[Wien] magnetc moment

[Andrei Postnikov]

Dear Svetlana,

not being an expert in Pd calculations, I dare nevertheless offer
some very general considerations:

1. If you get a magnetic moment on pure Pd, whether genuine or
by calculation error, it comes from a very narrow peak which resides
just at the Fermi level in the nonmagnetic case, so that the electronic
structure is very unstable against removing (splitting) this peak
from the E_f. That's why your results are so sensitive to the number
of k-points. If you really care about absolute convergency of
your results in k-mesh, you may try introducing a (small) broadening,
extrapolating it to zero simultaneously :-) with increasing the number
of k-points.

2. But there are other issues which might be important:
You get magnetism in GGA and not in LDA. I'd expect a dependency of
results on how "good" your GGA is, i.e., on GMAX.
I'd advise you to go beyond 16. in order to test the sensitivity
of your results on this parameter.

3. Also spin-orbit interaction might play a role. I can imagine
(without actually checking it) your "sensitive" peak to become
less "peaky", so that magnetism won't survive even in GGA...
Can it be that nobody tested it so far?

Hopefully this helps,
good luck,

Andrei

 +-- Dr. habil. Andrei Postnikov ----- Tel. +49-541-969.2377 -- Fax .2351 ---+
 | Universitaet Osnabrueck - Fachbereich Physik, D-49069 Osnabrueck, Germany |
 +-- apostnik at uos.de --------- http://www.home.uni-osnabrueck.de/apostnik/ --+

On Thu, 23 Sep 2004, Kulkova S.E. wrote:

| Dear Peter and Wien users,
|
| Several months ago my student calculated spin  magnetc moment of bulk fcc Pd
| versus lattice constants (7.2 -9.0 a.u.).
| We used GGA and LDA.
| In the case 7.3 a.u.  a liiitle bit lower than experimental value  (7.345
| a.u.) we found the spin magnetic moment for GGA (13) but it was absent in
| LDA.
| Our question is not connected with GGA and LDA but with peculiarities of
| calculations.
|
| Below you can see some results for  a=7.3 a.u. (GGA)
|
| For 1024 k-pont (47 in IRBZ)  the moment is 0.248 mu_B.
| We increased with "x  kgen"  number of k-points up to 4096(145) and continue
| convergence (charge).
| The moment was  0.180 mu_B.
| We repeat this procedure further for 10000 k and 20000k and had 0.185 and
| 0.186 mu_B, respectively.
:w
|
| If we started with 4096 k-points or (10000 and 20000) at the first step
| instead  consiquent increasing of k-points we had
| 47 (1024) - 0.258 mu_B
| 145 (4096) - 0.182 mu_B
| 286 (10000) - 0.221 mu_B
| 560 (20000) - 0.231 mu_B
|
| Note that these results were obtained within Wien_97 and GGA
|
| We checked Wien_2k_01 and
| for 47 k-point (1024) we had 0.137 mu_B.
|
| For Wien2k_08 our results are the following
| 47 (1024) - 0.199 mu_B
| 145 (4096) - 0.196 mu_B
| 286 (10000) - 0.221 mu_B
| 560 (20000) - 0.231 mu_B
| (the procedure with consiquent increasing of k-points using kgen)
|
| When we introduced 20000 k-points as the first step (in init_lapw) we had
| 0.2295 mu_B.
|
| The question is why results depend on the procedure of k-points increasing too
| much?
| Why we could not reach magnetic moment convergence?
|
| We used RKMAX=9 and Gmax=16 in all cases considered.
|
| The test calculation for Ni with Wien_2k_08 was OK in our PC Pentium IV
| computer.
|
| Any suggestion are welcome.
| Thank you in advance
|
| Best regard
| Svetlana Kulkova
| _______________________________________________
| Wien mailing list
| Wien at zeus.theochem.tuwien.ac.at
| http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
|
|