[CP2K-user] [CP2K:20087] Re: AUTO_BASIS in RI_MP2 non-PBC

[Ralf Frischmann]

Hi Frederick, hi Augustin,

thanks again for the helpful clarification.

A kind of "quick fix" could also be to force the AUTO_BASIS generator (by 
keyword?) to obey the primitive Gaussian "rule", i.e. to mimic how the e.g 
the RI basis in the RI_DZVP-GTH file (and other RI basis files as well) 
have been generated. But I have no idea how difficult that actually would 
be.

Regards,

Ralf



 



On Tuesday, April 2, 2024 at 9:56:16 AM UTC+2 Augustin Bussy wrote:

> Hi Ralf, hi Frederick,
>
> indeed, there is in principle no reason for the calculation to behave 
> differently depending on the way the RI basis set is generated. Looking at 
> the code (
> https://github.com/cp2k/cp2k/blob/f529a546d898301ce447cb1d79a747c5b45da7c9/src/mp2_eri.F#L1320C1-L1321C14), 
> we see that with "ERI_METHOD OS" the program expects the RI basis to have a 
> single Gaussian primitive per set. This happens to be the case for 
> RI_DZVP-GTH, but not for the automatically generated one.
>
> This is clearly an oversight from the original MP2 implementation. I 
> think, either the code should issue a proper error message, or this piece 
> of code should be rewritten in a more general manner (probably using 
> Libint).
>
> Best,
>
> Augustin
>
>
> On Tuesday 2 April 2024 at 09:03:29 UTC+2 Ralf Frischmann wrote:
>
>> Hi Frederick,
>>
>> thanks for the quick reply. 
>>
>> With "ERI_METHOD GPW" there is no crash in the _auto case anymore :-)
>>
>> However, the explicit (non-auto) RI_AUX basis case input does finish 
>> properly with "ERI_METHOD OS" and also gives the correct MP2 E_corr of 
>> about -0.199x, so it is not 100% clear to me how switching to an 
>> auto-generated RI basis can then cause that kind of failure. In other 
>> words, the outcome of the ERI_METHOD compute part should in my limited 
>> understanding *not depend *on how the RI_AUX basis has been initially 
>> "generated", i.e. read from a basis set file or generated numerically from 
>> scratch via AUTO_BASIS.
>>
>> Regards,
>>
>> Ralf
>>
>>
>>
>>     
>>  
>>
>>
>> On Tuesday, April 2, 2024 at 8:13:58 AM UTC+2 Frederick Stein wrote:
>>
>>> Dear Ralf,
>>> Can you switch from ERI_METHOD OS to ERI_METHOD GPW ?
>>> Frederick
>>>
>>> Ralf Frischmann schrieb am Dienstag, 2. April 2024 um 06:59:45 UTC+2:
>>>
>>>> Dear CP2K fellows,
>>>>
>>>> just came across the following issue:
>>>>
>>>> Trying to run a slightly modifed version (input files attached) of the 
>>>> regtest 
>>>>
>>>>
>>>> *tests/QS/regtest-ri-mp2/RI_MP2_H2O.inp *
>>>>
>>>> with the *cp2k-2024.1-Linux-gnu-x86_64.ssmp* executable actually works 
>>>> fine.
>>>>
>>>> However, upon enabling the AUTO_BASIS option and removing the RI_AUX 
>>>> basis set (see  H2O_auto.inp input file), the calculation crashes. 
>>>>
>>>> Any help on what is going wrong would be greatly appreciated*. *Thanks 
>>>> a lot in advance!
>>>>
>>>> Ralf
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>

-- 
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/3f0a7fc9-2efd-4dda-a4c7-4cd2dbc0fd33n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20240402/41de333c/attachment.htm>