Understanding the origin, formation, and evolution of the first galaxies is one of the major questions of astrophysics. Identified by the Astro2010 decadal review as a key science priority, the period of “Cosmic Dawn” encompasses the formation of the first stars, galaxies, and black holes that are responsible for First Light in the Universe. These luminous sources produce UV and X-ray radiation that radically alter the properties of the diffuse neutral hydrogen gas that constituted the majority of the pre- and intergalactic medium (IGM) at early times. It is important that we develop observational probes of this time to enable valuable insights into the nature of the first objects, processes of galaxy formation, reionization.

Low frequency radio observations of the redshifted 21 cm line of neutral hydrogen (Madau et al. 1997; Furlanetto et al. 2006) are a unique avenue for observationally constraining the radiative properties of the first luminous objects and the reionization history of the Universe. The 21 cm line occurs due to the hyperfine splitting of the hydrogen ground state due to the interaction between proton and electron spins and has a rest-frame frequency of 1.4 GHz.

The theoretical evolution of the global 21 cm signal, calculated using the models of Mesinger, Ferrara, & Spiegel (2013), is shown in Figure 1. After thermal decoupling at z = 200, the gas cools adiabatically, becoming cooler than the CMB. Atomic collisions are effective at coupling spin and gas temperatures while the universe is dense, leading to a 21 cm absorption signal at high-redshift (not shown in Figure 1). Collisional coupling becomes ineffective as the Universe expands, diluting the gas, and the spin temperature slowly re-equilibrates with the CMB. However, once the first stars form, they produce a background of UV photons that again couple spin and gas temperatures leading to a second absorption regime at z~20. After the first stars die, they leave behind compact objects, such as black holes and neutron stars that, through accretion, generate X-rays. Along with emission from early quasars and stars, these X-rays travel long distances depositing their energy as heat. This raises the IGM temperature (around z~15) and eventually leads to a 21 cm emission signal. As further galaxies form, ionizing radiation leads to large ionized bubbles that begin the process of reionization (z < 15). Ultimately, all neutral hydrogen in the diffuse IGM is ionized and the 21 cm signal ends (z~6).

This model represents our current theoretical expectation for the evolution of the thermal history of the early IGM and the resulting 21 cm signal. However, it is based upon the extrapolation of the observed properties of local galaxies and X-ray sources applied to the very first galaxies. Realistically our estimates of X-ray and UV emissivity may be off by orders of magnitude since the relative roles of stellar and black hole populations are not known (Volonteri & Gnedin 2009) and, while simulating the first stars is possible (Yoshida et al. 2008; Turk et al. 2009), accurately modeling the overall properties of the first galaxies is much more difficult. Further, clumpiness of the IGM and the role of dense clumps as sinks of ionizing and heating photons is unresolved (Iliev et al. 2007; McQuinn et al. 2007). The consequence of the theoretical uncertainties is considerable variation in the redshift of the events outlined above, so that measurement of the 21 cm global signal would provide a powerful and entirely new probe of the radiative properties of early sources.


As of 2016, constraints on the timing and during of reionization can be derived from several observational techniques. Cosmic Microwave Background experiments yield integral probes on reionization through two primary effects. CMB photons travelling from the surface of last scattering are Thomson scattered by free electrons in the IGM after reionization, giving rise to an optical depth that can be used to estimate the redshift of reionization. The latest Planck Collaboration et al. (2015) results yield a reionization redshift of zr = 8.8 +/-1.5. The South Pole Telescope (SPT) has produced limits on reionization duration (Zahn et al. 2012) based on the kinetic Sunyaev-Zeldovich (kSZ) effect in which CMB photons are Doppler shifted by the bulk streaming of free electrons. For patchy reionization scenarios, this effect yields excess power at high-l. The SPT reported that reionization must have lasted less than dz < 7.9. Using the more optimistic interpretation of Liu et al. (2013) yields a limit of dz < 2.64, although they note that the interpretation is not straightforward and that inclusion of more general reionization models in the analysis may weaken the constraint. Meanwhile, high-redshift galaxy observations (summarized in Robertson et al. 2013, 2015) including Lyman-α emitter and Lyman break galaxy evolution, Lyman-α forest transmission, quasar near-zones, and gamma-ray burst damping wing absorption indicate that reionization was largely complete by z=6.5.

First Stars and Black Holes

The first stars (PopIII) likely play a critical role in galaxy formation. They are predicted to be massive (Haiman et al. 1996, Bromm et al. 1999, Abel et al. 2002) and short-lived. They significantly affect the conditions for subsequent star formation through their photo-ionizing UV emission, metal enrichment, and thermal feedback (Bromm et al. 2009). Yet to date little evidence exists to support theoretical predictions beyond observed abundances of low-mass stars in our own Galaxy (Salvadori et al. 2007, Rollinde et al. 2009). Similarly, the existence of supermassive black holes by z=6 is well established (e.g. Fan 2006), but how they formed remains a mystery since sustained Eddington accretion or direct-collapse scenarios (Begelman et al. 2006, 2008) are required to explain their early presence—neither of which is constrained presently. JWST should help alleviate some of the uncertainty, but it will be limited to detecting PopIII galaxies and supernovae below z

Global redshifted 21 cm observations probe the probe the UV and X-ray radiative properties of the first stars and black holes through the shape and timing of the z~20 absorption feature. Mirocha et al. (2013) have shown that global 21cm observations can yield direct constraints on the Lyman-α background intensity and total heat deposition during First Light. They find that simple determination of the redshift of the 21 cm absorption feature will establish valuable limits on the early heating rate, in particular. More recently, they have extended their analysis and find that high-redshift global 21 observations can also constrain the growth rate of dark matter halos (Mirocha et al. 2015). They conclude that the “global 21-cm signal can in principle (i) identify the characteristic halo mass threshold for star formation at all redshifts z ≳ 15, (ii) extend z ≲ 4 upper limits on the normalization of the X-ray luminosity star formation rate (LX–SFR) relation out to z~20, and (iii) provide joint constraints on stellar spectra and the escape fraction of ionizing radiation at z~12”.

Foregrounds and Parameter Estimation

At the target frequencies, Galactic synchrotron emission accounts for 70% of the 100 to 10,000 K sky brightness temperature, depending on position and frequency. Free-free emission and discrete Galactic and extragalactic continuum sources contribute the remaining power (Bridle 1967). The total sky spectrum follows an extremely smooth, nearly power-law spectrum (Shaver et al. 1999; Rogers & Bowman 2008). Radio recombination lines (RRLs) in the interstellar medium (ISM) are an exception to the smooth spectrum, but they are extremely faint at high Galactic latitudes and occur at discrete, known frequencies that are naturally excised in observations by radio-frequency interference (RFI) filters.

Foreground subtraction in global redshifted 21 cm measurements is potentially limited by frequency-dependent (chromatic) structure in antenna beam patterns (Vedantham et al. 2014, Bernardi et al. 2015, Mozdzen et al. 2016). Chromatic beams couple angular structures in Galactic foreground emission to spectral structures that may not be removed by smooth functional forms. The current EDGES “blade” antenna has been chosen to minimize chromaticity compared to other dipole-based designs. We have shown with existing data that the foreground contribution can be fit at the ~mK level with a 5-term polynomial across either the high or low band, matching expectations based on simulations (Mozdzen et al. 2016).

The Earth’s ionosphere affects the propagation of radio waves, even above the plasma frequency (~30 MHz) through absorption of incoming radiation, as well as direct thermal emission from electrons in the ionosphere. For quiet night time conditions, these effects yield simple deviations from the true sky signal that can be described as power perturbations. We showed in Rogers et al. (2015) that our model of the ionospheric contributions is a good fit to data. Further, the ionospheric effects can typically be absorbed by standard polynomial terms in foreground models with 4 or more terms. Through simulation and data analysis, we have developed two foreground models, one is a simple power-law polynomial, the other is a sum of terms based on physically motivated power-law components, including spectral indices tuned for both ionospheric emission and absorption perturbations. Both models perform well and exhibit relative strengths at different sidereal times and for different polynomial orders.

Astrophysical information is derived from EDGES observations through parameter estimation in model fits of calibrated, integrated spectra. In addition to the five foreground terms discussed above, we simultaneously include a global 21 cm signal model and perform least-squares and/or Markov Chain Monte Carlo (MCMC) analyses to identify best-fits and confidence intervals. For high-band analysis, we have so far used a hyperbolic tangent with two free parameters (zr and dz) to model the signal contribution. In 2016, we are beginning to extend the sophistication of our parameter estimation pipeline to include numerically modeled signals using 21CMFast (Mesinger et al. 2011), as well as models from Dr. Anastasia Fialkov and others.