“…Note that this γ j is the operator appearing in the first-order absorbing boundary conditions used in [55]. Let us mention that the choice of b and of γ j may be more important when solving the problem with an iterative algorithm, using a sparse approximation as a preconditioner, in the spirit of [9]. In that case, as in domain decomposition methods [30], increasing the overlap b − a and using for γ j the approximation of the DtN map in normal incidence enable to speed up the convergence of iterative algorithms.…”