A constant upper luminosity limit of cool supergiant stars down to the extremely low metallicity of I Zw 18

TL;DR

The paper demonstrates that the upper luminosity limit of cool supergiants is effectively independent of metallicity down to $Z/Z_\odot \approx 1/40$, with a consistent ceiling around $\log (L/L_\odot) \approx 5.6$ across a diverse galaxy sample including I Zw 18. By combining HST and JWST photometry, neural-network or track-based identifications of cool SGs, and bolometric-correction-based luminosities, the authors compare observed distributions to BoOST evolutionary models and find an underprediction of very bright cool SGs at low $Z$, implying metallicity-independent late-stage mass loss. They discuss implications for He$^+$-ionizing emission, nitrogen enrichment at high redshift, and the possible wind-stripping pathway that could produce hot, He-rich stars with weak winds in extremely metal-poor environments. The results constrain massive-star evolution and feedback in the early universe, suggesting that hydrogen-rich layers are lost prior to collapse even at the lowest metallicities studied, thereby affecting the ionizing output and the remnants left behind.

Abstract

Stellar wind mass loss is often assumed to depend on their metallicity $Z$. Therefore, evolutionary models of massive stars at lower $Z$ are able to retain more of their H-rich layers and evolve into brighter cool supergiants (cool SGs; $T_\mathrm{eff} < 7$ kK). Surprisingly, in galaxies in the range $0.2 \lesssim Z / Z_\odot \lesssim 1.5$ previous studies did not find a $Z$-dependence of the upper luminosity limit $L_\mathrm{max}$ of cool SGs. Here, we study four extra galaxies with HST and JWST. Observations of the extremely low-$Z$ dwarf galaxy I Zw 18 from JWST allow us to go down to $Z / Z_\odot \approx 1/40$. For cool SGs in all studied galaxies including I Zw 18, we find a constant $L_\mathrm{max} \approx 10^{5.6}$L$_\odot$, similar to literature results for $0.2 \lesssim Z / Z_\odot \lesssim 1.5$. In I Zw 18 and the other studied galaxies, the presence of Wolf-Rayet stars has been claimed. Although we cannot rule out that some of them become intermediate-temperature objects, this paints a picture in which evolved stars with $L>10^{5.6}$L$_\odot$ burn He as hot He-rich stars down to extremely low $Z$. We argue that $Z$-independent late-phase mass loss would be the most likely mechanism responsible. Regardless of the exact stripping mechanism, for the Early Universe our results imply a limitation on black hole masses and a contribution of stars born with $M\gtrsim 30$M$_\odot$ to its surprisingly strong nitrogen enrichment. We propose a scenario in which single stars at low $Z$ emit sufficiently hard ionizing radiation to produce lines of He II and C IV. In this scenario, late-phase $Z$-independent mass loss produces hot He-rich stars. Due to the well-understood $Z$-dependence of radiation-driven winds of hot stars, a window of opportunity would open below 0.2Z$_\odot$, where self-stripped He-rich stars can exist without dense WR winds that absorb hard ionizing radiation.

Figures (7)

• Figure 1: Top panels: extinction-corrected color-magnitude diagrams of SMC sources, based on IR data (left) and optical 'VIS' data (right). Orange lines show the RSG cuts in the IR. Red markers indicate sources that we classified as cool SGs, and blue markers indicate other sources. The four grey lines are iso-luminosity lines (Eq. \ref{['eq:iso_l']}). These iso-luminosity lines represent $\log ( L / L_\odot ) = 6.0$, 5.5, 5.0, and 4.5, where the higher values are closer towards the top of the plots. At the top of each panel, gray ticks at various colors are labeled by the temperatures to which they correspond according to the adopted bolometric corrections. In the top left panel, following Nally24, the arrows parallel to various features indicate the main sequence (MS), red giant branch (RGB), oxygen (O) and carbon (C) AGB, and the RSGs. Bottom panels: luminosities calculated in this work (both in the visual as in the infrared) as a function of luminosities calculated in earlier work by Davies18. The dashed black line indicates where both luminosities would be identical.
• Figure 2: Same as the top panels of Fig. \ref{['fig:meth_test']}, but for VIS data of NGC 4395, NGC 5253, and NGC 300 (top panels) and IR and VIS data of I Zw 18 (bottom panels). Also, the purple line in the bottom panels represents an 8 M$_\odot$ MIST track with [Fe/H] = -1.75.)
• Figure 3: Luminosity distributions of observed sources (red lines) and sources in the theoretical population (light blue lines) in six different galaxies. The observed SMC and LMC sources are taken from Davies18, the rest is from this work. At the right side of each panel, we write the number of observed and model population sources with $\log (L / L_\odot) > 5.6$, and the associated Poisson probability to have as many observed sources or fewer (given the number of theoretically predicted sources).
• Figure 4: Third brightest cool SG in various galaxies as a function of galaxy metallicity. The data points with the labels HST, JWST, SMC IR, and SMC VIS are from this work. The data points labeled Davies+18 and McDon+22 are based on the work of Davies18 on the SMC and LMC, and the work of McDonald22 on M 31, respectively. We note that I Zw 18 and the SMC have multiple data points that are similar within 0.02 dex. The light blue horizontal lines show the third brightest cool SG in individual simulations. We perform 20 simulations per galaxy, and we visualize the number density of the outcomes of the simulation with violin plots. The grey ticks indicate the time-averaged cool SG luminosities of BoOST SMC models with three different initial masses, which are similar across the shown metallicity range.
• Figure 5: Similar to the bottom left panel of Fig. \ref{['fig:cmds_all']}, but only showing infrared data of I Zw 18, and zoomed in on the bluer side of the CMD. Here we show iso-luminosity lines that extend to bluer colors. The temperature ranges associated with the color intervals bordered by the black vertical lines are written in the plot. The green hatching highlights the empty region associated with $T_\mathrm{eff} < 20$ kK and $\log ( L / L_\odot ) > 5.6$. The red ticks indicate the absolute $F200W$ magnitudes at which Bortolini24 find a completeness ($f_\mathrm{compl}$) of 20%, 50%, 80%, and 95% in the densest region of I Zw 18, R1. With black crosses we show typical observational errors from Bortolini24 at absolute $F115W$ magnitudes of -8, -7, and -6, estimated also for R1.