The control program has two major functions: (1) to determine the overall direction of shift for each parameter, and (2) to determine the optimum magnitude (fraction of the shift) to be applied along the shift vector.
In an initial cycle of refinement, the overall gradient is obtained by combining the contributions from the crystallographic, stereochemical and any other terms and the direction of shift is obtained by the method of steepest descent. For second and subsequent cycles of refinement one can combine the overall gradient with the direction of the shift used in the previous cycle and determine the new shift direction by the conjugate-gradient technique. The control program also determines the optimal shift magnitude by searching along the shift direction as described previously.
A number of options exist for modifying the shift vector before the shift is applied. One can combine all or part of the structure into units within which all the atoms are treated identically (e.g. Table 3).
Table: Examples of parameter editing cards
It the atoms within an amino acid are combined, the temperature-factor shift applied to each atom will be the average of the individual shifts. Positional parameters can be treated as though the group were a rigid body. This is done by a least-squares fitting procedure. It is also possible to place upper and lower limits on the values of the thermal factors and the occupancies.