JWST Selects 108 Galaxies by Their Paschen Lines

The JWST Emission Line Survey has picked out 108 galaxies by their hydrogen Paschen lines alone, in two slices of cosmic noon. Seventy-seven sit at z ~ 1.50, selected in Paschen-alpha. Thirty-one sit at z ~ 2.65, selected in Paschen-beta.

All of them come from a single narrowband NIRCam mosaic of the COSMOS field, observed through the filter pair F466N/F470N. That pair does double duty, and the arithmetic of why is worth doing once. Paschen-alpha has a rest wavelength of 1.875 micrometres; Paschen-beta sits at 1.282. Stretch the first by the redshift of the nearer sample, 1 + 1.50, and it arrives at 4.69 micrometres. Stretch the second by 1 + 2.65 and it arrives at 4.68. Two different lines, in galaxies at two different distances, end up inside the same narrow near-infrared window.

So one set of filters catches both samples. Selection is by narrowband excess: a galaxy whose line falls inside the filter's passband shows up brighter there than its broadband colours predict, and a galaxy whose line falls outside it does not show up at all.

Both samples are, by construction, galaxies with a strong line to find. The ranges in the table below are what this filter pair can see, not a census of cosmic noon.

The selection has a blind spot built in: star formation buried deeply enough to produce no line at all never enters the sample. Probing that is what the far-infrared stacking later in the paper is for.

A line flux and a spectrum fit do not measure the same thing, and the survey leans on the gap between them. A Paschen flux counts ionising photons, and the stars that produce those are the most massive and shortest-lived, gone within a few million years. So the line reports star formation over roughly the last few million years. A stellar-population fit to broadband photometry reports an average over far longer, because it sums the light of every generation of stars still shining. The team measured both for every source. In most of them the line SFR exceeds the SED SFR, and the excess falls weakly as stellar mass rises. These are galaxies caught in short bursts, not running at a steady rate.

The clean way to hold the two numbers together is exposure time. The SED fit is a long exposure; the line is one flash frame. In a galaxy forming stars steadily the two agree. Where the flash frame is brighter, something started recently that has not been running long enough to show up in the average.

Active galactic nuclei were removed by matching the sample against X-ray and radio catalogues, because a black hole accreting at a galaxy's centre can light up a Paschen line with no star formation involved. Redshifts and stellar masses came from EAZY and STARDUST fits. Line SFRs used calibrations written specifically for Paschen. Nebular extinction was inferred from the EAZY Av, and it rises with both mass and star formation rate: heavier and busier galaxies carry more dust.

Stellar masses for the combined sample run from log(M*/Msun) 7.4 to 10.3, and SED-based star formation rates from 0.1 to 27 solar masses per year.

The Paschen-beta sample is the busier of the two. Its line SFRs run up to 47.9 solar masses per year against 22.0 for the Paschen-alpha sample, and its median UV rate is about eight times higher, even though both came out of the same mosaic. Line SFRs span 0.1 to 22.0 solar masses per year in one sample and 0.7 to 47.9 in the other. The mass range covers nearly three orders of magnitude, so any statement about the sample as a whole is an average over galaxies that are not much alike.

The far-infrared gives upper limits, not detections. The team stacked Herschel, SCUBA-2, AzTEC, MIRI and ALMA imaging at the positions of the selected galaxies, adding many individually invisible sources together in the hope that the sum rises above the noise. Beyond MIRI, nothing reached three sigma. ALMA Band 4 gave the tightest constraint of the whole exercise, and it is a boundary rather than a number: obscured star formation below 53 and below 44 solar masses per year for the two samples, and gas masses below 10^11.2 and 10^10.6 solar masses.

The unobscured star formation rates are 0.48 and 3.92 solar masses per year. An upper limit of 53 or 44 on the dust-obscured part does not establish that these galaxies are dust-poor; it cannot yet separate a sample with almost nothing hidden from one where the hidden fraction dominates.

The extinction correction comes from EAZY Av, and Av is a number the SED fit produces rather than a measurement of dust, so the line rates carry whatever error that fit carries. The gas masses are similar in kind. 10^11.2 and 10^10.6 solar masses are ceilings rather than figures for how much gas sits in these galaxies, and a ceiling that high is compatible with a wide range of real gas contents.

Turning those ceilings into detections is a matter of depth, and the depth required is not marginal. To see obscured star formation at the level of the unobscured medians in the table, the stacked limits would have to fall from 53 to roughly 0.5 solar masses per year for the Paschen-alpha sample, and from 44 to roughly 4 for Paschen-beta. Those are the depths that would decide whether these bursts are visible or buried.