No electronic convergence in Fe grain boundary with Yb solute
Posted: Sun Apr 17, 2022 12:18 pm
Dear Developers,
Please find attached files for a Fe grain boundary system with a single substitutional Yb atom, a type of system that I have found to consistently run into convergence issues (of both electronic and technical kind):
I have tried the RMM algorithm with/without spin polarisation (ISPIN = 1/2), davidson algorithm with and without spin, and davidson with no charge update (ICHARG = 12) with generally no success.
Directory containing example output is in Yb-11-d-6.5:
with spin polarisation, in RMM: electronic convergence fails
with spin polarisation, in DAVIDSON: crashes - the old and the new charge density differ BRMIX serious problems -> WARNING: Sub-Space-Matrix is not hermitian in DAV
without spin polarisation, in RMM: electronic convergence is never attained
without spin polarisation, in DAVIDSON: crashes - the old and the new charge density differ BRMIX serious problems -> WARNING: Sub-Space-Matrix is not hermitian in DAV
Rarely convergence is attained for specific structures, but I find that electronic convergence for these types of systems (Fe-GBs with 1 Yb atom) fails quite often in the manners listed above for different substitutional positions along the GB and also for different GB structures (not included). I am using vasp 5.4.4 with the PBE functional. I have tried on multiple HPC systems with similar results. I believe this has something to do with Yb specifically, since I experience no such issues for other lanthanides.
Any guidance/help is much appreciated.
Best, Han.
Please find attached files for a Fe grain boundary system with a single substitutional Yb atom, a type of system that I have found to consistently run into convergence issues (of both electronic and technical kind):
I have tried the RMM algorithm with/without spin polarisation (ISPIN = 1/2), davidson algorithm with and without spin, and davidson with no charge update (ICHARG = 12) with generally no success.
Directory containing example output is in Yb-11-d-6.5:
with spin polarisation, in RMM: electronic convergence fails
with spin polarisation, in DAVIDSON: crashes - the old and the new charge density differ BRMIX serious problems -> WARNING: Sub-Space-Matrix is not hermitian in DAV
without spin polarisation, in RMM: electronic convergence is never attained
without spin polarisation, in DAVIDSON: crashes - the old and the new charge density differ BRMIX serious problems -> WARNING: Sub-Space-Matrix is not hermitian in DAV
Rarely convergence is attained for specific structures, but I find that electronic convergence for these types of systems (Fe-GBs with 1 Yb atom) fails quite often in the manners listed above for different substitutional positions along the GB and also for different GB structures (not included). I am using vasp 5.4.4 with the PBE functional. I have tried on multiple HPC systems with similar results. I believe this has something to do with Yb specifically, since I experience no such issues for other lanthanides.
Any guidance/help is much appreciated.
Best, Han.