Massive star formation in Wolf-Rayet galaxies IV. Colours, chemical-composition analysis and metallicity-luminosity relations

Aims. We have performed a comprehensive multiwavelength analysis of a sample of 20 starburst galaxies that show a substantial population of very young massive stars, most of them classified as Wolf-Rayet (WR) galaxies. In this paper, the forth of the series, we present the global analysis of the derived photometric and chemical properties. Methods. We compare optical/NIR colours and the physical properties (reddening coefficient, equivalent widths of the emission and underlying absorption lines, ionization degree, electron density, and electron temperature) and chemical properties (oxygen abundances and N/O, S/O, Ne/O, Ar/O, and Fe/O ratios) with previous observations and galaxy evolution models. We compile 41 independent star-forming regions - with oxygen abundances between 12 + log(O/H) = 7.58 and 8.75-, of which 31 have a direct estimate of the electron temperature of the ionized gas. Results. According to their absolute B-magnitude, many of them are not dwarf galaxies, but they should be during their quiescent phase. We found that both c(H beta) and W-abs increase with increasing metallicity. The differences in the N/O ratio is explained assuming differences in the star formation histories. We detected a high N/O ratio in objects showing strong WR features (HCG 31 AC, UM420, IRAS 0828+2816, III Zw 107, ESO 566-8 and NGC 5253). The ejecta of the WR stars may be the origin of the N enrichment in these galaxies. We compared the abundances provided by the direct method with those obtained through empirical calibrations, finding that (i) the Pilyugin method is the best suited empirical calibration for these star-forming galaxies; (ii) the relations provided by Pettini & Pagel (2004, MNRAS, 348, 59) give acceptable results for objects with 12 + log(O/H) > 8.0; and (iii) the results provided by empirical calibrations based on photoionization models are systematically 0.2-0.3 dex higher than the values derived from the direct method. The O and N abundances and the N/O ratios are clearly related to the optical/NIR luminosity; the dispersion of the data is a consequence of the differences in the star-formation histories. The L-Z relations tend to be tighter when using NIR luminosities, which facilitates distinguishing tidal dwarf galaxies candidates and pre-existing dwarf objects. Galaxies with redder colours tend to have higher oxygen and nitrogen abundances. Conclusions. Our detailed analysis is fundamental to understand the nature of galaxies that show strong starbursts, as well as to know their star formation history and the relationships with the environment. This study is complementary -but usually more powerful- to the less detailed analysis of large galaxy samples that are very common nowadays.