[Pw_forum] CP simulation with ensemble-dft algorithm

[Giovanni La Penna]

Dear Pascal and QE-users,

 In the *.evp (and in stdout) of CP, ekinc (Kel) and actual T are printed,
as, respectively, columns 2 and 4 (in *.evp). Kel is the fictitious
kinetic energy
of electronic variables.
Empirically, I found that:

Kel/Kion < 1/10

for a reliable CP-MD run, with Kion= 3/2 Ndeg R T . Since Kel is
(please confirm) in Ha,
R must be converted in Ha/K.
Even though the word reliable may be an opinion, when the ratio
becomes larger, then Kel starts diverging very quickly (while in the
other case only a slow drift occurs).
I would be glad if other CP users could confirm this empirical observation.

                      Giovanni La Penna (Cnr-Iccom, Italy)