Convergence problem with respect to ENCUT

[bhakti]

Dear VASP users,

I am studying few systems and their interactions with other molecules, which are of industrial interest.
While doing this study, I came across some interesting problem of very slow convergence of the total energy of system with respect to ENCUT value.
As I started calculation at higher ENCUT, the total energy of system goes on increasing compared to previous cutoff. Which is bit odd, because, in principle higher the basis more accurate are the results, at least it should converge to some constant value.
With increase of plane waves a constant warning appears in the outfile saying "WARNING: PSMAXN for non-local potential too small". For this particular warning VASP users suggest to decrease the ENCUT value in INCAR file. Since my main interest was to see convergence wrt ENCUT value these settings were useless for my calculations.
Along with this warning another warning also appears in the outfile saying "Your FFT grids (NGX,NGY,NGZ) are not sufficient for an accurate calculation. The results might be wrong good settings for NGX NGY and NGZ are 88 87 and 183", with the help of this, I re sized the FFT mesh with the suggested values and now the results are converging wrt ENCUT once again well. (although first warning still appears in this new outfile )

Do these new energy values obtained with new settings are reliable?
Do I need to repeat previous calculations of lower cutoff with this redefined FFt mesh size? (since my aim is to study convergence and interaction with respect to cutoff)

Thanking you all in advance,

bhakti

[admin]

1) it may well happen that -although ENCUT is increased- the total energies become slightly less negative, due to the atomic convergence corrections (please have a look at the accuracy.pdf lecture of the VASP-workshop)
2) PSMAXN warning: the parameter is given in the POTCAR file. in principle, please use potentials of similar hardness (for different atoms) and -if ENCUT is increased- switch to harder potentials if possible.
3) re-sized FFT mesh: yes, the results are reliable if you explicitely use the suggested FFT-meshes
The FFT meshes are adjusted to the required cutoff radius, therefore please use the suggested meshes for lower cutoffs (if you want to set them explicitely).