\documentclass[reqno]{amsart} \usepackage[notref,notcite]{showkeys} \usepackage{color} \AtBeginDocument{{\noindent\small \emph{Electronic Journal of Differential Equations}, Vol. 2010(2010), No. 171, pp. 1--8.\newline ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu \newline ftp ejde.math.txstate.edu} \thanks{\copyright 2010 Texas State University - San Marcos.} \vspace{9mm}} \begin{document} \title[\hfilneg EJDE-2010/171\hfil Positive solutions] {Positive solutions for singular Sturm-Liouville boundary value problems on the half line} \author[J. Xu, Z. Yang\hfil EJDE-2010/171\hfilneg] {Jiafa Xu, Zhilin Yang} % in alphabetical order \address{ Department of Mathematics, Qingdao Technological University, Qingdao, Shandong Province, China} \email[Jiafa Xu]{xujiafa292@sina.com} \email[Zhilin Yang]{zhilinyang@sina.com} \thanks{Submitted April 1, 2010. Published November 30, 2010.} \thanks{Supported by 10871116 and 10971179 from the NNSF of China, and ZR2009AL014 from \hfill\break\indent the NSF of Shandong Province of China} \subjclass[2000]{34B40, 34B16, 47H07, 47H11, 45M20} \keywords{Sturm-Liouville problem on the half line; positive solution; \hfill\break\indent fixed point index; spectral radius} \begin{abstract} This article concerns the existence and multiplicity of positive solutions for the singular Sturm-Liouville boundary value problem \begin{gather*} (p(t)u'(t))'+h(t)f(t,u(t))=0,\quad 00$ on $(0,\infty)$ and $\int_0^{\infty}\frac{{\rm\,d}s}{p(s)}<\infty$, $a,b,c,d\geq 0$ with $\rho:=bc+ad+ac\int_0^{\infty}\frac{{\rm\,d}s}{p(s)}>0$. Boundary value problems on the half line arise in studying radially symmetric solutions of nonlinear elliptic equations and in various applications, such as an unsteady flow of gas through a semi-infinite porous media, theory of drain flows, and plasma physics (see for example \cite{Baxley,Chen,Hai,Tan}). This explains the reason that the last two decades have witnessed an overgrowing interest in the research of such problems, with many papers in this direction published. We refer the interested reader to \cite{Agarwal,Liu1,Liu2,Liu3,Lian1,Lian2, Ma3,Sun,Tian1,Tian2,Yan,Zhang,Zhang1} and references therein. In \cite{Lian2}, by using fixed point theorems in a cone, Lian et al. considered problem \eqref{problem} and obtained a set of sufficient conditions that guarantee existence, uniqueness, and multiplicity of positive solutions for \eqref{problem}. An interesting feature in \cite{Lian2} is that the nonlinearity $f$ may be sign-changing. In \cite{Zhang}, by using fixed point index theory, Zhang et al. studied the existence of positive solutions for \eqref{problem} with $h(t)f(t,u(t))$ replaced by the semipositone nonlinearity $f(t,u(t))+q(t)$, generalizing and improving some results due to Liu \cite{Liu2} and Zhang et al. \cite{Zhang1}. Motivated by the works cited above, we discuss the existence and multiplicity of positive solutions for \eqref{problem}. We use fixed point index theory to establish our main results based on a priori estimates derived by utilizing spectral properties of associated linear integral operators. This means that both our methodology and results in this paper are different from those in \cite{Liu2, Lian1,Lian2,Zhang,Zhang1}. The article is organized as follows. Section 2 contains some preliminary results, including spectral properties of two linear integral operators. In Section 3, we state and prove our main results. Four examples are given in Section 4 to illustrate applications of Theorems \ref{thm1}--\ref{thm3}. \section{Preliminaries} Let $E=\{u\in C(\mathbb{R}^+) :\lim_{t\to \infty}u(t) \text{ exists}\}$ be equipped with the supremum norm $\|\cdot\|$ and $P=\{u\in E: u(t)\geq 0, t\in \mathbb{R}^+\}$. Then $(E,\|\cdot\|)$ is a real Banach space and $P$ a cone on $E$. For simplicity, we denote $\xi(t)$ and $\eta(t)$ by \[ \xi(t):=b+a\int_0^t \frac{{\rm\,d}s}{p(s)},\quad \eta(t):=d+c\int_t^{\infty} \frac{{\rm\,d}s}{p(s)}. \] Clearly, $c\xi(t)+a\eta(t)=\rho$. Let \begin{equation}\label{Green} G(t,s):=\frac{1}{\rho}\begin{cases} \xi(t)\eta(s),& 0\leq t\leq s<\infty,\\ \xi(s)\eta(t),& 0\leq s\leq t<\infty.\\ \end{cases} \end{equation} We assume the following conditions hold throughout this article. \begin{itemize} \item[(H1)] $h$ is Lebesgue measurable and nonnegative on $\mathbb{R}^+$, with $\int_0^{\infty}G(s,s)h(s){\rm\,d}s \in (0,\infty)$. \item[(H2)] $p\in C(\mathbb{R}^+)\cap C^1(0,\infty)$, with $p>0$ in $(0,\infty$) and $\int_0^{\infty}\frac{{\rm\,d}s}{p(s)}<\infty$. \item[(H3)] $f\in C(\mathbb{R}^+\times\mathbb{R}^+, \mathbb{R}^+)$ is bounded on $\mathbb{R}^+\times [0,R]$ for every $R>0$. \end{itemize} \begin{lemma}\label{lemma1} Let {\rm (H1)-(H3)} hold and $G(t,s)$ be given in \eqref{Green}. Then \eqref{problem} is equivalent to the fixed point equation $u=Au$, where $A:P\to P$ is defined by \begin{equation}\label{A} (Au)(t):=\int_0^\infty G(t,s)h(s)f(s,u(s)){\rm\,d}s, t\in \mathbb R^+.\end{equation} \end{lemma} \begin{lemma}[\cite{Liu3,Lian2,Zhang}] \label{continous} Let {\rm (H1)-(H3)} hold. Then $A:P\to P$ is a completely continuous operator. \end{lemma} \begin{remark} \label{rmk2.1} \rm Note that $G$ satisfies the following properties, which are obtained from the monotonicity of $\xi$ and $\eta$: \begin{itemize} \item[(1)] $G(t,s)$ is continuous and $0\leq G(t,s)\leq G(s,s)$ for all $t,s\in \mathbb{R}^+$, \item[(2)] $G(t,s)\geq \gamma(t) G(s,s)$ for all $t,s\in \mathbb{R}^+$, where \begin{equation}\label{gamma} \gamma(t):=\min\big\{\frac{\xi(t)}{\xi(\infty)}, \frac{\eta(t)}{\eta(0)}\big\}, \quad t\in\mathbb{R}^+. \end{equation} \end{itemize} \end{remark} \begin{lemma}[\cite{Guo_1}] \label{lemma2} Let $\Omega \subset E$ be a bounded open set and $A:\overline{\Omega}\cap P\to P$ a completely continuous operator. If there exists $u_0\in P\setminus\{0\}$ such that $u-Au\neq \mu u_0$ for all $\mu \geq 0$ and $u\in \partial\Omega\cap P$, then $i(A,\Omega\cap P,P)=0$ , where $i$ indicates the fixed point index on $P$. \end{lemma} \begin{lemma}[\cite{Guo_1}] \label{lemma3} Let $\Omega \subset E$ be a bounded open set with $0\in \Omega$. Suppose $A:\overline{\Omega}\cap P\to P$ is a completely continuous operator. If $u\neq \mu Au$ for all $u\in \partial\Omega\cap P$ and $0\leq \mu\leq 1$, then $i(A,\Omega\cap P,P) = 1$. \end{lemma} To establish our main results, we need two extra completely continuous linear operators $T$ and $S$, defined by \begin{equation} \label{T} (Tu)(t):=\int_0^\infty G(t,s)h(s)u(s){\rm\,d}s,\quad u\in E, \end{equation} \[ (Sv)(s):=\int_0^\infty G(t,s)h(s)v(t){\rm\,d}t,\quad v\in L(\mathbb{R}^+). \] Condition (H1) implies $S: L(\mathbb{R}^+)\to L(\mathbb{R}^+)$ maps every nonnegative function in $L(\mathbb{R}^+)$ to a nonnegative function. Note that $T$ can be viewed as the dual operator of $S$ for the reason that $T$ can be extended to a bounded linear operator $\overline T: L^\infty(\mathbb{R}^+)\to L^\infty(\mathbb{R}^+)$, satisfying $\overline T(L(\mathbb{R}^+))\subset E$. It is easy to prove that the spectral radius of $T$, denoted by $r(T)$, is positive. Now the well-known Krein-Rutman \cite{Krein} theorem asserts that there exist a $\varphi\in P\backslash \{0\}$ and a nonnegative $\psi\in L(\mathbb{R}^+)\setminus\{0\}$ such that $T\varphi=r(T)\varphi$ and $S\psi=r(T)\psi$, which can be written as \begin{equation}\label{Krein-Rutman-1} \int_0^{\infty} G(t,s)h(s)\varphi(s){\rm\,d}s=r(T)\varphi(t),\end{equation} and \begin{equation}\label{Krein-Rutman-2} \int_0^{\infty} G(t,s)h(s)\psi(t){\rm\,d}t=r(T)\psi(s). \end{equation} Note that $\psi$ may be required to satisfy \begin{equation}\label{integral} \int_0^{\infty}\psi(t){\rm\,d}t=1. \end{equation} \begin{remark} \label{rmk2.2} \rm Let $\lambda_1=1/r(T)>0$. Then \eqref{Krein-Rutman-1} can be written in the form \begin{equation}\label{eigenvalue}\begin{gathered} (p(t)\varphi'(t))'+\lambda_1h(t)\varphi(t)=0,\quad 00$. Then $T(P)\subset P_0$ and, in particular, $\varphi\in P_0$ . \end{lemma} \section{Main results} Let $B_\delta :=\{u \in E: \|u\| <\delta\}$ for $\delta> 0$. \begin{theorem}\label{thm1} Let {\rm (H1)-(H3)} hold. Suppose \begin{gather}\label{sublinear-1} \liminf_{u\to 0^+}\frac{f(t,u)}{u}> \lambda_1, \\ \label{sublinear-2} \limsup_{u\to \infty}\frac{f(t,u)}{u}< \lambda_1 \end{gather} uniformly for $t\in \mathbb{R}^+$. Then \eqref{problem} has at least one positive solution. \end{theorem} \begin{proof} By \eqref{sublinear-1}, there exist $r>0$ and $\varepsilon>0$ such that \[ f(t,u)\geq (\lambda_1+\varepsilon)u,\quad \forall t\in \mathbb{R}^+,\; u\in [0,r], \] and thus for any $u\in \overline{B}_r\cap P$, we have \begin{equation}\label{Inequ-A-1} (Au)(t)\geq (\lambda_1+\varepsilon) \int_0^{\infty}G(t,s)h(s)u(s){\rm\,d}s =(\lambda_1+\varepsilon)(Tu)(t),\quad t\in \mathbb{R}^+. \end{equation} Next we shall show that \begin{equation}\label{notequ-1} u-Au\neq \mu \varphi,\quad \forall u\in \partial B_r \cap P, \mu\geq 0, \end{equation} where $\varphi$ is defined by \eqref{Krein-Rutman-1}. If the claim is false, then there exist $u_1\in \partial B_r \cap P, \mu_1\ge 0$ such that $u_1-Au_1= \mu_1 \varphi$. Thus $u_1=Au_1+\mu_1 \varphi\in P_0$ by Lemma \ref{lemma4} and $u_1\geq Au_1$. Combining the preceding inequality with \eqref{Inequ-A-1} (replacing $u$ by $u_1$) leads to \begin{equation}\label{Inequ-A-2} u_1(t)\ge (\lambda_1+\varepsilon) \int_0^{\infty}G(t,s)h(s)u_1(s){\rm\,d}s, t\in\mathbb{R}^+. \end{equation} Multiply the above by $\psi(t)$ and integrate over $\mathbb{R}^+$ and use \eqref{Krein-Rutman-2} and \eqref{integral} to obtain \[ \int_0^{\infty}u_1(t)\psi(t){\rm\,d}t \ge\lambda_1^{-1}(\lambda_1+\varepsilon) \int_0^{\infty}u_1(t)\psi(t){\rm\,d}t, \] so that $\int_0^{\infty}u_1(t)\psi(t){\rm\,d}t=0$. Recalling $u_1\in P_0$, we have $u_1\equiv 0$, a contradiction with $u_1\in \partial B_r \cap P$. As a result, \eqref{notequ-1} holds. Now Lemma \ref{lemma2} implies \begin{equation}\label{index1-0} i(A,B_r\cap P,P)=0. \end{equation} On the other hand, by \eqref{sublinear-2} and (H3), there exist $0<\sigma<1$ and $M>0$ such that \begin{equation}\label{Inequ-A-3} f(t,u)\leq \sigma\lambda_1u+M,\quad \forall u\geq 0,t\in \mathbb{R}^+. \end{equation} We shall prove that the set \begin{equation} \mathcal{M}_1 :=\{u\in P: u=\mu Au, 0\leq \mu\leq 1\}. \end{equation} is bounded. Indeed, for any $u_2\in\mathcal M_1$ we have by \eqref{Inequ-A-3} \[ u_2(t)\leq \sigma\lambda_1\int_0^{\infty}G(t,s)h(s)u_2(s){\rm\,d}s +u_0(t)=\sigma\lambda_1(Tu_2)(t)+u_0(t), \] where $u_0\in P$ is defined by $u_0(t)=M\int_0^{\infty}G(t,s)h(s){\rm\,d}s$. Notice $r(\sigma\lambda_1T)=\lambda_1\sigma r(T)<1$. This implies $I-\sigma\lambda_1 T$ is invertible and its inverse equals \[ (I-\sigma\lambda_1 T)^{-1}=I+\sigma\lambda_1 T +\sigma^2\lambda_1^2 T^2+\dots+\sigma^n\lambda_1^n T^n+\dots. \] Now we have $(I-\sigma\lambda_1 T)^{-1}(P)\subset P$ and $u_2\leq (I-\sigma\lambda_1 T)^{-1}u_0$. Therefore, $\mathcal{M}_1$ is bounded. Choosing $R>\max\{r,\sup\{\|u\|: u\in\mathcal M_1\}\}$, we have by Lemma \ref{lemma3} \begin{equation}\label{index1-1} i(A,B_{R}\cap P,P)=1. \end{equation} Now \eqref{index1-0} and \eqref{index1-1} imply \begin{equation} i(A,(B_{R}\setminus \overline{B}_r)\cap P,P)=i(A,B_{R}\cap P,P)-i(A,B_r\cap P,P)=1. \end{equation} Thus the operator $A$ has at least one fixed point on $(B_{R}\setminus \overline{B}_r)\cap P$ and hence \eqref{problem} has at least one positive solution. The proof is completed. \end{proof} \begin{theorem}\label{thm2} Let {\rm (H1)-(H3)} hold. Suppose \begin{equation}\label{superlinear-1} \liminf_{u\to \infty}\frac{f(t,u)}{u}> \lambda_1 \end{equation} and \begin{equation}\label{superlinear-2} \limsup_{u\to 0^+}\frac{f(t,u)}{u}< \lambda_1 \end{equation} uniformly for $t\in \mathbb{R}^+$. Then \eqref{problem} has at least one positive solution. \end{theorem} \begin{proof} By \eqref{superlinear-1} and (H3), there exist $\varepsilon > 0$ and $b> 0$ such that \begin{equation}\label{fsuper} f(t,u)\geq (\lambda_1 + \varepsilon)u - b, \quad \forall u\geq 0, \; t\in \mathbb{R}^+. \end{equation} This implies \begin{equation}\label{Ineq-2} (Au)(t)\ge(\lambda_1+\varepsilon)\int_0^{\infty} G(t,s)h(s)u(s){\rm\,d}s-b\int_0^{\infty}G(t,s)h(s){\rm\,d}s \end{equation} for all $u\in P$ and $t\in \mathbb{R}^+$. We shall prove that the set \begin{equation}\label{notequ-2} \mathcal{M}_2:=\{u\in P: u=Au+ \mu\varphi, \mu\geq 0\}. \end{equation} is bounded, where $\varphi\in P$ is given by \eqref{Krein-Rutman-1}. Indeed, if $u\in \mathcal{M}_2$, then we have $u\geq Au$ by definition and $u\in P_0$ by Lemma \ref{lemma4}. This together with \eqref{Ineq-2} leads to \[ u(t)\ge (\lambda_1+\varepsilon)\int_0^{\infty} G(t,s)h(s)u(s){\rm\,d}s-b\int_0^{\infty}G(t,s)h(s){\rm\,d}s, \quad t\in\mathbb{R}^+. \] Multiply the above by $\psi (t)$ and integrate over $\mathbb{R}^+$ and use \eqref{Krein-Rutman-2} and \eqref{integral} to obtain \[ \int_0^{\infty} \psi(t)u(t){\rm\,d}t \geq(\lambda_1 + \varepsilon)\lambda_1^{-1}\int_0^{\infty} \psi(t)u(t){\rm\,d}t-b\lambda_1^{-1}, \] so that $\int_0^{\infty} \psi(t)u(t){\rm\,d}t\leq b\varepsilon^{-1}$ for all $u\in\mathcal M_2$. Recalling $u\in P_0$, we obtain $\|u\|\leq (\varepsilon\omega)^{-1}b$ for all $u\in \mathcal M_2$, and thus $\mathcal M_2$ is bounded, as required. Taking $R>\sup\{\|u\|: u\in\mathcal M_2\}$, we have \begin{equation} u-Au\neq \mu\varphi, \quad \forall u\in \partial B_R\cap P,\; \mu\geq 0. \end{equation} Now Lemma \ref{lemma2} yields \begin{equation}\label{index2-0} i(A,B_R\cap P,P)=0. \end{equation} By \eqref{superlinear-2}, there exist $ r\in (0, R)$ and $\sigma\in (0,\lambda_1)$ such that \[ f(t,u) \leq (\lambda_1-\sigma)u, \quad \forall 0 \leq u \leq r,\; 0\leq t<\infty, \] so that \begin{equation}\label{Inequ-1} (Au)(t)\leq (\lambda_1-\sigma)\int_0^\infty G(t,s)h(s)u(s){\rm\,d}s \end{equation} for all $u\in\overline B_r\cap P$ and $t\in \mathbb{R}^+$. We claim that \begin{equation}\label{hom} u\neq \mu Au, \quad \forall u\in \partial B_r\cap P,\quad 0\leq \mu\leq 1. \end{equation} Suppose, to the contrary, there exist $u_1 \in \partial B_r \cap P$ and $\mu_1 \in[0,1]$ such that $u_1 = \mu_1Au_1$. Then we have $u_1\in P_0$ by Lemma \ref{lemma4} and \[ u_1(t)\leq (\lambda_1-\sigma)\int_0^\infty G(t,s)h(s)u_1(s){\rm\,d}s, \quad t\in\mathbb{R}^+ \] by \eqref{Inequ-1}. Multiply the above by $\psi (t)$ and integrate over $\mathbb{R}^+$ and use \eqref{Krein-Rutman-2} and \eqref{integral} to obtain \[ \int_0^{\infty} \psi(t)u_1(t){\rm\,d}t \leq \frac{\lambda_1-\sigma}{\lambda_1} \int_0^{\infty} \psi(t)u_1(t){\rm\,d}t, \] so that $\int_0^{\infty} \psi(t)u_1(t){\rm\,d}t=0$. Recalling $u_1\in P_0$, we obtain $u_1= 0$, a contradiction with $u_1\in \partial B_r\cap P$. As a result, \eqref{hom} is true. Now Lemma \ref{lemma3} yields \begin{equation} \label{index2-1} i(A,B_r\cap P,P)=1. \end{equation} Combining \eqref{index2-0} and \eqref{index2-1} gives \begin{equation} i(A,(B_R\setminus \overline{B}_r)\cap P,P)=i(A,B_R\cap P,P)-i(A,B_r\cap P,P)=-1. \end{equation} Consequently the operator $A$ has at least one fixed point on $(B_R\setminus \overline{B}_r)\cap P$, and hence \eqref{problem} has at least one positive solution. The proof is completed. \end{proof} \begin{theorem}\label{thm3} Let {\rm (H1)-(H3)} hold. Suppose that $f(t,u)$ satisfies \eqref{sublinear-1} and \eqref{superlinear-1}. Moreover, $f(t,u)$ is nondecreasing in $u$, and that there exists $N> 0$ such that \begin{equation}\label{nondecreasing} f(t,N)<\frac{N}{\kappa},\quad\text{a.e. } t\in\mathbb{R}^+, \end{equation} where $\kappa:=\int_0^{\infty} G(s,s)h(s){\rm\,d}s>0.$ Then \eqref{problem} has at least two positive solutions. \end{theorem} \begin{proof} The monotonicity of $f$ implies that for all $u\in \overline B_N\cap P$ and $t\in \mathbb{R}^+$, we have \begin{equation} (Au)(t)=\int_0^{\infty} G(t,s)h(s)f(s,u(s)){\rm\,d}s < \int_0^{\infty} G(s,s)h(s)\frac{N}{\kappa}{\rm\,d}s= N, \end{equation} so that $\|Au\|< \|u\|$ for all $u\in \partial B_N\cap P$. A consequence of this is \[ u\neq \mu Au, \forall u\in \partial B_N\cap P, 0\leq \mu\leq 1. \] Now Lemma \ref{lemma3} implies \begin{equation}\label{index3-1} i(A,B_N\cap P,P)=1. \end{equation} On the other hand, in view of \eqref{sublinear-1} and \eqref{superlinear-1}, we may take $R>N$ and $r\in (0,N)$ so that \eqref{index1-0} and \eqref{index2-0} hold (see the proofs of Theorems \ref{thm1} and \ref{thm2}). Combining \eqref{index1-0}, \eqref{index2-0} and \eqref{index3-1}, we arrive at \[ i(A,(B_R\backslash \overline{B}_{N})\cap P,P)=0-1=-1,\ \ i(A,(B_N\backslash \overline{B}_{r})\cap P,P)=1-0=1. \] Consequently the operator $A$ has at least two fixed points, one on $(B_R\backslash \overline{B}_{N})\cap P$ and the other on $(B_N\backslash \overline{B}_{r})\cap P$. Hence (1.1) has at least two positive solutions. The proof is completed. \end{proof} \section{Examples} In this section, we provide four examples to illustrate applications of Theorems \ref{thm1}--\ref{thm3}. Let us consider the boundary value problem \begin{equation}\label{example} \begin{gathered} ((1+t^2)u'(t))'+t^{-1/2}e^{-t}f(t,u)=0,\quad 0