6 Comments and conclusions

In this paper we have presented the first results obtained with the spectropolarimetric mode of THEMIS, based on observations made in August, 1998. The main interest of these results is that they provide a practical description of the state of the instrument. We describe the particular features of the raw data and the methods we have applied to reduce it. A particular attention has been given to the geometrical corrections. The reason for this can be described as one of the most important laws in polarimetry: the two images we compare to extract the polarization signal must be completely identical except for this polarization signal. Any other difference between them will contaminate the result. In particular, each pixel in one image must correspond to the same wavelength and spatial point in the sun as the comparison pixel in the other image. We describe the two methods which have been used for that purpose, with mainly identical results and precisions (up to several hundredths of pixel when the full profile was considered).

We have also detected some systematic errors in the instrumental setup. The most severe one was the difference in equivalent width of all the spectral lines when travelling through one or the other orthogonally polarized path. This and other errors are expected to disappear in the following campaigns, as the instrument becomes more and more tested. Apart from systematic errors, and independently of them, we have measured the noise in the data. Three different procedures have been used to estimate closely related noises. By means of statistics we have measured the photon noise and the polarimetric sensitivity on the continuum. For this last point, special attention was put into ensuring that the compared images were as identical as possible and that any residual signal could only come from noise. The FeI line at 5576 Å, insensitive to magnetic field, was very useful to this aim. We have also estimated the polarimetric sensitivity in the line while scanning active regions. The presence of different polarization signals, in amplitude, but also in form, makes it difficult to apply plain statistics. We relied on a more powerful technique, Principal Component Analysis (PCA), to detect the noise and separate it from the signal. A brief practical description of this technique and its present use has been given; however the interested reader should refer to the appropriate literature for a thorough understanding (Press et al. 1988; Golub & Van Loan 1993; Rees et al. 1999).

The three noise measurements are summarized in Table 3. They are coherent with each other, the highest one being that of the polarimetric sensitivity in the line, where less light is detected and for which a lot of manipulations have been done during the data reduction. In all the cases the noise is kept in a few units of 10^-3 of the intensity of the continuum.

Acknowledgements

V. Bommier participated in these first spectropolarimetric observations at THEMIS and her carefully kept notebook has been of great help during the data reduction. She has also shared with the authors her experience with her own data taken almost simultaneously. We are indebted to her. Special thanks for their help during the observations are also due to THEMIS telescope operators.