The number of options is quite large and some of them are suggested here.
The CFOM figure of merit is in practice an estimate of the probability that a trial solution is correct. If the highest CFOM does not correspond to the correct solution, subsequent CFOM's may be successful.
A special problem arises when several nearly equivalent CFOM's are available for different trials. Then the individual FOM's have to be carefully considered by the user. In particular the additional R(x,y) figure of merit Cascarano, Giacovazzo and Guagliardi, 1992 can also provide useful suggestions.
The magic integer procedure is an effective and simple way for starting a multisolution process. Second representation formula for triplet invariants proved much more powerful than first representation formula therefore it is used as a default for estimating triplets. In case of failure, the user can increase the number of symbols, or use the RANDOM procedure.
The use of 1- and 2- phase seminvariants are not default choices (they require additional computing time and often they are not necessary for the success of SIR92). If symmetry is rather high, several seminvariants can be reliably estimated: some of them will actively be used in phase expansion and refinement process, others will be used as figures of merit. Sometimes the additional estimation of one or few one-phase seminvariants is enough for solving crystal structures.
A check for possible pseudotranslational symmetry is always made by SIR92: if PSEUDO keyword is not activated then information is given but not used in the phasing process. If a pseudotranslational symmetry has been found by the normalization routine a new run of SIR92 may be started including the keyword PSEUDO.
Sometime a molecular fragment is located in the E-map, or it is available from other sources ( i.e.,Patterson,... ), which is too small for recovering the complete crystal structure via the usual Least-Squares - Fourier methods; then the keyword PARTIAL may be activated in the normalization routine.
Perhaps because of a faulty data collection strategy, weak reflections may not be included in diffraction data.This lack of information influences both the normalization process (scale and overall thermal factors are affected by systematic errors; the experimental E-distribution is often non-centric even when the crystal structure is centrosymmetric) and the estimation of 1- and 2-phase seminvariants and of triplet and quartet invariants: in particular, a reduced number of negative triplets (via p-10 formula) and of negative quartets is calculated. Success in the structure solution may be obtained if weak reflections are introduced in the data set.
Even though there are still a large number of options available three are quoted for special cases:
a) the value of NREF (number of reflections actively used in the phasing process) is fixed by SIR92. For some special structures the ratio "number of active triplets/NREF" is too small (less than 10). Larger values of NREF may improve the phasing procedure.
b) high (or low) resolution reflections may occasionally play a too important role in the first steps of convergence and divergence process. Fixing a thermal factor lower (or larger) than that provided by the normalization routine may successfully change the convergence and divergence process.
c) an