Reionization and Cosmic Dawn

Hydrogen reionization is the epoch when ultraviolet photons produced by the first galaxies ionized almost all of the hydrogen in the Universe. (The vast majority of the atoms in the Universe are hydrogen.) Astrophysicists think that this process happened when the Universe was several hundred million years old (a few percent of its current age), but exactly when and how it happened is something that we are struggling to answer.  I have worked on modeling this epoch (the adjacent image is from one of my simulations of this process) as well as on developing different observational methods that can be used to study it.  For example, I have thought about how reionization affects the statistical properties of Lyman-alpha emitting galaxies, absorption in the spectra of gamma ray bursts, and intensity fluctuations in the cosmic microwave background radiation. For a review of  different techniques to study hydrogen reionization see this link.  More recently I have thought about how the reionization process suppresses the formation of dwarf galaxies, and the thermal state of intergalactic gas (see below), and diffuse Lyman-alpha emission. Click here for movies generated from reionization simulations from my grad student days.

Relic temperature fluctuations from hydrogen reionization

Recent observations of the z >5 Lyman-alpha forest show large-scale spatial variations in the intergalactic Lyman-alpha opacity that grow rapidly with increasing redshift. Previous studies have attempted to explain this excess with spatial fluctuations in the ionizing background, but found that this required either extremely rare sources or problematically low values for the mean free path of ionizing photons.  In collaboration with Anson D’Aloisio [this study’s lead] and Hy Trac, we showed that opacity fluctuations could arise from residual spatial variations in temperature that are an inevitable byproduct of a patchy and extended reionization process.  (Click here for the paper link.)  We showed that if the entire excess is due to temperature variations alone, the observed fluctuation amplitude favors a late-ending but extended reionization process that was half complete by z~9 and that ended at z~6. In this scenario, the highest opacities occur in regions that were reionized earliest, since they have had the most time to cool, while the lowest opacities occur in the warmer regions that reionized most recently. This correspondence potentially opens a new observational window into patchy reionization.  Below is an illustration of a model for the distribution of reionization redshifts (left panel), and skewers through this model showing the reionization redshift and the transmission in the Lyman-alpha forest (right panel).  Regions reionized more recently are hotter (~104K), which results in more neutral hydrogen and more transmission, than regions ionized much earlier on (which can be a factor of five or so times cooler).  The size of the fluctuations in this model turn out to explain very well the observed opacity fluctuations and their redshift evolution, despite this model having essentially no freedom in describing the post-reionization evolution.

The competing explanation for these fluctuations is that they owe to spatial inhomogeneities in the hydrogen ionizing background. Fluctuating ionizing background models require either extremely rare sources (quasars to dominate the ionizing background) or for the mean free path of ionizing photons to be a smaller than expected.  We also examined the plausibility of these alternatives (here and here). An update (8/18): A recent observation led by George Becker of a dearth of Lyman-alpha emitting galaxies in a z~5.7 Gunn-Peterson trough suggests ionizing background fluctuations dominate the fluctuations and not temperature (see my Budapest talk here), although I think there is also the possibility of substantial neutral gas in this high-redshift trough.

An update (8/18): The above study assumed temperatures of 20,000K and 30,000K motivated by some calculations I had done in an earlier paper.  The IGM temperature not only impacts IGM opacity fluctuations, but the widths of absorption lines, which is measurable in the Lyman-alpha forest.   Thus, it is worth quantifying what temperatures reionization imparts.  (It is also unclear if simulations capture scales necessary to capture the temperature, which naively seems to require resolving the kiloparsec widths of ionization fronts.)  We finally got around to quantifying the exact value of the temperature.  There is a common misperception that the temperature is determined by the spectrum of the sources.  However, the sources inject via photoionization much more heat per baryon than the final temperature.  This causes the IGM to cool very efficiently in the ionization front, where there are a comparable number of neutrals and free electrons, resulting in collisional cooling being maximally efficient.  The more injected energy from photoionization, the more efficient it cools since cooling is exponentially sensitive to temperature.  The end result is that for plausible spectra, the temperature is most sensitive to the velocity of the ionization front (which sets  the duration over which the gas can cool).  The panel below shows how the temperature of the ionization front (the numbers are in degrees Kelvin) depends on its velocity and spectral index.  We
find that for plausible spectral indices (<2 for galactic stellar radiation) and plausible ionization front speeds (103-104 km/s), the resulting temperature depends principally on the velocity of the ionization front and not the spectral index, contrary to the pervading view!   Below is a comparison  of a calculation that measures the velocity of the ionization front in a simulation and puts down the temperature using just the ionization front velocity to the actual simulated temperature.  The ionization fronts accelerate as reionization progresses, imparting more heat at the end than at the beginning.   


Helium II reionization (AKA the reionization of HeII)

I have also studied the reionization of helium, the second-most abundant element in the Universe (constituting 25% of the total baryonic mass).  We think that radiation from accreting supermassive black holes -- objects astronomers call ``quasars’’ and the brightest objects in the Universe -- doubly ionized the intergalactic helium when the Universe was 20% of its current age.  The advantage of studying helium reionization compared to hydrogen reionization is that there is far more data about the state of the Universe at the epoch we think this process occurred.  The panel on the right shows a 430 comoving Mpc slice (~300 million light years across) through a model of the Universe  from one of my simulations of this process.  I have contributed to bolstering the argument that this process was ending when the Universe was a couple billion years, at z=3.  In addition, Eric Switzer and I identified two new observables (absorption from the 584A line of neutral helium and 8.7 GHz hyperfine absorption of singly ionized 3He) that have the potential to reveal much more detail regarding this epoch than other known methods.

High redshift 21cm radiation from reionization

Hydrogen (which constitutes 92% of all atoms in the Universe) emits at a fairly long wavelength of 21 centimeters, as least if it is atomic.  During and before hydrogen reionization, much of the hydrogen in the Universe was atomic and, thus, emitting copiously at this wavelength.  (Okay, copious may be an overstatement as the timescale for an excited atom to emit via this transition is an excruciating 10 million years!)  Efforts across the globe are gearing up to detect 21 cm emission from the first billion years in the Universe.  This observable has the potential to reveal unprecedented information about reionization, and it is really the only signal we can detect from when the Universe was between 5 and 300 million years old (during the Universe’s ``adolescence’’).   I have been involved in quantifying the sensitivity of potential instruments to this signal, in exploring design optimizations for these observatories, and in understanding the physics that 21cm observations can constrain.  I have been or currently am involved on the theoretical side with several of these efforts: the Murchison Widefield Array (MWA) in western Australia, the Precision Array for Probing the Epoch of Reionization (PAPER) in South Africa, as well as a newly commissioned instrument, the Large-aperture Experiment to Detect the Dark Ages (LEDA), in New Mexico.


High redshift 21cm radiation from the first stars


It has recently been realized by Tseliakhovich & Hirata (2010) that the baryons were moving supersonically with respect to the dark matter after recombination and prior to when the Universe was reheated by astrophysical sources.  The size of this velocity difference varies spatially in the Universe, with it being supersonic in most of the volume and reaching Mach numbers as high as several in select locations at 20<~z<200.  If the fluctuations in the 21cm background coupled to this velocity difference, even in a very weak manner, this could result in much larger and more distinctive spatial correlations than previously thought.  Somewhat disappointingly, we showed that this coupling could generate only an order unity enhancement in the 21cm signal (although at the most observable redshifts and length scales, the enhancement is likely even smaller). 

The above figure shows how a typical difference in this relative velocity impacts the gas in a 3x106 Msun halo -- a halo that is likely to form stars before reionization --, from a cosmological simulation run with the Enzo code.  The panel on the left does not have any relative velocity whereas in the panel on the right the baryons are moving with Mach number of 1.8 relative to the dark matter (the RMS Mach number in our universe), but otherwise the initial conditions in the two simulations are the same.  The gas in this halo is significantly impacted by the relative velocity.  For movies illustrating the impact of the relative velocity click here. These simulations use cosmological initial conditions generated with the code available here.


The 21cm signal from reionization may be perturbative

Observations of redshifted 21cm targeting reionization are most sensitive to the power spectrum of this signal.  Indeed, the power spectrum should be detected well before any other statistic is measured.  The power spectrum is fairly far removed from an image large patches of neutral gas -- what we would most want to measure, but will lack the sensitivity to do.  A 21cm power spectrum measurement will then result in a large interpretational challenge.  It is not clear what information about reionization such a measurement constrains. Motivated by the fact that the cross correlation coefficient between the linear density field and the 21cm signal in simulations is not small on many of the scales potentially probed by 21cm instrumental efforts, we developed an effective perturbation theory (and, equivalently, a bias expansion) for the inhomogeneous 21cm radiation field from reionization. 


The panels above illustrate the success of this expansion.  The lefthand panels consider the ionization fraction overdensity, and the righthand ones consider the 21cm signal. The upper panels feature the power spectrum of these quantities, with the filled dots showing this signal in a simulation of reionization and with the solid curves showing the best-fit model (fitting three parameters of varying importance). The lower panels feature the power spectrum of the model error, divided by the simulation power spectrum. (The thick curves fit the model to modes as large as kmax  = 0.2 h Mpc-1 , and the thin to modes as large as kmax  = 0.4 h Mpc-1 .) The black solid histogram in the upper-right panel is the forecasted z  = 8 sensitivity of HERA, the next generation 21cm array. Over much of the wavenumber range probed by HERA (and the same holds for other efforts), the signal is well described by this perturbation theory, particularly for the first 70% of reionization in this model.

This result provides an understanding of the potential signal shapes that had been missing. We find that the observable signal can often be described with 2ish bias coefficients that can be interpreted in terms of the source galaxy’s bias, the average neutral fraction, and the characteristic size of ionized regions.