BigDFT

Non-orthorhombic cell support of Poisson solver

Asked by Juha Ritala

Dear developers,

I am interested in updating the implementation of the wavelet Poisson solver in CP2k to support monoclinic cells with surface boundary conditions. If I have understood correctly, the version of the Poisson solver used in BigDFT should have this support. I tried to adapt the changes I found in Surfaces_Kernel subroutine (using the metric of monoclinic cell when calculating mu1), but I still get incorrect results. Are there some other changes that I may have missed?

Thanks,
Juha

Question information

Language:
English Edit question
Status:
Solved
For:
BigDFT Edit question
Assignee:
No assignee Edit question
Solved by:
Juha Ritala
Solved:
Last query:
Last reply:
Revision history for this message
Luigi Genovese (luigi-genovese) said : 		#1

Dear Juha,

the answer depends on the original version of the Solver that was implemented in CP2K. I do not have enough details to help you.

However, in its present version, our Poisson Solver can be compiled, installed and linked independently from BigDFT.

I might suggest to update CP2K such as to use the ISF Poisson solver as external library. This will make easier for us to interact for correcting potential problems that might arise.

Do not hesitate

Many thanks

Luigi

Revision history for this message
Juha Ritala (jritala) said : 		#2

Dear Luigi,

I was hoping to do a quick modification to the existing implementation of the solver in CP2k because the only changes I found were in the Surfaces_Kernel subroutine. The version of the wavelet solver in CP2k seems to be dated back to 2007.

It would certainly be sensible to link the independent Poisson solver to CP2k though. There is no point in trying to keep the embedded one up to date with newest version. I will check how much work this would need and see if my motivation is high enough for doing it.

Thanks,
Juha

Revision history for this message
Luigi Genovese (luigi-genovese) said : 		#3

Dear Juha,

I fear it is wise to really proceed forward with a separate integration.
The version of 2007 is not ready for non-orthorhombicity and several modfications has been done to allows for this feature.

Do not heistate in contacting us should you have further questions. On our side, we are going to release a version in the forthcoming weeks where the non-orthorhombic system are explicitly tested in the non-regression programs of the Poisson Solver. THerefore potential problems might be more easily solved.

Best

Luigi

Revision history for this message
Juha Ritala (jritala) said : 		#4

Ok, so there is no point in trying to modify the old version. It sounds encouraging that the non-orthorhombic systems will be fully tested and supported. I will try and link the standalone version of the Poisson solver to CP2k then and report back if I encounter any problems.

With many thanks,
Juha