The slope and scatter of the star forming main sequence at z~5 : reconciling observations with simulations

TL;DR

This study probes the star-forming main sequence at z≈4–5 using 316 Hα emitters from JWST ALT behind Abell 2744, extending to M⋆≈10^6–10^10 M⊙. By accounting for the flux-limited selection with a Bayesian framework, the authors derive a SFMS slope of α ≈ 0.59^{+0.10}_{−0.09} and a normalization ms_norm ≈ 1.44 dex at log(M⋆/M⊙)=10.5, with a modest mass-dependent intrinsic scatter that is not decisively constrained. They find that traditional SFR calibrations, non-Gaussian scatter, and especially dust attenuation at high mass can push α toward unity, indicating that systematic uncertainties largely shape the inferred relation. The results reveal tension with simulations predicting α≈1 and highlight the importance of dust, calibration choices, and potential low-SFR tails in shaping the observed SFMS, motivating new observations to disentangle these effects and constrain star-formation histories in the early universe.

Abstract

Galaxies exhibit a tight correlation between their star-formation rate and stellar mass over a wide redshift range known as the star-forming main sequence (SFMS). With JWST, we can now investigate the SFMS at high redshifts down to masses of $\sim10^6$ M$_{\odot}$, using sensitive star-formation rate tracers such as H$α$ emission -- which allow us to probe the variability in star formation histories. We present inferences of the SFMS based on 316 H$α$-selected galaxies at $z\sim4$-$5$ with $\log(\rm M_\star/M_\odot) = 6.4$ -$10.6$. These galaxies were identified behind the Abell 2744 lensing cluster with NIRCam grism spectroscopy from the ``All the Little Things'' (ALT) survey. At face value, our data suggest a shallow slope of the SFMS (SFR $\propto \mathrm{M}_\star^α$, with $α=0.45$). After correcting for the H$α$-flux limited nature of our survey using a Bayesian framework, the slope steepens to $α= 0.59^{+0.10}_{-0.09}$, whereas current data on their own are inconclusive on the mass dependence of the scatter. These slopes differ significantly from the slope of $\approx1$ expected from the observed evolution of the galaxy stellar mass function and from simulations. When fixing the slope to $α=1$, we find evidence for a decreasing intrinsic scatter with stellar mass (from $\approx 0.5$ dex at M$_\star=10^8$ M$_\odot$ to $0.4$ dex at M$_\star=10^{10}$ M$_\odot$). This tension might be explained by a (combination of) luminosity-dependent SFR(H$α$) calibration, a population of (mini)-quenched low-mass galaxies, or underestimated dust attenuation in high-mass galaxies. Future deep observations across facilities can quantify these processes, enabling better insights into the variability of star formation histories.

Figures (13)

• Figure 1: F356W grism spectra for six H$\alpha$ emitters in our sample. On top of each panel we report the stellar mass of the galaxy. The shaded region shows the uncertainty on the flux and the 1.5"$\times$1.5" insets show false-color rest-frame optical RGB images constructed from NIRCam F115W/F200W/F356W. The green and brown lines highlight the H$\alpha$$\lambda6564.6$ and [Nii] $\lambda \lambda 6549.9, 6585.4$ wavelengths, respectively.
• Figure 2: Projected separation between all pairs of galaxies in our sample with $\Delta v < 1000$ km s$^{-1}$. Dashed vertical line at 0.8", which corresponds to $\sim 5.5$ kpc at $z=4.3$, is our choice for the maximum angular separation among components belonging to the same system.
• Figure 3: Observed H$\alpha$ fluxes as a function of redshift. In gray is the parent H$\alpha$ sample (N = 512), while in pink we highlighted the robust sample (N = 316, observed H$\alpha$ flux $> 10^{-18}$ erg s$^{-1}$ cm$^{-2}$ and $\mu \leq 2.5$). Red hexagons highlight the 6 confirmed Broad Line H$\alpha$ emitters (BLH$\alpha$), with $v_{\rm FWHM} > 1000$ km/s, presented in Matthee+2024_env. The histograms show the redshift and H$\alpha$ flux distributions for both the parent and robust samples.
• Figure 4: The probed parameters space in the SFR-M$_\star$ plane. The parameter space probed by ALT (pink) compared to that of the CONGRESS survey (green). Both surveys cover H$\alpha$ emission in the redshift range $z=3.7-5.1$. CONGRESS Egami+2023jwst.propCovelo-Paz+2025 employs F356W grism in the GOODS-North field and captures galaxies with a median stellar masses higher than those from ALT, while ALT extends the parameter space approximately 2 orders of magnitude lower in stellar mass.
• Figure 5: Obscured star formation as a function of stellar mass. The fraction of dust attenuated H$\alpha$ emission, i.e. the ratio between the obscured and total H$\alpha$ luminosity, as a function of stellar mass for the ALT sample. In gray is the parent H$\alpha$ sample, while highlighted in pink is the robust one. Median values in each mass bin are shown as filled circles. Empty markers show obscured H$\alpha$ fraction once we account for enhanced dust attenuation in high-mass galaxies (see Sect. \ref{['sec:test_dustatt']}). As a comparison, we show the fraction of obscured star formation for galaxies at $2<z<2.5$ from Whitaker+2017 (blue), at $0.7 < z < 2$ from Shivaei+2024 (green) and at $z\sim 4.5$ ALPINE galaxies from Fudamoto+2020_alpine (orange), with squares showing individual FIR continuum detections at $4<z<5$ and triangles the 3$\sigma$ upper limits for IR non detections. Filled squares show stacks in 2 mass bins at $z \sim 4.5$.