[CPMD-list] Energy components in PIMD

[Łukasz Walewski]

Dnia 29-04-2008, Wt o godzinie 07:26 -0400, Axel Kohlmeyer pisze:
> On Tue, 29 Apr 2008, ukasz Walewski wrote:
> 
> LW> Dear All,
> 
> dear lukasz,

Dear Axel,

Thank you for the rapid reply.

> LW> Please note that I use termostat both for ions and electrons, so I would
> LW> expect kinetic energy to by conserved in such a job.
> 
> see... i guess here is the basic mistake. have you done any simulations
> _without_ a thermostat? if you have them on both electrons and ions
> you are _completely_ blind as to whether you are actually simulating
> reality or just generating entropy.

Oh, yes! I forgot to mention: of course I did the CPMD (non-PIMD) NVE
(http://bioexploratorium.pl/ljw/cpmd/nve/) and NVT
(http://bioexploratorium.pl/ljw/cpmd/nvt/) test simulations (input files
together with the plots).

> so before doing anything else you should verify the following:
> 
> can your input conserve energy for a non-PIMD run?
> ...and then you have to check whether you can maintain
> adiabacity with the massive thermostat on the ions only.

This is exactly what I did. The NVE CPMD simulations do conserve total
energy very well (at least for 1ps runs). Thermostatting the ions makes
the fictitious kinetic energy of the electrons drift after 3ps of
simulation (http://bioexploratorium.pl/ljw/cpmd/nvt/ekinc.png) and this
is what I realized already after performing some PIMD tests (the fault
of my impatience).

> quite a few times, the no-thermostat CP-dynamics is only
> meta-stable and when you switch to massive N-H chains you
> excite DOFs that will make the CP dynamics lose it.
> 
> then you should try PIMD without a thermostat on the 
> electrons. that has to be reasonably stable as well.

This is what I also did already (again - sorry not to mention this
beforehand). The plots here http://bioexploratorium.pl/ljw/pimd/tests/
contain three panels: bottom - no thermostats, different electron mass
(bottom.inp); middle - thermostat on electrons (middle.inp); top -
thermostat both on electrons and the ions, different time steps
(top.inp).

> when switching to PIMD you have to realize that your
> "atoms cores" become lighter and move faster, so you
> will most likely have to reduce the time step. it
> is also quite tricky to make sure that the PI-DOFs
> and the fictitious electron DOFs don't couple.

Yes, that is right.

> LW>   ANGSTROM
> LW>   SYMMETRY
> LW>     8
> LW>   CELL ABSOLUTE
> LW>     13.0 13.0 8.0 0.0 0.0 0.0
> LW>   CUTOFF
> LW>     30.0
> 
> hmmmm.... 30ry cutoff?
> 
> 
> [...]
> LW> 
> LW> &ATOMS
> LW> *N_VDB_BLYP.psp FORMATTED
> 
> ...and ultra-soft pseudopotentials.
> i doubt that this will conserve energy
> well even for classical atom cores.
> at the very least you have to crank
> up the density cutoff (or dual) to
> have a tight enough real space grid
> to accommodate the augmentation charges.

This is what I did not know. I performed a lot of single point tests
with different cutoff and cell size and decided that 30 Ry is a "safe"
value for electronic structure calcs with USPP's. Is it all different
when it comes to dynamics ?

Regards,
Lukasz