[Wien] Best RKmax

[Gavin Abo]

A comment:

I'm not familiar with your increase RKmax until the "energy start to 
increase again".

 From what I have seen or read, you want to increase RKmax until the 
"parameter you are interested in has converged".  That parameter might 
be total energy, EEL spectrum intensity, or another parameter.

For example, you might want to refer to Figure 1(b) in [1]. In the 
figure, it can be seen that the total energy decreases rapidly as RKmax 
increases from about 3.5 to 5.5 (i.e., has not converged). However, it 
can be seen that the total energy has converged at about -302.5 Ry for 
RKmax above 5.5.  From this, they choose a suitable value of 7 for 
RKmax.  I suspect because RKmax of say 6 was too close to the transition 
in energy (between non-convergence and convergence) and the slight 
increase in accuracy of the energy for RKmax greater than 7 looks like 
it was not worth the additional time needed to perform the calculations 
(i.e., the shown increase in "time per scf cycle").

[1] R. J. Nicholls and A. J. Scott, Practical approaches to the accurate 
modelling of EELS edges using density functional theory, J. Phys.: Conf. 
Ser. 126, 012038 (2008). http://dx.doi.org/10.1088/1742-6596/126/1/012038

On 3/6/2015 6:43 AM, Hannan.Sadek at uv.es wrote:
> To determine the suitable size of the basis set for the case that I'm
> calculating.
>
> Yes, you can increase RKmax as far as you want.
>
> The question is:  why would one want to do that ?
>
> On 03/06/2015 10:10 AM, Hannan.Sadek at uv.es wrote:
>> Hello
>>
>> I'm using the last version of WIEN2k. And it is explained in the user
>> guide that the best value of RKmax can be obtained by starting from the
>> recommended minimal value and then increasing it in steps of eg. 0.5
>> until we get the point at which the energy start to increase again as a
>> function of the used RKmax values. Also we have to note that the value
>> of RKmax cannot be set arbitrarily large and can't be lager than 11. So
>> my question is if I reached this value(Rkmax=11) and still don't have
>> the expected behavior of the energy(the energy doesn't start to increase
>> again)? can I use values larger than 11 in that case?
>>
>> Best regards
>> Hannan