Raising the observed metallicity floor with a 3D non-LTE analysis of SDSS J102915.14+172927.9 star

Context. The first stars marked the end of the cosmic dark ages, produced the first heavy elements, and set the stage for the formation of the first galaxies. Accurate chemical abundances of ultra metal-poor stars ([Fe/H] < -4) can be used to infer the properties of the first stars and thus the formation mechanism for low-mass second-generation stars in the early Universe. Spectroscopic studies have shown that most second-generation stars are carbon enhanced. A notable exception is SDSS J102915.14+172927.9, which is the most metal-poor star known to date, largely by virtue of the low upper limits of the carbon abundance reported in earlier studies.Aims. We re-analysed the composition of SDSS J102915.14+172927.9 with the aim of providing improved observational constraints on the lowest metallicity possible for low-mass star formation and constraining the properties of its Population III progenitor star.Methods. We developed a tailored three-dimensional model atmosphere for SDSS J102915.14+172927.9 with the Stagger code, making use of an improved surface gravity estimate based on the Gaia DR3 parallax. Snapshots from the model were used as input in the radiative transfer code Balder to compute 3D non-local thermodynamic equilibrium (non-LTE) synthetic spectra. These spectra were then used to infer abundances for Mg, Si, Ca, Fe, and Ni as well as upper limits on Li, Na, and Al. Synthetic 3D LTE spectra were computed with Scate to infer the abundance of Ti and upper limits on C and N.Results. In contrast to earlier works based on 1D non-LTE corrections applied to 3D LTE results, we are able to achieve ionisation balance for Ca I and Ca II when employing our consistent 3D non-LTE treatment. The elemental abundances are systematically higher than those found in earlier works. In particular, [Fe/H] is increased by 0.57 dex, and the upper limits of C and N are larger by 0.90 dex and 1.82 dex, respectively.Conclusions. We find that Population III progenitors with masses 10-20 M-circle dot exploding with energy E < 3 x 10(51) erg can reproduce our 3D non-LTE abundance pattern. Our 3D non-LTE abundances are able to better constrain the progenitor mass and explosion energy as compared to our 1D LTE abundances. Contrary to previous work, we obtain higher upper limits on the carbon abundance that are 'marginally consistent' with star formation through atomic line cooling, and consequently, these results prevent us from drawing strong conclusions about the formation mechanism of this low-mass star.