Recent high-resolution interferometric images of submillimetre galaxies (SMGs) reveal fascinatingly complex morphologies. This raises a number of questions: how does the relative orientation of a galaxy affect its observed submillimetre emission, and does this result in an `orientation bias' in the selection and analysis of such galaxies in flux-limited cosmological surveys? We investigated these questions using the \textsc{Simba} cosmological simulation paired with the dust radiative transfer code \textsc{Powderday}. We selected eight simulated SMGs ($S_{850}\gtrsim2$ mJy) at $z = 2$, and measured the variance of their `observed' emission over 50 random orientations. Each galaxy exhibits significant scatter in its emission close to the peak of the thermal dust emission, with variation in flux density of up to a factor of 2.7. This results in an appreciable dispersion in the inferred dust temperatures and infrared luminosities ($16^{\mathrm{th}}-84^{\mathrm{th}}$ percentile ranges of 5\,K and 0.1\,dex, respectively) and therefore a fundamental uncertainty in derived parameters such as dust mass and star formation rate ($\sim$30% for the latter using simple calibrations). Using a Monte Carlo simulation we also assessed the impact of orientation on flux-limited surveys, finding a bias in the selection of SMGs towards those with face--on orientations, as well as those at lower redshifts. We predict that the orientation bias will affect flux-limited single-dish surveys, most significantly at THz frequencies, and this bias should be taken into account when placing the results of targeted follow--up studies in a statistical context.

The dust-to-stellar mass ratio ($M_{\rm dust}$/$M_{\rm \star}$) is a crucial yet poorly constrained quantity to understand the production mechanisms of dust, metals and stars in galaxy evolution. In this work we explore and interpret the nature of $M_{\rm dust}$/$M_{\rm \star}$ in 300 massive ($M_{\star}>10^{10}M_{\odot}$), dusty star-forming galaxies detected with ALMA up to $z\approx5$. We find that $M_{\rm dust}$/$M_{\rm \star}$ evolves with redshift, stellar mass, specific SFR and integrated dust size, differently for main sequence and starburst galaxies. In both galaxy populations $M_{\rm dust}$/$M_{\rm \star}$ rises until $z\sim2$ followed by a roughly flat trend towards higher redshifts. We show that the inverse relation between $M_{\rm dust}$/$M_{\rm \star}$ and $M_{\star}$ holds up to $z\approx5$ and can be interpreted as an evolutionary transition from early to late starburst phases. We demonstrate that $M_{\rm dust}$/$M_{\rm \star}$ in starbursts mirrors the increase in molecular gas fraction with redshift, and is enhanced in objects with the most compact dusty star-formation. The state-of-the-art cosmological simulation SIMBA broadly matches the evolution of $M_{\rm dust}$/$M_{\rm \star}$ in main sequence galaxies, but underestimates it in starbursts. The latter is found to be linked to lower gas-phase metallicities and longer dust growth timescales relative to data. Our data are well reproduced by analytical model that includes recipes for rapid metal enrichment, strongly suggesting that high $M_{\rm dust}$/$M_{\rm \star}$ is due to fast grain growth in metal enriched ISM. Our work highlights multifold benefits of using $M_{\rm dust}$/$M_{\rm \star}$ as a diagnostic tool for: (1) separating main sequence and starburst galaxies until $z\sim5$; (2) probing the evolutionary phases of dusty galaxies, and (3) refining the treatment of dust life cycle in simulations.