by scf calculation and change the value of cutoff you can optimize that.

You need to ensure that the properties you're interested in are well-converged with respect to the cut-off energy. The best way to do that is to perform a series of calculations with different cut-off energies, and see how your results change. Higher cut-off energies are always more accurate, but they take more computer memory and time; any simulation is always a compromise between the accuracy you get (higher is better) and the computational resources needed to achieve it (lower is better).

As a starting point, the first few lines of the pseudopotential files contain some estimates of cut-off energies. The first two ("coarse" and "medium") are not suitable for real calculations, they are mostly for quick prototype simulations, but anything from "fine" and above could be sufficient.

Sometimes the property you're really interested in might be very computationally demanding. For example, if you want the phonon dispersion, it would be very time- consuming to compute a full dispersion for each cut-off energy. In these cases, you should think about what the property really depends on, and whether there is a simpler way or proxy property which you can converge instead. In the example of the phonons, the usual proxy is the stress on the simulation cell, since this governs the behaviour of the long-wavelength acoustic phonons.

Hope that helps,

Phil Hasnip

(CASTEP Developer)

Philip James Hasnip

Thank you for your response.