19990731: ``long equilibrated'' version of 19990614

GOAL

``Long equilibrated'' version of 19990614, to see if the starting configuration is really important or rather the equilibration time rules.

Results

The starting configuration of the 19990614 calculation was quite different from the homogeneous 19990727 19990729 19990730 ones (it was not created with the dlsub utility: it contained oxygen vacancies in nearest neighbor position and so on; see the RDF plots below).

Now we run the same 19990614 configuration, but with the long equilibration period (50000 time steps): to see the relative importance of the starting configuration and the equilibration time.

It appears that the ``nature'' of the starting configuration really makes a difference: notwithstanding the long equilibration period for this calc, the oxygen MSD plot is still different from those obtained for 19990727 and 19990729. Actually, the long equilibrated MSD plot has a slope which is even smaller than that of the short equilibrated one.

This is the current situation regarding the influence of the starting configuration upon the oxygen MSD plot (and hence the oxygen diffusion coefficient). In the following plot, in all the calculations a long equilibration time was set (50000 time steps). 19990727 and 19990729 have the same starting configuration (i.e. no oxygen vacancies and/or cerium three impurities in nearest neighbor position), 19990731 has oxygen vacancies and/or cerium three impurities in nearest neighbor position (the original calculation (19990614) was set up without any control on the creation of the oxygen vacancies and the cerium three impurities).

The following is the RDF plot for the relevant defect pairs in the starting configuration, taken from 19990712 and replotted here for convenience.

The correspondence between short and long equilibrated caclulations is as follows:

short	long
19990614	19990731
19990702	19990729
19990707	19990727

So, up to now, it looks like two factors are important with regard to the reproducibility of the oxygen diffusion coefficient: the equilibration time and the starting configuration.

The equilibration time may change dramatically the MSD plot obtained from exactly the same starting configuration (the clear example of this effect is the comparison of 19990727 with 19990709).

On the other hand, even with a long equilibration time, the starting configuration, described in terms of RDF for the relevant defect pairs, basically $(V_{O}^{\cdot\cdot}-V_{O}^{\cdot\cdot})$ $(V_{O}^{\cdot\cdot}-Ce_{Zr}^\prime)$ and $(Ce_{Zr}^\prime-Ce_{Zr}^\prime)$, may have a substantial effect. The clear example in this regard is the series 19990727 19990729 and 19990731. As can be seen from the RDF plots above, in the first two calculations there are no tight pairs while in 19990731 nearest neighbor associates of all three kinds of defects are present. This is reflected in the MSD plots: 19990727 and 19990729 give nearly the same MSD plot for oxygen, while that of 19990731 has a clearly different slope.