The domain of convergence for the iterative solution of nonlinear second order boundary value problems (Q1078737)

The domain of convergence for the so-called Picard iteration scheme of nonlinear, two-point, second order boundary value problems is determined. The procedure followed replaces the boundary value problem by an equivalent system of integral equations which represents a fixed point problem in a suitably chosen function space. The iteration scheme carried over into this space is known to converge to the uniquely defined solution whenever the spectral radius of a certain comparison operator is less than one. The spectral radius of this operator can be determined by solving an eigenvalue problem which appears as a single integral equation. It turns out that on certain subintervals, the eigenelements of this equation satisfy linear differential equations. Putting together the solutions of these equations yields a family of continuously differentiable functions on the total interval, from which the eigenelements with their associated eigenvalues can be selected. For two well known examples, one with Dirichlet and the other with mixed Dirichlet-Neumann boundary conditions, the domain of convergence is determined as a function of the Lipschitz constants involved, and the results are compared with the best ones obtained so far.