[Pw_forum] on variable cell optimization

[Paolo Giannozzi]

On Friday 28 January 2005 16:58, Serge Nakhmanson wrote:

>      Entering Dynamics;  it =    28   time =  0.00131 pico-seconds
> [...]
> Nothing is moving, nothing is changing.

in 1.31 fs very little is happening, for sure

> Or is my cell too heavy (see below)?

you used the default value for the time step (1 a.u.=0.0484fs): it is 
way too small. There shouldn't be a default value for the time step
in the first place. There is one for the cell mass, and it should be good 
for most purposes.

> I've seen people asking questions about a good value for cell mass
> but no answers have been given. 

Apparently nobody gives a damn about variable-cell relaxation, or maybe
those who give a damn don't give a hint. I don't give a damn either but I'll
try anyway to give a hint. Set a time step following the usual MD criterion:
a few percent of the inverse of the largest typical vibrational frequency
of the system. For relaxation, it is usually convenient to set all masses 
to the same value. Set the cell mass to the sum of atomic masses in the 
cell (this is more or less the default, by the way). I guess that the cell
dynamics is typically slower than the nuclear dynamics, so it might be
convenient to reduce the cell mass somewhat. Since however I never did 
such a calculation, this might be all wrong.

> BTW, what is the UMA unit ( cell mass =  36.00000 UMA )?

atomic mass units, I guess: mass of H=1

P.

-- 
Paolo Giannozzi             e-mail:  giannozz at nest.sns.it
Scuola Normale Superiore    Phone:   +39/050-509876, Fax:-563513 
Piazza dei Cavalieri 7      I-56126 Pisa, Italy