Error(seceqnfv): diagonalisation failed

exciting
1

Hi Exciting Community, I am new user of the code. I am facing a diagonalization error as below:

Error(seceqnfv): diagonalisation failed
ZHPGVX returned INFO = 914
The leading minor of the overlap matrix of order 448
is not positive definite
Order of overlap matrix : 466

Elapsed time = 12m17s

Please look below[1] for the corresponding output files.

[1]

2

Dear Jatin19k,

Your structure seems to be very “compressed”. Could you check if the structure input data are given correctly in atomic units (Bohr)? I have the feeling you were using Angstrom instead…

Your sincerely,

Pasquale Pavone (exciting team)

3

Dear Pasquale,

Yes, you are correct. I am pulling the systems from VASP which uses Angstrom.
Is there any utility that I can use (inside/outside of Exciting) by which I can convert my structures from Angstrom to Bohr. There are around 20 structures so can’t do it by hand calculations.

Thank you.

4

Can you give more details on what you’re using to get the structures from VASP into exciting’s XML input?

Cheers,
Alex

5

So far I just manually copy pasted the coordinates from POSCAR file of VASP into xml file, since I am just setting up the Exciting framework.

Thank you.

6

I think you should be able to use atomic simulation environment to read POSCAR into their Atoms() object (and convert the units).

Currently you can also use the ASE calculator to run exciting calculations, however it’s a bit of a black box. As of next week, excitingtools will be installable via pip, which will allow more flexibility for creating exciting input files with python and ASE.

If you give this thread a poke in a week, I will update with the details but in the meantime, I think you can us ASE (in python) to convert the files, then write your own routine to write the XML string, or try running exciting calculations with ASE

Cheers,
Alex

7

Thanks Alex, now I am using scaled coordinates instead converting units. But now I am facing muffin tin spheres overlap error. If I reduces the rmt just enough to get rid of it then it starts giving diagonalization error similar to the 1st post. Any hint, what maybe going on?

Error(checkmt): muffin-tin spheres overlap between
species 2 atom 4 and
species 1 atom 1

Sum of muffin-tin radii : 2.0000
Distance between atoms : 1.7946

8

Hmm, I am slightly confused actually. I think exciting only takes positions in Bohr, however the manual suggests coord are in fractional coordinates. @PasqualePavone could you clarify?

Please try it in Bohr, with the default MT radii and let me know.

Cheers,
Alex

9

It worked with fractional coordinates itself, turns out I was not doing them correctly at first place. However I need to check the structure of this ionic relaxation job. Is there any way I can visualize frame by frame animation/structure directly from the Exciting output files.

Thank you.

10

You can have a look to the following attribute options of the element relax ([Relax - exciting]):

  • history
  • historyformat

Best,
Pasquale Pavone (exciting team)

11

Thanks it is working, is there any way to generate this file when I am just performing SCF cycle and not relaxation? Additionally, sometime it skips providing space between adjacent columns for some of the xyz file entries and the parser code gets confused and refuses to open file.

I am working on a smaller system for now so was able to manually edit the file but can be a difficult for a bigger structure.

Thank you.