Probing galaxy evolution from $z=0$ to $z\simeq10$ through galaxy scaling relations in three L-Galaxies flavours

We present a comprehensive examination of the three latest versions of the L-Galaxies semi-analytic galaxy formation model, focusing on the evolution of galaxy properties across a broad stellar mass range ($10^7\:{\rm M}_{\odot}\lesssim{M_\star}\lesssim10^{12}\:{\rm M}_{\odot}$) from $z=0$ to $z\simeq10$. This study is the first to compare predictions of L-Galaxies with high-redshift observations well outside the original calibration regime, utilising multiband data from surveys such as SDSS, CANDELS, COSMOS, HST, JWST, and ALMA. We assess the models' ability to reproduce various time-dependent galaxy scaling relations for star-forming and quenched galaxies. Key focus areas include global galaxy properties such as stellar mass functions, cosmic star formation rate density, and the evolution of the main sequence of star-forming galaxies. Additionally, we examine resolved morphological properties such as the galaxy mass-size relation, alongside core $(R<1\,{\rm{kpc}})$ and effective $(R<R_{\rm{e}})$ stellar-mass surface densities as a function of stellar mass. This analysis reveals that the L-Galaxies models are in qualitatively good agreement with observed global scaling relations up to $z\simeq10$. However, significant discrepancies exist at both low and high redshifts in accurately reproducing the number density, size, and surface density evolution of quenched galaxies. These issues are most pronounced for massive central galaxies, where the simulations underpredict the abundance of quenched systems at $z\geq1.5$, reaching a discrepancy of a factor of 60 by $z\approx3$, with sizes several times larger than observed. Therefore, we propose that the physical prescriptions governing galaxy quenching, such as AGN feedback and processes related to merging, require improvement to be more consistent with observational data.