“To do what we do, you have to be hopelessly optimistic, but also brutally realistic.”

must identify them and strip them out.
No images or catalogs of sources exist for this poorly studied part of the spectrum; the teams must map it out themselves before they can discount it from their data. “After subtracting all the foregrounds, the signal- to-noise ratio is still one-tenth. You have to understand the noise [and] find out ways to
quantify it,” says ASTRON’s Brentjens.
Once that’s done, investigators will be rewarded not with an image of the neutral

hydrogen at the EoR, but rather a power spectrum: a statistical analysis of how the ra- dio signal varies across the sky. It will reveal whether the biggest variations occur over small distances or large ones—whether the bubbles of ionized gas were small, the handi- work of individual stars, or galaxy-sized cavi- ties. The teams should also be able to watch reionization unfold over time. The EoR may have lasted millions of years; 21-cm radia- tion from earlier in its history will have trav- eled farther and thus will be stretched out to a longer wavelength and a lower frequency than later radiation. So a signal detected at 140 MHz will be from an earlier time than one at 160 MHz.
As interferometers are sensitive to differ- ences, the middle of the EoR—when half the universe is neutral and half ionized—will produce the strongest signal. So the teams will be scanning the frequencies for a signal that has a peak and then drops off farther into the past (when more of the universe was neutral) and farther toward the present (when more of the universe was ionized). “These telescopes hope to learn two basic things: when the EoR happened and how long it lasted,” Bowman says. “That should be easy to read when they detect a signal.”
All three teams are optimistic that they will soon get their first glimpse of the EoR. “We’re getting pretty close to what theorists predict the signal level is, and we expect to do two or three times better with the data that is coming in right now,” De Bruyn says. He hopes LOFAR will get a “first-order result” by next year. The PAPER and MWA teams are similarly hopeful. But MWA’s Bowman adds that those projections are all based on theoretical models of the EoR sig- nal. “If there is no detection by 2020,” he ac- knowledges, “that will be a disappointment for the community.”

Because of the amount of signal process- ing required and the many different assump- tions that underlie the calculations, “there’ll be no eureka moment. It’ll be hard to con- vince ourselves [of the detection],” Brentjens says. Even harder will be convincing the rival teams. “I worry about this a lot,” says Aaron Parsons of UC Berkeley, who is head of PAPER. “I hope journal editors are very careful. It’s very important that papers are reviewed by people who are really knowl- edgeable. And we have to be very careful not to overstate claims.”
A confirmed and reliable signal from the time of reionization could amount to what some researchers are calling “a COBE mo- ment” for astrophysics. COBE was the NASA satellite that, in 1992, revealed the size of fluctuations in the microwave background and opened a floodgate of results in cos- mology. A glimpse of the EoR would give astrophysicists their own origins story and a starting point for studying the very first things to shine.
Knocking out an electron and ionizing hydrogen takes quite a lot of energy, so any potential ionizing source needs to produce a lot of photons at high energies—ultraviolet or higher. It’s expected that the first stars to

form in the universe were unlike any that ex- ist now because they were made of almost pure hydrogen, without any of the heavier elements that were forged inside stars as the universe aged. Pure hydrogen stars, known as population III stars, should grow to enor- mous size before their internal furnaces ignite—hundreds or even thousands of times as massive as our sun. Big stars burn bright, hot, and fast, making them a perfect source of ionizing radiation. But do they form in isolation, or does dark matter draw the hy- drogen into galaxies first? Or were bigger, more powerful sources such as quasars— hugely luminous galactic nuclei centered on supermassive black holes—the engine of reionization? Theorists speculating about the EoR have also invoked more exotic driv- ers, including decaying dark matter and cos- mic strings. “We’re shooting in the dark. We have no idea what they are,” Zaroubi says.
The existing arrays probably won’t be able to answer all of these questions. “To really understand how the first stars form and what early galaxies were like needs the next generation of instrument,” Parsons says.
The LOFAR team hopes to build up its ar- ray with more stations and faster computing. But the PAPER and MWA teams are joining forces to build a new, more powerful instru- ment called HERA, the Hydrogen Epoch of Reionization Array. HERA’s antennas will be static wire-mesh dishes pointing straight up. The joint team has won $2 million to build a test array of 37 dishes in the Karoo, using PAPER’s infrastructure. This alone will have

The Square Kilometre Array (left, in an artist’s conception) will use dishes and static antennas to pick up different radio frequencies. LOFAR (above) mingles low-frequency antennas (brown specks) with higher frequency ones (inside dark tiles).
up to three times the sensitivity of PAPER, Parsons says. Then the team will seek up to $20 million to build an array of 350 dishes by 2019. “We’ll turn the tables on theorists and really start to drive theory, really ad- vance our understanding,” he says.
Looming on the horizon is the next gen- eration: the Square Kilometre Array (SKA). This enormous international project will be built mostly in South Africa, starting in 2018, and will target everything from galaxy evo- lution to signals from extraterrestrial intelli- gence. But part of the array, to be sited at the Murchison Radio-astronomy Observatory, home of MWA, will collect low-frequency ra- diation with a quarter of a million antennas spread over 100 kilometers. With its huge collecting capacity, SKA will be able to move beyond statistical observations and produce images. “We’ll see the structures themselves directly. That’s a huge step,” Zaroubi says.

But first the rival teams need to catch that first glimpse of early light. They will have to overcome radio interference, computing challenges, and the deafening noise—and they must hope that theoretical models of the EoR signal are correct. Says Zaroubi: “To do what we do, you have to be hopelessly op- timistic, but also brutally realistic. You need both sides.” ■