ABINIT Discussion Forums

Hi everybody,
I am attempting to generate a PAW file for Gallium including 3s, 3p, 3d, 4s, 4p states. I include one more partial waves for each angular momentum, so that I get 3 partial waves for each angular momentum (I need more than 2 partial waves per angular momentum, since I plan to carry out some GW calculations).
I am facing two kinds of WARNINGs with my simple test on GaAs using abinit:
The first WARNING is:

mkdenpos : WARNING -
Density went < 0 at 9398 points
and was set to 1.00E-14. Lowest was -0.22E+01.
Likely due to too low boxcut or too low ecut for pseudopotential core charge.

I tried a big pawecutdg (70 Ha) to get rid of this, but still I got this message, since I don't know what means these numbers I don't know if I can get rid of this WARNING.
My second WARNING is:

print_ij: WARNING -
The matrix seems to have high value(s) !
( 1 components have a value greater than 50.0).
It can cause instabilities during SCF convergence.
Action: you should check your atomic dataset (psp file)
and look for "high" projector functions...

I know that I should reduce the high of the projector functions but this is not so easy, so I would like to know, if some people have some experiences with that,
how much this WARNING will be relevant for my next calculation (GW), since presently I got this WARNING but the selfconsistency ended to converge?
Thanks a lot in advance for any help.
Best regards
Pierre-Yves

Hello Pierre-Yves,

For warning 1 ... negative density is quite disturbing. You increased pawecutdg - not a bad idea but what's the ecut of your PAW pseudo? Test by calculating the total energy of a plain Ga bulk with increasing ecut values (let's say from 8 to 28 step 2) and look for variations <1mHa or <0.1mHa. If ecut should be set to more than 30, you'll probably have to optimize the pseudo.

For warning 2 ... it is difficult to predict the adverse effect of an unstable condition on your specific GW calculation. At first for the pseudo, choose a radius for augmentation wisely ; check for ghosts in logderiv file and adjust the reference energies. You should have wfn components < 15 and, in order to attain this condition, play with the pseudization scheme (Bloechl = easy, low components, poor ecut, Vanderbilt = higher component, better ecut, polynom2 rrkj ... = risk of high components, best ecut). so if speed is not your goal i.e. you can live with a no-so-optimal ecul, try Bloechl or Vanderbilt; plot the wfnX files to spot high components. Play with the vloc scheme (Bessel = easy but you'll have to reduce the matching radii, TM or ultrasoft = better, small adjustments of the radii). It's an iterative process, keep your mojo at hand: after selecting more efficient schemes, you'll have to (slightly) adjust the matching radii and reference energies and watch the effects on the logderiv/wfn plots. Then estimate ecut and check accuracy with a couple of Abinit test runs.
Keep pawecutdg at about 2.5 X ecut.

Kind regards,

Alain

Dear Alain,
Thanks a lot for your answer,
Actually for the moment I checked my PAW pseudopotential on GaAs only (since this material is much more known than pure Ga), the total energy is converged within 1mHa at a 26 Ha cutoff. First, I was afraid by such a high cutoff, since following the tutorial I thought that above a 20 Ha cutoff the PAW pseudopotential was bad, but from what you said I guess this hold only for PAW including only valence states.
But still I got this WARNING at the beginning of the scf cycle:

mkdenpos : WARNING -
Density went < 0 at 9398 points
and was set to 1.00E-14. Lowest was -0.22E+01.
Likely due to too low boxcut or too low ecut for pseudopotential core charge.

After some steps of the scf it disappears but still it's awkward this negative density. Should I play with the cutoff r_shape or r_core to remove this message?

Now for the second WARNING message, it seems that I should just give a try of a GW computation... I think that now I have more or less a scheme to follow to produce my pseudo.
However I have one last question:
do you know what are the files wfn9 and above when I have only 8 components (3 components for s, 3 components for p and 2 components for d)?

Thanks in advance for your answer,
Best regards.
Pierre-Yves