[QE-users] how to set the electron thermostat fnosee in CP molecular dynamics

[Eduardo Menendez]

Hi fellows,
I am exploring the Car Parrinello dynamics for a system with hydrogens:
CH3NH3PbI3. With N-H vibrations faster than 3000 cm^-1 and a
fluctuating bandgap that may be as small as 1.5 eV , I guess I will need to
use a thermostat for the electrons, in order not to use a very small emass
and dt.
Estimating a minimum electronic frequency sqrt(2Eg/emass), using the
default emass=400, I get omega_min~0.016 atomic units. For the N-H
vibrations, the frequency is 3000 cm^ -1~ 0.014 atomic units .
I read elsewhere than a good frequency for the electron thermostat is 2-3
times the maximum phonon frequency (90THz), this led me to the value
fnosee=270.0
My guess is 270 times larger than the default value fnosep=1.D0. Running in
a parameter space different to what is tested is generally not a good
idea.  Am I missing something ?

A related, but independent question is about the ion thermostat.
The vibrational frequencies of this system are very well separated: ~3000
cm^-1 for N-H and C-H,
900-1200cm^-1 for CH3NH3 molecular vibrations, and <300 cm^-1 for others.
Hence, should I set an intermediate fnosep=30.0 (1000 cm^-1), or should I
set a Nose-Hoover chain with frequencies close to every phonon band, e.g.,
fnosep=90.0 30.0  10.0 3.0

Thanks a lot,

Eduardo Menendez Proupin
University of Chile
www.gnm.cl/emenendez
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20181014/cade3902/attachment.html>