Problem: Singular value decomposition failed

[Qiang Fu]

Dear Experts,

We performed a mGGA(TB09LDA) calculation for the transport properties of a doped-semiconductor/metal interface. The basis set of the metallic element was set to be LDABasis.Element_DoubleZeta, while for other elements LDABasis.Element_DoubleZetaPolarized were used. The bias was set to be 1.0V(Left electrode: +0.5V and Right electrode: -0.5V).

In the calculation, the three parts: Left Electrode Calculation, Right Electrode Calculation, and Equivalent Bulk finished well. However, in the following part, we encounter the following error message:

in the output file:
--------------------------------------------------------------------------------
rank 15 in job 1  g071.dcsc.fysik.dtu.dk_39347   caused collective abort of all ranks
  exit status of rank 15: killed by signal 9
rank 13 in job 1  g071.dcsc.fysik.dtu.dk_39347   caused collective abort of all ranks
  exit status of rank 13: return code 134
rank 10 in job 1  g071.dcsc.fysik.dtu.dk_39347   caused collective abort of all ranks
  exit status of rank 10: return code 134
rank 9 in job 1  g071.dcsc.fysik.dtu.dk_39347   caused collective abort of all ranks
  exit status of rank 9: return code 134
rank 6 in job 1  g071.dcsc.fysik.dtu.dk_39347   caused collective abort of all ranks
  exit status of rank 6: killed by signal 9
rank 5 in job 1  g071.dcsc.fysik.dtu.dk_39347   caused collective abort of all ranks
  exit status of rank 5: return code 134
rank 3 in job 1  g071.dcsc.fysik.dtu.dk_39347   caused collective abort of all ranks
  exit status of rank 3: return code 134
rank 2 in job 1  g071.dcsc.fysik.dtu.dk_39347   caused collective abort of all ranks
  exit status of rank 2: return code 134
rank 0 in job 1  g071.dcsc.fysik.dtu.dk_39347   caused collective abort of all ranks
  exit status of rank 0: return code 134
--------------------------------------------------------------------------------

in the log.err file
-----------------------------------------------------------------------------------------------------------------------------------
** Back Engine Exception : Singular value decomposition failed
** Location of Exception : mathutils.cpp:870


/home/opt/common/QuantumWise/atk-13.8.0/atkpython/bin/atkpython: line 7:  7673 Aborted                 $EXEC_DIR/atkpython_exec $*
/home/opt/common/QuantumWise/atk-13.8.0/atkpython/bin/atkpython: line 7:  7670 Aborted                 $EXEC_DIR/atkpython_exec $*
-----------------------------------------------------------------------------------------------------------------------------------

Could you please give us some advice to fix the problem? Many thanks in advance.

It is worth noting that calculations at a lower bias is running well now (has not finished). In addition, calculation at the same bias (1.0V) but using PBE functional finished successfully.
We use Atomistix ToolKit 13.8.0 for our calculations.

Have a nice day!
Best regards,
Qiang

[Umberto Martinez]

It seems like a network error.

[Qiang Fu]

It seems like a network error.

Thank you very much 

[Anders Blom]

I'm not so sure about that. The abort of the MPIs is caused but the C++ runtime error, which is an unusual one. Try keeping a fixed value of the c parameter in the TB09 functional instead, and see if that helps. It would be most appropriate to do so anyway. The value should be found from the bulk calculation of the material, or the most important material in the system in case you have several.

[Qiang Fu]

I'm not so sure about that. The abort of the MPIs is caused but the C++ runtime error, which is an unusual one. Try keeping a fixed value of the c parameter in the TB09 functional instead, and see if that helps. It would be most appropriate to do so anyway. The value should be found from the bulk calculation of the material, or the most important material in the system in case you have several.

Thank you very much for your suggestion, Anders. I did not have a good understanding until yesterday. Your advice is very insightful. Now I realize that our setting for mGGA calculations was not appropriate. We will make the correction.