2012A&A...542A..45K

The commonly accepted mass-luminosity relation of central stars of planetary nebulae (CSPNs) might not be universally valid. While the atmospheric parameters T_eff, logg, and the He abundances have in the past been determined using fits to photospheric H and He absorption lines from plane-parallel model atmospheres, the masses and luminosities could not be derived independently this way, and were instead taken from theoretical evolutionary models. Improvements in hydrodynamically selfconsistent modelling of the stellar atmospheric outflows now allow using fits to the wind-sensitive features in the UV spectra to consistently determine the stellar radii, masses, and luminosities without assuming a mass-luminosity relation. Recent application to a sample of CSPNs in an earlier paper of this series raised questions regarding the validity of the mass-luminosity relation of central stars of planetary nebulae. The results of the earlier UV analysis are reassessed by means of a simultaneous comparison of both the observed optical and UV spectra with corresponding synthetic spectra. The synthetic optical and UV spectra are computed simultaneously from the same atmospheric models, using the same model atmosphere code. Synthetic spectra for the two central stars NGC 6826 and NGC 2392 are computed using parameter sets from two different published analyses to check their compatibility to the observations. Using the different published stellar parameter sets, derived on the one hand by a consistent UV analysis, and on the other hand from fits to optical H and He lines, we calculate corresponding optical and UV spectra with our model atmosphere code. We have improved this model atmosphere code by implementing Stark broadening for hydrogen and helium lines, thus allowing us to obtain consistent optical H and He line profiles simultaneously with our state-of-the-art modelling of the UV-spectrum. Optical line profiles computed with the consistent parameter sets from the UV analysis yield line profiles with good agreement to the observations (with small discrepancies in the emission lines corresponding to about a factor of one half in the mass loss rate). Spectra computed with the stellar parameter sets from the optical analysis in the literature and corresponding consistent wind parameters, however, show large discrepancies to the observed spectra, in the optical as well as in the UV. We conclude that the published optical analyses give good fits to the observed spectrum only because the wind parameters assumed in these analyses are inconsistent to their stellar parameters. By enforcing consistency between stellar and wind parameters, stellar parameters are obtained which disagree with the core-mass-luminosity relation for the two objects analyzed. This disagreement is also evident from a completely different approach: an investigation of the dynamical wind parameters.