CANDELS Detects First Light Galaxies

This is a guest post by Ketron Mitchell-Wayne, graduate student at the University of California-Irvine. He and other CANDELS team member recently published a paper in Nature. This paper was the subject of a recent press release. Here, Ketron describes the project and the role that he played.

The cosmic extragalactic background light is a product of many different component emissions throughout all cosmic times. Recent CANDELS observations have opened up a new window of opportunity for measuring this cosmic background light at optical and near infrared wavelengths. We have assembled Hubble frames taken over a 10 year period and mosaiced them to produce some of the deepest images suitable for such a study. With the mosaics, we can study this diffuse, clumpy light that resides behind all the resolved stars and galaxies in the mosaics. With statistics, we have attributed a fraction of this diffuse background to the first light galaxies during reionization. Here's a short summary of the work that I did, over the course of two years, in order to make these very interesting measurements.

My main job for this paper was generating the mosaics and making the statistical measurements. I started working on the data reduction in the summer of 2013 and have spent the better part of the last two years working on the project. Anton Koekemoer had a data reduction pipeline set up for all the incoming CANDELS data, but I wanted to incorporate archival data in our analysis too. So I had a number of reduction steps to complete on thousands of frames, even before making the mosaics (which is in itself very difficult).

Once we had mosaics in multiple bands (left panel of Figure 1), I generated a source mask. We want to isolate the background light signal, so foreground stars and galaxies need to be removed from the image. The dark areas in the second panel of Figure 1 is the source mask (just zeros in the array).

Figure 1: These three panels show different components of near-infrared background light. The one on the left is a mosaic of images taken, the one in the middle shows the intrahalo light seen when masking out all the stars and galaxies, and the one on the right shows the signature of the first galaxies. Credit: Ketron Mitchell-Wynne / UCI

At this point I could start making statistical measurements of the background light in the mosaiced, source-subtracted maps. The methods we used aren't new, but much of the dataset was. We used a very similar method to what was used in the Cosmic Microwave Background (CMB) studies. We look at "empty pixels" (what's left over after source removal) and measure whether or not some group of pixels in one part of the image is correlated with another group of pixels in a different part of the image. This is the angular power spectrum, which quantifies these correlations, as a function of angular scale. This is exactly what the CMB team did to measure the microwave background power spectrum, which is paramount in our understanding of cosmology. 

Figure 2: The brightness of the near-infrared background light as a function
of wavelength. Our new Hubble measurements are highlighted in orange.
The components from the "intrahalo light" (shown above in middle panel)
and the first light galaxies (right panel above) are shown as the blue
and red line, respectively.

I made maps in five different wavelength ranges, or "bands": 0.6, 0.7, 0.85, 1.25 and 1.6 microns. The shortest band is in the yellow range of visible light, and the longest two are in the near-infrared (NIR), which our eyes aren't sensitive to. This wavelength range (1 micron) is special because it is sensitive to Lyman break signatures with a multi-wavelength study, and it is the wavelength at which we expect a signal from the reionization epoch. Figure 2 shows the brightness of the background light in each of these bands. Each of the bands has a common component - what we call "intrahalo light" - which is the light emitted by stars which have been tidally stripped from their host galaxies via mergers or interactions. But in Figure 2 you can see that the brightness drops significantly from the two NIR bands to the shorter bands. We think this is because the NIR bands are picking up, in addition to intrahalo light, a high-redshift signal from the first light galaxies. Because the photons from the reionization era have been redshifted by a factor of about 10, we expect their signal to peak between 0.9 and 1.1 microns, with no shortward contribution below the Lyman break at about 0.8 microns.

We're studying the background light, which traces emission from many different kinds of sources over all cosmological times. So we don't have a direct image of only the first galaxies. With sophisticated modeling, we were able to separate the different component emissions, and isolate the signal from the first galaxies. So what we have, via statistical methods, is a description of the astrophysical environment 500 million years after the big bang. The third panel in Fig 1 is a reconstruction of what they would look like based on our statistical measurements. Cosmological theory suggests that these first light galaxies are the progenitors to our milky way, and all other evolved galaxies.