[Wien] Ghostbands: pushed energy range in case.in1 to 6.3, does this mean there is a problem?

Peter Blaha pblaha at theochem.tuwien.ac.at
Sat Mar 7 08:24:09 CET 2015


There is a misunderstanding: No, don't take a "mean" energy.

Check also the corresponding charge. When it is large, it is a major
component and needs an energy parameter close to this energy.

>> In case.scf2 you can find under the line :EPH and :EPL
>> the "mean" energy of the P-s states. If they are not close to   -0.73
>> (thats where you expand P-s), change the corresponding input value.
>
> For P-s, :EPL and :EPH are -1.34 and -0.43, mean of -0.89, fairly close to
> -0.73?

Together with the large P-s charge in :EPL it tells you, that you should
lower the P-s parameter in case.in1 to -1.34
Whether one sets in addition a second l=0 Eparameter in case.in1 depends on
the E-separation between these EPL and EPH values, corresponding charges
and the sphere radii (the larger the spheres: more probably yes).
Here you have 0.9 Ry difference, but presumably the P-s charge in the upper E-window is
very small and you have a very small spheres. Setting the two
energies to those values might lead to ghostbands and at least at the beginning you
moved one E-parameter up to +6 Ry. As I said previously, you may test it at the end
and set the second Al-s line to 0.30 (no search), so that the actual E-parameter will be EF+0.2

> Other than P-s they are not close.  Al-s has -7.24, -0.34, mean of -3.79,
> case.in1 has -7.65.
> O-s has -1.21, -0.30, mean is -0.75, while case.in1 has -1.46.  Similarly
> for Al-p and P-p.

Again, there should be l=0 E-parameters close to -7.24 and 1.21, respectively.
In addition there should be definitely a second Al-s line at 0.30, since there are
"real" Al 3s states in the valence region.

> I had mostly read about a supercell with one full core hole.  Some of these
> are certainly cells where I do not want to build a larger supercell than I
> have to.  Is the HALF a core hole a better choice?

Are you interested in XPS or in XAS. This is a VERY different process where the
excited electron leave the bulk or stays whithin the bulk.

For XPS you are interested just in ONE number (Al-2p ionizationpotential) and
Slaters transition state concept with  half a core hole applies.

For XAS you want to simulate a spectrum for a system with one core hole and an additional e- in
the valence band. Use (at least for insulators) a full core hole.
>
> Do I understand correctly that whether I use a HALF core-hole or a full one,
> I then do minimization of the ionic positions again?

No. Electronic spectroscopy is a very fast process and the ions have no time to move
around.

>
> The user-guide says "The energy cut-off specified in lstart during init lapw
> (usually -6.0 Ry) defines the separation
> into core- and band-states (the latter contain both, semicore and valence)."
> How do I get the Al 2p state into the core?
> Do I have to change the cut-off and use .lcore, or is there some way to move
> just the Al 2p state into the core?

Besides an Energy (-X Ry), you can also specify a charge localization criterium
(like 0.999), which will put all states with less charge inside sphere as valence.
Checkout case.outputst to see how much charge each state has inside sphere:
           E-up(Ry)      E-dn(Ry)   Occupancy   q/sphere  core-state
   1S      -3.801989     -3.785331  1.00  1.00    0.9922  F
   2S      -0.236724     -0.003329  1.00  0.00    0.0675  F


>
> PS There was no "reply" button in the archive except "Reply via email".  I
> could not find an answer as to how to reply to a post in either the Mail
> Archive FAQ, or the WIEN mail archive.
>
> Thanks,
>    David
>
> ---------------------------------------------------------------------------
> At Thu, 05 Mar 2015 22:41:38 -0800, Peter Blaha wrote
> I think you have solved the problem very well.
>
> Due to the small P sphere and the fact, that P-s states are relatively
> high in energy, the two linearization energies must be quite well
> separated. (An alternative would have been to simply remove the second
> l=0 line for P and change to "3" lines only:
>     0.30    3  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
>    1    0.30      0.000 CONT 1
>    1   -8.83      0.001 STOP 1
>    0   -0.73      0.002 CONT 1
>
> Two more checks towards the "end of the scf cycle":
> In case.scf2 you can find under the line :EPH and :EPL
> the "mean" energy of the P-s states. If they are not close to   -0.73
> (thats where you expand P-s), change the corresponding input value.
>
> If the energy of the P-s states has gone down in energy at the end of the
> scf-cycles, you may
> checkout if you can go down with this second E-s input line from +6 back to
> 2.0
> or even
> back to 0.3  (sometimes such problems are temporary).
>
> PS: If you are interested in Al-2p XPS you should do Slaters transition
> state !
> Put Al 2p into the core and introduce HALF a core hole (compensated by a
> background).
> This gives much better core-eigenvalues that the plain DFT groundstate
> eigenvalues,
> typically lt. 1 % error as compared to 10 % error in comparison with
> experiment.
> In addition, final state screening effects are better accounted for.
>
>
> Am 06.03.2015 um 00:44 schrieb David Olmsted:
>
>
> Ghostbands: pushed energy range in case.in1 to 6.3, does this mean there is
> a problem?
>
> WIEN2k_14.2 (Release 15/10/2014)
> Quad-Core AMD Opteron(tm) Processor 2378
> Linux cluster
> Intel 11.1 compilers with mkl.
>
> The purpose of my computation is to compare predicted XPS spectra for Al 2p
> electon
> for different environments of the Al atom in the Al-P-O-H system.
>
> User: beginner!  My first time using WEIN2k.  Moderate amount of VASP work.
>
> Issue: ghostbands
>
> GGA-PBE, 48 atoms, K-mesh 6x6x4, no shift.  Not spin-polarized.
> Initial cell and positions from relaxed GGA-PBE using VASP, same K-mesh.
>
> RMT from w2web StructGen (3% reduction)
>     H  0.63
>     O  1.17
>     P  1.34
>     Al 1.72
>
> RKmax 3.5 to get "effective RKmax" of 6.5 for O.
>
> rmt(min)*kmax =    3.50000
> gmin =   11.11111
> gmax =   20.00000
>
> ------- metavar_v.in0
> TOT  XC_PBE     (XC_LDA,XC_PBESOL,XC_WC,XC_MBJ,XC_REVTPSS)
>
> NR2V      IFFT      (R2V)
>     64 120 108    1.00  1    min IFFT-parameters, enhancement factor, iprint
> ---------
>
> For default -6 Ry cutoff for core states, charge was leaking out of RMT
> sphere
> for P 2p states.  Final iteration in metavar_v.outputst:
>             14         350
>      14   1.85E-07    1.884765E+00   -8.645384E-01   -8.645386E-01   1.72E-07
> -1.67E-08    1.707034E-01    1.707034E-01
>     1S      -153.17082     -153.17082
>     2S       -12.78682      -12.78682
>     2P*       -9.19366       -9.19366
>     2P        -9.12626       -9.12626
>     3S        -1.02668       -1.02668
>     3P*       -0.40735       -0.40735
>     3P        -0.40342       -0.40342
>
> Cutoff set to -9.2 Ry.  (Also tried leaving it at -6.0 Ry and touching
> .lcore.  Similar results.)
>
> ===============================================================
> ---------------------------- Question -------------------------
> ===============================================================
>
> With the original case.in1 file, had messages for the P atom, L=0:
> (All these messages are from the first run of LAPW2.)
>
> metavar_v.scf2_1:   QTL-B VALUE .EQ. 4951.54243 in Band of energy  -6.46139
> ATOM=    2  L=  0
>
> increased 0.3 to 2.3 in case.in1, now:
>
>      QTL-B VALUE .EQ. 1347.97207 in Band of energy  -4.71553  ATOM=    2  L=
> 0
>
> increased it 4.3
>
>      QTL-B VALUE .EQ.  602.53449 in Band of energy  -2.14697  ATOM=    2  L=
> 0
>
> When I increased it to 6.3, no complaints.
>
> The initial scf run has completed with no warnings; the position
> minimization is still running.
>
> In the mailing list search, there are suggestions to increase the (upper)
> energy range to
> 1.3 or "even 2.0" Ry.  That makes me worry about the fact that I had to
> increase it to a much
> larger value.  Does this mean something is going wrong?
>
> -------------------------  End of question ------------------------
> =========== case.in1 =============================================
> WFFIL  EF=.1268392143   (WFFIL, WFPRI, ENFIL, SUPWF)
>     3.5       10    4 (R-MT*K-MAX; MAX L IN WF, V-NMT
>     0.30    4  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    0    0.30      0.000 CONT 1
>    0   -7.65      0.001 STOP 1
>    1    0.30      0.000 CONT 1
>    1   -4.81      0.001 STOP 1
>     0.30    4  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    1    0.30      0.000 CONT 1
>    1   -8.83      0.001 STOP 1
>    0   -0.73      0.002 CONT 1
>    0    6.30      0.000 CONT 1
>     0.30    3  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    0   -1.46      0.002 CONT 1
>    0    0.30      0.000 CONT 1
>    1    0.30      0.000 CONT 1
>     0.30    3  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    0   -1.46      0.002 CONT 1
>    0    0.30      0.000 CONT 1
>    1    0.30      0.000 CONT 1
>     0.30    3  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    0   -1.46      0.002 CONT 1
>    0    0.30      0.000 CONT 1
>    1    0.30      0.000 CONT 1
>     0.30    3  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    0   -1.46      0.002 CONT 1
>    0    0.30      0.000 CONT 1
>    1    0.30      0.000 CONT 1
>     0.30    3  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    0   -1.46      0.002 CONT 1
>    0    0.30      0.000 CONT 1
>    1    0.30      0.000 CONT 1
>     0.30    3  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    0   -1.46      0.002 CONT 1
>    0    0.30      0.000 CONT 1
>    1    0.30      0.000 CONT 1
>     0.30    1  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    0    0.30      0.000 CONT 1
>     0.30    1  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    0    0.30      0.000 CONT 1
>     0.30    1  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    0    0.30      0.000 CONT 1
>     0.30    1  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global
> APW/LAPW)
>    0    0.30      0.000 CONT 1
> K-VECTORS FROM UNIT:4  -12.2       1.5   250   emin / de (emax=Ef+de) /
> nband #red
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>

-- 
-----------------------------------------
Peter Blaha
Inst. Materials Chemistry, TU Vienna
Getreidemarkt 9, A-1060 Vienna, Austria
Tel: +43-1-5880115671
Fax: +43-1-5880115698
email: pblaha at theochem.tuwien.ac.at
-----------------------------------------


More information about the Wien mailing list