[Wien] Error in spaghetti (irrep?)

[Peter Blaha]

TOLDG of 0.5E-3 is meaningless.

Degenerate eigenvalues should be degenerate to about 2.E-8, thus maybe my
recommendation to increase TOLDG slightly.

A few more points: ONLY at Gamma and A (maybe I missed some others) there
are 2-dimensional representations (degenerate eigenvalues). At your knum=7
all eigenvalues are non-degenerate and do not need to be identical.
At Gamma you have eg:
                   
     -0.0384385   -0.0275091   -0.0275090   -0.0252905   -0.0242101
     -0.0242100   -0.0212862   -0.0212862   -0.0190025   -0.0186596

2+3, 5+6 and 7+8 should most likely be degenerate, but only 7+8 are.
TOLDG should be set such that also 2+3 are "degenerate".  

In order to test spaghetti, I need case.struct, case.output1up, 
case.outputirrup,case.irrepup and case.insp

Please send some "non-parallel" test.

> Increasing TOLDG up to 0.5E-3, the number of errors in spaghetti are reduced
> systematically. However, even for TOLDG =0.5E-3, all errors do not disapperar.
> Is this a reasonable value for TOLDG?
> 
> For each k-point at which an error occurs in the new plot, there are two
> eigenvalues close together which 'float' alone and are not 'joined' to other
> eigenvalues in the spaghetti plot.
> In outputir, these points are skipped (they correspond to the incorrect points
> in spaghetti):
> Spin up:   knum=7, bnd 17,   knum=14, bnd 17,   in outputirup_1
> Spin dn:   knum=6, bnd 17,   knum=9, bnd 17,    in  outputirdn_1  and
> knum=10, bnd17   in outputirdn_4
> Other bands have =?? (eg SpinUP band 44 of knum 7 and band 48 of knum 14) most
> likely due to the earlier point's problem.
> 
> For knum=7 (eigenvalues below) band 17 (-0.9725160) differs by 0.0004 from its
> previous eigenvalue, while band 44 (-0.0969005) differs by -0.0000012 which
> seems strange with TOLDG=0.5E-3?
> All the other differences are of similar size and smaller than 0.5E-3. For the
> spin down point knum=10, bnd 17 in outputirdn_4, the difference is 0.4E-6.
> 
>     K=   0.15000   0.00000   0.00000             (knum=7)
>     -3.8395122   -2.1050534   -2.1050530   -2.1023162   -1.0487342
>     -1.0139718   -1.0130222   -1.0122362   -1.0075969   -0.9951106
>     -0.9878561   -0.9783544   -0.9770046   -0.9769033   -0.9748384
>     -0.9729166  *-0.9725160*  -0.9616254   -0.9609163   -0.9583208
>     -0.1549498   -0.1527483   -0.1464105   -0.1450850   -0.1439287
>     -0.1379788   -0.1367145   -0.1312673   -0.1306861   -0.1283666
>     -0.1260235   -0.1256493   -0.1245272   -0.1151976   -0.1142609
>     -0.1137208   -0.1116372   -0.1089792   -0.1078344   -0.1063183
> 
>     -0.1057823   -0.1011095   -0.1005828  *-0.0969005*  -0.0968993
>     -0.0921372   -0.0910378   -0.0881868   -0.0881706   -0.0838515
> 
> I am sending the new spaghetti files (for TOLDG=0.5E-3) and output files
> directly to your e-mail.


                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------