Cholesky failed: the matrix is not positive definite or ill-conditioned.

[David T]

Dear all

I am trying to optimize a structure using cp2k v 3.0 svn:16458,
I had problems with the subject error so that I added Cholesky off and remove OT from the output.
However, despite this cp2k still crashed after 1 cycle complaining that the matrix is not positive or ill-conditioned.

I've just started using CP2K so I am not really comfortable in playing with parameters. For example
I read in a previous post that one solution could be changing the basis set but I would rather avoid that if possible.

Probably more important I would like to understand why this happens despite activating the Cholesky off option

Any suggestion? Is there any other way for fixing this problem?

Please find in attachment the input, output and structure I am trying to optimise and please apologise if I am doing something silly.

Cheers
Davide

[Matt W]

Hi,

given the basis sets you are using (MOLOPT), I would strongly suspect instead that the reason for the error is a bad geometry. Possibly in the cell definition?

Matt

[David T]

Hi Matt
thanks for answering.
I was also worried about that so I tried optimising with any problems using plane-wave code (vasp and espresso) and it does work excluding geometrical issue.


However, your message "illuminate me" so that I removed the hexagonal symmetry, it does work now.

[S Ling]

Hi

The hexagonal symmetry in CP2K requires the gamma angle to be 60. In your case, you had 120, and if you specify "SYMMETRY HEXAGONAL", the code will overwrite your definition, and you ended up with a wrong geometry.

SL

--
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@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

[David T]

Thanks
that make sense.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.

[shixun sun]

hi David, Could you tell me how to remove the hexagonal symmetry, I have the same problem now. Appreciate for your help.
在 2016年9月7日星期三 UTC+8上午12:19:28，David T写道：