5 Results

Table 3 gives the stellar parameters and abundances derived for the four programme stars in comparison with H II-region data from Garnett (1999). Consistency among the abundances is generally very high with the rms scatter confined to well below 0.1dex. Also included is the main sequence star D1 in the LMC cluster NGC1818 already analysed in Paper I. For the latter star, the abundances of He, C and N are redetermined on the basis of all available lines and utilizing the new model atoms. Where available, the abundances are compared with measurements from LMC H II regions as carried out/compiled by Garnett (1999). Finally, we also include the Solar System composition (Grevesse & Sauval 1998) as a reference. Note that a recent redetermination of the solar oxygen abundance from [O I] 6300 based on 3D hydrodynamical computations (Allende Prieto et al. 2001) yields $\rm\varepsilon(O)=8.69\pm 0.05$. Holweger (2001) confirms this trend towards a lower value by deriving $\rm\varepsilon(O)=8.74\pm 0.08$ from eight optical and IR lines taking into account effects due to NLTE and granulation. Thus the difference between oxygen abundances from stars in Orion and the Sun reduces to 0.1dex at most (cf. Gummersbach et al. 1998), but important consequences can also be expected for the evaluation of the [$\alpha$/Fe] ratios in the MCs (see below).

Figure 2 shows the positions of the programme stars in a $\log\,T_{\rm eff}{-}\log\, g$ parameter space, sometimes referred to as a Kiel diagram. Note that the four MS stars nicely span the full width of the main sequence in the B-star temperature regime. This fact allows us to potentially address chemical evolution of certain elements in these stars during their MS lifetime: no such evolution is, however, found (cf. Table 3 in which the stars are ordered according to their gravity).

Figure 3 gives examples of profile fits for the most important species of this study: H, CNO, Si and Mg. NLTE effects in iron are currently under investigation. Thus, for the time being, $\varepsilon$(Fe) based on the few Fe III lines ( $W_\lambda<40$mÅ) visible between 4000Å and 5000Å is given assuming LTE.

5.1 Pristine CNO abundances in the LMC

In comparing the stellar CNO abundances with the nebular ones, one immediately notices several features: the oxygen abundances are in excellent agreement, while both nitrogen and carbon are somewhat more abundant in the stars. As carbon is one of the principle dust-forming elements, interstellar dust depletion is the first reason that comes to mind. Though not of high statistical significance, this offset of 0.16 dex could be interpreted as a manifestation of the fraction of carbon locked up in interstellar dust. Unfortunately, the error limits on both sides prevent us from drawing firm conclusions on this matter. Another observation, however, may lend support to this scenario: a very similar offset is found between the gas-phase carbon abundance of Orion ($\simeq$8.4, cf. Garnett et al. 1995) and that of the Sun between 8.52 (Grevesse & Sauval 1998) and 8.59 according to Holweger (2001). More work is clearly needed here.

Conversely, nitrogen is believed not to deplete into the dust phase in significant amounts (Mathis 1996). An enhanced nitrogen abundance in B stars could then be taken as an indication of mixing (internal or external), raising the original (nebular) abundance from which the stars formed some million years ago. The high internal consistency among the $\varepsilon$(N) determinations in the programme stars speaks against this scenario as a likely explanation. Besides, the residual offset with respect to the nebular abundance is hardly of high significance considering the accuracy of 0.2 dex for each of the data sets. We therefore take this result as strong evidence in favour of the extraordinarily low present-day nitrogen abundance of $\varepsilon{\rm (N)}\simeq7.0$. On the square-bracket scale this corresponds to $\rm [N/H]\le -0.92$ at a metallicity of $\rm [Fe/H]\simeq-0.2$ or $\rm [N/Fe]\simeq-0.7$. $\rm [N/Fe]\simeq 0$ is only reached after the first dredge-up on the red-giant branch, as already emphasized by Venn (1999a).

Finally, we comment on oxygen-related matters: there can be little doubt that the present-day LMC oxygen abundance is close to 8.40. With a mean nitrogen abundance of 7.01, our mean oxygen value of 8.37 results in $\rm\log(N/O)=-1.36$. Therefore, even though the metallicity of the LMC is not so different from solar, its N/O ratio is among the lowest found in the local universe, as derived from Galactic and extragalactic nebular and stellar abundances (cf. Henry et al. 2000). Indeed, within the chemical evolution model favoured by Henry et al., which attributes nitrogen production entirely to low and intermediate-mass stars, this implies that the present-day nitrogen abundance of the LMC is still dominated by primarily produced nitrogen, which is consistent with the metallicity-dependent nitrogen yields of van den Hoek & Groenewegen (1997).

Figure 4: Carbon and nitrogen profile fits for NGC1818/D1: the best-fit value is compared to the mean value as derived from the four MS stars in NGC2004. The thin line represents the LTE utility SYNSPEC ( $T_{\rm eff}=25\,000\,{\rm K}, \log\,g=4.0$, $\xi = 10$km s-1, solar composition, cf. http://feros.lsw.uni-heidelberg.de/cgi-bin/websynspec.cgi) which is used to locate the continuum. Note that some of the line strengths are greatly overestimated under LTE conditions (e.g. C II 4616/4619 and 4625).

  
5.2 The $\alpha$-elements

Due to a state-of-the-art redetermination of the solar oxygen abundance (see above) it has become unclear what [O/Fe] ratio $\rm\varepsilon(O)=8.40$ corresponds to. In fact, this question concerns not only oxygen, but all other $\alpha$-elements (Ne, Mg, Si, S, Ca and possibly Ti) in the sense that oxygen is often taken to be indicative of the behaviour of these elements. Thus the whole question of whether or not the [$\alpha$/Fe] ratios in the MCs are vastly different from those found in the Galactic thin disk at the appropriate metallicities must be re-evaluated in the light of these recent developments: from the three $\alpha$-elements we have access to, we derive $\rm [\alpha/Fe]\simeq-0.15$ using the new solar oxygen abundance. Even if we account for a lower NLTE Fe abundance, [$\alpha$/Fe] is still significantly lower than in the Galactic thin disk: +0.10 at $\rm [Fe/H]\simeq-0.3$ (Fuhrmann 1998, 2000).

For silicon we find a relatively clear offset of stellar silicon abundances relative to that from H II regions of 0.4dex which could be significant. The LMC giants of Paper I show a similar offset ($\simeq$0.3dex) which can most naturally be explained as the fraction of silicon locked up in interstellar grains (silicates).

5.3 Iron

On average, the iron abundances presented in Table 3 are merely 0.17dex below the solar value which is quite a high value for LMC objects. We reemphasize, however, that LTE results are given in Table 3. First results of the new NLTE computations indicate that the LTE results constitute upper limits to the NLTE abundances. Thus, the high iron abundances is most likely caused by the inappropriateness of LTE for stars as hot as 25000K. For the time being, we account for this fact by assigning larger error bars ($\pm$0.3dex).

5.4 Isochrone ages

In comparison with evolutionary tracks, the stellar parameters may serve to constrain the cluster age which has only been determined from colour-magnitude diagrams (CMD) so far. Biases are certainly present in both cases: MS fitting in CMDs will be prone to Be stars and rapid rotators "above'' the MS mimicking a hotter turnoff and artificially lowering the derived age (cf. Grebel et al. 1996). Additionally, (differential) reddening is always an issue. In the case of spectroscopic ages derived from individual stars, unknown cluster membership might introduce a bias, not even the direction of which is known. Advantageously, reddening does not enter.

Having shown that our programme stars are bona-fide unmixed objects, we are hopeful that the derived ages are at least not affected by rotation. As can be seen from Fig. 2, the ages scatter substantially, e.g. C9 is fully compatible with age zero. It is unclear whether or not this hints at multiple star formation episodes for NGC2004, as proposed by Caloi & Cassatella (1995) based on IUE data. Our best estimate comes from the most evolved MS target D15 close to the turnoff where the age resolution is highest: $19\,$Myr $\leq\,t_{\rm NGC\,2004}\,\leq\,24\,$Myr. A CMD analysis by Bencivenni et al. (1991) points towards a much lower age of 8Myr confirming an earlier result by Hodge (1983). Note, though, that these estimates are based on B-V colours which are a rather insensitive indicator of turnoff temperatures and prone to contamination with Be stars. More recently, Keller et al. (2000) utilized the HST filters F160BW (far UV), F555W ("v'') and F656N (H$\alpha$), separated the Be population and derived an age for the cluster of $16\pm 4$Myr, in much better agreement with our result.

5.5 A fresh look at NGC1818/D1

Having redetermined the He, C and N abundance of this star, it is worthwhile taking another look at it: even though both $\varepsilon$(He) and $\varepsilon$(N) have decreased due to better statistics and more sophisticated modelling, the signature indicative of mixing with CN-cycled material remains intact. More importantly, we now have pristine B star abundances to compare it to (see Fig. 4).

The agreement in the O, Mg, Si and Fe abundance with NGC2004/C16 (the object closest to D1 in terms of stellar parameters) is remarkable, yet both carbon and nitrogen are offset by $\geq$0.3 dex in opposite directions. We refrain from rephrasing the discussion already presented in Paper I, but simply stress that NGC1818/D1 remains a prime candidate for a MS B stars that has undergone rotational mixing. We urge observations of boron to be performed for this star which according to Fliegner et al. (1996) have the potential of distinguishing between external contamination (e.g. binary interaction) and an internal process like rotational mixing.