Forrtl: error (76): Abort trap signal

[markperri]

Dear Psi4,

I’m trying to optimize Cr(CO)6. It starts the optimization and then just dies. I’m running it in a singularity container, but other jobs run fine. I’ve attached the input and output file. The error I get is:

forrtl: error (76): Abort trap signal
Image              PC                Routine            Line        Source
libgdma.so         00007F08E069910F  for__signal_handl     Unknown  Unknown
libpthread-2.28.s  00007F08F4C08C20  Unknown               Unknown  Unknown
libc-2.28.so       00007F08F486837F  gsignal               Unknown  Unknown
libc-2.28.so       00007F08F4852DB5  abort                 Unknown  Unknown
core.cpython-38-x  00007F08E2BB39B0  Unknown               Unknown  Unknown
core.cpython-38-x  00007F08E53C9C6D  Unknown               Unknown  Unknown
core.cpython-38-x  00007F08E54D21E2  Unknown               Unknown  Unknown
core.cpython-38-x  00007F08E509FD69  Unknown               Unknown  Unknown
python3.8          000055CCC1AD7834  PyCFunction_Call      Unknown  Unknown
python3.8          000055CCC1A98EA5  _PyObject_MakeTpC     Unknown  Unknown
python3.8          000055CCC1AF76F8  _PyEval_EvalFrame     Unknown  Unknown
python3.8          000055CCC1AC49C9  _PyEval_EvalCodeW     Unknown  Unknown
python3.8          000055CCC1AC585F  _PyFunction_Vecto     Unknown  Unknown
python3.8          000055CCC1A72CC7  Unknown               Unknown  Unknown
python3.8          000055CCC1AC49C9  _PyEval_EvalCodeW     Unknown  Unknown
python3.8          000055CCC1B24383  PyEval_EvalCode       Unknown  Unknown
python3.8          000055CCC1B248D2  Unknown               Unknown  Unknown
python3.8          000055CCC1B34982  Unknown               Unknown  Unknown
python3.8          000055CCC1B3621D  PyRun_StringFlags     Unknown  Unknown
python3.8          000055CCC1B0F053  Unknown               Unknown  Unknown
python3.8          000055CCC1A9C61D  Unknown               Unknown  Unknown
python3.8          000055CCC1A72E14  Unknown               Unknown  Unknown
python3.8          000055CCC1AC49C9  _PyEval_EvalCodeW     Unknown  Unknown
python3.8          000055CCC1B24383  PyEval_EvalCode       Unknown  Unknown
python3.8          000055CCC1B248D2  Unknown               Unknown  Unknown
python3.8          000055CCC1B34982  Unknown               Unknown  Unknown
python3.8          000055CCC1B36429  Unknown               Unknown  Unknown
python3.8          000055CCC1B3660F  PyRun_SimpleFileE     Unknown  Unknown
python3.8          000055CCC1B36A09  Py_RunMain            Unknown  Unknown
python3.8          000055CCC1B36CA9  Py_BytesMain          Unknown  Unknown
libc-2.28.so       00007F08F4854493  __libc_start_main     Unknown  Unknown
python3.8          000055CCC1B0F1C7  Unknown               Unknown  Unknown

In the optimization section it’s got: 3 R(1,4) = 1.915766 0.377776 321.034773 322.950540
Does that mean it’s increasing that bond length by 321 A?

Thanks,
Mark

CrCO6.in.txt (978 Bytes)
CrCO6.out.txt (31.8 KB)

[jmisiewicz]

Correct, it’s increasing the bond length by 321 A. This is, as you might imagine, not a sane thing to do. We’re working on some changes to the optimizer that will hopefully prevent this. In the meantime, try the keyword ensure_bt_convergence true.

Obligatory @Rollin_King ping.

[Rollin_King]

Yikes! That’s an interesting test, with the multitude of almost 180 angles. I would try just optimizing in Cartesian coordinates.

opt_coordinates = cartesian

[markperri]

Thanks! opt_coordinates cartesian fixed it.

ensure_bt_convergence true tried really hard, but ended up with: Unable to back-transform even 1/10th of the desired step rigorously.

Can I set ensure_bt_convergence true in all runs, just in case it might help? Does it cause any problems / slow downs?

Thanks,
Mark

[jmisiewicz]

Marked as solved.

I have ensure_bt_convergence on by default. There may be some corner cases where it causes slow downs, but in my experience, that’s never been a problem.