magnetic tunnel junction

[sahil]

Hiiii i am new to atk vnl.. i am working on the cofeb/mgo/cofeb magneto tunnel junction for this purpose i made the structure using replacing the fe atom with co and at the interface with mgo i attach single boron atom but due to this the bond between the boron and other atom(fe or co) was borken hence i reduce the distance (2.177A) as shown in fig 1.
then i optimize the bulk structure with forces and stress .05. which takes a long time. i select my k point 6,6,3 for this optimization. and for better convergence i kept electron temp 1200K.
but i have following doubt  
1> is it a correct structure?
2> i am facing the problem with convergence in the optimization process as below...
# Warning: The calculation did not converge to the requested tolerance!        #

how could i handle this..
3> now after optimization and during the parallel and anti-parallel confg how may i decide the electrode length in device configuration?

4> and at last how to study about Mulliken population and k-parallel and antiparallel transmission curve for majority and minority spin..

5>How can i make my calculation faster with rough result?
please tell me any one.....  

[Anders Blom]

Your procedure seems reasonable, and since you do optimize it, the initial structure doesn't have to be perfect. The convergence problems may be because of too low mesh cut-off, Fe needs a higher value than default (try to double it, at least). 6x6 is also on the low side for the k-points.

There is a mini-tutorial on how to determine the electrode length. It think 4 layers is enough, 6 would be really good. But note that you should keep the "electrode copy" constrained in the relaxation. So, if you plan to have 4 electrode layers, those atoms must be fixed already in the bulk relaxation, unless you plan to add them on later, i.e. to buffer the configuration out with those 4 layers.

This calculation will benefit immensely from parallelization, it will be 10 times faster if you run it on 16 nodes or so, and perhaps 3-4 times faster on 5 nodes.

You should study the FeMgO tutorial, it has a lot of advise on how to analyze the parallel/antiparallel systems.

[sahil]

thank you for your reply sir....

I have optimized my image, and after optimization the electrode length was automatically set.  now i am working on device configuration. Here i want to plot the graph between the conductance ratio(parallel conductance/ anti parallel conductance) to "per layer thickness of cofeb"

but how can i draw this graph? 


 

[Anders Blom]

You mean the TMR as a function of number of layers MgO (or FeCoB)? That's a simple plot of a few numbers (TMR) vs. another few number (layers), just use any plotting program, and copy down the results of each calculation in a file, you don't need a fancy plotting solution for that.

[sahil]

means i have to run the TMR calculation by changing the length of electrode or insulator manually.. ok

sir i have one more question

I have studied your tutorial regarding Fe-Mgo-Fe. In chapter 4 of that tutorial you make your spin "-1" for 14 to 23 atom(from second last insulating layer when moving left to right)for anti-parallel case. but i calculated my total energy calculation form 12-23 and 14-23 with "-1" spin. This both of calculation gives same "total energy" result. then how u decide that the  "-1" spin will be only from 14-23.  

and in another question  i have calculate the density of state for cofe/mgo/cofe foe parallel case(all atom have "+1" spin). this graph i am attaching. please tell me is it correct or not? because characteristic of this curve is some like strange.

Thanks 

[Anders Blom]

The spin you set in the input script is only the initial spin, the starting guess. The calculation will proceed find a self-consistent solution which can be quite different from the initial spin - you would find the same solution also by setting all initial spins to -0.9 instead of -1, etc.

The plot probably just needs to be zoomed in energy; you have a very flat band at a large energy, this is irrelevant, you should just zoom in on the energy region of interest, around the Fermi level.

[sahil]

sir is it necessary to optimize same structure again when i worked for different temperature???
  

[Anders Blom]

It would be, if we had real temperature effects included in the calculations. As it is now, the temperature only enters as a numerical technique to smoothen the Fermi distribution, and in this way it has no real physical meaning. Hence, no, you do not need to reoptimize, however you should also be very aware that changing the "temperature" will generally not produce the effect of changing the thermal properties of the system.

There is one exception to this: when we set different electron temperatures on the electrodes in the transmission calculation, we can use this to simulate a hot electron (equillibrium) distribution. But still there are no phonon effects included, and also in this case the geometry would not change.

[sahil]

please tell me anyone which type of error is this?

Traceback (most recent call last):
  File "c:\users\user2\appdata\local\temp\6727543481552833.py", line 248, in <module>
    nlsave('D:/paper fig/cofeb_anti28feb12k.nc', transmission_spectrum)
  File ".\zipdir\NL\IO\NLSaveUtilities.py", line 154, in nlsave
  File ".\zipdir\NL\IO\IOUtilityFunctions.py", line 543, in createNetCDFFile
  File ".\build\atkpython\lib\shutil.py", line 88, in copy
  File ".\build\atkpython\lib\shutil.py", line 54, in copyfile
  File ".\build\atkpython\lib\shutil.py", line 30, in copyfileobj
IOError: [Errno 28] No space left on device
 
this comes when i tried to perform "anti parallel spin" case. and due to this my k parallel graph got affect. how could i ignore that

[Anders Blom]

I think the error message says it all:

IOError: [Errno 28] No space left on device

Your disk is full.

[sahil]

how can i apply the magnetic field on MTJ of different value? and often sometime i tried to find out the change in TMR by changing the insulator length? but manually it takes a long time because i have to optimize my structure first for accurate result, after that i work on device configuration. is their any other method of doing this???  

[zh]

how can i apply the magnetic field on MTJ of different value? and often sometime i tried to find out the change in TMR by changing the insulator length? but manually it takes a long time because i have to optimize my structure first for accurate result, after that i work on device configuration. is their any other method of doing this???  

It is not available to apply the magnetic field.  Of course, you had better do such way to check the effect of insulator length on the TMR. Usually, no short-cut way to take in this problem.

[Nordland]

It looks like the usual result for Fe-MgO-Fe.

[sahil]

hiiii
thanks for your reply 
sir i have found the k parallel transmission curve but i want to analyze it in 3D for better appearance...
but i am not good in programming at python 
so please may you send me a simple script of analysing it in 3D....
i am also attaching my 2D graph with this.... 

reply please

[Anders Blom]

There is no immediate support for that kind of 3d plot in matplotlib (anymore... there used to be), so you would have to resort to other software packages, like Matlab, GNUPlot, perhaps even Excel. The data you need is T(E,kx,ky) which is a matrix for a given E, as function of k. The is available in the script as the variables T_uu and T_dd (I assume you were following the tutorial), while the k-values are in the variables K_A and K_B. You would need to export those (print them to a file) in a format the other software understands.