Early galaxies crossed a dust threshold at redshift 8.9

The dust content of early galaxies stops following a single trend at redshift 8.9, about 570 million years after the Big Bang. Below that redshift, meaning later in cosmic time, dust rises steeply as galaxies become more chemically enriched. Above it, the dust is mostly grains condensed in supernova ejecta. An analysis of dust masses from JWST and sub-millimetre observations reads the break as a change in which of two suppliers dominates.

At z = 8.9, JWST is seeing galaxies as they were 570 million years after the Big Bang, four per cent of the universe's 13.8-billion-year history. These are systems where the first generation of massive stars has already injected metals into the gas, and where the cold clouds have not had time to grow most of their dust on site.

Dust, to an astronomer, means solid grains rather than gas molecules. Grains absorb ultraviolet and blue starlight and re-emit that energy in the infrared, which is why the same galaxy can be faint or missing in JWST's near-infrared spectrograph and bright to ALMA's sub-millimetre receivers. Dust mass decides how you read almost everything else about a distant galaxy: its star-formation rate, its stellar mass, how much of its light you are failing to see.

Two ways to make a grain, and why one takes over

Supernovae make dust because their ejecta cool. As the expanding material thins, heavy elements condense into solids, and those grains are mixed into the surrounding gas. The supply is per event: it counts how many massive stars have exploded, not how enriched the gas around them has become. That part is standard stellar accounting rather than a result of the new analysis.

Interstellar growth works by accretion. In a cold, dense cloud, grains sit in a bath of gas-phase atoms, and a metal atom that strikes a grain sticks to it. The rate is set by how much metal is still in the gas phase rather than locked up in grains, and growth shuts itself off as those metals are consumed.

Both processes run at all times. What changes across the transition is which of them can build mass fast enough to matter. For accretion the relevant quantity is the time a grain needs to collect a mantle worth a decent fraction of its own mass, and that time is set by how often metal atoms arrive at the grain surface. It falls as the density of gas-phase metals rises and as the cloud grows denser; halve the metals available in the gas and the growth time doubles. Against that sits the time the galaxy has had, which its redshift fixes. The analysis finds the comparison flipping near z = 8.9. Above it, where the gas is metal-poor, the accretion time is longer than the few hundred million years available, so grains grown in clouds stay a negligible part of the dust budget. Below it, with more metals in the gas, the accretion time drops beneath the elapsed time and cloud-grown grains overtake the supernova supply. Every further generation of stars raises the metal abundance that growth feeds on, so the lead widens.

A change point is not a switch. It is the redshift at which the slope of a trend changes, and the fitted break is sharp because the model is. Physically the transition should be gradual and metallicity-dependent, beginning in the most enriched systems and spreading as enrichment continues. Supernova dust kept forming after z = 8.9; what changed is that it stopped being the main supply.

Finding a break in a trend that includes upper limits

Dust masses here come mostly from fitting JWST NIRSpec spectrophotometry, and partly from sub-millimetre photometry. Some of those measurements are upper limits rather than numbers, and the analysis handles them as censored data: the limits are kept in the fit rather than discarded. My reading of why that matters is that the galaxies with a measurable dust mass are, by construction, the dusty ones, and a fit to detections alone would tilt; the analysis itself reports only that the censored method was used. The model is a trend with a break at an unknown redshift, and the likelihood peaks at z ≈ 8.9, both for dust mass and for the dust-to-stellar mass ratio.

What it changes for the next dust mass you read

If the interpretation holds, a dust mass quoted for a galaxy at z = 10 is closer to a measurement of that galaxy's supernova history than of its accumulated enrichment. At low redshift, more dust means a more evolved, more metal-rich galaxy. Above the transition that reading fails: the dust-to-stellar mass ratio stops being a proxy for maturity, and the comparison you would instinctively make with a local galaxy does not hold, because the two numbers are produced by different processes with different dependence on mass and star-formation history.

It also means the dust-to-gas ratio of a galaxy above the transition is not a scaled-down version of one below it. The assumption that the same machinery made both fails somewhere between z = 9 and z = 8.

What is still open

The characteristic metallicity above which growth becomes efficient is invoked by the models rather than measured, and pinning down its value would be the cleanest test of the picture. Ultraviolet attenuation is consistent with a break at the same epoch, but would not have required one on its own, so the case rests mostly on dust masses from spectrophotometric fitting, which depend on assumed geometry and star-formation history. Sub-millimetre dust masses carry a further assumption: the inferred mass depends on dust temperature and emissivity, and if either evolves with redshift, part of the trend moves with it. The analysis also checks whether enrichment by the first, metal-free stars changes anything. That can alter the earliest chemical history while leaving the timing of the dust transition about where it is, which makes it a robustness check rather than a result.

One measurement would test the picture: the metallicity at which the dust-to-metals ratio, the fraction of a galaxy's metals locked into grains, starts to climb. That would turn the threshold from an assumption into a number.