5. The million K dimming
To investigate the nature of the million K dimming, we highlight an area in the corona, defined by the box in the bottom-left frame of Fig. 6. This box covers 290 CDS pixels, an area approximately 20 arcsec by 60 arcsec which is well above the limb. This area is chosen in order to minimise the influence of any variability in low lying coronal loops. We are simply attempting to measure the degree of the dimming of the corona and estimate the mass loss or temperature change that this may signal.
For the pixel group within the defined box, we sum the line intensities for the six emission lines recorded. We apply instrumental corrections and calibrations and fit the lines and remove background for each of the 16 rasters. Figs. 8 and 9 show the intensity time plots resulting from this. For each curve, we plot the intensity (photons s-1 for the pixel group) against time from the start of the day.
Errors in the two plots are calculated in the following way. There is a Poisson noise associated with the total number of photons which interact with the CDS detector - including the background. In addition, the fluctuation in the amount of amplification in the detector system provides an effect which is comparible to the Poisson noise. Thus, for each pixel group, we sum the total number of counts (N) and estimate a sqrt(2N) error.
The Mg IX emission shows the most dramatic change, corresponding to the dimming shown earlier. The intensity in this emission line falls by a factor of 2 over the 12.5 hour period. Indeed, the intensity has fallen by a factor 1.75 within the first 6 hours, i.e. by 16:00 UT. The decay in intensity is apparent from the start of the observation which suggests that we can only put a limit on the onset of the dimming, that is, it started at, or earlier than, 10:02 UT. It is interesting to note that the emission had fallen by about 30% by the time the CME was crossing the occulting disc (at about 12:30 UT).
Surprisingy, the Si X emission line shows a remarkably constant intensity despite the temperature being only marginally above that of the Mg IX ion. Mg IX has a characteristic temperature of 1,000,000 K and Si X has a characteristic temperaure of 1,300,000 K. The error bars indicate that this is a real effect rather than the emission line being swamped by background noise.
The He I, O V and Fe XVI emission line intensity curves are shown in Fig. 9. The He I and O V intensities (upper and middle panels) remain effectively constant, when one considers the large error bars which result from the low He I and O V intensities off the limb. However, there is a marginally significant but stready rise in intensity in the last 4-5 images for both ions, from 20:00 UT, though this is some 10 hours after the CME onset.
The Fe XVI data show more significant variation, with one rather bright reading at 13:30 UT and a brightening in the 17:00-20:00 UT portion. However, these variations are marginally significant. Given the intensity and the calculated error, there is clearly no significant decrease in intensity to match that detected in the Mg IX data, i.e. no coronal dimming at the 2 million K level. Such a dimming would be clear in these data. There is no increase in Fe XVI intensity (or Si X intensity) as the Mg IX curve decreases between 10:00 and 16:00 UT, which would be a signal of heating.
The lack of intensity decreases or increases from ions other than Mg IX suggest strongly that the dimming is caused by the loss of plasma specifically at about 1 million K from the corona rather than a change of temperature in the corona. Indeed, the lack of decline of the Si X data, in particular, suggests that the temperature of the plasma which is decreasing in density is pretty well confined to 1,000,000 K.
Fig. 10 shows the ratio of the intensities in the Si X 356 and 347 Å lines for the same pixel group over the same period. They show a value of about 1.1 which is consistent with a log10 density of 9.0 [cm-3] (Mason et al., 1997). The error bars are large but the data points appear to display a modest decline. The decline of the ratio is barely significant but the best fit to the data suggests a decline from 1.2 to 1.08 in 13 hours, which, according to Mason et al. (1997) represents a fall in log10 density from 9.2 to 9.0. This is a 37% decrease in density. Clearly, this figure has to be regarded with some degree of uncertainty. The decrease is barely significant and the high value of the first data-point has a large influence on the least-square fit. However, the high-value of the first data-point may be relevant.
The decrease in intensity of the Mg IX line was of order a factor of 2. Since the intensity is proportional to the square of the density, this corresponds to an apparent density decrease of about 40%.
The intriguing point is the fact that the density decrease suggested by the best fit to the Si X density ratio is consisent with the density decrease indicated by the Mg IX intensity, yet the Si X intensity does not suggest a density decrease at 1,300,000 K. Given the size of the error bars in Fig. 10, one has to conclude that the Mg IX data gives strong evidence for the loss of 1,000,000 K but that the Si X data provides evidence for no significant loss at 1,300,000 K.
Can such a decrease in density of the 1,000,000 K plasma provide enough material for the CME? To assess this, we provide an order of magnitude estimate of the amount of missing mass. The area under consideration is 20 arcsec x 60 arcsec and we detect evidence for a 40% density decrease. Let us assume that the emission (for which the plasma is optically thin) comes from a region as deep as it is wide, i.e. 20 arcsec, and let us take our density of 109 cm-3 as the initial density. The mass loss, suggested by the dimming, is 6.5 x 109 kg for the area boxed in Fig. 6. This area clearly represents only a fraction of the area which displays dimming. A rough, but realistic, factor by which we could multiply this mass to cater for the total area of dimming would be of order 5, thus suggesting a total `lost' mass of 3.3 x 1010 kg.
As mentioned above, the calculated mass for the CME in question is (5 1) x 1010 kg. We do stress that this is a very small CME but it is interetesting to note that the estimated missing mass is almost 70% the mass of the CME. Given the error on the CME mass calculation and the missing mass calculation from the EUV dimming, we cannot rule out the suggestion that all of the CME mass is accounted for by the EUV dimming, i.e. the CME source region is the dimming region of the low corona. In addition, it must be remembered that the CME span and the span of the dimming region are identical.
Despite this, it is clear that the total mass of most CMEs will not be provided by portions of the low corona observed to be dimming in the EUV. For example, a significant fraction of the mass of many CMEs is contained in erupting prominence material. This event does not involve a prominence eruption, as is the case for, maybe, 30% of all CMEs. If a prominence eruption had been detected, it would have been seen as an ascending structure in the cooler EUV lines, such as He I or O V.
Thus, the dimming may account for much of the mass of the `outer' CME shell, perhaps for all CME events, but the prominence material is a different matter, and would be seen in cooler lines. Here, we show just one event, and the general properties of the EUV signatures of CME onsets, such as dimming, need to be established for many tens of events in order to understand the full picture.
© European Southern Observatory (ESO) 2000
Online publication: June 20, 2000