[QE-users] Density Guess and wfc guess

[Lorenzo Paulatto]

> 
> This is the question. From my logic, the best speed up is when use the 
> neutral system wfc set, and the density is constructed with this wfcs, 
> but using the correct number of electrons to the charged system, then 
> this density is the density for the freezed neutral system plus the 
> substraction/addition electrons to charge the system (the orbital density).
> How can i do this? Is this a good idea?
> 
> I propound this, because the 3 cases for starting commands are not the 
> best approach to charged systems.
> 

If you do the initial and the second calculation with the same number of 
bands (set nbnd by hand) the approach could work, although I'm not sure 
what the code would do in practice.

However, from my experience, the speedup of this kind of tricks is 
usually negligible with respect to the human time spent setting thing 
up. The reason is that convergence is usually exponential i.e. at each 
iteration the accuracy is multiplied by a factor (hopefully <1) which is 
more or less constant. Even if you start with a better guess, you may 
only save a couple iterations

hth

-- 
Lorenzo Paulatto - Paris