CIF handling should be optimized (I)

brockdyer03 · January 29, 2026, 1:08am

Turns out Vesta is free and I was confusing it with the paid biomed software by the same name.

Anyways, the default for Vesta when you open up the file NaCl-Halite.cif (from Avogadro actually) is this view:

You can then go to Edit>Bonds... to open a dialog, where you can then select the bond and pick Do not search atoms beyond the boundary

If you want to display what Avogadro displays, e.g. only the sites which are not translationally equivalent via a +1 shift in the <100> family, you need to open up the Boundary menu and manually specify range of fractional coordinates

So it kinda seems like Vesta doesn’t even want to show you anything but the full unit cell with redundant atoms. I think that the default for Avogadro should be to show the full unit cell with redundant atoms right on file load, except for the case that it is a molecular crystal (not sure what the right heuristic is for that though), where it should show one molecular unit.

I’d then additionally say that maybe the input generators should display a warning when a user tries to generate an input with the translationally equivalent images?

Thomas · January 29, 2026, 2:01pm

Present day Avogadro 1.102.1 has the side panel display types with an entry Ball and Stick which, by tap on the later, opens its own sliders and radio buttons the program remembers for future use. By similar token, the entry Crystal Lattice could be amended with display options once set and kept until additional intervention by the user. Influenced by CCDC’s Mercury, entries here could be:

[ ] packing of a unit cell with
  [ ] atoms which fit
  [ ] atoms of molecules whose centroids fit
  [ ] atoms of molecules where any atoms fit
  [ ] atoms of molecules where all atoms fit
[ ] display of the asymmetric unit

while Avogadro would default to the representation of the whole molecule. Here, “whole molecule” might require a visual completion of the representation for instance if asymmetric unit and symmetry operator (e.g., centre of inversion, axis of rotation) yields a molecule as one would write into a .sdf file. This is what e.g., codcif2sdf of cod-tools (DebiChem package, GitHub repository) provides, (including bond orders, which are not a requirement in .cif files). If a molecules appears to reach beyond the borders of the parallelepiped (example COD7060476)

codcif2sdf “reconstructs” as few sane molecules as deemed necessary for the .sdf (for this particular model in P2_1/c, one expects only one molecule in the .sdf). Note there is some overlap of cod-tool’s codcif2sdf and complementary cif_molecule, too.

Within the suggested packing options, the four levels equally shall mutually be exclusive of each other. The preferred default is to include atoms of molecules whose centroids fit. Avogadro’s current “fill cell” is closer to cod-tools’ cif_fillcell (or Jmol’s load input.cif {1 1 1}) to consider atoms which fit at all the parallelepiped. Thus a subsequent search for bonds by Avogadro is affected, too.

I didn’t venture out cod-tools’s scripts and potential interdependence to account for. I don’t expect Avogadro 2.0 to fill each of these desiderata, too.

A note on / about sodium chloride: the default display with JSmol of the corresponding model equally displays the unit cell virtually partially filled:

which is the result of the database’s interface running a load cif/1/00/00/1000041.cif {1,1,1}; hide symmetry instead of a load cif/1/00/00/1000041.cif {1,1,1}; on the website’s console:

A note on Jmol: the different display of e.g. load input.cif; vs for instance a load input.cif {1 1 1} equally affects the generation of an input file for Gaussian (tools → Gaussian).