How TEMPOS and FEAST Are Rewriting Models of Early Galaxies

Early galaxies are often interpreted with simplified feedback prescriptions: stellar winds are assumed to scale smoothly with metallicity, and young star clusters are assigned a common time to break out of their natal clouds. The TEMPOS and FEAST surveys challenge both shortcuts. Together, they point to feedback that depends strongly on elemental composition in massive stars and cluster mass in stellar nurseries.

The two surveys probe different stages of the same feedback cycle

TEMPOS—the Treasury of Extremely Metal-Poor O Stars—uses Hubble ultraviolet spectroscopy to build a homogeneous atlas of 29 massive O stars in nearby galaxies with metallicities below about one-fifth solar. The survey adds new Cosmic Origins Spectrograph observations of 12 very metal-poor O stars to archival data.

FEAST—Feedback in Emerging extrAgalactic Star clusTers—combines JWST infrared data, which can see into dusty birth clouds, with Hubble ultraviolet and optical observations of exposed clusters. Its four-galaxy census contains nearly 9,000 young clusters and tracks their transition from embedded to optically visible. 115

TEMPOS addresses how massive stars lose mass and shape their ultraviolet spectra. FEAST addresses when newly formed clusters clear their surrounding gas and dust. Both processes affect how radiation, momentum, and eventually heavy elements move through a young galaxy.

TEMPOS: metallicity is not a complete description of stellar winds

The TEMPOS results report unexpectedly weak winds among massive stars in extremely metal-poor nearby dwarf galaxies, based on observations of 29 stars across six systems. The key implication is that the familiar practice of using a galaxy’s overall metallicity—or gas-phase oxygen abundance—as a stand-in for wind strength can be inadequate at the lowest abundances.

Radiatively driven winds depend on the metal ions that absorb stellar radiation, and iron is especially important to that process. Existing ultraviolet work on extremely metal-poor galaxies has found stellar iron abundances below 10% of solar while gas-phase oxygen can imply a different chemical picture, including enhanced oxygen relative to iron. In other words, two galaxies with similar oxygen measurements may not host massive stars with identical wind properties.

The reported TEMPOS trend is therefore best understood as a warning against simple extrapolation: at very low abundance, wind behavior may change more abruptly than a smooth metallicity scaling predicts. The exact strength and universality of any threshold will require further modeling and comparison across larger samples.

Why weaker winds matter

A weaker wind means a massive star sheds less material and angular momentum before it dies. That can alter model predictions for its final mass, surface composition, lifetime, interactions with a binary companion, and the relative likelihood of different core-collapse or black-hole-forming outcomes. These are physical consequences for stellar-evolution calculations—not direct measurements of individual supernova fates by TEMPOS.

It also matters for galaxy-scale models. Stellar winds provide mechanical feedback before the first supernovae, while ultraviolet spectra from massive stars set the ionization conditions that produce many of the nebular lines JWST observes. A model calibrated with the wrong iron abundance or wind prescription can misestimate stellar ages, masses, ionizing spectra, and inferred star-formation properties.

FEAST: massive clusters clear their birth clouds sooner

FEAST finds that the time required for clusters to emerge from natal gas and dust is mass-dependent. Across the Hubble–JWST sample, the overall transition from embedded to exposed lasts about 6 million years. More massive clusters, above roughly 5,000 solar masses, emerge in about 5 million years, compared with about 7 million years for clusters near 1,000 solar masses; a broader analysis describes the lower-mass comparison as roughly 7–8 million years. 5

This is a meaningful change from models that assign every young cluster the same embedded-phase duration. More massive clusters appear able to remove or disperse their surrounding material earlier, exposing their stars to view at ultraviolet and optical wavelengths. NASA’s summary of the four-galaxy study likewise reports that more massive clusters emerge more quickly from their birth clouds. 1

Earlier UV visibility is not automatically higher intrinsic UV production

The distinction matters. Faster cloud clearing does not necessarily mean a cluster’s stars emit more ultraviolet light at the source. It means less surrounding gas and dust remains to obscure, absorb, or locally confine that radiation.

As a result, the timing of UV radiation reaching the surrounding interstellar medium can change. That can influence nebular emission, dust attenuation, the geometry of ionized gas, chemical enrichment, and—if low-density pathways exist—the fraction of ionizing photons that escape a galaxy. FEAST was designed specifically to constrain emergence times as a function of mass and environment for feedback and star-formation models. 1615

What changes in models of first-billion-year galaxies

The main revision is not that early galaxies had a single new feedback setting. It is that two commonly fixed ingredients should become variable.

1. Use iron-sensitive stellar-wind and spectral models

Population-synthesis and spectral-energy-distribution models should avoid treating oxygen abundance alone as a universal predictor of massive-star winds. Incorporating stellar iron abundance and variations in oxygen-to-iron ratio can improve predictions for ultraviolet continua, wind features, ionizing radiation, and the nebular lines used to infer physical properties from JWST spectra.

2. Make birth-cloud clearing dependent on cluster mass

Galaxy simulations and SED-fitting frameworks can replace a universal embedded-phase timescale with mass-dependent emergence prescriptions. The FEAST measurements provide an empirical scale: a mean of about 6 Myr, with faster clearing for the more massive clusters in the sample. 5

That adjustment can change predicted dust attenuation, nebular covering fractions, feedback timing, and the apparent ages of young stellar populations.

3. Separate mechanical feedback from radiative exposure

The combined picture is more nuanced than “stronger feedback” or “weaker feedback.” Very metal-poor massive stars may inject less mechanical energy through winds than models expect, while the most massive clusters may nevertheless reveal their radiation sooner by clearing local material more rapidly. Both effects can shape how UV-bright a young galaxy appears and how effectively its radiation interacts with surrounding gas.

A calibration for JWST—not a direct view of the first galaxies

Both surveys study nearby systems. TEMPOS targets local low-metallicity analogues, and FEAST studies clusters in four nearby galaxies. 15 They are therefore not direct observations of stars or clusters from the universe’s first billion years.

Their value is that they anchor parts of early-galaxy modeling that have often relied on extrapolation. TEMPOS supplies a much-needed ultraviolet reference set for massive stars at low metallicity, while FEAST supplies measured cloud-clearing times across cluster masses. Applied carefully—and tested against JWST spectra at high redshift—those calibrations can make estimates of stellar populations, dust, feedback, and ionizing radiation more physically grounded.