We compare ultraviolet (UV) spectra of the recent soft X-ray transients XTE J1118+480 and XTE J1859+226. The emission line strengths in XTE J1118+480 strongly suggest that the accreting material has been CNO processed. We show that this system must have come into contact with a secondary star of about 1.5 M⊙, and an orbital period ∼15 h, very close to the bifurcation value at which the nuclear and angular momentum loss time-scales are similar. Subsequent evolution to the current period of 4.1 h was driven by angular momentum loss. In passing through a period of 7.75 h the secondary star would have shown essentially normal surface abundances. XTE J1118+480 could thus represent a slightly later evolutionary stage of A0620-00. We briefly discuss the broad Lyα absorption wings in XTE J1118+480.
The recently discovered soft X-ray transient (SXT) XTE J1118+480 (=KV UMa; Remillard et al. 2000) lies at high galactic latitude, close to the Lockman Hole in the local interstellar medium (ISM). The exceptionally low interstellar absorption permits unprecedented wavelength coverage (Mauche et al. 2000; Hynes et al. 2000b; McClintock et al. 2001b; Chaty et al., in preparation).
A low-amplitude photometric modulation with a period of 4.1 h was reported by Cook et al. (2000) and Uemura (2000). Dubus et al. (2000) present spectroscopy showing an emission line radial velocity ‘S-wave’ modulation at a period close to this. This period was recently confirmed as orbital, and a mass function of f(M)≈6 M⊙ was measured by McClintock et al. (2001a) and Wagner et al. (2001). Therefore the accretor clearly has to be a black hole.
We present the UV line spectrum of XTE J1118+480, which we compare with that of another recent SXT, XTE J1859+226. Full analyses of these Hubble Space Telescope (HST) spectra will be presented elsewhere (Hynes et al. 2002, and also in preparation; Chaty et al., in preparation). Here we focus on the prior evolution of XTE J1118+480.
We obtained HST spectra of XTE J1859+226 near the peak of outburst on 1999 October 18 and 27 and November 6. At each epoch, far-UV spectra were obtained using the low-resolution G140L grating and wide (0.5-arcsec) slit, yielding wavelength coverage of 1150–1730 Å and resolution ∼2.1 Å. We used the standard pipeline data products, except that the spectral extraction was done by hand using the iraf implementation of optimal extraction. This gave a significantly cleaner removal of geocoronal emission lines (principally Lyα and O i 1302 Å). After extracting the one-dimensional spectra (with the pipeline wavelength and flux calibrations), we resampled them on to a common wavelength grid and took an exposure-time weighted average.
For XTE J1118+480, we obtained high-resolution (echelle) spectra using the E140M grating and 0.2-arcsec2 aperture on 2000 April 8, 18 and 29, May 28, June 24, and July 8. The pipeline extraction was adequate, although a few high pixel values had to be replaced by hand. The extracted spectra were rebinned into 0.2-Å bins and averaged with exposure-time weighting. The process is clearly not perfect. In particular, there are a number of abrupt steps in the spectrum at longer wavelengths. These artefacts, at the ≲5 per cent level, arise from inconsistencies in the relative flux calibration of adjacent echelle orders.
Fig. 1 gives the UV spectra of XTE J1118+480 and XTE J1859+226. Both show interstellar absorption features. In XTE J1118+480 these appear sharper because the spectral resolution was higher.
The difference in the equivalent widths of the emission lines in the two systems is striking (Fig. 1 and Table 1; see also Haswell et al. 2000). In XTE J1859+226, C iv is the strongest line, with N v, C iii, O v, Si iv, and He ii also strongly present. These emission line strengths are roughly as expected for an X-ray bright LMXB (cf. Sco X-1; Kallman, Boroson, & Vrtilek 1998). Compared to cataclysmic variables (CVs), the N v/C iv ratio is relatively large, while Si iv/C iv is small, as expected for a high ionization parameter (Mauche, Lee & Kallman 1997). In contrast, for XTE J1118+480, the carbon and oxygen lines are undetectable, while the N v emission appears enhanced. A similar anomalous emission line spectrum is seen in the magnetic CV AE Aqr, in which the N v is much stronger than the C iv line (Jameson, King & Sherrington 1980; Eracleous et al. 1994).
The N iv 1718-Å line would add a valuable additional constraint on the interpretation of the line spectrum in terms of the photo-ionization conditions, so we carefully assessed our XTE J1118+480 spectra to determine whether it was present. Unfortunately, our well-exposed E140M echelle spectrum stops at 1710 Å and our E230M echelle spectrum suffers from shorter total exposure and the low sensitivity of the NUV MAMA in this region. 1718 Å falls right on the boundary between the two STIS MAMA detectors, and consequently neither is optimized at this wavelength. We find no strong (≳3×10−14 erg s−1 cm−2) line in the spectrum, but this upper limit provides a weak constraint due to the poor quality of our data at this wavelength.
The Lyα absorption in XTE J1118+480 (Fig. 1) clearly has a sharp core, which is probably interstellar, with very broad wings. The continuum slope is well constrained by the spectrum longwards of 1280 Å. Extrapolating this continuum is suggestive of broad Lyα extending from ≳1160 to ≲1280 Å. The combination of noise, N v emission and absorption features hinder measurement of the Lyα wings, but they definitely extend at least from ∼1180 to ∼1230 Å.
The broad absorption may be damped Lyα from the H i in our galaxy (Bowen, private communication). If we assume, instead, that it is due to absorbing gas executing Keplerian motion around the black hole, and take the largest width suggested by our data, the full width implies velocities of ∼0.05c and absorbing material at distances as close as Rin∼200RSch. Taking the securely determined linewidth would imply absorption from Keplerian material at Rin∼500RSch.
An independent line of argument based on the shape of the spectral energy distribution, and dependent on the value of the neutral hydrogen column density, NH, has been used to estimate the disc inner radius (Hynes et al. 2000b; McClintock et al. 2001b; Chaty et al., in preparation). Hynes et al. (2000b) estimate NH=0.75×1020 cm-2, a choice which suggests the disc is terminated at ≳103RSch. This is inconsistent with the Keplerian interpretation of the broad Lyα absorption. Alternatively, informed by the Chandra data, McClintock et al. (2001) suggest that NH could be as high as 1.3×1020 cm-2, with Rin≳65RSch. Chaty et al. (in preparation) adopt NH=1.0×1020 cm-2, which leads to Rin∼350RSch. The latter two values are consistent with the Keplerian interpretation of the broad Lyα absorption.
The feature near 1425 Å does not have an obvious identification and we believe it may be spurious, although it does not correspond to any known instrumental artefacts (Sahu, private communication). This issue is under investigation.
Ionization differences versus abundance anomalies
The C iv 1550 Å and the N v 1240 Å lines are both resonance lines of lithium-like ions, and are produced under essentially the same physical conditions. Kallman & McCray (1982, hereafter KM82) present theoretical models of compact X-ray sources that predict the ionization structure expected in a wide variety of astrophysical situations, including galactic X-ray binaries. KM82 present eight distinct models; in each one the source luminosity L, the gas density n and the spectral shape are fixed. For each model the output includes the dependence on ionization parameter,
(where R is the distance from the central source) of the abundances (as a fraction of the total elemental abundance) of the ions of common elements.
KM82 consider (i) three species detected in both XTE J1118+480 and XTE J1859+226: He ii, Si iv and N v; and (ii) three species detected only in the spectrum of XTE J1859+226: C iv, C iii and O v. Strikingly similar ionization parameters were required for the two sets of lines in all models. In almost all cases the presence of N v places the tightest constraint on ξ, and almost invariably this is encompassed within the range producing C iv. In the minority of cases where there exists an ionization parameter which produces N v and not C iv, this range also produces O v. Consequently, KM82 give no set of photoionization parameters which would predict production of N v in XTE J1118+480 while suppressing the C iv and O v.
The X-ray spectrum of XTE J1118+480 is clearly different from that of XTE J1859+226: the former is an extended power law (Hynes et al. 2000b; McClintock et al. 2001b; Frontera et al. 2001; Chaty et al., in preparation ) with a cut-off at ∼100 keV indicative of the low, hard state, and is often attributed to an advection-dominated accretion flow (ADAF; e.g. Esin et al. 2001); while in the early stages of the outburst the latter was dominated by the soft thermal blackbody disc spectrum which turned over by ∼4 keV (Hynes et al. 2002), typical of the high, soft state. Ho et al. (2000) note that the characteristically harder ionizing spectrum of an ADAF lowers the effective ionization parameter, and hence favours the production of lower ionization species. Hence, if anything, the differences in the X-ray spectra should cause C iv to be relatively prominent (compared with N v) in XTE J1118+480, rather than being suppressed as we observe.
Hamann & Ferland (1992) used the observed C iv 1550 Å and N v 1240 Å lines to estimate the N/C abundance ratio in high-redshift quasi-stellar objects (QSOs). Their photionization calculations show that for a broad range of ionization parameters the N v1240 Å/C iv 1550 Å line ratio is lower at lower metallicities even when the N/C abundance ratio is kept constant. This means that, if anything, we should expect the N v 1240 Å line to be suppressed relative to C iv 1550 Å in XTE J1118+480, which is a halo object, and consequently might be expected to have a lower metallicity than that of XTE J1859+226.
A definitive abundance analysis for XTE J1118+480 would require exact knowledge of the geometry, densities and ionizing spectrum in the regions emitting the UV lines. While Hynes et al. (2000a, and also in preparation) gives some information on the geometry, and Chaty et al. (in preparation) gives good coverage of the X-ray spectrum, the EUV spectrum remains open to significant uncertainty (compare the range of dereddened EUV spectra in Hynes et al. 2000b; Merloni et al. 2000; McClintock et al. 2001b; Esin et al. 2001; and Chaty et al., in preparation). Furthermore the range of densities present in the UV line emitting gas is difficult to determine. Consequently a quantitative abundance determination of the type carried out in stellar photospheres, where the physical conditions are well-known, is not possible.
Hence it is impossible to rule out definitively a contrived spectrum and gas distribution which would produce the set (i) lines while suppressing the set (ii) lines. However, as
(1) set (i) and set (ii) ions require essentially identical ranges of ξ,
(2) the ionizing spectrum of XTE J1118+480 might be expected to favour C iv relative to N v, and
(3) the lower metallicity expected for XTE J1118+480 might be expected to boost the C iv1550 Å/N v 1240 Å line ratio,
by far the simplest explanation of our observed UV line spectra is that a substantial underabundance of carbon is present in the surface layers of the mass donor star in XTE J1118+480.
In XTE J1118+480 (as in AE Aqr), the underabundance of carbon compared to nitrogen strongly suggests that the material in the accretion flow is substantially CNO processed (Clayton 1983). If the CNO bi-cycle achieves equilibrium, it converts most CNO nuclei into 14N. At the lower temperatures typical in ∼1.5 M⊙ MS stars (which we shall suggest as the progenitor of the companion in XTE J1118+480) this conversion is much more at the expense of 12C than 16O, and in fact 17O increases. Hence the oxygen abundance is not expected to decrease by much in XTE J1118+480 and this argument does not of itself explain the observed weakness of O v 1371 Å. However C iv1550 Å N v 1240 Å are resonance lines, whose ratio must be regarded as a robust indicator of conditions in the line emitting gas, whereas O v 1371 Å is a subordinate line, which may not be formed efficiently in a low density photoionized gas. We thus regard the spectrum of XTE J1118+480 as indicating CNO processing.
The evolution of XTE J1118+480
The spectrum shown in Fig. 1 implies that the companion star in XTE J1118+480 must be partially nuclear-evolved and have lost its outer layers, exposing inner layers which have been mixed with CNO-processed material from the central nuclear-burning region.
Mass transfer must therefore have been initiated from a somewhat evolved and sufficiently massive donor of M2i≳1.5 M⊙, and thus with an initial period Pi≳12 h. The main difficulty in understanding the current status of XTE J1118+480 is to explain its observed short orbital period; in general, significant nuclear evolution is associated with a period increase, rather than the inferred decrease.
There are two ways in which this decrease could have occurred, listed below.
(i) The system could have come into contact with initial secondary/primary mass ratio qi=M2i/M1i≳1 and undergone a phase of thermal time-scale mass transfer until q≲1. At the end of this phase the period would have either decreased or not increased greatly. Subsequent orbital angular momentum loss could then pull the system in to the observed Porb=4.1 h. This case is similar to the likely evolution of AE Aqr (Schenker et al., in preparation).
(ii) The system could have come into contact with qi<1 already, and subsequent evolution could then be driven by angular momentum loss towards shorter periods. This case resembles that proposed for A0620–00 by de Kool, van den Heuvel & Pylyser (1983), with a severe additional constraint: the donor must be sufficiently evolved to provide the observed surface abundance ratios. This requires a near–equality of the nuclear and orbital angular momentum loss time-scales tnuc, tAML (cf. Fig. 2). Put another way, Pi must have been very close to the ‘bifurcation period’ defined by Pylyser & Savonije (1988).
Case (i) above can be ruled out by the current system parameters. At the end of thermal time-scale mass transfer we must have q≲1, which would require a donor mass M2∼6 M⊙ and an orbital period ∼15 h. However, for such a system tnuc≪tAML (cf. Fig. 2), implying evolution to longer periods, in stark contrast to the current 4.1 h. Put another way, there is an upper limit on the binary mass M=(1+q-1)M2 for Case (i) evolution to give nuclear-evolved donors at short orbital periods. The near equality tnuc∼tAML can only realistically hold if magnetic braking dominates angular momentum loss, so we require M2≲1.5 M⊙ at the end of the thermal time-scale episode. However, because q∼1 there, we must have total binary mass M≲3 M⊙ in Case (i) evolutions.
We are therefore left only with Case (ii) above. The secondary mass when the system came into contact is now constrained to be close to 1.5 M⊙: CNO processing excludes significantly lower masses, while the requirement for tAML≈tMB>tnuc excludes significantly higher masses. Figs 3 and 4 show the evolution of such a system: a 1.5 M⊙ MS star was allowed to evolve and fill its Roche lobe in a 15-hr binary with a 7-M⊙ black hole. The predicted mass transfer rate throughout the evolution, including its current value 3×10-10 M⊙ yr-1, shows that the system would indeed have appeared as a soft X-ray transient, according to the irradiated disc criterion for black hole systems given by King, Kolb & Szuszkiewicz (1997). It should be noted that the change in C/N is mostly due to a drastic depletion of 12C, supplemented by a more modest increase in 14N. The total O abundance at the surface has only decreased very little; in fact the O/N ratio is reduced by only a factor of 3 to 5 at most, predominantly due to the increase of N. Interestingly, this model also passes through Porb=7.75 h at around M2≈0.6 M⊙, showing much weaker N enhancements (i.e. almost normal abundances, cf. Fig. 3). Thus it can also be considered a reasonable model for A0620–00. A subsequent paper (Schenker et al., in preparation) will explore this and related evolutions in detail.
Finally we can compare further properties of the secondary in our model to observations: Fig. 4 shows the evolution of effective temperature with orbital period, together with a simple mapping to spectral types. The procedure is based on a conversion of observed colours to spectral types (Beuermann et al. 1998) and a set of non-grey stellar atmospheres (Hauschildt, Allan & Baron 1999) providing the colours for each set of stellar boundary conditions. For solar metallicity the resulting SpT turns out to be only a function of effective temperature with a very weak dependence on surface gravity. However, this mapping should only be considered a first estimate, as the evolutionary code in its current form still uses grey atmospheres. Furthermore the whole conversion is based on an observed set of unevolved stars and theoretical zero-age main-sequence models, i.e. no fully self-consistent application to a partially evolved MS star such as shown in Fig. 4 is possible for the time being. Allowing for the uncertainties described above, our model may be slightly too cool [M2 rather than K5-M1 as mentioned in McClintock et al. (2001a), or K7-M0 by Wagner et al. (2001)]. Similar evolutionary tracks of strongly evolved CVs are known to get hotter again (Baraffe & Kolb 2000) at short periods, so a confirmed secondary spectral type would provide insight as to the state of nuclear evolution in the donor star of XTE J1118+480. Fig. 4 shows that, in any case, the donor will currently appear to be close to the ZAMS, and thus that the distance estimate derived by McClintock et al. (2000) on this basis is likely to be quite good.
We have compared the UV spectra of XTE J1118+480 and XTE J1859+226. The former shows strong evidence of CNO processing, which tightly constrains the evolution of the system. XTE J1118+480 must have reached contact with the donor star having a significantly nuclear-evolved core first, at a period where nuclear and angular momentum loss time-scales were comparable. This in turn constrains the end point of the earlier common envelope phase: immediately after the common envelope phase the system had a wide enough separation that significant nuclear evolution could occur before contact was achieved. In a future paper we will investigate such evolutions systematically.