DEGAS 2 | Previous | Next

Now, with the treatment of low energy reflections in DEGAS 2 revised to reasonably approximate the extrapolation used in EIRENE, we can add the other molecular dissociation reactions. At this point, both codes have essentially the same CX, ionization, and dissociation reaction rates.

Because of the small, persistent discrepancies in the treatment of low energy reflected atoms, we'll take the goal for "good agreement" to be 5%, as long as that covers the statistical errors. To demonstrate that agreement, we'll show here plots of the fractional difference, defined like:

abs(DEGAS 2 value - EIRENE value) / (DEGAS 2 value + epsilon)

Indeed, at the resolution available in these JPEG's it would be difficult to discern any difference between the densities computed by the two codes. So, we'll present just one plot of the density along with the fractional difference.

Both the D and D2 densities agree within the 5% range near the target (again recall that the noise level in the EIRENE runs increases rapidly with distance from the target):

In these simulations, neither code is actually tracking D2+; it is dissociated as soon as it is formed. Scoring the D2+ density in this case is tricky. But, it is at least being done the same way in the two codes. The error bars are much larger here (note the relatively small densities); greater than 15% everywhere. Hence, the fractional differences are quite a bit more than with D and D2, but still consistent with the computed errors.

Now we compare the sources which will be transferred to the fluid plasma codes. In EIRENE, these data are written to the FORT.32 file. In DEGAS 2, they are contained in the output_2D_coupling array (in the complete netCDF output file) as well as in the sources.out text file. Note that the EIRENE data in this file have been normalized to the ion current to the target (presumably so the plasma code can rescale the results as needed). For these purposes, that scaling factor will be reapplied. Note also that the plots to be shown here are volume integrated. Hence, the volume of the computational zones must be taken into account when physically interpreting these results (rather, we'll focus here on the numbers themselves).

The particle sources are in roughly the same shape as the D and D2 densities: easily agreeing within the prescribed 5% level:

With numerically small values near the target, the electron energy source yields fractional differences well in excess of 5% there. However, the two codes agree to better than 5% where the energy source (sink) is significant:

There are three problems with the ion energy and momentum sources:

- Because of the tight CX coupling between the neutrals and the ions, these source terms represent the difference between two similar numbers. Hence, the intrinsic errors are considerably larger than for the other quantities presented here.
- Up to this point, there has been a problem computing the relative standard deviations for negative quantities. These plots were generated before the problem was fixed. The upshot is that where the relative standard deviation is less than one, it can be believed. In the future, we will have legitimate values everywhere, although the relative standard deviation could still be very large in places if the mean is smaller than the standard deviation.
- EIRENE approximates the scoring of these quantities: it essentially assumes that sigma-v is a constant. By using the collision estimator of these sources for CX reactions within DEGAS 2, we can easily duplicate these expressions. The result is the approximation is a very good one; the DEGAS 2 sources are virtually indistinguishable from the more exact ones. We'll come back to this point later when we can compare scoring of these sources by collision estimator with that done by track-length (the latter can't be done with the EIRENE CX data). For now, though, I'll ignore this (potential) problem.

As with the other comparisons on this page, the DEGAS 2 and EIRENE plots are visually identical. The fractional differences are not as small for the reasons just given. We include here the DEGAS 2 - computed relative standard deviations; the red areas are where the relative standard deviations are > 1; in fact, at some of these points the data list the value as infinite. Roughly speaking, the r.s.d.'s given by EIRENE are similar.

EIRENE computes only the source of parallel momentum; it's not clear if it tracks its variance. DEGAS 2 computes the full momentum source vector. Since the magnetic field used here is a constant, it's easy to compute the resulting parallel momentum source. Based on investigations performed on later runs, we know that the error bars here are somewhat larger (roughly 20% in the best regions) than for the ion energy source:

DEGAS 2 | Previous | Next