An Improved Differential Evolution (IDE) algorithm is proposed for optimization problems. With the novel mutant operation adopting sub-optimal individual, the convergence of Differential Evolution (DE) algorithm is accelerated without increasing the risk of premature. Five typical test functions are minimized using DE and IDE algorithms and the results show the superior performance of IDE algorithm. Furthermore, the algorithm is applied to pattern synthesis of two antenna arrays. Broad nulls are formed in radiation pattern of a linear array to suppress broad-band interferences. In a microstrip patch array, the sidelobe level of array is decreased about 12.9dB and the mainlobe can scan to the desired angle.