Discovery of Extremely Metal-Poor Stars in Milky Way dSphs

A diffuse halo of stars and dark matter surrounds the Milky Way.  Most astronomers believe that some of the oldest stars in the Universe—nearly as old as the Universe itself—inhabit the stellar halo.  Part of the reason to expect these stars are so old is that they contain very little metals.  ("Metals" means every element except hydrogen and helium.)  Metals exist in the Universe because nuclear burning in stars processed hydrogen and helium into heavier elements.  That means that the first generations of stars had no metals, or at least very small amounts of metals.

One widely accepted theory of the formation of the Milky Way's stellar halo posits that small "building blocks" called dwarf spheroidal galaxies (dSphs) fell into each other by their mutual gravitational attraction.  Those building blocks are no longer distinct entities but instead a single enormous diffuse cloud of stars that surrounds the Milky Way.  However, some dSphs still exist today.  In the hierarchical assembly model, these dSphs would be the rare handful of latecomers to the Milky Way system which have not yet suffered the gravitational interactions that pull the small galaxies apart and melds them into the stellar halo.

Recently, some astronomers claimed that the stars in the halo are different enough from the stars in the surviving dSphs to cast doubt on the hierarchical assembly model.  Specifically, a small fraction of stars in the halo are extremely metal-poor.  Some have one ten thousandth the amount of metals that the Sun has.  Some have even less.  If dSphs built the stellar halo, then even the surviving dSphs should also have some extremely metal-poor stars.  The claim is that, on the other hand, some surviving dSphs—specifically the large, luminous ones—do not contain stars that are quite so metal-poor.

By measuring the metal content of very tiny, very faint surviving dSphs with a new technique, my group has discovered that certain dSphs do in fact contain stars nearly as metal-poor as those in the stellar halo.  The cumulative distribution of metallicity ([Fe/H]) shown above demonstrates that the extremely metal-poor tail of the faintest surviving dSphs has a shape similar to that of the Milky Way stellar halo after all.  The discrepancy at [Fe/H] > -2.3 is a result of excluding the more luminous surviving dSphs, whose stars are on average more metal-rich than the very faint dSphs represented by the black line.

While the discovery of extremely metal-poor stars in some surviving dSphs bolsters the hierarchical assembly model, it does not prove that all dSphs contain extremely metal-poor stars.  Therefore, the claim that the luminous dSphs are devoid of very metal-poor stars may be accurate.  The discovery of very metal-poor stars in the faint dSphs warrants a new look at the luminous dSphs with the same technique.  Presently, it is unclear whether the hierarchical assembly model depends on the existence of extremely metal-poor stars in luminous dSphs.  It is possible that the extremely metal-poor tail of the halo metallicity distribution formed from tiny dSphs alone.  Future work will apply the same technique used to discover the extremely metal-poor stars in small, faint dSphs to the stars in large, luminous dSphs.

Metallicities and Alpha Enhancements of Red Giants with Medium Resolution Spectra

Detailed tests of theories of hierarchical structure formation require chemical abundances of many stars at large distances. In additional to overall metallicity, models predict the alpha  enhancement of different dynamical components of the Milky Way and M31.  High dispersion spectroscopy can probe very few stars in the Milky Way satellite galaxies and no stars in M31. 

My dissertation employs a technique using medium resolution spectroscopy and spectral synthesis to extract metallicity and [α/Fe] for large samples of individual stars at large distances.  I have found that R ~ 6000 spectra reproduce high resolution [Fe/H] and [α/Fe] measurements of Galactic globular cluster stars. At signal-to-noise ratios for which Milky Way satellites are easily accessible, the typical error on metallicity is 0.15 dex, and the typical error on [α/Fe] is 0.2 dex. Unlike empirical metallicity estimators, such as the equivalent width of the Ca II triplet, this synthetic method does not depend on the metallicity range or alpha enhancement of calibrators.

This plot shows the mean metllaicities of seven Milky Way globular clusters.  The horizontal axis is the result from high dispersion spectroscopy, and the horizontal axis is my result from DEIMOS spectra.  Three clusters (NGC 2419, M79, and NGC 7492) have fewer than five stars observed in high resolution spectroscopy.  The measurements with DEIMOS are based on 32, 16, and 21 stars, respectively.  Despite the lower spectral resolution, the DEIMOS results may be more precise.

Confirming the Identities of z = 5 Lyα-Emitting Galaxies via Spectral Coaddition

DEEP2 target galaxies must be detected in the B, R, and I photometric bands. However, the occasional untargeted galaxy falls serendipitously on a DEIMOS spectroscopic slit. Very rarely, these "serendips" are very high redshift galaxies with Lyα emission visible in DEIMOS. The trick is to prove that the line really is Lyα.

One way to build confidence in the candidate sample is spectral coaddition, or stacking. A set of true Lyα spectra exhibits different properties than a set of impostors. For example, if we assume that the single lines are Hα, the most likely interloper, we would expect to see [NII] (two dashed lines in the left panels) and [SII] (two dashed lines in the right panels) emission. The top subset shows no clear [NII] or [SII] emission whereas the bottom subset does. Therefore, we have established confidence in identifying the top subset as Lyα emitters while proving that a large fraction of the bottom subset is actually Hα emitters.

Determining the Redshifts of Single-Line Emission Galaxies

The Deep Extragalactic Evolutionary Probe (DEEP2) has an 80% success rate at determining the redshifts of its 50,000 targeted galaxies.  One of the primary failure modes is the detection of only one emission line.  Ordinarily, spectroscopic redshifts rely on the ability to identify a pattern of lines.  However, photometric measurements can supply enough additional information to identify the single line and hence obtain a very precise spectroscopic redshift.

This project compares the optical colors and magnitudes of single-line galaxies to the same properties of a training set of galaxies with traditionally measured redshifts.  For each single line galaxy, we assign a probability to each of four possible line identities, and we modify those probabilities by priors, or selection rules.  For example, the single line may not be [OII] unless the line width is broad enough to allow the two lines of the [OII] doublet to blend together.  If the probability of one line identity clearly dominates over the other three, we have successfully determined the line identity and hence the galaxy's redshift.

In this example, a DEEP2 galaxy displays a single line at 8425 Å,  Its B-R and V-I colors place it squarely in the locus of galaxies at redshifts such that Hα is observed near 8425 Å.  Therefore, we can unambiguously identify the single line as Hα, putting this galaxy at z = 0.284.