The non-linear infrared-radio correlation of low-z galaxies: implications for redshift evolution, a new radio SFR recipe, and how to minimize selection bias

The infrared-radio correlation (IRRC) underpins many commonly used radio luminosity-star formation rate (SFR) calibrations. In preparation for the new generation of radio surveys we revisit the IRRC of low-$z$ galaxies by (a) drawing on the best currently available IR and 1.4 GHz radio photometry, plus ancillary data over the widest possible area, and (b) carefully assessing potential systematics. We compile a catalogue of $\sim$9,500 z $<$ 0.2 galaxies and derive their 1.4 GHz radio ($L_{\mathrm{1.4}}$), total IR, and monochromatic IR luminosities in up to seven bands, allowing us to parameterize the wavelength-dependence of monochromatic IRRCs from 22-500 $\mu$m. For the first time for low-$z$ samples, we quantify how poorly matched IR and radio survey depths bias measured median IR/radio ratios, $\overline{q}_{\mathrm{TIR}}$, and discuss the level of biasing expected for low-z IRRC studies in ASKAP/MeerKAT fields. For our subset of $\sim$2,000 high-confidence star-forming galaxies we find a median $\overline{q}_{\mathrm{TIR}}$ of 2.54 (scatter: 0.17 dex). We show that $\overline{q}_{\mathrm{TIR}}$ correlates with $L_{\mathrm{1.4}}$, implying a non-linear IRRC with slope 1.11$\pm$0.01. Our new $L_{\mathrm{1.4}}$-SFR calibration, which incorporates this non-linearity, reproduces SFRs from panchromatic SED fits substantially better than previous IRRC-based recipes. Finally, we match the evolutionary slope of recently measured $\overline{q}_{\mathrm{TIR}}$-redshift trends without having to invoke redshift evolution of the IRRC. In this framework, the redshift evolution of $\overline{q}_{\mathrm{TIR}}$ reported at GHz frequencies in the literature is the consequence of a partial, redshift-dependent sampling of a non-linear IRRC obeyed by low-$z$ {\it and} distant galaxies.