The dawn of disks: unveiling the turbulent ionised gas kinematics of the galaxy population at $z\sim4-6$ with JWST/NIRCam grism spectroscopy

TL;DR

This study uses JWST/NIRCam slitless spectroscopy to assemble a statistically robust sample of 272 Hα emitters at z ≈ 3.9–6.5, enabling spatially resolved ionised-gas kinematics across a wide stellar-mass and SFR range. A novel forward-modelling framework (geko) combines grism data with NIRCam imaging and Prospector-derived stellar populations to jointly infer morphology, rotation, and velocity dispersion, yielding σ0 ≈ 100 km/s and v/σ0 ≈ 1–2 on average, with large intrinsic scatter. The results show a mild redshift evolution toward higher turbulence and lower rotational support at earlier times, a strong σ0–SFR/ΣSFR coupling, and only a modest increase in the fraction of rotationally supported systems with cosmic time, implying that disks do not dominate the z > 4 population. These findings support a picture where galaxies experience a turbulent, rapidly evolving phase early on, with disks gradually assembling and stabilizing by cosmic noon, and they highlight the power of grism-based population kinematics to inform galaxy evolution models. The work also emphasizes the need for multi-wavelength follow-up (e.g., JWST/NIRSpec, ALMA) to fully map 2D kinematics and disentangle turbulence drivers such as gravitational instabilities, feedback, and mergers.

Abstract

Recent studies of gas kinematics at high redshift have reported disky systems which appear to challenge models of galaxy formation, but it is unclear whether they are representative of the underlying galaxy population. We present the first statistical sample of spatially resolved ionised gas kinematics at high redshift, comprised of $272$ H$α$ emitters in GOODS-S and GOODS-N at redshifts $z\approx3.9-6.5$, observed with JWST/NIRCam slitless spectroscopy and imaging from JADES, FRESCO and CONGRESS. The sample probes two orders of magnitude in stellar mass ($\log (M_{\star}[\mathrm{M}_{\odot}])\approx8-10$) and star formation rate ($\text{SFR}\approx0.3-100\thinspace M_{\odot}/$yr), and is representative down to $\log(M_{\star}[\mathrm{M}_{\odot}])\approx 9$. Using a novel inference tool, $\texttt{geko}$, we model the grism data to measure morphological and kinematic properties of the ionised gas, as probed by H$α$. Our results are consistent with a decrease of the rotational support $v/σ_0$\ and increase of the velocity dispersion $σ_0$ with redshift, with $σ_0\approx100$ km/s and $v/σ_0\approx1-2$ at $z\approx3.9-6.5$. We study the relations between $σ_0$, and $v/σ_0$, and different star formation tracers and find a large scatter and diversity, with the strongest correlations between $σ_0$ and SFR and SFR surface density. The fraction of rotationally supported systems ($v/σ_0>1$) slightly increases with cosmic time, from $(36\pm6)\%$ to $(41\pm6)\%$ from $z\sim 5.5$ to $z\sim 4.5$, for galaxies with masses $9<\log(M_{\star}[\mathrm{M}_{\odot}])<10$. Overall, disks do not dominate the turbulent high-redshift galaxy population in the mass range probed by this work. When placed in the context of studies up to cosmic noon, our results are consistent with a significant increase of disk-like systems with cosmic time.

Figures (62)

• Figure 1: Selection of our sample in the plane of H$\alpha$ S/N and $|\text{PA}_{\text{morph}}|$, colour-coded by stellar mass. The discarded region highlights the parameter space in which galaxies are discarded due to low S/N and/or a PA that is too close to the dispersion direction ($\rm PA = 90^{\circ}$). The $\mathrm{S/N}=20$ and $|\text{PA}_{\text{morph}}| = 60^{\circ}$ cut-off for the gold sample is highlighted with the red dashed line. The small dots represent the unresolved sample of galaxies that have a spatial resolution below the FWHM of the F444W PSF and/or for which we do not measure a resolved velocity gradient. Based on the selection cuts described in Sec. \ref{['sec:sample-selection']}, we obtain 41 galaxies in the gold sample, 132 in the silver sample, and 99 in the unresolved sample (see Tab. \ref{['tab:sample-selection']}).
• Figure 2: Star-formation rates (SFR$_{10}$) and stellar masses ($M_{\star}$) of our sample, derived with Prospector, colour-coded by the spectroscopic redshifts. The SFRs are averaged over 10 Myr. We compare our sample to SFMS prescriptions from McClymont:2025aa and Simmonds et al (in prep.). Our sample is representative of the star-forming galaxy population at $M_{\star}>10^{9}~M_{\odot}$ (indicated by the vertical dashed line). Below this stellar mass, because our sample selection is based on S/N in H$\alpha$ (Fig. \ref{['fig:sample-selection']}), it is biased toward high SFRs relative to the SFMS. The discarded merger sample (green stars) is evenly spread across the parameter space and discarding it does not bias our results.
• Figure 3: Distribution of our sample in the stellar mass ($M_{\star}$)-redshift plane. Our gold sample spans a wide range in stellar mass, but lies preferentially at $z<5$. There are overall fewer galaxies at $z>5$, with only few low mass systems ($M_{\star}\lesssim10^{9}~M_{\odot}$) falling in the gold and silver samples. Overall, our sample probes the high mass end of photometric candidates, as shown by the green contours representative of the JADES sample from Simmonds:2024ab, which is expected from our S/N cut. We colour-code galaxies in the gold and silver samples by their effective radius $r_{\rm e}$ in the rest-frame near-UV, highlighting that although more massive galaxies are typically larger, at fixed stellar mass, galaxies span a wide range of sizes.
• Figure 4: Example science products obtained with geko for a galaxy in our gold sample. In the top panels, we show the observed grism data, the best-fit model, and the corresponding residuals computed with the $\chi$ metric. The plotting range is $\chi = [-5, - 5]$. We also highlight the best fit effective diameter $D_e = 2r_{\rm e}$ for the H$\alpha$ emission, and the (projected) diameter of the $\text{S/N}>3$ observed map. In the bottom panel, we show the derived best-fit velocity field, velocity dispersion field (which has the same velocity scale as the velocity field plot), and the intrinsic H$\alpha$ emission map. In all of the panels on the left side, we show the scale in arcseconds (black) along with the velocity centroid (black cross). In each row, we show the physical meaning of the axes (spatial vs dispersion).
• Figure 5: Prior (pink) and posterior (blue) distributions inferred with geko for the same gold sample galaxy as in Fig. \ref{['fig:geko-summary']}, for the free parameters of our model (Tab. \ref{['tab:priors']}). From these, we derive the posterior distributions for the rotational velocity at $v_{\rm re}$ and the rotational support $\text{v}/\sigma_0$, whose best fit values and uncertainties, computed from the $16^{\rm th}$ and $84^{\rm th}$ quantiles, are shown on the top right corner.