2. SUMER observations
The SUMER instrument (Wilhelm et al. 1995) uses an off-axis parabolic mirror to focus radiation on the interchangeable entrance slit of a spectrometer. In the spectrometer a parabolic mirror is used to collimate the light from the entrance slit. The collimated beam is then directed to a plane scan mirror which deflects it on to a spherical concave grating whose movement is coupled to the scan mirror in such a way as to ensure the maintenance of the sharp focus of the image. Rotating the scan mirror changes the angle of incidence of the light on to the diffraction grating and hence the wavelength range of the dispersed radiation. Two alternative two-dimensional detectors are positioned in the focal plane of the grating to collect stigmatic images of the entrance slit. Microchannel plates with crossed delay lines are used for the detectors. Along the dispersion axis, the central 511 of the 1024 pixels are coated with KBr with the remaining pixels bare (there are also L attenuators at the detector edge) to allow optimisation of effi ciencies and the 27.0 mm detector length allows sampling of the spectrum in approximately 40 Å segments. The effective wavelength coverage of the detector A used for these observations was from 788 Å to 1610 Å in first order and from 500 Å to 805 Å in second order. The lower limit of 500 Å is essentially determined by the sensitivity imposed by the normal incidence reflections in the instrument.
2.1. The OPAC observing sequences
The observations were carried out on 1996 September 7 using the observing sequences OPAC-1, OPAC-2, OPAC-3 and OPAC-4. In these sequences the 120 arc sec entrance slit was placed on the east limb with the slit length oriented north - south on the Sun. The spatial raster was east-west and composed of 18 exposures of 100 s duration with 1.9 arc sec steps between them. After each exposure the full 44Å by 120 arc sec image was received. The central wavelengths of the spectral intervals and the slit widths were chosen so that the multiplets to be studied were optimally exposed.
Multiplets with high count rates were observed through the narrow, 0.3 arc sec slit in the sequence OPAC-1. The key multiplets measured in this sequence were O III - (832.7 Å), N III - (991.6 Å), Ne VI - (992.6 Å), C II - (1036 Å), N II - (1085.7 Å), C III - (1175 Å) and Si III - (1298.0 Å).
The sequence OPAC-2 covered weaker multiplets with wavelengths shortward of 1350 Å. The multiplets measured in this sequence were C II - (904.48 Å), C II - (1010 Å), Si IV - (1128 Å), C II - (1335 Å) and N III - (684 Å, 2nd order).
The OPAC-3 and OPAC-4 sequences covered multiplets longward of 1350 Å. These multiplets were observed both on the bare and the KBr parts of the detector because beyond 1350 Å the sensitivity of the bare drops off very rapidly. By placing the spectrum on the bare and the KBr it is possible to separate the first and second order line contributions (c.f. Curdt et al. 1997). The key multiplets measured in these sequences were O V - (760 Å, 2nd order), N III - (763 Å, 2nd order), N III - (772 Å, 2nd order), O IV - (787 Å, 2nd order) and the O IV - (1397.2 Å, 2nd order), O III - (703 Å, 2nd order) and S VI - (706.4 Å, 2nd order).
2.2. Spectral fitting
The spectral observations were fitted with Gaussian profiles by a procedure (ADAS602) described by Brooks et al. (1999).
2.3. Multiplet selection and spectral fitting
A limited set of multiplets have been chosen from the full OPAC data for detailed examination. The principal selection objectives were firstly to classify the spectral lines emitted by carbon, nitrogen and oxygen ions on the solar disk in quiet sun conditions according to how much they are affected by opacity and secondly in the same solar observation conditions, to establish the trend of optical thickness of like lines along isoelectronic sequences. The OPAC data sets allow examination in principle of several multiplets of the C II , C III , N II , N III , O III and O IV spectra. In practice, blending and overlapping of multiplet components greatly restrict the number of multiplets amenable to detailed study. A summary is given in Table 1 of the multiplets assessed. Four of these were then studied in detail.
2.3.1. C III - (1175 Å)
The spectral interval for this multiplet is shown in Fig. 1a. The key multiplet components for opacity studies and the intensity ratios conventionally used are 2-2 and 1-2 with ratio I(2-2)/I(1-2) (where the multiplet components are identified by the J quantum numbers of the relevant levels with the first J-value that of the lower level) and 0-1 and 2-1 with ratio I(0-1)/I(2-1). The principal difficulty is the overlap of the 1-1 and 2-2 components which requires care with handling by the spectral fitting program. A series of fittings were conducted exploring the imposition of constraints on widths and centroid position to seek improvement of the fit reliability. Improvement was assessed by the reduction in the uncertainties on the counts estimated by the fitting program. The optimal fit was achieved by allowing the centroid positions of lines 2-1, 1-0, 0-1 and 1-2 to float as the program's ability to located them is reliable, but to constrain the positions of the other lines relative to lines 2-2 and 1-1, allowing them to track their shift at the limb (see Sect. 2.5 and Fig. 7b).
The intensity ratios I(2-2)/I(1-2) and I(0-1)/I(2-1) are plotted against magnitude of the distance from sun centre in Fig. 1b. The latter ratio remains close to its A-value ratio indicating insensitivity to opacity at all raster positions across the limb. The individual lines of the latter ratio are in fact thin on disk and thicker at the limb but with very similar opacities. The former ratio approaches the optically thin value towards the disk, but deviates significantly on traversing the limb as the geometrical path length along the line of sight through the emitting layer lengthens. At raster positions above the limb, the ratio appears to return to the on-disk ratio rather than the optically thin ratio and is probably evidence of scattered light within the instrument. This movement in the ratio occurs at the same raster positions as the wavelength limb shift (see Sect. 2.5). Evidently the 2-2 line is nearly optically thin on disk but becomes thicker at the limb.
2.3.2. C II - (1036 Å)
For this multiplet the components for opacity studies are 1/2-1/2 and 3/2-1/2 with the intensity ratio I(3/2-1/2)/I(1/2-1/2). There is no overlapping line problem in this case. However, there is a strong O VI line to the red of the 3/2-1/2 component which caused some difficulties in fitting the weaker C II features. As well as the O VI line there are two very small unidentified lines on either side of the C II multiplet. The spectral interval is shown in Fig. 2a. The two weak features became significant after the twelfth raster position, that is on crossing to above the limb. A consequence of this was that the autolooping option on the ADAS602 fitting procedure could not be used for the whole raster scan. The looping was then split into three sets of twelve, six and one raster position with five, three and three fitted features respectively. There was a small amount of line bending, so a line straightening program was run on all the raster positions. Various fitting options were performed in order to achieve a fit with the minimum uncertainties in the estimate of the counts for the C II and O VI lines.
The I(3/2-1/2)/I(1/2-1/2) intensity ratio is shown in Fig. 2b and shows substantial deviation from the theoretical optically thin limit on disk. The ratio is relatively constant on crossing the limb and then moves towards the optically thin ratio above the limb. Both lines do not show much limb brightening confirming that they are optically thick on disk. This contrasts with the C III case since the more substantial self absorption orthogonal to the layer here affects local emitting ion population structure. The evidence for scattered light above the limb in Fig. 2b is less clear. The error bars are large due to O VI intensity dominating the weakening C II intensities in this region.
2.3.3. C II - (904.48 Å)
The key multiplet components used for opacity studies are 3/2-1/2 and 1/2-1/2 with intensity ratio I(3/2-1/2)/I(1/2-1/2) and 3/2-3/2 and 1/2-3/2 with intensity ratio I(3/2-3/2)/I(1/2-3/2). The background intensity level is much higher for this multiplet than for the previous two and the 3/2-3/2 and 1/2-1/2 lines are strongly overlapping. However, the fitting procedure was able to resolve the components without difficulty. There are other adjacent line features, two to the red and three to the blue. Of these, one on each side blends into the C II multiplet. These features do not become significant until the fifteenth raster position just beyond the limb. The spectral interval is shown in Fig. 3a where the relatively high background is due to the Lyman continuum of hydrogen. There was a limb wavelength centroid shift (see Sect. 2.3.3) for the C II lines but its onset was higher above the limb than in the C III case. As before, error bars arise from the composite fitting error of the two component lines. The intensity ratio plots are shown in Fig. 3b. The solid lines show the corresponding A-value ratios. The graphs in this case contrast optically thick on disk and optically thin at limb ratios and are similar to ratios already illustrated in Fig. 1b and Fig. 2b. The ratio again suggests that scattered light is a major component of the off-limb observations with an onset in raster position close to that of the wavelength limb shift. The ratio's return to the on-disk values is not clear but it does show a marked deficit compared to the optically thin ratio well above the limb.
2.3.4. N III (684 Å, 2nd order)
The multiplet at 684 Å was observed in second order around 1368 Å. The spectral interval is shown in Fig. 4a. The key multiplet components used for opacity studies are 1/2-1/2 and 3/2-1/2 with the intensity ratio I(1/2-1/2)/I(3/2-1/2), and 3/2-3/2 and 1/2-3/2 with I(3/2-3/2)/I(1/2-3/2). Severe blending affects all but the 3/2-1/2 component. Attempts to separate the other components from their blended neighbours were largely unsuccessful in the first instance. However, a series of fittings were conducted to obtain the results. First, the 3/2-1/2 component was fitted separately in the interval 500 to 527 pixels in order to obtain estimates of its width. The width should be the same for each member of the multiplet. The interval 555 to 585 pixels was then fitted, fixing the widths of all lines to be the same as the 3/2-1/2 component. This allowed identification of the 1/2-3/2 component by its position (from a rough wavelength calibration using the values of Wilhelm et al. 1995) and its intensity relative to the 3/2-1/2 component. The positions of all lines were then fixed (along with the widths) and the fitting redone. Using the newly found pixel position of the 1/2-3/2 component, an improved wavelength calibration was calculated to obtain pixel positions for the 3/2-3/2 and 1/2-1/2 components. Poor separation of these components created a problem that was resolved by subtracting the background, estimated from the fit to the interval 555 to 585, from the intervals 500 to 527 and 526 to 556. Then the 3/2-1/2 component was fitted again with its position fixed and the updated values of its width used to fix the widths of all the lines within the 526 to 556 pixel interval. Comparison of the estimated positions and estimated intensities of the lines with their expected positions and expected intensities allowed their probable identification through the majority of the raster position datasets. However, when the lines were too weak off limb they could not be unambiguously separated and hence two unrealistic points are missing from Fig. 4b. The estimated positions were then used to fix the line positions and the fitting redone with the widths also fixed. This was our best attempt at the spectral fitting and produced the results shown in Fig. 4b. Due to the great attention paid to the fitting and its optimisation by analysis of estimated errors the error bars overplotted are very small. This reflects the success of the method described above in producing a `good' fit. However, the appearance of up to 9 unidentified lines in the 526 to 556 interval reduces our confidence in whether the lines have truly been identified correctly. In the figure, the I(3/2-3/2)/I(1/2-3/2) ratio is the solid line with error bars and the straight solid line is the corresponding theoretical A-value ratio. The I(1/2-1/2)/I(3/2-1/2) ratio is the dotted line with error bars and the straight dotted line is its theoretical A-value ratio. Note the discontinuity for the reason described above. The error bars arise from the composite fitting error of the two component lines. The erratic results for the multiplet ratios preclude further deductive analysis in this case.
2.4. Variability along the slit
The analysis in this paper uses spectral data summed along the slit to deduce opacities. The OPAC observations spanned quiet sun regions, but there are significant random short term local brightness variations along the slit presumably due to the network. These are also on a short spatial scale so that such variations at adjacent raster positions are uncorrelated. We have sought to assess the implications of this variability for opacity deductions using branching ratios. In optically thin plasma, the pure branching ratios are independent of such variations. In the optically thick case, line emission is observed essentially from plasma layers up to unit optical depth from the observer. The contributing geometrical depth is then self adjusting through brightness variations, however caused, and so again the ratios would be expected to be insensitive. The sensitivity of the line ratios to brightness variability is when the optical depth of the emitting layer is of order unity. To investigate this, at each raster position the brightest 60 and weakest 60 pixels were separately summed and the line ratio variations for each as a function of raster position plotted separately. Results for the C II - (1036 Å) and the C III - (1175 Å) spectral intervals are shown in Fig. 5. In Fig. 5a, on disc where the ratio is optically thick, the bright and weak pixel cases agree. On crossing the limb, there is a tendency for the bright pixel case to indicate greater optical depth but the error bars are large. In Fig. 5b, where the I(2-2)/I(1-2) ratio is nearly thin on disk, the brighter pixel group is more optically thick and this difference is enhanced on crossing the limb. The surface plots of Fig. 6 show that on disk there is a large variability in the observed fluxes for both C II and C III . Although the C II fluxes are clearly the more variable (c.f. the strong bright point on disk in Fig. 6a), there is less variation in opacity on disk for the C II data than for C III . Note that Fig. 6a refers to a different C II multiplet from Fig. 5a but the trend is the same for both.
It is to be noted also that, in the thinning atmosphere above the limb, there is a weak tendency for the bright pixel case to be apparently less optically thick. However, the above limb results do not return to the optically thin ratio and we interpret this as the actual above limb emission being dominated by scattered light from the disk. The brighter pixel case, having more intense lines, might be expected to be less dominated by scattered light. Overall the brightness ratio between the bright and weak pixel cohorts can be as much as a factor 2. The present results show that the brightness variation is primarily due to density variation. In conclusion, deduced on-disk optical depths are modestly influenced by the size of the area integrated over but the broad interpretation of our results, when summing over the whole slit, remains valid for the opacity study.
2.5. Anomalous wavelength shifts
In the spectral fitting alluded to in Sect. 2.2, aside from the usual expected data corrections such as flat fielding, line bending and line tilting etc., for which there are established correction procedures, an anomaly was found in that there was an apparent wavelength shift of uncertain origin on crossing the limb.
Using the spectral fitting program with the C III - (1175 Å) multiplet, spectral positions (pixel number) of the components of the multiplet were plotted as a function of raster number (see Fig. 7a). There was clear evidence of a shift of 1-2 pixels systematically in each component as the limb was crossed. This substantially exceeds the solar rotation shift (0.003 pixel). It creates a problem in automatic use of the spectral fitting program when it cycles on the basis of fixed positions for each component. The situation is a sensitive one with overlapped components, since fixed relative positioning is necessary for separation. Fig. 7a shows the systematic shift in the well resolved components (2-1, 1-2) and a spurious separation distortion from the fitting program at the limb ( raster index 9) on the overlapped components (2-2, 1-1). We evaluated a pixel shift vector based on the average shift of the well separated 1-2 and 2-1 components at each raster point. This vector was then used with the looping fitting program for the overlapped components. The correction is shown convincingly in Fig. 7b. The limb shift varied in magnitude (up to 2.0 pixels) and onset in raster position between different multiplets. The origin of these shifts is unclear. Differential heating of the SUMER instrument with pointing is considered unlikely. We conjecture that they may arise from mass flows giving Doppler shifts in the observational line of sight. In the detailed line ratio studies of Sect. 2.3 it was pointed out that there are indications of scattered light dominating spectral intensities above the limb. It appears also that the onset of the wavelength shift coincides with the raster position at which scattered light dominates and not with the precise position of the limb. If so this implies that the scattered light (most probably a whole disk integral effect) has a blue shift relative to our limb observations of up to 20 km .
Fig. 8 shows the shift which was observed in the C II - (904.48 Å) multiplet indicative of similar velocities as in the C III case.
No systematic shift was found in the fitted centroid positions for both the C II - (1036 Å) and N III (684 Å) multiplets, neither is there clear evidence for scattered light in the ratios though in the latter case the ratios are, as stated in Sect. 2.3.4, erratic and we are uncertain as to their interpretation. In the C II case it is possible that the strong O VI line which dominated the fit to the C II 1036 Å multiplet components beyond the limb (see Sect. 2.3.2), masks both these effects if they are present.
© European Southern Observatory (ESO) 2000
Online publication: June 5, 2000