Structures of copper(II) and manganese(II) di(hydrogen malonate) dihydrate; effects of intensity profile truncation and background modelling on structure models
Ath. Lenstra et On. Kataeva, Structures of copper(II) and manganese(II) di(hydrogen malonate) dihydrate; effects of intensity profile truncation and background modelling on structure models, ACT CRYST B, 57, 2001, pp. 497-506
The crystal structures of the title compounds were determined with net inte
nsities I derived via the background-peak-background procedure. Least-squar
es optimizations reveal differences between the low-order (0 < s < 0.7 Angs
trom (-1)) and high-order (0.7 < s < 1.0 Angstrom (-1)) structure models. T
he scale factors indicate discrepancies of up to 10% between the low-order
and high-order reflection intensities. This observation is compound indepen
dent. It reflects the scan-angle-induced truncation error., because the app
lied scan angle (0.8 + 2.0 tan theta)degrees underestimates the wavelength
dispersion in the monochromated X-ray beam. The observed crystal structures
show pseudo-l-centred sublattices for three of its non-H atoms in the asym
metric unit. Our selection of observed intensities (I > 3 sigma) stresses t
hat pseudo-symmetry. Model refinements on individual data sets with (h + k
+ l) = 2n and (h + k + l) = 2n + 1 illustrate the lack of model robustness
caused bv that pseudosymmetry. To obtain a better balanced data set and thu
s a more robust structure we decided to exploit background modelling. We de
scribed the background intensities B( (H) over right arrow) with an 11th de
gree polynomial in theta. This function predicts the local background b at
each position (H) over right arrow and defines the counting statistical dis
tribution P(B), in which b serves as average and variance. The observation
R defines P(R). This leads to P(I) = P(R)IP(B) and thus I = R - b and sigma
(2)(I) = I So that the error sigma (l) is background independent. Within t
his framework we reanalysed the structure of the copper(II) derivative. Bac
kground modelling resulted in a structure model with an improved internal c
onsistency. At the same time the unweighted R value based on all observatio
ns decreased from 10.6 to 8.4%. A redetermination of the structure at 120 K
concluded the analysis.