4 H$\alpha$ line variations

4.1 Residual H$\alpha$-emission

Residual H$\alpha$ spectra of UX For and AG Dor are generated by the subtraction of a synthesized inactive binary-star spectra. The resulting spectra can be considered as the chromospheric spectra of the object stars, and are shown as phase images in the top row of Fig. 5. The emission from both stellar components in both binary systems is seen as bright sinusoidal stripes (the horizontal grey zones mark holes in the phase coverage). The phase image of AG Dor also shows weak dark sinusoidal lines at the wavelengths of nearby absorption lines. This is due to an incomplete reproduction of the depths of some photospheric lines and could be due to a relative abundance mismatch between the object and the reference spectra.

Further detail is investigated in two ways, firstly, by measuring equivalent widths (EW) of the residual H$\alpha$-emission lines and, secondly, by subtracting the one residual spectrum with the least emission-EW(we call these the normalized residual spectra). The remaining emission, if any, must be due to non-stationary features in the stellar chromosphere, like plages, flares, bright points and others.

Figure 5: Spectral phase images from the residual spectra of a) UX For and b) AG Dor. The H$\alpha$ emission from both stellar components is seen as two sinusoidal stripes. For both binary systems the emission of the primary is stronger than that of the secondary. The normalized residual phase images are shown in c) for UX For and in d) for AG Dor, after subtracting the one residual spectrum with the smallest emission equivalent width. This should leave the non-stationary part of the chromospheric emission as the remainder

4.2 UX For

Double-Gaussian fits to the residual spectra with IRAF's splot routine yield average EW values of $588\pm73$ mÅ for the primary and $407\pm42$ mÅ for the secondary. The range of individual values is between 498-784 and 250-477 mÅ for the primary and secondary, respectively. The scatter of a single measurement is $\sim$20%. Due to severe blending of the residual emission profiles, the EW values around phase 0.25 and 0.75 are not reliable and were not included in the calculation of above values.

To build the normalized residual spectra, we subtract the residual spectra with the spectrum from phase 0 $.\!\!^{\scriptscriptstyle\rm p}$6101. In this spectrum the two H$\alpha$ emission peaks are the weakest within our data set, thus presumably representing the least-active phase, and they are well separated. Figure 5c plots the normalized spectra versus phase. Variations are seen in the stronger-emission component, e.g. between phase $\approx$0 $.\!\!^{\scriptscriptstyle\rm p}$4 and 0 $.\!\!^{\scriptscriptstyle\rm p}$6. These variations may be due to additional emission from plages or unresolved flares. There is also some evidence for additional emission between the two H$\alpha$ components (the zero-velocity center) around phase 0 $.\!\!^{\scriptscriptstyle\rm p}$5. This emission might be enhanced by errors in setting the continuum, as it is poorly defined between the primary and secondary H$\alpha$-feature. However, tests with various continuum settings showed that it could not account for the total residual emission around 0 $.\!\!^{\scriptscriptstyle\rm p}$5. At that phase the intensity of the residual emission exceeds the signal-to-noise level at least by a factor of five. We suspect that it is caused by neutral hydrogen left over from coronal-mass ejections and "stored'' near the inner Lagrangian point. A similar scenario was suggested for other chromospherically active binaries, e.g. for V471 Tau (Young et al. 1991).

The presence of circumstellar material or mass flow due to prominences or flares may dilute (or even supress) the rotationally modulated signature. For both components of UX For, no clear phase dependency of the H$\alpha$-variations could be found, but due to the limited phase coverage and signal-to-noise ratio of our spectra we can not fully exclude that H$\alpha$ is rotationally modulated. On the Sun, we see that eruptive prominences change their shape on timescales of hours or even minutes, with radial velocities of several hundred kms^-1 (e.g. Tandberg-Hanssen 1995). If the H$\alpha$-variations observed on UX For are due to eruptive prominences or flares, then it is likely that the relatively long exposure times of 20 min or more cannot resolve this activity.

4.3 AG Dor

AG Dor's residual H$\alpha$ emission shows only small or no variations. EW values were measured by fitting a double Gaussian to the residual emission lines. These EW values are only reliable when the separation of the emission lines is large enough so that no severe blending occurs. The average EW for AG Dor is $428\pm27$ mÅ and $168\pm32$ mÅ for the primary and secondary, respectively, and the range of individual values is 387-511 and 121-241 mÅ for the primary and secondary. The error per single measurement remains within 20%. Spectral phases at $0.25\pm0.07$ and $0.75\pm0.07$ were excluded.

The spectra, after subtraction of the residual spectrum at phase 0 $.\!\!^{\scriptscriptstyle\rm p}$0178, are shown in Fig. 5d. As in the case of the EW method, we found no rotational modulation of the H$\alpha$ emission. The only noticeable variation is a transient absorption feature at phase 0 $.\!\!^{\scriptscriptstyle\rm p}$2-0 $.\!\!^{\scriptscriptstyle\rm p}$3 (observed in the same night). It passes through the combined H$\alpha$-line from shorter to longer wavelengths, i.e. along the orbit of the secondary, but at a steeper angle, i.e. a higher velocity. We suspect that it might be caused by a circum-stellar cloud around the secondary, e.g. a corotating prominence, but because it occured only once and exactly at a time of inferior conjunction (where the degree of blending is most severe), we can not claim that it is a real detection.

Figure 6: The normalized goodness of fit for AG Dor (from Doppler imaging) as a function of a) the stellar inclination and b) the projected rotational velocity. The minima indicate the chosen "best-fit'' values