[CP2K:8370] BAND_STRUCTURE k-point unit conversion bug leads to bogus band structures (CP2K 4.1)

[hut... at chem.uzh.ch]

Hi Ari

thank you for reporting this issue. I will look at it as soon as possible.

best

Juerg
--------------------------------------------------------------
Juerg Hutter                         Phone : ++41 44 635 4491
Institut für Chemie C                FAX   : ++41 44 635 6838
Universität Zürich                   E-mail: hut... at chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp... at googlegroups.com wrote: -----To: cp... at googlegroups.com
From: Ari Paavo Seitsonen 
Sent by: cp... at googlegroups.com
Date: 11/09/2016 11:22AM
Subject: Re: [CP2K:8370] BAND_STRUCTURE k-point unit conversion bug leads to bogus band structures (CP2K 4.1)

Hello,

  Talking about the k points and their units - would it be possible to have an option for 'MACDONALD' scheme of k points where the "kp_shift" would also be given in units of the reciprocal lattice vectors - they are natural due to the algorithm for their construction and easier - or even an option/short-cut that would enable giving Gamma-point centred k point sets, as these are needed in hexagonal cells. Now the shift depends on and thus changes with the chosen number of grid points.

  A side note - in the documentation there appears to be a small error, namely the shift is not mentioned together with the 'MACDONALD' scheme, yet the code crashes if one only provides the three integers mentioned in the document (this is correct for the scheme 'MONKHORST-PACK'). And the symmetries apparently always fail with the 'MACDONALD' scheme - as said, this is basically the compulsory scheme in hexagonal cells. And if I may note, I tried to calculate a hexagonal cell together with the 'MACDONALD' method, with a shift adjusted so that Gamma point is included in the set, but I obtain a strange reduction of symmetry in the forces (the three-fold symmetry that is present due to the lattice is not obeyed laterally) - any idea? I know that this is a new feature, maybe I should just be patient. :)

    Greetings from Paris,

       apsi

2016-11-09 10:14 GMT+01:00  <hut... at chem.uzh.ch>:
Hi



I updated the code for the k-point input. I hope I did catch

all the error. I only made some minimal tests, so I am still

not sure that it is correct.



regards



Juerg Hutter

--------------------------------------------------------------

Juerg Hutter                         Phone : ++41 44 635 4491

Institut für Chemie C                FAX   : ++41 44 635 6838

Universität Zürich                   E-mail: hut... at chem.uzh.ch

Winterthurerstrasse 190

CH-8057 Zürich, Switzerland

---------------------------------------------------------------



-----cp... at googlegroups.com wrote: -----To: cp2k <cp... at googlegroups.com>

From: Eric Hermes

Sent by: cp... at googlegroups.com

Date: 11/01/2016 09:35PM

Subject: [CP2K:8332] BAND_STRUCTURE k-point unit conversion bug leads to bogus band structures (CP2K 4.1)



Hello,

I am working with ASE to develop routines for the various DFT calculators we support for simple, automatic generation of band structures. I have run into a bug in CP2K's SPECIAL_POINT specification, wherein the specified k-points appear to undergo a redundant B_VECTOR -> CART_BOHR unit conversion. What I mean by that is if I pass the k-points as multiples of the reciprocal unit vectors (i.e. UNIT B_VECTOR, SPECIAL_POINT 0.5 0.25 0.75 for the W point of the diamond lattice), this is multiplied by the reciprocal unit cell vector *twice*. If I pass the points in units of 2pi/Bohr (i.e. UNIT CART_BOHR, SPECIAL_POINT 0.049 0.098 0.000 for a primitive Si diamond lattice with lattice constant a=5.4 angstrom), the k-points are multiplied by the reciprocal unit cell vector *anyway*, giving the same results as before. I can "trick" the system into providing the correct result by saying that I am passing the k-points in 2pi/Bohr, but actually passing the k-points in multiples of the reciprocal unit cell vectors (i.e. UNIT CART_BOHR, SPECIAL_POINT 0.5 0.25 0.75). This results in both the output file and the band structure file printing the exact k-points that I specified (i.e. the k-points in multiples of the reciprocal unit vectors), and the resulting band structure matches that of other DFT codes.

Here is an example of the input file I am using that expresses the bug: https://gist.github.com/ehermes/7ef3d7249b99eeafb34bb14e6c51fd48This input file was generated by (a modified version of) ASE, and it produces the incorrect band structure. If line 31 is changed to read "UNITS CART_BOHR", the resulting band structure is correct and matches the band structure produced by other codes (GPAW, VASP, etc).

I have taken a look at the qs_band_structure.F source file, but the origin of the bug is not obvious to me. I suspect it might be related to line 197, which multiplies each kpoint by the transpose of the unit cell vector, but I am not certain.







--



You received this message because you are subscribed to the Google Groups "cp2k" group.



To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns... at googlegroups.com.



To post to this group, send email to cp... at googlegroups.com.



Visit this group at https://groups.google.com/group/cp2k.



For more options, visit https://groups.google.com/d/optout.





--

You received this message because you are subscribed to the Google Groups "cp2k" group.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns... at googlegroups.com.

To post to this group, send email to cp... at googlegroups.com.

Visit this group at https://groups.google.com/group/cp2k.

For more options, visit https://groups.google.com/d/optout.




-- 
-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-
  Ari Paavo Seitsonen / Ari.P.S... at iki.fi / http://www.iki.fi/~apsi/
    Ecole Normale Supérieure (ENS), Département de Chimie, Paris
    Mobile (F) : +33 789 37 24 25    (CH) : +41 79 71 90 935





-- 

You received this message because you are subscribed to the Google Groups "cp2k" group.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns... at googlegroups.com.

To post to this group, send email to cp... at googlegroups.com.

Visit this group at https://groups.google.com/group/cp2k.

For more options, visit https://groups.google.com/d/optout.