Author Topic: the converge problem when using gatedtwoprobemethod  (Read 7300 times)

0 Members and 1 Guest are viewing this topic.

Offline wring

  • Regular QuantumATK user
  • **
  • Posts: 24
  • Reputation: 0
    • View Profile
the converge problem when using gatedtwoprobemethod
« on: August 25, 2009, 03:39 »
    when I apply 1 voltage gated bias to the central region, the scf calculation is not coverged. Why it happens?  The possible reason.

Offline zh

  • Supreme QuantumATK Wizard
  • *****
  • Posts: 1141
  • Reputation: 24
    • View Profile
Re: the converge problem when using gatedtwoprobemethod
« Reply #1 on: August 25, 2009, 09:12 »
Sometimes it could happen because the applied gate voltage affect the electrostatic potential in the scattering region. There are some tips listed in the manual, please refer to them first:  http://www.quantumwise.com/documents/manuals/ATK-2008.10/chap.tipstricks.html#sect1.tipstricks.twoprobecalculations
« Last Edit: August 25, 2009, 13:39 by zh »

Offline wring

  • Regular QuantumATK user
  • **
  • Posts: 24
  • Reputation: 0
    • View Profile
Re: the converge problem when using gatedtwoprobemethod
« Reply #2 on: August 31, 2009, 05:20 »
    In the parameters of twoProbeAlgorithmParameters, the parameters of electrode_constraint may be set up ElectrodeConstraints.Off or ElectrodeConstraint.RealSpaceDensity. The difference between the two setting? In the procedure, how to make up the two different parameters? In my opinion, the electrode density is calculated to be identical to the bulk calculations.
     Thanks!

Offline zh

  • Supreme QuantumATK Wizard
  • *****
  • Posts: 1141
  • Reputation: 24
    • View Profile
Re: the converge problem when using gatedtwoprobemethod
« Reply #3 on: August 31, 2009, 07:02 »
Some explanation for the difference between "ElectrodeConstraints.Off" and "ElectrodeConstraint.RealSpaceDensity" are given in the manual: http://www.quantumwise.com/documents/manuals/ATK-2008.10/ref.twoprobealgorithmparameters.html#ref.twoprobealgorithmparameters.notes

If the surface layers in the scattering region are not thick enough so that the interaction between the electrode and the conductor in scatter region is strong and the charge density between the electrode and the scattering region is redistributed, it is better to adopt "ElectrodeConstraints.Off". Otherwise, the surface layers are very thick enough so that the screening effect is satisfied, "ElectrodeConstraints.Off" and "ElectrodeConstraint.RealSpaceDensity" would give identical results.


A script is shown there for a calculation on a two-probe system performing first a constrained and afterwards an unconstrained self-consistent calculation.


Offline wring

  • Regular QuantumATK user
  • **
  • Posts: 24
  • Reputation: 0
    • View Profile
Re: the converge problem when using gatedtwoprobemethod
« Reply #4 on: August 31, 2009, 11:00 »
I think if we want to use the option "ElectrodeConstraint.RealSpaceDensity", then I think the scattering region should have such a structure:

"one left electrode cell + left surface layers + molecule + right surface layers + one right electrode cell"

so that we can replace the charge density in the "one left electrode cell" and "one right electrode cell" in the scattering region by that of the bulk lead. Here if the left and right surface layers are thick enough,  then "ElectrodeConstraint.RealSpaceDensity" and ""ElectrodeConstraint.Off" will give the same results, if not, then maybe big difference exists.

Is my understanding right?  If my understanding is right and if the surface layers(here I mean, the total lenghth of "one left electrode cell + left surface layers" in the scattering region)  are not as thick as one electrode cell, then I think we can not use the option "ElectrodeConstraint.RealSpaceDensity" at all.  In this case, I think only "ElectrodeConstraint.Off" can be used.  Am I right?
« Last Edit: August 31, 2009, 11:03 by wring »

Offline zh

  • Supreme QuantumATK Wizard
  • *****
  • Posts: 1141
  • Reputation: 24
    • View Profile
Re: the converge problem when using gatedtwoprobemethod
« Reply #5 on: September 1, 2009, 04:31 »
I think if we want to use the option "ElectrodeConstraint.RealSpaceDensity", then I think the scattering region should have such a structure:

"one left electrode cell + left surface layers + molecule + right surface layers + one right electrode cell"

so that we can replace the charge density in the "one left electrode cell" and "one right electrode cell" in the scattering region by that of the bulk lead. Here if the left and right surface layers are thick enough,  then "ElectrodeConstraint.RealSpaceDensity" and ""ElectrodeConstraint.Off" will give the same results, if not, then maybe big difference exists.

Is my understanding right?  If my understanding is right and if the surface layers(here I mean, the total lenghth of "one left electrode cell + left surface layers" in the scattering region)  are not as thick as one electrode cell, then I think we can not use the option "ElectrodeConstraint.RealSpaceDensity" at all.  In this case, I think only "ElectrodeConstraint.Off" can be used.  Am I right?
You understanding is wrong. The scattering region does not include  the electrode regions. Actually, the surface layers in the scattering region have the same periodicity (in the directions perpendicular to the transport direction) with the electrode regions. The surface layers act as the screening effect to avoid the interaction between electrodes and molecule.


Offline xhsh

  • Heavy QuantumATK user
  • ***
  • Posts: 32
  • Reputation: 0
    • View Profile
Re: the converge problem when using gatedtwoprobemethod
« Reply #6 on: September 2, 2009, 09:29 »
I think if we want to use the option "ElectrodeConstraint.RealSpaceDensity", then I think the scattering region should have such a structure:

"one left electrode cell + left surface layers + molecule + right surface layers + one right electrode cell"

so that we can replace the charge density in the "one left electrode cell" and "one right electrode cell" in the scattering region by that of the bulk lead. Here if the left and right surface layers are thick enough,  then "ElectrodeConstraint.RealSpaceDensity" and ""ElectrodeConstraint.Off" will give the same results, if not, then maybe big difference exists.

Is my understanding right?  If my understanding is right and if the surface layers(here I mean, the total lenghth of "one left electrode cell + left surface layers" in the scattering region)  are not as thick as one electrode cell, then I think we can not use the option "ElectrodeConstraint.RealSpaceDensity" at all.  In this case, I think only "ElectrodeConstraint.Off" can be used.  Am I right?
You understanding is wrong. The scattering region does not include  the electrode regions. Actually, the surface layers in the scattering region have the same periodicity (in the directions perpendicular to the transport direction) with the electrode regions. The surface layers act as the screening effect to avoid the interaction between electrodes and molecule.



I think you did not understand wring's question.  Let's suppose the electrode supercell is "ABCABCABC" and the central region is "ABC+molecule+CABC",  then since in the central region, the surface layers "ABC" or "CABC" are not as long as the electrode supercell "ABCABCABC", we can not replace the charge density of the surface layers in the central region by the one in the electrode supercell.

So, if the surface layers are shorter than the electrode supercell, we can not use the option "ElectrodeConstraint.RealSpaceDensity".

Offline wring

  • Regular QuantumATK user
  • **
  • Posts: 24
  • Reputation: 0
    • View Profile
Re: the converge problem when using gatedtwoprobemethod
« Reply #7 on: September 2, 2009, 09:35 »
   You don't understand my question.
   I understand that the surface layers act as the screening effect to avoid the interaction between electrodes and molecule, so the surface layers must be as long as the electrode supercell. In that case,  "ElectrodeConstraint.RealSpaceDensity" and "ElectrodeConstraint.Off" are the same.
   Thanks.

Offline zh

  • Supreme QuantumATK Wizard
  • *****
  • Posts: 1141
  • Reputation: 24
    • View Profile
Re: the converge problem when using gatedtwoprobemethod
« Reply #8 on: September 2, 2009, 10:29 »
 the surface layers "ABC" or "CABC" are not as long as the electrode supercell "ABCABCABC", we can not replace the charge density of the surface layers in the central region by the one in the electrode supercell.

The problem is that "ElectrodeConstraint.RealSpaceDensity" or "ElectrodeConstraint.off" just allows user to concern about the charge density in electrode during the self-consistent calculation of two-probe system.  As explained in the manual, "ElectrodeConstraints.RealSpaceDensity" means that the real-space density in the electrode region is constrained to be identical to the one of the bulk calculation for the electrodes, while "ElectrodeConstraints.Off" means that the calculation is performed without constraints on the electrodes. Somehow, it seems too concise for the explanation of "ElectrodeConstraints.Off". We may think of it further. On the other hand, it also means that when "ElectrodeConstraints.Off" is adopted the charge densities in electrodes should be obtained by means of a self-consistent manner together the self-consistent calculation of two-probe system.  

Offline xhsh

  • Heavy QuantumATK user
  • ***
  • Posts: 32
  • Reputation: 0
    • View Profile
Re: the converge problem when using gatedtwoprobemethod
« Reply #9 on: September 2, 2009, 11:25 »
 the surface layers "ABC" or "CABC" are not as long as the electrode supercell "ABCABCABC", we can not replace the charge density of the surface layers in the central region by the one in the electrode supercell.

The problem is that "ElectrodeConstraint.RealSpaceDensity" or "ElectrodeConstraint.off" just allows user to concern about the charge density in electrode during the self-consistent calculation of two-probe system.  As explained in the manual, "ElectrodeConstraints.RealSpaceDensity" means that the real-space density in the electrode region is constrained to be identical to the one of the bulk calculation for the electrodes, while "ElectrodeConstraints.Off" means that the calculation is performed without constraints on the electrodes. Somehow, it seems too concise for the explanation of "ElectrodeConstraints.Off". We may think of it further. On the other hand, it also means that when "ElectrodeConstraints.Off" is adopted the charge densities in electrodes should be obtained by means of a self-consistent manner together the self-consistent calculation of two-probe system.  

I understand the meaning of the two options "ElectrodeConstraint.RealSpaceDensity" or "ElectrodeConstraint.off". I am just stressing that if we want to use "ElectrodeConstraint.RealSpaceDensity", the surface layers should be at least as long as the electrode supercell. However, if we use "ElectrodeConstraint.off", then there is no such requirement needed. I think this is also what wring is worrying about. Please just confirm this. Thanks.



Offline zh

  • Supreme QuantumATK Wizard
  • *****
  • Posts: 1141
  • Reputation: 24
    • View Profile
Re: the converge problem when using gatedtwoprobemethod
« Reply #10 on: September 2, 2009, 12:00 »
I am just stressing that if we want to use "ElectrodeConstraint.RealSpaceDensity", the surface layers should be at least as long as the electrode supercell. However, if we use "ElectrodeConstraint.off", then there is no such requirement needed. I think this is also what wring is worrying about. Please just confirm this. Thanks.
No such requirement for "ElectrodeConstraint.RealSpaceDensity" or "ElectrodeConstraints.Off" because the code does not replace the charge density in electrode with the counterpart in scattering region. The requirement for the thickness of surface layers in scatter region just originates from the screening effect.