3D NLTE analysis of the most iron-deficient star, SMSS0313-6708

Abstract

Context. Models of star formation in the early universe require a detailed understanding of accretion, fragmentation and radiative feedback in metal-free molecular clouds. Different simulations predict different initial mass functions of the first stars, ranging from predominantly low-mass (0.1-10 MȮ), to massive (10-100 MȮ), or even supermassive (100-1000 MȮ). The mass distribution of the first stars should lead to unique chemical imprints on the low-mass second and later generation metal-poor stars still in existence. The chemical composition of SMSS0313-6708, which has the lowest abundances of Ca and Fe of any star known, indicates it was enriched by a single massive supernova. Aims. The photospheres of metal-poor stars are relatively transparent in the UV, which may lead to large three-dimensional (3D) effects as well as departures from local thermodynamical equilibrium (LTE), even for weak spectral lines. If 3D effects and departures from LTE (NLTE) are ignored or treated incorrectly, errors in the inferred abundances may significantly bias the inferred properties of the polluting supernovae. We redetermine the chemical composition of SMSS0313-6708by means of the most realistic methods available, and compare the results to predicted supernova yields. Methods. A 3D hydrodynamical Staggermodel atmosphere and 3D NLTE radiative transfer were applied to obtain accurate abundances for Li, Na, Mg, Al, Ca and Fe. The model atoms employ realistic collisional rates, with no calibrated free parameters. Results. We find significantly higher abundances in 3D NLTE than 1D LTE by 0.8 dex for Fe, and 0.5 dex for Mg, Al and Ca, while Li and Na are unaffected to within 0.03 dex. In particular, our upper limit for [Fe/H] is now a factor ten larger, at [Fe/H] < -6.53 (3σ), than previous estimates based on < 3D > NLTE (i.e., using averaged 3D models). This higher estimate is due to a conservative upper limit estimation, updated NLTE data, and 3D-< 3D > NLTE differences, all of which lead to a higher abundance determination. Conclusions. We find that supernova yields for models in a wide range of progenitor masses reproduce the revised chemical composition. In addition to massive progenitors of 20-60 MȮ exploding with low energies (1-2 B, where 1 B = 1051 erg), we also find good fits for progenitors of 10 MȮ, with very low explosion energies (<1 B). We cannot reconcile the new abundances with supernovae or hypernovae with explosion energies above 2.5 B, nor with pair-instability supernovae.