Final Activity Report Summary - STARFORM (Sequential Self-Propagating Star Formation) The aim of this project was to study sequential self-propagating star formation. In particular, we wanted to test the so-called collect-and-collapse model by means of numerical simulations running on large computing facilities. The collect-and-collapse model is based on the well-known fact that massive stars, i.e. stars between ten and a hundred times more massive than the Sun, inject a tremendous amount of energy into the ambient interstellar gas in the form of: 1. ionising ultra-violet radiation 2. powerful stellar winds and 3. energy released when they finally explode as supernovae. This typically leads to the formation of a bubble of hot gas, which expands with supersonic velocity and sweeps up the surrounding interstellar gas into a dense thin shell. Once this shell becomes sufficiently massive, it becomes gravitationally unstable and breaks up into fragments which then collapse to form new stars. It was important to understand the fragmentation process in order to identify the conditions under which the formed fragments were large. This was because large fragments were likely to condense into massive new stars, and, in case this happened, the entire scenario repeated itself. In this way the star formation propagated as an infection until it consumed the entire molecular cloud. We simulated for the first time the gravitational instability of an expanding shell in three dimensions, using an adaptive mesh refinement hydrodynamic (AMR) code which took self-gravity into account. We used the publicly available AMR code Flash, augmented by an algorithm for computing the gravitational field which we developed as part of this project. This algorithm was a key element of the code, because the gravitational field had to be recomputed at each time step as the simulation evolved, and this was very compute-intensive. We run the code on the new MERLIN super-computer of the Cardiff University advanced research computing at Cardiff (ARCCA) service. We found that the pattern of fragmentation depended strongly on the pressure in the ambient interstellar medium which confined the shell. If this pressure was low fragmentation was slow and resulted in large fragments which would spawn large stars, hence triggering another generation of star formation. On the other hand, if the confining pressure was high, it led to fast fragmentation and produced many small fragments which would spawn low-mass stars, incapable of triggering further star formation. This result was not predicted by previous analytical work, which was based on the approximation of an infinitesimally thin shell. We also compared our results with simulations by our collaborators. These simulations were based on a completely different numerical technique, called smoothed particle hydrodynamics, but produced almost identical results, and this gave us confidence that the utilised codes were very reliable. In addition to the collect-and-collapse model, we discovered and simulated a very different mechanism which triggered second-generation star formation. In this mechanism, the second generation of star formation occurred inside the earlier generation, instead of its periphery. This was because, in an extremely dense cluster of stars like the super star clusters observed in starburst galaxies, the gas in the stellar winds and supernova ejecta from the first generation of stars might become thermally unstable, due to strong radiative emission and the resultant rapid cooling. This caused the gas to collapse into new stars. The interesting aspect of this model was that it was able to explain the fact that, often, several generations of stars were observed in globular clusters. This had been a long-standing puzzle in astronomy. We developed a two-dimensional model of super star cluster winds and used it to quantify the fraction of mass ejected by massive stars which remained inside the cluster and was therefore available for secondary star formation. We finally developed a code which calculated spectral line profiles of emission from the gas in simulated super star clusters and found that these profiles were similar to the ones observed in super star clusters.