Photometric Redshifts And The History Of Star Formation

Jupsat Pro Astronomy Software

Secrets of the Deep Sky

Get Instant Access

A generalization of this approach, tried in the 1980s by Loh and Spillar, but refined enough for widespread use only in the mid-1990s, is determination of so-called photometric redshifts. Galaxy spectra have only a limited range of properties, being to a good approximation a one-parameter family and to an excellent approximation a two-parameter family. Photometric measures of the galaxy's brightness through sets of broad-band filters, if they have high enough precision covering a wide wavelength baseline, can simultaneously determine the kind of galaxy spectrum and its redshift, as long as the galaxy spectrum is represented within the range of fitting functions. This works reasonably well for optical data for z > 2 and 2.8 < z < 7, while near-infrared data are crucial for getting acceptable errors for z = 2-2.8 and and at very high redshifts when no optical flux remains due to Lyman absorption. Both the algorithms for retrieving photometric redshifts and the general confidence in the technique got a major boost from data collected in the Hubble Deep Field, in which a major international effort yielded a large set of redshifts of very faint galaxies to act as a training and testing set for the photometric technique. From photometric red-shifts, we now have in hand samples of thousands of galaxies beyond z = 2 with known colors and luminosities.

With statistically significant samples of galaxies at redshifts up to z = 4, we can start to address some of the basic questions about galaxy evolution. One of the most important, with impacts ranging from the overall metallicity of galaxies to the timing of galaxy evolution and the ionization of the intergalactic medium, is the cosmic starformation history. This may be expressed, for example, as the spatially averaged rate of star formation in solar masses per unit volume, as a function of redshift. The resulting plot is often known as a Madau diagram, or Madau-Lilly diagram, after the extensive discussion in a 1996 paper by Piero Madau and collaborators. That original derivation used photometric redshifts and UV luminosities from the Hubble Deep Field to imply a rapid increase in star-formation rate as we look back in redshift to about z = 1, a broad peak near z = 2, and a decline beyond that (while of course the actual behavior in time is the reverse of the redshift trends).

It was encouraging that the mean value of star formation over time from this work was close to the average expected from the metal content of present-day bright galaxies. Lennox Cowie and coworkers recognized as early as 1988 that a connection exists between the ultraviolet light emitted from galaxies and the rate of production of heavy elements, since the same stars that produce most of the heavy elements are important contributors to ultraviolet radiation. Thus, the typical metal content of the Universe today (as judged from bright galaxies such as our own) implies a total surface brightness, or sum from individual galaxies, in the ultraviolet. This total is slightly greater than the integral of the Madau et al. (1996) star-formation history, so at least we have a consistency check.

Other techniques, however, have indicated that the early history of star formation was more active and extensive than this. At issue is how much star formation was missed by the sample in the Hubble Deep Field, either because of internal dust absorption or because it took place in environments which were too dim to appear in the catalog. Several analyses, some starting from the same data, show star formation being either comparably active or actually more intense to z > 4 as compared with z = 2. Furthermore, recent results on the intergalactic medium show that the Cowie limit on cumulative star formation should be higher by a poorly known factor, since substantial masses in heavy elements were expelled by galaxies at early epochs and were not accounted for in that original value. This material is seen today as the metals in the hot gas in clusters of galaxies, and as the highly-ionized absorption seen from the intergalactic medium in O vi and similar absorption lines.

In view of its importance, and discrepancies in addressing this issue, a variety of techniques must be brought to bear in tracing the cosmic history of star formation. Infrared observations are particularly important, being insensitive to dust absorption and in fact relying on it by measuring the total amount of starlight which has been absorbed to heat interstellar dust. In our neighborhood, strong far-infrared emission is a hallmark of star-forming galaxies, and submillimeter observations have traced IR-selected galaxies to objects so faint that their redshifts cannot be accurately measured yet. Mid-infrared surveys of galaxies are an important goal of Hubble's successor, the James Webb Space Telescope (JWST), and the international millimeter-astronomy array ALMA will excel at detecting the dust from very high-redshift objects. In our vicinity the GALEX satellite has nearly completed its survey of the entire sky in the ultraviolet, for the first deep and unbiased census of unobscured star formation in the local universe (which is why the name stands for Galaxy Evolution Explorer). The current status of our knowledge of cosmic star-forming history is summarized in Figure 5.6, including results from UV- and IR-based surveys.

0 12 3 4 redshift z

0 12 3 4 redshift z

Figure 5.6. Cosmic star-forming history. This is a schematic version of the "Madau diagram", showing the comoving density of star formation over cosmic time. The lower curve is evaluated from the amount of star formation actually found from nearby surveys and the ultraviolet emission from high-redshift galaxies in the Hubble Deep Fields. The upper curve indicates the amount of star formation inferred from infrared observations and variously corrected analyses of the Deep Field data. In both cases, the inferred overall rate of star formation requires extrapolation upward by a factor near 40 to allow for the dimmer stars which are dominant by mass but unimportant in the luminosity of star-forming systems.

5.4.1 Host galaxies of gamma-ray bursts

In a more speculative vein, it may become possible to trace the history of star formation with such exotica as gamma-ray bursts. With substantial evidence that these represent beamed emissions from a species of supernova, they form an easy-to-detect subsample of massive star deaths which we can already see to large redshifts. As a newly applicable tracer of cosmic history, their story should perhaps be set out in detail.

One of the four Great Observatories launched by NASA, the Compton Gamma-Ray Observatory or CGRO marked significant advances in our ability to study the universe at high energies, in several ways. It carried detectors sensitive to hard X-rays and gamma rays over the thousandfold range from 30 keV to 30 GeV. Compton operated from 1991 to 2000, delivering on its promise to uncover new facets of the violent high-energy Universe. It detected numerous AGNs at gamma-ray energies, yielding new evidence of relativistic motion in their jets, and provided a census of repeating high-energy sources in the Milky Way. The mission also gave one of the key clues to the long-standing question of gamma-ray bursts, one which shows them to trace the history of star formation in an unexpected way.

The discovery of gamma-ray bursts was a byproduct of the Cold War. With the Limited Test-Ban Treaty, the nuclear powers of 1963 agreed to refrain from atmospheric (and space-borne) tests of nuclear explosions. Verification of compliance was a significant issue in ratification, to be carried out with what later became well known as "national technical means'' (i.e., satellites). The USA launched the Vela series of satellites starting in 1963, on high 112-hour orbits which not only insured uninterrupted coverage of the eastern hemisphere, but would also detect clandestine nuclear tests on the lunar farside through emission from the expanding debris cloud. The second batch of Vela spacecraft, launched beginning in May 1969, carried improved detectors. Starting two months later, four of these Vela satellites detected unexpected bursts of gamma rays—not from Earthly nuclear arms, but from random directions in deep space.1 The large orbits allowed crude directional determination, from the arrival times of the bursts as measured by different satellites. This allowed Ian Strong to analyze data collected by Ray Klebesadel at Los Alamos to show that they came from neither the Sun nor the Earth, thus representing a new and exciting astronomical phenomenon (Klebesadel et al., 1973). Over a ten-year period, these satellites detected 73 cosmic gamma-ray bursts (with one retroactively found in 1967 data that had been insufficient on its own to warrant detailed analysis).

Models for the production of these bursts ran a wide gamut, since the only information available was the rate, total detected energy (fluence), and crude spectral shape of the bursts. Neutron stars were popular sites, since their deep potential wells could allow the liberation of vast amounts of gravitational energy. There were

1 Vela 6911, far outlasting its design lifetime, may have finally done its designed job on September 22, 1979, when an optical flash was seen over the Indian Ocean. This has been widely rumored to be the test of a South African nuclear device, although open sources do not allow a firm conclusion.

calculations of the impact of asteroids with neutron stars, and of neutron-star mergers at the end of the orbital decay of a binary system. The general argument for burst production near neutron stars seemed compelling enough that by the launch of the Compton Observatory, most workers expected its results to trace the structure of neutron stars in the Galaxy.

Reality was to prove quite different. One of the most intriguing early results from CGRO was that the statistical properties of its large catalog of gamma-ray bursts confounded all expectations based on events involving galactic neutron stars. Its burst-detection system (BATSE, the Burst and Transient Source Experiment) incorporated eight detectors at the corners of the spacecraft structure, to record bursts from any direction not occulted by the Earth. The BATSE detectors could determine the position of each burst in the sky to only a few degrees' accuracy, but that was enough to show no concentration in the galactic plane or the galactic center. In fact, through the end of the nine-year mission, as a total of 2704 bursts were observed, the angular distribution remained as random as statistical tests could show. This in itself only meant that we were observing the sources within some volume that made them isotropic. In principle, sources among the nearest stars, the Oort Cloud, distant galaxies, or an encircling fleet of invading starships would satisfy this requirement. But even more telling was the distribution of bursts in intensity, in the same kind of log N-log S diagram that had proven so powerful in establishing the evolution of quasars and radio galaxies. This time, we saw too few faint bursts compared with bright ones, a situation that implies that we are somehow centered in the distribution. This very non-Copernican situation could be satisfied by some kind of very large and hollow halo of sources centered on the Milky Way, so large that we would be nearly centered in it but not yet seeing any similar halo around Andromeda, or by a cosmologically evolving population, in which we would be central in time (equiva-lently in redshift). To extragalactic astronomers, the log N-log S diagram fairly screamed "cosmologically distant''.

Demonstrating this required an additional leap. It had been clear from early in the study of gamma-ray bursts that they do not have bright or otherwise obvious counterparts at other wavelengths, and faint counterparts cannot be found until the source's localization is better than the inverse of the number density of comparably faint sources (say, in the optical band). Individual detectors still have poor angular precision, and arrival-time triangulation from interplanetary spacecraft worked for too few bursts to be helpful (i.e., no clear counterparts emerged for classical bursts). This finally changed through the use of the X-ray tails of burst spectra. Strong bursts produce enough lower-energy radiation to be detected with imaging X-ray telescopes, easily bringing positional accuracies of an arcminute within reach. A field of view that small can be efficiently searched for variable (fading) objects, a strong hint that an optical object might be associated with transient gamma-ray emission. To carry this out successfully required the combined gamma-ray and X-ray detectors on the Italian/Dutch BeppoSAX satellite, launched on April 30, 1996. Its imaging X-ray telescope delivers 1' resolution in the 0.1-10 keV band. The burst observed on February 28, 1997 (known as GRB 970228) was the first to be located well enough to find an unambiguous optical counterpart, whose redshift proved to be significant

Figure 5.7. The sky distribution of gamma-ray bursts detected by the Compton Gamma-Ray Observatory, from the Fourth BATSE Catalog. This map is shown in galactic coordinates, with the galactic center in the middle. The equal-area projection makes it clear that the mean number per unit sky area is the same over the whole celestial sphere, with no concentrations toward the galactic center of the galactic plane, such as were expected if these bursts came from galactic stellar remnants. Individual bursts have a location error typically 2-3°, which may be compared with the grid lines spaced at 30° intervals. (These data are available from http:Hgamma.ray. msfc.nasa.gov/batse/grb/catalog/current/)

Figure 5.7. The sky distribution of gamma-ray bursts detected by the Compton Gamma-Ray Observatory, from the Fourth BATSE Catalog. This map is shown in galactic coordinates, with the galactic center in the middle. The equal-area projection makes it clear that the mean number per unit sky area is the same over the whole celestial sphere, with no concentrations toward the galactic center of the galactic plane, such as were expected if these bursts came from galactic stellar remnants. Individual bursts have a location error typically 2-3°, which may be compared with the grid lines spaced at 30° intervals. (These data are available from http:Hgamma.ray. msfc.nasa.gov/batse/grb/catalog/current/)

at z = 0.695. As of early 2007, there are 118 redshifts for counterparts of so-called long bursts (the short bursts might still be a different class altogether, since we have as yet only a handful of counterparts suggesting a different environment than the more familiar long bursts). The burst afterglows seen optically show no distinguishing emission lines, but redshifts can be estimated quite closely when the onset of the Lyman a forest appears against their continuum radiation, or a lower limit is set when there is a metal-line system from the gas around some galaxy along the lines of sight. Many of these redshifts properly pertain to the host galaxy, observed after the burst radiation has faded. After three decades of mystery, we finally know where gamma-ray bursts are. What can we ascertain about what exactly makes them?

The large redshifts for the host galaxies of gamma-ray bursts imply enormous radiated powers for these sources unless their radiation is relativistically beamed. In fact, production of such large amounts of gamma radiation almost certainly means that beaming is important in what we see, so the question is more precisely whether this beaming is confined to certain directions. If so, the energy requirements are relaxed accordingly. One model finding considerable favor with regard to physical plausibility involves a hypernova—the collapse of a high-mass star in which a central black hole forms, surrounded by a transient and very dense accreting disk of former stellar matter. This configuration would then launch relativistic jets along the poles, much as seen for accreting neutron stars and black holes from SS 433 to quasars, and a gamma-ray burst would be near the direction of each jet. Indeed, one gamma-ray burst seems to have been associated with the explosion of a peculiar supernova, GRB 980425 at z = 0.0085 which was associated with SN 1998bw in the galaxy ESO 184-G82. The connection with supernovae has been further strengthened by X-ray measurements of the afterglow of burst 011211, which detected spectral lines from magnesium, silicon, and other elements expected to be produced in a core-collapse supernova, blueshifted by about 0.1c with respect to the parent galaxy (itself at z = 2.14).

The gamma radiation from such a supernova would be associated with initial breakout of the jet through the photosphere, when we would see the shocked material, particularly internal shocks within the jet, Doppler-boosted by a factor 7 « 100. The optical and IR afterglows in this model come as the jet material cools and is decelerated by entraining stellar and wind material. In a simple geometric picture, the fading pattern of these afterglows offers a way to distinguish between this jet picture and one in which some enormous explosion produces a relativistically expanding fireball. As relativistically outflowing material (in whatever geometry) decelerates, its beaming angle « 1/7 increases, so that we observe radiation from a larger fraction of the whole outflow. For this reason, we will see a rate of fading which is the combination of each volume element sending us less radiation and seeing more volume elements. If the radiating region is a jet, the fading will accelerate once the beaming angle becomes comparable with the cone angle of the jet, since there is no additional volume to have radiation beamed from. Thus, in the simple jet scheme, we expect a break in the flux decline of optical and IR counterparts. There have been several (though not all) afterglows which do show such breaks in fading, with a much more rapid drop in flux starting a few days after the initial outburst was observed.

The high typical redshifts of the gamma-ray bursts, and the fact that these had to be relatively bright bursts to be localized and identified for redshift measurements, make these objects very attractive in studying galaxy evolution. This small sample already has a median redshift 1.0, a depth achieved for optical galaxy surveys only by going very deep (in fact, deeper than any existing spectroscopic samples outside the Hubble Deep Fields). Being able to measure even statistical properties of much fainter bursts could trace their occurrence far into the early Universe, and, if we can understand how they relate to the history of star formation, provide a completely new window on the earliest epochs of galaxy assembly.

The hypernova scheme implies that gamma-ray bursts trace the death rate of some kinds of stars. If these kinds of supernova are generally like type II, these must be massive and short-lived stars. On the other hand, if they require that a compact stellar remnant be pushed over its stability limit by accretion or merger with a companion, the progenitors could take cosmologically long times to reach the burst event. Some information on the kinds of stars involved comes from studies of the host

Figure 5.8. The fading afterglow of gamma-ray burst 991216, seen in these near-infrared (1.2-2.2 micron) images taken 2 and 4 days after burst arrival. This burst was found to have a redshift z — 1.02, and the very rapid fading of its optical/infrared afterglow suggested emission from a decelerating relativistic jet. (Data by the author using the 2.4-meter Hiltner telescope of the MDM Observatory, through the good offices of Ohio State University.)

Figure 5.8. The fading afterglow of gamma-ray burst 991216, seen in these near-infrared (1.2-2.2 micron) images taken 2 and 4 days after burst arrival. This burst was found to have a redshift z — 1.02, and the very rapid fading of its optical/infrared afterglow suggested emission from a decelerating relativistic jet. (Data by the author using the 2.4-meter Hiltner telescope of the MDM Observatory, through the good offices of Ohio State University.)

Figure 5.9. Histogram of secure gamma-ray burst redshifts, as collected by Jochen Greiner at http://www.mpe.mpg.de/~jcg/grbgen.html. Rapid identification and followup—made possible by the Swift mission—have been important in extending this to high values. No other known tracer has such a high characteristic redshift, or such a broad range, for flux-limited samples of the brightest examples.

Figure 5.9. Histogram of secure gamma-ray burst redshifts, as collected by Jochen Greiner at http://www.mpe.mpg.de/~jcg/grbgen.html. Rapid identification and followup—made possible by the Swift mission—have been important in extending this to high values. No other known tracer has such a high characteristic redshift, or such a broad range, for flux-limited samples of the brightest examples.

Figure 5.10. The host galaxy of a gamma-ray burst. This Hubble image of the fading afterglow of the first burst with an identified optical counterpart, GRB 970228 at z = 0.695, shows the host galaxy as well as the optical afterglow itself, on the edge of the galaxy image. The galaxy is detected over a region about one arcsecond in diameter, and could plausibly be a spiral with the burst projected about 3.5 kiloparsecs from the center. (This image is from data retrieved from the NASA Hubble Space Telescope Archive, with Andrew Fruchter as the original principal investigator.)

Figure 5.10. The host galaxy of a gamma-ray burst. This Hubble image of the fading afterglow of the first burst with an identified optical counterpart, GRB 970228 at z = 0.695, shows the host galaxy as well as the optical afterglow itself, on the edge of the galaxy image. The galaxy is detected over a region about one arcsecond in diameter, and could plausibly be a spiral with the burst projected about 3.5 kiloparsecs from the center. (This image is from data retrieved from the NASA Hubble Space Telescope Archive, with Andrew Fruchter as the original principal investigator.)

galaxies, especially from HST imaging. Most host galaxies, when one is detected, have a significantly clumpy structure and blue colors (Figure 5.10), linking the GRBs with star-forming regions.

Detailed study of the locations of afterglows in these host galaxies has shown that they roughly trace blue and ultraviolet starlight. This suggests that the progenitors were short-lived (as for core-collapse supernovae), not having had time to diffuse far through the galaxy since being produced in UV-bright star-forming regions. This supports the idea that GRBs can be used to trace the formation history of massive stars. Given the large redshifts found for relatively bright bursts, the fainter ones would be probing much higher redshifts, perhaps right up to the time of the first significant star formation. That would be a remarkable harvest indeed from the statistics of gamma-ray bursts.

Even more remarkable is the possibility that we are already observing gamma-ray bursts produced by the deaths of massive first-generation stars. These are predicted to explode in very powerful "hypernovae", and, if they produce narrow-beamed bursts at a relative rate similar to that required for type II supernovae in the later Universe, many of the bursts now being detected could come from these objects at z > 10. If this is so, we would expect to see their afterglows, as well as the hypernova emission itself, with deep observations further into the infrared than are typical for today. Lyman a and continuum absorption means that they are invisible at wavelengths shortward of 0.12(1 + z) microns. The flux levels we expect for the fading hypernovae are accessible for deep, targeted observations with existing equipment for z < 17, provided we have a position accurate enough to warrant spending large amounts of time on 8-10 m telescopes. At larger redshifts, spaceborne cryogenic instruments will be required, as atmospheric absorption, and especially emission, become rapidly worse with wavelength. Some bursts found at large red-shifts (such as 050904 at z = 6.3) have been quite bright in gamma rays, indicating that we may already be detecting similar events to much higher redshift. There is a real possibility that the non-Euclidean dropoff in the counts of bursts versus intensity is a measurement of the first appearance of stars in the Universe.

A significant improvement in the localization rate, and hence statistics, of bursts came with the 2004 launch of Swift. This satellite's name derives from the goal of "catching gamma-ray bursts on the fly'', generating identifications rapidly using on-board instruments for rapid response by ground-based facilities. This mission is deeply international, with co-investigators representing 40 institutions in 9 countries. Bursts are initially detected using a coded-mask detector (BAT) sensitive to hard X-rays (up to 150 keV). A coded mask, in which different parts of the detector see different subsets of the field of view, offers a fast and robust way to localize transient sources. This technique is very forgiving of partial detector failures, and allows fast algorithms for on-board estimation of the source location. This instrument yields immediate positions with a resolution of 22 arcminutes, accurate enough for the satellite to slew autonomously so as to place a burst within the field of view of an imaging X-ray telescope (XRT) and UV/optical telescope (UVOT). The XRT has 15'' spatial resolution, and is sensitive over the softer energy range 0.2-10 keV. The UVOT by itself can yield superb locations for optical counterparts, with 0.5'' pixels spanning a 17' field of view illuminated by a 30-cm primary mirror. A significant number of bursts will be seen serendipitously in the narrow fields of UVOT and XRT, independent of targeting from the gamma rays. Such an event had occurred prior to Swift only once in the history of GRB research, with the measurement of an optical flash by the ROTSE experiment in New Mexico while the gamma-ray burst was still in progress.

The evidence connecting gamma-ray bursts to supernovae is strong, although much remains to be done in understanding what kinds of supernovae from what kinds of progenitors. The data in hand indicate that these events may be the first to show us the earliest generations of stars, as our ability to observe them extends more deeply in other wavebands.

Was this article helpful?

0 0
Telescopes Mastery

Telescopes Mastery

Through this ebook, you are going to learn what you will need to know all about the telescopes that can provide a fun and rewarding hobby for you and your family!

Get My Free Ebook


Post a comment