cannot use "device from bulk"

[lknife]

Dear all,

I finished the optimization of the central region of a device as described in tutorial "Advanced device relaxation" http://docs.quantumwise.com/tutorials/device_relaxation/device_relaxation.html

In consideration of the time consuming, I used a relatively shorter central region for the optimization process. The original length of the central region is 20.0871 angstom which can be converted to device configuration using "device from bulk" plugin. However, after the optimization, the length of the central region reduced a little to 19.9862 angstrom and the "device from bulk" plugin cannot work on this optimized central region. I wonder if the minimum length for "Device from bulk" plugin is 20 angstrom so that it doesn't work on the optimized (shorter than 20) one.

Please see the attached files: before_opt.py and after_opt.py are the two configurations before and after the optimization. the file 1HB-TdB*.py is the .py file for the optimization calculation.

Can anyone here help me with this problem?
Thank you very much for your time and kind help!

[Jess Wellendorff]

You have most likely not sufficiently constrained the central region ends during the geometry optimization. Your central region is very short for an interface, so I would suggest to double the length so that you can constrain more atoms during the relaxation. Otherwise the symmetry of the lattice towards the electrodes is broken, and the Device From Bulk will not be able to attach electrodes.

[lknife]

Thank you very much for your reply! I used the BBR method to constrain the outermost molecular of the central region (each contains 6 atoms, keep the molecular rigid). I used the short central region since I want to save computational time. The optimization of this short central region took me about 1 day with 64 MPI processes. If the length of the central region increases, I am afraid that I cannot finish the optimization procedure within the time limit of the computer cluster (5 days). The trajectory of the calculation seems to have little help on the subsequent calculation. I once used the trajectory files to restart the calculation, but it did not help a lot.

[Jess Wellendorff]

Perhaps you can speed up the the central region relaxation by doing it in 2 steps:
1) use a coarse k-point grid and perhaps smaller density mesh cutoff, use force criterion 0.2 eV/Å,
2) start from pre-relaxed structure, use more k-points and restore the density mesh curoff and lower force tolerance.
Just a suggestion.

[lknife]

Thank you very much for your valuable advises!