Positron calculations fail with forrtl: severe (174): SIGSEGV, segmentation fault occurred

[rajaraman]

Dear all,
I have been doing positron calculations in vacancies in various systems.
My recent system FeB gave lot of trouble in optimizing and completing runs.
system is 95 atom FeB supercell with B vacancy.
Run supposed to self consistently relax the structure taking into account positron induced forces.
After completing the electronic convergence step my job fails at start of positron convergence

potentials are generated from JTH Input files and groundsate lattice parameters and magentic moments are fine.
simple two-step positron calculations go through without problem even for both Fe and B vacancies.
failure occur for self-consistent positron calculations

tried various combinations and job keep failing at positron step.

Attached input and output files
log file is big and given below is the fail point

Where am I going wrong?
advanced thanks
-rajaraman

------------------
Tail of out file
ETOT 57 -6115.2796550569 7.27E-07 8.17E-08 2.43E-04 4.5E-05 9.72E-03 54.83

At SCF step 57, etot is converged :
for the second time, diff in etot= 7.266E-07 < toldfe= 1.000E-06

TC-DFT STEP 2 - POSITRONIC GROUND-STATE CALCULATION IN PRESENCE OF ELECTRONS AND IONS

Failure part in log file
Component down:
25.99451 -6.25429 0.00067 -0.00017 -0.00005 -0.00065 0.00016 0.00005 -0.00006 -0.00098 0.00058 -0.00009 ...
-6.25429 1.53067 -0.00014 0.00002 0.00001 0.00005 -0.00011 -0.00002 0.00001 0.00012 -0.00008 0.00001 ...
0.00067 -0.00014 21.11825 -0.00076 -0.00006 -28.30960 0.00070 0.00005 -0.00007 -0.00101 -0.00080 0.00002 ...
-0.00017 0.00002 -0.00076 21.11983 -0.00007 0.00070 -28.31112 0.00007 0.00002 -0.00014 0.00076 -0.00010 ...
-0.00005 0.00001 -0.00006 -0.00007 21.11975 0.00005 0.00007 -28.31109 0.00116 0.00002 -0.00004 -0.00008 ...
-0.00065 0.00005 -28.30960 0.00070 0.00005 38.26167 -0.00016 -0.00005 0.00007 0.00084 0.00061 -0.00001 ...
0.00016 -0.00011 0.00070 -28.31112 0.00007 -0.00016 38.26240 -0.00005 -0.00001 0.00018 -0.00042 0.00009 ...
0.00005 -0.00002 0.00005 0.00007 -28.31109 -0.00005 -0.00005 38.26237 -0.00084 -0.00001 0.00002 0.00018 ...
-0.00006 0.00001 -0.00007 0.00002 0.00116 0.00007 -0.00001 -0.00084 10.18214 0.00007 0.00007 -0.00058 ...
-0.00098 0.00012 -0.00101 -0.00014 0.00002 0.00084 0.00018 -0.00001 0.00007 10.18099 0.00019 0.00000 ...
0.00058 -0.00008 -0.00080 0.00076 -0.00004 0.00061 -0.00042 0.00002 0.00007 0.00019 10.18183 -0.00016 ...
-0.00009 0.00001 0.00002 -0.00010 -0.00008 -0.00001 0.00009 0.00018 -0.00058 0.00000 -0.00016 10.18273 ...
... only 12 components have been written...


ITER STEP NUMBER 58
vtorho : nnsclo_now=1, note that nnsclo,dbl_nnsclo,istep=-3 0 58
You should try to get npband*bandpp= 504
For information matrix size is 3202
forrtl: severe (174): SIGSEGV, segmentation fault occurred

[rajaraman]

I found that "optstress 1" is giving me this error.
I switched on stress calculation to first see how the vacancy relax without positron (normal structural relaxation run).
Once I put it off with "optstress 0", positron optimisation steps are going through.

But testing with smaller supercell as given in positron tutorial-4 (https://docs.abinit.org/tutorial/positron/index.html) does not make any issue with optstress being "on" or "off".

However "Segmentation fault" recurrs when I use 64 atom supercell for silicon in our parallel machine with "optstress 1".
May be some thing todo with our parallel environment and my compilation.

Any way for positron calculations We dont need the stress calculations as long as our super cell is big enough.
I will get back if this issue resurfaces again with "optstress 0".

Thanks
-rajaraman