Dear Abdullah Bin Faheem,
there is nothing wrong. What you are observing is not due to incomplete equilibration, but
rather a consequence, of incomplete SCF convergence. Beside time-step errors, which in
your case may be non-negligible due to a discretized time-step of 1.0 fs, the so-called
„conserved quantity“ is in a numerical calculation only conserved at perfect SCF convergence.
Hence, reducing the integration time-step and more importantly tightening the EPS_SCF
should lead to a better conservation of the conserved quantity. However, in a numerical
calculation the latter will never be perfectly conserved and it is up to you to decided what
level gives you the best efficiency / accuracy compromise.
I personally, would suggest 0.5-0.75fs for the time-step and 1e-7 - 1e-6 for EPS_SCF when
using OT. I have commented on the differences between TD and OT wrt to EPS_SCF in an
earlier posting.
Cheers,
Thomas