Observations and Data Reduction

**Figure:** A sample plot of the Si VIII spectral region in the coronal hole. Line fits are given by the dotted line.
$\scalebox{0.5}{\includegraphics [2.5cm,13cm][19.5cm,25cm]{/home/jgd/tex/papers/solar/FeVIII_fig1.ps}}$

Briefly, the data was acquired above a north polar coronal hole (NPCH) on 4 November 1996 and 10 December 1996. Sequences in the NPCH comprised of a temporal series of spectra taken at the same pointing but at successive times. The images were taken with the 1 x 300 and 4 x 300 arc sec slits. Using the individual spectra a summed spectrum was obtained. Analysis at different positions along the slit will give us information at different heights above the coronal hole. Details on the reduction procedures can be found in BTDW, where they have studied up to 1.25 $R_{\odot}$ off the limb. In the present paper we extend the analysis further off the limb up to 1.38 $R_{\odot}$ .

For each line a Gaussian fit (Fig. 1) was applied using the Genetic Algorithm (GA) of Charbonneau (1995). A single Gaussian fit was applied for the analysis of each line. A complete examination of the reliability of GA with respect to other algorithms was performed by McIntosh et al. (1998). An estimation of the errors in the derived parameters was obtained using the GA-derived parameters as an input for a ``classical'' IDL-CURVEFIT procedure (Peter, 1999; Peter & Judge, 1999).

Using the new algorithms we re-analyze the dataset already presented by BTDW. For the NPCH, we add 3 more points at 285, 325 and 370 arcsec above the solar limb with respect to the data presented in BTDW. We find that our previous estimate of errors around 0.5 km s^-1 for the non-thermal velocity obtained with the 1 x 300 slit and of 1.0-1.5 km s^-1 for those obtained with the 4 x 300 slit, are correct. Slightly larger errors are present towards the end of each slit (see $\S$ 3 and Fig. 4. later). For the new 3 points, errors of 4, 5 & 7 km s^-1 were derived. For the electron density, the error arises not only from the measured ratio of the two lines but also from the errors in the atomic CHIANTI (Dere et al. 1997) database that we have used. The expected error coming from the database is estimated to be from 12% to 15% (see Laming et al., 1997). The propagation of this error in the ratio through the CHIANTI database was simply performed introducing the value of the ratio and the value of the ratio plus/minus the error. The resulting relative error was summed quadratically with the error of 15% estimated before. The final error was practically dominated by the errors in the database up to 200 arcsec (substantially the entire dataset presented in BTDW). Larger errors are again present for the new 3 points (see Fig. 2, inset).

Inspection of the spectral region around the Si VIII lines revealed a strong feature at 1442.47Å, (identified as a second order line at 721.23Å from a comparison of off-limb data taken on and off the KBr coated part of the detector; Feldman et al. 1997). The intensity variation of this line compared to Si VIII 1445.75Å and 1440.49Å shows a similar fall-off implying a similar temperature of formation. No identification was given by Feldman et al. (1997) of this line, although they did suggest it to be a hot coronal feature, perhaps due to Fe VIII. We are unable to verify its identification due to incomplete energy levels for Fe VIII.