The ABINIT Users Mailing List ( CLOSED )

[Damien Caliste <damien.caliste@cea.fr>]

Hello,

   I'm looking at the code of molecular dynamics as implemented inside
ABINIT and I've two (unrelated) questions:
- in the velocity-verlet (ionmov=6) simple approach, the forces that
are used are preconditioned forces (coming from product between the
hessian and the reduced gradients, where the hessian itself is
approximated by the metric in reciprocal space). It seems to me that
using this preconditioning breaks the time evolution and the constancy
of (Ekin + Epot), doesn't it ?
- in the Nose-Hoover algorithm (ionmov=8), why the coordinates are
updated using an acceleration calculated from forces of previous
iteration (finose are computed from fcart_mold) ? I would have written:
x(t+dt) = x(t) + v(t)dt+0.5dt*dt*a(t) and not a(t-dt). What's wrong
with my guess ? Is it from the equations of motion of the Nose-Hoover
formalism ? But even there, a(t) = f_m(t) - dXi/dt.v(t) and not f_m(t -
dt) ?

   Besides, is there a documentation explaining how the Newton-Raphson
scheme is applied to Xi later in the routine ?

   Thanks for your help.

Damien.