Figure 1: The telescopes we use for 21 cm analysis: Top : The recently upgraded core of the MWA in 2016. Bottom: HERA, under construction in 2018

Figure 1: The telescopes we use for 21 cm analysis. Top: the recently upgraded core of the MWA in 2016. Below: HERA, under construction in 2018 (click to enlarge)

Have you ever wondered why the universe looks the way it does, with isolated galaxies loosely arranged like pearls on the cosmic web? The key to understanding the structure of the universe lies in finding out just how the universe formed after the Big Bang.

Immediately after the Big Bang, the universe was so small and hot that protons, neutrons, and electrons couldn’t stick together. After some time, the universe expanded and cooled enough to form a fog of neutral hydrogen. From there, some of the gas coalesced and formed the first stars, which burned away the primordial neutral hydrogen fog and formed the seeds of what has become our universe.

Exactly how this happened is still a mystery, but there is hope in solving this by looking deep into the past using 21 cm radio light emitted by the neutral hydrogen fog. Observing this light allows us to peer into the Cosmic Dawn of the first stars, and from there start to understand why our universe and galaxies are structured the way they are.

However, observing the faint radio emission from 13 billion years ago is difficult. One challenge is that the 21 cm light from the cosmic dawn has been red-shifted by the expansion of the universe to about 150 HERA, under construction in 2018MHz, right in the middle of the radio broadcast spectrum, so any human activity, especially FM radio and TV signals, can contaminate the measurements. Another is that emissions from our galaxy and other galaxies get in the way, with brightnesses 100,000 times brighter than the predicted emission from the first galaxies. Observing the faint radio emission from the formation of our universe’s first stars and galaxies is one the grand challenges of modern astrophysics.

Figure 2: Members of our group using the big table in the meeting room of the Data Science Studio to look at data in the form of power spectrum jackknives. Each page represents half an hour of integration, about ¼ of a petabyte of data are represented on the table. The spatial organization (and reorganization) of the plots help us identify the sources of faint systematics.

Figure 2: Members of our group using the big table in the meeting room of the Data Science Studio to look at data in the form of power spectrum jackknives. Each page represents half an hour of integration, about ¼ of a petabyte of data are represented on the table. The spatial organization (and reorganization) of the plots help us identify the sources of faint systematics.

Using advanced statistical techniques we helped pioneer, it is in principle possible to separate the faint primordial signal from the bright contamination. However, isolating the signal in practice requires unprecedented precision in our instrumental calibration and data analysis. Our group has helped to design and build two radio telescopes (Figure 1) to pursue this signal – the Murchison Widefield Array (MWA) in Western Australia (a place with very few people) and the Hydrogen Epoch of Reionization Array (HERA) in rural South Africa – and we have developed one of the world’s most precise data analysis software suites to achieve the unprecedented spectral-spatial dynamic range needed to see the first stars and galaxies.

Our analysis approach is intrinsically data-driven and iterative: we strive to understand what the data are telling us about our instrument and our foregrounds and we incorporate those lessons into our models and analysis software. This often requires looking at a lot of data (Figure 2)! As our models and analyses improve, we hunt for and discover more and more subtle analysis and instrumental systematics to address. Testing the often extremely subtle effects of software and instrument model changes is critical to our iterative approach, and we have developed a number of innovative techniques for precision analysis of vast data sets (Figure 3).

Our data-driven approach has yielded some of the best limits to date on the 21 cm signal from the first galaxies, and we are excited about the potential for these observations to improve dramatically in the next few years as HERA is fully built and as the MWA is upgraded and improved.

Figure 3: An example of a full pipeline integration test showing a change in the code and the effect of the code change on a primary 21 cm cosmology figure of merit. The plot shows a difference between the two-dimensional power spectrum before and after the code change. The 2D power spectrum is an extremely useful metric because it separates effects along instrumentally meaningful axes, helping researchers to identify how instrument and analysis systematics interact and affect the final measurement.

Figure 3: An example of a full pipeline integration test showing a change in the code and the effect of the code change on a primary 21 cm cosmology figure of merit. The plot shows a difference between the two-dimensional power spectrum before and after the code change. The 2D power spectrum is an extremely useful metric because it separates effects along instrumentally meaningful axes, helping researchers to identify how instrument and analysis systematics interact and affect the final measurement.

Our research team members are:

  • Ruby Byrne: PhD student, Department of Physics, University of Washington
  • Mike Wilensky: PhD student, Department of Physics, University of Washington
  • Ian Sullivan: Research Scientist, Department of Astronomy and Dirac Institute, University of Washington
  • Bryna Hazelton: Research Scientist, Department of Physics and the eScience Institute, University of Washington
  • Miguel Morales: Associate Professor, Department of Physics, University of Washington
  • Previous PhD students with our group include: Nichole Barry (now at University of Melbourne, Australia), Patricia Carroll and Adam Beardsley (now at Arizona State University)