3/4: Options for σ(Fo2) in SADABS
In SADABS, intensity standard uncertainties [i.e., su or σ(Fo2)] are assigned as follows:
where the σ(I) are from SAINT and account for counting statistics and background etc. SADABS provides twelve ways to specify K and g, as follows:
[1] K = 1, refine overall g
[2] K = 1, refine all g
[3] refine overall K and overall g
[4] refine overall K and all g
[5] refine all K and overall g
[6] refine all K and all g
[7] refine overall K, input fixed g
[8] refine all K, input fixed g
[9] input fixed K, refine overall g
[10] input fixed K, refine all g
[11] input fixed K and g
The default is option [5], which as we have seen, sometimes results in SHELXL weights with larger than expected b values. The SADABS *.abs file lists assigned (or refined) values of K and g. The K parameter is a scale factor applied to σ(I), so it is similar in scope to sm in the (expanded) HKLF command. Refinement of K is not optimal in all cases, as it can refine to too small of a value. Options [9], [10], and [11], however, allow input of a fixed K. For a reasonable first guess at K, we could try the sm value obtained on the HKLF test above. What then for parameter 'g'? The g value is simply the reciprocal of the limiting value of I/σ(I) for a reflection of infinite intensity. In visual terms, it sets the upper limit of the Diederichs plot (Diederichs, 2010). Thus, a default SADABS run is intended to give a value for g that is optimal for the dataset at hand. At the very least, it's the best (only!) estimate we have. Attempts to refine g (options [9] or [10]) would alter the limiting I/σ(I), which without good reason, is probably a bad idea (note: this is also the reason why we should not use the sm parameter on the HKLF line to re-scale the σ(Fo2) values). This all points to option [11] as the appropriate choice. This is again best illustrated by example, so on to Part 4.