Today when I am relaxing the atoms I found the size of
lattice shrank grammatically.
Shrank compared to what?
Be ready that when you change the Energy Cutoff the equilibrium lattice
constant would change.
Meanwhile, the forces on Ti atoms are not balanced at all.
Forces in bulk unit cell should be strictly zero by symmetry. So if you
don't achieve this, it means that your grid is far from being converged
(even if k-grid is fine).
I would agree that 6x6x6 for a unit cell with 2 atoms might be too
small. Try using k-grid cutoff instead, and use at least 25A cutoff for
metals.
with a lot negative modes.
Vibrational properties can't be reproduced if forces are even a little
bit spoiled (i.e. with bad energycutoff)
> 2) Instead of MP smearing, try using the good old Fermi-Dirac.
This is true. There is some problem with MP in SIESTA.
> 3) What happens to your calculation if you just set your orbital radii
> to zero and use an energy shift of, say, 150 meV - in other words, if
> you let the energy shift decide what the orbital radii will be? Also,
> I don't get why you have a charge of 1.91 in your basis set when in
> metallic Ti one would not expect that at all.
This may give you not very well optimized basis but it should not create
negative modes in phonons.
