Astron. Astrophys. 363, 1145-1154 (2000)

2. Observations, data reduction and analysis

2.1. Observational data

The observations have been carried out using the SUMER Spectrometer (Solar Ultraviolet Measurements of Emitted Radiation) onboard the Solar and Heliospheric Observatory, SOHO. This instrument allows high-resolution solar observations over a broad wavelength range of approximately 500-1610 Å (extreme ultraviolet). The detectors have opaque KBr photocathode material deposited over the central half of their areas, while the two sides consist of bare microchannel plates. A detailed description of the SUMER instrument, its specifications and capabilities is provided by Wilhelm et al. (1995). Our observations were recorded using detector B and a long slit which covers of the solar disk.

Three sets of observations were used. The first set (Joint Observing Programme JOP055) was obtained between the 10th and 17th of December 1996. This JOP was run 12 times. Each time, 14 different spectral frames were recorded and read out (1024 spectral 360 spatial pixels), each exposed for 300 s with the slit position at the central meridian, where it crossed the boundary of either the northern or the southern coronal hole. As the slit location is very close to the pole, no rotation compensation was necessary. The 14 frames cover a large part of the spectrum between 730 and 1420 Å, including more than 70 identified and relatively unblended spectral lines.

The second set (JOP55_TR), taken on the 6th of September 1997, is almost identical to the first set, but consists of only two series of 14 frames each. The slit crossed the southern coronal hole and was slightly offset from the central meridian, to a position where the coronal hole extension was favourable at the time of the observations.

The third set consists of series of 12 different spectral frames each ( pixels) taken during the SOHO roll manoeuver on the 20th of March 1997. Series of frames were obtained every 30 degrees along the limb, with a larger number of exposures being made at the equator (total number of 190 frames, exposure time: 150 s). During the roll manoeuvre the slit was always oriented radially instead of being parallel to the N-S axis. This data set includes about 50 identified and relatively unblended spectral lines.

Fig. 1 schematically represents the locations of the slit on the solar disk during the different sets of observations.

Fig. 1. Position of the different observation sets on the Sun: JOP055, JOP55_TR and Roll data

2.2. Data reduction

The following corrections have been made to the data before the analysis: A flat-field correction was applied using the flat-field image taken closest to the date of the respective observation. The geometrical distortion introduced by the detector has been reduced using the procedure supplied by the SUMER consortium. This routine compensates the pin cushion distortion of the image and, at the same time, corrects to a first approximation the inclination of the spectral lines with respect to the detector columns due to the alignment error between detector and grating. To obtain the correct locations on the Sun for the different spectral lines, their position along the slit on the detector has been measured as a function of wavelength and the displacement and magnification of the spectrometer have been compensated for. The data frames had been carefully selected beforehand to avoid intense lines, which would require corrections of saturation effects of the SUMER detectors, such as dead time or gain depletion.

2.3. Spectral line selection

The spectral ranges covered by the JOP55 and the roll observations do not exactly coincide. We selected 10 spectral regions (common to all data sets) containing interesting spectral lines for further analysis. Fig. 2 shows examples of three of the retained frames (spectra averaged in the spatial direction of each frame are plotted). They include coronal spectral lines, of Ne VIII and Fe XII, transition region lines, of N IV, O V and N V, as well as chromospheric lines, of O I, C I, Ni II and Si II .

Fig. 2. Examples of analysed SUMER spectra. The upper two frames were recorded mainly on the potassium bromide (KBr) photocathode (within the two vertical lines) while the lowest frame shows a spectrum partially recorded on the bare microchannel plate and partially on KBr (separated by the vertical line). The spectra are averaged over the slit length. The analysed lines are identified.

A list of all the spectral lines retained for our analysis is given in Table 1. The formation temperatures of the corresponding ions have been taken from Arnaud & Rothenflug (1985). The formation temperature of lines of neutral species need to be treated with caution. More details can be found in Wilhelm & Inhester (2000, private communication).

Table 1. Analysed spectral lines with the formation temperature of the corresponding ion and the number of Gaussian components used for the fit.

Although other authors (Chae et al. 1998; Peter & Judge 1999) have pointed out problems with the S VI line at 933.39 Å (anomalous center-to-limb variation), the results we obtained with this line are consistent with those obtained from other spectral lines. The parameter values derived for this line have therefore been retained when discussing the results.

2.4. Data analysis

The lines were identified using the line lists of Curdt et al. (1997) and Sandlin et al. (1986). For this analysis we selected lines that cover a wide range of formation temperatures. In addition, to facilitate data interpretation, lines with known strong blends have been avoided whenever possible. Line parameters are determined from fits of a Gaussian plus linear background to the spectrum at each spatial position. In cases in which a dominant line has one or more minor blends a routine fitting multiple Gaussians plus background has been employed to separate out the contributions of the blends. The number of Gaussian components used for the fit to each spectral line is listed in the last column of Table 1. Some lines, like C III 1175.71 Å have well separated components which allows us to apply multi-Gaussian fitting with success. Other lines such as Si II at 1309.28 Å have blends that are so well hidden and are so strongly variable relative to the line of interest that multi-Gaussian fits gave inconsistent results. For this reason, we do not consider both Si II lines in our sample for further analysis. A particularly difficult, but important case is that of the Fe XII line at 1242.01 Å. In coronal holes this line has blends that are almost as strong as the line itself. Consequently, it was quite difficult to fit, but a 4-Gaussian gave reasonable results, although with lower accuracy than the other lines. Due to its importance it has nevertheless been retained for the analysis.

The intensity, shift and width parameters of the selected lines were obtained at each spatial position from the best-fit Gaussians. This method turned out to be more precise for this kind of data than the moments method described by Doyle et al. (1997), which we used in a previous analysis (Stucki et al. 1999). For the analysis, we used the integrated intensity over the Gaussian fit to the spectral line profile.

The reference value for the shift has been chosen for each line separately as the mean value of the wavelength of the peak of the line profile averaged over the length of the slit. Thus the choice of reference for the shifts is not related to an absolute wavelength scale and has no influence whatsoever on the results. Since the shift values are unaffected by any instrumental drift, we can derive relative shifts to a precision of about 1 kilometer per second depending upon the quality of the Gaussian fit. Although the slit often crossed the limb, only on-disk data are analysed.

The position along the slit of the boundary between the coronal hole and the quiet Sun has been deduced from whole-Sun images of the Fe XII 195 Å line of EIT (Extreme-ultraviolet Imaging Telescope, Delaboudinière et al. 1995) taken on the corresponding days.

An example of the variation of the line parameters along the slit is shown for the N IV (765.15 Å) line in Fig. 3. In the illustrated case we notice the presence of limb-brightening. Also, the line is on average blue-shifted inside the hole (relative to the average along the slit outside the hole) and somewhat broader. From this data set alone it is impossible to distinguish how much (1) the presence of the coronal hole, (2) the center-to-limb variation, and (3) intrinsic variability and spatial structure of the solar radiation contribute to the differences between the parameters in the hole and outside it.

Fig. 3. An example of each of the fit parameters (intensity, shift and width) using the spectral profiles of N IV at 765.15 Å, obtained from the first set of observations (JOP055) as a function of distance from the limb. The coronal hole boundary is located approximatively from the limb. It is marked by the vertical line near the center of the frame. The vertical solid line on the left, representing the position of the limb, is located just outside the limits of the image on the detector (dashed vertical lines) in this dataset. The scale at the top of each frame indicates , where is the heliocentric angle. Negative shifts signify blue shifts.

In order to distinguish between these three sources we need, on the one hand, better statistics (i.e., more frames containing the same spectral lines), and, on the other hand, also quiet-Sun data covering the same heliocentric angle as the coronal hole data (we define , where is the heliocentric angle). The roll data partly fulfill these requirements. By comparing the line parameters averaged over all profiles arising in the coronal hole with averaged line parameters from the quiet Sun at the same µ values it is in principle possible to distinguish between the center-to-limb variation and the hole-non-hole difference.

It is, however, still possible that we have not averaged over a sufficient number of data sets to reduce intrinsic solar variability to an acceptable degree. An idea of this residual intrinsic spatial variation of the line parameters is given by Fig. 4 and Fig. 5. They show the variations of line parameters along the slit for the N IV and Ne VIII lines, averaged over all quiet-Sun spectra as well as over all central meridian spectra of the roll data. The center-to-limb variation can be clearly seen in the intensity. Due to the relatively few spectra (three for the meridian and nine for the quiet Sun) the remaining variation of the intensity after the averaging is still large, in particular for N IV . By considering all available data (JOP055 and roll data) this variation can be further reduced. However, then the problem arises that the instrumental parameters underlying the data obtained along the meridian (containing the coronal hole) and those at other locations at the limb (quiet Sun; roll data) are not exactly the same. In order to counter this we also compare the line parameters obtained along the meridian, but outside the coronal hole, with those from the same µ values at other locations along the limb.

Fig. 4. Line parameters (intensity, shift and width) of the spectral profiles of N IV at 765.15 Å obtained in the third set of observations (roll data), averaged over all available data (nine images) in the quiet Sun (right frame) and averaged over all central meridian data (left frame, three images).

Fig. 5. Same as Fig. 4 for Ne VIII at 770.43 Å

In total we had about 25 exposures of each spectral line at our disposal. In each exposure, we averaged the fit parameters in a sector corresponding to the position of the coronal hole ("small µ sector"), which means from to , and in a sector corresponding to the quiet Sun ("large µ sector"), from to . Those values were then averaged over all exposures taken at the meridian (containing northern or southern coronal hole regions), and over all exposures taken at other locations (roll data taken around the disk). Note that the actual coronal hole boundary in each exposure along the meridian (obtained from EIT Fe XII images) is used to distinguish between the small and large µ sectors, so that the µ values given above are only averages. We have tested whether inaccuracies in determining the coronal hole boundary may influence our results by introducing a "no-man's-land" of about width between the small and large µ sectors. The line parameters of these pixels are not counted to either sector. We found that the results do not depend in any significant way on the presence or absence of such a "no-man's-land". The line parameters found are then related to the temperatures of maximum abundance of the ions taken from Arnaud & Rothenflug (1985).

© European Southern Observatory (ESO) 2000

Online publication: December 5, 2000
helpdesk@link.springer.de