum_{i=1}^{m-2}\alpha_i<1$. Thus, \eqref{e2.7} holds. It follows from \eqref{e2.6} and \eqref{e2.7} that the algebraic multiplicity of the eigenvalue $\frac{\beta}{\lambda_n}$ is 1. The proof is complete. \end{proof} \begin{lemma} \label{lm6} Suppose that (H0) and (H1) hold and $y\in P\backslash\{\theta\}$ is a solution of the boundary-value problem \eqref{e1.1}. Then $y\in \stackrel{\circ}{P}$. \end{lemma} \begin{proof} Since $y''(t)=-f(y(t))\leq 0$ for $t\in [0,1]$, then $y$ is a concave function on $[0,1]$. For all $i\in\{1,2,\dots, m-2\}$, we have from the concavity of $y$ that $$ y(t)\leq \frac{y(1)-y(\eta_i)}{1-\eta_i}(t-1)+y(1), \quad t\in [0,\eta_1] $$ that is $ y(t)(1-\eta_i)\leq (y(1)-y(\eta_i))(t-1)+y(1)(1-\eta_i)$, $t\in [0,\eta_1]$. This together with the boundary condition $y(1)=\sum_{i=1}^{m-2}\alpha_iy(\eta_i)$ implies \begin{equation} \begin{aligned} y(t)&\leq y(1)\frac{\sum_{i=1}^{m-2}\alpha_i(1-\eta_i)+ (1-\sum_{i=1}^{m-2}\alpha_i)(1-t)}{\sum_{i=1}^{m-2}\alpha_i(1-\eta_i)} \\ &\leq y(1)\frac{\sum_{i=1}^{m-2}\alpha_i(1-\eta_i)+ (1-\sum_{i=1}^{m-2}\alpha_i)}{\sum_{i=1}^{m-2} \alpha_i(1-\eta_i)}\\ &= y(1)\frac{1-\sum_{i=1}^{m-2}\alpha_i\eta_i }{\sum_{i=1}^{m-2}\alpha_i(1-\eta_i)},\quad t\in [0,\eta_1]. \end{aligned} \label{e2.11} \end{equation} From the concavity of $y$ and this inequality, we have \begin{equation} y(t)\leq \frac{y(\eta_1)}{\eta_1}t\leq \frac{y(\eta_1)}{\eta_1}\leq y(1)\frac{1-\sum_{i=1}^{m-2}\alpha_i\eta_i }{\sum_{i=1}^{m-2}\alpha_i(1-\eta_i)\eta_1} , t\in [\eta_1,1]. \label{e2.12)} \end{equation} From this inequality and \eqref{e2.11} it follows that $$ y(1)\geq\frac{\sum_{i=1}^{m-2}\alpha_i(1-\eta_i) \eta_1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}\|y\|_0. $$ Since $y$ is a concave function on [0,1], we have \begin{equation} y(t)\geq (y(1)-y(0))t=y(1)t\geq\frac{\sum_{i=1}^{m-2} \alpha_i(1-\eta_i) \eta_1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}\|y\|_0t, \quad t\in [0,1].\label{e2.13} \end{equation} Consequently, $$ y'(0)=\lim_{t\to 0}\frac{y(t)}{t}\geq \frac{\sum_{i=1}^{m-2}\alpha_i(1-\eta_i) \eta_1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}\|y\|_0>0. $$ Then there exist $\varepsilon>0$ and $\tau_1>0$ such that \begin{equation} y'(t)>\tau_1,\forall t\in [0,\varepsilon].\label{e2.14} \end{equation} By \eqref{e2.13}, there exists $\tau_2>0$ such that \begin{equation} y(t)>\tau_2,\quad \forall t\in [\varepsilon,1]\label{e2.15} \end{equation} Let $\tau=\min\{\tau_1,\tau_2\}$. Then by \eqref{e2.14} and \eqref{e2.15}, we obtain $u(t)\geq 0$, $t\in[0,1]$ for any $u\in E$ with $\|u-y\|<\tau$. Therefore, $B(y,\tau)\subset P$ and $y\in \stackrel{\circ}{P}$, where $B(y,\tau)=\{x\in E:\|x-y\|<\tau\}$. The proof is complete. \end{proof} By \cite[Lemmas 2.3.7, 2.3.8]{d1}, we have the following Lemma. \begin{lemma} \label{lm7} Let $A:P\mapsto P$ be completely continuous, Suppose that $A$ is differentiable at $\theta$ and $\infty$ along $P$ and 1 is not an eigenvalue of $A'_+(\theta)$ and $A'_+(\infty)$ corresponding to a positive eigenfunction. \begin{itemize} \item[(1)] If $A'_+(\theta)$ has a positive eigenfunction corresponding to an eigenvalue greater than 1, and $A\theta=\theta$. Then there exists $\tau>0$ such that $i(A, P\cap B(\theta, r), P)=0$ for any $00$ such that $i(A, P\cap B(\theta, R), P)=0$ for any $R>\varsigma$. \end{itemize} \end{lemma} \begin{lemma} \label{lm8} Suppose that (H0)--(H3) hold. Then \begin{itemize} \item[(1)] There exists $C_0>r_0>0$ such that for any $0C_0$ such that for any $R\geq R_0$, $$i(A, P\cap B(\theta,R),P)=0,\quad i(A, -P\cap B(\theta,R),-P)=0.$$ \end{itemize} \end{lemma} \begin{proof} We prove only conclusion (1). The same way, conclusion (2) can be proved. First we claim that $K(P)\subset P$ and $K(-P)\subset -P$. Let $x\in P$ be fixed and $y= Kx$. Obviously, $y\in C^1[0,1]$. By direct computation, we have \begin{equation} \begin{aligned} y(1)&=\frac{1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i} (\sum_{i=1}^{m-2}\eta_i\int^1_0(1-s) x(s)ds -\sum_{i=1}^{m-2}\alpha_i\int^{\eta_i}_0(\eta_i-s) x(s)ds)\\ &\geq \frac{1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}\sum_{i=1}^{m-2}\alpha_i \int^{\eta_i}_0(1-\eta_i)sx(s)\,ds \geq 0\,. \end{aligned} \label{e2.16} \end{equation} It follows from Lemma \ref{lm3} that \begin{gather} y''(t)=-x(t)\leq 0,\quad \forall t\in [0,1].\label{e2.17}\\ y(0)=0, \quad y(1)=\sum_{i=1}^{m-2}\alpha_i y(\eta_i)\label{e2.18} \end{gather} By \eqref{e2.17}, we see that $y$ is a concave function on [0,1]. Then the boundary condition \eqref{e2.16} and \eqref{e2.18} mean that $y(t)\geq 0$ for $t\in [0,1]$. Therefore, $y\in P$, and so $K(P)\subset P$, $K(-P)\subset (-P)$. Since $xf(x)> 0$ for $x\in \mathbb{R}\backslash\{0\}$, then we see that $A(P)\subset P$ and $A(-P)\subset (-P)$. It follows from Lemmas \ref{lm4} and \ref{lm5} that $A'_+(\theta)=\beta_0 K$, $\beta_0/\lambda_1$ $(>1)$ is an eigenvalue of the linear operator $\beta_0 K$ and the eigenfunction corresponding to $\frac{\beta_0}{\lambda_1}$ is $$ y(t)=C\sin t\sqrt{\lambda_1},\quad t\in [0,1], $$ where $C$ is an arbitrary positive constant and $\lambda_1$ is the smallest positive solution of the equation $\sin\sqrt{x}=\sum_{i=1}^{m-2}\alpha_i \sin \eta_i\sqrt{x}$. Since $$ \lim_{x\to 0}\frac{\sin\sqrt{x}-\sum_{i=1}^{m-2}\alpha_i \sin \eta_i\sqrt{x}}{\sqrt{x}}=1-\sum_{i=1}^{m-2}\alpha_i\eta_i>0\,, $$ there exists $\delta_0\in (0,1)$ small enough such that $$ \frac{\sin\sqrt{\delta_0}-\sum_{i=1}^{m-2}\alpha_i \sin \eta_i\sqrt{\delta_0}}{\sqrt{\delta_0}} \geq\frac{1}{4}(1-\sum_{i=1}^{m-2}\alpha_i\eta_i)>0. $$ On the other hand, $$ \sin\sqrt{\pi^2}-\sum_{i=1}^{m-2}\alpha_i \sin \eta_i\sqrt{\pi^2}=-\sum_{i=1}^{m-2}\alpha_i \sin \eta_i\pi<0. $$ Then, by the intermediate-value principle, $\lambda_1\in (\delta_0, \pi^2)$. Consequently, $$ y(t)=C\sin t\sqrt{\lambda_1}\geq 0,\quad t\in [0,1]. $$ It follows from Lemma \ref{lm7} that there exists $\tau_0>0$ such that $i(A, P\cap B(\theta, r),P)=0$ for any $00$ such that $i(A, -P\cap B(\theta, r),-P)=0$ for any $00$ $$ \deg (I-A, B(x_0, r), \theta)=(-1)^k, $$ where $k$ is the sum of the algebraic multiplicities of the real eigenvalues of $A'(x_0)$ in $(1, +\infty)$. \end{lemma} \begin{lemma} \label{lm10} Let $A$ be a completely continuous operator which is defined on all $E$. Assume that 1 is not an eigenvalue of the asymptotic derivative. The completely continuous vector field $I-A$ is then nonsingular on spheres $S_\rho=\{x|\|x\|=\rho\}$ of sufficiently large radius $\rho$ and $$ \deg(I-A, B(\theta, \rho), \theta)=(-1)^k, $$ where $k$ is the sum of the algebraic multiplicities of the real eigenvalues of $A'(\infty)$ in $(1, +\infty)$. \end{lemma} \section {Proof of main Theorem} \begin{proof}[Proof of Theorem \ref{thm2.1}] From Lemma \ref{lm3}, a function $y$ is a solution of the boundary-value problem \eqref{e1.1} if and only if $y$ is a fixed point of the operator $A$. By (H3), we have for any $x\in E$, $\|x\|=C_0$, \begin{align*} &|(Ax)(t)|\\ &\leq\frac{1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i}\int^1_0(1-s)\max_{s\in [0,1]}|f(x(s))|ds +\int^1_0(1-s)\max_{s\in [0,1]}|f(x(s))|ds\\ &\quad +\frac{1}{1-\sum_{i=1}^{m-2}\alpha_i\eta_i} \sum_{i=1}^{m-2}\alpha_i\int^{\eta_i}_0(\eta_i-s)\max_{s\in [0,1]}|f(x(s))|ds\\ &<\frac{2(1-\sum_{i=1}^{m-2}\alpha_i\eta_i)}{5-\sum_{i=1}^{m-2}\alpha_i\eta_i} \Big(\frac{1}{2(1-\sum_{i=1}^{m-2}\alpha_i\eta_i)} +\frac{1}{2}+\frac{\sum_{i=1}^{m-2}\alpha_i\eta_i} {2(1-\sum_{i=1}^{m-2}\alpha_i\eta_i)}\Big)C_0\\ &\leq \frac{2C_0}{5-\sum_{i=1}^{m-2}\alpha_i\eta_i},\quad t\in [0,1]\,. \end{align*} Therefore, \begin{equation} \|Ax\|_0<\frac{2C_0}{5-\sum_{i=1}^{m-2}\alpha_i\eta_i} .\label{e3.1} \end{equation} Similarly, we can show that for any $x\in E$, with $\|x\|=C_0$, \begin{equation} \|(Ax)'\|_0<\frac{3-\sum_{i=1}^{m-2}\alpha_i\eta_i} {5-\sum_{i=1}^{m-2}\alpha_i\eta_i}C_0.\label{e3.2} \end{equation} It follows from \eqref{e3.1} and \eqref{e3.2} that $\|Ax\|0$ such that $B(x,\delta_x)\subset P\cap (B(\theta,R_1)\backslash\overline{B(\theta, C_0)})$. Let $O_1=\bigcup_{x\in S}B(x, \delta_x)$. Then, we have $O_1\subset P\cap (B(\theta, R_1)\backslash\overline{B(\theta,C_0)})$. By \eqref{e3.12} and the excision property of the fixed point index, we have \begin{equation} i(A, O_1, P)=-1.\label{e3.16} \end{equation} By the definition of the fixed point index, we have \begin{equation} i(A, O_1, P)=\deg(I-A\cdot r, B(\theta, \bar{R})\cap r^{-1}(O_1),\theta),\label{e3.17} \end{equation} where $r:E\mapsto P$ is an arbitrary retraction and $\bar{R}$ is a large enough positive number such that $O_1\subset B(\theta, \bar{R})$. Now, we assume that $y^*\in B(\theta,\bar{R})\cap r^{-1}(O_1)$ such that $y^*=A\cdot r(y^*)$. Since $r:E\mapsto P$ and $A:P\mapsto P$, then $y^*\in P$, and so $y^*=ry^*\in O_1$. Therefore, $y^*\in O_1$ whenever $y^*\in B(\theta, \bar{R})\cap r^{-1}(O_1)$ is a fixed point of the operator $A\cdot r$. Then, by the excision property of the degree we have \begin{equation} \deg(I-A\cdot r, B(\theta, \bar{R})\cap r^{-1}(O_1),\theta) =\deg (I-A, O_1,\theta).\label{e3.18} \end{equation} By \eqref{e3.16}-\eqref{e3.18}, we have \begin{equation} \deg (I-A, O_1,\theta)=-1.\label{e3.19} \end{equation} Similarly, by \eqref{e3.13}-\eqref{e3.15}, we can show that there exist open sets $O_2,\ O_3$ and $O_4$ such that \begin{gather} O_2\subset P\cap(B(\theta, C_0)\backslash\overline{B(\theta, r_1)}),\nonumber\\ O_3\subset -P\cap(B(\theta, C_0)\backslash\overline{B(\theta, r_1)}),\nonumber \\ O_4\subset -P\cap(B(\theta, R_1)\backslash\overline{B(\theta, C_0)}),\nonumber\\ \deg (I-A, O_2, \theta)=1,\label{e3.20}\\ \deg (I-A, O_3, \theta)=1,\label{e3.21} \\ \deg (I-A, O_4, \theta)=-1.\label{e3.22} \end{gather} It follows from \eqref{e3.5}, \eqref{e3.6}, \eqref{e3.20} and \eqref{e3.21} that $$ \deg(I-A, B(\theta, C_0)\backslash(\overline{O_2}\cup\overline{O_3} \cup\overline{B(\theta, r_1)}), \theta)=1-1-1-1=-2. $$ This implies that $A$ has at least one fixed point $x_5\in B(\theta, C_0)\backslash(\overline{O_2}\cup\overline{O_3}\cup\overline{B(\theta, r_1)})$. Similarly, by \eqref{e3.5}, \eqref{e3.7}, \eqref{e3.19} and \eqref{e3.22}, $$ \deg(I-A, B(\theta,R_1)\backslash(\overline{O_1}\cup\overline{O_4} \cup\overline{B(\theta, C_0)}), \theta)=1-1+1+1=2. $$ This implies that $A$ has at least one fixed point $x_6\in B(\theta, R_1)\backslash(\overline{O_1}\cup\overline{O_4}\cup\overline{B(\theta, C_0)}) $. Obviously, $x_5$ and $x_6$ are two distinct sign-changing solutions of the boundary-value problem \eqref{e1.1}. The proof is complete. \end{proof} By the method used in the proof of Theorem \ref{thm2.1}, it is easy to show the following four corollaries. \begin{corollary} \label{coro1} Suppose that (H0), (H1) and (H3) hold, and that there exists positive integer $n_0$ such that $\lambda_{2n_0}<\beta_0<\lambda_{2n_0+1}$. Then the boundary-value problem \eqref{e1.1} has at least one sign-changing solution. Moreover, the boundary-value problem \eqref{e1.1} has at least one positive solution and one negative solution. \end{corollary} \begin{corollary} \label{coro2} Suppose that (H0), (H1) and (H3) hold, and that there exists positive integer $n_1$ such that $\lambda_{2n_1}<\beta_1<\lambda_{2n_1+1}$. Then the conclusion of Corollary \ref{coro1} holds. \end{corollary} \begin{corollary} \label{coro3} Suppose that (H0) and (H1) hold, $\beta_0>\lambda_1$, $\beta_1<\lambda_1$ (or $\beta_0<\lambda_1$, $\beta_1>\lambda_1$). Then the boundary-value problem \eqref{e1.1} has at least one positive solution and one negative solution. \end{corollary} \begin{corollary} \label{coro4} Suppose that (H0), (H1) and (H3) hold, $\beta_0>\lambda_1$, $\beta_1>\lambda_1$. Then the boundary-value problem \eqref{e1.1} has at least two positive solutions and two negative solutions. \end{corollary} \noindent{\bf Remark.}\; %3. In Theorem \ref{thm2.1}, we show not only the existence of multiple sign-changing solutions, but also the existence of multiple positive solutions and negative solutions. Obviously, we can employ this method to show the existence of sign-changing solutions for other nonlinear boundary-value problems. \subsection*{Acknowledgment} The author wishes to express his sincere gratitude to the anonymous referees for their helpful suggestions in improving the paper. \begin{thebibliography}{00} \bibitem{d1} Guo Dajun, V. Lakshmikantham, \emph{Nonlinear Problems in Abstract Cones}, Academic press, Inc, New York, 1988. \bibitem{d2} Guo Dajun, \emph{Nonlinear Anylysis}, Shandong Sci. $\&$ Tec. Press, Jinan, 2001. \bibitem{f1}W. Feng and J. R. L. Webb, \emph{Solvability of a m-point nonlinear boundary-value problem with nonlinear growth}, J. Math. Anal. Appl., 212(1997),467-480. \bibitem{g1} C. P. Gupta, S. I. Trofimchuk, \emph{Existence of a solution of a three-point boundary-value problem and spectral radius of a related linear operator}, Nonlinear Anal., 34(1998):489-507. \bibitem{g2} C. P. Gupta, S. I. Trofimchuk, \emph{A sharper condition for the solvability of a three-point second order boundary-value problem}, J. Math. Anal. Appl., 205(1997),586-597. \bibitem{k1} M. A. Krasnosel'ski$\breve{1}$, P. P. Zabre$\breve{1}$ko, \emph{Geometrical Methods of Nonlinear Analysis}, Springer-Verlag, Berlin Heidelberg, New York, Tokyo, 1984. \bibitem{l1} X. Liu, \emph{Nontrivial solutions of singular nonlinear m-point boundary-value problems}, J. Math. Anal. Appl.,284(2003),576-590. \bibitem{l2} Liu Z. L., \emph{On Dancer's conjeture and multiple solutions of elliptic partial differential equations}, Northeast Mathematics J., 9(1993)388-394. \bibitem{m1} R. Ma, \emph{Existence of solutions of nonlinear m-point boundary-value problems}, J. Math. Anal. Appl., 256, 556-567(2001). \bibitem{m3}R. Ma, \emph{Positive solutions of a nonlinear three-point boundary-value problem}, Electronic Journal of Differential Equations, Vol. 1998(1998) No. 34, 1-8. \bibitem{m4} M. Moshinsky, \emph{Sobre los problemas de condiciones a la frontiera en una dimension de caracteristicas discontinuas}, Bol. Soc. Mat. Mexicana 7 (1950) 1¨C25. \bibitem{r1} Youssef N. Raffoul, \emph{Positive solutions of three-point nonlinear second order boundary-value problem}, EJQTDE, 2002, Vol. 15, 1-11. \bibitem{t1} S. Timoshenko, \emph{Theory of Elastic Stability}, McGraw-Hill, New York, 1961. \bibitem{w1} J. R. L. Webb, \emph{Positive solutions of some three point boundary-value problems via fixed point index}, Nonlinear Analysis, 47(2001), 4319-4332. \bibitem{x1} X. Xu, \emph{Positive Solutions for Singular m-point Boundary Value Problems with Positive Parameter}, J. Math. Anal. Appl.,291(2004)352-367. \bibitem{x2} X. Xu, \emph{Multiplicity Results for Positive Solutions of Some Semi-positone Three-point Boundary Value Problems}, J. Math. Anal. Appl., 291(2004)673-689. \bibitem{y1} Qing Liu Yao, \emph{Existence and multiplicity of positive solutions for a class of second order three-point nonlinear boundary-value problems}, Acta Mathematica Sinica, Vol. 45, No 6, 2002. \end{thebibliography} \end{document}