QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: Cyrille on November 3, 2015, 13:08

Title: Problem with OMX
Post by: Cyrille on November 3, 2015, 13:08

Dear QW experts

My goal is to do transport calculations with SOC (to calculate AMR) but before doing complicated calculations I have started by doing some simple tests on an infinite Iron wire. The standard DFT calculations work fine and the corresponding band-structure looks OK (with correct band splitting due to SOC)

The problems start when I do device calculations.

Since with the new atk version it seems possible to start a SOC calculation from a previous collinear calculation I have used OMX pseudopotential but for a collinear calculation:

IronBasis = OpenMXBasisSet(
    element=PeriodicTable.Iron,
    filename="openmx/pao/Fe6.0S.pao.zip",
    tier=0,
    pseudopotential=NormConservingPseudoPotential("normconserving/upf2/Fe_CA13S.upf.zip"),
    )
#----------------------------------------
# Exchange-Correlation
#----------------------------------------
exchange_correlation = LSDA.PZ

My system is made of L and R electrodes and a Device.
The L and R are identitcal and made of 3 wire atoms while the Device is made of 7 atoms.

When I run the calculation the standard DFT part (equivalent Bulk) runs smoothly but when the code enters into the true NEGF loop of the Device then everything goes bananas! At the end I get zero charge and zero transmission (and no crash..)

The same calculation with FHI pseudo+basis set runs fine..... I have the feeling that something is going wrong with OMX....

I have attached my input python file.

Thanks in advance to tell me what is going wrong

Cyrille

PS: I am using:
+------------------------------------------------------------------------------+
|                                                                              |
| Atomistix ToolKit 2015.rc1 [Build 40ab357]                                   |
|                                                                              |
+------------------------------------------------------------------------------+

Title: Re: Problem with OMX
Post by: Umberto Martinez on November 3, 2015, 13:39

Use the latest version of ATK 2015.0. http://quantumwise.com/products/download
rc1 is not suported and may contain bugs with respect to your problem.

Also, you may find this tutorial useful: http://docs.quantumwise.com/tutorials/spin_orbit_silicon.html

Title: Re: Problem with OMX
Post by: Cyrille on November 3, 2015, 13:48

OK thanks I will try the very last "stable" version and report if it works:-)

Cyrille

Title: Re: Problem with OMX
Post by: Cyrille on November 3, 2015, 13:57

Ups I get exactly the same problem.
I include my input python and log file

Cyrille

Title: Re: Problem with OMX
Post by: Cyrille on November 3, 2015, 15:36

By the way I could not find any example with SOC and transport..
I found example with non-colinear magnetism and transport or SOC in standard DFT but no examples of SOC+Transport.

My Fe wire is definitely behaving incorrectly. It seems to be related to the OMX pseudo+basis in the NEGF scf loop (even in the collinear calculation).

Cyrille

Title: Re: Problem with OMX
Post by: Jess Wellendorff on November 4, 2015, 16:40

Agree, it looks strange. Will need more than 5 mins to look into this...