An extension of the method of quasilinearization (Q1333352)

Let \(\alpha_ 0\), \(\beta_ 0 \in C^ 1 [J,R]\) be such that \(\alpha_ 0 (t) \leq \beta_ 0 (t)\) on \(J=[0,T]\). Define the closed set \(\Omega = \{(t,u) : \alpha_ 0 (t) \leq u \leq \beta_ 0(t)\), \(t \in J\}\) and consider the initial value problem (1) \(u' = F(t,u)\), \(u(0) = u_ 0\), where \(F(t,u) = f(t,u) + g(t,u) \in C [\Omega, R_ +]\). The author proves the following Theorem. Assume that (A1) \(f,g\) are uniformly convex and concave in \(u\) for \(t \in J\) respectively and \(g_ u (t,u) \leq 0\) in \(\Omega\), (A2) \(\alpha_ 0\), \(\beta_ 0\) are coupled lower and upper solutions of (1) on \(J\), i.e., \(\alpha_ 0' \leq f (t,\alpha_ 0) + g(t, \beta_ 0)\), \(\alpha_ 0 (0) \leq u_ 0\), \(\beta_ 0' \geq f(t, \beta_ 0) + g(t, \alpha_ 0)\), \(\beta_ 0 (0) \geq u_ 0\). Then there exist monotone sequences \(\alpha_ n (t)\), \(\beta_ n (t)\) that converge uniformly to the unique solution of (1) on \(J\). Moreover, the convergence of these solutions is quadratic. This result extends the method of quasilinearization to a larger class of initial value problems.