A quasilinearization approach for two point nonlinear boundary value problems on time scales (Q2566475)

The authors consider the separated boundary value problems (SBVP) \[ \begin{gathered} -(p(t)x^{\triangle})^{\triangle} + q(t)x^{\sigma} = f(t,x^\sigma)+g(t,x^{\sigma}), \quad t \in [a,b]^{\kappa^2}, \tag{e1}\\ x(a) = A, x(b) = B, \tag{e2} \end{gathered} \] and \[ \begin{gathered} -(p(t)x^{\triangle})^{\nabla} + q(t)x = f(t,x)+g(t,x), \quad t \in [a,b], \tag{e3}\\ x(\rho(a)) = A, x(\sigma(b)) = B. \tag{e4} \end{gathered} \] Following the introductory section, in Section 1, several background theorems from calculus on time scales are presented. Section 2 provides the method of upper and lower solutions. In Section 3, the method of quazilinearization is applied to the above boundary value problems under the assumptions that \(f\) and \(g\) are \(k\)-hyperconvex and \(k\)-hyperconcave functions, respectively. An application of the method of quazilinearization yields the following existence results: Theorem. Assume that \(\alpha_0\) and \(\beta_0\) are, respectively, lower and upper solutions of the SBVP (1), (2) on \([a,b]\) and \(f,g \in C^{4}([a,b]^{\kappa^2} \times [\alpha_0,\beta_0]_1)\), where \([u,v]_1 = \{ w \in D_1: u \leq w \leq v\}\) with \[ \begin{aligned} D_1=\{x \in C([a,b]): \quad &x^{\Delta} \text{ is continuous and }px^{\Delta} \text{ is delta-differentiable on }[a,b]^{\kappa} \text{ and} \\ &(px^{\Delta})^{\Delta} \text{ is rd-continuous on }[a,b]^{\kappa^2}\}. \end{aligned} \] If (i) \(f\) is \(3\)-hyperconvex such that \(f^{(1)}(t,x) < 0\) and \(f^{(i)}(t,x) \leq 0\) on \([a,b]^{\kappa^2} \times [\alpha_0,\beta_0]_1\) for \(i =2,3\), and (ii) \(g\) is \(k\)-hyperconcave such that \(g^{(i)}(t,x) \leq 0\) on \([a,b]^{\kappa^2} \times [\alpha_0,\beta_0]_1\) for \(1 \leq i \leq 3,\) then there exist monotone sequences \(\{\alpha_n\}\) and \(\{\beta_n\}\) converging uniformly in \([\alpha_0,\beta_0]_1\) on \([a,b]\) to the unique solution \(u\) of the SBVP (1), (2). The rate of convergence of the sequences \(\{\alpha_n\}\) and \(\{\beta_n\}\) is cubic. If \(g=0\), then the rate of convergence is \(4\). Generalizations of the above result on rate \(k\) are given without proof. Some existence results for the SBVP (3), (4) are also provided.