Nebular dominated galaxies: insights into the stellar initial mass function at high redshift

TL;DR

This study analyzes a z=5.943 galaxy (GS-NDG-9422) with a strong Balmer jump and a notable UV turnover, using JWST/NIRSpec and NIRCam data to dissect its nebular conditions and continuum. Emission lines are well reproduced by a young, metal-poor stellar population, with AGN activity unlikely, and the Balmer jump confirms a significant nebular continuum contribution. The UV turnover, unlikely to arise from a DLA, is best explained by a dominant two-photon nebular continuum powered by an extremely hot ionizing spectrum, implying a top-heavy IMF or exotic hot-star populations. These findings suggest that some high-redshift galaxies may form disproportionately many very massive stars, influencing the ionizing photon budget and early galaxy evolution, and motivate the search for additional similar objects to test IMF variations in the early universe.

Abstract

We identify a low-metallicity ($12+\log({\rm O}/{\rm H})=7.59$) Ly$α$-emitting galaxy at $z=5.943$ with evidence of a strong Balmer jump, arising from nebular continuum. While Balmer jumps are sometimes observed in low-redshift star-forming galaxies, this galaxy also exhibits a steep turnover in the UV continuum. Such turnovers are typically attributed to absorption by a damped Ly$α$ system (DLA); however, the shape of the turnover and the high observed Ly$α$ escape fraction ($f_{\rm esc,Lyα}~\sim27\%$) is also consistent with strong nebular two-photon continuum emission. Modelling the UV turnover with a DLA requires extreme column densities ($N_{\rm HI}>10^{23}$ cm$^{-2}$), and simultaneously explaining the high $f_{\rm esc,Lyα}$ requires a fine-tuned geometry. In contrast, modelling the spectrum as primarily nebular provides a good fit to both the continuum and emission lines, motivating scenarios in which (a) we are observing only nebular emission or (b) the ionizing source is powering extreme nebular emission that outshines the stellar emission. The nebular-only scenario could arise if the ionising source has `turned off' more recently than the recombination timescale ($\sim$1,000 yr), hence we may be catching the object at a very specific time. Alternatively, hot stars with $T_{\rm eff}\gtrsim10^5$ K (e.g. Wolf-Rayet or low-metallicity massive stars) produce enough ionizing photons such that the two-photon emission becomes visible. While several stellar SEDs from the literature fit the observed spectrum well, the hot-star scenario requires that the number of $\gtrsim50~{\rm M}_\odot$ stars relative to $\sim5-50~{\rm M}_\odot$ stars is significantly higher than predicted by typical stellar initial mass functions (IMFs). The identification of more galaxies with similar spectra may provide evidence for a top-heavy IMF at high redshift.

Figures (17)

• Figure 1: Top: 2D Prism/CLEAR spectrum of GS-NDG-9422. Middle: 1D Prism/CLEAR spectrum of GS-NDG-9422 shown in $f_\nu$ (upper middle) and $f_\lambda$ (lower middle). Blue squares show 0.15" radius aperture photometry from JWST/NIRCam medium- and broad-band imaging, while orange squares show the predicted photometry obtained by convolving the Prism spectrum with the NIRCam filter transmission profiles. Bottom left: Three-colour image ($F090W$, $F200W$, $F444W$) of GS-NDG-9422 showing the positioning of the three NIRSpec micro-shutters across the three nod positions. Bottom centre: Zoom-in of the region surrounding [O iii] $\lambda$5007 in the G395M spectrum. Bottom right: Zoom-in of the region surrounding H$\alpha$ in the G395H spectrum.
• Figure 2: GS-NDG-9422 plotted onto several line ratio diagnostic diagrams. Light blue points show model predictions for star-forming regions from Gutkin2016, while pink points show AGN model predictions from Feltre2016 across a large range of metallicities. Navy (star-forming) and red (AGN) points show the subsets of these model grids that have $Z=0.001$ ($Z\approx 0.07 Z_\odot$; $12+\log ({\rm O}/{\rm H})\approx7.54$) which is the closest grid value to that measured for GS-NDG-9422 (see Section \ref{['sub:chem_abun']}). Top left: The strong upper limit on [S ii] $\lambda\lambda$6716, 6731 positions GS-NDG-9422 well below the theoretical maximum starburst limit from Kewley2001 (black line). Top right: The weak detection of He ii$\lambda$4686 is also below the maximum starburst limit from ShiraziBrinchmann2012 (black line). Strong He ii-emitting star-forming galaxies from ShiraziBrinchmann2012 and Senchyna2017 are also shown as orange and green diamonds respectively. We also show measurements from Hanny's Voorwerp, suggested to be a quasar light echo (Lintott2009; purple line-connected points), which we discuss in Section \ref{['sub:geometry']}. Points of increasing size indicate larger offset from the quasar ranging from 13 to 31 kpc. Bottom left: GS-NDG-9422 lies beyond the AGN region defined by Mingozzi2024 (black lines), and is more coincident with the star-forming models than AGN models. He ii-emitting star-forming galaxies at $z\sim2.5-4$ from Saxena2020_HeII are shown for comparison as orange squares. Bottom right: GS-NDG-9422 lies at the tip of the parameter space covered by AGN models, while it is well within the bounds of that covered by star-forming models.
• Figure 3: Observed Balmer decrements compared to theoretical values at different temperatures and densities. Green squares give the measured ratios from the Prism, while diamonds give ratios measured from gratings, where available. Solid black lines give the theoretical ratio for case B recombination at $T_e=10^4$ K and $n_e=100$ cm$^{-3}$. Coloured points show theoretical ratios for a range of temperatures ($10^4\leq T_e \leq 2.5 \times 10^4$ K; colour) and densities ($n_e=10$, 100, 1000 cm$^{-3}$; marker size). The measured H$\alpha$/H$\beta$ is somewhat lower than the theoretical value, but we note that H$\alpha$ may be underestimated in the prism spectrum (Table \ref{['tab:line_EWs']}).
• Figure 4: Prism spectrum of GS-NDG-9422 (black) compared with the best fit models using a standard SSP (magenta) accounting for both the $z=6$ IGM opacity (dashed yellow) and a DLA with column density of $10^{23.1}\ {\rm cm^{-2}}$ (dotted cyan).
• Figure 5: Top: Zoom-in on UV turnover for models plotted in Figure \ref{['fig:best_fit_bpass_full']}. Middle: Fitting to GS-NDG-9422 using a DLA model with a 70 % covering fraction allows Ly$\alpha$ escape, but implies an extremely high column density. Bottom: Comparison of implied column density with known DLAs.