6 Conclusions

We have presented detailed chemical abundances for four MS B stars in NGC2004. This study doubles the number of MS B stars in the LMC for which accurate abundances have been derived by means of high-resolution spectroscopy. None of the stars shows any indication of effects due to binary interaction or rotational mixing allowing us to derive what we consider to be the first data set of unaltered present-day CNO abundances from LMC stars.

• We confirm the extraordinarily low LMC nitrogen abundance previously found from H II-region studies to within 0.1 dex. This value is more than 0.5dex below average values found in the Galactic thin disk at LMC metallicities (Liang et al. 2001, cf. their Fig. 10). It implies an enrichment history for the two environments different from one another and - within the chemical evolution model of Henry et al. (2000) - a dominance of primary nitrogen production in the LMC until today.
• In our programme stars, carbon is 0.16dex above the nebular value. While this is the direction in which dust depletion would act, the offset found is hardly significant considering the error limits on both sides. More work is needed to reduce both random and systematic errors to draw definitive conclusions on dust depletion fractions.
• The B-star oxygen abundance (also from B giants, cf. Paper I) is in excellent agreement with the nebular value. In the absence of systematic offsets between the two data sets, little nebular oxygen seems to be tied up in grains at LMC metallicities.
• The stellar silicon abundance is offset from the nebular one by 0.4dex. This might imply that a major fraction of the interstellar silicon in the LMC is bound in grains.

This work illustrates the potential of abundance analyses of hot stars based on high-quality spectra and sophisticated input physics. With modern multi-object spectrographs (e.g. GIRAFFE on VLT UT2 going on-line in 2002) we will be able to analyse some 100 slowly rotating B stars in each Galactic and MC cluster like NGC2004. This will not only result in reduced random errors (which we will be able to address truly statistically for the first time), but also prove or refute the importance of rotational mixing for this class of stellar objects. In particular, a sequence consisting of Galactic, LMC and SMC B stars could settle the rôle metallicity, rotation and mass loss play in hot-star evolution. As for the H II-region data, more objects will have to be observed to lower the errors on that side as well.

Systematic errors - arguably more important than random ones at the current level of accuracy - will have to be addressed by an integral approach: by studying the stellar and gas component of H II regions in tandem. Both fields of research would undoubtedly profit from such an endeavour.

Acknowledgements

AJK wishes to thank I.Appenzeller for making this work possible financially: the hospitality of the LSW during April and May 2001 was well-appreciated. We thank the anonymous referee for valuable comments on the initially submitted version.

SCK's work at LLNL was performed under the auspices of the U.S. Department of Energy by the University of California Lawrence Livermore National Laboratory under contract No. W7405-ENG-48.