2 IUE observations of high-mass X-ray binaries

Our sample comprises five HMXBs; only for HD 77581/Vela X-1 and HD 153919/4U1700-37 high-resolution IUE spectra will be discussed. The system parameters (Table 1) are collected from the literature. Vela X-1 is an X-ray pulsar, identifying the compact object as a neutron star. For 4U1700-37 no X-ray pulsations have been detected, but the X-ray spectrum suggests that the X-ray source is a neutron star (White & Marshall 1983; Reynolds et al. 1999). Both HMXBs are wind-fed systems (Kaper 1998) and have relatively modest X-ray luminosities. Cyg X-1 is a well-known black-hole candidate. The other two sources, SMC X-1 and LMC X-4 are located in the Magellanic Clouds. The low metallicity in the Magellanic Clouds and the occcurence of Roche-lobe overflow due to their tight orbits account for their high X-ray luminosities. Sk-Ph/LMC X-4 contains a (sub)giant rather than a supergiant. The short pulse period of both SMC X-1 and LMC X-4 suggests that these neutron stars are surrounded by an accretion disk. Periods of 60 and 30 days in the lightcurves of respectively SMC X-1 (Wojdowski et al. 1998) and LMC X-4 (Heemskerk & van Paradijs 1989) are interpreted as the precession period of this disk.

The observation logs of the IUE spectra are listed in Table 2. Some spectra were included in previous studies (Dupree et al. 1978, 1980; Treves et al. 1980; Bonnet-Bidaud et al. 1981; van der Klis et al. 1982; Sadakane et al. 1985; Hammerschlag-Hensberge et al. 1990; Kaper et al. 1993), but many are used here for the first time. The low-resolution spectra (HDE 226868/Cyg X-1, Sk 160/SMC X-1 and Sk-Ph/LMC X-4) are sampled on a grid of 1 Å per point. The IUEDR software package (Giddings 1983) provided by STARLINK was used for data reduction. We refer to Kaper et al. (1993) for details on the data reduction of the high-resolution spectra of HD 77581/Vela X-1 and HD 153919/4U1700-37, sampled on a grid of 0.1 Å per point. The determination and subtraction of the inter-order background light has been improved for the high-resolution IUE final archive spectra in the INES system (http://ines.vilspa.esa.es) using NEWSIPS (González-Riestra et al. 2000), which may be important for the N  V resonance line around 1240 Å.

The exposure times are typically 45, 37, 45, 150 and 30 min for HDE 226868/Cyg X-1, Sk 160/SMC X-1, Sk-Ph/LMC X-4, HD 77581/Vela X-1 and HD 153919/ 4U1700-37, respectively, corresponding to a spread in orbital phase of 0.006, 0.007, 0.022, 0.012 and 0.006 per spectrum, respectively. Average X-ray eclipse ($\phi =0$) spectra, which should be representative of the "undisturbed'' stellar wind of the OB companion, are presented in Appendix A. The continuum variability (which becomes apparent when normalising the spectra), known as ellipsoidal variations, is due to the "pear-like'' shape of the supergiant filling its Roche lobe. These variations are discussed in Appendices B and C. The method we used for our error and variability analysis is based on the calculation of covariances and is described in Appendix D.

The strongest variability is detected in the ultraviolet ²S-²P⁰ resonance doublets that are formed in the stellar wind; e.g. C  IV ( $\lambda_0=1548.20$& 1550.774 Å, $\Delta v=498$ kms^-1), Si  IV ( $\lambda_0=1393.755$ & 1402.77 Å, $\Delta v=1927$ kms^-1), and N  V ( $\lambda_0=1238.821$ & 1242.804 Å, $\Delta v=961$ kms^-1). The orbital modulation of the mean flux in these lines is shown in Figs. 1 (HDE 226868/Cyg X-1, Sk 160/SMC X-1 and Sk-Ph/LMC X-4) and 2 (HD 77581/Vela X-1 and HD 153919/4U1700-37). The error bars follow from the covariance-based procedure. In the luminous (Roche-Lobe overflow) X-ray sources Cyg X-1, SMC X-1 and LMC X-4, the X-rays ionize most of the stellar wind, leaving only a shadow wind unaffected. In the fainter (wind-fed) X-ray sources Vela X-1 and 4U1700-37 the stellar wind is much less disturbed.

2.1 Cyg X-1

In HDE 226868/Cyg X-1 (Fig. 1) all three lines show clear modulation of the strength of the P-Cygni absorption with orbital phase due to the HM-effect. For C  IV and Si  IV this was already observed by Treves et al. (1980). The N  V line shows the HM-effect too, although less pronounced. The modulation is very strong and not confined to orbital phases near $\phi =0.5$: apparently, a large fraction of the stellar wind is strongly ionized by the X-ray source, suggesting the presence of a shadow wind in this system. The low resolution does not allow a detailed study of the wind velocity structure, but it is clear that the modulation of the absorption is visible up to the highest wind velocities.

2.2 LMC X-4

In Sk-Ph/LMC X-4 (Fig. 1) the HM-effect is seen in all three lines, in agreement with van der Klis et al. (1982). The N  V profile is completely absent at $\phi =0.5$, whereas part of the C  IV line remains visible. High-resolution HST GHRS/STIS spectra (Boroson et al. 2001) show that the remaining C  IV absorption at $\phi =0.5$ is mostly due to the intervening interstellar medium. These HST data also show some N  V absorption to persist around $\phi =0.5$, which is probably photospheric. The strength of the P-Cygni lines as a function of orbital phase indicate that the stellar wind is confined to the X-ray shadow behind the OB star. The orbital modulation of the C  IV and N  V profiles demonstrates a complexity beyond that of a smooth single wave, indicating a more complex mass flow. This has been confirmed with recent HST GHRS/STIS spectra (Boroson et al. 1999; Kaper et al. in preparation).

Figure 1: Resonance lines of C  IV (left), Si  IV (middle) and N  V (right) for HDE 226868/Cyg X-1 (top), Sk 160/SMC X-1 (middle) and Sk-Ph/LMC X-4 (bottom). Each upper panel displays the average spectrum near $\phi =0$ (drawn) and $\phi =0.5$ (dotted). The short-dashed vertical line marks the restwavelength, while the two long-dashed lines mark the boundaries of the region in which the mean flux is determined and plotted in the lower panel against orbital phase. The dotted horizontal line in each lower panel represents the mean flux level in the $\phi =0$ average spectrum.

Figure 2: Same as Fig. 1, but now for HD 77581/Vela X-1 (top) and HD 153919/4U1700-37 (bottom).

2.3 SMC X-1

In Sk 160/SMC X-1 (Fig. 1) the HM-effect is seen in all three lines (see also van der Klis et al. 1982). The shape and amplitude of the orbital modulation of the mean flux suggests that absorption is present in a tight orbital phase range around $\phi =0$ only, i.e. a shadow wind. This is consistent with the high X-ray luminosity of SMC X-1 (Hutchings 1974; van der Klis et al. 1982).

2.4 Vela X-1

In HD 77581/Vela X-1 (Fig. 2) the HM-effect is observed in the C  IV and Si  IV lines (Dupree et al. 1980) and the N  V line (see Kaper et al. 1993). The severely saturated C  IV line shows less orbital modulation than the just saturated Si  IV line; remarkably, the N  V line shows the inverse behaviour. The variability in the UV lines of HD 77581/Vela X-1 will be studied in more detail in Sect. 4.1.

2.5 4U1700-37

In HD 153919/4U1700-37 (Fig. 2) the P-Cygni lines are very strong and saturated. The N  V line does not show any significant variability with orbital phase. The C  IV and Si  IV absorption lines are possibly slightly weaker very near $\phi =0.5$. This would indicate a very small Strömgren zone, if the variability is due to the HM-effect. The variability in the C  IV line results from changes in the blue edge of the P-Cygni profile, while the variability in the Si  IV line originates at wind velocities of about -900 kms^-1. However, the UV spectrum is affected by variable Raman scattered emission lines which show a similar orbital-phase dependence (Kaper et al. 1990, 1993).