2. Observations and data analysis
Our observations cover the period 1996 September 10-October 5, which includes the fourth passage of the active region on the disk, including its appearance around the east limb early in that period, and its disappearance around the west limb late in the period. Fig. 1 shows two images from SXT during the period, with west to the right and south downward. Fig. 1a shows the region about two days' rotation behind the east limb and Fig. 1b shows it close to disk-center passage. Both images are 5.34 s exposures with SXT's AlMg filter and have pixel resolution. Fig. 1b shows that the region is very extended, where the 50%-intensity level covers approximately At times during our observation period there also seem to be connections between the brightest parts of the region and surrounding regions, in particular those further to the south.
Yohkoh 's BCS (see Culhane et al. 1991and Lang et al. 1992for overviews) consists of four channels, covering the resonance lines and principal satellite lines of H-like iron, (FeXXVI , nominally covering the wavelength range 1.7636-1.8044 Å); He-like iron (FeXXV , 1.8298-1.8942 Å), He-like calcium (CaXIX , 3.1631-3.1912 Å), and He-like sulfur (SXV , 5.0160-5.1143 Å). Only the SXV channel covers a low enough energy range to observe emissions from non-flaring active regions routinely, and therefore our BCS work here is confined to SXV results.
Even in SXV , the flux from the late-phase active region presented here is so low that we can obtain useful spectra only after extremely long integrations. Also, since the region does not show short time-scale activity (see Fig. 3, introduced below), there is no need for high time cadence. Accordingly, spectra for the data here are for time intervals ranging from 3120 s to 29 448 s in order to obtain statistically-significant spectra.
As in our previous low-flux spectral studies (e.g., Sterling et al. 1997), we addressed the issue of the background in SXV by obtaining spectra at times when there were no active regions on the Sun. We used data accumulated for some 27 000 s on 1996 September 13 for the background, when the active region was on the far side of the Sun prior to the disk passage of this study. The resulting background spectrum is similar to that found by Sterling et al. (1997) in its magnitude, its variation with time, and its wavelength distribution. For example, a background spectrum integrated for about the same length of time from 1996 October 7, which was a time when the region was behind the Sun after the disk passage of this study, has a magnitude within 15% of that of the September 13 background spectrum over the full wavelength range of the SXV channel. Fig. 2a shows the September 13 background spectrum (multiplied by a "background multiplication factor," discussed below, of 0.7), as the lower of the two features in the figure. In the same panel, the upper feature is a spectrum integrated for 21 384 s from a time period when the active region is on the disk. Prominent spectral lines are visible in the upper spectrum, and these lines are strikingly absent in the background spectrum. Since the two spectra were accumulated for a comparable amount of time, this clearly indicates that virtually all the emission-line features in the upper spectrum originate from the weak soft X-ray flux of the active region itself.
There is a wavelength dependence of the background spectrum, which shows a broad intensity peak at wavelengths just short of 5.10 Å. This structured background causes spectra from weak sources to be deformed, and removing some factor times the average background spectrum often improves the shape of the observed spectrum compared to theoretical spectra, as discussed in Sterling et al. (1997). This "background multiplication factor," however, must be estimated, since we do not have a good understanding of the absolute level of the background at a specified time. For each of the spectra we remove 0.6-1.3 (0.8 on average) times the average background spectrum, where we use the expected near-continuum level fluxes near 5.06 Å and 5.08 Å to estimate the background multiplication factor. This estimate is made visually, but our resulting resonance line-to-continuum flux ratios are similar to those found during periods of more intense flux and during flares (Fludra et al. 1993), giving us confidence that our resulting background-subtracted continuum levels are reasonable. We will allow for variations in this estimate of the background in our determination of the uncertainties of the electron temperatures from the spectra below.
Fig. 2b shows as a histogram the observed spectrum of Fig. 2a, with the background spectrum of Fig. 2a removed. Overlaid as the solid line is a synthetic SXV spectrum formed using atomic data from Harra-Murnion et al. (1996), and data on ionization states from Arnaud & Rothenflug (1985). Be-like satellite lines are not included in the fits, but their contribution is not large for the temperatures we deduce from other lines (see Doschek et al. 1996). An isothermal, single temperature component is assumed here and in all our fits (we discuss this assumption in Sect. 3 below), and we obtain the synthetic spectrum that best fits an observed spectrum by seeking the minimum value over the approximate wavelength ranges of 5.025-5.055 Å and 5.094-5.110 Å. These ranges include lines which are most sensitive to variations in electron temperature, namely, the resonance line (the w-line in the notation of Gabriel 1972) near 5.04 Å, a group of satellite lines including the dielectronic n @ lines near 5.05 Å, and a blended complex consisting of two dielectronic lines (lines j and k) and a forbidden line (line z) near 5.10 Å. We assume Voigt profiles for the spectral lines in generating the synthetic spectra. Fits to the observed spectra yield information on the flux, emission measure, electron temperature, and excess line broadenings. In Fig. 2b, the best-fitting spectrum has an electron temperature, , of 2.7 MK. Fig. 2c shows a background-subtracted spectrum from a different time, showing characteristics of slightly cooler plasma at 2.4 MK (the ratio of the w-line to the complex at 5.10 is reduced in the lower- spectrum).
We also calculate from SXT using the filter-ratio method (Tsuneta et al. 1991). We used images from the full-frame mode of SXT obtained with the Al.1 and the AlMg filters of exposure times 2.67 and 5.34 s, respectively. We selected several image pairs per day. The cadence of these full-frame images pairs is not great enough to allow us to integrate the SXT images over the approximate time periods of the accumulated BCS spectra. SXT's partial-frame mode has much higher cadence, but the field of view of the partial frame images is often less than the spatial extent of the active region itself. We therefore opt to use the full-frame images for our analysis, where we can select a sub-image of size large enough to encompass the entire active region. We used sub-images of 810" 810", approximately centered on the region, and we integrate the flux over the entire sub-image in calculating the temperatures.
© European Southern Observatory (ESO) 1999
Online publication: June 17, 1999