Wednesday, 21 March 2018

21cm to dark matter

The EDGES discovery of the 21cm absorption line at the cosmic dawn has been widely discussed on blogs and in popular press. Quite deservedly so.  The observation opens a new window on the epoch when the universe as we know it was just beginning. We expect a treasure trove of information about the standard processes happening in the early universe, as well as novel constraints on hypothetical particles that might have been present then. It is not a very long shot to speculate that, if confirmed, the EDGES discovery will be awarded a Nobel prize. On the other hand, the bold claim bundled with their experimental result -  that the unexpectedly large strength of the signal is an indication of interaction between the ordinary matter and cold dark matter - is very controversial. 

But before jumping to dark matter it is worth reviewing the standard physics leading to the EDGES signal. In the lowest energy (singlet) state, hydrogen may absorb a photon and jump to a slightly excited (triplet) state which differs from the true ground state just by the arrangement of the proton and electron spins. Such transitions are induced by photons of wavelength of 21cm, or frequency of 1.4 GHz, or energy of 5.9 𝜇eV, and they may routinely occur at the cosmic dawn when Cosmic Microwave Background (CMB) photons of the right energy hit neutral hydrogen atoms hovering in the universe. The evolution of the CMB and hydrogen temperatures is shown in the picture here as a function of the cosmological redshift z (large z is early time, z=0 is today). The CMB temperature is red and it decreases with time as (1+z) due to the expansion of the universe. The hydrogen temperature in blue is a bit more tricky. At the recombination time around z=1100 most proton and electrons combine to form neutral atoms, however a small fraction of free electrons and protons survives. Interactions between the electrons and CMB photons via Compton scattering are strong enough to keep the two (and consequently the hydrogen as well) at equal temperatures for some time.  However, around z=200 the CMB and hydrogen temperatures decouple, and the latter subsequently decreases much faster with time, as (1+z)^2. At the cosmic dawn, z~17, the hydrogen gas is already 7 times colder than the CMB, after which light from the first stars heats it up and ionizes it again.

The quantity directly relevant for the 21cm absorption signal is the so-called spin temperature Ts, which is a measure of the relative occupation number of the singlet and triplet hydrogen states. Just before the cosmic dawn, the spin temperature equals the CMB one, and as  a result there is no net absorption or emission of 21cm photons. However, it is believed that the light from the first stars initially lowers the spin temperature down to the hydrogen one. Therefore, there should be absorption of 21cm CMB photons by the hydrogen in the epoch between z~20 and z~15. After taking into account the cosmological redshift, one should now observe a dip in the radio frequencies between 70 and 90 MHz. This is roughly what EDGES finds. The depth of the dip is described by the formula:
 As the spin temperature cannot be lower than that of the hydrogen, the standard physics predicts TCMB/Ts ≼ 7 corresponding  T21 ≽ -0.2K. The surprise is that EDGES observes a larger dip, T21 ≈ -0.5K, 3.8 astrosigma away from the predicted value, as if TCMB/Ts were of order 15.

If the EDGES result is taken at face value, it means that TCMB/Ts at the cosmic dawn was much larger than predicted in the standard scenario.  Either there was a lot more photon radiation at the relevant wavelengths, or the hydrogen gas was much colder than predicted. Focusing on the latter possibility, one could imagine that the hydrogen was cooled due to interactions with cold dark matter  made of relatively light (less than GeV) particles. However, this idea very difficult to realize in practice, because it requires the interaction cross section to be thousands of barns at the relevant epoch! Not picobarns typical for WIMPs. Many orders of magnitude more than the total proton-proton cross section at the LHC. Even in nuclear processes such values are rarely seen.  And we are talking here about dark matter, whose trademark is interacting weakly.   Obviously, the idea is running into all sorts of constraints that have been laboriously accumulated over the years.     
One can try to save this idea by a series of evasive tricks. If the interaction cross section scales as 1/v^4, where v is the relative velocity between colliding matter and dark matter particles, it could be enhanced at the cosmic dawn when the typical velocities were at its minimum. The 1/v^4 behavior is not unfamiliar, as it is characteristic of the electromagnetic forces in the non-relativistic limit. Thus, one could envisage a model where dark matter has a minuscule electric charge, one thousandth or less that of the proton. This trick buys some mileage, but the obstacles remain enormous. The cross section is still large enough for the dark and ordinary matter to couple strongly during the recombination epoch, contrary to what is concluded from precision observations of the CMB. Therefore the milli-charge particles can constitute only  a small fraction of dark matter, less then 1 percent. Finally, one needs to avoid constraints from direct detection, colliders, and emission by stars and supernovae.  A plot borrowed from this paper shows that a tiny region of viable parameter space remains around 100 MeV mass and 10^-5 charge, though my guess is that this will also go away upon a more careful analysis.

So, milli-charge dark matter cooling hydrogen does not stand scrutiny as an explanation for the EDGES anomaly. This does not mean that all exotic explanations must be so implausible. Better models are being and will be proposed, and one of them could even be correct. For example, models where new particles lead to an injection of additional 21cm photons at early times seem to be more encouraging.  My bet? Future observations will confirm the 21cm absorption signal, but the amplitude and other features will turn out to be consistent with the standard 𝞚CDM predictions. Given the number of competing experiments in the starting blocks, the issue should be clarified within the next few years. What is certain is that, this time,  we will learn a lot whether or not the anomalous signal persists :)


neo said...

can you comment on Stacy McGaugh's explanation of an all baryon no dark matter universe as the explanation?

Strong Hydrogen Absorption at Cosmic Dawn: the Signature of a Baryonic Universe
Stacy McGaugh

andrew said...

The link to the paper mentioned by neo is

Unknown said...

Stacy's paper observes that one would get the correct amplitude of the EDGES signal if the total matter density in the universe were equal to the baryonic density measured in CMB and BBN, OmegaM = OmegaB = 0.05. This is an interesting fact to keep in mind. However, the challenge here is not to explain one data point, but to provide an explanation consistent with all existing experimental constraints. Just fixing OmegaM = OmegaB = 0.05 without modifying other physics does not give you a complete model satisfying the existing constraints. So the paper is not at the same footing as the other ones.

neo said...

What other existing constrains do you have in mind? There was another paper which shows a chart how incredibly difficult it is to get that value from any DM value. Andrew can provide the link.

Unknown said...

CMB and BBN in the first place. Then the entire cosmological history. OmegaM = OmegaB does not work unless you add more particles and/or interactions. Then you have to satisfy constraints on the particles and forces you introduced (just like DM models with milli-charged particles have to satisfy the supernovae and direct detection constraints). You need a complete model, otherwise you cannot even fail.

neo said...

so you've not found any problems with McGaugh's equations in the paper I ref and Andrew linked, good.

regarding BBN - dark matter has lithium 7 problem ref

this paper suggests EDGES eliminates most of the parameter space for viable dark matter models

btw was there any new LHC physics results presented in Moriond EW 2018

Anonymous said...

Hi, thank you for this very interesting post. I have a very basic doubt regarding the interpretation of the results and the physical process involved. If you could take a few minutes to resolve it I would be most grateful.

As I understand, the absorption of Lyman alpha photons from the first stars by the neutral hydrogen drives electronic transitions that ultimately *increase* the number of hydrogen atoms in which the electron is in the higher energy level of the fundamental level hyperfine structure. This is what is called Lyman alpha pumping, I guess. In the equation for T21 this would correspond to lowering the Ts - i.e., Ts is lower when the higher energy is more populated (weird) - to produce an absorption (negative value for T21). However, pumping should make less hydrogen atoms available to absorb 21cm photons from the CMB and thus absorption of such photons should decrease instead of increasing to form the now observed dip at the expansion corrected frequency between 70-90 MHz.

I have some key idea wrong somewhere in this argument but despite some effort have been unable to pin it down.

Thank you in advance for your time.

Best regards,

Das Wildschwein said...

Luis: the Lyman alpha pumping in this case DECREASES the number of Hydrogen atoms in the triplet state and hence decreases the spin temperature. This is because the Baryon kinetic energy is colder than the CMB and the Lyman alpha "colour temperature" is coupled to the Baryon kinetic energy due to recoils. The spin temperature starts at the CMB temperature before significant numbers of Lyman alpha photons are produced. See 1109.6012.

Edwin Steiner said...

Dear Jester,

Thank you for posting this nice summary. It helps to understand the context of this exciting measurement.

Looking at the paper, I noticed that the improvement in residual RMS they get does not look too impressive for such a large signal. The simultaneous match giving residuals in Fig 1c reduces RMS residuals by 0.062 K, while I estimate that adding back the (centered) signal would change RMS by roughly 0.15 K (assuming zero correlation of signal and residuals in Fig 1c). I think this indicates that a significant part of the signal "comes" from the change of foreground parameterization, i.e. the correlation of the signal with the difference of foreground models between Fig 1b and Fig 1c is rather large. That this is plausible is confirmed by Extended Data Fig 8, which shows that the foreground models actually are not too bad at matching the signal. Unfortunately they did not plot the difference in foreground models between Fig 1b and 1c, as far as I see.

Therefore I agree with you that a reduction of the signal amplitude with further analysis would not be unexpected. I am a bit surprised that the authors are so confident that they can exclude the Standard Model prediction.

Another concern I'd have is that the signal is mostly "in a low Fourier mode" of the observed brightness temperature function as it is wide and close to sinusoidal, so it is not too distinctive. They checked that there is no similar signal in the 90-200 MHz range, though, so that diminishes this concern a bit.

Anonymous said...

Das Wildschwein, thank you for clearing this up for me and for the paper reference.



Unknown said...

Thanks Wildschwein and Louis. I haven't yet studied the details of how the spin temperature evolves, so I only wrote a vague sentence in the post. The reference Wildschwein mentions looks very useful to understand it better.

Anonymous said...

Regarding the photon-injection way, should not it lead to photon over-abundance at the surrounding regions? It seams that at the respective CMB tail, at cca 1.4 GHz this && that suggest otherwise. Even though this says that there is some rise detected, though at above frequencies.
May be that the (over)decrease only occurs when the hydrogen line is measured, at the Dawn, now, or for any suitable z value.