Cosmic Dawn

Cosmic Dawn refers to the time in the Universe’s past when the first stars and galaxies burst into life. We know such a time exists, simply because we know there was a time without stars in the early Universe, and we know there are stars now. At some point, the first stars came into being!

The “Epoch of Reionization” (or EoR) refers to a period that probably occurred just after Cosmic Dawn. Unlike the Cosmic Dawn, which is defined by when the first stars started to shine, the EoR is defined by what the first galaxies did to the surrounding gas — their radiation ionized the gas (i.e., gave the Hydrogen atoms enough energy that their single electron was freed from the nucleus). Again, we know this happened at some point, because the early Universe definitely had predominantly non-ionized Hydrogen, and the current Universe predominantly has ionized Hydrogen. However, exactly how it happened is up for some debate — was it early galaxies, or early black holes, or something else that caused the ionization?

The Standard Picture

The standard picture goes like this: the early Universe (just after “recombination”, or the Cosmic Microwave Background) had essentially three components: a whole bunch of dark matter (about 85% of the stuff), a lot of neutral Hydrogen (i.e. simple atoms with a single proton and single electron), and a tiny bit of Helium. Everything was cold and neutral and really not doing much at all. There was about the same amount of stuff everywhere. But small differences can end up making a big difference — regions of the Universe that were slightly overdense compared to the rest had stronger gravitational pull, and pulled surrounding material in, thereby becoming more overdense and pulling more in, and… well, you get the picture. This runaway process of “gravitational collapse” led to the eventual formation of “dark matter halos” — essentially self-contained clumps of dark matter.

Of course, the gas (remember, the “gas” really just means “Hydrogen” at this point) was also gravitationally attracted, and got pulled into the same halos. In fact, because gas interacts with itself (i.e. two gas particles can bump each other, unlike dark matter, probably) it tends to fall down further into the halo because it has a way to lose its energy. Eventually, a whole lot of Hydrogen congregates in a fairly small space, all jostling together, and voila, you have a star!

These earliest stars — composed almost purely of Hydrogen — were kinda unique. Stars today contain an abundance of “metals” (for an astronomer, “metals” are basically any element heavier than Helium). We call this first generation “Population III” stars, and we know very little about them (having never actually observed one), except that we expect them to be huge, hot and short-lived. The key thing is that the process of burning their fuel via nuclear fusion created heavier elements. when they finished burning up their fuel, they would have exploded as supernovae, and released all those heavier elements (like Carbon) back in the surrounding gas, ready for the next generation of stars to use.

We’ve been thinking about the stars themselves, but what about the surrounding gas? On the whole, it’s still just sitting around being neutral hydrogen. However, once the stars form, and multiple stars clump together into galaxies, the radiation they produce starts to penetrate the surrounding gas. This radiation is produced at different energies — some of the photons are high-energy X-rays, and some are lower-energy UV photons. X-rays tend to permeate the gas and get further out before they really interact. When they do, they heat up the gas. UV photons, on the other hand, can be at just the right energy to bump an electron off a Hydrogen atom, causing it to ionize. What we expect, then, is for pockets of gas around individual stars and galaxies to become ionized, and for these pockets to steadily grow in size as more radiation is emitted. The pockets will grow and start to overlap, and eventually, all of the gas will be ionized. This transition from essentially none of the gas being ionized, through the patchy reionization process, to the eventual full reionization of the Universe, actually probably happens fairly rapidly in the grand scheme of things. We’re fairly confident (due to the Lyman-alpha forest and other observables) that it concluded about 12.7 billion years ago (or about 1 billion years after the Big Bang). But all of these details are still up for debate — we haven’t yet observed it!

Observing the Cosmic Dawn and EoR

Often when people ask me what I do, I tell them “I search for the first stars in the Universe”. I say that mostly because it’s fairly understandable (and sounds much cooler than “I write computer programs to process scientific data”). In reality, we have little to no hope of observing the first stars directly. The light from those stars has been traveling for 13 billion years or more — the distance it has traveled is astonishingly large! That makes the stars incredibly faint to us, far too faint for even our largest telescopes.

But that doesn’t mean we have to give up. Although we can’t see those tiny pin-prick stars, we do have a hope of seeing radiation from the vast sea of gas surrounding them! The key here is that neutral hydrogen atoms (recall, that means hydrogen with a proton and electron) every now and then randomly emit a tiny bit of radiation when the electron flips its “spin” state to match the proton’s (the inverse happens — i.e. it absorbs radiation — when it flips to a non-matching spin state). Due to the magic of nature, or quantum mechanics, this emitted photon always has the same energy, and therefore frequency/wavelength — close to 21 cm. Fortunately for us, this is one of the only atomic transitions that we expect to emit at this wavelength.

This “hyperfine spin-flip transition” occurs incredibly rarely. On the other hand, there are a lot of hydrogen atoms in the Universe, and so we still expect to see quite a bit of 21 cm radiation. Now, as the Universe expands, so does the radiation. In fact, radiation emitted during the EoR should have stretched out by a factor of about 10 by the time we see it, and therefore it should be arriving at wavelengths of well more than a metre. These wavelengths are firmly in the low-frequency radio regime — about 100 MHz. Coincidentally, so are popular radio stations (this is a big problem!).

In fact, it’s more interesting than that. Since we know the exact wavelength the radiation was emitted at, if we observe such radiation at a given wavelength (say 2.1 m) we know exactly how much the Universe has expanded since that radiation was emitted (10x). Since we have a pretty good idea about how Universal expansion occurs throughout cosmic time, this means we can place a time on that emission.

By looking at the sky with large radio telescopes, tuned to frequencies around 100 MHz, we expect to receive this radiation from the neutral hydrogen gas in the early Universe. But this is not really the exciting thing. The exciting thing is that only neutral hydrogen emits this radiation. This makes sense, because ionized hydrogen doesn’t have an electron to spin-flip! Thus, we expect that early in the Universe, we should see this radiation, and later in the Universe’s evolution, we should no longer see it. Since the observing wavelength corresponds to the time of emission, we can rather say that we should see this radiation on large wavelengths, but not on short wavelengths. In fact, the wavelength at which we stop seeing it tells us when the EoR occurred (in principle).

Easy. Done, right?

Well, there are all kinds of complications that make this really hard to do in practice.

Challenges

There are two big problems when trying to detect the reionization of hydrogen in this way. Firstly, the signal itself is still super weak, despite all those Hydrogen atoms. To have a chance of detecting it requires really big telescopes looking for a long time. This in turn means huge volumes of data (we’re talking petabytes).

Secondly, and more importantly, while I said that the Hydrogen spin-flip transition was unique in emitting at 21 cm, that was only true of atomic transitions. There are other processes which also emit radiation, and can easily emit at 21 cm or longer wavelengths. These processes happen in all galaxies, including our own. Obviously, our own Galaxy is a lot closer to us than the gas around the first stars. Our Galaxy radiates extremely strongly at 100 MHz, about 10,000 times more strongly than the radiation we expect from the early Universe. At the same time, we thought it was a good idea to setup radio stations at about 100 MHz. These broadcast even more strongly than the Galaxy (which is why you can hear them when you tune in from you car, rather than hearing the Galaxy, which would sound either weird or boring). So, how can we tell that what we’re seeing is in fact from the early Universe, not our own Galaxy or Hits FM?

Well, as for the radio stations (we call this “Radio Frequency Interference” or RFI), we just build our telescopes as far away from civilization as we can (like Western Australia, the South African desert, or, one day, on the far side of the moon). Even there, the reflection of RFI off planes and back into our telescopes is still so strong that we usually can’t use much of the data we get from 90-100MHz. At this stage, it’s so hard to use that we usually don’t even try — we just use data at other frequencies. Luckily, the bulk of RFI is concentrated in this 10MHz band.

As for the Galaxy (and, indeed, all the other galaxies), there is one key fact we can use to separate the cosmological signal: the galactic emission is very smooth in frequency. That is, if you were to plot the amount of radiation from the galaxy against frequency, the line you’d get would be a nice smooth curve. The power at one frequency is always similar to the power at close-by frequencies (it’s roughly a power-law, with an index of -2.5, if that means anything to you). The reason for this is physical — the kind of physical processes that output radiation at 100 MHz in our Galaxy (synchrotron and free-free emission) do so randomly in frequency, and on average, it turns out to be smooth.

But recall that the cosmological emission is of an entirely different kind — it is an atomic transition that can only occur at a particular frequency. Each observed frequency then corresponds to a different time and place in the history of the Universe. But we expect that at one place, we might have a whole bunch of neutral hydrogen, while in another, there might be a star, and all the surrounding hydrogen is neutral (and we don’t get any emission). Thus, as a function of frequency, we do not expect the cosmological emission to be smooth. We can extract the cosmological signal by looking for the lumpy parts of the spectra.

Problematically, even if the Galactic emission really is super smooth, we don’t see it with a perfect telescope. The telescope itself can impart lumpiness onto the spectrum, making it impossible to tell which part is cosmological, and which is just the smooth Galaxy being made slightly lumpy by the telescope. On the other hand, we can introduce lumpiness in our own analysis as we try to correct for the telescope’s imperfections.

Building an extremely smooth, well-understood telescope is the main goal of the field right now, and there has been considerable progress over the last decade. On the other hand, developing our statistical analyses to ensure we don’t confuse the cosmic signal with our own Galaxy has also come a long way (and is my main focus).

Hopefully, it won’t be long before I can say we’ve finally seen it!

Steven G. Murray
Steven G. Murray
Marie Sklodowska-Curie Fellow

Astrophysics, code, math.