Random Approaches to the Phase Problem

Abstract

A weakness of the original multisolution methods was that they relied on a very small base of initial phases so that some early phase indications were unreliable. If in the early stages there were a few phase relationships which held poorly then this could throw the phasing into confusion no matter what the starting phase set. Thus the pattern of phase relationships, for every starting phase set, could indicate that the phases of two particular reflexions differed by π when in fact they were almost equal. It was early realised that such difficulties could be obviated by having a much larger starting set and here we shall be examining various ways in which this goal was sought.