poor geometry convergence even with accurate parameters

[cpp6f]

I have a problem that the geometry convergence slows down considerably once the forces are < 0.1 eV/A. I am using all the accurate parameters:

PREC = ACCURATE
LREAL = FALSE
ADDGRID = TRUE
EDIFF = 1e-6
NELMIN = 8

IBRION = 1
POTIM = 0.2
ENCUT = 396
EDIFFG = -0.01

I am using the USPP for the ions. Why does the convergence slow down so much? Shouldn't it speed up since the model becomes more linear close to the minimum?

Also, increasing ENCUT doesn't affect the convergence either - and it shouldn't - I can converge the geometry of H2 using a minimal gaussian basis to extraordinary accuracy, even though the basis is poor.

Just for reference, the systems I am studying are adsorbates on a metal surface.

Thanks for any suggestions.

[admin]

whether ENCUT is sufficiently high depends on the POTCARs of the ions you use, please set it at least to 1.3*max(ENMAX).
maybe you got stuck in a local minimum?
please try a different optimization algorithm to check whether the convergence can be improved..