News & Views | Published:


Galaxies in from the cold

Nature volume 457, pages 388389 (22 January 2009) | Download Citation

Computer simulations of the cosmos suggest that cold streams of gas could underlie the unexpectedly high star-formation activity of many massive galaxies found to exist a few billion years after the Big Bang.

Recent surveys of galaxies1 have found evidence that galaxies with masses comparable to or greater than that of the Milky Way were already present in large numbers about 3 billion years after the Big Bang. What's more, a significant fraction of these massive galaxies seem to have been gas-rich, rotating disks in which stars formed at a rate of up to 150 solar masses per year, 50 times the rate in the present-day Milky Way2,3. Most of these star-forming galaxies do not seem to be the aftermath of mergers of smaller systems2, and are found to produce stars steadily over a long period of time4 — characteristics that are at odds with the prevailing view of how galaxies form and evolve. So, if mergers are not the cause, what else could trigger the formation of stars in these massive galaxies? Building on earlier work5,6,7, Dekel et al.8 (page 451 of this issue) use cosmological simulations7 to show that the galaxies might grow and form stars as a result of being fed by rapid, narrow streams of cold gas.

In the classical picture of their formation9,10, galaxies are created when gas cools and collects at the centres of collapsing haloes of dark matter. They then evolve to form larger galaxies by merging with smaller galaxies. But the fundamental properties of atomic cooling divide evolving galaxies into two branches. Galaxies with a dark-matter mass below a critical value of about half the mass of the Milky Way (about 500 billion solar masses) would have grown rapidly through accretion of cold gas, resulting in sustained (or continuous) star formation. In contrast, massive galaxies would have grown at a much slower pace (determined by the gas cooling rate), and bursts of star formation would have occurred only when parent galaxies of comparable masses underwent intense, rapid mergers (known as major mergers).

The classical picture thus predicts that active star-forming galaxies of low mass should already have been abundant at early epochs in the history of the Universe, whereas massive galaxies should have assembled later through mergers of smaller systems. But Dekel et al.8 challenge this picture using substantially improved, high-resolution hydrodynamical simulations7 representing a large volume of the cosmos.

Perhaps the most notable aspect of Dekel and co-workers' study is the indication that, at early epochs (redshift ≥ 2), haloes with a mass substantially above the average for that epoch tend to form at the dense nodes of the 'cosmic web' of dark matter, which comprises long filaments of denser gas connecting these nodes. As a result, much of the gas in the filaments remains cold and flows at high speed from large distances deep into the halo, near the evolving disk in which stars will subsequently form. Under these conditions, massive galaxies above the critical mass can grow rapidly and steadily.

Dekel et al. find, however, that two other requirements must be met to explain the high star-formation rates of massive, early-epoch galaxies. The first is that the accretion of material must be largely gaseous, with only a small fraction of stars involved. The second is that the conversion of accreted gas into stars must be highly efficient. Because most of the gas in the streams is smooth or exists in low-mass clumps and smaller (satellite) galaxies, the first criterion means that the star-formation efficiency in the streams must be low. Qualitatively, this might be plausible if massive stars that did manage to form in the streams inject energy back into the interstellar medium through supernova explosions and stellar winds. Such 'stellar feedback' can thus disperse the surrounding gas and halt further star formation in the streams, and it would do so more effectively in lower-mass systems because of their lower gravitational binding energy. In addition, the fraction of dense molecular gas required to form stars may be lower in these lower-mass clumps, which contain fewer heavy elements and have less shielding against destructive ultraviolet radiation from hot stars and the intergalactic radiation field.

Although star formation in present-day galaxies is inefficient, the necessary high star-formation rates at early epochs might be reached if star formation consumes piled-up gas at the same rate as it is deposited. This requires that the fraction of gas ejected from the evolving massive galaxy by stellar winds and supernovae is modest, which might be at odds with observations11. Alternatively, or additionally, a higher efficiency of star formation or a different distribution of stellar masses may be required in the early-epoch systems compared with those in the present-day Universe. Dekel and colleagues' simulations cannot yet resolve the complex interaction between the gas inflow and the disk, which might answer many of the remaining questions. Much more detailed simulations will be required to correctly model the radiation, energy balance, dynamical structure and star formation in the embryonic disks.

It is nevertheless tempting to conclude that the cold streams hypothesized by Dekel et al.8 can explain the formation of the early-epoch massive disks. If so, what happens next in the process of galaxy formation? Gas-rich, turbulent disks are unstable, and prone to fragmentation and the formation of massive star-forming clumps of gas12, in agreement with observations3. Dynamical friction then forces the clumps to spiral rapidly into the centre of the galaxy, forming a central bulge surrounded by a remnant disk12, whose present-day relic may be the old 'thick disk' component seen in nearby galaxies. Gas accretion decreases naturally at later epochs, perhaps aided by the energy injection of massive black holes that form at the galactic centre. Disk turbulence then subsides, and a maturing thin disk can plausibly grow at redshift ≤1. Depending on whether or not a major merger occurs during this period, the end-product might be a massive elliptical or disk galaxy. The work by Dekel and co-workers may turn out to be a decisive stepping stone in elucidating the origin of these massive galaxies.


  1. 1.

    et al. Astron. Astrophys. 459, 745–757 (2006).

  2. 2.

    et al. Astrophys. J. 682, 231–251 (2008).

  3. 3.

    et al. Nature 442, 786–789 (2006).

  4. 4.

    et al. Astrophys. J. 670, 156–172 (2007).

  5. 5.

    & Mon. Not. R. Astron. Soc. 368, 2–20 (2006).

  6. 6.

    , , & Mon. Not. R. Astron. Soc. 363, 2–28 (2005).

  7. 7.

    , & Mon. Not. R. Astron. Soc. 390, 1326–1338 (2008).

  8. 8.

    et al. Nature 457, 451–454 (2009).

  9. 9.

    & Mon. Not. R. Astron. Soc. 179, 541–559 (1977).

  10. 10.

    & Mon. Not. R. Astron. Soc. 183, 341–358 (1978).

  11. 11.

    Astrophys. J. 674, 151–156 (2008).

  12. 12.

    , & Astrophys. J. 670, 237–248 (2007).

Download references

Author information


  1. Reinhard Genzel is at the Max Planck Institute for Extraterrestrial Physics, 85748 Garching, Germany, and in the Department of Physics, University of California, Berkeley, USA.

    • Reinhard Genzel


  1. Search for Reinhard Genzel in:

About this article

Publication history




By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing