Topology issues
In order to re-enable the automatic topology creation with MatID, we should improve our testing and fix any outstanding issues. The goal is to have a version running that fairly confidently can identify 2D materials and surfaces in the entries - focus should be on minimizing the amount of false positives.
-
Fix the issue with cluster merging. Currently, MatID is too eager in merging two clusters together: the merging threshold should be increased or an additional check for the amount of shared species between the merged clusters should be checked. -
Fix issue with remainder handling when two clusters are merged. -
Ensure that the reported conventional cells are physically sensible: -
Cell should have the correct dimensionality -
Atoms in cell should not "overlap"
-
-
Add check that prevents clusters with "complicated" cells from being reported. These are more likely to be false-positives. -
Add check for the material id: only if it is one of the top 50k most occurring materials, it is accepted. This will dramatically reduce false-positives. -
Fixed issue with "dangling"/non-bonded atoms sometimes being left as part of a region. -
Fix issue when processing unwrapped systems (=coordinates well outside cell in system with full pbc)