Different Values with Wien2K

[Arles Gil Rebaza]

Hi, Exciting User's.. 

I'm doing a calculation on the magnetic moment of Fe4C (has two atoms of Fe distinguishable) ,
I perform calculations with Exciting and the values I find no magnetic (or a very small magnetic moment) but in other papers that used the Wien2k (or LMTO) reported that the material is magnetic (close to 2.67 mu_B and 1.704 mu_B  for each iron ).  This is my input file

spinpol
.true.

xctype    20

gmaxvr    25.0

nempty   10

atoms
   3                                    : nspecies
  'Fe.in'                               : spfname
   1                                    : natoms; atposl, bfcmt below
  0.0000  0.0000  0.0000    0.0000  0.0000  0.0000
  'Fe.in'                               : spfname
   3                                    : natoms; atposl, bfcmt below
  0.0000  0.5000  0.5000    0.0000  0.0000  0.0000
  0.5000  0.5000  0.0000    0.0000  0.0000  0.0000
  0.5000  0.0000  0.5000    0.0000  0.0000  0.0000
  'C.in'                                : spfname
   1                                    : natoms; atposl, bfcmt below
  0.5000  0.5000  0.5000    0.0000  0.0000  0.0000

ngridk
  20  20  20

.----------------
In the last interaction in INFO.OUT

Moments :
interstitial                :   0.7544219510E-15
muffin-tins
  species :    1 (Fe)
   atom    1                 :   0.2254784591E-15
  species :    2 (Fe)
   atom    1                 :  -0.1069676614E-15
   atom    2                 :  -0.1069676614E-15
   atom    3                 :  -0.1069676614E-15
  species :    3 (C)
   atom    1                 :   0.1764632565E-16
total in muffin-tins        :  -0.7777819944E-16
total moment                :   0.6766437516E-15

What changes should I do if I make any error in the file "exciting.in"  .!!!   or any suggestions  ..!!

Greetings...!!!

Arles V.

[francesco cricchio]

Dear Arles,

you need to break the symmetry with a magnetic field , otherwise your system will stay paramagnetic with zero moment.

The most efficient way to do this is to use a large field , let's say 10, and then use the flag reducebf=0.5, so the field will get halved at every iteration.
By the way , is there a particular reason for which gmaxvr in your input is so high ?

So for example

spinpol
.true.

reducebf
0.5

bfieldc
0.0 0.0 10.0

xctype 20

gmaxvr 25.0

nempty 10

atoms
3 : nspecies
'Fe.in' : spfname
1 : natoms; atposl, bfcmt below
0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
'Fe.in' : spfname
3 : natoms; atposl, bfcmt below
0.0000 0.5000 0.5000 0.0000 0.0000 0.0000
0.5000 0.5000 0.0000 0.0000 0.0000 0.0000
0.5000 0.0000 0.5000 0.0000 0.0000 0.0000
'C.in' : spfname
1 : natoms; atposl, bfcmt below
0.5000 0.5000 0.5000 0.0000 0.0000 0.0000

ngridk
20 20 20

All the best
Francesco

[Arles Gil Rebaza]

Hi Francesco,

I put a value to the magnetic field, (from 0.01 to 0.2) but the values were unsatisfactory

B_field     Mom. Mag. Total
-----------------------------------------------------
0.01       -0.1599056663E-02
0.03       -0.4797265122E-02
0.05       -0.7995527096E-02
0.07       -0.1119392476E-01
0.09       -0.1439246556E-01
0.11       -0.1759119441E-01
0.12       -0.1919064789E-01
0.13       -0.2079016800E-01
0.14       -0.2238975961E-01
0.15       -0.2398943002E-01
0.16       -0.2558918301E-01
0.17       -0.2718902471E-01
0.18       -0.2878896092E-01
0.19       -0.3038899619E-01
0.20       -0.3198914912E-01

I did't know I had to use a high field,  so high that the field should be to break the symmetry...??
There is some convergence criterion ..?

Thanks for your help and I will make the change that you say...!!!

Greetings...!!!

Arles V.

[francesco cricchio]

Hi again,

In principle you break the symmetry between spin up and down channel with any field.

If you know you're looking for an high moment solution ( let's say Fe with moment 2 ), by applying an high magnetic field you'll get a large moment after few iterations, i.e. less iterations to converge. Then by halving the field ( reducebf=0.5 ) the system should stay there if that is your true energy minimum.
If you use a small field the system might get stuck in a local energy minimum, or simply take many more iterations to converge.
I suspect your low moment solution is a local minimum.

You can verify what I say by trying a large field (10) and reducebf=0.5 in the Fe-bcc example, and you'll note the number of iterations will be consistently smaller than by using the small field 0.1

On the other side if you're looking for a small moment solution, it is not wise to use a large field since you'll start with a moment too large.

So you could always try both approaches and see if the system converges to the same solution, if not then you compare total energies and find the true minimum.

In principle the code should find always the true energy minimum automatically, with any field, but sometimes, if there are two solutions very similar in energy, that will be a  difficult task.

Good luck

Francesco

[Arles Gil Rebaza]

Thank you very much for your advice, I will calculate as you said to me, I informed on my results..!!

Arles V.