Abstract
The luminosity function (LF) of young star clusters is often approximated by a power law function. For clusters in a wide range of galactic environments this has resulted in fit indices near -2, but on average slightly steeper. A fundamental property of the -2 power law function is that the luminosity of the brightest object (L_max) scales linearly with the total number of clusters, which is close to what is observed. This suggests that the formation of Young Massive Clusters (YMCs) is a result of the size of the sample, i.e. when the SFR is high it is statistically more likely to form YMCs, but no particular physical conditions are required. In this contribution we provide evidence that the LF of young clusters is not a -2 power law, but instead is curved, showing a systematic decrease of the (logarithmic) slope from roughly -1.8 at low luminosities to roughly -2.8 at high luminosities. The empirical LFs can be reproduced by model LFs using an underlying cluster IMF with a Schechter type truncation around M*=2x10^5 M_sun. This value of M* can not be universal since YMCs well in excess of this M* are known in merging galaxies and merger remnants. Therefore, forming super massive clusters (>10^6 M_sun) probably requires conditions different from those in (quiescent) spiral galaxies and hence is not only the result of a size-of-sample effect. From the vertical offset a cluster formation efficiency of ~10% is derived. We find indications for this efficiency to be higher when the SFR is higher.