A novel equation for array beam pattern synthesis is presented. The expected pattern total power is set to a fixed value, under this constraint, the difference between the main beam total power and the other part total power is maximized to achieve the array pattern synthesis. To solve the proposed pattern synthesis equation, which cannot be solved by any published traditional method, based on the Lagrange multiplier method, it is transformed into a new equation in which the array element excitation vector is the eigenvector of a matrix. Thus, the array element excitation vector can be obtained by matrix eigenvalue decomposition. Three array architectures using the method are taken as examples to show its advantages. Simulation results of the examples show that the method has better performance than other methods. The method is a noniterative approach, so it requires less computational volume to complete the pattern synthesis process. In addition, through eigenvalue decomposition, multiple eigenvectors can provide other different solutions of array elements current excitations, which can be applied in different areas.