S matrix not positive definite.
I am using Quantum Espresso code. For some calculations the crystal structure is define as fcc lattice with four basis. For some other calculations ,I define the same crystal as sc lattice with 16 basis. For fcc lattice,the scf converges while with the same parameters for sc lattice, the scf calculations does not converges and shows the following error "S matrix not positive definite". Please suggest me where I am going wrong.
This commonly happens when doing tight-binding calculations on crystals when using diffuse basis functions and too small a distance cutoff for truncating overlap and Hamiltonian matrix elements. If you increase the distance cutoff, the problem should be resolved.
Respected Sir Hughbanks, thank you so much for your suggestion. In quantum espresso code , I have kinetic energy cut off and not the distance cutoff. Do they mean the same ? And I shall try increasing the kinetic energy cut off as suggested by you. Thank you once again sir.
change pseudopotential, sometimes pseudopotental files r corrupted.
As previously explained ( By The developers itself. This is toughest error to identify and resolve)
1. Could be serious error in data, such as bad atomic positions or bad crystal structure/supercell; ---> check your output file ( .axsf file if you are QE) using xcrysden or other software.
If images are very very close , NEB calculation will not work.
2. a bad pseudopotential, typically with a ghost, or a USPP giving non-positive charge ---> I don't know what is ghost states, i heard of that in papers and in community discussion. In Some cases Pseudopotential can't changes since that that psedoP is used for all the corresponding set of calculation
3. Run with less processors + Use intermediate images from crashed result + Increase cut off of E. is another resolution. ( Might work)
4. Also Run with 'cg' diagonalization. I don't know reason. But i say it from experience. Less chance for this error with 'cg'.
5. After all the try even if fails, Try restart again with same parameters . Believe me it works for me
If any one have better idea, please explain. Thank you.
I had the same problem. I found the solution checking my output file as Ajmalghan Muthali said. In my case, I forgot to specify units in cell_parameters card in input file.
All the responses and suggestions given above are correct - depending on the origin of the run-time error that caused the crash. In my case, setting the correct unit of the Cell_Parameter and Atomic_Positions keywords was sufficient to resolve this error. Good luck with the debugging!
I encountered this problem when trying to run an scf calculation on an HPC cluster, while the same input file runs perfectly fine on my laptop.
I ran the calculation on my laptop in serial execution just using pw.x < input but doing the same on the cluster produced this error. I was able to fix it by using parallel execution like mpirun -np 32 pw.x < input.
I encountered this " S matrix not positive definite " error. The problem was fixed by increasing the kinetic energy cutoff even with Optimized Norm-Conservinng Vanderbilt Pseudopotentials (*.ONCV_PBE-1.2.upf).
Thank you, Increasing kinetic energy cutoff worked for me!
Try increasing diago_david_ndim in electron card to 4 or even 8. Increasing this parameter has worked for me every time.
Check if there is an overlap of two atoms, this might happen when you are optimizing the lattice parameter. You may open your input file in BRUAI and check for the overlap.
I encountered the same problem when I was trying to study the heterostructure. I found that the error was due to overlapping images. Try to open your structure in VESTA and look closely, you may be able visualize the overlap.
One possible solution for this problem is to remove periodicity from your strcuture along lattice vectors. In my case I removed the perodicity along a and b direction.
PS: if you are using heterostructures, then you should run the structures separately in the same cell. For example if you have Mgo/CuO hetero-structure try to run Mgo and Cuo separately with heterostructure cell dimension, this will help you in identifying which structure is causing problem. (Credits: Prof. Krishna balasubramanian, IIT Delhi)
If you want to understand what this error means, you might find some of the information here useful:
https://www.researchgate.net/post/What-is-the-solution-to-solve-the-error-Error-EDDDAV-Call-to-ZHEGV-failed-Returncode-29-2-56/1
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Materials Science