A surprising chill before the cosmic dawn

An experiment to estimate when stars began to form in the Universe suggests that gas temperatures just before stars appeared had fallen well below predicted limits, and that dark matter is not as shadowy as was thought.
Lincoln Greenhill is in the Radio and Geoastronomy Division, Smithsonian Astrophysical Observatory, Cambridge, Massachusetts 02138, USA, and in the Department of Astronomy, Harvard University, Cambridge.

Search for this author in:

The first stars to form generated copious fluxes of ultraviolet radiation that suffused the early Universe — a phenomenon referred to as the cosmic dawn. Many calculations have been performed to estimate when this occurred1, but no data-driven constraints on the timing have been available. In a paper in Nature, Bowman et al.2 report what might be the first detection of the thermal footprints of these stars, tracking back to 180 million years after the Big Bang.

Less than one million years after the Big Bang, the Universe consisted of atomic gas (chiefly hydrogen) and a form of matter that outweighs regular matter by more than five times3 but has yet to be seen directly. Measurements over decades have indicated that, oddly enough, this ‘dark’ matter interacts with itself and with regular matter only through the action of gravity. It was mainly the gravity of dark matter that amplified small, localized density perturbations in the Universe shortly after the Big Bang to generate the first large-scale structures. But it was the hydrogen within these perturbations that collapsed piecemeal to form stars, bringing about the cosmic dawn.

The observable thermal footprints of early stars derive from small variations in the ratio of the number of interstellar hydrogen atoms found in two particular energy states; a transition between these states causes a photon to be emitted or absorbed at a characteristic radio frequency. The ratio reflects the degree of excitation of the hydrogen, and can be expressed as a temperature, known as the atomic spin temperature (TS).

At early times, when the Universe was relatively small and mean gas density was high, collisions between atoms were frequent. TS was therefore the same as the kinetic temperature of the gas (TG), an indicator of the energy available to excite atoms through collisions. By the time stars began to form, the Universe had expanded. Both TG and mean gas densities had fallen, and collisions were infrequent, allowing TS to drift upward to the temperature of the radiation (TR) left over from the Big Bang (Fig. 1). TR also fell as the Universe expanded, but not as quickly as TG.

Figure 1 | Temperature changes during the evolution of the Universe. The first two phases of the Universe were the ‘dark age’, before stars formed (grey), and the cosmic dawn (yellow), when clouds of hydrogen in early structures collapsed to form stars. The temperature of radiation (TR) left over from the Big Bang has declined slowly over time. The spin temperature (TS) of hydrogen that has not formed stars reflects the excitation state of the hydrogen atoms (solid blue line shows previous estimates of TS based on models). Bowman et al.2 use observations to estimate TS, and find that it dropped to lower values (red solid line) than predicted by models. Barkana7 proposes that this could be evidence for a previously unrecognized, non-gravitational interaction between normal and dark matter. Such interactions would mean that the ‘kinetic’ temperature of gas (TG) in the Universe also dropped to a lower minimum (red dashed line) than is predicted by known physics (blue dashed line). (Adapted from ref. 12.)

A long-standing theory4,5 that still awaits testing predicts that absorption of UV radiation from early stars by nearby clouds of hydrogen could have driven TS back down to TG, but not lower. In other words, the cosmic dawn would make the gas seem colder when observed at radio frequencies. This would create an absorption feature in the spectrum of the background radiation left over from the Big Bang.

Bowman et al. now report the possible detection of just such an absorption signal. The authors measured TS, averaged over much of the sky and over a contiguous range of radio frequencies; each frequency provides a window on a different time in the Universe’s past. The measurement is very difficult because it must be performed using an extremely well-calibrated VHF radio antenna and receiver, to enable the weak cosmological signal to be separated from much stronger celestial signals and from those within the electronics systems of the apparatus used.

The putative absorption signal extends over a wide frequency range, one end of which looks as far back as 180 million years after the Big Bang, in good agreement with theoretical predictions6. Remarkably, however, the peak amplitude of the absorption is two to three times larger than predicted by the most optimistic models, and the absorption profile is flat-bottomed, rather than curvilinear and Gaussian-like, which is also at odds with models.

So how can the differences from the models be explained? In another paper in Nature, Barkana7 argues that models could achieve the reported signal amplitude and profile if non-gravitational interactions — like those that occur between charged particles — occur between dark matter and normal-matter particles, and if the dark-matter particles have relatively low masses and velocities that are less than the speed of light. The effects of variously hypothesized types of dark matter have been calculated previously8,9, but only those in which dark matter and normal matter scatter each other increase the magnitude of the absorption signature. The idea that a detectable radio signal from the cosmic dawn can be connected to the particle properties of dark matter suggests a potentially revolutionary angle for exploring fundamental physics.

Bowman and colleagues’ claim to have detected the long-sought absorption signal is bolstered by myriad tests in which the authors altered their experimental hardware or data analysis, in a concerted effort to identify systematic errors that might be responsible for the measured signal. The tests included repeating the data acquisition and analysis using a duplicate antenna at a second, nearby location; orienting the antenna at different angles with respect to the compass; and changing the ways in which the antenna is isolated from the ground. Other tests focused on switching various facets of the data calibration on and off.

However, the most stringent test will be to compare the current results with those to come from independent experiments also aimed at detecting the cosmic-dawn signal10,11. I hope that the unexpected amplitude and line shape of the reported absorption signal is indeed a hard-won breakthrough that reveals evidence of unexpected physics. But it is possible that systematic errors have escaped detection by the tests that were run. Two extensions to the reported tests include using circuitry that more precisely imitates the antenna than Bowman and colleagues’ circuitry when attached to the receiver during performance evaluation and calibration, and the cross-checking of performance models for the antenna (which are currently based on computer simulations of antenna electromagnetics) with field measurements made when narrowband or sinusoidal signals are broadcast near the antenna.

Bowman and co-workers’ report will be recognized as a milestone for this nascent experimental field: the first reputable claim of a much-anticipated detection. The follow-up will not be limited to ever finer interpretations of increasingly accurate one-dimensional spectra. Studies of the cosmic-dawn signal using interferometers (arrays of antennas) could describe the 3D structure of the Universe at that time and, by extrapolation, during the primordial ‘dark age’ when large-scale structure in the Universe first formed. One of Barkana’s particularly notable predictions is that, if non-gravitational interactions between normal and dark matter do exist, then the absorption signal detectable by interferometers could be stronger and more distinctive than had been predicted. It would encode the spatial fluctuations of matter density that occurred during the dark age, rather than just gas temperature, thus presenting new opportunities for tests of fundamental physics.

Nature 555, 38-39 (2018)

doi: 10.1038/d41586-018-02310-9

Updates & Corrections

  • Correction 09 March 2018: In the News & Views article 'A surprising chill before the cosmic dawn' by Lincoln Greenhill (Nature 555, 38–39; 2018), the seventh paragraph incorrectly said that a putative absorption signal looks back as far as 180 million years ago. This should have said "as far back as 180 million years after the Big Bang".


  1. 1.

    Pritchard, J. R. & Loeb, A. Rep. Prog. Phys. 75, 086901–086935 (2012).

  2. 2.

    Bowman, J. D., Rogers, A. E. E., Monsalve, R. A., Mozdzen, T. J. & Mahesh, N. Nature 555, 67–70 (2018).

  3. 3.

    Ade, P. A. R. et al. Astron. Astrophys. 571, A16 (2014).

  4. 4.

    Wouthuysen, S. A. Astrophys. J. 57, 31–32 (1952).

  5. 5.

    Field, G. B. Astrophys. J. 129, 536–550 (1959).

  6. 6.

    Cohen, A., Fialkov, A., Barkana, R. & Lotem, M. Mon. Not. R. Astron. Soc. 472, 1915–1931 (2017).

  7. 7.

    Barkana, R. Nature 555, 71–74 (2018).

  8. 8.

    Evoli, C., Mesinger, A. & Ferrara, A. J. Cosmol. Astropart. Phys. 2014, 024 (2014).

  9. 9.

    Tashiro, H., Kadota, K. & Silk, J. Phys. Rev. D 90, 083522 (2014).

  10. 10.

    Price, D. C. et al. Preprint at (2017).

  11. 11.

    Singh, S. et al. Preprint at (2017).

  12. 12.

    Pritchard, J. & Loeb, A. Phys. Rev. D 78, 103511–103516 (2008).

Download references

Nature Briefing

Sign up for the daily Nature Briefing email newsletter

Stay up to date with what matters in science and why, handpicked from Nature and other publications worldwide.

Sign Up