CP2K QMMM Calculation seems to randomly crash

52 views
Skip to first unread message

Tomo Oka

unread,
Dec 13, 2023, 12:14:49 PM12/13/23
to cp2k
Dear all,

I have encountered a very disturbing phenomena. 

My system is fairly simple: One Na+ in water (up to 800 water molecules). I generated the system using tleap. The plan is the following:

Since full semi empirical MD with XTB is costly for 801 atoms, I tried to calculate MD with both conventional QMMM and abfQMMM. I do acknowledge, that conventional QMMM will not work since the QM atoms in the QM box change. Therefore abfQMMM seems to be the solution. In any case, I tried conventional QMMM first.

1) conventional QMMM
Following the normal procedure, I did first minimization without QMMM, and it crashed for most of the time. Somehow it depends on the box size and the binary I execute. With any executable ssmp, psmp, popt, sopt using CPU only, it crashed. The only successful run was with the local_cuda executable sopt. Sometimes it also did not work. 

I run the cp2k.pdbg and cp2k.sdbg. The error message when I run cp2k.sdbg was:

At line 766 of file /usr/local/cp2k/src/force_fields_ext.F
Fortran runtime error: Pointer argument 'amb_info' is not associated

Error termination. Backtrace:
#0  0x2183c84 in __force_fields_ext_MOD_read_force_field_amber
at /usr/local/cp2k/src/force_fields_ext.F:766
#1  0x1d8aa45 in __force_fields_MOD_force_field_control
at /usr/local/cp2k/src/force_fields.F:161
#2  0x19d5df9 in __fist_environment_MOD_fist_init
at /usr/local/cp2k/src/fist_environment.F:176
#3  0x184bc1b in __fist_main_MOD_fist_create_force_env
at /usr/local/cp2k/src/fist_main.F:98
#4  0x15ed06b in __f77_interface_MOD_create_force_env
at /usr/local/cp2k/src/f77_interface.F:786
#5  0x473b25 in cp2k_run
at /usr/local/cp2k/src/start/cp2k_runs.F:308
#6  0x475a19 in __cp2k_runs_MOD_run_input
at /usr/local/cp2k/src/start/cp2k_runs.F:975
#7  0x46b112 in cp2k
at /usr/local/cp2k/src/start/cp2k.F:379
#8  0x46b25d in main
at /usr/local/cp2k/src/start/cp2k.F:44

The sporadic errornous and successful runs continue as well in the following NVT and NPT equilibration with QMMM. For me the error message seems to originate in the amber force field parser.

2) abfQMMM. No matter what executable I called, everytime it crashed and the same error message appears. I enlarged the box, still no success.

Best regards,

T. Oka


Reply all
Reply to author
Forward
0 new messages