Key points are not available for this paper at this time.
An extension to the peaklist optimization procedure is proposed, in which one overall phase refinement cycle consists of tangent expansion, E-map, peaksearch and elimination of peaks to achieve a maximum correlation coefficient between E o and E c . This procedure appears to be able to solve large structures from random phases given data to atomic resolution. The power of the method can be substantially increased by starting with slightly better than random phases, obtained for example from threefold Patterson vector superposition minimum functions or rotation searches using a fragment of known geometry. These two sources of phase information require expansion of the data to the space group P1, which also appears to be a useful strategy when starting from random phases. This real/reciprocal space recycling procedure was successful in solving two small known proteins and three unknown 200+-atom small-molecule structures. An investigation of the influence of the resolution on the peaklist optimization algorithm shows that there is a marked deterioration in the effectiveness as the resolution of the data is truncated, the deterioration being particularly marked between 1.2 and 1.3 A
Sheldrick et al. (Tue,) studied this question.