JWST counts faint galaxies to trace a busy era of star formation

A hydrogen signal from cosmic noon

Much of the Universe’s stellar growth happened during “cosmic noon,” the period spanning roughly redshifts 1 to 3. Measuring that growth means counting not just conspicuous galaxies, but also the fainter populations that contribute to the total.

Qianqiao Zhou and colleagues investigate this population using observations from the James Webb Space Telescope’s CANUCS program. Their study measures how common galaxies are at different hydrogen-line luminosities in two redshift intervals, centered near 1.3 and 2.0. Higher redshift means that cosmic expansion has stretched the light further, revealing an earlier epoch.

The work is an observational analysis supported by computational modeling and simulations of measurement effects—not a laboratory experiment or a simulation predicting galaxy formation. It is an arXiv preprint submitted to The Astrophysical Journal; the supplied record does not establish completed peer review.

Counting galaxies through natural lenses

The team uses NIRISS slitless spectroscopy, which spreads light from objects across the field into spectra without selecting individual targets with slits. They search for Hα, a hydrogen emission line tracing gas energized by young, massive stars.

Their observations cover five massive galaxy clusters and five associated parallel pointings. The clusters act as gravitational lenses, magnifying background galaxies and making intrinsically faint objects detectable. Lens models are needed to recover their luminosities and the actual volume surveyed.

After automated selection and visual inspection of images and spectra, the researchers retain 762 emitters: 433 between redshifts 1.1 and 1.5, and 329 between 1.8 and 2.2. They exclude severely contaminated spectra and remove one source with broad Hα emission associated with an active galactic nucleus. They also identify nine candidate multiply imaged systems, an important complication when lensing can make one galaxy appear more than once.

NIRISS cannot resolve Hα cleanly from neighboring nitrogen emission lines at this resolution. The paper’s reported “Hα” luminosity functions therefore describe blended Hα plus nitrogen emission. Estimating star formation requires a further correction for nitrogen, as well as dust.

What could the survey have missed?

A faint galaxy is not equally detectable everywhere. Nearby spectra can overlap, noise varies across an image, and lensing magnification changes with position and distance.

The researchers calculate an accessible survey volume separately for each source, accounting for these effects and its measured line width. They also generate 20,000 mock spectra per source across different positions and noise realizations. These computational tests estimate detection completeness and flux bias: near a detection threshold, noise can preferentially promote sources whose measured flux fluctuates upward.

The analysis reaches intrinsic line luminosities of about 10^40.6 erg per second near redshift 1.3 and 10^41 erg per second near redshift 2.0, with reported completeness above 80%.

Another uncertainty comes from cosmic variance: different small patches of sky contain different concentrations of galaxies. The authors estimate relative cosmic variance of roughly 20–30% for individual fields, falling to about 10% when the ten survey regions are combined. Multiple sightlines reduce this problem rather than eliminate it.

A tentative change in the faint population

The team fits a luminosity function describing the number of galaxies at each luminosity. Its faint-end slope measures how rapidly galaxy abundance increases toward lower luminosities.

They obtain slopes of −1.51, with uncertainties of +0.06/−0.04, near redshift 1.3, and −1.42, with +0.09/−0.06, near redshift 2.0. The central values suggest a slightly shallower slope at the earlier epoch, but the measurements overlap within their uncertainties. They do not establish a clear evolutionary change.

The bright end is less securely measured because the survey contains few very luminous sources. Its fitted behavior depends substantially on prior constraints adopted from earlier studies.

Dust changes the inferred total

Assuming the same one-magnitude Hα dust attenuation at both epochs, the authors estimate star-formation-rate densities of 0.103 and 0.149 solar masses per year per cubic megaparsec of comoving volume, respectively. These totals come from integrating fitted star-formation-rate distributions down to 0.4 solar masses per year—not directly measuring every galaxy.

Dust assumptions make a substantial difference. Using alternative attenuation estimates from stacked hydrogen-line measurements changes the redshift-2 result to approximately 0.100–0.164 in the same units. That range illustrates sensitivity to the correction; it is not a single statistical confidence interval.

The study strengthens the faint-galaxy census during cosmic noon. Its central limitation remains the conversion from observed emission to total star formation: dust, blended lines, fitted distributions and integration limits all influence the answer.