Grid for Broyden might be to small - warning

[RChattopadhyay]

Hello,

I am running single point calculations to find out optimum values of k-points and energy cutoff that meets a convergence criteria. The calculations run successfully but with a warning - WARNING: grid for Broyden might be too small. How can I suppress this warning ? Does this have significant effects on forces and energies calculated for insulators ? My INCAR file is as follows -

SYSTEM = Rw_energy_convergence

METAGGA = R2SCAN
NCORE= 6
ISPIN = 1 # Spin-relaxation: 1 = OFF, 2 = ON
ISYM = 1 # Symmetry: 0 = no-sym; 1 = search for sym

### Electronic Relaxation
EDIFF = 1E-6 #
NELM = 300 # Maximum relaxation steps

### Ionic Relaxation
EDIFFG = -0.01 # neg = forces; pos = energy
NSW = -1 # -1 = single-point; 50 for relaxation
ISIF = 2 # 2 = only ions; 4 = ions+shape; 5 = shape
IBRION = -1 # 1 = quasi newton method; 2 = CG; -1=single point

### Plane Waves
PREC = Accurate # Accurate, Normal
ENCUT = 1300 # Kinetic Energy Cut-off
KSPACING = 0.23
KGAMMA = .TRUE.

### MetaGGA
LASPH = .TRUE. #recommened for metagga
LDIAG = .TRUE. #generally recommened for ALGO=Fast
LMIXTAU = .TRUE. #recommended if IALGO=38. ALGO=Fast includes some sweeps of IALGO=38

### FFT Grid
ADDGRID = .TRUE.
LREAL = Auto #generally recommended for systems with more than 30 atoms

### Matrix Diagonalization
ALGO = Fast

### DOS related
ISMEAR = -5 # tetrahedral method
SIGMA = 0.1
LORBIT = 11 # atom-, l-, ml-projected
NEDOS = 2000

I thank you in advance for your assistance.

[ferenc_karsai]

Please upload the full calculation according to the forum guidelines.

[RChattopadhyay]

Hello,

I did not know about the guidelines of the forum. I am sorry for my ignorance.

I have uploaded all the relevant input and output files of the calculation. I find that VASP issues a warning - WARNING: grid for Broyden might be too small on some of the scf steps. I tried modifying the AMIX flag (based on a previous discussion thread in the forum) but this does not seem to remove the warning. However, I found that the change in total energy is of the order of 10-4 meV/atom when the AMIX flag is changed. This is around 3 orders of magnitude smaller than the convergence cutoff I am aiming for (wrt ENCUT and KSPACING). But still I am curious to know if this warning may be removed before I perform production calculations.

I thank you for your assistance and suggestions.

[ferenc_karsai]

I've played around with the parameters in your calculations.
As you wrote, changing the mixing parameters will not help. Also increasing ENCUT, PREC etc. which result in much more grid points did not help.

You have set LMIXTAU = .TRUE.. I guess you do this because it is written in the wiki to do that for IALGO=38 and meta gga's.
I tried to remove that tag and then the WARNING disappears, but I don't know why. That tag is important if your calculations struggle to converge.
I could easily converge the calculation without that tag.
So you should try omitting that tag if you have no convergence problems.