Computing positive fixed-points of decreasing Hammerstein operators by relaxed iterations (Q5939756)

The authors study the iterative solution of abstract nonlinear operator equations of the type \[ u(x)= KNu(x),\quad x\in \Omega, \] where \(\Omega\) is a compact Hausdorff space, \(K\) is a linear completely continuous operator mapping the positive cone of the space of continuous real functions on \(\Omega\), into itself, and \(Nu(x):= 1/\sigma(x, u(x))\) is a Nemytskij operator with \(\sigma(\cdot,\cdot)\) being continuous and strictly positive, and -- with respect to the second argument -- being nondecreasing and sublinear. Examples are integral equations that arise in nuclear physics, in the theory of radiative transfer, and Chandrasekhars \(H\)-equation modelling some atmospheric heat transfer. The equation is solved numerically by so-called stationary underrelaxation of Picard and updated Picard methods, and of Jacobi and Gauss-Seidel methods. Existence of a unique positive solution and global convergence are proved in a framework of fixed-point approximations for positive decreasing operators in ordered Banach spaces. Moreover, the methods above are compared with Newton-type methods.