Discussion List Archives

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Ambiguity in atom_site.disorder_group value -1

On 20/10/2022 00:01, Robert Hanson via coreDMG wrote:
Maybe more to the point, here. My query is really not about documentation language. What I'm interested in is the way the -1 value should be interpreted. ...

What I would be interested in is some way in the CIF to be able to describe this "localness" of the disorder. That, for example, the results of symop 1 and symop 2 are paired. And, likewise, in this case I presented, symops paired as [1,2], [3,4], [5,6], and [7,8]. Wouldn't this be useful information?

Perhaps something like:

_local_disorder_assembly  # matches atom_site_disorder_assembly
_local_disorder_group  # matches atom_site_disorder_group
_local_disorder_assembly_symmetry_operation_set   #matches list of space_group_symop_id
1 A -1 1,2
2 A -1 3,4
3 A -1 5,6
4 A -1 7,8
Would that be a reasonable feature request?


I often find it difficult to understand all the ramifications of a discussion such as this purely in the abstract, so I looked for a real-world example to see how this is currently treated in the literature. The example I have found was published in Acta Cryst. E: https://journals.iucr.org/e/issues/2022/11/00/jy2022/

The ATOM_SITE loop for this structure appears in the CIF as













    Ni Ni1 0.500000 0.11950(2) 0.250000 0.01509(10) Uani d 1 . .

    N N1 0.54965(7) 0.11278(9) 0.10148(12) 0.0185(2) Uani d 1 . .

    C C1 0.57231(7) 0.08251(10) 0.01406(14) 0.0165(2) Uani d 1 . .

    S S1 0.60344(2) 0.03798(3) -0.11091(4) 0.01977(10) Uani d 1 . .

    N N11 0.58540(6) 0.24076(9) 0.38595(12) 0.0181(2) Uani d 1 . .

    C C11 0.62454(8) 0.22828(11) 0.53994(15) 0.0204(3) Uani d 1 . .

    H H11 0.607248 0.168907 0.584775 0.025 Uiso calc 1 . .

    C C12 0.68901(8) 0.29758(12) 0.63729(16) 0.0244(3) Uani d 1 . .

    C C13 0.71270(9) 0.38429(13) 0.56974(18) 0.0297(3) Uani d 1 . .

    H H13 0.756660 0.433386 0.631780 0.036 Uiso calc 1 . .

    C C14 0.67207(10) 0.39913(13) 0.41148(19) 0.0309(3) Uani d 1 . .

    H H14 0.687354 0.458893 0.363972 0.037 Uiso calc 1 . .

    C C15 0.60876(9) 0.32557(12) 0.32326(16) 0.0240(3) Uani d 1 . .

    H H15 0.581066 0.335802 0.214608 0.029 Uiso calc 1 . .

    C C16 0.73170(9) 0.27472(15) 0.80794(17) 0.0336(3) Uani d 1 . .

    H H16A 0.696992 0.226009 0.835783 0.050 Uiso calc 1 . .

    H H16B 0.741856 0.346916 0.864123 0.050 Uiso calc 1 . .

    H H16C 0.784063 0.236346 0.835127 0.050 Uiso calc 1 . .

    N N21 0.4958(3) 0.5343(3) 0.0380(4) 0.0525(8) Uani d 0.5 A -1

    C C21 0.4990(2) 0.5882(3) 0.1369(4) 0.0366(7) Uani d 0.5 A -1

    C C22 0.4981(19) 0.6557(4) 0.238(3) 0.059(3) Uani d 0.5 A -1

    H H22A 0.511446 0.613590 0.333381 0.089 Uiso calc 0.5 A -1

    H H22B 0.443345 0.689575 0.200183 0.089 Uiso calc 0.5 A -1

    H H22C 0.538683 0.716083 0.258413 0.089 Uiso calc 0.5 A -1


Figure 5 of the paper shows disordered acetonitrile molecules (for example in the middle of the unit cell as viewed), and is nicely reproduced by Jmol when all symmetry operations of the space group

_space_group_name_H-M_alt     'C 1 2/c 1'

_space_group_name_Hall     '-C 2yc'

are applied:




If one looks at the area where the disorder occurs (around the acetonitrile molecule) in Jmol, showing superimposition of all symmetry-generated copies, one gets the following view:



Consider the green and purple pair; they are related through an inversion point (which Jmol can render as the little yellow sphere):



Let me call this configuration “nose-to-nose” (as a fanciful description of the shape and orientation of the molecules in this view). Likewise the orange and turquoise pair are related by inversion (this I’ll call “shoulder-to-shoulder”):



while the orange and purple are related by a c-glide plane:



Let me call this “nose-to-shoulder”. Now, note that Figure 1 of the original publication suppresses some of the disorder, and shows a neater arrangement of the acetonitrile molecules:



At first I thought this was a selection of “nose-to-nose” arrangements, but along this axis that’s not easy to tell – in projection all the possible configurations show a similar shape. However, it demonstrates that authors may feel a need to select among the disordered possibilities.


So the questions that come to my mind are:


(1) How did the authors produce their “tidy” Fig. 1? They cite the software they have used as Computer programs: CrysAlis PRO (Rigaku OD (2021). CrysAlis PRO. Rigaku Oxford Diffraction), SHELXT2014/5 (Sheldrick, G. M. (2015a). Acta Cryst. A71, 3-8), SHELXL2016/6 (Sheldrick, G. M. (2015b). Acta Cryst. C71, 3-8), DIAMOND (Brandenburg, K. & Putz, H. (1999). DIAMOND. Crystal Impact GbR, Bonn, Germany) and publCIF (Westrip, S. P. (2010). J. Appl. Cryst. 43, 920-925). I suspect that the visualization software DIAMOND was what they used for the figure.


(2) Is the selected pairing purely a rendering choice? I.e. the author can perhaps use the visualization software to show any one or more of the eight possible locations of the symmetry-transformed atoms. This is what Jmol is currently doing, although its right-click menu allows only each individual symmetry operation to be rendered, and not  pairs or other combinations. (It does, though, allow the superposition of all the symmetry operations.)


I note that in this example, if you consider the specific “pocket” that I have singled out, it is populated by the imposition of four of those symmetry operations, but not by the other four. (I suppose this is implied by the site occupancy factor of 0.5 for these atoms.) So if I (as an author) wanted to show all the pockets populated by particular orientations of the acetonitrile molecule, I would choose the half of the symmetry operations that would combine to provide me with my preferred rendering.


If I am not a crystallographically sophisticated user (which I am not), I might find this confusing or difficult to show. Could Jmol (without additional guidance) offer a choice of renderings which would populate the extended structure views with sets of symmetry-generated but not overlapping mates? E.g. in my example below, I have selected the symmetry operations that Jmol labels internally as A (yellow in this figure), D (red), B (green) and F (white).


And if I wanted to import this view into, say, DIAMOND, then Jmol could write the specific choices into a table such as Bob suggests, so that DIAMOND (or Mercury) would reproduce that view. This is certainly a use case for selecting particular disorder configurations, although it’s for the benefit of rendering programs rather than a physico-chemical description of the crystal structure.



(3) … which leads me to my third question, aimed at the crystallographers on the list. Are there refinement strategies (in SHELX or other software) that allow you to constrain the local symmetries at different locations? I.e. in my terminology, if the pocket at the centre of the unit cell contains “nose-to-nose” molecules, can you constrain the neighbouring pockets to be “shoulder-to-shoulder”? Would you even want to be able to do this? Are there any experimental features in the diffraction images that would give you some clue as to the distribution of those orientations? I would guess that cases like this are fairly rare – energetically the different orientations within the pocket must be very similar, but perhaps there are weak non-bonding interactions that do bias particular combinations. Might every pocket contain a “nose-to-shoulder” arrangement?


If these structural arrangements are amenable to discovery (or enforcement in refinement), then perhaps Bob’s _local_disorder_... items do indeed have a place in the crystal structure description.



coreDMG mailing list

[Send comment to list secretary]
[Reply to list (subscribers only)]