Has light from the first stars after the big bang been detected?


Published: 13 April 2018 (GMT+10)

“Astronomers detect light from the Universe’s first stars” is the headline of a Nature news article, which appeared 28 February 2018.1 It relates to observations made by a team of astronomers led by Judd Bowman of Arizona State University in Tempe. The team published their results in Nature the same week.2 According to Bowman,

This is the first time we’ve seen any signal from this early in the Universe, aside from the afterglow of the Big Bang.

They used a small radio-telescope situated in the Western Australian desert, far away from human settlement to minimize interference from radio signals generated by human technology (figure 1). The antenna was tuned to a waveband of about 78 MHz, which is at the low end of FM radio, so isolation from human-generated radio signals was essential.

Figure 1. The small radio telescope in Western Australia used to detect evidence of light allegedly from the Universe’s first stars. Credit: CSIRO

To understand what the astronomers interpret from this research I quote an editorial summary from Nature:3

As the first stars heated hydrogen in the early Universe, the 21-cm hyperfine line—an astronomical standard that represents the spin-flip transition in the ground state of atomic hydrogen—was altered, causing the hydrogen gas to absorb photons from the microwave background. This should produce an observable absorption signal at frequencies of less than 200 megahertz (MHz). Judd Bowman and colleagues report the observation of an absorption profile centred at a frequency of 78 MHz that is about 19 MHz wide and 0.5 kelvin deep. The profile is generally in line with expectations, although it is deeper than predicted. An accompanying paper by Rennan Barkana suggests that baryons were interacting with cold dark-matter particles in the early Universe, cooling the gas more than had been expected.

Let’s look at this in two stages. What was observed and what is the interpretation of the recorded data.

What was observed

Astronomers recorded the spatially averaged, all-sky radio-frequency emissions received at their antenna for all frequencies between 50 MHz and 100 MHz. This was filtered and averaged over hundreds of hours. It was then represented as a brightness temperature (figure 2a).

The astronomers used an antenna tuned to a region less than 200 MHz because of cosmological theory. The theory dictates/predicts that the so-called ‘cosmic dawn’ occurred several hundred million years after the alleged big bang. This is the period in the big bang story that when sufficient neutral hydrogen clumped together the first stars turned on. According to standard ΛCDM cosmology, the ‘cosmic dawn’ ended at a redshift of about 20 and started some 100 million years before that.

The neutral atomic hydrogen of those first stars allegedly absorbed radiation from the big bang fireball (now called the cosmic microwave background (CMB) radiation). This resulted in spin-flips of the atomic hydrogen atom when the atoms absorbed radiation at a wavelength of 21 cm, or about 1420 MHz. As a result, the researchers looked for a 21-cm absorption dip in the sky-averaged radio-spectrum. But if you want to see this absorption dip today after the universe has allegedly expanded by a factor of about 20 you have to look down in the FM radio frequency band. Roughly said: divide 1420 MHz by 20 equals approximately 70 MHz. That means the 21-cm absorption feature has been redshifted down to about 70 MHz for a period in the big bang history characterized by a redshift z =19.4

The astronomers had to be sure the source of this radiation they were observing, together with its absorption feature, was not from Earth or the Galaxy itself. Both have radio emissions at these frequencies at much higher intensities than what was expected to be seen from the ‘cosmic dawn’. So they spent several years with a second antenna making sure it was not interference from Earth or the Galaxy.

And despite being a puny 0.1% drop in the radiation, it was still twice the magnitude predicted. The finding was so stark that the researchers spent two years checking that it didn’t come from an instrumental effect or noise. They even built a second antenna and pointed their instruments at different patches of sky at different times. “After two years, we passed all of these tests, and couldn’t find any alternative explanation,” says Bowman. “At that point, we started to feel excitement.1
Figure 2. Summary of detection. a, Measured spectrum for the reference dataset after filtering for data quality and radio-frequency interference. The spectrum is dominated by Galactic synchrotron emission. b, c, Residuals after fitting and removing only the foreground model (b) or the foreground and 21-cm models (c). d, Recovered model profile of the 21-cm absorption, with a signal-to-noise ratio of 37, amplitude of 0.53 K, centre frequency of 78.1 MHz and width of 18.7 MHz. e, Sum of the 21-cm model (d) and its residuals (c). Reproduced from Fig. 1 of Bowman et al.2

The researchers modelled all known foreground emissions from the Galaxy and Earth and subtracted those from the measured data to obtain the remaining spectrum from the alleged ‘cosmic dawn’ (figure 2, reproduced from the Nature paper).2 Because of the extremely large temperature range (1,000–5,000 K) of the measured data (figure 2a), to find the expected feature with only a ‘dip’ of 0.5 K it was necessary to subtract models of the 21-cm feature (figure 2d) and all expected foreground sources (the Galactic synchrotron spectrum and Earth’s ionosphere) until the residuals were minimized. This eventually reduced the root-mean-squared (r.m.s) residuals to 0.025 K (figure 2c). Assuming nothing was missed in the modelling, they concluded that the dip (figure 2d) was a real signal buried in the all-sky radio emission data (figure 2a).

I have it on good authority that the team was very careful not to make too rash a claim. For this reason, they spent so much time trying to verify that the ‘dip’ feature (figure 2e) was not from some spurious source. Bowman et al. argue that the profile of the ‘dip’ centred around 78 MHz is what was expected. But they were surprised that the feature had an extended flat bottom (figure 2d), and they had to introduce that into the model before they could achieve flat residuals across the region of interest (figure 2c). I have no doubt that they were very careful getting this result.

Could some other source be the origin of the feature? The authors state that “astrophysical phenomena (such as radiation from stars and stellar remnants) are unlikely to account for this discrepancy”.2 But there is no way one could definitely know that.

What is the interpretation

Regarding this experiment, I once asked an astronomer, how could he be sure that the 100 MHz radiation he was looking for was redshifted light from the first stars of the ‘cosmic dawn’? After all, could there not be other astrophysical sources? He replied, “where else could it come from?” Such is the mindset. The big bang is believed to be true, so once the Earth and the Galaxy are eliminated it could be from nowhere else.

The astronomers saw only a tiny 0.1% dip in the spectrum of the 100 MHz band radiation coming from the sky (compare figure 2d with figure 2a). They see two features in that tiny dip—the beginning of the dip and the ending of the dip. Assuming big bang expanding universe cosmology, they precisely calculate4 when, in the big bang timeline, these features occur by assuming the dip results from the physics of neutral hydrogen, though greatly redshifted. From this they infer that the ‘cosmic dawn’ started 180 million years after the big bang and concluded 70 million years later.

Radiation from this period arrives stretched out by the expansion of the Universe, meaning the band in which the signal was found gives away its age. This allowed the team to date the latest onset of the cosmic dawn to 180 million years after the Big Bang. The signal’s disappearance gives away a second milestone—when more-energetic X-rays from the deaths of the first stars raised the temperature of the gas and turned off the signal. Bowman’s team puts that time around 250 million years after the Big Bang.1

Is this verifiable? The astronomers expect that it will be, once more measurements are made by others using much larger telescopes with better precision. But that would only be measuring the same phenomenon. It is not getting different information on it by an independent method. I have been informed though that work is in progress to do just that—to use a different, more precise method, and to look across a wider region of frequencies as well as spatial position.

A more fundamental question might be, is there an independent method of verifying that the ‘dip’ seen centred at 78 MHz is a highly redshifted version of neutral hydrogen 21-cm absorption by starlight? Could it not be due to some other astronomical source which is not even at the alleged redshift? And how can you infer the 1420 MHz radiation absorbed by the hydrogen atoms was highly redshifted light from the big bang fireball?

One of the problems with this is that astronomers do not actually measure the redshift of the source, like is routinely done in measurements on stars and galaxies. In this case they see a small dip in the FM radio-frequency spectrum coming from the sky and they simply take for granted that it results from the assumed phenomenon. They do not question from where else this could arise. So they then use standard cosmological theory to calculate what redshift 78 MHz represents for a 1420 MHz absorption feature in the rest frame of a putative star in the alleged ‘cosmic dawn’ epoch. The result aligns quite well with their expectations, but of course that can be adjusted slightly depending on the result. And at that epoch, where else could the 1420 MHz radiation come from, as they believe the first stars were bathed in CMB radiation at that redshift? There was, supposedly, nothing else there then.

However, as noted by Charles Bennett in his letter to the editors of Science, and with which the editors agreed, “… science doesn’t ‘prove’ theories. Scientific measurements can only disprove theories or be consistent with them. Any theory that is consistent with measurements could be disproved by a future measurement.”5 In fact, concluding that an expected observation ‘proves’ a theory to be correct is known in formal logic as the fallacy of affirming the consequent. Thus, assuming that this observation categorically verifies that the assumed source of the phenomenon has been established, without first eliminating all other possible sources, is something of an overreach. It is similar to what happened with the BICEP2 announcement concerning the alleged detection of gravitational waves from the end of the alleged cosmic inflation period that had to be subsequently rescinded.6,7

Moreover, the depth of the ‘dip’ does not agree with expectation. It is twice what was expected. So how do they explain that? Well, in cosmology today that is easy. Just invoke dark matter. It is needed to get stars to form in the first place so there must have been plenty of it around. But it presents another problem.

Dark matter (up until now) by definition does not interact with normal (baryonic) matter. So they have to propose that there is again new physics here and that the dark matter in the early universe did interact with normal matter.

… of the proposed extensions to the standard model of cosmology and particle physics, only cooling of the gas as a result of interactions between dark matter and baryons seems to explain the observed amplitude.2

An accompanying paper by an Israeli physicist Rennan Barkana proposed the dark solution.8

… this absorption can be explained by the combination of radiation from the first stars and excess cooling of the cosmic gas induced by its interaction with dark matter.

It is only by modelling that such a statement can be made. No first stars are directly observed. Assuming it is cosmological in origin, it is the cosmic radio-frequency background spectrum that is observed, and it is inferred to arise from the first stars for the waveband near the 78 MHz absorption feature. Now it is necessary to assume that hydrogen gas was then cooled by the presence of dark matter particles, when all other inferred ‘observations’ of dark matter have meant that it does not interact with baryons (normal matter), or when it does it is extremely weakly interacting—so weak, in fact, that it has not been detectable in any laboratory experiment to date. The searched-for particles are even called WIMPs (Weakly Interacting Massive Particles).

The theorist Barkana proposed that

… the dark-matter particle is no heavier than several proton masses, well below the commonly predicted mass of weakly interacting massive particles.4

So it is a new type of dark matter, and it is quite cold, and able to cool the surrounding neutral hydrogen in the ‘cosmic dawn’. He proposes that “21-centimetre cosmology can be used as a dark-matter probe”, meaning this sort of ‘dip’ in the FM frequency band can be used to infer the existence of a new type of dark matter and to measure how energetic those particles are. However, this is all without ever detecting them in a laboratory experiment.

It was probably unfortunate that Nature published Barkana’s speculative paper along with the research that made the observations of the ‘dip’. The latter is based on real science whereas the dark matter interpretation is speculative at best. I am sure the theorists are now churning out new papers by the minute to explain this feature. Perhaps it would be better if they waited though until better data becomes available in the future.


The standard big bang ΛCDM model is assumed. The absorption ‘dip’ in the observed 100 MHz radiation, assumed to be from the cosmic background, is assumed to be the redshifted 21-cm line of neutral hydrogen and that is assumed to be bathed in light from primordial stars. In addition, the depth of the absorption line is assumed to have been modified by the presence of a new type of, even more speculative form of, dark matter; not the elusive WIMP that has been sought for in local laboratory experiments. This has led to the proposal of a dark matter particle being necessarily cold and of much lower mass than the putative WIMP.

I have written that cosmic inflation is built on a “house of cards”.9 This alleged ‘detection’ of light from the first stars is also assumption built on assumption. Pull out just one and the whole house collapses. Or could it turn into a whole new realm of cosmology? That is, something like the Emperor’s new clothes where no direct evidence is needed? We’ll have to wait and see.


  1. Gibney, E., Astronomers detect light from the Universe’s first stars, Nature News, nature.com/articles/d41586-018-02616-8, 28 February 2018. Return to text.
  2. Bowman, J.D., Rogers, A.E.E., Monsalve, R.A., Mozdzen, T.J., and Mahesh, N., An absorption profile centred at 78 megahertz in the sky-averaged spectrum, Nature 555: 67–70, 1 March 2018. Return to text.
  3. Editorial Summary found at www.nature.com/articles/nature25792 Return to text.
  4. More precisely the expected frequency = 1420 MHz/(1+z) where z is the source redshift. Return to text.
  5. Bennett, C.L., “Science Title Misstep,” Science 332:1263, 2011. Return to text.
  6. Hartnett, J.G., The authors of the claimed biggest astrophysics discovery of the century admit they may have been wrong, 3 July 2014. Return to text.
  7. Hartnett, J.G., New study confirms BICEP2 detection of cosmic inflation wrong, February 5, 2015. Return to text.
  8. Barkana, R., Possible interaction between baryons and dark-matter particles revealed by the first stars, Nature 555: 71–74, 1 March 2018. Return to text.
  9. Hartnett, J.G., Cosmic Inflation: Did it really happen? September 11, 2015. Return to text.

Helpful Resources

Taking Back Astronomy
by Jason Lisle
US $17.00
Dismantling the Big Bang
by Alex Williams, John Hartnett
US $17.00
Comment on this article

Reader’s comments

Russ H.
Thanks for helping me decipher this news, Dr. Hartnett. I might have put more faith in the team's interpretation if they had explained the 18.7 MHZ width of the dip, and its flatness. I would expect an absorption line to have a Gaussian (more rounded) shape and be much narrower. The dip's width is a large percentage (24%) of its center frequency, which seems remarkably broad to me. Did I miss some explanation in the articles (yours or theirs)?
John Hartnett
Dr Humphreys, You make a very good observation. The research team tried a standard Gaussian functional form in their fitting function for the line shape. But they found that they had to add a broadening function to the shape of the 'dip' in order to minimise the residuals. This really surprised them as it was far from their expectations. They might now expect that in the future when they implement an interferometric survey with a wider frequency range and with spatial information that they can get a more accurate line shape. However it might also show other unexpected results.
Peter T.
I love reading your articles! This science reminded me of myself when I was a child. I would look up in the sky and study the clouds, and lo and behold, if I wanted to see the shape of a horse in the clouds, if I looked long and hard enough, wow there it was!. The only science that I trust doesn't set what the outcome is and then measure to make the data fit, so it meets expectations! I wish that the majority of the "enlightened" generation of Scientists that exist in the world today would use their talent on that which is known, and help to engineer solutions to today's world problems, such as ridding hunger and poverty, but unfortunately Jesus has already told us that the poor will always be with us. So, I guess the secular scientists will keep making up their sometimes outrageous theories, which can never be proven, but will have Governments all over the world throwing money at them to help prove their "snake oil". I still pray for the Lord to have mercy on all of us as our country continues to turn its back on Him.
Lenny L.
It's incredible how the science community put so much faith in minute particles "billions and billions" (in the cadence of the late Carl Sagan) of light years away, predict exactly how it happened and laugh at us for faith in things unseen. I thought "Science" encompassed the repeatable and observational premise in its model. Can they repeat the Big Bang to test their hypotheses? Of course not! Can they observe things light years away and make a pronouncement? Of course not! Without even waxing theological against their argument, can they honestly say that in a few thousand years when technology further refines itself to even higher levels that there current findings may be so far off base, their assumptions may be the jokes of that generation of scientists. But the one thing that IS certain, they have "faith"--and nothing more--that they are right. The same as we do . . .
Thomas C.
I would be interested in a second experiment exploring the same phenome from a space platform. That way the hydrogen content of atmospheric water could be ruled out.
John G Hartnett
The measurements from a space platform is not very relevant because the radiation from hydrogen in the atmosphere is at much higher frequencies. The spin-flip of neutral hydrogen produces absorption at 21 cm (1420 MHz). The 'dip' was observed at 78 MHz, and their assumption is that that is the result of neutral hydrogen but greatly redshifted from 1420 MHz --> 78 MHz.
Joe B.
I have just a comment about the wave-shape of the ‘apparent’ absorption pattern. It looks more like that of a two-pole band-stop filter. I had to dig to find the dimensions of the antenna (here: arXiv:1510.03477v1 [astro-ph.IM]) and it’s tuned to about twice the frequency of the center of the dip (~78MHz) as a ½ wavelength dipole. I did not see any references to how the performance of the antenna behaves in the traditional sense (measure it as a radiator and generate the appropriate Smith charts) or how it behaved installed on site, but shielded (small Faraday cage connected to the ground plane would do). I am not sure how they can ascertain how much of what they are seeing is system noise, random background noise or even environmental noise and if the apparent absorption is actually real. Simulations and laboratory measurements are not enough (IMHO).