[Wien] non-rectangular cell

Tianjiao Zhang tjzhang at phys.ntu.edu.tw
Sat Jun 19 06:48:47 CEST 2004


Hello,
 Here I wish to explore further about the convergence.  According to the 
charge distance :DIS, I really don't think the charge density converged 
well as Mahbube said in the original email. It is actually flunctuating. 
Also I can not understand well Torsen's reply
....
 This is actually how it is supposed to work, since a self-consistent 
convergence has to fluctuate a bit in order to assure that you end up 
with a good solution.
 ....

 I am experiencing a similar problem. In my calculation of magnetic 
properties, during more than 100 iterations,  I just found that the 
charge distance :DIS was flunctuating about 0.0003.   As I chose a stop 
criteria -cc 0.0001, the calculation did not stop. And I don't think it 
is a very strict criteria.  I  did some similar  runs but with different 
mixing parameters and could not bring down the :DIS value further. The 
calculation seems to be quite stable and not crash.
 Later, I used the "-in1new -N " option in runsp_lapw and was able to 
bring down the   :DIS value  one order smaller !  
 Although the manual of WIEN does not say which criteria should be the 
final judge about convergence, the guidance in my calculation is that 
:DIS is the ultimate factor as that is what the word "self-consistency" 
means and the practice of DFT relies on. So I really expect it becomes 
smaller and smaller as iterations go on.
 Here I'd like to know what makes up the inherent contribution to the 
fluctuation of charge distance :DIS. Hope it may be helpful  to choose a 
stop criteria with less ambiguity.
(1)  Are there any ideas /references about studying contributions to 
:DIS from imcomplete basis set similar to the philosophy that Pulay 
force is treated and corrected in calculation ?  Will an incomplete 
basis can actually get :DIS=0 and we should not call it as "incomplete" 
all all ?
(2) How effective is a smaller searching deltaE in case.in1 (and 
case.in2 ?) in shrinking the magnitude of flunctuation in :DIS ?
(3) Are there any explicit algorithms build in WIEN to exploit the 
fluctuation ? I read through Stefaan's step-by-step introduction but I 
am not sure whether it is mentioned there. Will the approximation in 
Equ. (4.1)  ( lapw appromimation of APW ) give significant contribution ? 
(4) How much restraints should we keep to ourselves so that we are aware 
that we are away from the inherent numerical errors brought by limited 
digits in storing variables ?  For example, does a charge distance 
:DIS=0.00000001 make sense ?
  In my calculation, I used modified TETRA method to do Ef and k space 
integration ( the default setup generated with init_lapw).

Sincerely,
Tianjiao

 

Mahbube Hortamani wrote:

>unfortunately, this is the result of  *.scf_mini. I should say one of my previous
>attachment was not correct, I send the correct one with this mail
>I tried to use of P lattice in several different cases. In some cases it
>works and with minimization of force, the energy also becomes minimal.
>but for most of them the energy rises up.
>>From one case that I have problem, I attached  the first
>structure file which I generated and the last file that was generated by Wien.
>Unfortunately, *.scf_mini file is too big and I cannot send it with mail but If
>you need  more detail I can give you any files which you want.
>thanks in advance,
>Mahbube.
>
>best regards,
>Mahbube.
>
>Peter Blaha wrote:
>
>  
>
>>Is this the output of
>>
>>grep :ENE case.scf    or
>>
>>grep :ENE case.scf_mini  ????
>>
>>In the first case, this is ok, as explained before.
>>
>>However, in the latter, this is not ok, since during minimization of
>>forces also the energy (as given in case.scf_mini) should become minimal.
>>
>>I just noticed: Your struct file is a CXZ monoclinic lattice.
>>
>>Are you aware of the famous "possible monoclinic CXZ bug" ???
>>Have you checked your results ?
>>
>>Most likely you should do this in a P lattice (unfortunately with twice
>>as many atoms...)
>>
>>    
>>
>>>Dear Wien2k users,
>>>I want to calculate the total energy of the relaxed system with non
>>>rectangular cell (monoclinic).
>>>During the force minimization we have fluctuation in the total energy ,
>>>at the first energy goes down then rises up! The force and charge
>>>density were converged very well.
>>>
>>>:ENE  : ********** TOTAL ENERGY IN Ry =       -23195.781714    :DIS  :
>>>CHARGE DISTANCE       0.0004508
>>>:ENE  : ********** TOTAL ENERGY IN Ry =       -23195.806552    :DIS  :
>>>CHARGE DISTANCE       0.0003372
>>>:ENE  : ********** TOTAL ENERGY IN Ry =       -23195.825306    :DIS  :
>>>CHARGE DISTANCE       0.0009501
>>>:ENE  : ********** TOTAL ENERGY IN Ry =       -23195.827936    :DIS  :
>>>CHARGE DISTANCE       0.0005407
>>>:ENE  : ********** TOTAL ENERGY IN Ry =       -23195.830196    :DIS  :
>>>CHARGE DISTANCE       0.0006952
>>>:ENE  : ********** TOTAL ENERGY IN Ry =       -23195.829836    :DIS  :
>>>CHARGE DISTANCE       0.0006057
>>>:ENE  : ********** TOTAL ENERGY IN Ry =       -23195.827175    :DIS  :
>>>CHARGE DISTANCE       0.0006018
>>>:ENE  : ********** TOTAL ENERGY IN Ry =       -23195.826570    :DIS  :
>>>CHARGE DISTANCE       0.0004550
>>>
>>>      
>>>





More information about the Wien mailing list