Cholesky decompose failed

[Nimali Rathnayake]

Dear All,

I am running a system with CP2K. When I was using QZV2P basis set and the simulation crashed with the following error: I am running this geometry optimization calculation for 1 layer graphene layer, which has the 3.35A d-space.

Cholesky decompose failed: the matrix is not positive definite or 

                         ill-conditioned.

                                                                      fm/cp_fm_cholesky.F:94

I am wondering what would be the reason for this error message. 

I am very thankful, if you can help me to solve this

The input and output files are attached.

--

Thank you,

Best Regards,

Nimali

[Espen Tangen]

Hi all. Please help.

Seems that I am at the very start again. I have compiled binaries popt and psmp both with intel 2017.0.4 compiler and everything works fine for test H2O-32 and H2O-128 with single node running mpirun. For some odd reason, I get either the previously reported error or the above mentioned error when I try to run mpirun or srun on two nodes.

This holds for both popt and psmp. Funny thing is, for intel compiler 2018.0.1 release this vanishes for popt but persists for psmp.

Anyone who may point me in the right direction?



Et.

[Nimali Rathnayake]

Thank you Travis.

Kind regards,

Nimali

On Tue, May 14, 2019 at 3:14 AM Travis <polla...@gmail.com> wrote:

Large basis sets have overlap issues. Use a smaller basis set or adjust EPS_EIGVAL.

--
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 post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/acede0e2-edde-4e16-b4c7-cdbbbb1fea44%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.