[CPMD-list] HGH type pseudopotentials

[Juerg Hutter]

Dear Howard

> Dear Prof. Hutter and all:
>
> I used HGH type pesudopotentials to simulate transtion metals. I found that
> the HGH psedopotentials require larger Ecut to achieve full convergence. For
> instance, in the case of Cu (among other metals), an accutrate energy
> calculation requires an energy cutoff of 130 to 150 rydberg, which is few
> times more than other pseudopoentals like TM and USPP. As a result, the

That's correct and a typical behavior of this type of PP.

> enormous numbers of planewaves needed by HGH make the simulation
> prohibitively slow. As such, I think HGH has no apparent advantages over

This depends on your computing resources. High cutoffs might be a major
drawback on small computers and workstations but might be rather
irrelevant on massively parallel computers and vector processors.
Simple PP (no USPP) and high cutoffs provide better parallelisation
and vectorisation such that often these jobs run at least as fast
on high-end computers as low cutoff jobs.

> other types. Would you please comment on this?

Dual-space PP were developed to be optimally transferable and have
optimal properties for real-space algorithms. This can only be achieved
if smoothness in Fourier space is sacrificed.

Dual-space PP are often used in NON-plane wave based codes or for
testing purposes were smoothness is not of importance.

regards

Juerg Hutter