ISHIFT=4 more neighbours in slab than bulk

[nedtaylor]

More neighbours found in the slab than in the bulk when using ISHIFT=4.

Likely an issue with unusually large compression, or angular mismatch.

Could this really be the case though? Isn't the bulk DON formed from a transformed version of the bulk, instead of the true bulk?

POSCARs:
up_term.vasp
POSCAR.mp-352_HfO2
POSCAR.mp-985829_HfS2

DONs:
full_broken_DON.dat zoom_broken_DON.dat

Input file:
param.in

Output:
out.out

[nedtaylor]

Possibly an issue relating to how the transform match is performed. We are currently applying the match to trying to match a with a for lower and upper lattice before they are transformed (but keeping c unchanged as it should be perpendicular to the strain). As such, it doesn't really work because only in a special case does a match onto a.

[nedtaylor]

No, I was wrong before. Around line 410 of src/interfaces.f90, we find that the transformed matching supercell is stretched and compressed (other than the new c axis as this will not be deformed). Then it is transformed back into its original primitive cell. In doing so, the deformations of the supercell should be mapped back onto the primitive cell.

As such, I am not sure of the issue. It is clear that the cells are different, as observation of the atom in question shows that all bonds have shrunk massively in the slab when compared to the primitive. But the cause of this I am as of yet unsure. There is not much of an angular mismatch, so I'd be surprised that that could cause it.
bulk_up_term.vasp

[nedtaylor]

On comparing the mapped POSCAR to the original, you see massive angular change. Original angles are 90 99 90 (degrees). Mapped angles are 98 91 84 (degrees).

[nedtaylor]

Reduce mbond_maxlen to 6 from 10 and you get it working. Seems that looking too far away somehow caused issues. Could be related to the number of supercells the DON generates based on dist_max... If so, need to fix that as it should be looking at the smallest distance by using sums of the lattice vectors, not just any one lattice vector.

[nedtaylor]

Should be fixed now by taking into account the obtuse lattice issue.

[nedtaylor]

Mostly fixed in 1.0.2 release

Related issue: hepplestone/artemis#14