|
igor
Registered: 12/03/07
Posts: 3
|
 Thursday, January 07 2010 @ 11:54 AM CET |
|
Has anyone ever tried to make RSPt simulations with a magnetic field???
I have been trying in the last weeks but I could not get any reliable results. In particular I have found that for extremely small fields RSPt seems to converge, but as soon as the field reaches a few mRy the program goes crazy.... this is independent on the size of the mixing (I have tried many possible configurations, up to pmix=0.0001 and spin_mix=0.00005). My test cases have been bcc Fe and fcc Fe.
Moreover I guess that in ud2tm0.f the condition (fiestr .gt.0) should be replaced with (fiestr .ne.0).
Igor |
|
|
|
|
| |
Admin
Registered: 10/12/06
Posts: 130
|
| Thursday, January 07 2010 @ 12:47 PM CET |
|
1: Remember that the Bohr magneton is of the order of 4*10^-6 Ry/T so a few mRy is a wopper of a field. It could be that there is no stable solution for that kind of field. Do you think that the solutions for lower fields look OK? When the magnetic energy is of the same order as the energy scale of the electronic part of the problem, the SE is not a good description of the problem anymore as magnetic effects are intrinsically relativistic.
2: I think that ud2tm0.F should be corrected but I have to look closer at the problem. I guess (always a bad thing) that the author never thought about negative fields.
/Admin |
|
|
|
|
| |
igor
Registered: 12/03/07
Posts: 3
|
| Thursday, January 07 2010 @ 03:50 PM CET |
|
| What I am trying to do is Fixed Spin Moment calculations, and the simplest way is just to force the magnetic moment to vary by applying an external magnetic field. This is why I need magnetic fields which are so big. It is just a constraining field to force the magnetic moment to assume a well definite value. In any case I am trying to reduce the steps between the fields, maybe this will increase the convergence... |
|
|
|
|
| |
Admin
Registered: 10/12/06
Posts: 130
|
| Thursday, January 07 2010 @ 04:35 PM CET |
|
We ought to have a more stable mechanism for that. Any volunteers?
/Admin |
|
|
|
|
| |
Torbjorn
Registered: 11/03/06
Posts: 75
|
| Thursday, January 07 2010 @ 06:49 PM CET |
|
Hello,
I have made an attempt to do this using a PID regulator. I have not been able to find parameters that keeps this stable, however. But it is under way in my branch for merging the SIC code into trunk.
Torbjörn |
|
|
|
|
| |
Torbjorn
Registered: 11/03/06
Posts: 75
|
| Tuesday, February 02 2010 @ 10:39 PM CET |
|
Hello again!
Now the long awaited fix spin moments are in the trunk. No documentation written yet, but there is a sample input file in the documentation/inputs directory. It is tricky to get it stable, but do a readup on how a PID controller works and you should be able to use it. I have managed to do some funky stuff like taking the self-consistent moments of bcc Fe and reversing the sign of the interstitial moment. It seems that constraining site moments or the interstitial moment is more stable than constraining the total moment.
Good luck!
Torbjörn |
|
|
|
|
| |
Torbjorn
Registered: 11/03/06
Posts: 75
|
| Sunday, February 07 2010 @ 06:25 PM CET |
|
Hello again!
Now the manual is also updated with information about the fixed spin moment calculations.
Torbjörn |
|
|
|
|
| |
