nexus tile may duplicate atom position

[Paul-St-Young]

I have attached an example where nexus.Structure::tile produced a supercell with two atoms on top of each other. @jtkrogel do you see an easy fix for this?

I found a workaround by adding "use_ase" flag to tile(). May I PR this as a feature?

diff --git a/nexus/lib/structure.py b/nexus/lib/structure.py
index a374ca728..9f2d6143d 100755
--- a/nexus/lib/structure.py
+++ b/nexus/lib/structure.py
@@ -2947,6 +2947,7 @@ class Structure(Sobj):
 
 
     def tile(self,*td,**kwargs):
+        use_ase            = kwargs.pop('use_ase',False)
         in_place           = kwargs.pop('in_place',False)
         magnetic_order     = kwargs.pop('magnetic_order',None)
         magnetic_primitive = kwargs.pop('magnetic_primitive',True)
@@ -2979,8 +2980,18 @@ class Structure(Sobj):
 
         self.recenter()
 
+        # get atomic positions "pos"
+        if use_ase:
+          from ase import Atoms
+          from ase.build import make_supercell
+          atext = ''.join(self.elem)
+          pbc = [True if b == 'p' else False for b in self.bconds]
+          a0 = Atoms(atext, cell=self.axes, positions=self.pos, pbc=pbc)
+          a1 = make_supercell(a0, tilematrix)
+          pos = a1.get_positions()
+        else:
+          pos  = self.tile_points(self.pos,self.axes,tilematrix,tilevector)
         elem = array(ncells*list(self.elem))
-        pos  = self.tile_points(self.pos,self.axes,tilematrix,tilevector)
         axes = dot(tilematrix,self.axes)
 
         center   = axes.sum(0)/2

tile.zip

[prckent]

Better to fix the bug in tile. This should never happen.

[jtkrogel]

I'm looking at this now. Apparently performing the shear operations to obtain the diagonal tiling can depend on the order in which you perform them. I will explore whether an alternative based on row reduction and pivoting would do the right thing.

[jtkrogel]

@Paul-St-Young, please see #1432. Can you confirm that the fix works for you?

[Paul-St-Young]

Yes, I can confirm the problem is fixed. Please go ahead and close this issue once #1432 is merged.
Thanks!

[jtkrogel]

Resolved by #1432