Equações integrais via teoria de domínios: problemas direto e inverso

Abstract

We present a study in Domain Theory of integral equations of the form

f (x) = h(x)+g Z b(x) a(x) g(x, y, f (y))dy

for a0 ≤ a(x) ≤ b(x) ≤ b0 with h, a, b defined for x ∈ [a0,b0] and g defined for x, y ∈ [a0,b0], in which the right-hand side defines a contraction on the metric space of continuous realvalued functions on [a0,b0]. The starting point of this work is to revisit Interval Analysis in Domain Theory for the initial-value problem in ordinary differential equations where a solution is expressed as a fixed point of the Picard operator. With the set of real numbers interpreted as the interval domain, real-valued functions are extended to work in the space of interval-valued functions of the real variable domain. In particular, the vector field is extended in the second argument. Under these conditions, for the first time Fredholm and Volterra integral equations have solutions expressed as fixed points of a contraction mapping in terms of the splitting on interval-valued functions of the real variable domain. The measurement for interval-valued functions of the real variable domain is considered where we can asssess the convergence properties of the interval integral operator by means of the informatic derivative. The proposed techniques are applied to more general methods in ordinary differencial equations (ODEs) and partial differential equations (PDEs). For the first time, an algorithm is proposed to provide solutions to the inverse problem for Odinary Differential Equation where, given a function f , it is found a vector field v that defines a Picard operator which maps the solution f as close as possible to itself, such that the ODE f ′(x) = v(x, f (x)) admits f as either an exact or, as closely as desired, an approximate solution.