19990729: same as 19990702, but with longer equilibration period

GOAL

The ``long equilibrated'' 19990727 result compares very well with the ``short equilibrated'' 19990702 one (see 19990727): re-run the 19990702 calculation on t3e with the same long (50000 steps) equilibration time to see if the two ``long equilibrated'' results still agree.

Results

Differently from what happened for 19990709 vs 19990727, in the present case the long equilibration time does not produce appreciable changes.

However, this is good news for us, since we may now compare 19990727 and 19990729. In fact, these two calcs are now fully comparable: they have been performed on the same machine (t3e) and all other DL_POLY parameters are the same. The two runs differ only in the starting configuration: however, the two different configurations are qualitatively similar, in the sense that both were created with the dlsub code so as to avoid any $(Ce_{Zr}^\prime{}Ce_{Zr}^\prime)^{\prime\prime}$, $(Ce_{Zr}^\prime{}V_O^{\cdot\cdot})^\cdot$ or $(V_O^{\cdot\cdot}V_O^{\cdot\cdot})^{\cdot\cdot\cdot\cdot}$ tight pair (i.e. in nearest neighbor position).

At this point, we can compare the two (19990702 19990707) and (19990727 19990729) pairs of calculations, hereinafter called ``short'' and ``long'' equilibrated, respectively.

The two calcs of each pair are fully comparable, i.e. the input was identical, apart from the starting configuration, which however was ``standardized'' in the dlsub code sense.

The difference between the short and the long equilibrated pairs of calculations is just the equilibration period, which was $5000$ time steps for the short pair and ten times longer ($50000$ time steps) for the long pair.

Another difference is that the short pair was run on utsj90 while the long one was run on t3e, but I'm now becoming confident that there should not be problems in comparing calculations run on these two machines; see 19990709 and 19990728

From the obtained results it looks like the equilibration period makes a BIG difference:

In the light of these results I would say the following:

• my previous conclusions about the (un-)correlation between the oxygen MSD plots and the starting configuration (see 19990712) have to be reconsidered. In particular:
  • run some other calculations fully comparable with 19990727 and 19990729 (i.e. $50000$ time steps equilibration time) but with a different standardized initial configuration, to see if the oxygen MSD plot remains reproducible (19990730 is one such calc, currently on the way...)
  • re-run 19990614 and/or 19990617 with the long equilibration time and see what happens: the oxygen MSD may come out different or not; in the first case, we can say that the starting config has importance, in the second we can say that the equilibration time is important.
• tune some other parameter to see if it is possible to get better (i.e. more linear) MSD plots