Repository navigation
Geometry constraints and dummy atoms: make them discoverable #306
Description
From the 9.0.0 release test (Linux, 2026-10-12). Follow-up to #298 (dummy atoms work). The workflow is clunky:
- The dummy atom X in the periodic table needs a tooltip ("Dummy atom: no electrons, no basis; for constraints and symmetry").
- Constraints are reached from the Calculation Editor (Constraints button, only after choosing run type Geometry), which opens the Builder; they belong in the Builder itself (e.g. Tools › Geometry Constraints), usable at any time.
- Atom selection for a constraint takes Shift+click; Ctrl+click does nothing. Both should work, as elsewhere.
- After selecting atoms it is not clear that Constrain has to be pressed again in the Geometry Constraints dialog; the dialog should say what to do next and show the constraints made.
Activity
mattasplund commentedon Oct 12, 2026
When I had them write that part for me years ago, it was clunky, but I was happy to have the capabililty. I got used to it, and my labs explained it clearly enough that I had forgotten that it was tricky.
One additional sanity check that might be nice ...
When students select an angle or torsion constraint they sometimes pick them in the wrong order; say the center atom first and then the two end atoms. I don't know if the system could hint at selecting them in the correct order.
ohlincha commentedon Oct 12, 2026
More from the release test (Andy, 2026-10-12):
Constraints and run types. Constraints belong to the structure and are defined in the Builder at any time. A calculation uses them only if its run type can (geometry optimisation, GeoVib, scans). Otherwise they are kept with the structure but not used, and ECCE says so: the Calculation Editor shows e.g. "2 constraints defined: used only by geometry optimisations", Verify notes it, and the input writer leaves them out. Switching the run type to Geometry later applies them without setting them again.
Placing a dummy atom is hard. With X chosen in the Build panel, a single click in empty space does nothing (a single click only adds at an open valence, and NH3 has none); a double click places X at the clicked screen point but at an unrelated depth (here 4.5 Å behind the molecule). Needed: (a) select atoms → "Add dummy atom at their centre" (the centroid of NH3's three H lies on the C3 axis; also ring centroids, bond midpoints); (b) a double click places a new atom at the depth of the molecule's centre.
The Geometry Constraints dialog for NWChem does not look right — screenshot to follow.
ohlincha commentedon Oct 12, 2026
Update: placing X works as expected in an NWChem calculation's Builder (single click puts it where intended). The failed placement above was in an ECCE-QM calculation, and ECCE-QM does not support dummy atoms. Needed there: when the calculation's code cannot use dummy atoms, choosing X in the Builder says so ("ECCE-QM does not support dummy atoms") instead of silently doing nothing on a click and placing the atom at an odd depth on a double click. The centroid tool ("Add dummy atom at the centre of the selection") remains useful but is not urgent.
mattasplund commentedon Oct 12, 2026
The way I would naturally do this would be to treat it as an atom. So, make the N atom tetrahedral, and add an X atom as a terminal atom. This treats it in the most natural way.
NWChem: a fixed atom no longer makes bond/angle/torsion constraints v…
Geometry constraints are defined in the Builder, at any run type (#306)
Ctrl+click adds to the selection as Shift+click does (#306)
Dummy atoms: X attaches to an open site, says so where a code cannot …
Builder: Tools > Add Dummy Atom... places X from the selected atoms (#…
ohlincha commentedon Oct 13, 2026
In 9.0.1:
Known gap: an undo that re-creates atoms drops the constraints on them without a message.