center's electrode layers less than left(right) electrode layers are reasonable?

[fanjiaping]

   
    as we know, the left electrode and the right electrode were repeated from the centered

region. but when I read the paper ,I found some devices had more electrod layers than the

center's.I want to know if the device is reasonable and how to build the device as what is in

the addtional figure.
          thank you very much!!!

[Anders Blom]

This depends a bit on the algorithm used. If the "electrodes" are truly electrodes meaning they have a bulk-like Hamiltonian and effective potential, then I would be very sceptical that 2 layers can screen the potential from the scattering region, esp. at finite bias.

[fanjiaping]

     I am sorry that I didn't understand your answer clearly! could you show me more detail about that?
    thank you!

[Anders Blom]

I believe you will find the upgrade guide for 10.8 very useful. It describes the two-probe algorithm in more details (both the 2008.10 and 10.8 versions, which are a bit different).

[fanjiaping]

        I think the new version(2010 10.8or2010.10.2) can't build the model like the additional picture,isn't it?

[Anders Blom]

Sure you can

[fanjiaping]

           Could you show me more details about how to build the model.
The steps I built the model as:
First: select the Al from the database, cleave the al(1,0,0), then choose the repetition indices as (4,4,3)，finally delete the third lay. The model looks like that in the picture 1(1.PNG).
Second: adjust the distance and insert the C60 molecule. The object looks like that in picture 2.(2.PNG)(Also there is a model document)
Third: I drag-to-drop the document to the device from the bulk which was in the custom module.
Then I want to make the four electrode lays, but I can’t achieve it.
Is there anything wrong?
Thank you for your guiding!  

[Anders Blom]

Good job - you're almost there I think!

It's important to design the electrode and surface fist, before inserting the molecule. So if you wanted 4 layers in the electrode, you should have already 4 surface layers on each side from the beginning (and ideally 5 to make it easier for the code to find the repetition in order to make the electrode).

[fanjiaping]

  But now the problem I met was that if I set four or five metallic lays in center region, I can obviously get four electrode layers on both sides but can’t improve the k point parameter (it works out of computer memory (12Gb)).  (figure 1 contains four electrode layers in left electrode region, four Al layers in the left of central region and five Al layers in the right of central region, four electrode layers in the right electrode region ). There is no choice but to decrease the metallic layers. From your latest answer, I think the one you figure out is shown in figure 1. however, the one I want to build up is figure 2. so I will appreciate if you can show me the way to build up the model as that in figure2.

[Anders Blom]

It is correct that the system in Fig. 2 cannot be built in ATK as it is drawn, but what you have to remember is that in some codes (incl. ATK 2008.10) the electrodes are part of the two-probe calculation, while in ATK 10.8 they are not. So in fact you have fewer atoms in your two-probe calculation (only 4/5 layers left/right) in Fig 1 than in Fig 2 which has 6/7. Unless one assumes the electrodes in Fig 2 to be completely bulk-like, but that's a very crude approximation (too short screening)...

If you have memory problems, please make sure you are not running several MPI processes on the same node. Only one process per node is the rule, otherwise you use too much RAM.

[fanjiaping]

   Each time I run only one task in one computer, which is with the RAM of 12GB. I didn’t install the MPI in my computer because I use the trial license version of the program. Does the MPI play a crucial role in controlling the calculation process? How to choose a suitable memory for this kind of calculations? And I adjusted only the k point parameter as (3 3 100) and others I use the default parameters. I also attached the script_log file because I don’t know how to deal with the error information given in this file. Could you help to give me a solution and also tell me why this problem appeared if it is possible?
Thank you for your kind help!

[Anders Blom]

Are you perhaps running this on Windows? Even if your computer has 12 Gb, ATK for Windows is a 32-bit application and can thus only address slightly less than 3 Gb of RAM. To use the full capacity of the machine, a 64-bit ATK is needed. A 64-bit Windows version will probably be ready later this year (hopefully; we would have made one already if it were not for some libraries we use that are not yet ready), while a 64-bit Linux version is already available.

[fanjiaping]

Thank you for your swift answer. Actually I am using the 64-bit Linux to run my task. The version of the ATK I used is the 200 10 2, which is also 64-bit, I think, because the 32-bit ATK cannot be installed in the 64-bit Linux. And sometimes I can see the task use the RAM which is more than 6Gb. But everytime the same error appeared in my computer just like what I said yesterday. Are there any another reasons which may cause this problem, which has really troubled me for a long time?

[Anders Blom]

Version 2010.02 has several issues with memory handling. These have been addressed in version 10.8, which therefore uses much less memory. So, that should be your next attempt