H-II GALAXIES VERSUS PHOTOIONIZATION MODELS FOR EVOLVING STARBURSTS

We have constructed a grid of models representing an H II region produ ced by an evolving starburst embedded in a gas cloud of the same metal licity, The models were produced with the spectral energy distribution from a stellar evolutionary synthesis code as input for a photoioniza tion code that computes the emission-line strengths and equivalent wid ths. Stellar evolution was assumed to proceed according to the models of Maeder. The radiation field was computed using the Kurucz model atm ospheres, supplemented by the expanding non-LTE atmospheres of Schmutz et al. for stellar evolutionary phases with strong winds, making a si gnificant improvement over previous works using classical static, plan e-parallel model atmospheres. Models for stellar interiors and atmosph eres being still in a phase of continuous improvement, our population synthesis models reflect the state of the art in 1995. The models were used to analyze a sample of 100 H II galaxies for which both the H be ta equivalent widths and the [O III] lambda 4363 line intensities were available (the latter allowing a direct determination of the oxygen a bundances based on measured electron temperatures). Because of these s election criteria, the results of our study are restricted to metal-po or objects with metallicities less than about one-half solar. The conf rontation of models with observations is presented in six diagnostic d iagrams involving hydrogen and oxygen lines. Our approach is in many r espects much more constraining for the models than previous studies on H II regions ionized by evolving starbursts. We found that the standa rd starburst model (instantaneous burst of star formation with a Salpe ter initial mass function and an upper cutoff mass of 100 M.) reproduc es the observational constraints provided by the nebular emission line s extremely well if selection effects are taken into account. Models w ith a unique initial mass function are consistent with essentially all observational constraints over a metallicity range from similar to 0. 025 to similar to 0.25 Z.. In contrast, models with a Salpeter-type in itial mass function truncated at 50 M. are not consistent with the obs ervations: they violate the observed distribution of H beta equivalent widths. The mean effective temperature of the ionizing star cluster d eclines from about 50,000 to 40,000 K during the time when the line [O III] lambda 4363 is strong enough to be measurable. Within the framew ork of our models, and in the abundance range where comparisons were m ade with observations, there is no significant evidence for a variatio n of the star cluster mean effective temperature with metallicity, oth er than the one generated by the metallicity-dependent stellar atmosph eric and evolutionary models. A very narrow range in ionization parame ters is required to reproduce the observed line ratios. This should se t limits on the dynamical evolution of giant H II regions. We find a l arge fraction of H II galaxies having [O I] lambda 6300/H beta ratios larger than 0.02. Even models with the lowest ionization parameters co nsidered do not produce these large ratios. An approximate estimate of the mechanical energy released by winds and supernovae during later p hases of the starburst leads to the suggestion that the [O I] lambda 6 300/H beta ratio-in contrast to other line ratios studied-is significa ntly affected by shocks. The small spread in the free parameters neces sary to reproduce the emission-line properties of metal-poor H II gala xies allows us to propose a new indicator of the starburst age: the [O III] lambda 5007 equivalent width is quite robust and can be used up to larger ages than the traditional H beta equivalent width for high s ignal-to-noise spectra. This indicator should also prove useful for lo w signal-to-noise spectra of star-forming galaxies at higher redshift, because of the large value of [O III] lambda 5007/H beta in starburst s younger than 5 Myr.