ABSTRACT
We present new constraints on the star formation histories of the ultra-faint dwarf (UFD) galaxies, using deep photometry obtained with the Hubble Space Telescope (HST). A galaxy class recently discovered in the Sloan Digital Sky Survey, the UFDs appear to be an extension of the classical dwarf spheroidals to low luminosities, offering a new front in efforts to understand the missing satellite problem. They are the least luminous, most dark-matter-dominated, and least chemically evolved galaxies known. Our HST survey of six UFDs seeks to determine if these galaxies are true fossils from the early universe. We present here the preliminary analysis of three UFD galaxies: Hercules, Leo IV, and Ursa Major I. Classical dwarf spheroidals of the Local Group exhibit extended star formation histories, but these three Milky Way satellites are at least as old as the ancient globular cluster M92, with no evidence for intermediate-age populations. Their ages also appear to be synchronized to within ∼1 Gyr of each other, as might be expected if their star formation was truncated by a global event, such as reionization.
Export citation and abstract BibTeX RIS
1. INTRODUCTION
Although the lambda cold dark matter paradigm is consistent with many observable phenomena, discrepancies arise at small scales. One of the most prominent issues is that it predicts many more dark-matter halos than are actually seen as dwarf galaxies (e.g., Moore et al. 1999). A possible solution has arisen with the recent discovery of additional satellites around the Milky Way (e.g., Willman et al. 2005; Zucker et al. 2006; Belokurov et al. 2007) and Andromeda (e.g., Zucker et al. 2007) in the Sloan Digital Sky Survey (York et al. 2000) and other wide-field surveys (e.g., McConnachie et al. 2009).
The newly discovered ultra-faint dwarf (UFD) galaxies appear to be an extension of the classical dwarf spheroidals (dSphs) to lower luminosities (MV ≳ −8 mag). UFD luminosities are comparable to those of globular clusters, but one distinction in the former is the presence of dark matter. Even massive globular clusters have mass-to-light ratios (M/LV) of ∼2 (e.g., Baumgardt et al. 2009; van de Ven et al. 2006), precluding significant dark matter. In contrast, all known dwarf galaxies have higher M/LV (Kalirai et al. 2010 and references therein). UFD kinematics are clearly dark-matter-dominated, with M/LV > 100 (e.g., Kleyna et al. 2005; Simon & Geha 2007; Muñoz et al. 2006), even where velocity dispersions have been revised downward (e.g., Koposov et al. 2011). The inferred dark-matter densities of dwarf galaxies suggest a high-redshift collapse for both classical dSphs and UFDs (z ∼ 12; Strigari et al. 2008), but the dSphs apparently continued to evolve (Orban et al. 2008; Weisz et al. 2011). In contrast, the UFDs are the least chemically evolved galaxies known, with abundance patterns that imply their star formation was brief (Frebel et al. 2010) and individual stellar metallicities as low as [Fe/H] = −3.7 (Norris et al. 2010). The strict conformance to a metallicity–luminosity relation for all Milky Way satellites limits the amount of tidal stripping to a factor of ∼3 in stellar mass (Kirby et al. 2011). Therefore, UFDs are not tidally stripped versions of classical dSphs (see also Peñarrubia et al. 2008; Norris et al. 2010).
As one way of solving the missing satellite problem, galaxy formation simulations assume that UFDs formed the bulk of their stars prior to the epoch of reionization (e.g., Tumlinson 2010; Muñoz et al. 2009; Bovill & Ricotti 2009; Koposov et al. 2009). Mechanisms that could drive an early termination of star formation include reionization, gas depletion, and supernova feedback. Using the Hubble Space Telescope (HST), we are undertaking a deep imaging survey of UFDs that reaches the old main sequence (MS) in each galaxy, yielding high-precision color–magnitude diagrams (CMDs) that provide sensitive probes of their star formation histories. The program includes Hercules, Leo IV, Ursa Major I, Bootes I, Coma Berenices, and Canes Venatici II. Here, we give preliminary results for the first three galaxies.
2. OBSERVATIONS AND DATA REDUCTION
We obtained deep optical images of each galaxy (Table 1) using the F606W and F814W filters on the Advanced Camera for Surveys (ACS). These filters efficiently enable a high signal-to-noise ratio (S/N) on the stellar MS and facilitate comparison with other HST programs exploring the Local Group. The image area and depth were chosen to provide a few hundred stars on the upper MS, with an S/N of ∼100 near the turnoff. Our image processing includes the latest pixel-based correction (Anderson & Bedin 2010; J. Anderson 2012, in preparation) for charge-transfer inefficiency (CTI). All images were dithered to mitigate detector artifacts and enable resampling of the point-spread function (PSF). We co-added the images for each filter in a given tile using the IRAF drizzle package (Fruchter & Hook 2002), including masks for cosmic rays and hot pixels, resulting in geometrically correct images with a plate scale of 0.03'' pixel−1 and an area of approximately 210'' × 220''.
Table 1. Observations
| Exposure Per Tile | 50% Complete | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| R.A. | Decl. | (m − M)V | E(B − V) | MV | [Fe/H] | F606W | F814W | F606W | F814W | |||
| Name | (J2000) | (J2000) | (mag) | (mag) | (mag) | 〈[Fe/H]〉a | rmsb | Tiles | (s) | (s) | (mag) | (mag) |
| Hercules | 16:31:05 | +12:47:07 | 20.92 ± 0.06 | 0.08 ± 0.02 | −6.2c | −2.41 | 0.6 | 2 | 12,880 | 12,745 | 29.1 | 29.1 |
| Leo IV | 11:32:57 | −00:31:00 | 21.15 ± 0.08 | 0.05 ± 0.02 | −5.8d | −2.54 | 0.9 | 1 | 20,530 | 20,530 | 29.1 | 29.2 |
| Ursa Major I | 10:35:04 | +51:56:51 | 20.11 ± 0.04 | 0.04 ± 0.02 | −5.5e | −2.18 | 0.7 | 9 | 4,215 | 3,725 | 28.4 | 28.4 |
Notes. aKirby et al. (2011), based on Simon & Geha (2007) spectroscopy. bFor stars with <0.3 dex uncertainties. cSand et al. (2009). dde Jong et al. (2010). eMartin et al. (2008).
Download table as: ASCIITypeset image
We performed both aperture and PSF-fitting photometry using the DAOPHOT-II package (Stetson 1987), assuming a spatially variable PSF constructed from isolated stars. The final catalog combined aperture photometry for stars with photometric errors <0.02 mag and PSF-fitting photometry for the rest, all normalized to an infinite aperture. Due to the scarcity of bright stars, the uncertainty in the normalization is 0.02 mag. Our photometry is in the STMAG system: m = −2.5 × log10fλ − 21.1. For galaxies with multiple tiles, the ridge line in each tile was compared to that in other tiles; we then made small (∼0.01 mag) color adjustments to force agreement between tiles, ensuring the full catalog for a galaxy did not have larger photometric scatter than that in any tile. The catalogs were cleaned of background galaxies and stars with poor photometry, rejecting outliers in χ2 (from PSF fitting), PSF sharpness, and photometric error. We also rejected stars with bright neighbors or falling within extended background galaxies. We performed artificial star tests to evaluate photometric scatter and completeness, including CTI effects, with the same photometric routines used to create the photometric catalogs.
The CMD of each UFD galaxy is shown in Figure 1. Each UFD is clearly dominated by an ancient population. Field contamination is significant, but is particularly high in the vicinity of the red giant branch (RGB). The expected number of RGB stars can be estimated from the number of subgiant branch (SGB) stars, given the relative lifetimes implied by isochrones. The implied contamination varies from roughly 30% at 1 mag above the SGB to twice that at 2 mag above the SGB, where this contamination increases the apparent RGB breadth. In Hercules and Leo IV, the star counts are a bit lower than expectations from ground surveys (R. R. Muñoz et al. in preparation), while the counts in Ursa Major I are significantly lower, but the sample is large enough to judge its population relative to those of the other two galaxies.
Figure 1. CMD of each UFD (black points), with axes shifted to ease comparisons. For reference, we show the empirical ridge line for the MS, SGB, and RGB in M92 (dark blue curve), along with the HB locus in M92 (blue points). The M92 fiducial has been placed at the distance and reddening for each galaxy, matching the luminosity of HB stars (highlighted in red) and the color of the lower RGB stars in each UFD (accounting for the metallicity distribution). The light blue band is bounded by isochrones within 1 Gyr of M92's age (13.7 Gyr), at the M92 metallicity ([Fe/H] = −2.3), showing excellent agreement with the M92 ridge line. Because the UFD metallicities extend much lower than those in globular clusters, we also show an isochrone at the age of M92 but at the low extreme of the metallicity distribution ([Fe/H] = −3.2; dark green curve), bounded by isochrones offset by 1 Gyr (light green band). A blue straggler sequence is present in each galaxy, but there are no stars significantly younger than M92.
Download figure:
Standard image High-resolution image3. ANALYSIS
3.1. Comparison to M92
As one of the most ancient, metal-poor, and well-studied globular clusters, M92 is an important empirical fiducial for the UFD galaxies. It was previously observed by Brown et al. (2005) with ACS, using the same filters employed here, as one of six empirical templates for deep CMD analyses of Andromeda (e.g., Brown et al. 2006). Since then, estimates of the M92 distance modulus have generally drifted ∼0.1 mag higher, and estimates for the metallicity have generally decreased by ∼0.2 dex. We assume (m − M)o = 14.62 mag, which is the mean of Paust et al. (2007; 14.60 ± 0.09 mag), Del Principe et al. (2005; 14.62 ± 0.1 mag), and Sollima et al. (2006; 14.65 ± 0.1 mag). We assume E(B − V) = 0.023 mag (Schlegel et al. 1998) and [Fe/H] = −2.3 (Harris 1996). Because the UFD metallicity distribution extends much lower than that of any globular cluster, the age determination must rely upon theoretical isochrones (see Section 3.2), but a comparison to M92 is still instructive.
In our analysis, the largest uncertainties are the distances of the UFDs relative to the globular cluster calibrators of our isochrone library, with emphasis on M92. We determined apparent distance moduli by minimizing the luminosity differences between the UFD horizontal branch (HB) stars (Figure 1, red points) and the horizontal region of the M92 HB locus (blue points). We determined the foreground extinction by minimizing the color differences between UFD stars on the lower RGB and the RGB color distribution implied by the spectroscopic metallicities associated with each galaxy (Simon & Geha 2007; Kirby et al. 2008), anchored to M92 (which is near the mean metallicity of each UFD). The uncertainties in our determinations of distance and reddening are from the standard deviation in Monte Carlo realizations of these fits, excluding any uncertainties in the distance and reddening of M92. We show our derived apparent distance moduli and reddening values in Table 1, but note comparisons to the recent literature here. For Hercules, Sand et al. (2009) found (m − M)o = 20.625 ± 0.1 mag and assumed E(B − V) = 0.064 mag (Schlegel et al. 1998), implying (m − M)V = 20.8 mag. For Leo IV, Moretti et al. (2009) found (m − M)o = 20.94 ± 0.07 mag and E(B − V) = 0.04 ± 0.01 mag using RR Lyrae stars (associated with an old population), implying (m − M)V = 21.06 mag. For Ursa Major I, Okamoto et al. (2008) found (m − M)o = 19.93 ± 0.1 mag and assumed E(B − V) = 0.019 mag (Schlegel et al. 1998), implying (m − M)V = 19.99 mag. Note that if we were to assume the shorter distances and/or lower reddening in the literature, the UFD ages relative to M92 would increase. The observed UFD CMDs would not change, but the M92 ridge line would shift brighter and bluer, making the MS turnoff in each galaxy appear fainter and redder (and thus older) relative to that in M92, which would significantly violate the age of the universe (13.75 ± 0.11; Jarosik et al. 2011) unless other factors (e.g., CNO abundances) were adjusted accordingly.
With these caveats, we show the comparison of each UFD to M92 in Figure 1. HB morphology is most sensitive to metallicity (but also age and the abundances of He and CNO), with the HB color distribution shifting toward the blue at lower [Fe/H]. The colors of the HB stars in each UFD are consistent with their metallicities, in the sense that 〈[Fe/H]〉 is slightly higher in Ursa Major I than in Hercules or Leo IV (Table 1), and its HB stars are somewhat redder, but this may be due to small number statistics. The RGB appears slightly wider than one would expect from the photometric errors and metallicity distribution, but this is uncertain, given the field contamination. The MS turnoff and SGB in each UFD extends slightly brighter and bluer than those features on the M92 ridge line. If we did not know the metallicity distributions, we would conclude that the brighter and bluer stars have ages a few Gyr younger or metallicities up to ∼1 dex lower than M92. However, we know the UFD metallicities extend below [Fe/H] = −3, so these stars are consistent in age with M92, as we show below.
3.2. Comparison to Theoretical Isochrones
Because the UFDs are among the most metal-poor populations in the local universe, we cannot rely solely upon empirical stellar population templates (e.g., globular clusters) to analyze them. We must extend these empirical templates to lower metallicities using theoretical models. For this purpose, we employ the Victoria-Regina isochrones (VandenBerg et al. 2012).
Compared to the previous version of the code (VandenBerg et al. 2006), the current version includes, among other things, the effects of He diffusion, the latest improvements to the H-burning nuclear reaction rates, and an update to the Asplund et al. (2009) solar metals mixture. The net effect of these changes is to reduce the age at a given turnoff luminosity by ∼0.5 Gyr. Our purpose here is to use these isochrones to derive ages relative to those of the globular clusters, but these effects should be kept in mind when considering absolute ages.
Brown et al. (2005) produced a set of empirical population templates in the ACS F606W and F814W filters, spanning a wide metallicity range, based upon Galactic globular and old open clusters. They then calibrated a transformation of the Victoria-Regina isochrones into the same filters, producing agreement at the 1% level. Here, we have adjusted this transformation by 0.03 mag in color. This adjustment accounts for the new version of the Victoria-Regina isochrones and the current state of the literature regarding cluster parameters. These isochrones match the ACS CMDs of 47 Tuc at an age of 12.3 Gyr, NGC 6752 at an age of 13.2 Gyr, and M92 at an age of 13.7 Gyr. As discussed in Section 3.1, M92 is the most relevant calibrator, but the other clusters demonstrate that the isochrones are valid over a wide range of [Fe/H] (−0.7 to −2.3). We stress that these absolute ages depend critically upon the assumed distance moduli and oxygen abundances.
In Figure 1, we show the expected location of stars in the ACS CMDs for ages within a Gyr of the M92 age (i.e., 12.7–14.7 Gyr) at two different metallicities: that of M92 (i.e., [Fe/H] = − 2.3; blue curve) and the low extreme of the metallicity distribution in these UFDs ([Fe/H] = − 3.2; green curve). It is clear from the comparison that the luminosity of the MS turnoff and SGB in each galaxy is consistent with ancient metal-poor stars. The few stars immediately brighter and bluer than the old MS turnoff are blue stragglers, which are ubiquitous in old populations. If these were young stars, each UFD would have a star-formation history consisting of two delta functions in age, with a few percent of the stars at ∼4 Gyr and the rest ancient; continuous low-level star formation would produce intermediate-age MS and SGB stars that are not observed.
Although the MS turnoff becomes bluer and brighter at both younger ages and lower metallicities, this age–metallicity degeneracy can be broken by simultaneously fitting the MS, SGB, and RGB in a CMD. However, the scarcity of RGB stars and multitude of field stars precludes this type of fit here. Instead, we rely upon the observed metallicity distribution from spectral synthesis of neutral Fe lines (Kirby et al. 2008; based upon Keck spectroscopy from Simon & Geha 2007). For the current analysis, we will use the relatively well-populated CMD of Hercules as an example, employing the 15 RGB stars with low spectroscopic errors (<0.3 dex) observed to date. We constructed a library of CMD probability clouds representing simple stellar populations spanning −1.4 > [Fe/H] > − 3.2 and ages of 10–16 Gyr. The probability clouds were populated using the Victoria-Regina isochrones, a Salpeter initial mass function (IMF), and the photometric scattering and completeness from the artificial star tests. Synthetic CMDs constructed from this library were fit to the observed CMD by allowing the ages of the components to float while holding the metallicity distribution fixed, using a downhill simplex to minimize the Maximum Likelihood statistic of Dolphin (2002). The fits were repeated with a fine grid of starting conditions to avoid a spurious result driven by local minima in the fit. Uncertainties in the derived age were determined by fitting repeated Monte Carlo realizations of the observed CMD. The CMD was fit in the vicinity of the MS turnoff (24.0 < mF814W < 25.7 mag), given its sensitivity to age, while excluding the lower MS stars that provide no age leverage. Note that the mass range spanned by our fit is only ∼0.03 M☉, and thus insensitive to the assumed IMF. We find a mean age of 13.6 Gyr for Hercules. Formally, the statistical uncertainty is 0.2 Gyr, consistent with the age of M92 (13.7 Gyr). The systematic uncertainties from distance and [O/Fe] are each potentially as large as 0.6 Gyr.
To demonstrate the sensitivity of these CMDs to age, we show in Figure 2 three simulated CMDs for Hercules. Each CMD assumes the observed spectroscopic metallicity distribution, under three age assumptions: 1 Gyr younger than M92 (left panel), the same age as M92 (middle panel), and 1 Gyr older than M92 (right panel). It is clear that the best match to the observed Hercules CMD (Figure 1) is the middle panel, where the populations have the same age as M92. Although the population statistics are best in Hercules, the similarity of the Leo IV and Ursa Major I CMDs implies that they also host populations of approximately the same age. None of these galaxies appears to host a significant population of stars younger than M92.
Figure 2. Simulated CMDs for Hercules (MS–RGB) under three assumptions: a population 1 Gyr younger than M92 (left panel), a population at the age of M92 (middle panel), and a population 1 Gyr older than M92 (right panel). To ease comparison with the observed Hercules CMD, these simulations are plotted on the same scale as Figure 1 and include the same empirical and theoretical fiducials (colored curves). The middle panel best matches the observed Hercules CMD, given the luminosity distribution of the SGB stars and the color of the MS turnoff (bluest point on the MS).
Download figure:
Standard image High-resolution image
In the Hercules spectroscopy, approximately 20% of the RGB stars fall at [Fe/H] = −1.7 ± 0.3. In the Hercules CMD (Figure 1), the stars immediately below the M92 SGB, especially near the MS, are consistent with such stars at ages within a Gyr of M92 (cf. Figure 2), although the exact number is uncertain, given the field contamination. However, if any of these stars at [Fe/H] > − 2 were as much as 2 Gyr younger than M92, they would be difficult to discern in the Hercules CMD, because the SGB for such stars would fall within the dominant SGB observed for Hercules (i.e., within the region bounded by the blue and green fiducials of Figure 1). Thus, while it is clear that most stars in Hercules are as old as those in M92, we cannot at this time rule out a minority subpopulation of stars (up to ∼10% of the total population) that are 1–2 Gyr younger at the high end of the Hercules metallicity distribution.
4. DISCUSSION
The three UFD galaxies here are among the more distant of those known in the Milky Way system, at 100–150 kpc, so ground-based CMDs of each implied they were old (>10 Gyr), but could not put tight constraints on their ages (e.g., Sand et al. 2009; Sand et al. 2010; Okamoto et al. 2008, 2012; Adén et al. 2011). Our HST observations reach well below the MS turnoff in each galaxy, revealing that all three host truly ancient metal-poor populations. The majority of the stars in each galaxy must have ages within 1 Gyr of M92's age, with younger ages strongly ruled out, although we cannot exclude a trace population of stars 1–2 Gyr younger than M92 at the high end of the metallicity distribution.
Two UFD galaxies (Bootes I and Coma Berenices) are significantly closer (44–66 kpc), such that ground-based CMDs can place constraints on their ages approaching those possible with HST (Muñoz et al. 2010; Okamoto et al. 2012). These CMDs also imply ages approximately as old as M92, although the use of distinct bandpasses hampers accurate comparisons to our observations. Our HST survey includes these galaxies, enabling accurate age measurements in a significant sample of UFDs, with all observed in the same photometric system as each other and the most ancient globular clusters.
If we include Bootes I and Coma Berenices, it seems likely that at least five UFD galaxies have ages consistent with that of the oldest known globular cluster (M92), with no evidence for significantly younger populations. This is in striking contrast to any other galaxy class in the local universe. Our external vantage point for Andromeda enables accurate ages throughout its halo, using the same instrument and techniques described here, yet all such measurements to date have found an extended star formation history, with significant numbers of stars younger than 10 Gyr (Brown et al. 2006, 2008). An HST survey of 60 dwarf galaxies within 4 Mpc found that most formed the bulk of their stars prior to z ∼ 1, but none were consistent with a purely ancient population (Weisz et al. 2011). The UFDs may be the only galaxies where star formation ended in the earliest epoch of the universe. If so, the apparent synchronicity to their star formation histories suggests a truncation induced by a global event, such as reionization (13.3 Gyr ago; Jarosik et al. 2011). These UFDs were likely the victims of reionization, rather than the agents, given the small numbers of stars available to produce ionizing photons.
Support for GO-12549 was provided by NASA through a grant from STScI, which is operated by AURA, Inc., under NASA contract NAS 5-26555. E. N. K. acknowledges support by NASA through Hubble Fellowship grant 51256.01 from STScI. A. R. acknowledges support from ASI via grant I/009/10/0. R. R. M. acknowledges support from the GEMINI-CONICYT Fund, allocated to the project No. 32080010.
Footnotes
- *
Based on observations made with the NASA/ESA Hubble Space Telescope, obtained at STScI, which is operated by AURA, Inc., under NASA contract NAS 5-26555.