by George Rhee
We can use galaxies to map the rise and fall of star formation over 95% of the history of the universe. Charting the cosmic history of star formation tells us how galaxies formed and how the universe evolved with time. As we push further back we can hope to see the epoch of reionization, the 600 million year era when the first stars and galaxies formed and ionized the neutral hydrogen in the universe.
In this chapter we present the methods used to detect the highest redshift galaxies, we will show examples of galaxies that have been discovered with Hubble Space Telescope observations. We then show how we can study the cosmic history of star formation. Our Milky Way galaxy forms stars at a rate of three solar masses per year, but the results indicate that galaxies formed stars at a much higher rate in the past. We discuss in the final section attempts to find out if the known number of distant galaxies is sufficient to ionize the universe or whether we have to probe even further back in time to discover the first galaxies to form after the Big Bang at the end of the dark ages.
Fig. 9.2Computer simulation shows two galaxies colliding. Color indicates temperature, and brightness indicates the gas density. When the central black holes merge, a quasar is ignited, pushing gas outward (Credit: Tiziana Di Matteo, Carnegie Mellon University)
When Did the First Galaxies Form?
Observations of massive elliptical galaxies in our neighborhood suggest they formed a long time ago since they are composed of old stars that have an age comparable to the age of the universe. In contrast, lower mass galaxies such as dwarf galaxies and spiral galaxies are still forming stars and contain old and young stars. However our theories of dark matter structure formation (Chap. 6) predict that the lower mass halos form first and the more massive ones later by mergers.
A second problem is that the ratio of gas to dark matter in galaxies is several times smaller than the ratio of gas to dark matter in the universe as a whole. This can only mean that gas must have been expelled from galaxies, but how? We see evidence for winds expelling gas from galaxies at redshifts of two and lower. We think these winds are produced by the radiation from young stars in star-forming regions of galaxies, although supernova explosions and black holes also contribute. Figure 9.2 illustrates how black holes might expel gas during a merger of two galaxies. These processes are known as feedback processes and are being modeled using computer simulations.
The current picture is that massive elliptical galaxies form in two stages. At high redshift gas enters dark matter halos along filaments known as cold streams and turns into stars at the center of the halo. Eventually the buildup of hot gas through feedback processes shields the halo from further gas inflows. These compact young galaxies are dominated by young blue stars. They eventually age and merge to produce more massive elliptical galaxies in the second stage of galaxy formation. To see if these ideas have any validity we must turn to observations.
Searching for the First Galaxies
We have opened a window into the first billion years of the universe to see the beginning of the growth and evolution of cosmic structure. The scientific motivation for this difficult work is to measure the cosmic history of star formation at the earliest times and to detect the objects that reionized the universe. But how are we to detect these very faint galaxies at the edge of time? We can’t easily identify them in a single image of the sky because the vast majority of galaxies have redshifts much less than five. We need some way to sort the most distant objects from the nearby ones. Three methods are discussed below.
The first method makes use of the fact that hydrogen in high redshift galaxies and their surroundings absorbs almost all the short wavelength ultraviolet light. This method selects galaxies that contain mostly young stars and have little dust.
A second method makes use of two features in the spectra or colors of galaxies that are evidence of the presence of stars older than half a billion years. These features are known as the Balmer break and the 4,000 Å break. These features at redshifts of five and higher get redshifted to infrared wavelengths that are accessible from space. We must await the launch of the James Webb Telescope to arrive at reliable results for galaxies at redshifts of five and above containing older stars.
The third method uses far infrared and sub-mm emission from dust in high redshift galaxies. The very first galaxies and stars are not expected to contain dust since the chemical elements contained in dust are only produced inside stars. However we do not know at what stage dust makes its first appearance in the observable universe. This method is still in its infancy at high redshifts but new results are expected from a new telescope, the Atacama Large Millimeter/submillimeter array which has started carrying out observations high in the Andes.
Star Forming Galaxies at High Redshift: Lyman-Break Galaxies
The Lyman-break method selects high redshift galaxies that are forming lots of young stars. Some of the newborn stars are more massive than the Sun and emit large amounts of light at ultraviolet and blue wavelengths. As we discussed in Chap. 8, the high redshift universe contains enough neutral hydrogen to absorb ultraviolet light that may be emitted by these galaxies. The details of absorption of light by atomic hydrogen depend on the density of those atoms. At redshifts less than three, the hydrogen atoms
Fig. 9.3The near-infrared spectrum (black line) of a distant quasar at a redshift of 7. 1. The quasar is observed only 0.77 billion years after the Big Bang, but it has an intrinsic spectrum very similar to the averaged spectrum of lower redshift quasars (red line). The key difference is that shortward of a wavelength of about 1 μm corresponding to an emitted wavelength of λ r e s t ≃ 1, 216 Å there is no light. The larger amount of neutral hydrogen along the line-of-sight has absorbed all the ultraviolet light emitted by the quasar. This sudden drop in emission at wavelengths shorter than of Ly-α enables us to select rare extreme-redshift quasars such as this, but also more numerous “Lyman-break galaxies” at redshifts z > 5 (Credit: Reprinted by permission from Macmillan Publishers Ltd: Nature, 474, 7353, A luminous quasar at a redshift of z = 7.085, Daniel J. Mortlock et al. copyright (2011))
absorb light that has a wavelength shorter than 912 Å. As we go to redshifts greater than five, the absorption starts at wavelengths of 1,216 Å the wavelength of the Lyman-α line. We can use this effect to identify high redshift galaxies. This technique is illustrated in Figs. 9.3–9.5. Figure 9.3 shows the spectrum of a quasar at a redshift of 7.1. The light is emitted by fast moving hot gas surrounding a massive black hole. The point to note is that all the light emitted by the quasar at wavelengths shorter than 1,216 Å is missing because it has been absorbed by neutral hydrogen atoms. Bear in mind that the light emitted by the stars in the ultraviolet at 1,216 Å gets redshifted to a wavelength of 1 μm in the infrared by the time it reaches our telescope. The result is that the light detected at wavelengths shorter than 1 μm is missing in Fig. 9.3.
We can select galaxies in images using this effect as shown in Fig. 9.4. We measure the colors of galaxies by taking images of them with filters placed in front of the camera that let through only light of a specific range of wavelengths. The light profile or throughput of these filters is shown in the figure. For example, the
Fig. 9.4The Lyman-break method for finding galaxies at redshifts of about three. The sharp break expected at wavelengths shorter than 912 Å in the spectrum of a galaxy dominated (in the UV spectrum) by massive, young stars is shown (a top). The break is accentuated by absorption both in the galaxy hosting the stars (a middle) and in the intervening intergalactic medium (a bottom). The galaxy redshift is 3.1 in this example, bringing the feature into the optical window observable from the ground. Shown in a Bottom are broad-filter passbands that can be used to find Lyman-break galaxies in the vicinity of redshift z = 3. (b) An example of images taken through these filters. The circled galaxy is seen clearly through the red and green filters, but it disappears completely through the UV filter. Only a few percent of all comparably faint galaxies w
ill behave in this way (Credit: Charles C. Steidel, Observing the epoch of galaxy formation, Proc. Natl. Acad. Sci. USA, Vol 96, issue 8, April 13 1999, copyright (1999) National Academy of Sciences, U.S.A)
filter labeled R lets through light in the wavelength range 6,000–7,500 Å. For a redshift three galaxy the light with a wavelength less than 912 Å has been completely absorbed by neutral hydrogen, however if any light were present we would see it redshifted into the short wavelength visible band known as the U-band. But it is not present. The consequences of this are that when we take a series of images in the various filters, the image of the galaxy in the U-filter will not be present. Such galaxies are referred to as U-band dropouts. The U-dropouts are galaxies in the redshift range two to three. Galaxies at higher redshifts will be observed as dropouts in longer wavelength filters. For example B-dropouts are galaxies in the range three to four because the B-filter lets through light of longer wavelengths than the U-filter. One of the reasons for building the James Webb Space Telescope is to push the dropout technique into the infrared to find even more distant galaxies.
Why not simply obtain spectra of galaxy candidates and decide the redshift from the many emission and absorption features that are seen in the spectra? This is difficult because these galaxies are very faint. The galaxies are barely visible in images (Fig. 9.5) and obtaining a spectrum requires much more telescope time than obtaining an image. Figure 9.6 shows the thousands of galaxies detected in the Hubble Ultra Deep Field image. We cannot tell by inspecting this image which galaxies have the highest redshifts but by taking an image with different filters we can use the dropout technique to select the most distant galaxies ever seen with a telescope.
As we search for ever more distant galaxies, the dropout technique shifts to longer and longer wavelengths eventually moving out of the visible and into the infrared part of the spectrum. Using the infrared filters on the Hubble Space Telescope we can push the dropout technique out to redshifts of seven as illustrated in Fig. 9.5. This is one reason the study of the youngest and most distant galaxies depends on sensitive infrared detectors on large aperture telescopes on the ground or in space.
The dropout technique is not completely reliable. At a redshift of six, the technique produces one in four misidentifications; galaxies at lower redshifts that are mistaken for higher redshift galaxies.
There is also a population of galaxies known as Lyman-alpha blobs that are the opposite of the dropouts, these galaxies actually emit Lyman alpha radiation. These galaxies do have in common with the dropouts that they are star-forming galaxies. Why are not all star forming galaxies Lyman-alpha emitters? It appears that this depends on the details of the structure of these galaxies.
Star-Forming Galaxies at High Redshift: Sub-millimeter Galaxies
Emission in the far infrared that is not associated with the cosmic background radiation comes from galaxies that contain dust. The dust re-emits light in the infrared that it has absorbed at optical wavelengths. In fact one can show that for every two photons of light emitted by stars or by black holes one photon has been absorbed by dust and re-radiated as infrared and sub-millimeter emission by the dust. Since dust is closely associated with star-forming regions in galaxies, we can use dust emission as a tracer of star formation.
The dust emission spectrum has an interesting property that observers of distant galaxies can exploit. The intensity falls steeply from a peak of wavelength of one tenth of a millimeter down to about 2 mm. The emitted energy drops by a factor of about 1,000. Consequently sub-millimeter galaxies can appear brighter in our telescopes when observed at increasingly high redshifts. This is because for a fixed observing wavelength the emitted wavelength decreases with increasing redshift. These simple calculations show that the millimeter wavelength range is a good place to search for very distant galaxies that are actively forming stars.
One of the first cameras to operate at these wavelengths was the SCUBA (Sub-millimeter Common User Array) on the James Clerk Maxwell Telescope at Mauna Kea observatories in Hawaii. The first studies made with this camera revealed the existence sub-millimeter galaxies. The location of sub-millimeter galaxies on the sky could not be determined to high enough accuracy to see if optical counterparts to these galaxies exist in the Hubble Deep Fields. By combining SCUBA observations with infrared and radio wavelength observations it has become possible to identify these galaxies optically and take their spectra at visible wavelengths using the Keck telescopes.
Fig. 9.5Four candidate galaxies that are likely to have redshifts of 7 and thus have emitted their light when the universe was just 750 million years old. Each of the four candidate high-redshift galaxies are presented in a distinct row. All four candidate galaxies are shown using images at each of five different wavelengths. The three columns starting from the left are images taken with filters that select colors in visible light with bluer light on the left and redder on the right. The last two columns are images taken with two infrared filters. The galaxies aret all detected in the infrared but remain completely undetected at wavelengths shorter than red visible light. This abrupt drop-off in light emission is characteristic of star-forming galaxies at high redshifts and occurs due to the absorption of light by the large amounts of neutral hydrogen in the universe at early times. Astronomers use the presence of this break to find high-redshift galaxies. The present sources were found in the faintest images taken with the Hubble Space Telescope (Credit: Rychard Bouwens)
Keck telescope data revealed that the sub-millimeter galaxies are at redshift two on average (i.e. the light was emitted at a time when the universe was about 3 billion years old). These galaxies occurred more frequently in the past and make a significant contribution to the star formation at high redshift.
Fig. 9.6The Hubble Ultra Deep Field. There are about 10,000 galaxies in this image. Using the Lyman break or dropout techniques described in the text we can search for the most distant galaxies out to redshifts as high as ten using these data (Credit: NASA/STScI)
It is possible that the sub-millimeter galaxies are merging galaxies that contain alot of gas. This gas is then used up in a huge burst of star formation. The sub-millimeter galaxies and the Lyman-break galaxies have huge rates of star formation from several hundred to several thousand solar masses of stars formed per year. Our Milky Way galaxy by comparison forms stars at a rate of a few solar masses per year.
Older Galaxies at High Redshift: The Distant Red Galaxies
We discuss below the tools for discovering old galaxies at high redshifts and the implications of these discoveries. By comparing observed galaxy spectra to models constructed from stellar spectra one can learn something about the ages of galaxies. A galaxy spectrum consists of the light from lots of stars of different mass and age, each with its own spectrum. If we imagine a bunch of stars all forming at the same time (the simplest assumption) we can predict how the spectrum of such a galaxy will change as it gets older. We know that the more massive hot blue stars run out of hydrogen soonest and explode as supernovae. As the population ages and the blue stars contribute less light, the galaxy spectrum changes. Figure 9.7 shows how the spectrum of a galaxy is expected to change as its population of stars age.
Fig. 9.7Simulated galaxy spectra following a billion year (Gyr) burst of star formation. Ages are marked in Gyr on each spectrum. Note the dramatic change in the relative amounts of flux blueward and shortward of 4,000 Å as the galaxy ages. The galaxy will appear to get progressively redder as it ages (Credit: Stephen Serjeant, Observational Cosmology, Cambridge University Press, which has been adapted from Fig. 1 in Rocca-Volmerange & Guiderdoni, 1988, A&AS, 75, 93, reproduced with permission ©ESO)
There is a feature in the blue part of the spectrum of galaxies that we can use to detect galaxies with older stars. This feature is called the 4,000 Å break. 4,000 Å refers to wavelength of the blue light where the feature occurs. The feature is caused in part by the absorption of light by ionized calcium. The feature is also more pron
ounced because the light from old stars is dropping off towards the blue.
Galaxies with an old population of stars have been identified at high redshift using the 4,000 Å break. At a redshift of three this feature is redshifted into the infrared part of the spectrum (1.6 μm). We can select older galaxies in the redshift range 2 < z < 3 by imaging them at 1.24 and 2.2 μm using standard infrared filters. Spectra of galaxies selected in this manner do not show emission lines and appear to have an old population of stars.
Figure 9.7 illustrates how the spectrum of a galaxy can change with time due to the effects of an aging stellar population. Figure 9.8 shows a model of a redshift seven galaxy spectrum. The figure shows the colors of the light that a redshift seven galaxy might have when observed by the Spitzer and Hubble space telescopes.
Fig. 9.8The redshifted light anticipated from a young galaxy at redshift seven. The emitted ultraviolet light can be sampled by the red color and near-infrared filters on the Hubble Space Telescope shown in blue and green in the Figure. The longer wavelengths can be probed with filters aboard an infrared satellite called Spitzer. The Spitzer filters are shown in red in the figure. The galaxy light shows the sharp drop at λ r e s t = 1, 216 Å due to the strong absorption by intervening neutral hydrogen anticipated at this redshift. Longward of this “Lyman-break” the galaxy light shown is simply that of all the stars added together. The features known as the Balmer break and the 4,000 Å break can help estimate the age of the stars in the galaxy. As the galaxy ages these feature become more pronounced as shown in Fig. 9.7. The gap between the Hubble Space Telescope and the Spitzer Space Telescope filter coverage will not be covered from space until the advent of the James Webb Space Telescope (Credit: A. B. Rogers and J. S. Dunlop)