The Kramers-Kronig method finds a pair of *f*'(*E*) and *f*''(*E*) functions
which are a self-consistent Kramers-Kronig transform pair, and which also
agree with the data. For centrosymmetric systems with crystallographic
weights, , this method uses the intensity fitting function,

where and , and where the energy dependent instrument
correction function .

To obtain starting values for , , *b*, and *m*, *f*' and
*f*'' are estimated using Cromer-Liberman and values, and
Eqn. 15 is fit to the data. Then the functions *f*' and *f*''
are obtained by using an iterative procedure: Equating the fitting function
to the data, the resulting quadratic equation is solved for *f*'. This *f*'
is then Kramers-Kronig transformed to obtain a corresponding *f*''. With
these new values for *f*' and *f*'', the data is refit to obtain new values
of , , *etc*., and in turn a new value for *f*'. This
process usually converges after three or four iterations.

The iterative Kramers-Kronig method has several advantages. First, little
knowledge of the crystal structure is required to obtain the *f*' and *f*''
functions. Second, the iterative method properly accounts for the
diffracted intensity's dependence on the real and imaginary parts of the
anomalous amplitude, and also properly accounts for its dependence on the
square of the fine structure. Note that the direct method for isolating
the fine structure, presented in the previous subsection, neglected the
square of the fine structure.