[CP2K-user] Fermi energy too high?

[Phil G.]

Dear Matt,

thank you for the suggestions and after centering the slab in y-direction 
and turning off the surface dipole correction, the program finally runs and 
I get the result, but there are some error messages such as:

 *** WARNING in pw/ps_wavelet_methods.F:236 :: Density non-zero on the ***
 *** edges of the unit cell: wrong results in WAVELET solver           ***

    89 Broy./Diag. 0.10E+00    2.3     0.00000880     -3020.8753805067  
4.85E-05

  *** SCF run converged in    89 steps ***

I have chosen the cell length of 40 angstroms in Y direction (slab length 
in y-direction is about 27.6 angstroms).
For the LiNbO3 slab consisting of 14 trilayers as stated in the message of 
9th January, I obtain the result of the Fermi energy:

  E_F = 11.174 eV    (in comparison to the -0.8516 eV in case b) )

This is unrealistic...should I have to enlarge the cell length in 
y-direction or should I turn on the dipole correction?

Phil



Am Dienstag, 15. Januar 2019 12:31:22 UTC+1 schrieb Matt W:
>
> Ah, OK. The extended FFT lengths only works with FFTW not with the wavelet 
> FFT.
>
> You do not need such a large vacuum with the wavelet solver as it is 
> genuinely non-periodic. Place the slab in the center and 5A of vacuum 
> either side should be sufficient - allow 10 A either side to get a clear 
> decay to vacuum level(s). You will get two vacuum levels if you have  a 
> dipole. I don't know if wavelet will work with the dipole correction, I'd 
> turn it off to start with.
>
> Matt
>
> On Tuesday, January 15, 2019 at 8:52:35 AM UTC, Phil G. wrote:
>>
>> Hello again,
>>
>> have tried some attempts to start calculation with WAVELET poisson 
>> solver, but all attempts have failed due to following error messages:
>>
>> 1)  the FFT in the x direction is not allowed
>>      n01 dimension         154
>>      (pw/ps_wavelet_util.F:358)
>>
>>      ===== Routine Calling Stack ===== 
>>
>>            13 S_FFT_dimensions
>>            12 RS_z_slice_distribution
>>            11 ps_wavelet_create
>>      the FFT in the x direction is not allowed
>>      n01 dimension         154
>>            10 pw_poisson_rebuild
>>             9 pw_poisson_solve
>>             8 qs_ks_build_kohn_sham_matrix
>>             7 rebuild_ks_matrix
>>      the FFT in the x direction is not allowed
>>      n01 dimension         154
>>      the FFT in the x direction is not allowed
>>      n01 dimension         154
>>      the FFT in the x direction is not allowed
>>      n01 dimension         154
>>      the FFT in the x direction is not allowed
>>      n01 dimension         154
>>             6 qs_ks_update_qs_env
>>             5 scf_env_do_scf_inner_loop
>>             4 scf_env_do_scf
>>             3 qs_energies
>>             2 qs_forces
>>             1 CP2K
>>
>>
>> 2)  after that I turn off the command EXTENDED_FFT_LENGTHS, then:
>>
>>     Index to radix array not found.
>>     (pw/fft_tools.F:293)
>>
>>      ===== Routine Calling Stack ===== 
>>
>>             6 pw_grid_setup
>>             5 pw_env_rebuild
>>             4 qs_env_rebuild_pw_env
>>             3 qs_env_setup
>>             2 qs_init_subsys
>>             1 CP2K
>>
>>
>> That's strange and I don't know what to do. 
>> In my input file there are some info about commands:
>>
>> [...]
>>     SURFACE_DIPOLE_CORRECTION .TRUE.
>>     SURF_DIP_DIR Y
>> [...]
>>     &MGRID
>>       CUTOFF 600
>>       NGRIDS 5
>>       REL_CUTOFF 50
>>     &END MGRID
>> [...]
>>     &POISSON
>>       POISSON_SOLVER WAVELET
>>       PERIODIC XZ
>>     &END POISSON
>> [...]
>>
>> [...]
>>     &CELL
>>       A    5.148      0.0         0.0
>>       B    0.000      100.0      0.0
>>       C    0.0          0.0         8.9166         
>>       PERIODIC XZ
>>     &END CELL
>> [...]
>>
>> Phil
>>
>>
>> Am Donnerstag, 10. Januar 2019 10:56:24 UTC+1 schrieb Phil G.:
>>>
>>> Dear Matt,
>>>
>>> how can find the potential in the vacuum (which type of potential? 
>>> potential energy or electric/electrostatic potential?) ?
>>> For the case of electric/electrostatic potential, there is a flat curve 
>>> with a step near the vacuum center as a consequence of dipole correction in 
>>> Z direction, while in the bulk slab there is a periodic curve. 
>>>
>>> I will try to use the wavelet solver with PERIODIC XY.
>>>
>>> Phil
>>>
>>> Am Mittwoch, 9. Januar 2019 14:05:04 UTC+1 schrieb Matt W:
>>>>
>>>> Hello again,
>>>>
>>>> did you find the potential in the vacuum and align to that? You need to 
>>>> set a reference to get absolute values.
>>>>
>>>> You could also try using the wavelet solver 
>>>>
>>>> &POISSON
>>>>    PSOLVER WAVELET
>>>>    PERIODIC XZ
>>>> &END
>>>>
>>>> and PERIODIC XZ  in the &CELL section. The Y direction must be the 
>>>> non-periodic one. That gives an absolute reference (if there is no dipole 
>>>> in the cell otherwise you need the  dipole correction switched on).
>>>>
>>>> Matt
>>>>
>>>> On Wednesday, January 9, 2019 at 8:18:58 AM UTC, Phil G. wrote:
>>>>>
>>>>> Dear Matt,
>>>>>
>>>>> thank you for your reply and good suggestions. Now I have let 
>>>>> different LiNbO3 slab systems to be calculated:
>>>>>
>>>>> a) 14 trilayer system as from Sanna et al., *Appl. Surf. Sci.* *301* 
>>>>> (2014), 70-78 with Nb-O3-Li2 surface termination on the one side of the 
>>>>> slab and Li-O surface termination on the other side. Vacuum space of at 
>>>>> least 40 Angstroms was included. The bulk region was already 
>>>>> geometry-optimized and bulk atoms were fixed in the inner 6 trilayers. 
>>>>> Geometry optimization on the whole slab system was performed and then the 
>>>>> pdos of the system was calculated and plotted for every atom layers.
>>>>> Result: E_F = 0.1552 eV  (fermi energy is overall constant, in every 
>>>>> atom layers)
>>>>>
>>>>> b) the same as a), but the bulk region was not already 
>>>>> geometry-optimized before. Geometry optimization was performed and 
>>>>> calculation of pdos.
>>>>> Result: E_F = - 0.8516 eV
>>>>>
>>>>> c) the same as b), but 26 trilayers instead of 14 trilayers. Geometry 
>>>>> optimization and calculation of pdos were performed.
>>>>> Result: E_F = 2.3372 eV
>>>>>
>>>>>
>>>>> So, I am wondering why these values differ so much. Should I need band 
>>>>> structure calculation of the bulk LiNbO3 in order to find the global 
>>>>> valence band edge maximum (with KPOINT calculation)?
>>>>>
>>>>> Kind regards,
>>>>>
>>>>> Phil
>>>>>
>>>>>
>>>>> Am Freitag, 14. Dezember 2018 17:41:03 UTC+1 schrieb Matt W:
>>>>>>
>>>>>> In a periodic system the zero of the one electron levels is 
>>>>>> arbitrary. If you need a reference you need to run a slab calculation with 
>>>>>> vacuum or try to align semi-core states to something.
>>>>>>
>>>>>> Matt
>>>>>>
>>>>>> On Friday, December 14, 2018 at 4:33:13 PM UTC, Phil G. wrote:
>>>>>>>
>>>>>>> Dear people and experts of CP2K,
>>>>>>>
>>>>>>> after the geometry optimization of the lithium niobate (LiNbO3) unit 
>>>>>>> cell I would like to obtain pdos in order to determine the band gap and 
>>>>>>> Fermi energy of the bulk system.
>>>>>>> After the calculation with ENERGY_FORCE I got pdos files of the 
>>>>>>> three atoms (indexing depends on the z-position of the atoms) and I'm 
>>>>>>> wondering about the value of Fermi energy: E_F = 0.300174 a.u. which is 
>>>>>>> 8.168 eV. Is that not too high? And which energy has the value 0 and what 
>>>>>>> is the reference? What is the Fermi energy defined in the language of CP2K?
>>>>>>> The energy band gap (HOMO-LUMO gap) of 3.62 eV agrees well with 
>>>>>>> experimental values of 3.7 to 3.9 eV. But I cannot imagine that Fermi level 
>>>>>>> has too high energy values.
>>>>>>>
>>>>>>> Has anyone an idea what is the reason for such high Fermi energy 
>>>>>>> values?
>>>>>>>
>>>>>>> Here the input and output files are attached here.
>>>>>>>
>>>>>>> Kind regards,
>>>>>>>
>>>>>>> Phil
>>>>>>>
>>>>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190115/8857a63c/attachment.htm>