Runaway collisions in dense clusters may lead to the formation of supermassive black hole (SMBH) seeds, and this process can be further enhanced by accretion, as recent models of SMBH seed formation in Population III star clusters have shown. This may explain the presence of supermassive black holes already at high redshift, $z>6$. However, in this context, mass loss during collisions was not considered and could play an important role for the formation of the SMBH seed. Here, we study the effect of mass loss, due to collisions of protostars, in the formation and evolution of a massive object in a dense primordial cluster. We consider both constant mass loss fractions as well as analytic models based on the stellar structure of the collision components. Our calculations indicate that mass loss can significantly affect the final mass of the possible SMBH seed. Considering a constant mass loss of 5% for every collision, we can lose between 60-80% of the total mass that is obtained if mass loss were not considered. Using instead analytical prescriptions for mass loss, the mass of the final object is reduced by 15-40%, depending on the accretion model for the cluster we study. Altogether, we obtain masses of the order of $10^4M_{\odot}$, which are still massive enough to be SMBH seeds.

To explain the observed population of supermassive black holes at z~7, very massive seed black holes or, alternatively, super-Eddington scenarios are needed to reach final masses of the order of 10^9 solar masses. A popular explanation for massive seeds has been the direct collapse model, which predicts the formation of a single massive object due to the direct collapse of a massive gas cloud. Simulations over the last years have however shown that such a scenario is very difficult to achieve. A realistic model of black hole formation should therefore take fragmentation into account, and consider the interaction between stellar-dynamical and gas-dynamical processes. We present here numerical simulations pursued with the AMUSE code, employing an approximate treatment of the gas. Based on these simulations, we show that very massive black holes of 10^4-10^5 solar masses may form depending on the gas supply and the accretion onto the protostars.