optimize geometry

[tara95]

hi
please  explain to me that,
Is it necessary  to optimize the k-points and cut-off energy parameters before optimize geometry?

[Anders Blom]

Generally yes, and most often the optimization will converge faster if you have sufficient accuracy, because the forces will be more accurate, and the forces are actually more sensitive to settings like k-points than the total energy is. However, sometimes it can save some time to use slightly lower accuracy settings for the first part of the optimization, maybe converge to higher force tolerance like 0.05 eV/A, and then use finer settings to converge the last bit to 0.01 eV/A or so. But the safest option is to ensure good accuracy first, and then run the optimization with that.

[tara95]

thank you
Another question;
Which basis set(HGH-FHI-OMX) is more suitable if I use GGA for optimize geometry? And should optimized parameters be used to band structures?

[Jess Wellendorff]

The choice of pseudopotentials+basis set depends in general on the system under investigation. So it's often a matter of testing. Generally speaking, OMX will be better than FHI, but also computationally more demanding. And yes, once you chose a set of parameters, stick to those for subsequent band structure calculation also.

[Anders Blom]

OMX is harder, you need a considerably larger mesh cut-off for those, but I agree they are often (not always) of better quality than FHI.
I would however recommend SG15 as a general-purpose choice.

[tara95]

thank you so much

[tara95]

Hi
This warning  is given in the calculation of the I-V curve device
( the computed multigrid is greater than the required accuracy.
computed residual : 1.07837e-09
Required accuracy  : 1.00000e-12)
 
I do not know what parameters need to be changed now.

[Daniele Stradi]

Hi,
you should increase the density_mesh_cutoff until the error disappears.
Regards,
Daniele.

[tara95]

thank you
Another question
In a paper, source and drain regions are heavily doped to be n-type with the density of 1.5×10^13/cm2 for  InSe monolayer,but unit in ATK  is e/atom or e/cm3.Now how do I apply it to the software.

[Petr Khomyakov]

You have to convert the doping concentration of 1.5x10^13 per cm^2 to that per atom in a unit cell, and use the obtained number to set the doping concentration in e/atom in the ATK. 

[tara95]

thank you
please explain how to change unit cell of InSe to build armchair and zigzag inse  nanoribbon.

[Ulrik G. Vej-Hansen]

The nanoribbon plugin in the Builder allows you to change the elements, so that would be my suggested way of doing it.