We study the origin of the diffuse stellar component (DSC) in 117 galaxy clusters extracted from a cosmological hydrodynamical simulation. We identify all galaxies present in the simulated clusters at 17 output redshifts, starting with z = 3.5, and then build the family trees for all the z = 0 cluster galaxies. The most massive cluster galaxies show complex family trees, resembling the merger trees of dark matter haloes, while the majority of other cluster galaxies experience only one or two major mergers during their entire life history. Then, for each diffuse star particle identified at z = 0, we look for the galaxy to which it once belonged at an earlier redshift, thus linking the presence of the DSC to the galaxy formation history.The main results of our analysis are as follows. (i) On average, half of the DSC star particles come from galaxies associated with the family tree of the most massive galaxy (bright cluster galaxy -hereafter BCG), one quarter comes from the family trees of other massive galaxies and the remaining quarter from dissolved galaxies. That is, the formation of the DSC is parallel to the build-up of the BCG and other massive galaxies. (ii) Most DSC star particles become unbound during mergers in the formation history of the BCGs and of other massive galaxies, independent of cluster mass. Our results suggest that the tidal stripping mechanism is responsible only for a minor fraction of the DSC. (iii) At cluster radii larger than 250 h −1 kpc, the DSC fraction from the BCG is reduced and the largest contribution comes from the other massive galaxies; in the cluster outskirts, galaxies of all masses contribute to the DSC. (iv) The DSC does not have a preferred redshift of formation: however, most DSC stars are unbound at z < 1. (v) The amount of DSC stars at z = 0 does not correlate strongly with the global dynamical history of clusters, and increases weakly with cluster mass.

We present a new multiphase subresolution model for star formation and feedback in smoothed particle hydrodynamics (SPH) numerical simulations of galaxy formation. Our model, called MUPPI (MUlti-Phase Particle Integrator), describes each gas particle as a multiphase system, with cold and hot gas phases, coexisting in pressure equilibrium and a stellar component. Cooling of the hot tenuous gas phase feeds the cold gas phase. We compute the cold gas molecular fraction using the phenomenological relation of Blitz & Rosolowsky between this fraction and the external disc pressure, which we identify with the SPH pressure. Stars are formed out of molecular gas with a given efficiency, which scales with the dynamical time of the cold phase. Our prescription for star formation is not based on imposing the Schmidt-Kennicutt relation, which is instead naturally produced by MUPPI. Energy from supernova explosions is deposited partly into the hot phase of the gas particles, and partly to that of neighbouring particles. Mass and energy flows among the different phases of each particle are described by a set of ordinary differential equations which we explicitly integrate for each gas particle, instead of relying on equilibrium solutions. This system of equations also includes the response of the multiphase structure to energy changes associated to the thermodynamics of the gas. Our model has an intrinsically runaway behaviour: energy from supernovae increases gas pressure which increases in turn the star formation rate through the molecular fraction. This runaway is stabilized in simulations by the hydrodynamic response of the gas: when it receives enough energy, it expands thereby decreasing its pressure. We apply our model to two isolated disc galaxy simulations and two spherical cooling flows. MUPPI is able to reproduce the Schmidt-Kennicutt relation for disc galaxies. It also reproduces the basic properties of the interstellar medium in disc galaxies, the surface densities of cold and molecular gas, of stars and of star formation rate, the vertical velocity dispersion of cold clouds and the flows connected to the galactic fountains. Quite remarkably, MUPPI also provides efficient stellar feedback without the need to include a scheme of kinetic energy feedback.