[Wien] optimizing so on converged sp

[Stefaan Cottenier]

For such situations, you should have done the following first:

* in your optimize.job, save with the -a switch (then you have really  
all necessary files for a restart, also the case.in1 with the  
linearization energies for each structure)

* you should have done initso_lapw first, to have the case.struct and  
symmetries that correspond to spin-orbit coupling, even if you want to  
do a series without spin-orbit first.

Having done this, you can now add a restore_lapw command in your  
optimize.job, such that the converged cases without -so are used as a  
start.

As you probably did not do initso_lapw first, you are a bit in  
trouble. Adding a restore command is still possible, but you also need  
to execute initso_lapw for every structure. Either you try to do that  
inside optimize.job in a kind of batch mode (never done that, is  
probably possible), or you do it first manually (probably easier).

Stefaan



Quoting mansore ilkhani <ilkhaniirn at yahoo.com>:

> Dear WIEN group,
> I did run optimize.job script for a magnetic case using runsp_lapw   
> –i 300 –cc 0.00001 –in1new 1 and plotted an energy curve versus   
> volume. Now I am going to gradually add –so flag to include   
> spin-orbit coupling (SOC) to study the effect of SOC on the   
> structural properties of my case. I would use my last converged   
> results. However, it seems that I cannot simply add the –so flag to   
> the above runsp_lapw command in the optimize.job. I cannot use my   
> last result only by adding -so due to the fact that my current   
> *.clmsum is corresponded to the last structure of the optimization.   
> Thus if I only add –so and then run again the optimize.job, then the  
>  current *.clmsum of the last structure will be used by LAPW0 for  
> the  first structure of the optimization. Obviously this is not  
> suitable,  for the lattice parameter of the first structure is  
> smaller than the  lattice parameter of the last structure. Similarly  
> the current *.in1  file is corresponded to the last structure.  
> Therefore, the
>  performed linearization energy for the case of last structure of   
> the optimization will also be used for the first structure with most  
>  likely different linearization energy. Thus one does expect to   
> encounter ghost band problem, as I do. In this case one prefers to   
> run x dstart –up/dn and copy *.in1_orige to *.in1, which destroys   
> our last converged results.     This problem will be more serious,   
> because I would also gradually add –orb flag to include LDA+U   
> interaction.
>     I appreciate any comments from you.
>   It woul not be a good idea to include both –so and –orb from   
> scratch forgetting about all our obtained results without these two   
> flags, would it be?
>
> best regards,   M. Ilkhani.
>
>
>
> ---------------------------------
> Ahhh...imagining that irresistible "new car" smell?
>  Check outnew cars at Yahoo! Autos.



-- 
Stefaan Cottenier
Instituut voor Kern- en Stralingsfysica
K.U.Leuven
Celestijnenlaan 200 D
B-3001 Leuven (Belgium)

tel: + 32 16 32 71 45
fax: + 32 16 32 79 85
e-mail: stefaan.cottenier at fys.kuleuven.be


Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm