6 Conclusion

This paper presents a detailed model of the compact Galactic H II region G29.96 for which high quality data imagery and spectroscopy is available at both infrared and radio wavelengths, including recent ISO observations. The model, which is based on the photoionization code NEBU and state-of-the-art stellar atmosphere models, reproduces most of the observations, with the exception of a few points known to be less accurately measured. The radio and infrared data on G29.96 are best reproduced by a 2-density component model nebula, with a diffuse ( $n_{\rm H}\sim 600$ cm^-3) extended (1 pc) halo surrounding a dense ( $n_{\rm H}\sim 50\,000$ cm^-3) compact (0.1 pc) core.

Using CoStar stellar atmosphere models we derived an effective temperature of $\sim$ 30⁺²_-1 kK. Adopting more recent non-LTE line blanketed atmospheres with stellar winds, a somewhat higher $T_{{\rm eff}}$ $\sim$ 32-38 kK is found. This temperature range is compatible with all observational constraints. For $T_{{\rm eff}}$ $\sim$ 33-36 kK compatibility is also obtained with the K-band spectral type O5-O8 determined by (Watson & Hanson 1997) when recent downward revisions of the effective temperature scale of O stars (Martins et al. 2002) are taken into account.

We explored the effect of varying the different model parameters on the predictions. The main sources of uncertainty in determining the abundances are the fluxes of the hydrogen recombination lines and the geometry of the dense compact core.

The derived elemental abundances are in agreement with the lowest values found in previous studies. The most robust results are N/O and Ne/S which are 3.5 and 1.3 times the solar values, respectively.

From the reanalysis of the different available observational constraints (see Sect. 5.7) it is not surprising that several earlier studies reached apparently conflicting results on the spectral type or $T_{{\rm eff}}$ of the ionizing source of G29.96. This is mostly due to the following facts. First, proper photoionization models must account for the dependence of nebular lines on several parameters, including the ionization parameter, geometry etc. Second, consistent predictions from stellar models regarding ionizing fluxes, $T_{{\rm eff}}$ scales etc. should be used taking into account recent progress made with fully line blanketed non-LTE atmosphere models. Last, but not least, as the available K-band spectroscopy does not allow a determination of the luminosity class, allowance should be made for possible variations of the evolutionary state of ionizing sources in compact H II regions. In view of these issues one may question whether mid-IR analysis of compact H II regions yield intrinsically different estimates of $T_{{\rm eff}}$ compared to spectral types or other constraints as suggested by Hanson et al. (2002). Our detailed analysis of G29.96 indicates the opposite.

The age of the ionizing star required by our model is $\approx$ $3\times 10^6$ yr, much older than the expected lifetime of UCHII regions. This could indicate that G29.96 is not excited by a bona fide young massive star. Instead the ionizing star creating today the H II region G29.96 may have left its birthplace, exciting gas further out. This matter could still be part of the larger parental cloud from which the stellar cluster associated with G29.96 was formed.

Acknowledgements

We thank the referee for useful questions and comments. CM thanks D. Péquignot for helpful discussions on photoionization models and Ryszard Szczerba for discussions on dust. DS thanks Alan Watson, Margaret Hanson, and Yuri Izotov for useful discussions and Margaret Hanson for sharing data before publication. We thank Thierry Lanz for sending us line blanketed TLUSTY model atmospheres before publication, and John Hillier for making his atmosphere code CMFGEN available.