[Pw_forum] Pseudopotential troubles

[Miguel Martínez Canales]

Dear people,

I'm interested in studying calcium at high pressure, close to 1Mbar. For 
that, I downloaded the available pseudopotentials in the web and did a few 
quick convergence tests.

I found that the LDA pseudopotential had a notable energy jump (4 mRy) when 
going from a cutoff of 150 rydbergs to 200. I wasn't expecting this, due to 
a not so small r_c. Could be my fault, though. The pbe pseudopotential 
seems to predict a too small fcc->bcc transition pressure. What's more, 
running a phonon calculation of the simple cubic structure (experimentally 
stable, and in my calculations preferred over fcc and bcc) yields some 
instabilities. I assumed this could be related to the pseudopotential, 
though the pbe pseudopotential predicts a nice bandstructure and stable 
phonons (just a quick look) at 0 GPa. Thus, I intended to "cook" my 
pseudopotential.

Here's the catch. I can't produce a Ca pseudopotential. Doing a RRKJ 
pseudization with a 4s or 4p r_c under 2.66 a_0 results in "error: too many 
nodes". Changing rho0 to 0.01 or 0.02 or 0.1 or even 1.0 doesn't solve it. 
Any clues? 4s and 4p nodes are both under 2 bohrs. To use 
Ca.pbe-nsp-van.UPF rcuts I need to force Troulliers-Martins pseudization. I 
can put one of those rcuts under 2.66 by pseudizing that l channel, but I 
want both rcuts small because my interest lies in high pressure. I've noted 
that if I change the grid settings to the Mg example one, I can reduce the 
4p rcut slightly, but have to increase the 4s rcut to 2.68.

Anyway, I have another more profound question. I don't understand why this 
pseudopotential input is terribly flawed (its result is). I'd be glad if 
some of you more knowledgeable people kindly pointed my errors.

  &input
      title='Ca',
      zed=20.0,
      xmin=-6.0,
      dx=0.0200,
      rmax=80,
      rel=1,
      beta=0.3,
      rlderiv=2.7,
      eminld=-4.0,
      emaxld=4.0,
      deld=0.02d0,
      nld=4,
      config='[Ar] 3d0 4s2 4p0',
      iswitch=3,
      dft='PBE',
  /
  &inputp
    pseudotype=3,
    lloc=-1,
    rcloc=2.1,
    tm =.false.,
    nlcc =.true.,
    rho0 =0.01,
    file_pseudopw='Ca.pbe-van.UPF',
    file_screen='ld1ps.screen'
  /
3
3D  3  2  0.00  0.00  2.40  2.40
4S  1  0  2.00  0.00  2.68  2.68
4P  2  1  0.00  0.00  2.59  2.59

Another question I have is how to add additional projectors per orbital, 
such as the ones used in Paolo Gianozzi's pbe pseudopotential.

Thank you very much,

Miguel

PS: I'm also thinking on including an empty 4s orbital. ¿Any comments?

----------------------------------------
  Miguel Martí­nez Canales
      Dto. Física de la Materia Condensada
      UPV/EHU
      Facultad de Ciencia y Tecnología
      Apdo. 644
      48080 Bilbao (Spain)
  Fax:  +34 94 601 3500
  Tlf:  +34 94 601 5437
  ----------------------------------------