1 Introduction

The generally accepted explanation for photospheric variations of chromospherically active binary stars is that their surfaces are covered with cool starspots. There are only very few stars that are chromospherically active and have no photospheric rotational variability due to starspots (e.g. the Hertzsprung-gap giants 31 Comae and $\psi ^3$ Psc; Bopp et al. 1988; Strassmeier et al. 1994). The vast majority, however, shows chromospheric and photospheric activity (along with coronal activity if measured) that suggest an organised magnetic field as the mutual cause because in-phase rotational modulation is usually detected at all three atmospheric levels. Just to name a few examples, Rodonò et al. (1987) presented evidence of a close spatial correlation between spots and plage-like features on the RS CVn binary II Peg, as on the Sun, while Catalano et al. (1996) found a systematic lag of 30$^\circ$- 50$^\circ$ between plages and spots on a set of five very active RS CVn binaries. Doyle et al. (1989) found evidence for a rotational modulation in chromospheric as well as transition-region lines. Our disk resolved Doppler-imaging data of RS CVn binaries confirm the spatial relation of spots with Ca II H and K emission regions as well as with H$\alpha$ emission and absorption (e.g. for the K0III-IV binary HU Vir; Strassmeier 1994; Hatzes 1998). Furthermore, Collier Cameron & Robinson (1989; and subsequent papers) observed transient absorption features in H$\alpha$ due to prominences corotating with the star.

A large piece of information about the solar chromosphere comes from observations of the Balmer H$\alpha$ line. Analogous to the Sun, we expect that stellar plages are seen in the H$\alpha$ line as emission components while prominences show up as transient absorption features. H$\alpha$ contains information about the photosphere, the chromosphere, and about activity tracers such as plages, prominences and macroscopic velocity fields (see e.g. Oliveira & Foing 1999). H$\alpha$ is easy to observe thanks to efficient detectors and spectrographs in the red part of the spectrum. Among the class of chromospherically active binaries, only the most active systems show H$\alpha$ as a pure emission line. Commonly, it is seen in absorption partly filled in by chromospheric emission. The amount of filling-in often varies and is used to detect rotational modulation from the stellar chromosphere, preferably for red dwarf stars that become relatively faint in the Ca II H and K region (e.g. Giampapa et al. 1989).

For the present study, we chose two binary systems with rapidly-rotating G-K-dwarf components. Both systems, UX For (G6+K0) and AG Dor (K0+K4), show H$\alpha$ from the primary star as an absorption line filled in by chromospheric emission while the H$\alpha$-line of the secondary star is in pure emission. The first system, UX For = HD17084, shows strong Ca II H and K emission lines, moderate Li abundance and strong 6-cm radio emission (Lloyd-Evans & Koen 1987; Henry et al. 1996; Randich et al. 1993; Strassmeier et al. 1993). It was also detected by the ROSAT and EUVE satellites (Pye et al. 1995; Bowyer et al. 1996). Bidelman & MacConnell (1973) determined a spectral type of G5IV while Houk (1982) listed G5-8V + (G). Lloyd-Evans & Koen (1987) found a photometric period of 0.957 days. The light-curve amplitude as well as the mean brightness are variable. No colour variation was detected. Based on the combined colours, Cutispoto (1998) derived a spectral classification of G6V + K3V.

The second system, AG Dor = HD26354, shows only weak Ca II H and K emission lines (Bidelman & MacConnell 1973; Houk & Cowley 1975) and a low lithium abundance of less than solar (Pallavicini et al. 1992). However, it was detected by the EUVE satellite (Bowyer et al. 1996). Lloyd-Evans & Koen (1987) reported light modulations with a period of 2.533 days and an amplitude of 0 $.\!\!^{\rm m}$05 in V, in agreement with the orbital period. Cutispoto (1998) found a V-amplitude of about 0 $.\!\!^{\rm m}$04 in February 1992, the smallest so far observed, and only very weak colour variations. At this time, AG Dor was brighter than in any previous epoch, confirming the presence of long-term variability in the global spottedness (Cutispoto 1998). Amado et al. (1999) determined spot temperatures by making use of the light-curve spot-modelling technique and the molecular TiO bandhead strength. They found a photospheric temperature of 5000K and spot temperatures of 4000-4600K. Houk (1978) determined the spectral type as K1Vp. A refined broad-band colour-disentangling technique along with accurately measured combined colours led Cutispoto to a classification of K1-2V + K6V (Cutispoto 1992) and K1V + K5V (Cutispoto 1998), in good agreement with the spectroscopically determined classification from Houk (1978). However, Balona (1987) found no evidence for a K5-6V secondary component in his radial-velocity spectra. Such a secondary would be expected to be $\Delta V\approx$1 $.\!\!^{\rm m}$5 fainter than a K1-2 primary and should be visible in AG Dor's spectra.

Our study aims to detect a correlation between activity in the H$\alpha$-chromosphere and activity in the stellar photosphere. First, we determine the absolute parameters and present new double-lined orbits for both systems. Then, we disentangle the two spectra for each of the binary systems and extract H$\alpha$ equivalent widths and photospheric absorption line profiles. While the former are used to construct dynamic chromospheric spectra, the latter are used - at least for AG Dor - to obtain its first photospheric Doppler image. For UX For, S/N turned out to be too low for Doppler imaging.