I am working on DFT calculations on halide perovskites using VASP. My system consists of 56 atoms and is reported to possess monoclinic symmetry. I am performing k mesh optimization for this system using following INCAR (eg. at different k-meshes 1x1x1, 2x2x2 etc.)
ISTART = 0
ICHARG = 2
ISPIN = 2
# ENCUT = 363.00 eV
EDIFF = -0.000001
EDIFFG = -0.0001
IALGO = 38
IBRION = 1
POTIM = 0.5
NSW = 150
PREC = MEDIUM
LREAL = Auto
ADDGRID = T
NPAR = 8
However, my system is not getting converged even after these many ionic iterations. I tried copying CONTCAR to POSCAR and re-running vasp but at the end the OUTCAR isn't showing "reaching structural energy minimization". How can I obtain converged energy for this system? Can anyone please suggest the solution for this ?
Dear Shruthi Nair ,
A few questions:
What type of convergence problem are you experiencing, SCF or geometry? If the former, then there are a bevy of possible ways to improve convergence. If the latter, see below.
What type of halide perovskite are you working on? If it's hybrid organic-inorganic, their potential energy surface near the minimum is very flat because molecular reorientations only slightly change the energy. In other words, convergence of the total energy is hard in this system.
How many steps have you run? It's entirely possible that you just have to keep plugging away.
How close is your calculation to being converged? You could try switching to a different algorithm to relax the ions, e.g. IBRION = 2 (conjugate gradient, which is quite robust). You could also try switching to a force convergence criterion, e.g. EDIFFG = -0.03 (you can interpret this as 0.03 eV/Å/atom).
The bottom-line is that it can be computationally expensive to study the halide perovskites as they are big and slow-to-converge. SCF shouldn't be a problem though because they are insulators.
Best,
Rob
As Robert Wexler pointed out, I suggest you to change the " EDIFFG " of your calcualtion to a littile higher value such as "EDIFFG = -0.01" and change the EDIFF = -0.000001 to 1.0e-05 or 1.0e-04. Most probably, It will try to converge your calculation.I am thinking that it is not related to a serious potential energy surface issue. How ever, Robert Wexler comrehensively discussed about your issue in different contexts. So, if these seetings not solve your problem, please look into his other suggestions.
Robert Wexler Thank you sir for the answer. Actually, I am not working on hybrid perovskites, it's inorganic. As you suggested I tried using EDIFF=-0.0001 eV and EDIFFG=-0.003eV and my problem got solved.
Thank you , your answer was really helpful.
Naga Venkateswara Rao Nulakani Thank you for your help.
I tried changing the energy and force convergence criteria and my problem got solved.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Physical Phenomena
- Luminescence
- Fluorescence