[Pw_forum] Query about restarting vc-relax

[Phillip English]

Dear users and developers,

 

I'm new to using PWscf, and I've been attempting to run some basic
calculations of TiO2 and its two predominant polymorphs, rutile and anatase.
I'm having no trouble running 'scf', 'relax', and 'vc-relax' calculations
for the rutile polymorph, as it finishes within one period of the time
allocated to me on our supercomputers. However, anatase takes a little bit
longer, so it is being cut off mid-calculation by the queueing system. From
what I understand, this is what the restart_mode='restart' flag is for, so I
have tried using that to restart my calculation.

 

Before I restarted the calculation, the energy was converging nicely towards
a value I expected. However as soon as I restarted, the energy blew out.
Looking at the cell parameters, they too blew out, and visualising the
structure showed that the cell was indeed expanding quite rapidly. When I
looked at the output more closely, it seemed like the starting cell
parameters for the restarted calculation were still being calculated by the
program according to the specified 'ibrav' and 'celldm' values, rather than
restarting with the cell parameters that had relaxed from the previous
calculation. This is what I supposed caused the 'explosion'. It seems
strange to me that while the atomic positions are read in correctly, the
cell parameters are not.

 

So, my question after that long preamble; do you need to manually re-enter
the cell parameters using the 'CELL_PARAMETERS' flag when restarting a
vc-relax calculation? If this isn't the case, does anyone have any ideas
about what might have caused the behaviour described.

 

Apologies if this has been covered before, I used the search function and a
google search on the mailing list archives to try and find the answer but
couldn't.

 

Cheers,

 

Phillip English

 

PhD Student

Curtin University, Perth

Western Australia

Australia

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20080417/4ed78bba/attachment.html>