Young populous clusters like NGC2004 in the Large Magellanic Cloud (LMC) are ideal objects to study both the current state of chemical evolution of the LMC and the physics of hot stars at a metallicity a factor of two below that of the Sun.

The study of stellar absorption spectra yields chemical abundances with an accuracy comparable to that achieved from the analysis of nebulae such as H II regions. However, both sources are subject to physical processes which alter the abundances of certain chemical species in a systematic fashion. In most cases, the direction of these systematic effects is known, yet their extent is often uncertain. In particular, the scaling of these effects with metallicity is poorly known. In H II regions, it is mostly dust formation and depletion onto grains which will result in lower limits on the present-day (pristine) abundances. As for B stars, rapid rotation can lead to contamination of the atmospheres with CN-cycled material from the core resulting in He and N enrichment and a corresponding C (and - to a lesser extent - O) depletion (Fliegner et al. 1996; Meynet & Maeder 2000; Heger & Langer 2000). The onset of "rotational mixing'' is a function of the rotation rate, stellar mass and metallicity. Rotating models of hot stars indicate that it can take place very early on, i.e. whilst on the main sequence (MS). Furthermore, one can expect rotational mixing to be more efficient in metal-poor environments (cf. Maeder & Meynet 2001), as metal-deficient stars have smaller radii for a given mass and thus rotate faster for a given angular momentum.

Table 1: Chemical abundances of bona-fide MS B stars in the LMC in comparison with H II-region data. $\varepsilon$ (X):= $\log$ ( $n_{X}/n_{\rm H}$ )+12. Typical errors are 0.2-0.3dex.
star PS $34{-}16^{{\rm a}}$ LH 104-24 $^{{\rm a}}$ NGC1818/D1 $^{{\rm b}}$ H II $^{{\rm c}}$

$\varepsilon$ (C) 7.10 7.50 7.83 7.90

$\varepsilon$ (N) 7.50 7.70 7.59 $^{{\rm d}}$ 6.90

$\varepsilon$ (O) 8.40 8.50 8.46 8.40

$\varepsilon$ (Mg) 7.00 7.40 7.35 -

$\varepsilon$ (Si) 7.00 7.40 7.10 6.70

$\varepsilon$ (Fe) 7.20 - 7.34 -

**Table 1:** Chemical abundances of bona-fide MS B stars in the LMC in comparison with H II-region data. $\varepsilon$ (X):= $\log$ ( $n_{X}/n_{\rm H}$ )+12. Typical errors are 0.2-0.3dex.
star	PS $34{-}16^{{\rm a}}$	LH 104-24 $^{{\rm a}}$	NGC1818/D1 $^{{\rm b}}$	H II $^{{\rm c}}$
$\varepsilon$ (C)	7.10	7.50	7.83	7.90
$\varepsilon$ (N)	7.50	7.70	7.59 $^{{\rm d}}$	6.90
$\varepsilon$ (O)	8.40	8.50	8.46	8.40
$\varepsilon$ (Mg)	7.00	7.40	7.35	-
$\varepsilon$ (Si)	7.00	7.40	7.10	6.70
$\varepsilon$ (Fe)	7.20	-	7.34	-

$^{{\rm a}}$ From Rolleston et al. (1996), $^{{\rm b}}$ from Korn et al. (2000),
$^{{\rm c}}$ From Garnett (1999), $^{{\rm d}}$ revised in this publication.

Among a sample of B stars only the fastest rotators ( $v_{\rm rot}>$ 200 km s^-1) are expected to display contaminated CNO abundances (Heger & Langer 2000; Meynet & Maeder 2000). If one is interested in the pristine present-day abundance pattern of these elements, it is therefore sensible to look for and analyse the least evolved objects accessible.

So far, the faintness of unevolved MC B stars ( $m_{V}>15^{\rm m}$ ) has prevented an extensive confrontation of the H II-region CNO data with that from B dwarfs. Four stars have been studied in detail, three by Rolleston et al. (1996) and one by Korn et al. (2000, hereafter Paper I). Of the three stars of Rolleston et al. (1996) one (PS 34-144) is classified as a He-weak star which we will therefore disregard in what follows. The CNO abundances of the remaining three MS stars are confronted with the most recent measurements from H II regions in Table 1: While the consistency among determinations of O abundances is very high, there is a significant positive offset ( $\simeq$ 0.7 dex) of B-star N abundances with respect to the H II-region value. It is a priori unclear which of the two data sets reflects the pristine LMC N abundance. There are many possible explanations for the peculiar behaviour of N: either the H II-region value is systematically too low (note that it is - like all the other H II-region data - exclusively based on 30 Dor and N44C, cf. Garnett 1999), the B-star value systematically too high or we see the imprint of a physical process which alters one of the two data sets. Since dust depletion of nitrogen seems hard to envision (Mathis 1996), it looks as if the B stars are to blame. And indeed, it might be that - due to the limitations imposed on us by the use of 4m-class telescopes up to 1999 - choosing the brightest targets compatible with MS colours resulted in preselecting stars that are preferentially rapid rotators as rapidly rotating stars will evolve to higher luminosities (Fliegner et al. 1996). Whatever the scenario, we regard it as worthwhile to push the limit further towards the zero-age MS to see whether the N abundances already found are representative of LMC B stars in general.

In contrast to the paucity of data on dwarfs, many MC supergiants - from spectral type O to K - have been studied in detail (e.g. Haser et al 1998: O-type, Paper I: B-type, Venn 1999b: A-type, Andrievsky et al. 2001: F-type, Hill & Spite 1999: K-type). Obviously, in trying to understand the CNO pattern of these evolved objects the problem lies in separating all contributing signatures: abundance changes due to dredge-up episodes in the red-giant phase and potentially rotational mixing in earlier phases when the rotation rate was still high. This is particularly true for stars located in the so-called Blue Hertzsprung Gap (between the MS and the region of blue-loop excursions, cf. Fitzpatrick & Garmany 1990), whose evolutionary status is still highly uncertain. As in the MS B stars already discussed, the most significant signature is that of nitrogen: Among SMC A-type supergiants Venn (1999b) found a large scatter in NLTE (non local thermodynamic equilibrium) N abundances ( $6.9\leq\varepsilon{\rm (N)}\leq7.7$ ) which is hard to explain in the framework of the first dredge-up alone and argues in favour of rotational mixing in earlier evolutionary phases as the most likely cause. As far as abundances from other elements with less pronounced signatures are concerned, one has to worry about how severely they might be affected by systematic effects when comparing different spectral types: not in all cases are the assumptions made (plane-parallel geometry, stationarity, LTE) as justified as in the case of dwarfs. For example, the carbon abundance from K supergiants seems to be systematically higher than in the hotter stars. On the whole, the general picture drawn above is, however, supported by all spectral types.

All in all, rotational mixing could help to resolve the discrepancies outlined above: adding rapid rotation, a fraction of any young stellar population will display modified CNO abundances, widen the MS, populate regions of the HRD which are traversed quickly by non-rotating stars and lead to enhanced scatter in N after the first dredge-up by mimicking a distribution of initial abundances.

**Table 2:** Observation log for the programme stars. Nomenclature and visual magnitudes from Robertson (1974). $t_{\rm exp}$ denotes the exposure time, m/y the month and year of the observation. The line-free wavelength ranges 4202-4210, 4402-4409, 4493-4500 and 4680-4695 Å were used to estimate the S/N ratios (1 $\sigma$ ) of the rebinned spectra. Appended to this table are the two luminosity class III stars in NGC2004 from Paper I observed with CASPEC.
object in	m_V	$t_{\rm\,exp}\$	S/N	m/y	$v \sin i$
NGC2004	[mag]	[h]			[km s^-1]
B18	14.8	1.0 + 1.0	140	12/00	130
C8	14.9	1.0	110	11/00	100
C9	15.8	1.4 + 1.4	100	11/00	60
C16	14.7	1.0	130	11/00	60
D3	15.8	1.4 + 1.4	120	12/00	70
D15	15.1	1.5	120	12/00	45
B15	14.2	3.0 + 3.0	70	12/89	25
B30	13.8	$3 \times 2.0$	150	11/87	30

1 Introduction