POSCAR: Import-Export Issues

[markmick]

When I import the attached POSCAR-before file, the hexagonal unit cell immediately changes into another unit cell, that is then included when I Export the POSCAR.  I want to keep the original unit cell intact during Export. 

Is there an option to keep VNL from changing the unit cell?  The POSCAR-before file has the correct cell, and I want to keep that one, but I don't understand why VNL is changing it.

Thanks,
Mark

[Petr Khomyakov]

Hi Mark,

VNL has not changed your unit cell, i.e., it is still hexagonal. The lattice vectors are also exactly the same as you defined them originally. In the POSCAR_after, VNL has just rotated the original Cartesian coordinate system about the c-axis. Since the atom positions are given with respect to the lattice vectors (i.e., in 'Direct'), this rotation had no effect on the atom coordinates given in the POSCAR file.

Note that you may also see this rotation as a rotation of the whole lattice and its lattice vectors (defined in POSCAR_before) about the c-axis.

Regards,
Petr

[Anders Blom]

A bug fix was issued about a month ago, so if you have slightly old version of the VASP plugin, you may have some issues. Please ensure you run version 1.13.

[markmick]

I have 2015.1 and still experiencing this issue.  Although the modified unit cell is technically correct, it crashes on VASP.  The "before" version runs perfectly fine on VASP, while the "after" version never converges, even after a 4-hour run.  How do I prevent VNL from changing the axis?

Mark

[Anders Blom]

I understand you have 2015.1, but you don't have the latest plugin for VASP files. Open Help>AddOn Manager and go down to the VASPPlugins line. You will see there is an update available (ver 1.13). Update to that, restart VNL, and the cell will now be respected when importing.

[Petr Khomyakov]

Hi Mark,

I have imported your POSCAR_before into VNL, and then exported the corresponding structure to POSCAR. VNL does export it in the original format. So, this is a plugin version issue indeed. I assume that you did not manually change the lattice type from 'Unit Cell' to 'Hexagonal' in 'Lattice Parameters' after importing POSCAR_before. This is when VNL would convert it in accordance with its internal convention for a hexagonal lattice.

For your information, VASP should work without any problem for the hexagonal lattice vectors as defined in POSCAR_after, at least according to my experience. In fact, this definition is considered as conventional for hexagonal lattices, see Setyawan&Curtarolo, Computational Materials Science 49 (2010) 299–312.

Regards,
Petr

[Anders Blom]

Correct, however there was a real bug in 1.12 where it converted the lattice but forgot to correctly convert the atom positions, so it may be that the atomic positions in your POSCAR_after are not correct if you used the older plugin.

[Petr Khomyakov]

The POSCAR_after (generated presumably with an older plugin) enclosed in the original post has correct atom position coordinates, which are identical to that in the POSCAR_before file as it should be in the 'Direct' (lattice vectors) coordinate system, no matter how it is rotated. This might still have been an issue if the original coordinates were given in 'Cartesian'. So, using a bug-free plugin (1.15) is the best option to avoid any import-export issues indeed.   
VASP convergence problem is another issue, which I believe has nothing to do with the VASP plugin in VNL, at least in this particular case.