Author Topic: Error running DDOS/PLDOS analysis on QuantumATK 2019.12  (Read 3026 times)

0 Members and 1 Guest are viewing this topic.

Offline athiyag

  • New QuantumATK user
  • *
  • Posts: 4
  • Country: us
  • Reputation: 0
    • View Profile
Hello everyone,

I am having an issue running DDOS/PLDOS calculations from converged hdf5 device simulations (analysis from file) on my University supercomputer cluster. The primary error message seems to be:

** Back Engine Exception : Not implemented
** Location of Exception : foldedselfenergycalculator.h:91

I remember running the same script a few days ago without any error (the same PLDOS analysis of a different .hdf5 file). Could you please point to possible sources of error or how I could rectify them?

I had encountered this issue earlier and rectified some of them by "centering" the device atoms and re-running them which is curious. However, this hack works only for some models and not for all of them. The error started to crop up again after a few more runs. This happens only with a few files.

Earlier posts in the forum say the error could be due to the fact that some atoms are too close to each other or overlapping. Is there a way to fix this issue or is it inherent in my model?

Thank you.

Offline Pieter Vancraeyveld

  • Regular QuantumATK user
  • **
  • Posts: 22
  • Country: dk
  • Reputation: 1
    • View Profile
Re: Error running DDOS/PLDOS analysis on QuantumATK 2019.12
« Reply #1 on: June 10, 2020, 08:56 »
Can you share the input scripts and log output of the device and PLDOS calculations?

Offline athiyag

  • New QuantumATK user
  • *
  • Posts: 4
  • Country: us
  • Reputation: 0
    • View Profile
Re: Error running DDOS/PLDOS analysis on QuantumATK 2019.12
« Reply #2 on: June 10, 2020, 18:38 »
Hi Pieter,

Please find attached.

Thank you

Offline Pieter Vancraeyveld

  • Regular QuantumATK user
  • **
  • Posts: 22
  • Country: dk
  • Reputation: 1
    • View Profile
Re: Error running DDOS/PLDOS analysis on QuantumATK 2019.12
« Reply #3 on: June 11, 2020, 13:25 »
Hi Aanand, You have run into the issue that DDOS currently does not support devices with transverse electrode repetitions. We will make sure to address this in the next release. In the meantime, you can run PLDOS using the LDOS method as follows:
Code
projected_local_density_of_states = ProjectedLocalDensityOfStates(
    configuration=configuration,
    method=LocalDeviceDensityOfStates,
    energies=numpy.linspace(-4, 3, 351)*eV,
    kpoints=kpoint_grid,
    contributions=All,
    self_energy_calculator=RecursionSelfEnergy(),
    energy_zero_parameter=AverageFermiLevel,
    infinitesimal=1e-06*eV,
    density_mesh_cutoff=75.0*Hartree,
    processes_per_energy=All,
    )
Does this solution work for you? Sorry for the inconvenience, Pieter

Offline athiyag

  • New QuantumATK user
  • *
  • Posts: 4
  • Country: us
  • Reputation: 0
    • View Profile
Re: Error running DDOS/PLDOS analysis on QuantumATK 2019.12
« Reply #4 on: June 15, 2020, 04:13 »
Hi Pieter,

Thank you for the solution. However, this worked only for one configuration and failed for three others. I have had this issue before (although the scripts are seemingly the same). I am attaching the input and log files for the failed run in this post and the same for the successful run in the next post.

Offline athiyag

  • New QuantumATK user
  • *
  • Posts: 4
  • Country: us
  • Reputation: 0
    • View Profile
Re: Error running DDOS/PLDOS analysis on QuantumATK 2019.12
« Reply #5 on: June 15, 2020, 04:47 »
Hi Pieter,

Sorry, for some reason, I am not able to attach the files of the successful run.

Offline Pieter Vancraeyveld

  • Regular QuantumATK user
  • **
  • Posts: 22
  • Country: dk
  • Reputation: 1
    • View Profile
Re: Error running DDOS/PLDOS analysis on QuantumATK 2019.12
« Reply #6 on: June 30, 2020, 12:01 »
Hi Aanand,

The reason that you can run some, but not all of the configurations, is related to the electrodes. Some of your devices have transverse electrode repetitions, while others don't. DeviceDensityOfStates does not support all definitions in Q-2019.12 and we will address this for the R-2020.09 release.

For now, you can pass method=LocalDeviceDensityOfStates to the ProjectedLocalDensityOfStates definition in your scripts. You might considering lowering the density_mesh_cutoff passed to ProjectedLocalDensityOfStates to improve performance.

Regards,
Pieter