Shape optimization for radar cross sections by a gradient method

A gradient method for optimization of shape and materials for radar cross section (RCS) reduction is derived from Maxwell's equations. The method uses the adjoint problem and finds the derivatives of the RCS with respect to all design parameters from a single solution of the scattering problem. The method is tested in two-dimensional test problems to minimize the RCS in specified angular intervals at a single frequency, and finds configurations with strongly reduced RCS in a small number of iterations. In the absence of other constraints, the optimal shapes of perfectly electrically conducting (PEC) scatterers have sharp corners pointing in the directions where the RCS is minimized and shapes optimized at a single or small number of frequencies exhibit corrugations with about half the wavelength of the incident wave. The corrugations can be suppressed by means of penalty functions, and this gives only moderate increases of the RCS. After such regularization, the optimal shapes agree well with expectations from geometrical optics; they have almost flat surfaces whose normals lie outside the intervals chosen for RCS optimization, and sharp edges at locations where the surface normal points into the intervals in which the RCS is minimized. Shape optimization of PEC wing profiles aiming at both good aerodynamical properties and low RCS show give conflicting requirements on the shape. Copyright © 2004 John Wiley & Sons, Ltd.