PDF - rigid bodies and correlated motion

danweber

Dear all,

how does one best combine rigid bodies and correlated motion?

I recently started using Topas v6 for PDF refinements. The structure is an very symmetric aromatic organic radical anion in a lattice of K+ cations. The degree of freedom of the organic can nicely be reduced by using a rigid body, which captures most features of the PDF pattern reasonably well (best R_wp about 18 %).

When only a delta1 correlated motion is being used (r-dep peak sharpening with 1/r dependence), the two first probability peaks in G(r) are overly high, and the fit for the rest of the pattern becomes bad, see:

When I additionally introduce a "pdf_for_pairs" for the first neighbors of the atoms within the ring of the anion radical, the pattern looks great for r > 2 Å, but the first two peaks are not captures, see image below. The first two peaks should be part of the pattern though, as they represent the distance of the atoms within the ring molecule.

Did anyone experience similar issues? What could be possible suggestions how to remedy this?

Best, Dan

johnsoevans

Others will know more about this than me, but my general answer is that this is hard!
pdf.inc has the PDFGui-type correlated motion correction that it looks like you're using. There are the various other r-dependent macros that Phil Chater put together and distributed. Most of these are probably only useful for relatively minor corrections.
With molecular systems the difficulty is separating intramolecular and intermolecular distances. In your "rigid" case the former will give very sharp peaks and the latter broad.
In cases where intermolecular separations are longer than intramolecular you might be able to fit different r ranges of the pattern simultaneously with an equated structural model (two separate xdds) and different beq's.
When inter and intramolecular separations overlap in r you'd have to use pdf_for_pairs.
Depending on your structure it might be possible to use a lower symmetry spacegroup to help you separate things. e.g. if a subgroup has two molecules in the asymmetric unit not one, you could have different site labels on the different molecules and separate the interactions with different pdf_for_pairs commands. e.g. have site labels C_a for your orginal molecule and C_b for the second molecule. You could write equations to equate positions of C_a and C_b to maintain the hgher symmetry.
I think pdf_only_eq_0 effectively tries to do something similar. I haven't used it in anger, but I think it will only work if your molecule is built from symmetry unique sites.
Another option would be to try a big-box approach where you have a supercell containing multiple molecules. If you allow their positions and rotations to distribute you can potentially describe the mixture of sharp and broad peaks simultaneously. Again, that's non-trivial!