Observations and data reduction

Spectroscopic observations used in this study were obtained with the new, cross-dispersed spectrograph ELODIE at the 1.93-m telescope at Observatoire de Haute-Pro- vence (Baranne et al. 1996). The detector used is a thinned Tk1024 CCD, with 1024x1024 elements of size 24 $\mu$m².

The observations were obtained during three consecutive nights from the 3rd to 5th of August 1994 (see Tables 1 and 2 in Chadid & Gillet 1996, Paper II). The spectral domain ranges from 3900Å to 6800Å with a resolving power $R\simeq42,000$, a signal-to-noise ratio S/N around 50 for an exposure time between 5-10mn giving a time resolution around 1% of the pulsation period (13h 36mn). The description of the data reduction can be found in Paper II. Yet, the phasing of the observations in Paper II are only approximative because we did not have simultaneous photometry. After, we corrected this phasing (Chadid & Gillet 1997). The blazhko phase of these observations is $\psi=24.98$ while the pulsation phase is approximately similar with those of Paper II (shift of 0.04).

Heliocentric radial velocities were determined from a gaussian fit on the whole profiles, except for H$\alpha$, because due to well-marked Stark wings and blends, only the H$\alpha$-core was considered. Moreover, bacause during the hump phase, hydrogen and some metallic lines show a doubling line structure, consequence of the shock wave propagation (see Chadid & Gillet 1996a, 1997). The bottom of the line forming region is rising (blue component) due to the strong outward wave, while its top is still falling (red component) according to the ballistic motion. Hence, since we are interested in the wave propagation along the radial axis, we only constructed the hydrogen and some metallic velocity curves associated with the blue components. This latter was measured using the Munich Image Data Analysis System (MIDAS).