[CP2K-user] [CP2K:12058] Re: Cell Optimization does not work in QM

Matt W mattwa... at gmail.com
Thu Aug 8 11:58:50 UTC 2019


You are not using K points, so you will need a 2x2x2 cell to get something 
sensible.

Matt

On Thursday, August 8, 2019 at 11:54:20 AM UTC+1, Pierre Cazade wrote:
>
> Hi Matt,
>
> Thanks for the tip about the cell_ref, I'll give it a try. Otherwise, I 
> define the symmetry and I keep it fixed. I also keep the angles fixed for 
> the first optimization.
>
> A new calculation I've just run with the DFT parameters from Dirk's input 
> file yields similar results as I obtained before:
>
>  CELL| Vector a [angstrom]:       4.251     0.000     0.000    |a| =       
> 4.251
>  CELL| Vector b [angstrom]:       0.000     7.405     0.000    |b| =       
> 7.405
>  CELL| Vector c [angstrom]:      -2.287     0.000     5.298    |c| =       
> 5.771
>
> I started with this however:
>
> &CELL
>       ABC [angstrom]   5.078       6.192     5.387
>       ALPHA_BETA_GAMMA [deg] 90.00 113.35  90.00
>       SYMMETRY MONOCLINIC
>       PERIODIC XYZ
> &END CELL
>
> I know from VASP the optimized unit cell is:
>
>    a                  b             c         alpha          beta          
> gamma
>  5.12269  6.41270  5.41981   90.0000    112.7275     90.0000
>
> Of course, I do not expect two different software using different methods 
> and different pseudopotentials to give the exact same result but they 
> should be close enough. The contraction of the A vector and the expansion 
> of the B vector are absurd. I will try the ref_cell, it might help.
>
> Regards,
> Pierre
>
>
> On 08/08/2019 11:38, Matt W wrote:
>
> Standard inputs are in Angstroms for a coord section. 
>
> The most common problem is that the planewave basis changes as the cell 
> optimizes. You can use a constant size of plane wave basis by specifying a 
> reference cell, something like this. Where the CELL_REF contains a 
> reference cell that should be a bit bigger that the expected largest cell 
> after optimization.
>
>     &CELL
>       ABC 5.620 5.620 5.620
>       &CELL_REF
>         ABC 5.620*1.1 5.620*1.1 5.620*1.1
>       &END
>     &END CELL
>
> If your cell has symmetry (orthorhombic, hexagonal etc) you can specify 
> this to be kept. Similarly first optimizing just the volume and keeping the 
> angles constant, then relaxing everything can be more efficient.
>
> Also I'd stick with BFGS rather than LBFGS unless you system is huge (> 
> 2000 atoms)
>
> Maybe this helps a little, 
>
> Matt
>
>
> On Thursday, August 8, 2019 at 10:46:18 AM UTC+1, Dr. Dirk Buddensiek 
> wrote: 
>>
>> Hi Pierre,
>>
>> good luck. Concerning distance units: It might be that there are in Bohr 
>> for MP2. But you can set the units explicitly in the input deck.
>> Dirk
>>
>> Am Mittwoch, 7. August 2019 17:05:56 UTC+2 schrieb Pierre-André Cazade: 
>>>
>>> Hello, 
>>>
>>> I am rather new to CP2K though I am experienced with molecular modelling 
>>> both quantum and classical, and with a wide range of software. It is quite 
>>> frustrating that it seems impossible to get CP2K to perform a simple 
>>> CELL_OPT of a beta glycine crystal (20 atoms, monoclinic) with PBE. Such a 
>>> calculation works fine with VASP, so I am quite surprised the same 
>>> calculation goes astray with CP2K. Basically, the B unit cell vector 
>>> increases drastically up to absurd values. Needless to say, this is a test 
>>> system as I wish to use CP2K for its linear scaling DFT to model protein 
>>> crystals. I have tried to play with CUTOFF and REL_CUTOFF in the MGRID 
>>> section up to 1400 and 80 respectively. I also tried GPW and GAPW methods, 
>>> all electrons or GTH pseudo. Nothing works. Please, if anyone could have a 
>>> look at the attached input file "cell_opt_qm.inp" and point what's wrong in 
>>> it, I would be grateful.
>>>
>>> Furthermore, I have tried LS approach with DFTB on a 5x5x5 crystal of 
>>> beta glycine and do not have much more success. First, the pressure in the 
>>> crystal is insanely high and will lead again to a 20 folds increase of the 
>>> lattice vectors (with CG). Then, with LBFGS method, the system stops after 
>>> 3 to 4 cycles with a message telling LBFGS convergence criteria were 
>>> reached, whereas the system is clearly not. Again, I would be grateful if 
>>> anyone could tell me what's wrong with my system ("cell_opt2_qm.inp" with 
>>> "betacry.pdb" for the coordinates). Ideally, my goal would be to be able to 
>>> run large complex systems with the LS approach and with any DFT/DFTB method.
>>>
>>> Thank you in advance for your help.
>>>
>>> Regards,
>>> Pierre
>>>
>> -- 
> 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 c... at googlegroups.com <javascript:>.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/cp2k/949d34b8-c700-43ac-a1af-c0b62887d785%40googlegroups.com 
> <https://groups.google.com/d/msgid/cp2k/949d34b8-c700-43ac-a1af-c0b62887d785%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
>
> -- 
> Dr Pierre Cazade, PhD
> AD3-023, Bernal Institute,
> University of Limerick,
> Plassey Park Road,
> Castletroy, co. Limerick,
> Ireland
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190808/60536ee3/attachment.htm>


More information about the CP2K-user mailing list