[Wien] question about tetra's choice of Emax
pieper
pieper at ifp.tuwien.ac.at
Thu Nov 17 13:47:43 CET 2016
A thought when reading your description of the cases that might give
troubles:
> Of course it could simply be that you have so steep bands that one
> does not catch an empty band within 1.5+EF Ry. In this case one has to
> increase EMAX in case.in1. And yes, if this really happens, one should
> give a :WAR ...(and I don't think this is there yet ..., but I would
> need a case where this happens).
It seems in hindsight I of all people should have been able to point you
in the direction of ferromagnetic metals for examples where this might
happen. A broad band (4s) and a narrow one (4d) at E_F are Herring's
(1966) basic ingredients for itinerant magnetism.
You have small exchange in the broad 4s-band and a large one in the
narrow 4d-band. One might (correctly) expect Cu with 11 electrons for
these bands to be a happy paramagnet: the spins of 10 e in the d-band
compensate, and the exchange for the one in the 4s is too small to
redistribute anything from there, 4s is spin balanced. Furthermore it
costs too much energy to increase the 4s population above half full, the
exchange gain in an unbalanced 4d is not sufficient. For Ni with 10
electrons available the cost to increase the 4s population from zero is
smaller. The 4d exchange is large enough to gain energy from an
unbalanced arrangement in an unfilled 4d-band and put 0.6 e into 4s. 4s
remains balanced as in Cu, but exchange puts 5 of the 9.4 4d-electrons
up, 4.4 down, resulting in the observed 0.6 muB. The same argument works
for Co, and to my knowledge for Fe, though its less accurate there, and
it breaks down for Mn. So maybe fcc-Fe and fcc-Ni are other examples - I
didn't try.
Thanks again,
Martin
---
Dr. Martin Pieper
Karl-Franzens University
Institute of Physics
Universitätsplatz 5
A-8010 Graz
Austria
Tel.: +43-(0)316-380-8564
Am 17.11.2016 07:46, schrieb Peter Blaha:
> Well, besides the :WAR nings, this is a good example for the problem.
>
> Apparently, at Gamma there is a huge "gap" and you find only 9
> eigenvalues (bands) up to 2.1 Ry (while on other k-points there are up
> to 12 eigenvalues.
> Thus, even the output in scf2 is "wrong", the band-range maximum of
> band #10 is not 2.08 but will be "above" 2.1 (1.5+EF).
>
> The TETRA method always integrates bandwise, ie. each band
> individually and I need an energy at EVERY k-point to do this
> integration. In lapw2 there is a crude fix for "missing eigenvalues"
> (it creates an arbitrarily large artificial eigenvalue), but in TETRA
> it is not done.
>
> It might be save to calculate the "DOS" (and its integral) to some
> higher energy, but it is hard to estimate up to which value it is
> still save, since I would have to know the band dispersion. It could
> be that (in your case ?) only near Gamma a few k-points miss the 10th
> band, but in principle it could also be opposite. Since a reliable
> E-limit ("1.5-1.9 ?? in your case") is hard to estimate, I've chosen
> the "save" way.
>
>
> Am 17.11.2016 um 00:55 schrieb pieper:
>> OK, my mistake: I didn't check the scf for :WARN - which I should have
>> done, at least after running into the problem.
>>
>> Just for the record a brief account what happened:
>>
>> ---------- Cofcc.struct --------
>> fcc-Co
>> F LATTICE,NONEQUIV.ATOMS: 1
>> MODE OF CALC=RELA unit=ang
>> 6.637022 6.637022 6.637022 90.000000 90.000000 90.000000
>> ATOM 1: X=0.00000000 Y=0.00000000 Z=0.00000000
>> MULT= 1 ISPLIT= 2
>> Co NPT= 781 R0=0.00005000 RMT= 2.3300 Z: 27.0
>> LOCAL ROT MATRIX: 1.0000000 0.0000000 0.0000000
>> 0.0000000 1.0000000 0.0000000
>> 0.0000000 0.0000000 1.0000000
>> 48 NUMBER OF SYMMETRY OPERATIONS
>> .
>> .
>> .
>> (list of sym ops)
>> -------------------------------
>>
>> Initialization via w2web, quick version, spin polarized marked, 10000
>> k
>> vectors entered (probably not significant), everything else left open
>> (defaults). Results without any complaints in
>>
>> --------- Cofcc.in1 -----------
>> WFFIL EF=.600081718450 (WFFIL, WFPRI, ENFIL, SUPWF)
>> 7.00 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)
>> 1 0.30 0.000 CONT 1
>> 1 -4.57 0.001 STOP 1
>> 2 0.30 0.005 CONT 1
>> 0 0.30 0.000 CONT 1
>> K-VECTORS FROM UNIT:4 -9.0 1.5 35 emin / de (emax=Ef+de)
>> /
>> nband
>> ------------------------------
>>
>> scf started in w2web without any changes in the menu. It converges in
>> 11
>> cycles.
>>
>> When I ploted the DOS from w2web I ran into the problem of Emax.lt.E_F
>>
>> Looking into the scf's is a good idea:
>>
>> --------- Cofcc.scf1up --------
>>
>> ATOMIC SPHERE DEPENDENT PARAMETERS FOR ATOM Co
>> :e__0001: OVERALL ENERGY PARAMETER IS 0.4003
>> OVERALL BASIS SET ON ATOM IS LAPW
>> :E1_0001: E( 1)= 0.4003
>> APW+lo
>> :E1_0001: E( 1)= -3.7752 E(BOTTOM)= -3.806 E(TOP)= -3.744 1
>> 2 175
>> LOCAL ORBITAL
>> :E2_0001: E( 2)= 0.3944 E(BOTTOM)= 0.168 E(TOP)= 0.621 0
>> 1 124
>> APW+lo
>> :E0_0001: E( 0)= 0.4003
>> APW+lo
>>
>> K= 0.00000 0.00000 0.00000 1
>> :RKM : MATRIX SIZE 39LOs: 12 RKM= 6.24 WEIGHT= 1.00 PGR:
>> EIGENVALUES ARE:
>> :EIG00001: -3.7586332 -3.7586332 -3.7586332 -0.0750175
>> 0.3973462
>> :EIG00006: 0.3973462 0.3973462 0.4756761 0.4756761
>> ********************************************************
>>
>> :KPT : NUMBER OF K-POINTS: 286
>> ----------------------------------------------------------
>>
>> and
>>
>> ------------Cofcc.scf2up --------------------------------
>> :GAP : -99999. Ry = -9999. eV ( metallic )
>> Bandranges (emin - emax) and occupancy:
>> :BAN00001: 1 -3.782207 -3.758633 1.00000000
>> :BAN00002: 2 -3.774241 -3.758633 1.00000000
>> :BAN00003: 3 -3.767467 -3.758633 1.00000000
>> :BAN00004: 4 -0.075018 0.308043 1.00000000
>> :BAN00005: 5 0.252261 0.400466 1.00000000
>> :BAN00006: 6 0.341921 0.531620 1.00000000
>> :BAN00007: 7 0.397346 0.559086 1.00000000
>> :BAN00008: 8 0.439501 0.560197 1.00000000
>> :BAN00009: 9 0.472444 1.130465 0.32393899
>> :BAN00010: 10 1.052748 2.086705 0.00000000
>> :BAN00011: 11 1.362791 2.100100 0.00000000
>> :BAN00012: 12 1.460431 2.100017 0.00000000
>> Energy to separate low and high energystates: -0.12502
>>
>>
>> :NOE : NUMBER OF ELECTRONS = 15.000
>>
>> :FER : F E R M I - ENERGY(TETRAH.M.)= 0.6000786246
>> :GMA : POTENTIAL AND CHARGE CUT-OFF 12.00 Ry**.5
>>
>>
>>
>>
>> :POS001: ATOM 1 X,Y,Z = 0.00000 0.00000 0.00000 MULT= 1 ZZ=
>> 27.000 Co
>>
>> LMMAX 5
>> LM= 0 0 4 0 4 4 6 0 6 4
>>
>> :CHA001: TOTAL VALENCE CHARGE INSIDE SPHERE 1 = 7.9401 (RMT=
>> 2.3300 )
>> :PCS001: PARTIAL CHARGES SPHERE = 1 S,P,D,F, D-EG,D-T2G
>> :QTL001: 0.2277 3.2151 4.4719 0.0180 0.0000 0.0000 0.0000 1.8125
>> 2.6594
>> 0.0000 0.0000 0.0000
>> Q-s-low E-s-low Q-p-low E-p-low Q-d-low E-d-low Q-f-low
>> E-f-low
>> :EPL001: 0.0001 -3.7727 2.9953 -3.7686 0.0004 -3.7693 0.0002
>> -3.7677
>> Q-s-hi E-s-hi Q-p-hi E-p-hi Q-d-hi E-d-hi Q-f-hi
>> E-f-hi
>> :EPH001: 0.2275 0.1987 0.2198 0.3397 4.4715 0.4145 0.0179
>> 0.4072
>>
>> :CHA : TOTAL VALENCE CHARGE INSIDE UNIT CELL = 8.323939
>>
>> :SUM : SUM OF EIGENVALUES = -9.209154185
>>
>>
>>
>> QTL-B VALUE .EQ. 22.75945 in Band of energy 1.99144 ATOM= 1
>> L= 2
>> Check for ghostbands or EIGENVALUES BELOW XX messages
>> Adjust your Energy-parameters for this ATOM and L (or use -in1new
>> switch), check RMTs !!!
>>
>>
>> NBAND in QTL-file: 9
>> ----------------------------------------------------------------------------
>>
>>
>> So yes, there are unoccupied bands above no.9, but there is a qtl
>> problem with single atom of the structure ...
>>
>> grep for :WAR in the scf yields the warnings one clearly should not
>> ignore:
>>
>> :WARN : You should change the E-parameter for this atom and L-value in
>> case.in1 (or try the -in1new
>> :WARN : QTL-B value eq. 3.98 in Band of energy 0.64376 ATOM= 1
>> L= 2
>> :WARN : QTL-B value eq. 2.97 in Band of energy 0.29294 ATOM= 1
>> L= 2
>>
>>
>> ---
>> Dr. Martin Pieper
>> Karl-Franzens University
>> Institute of Physics
>> Universitätsplatz 5
>> A-8010 Graz
>> Austria
>> Tel.: +43-(0)316-380-8564
>>
>>
>> Am 16.11.2016 16:06, schrieb Peter Blaha:
>>> The question is: what does your scf2up/dn files (and the scf1up/dn
>>> files say ?
>>>
>>> In case.in1 there should be an EMAX of 1.5 ?
>>>
>>> and in scf1up/dn there should be eigenvalues up to 1.5+EF ??
>>>
>>> and in scf2up/dn there is a list of "bandranges". Are there "empty"
>>> band ranges (band which have zero occupation) ??
>>>
>>> Of course it could simply be that you have so steep bands that one
>>> does not catch an empty band within 1.5+EF Ry. In this case one has
>>> to
>>> increase EMAX in case.in1. And yes, if this really happens, one
>>> should
>>> give a :WAR ...(and I don't think this is there yet ..., but I would
>>> need a case where this happens).
>>>
>>> Regards
>>>
>>> On 11/16/2016 12:25 PM, pieper wrote:
>>>> Thanks for the quick response!
>>>>
>>>> Sorry, I did not fully catch the problem of consistency between what
>>>> is
>>>> calculated in lapw1 and integrated in tetra.
>>>>
>>>> However, your PS and PPS leave me worried about where things go
>>>> sideways
>>>> with something as simple as fcc Co should be.
>>>>
>>>> As for the version, my VERSION file says
>>>> WIEN2k_14.2 (Release 15/10/2014)
>>>> which to my knowledge is not too bad? It does not include the
>>>> dynamical
>>>> Emax in lapw1?
>>>>
>>>> With that version, kgen 10000 vectors, spin-resolved, and everything
>>>> else default in w2web, I get from tetra something like the file
>>>> below.
>>>> Emax is apparently changed to a value below E_F.
>>>>
>>>> I understand that this might lead to problems already in the scf
>>>> because
>>>> the wrong states are occupied, but in that case again I cannot rely
>>>> on
>>>> the DOS, even when I increase Emax in .in1 and run lapw1 ... I have
>>>> to
>>>> re-run the scf, correct?
>>>>
>>>>
>>>> ----- Co-fcc.outputtup ------
>>>> Co-fcc #
>>>>
>>>> IAV : 0
>>>> NPRINT : 1
>>>> 2 CASES FOR DOS : ATOM L
>>>>
>>>> cc-Co
>>>> LATTICE CONST.= 6.79300 6.79300 6.79300 FERMI ENERGY= 0.56702
>>>> 48 <; NMAT <; 63 SPIN=2 NATO= 2
>>>> JATOM 1 MULT= 1 ISPLIT= 2 tot,0,1,2,D-eg,D-t2g,3
>>>> CASE 1 : ATOM NUMBER 0 COLUMN READ 0 DOSTYPE=total-DOS
>>>> CASE 2 : ATOM NUMBER 1 COLUMN READ 1 DOSTYPE= 1:total
>>>> We will add 0 DOS-cases together:
>>>> BAND LIMITS OF BAND 1 ARE -3.81359 -3.79544
>>>> BAND LIMITS OF BAND 2 ARE -3.80710 -3.79544
>>>> BAND LIMITS OF BAND 3 ARE -3.80235 -3.79544
>>>> BAND LIMITS OF BAND 4 ARE -0.07833 0.28402
>>>> BAND LIMITS OF BAND 5 ARE 0.24033 0.37076
>>>> BAND LIMITS OF BAND 6 ARE 0.32045 0.48759
>>>> BAND LIMITS OF BAND 7 ARE 0.37076 0.51438
>>>> BAND LIMITS OF BAND 8 ARE 0.40511 0.51502
>>>> BAND LIMITS OF BAND 9 ARE 0.43442 1.04422
>>>> EMAX reduced due to lower HIGHEST BAND-minimum
>>>> EMIN, DE, EMAX : -0.50000 0.00200 0.43442
>>>>
>>>> EMIN= -0.50000 EMAX= 0.43442 EFACTR= 499.99996948 ESTEP =
>>>> 0.00200
>>>> ENERGY BAND 1 THROUGH 9 ENERGY CHANNEL: 1 TO 468
>>>> NUMBER OF K-POINTS: 165
>>>> NUMBER OF TETRAHEDRONS: 693
>>>> # BAND 9
>>>> #EF= 0.56702 NDOS= 2 NENRG= 468 Gaussian bradening:
>>>> 0.00300
>>>> NUMBER OF ELECTRONS UP TO EF : 0.0000
>>>>
>>>> DOS in states/Ry/spin
>>>> smearing 1 0.398942274850643 2.00000000000000
>>>> smearing 1 0.398942274850643 2.00000000000000
>>>> # ENERGY 0 total-DOS 1 1:total
>>>> -0.50000 0.00 0.0000 0.00 0.0000
>>>> -0.49800 0.00 0.0000 0.00 0.0000
>>>> -0.49600 0.00 0.0000 0.00 0.0000
>>>> .
>>>> .
>>>> .
>>>> ------------------------------------------------------------------
>>>>
>>>>
>>>>
>>>> ---
>>>> Dr. Martin Pieper
>>>> Karl-Franzens University
>>>> Institute of Physics
>>>> Universitätsplatz 5
>>>> A-8010 Graz
>>>> Austria
>>>> Tel.: +43-(0)316-380-8564
>>>>
>>>>
>>>> Am 16.11.2016 07:41, schrieb Peter Blaha:
>>>>> The DOS is calculated up to a value for which we can guarantee that
>>>>> the DOS is correct and complete.
>>>>>
>>>>> Of course, the DOS up to the highest band-maximum would be
>>>>> non-zero,
>>>>> but there is a chance that some (maybe a lot) of DOS is missing and
>>>>> a
>>>>> user would not notice this and "misintewrprete" this wrong DOS.
>>>>>
>>>>> Now he can trust that the calculated DOS is ok, and if he needs
>>>>> higher
>>>>> DOS, he has to increase the emax in case.in1 and/or in case.int.
>>>>>
>>>>> PS: You are probably using an older WIEN2k version, because now we
>>>>> use
>>>>> a "dynamical" Emax in case.in1, which takes the actual EF into
>>>>> account. But of course in cases of very steep bands above EF, the
>>>>> default in1 file may still be insufficient.
>>>>>
>>>>> PPS: In your situation it could even be, that the scf calculation
>>>>> is
>>>>> "wrong", since you occupied the wrong bands ....
>>>>>
>>>>> Am 15.11.2016 um 19:12 schrieb pieper:
>>>>>> Hello, mailing list,
>>>>>>
>>>>>> yesterday I had for the first time some dispute with the way tetra
>>>>>> automatically chooses its input energy range. I would like to
>>>>>> understand
>>>>>> why the particular automatic choice of Emax was introduced.
>>>>>>
>>>>>> Until then it worked so well that I didn't even notice that Emax
>>>>>> is
>>>>>> automatically adjusted, but then I wanted to illustrate Wien2k by
>>>>>> calculating the example fcc Ni. I took Co instead of Ni, lattice
>>>>>> constant adjusted to 6.637, RMT to touching spheres. I used w2web
>>>>>> for
>>>>>> initialiazation and initiating scf, kgen with 10000 vectors,
>>>>>> spin-polarized in an FM starting configuration, no spin-orbit
>>>>>> coupling.
>>>>>>
>>>>>> Then I chose DOS from the Tasks menu, skipped the optional steps
>>>>>> lapw1
>>>>>> and qtl, used lapw2 -qtl, configured .int to calculate just the
>>>>>> total
>>>>>> DOS (the default). When I proudly presented the result of dosplot
>>>>>> to a
>>>>>> visitor the plot ended below the Fermi energy ...
>>>>>>
>>>>>> Increasing Emax in .int did nothing, as the experts probably could
>>>>>> have
>>>>>> told me beforehand. The (in this case in my view annoying)
>>>>>> automatic
>>>>>> choice of Emax that kicks in is indicated in the header of
>>>>>> .outputtup
>>>>>> (as well as in the user guide - and yes, I know one should read
>>>>>> it):
>>>>>>
>>>>>> .
>>>>>> .
>>>>>> .
>>>>>> BAND LIMITS OF BAND 8 ARE 0.40511 0.51502
>>>>>> BAND LIMITS OF BAND 9 ARE 0.43442 1.04422
>>>>>> EMAX reduced due to lower HIGHEST BAND-minimum
>>>>>> EMIN, DE, EMAX : -0.50000 0.00200 0.43442
>>>>>>
>>>>>> The problem in this case is that the 'lower highest Band-minimum'
>>>>>> is
>>>>>> BELOW the Fermi-energy:
>>>>>>
>>>>>> ******** EF and DOS at fermi level *******
>>>>>> 0.56702 0.00 0.00
>>>>>>
>>>>>> So one MUST go back to the (in principle optional) lapw1 step with
>>>>>> some
>>>>>> larger Emax in .in1 (as indicated in the DOS menu and in the UG).
>>>>>> Increasing Emax there appears to me a little clumsy since one has
>>>>>> to
>>>>>> guess an Emax that will generate a band with a band minimum above
>>>>>> E_F.
>>>>>>
>>>>>> Why not have tetra choose Emax as the minimum of the Emax input in
>>>>>> .int
>>>>>> and the highest Band-MAXIMUM?
>>>>>>
>>>>>>
>>>>>> ---
>>>>>> Dr. Martin Pieper
>>>>>> Karl-Franzens University
>>>>>> Institute of Physics
>>>>>> Universitätsplatz 5
>>>>>> A-8010 Graz
>>>>>> Austria
>>>>>> Tel.: +43-(0)316-380-8564
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> 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
>>>> _______________________________________________
>>>> 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
>> _______________________________________________
>> 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
More information about the Wien
mailing list