\documentclass[reqno]{amsart} \usepackage{hyperref} \usepackage{graphicx} \AtBeginDocument{{\noindent\small Eighth Mississippi State - UAB Conference on Differential Equations and Computational Simulations. {\em Electronic Journal of Differential Equations}, Conf. 19 (2010), pp. 135--149.\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} \setcounter{page}{135} \title[\hfilneg EJDE-2010/Conf/19/\hfil short Population models] {Population models with nonlinear boundary conditions} \author[J. Goddard II, E. K. Lee, R. Shivaji \hfil EJDE/Conf/19 \hfilneg] {Jerome Goddard II, Eun Kyoung Lee, Ratnasingham Shivaji} % in alphabetical order \address{Jerome Goddard II \newline Department of Mathematics and Statistics\\ Center for Computational Sciences\\ Mississippi State University\\ Mississippi State, MS 39762, USA} \email{jg440@msstate.edu} \address{Eun Kyoung Lee \newline Department of Mathematics\\ Pusan National University\\ Busan 609-735, Korea} \email{eunkyoung165@gmail.com} \address{Ratnasingham Shivaji\newline Department of Mathematics and Statistics\\ Center for Computational Sciences\\ Mississippi State University\\ Mississippi State, MS 39762, USA} \email{shivaji@ra.msstate.edu} \thanks{Published September 25, 2010.} \thanks{E. K. Lee was supported by grant NRF-2009-353-C00042 from the National Research \hfill\break\indent Foundation of Korea} \subjclass[2000]{34B18, 34B08} \keywords{Nonlinear boundary conditions; logistic growth; positive solutions} \begin{abstract} We study a two point boundary-value problem describing the steady states of a Logistic growth population model with diffusion and constant yield harvesting. In particular, we focus on a model when a certain nonlinear boundary condition is satisfied. \end{abstract} \maketitle \numberwithin{equation}{section} \newtheorem{theorem}{Theorem}[section] \section{Introduction} Consider the Logistic growth population dynamics model with nonlinear boundary conditions: \begin{gather} \label{1.1} u_t = d\Delta u + au - bu^2 - c h(x) \quad \text{in } \Omega,\\ \label{1.2} d \alpha(x, u)\frac{\partial u}{\partial \eta} + [1 - \alpha(x, u) ]u = 0 \quad \text{on }\partial \Omega, \end{gather} where $\Omega$ is a bounded domain in $\mathbb{R}^n$ with $n \geq 1$, $\Delta$ is the Laplace operator, $d$ is the diffusion coefficient, $a, b$ are positive parameters, $c \geq 0$ is the harvesting parameter, $h(x):\overline{\Omega} \to \mathbb{R}$ is a $C^1$ function, $\frac{\partial u}{\partial \eta}$ is the outward normal derivative, and $\alpha(x, u):\Omega \times \mathbb{R} \to [0, 1]$ is a nondecreasing $C^1$ function. The parameter $c \geq 0$ represents the level of harvesting, $h(x) \geq 0$ for $x \in \Omega$, $h(x) = 0$ for $x \in \partial \Omega$, and $\|h\|_\infty = 1$. Here $c h(x) $ can be understood as the rate of the harvesting distribution. The nonlinear boundary condition \eqref{1.2} has only been recently studied by such authors as \cite{Cantrell2002g31, Cantrell2003gsb, Cantrell2007g33}, among others. Here \[ \alpha(x, u) = \alpha(u) = \frac{u}{u - d \frac{\partial u}{\partial \eta}} \] represents the fraction of the population that remains on the boundary when reached. For the case when $\alpha(x, u) \equiv 0$, \eqref{1.2} becomes the well known Dirichlet boundary condition. If $\alpha(x, u) \equiv 1$ then \eqref{1.2} becomes the Neumann boundary condition. Here we will be interested in the study of positive steady state solutions of \eqref{1.1}--\eqref{1.2} when $d = 1$ and \[ \alpha(x, u) = \frac{u}{u + 1} \quad \text{on }\partial \Omega. \] Hence, we consider the model \begin{gather} \label{1.5} -\Delta u = au - bu^2 -ch(x) =: f(x, u) \quad \text{in }\Omega,\\ \label{1.6} u [\frac{\partial u}{\partial \eta} + 1 ] = 0 \quad \text{on }\partial \Omega. \end{gather} We will present the results of the case when $n = 1$, $\Omega = (0, 1)$, and $h(x) \equiv 1$. Thus, we study the nonlinear boundary-value problem \begin{gather} \label{1.11}-u'' = au - bu^2 - c, \quad x \in(0, 1),\\ \label{1.12} [-u'(0) + 1 ]u(0) = 0,\\ \label{1.13} [u'(1) + 1 ]u(1) = 0. \end{gather} It is easy to see that analyzing the positive solutions of \eqref{1.11}--\eqref{1.13} is equivalent to studying the four boundary-value problems \begin{gather} \label{1.14}-u'' = au - bu^2 - c, \quad x \in(0, 1),\\ %\label{1.15} u(0) = 0,\quad \label{1.16} u(1) = 0; \end{gather} \begin{gather} \label{1.23}-u'' = au - bu^2 - c, \quad x \in(0, 1),\\ %\label{1.24} u(0) = 0, \quad \label{1.25} u'(1) = -1; \end{gather} \begin{gather} \label{1.20}-u'' = au - bu^2 - c, \quad x \in(0, 1),\\ %\label{1.21} u'(0) = 1, \quad \label{1.22} u(1) = 0; \end{gather} \begin{gather} \label{1.17}-u'' = au - bu^2 - c, \quad x \in(0, 1),\\ %\label{1.18} u'(0) = 1, \quad \label{1.19}u'(1) = -1. \end{gather} Hence, the positive solutions of these four BVPs are the positive solutions of \eqref{1.11}--\eqref{1.13}. Notice that if $u(x)$ is a solution of \eqref{1.23}--\eqref{1.25} then $v(x) := u(1 - x)$ is a solution of \eqref{1.20}--\eqref{1.22}. Thus, it suffices to only consider \eqref{1.14}--\eqref{1.16}, \eqref{1.23}--\eqref{1.25}, and \eqref{1.17}--\eqref{1.19}. The structure of positive solutions for \eqref{1.14}--\eqref{1.16} is known (see \cite{Collins2004g27} and \cite{Ladner2005g29}) via the quadrature method introduced by Laetsch in \cite{Laetsch1970g23}. We develop quadrature methods in Section 2 to completely determine the bifurcation diagram of \eqref{1.11}--\eqref{1.13}. In Section 3 we use Mathematica computations to show that for certain subsets of the parameter space, \eqref{1.11}--\eqref{1.13} has up to exactly 8 positive solutions. For higher dimensional results, in the case when $\alpha(x, u) = 0$ on $\partial\Omega$ (Dirichlet boundary conditions) see \cite{Oruganti2002g10}, and for the case when $\alpha(x, u) = \frac{u}{u + 1}$ on $\partial\Omega$ see recent work in \cite{goddard2009gM3}. \section{Results via the quadrature method} \subsection{Positive solutions of \eqref{1.14}--\eqref{1.16}} In this section we summarize the known results (see \cite{Oruganti2002g10}) for positive solutions of \eqref{1.14}--\eqref{1.16}. Consider the boundary value problem: \begin{gather} \label{2.1}-u'' = au - bu^2 - c =: f(u), \quad x \in(0, 1),\\ %\label{2.2} u(0) = 0, \quad \label{2.3}u(1) = 0. \end{gather} \begin{figure}[ht] \begin{center} \includegraphics[width=0.5\textwidth]{fig1} % pde3_u_DBC \end{center} \caption{Typical solution of \eqref{2.1}--\eqref{2.3}} \label{fig2.1} \end{figure} It is easy to see that positive solutions of \eqref{2.1}--\eqref{2.3} must resemble Figure \ref{fig2.1} where $\ell_i$ for $i = 1, 2$ are the positive zeros of $f(u)$. The following theorem details the structure of positive solutions of \eqref{2.1}--\eqref{2.3} for the case when $b = 1$: \begin{theorem}[\cite{Collins2004g27,Oruganti2002g10}] \label{Thm2.1} \begin{itemize} \item[(1)] If $a < \lambda_1$ then \eqref{2.1}--\eqref{2.3} has no positive solution for any $c \geq 0$. \item[(2)] If $\lambda_1 \leq a < \lambda^*$ (some $\lambda^* > \lambda_1$) then there exists a $c_0 > 0$ such that if \begin{itemize} \item[(a)] $0 \leq c < c_0$ then \eqref{2.1}--\eqref{2.3} has 2 positive solutions. \item[(b)] $c = c_0$ then \eqref{2.1}--\eqref{2.3} has a unique positive solution. \item[(c)] $c > c_0$ then \eqref{2.1}--\eqref{2.3} has no positive solution. \end{itemize} \item[(3)] If $a > \lambda^*$ then there exist $c_0, \tilde{c} > 0$ such that if \begin{itemize} \item[(a)] $\tilde{c} < c < c_0$ then \eqref{2.1}--\eqref{2.3} has 2 positive solutions. \item[(b)] $0 \leq c < \tilde{c}$ or $c = c_0$ then \eqref{2.1}--\eqref{2.3} has a unique positive solution. \item[(c)] $c > c_0$ then \eqref{2.1}--\eqref{2.3} has no positive solution. \end{itemize} \end{itemize} \end{theorem} Figure \ref{fig2.2} illustrates this theorem. \begin{figure}[ht] \begin{center} \includegraphics[width=0.45\textwidth]{fig2a} % PvsC_DBC_b_1_a_10_P} \includegraphics[width=0.45\textwidth]{fig2b} % PvsC_DBC_b_1_a_40_P} \end{center} \caption{$a = 10$, $b = 1$ (left), and $a = 40$, $b = 1$ (right)} \label{fig2.2} % \label{fig2.3} \end{figure} \subsection{Positive solutions of \eqref{1.23}--\eqref{1.25}} In this subsection, we adapt the quadrature method in \cite{Laetsch1970g23} to study \begin{gather} \label{2.23}-u'' = au - bu^2 - c =: f(u), \quad x \in(0, 1),\\ %\label{2.24} u(0) = 0, \quad \label{2.25}u'(1) = -1. \end{gather} Now, define $F(u) = \int_0^{u} f(s) ds$, the primitive of $f(u)$. Since \eqref{2.23} is an autonomous differential equation, if $u(x)$ is a positive solution of \eqref{2.23} with $u'(x_0) = 0$ for some $x_0 \in (0, 1)$ then $v(x) := u(x_0 - x)$ and $w(x) := u(x_0 + x)$ both satisfy the initial value problem, \begin{gather} \label{IVP7}-z'' = f(z)\\ \label{IVP8}z(0) = u(x_0)\\ \label{IVP9}z'(0) = 0 \end{gather} for all $x \in [0, d)$ where $d = min\{x_0, 1 - x_0 \}$. As a result of Picard's existence and uniqueness theorem, $u(x_0 - x) \equiv u(x_0 + x)$. Thus, if we assume that $u(x)$ is a positive solution of \eqref{2.23}--\eqref{2.25} then it is symmetric around $x_0$ with $\rho := \|u\|_\infty = u(x_0)$. This implies that $u'(x_0) = 0$, $u'(x) > 0;\ [0, x_0)$, and $u'(x) < 0;\ (x_0, 1]$. Using symmetry about $x_0$, the boundary conditions \eqref{2.25}, and the sign of $u''$ given by $f(u)$ we see that positive solutions of \eqref{2.23}--\eqref{2.25} must resemble Figure \ref{fig2.4}, where $\rho = \|u\|_\infty$ and $q = u(1)$. This implies that $\ell_1 < \rho < \ell_2$ and $0 \leq q < \rho$ where $\ell_i$, $i = 1, 2$ are the zeros of $f(u)$. \begin{figure}[ht] \begin{center} \includegraphics[width=0.5\textwidth]{fig3} % pde3_u_DSC} % fig3 \end{center} \caption{Typical solution of \eqref{2.23} -\eqref{2.25}} \label{fig2.4} \end{figure} Multiplying \eqref{2.23} by $u'$ gives \begin{equation} \label{2.7} -u'u'' = f(u)u' \end{equation} Integration of \eqref{2.7} with respect to $x$ gives, \begin{equation} \label{2.8} -\Big( \frac{[u'(x)]^2}{2}\Big) = [ F(u(x))] + K. \end{equation} Substituting $x = 1$ and $x = x_0$ into \eqref{2.8} yields, \begin{gather} \label{2.9}-K = F(q) + \frac{1}{2}\\ \label{2.10}K = -F(\rho). \end{gather} Combining \eqref{2.9} and \eqref{2.10}, we have \begin{equation} \label{2.11a}F(\rho) = F(q) + \frac{1}{2}. \end{equation} Substituting \eqref{2.10} into \eqref{2.8} yields, \begin{equation} \label{2.11} -\Big( \frac{[u'(x)]^2}{2}\Big) = [ F(u(x))] - F(\rho). \end{equation} Now, solving for $u'$ in \eqref{2.11} gives \begin{gather} \label{2.12}u'(x) = \sqrt{2}\sqrt{F(\rho) - F(u(x))}, \quad x \in [0, x_0],\\ \label{2.13}u'(x) = -\sqrt{2}\sqrt{F(\rho) - F(u(x))}, \quad x \in [x_0, 1]. \end{gather} Integrating \eqref{2.12} and \eqref{2.13} with respect to $x$ and using a change of variables, we have \begin{gather} \label{2.14}\int_0^{u(x)}\frac{ds}{\sqrt{F(\rho) - F(s)}} = \sqrt{2}x, \quad x \in [0, x_0],\\ \label{2.15}\int_{\rho}^{u(x)}\frac{ds}{\sqrt{F(\rho) - F(s)}} = -\sqrt{2}(x - x_0), \quad x \in [x_0, 1]. \end{gather} Substitution of $x = x_0$ into \eqref{2.14} and $x = 1$ into \eqref{2.15} gives \begin{gather} \label{2.16}\int_0^{\rho}\frac{ds}{\sqrt{F(\rho) - F(s)}} = \sqrt{2}x_0\\ \label{2.17}\int_{\rho}^{q}\frac{ds}{\sqrt{F(\rho) - F(s)}} = -\sqrt{2}(1 - x_0). \end{gather} Finally, subtracting \eqref{2.17} from \eqref{2.16}, yields \begin{equation} \label{2.18}\int_0^{\rho}\frac{ds}{\sqrt{F(\rho) - F(s)}} + \int_q^{\rho}\frac{ds}{\sqrt{F(\rho) - F(s)}} = \sqrt{2}, \end{equation} or equivalently, \begin{equation} \label{2.19} 2\int_0^{\rho}\frac{ds}{\sqrt{F(\rho) - F(s)}} - \int_0^{q}\frac{ds}{\sqrt{F(\rho) - F(s)}} = \sqrt{2}. \end{equation} We note that in order for $\int_0^{\rho}\frac{ds}{\sqrt{F(\rho) - F(s)}}$ to be well defined, $F(\rho) > F(s)$ for all $s \in [0, \rho)$. Moreover, the improper integral is convergent if $f(\rho) > 0$. Thus, for such a positive solution to exist, $f(u)$ and $F(u)$ must resemble Figure \ref{fig2.5}, where $\mu_1$, $\ell_i$, and $\theta_i$ are the zeros of $f'(u)$, $f(u)$, and $F(u)$ respectively for $i = 1, 2$. \begin{figure}[ht] \begin{center} % Picture of f(u) \includegraphics[width=0.45\textwidth]{fig4a} % pde3_fu_1} % \includegraphics[width=0.45\textwidth]{fig4b} % pde3_FFu_1} % \end{center} \caption{Graph of $f(u)$ (left), and of $F(u)$ (right)} \label{fig2.5} % \label{fig2.6} \end{figure} From Figure \ref{fig2.5}, we note that if $\rho \in (\theta_1, \ell_2)$ then both of these conditions hold and the integrals in \eqref{2.19} are well defined. From this and letting $c_1 := \frac{3a^2}{16b}$ and $c_2: = \frac{a^2}{4b}$, we can arrive at the following result. \begin{theorem}\label{Thm2.2} If $c > c^*(a, b)$ then \eqref{2.23}--\eqref{2.25} has no positive solution, where $c^*(a, b) = \min\{c_1, c_2\} = \frac{3a^2}{16b}$. \end{theorem} Further, since $x_0 \in (0, 1)$ is fixed for each $\rho > 0$, we need a unique $q < \rho$ corresponding to each $\rho$-value such that \eqref{2.11a} is satisfied. Otherwise, uniqueness of solutions to the initial value problem, \eqref{IVP7}--\eqref{IVP9}, would be violated. Let \[ H(x) := F(x) + \frac{1}{2}. \] It follows that $H'(x) = -bx^2 + ax -c$, $H(0) = 1/2$, and $H'(0) = -c < 0$. In order for a unique $q < \rho$ to exist such that $H(q) = F(\rho)$, $H(x)$ must have the following structure in Figure 5, where $H'(\ell_2) = 0$. So, for such a unique $q < \rho$ to exist $F(\rho) > 1/2$. \begin{figure}[ht] \begin{center} \includegraphics[width=0.5\textwidth]{fig5} % pde3_hu_1 \end{center} \label{fig2.7} \caption{Graph of $H(x)$} \end{figure} Since $\rho \in (\theta_1, \ell_2)$, for this to be true we will need $H(\ell_2) > 1/2$. In fact, if \begin{equation} \label{2.20} F(\ell_2) > \frac{1}{2} \end{equation} then clearly for $\rho \in (\theta_1, \ell_2)$ with $\rho \approx \ell_2$ we have $F(\rho) > 1/2$. It is easy to see that \eqref{2.20} will be satisfied if (solving using Mathematica) \begin{align*} c < c_3 &:= \frac{9a^2}{144b} - \frac{9(a^4 - 96ab^2)} {144b \Big( -a^6 - 240a^3b^2 + 16\big(72b^4 + \sqrt{3}\sqrt{b^2(a^3 + 12b^2)^3}\big)\Big)^{1/3}}\\ &\quad - \frac{9}{144b}\Big( -a^6 -240a^3b^2 + 16\big(72b^4 + \sqrt{3}\sqrt{b^2(a^3 + 12b^2)^3}\big) \Big) \end{align*} and for $c_3$ to be positive (again using Mathematica) \[ a > a_0 := \sqrt[3]{3 b^2} \] both hold. This leads to the following results. \begin{theorem}\label{Thm2.6} If $a \leq a_0$ then \eqref{2.23}--\eqref{2.25} has no positive solution for any $c \geq 0$. \end{theorem} \begin{theorem}\label{Thm2.7} If $a > a_0$ then there is a $c^*(a, b) \leq min\{c_1, c_2, c_3\}$ such that for $c \geq c^*$ \eqref{2.23}--\eqref{2.25} has no positive solution. \end{theorem} We now state and prove the main theorem of this subsection. \begin{theorem}\label{Thm2.8} If $a > a_0$ and $c < c^*(a, b)$ then there is a unique $r(a,b,c) \in (\theta_1, \ell_2)$ such that $F(r) = 1/2$ and \[ G(\rho) := 2\int_0^{\rho}\frac{ds}{\sqrt{F(\rho) - F(s)}} - \int_0^{q}\frac{ds}{\sqrt{F(\rho) - F(s)}} \] is well defined for all $\rho \in [r, \ell_2)$ where $q < \rho$ is the unique solution of $F(\rho) = H(q)$. Moreover, \eqref{2.23}--\eqref{2.25} has a positive solution, $u(x)$, with $\rho = \|u\|_\infty$ if and only if $G(\rho) = \sqrt{2}$ for some $\rho \in [r, \ell_2)$. \end{theorem} \begin{proof} Let $a, b > 0$ s.t. $a > a_0$ and $c \in [0, c^{*}(a, b))$. From the preceding discussion, it follows that if $u$ is a positive solution to \eqref{2.23}--\eqref{2.25} with $\rho = \|u\|_\infty$ then $G(\rho) = \sqrt{2}$. Next, suppose $G(\rho) = \sqrt{2}$ for some $\rho \in [r, \ell_2)$. Define $u(x):(0, 1) \rightarrow \mathbb{R}$ by \begin{gather} \label{2.21}\int_0^{u(x)}\frac{ds}{\sqrt{F(\rho) - F(s)}} = \sqrt{2}x, \quad x \in [0, x_0],\\ \label{2.22}\int_{\rho}^{u(x)}\frac{ds}{\sqrt{F(\rho) - F(s)}} = -\sqrt{2}(x - x_0), \quad x \in [x_0, 1]. \end{gather} Now, we show that $u(x)$ is a positive solution to \eqref{2.23}--\eqref{2.25}. It is easy to see that the turning point is given by $x_0 = \frac{1}{\sqrt{2}}\int_0^{\rho}\frac{ds}{\sqrt{F(\rho) - F(s)}}$. The function, $\int_0^{u}\frac{ds}{\sqrt{F(\rho) - F(s)}}$, is a differentiable function of $u$ which is strictly increasing from $0$ to $x_0$ as $u$ increases from $0$ to $\rho$. Thus, for each $x \in [0, x_0]$, there is a unique $u(x)$ such that \begin{equation} \label{2.23a}\int_0^{u(x)}\frac{ds}{\sqrt{F(\rho) - F(s)}} = \sqrt{2}x \end{equation} Moreover, by the Implicit Function theorem, $u$ is differentiable with respect to $x$. Differentiating \eqref{2.23a} gives \[ u'(x) = \sqrt{2[F(\rho) - F(u)]}, \quad x \in [0, x_0]. \] Similarly, $u$ is a decreasing function of $x$ for $x \in [x_0, 1]$ which yields, \[ u'(x) = -\sqrt{2[F(\rho) - F(u)]}, \quad x \in [x_0, 1]. \] This implies \[ \frac{-(u')^2}{2} = F(\rho) - F(u(x)). \] Differentiating again, we have $-u''(x) = f(u(x))$. Thus, $u(x)$ satisfies \eqref{2.23}. Now, from our assumption, $G(\rho) = \sqrt{2}$, it follows that $u(0) = 0$ and $u(1) = q(\rho)$. Since $F(\rho) = H(q(\rho)) = F(q) + \frac{1}{2}$, we have that $u'(1) = -\sqrt{2[F(\rho) - F(q)]} = -1$. Hence, the boundary conditions \eqref{2.25} are both satisfied. \end{proof} \subsection{Positive solutions of \eqref{1.17}--\eqref{1.19}} A similar quadrature method can be adapted to study \begin{gather} \label{2.4}-u'' = au - bu^2 - c =: f(u), \quad x \in(0, 1),\\ %\label{2.5} u'(0) = 1, \quad \label{2.6}u'(1) = -1. \end{gather} Again, define $F(u) = \int_0^{u} f(s) ds$, the primitive of $f(u)$. Using a similar argument as before, symmetry about $x_0$, the boundary conditions \eqref{2.4}--\eqref{2.6}, and the sign of $u''$ given by $f(u)$ ensure that positive solutions of \eqref{2.4}--\eqref{2.6} must resemble Figure 6, where $\rho = \|u\|_\infty$ and $q = u(0) = u(1)$. Clearly, $x_0 = 1/2$ in this case. \begin{figure}[ht] \begin{center} \includegraphics[width=0.5\textwidth]{fig6} % pde3_u_NBC} % \end{center} \label{fig2.8} \caption{Typical solution of \eqref{2.23}--\eqref{2.25}} \end{figure} Through an almost identical approach as the one in Section 2.2, we can prove the following results. \begin{theorem}\label{Thm2.3} If $a \leq a_0$ then \eqref{2.4}--\eqref{2.6} has no positive solution for any $c \geq 0$. \end{theorem} \begin{theorem}\label{Thm2.4} If $a > a_0$ then there is a $c^*(a, b) \leq \min\{c_1, c_2, c_3\}$ such that for $c \geq c^*$ \eqref{2.4}--\eqref{2.6} has no positive solution. \end{theorem} We now state the main theorem of this subsection. \begin{theorem}\label{Thm2.5} If $a > a_0$ and $c < c^*(a, b)$ then there is a unique $r(a,b,c) \in (\theta_1, \ell_2)$ such that $F(r) = \frac{1}{2}$ and \[ \widetilde{G}(\rho) := 2\int_0^{\rho}\frac{ds}{\sqrt{F(\rho) - F(s)}} - 2\int_0^{q}\frac{ds}{\sqrt{F(\rho) - F(s)}} \] is well defined for all $\rho \in [r, \ell_2)$ where $q < \rho$ is the unique solution of $F(\rho) = H(q)$. Moreover, \eqref{2.4}--\eqref{2.6} has a positive solution, $u(x)$, with $\rho = \|u\|_\infty$ if and only if $\widetilde{G}(\rho) = \sqrt{2}$ for some $\rho \in [r, \ell_2)$. \end{theorem} \noindent \textbf{Remark.} See \cite{Ladner2005g29} where Ladner et al. adapted the quadrature method to study the case when $\alpha(x, u) = \frac{u}{a}$ on $\partial\Omega$. Also, see \cite{Goddard2009gM1} where the quadrature method was adapted to study the case with a Strong Allee effect and $\alpha(x, u) = \frac{u}{b}$ on $\partial\Omega$. \section{Computational results} \subsection{Positive solutions of \eqref{1.23}--\eqref{1.25} and \eqref{1.20}--\eqref{1.22}} We are particularly interested in the case when $b = 1$. From Theorem \ref{Thm2.8}, we plot the level sets of \begin{equation} \label{3.2}G(\rho) - \sqrt{2} = 0 \end{equation} for $a > \sqrt[3]{3}$ and $\rho \in [r, \ell_2)$. By implementing a numerical root-finding algorithm in Mathematica we were able to solve equation \eqref{3.2}. Explicit formulas were used to calculate the unique $r = r(a,b,c)$ and $q = q(\rho)$ values. Note that these computations are expensive due to the natural of the improper integral equations involved. Figures \ref{fig3.1} - \ref{fig3.5} depict several level sets plotted within $[r, \ell_2) \times [0, c^*]$. In what follows, the green curve represents $\rho$ vs $c$ while the upper and lower branches of the dotted black curve represent $\ell_2$ and $r$, respectively. The green curve's lower branch begins to shrink for $a \geq 10.1388$. This is due to the fact that solutions of \eqref{3.2} are outside of $[r, \ell_2)$. The bifurcation diagrams also indicate the following results. \begin{figure}[ht] \begin{center} \includegraphics[width=0.45\textwidth]{fig7a} % PvsC_DSC_and_SAB_b_1_a_6_P} % \includegraphics[width=0.45\textwidth]{fig7b} % PvsC_DSC_and_SAB_b_1_a_10_P} % \end{center} \caption{$a = 6$, $b = 1$ (left), and $a = 10$, $b = 1$ (right)} \label{fig3.1} %\label{fig3.2} \end{figure} \begin{figure}[ht] \begin{center} \includegraphics[width=0.45\textwidth]{fig8a} % PvsC_DSC_and_SAB_b_1_a_11_P} % \includegraphics[width=0.45\textwidth]{fig8b} % PvsC_DSC_and_SAB_b_1_a_40_P} % \end{center} \caption{$a = 11$, $b = 1$ (left), and $a = 40$, $b = 1$ (right)} \label{fig3.3} % \label{fig3.4} \end{figure} \begin{figure}[ht] \begin{center} \includegraphics[width=0.45\textwidth]{fig9} % PvsC_DSC_and_SAB_b_1_a_100_P} % \end{center} \caption{$a = 100$, $b = 1$ }\label{fig3.5} \end{figure} \begin{theorem}\label{Thm3.3} For $b = 1$, if $a < a_4$ (for $a_4 \approx 5.0407$) then \eqref{1.23}--\eqref{1.25} and \eqref{1.20}--\eqref{1.22} have no positive solution for any $c \geq 0$. \end{theorem} \begin{theorem}\label{Thm3.4} If $b = 1$ then $c_0(a) \rightarrow c^*(a)$ as $a \rightarrow \infty$. Furthermore, $\rho \rightarrow \ell_2$ as $a \rightarrow \infty$ where $u(x)$ is a positive solution to \eqref{1.23}--\eqref{1.25} or \eqref{1.20}--\eqref{1.22} with $\|u\|_\infty = \rho$. \end{theorem} \subsection{Positive solutions of \eqref{1.17}--\eqref{1.19}} Again, we are particularly interested in the case when $b = 1$. Recalling Theorem \ref{Thm2.5}, we plot the level sets of \begin{equation} \label{3.1}\widetilde{G}(\rho) - \sqrt{2} = 0 \end{equation} Using our numerical root-finding algorithm in Mathematica to solve equation \eqref{3.1} and explicit formulas to calculate the unique $r = r(a,b,c)$ and $q = q(\rho)$ values, level sets were plotted within $[r, \ell_2) \times [0, c^*]$. The blue curve breaks into two components somewhere around $a = 4.39$, with the lower component vanishing for $a > 10.1387$. This is due to the fact that the $\rho$-values, which are solutions of \eqref{3.1}, are outside of $[r, \ell_2)$. These bifurcation diagrams also indicate the following results. \begin{figure}[ht] \begin{center} \includegraphics[width=0.45\textwidth]{fig10a} % PvsC_NBC_and_SAB_b_1_a_4_P} % \includegraphics[width=0.45\textwidth]{fig10b} % PvsC_NBC_and_SAB_b_1_a_4p4_P} % \end{center} \caption{$a = 4$, $b = 1$ (left), and $a = 4.4$, $b = 1$ (right)} \label{fig3.6} %\label{fig3.7} \end{figure} \begin{figure}[ht] \begin{center} \includegraphics[width=0.45\textwidth]{fig11a} % PvsC_NBC_and_SAB_b_1_a_6_P} % \includegraphics[width=0.45\textwidth]{fig11b} % PvsC_NBC_and_SAB_b_1_a_10_P} % \end{center} \caption{$a = 6$, $b = 1$ (left), and $a = 10$, $b = 1$ (right)} \label{fig3.8} %\label{fig3.9} \end{figure} \begin{figure}[ht] \begin{center} \includegraphics[width=0.45\textwidth]{fig12a} % PvsC_NBC_and_SAB_b_1_a_11_P} % \includegraphics[width=0.45\textwidth]{fig12b} % PvsC_NBC_and_SAB_b_1_a_40_P} % \end{center} \caption{$a = 11$, $b = 1$ (left), and $a = 40$, $b = 1$ (right)} \label{fig3.10} % \label{fig3.11} \end{figure} \begin{theorem}\label{Thm3.1} For $b = 1$, if $a < a_1$ (for $a_1 \approx 2.8324$) then \eqref{1.17}--\eqref{1.19} has no positive solution for any $c \geq 0$. \end{theorem} \begin{theorem}\label{Thm3.2} If $b = 1$ then $c_0(a) \rightarrow c^*(a)$ as $a \rightarrow \infty$. Furthermore, $\rho \rightarrow \ell_2$ as $a \rightarrow \infty$ where $u(x)$ is a positive solution to \eqref{1.17}--\eqref{1.19} with $\|u\|_\infty = \rho$. \end{theorem} \subsection{Structure of Positive solutions to \eqref{1.11}--\eqref{1.13}} Combining results from the three cases, \eqref{1.14}--\eqref{1.16}, \eqref{1.23}--\eqref{1.25}, and \eqref{1.17}--\eqref{1.19} while recalling that the \eqref{1.23}--\eqref{1.25} case represents two symmetric solutions, we are able to completely determine the structure of positive solutions to \eqref{1.11}--\eqref{1.13}. As before, we are primarily interested in the case when $b = 1$. Comparison of nonexistence Theorems \ref{Thm2.1}, \ref{Thm2.6}, and \ref{Thm2.3} from Section 3 yields the following nonexistence result for \eqref{1.11}--\eqref{1.13}. \begin{theorem}\label{Thm4.1} If $a \leq \min[\sqrt[3]{3 b^2}, \lambda_1] $ then \eqref{1.11}--\eqref{1.13} has no positive solution for any $c \geq 0$. \end{theorem} Moreover, our computational results for the case $b = 1$ provide the following nonexistence result. \begin{theorem}\label{Thm4.2} For $b = 1$, if $a < a_1$ (for $a_1 \approx 2.8324$) then \eqref{1.11}--\eqref{1.13} has no positive solution for any $c \geq 0$. \end{theorem} Also, our computations indicate the following existence results for $b = 1$. For what follows, \eqref{1.14}--\eqref{1.16} is depicted in yellow, \eqref{1.23}--\eqref{1.25} and \eqref{1.20}--\eqref{1.22} both in green, and \eqref{1.17}--\eqref{1.19} in blue. \begin{theorem}\label{Thm4.3} For $b = 1$, if $a \in [a_1, a_2)$ (for some $a_2 > a_1$) (for $a_2 \approx 4.39$) then there exists a $C_0 > 0$ such that if \begin{enumerate} \item[(1)] $0 \leq c < C_0$ then \eqref{1.11}--\eqref{1.13} has exactly 2 positive solutions. \item[(2)] $c = C_0$ then \eqref{1.11}--\eqref{1.13} has a unique positive solution. \item[(3)] $c > C_0$ then \eqref{1.11}--\eqref{1.13} has no positive solution. \end{enumerate} \end{theorem} A bifurcation diagram of the case when $b = 1$ and $a = 4$ is shown in Figure \ref{fig3.12}. \begin{figure}[ht] \begin{center} \includegraphics[width=0.5\textwidth]{fig13} % PvsC_ALL_b_1_a_4_P} % \end{center} \caption{$\rho$ vs $c$ for $a = 4$, $b = 1$} \label{fig3.12} \end{figure} \begin{theorem}\label{Thm4.4} For $b = 1$, if $a \in [a_2, a_3)$ (some $a_3 \in (4.4, 5)$) then there exist $C_i > 0$, $i = 0, 1, 2$, such that if \begin{enumerate} \item[(1)] $0 \leq c \leq C_2$ or $C_1 \leq c < C_0$ then \eqref{1.11}--\eqref{1.13} has exactly 2 positive solutions. \item[(2)] $C_2 < c < C_1$ or $c = C_0$ then \eqref{1.11}--\eqref{1.13} has a unique positive solution. \item[(3)] $c > C_0$ then \eqref{1.11}--\eqref{1.13} has no positive solution. \end{enumerate} \end{theorem} Figure \ref{fig3.13} illustrates Theorem \ref{Thm4.4}. \begin{figure}[ht] \begin{center} \includegraphics[width=0.5\textwidth]{fig14} % PvsC_ALL_b_1_a_4p4_P} % \end{center} \caption{$\rho$ vs $c$ for $a = 4.4$, $b = 1$} \label{fig3.13} \end{figure} \begin{theorem}\label{Thm4.5} For $b = 1$, if $a \in [a_3, a_4)$ (for $a_4 \approx 5.0407$) then there exist $C_i > 0$, $i = 0, 1$, such that if \begin{enumerate} \item[(1)] $0 \leq c \leq C_1$ then \eqref{1.11}--\eqref{1.13} has exactly 2 positive solutions. \item[(2)] $C_1 < c \leq C_0$ then \eqref{1.11}--\eqref{1.13} has a unique positive solution. \item[(3)] $c > C_0$ then \eqref{1.11}--\eqref{1.13} has no positive solution. \end{enumerate} \end{theorem} Theorem \ref{Thm4.5} is illustrated in Figure \ref{fig3.14}. \begin{figure}[ht] \begin{center} \includegraphics[width=0.5\textwidth]{fig15} % PvsC_ALL_b_1_a_5p03_P} % \end{center} \caption{$\rho$ vs $c$ for $a = 5.03$, $b = 1$} \label{fig3.14} \end{figure} \begin{theorem}\label{Thm4.6} For $b = 1$, if $a \in [a_4, a_5)$ (for $a_5 = \pi^2)$ then there exist $C_i > 0$, $i = 0, 1, 2$, such that if \begin{enumerate} \item[(1)] $0 \leq c \leq C_2$ then \eqref{1.11}--\eqref{1.13} has exactly 6 positive solutions. \item[(2)] $C_2 < c < C_1$ then \eqref{1.11}--\eqref{1.13} has exactly 5 positive solutions. \item[(3)] $c = C_1$ then \eqref{1.11}--\eqref{1.13} has exactly 3 positive solutions. \item[(4)] $C_1 < c \leq C_0$ then \eqref{1.11}--\eqref{1.13} has a unique positive solution. \item[(5)] $c > C_0$ then \eqref{1.11}--\eqref{1.13} has no positive solution. \end{enumerate} \end{theorem} Theorem \ref{Thm4.6} is depicted in Figure \ref{fig3.15}. \begin{figure}[ht] \begin{center} \includegraphics[width=0.6\textwidth]{fig16} % PvsC_ALL_b_1_a_6_P} % \end{center} \caption{$\rho$ vs $c$ for $a = 6$, $b = 1$} \label{fig3.15} \end{figure} \begin{theorem}\label{Thm4.7} For $b = 1$, if $a \in [a_5, a_6)$ (some $a_6 \in (10, 10.1388)$) then there exist $C_i > 0$, $i = 0, 1, 2, 3$, such that if \begin{enumerate} \item[(1)] $0 \leq c < C_3$ then \eqref{1.11}--\eqref{1.13} has exactly 8 positive solutions. \item[(2)] $c = C_3$ then \eqref{1.11}--\eqref{1.13} has exactly 7 positive solutions. \item[(3)] $C_3 < c \leq C_2$ then \eqref{1.11}--\eqref{1.13} has exactly 6 positive solutions. \item[(4)] $C_2 < c < C_1$ then \eqref{1.11}--\eqref{1.13} has exactly 5 positive solutions. \item[(5)] $c = C_1$ then \eqref{1.11}--\eqref{1.13} has exactly 3 positive solutions. \item[(6)] $C_1 < c \leq C_0$ then \eqref{1.11}--\eqref{1.13} has a unique positive solution. \item[(7)] $c > C_0$ then \eqref{1.11}--\eqref{1.13} has no positive solution. \end{enumerate} \end{theorem} Figure \ref{fig3.16} shows the bifurcation diagram for $a = 10$, $b = 1$ along with Figure \ref{fig3.17}, which gives two small cross sections of the diagram. \begin{figure}[ht] \begin{center} \includegraphics[width=0.6\textwidth]{fig17} % PvsC_ALL_b_1_a_10_P} % \end{center} \caption{$\rho$ vs $c$ for $a = 10$, $b = 1$} \label{fig3.16} \end{figure} \begin{figure}[ht] \begin{center} \includegraphics[width=0.45\textwidth]{fig18a} % PvsC_ALL_b_1_a_10_ZM_P} % \includegraphics[width=0.45\textwidth]{fig18b} % PvsC_ALL_b_1_a_10_ZM_2_P} % \end{center} \caption{$\rho$ vs $c$ cross-sections for $a = 10$, $b = 1$} \label{fig3.17} \end{figure} \begin{theorem}\label{Thm4.8} For $b = 1$, if $a \in [a_6, a_7)$ (for $a_7 \approx 10.1388$) then there exist $C_i > 0$, $i = 0, 1, 2, 3$, such that if \begin{enumerate} \item[(1)] $0 \leq c \leq C_3$ then \eqref{1.11}--\eqref{1.13} has exactly 8 positive solutions. \item[(2)] $C_3 < c < C_2$ then \eqref{1.11}--\eqref{1.13} has exactly 7 positive solutions. \item[(3)] $c = C_2$ then \eqref{1.11}--\eqref{1.13} has exactly 6 positive solutions. \item[(4)] $C_2 < c < C_1$ then \eqref{1.11}--\eqref{1.13} has exactly 5 positive solutions. \item[(5)] $c = C_1$ then \eqref{1.11}--\eqref{1.13} has exactly 3 positive solutions. \item[(6)] $C_1 < c \leq C_0$ then \eqref{1.11}--\eqref{1.13} has a unique positive solution. \item[(7)] $c > C_0$ then \eqref{1.11}--\eqref{1.13} has no positive solution. \end{enumerate} \end{theorem} The bifurcation diagram for $a = 10.1, b = 1$ is depicted in Figures \ref{fig3.18} and \ref{fig3.19}. \begin{figure}[ht] \begin{center} \includegraphics[width=0.6\textwidth]{fig19} % PvsC_ALL_b_1_a_10p1_P} % \end{center} \caption{$\rho$ vs $c$ for $a = 10.1$, $b = 1$} \label{fig3.18} \end{figure} \begin{figure}[ht] \begin{center} \includegraphics[width=0.45\textwidth]{fig20a} % PvsC_ALL_b_1_a_10p1_ZM_P} % \includegraphics[width=0.45\textwidth]{fig20b} % PvsC_ALL_b_1_a_10p1_ZM_2_P} % \end{center} \caption{$\rho$ vs $c$ cross-sections for $a = 10.1$, $b = 1$} \label{fig3.19} \end{figure} \begin{theorem}\label{Thm4.9} For $b = 1$, if $a \in [a_7, a_8]$ (for $a_8=4\pi^2$) then there exist $C_i > 0$, $i = 0, 1, 2, 3$, such that if \begin{enumerate} \item[(1)] $0 \leq c < C_3$ or $C_2 \leq c < C_1$ then \eqref{1.11}--\eqref{1.13} has exactly 5 positive solutions. \item[(2)] $c = C_3$ then \eqref{1.11}--\eqref{1.13} has exactly 4 positive solutions. \item[(3)] $C_3 < c < C_2$ or $c = C_1$ then \eqref{1.11}--\eqref{1.13} has exactly 3 positive solutions. \item[(4)] $C_1 < c \leq C_0$ then \eqref{1.11}--\eqref{1.13} has a unique positive solution. \item[(5)] $c > C_0$ then \eqref{1.11}--\eqref{1.13} has no positive solution. \end{enumerate} \end{theorem} Figure \ref{fig3.20} shows the bifurcation diagram for $a = 11$, $b= 1$. \begin{figure}[ht] \begin{center} \includegraphics[width=0.6\textwidth]{fig21} % PvsC_ALL_b_1_a_11_P} % \end{center} \caption{$\rho$ vs $c$ for $a = 11$, $b = 1$} \label{fig3.20} \end{figure} \begin{theorem}\label{Thm4.10} For $b = 1$, if $a \in (a_8, \infty)$ then there exist $C_i > 0$, $i = 0, 1, 2, 3$, such that if \begin{enumerate} \item[(1)] $C_3 \leq c < C_2$ then \eqref{1.11}--\eqref{1.13} has exactly 5 positive solutions. \item[(2)] $0 \leq c < C_3$ or $c = C_2$ then \eqref{1.11}--\eqref{1.13} has exactly 4 positive solutions. \item[(3)] $C_2 < c \leq C_1$ then \eqref{1.11}--\eqref{1.13} has exactly 3 positive solutions. \item[(4)] $C_1 < c \leq C_0$ then \eqref{1.11}--\eqref{1.13} has a unique positive solution. \item[(5)] $c > C_0$ then \eqref{1.11}--\eqref{1.13} has no positive solution. \end{enumerate} \end{theorem} The bifurcation diagram for $a = 40, b = 1$ is shown in Figure \ref{fig3.21}. \begin{figure}[ht] \begin{center} \includegraphics[width=0.6\textwidth]{fig22} % PvsC_ALL_b_1_a_40_P} % \end{center} \caption{$\rho$ vs $c$ for $a = 40$, $b = 1$} \label{fig3.21} \end{figure} \begin{thebibliography}{00} \bibitem{Cantrell2002g31} Robert~Stephen Cantrell and Chris Cosner. \newblock Conditional persistence in logistic models via nonlinear diffusion. \newblock {\em Proceedings of the Royal Society of Edinburgh}, 132A:267--281, 2002. \bibitem{Cantrell2003gsb} Robert~Stephen Cantrell and Chris Cosner. \newblock {\em Spatial Ecology via Reaction-Diffusion Equations}. \newblock Mathematical and Computational Biology. Wiley, 2003. \bibitem{Cantrell2007g33} Robert~Stephen Cantrell and Chris Cosner. \newblock Density dependent behavior at habitat boundaries and the allee effect. \newblock {\em Bulletin of Mathematical Biology}, 69:2339--2360, 2007. \bibitem{Collins2004g27} A.~Collins, M.~Gilliland, C.~Henderson, S.~Koone, L.~McFerrin, and E.~K. Wampler. \newblock Population models with diffusion and constant yield harvesting. \newblock {\em Rose-Hulman Institute of Technology Undergradate Math Journal}, 5(2), 2004. \bibitem{goddard2009gM3} Jerome Goddard~II, Eun~Kyoung Lee, and R.~Shivaji. \newblock Diffusive logistic equation with non-linear boundary conditions. \newblock {\em Submitted.}, 2009. \bibitem{Goddard2009gM1} Jerome Goddard~II and R.~Shivaji. \newblock A population model with nonlinear boundary conditions and constant yield harvesting. \newblock {\em Submitted.}, 2009. \bibitem{Ladner2005g29} Tammy Ladner, Anna Little, Ken Marks, and Amber Russell. \newblock Positive solutions to a diffusive logistic equation with constant yield harvesting. \newblock {\em Rose-Hulman Institute of Technology Undergraduate Math Journal}, 6(1), 2005. \bibitem{Laetsch1970g23} Theodore Laetsch. \newblock The number of solutions of a nonlinear two point boundary-value problem. \newblock {\em Indiana University Mathematics Journal}, 20(1):1--13, 1970. \bibitem{Oruganti2002g10} Shobha Oruganti, Junping Shi, and R.~Shivaji. \newblock Diffusive logistic equation with constant yield harvesting. \newblock {\em Transactions of the American Mathematical Society}, 354(9):3601--3619, 2002. \end{thebibliography} \end{document}