Lyman-Break Galaxy

The galaxy you find by what is missing

Most ways to identify a distant galaxy hinge on something it does — an emission line, a thermal peak, a flat radio spectrum. The Lyman break is the opposite. It is an absorption feature, a piece of the spectrum that simply is not there. A star-forming galaxy is, at far-ultraviolet wavelengths, the integrated continuum of its hot, young O- and B-type stars; that continuum is a moderately smooth power law from a few hundred to a few thousand ångströms. Then, at exactly 912 Å, the spectrum falls off a cliff.

The reason is atomic. A photon with wavelength λ < 912 Å carries more than 13.6 eV — enough to ionise a hydrogen atom from its ground state. Any neutral hydrogen along the photon's path therefore absorbs it. Inside the galaxy that means clouds of cold and warm neutral gas surrounding the star-forming regions; outside the galaxy it means the intergalactic medium, particularly the dense web of cold neutral clouds that produce the Lyα forest in quasar spectra. Together they make the rest-frame far-UV continuum below 912 Å nearly opaque. The cliff is the Lyman break.

Now redshift the whole picture. At z = 3, the 912 Å edge slides to (1 + z) × 912 = 3648 Å — into the U-band. A galaxy that looks bright in the B-band and V-band is suddenly invisible in U. Put a filter set on a deep image, and z ≈ 3 galaxies appear as a striking signature: faint or absent in U, normal in B, normal in V. They are dropouts.

Steidel 1996: turning the trick into a method

The idea that the Lyman break could be used as a redshift indicator had floated around since the 1970s (Meier 1976; Partridge), but the technique only became practical with the arrival of large-format CCDs on 4-metre and 10-metre telescopes. In 1996 Charles Steidel, Mauro Giavalisco and collaborators published the paper that turned the trick into a method. They observed deep fields in U, G and R, defined a colour-colour selection box in (U−G) vs (G−R), and showed that virtually every galaxy inside the box was at z = 2.7 – 3.4 when followed up with Keck LRIS spectroscopy. The contamination rate was low; the success rate was high; and crucially, the technique scaled to thousands of galaxies per night of imaging.

The follow-up programme — what would become the Lyman-break galaxy sample — established the population. UV luminosity function. Clustering. Spectroscopic confirmation. Stellar masses from rest-frame optical photometry. By the early 2000s, LBGs at z ≈ 3 were the best-characterised distant galaxy sample in astronomy, and the "Steidel survey" had set the methodological template for everything since.

A dropout cookbook

The dropout selection is filter-set arithmetic. For a fixed redshift, the break is at a known observed wavelength, and the filter that sits just blueward of it is the dropout filter. As you push to higher redshift, you simply redshift the filter set.

Dropout filter	Approx. observed λ of break	Selected redshift	Typical detector / instrument
U-band (~3600 Å)	3650 Å	z ≈ 3	Keck LRIS, ground 4-m CCDs
B-band (~4400 Å)	4400 Å	z ≈ 4	HST ACS, ground 8-m
G / V-band (~5500 Å)	5500 Å	z ≈ 5	HST ACS, deep ground
R-band (~6500 Å)	6500 Å	z ≈ 6	HST ACS, Subaru SC
i-band (~7800 Å)	7800 Å	z ≈ 7	HST WFC3 / NIRCam
z-band (~9000 Å)	9000 Å	z ≈ 8	HST WFC3, JWST NIRCam
Y-band (~1.0 µm)	1.0 µm	z ≈ 10	JWST NIRCam
J-band (~1.25 µm)	1.25 µm	z ≈ 12	JWST NIRCam
H-band (~1.6 µm)	1.6 µm	z ≈ 15	JWST NIRCam (CEERS, JADES)

Each row is essentially the same physics — the 912 Å break, augmented at z ≳ 5 by Lyα forest absorption — viewed through a different filter pair. The redshift granularity is set by the filter width and the contrast between absorbed and unabsorbed wings; in practice, photometric redshifts from dropouts have σ_z ≈ 0.2 – 0.5, with spectroscopic follow-up tightening this to redshift accuracies of 10⁻³ or better.

The role of the IGM at z ≳ 2

The "Lyman break" in observed-frame photometry of high-z galaxies is rarely a clean 912 Å edge. It is in fact a compound feature. Above z ≈ 2, three pieces of absorption stack onto the rest-frame UV continuum:

• The Lyman limit (912 Å). Local to the galaxy: photoelectric absorption by neutral hydrogen surrounding star-forming regions. The cleanest expression of the "break".
• The Lyα forest (912 – 1216 Å). A statistical superposition of thousands of low-column-density H I clouds along the line of sight, each absorbing at its own redshifted Lyα wavelength. The forest depresses the integrated flux between Lyα and the Lyman limit (the "Lyα forest decrement").
• Damped Lyα systems (DLAs) and Lyman-limit systems (LLSs). A small number of high-column-density clouds (N_HI > 10¹⁷ cm⁻², up to 10²¹ for DLAs) can produce additional 912 Å edges at their own redshift, sometimes mimicking a galaxy break.

The empirical formula for the IGM transmission below Lyα (Madau 1995; updated by Inoue, Shimizu et al. 2014) handles all three at once, allowing photometric redshifts to be derived from the SED with the IGM "baked in." By z ≈ 6 the forest becomes so opaque that the flux below Lyα is consistent with zero — this is the Gunn-Peterson trough, and it is responsible for why dropouts at z > 6 show essentially complete extinction of the bluer band rather than a partial decrement.

Worked example: which band drops out at z = 4.5?

Take a galaxy at z = 4.5 with a standard star-forming SED. The Lyman limit sits at:

λ_obs(912 Å) = (1 + z) × 912 Å
            = 5.5 × 912 Å
            = 5016 Å    (between V and R)

And Lyα at 1216 Å sits at:

λ_obs(Lyα) = 5.5 × 1216 Å
          = 6688 Å    (deep in R)

Below 5016 Å (rest 912 Å), the galaxy is essentially dark — neutral hydrogen wins. Between 5016 and 6688 Å, the spectrum is depressed by the Lyα forest but not zero. Above 6688 Å, the continuum is mostly clean. Filter that through B, V, R, i, z:

• B (~4400 Å) — entirely below the break → dropout, faint or absent.
• V (~5500 Å) — straddles the break; partial flux.
• R (~6500 Å) — straddles Lyα; depressed but detected.
• i (~7800 Å) — above Lyα → full continuum, bright.
• z (~9000 Å) — full continuum, bright.

So at z = 4.5 the canonical signature is "drops out of B, detected from V redward" — a B-dropout. The same logic, with redder filters, gives R-dropouts at z ≈ 5 and z-dropouts at z ≈ 6 – 7.

What LBGs are actually like

The LBG sample at z ≈ 3 is now characterised in considerable detail. Median properties are:

Property	Typical value	Notes
Stellar mass M_*	10⁹ – 10¹⁰·⁵ M☉	Few times 10¹⁰ at the bright end; up to 10¹¹ for the rarest
Star-formation rate	10 – 100 M☉/yr	Bright end can reach 10³ M☉/yr (overlap with SMGs)
UV continuum slope β	−2.5 to −1.5	Sets the inferred dust extinction A_V ≈ 0 – 1.5
Stellar metallicity	0.1 – 0.5 Z☉	Sub-solar but already enriched
Lyα emission	~30 % strong emitters, ~30 % strong absorbers	Bimodal; selection effects matter
Sizes (rest-frame UV)	1 – 3 kpc	Compact, often clumpy
Outflow velocity (interstellar lines)	200 – 600 km/s	Galactic winds are nearly universal
Number density (bright LBGs)	~10⁻³ Mpc⁻³	At L* of the z=3 UV LF
Halo mass	10¹¹·⁵ – 10¹² M☉	From clustering — direct progenitors of Milky-Way-like haloes

Population-wide, LBGs are the building blocks of the today's Milky-Way-mass galaxies. The clustering measurements pin them to haloes near 10¹² M☉ at z ≈ 3 — exactly the haloes that, evolved forward in a ΛCDM simulation, produce L* galaxies today. They are not freaks; they are the L* galaxies of their epoch, caught in the act of assembling.

From dropout counts to cosmic star-formation history

The headline use of LBGs is the Madau-Lilly-Dickinson diagram — the cosmic star-formation-rate density ρ_SFR(z). The recipe is:

1. Count galaxies in a redshift bin (dropout selection).
2. Build the rest-frame UV luminosity function ϕ(L_UV).
3. Integrate L · ϕ(L) dL down to a faint-end limit → ρ_UV(z).
4. Correct for dust attenuation using the UV slope β → ρ_UV(intrinsic).
5. Convert to SFR using a stellar-population synthesis calibration → ρ_SFR(z).

For a Kroupa IMF and constant SFR for > 100 Myr, the conversion is approximately:

SFR (M☉/yr) ≈ 1.15 × 10⁻²⁸ × L_UV (erg/s/Hz)    [Kennicutt & Evans 2012]

The result, plotted over z = 0 – 10, is the celebrated rise of ρ_SFR from z = 0 to a peak near z ≈ 2 ("cosmic noon"), followed by a gentle decline toward higher z. LBG dropouts dominate the diagram from z ≈ 3 onward; JWST has now extended it confidently to z ≈ 10 and tentatively to z ≈ 14.

JWST and the first-galaxies frontier

Before JWST, the highest-z confirmed LBG was GN-z11 at z = 11.09 (Oesch et al. 2016, HST grism). JWST changed the scale of what was possible at one stroke. With NIRCam imaging across 1 – 5 µm and NIRSpec multi-object spectroscopy, the same Lyman-break logic — now with the break redshifted past 1 µm — became routine.

• CEERS (Cosmic Evolution Early Release Science) provided the first confirmed z = 8 – 11 sample within a few months of NIRCam first light.
• JADES (JWST Advanced Deep Extragalactic Survey) pushed to z = 13 – 14 with spectroscopic confirmations, including JADES-GS-z14-0 at z = 14.32 — the current spectroscopic record.
• UNCOVER and CANUCS exploited cluster lensing to find intrinsically faint but magnified dropouts at the same redshifts, sampling the faint end of ϕ(L_UV).
• COSMOS-Web mapped a 0.54-deg² field to find rare bright z > 10 galaxies whose comoving abundance was probably too small to catch in narrower surveys.

One striking JWST result is the apparent excess of UV-bright galaxies at z > 10 compared with pre-launch model predictions. The data suggest some combination of (a) higher-than-expected star-formation efficiency in early haloes, (b) burstier star-formation histories that boost short-timescale L_UV, (c) reduced dust attenuation, or (d) a top-heavy initial mass function. The eventual answer is still in flux as samples grow.

LBGs and reionization

Reionization — the transition of the IGM from neutral to ionised between z ≈ 6 and z ≈ 10 — required a source of ionising photons. The leading candidate is star-forming galaxies, and LBG counts are the way to test that. The condition for galaxies to maintain ionisation balance against recombinations is roughly:

ṅ_ion = ρ_UV × ξ_ion × f_esc > n_rec
       [photons s⁻¹ Mpc⁻³]

where ξ_ion is the ionising photon production per unit UV luminosity (calibrated from Hα and [O III] equivalent widths) and f_esc is the fraction of ionising photons that escape the galaxy into the IGM. JWST has measured ξ_ion in stacked z ≈ 6 – 8 LBGs to be ~10²⁵·⁵ erg⁻¹ Hz, slightly elevated above z ≈ 3 values. f_esc is the persistently uncertain term — direct constraints come from rare low-z analogues (the "Lyman-continuum leakers"). Current estimates are 5 – 20 %, with weight on roughly 10 %.

If those numbers hold, LBGs alone — extrapolated below the JWST faint limit — can produce more than enough ionising photons to reionise the universe by z ≈ 6, in agreement with Planck and quasar-Lyα-forest constraints. There is room for AGN or other sources, but they are not required.

Variants, complements and selection biases

• Lyα emitters (LAEs). Selected by strong Lyα emission rather than by the continuum break. LAEs and LBGs overlap but are not the same — LAEs are lower-mass, less dusty, and bluer on average. The two selections together give a fuller view of the high-z galaxy population.
• BX/BM galaxies (Steidel z ≈ 2). The same colour-colour technique extended to z = 1.5 – 2.5 using rest-UV photometry between U_n, G and R. They are LBGs' lower-z cousins and characterise "cosmic noon" star formation.
• "Distant red galaxies" (DRGs) and BzK-selected galaxies — complementary selections that pick up dusty and quiescent z ≈ 2 galaxies that LBG colour cuts miss.
• Submillimetre galaxies (SMGs). Strongly dust-obscured starbursts; their UV continuum is suppressed and they often fail LBG colour cuts. The two populations represent extremes of dust content in a single underlying SFR distribution.
• NB816 / NB921 narrowband Lyα surveys select LAEs at z ≈ 5.7 and 6.6 — useful as a redshift slice cross-check on dropout samples at the same epochs.

Common pitfalls

• Confusing the Lyman break with the Balmer/4000 Å break. The Balmer break is a much smaller, rest-frame-3700 Å feature in older, quiescent populations. At very high z, the redshifted Balmer break in IR photometry can mimic a Lyman break of a much lower-redshift dusty galaxy — the source of many JWST "lost" high-z candidates.
• Treating LBGs as a complete census. Dropout selection requires a UV-bright, low-extinction continuum. Dusty galaxies and old quiescent populations drop out of dropout samples. The bias matters at z = 1 – 3 where dusty SMGs dominate ρ_SFR.
• Ignoring IGM stochasticity. The Lyα forest is statistical; individual sightlines vary. Mean IGM transmission curves are fine for stacked populations, but individual photometric redshifts at z > 5 should be tested against forest variance.
• Underestimating contamination from Galactic stars and low-z interlopers. Cool M-dwarfs and L-dwarfs can mimic dropout colours; so can z ≈ 1.5 dusty galaxies. Robust dropout samples require a combination of multi-band SED fitting and morphological cuts.
• Forgetting the f_esc unknown for reionization budgets. Tightly constrained ρ_UV times an order-of-magnitude-uncertain f_esc gives an order-of-magnitude-uncertain ṅ_ion. The biggest leverage on the reionization budget is direct ionising-leakage measurements, not deeper continuum surveys.