Restarting a device configuration

[ATK-user-zuox]

hello,quantumwise staff : 1.i also met the restarting problems ,here is my confusion:    there are so many verisions of restart a bias dependent calculation ,but i can't get the uniform answer about it ,there is basically two answers  in the forum :    first solution ,

Code

# -------------------------------------------------------------
# IV Curve
# -------------------------------------------------------------
calculator=device_configuration.calculator()
device_configuration = nlread("ivcurve_selfconsistent_configurations_bias_0.nc",)[0]

     second solution,

Code

device_configuration.setCalculator(calculator)
nlprint(device_configuration)
device_configuration = nlread("ivcurve_selfconsistent_configurations_bias_0.nc")[0]
device_configuration.update(force_restart=True)
nlsave('CONFIG.nc', device_configuration)

            I don't know  the accuracy and differences between these two solutions  and in the

Code

 nlread("ivcurve_selfconsistent_configurations_bias_0.nc")[0]

,what is the exact meaning of

• and [-1],or when should i use
• and [-1],what is the exact meaning of 
  Code

  force_restart=True

  . 2.i am also very confused about the restarting procedure ,can you explain about it ?In another posthttp://quantumwise.com/forum/index.php?topic=1374.msg6752#msg6752 ,it tells me that the restarting procedure includes these steps : 1.) read Electrode 2.) read checkpoint-file  (== equiv. bulk) 3.) create Device from equiv. Bulk and electrodes 4.) attach a new device calculator 5.) update  Hope you can give us a detailed description of the equiv bulk and the  restarting procedure .

[zh]

It depends on how much information has been saved in your previous calculations.

[Jess Wellendorff]

1. There is no straight-forward way to restart an IVCurve calculation that was interrupted. This is because what the analysis really does is to run a range of SCF calculations followed by TransmissionSpectrum analysis. A rather complex operation that requires restarting from previous finite-bias calculations in order to ease convergence. A restart scheme for this is much more complicated than it may appear. In general, however, you can always start an analysis from the zero-bias calculations, e.g.

Code

device_configuration = nlread("ivcurve_selfconsistent_configurations_bias_0.nc")[0]

where "[ 0 ]" means that you want the first item saved in the file, in this case the zero-bias calculation. 2. Checkpoint files should be used only a last resort. Restarting from the saved zero-bias calculation is a more transparent approach.