QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: perfetti on April 20, 2012, 03:46

Title: Log file stopped
Post by: perfetti on April 20, 2012, 03:46

Dear Everyone,
        I am trying to calculate the transmission spectrum of a device on three parallel computers, and the calculation runs smoothly at first. However, at the two probe part, it almost converged at the last step, and it stopped there, without any progress any more.
         I have spent all day to get this result, and now it seemed to fail at the last step. And I don't know what's going on there.
         I just set the transmission parameters na=3 , and nb=3, the same as the k points in scf calculations. is that the reason that cause problem? Besides this I cannot imagine any other problem.
        Any advice is appreciated.

       

Title: Re: Log file stopped
Post by: kstokbro on April 20, 2012, 08:44

please post your input and output, without that it is very difficult for us to give you advice

Title: Re: Log file stopped
Post by: perfetti on April 22, 2012, 15:48

Dear Mr., I already sent them to your email, thanks.

Quote from: kstokbro on April 20, 2012, 08:44

please post your input and output, without that it is very difficult for us to give you advice

Title: Re: Log file stopped
Post by: Anders Blom on April 22, 2012, 23:18

In 12.2 the default method for computing the self-energies has been changed to Krylov, it was Direct before. The Direct method is more stable, but Krylov is a lot faster (about 10x in many cases). So if you don't have specific problems with the Krylov method (negative transmission being the problem one can see) use it and enjoy the improved performance.

If you calculation converged (meaning it saved the converged state to the NC file) you can break the script and make a separate script for the transmission, like

Code: python

device_configuration = nlread("file.nc",DeviceConfiguration)[0]
transmission_spectrum = TransmissionSpectrum(
    configuration=device_configuration,
    energies=numpy.linspace(-2,2,101)*eV,
    kpoints=MonkhorstPackGrid(3,3),
    self_energy_calculator=KrylovSelfEnergy(),
    )
[/python]