This paper reports the first narrow‑band measurement of the Hα luminosity function — the count of galaxies by Hα line brightness — reaching into the Epoch of Reionization at redshift z ≈ 6.1. Hα (hydrogen alpha) is a bright spectral line produced when ionized gas recombines and it directly traces very recent star formation. The authors use 39 Hα emitters found by the JWST Emission Line Survey (JELS) to estimate how common Hα emitters were when the Universe was less than a billion years old.

The team used James Webb Space Telescope (JWST) NIRCam narrow‑band imaging with filters centered near 4.7 μm (F466N and F470N). The survey covered about 63 square arcminutes in the COSMOS field. The imaging strategy gave about 6,000 seconds (6 ks) of exposure across the whole mosaic and up to 12 ks in the central 40% of the area, for a total of 43 hours of program time. Narrow‑band selection picks objects with excess light in the filter caused by emission lines, so it is a clean way to find Hα emitters based only on their line strength.

From the 39 objects, the authors build both the observed Hα luminosity function and versions corrected for dust. They find results broadly consistent with recent slitless spectroscopic JWST measurements, but the observed counts disagree with some cosmological hydrodynamical simulations. The paper suggests those differences probably come from how simulations model emission lines and dust, not necessarily from JWST observations. By combining datasets the authors constrain the faint‑end slope of the Hα luminosity function to be between about −1.79 and −1.62, although it remains unclear how that slope changes with redshift.

The authors convert the Hα luminosity function into a cosmic star‑formation rate density. They report log10(ρ_SFR_Hα / [M⊙ yr⁻¹ Mpc⁻³]) = −1.93 (+0.14 / −0.12) if the continuum and line suffer the same dust extinction, and −2.00 (+0.16 / −0.10) if the line is less obscured (they quote two values using continuum‑to‑line extinction ratios η_dust = A_cont(6563 Å)/A_Hα = 1 and 0.44, respectively). Both estimates agree, within uncertainties and standard assumptions about the faint luminosity cutoff and SFR calibration, with earlier measurements.