3. Chemical composition
Our analysis is based on line-blanketed, hydrogen-deficient model atmospheres, similar to those described by Asplund et al. (1997) but with a range of hydrogen abundances. In estimating the stellar parameters , log g and hydrogen abundance various ionization (Fe I /Fe II, Mg I /Mg II, Si I /Si II, Cr I /Cr II) and excitation equilibria ([O I ]/O I, Fe I, Fe II) together with the H and H line profiles (with line broadening data following Seaton 1990) have been used. The C/He ratio was determined from the C II and He I lines in the May spectra, which indicate C/He %. The same ratio had to be assumed for October when the lines were too weak to be utilized. The microturbulence parameter was estimated from Ti II, Fe I and Fe II lines of different strengths. The May spectra are characterized by K, log , and km s-1, while it had cooled significantly in October: K, log , and km s-1. In fact, the derived parameters are not consistent with a constant stellar luminosity but rather indicate a decrease by a factor of 4, which is not supported by the observed photometry. It could, however, be that hydrostatic equilibrium is inapplicable in May due to an expansion of the star or effects of turbulent pressure: a dynamical atmosphere can be mimicked by an underestimate of log g when assuming hydrostatic equilibrium. Indeed, with the May parameters the star is located at the classical Eddington limit (e.g. Asplund & Gustafsson 1996).
The analysis of the C I lines reveals the same inconsistency between theoretical and observed line strengths as for R CrB stars (Gustafsson & Asplund 1996; Lambert et al., in preparation): the strengths of weak lines predicted with the input C abundance are a factor of 4 stronger than observed (Fig. 1 and 2). It should be noted that no agreement between all -log g indicators could be achieved using consistent C abundance for the analysis. Naturally, this C I problem makes the absolute abundances uncertain but relative abundances are generally expected to be much less affected (Lambert et al., in preparation).
The derived LTE abundances for May and October are summarized in Table 1. More details on the analysis and atomic data (lines, gf, hfs, etc) as well as a comparison with V854 Cen will be given elsewhere. The weak Balmer lines certainly rule out a solar hydrogen abundance (Fig. 2). Note that the absolute abundances of most elements are effectively unchanged from May to October within the uncertainties (typically dex). Some elements, however, exhibit a marked change, for example, hydrogen declined as lithium and the light s -process elements increased in abundance by a factor of about 4 (Fig. 1). Also Sc, Ti, Cr and Zn seem to have increased during the timespan (Fig. 1). The general agreement between the May and October abundances for most elements suggests that the stellar parameters are not seriously in error, which could otherwise have resulted in spurious abundance effects. Besides Li, the abundances of elements showing variations are not very sensitive to the stellar parameters: the required K for either May or October to annul the abundance variations would be inconsistent with the -log g indicators and introduce other as severe changes (e.g. for Ca) less easily explainable; a different log g can not simultaneously explain all changes. It would also only aggravate the luminosity discrepancy. Hence, the few changes seem to be real. Furthermore, they are limited to elements expected to show alterations due to a final flash.
The metallicity of Sakurai's object is, judging from Fe, slightly below solar by 0.2 dex in mass fraction (0.9 dex if the input rather than the spectroscopic C abundance is adopted). The quantities [Si/Fe], [S/Fe], [Ca/Fe], and [Ti/Fe], which are 0.8, 0.6, 0.3 and 0.4 respectively, are, if unchanged from the star's birth, also indicative of a metal-poor star (Edvardsson et al. 1993). An isotopic ratio C/13 C is determined from the strong C2 (1-0) and (0-1) Swan bands (Fig. 2). The strengthening of the C2 bands due to the change in stellar parameters is clearly illustrated in Fig. 2.
It is of considerable interest to compare the compositions of Sakurai's object and FG Sge, another born-again candidate which has recently experienced R CrB-like visual declines. FG Sge resembles Sakurai's object in that it is strongly s -element enriched (Langer et al. 1974), as well as carbon-rich and poor in iron-group elements, except for Sc (Kipper & Kipper 1993). In FG Sge, however, the heavy s -elements are as overabundant as the light, and it has not yet been shown to be hydrogen-deficient. FG Sge may therefore have experienced a late shell flash as a luminous post-AGB star rather than a final flash as a white dwarf (Blöcker & Schönberner 1996).
Two of the outstanding aspects of the chemical composition of Sakurai's object are hallmarks of the R CrBs: H-deficiency and a high C content, but also other similarities in relative abundances exist (Lambert et al., in preparation; Rao & Lambert 1996; Lambert & Rao 1994). Except for the high Y/Fe other observed X/Fe ratios are similar to those found in R CrB stars. In particular it resembles the (relatively) H-rich V854 Cen (Asplund et al., in preparation). If, however, C/He is correctly estimated, it may sooner be related to objects such as V605 Aql, Abell 30 and 78 and the hot R CrB star V348 Sgr, which are also surrounded by planetary nebulae and have been proposed to be final flash candidates (Renzini 1990).
Similar abundance patterns as presented here for Sakurai's object have also been obtained in less detailed analyses by Shetrone & Keane (1997) and Kipper & Klochkova (1997). Shetrone & Keane's finding of a near normal H abundance is however very puzzling.
© European Southern Observatory (ESO) 1997
Online publication: June 30, 1998