abrupt termination of the cp2k md simulation with xtb

[Alexandra Tsybizova]

Dear cp2k community, 

I have been experiencing weird errors when I try to run md simulations with cp2k, when I use xtb method. The calculation seems to run fine for the first 130 steps or so, and then abruptly crashes with a signal 6 (SIGABRT). In the past, I used similar inputs to run dft trajectories, and they all finished normally. Browsing this forum I found that similar behavior (termination with a signal 6 (SIGABRT)) was reported by one of the users previously, (https://groups.google.com/g/cp2k/c/QWeC0KAgr2A) and in that case was solved by removing a duplicate line from the input. I checked my input for duplicate lines and could not find any, however. Could you help me identify where the problem might be? 

Thank you and best regards,

Alexandra 

[Anton Lytvynenko]

Dear Alexandra,

did you try to use the trunk version with some bugs fixed (it could be not the case of your particular problem, but anyway)?

Yours,

Anton

30.09.2022 10:00, Alexandra Tsybizova пише:

--
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 view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/95d68e6e-d7ba-408b-9389-3a10027ac4d8n%40googlegroups.com.

[Alexandra Tsybizova]

Dear Anton, 

Thank you for your answer. I am using cp2k v.8.2, that is what we have available on the cluster. I haven't tried the trunk version. 

Best regards,

Alexandra 

[Krack Matthias (PSI)]

Hi Alexandra

 

Your input runs fine for me (stopped only after 516 steps because of the time limit). Try to run the job again to see if the problem is reproducible.

 

HTH

 

Matthias

--

[Jürg Hutter]

Hi

on a side note. You spend all your CPU time for the orbital localization.
I would optimize the input parameters (e.g. max steps for localization).
I would also decrease the SCF convergence (make it tighter).

regards
JH

________________________________________

From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Alexandra Tsybizova <alexandra...@gmail.com>

Sent: Friday, September 30, 2022 10:00 AM
To: cp2k

Subject: [CP2K:17782] abrupt termination of the cp2k md simulation with xtb

--
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<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/95d68e6e-d7ba-408b-9389-3a10027ac4d8n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/95d68e6e-d7ba-408b-9389-3a10027ac4d8n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Alexandra Tsybizova]

Dear Matthias, 

thank you for trying out to run my input file. The problem is reproducible for me, but it sounds like the input file is not a problem. I will continue to investigate this further. 

Best regards,

Alexandra 

[mdsimula...@gmail.com]

HI Alexandra,

Have you considered trying DFTB+ (https://dftbplus.org/) ?  It has the xTB Hamiltonians and they have precompiled binaries which are super easy to use.

I've used it a few times for some periodic calculations and it works nicely.

 Cheers!

Frank