Dear All,

I am trying to run molecular dynamics simulations (Born-Oppenheimer and 
Car-Parrinello) of titanium at various temperatures using position and velocity 
input data from LAMMPS. I have successfully run such simulations with aluminum, 
but titanium has proved troublesome thus far. It seems that no matter what I 
do, the first conjugate gradient step experiences a missed minimum, converges, 
and then diverges on the second step (with ortho going bananas).

So far, I have tried reducing the timestep (to 1-5 au), increasing the number 
of max iterations for ortho (50-100), Gram-Schmidt orthogonalization, reducing 
the energy convergence threshold (10^-6), and two different pseudopotentials 
(pz-hgh and pz-sp-hgh), and different configurations from classical code. None 
of these variations have changed the result. Each time, there is a missed 
minimum (case 3) in CG before the first step converges followed by a second 
step that diverges.I even tried using the converged first step as a starting 
point for Car-Parrinello MD, but that still fails to converge due to ortho 
going bananas.

Can anyone help me understand what is going wrong and how I might fix this 
problem?

Thanks,
Nicholas Walker
Louisiana State University
_______________________________________________
Pw_forum mailing list
[email protected]
http://pwscf.org/mailman/listinfo/pw_forum

Reply via email to