Unreasonable CELL_OPT result obtained for TiO2

[Yike Huang]

Hello, cp2k developers and users,

I try to optimize primitive cell of rutile TiO2 by running CELL_OPT calculation at PBE+U(presently set to 0 eV)/DZVP level, but I find cell parameters deviate from experimental values quite quickly during calculation.

I have tried changing STRESS_TENSOR to both NUMERICAL and ANALYTICAL but it does not help, I also think it may be not the reason why I get this kind of strange result.

I attach my input script and CIF file below, are there any parameters in input script I forget?

In addition, anatase TiO2 CELL_OPT task also has such problem, but I remember that primitive cell of Co3O4 can be successfully optimized.

Thanks in advance.

Yike Huang

[Krack Matthias (PSI)]

Hi Yike

 

Your simulation cell is small and you use no k point sampling. Either you use k point sampling or you enlarge your simulation cell to at least a cell edge length of 12 A in each direction. I suggest the latter option for CP2K. Furthermore, you could reduce EPS_DEFAULT to 1.0E-12 and increase REL_CUTOFF to 60.

 

HTH

 

Matthias

 

--
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...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/11f85e5a-b3f6-45b2-b164-11f12644f358n%40googlegroups.com.

[Hasan Al-Mahayni]

Hi,

I’m gonna jump in if you don’t mind. I am trying to do cell optimization of  fcc copper. I used k points and got 3.65A  which is different from the 3.62 that was expected. I was wondering for your latter option (using a large simulation cell) , would Run CELL_OPT or GEO_OPT?

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/b6c6bd92f59a4a41b1155ab00ac8bfc4%40psi.ch.

[Yike Huang]

Thanks very much,  Matthias.

I will try both of them and check if I can get reasonable result.

Yike

[Krack Matthias (PSI)]

Hi

 

I guess you have been using DFT-PBE as level of theory for fcc Cu. In that case, a deviation of only 0.03 A for the relaxed lattice constant compared to experiment seems to me acceptable. DFT-GGA is known to give often too large lattice constants.

 

M.

 

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAECRhQz-SYS1NPxvZ_D5F40dOxKD84M%3DDE37m_sGs0-285Amzg%40mail.gmail.com.

[Hasan Al-Mahayni]

I see, Thank you!!

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/aebdfa971c58464cb9595f9b255d765c%40psi.ch.

--

Hasan Al-Mahayni

[Yike Huang]

Dear  Matthias,

I have tested these two methods. 

For the first &DIAGONALIZATION plus &KPOINTS method, I find it somewhat strange that even if I switch on &OUTER_SCF, SCF will still exist with non-converged warning, and by now I have no idea about it. So I turn to the second.

For the second that to set MULTIPLE_UNIT_CELL both in &CELL and &TOPOLOGY, it works quite well and I find for accuracy of structure prediction (bond length, cell parameters, etc.), it is even not necessary to use DFT+U to tune electron delocalization property of Ti (both anatase and rutile TiO2), although for band structure calculation, DFT+U is usually critical.
I share information above here and hope it will be useful for other users.

Thank you again.

Yike

在2021年5月16日星期日 UTC+2 下午6:46:31<Matthias Krack> 写道：