Hi all!
I am facing an issue in using the LocalPropertyDifference featurizer in Matminer. For some structures, I get an error
“ValueError: No Voronoi neighbours found for site - try increasing cutoff”
Now while LocalPropertyDifference has no cutoff parameter during initialization, it calls pymatgen’s local_env.py module where the default cutoff is 13 angstroms. I had to increase it to nearly 100 angstroms for it to find neighbors but my structure’s unitcell’s dimensions are less then 10 angstroms.
What could be the cause of this problem? Any help is appreciated.
TIA!
Yes, when I change pymatgen’s cutoff in local_env’s VoronoiNN class to a very large value (100 angstroms), it is able to find neighbors and compute the local property difference. However, the cutoff is way larger than the unitcell dimensions, which seems problematic.
Here’s the POSCAR:
Li P Sr
1.0000000000000000
4.3771194429999998 0.0000039350000000 0.0000000000000000
-2.1885613049999999 3.7906955210000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 8.0599966240000001
Li P Sr
2 2 2
Cartesian
0.0002341160000000 -0.0001352540000000 0.0000000000000000
0.0002047820000000 2.5270129059999999 4.0299983120000000
0.0002344720000000 -0.0001352980000000 4.0299983120000000
0.0002036330000000 2.5270134049999999 0.0000000000000000
2.1887762049999999 1.2634425069999999 2.0149991570000001
2.1887762049999999 1.2634425069999999 6.0449974670000000
FYI - I’ve been having this issue with a structure I’m working with - I was able to resolve it with a small perturbation to the structure (structure.perturb(0.001)), which didn’t affect my use case. It seems like it might be an issue in the underlying Voronoi routine.
Hello! I also met this error: “ValueError: No Voronoi neighbours found for site - try increasing cutoff”. Will a small perturbation to the structure influence the strucutre? I am stiil not understand the what exactly does this step do here. Can you give me some Pointers on this, can the cutoff be arbitrarily large, or how do I solve this problem? I am really appreciate for this. Thanks!