[CP2K-user] [CP2K:18887] Re: Problems with converging PW cutoff

[Léon Luntadila Lufungula]

Dear Lobna,

Thanks for the reply! 

I can see that the total energies may differ from those obtained with other 
QC software packages as the computational methods (implementation, basis 
set, etc.) might be different, but I would think that this does not 
influence the convergence of the total energy as a function of the cutoff. 

I could indeed compare the lattice parameters of a calculated anatase unit 
cell at different cutoffs with those found experimentally, but wouldn't it 
be computationally expensive to check as I would have to do CELL_OPT 
calculations for each cutoff as opposed to simple ENERGY calculations? 
Also, should thermal effects be taken into account in some way as CELL_OPT 
calculations are at 0K and the experimental CIF was recorded at RT?

You also mention that Ti requires an optimization of the basis set and 
pseudopotentials, but I don't know how I would go about doing this 
optimization, do you have any suggestions?

Kind regards,
Léon

On Monday, 5 June 2023 at 09:19:01 UTC+2 Lobna Saeed wrote:

> I had the same exact issues with the convergence !! exactly the same. I 
>> was also confused about the largest exponent as well. When I used this 
>> https://www.cp2k.org/faq:cutoff, I got a very high cutoff value which 
>> when I employed deviated very much from the experimental values. So may be 
>> you need to look at other values, other than the total energies. The total 
>> energies we get here are not by any mean realistic. When I compared the 
>> total energy values I got from VASP with the values we get from CP2K (for 
>> single elements), I got completely different results. So when I increased 
>> the basis sets I was able to get the same values for Al and Nb, other 
>> elements like Ti REALLY need optimization of the basis sets and 
>> pseudopotentials. Therefore you should try to look at other values other 
>> than the total energies (like the lattice parameters for example) to be 
>> able to deduce a final conclusion about the cutoff value.
>
>
> On Mon, Jun 5, 2023 at 6:59 AM Léon Luntadila Lufungula <
> Leon.luntad... at uantwerpen.be> wrote:
>
>> Dear all,
>>
>> Apparently I forgot to add my input file in my previous message, so I 
>> have included it as an attachment in this reply. I'm eagerly looking 
>> forward to your replies and thank you in advance.
>>
>> Kind regards,
>> Léon
>>
>> On Friday, 2 June 2023 at 15:44:38 UTC+2 Léon Luntadila Lufungula wrote:
>>
>>> Dear all,
>>>
>>>  
>>>
>>> I am currently switching from Quantum ESPRESSO (QE) to CP2K as my main 
>>> computational software package (currently using version 8.2) and I am 
>>> struggling a bit with the convergence tests for the PW cutoffs (REL_CUTOFF 
>>> and CUTOFF). My intended structures to investigate are moderate-sized 
>>> organic molecules adsorbed onto a (101) surface of anatase (TiO2) and as 
>>> such I am using an optimized structure from my QE calculations as a 
>>> representative system to do my convergence tests on (see attached file 
>>> 3mppa-md1.inp). I based my method on the method proposed by Prof. Hütter in 
>>> a previous post <https://groups.google.com/g/cp2k/c/ySUAYEgwmhc>, 
>>> whereby I first set REL_CUTOFF=CUTOFF and increase the value until I reach 
>>> convergence. The results I got are the following:
>>>
>>>  
>>>
>>> # Grid cutoff vs total energy
>>>
>>> # Date: Fri Jun  2 14:09:32 CEST 2023
>>>
>>> # PWD: /home/lluntadilal/cp2k/scf/dft/test/final/convergence/cutoff
>>>
>>> # Cutoff (Ry) | Relative cutoff (Ry) | Total energy (a.u.) | Total 
>>> charge (a.u.)
>>>
>>>           300                    300      -3303.4693919783 
>>> <(469)%20391-9783>          0.0000049144
>>>
>>>           400                    400      -3303.4678396250         
>>> -0.0000000118
>>>
>>>           500                    500      -3303.4658109937         
>>> -0.0000000003
>>>
>>>           600                    600      -3303.4675587607         
>>> -0.0000000003
>>>
>>>           700                    700      -3303.4671967038         
>>> -0.0000000003
>>>
>>>           800                    800      -3303.4663947747         
>>> -0.0000000003
>>>
>>>           900                    900      -3303.4659684557         
>>> -0.0000000003
>>>
>>>          1000                   1000      -3303.4657584430         
>>> -0.0000000002
>>>
>>>          1100                   1100      -3303.4651289391         
>>> -0.0000000002
>>>
>>>          1200                   1200      -3303.4650897922         
>>> -0.0000000003
>>>
>>>          1300                   1300      -3303.4651986119         
>>> -0.0000000002
>>>
>>>          1400                   1400      -3303.4651800964         
>>> -0.0000000002
>>>
>>>          1500                   1500      -3303.4650037920         
>>> -0.0000000002
>>>
>>>          1600                   1600      -3303.4650956961         
>>> -0.0000000002
>>>
>>>          1700                   1700      -3303.4651701336         
>>> -0.0000000002
>>>
>>>          1800                   1800      -3303.4651578599         
>>> -0.0000000002
>>>
>>>          1900                   1900      -3303.4651999129         
>>> -0.0000000002
>>>
>>>          2000                   2000      -3303.4651936486         
>>> -0.0000000002
>>>
>>>  
>>>
>>> I have several questions about this:
>>>
>>>    1. I would like to be able to give the energies of my structures 
>>>    accurate up to 0.01 kJ/mol which is about 1.10^-6 a.u. Unfortunately, I 
>>>    only reach this level of accuracy at a cutoff of 2000 Ry… This seems quite 
>>>    a large cutoff as I see similar calculations with cutoffs between 400-1200 
>>>    Ry and also because the charge is already converged at 500 Ry. Am I doing 
>>>    something wrong or is my criteria just too strict? Does this perhaps have 
>>>    anything to do with the fact that the relative position of the atoms to the 
>>>    grid points changes with the cutoff as mentioned by Prof. Hütter in his 
>>>    post? 
>>>    2. The faq <https://www.cp2k.org/faq:cutoff> and converging cutoff 
>>>    exercise 
>>>    <https://www.cp2k.org/events:2018_summer_school:converging_cutoff> 
>>>    state that the PW cutoff should be large enough to properly represent the 
>>>    Gaussian with the largest exponent, which in my system seems to be oxygen 
>>>    with an exponent around 10.4 for the DZVP-MOLOPT-SR-GTH basis set. So is it 
>>>    then correct to use such a complex structure (3-layer slab with adsorbate 
>>>    and 30Å vacuum width) with "complex" options (dipole and dispersion 
>>>    corrections) as input for my convergence tests or could I use a simpler 
>>>    structure (e.g. an oxygen molecule in a box) with more basic settings (no 
>>>    corrections) or would this be insufficient? 
>>>    3. I may be planning to use larger basis sets ("regular" DZVP, TZVP 
>>>    or TZV2P) of the MOLOPT family in the future and I have seen that the 
>>>    largest exponents for these sets is the same as that for the smaller 
>>>    DZVP-SR basis set (10.4 for oxygen). Is it correct then to assume that once 
>>>    I obtain optimized cutoffs for DZVP-SR, these can be kept when switching to 
>>>    any of the larger basis sets? 
>>>    4. Lastly, I would like to ask for some general feedback on the 
>>>    input for my slab calculation as this is my first calculation on CP2K and 
>>>    I'm still getting used to all the settings. Is it okay to use 3D 
>>>    periodicity with a large vacuum space and a dipole correction or is it 
>>>    better to use 2D periodicity with a corresponding Poisson solver (e.g. MT, 
>>>    analytic or wavelet)? Is it okay to use the WC functional with basis sets, 
>>>    pseudopotentials and D3 settings optimized for PBE? 
>>>
>>>  
>>>
>>> Any help would be greatly appreciated!
>>>
>>>  
>>>
>>> Kind regards,
>>>
>>> Léon
>>>
>>>  
>>>
>>> P.S. For those looking at my structure in a graphical environment and 
>>> noticing that some atoms are outside the unit cell. This is due to the fact 
>>> that my cell was monoclinic and I made the unit cell orthorhombic to save 
>>> computational time and trouble (some codes have trouble dealing with cells 
>>> that are not orthorhombic), but I did not yet wrap all my atoms into the 
>>> unit cell. However, I don't think this should be a problem under PBC.
>>>
>> -- 
>> 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 view this discussion on the web visit 
>> https://groups.google.com/d/msgid/cp2k/ee55ab54-b914-49fb-bd8d-f185f91b09een%40googlegroups.com 
>> <https://groups.google.com/d/msgid/cp2k/ee55ab54-b914-49fb-bd8d-f185f91b09een%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
>

-- 
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+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/e3f4936e-8d9d-4530-bd8b-85a98f484f2fn%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20230605/edb01dd7/attachment-0001.htm>