Number of kpoints needed for GMR structures

[baizq]

Dear colleagues,

I am now working on the transmission properties of GMR junctions. I have found in ATK manual(version 2008.10) that, for TMR structure Fe/MgO/Fe, it is always necessary to include large number of kpoints for transmission spectrum calculation (k-mesh 301x301 as in the tutorial).

Here is my question : what about GMR structures, say, Co/Cu/Co , or Co2MnSi/Ag/Co2MnSi which has demostrated the largest MR ratio experimentally?  Approximately how many kpoints are needed for such systems when calculating transmission spectrum or IV curve?

Manny thanks.

baizq

[Anders Blom]

This question can unfortunately not be answered without a careful analysis of the k-point dependent transmission coefficients. The reason you need so many points for FeMgO is the presence of very sharp resonances. If, for another case, the coefficients vary slowly with (kx,ky), you can can get by with fewer points.

Thus the only viable approach is to increase the number of points and look at how the transmission changes.

Note that the calculation of the transmission coefficients parallelizes linearly, so if you run it over 10 MPI nodes it takes roughly 10% of the time compared to a serial calculation. If the system is small compared to the memory available on each node, you can even consider MPI-parallelizing over the cores in order to maximize the node count.

[baizq]

Dear Blom,

Thanks for your rapid reply.

Actually I have found a tutorial on ATK website called 'Modeling of magnetic tunnel junctions', in which on the last page it shows an IV curve for Co/Cu/Co GMR spin valve. I just would like to know how you set the kpoints for that case, or would you please suggest a kpoint mesh size as a starting guess?

Besides, as you say, how to analyze the k-point dependent tranmission coefficient..could you please state it a little more specifically...

baizq

[Anders Blom]

I'm afraid that plot was made a very long time ago, and I don't have access to the scripts any more... Sorry.

For plotting k-point resolved transmission, see http://quantumwise.com/documents/tutorials/Fe-MgO-Fe/XHTML/chap.mgo.para.html#sect2.para.analysis.transmission (for ATK 10.8.2). There you clearly see the sharp peaks in the minority spin in the parallel case.