[Wien] very slow SCF convergence

Laurence Marks ldmarks at yahoo.com
Fri Aug 4 14:50:11 CEST 2006


See:

http://zeus.theochem.tuwien.ac.at/pipermail/wien/2006-July/007547.html

http://zeus.theochem.tuwien.ac.at/pipermail/wien/2005-May/005277.html

Also, be careful that the Ni RMT's are not too much
bigger than the O & C; the later will be small due to
the relatively close difference. Look at NiO for the
relative difference in size that you should have.

--- Mahbube Hortamani <hortaman at fhi-berlin.mpg.de>
wrote:

> Dear Wien user,
> 
> I have problem with SCF convergence of a system
> which is one CO molecule
> on top of a (2x2) Ni surface with 68 atoms per unit
> cell.
> The charge converges very very slow. For example, in
> the best case,
> charge convergence after 340 iterations is just
> 0.002 while the energy
> and magnetic moments are converged (see below). Can
> one trust to these 
> results for
> energy and magnetic moments?
> 
> grep_lapw :ITE '*scf'
> :ITE328:328. ITERATION
> :ITE329:329. ITERATION
> :ITE330:330. ITERATION
> :ITE331:331. ITERATION
> :ITE332:332. ITERATION
> :ITE333:333. ITERATION
> :ITE334:334. ITERATION
> :ITE335:335. ITERATION
> :ITE336:336. ITERATION
> :ITE337:337. ITERATION
> :ITE338:338. ITERATION
> :ITE339:339. ITERATION
> :ITE340:340. ITERATION
> 
> grep_lapw :DIS '*scf'
> :DIS  :  CHARGE DISTANCE       0.0019451
> :DIS  :  CHARGE DISTANCE       0.0019467
> :DIS  :  CHARGE DISTANCE       0.0019065
> :DIS  :  CHARGE DISTANCE       0.0018985
> :DIS  :  CHARGE DISTANCE       0.0019240
> :DIS  :  CHARGE DISTANCE       0.0019640
> :DIS  :  CHARGE DISTANCE       0.0019174
> :DIS  :  CHARGE DISTANCE       0.0021151
> :DIS  :  CHARGE DISTANCE       0.0023446
> :DIS  :  CHARGE DISTANCE       0.0023163
> :DIS  :  CHARGE DISTANCE       0.0022396
> :DIS  :  CHARGE DISTANCE       0.0023006
> :DIS  :  CHARGE DISTANCE       0.0020873
> 
> grep_lapw :ENE '*scf'
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101700
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101709
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101712
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101711
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101671
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101749
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101726
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101712
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101693
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101746
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101790
> :ENE  : ********** TOTAL ENERGY IN Ry =     
> -195118.101858
> 
> grep_lapw :MMI003 '*scf'
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79119
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79119
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79121
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79125
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79129
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79133
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79157
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79167
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79167
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79167
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79167
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79166
> :MMI003: MAGNETIC MOMENT IN SPHERE   3    =   
> 0.79165
> 
> I applied following treatments to get converged
> charge but charge 
> convergence is still slow!
> 
> 1) changing the mixing factor in *.inm file for the
> values of 0.2, 0.05
> and 0.02.
> 2) using "TEMP" switch with "eval 0.009 & 0.005" in
> *.in2 file to do E_F
> and k space integration, in addition to the "TETRA"
> method  ( the
> default setup).
> 3) changing muffin-tin radius of Ni.  (R_O=1 & R_C=1
> & R_Ni= 2.0, 2.3)
> 4) changing number of Ni layers in the slab (16 & 17
> layers).
> 5) reducing number of symmetry operations of system.
> 6) testing with/without CO molecule.
> 7) starting with very low RKmax to get *clm* and
> then increasing RKmax.
> for small RKmax (i.e. E_cut= 6.76 Ry) convergence is
> faster but for
> higher RKmax (E_cut= 13.66 Ry) is very slow.
> 8) using -in1new flag.
> 9) testing different Ni-C bound length (but this is
> not so important).
> 
> It might be helpful to know that for  bare Ni(2x2)
> surface, with 
> E_cut=13.66 Ry, the charge converges after 280
> iterations.
> 
> Does any body have experience with such problem
> and/or Ni surface?
> 
> P.S. this is real (not complex) and spin polarized
> ferromagnetic
> calculation. I used WIEN2k_03 version.
> 
> Many thanks in advance,
> Mahbube.
> 
> 
> -- 
>
---------------------------------------------------------------------------
> Mahbube Hortamani
> 
> Fritz-Haber-Institut
> der Max-Planck-Gesellschaft
> Abteilung Theorie
> Faradayweg 4-6
> D-14 195 Berlin-Dahlem / Germany
> 
> Phone  : ++49-30-8413 4820
> Fax    : ++49-30-8413 4701
> E-mail : hortaman at fhi-berlin.mpg.de
>
---------------------------------------------------------------------------
>      www:
> http://w3.rz-berlin.mpg.de/~hortaman/member/
>
------------------------------1-0.0735-11600-23.05-------------------------
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
>
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> 


L. D. Marks
Dept of Mat. Sci and Eng.
Northwestern University
Evanston IL
mailto:L-Marks at nwu.edu
http://www.numis.nwu.edu

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


More information about the Wien mailing list