Lyα Tomography of Extremely Metal-poor Domains

LATED Explorer

LATED (Li et al. 2026) is a photometric search for extremely metal-poor and metal-free star formation, from cosmic noon through reionization. It starts from a Lyα-emitter sample with slice redshift known, and looks for the signature of metal-free star formation: a very strong Hα together with a weak or absent [O III]. Two band colours measure it, and their redshift-dependent thresholds define the selection. This page gives those thresholds for every band set that can carry it, and derives them for any bands you choose.

x
=
m[O III]+Hβ band − mHα band
the O/H axis
y
=
mHα band − mcontinuum band
the EW(Hα) axis
Selected when x ≥ xmin(z) and y ≤ ymax(z)

Look up a redshift

Type a redshift, or click and drag along the window ladder below.

Criteria against redshift, as in the methodology paper: (a) the Hα-strength cut ymax(z), (b) the [O III]-weakness cut xmin(z), for the JWST wide-band windows, each with its primary continuum band. Dashed: extended variants for multi-band fields. Dots mark the chosen redshift; click or drag to move it.

Criteria for any band set

Per-redshift envelope of Pop III templates through chosen bands.

Bands
Redshift range
Pop III target locus

Loading model grids…
Band geometry and filter curves

Transmission normalised to each band's peak. Shaded columns are the wavelengths Hβ, [O III]λλ4959,5007 and Hα sweep across the chosen redshift range; dotted lines mark the 50 per cent (line inside a band) and 10 per cent (band free of a line) thresholds.

Derived criteria. Line: envelope on the redshift grid. Dots: the 25-row table. Dashed (released windows only): the published table. Hover for the template that sets each boundary.
Show CSV text

Selection plane

The three-panel layout of Figs 5–9. Solid corner: loosest redshift edge; dashed: tightest.

(a) Model library. Noiseless colours; contaminant classes sampled at 16 redshifts across the range. The arrows point towards lower O/H and higher EW(Hα).
(b) Noise realisation. mcont ~ U(23.5, 29), flux noise at these depths. Pass rates apply the colour cuts and S/N ≥ 5 in the Hα and continuum bands.
(c) Analytic grid. Flat-fν continuum with Case-B hydrogen lines (Te = 104 K, ne = 102 cm−3) scaled by EW(Hβ), and the [O III] doublet scaled by R3 = [O III]λ5007/Hβ.

What is computed

The two cuts come from the Pop III templates themselves. At every redshift on a Δz = 0.001 grid the page pushes each template through the real filter curves, then keeps the extremes of the resulting locus: xmin is the smallest x and ymax the largest y any template reaches, so the box holds all of them. Those envelopes are interpolated onto the 25 rows published as CSV files.

The photometry is the pipeline's: filter curves integrated on a ≤5 Å grid with the pivot-wavelength AB convention.

Where the continuum band is not redward of Hα, the locus is trimmed to instantaneous bursts ≤10 Myr: there the three bands alone cannot separate a red continuum from an Hα excess, and the evolved bursts sit in the strip the red-continuum interlopers occupy.

When a band set works

A band set is usable over the redshifts where the geometry holds:

  • Hα, and all of [O III]λλ4959,5007 and Hβ, sit at 50 per cent of peak transmission or more in their own bands, so each band collects the whole line.
  • Neither line band holds the other's line above 20 per cent of peak, so the two colours stay independent.
  • The continuum band stays below 10 per cent of peak for Hα, [O III], Hβ, [O II] and He I 10830, and sits redward of Hα where a band exists there, otherwise between the two lines.
  • Where two NIRCam wide bands overlap, a line landing in the crossover is measured cleanly by neither, so those redshifts are trimmed.

Lyα coverage is a separate question and is not applied here, so a range can be wider than a published window; below z ≈ 1.9 Lyα sits in the near-ultraviolet, beyond ground-based parent surveys. A narrow band is too narrow to hold the whole [O III]+Hβ complex, so narrow bands appear as Hα or continuum bands only.

Models behind the numbers

Target locus: Yggdrasil metal-free populations (Zackrisson et al. 2011; instantaneous bursts to 20 Myr, three IMFs, fcov = 0.5 and 1) together with the Cloudy Pop III grids of Nakajima & Maiolino (2022), with their log U = −0.5 corner left out.

The contaminant classes on the selection plane are shown for context only and never enter the cuts: FSPS populations (Conroy et al. 2009, 2010; Byler et al. 2017) for the metal-poor, star-forming, dusty and passive classes, the pristine black holes (PBHs: massive black-hole seeds born in essentially metal-free environments, such as direct-collapse and primordial black holes; the DCBH grid) of Nakajima & Maiolino (2022), the Little Red Dot templates of Zhang et al. (2025), and AGN line sets transplanted from SDSS DR17.

Filter curves come from the SVO Filter Profile Service. Data version .