confirmed the existence of this shift between the 100 K and room-temperature
structures, presumably governed by intermolecular interaction potentials.
There are limitations to automation of this approach including the following:
heavier atoms increase the problems associated with SCF convergence, so it is
generally applicable only to molecular organic crystal structures; unmodelled disorder often results in atom coordinates located between two alternative disorder sites,
with large displacement parameters which correspond to a minimum in the fit to the
crystallographic data, but do not correspond to an energy minimum for the system;
and finally modelled disorder requires manual intervention to consider each conformation in turn in order to validate.
5 Horizons: Analysis of Multiple Experiments
Understanding properties of and characterizing new materials will often rely on a
synthesis of data and observations from multiple experiments and theoretical calculations. In crystallographic analyses, use of restraints is one method for including
external information in a least squares fit.
An area which is gradually developing is combining multiple sources of data into
one model where multiple diffraction techniques have been employed – e.g. a
combined X-ray and neutron model can take advantage of the accurate unit cell
and heavy atom positions from X-ray diffraction data, while the neutron diffraction
data can determine the hydrogen atom parameters, the contrast between scattering of
elements with similar atomic number and magnetic ordering. This approach has been
successfully applied to analysis of powder diffraction experiments [50–52], in part
because instrument parameters are included in the refined model, making a multidata set refinement a simpler extension of existing software. For single-crystal
experiments, combined studies are possible using software packages such as Jana
[53], TOPAS-Academic [54] and GSAS-II [55] and have been used to determine the
distribution of neighbouring metal ions in high pressure cells, where neither technique alone provided sufficient information [56].
There is a tension in this combined approach, which hinders its routine application, because X-rays and neutrons probe different scattering density in the crystal
structure. As well as having different magnitudes, the two densities do not necessarily have coincident extrema (hydrogen has a negative scattering length, so in this
discussion, the extreme values have opposite signs: a maximum for X-ray scattering
density and a minimum for neutron scattering density). For chemical crystallography, large differences of between about 0.09 Å and 0.16 Å occur between the
positions of the extrema of scattering density of the hydrogen atom [2]. The extrema
correspond to the mean of the distribution of the nuclear position in a neutron
experiment, and the average of the electron positions in an X-ray experiment,
which may be offset from the nuclear position when the atom is bonded to another.
Powder diffraction tools have long incorporated multi-model fits to an experiment, in order to account for multiple material phases in a powder sample. In the
Recent Developments in the Refinement and Analysis of Crystal Structures
63
structures, presumably governed by intermolecular interaction potentials.
There are limitations to automation of this approach including the following:
heavier atoms increase the problems associated with SCF convergence, so it is
generally applicable only to molecular organic crystal structures; unmodelled disorder often results in atom coordinates located between two alternative disorder sites,
with large displacement parameters which correspond to a minimum in the fit to the
crystallographic data, but do not correspond to an energy minimum for the system;
and finally modelled disorder requires manual intervention to consider each conformation in turn in order to validate.
5 Horizons: Analysis of Multiple Experiments
Understanding properties of and characterizing new materials will often rely on a
synthesis of data and observations from multiple experiments and theoretical calculations. In crystallographic analyses, use of restraints is one method for including
external information in a least squares fit.
An area which is gradually developing is combining multiple sources of data into
one model where multiple diffraction techniques have been employed – e.g. a
combined X-ray and neutron model can take advantage of the accurate unit cell
and heavy atom positions from X-ray diffraction data, while the neutron diffraction
data can determine the hydrogen atom parameters, the contrast between scattering of
elements with similar atomic number and magnetic ordering. This approach has been
successfully applied to analysis of powder diffraction experiments [50–52], in part
because instrument parameters are included in the refined model, making a multidata set refinement a simpler extension of existing software. For single-crystal
experiments, combined studies are possible using software packages such as Jana
[53], TOPAS-Academic [54] and GSAS-II [55] and have been used to determine the
distribution of neighbouring metal ions in high pressure cells, where neither technique alone provided sufficient information [56].
There is a tension in this combined approach, which hinders its routine application, because X-rays and neutrons probe different scattering density in the crystal
structure. As well as having different magnitudes, the two densities do not necessarily have coincident extrema (hydrogen has a negative scattering length, so in this
discussion, the extreme values have opposite signs: a maximum for X-ray scattering
density and a minimum for neutron scattering density). For chemical crystallography, large differences of between about 0.09 Å and 0.16 Å occur between the
positions of the extrema of scattering density of the hydrogen atom [2]. The extrema
correspond to the mean of the distribution of the nuclear position in a neutron
experiment, and the average of the electron positions in an X-ray experiment,
which may be offset from the nuclear position when the atom is bonded to another.
Powder diffraction tools have long incorporated multi-model fits to an experiment, in order to account for multiple material phases in a powder sample. In the
Recent Developments in the Refinement and Analysis of Crystal Structures
63