Electron diffraction
XDS can process electron diffraction data as long as the general preliminaries for XDS are met. In brief, this means the crystal is illuminated with a (quasi-)parallel, non-moving beam while the crystal rotates with a constant velocity about a fixed rotation axis, and the detector is composed of planar segments. These are quite general conditions, making XDS very suitable for the processing of electron diffraction data. XDS can even correct for optical distortions of the electromagnetic lens system of the electron source, especially when they are in-plane with the detector and do not change during data collection [1]. Compared with X-ray diffraction, the crystallographer should be far more flexible in the acceptance of quality indicators.
Processing (REFINE= keywords)
The very short wavelength results in a very small maximum scattering angle 2θmax of less than 5°. With such, sin(x)≈x, with the result of some strong correlations [2]between parameters. It is usually not possible to refine BEAM and POSITION simultaneously. A recommended setting for XDS.INP:
REFINE(IDXREF)=BEAM AXIS ORIENTATION CELL
REFINE(INTEGRATE)=BEAM AXIS ORIENTATION
REFINE(CORRECT)=BEAM AXIS ORIENTATION CELL
When resolution is sufficiently high, at about dmin < 0.8Å, both CELL and POSITION can be refined (but not the BEAM):
REFINE(CORRECT)=AXIS ORIENTATION CELL POSITION
In other cases, the INCIDENT_BEAM_CORRECTION and UNIT_CELL_CONSTANTS can be copied from XDS_ASCII.HKL to XDS.INP, and only the POSITION is refined.
REFINE(CORRECT)=AXIS ORIENTATION POSITION
(AXIS and ORIENTATION are usually very robust to be refined).
Scaling
Scaling works best with high multiplicity. With electron diffraction data, this is not always granted, because the rotation range is at best 150° and can be limited to much less. With low symmetry spacegroups, this can result in low multiplicity. On the other hand, electron diffraction data are typically collected from a multitude of crystals, and a subset of well-fitting ones are merged to increase data quality and multiplicity. In such cases, the parameter SNRC=50 [3] should be added to XDS.INP to switch off scaling for individual data sets and only scale when merging data from multiple crystals.
Other parameters for XSCALE are usually not needed. It turns out that the scaling model in XDS and XSCALE is well suited for ED data, too.