Virtual-source approach, a design method for continuous-surface-relief (CSR) fan-out gratings, is described in detail. A local-search-based genetic algorithm that combines a canonical genetic algorithm (GA) with a local search method is used to the optimization design of CSR fan-out gratings. The proposed algorithm incorporates advantages of both GA and local search techniques. It can overcome the premature convergence of a canonical GA and shows a stronger ability to locate the global minimum. Numerical results demonstrate that CSR fan-out gratings with high diffraction efficiency and good uniformity can be obtained by using the algorithm proposed hereinbefore.