We develop an optimization method for stable geometries of atomic clusters.It combines adaptive genetic algorithm,which has the ability of global optimization,with a local optimization method proposed in this paper which is based on the Newton method.Geometry optimization of carbon clusters is used to test the method.Compared with adaptive genetic algorithm,it is found that the new local optimization method can find local extremum effectively.Compared with hybrid genetic algorithm,it shows that the method can jump out of local extremes.The method presents good stability in four optimizations of C₁₂.