As will be shown in Sect. 3, determination of the masses of both components in the system is hampered by the considerable uncertainties in the stellar radius of HD 153919. In order to address this problem, and in the light of dramatic advances in the sophistication of non-LTE model atmospheres we have decided to reanalyse both new and published ultraviolet to near-infrared spectroscopic and optical to mid-infrared photometric observations of HD 153919 in order to refine previous estimates of the stellar parameters.

$\begin{figure} \par\includegraphics[angle=-90,width=17cm,clip]{MS2500f1.eps} \end{figure}$

Figure 1: Plots of selected regions of the optical spectrum of HD 153919 (solid line) and best fit model (dotted line); parameters listed in Table 1.

To determine the stellar properties of HD 153919 we have utilised the non-LTE code of Hillier & Miller (1998) which solves the radiative transfer equation subject to the constraints of statistical and radiative equilibrium, in a spherical, extended atmosphere. Line blanketing is incorporated directly through the use of a super-level approach. We use a similar atomic model to that employed by Crowther et al. (2002) in their study of early O supergiants, including H I, He I-II, C III-IV, N III-V, O III-VI, Si IV, P IV-V, S IV-VI and Fe IV-VII. For extreme O supergiants, line blanketing and the strong stellar wind conspire to produce significant differences in stellar parameters relative to the standard plane-parallel hydrostatic results (see Crowther et al. 2002 for further details).

Our procedure is as follows. We adjust the stellar temperature and mass-loss rate of an individual model until the "photospheric'' He II $\lambda$ 4542 and He I $\lambda$ 4471 lines are reproduced. Simultaneously, we vary the total mass-loss rate until H $\alpha$ is also matched. The exponent of the $\beta$ -law is adjusted until the shape of H $\alpha$ is well reproduced - for HD 153919 we obtain $\beta\sim 1.3$ . The input atmospheric structure, connecting the spherically extended hydrostatic layers to the $\beta$ -law wind is achieved via a parameterized scale height, h (see Hillier et al. 2002 for details), for which h=0.001 yields a reasonable match to He I and Balmer line wings, consistent with $\log g=3.45{-}3.55$ . We adopt a terminal wind velocity of $v_{\infty}=1750$ km s^-1 (vL01; Howarth et al. 1997).

The formal solution of the radiative transfer equation yielding the final emergent spectrum is computed separately, and includes standard Stark broadening tables for H I, He I-II. Except where noted, these calculations assume a microturbulent velocity $v_{\rm turb}=10$ km s^-1. Hillier et al. (2002) also discuss the effect of varying $v_{\rm turb}$ in O star models. Additionally, we find good agreement with observations using $v \sin i=150$ km s^-1 (Howarth et al. 1997 derived 120 km s^-1).

It is extremely difficult to determine accurate He/H abundances in O supergiants as discussed by Hillier et al. (2002). Consequently, we adopt $\rm He/H=0.2$ by number, whilst C and N abundances are varied until diagnostic optical line profiles are reproduced. In Fig. 1 we present selected optical line profile fits to FEROS observations of HD 153919. Overall, agreement for $T_{\rm eff}=35$ kK is very good, with the exception of He II $\lambda$ 4686. He I $\lambda$ 4471 provides our main temperature constraint since other blue optical He I lines are weak or absent. Alternatively, we considered using He I $\lambda$ 5876 (or $\lambda$ 10830) together with the He II $\lambda$ 4686 line. However, this method (followed by Crowther & Bohannan 1997) yields significantly ( $\sim$ 4 kK) lower stellar temperatures, and suffers from inconsistencies involving the ionization balance of UV/optical metal lines. Therefore, we have greater confidence in our adopted diagnostics, which do not suffer from such problems.

$\begin{figure} \par\includegraphics[width=8.8cm,clip]{MS2500f2.eps} \end{figure}$

Figure 2: Plot of the observed (solid line and data points) and theoretical (dotted line) UV-mid-IR spectral energy distribution for HD 153919.

$\begin{figure} \par\includegraphics[width=18cm,clip]{MS2500f3.eps} \end{figure}$

Figure 3: Plots of selected regions of the UV spectrum of HD 153919 (solid line) and best fit model (dotted line); parameters listed in Table 1. Note that our model does not take into account the observed Raman-scattered emission lines (which are not of a photospheric origin) in the range 1400-1700 Å.

From spectral energy distribution fits to IUE spectrophotometry and Johnson photometry (Fig. 2), we derive $E_{B-V} = 0.53\pm 0.02$ mag. Alternatively, using the intrinsic colour of (B-V)₀=-0.30from the early O supergiant calibration of Schmidt-Kaler (1982), we derive E_B-V=0.55 from Johnson photometry of HD 153919 (V=6.54 and B-V=0.25, Bolton & Herbst 1976). Consequently, we adopt $E_{B-V}=0.54\pm0.02$ for the remainder of this work. We adopt a distance modulus of 11.4 mag to HD 153919 (Ankay et al. 2001), implying M_V=-6.53 mag. Our derived temperature provides a bolometric correction of -3.3 mag, therefore we obtain $\log (L/L_{\odot})=5.82$ for HD 153919, and thus R=21.9 $R_{\odot }$ . The derived mass-loss rate is $9.5\times 10^{-6}$ $M_{\odot }$ yr^-1, assuming that the H $\alpha$ line-forming region is not clumped. Moderate clumping would reduce this value by a factor of $\sim$ 2. Derived parameters are listed in Table 1, and are in reasonable agreement with those derived by vL01 through the analysis of ultraviolet resonance lines based on the Sobolev with Exact Integration (SEI) method.

Our primary nitrogen abundance diagnostics are N III $\lambda$ 4634-41 and $\lambda$ 4097, which together imply $\epsilon_{\rm N} = 9 \epsilon_{{\rm N},\odot}$ . With this value, the very weak N III $\lambda$ 5320-4 feature is well matched, but other lines in the vicinity of He II $\lambda$ 4542 are somewhat too strong (likely due to an incomplete treatment of N III quartet states in our models). Carbon is somewhat more difficult to constrain, with C III $\lambda$ 4647-51 well matched for $\epsilon_{\rm C} = 1.0 \epsilon_{{\rm C},\odot}$ . C IV $\lambda$ 5801-12 is well reproduced with this value, whilst C III $\lambda$ 5696 is too weak, implying a yet higher abundance. As discussed elsewhere (e.g. Hillier et al. 1998), oxygen is exceedingly difficult to constrain in mid-O supergiants due to lack of suitable optical diagnostics. The high nitrogen overenrichment is not easily explained via single star evolution, unless carbon (and to a lesser degree oxygen) is very depleted via the CN (or ON) cycle. Crowther et al. (2002) discuss similar problems for Magellanic Cloud O supergiants.

Turning to UV comparisons, we show rectified high resolution IUE spectroscopy of HD 153919 (phase 0.15) in Fig. 3 together with synthetic spectra. Overall, agreement for He II $\lambda$ 1640, C IV $\lambda$ 1550 and N IV $\lambda$ 1718 is reasonable, with predicted Si IV $\lambda$ 1393-1402 emission too weak adopting $\epsilon_{\rm Si} = 1.0 \epsilon_{{\rm Si},\odot}$ . Since X-rays are not explicitly considered in this study, the shocked UV N V $\lambda$ 1238-42 resonance doublet is predicted to be too weak. The sole prominent oxygen feature present in the UV (or optical) region is O IV $\lambda$ 1338-43, which is reasonably well matched with $\epsilon_{\rm O} = 0.5 \epsilon_{{\rm O, \odot}}$ , although we do not claim that this represents an accurate constraint.

Additional, powerful evidence in favour of our derived temperature is the good match between the synthetic Fe IV-V spectrum and observations, again with $\epsilon_{\rm Fe} = 1.0 \epsilon_{{\rm Fe}, \odot}$ . Figure 2 shows good agreement with the dominant Fe V "forest'' observed between $\lambda$ 1300-1600 in HD 153919, plus the weaker Fe IV in the $\lambda$ 1500-1800 region. Fe VI is not strongly predicted nor observed in the $\lambda$ 1200-1400 region (see Crowther et al. 2002 for further details).

While it is possible to explain the nitrogen enrichment in terms of rotational mixing it is impossible to produce carbon enrichment via this mechanism. Any carbon produced in the helium burning layers of the star has to pass through the hydrogen burning layers before reaching the surface where it will be converted to nitrogen. Therefore, any excess carbon in HD 153919 is therefore likely to result from mass transfer from the more evolved binary component prior to SN - this will be returned to in Sect. 4.

Given the presence of a compact companion for HD 153919, it is reasonable to ask whether the assumption of spherical geometry is justified - does the X-ray flux lead to significant departures from spherical symmetry for the ionisation of the wind (which in turn could lead to modifications in the line driving force)? Hatchett & McCray (1977) suggest that the X-ray emission will lead to a reduction of moderately ionised atoms in the wind (such as Si IV and C IV). Given that the ionised zone will move with the compact object we might expect to see orbital modulation in some of the wind UV resonance lines as the ionised zone passes in front and behind the stellar disc (or from the presence of a photo-ionization wake in the system, cf. Kaper et al. 1994). However, there is no convincing evidence for orbital modulation in the UV resonance line due to the Hatchett-McCray effect (e.g. Kaper et al. 1990, 1993); the small changes in line profiles with orbital phase are instead most likely due to Raman scattering of EUV photons generated by the X-ray source (Kaper et al. 1990, 1993). Additionally, the modeling was performed on the spectrum obtained during the X-ray eclipse to further minimise any possible effects of irradiation on the stellar wind (cf. Sect. 2.1).

Using a modified 2-dimensional Sobolev Exact Integration (SEI) code vL01 analyse the UV line variability and confirm that any Strömgren sphere caused by the presence of the X-ray source is rather small, and will have a negligible effect on the ionization structure and line driving of the wind (since the wind is dense and the ionizing flux low). Equally, the Strömgren zone does not extend to the surface of the star and so should not lead to a significant degree of X-ray heating of the stellar surface.

Phase resolved continuum observations (e.g. vL01; Hammerschlag-Hensberge & Zuiderwijk 1977; van Paradijs et al. 1978) constrain orbital variability to <4 per cent in the UV and 4-8 per cent in the optical, indicating that HD 153919 shows little departure from sphericity (possibly as a result of a large mass ratio). Note that continuum emission from the stellar wind is essentially negligible at wavelengths shorter than a few microns. Therefore, the lack of significant variability cannot be attributed to emission from the outer regions of the stellar wind "shielding'' a heavily perturbed stellar surface and/or inner wind from view.

The photometric variability further constrains any change in stellar temperature due to X-ray heating to less than the uncertainty in the stellar temperature derived from our NLTE modeling. Therefore, we have confidence that deviations from spherical symmetry in HD 153919 and/or the effects of X-ray irradiation are negligible for the purposes of spectroscopic modeling.

2 Determination of the stellar parameters for HD 153919

2.1 The complete dataset

2.2 Spectral analysis