See comments below:
*For TiC* (I already posted this query few month ago and got a quick
reply from Prof. Marks)
Reply from Prom Marks was
"You do not have a problem! The charge convergence is very sensitive
to numerical issues, so using -cc 0.0001 demands a lot and is probably
not a good idea"
I ran calculation via w2web and also with terminal with -cc 0.0001 (suggested
in UG) or -ec 0.0001.
It conversed again in 14 cycles.
Any comment why it is not conversing at 11 iterations?
This should be simply because your calculation takes 14 cycles to
satisfy your convergence criteria (-cc 0.0001 or -ec 0.0001).
I want it in with 11 iterations.
You could force it to stop in the 11th iteration by using the -i option
(e.g., run_lapw -i 11 -cc 0.0001), but then it would not reach the
convergence criteria (-cc 0.0001).
Not being able to reproduce 'exactly' the results ofthe examples like
TiC in the userguide is normal.
The TiC calculation in the usersguide should be from one of the old
WIEN2k versions. It was also likely done using an old version of ifort/mkl.
The reason it converges in 14 instead of 11 cycles is likely because a
a. You might have slightly different input parameters somewhere.
b. Changes to the WIEN2k code (fixes and improvements have been made
since then to programs like lapw1, lapw2, mixer [
http://www.wien2k.at/reg_user/updates/ ]; the examples might have been
produced without the -in1ef switch behavior, but it is now coded to
always use the -in1ef behavior [
c. Numerical resultdifferences (from rounding errors) due to hardware
and software (compiler and math libraries) [
, https://software.intel.com/en-us/forums/topic/299617 ,
Wien mailing list
SEARCH the MAILING-LIST at: