Fourier Refinement

Contraints
- Number of Peaks to Select: The default value is a function of the number of non-H atoms in the asymmetric unit, the presence of heavier atoms, the type of data, and the resolution of the data. Typically, the number of peaks requested for large structures is less than the number of atoms in these structures because of the likelihood that high thermal motion or disorder at some sites will decrease the number of atoms observable during early stages of the solution process. More details are given on the Parameters page.
- Minimum Interpeak Distance: This parameter specifies the minimum distance that will be permitted between any two atoms in the trial structure (or substructure). For "Basic" high-resolution data, this parameter has a default value of 1.0Å. For SIR and SAS substructure data, which are used at lower resolution, this value has a default of 3.0Å.
- Minimum Distance between symmetry-related peaks: In space groups where special positions occur, this parameter is used to eliminate peaks located near such positions. This feature is especially useful when it is known that atoms cannot occur at these sites (e.g. Se atoms in selenomethionine derivatives). For example, the default value of 3A, when applied to a structure in space group P222, will create a cylinder (with a diameter of 3A) of excluded volume about each of the rotation axes. Experience has shown that, without this constraint, there is a tendency for false minima to develop by building up density at the special positions.
- Number of special position peaks to keep: Enter an appropriate non-zero value here if it is known, a priori, that a certain number of atoms will be located on the special positions. The requested number of the largest peaks that volunteer at the special positions will then be retained.
Fourier Grid
- Fourier Grid Size: By default, this value is set to 1/3 of the maximum resolution of the data set. Coarser grids (as coarse as 2/3 of the maximum resolution) can often be used successfully for substructures. Significantly coarser grids greatly reduce the number of points at which the Fourier series must by calculated and, consequently, the running time of the SnB procedure can also be greatly reduced. However, the number and quality of correct sites obtained may not be as large in such cases. Therefore, if a coarse grid is used, it is advisable to perform a second, single-trial, run of SnB in which the peak output of the first job is used as a Model Structure.
Extra Cycles
- Perform extra cycles with more peaks?: In cases where a greatly reduced number of peaks are selected during most of the Shake-and-Bake cycles in order to increase the efficiency of the solution process, it is desirable to perform some additional cycles requesting that the number of peaks used be closer to the number of atoms expected. If a solution has already been obtained, the quality of this solution will improve considerably during the extra refinement cycles. In some cases, it has been essential to increase the number of correct peaks during the dual-space refinement before proceeding to Fourier refinement or the solution would otherwise be lost. The extra-cycle feature is automatically activated when the number of peaks selected during the preliminary cycles is less than 80% of the expected number of atoms.
- Number of Extra Cycles: By default, the number of extra cycles performed is equal to 10% of the normal (preliminary) cycles.
- Number of Peaks: By default, the number of peaks selected during the extra cycles is equal to the number of atoms expected.
Twice Baking (Fourier Refinement)
- Trials for E-Fourier Filtering: The number of reflections phased during the dual-space Shake-and-Bake cycles is a relatively small fraction of the total available (typically 1/6-1/8 for basic data). The number of reflections used in direct-method procedures is restricted because the invariant relationships are more reliable when they involve only those reflections with large |E| values. However, the severity of the resultant series termination reduces the quality of the electron-density maps. The quality of solutions can be increased by adding additional reflections and performing Fourier refinement since this type of refinement does not involve the invariants. This procedure has been termed "E-Fourier filtering" in the context of the program SHELXS. In the interest of computational efficiency, Fourier refinement is recommended only if the current trial structure is better (lower Rmin value) than the previous best trial structure. By default, Fourier refinement is not used for substructures since the reliability of the smaller magnitude differences is often suspect.
- Number of Cycles: By default, the number of Fourier refinement cycles performed (if any) is equal to 10% of the number of dual-space (Shake-and-Bake) cycles. The number of reflections phased in successive cycles is increased in equal increments until all the reflections that pass the minimum |E| cutoff (see below) are phased in the last cycle.
- Number of Peaks: It is reasonable to request more peaks here. As map quality improves, a larger number of peaks are reliable.
- Minimum |E|: Enter the minimum |E| value for reflections to be included for Fourier refinement. The minimum selected should give more reflections than the number used during the dual-space direct-method cycles, but it should not include weak reflections if they are unreliable.