Modeling the ionizing spectra of H ii regions: individual stars versus stellar ensembles

Aims. We study how IMF sampling affects the ionizing flux and emission line spectra of low mass stellar clusters. Methods. We performed 2 x 10^6 Monte Carlo simulations of zero-age solar-metallicity stellar clusters covering the 20 - 10^6 Mo mass range. We study the distribution of cluster stellar masses, Mclus, ionizing fluxes, Q(H0), and effective temperatures, Tclus. We compute photoionization models that broadly describe the results of the simulations and compare them with photoionization grids. Results. Our main results are: (a) A large number of low mass clusters (80% for Mclus = 100 Mo) are unable to form an H ii region. (b) There are a few overluminous stellar clusters that form H ii regions. These overluminous clusters preserve statistically the mean value of <Q(H0)> obtained by synthesis models, but the mean value cannot be used as a description of particular clusters. (c) The ionizing continuum of clusters with Mclus < 10^4 Mo is more accurately described by an individual star with self-consistent effective temperature(T*) and Q(H0) than by the ensemble of stars (or a cluster Tclus) produced by synthesis models. (d)Photoionization grids of stellar clusters can not be used to derive the global properties of low mass clusters. Conclusions. Although variations in the upper mass limit, mup, of the IMF would reproduce the effects of IMF sampling, we find that an ad hoc law that relates mup to Mclus in the modelling of stellar clusters is useless, since: (a) it does not cover the whole range of possible cases, and (b) the modelling of stellar clusters with an IMF is motivated by the need to derive the global properties of the cluster: however, in clusters affected by sampling effects we have no access to global information of the cluster but only particular information about a few individual stars.