Galaxy Evolution
Cold Accretion Streams: How Galaxies Feed Without Shock-Heating
Around 12 billion years ago, when the universe was barely a fifth of its current age, the most vigorously star-forming galaxies were being fed by narrow ribbons of gas at just ~10,000 K — cold enough to fall straight to the galactic center at 200-300 km/s without ever being heated to the million-degree virial temperature of the halo around them. These are cold accretion streams: dense, filamentary flows of near-primordial hydrogen that thread through a hot dark-matter halo and deliver fuel directly to a growing galaxy.
Cold accretion (also called "cold-mode" accretion or "cold flows") is one of the two fundamental channels by which galaxies acquire baryons. Unlike the classical picture in which infalling gas is shock-heated at the virial radius, cold streams stay cool because they cool radiatively faster than a stable accretion shock can form. The concept reshaped galaxy-formation theory after ~2003-2006 and explains why disks grow so fast at high redshift.
- TypeFilamentary cold-mode gas accretion onto galaxies
- RegimeHalo mass below ~10^12 M_sun (or any mass at z > ~2)
- Stream temperature~10^4 - 10^4.5 K (near virial temp of the gas, not the halo)
- EstablishedKeres et al. 2005; Dekel & Birnboim 2006
- Key criteriont_cool < t_comp (cooling beats compression -> no stable shock)
- Observed viaLyman-alpha blobs; low-metallicity absorbers in quasar sightlines
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
What cold accretion streams are
In the standard Lambda-CDM picture, galaxies form at the centers of dark-matter halos, and the baryons fall in with the dark matter. The textbook expectation from the 1970s (Rees & Ostriker, Silk, White & Rees) was that infalling gas passes through a virial accretion shock near the halo's edge, converting its gravitational infall kinetic energy into heat at the virial temperature T_vir ~ 10^6-10^7 K. It then radiates slowly and drips inward as a "cooling flow."
Cold accretion is the realization that this shock does not always form. When gas is dense and can radiate its energy away faster than the shock can process it, the gas simply keeps falling in cold — never reaching the virial temperature. Because the densest gas lies along the filaments of the cosmic web, this cold gas arrives not as a spherical shell but as a few narrow streams that punch through the (sometimes hot) halo and reach the central disk almost radially.
- Cold-mode: gas stays at ~10^4 K, fed along filaments.
- Hot-mode: gas is shock-heated to T_vir, then cools slowly.
The mechanism: why the shock fails to form
Whether a stable virial shock exists is a competition between two timescales evaluated just behind the shock front. The compression (or dynamical) time t_comp — roughly how fast infalling gas piles onto and pressurizes the shocked layer — versus the radiative cooling time t_cool = (3/2) n k_B T / (n_e n_H Lambda(T)), where Lambda(T) is the cooling function.
The stability criterion, worked out analytically by Birnboim & Dekel (2003), is essentially:
- If t_cool > t_comp -> the post-shock gas holds its pressure, a stable shock stands off near the virial radius -> hot mode.
- If t_cool < t_comp -> the gas radiates away the shock's thermal support faster than it can build up; no stable shock forms and gas free-falls in cold -> cold mode.
Because Lambda(T) for primordial-to-low-metallicity gas peaks strongly near 10^4-10^5 K (hydrogen and helium line cooling), cooling is extremely efficient at low temperatures and low halo masses. The condition t_cool < t_comp is met when the halo (and thus the shock-compressed gas) is below a threshold mass — the origin of the critical mass scale.
Key quantities and the critical mass
The defining number is the critical shock-heating halo mass. Combining the cooling function with the halo dynamics, Dekel & Birnboim (2006) found the transition sits at:
- M_shock ~ 6 x 10^11 M_sun (roughly 10^11.6-10^12 M_sun), nearly independent of redshift.
Below M_shock the halo cannot support a stable shock; above it, gas is shock-heated. Characteristic stream numbers at high redshift:
- Stream temperature: T ~ 10^4-10^4.5 K (set by cooling, not by T_vir).
- Infall speed near the disk: ~200-300 km/s, close to the halo circular velocity.
- Accretion rate onto a ~10^12 M_sun halo at z~3: ~100 M_sun/yr, typically split among ~3 streams.
- Density contrast of a stream vs. ambient halo gas: factor of ~10-100.
Worked estimate: the free-fall time from R_vir ~ 100 kpc for such a halo is t_ff ~ R_vir / V_c ~ (100 kpc)/(250 km/s) ~ 4 x 10^8 yr — so cold streams deliver fuel on a few-hundred-Myr timescale, far shorter than the Gyr-scale cooling of a hot atmosphere. Crucially, above M_shock cold streams can still penetrate a hot halo at high z, because the hot gas hasn't had time to grow dense enough to disrupt them.
How cold streams are observed
Cold streams are faint and diffuse, so detection is indirect and hard-won. The main tracers:
- Lyman-alpha emission (Ly-alpha blobs): gravitational energy released as gas falls in is radiated partly in Lyman-alpha. Goerdt et al. (2010) and Dijkstra & Loeb showed cold streams can power extended Ly-alpha blobs of ~10^43-10^44 erg/s seen around z~2-3 galaxies — though AGN and starburst photoionization are competing explanations, which keeps this contested.
- Absorption lines in background quasar/galaxy sightlines: cold streams appear as metal-poor, ~10^4 K H I absorbers with modest column densities and coherent kinematics. Surveys of the z~2-3 circumgalactic medium (e.g., Keck/KCWI, VLT/MUSE, and QSO-absorption studies) find abundant cool, low-metallicity gas consistent with inflow.
- Inflow signatures: redshifted absorption against a galaxy's own light, and lopsided/one-sided kinematics of circumgalactic gas.
Cosmological simulations — SPH codes like GADGET (Keres et al. 2005) and the moving-mesh AREPO, plus modern suites (FIRE, IllustrisTNG) — are where the streams were first "seen" and remain essential for interpreting the data.
How it differs from related phenomena
Cold accretion is easily confused with several neighbors:
- vs. classical cooling flows: a cooling flow starts from shock-heated hot gas that radiates and settles inward; cold streams never heat up in the first place. Cooling flows dominate above M_shock, in groups and clusters.
- vs. an accretion disk: the same word "accretion" but a different scale — an accretion disk is the ~AU-to-pc gas orbiting a compact object; cold streams are ~100 kpc galaxy-scale inflows.
- vs. galactic fountains / recycled winds: feedback-driven outflows can rain back down; that recycled material is often metal-enriched, whereas pristine cold streams are metal-poor (Z ~ 0.01-0.1 Z_sun).
- vs. mergers: cold streams are smooth gas accretion, distinct from the clumpy accretion of merging satellite galaxies (though streams also carry clumps).
The unifying variable is the shock stability criterion t_cool vs t_comp, which sorts all of these into cold-mode versus hot-mode regimes.
Significance and open questions
Cold streams solved several puzzles at once. They explain the observed galaxy bimodality — blue star-forming disks below M_shock, red-and-dead ellipticals above it (Dekel & Birnboim 2006) — because switching off cold fuel above the critical mass naturally quenches star formation. They also account for the surprisingly high star-formation rates and clumpy, turbulent disks seen in z~2 galaxies, which cold streams feed continuously and even help build angular momentum.
What's still debated:
- Do the streams survive? Kelvin-Helmholtz and thermal instabilities can shred streams as they cross the hot halo (Mandelker et al. 2016-2020); whether they arrive intact depends sensitively on resolution and magnetic fields.
- Are the Lyman-alpha blobs really cold streams, or AGN/starburst fluorescence? Direct, unambiguous imaging of a stream remains a holy grail — JWST and future 30-m-class telescopes are the best hope.
- Feedback and magnetic fields may suppress or channel streams in ways current simulations don't fully capture.
The most-cited landmark cases remain the SSA22 protocluster Lyman-alpha blobs and simulated cold-stream halos, which crystallized the idea that galaxies grow largely by being fed cold, not by cooling hot.
| Property | Cold-mode (cold streams) | Hot-mode (shocked halo) |
|---|---|---|
| Halo mass regime | M_halo < ~10^12 M_sun (M_shock ~ 6x10^11) | M_halo > ~10^12 M_sun |
| Gas temperature | ~10^4 - 10^4.5 K | ~10^6 - 10^7 K (virial temperature) |
| Geometry | Anisotropic filaments / streams | Quasi-spherical hot atmosphere |
| Virial shock | Absent or unstable (t_cool < t_comp) | Stable, sits near virial radius |
| Fueling of star formation | Direct, fast (~free-fall time) | Slow, via cooling flow from hot gas |
| Dominant epoch | High z (z > 2); low-mass halos today | Low z; massive halos, groups, clusters |
Frequently asked questions
What is a cold accretion stream in simple terms?
It's a narrow filament of relatively cool gas (~10,000 K) that flows straight into a young galaxy from the cosmic web, delivering fuel for star formation. Unlike the classical picture, this gas is never heated to the million-degree virial temperature of the surrounding halo. It stays cold because it radiates energy away faster than a shock can heat it.
Why doesn't the gas get shock-heated like in the standard model?
A stable virial shock only forms if the post-shock gas can hold its pressure, which requires the cooling time to be longer than the compression time (t_cool > t_comp). In low-mass halos, radiative cooling near 10^4-10^5 K is so efficient that gas radiates away the shock's heat almost instantly (t_cool < t_comp), so no stable shock forms and the gas keeps falling in cold.
What is the critical mass for cold versus hot accretion?
The transition halo mass is about 6 x 10^11 solar masses (roughly 10^12 M_sun), first derived by Dekel and Birnboim in 2006. Below it, halos are fed by cold streams; above it, infalling gas is shock-heated into a hot atmosphere. This threshold is nearly independent of redshift, though at high redshift cold streams can still penetrate even massive hot halos.
When and where does cold accretion dominate?
It dominates in two regimes: (1) all halos below ~10^12 M_sun at any epoch, and (2) high-redshift galaxies (z greater than ~2), where even massive halos are fed by cold streams because the hot gas hasn't yet built up enough density to disrupt them. Most of the vigorous disk growth in the early universe is powered this way.
How do astronomers actually detect cold streams?
Directly imaging them is extremely hard because they're faint. The main tracers are extended Lyman-alpha emission (giant Lyman-alpha blobs powered partly by gravitational infall), and metal-poor, ~10^4 K hydrogen absorption lines seen against background quasars or the galaxy's own light. Cosmological simulations (GADGET, AREPO, FIRE) are essential for interpreting these signals.
How is a cold accretion stream different from an accretion disk?
They share the word 'accretion' but operate on vastly different scales. An accretion disk is gas orbiting a compact object like a black hole or protostar, spanning AU to parsecs. A cold accretion stream is a galaxy-scale inflow spanning ~100 kiloparsecs that delivers gas from the cosmic web toward the galaxy, where it may eventually settle into a disk.