Self-consistent population spectral synthesis with FADO

Context. The field of galaxy evolution will make a great leap forward in the next decade as consequence huge effort by scientific community multi-object spectroscopic facilities. Various future surveys enormously increase number available spectra, providing new insights into unexplored areas research. To maximise impact such incoming data, analysis methods must also step up, extracting reliable information from spectra. It is therefore urgent to refine and test tools that are able infer properties medium- or high-resolution Aims. In this paper we aim investigate limits reliability different spectral synthesis estimation mean stellar age metallicity. These two quantities fundamental determine assembly history key its star formation history. main question work aims address which signal-to-noise ratios (S/N) needed reliably metallicity spectrum how depends on tool used model Methods. built set realistic simulated spectra containing nebular emission, reproducing limiting cases: constant rate an exponentially declining with single initial burst. We degraded synthetic these histories (SFHs) S/N analysed three widely codes, namely F ADO , S TECKMAP TARLIGHT assuming similar fitting set-ups same base templates. Results. For / N ≤ 5 all show large diversity results. find median differences light-weighted ∼0.1 dex, while shows higher value ∼0.2 dex. > 50 ∼0.03 dex (∼7%), factor 3 4 lower than 0.08 (∼20%) 0.11 (∼30%) obtained respectively. Detailed investigations best-fit for galaxies overestimated mass-weighted point towards inability purely models fit observed energy distribution around Balmer jump. Conclusions. Our results imply when enters phase high specific (sSFR) neglect continuum emission process has strong SFH codes used, even presence order 7% (F ), 20% (S 30% ) quantities, 60% quantities. More specifically, continuous both overestimate 2 within first ∼100 Myr This bias, stems obviously implies severe overestimation mass-to-light ratio mass. But mild contribution continuum, there still possibility misinterpret data poor quality observations. underlines once more importance self-consistent treatment implemented which, according our analysis, only viable route determination any high-sSFR at low redshift.