CLSEGV **** NOT PAIRING AT K

Alexander

Hello,

I get this error after optimising the geometry of one of solid solution structures:

[SYMMETRY ALLOWED INTERNAL DEGREE(S) OF FREEDOM: 6
TOTAL NUMBER OF PARTNERS = 2
NUMBER OF THE FIRST COMPONENT = 2
STAR VECTOR = 1
IRREDUCIBLE REPRESENTATION = 1
ERROR **** CLSEGV **** NOT PAIRING AT K = 7
----------------------------------------------------------------]

The optimization starts as expected, but after reaching convergence criteria and going for FINALRUN calculations stops. Using obtained geometry for OPTGEOM restart or FREQCALC results in this error. All other 8 solid solutions were optimized smoothly.

CrTaO4_SIC4_FREQ.out
CrTaO4_SIC4.f34
CrTaO4_SIC4.out

fbodo

Hi Alexander,
I took a look into your calculations and I performed a few tests on them.

As you said, the calculation seems to break at the end of the optimization as it enters in the FINALRUN. This is because when you perform a geometry optimization the bielectronic integrals are indexed based on the starting geometry and they are kept the same throughout the entire optimization process to avoid any numerical instabilities. With FINALRUN active, the code performs a new indexing that in your case leads to the issue you observed.

Working a bit with the OPTGEOM input you gave us I found two possible workarounds:

You can either deactivate the FINALRUN option, and if you feel safer you can perform a second optimization on the geometry you obtain


OPTGEOM
FINALRUN
0
END
[...]

An alternative option is instead to increase the TOLINTEG used in the optimization, and, in this case, I didn't encounter any issue using

[...]
TOLINTEG
10 10 10 10 20
[...]

For both cases I have attached an output below
opt_finalrun.out
opt_tolinteg.out

Alexander

Thank you!
I used optimized geometry with increased TOLINTEGs for starting FREQCALC calculation (with old TOLINTEGs) and everything works fine.
Still curious, why was that .f34 file bad for re-starting OPTGEOM or starting FREQCALC?

fbodo

I honestly don't know, because when I tried to use your .f34 with the input you provided the FREQCALC was starting without any issue.