Super-Eddington growth of the first black holes

The assembly of the first super massive black holes (SMBHs) at $z \gtrsim 6$ is still a subject of intense debate. If black holes (BHs) grow at their Eddington rate, they must start from $\gtrsim 10^4 \, M_\odot$ seeds formed by the direct collapse of gas. Here we explore the alternative scenario where $\sim 100 \, M_\odot$ BH remnants of the first stars grow at super-Eddington rate via radiatively inefficient slim accretion disks. We use an improved version of the cosmological, data-constrained semi-analytic model GAMETE/QSOdust, where we follow the evolution of nuclear BHs and gas cooling, disk and bulge formation of their host galaxies. Adopting SDSS J1148+5251 (J1148) at $z = 6.4$ as a prototype of luminous $z \gtrsim 6$ quasars, we find that $\sim$ 80% of its SMBH mass is grown by super-Eddington accretion, which can be sustained down to $z \sim 10$ in dense, gas-rich environments. The average BH mass at $z \sim 20$ is $M_{\rm BH} \gtrsim 10^4 \,M_\odot$, comparable to that of direct collapse BHs. At $z = 6.4$ the AGN-driven mass outflow rate is consistent with the observations and the BH-to-bulge mass ratio is compatible with the local scaling relation. However, the stellar mass in the central 2.5 kpc is closer to the value inferred from CO observations. Finally, $\sim 20$ of J1148 progenitors at $z=7.1$ have BH luminosities and masses comparable to ULAS J1120+0641, suggesting that this quasar may be one of the progenitors of J1148.