Title: UV to submillimetre luminosity functions of TNG50 galaxies

Abstract: We apply the radiative transfer (RT) code SKIRT on a sample of ~14000 low-redshift (z<= 0.1) galaxies extracted from the TNG50 simulation to enable an apples-to-apples comparison with observations. The RT procedure is calibrated via comparison of a subsample of TNG50 galaxies with the DustPedia observational sample: we compare several luminosity and colour scaling relations and spectral energy distributions in different specific SFR bins. We consistently derive galaxy luminosity functions for the TNG50 simulation in 14 broadband filters from UV to submillimetre wavelengths and investigate the effects of the aperture, orientation, radiative transfer recipe, and numerical resolution. We find that, while our TNG50+RT fiducial model agrees well with the observed luminosity functions at the knee (+/- 0.04 dex typical agreement), the TNG50+RT luminosity functions evaluated within 5R_1/2 are generally higher than observed at both the faint and bright ends, by 0.004 (total IR)-0.27 (UKIDSS H) dex and 0.12 (SPIRE250)-0.8 (GALEX FUV) dex, respectively. A change in the aperture does affect the bright end of the luminosity function, easily by up to 1 dex depending on the choice. However, we also find that the galaxy luminosity functions of a worse-resolution run of TNG50 (TNG50-2, with 8 times worse mass resolution than TNG50, similar to TNG100) are in better quantitative agreement with observational constraints. Finally, we publicly release the photometry for the TNG50 sample in 53 broadbands from FUV to submillimetre, in three orientations and four apertures, as well as galaxy spectral energy distributions.