[Pw_forum] forces & energy conservation in cp

[Antonio Tilocca]

dear users,

I was running some tests with the current (May, 7) cvs version of the cp
code, using the same input file, pseudos, etc as in a previous MD run with the
old (pre-qEspresso) cpv code.

Although energies and eigenvalues after the initial (fixed-ions)
electronic minimization are ok,
I have noticed some serious problems with total energy (constant
of motion econt) conservation in a subsequent NVE MD trajectory, which
were not present in the same run with the old code, and are not affected by the
choice of time step, fictitious mass, etc.

The comparison of the output files seems to point to a problem with the
force calculation: the forces on most ions match closely with the forces
obtained with the previous code, except for 2-3 ions whose z-component of
force calculated with the cvs version is up to 100 times larger than
what it should be.

I was wondering if someone else experienced similar problems?

thanks,

Antonio Tilocca