We introduce a novel method to optimize field decomposition for a mode matching technique. Using our method, expanded mode numbers can be minimized to achieve the desired digits of computational accuracy.