[CPMD-list] NOSE ELECTRONS

[Dave Sherman]

Dear group:

For Car-Parrinello MD simulations with a NOSE thermostat, we need to 
set the target fictitious kinetic energy of the electrons.  From 
examples I've seen, this is always some very small number (e.g., 0.001 
a.u.).  However, when I let my simulations run without
any thermostat, the EKINC fluctuates around a much larger value of, 
say, 35 a.u. Is this simply because I'm using ultrasoft 
pseudopotentials?   If I set my target fictitious kinetic energy to be 
what it naturally fluctuates around without a thermostat (e.g., 35 a.u. 
for 29H2O, 1Cu and 1 Cl atoms), it seems to work OK and I can get the 
simulation to chill-out at around 300 K +/- 20 K after 1500 timesteps.  
If I try a target of 0.01 a.u. with NOSE ELECTRONS it crashes right 
away.

My question then, is "Are my large fictitious kinetic energies simply a 
consequence of using ultrasoft pseudopotentials?".  If a system  really 
leaves the BO surface and all hell breaks loose, what is the symptom?  
If the total energy (EHAM) is stable, is everything OK?

Thanks for your help!

Dave Sherman
__________________
Dr. David M. Sherman
Reader in Geochemistry
Department of EarthSciences
University of Bristol
Bristol BS8 1RJ
TEL: 44-(0)117-954-5446
FAX:44-(0)117-925-3385
http://mineral.gly.bris.ac.uk