Shocked POststarburst Galaxy Survey. III. The Ultraviolet Properties of SPOGs

There is an observed bimodality in morphology (Hubble 1926; Strateva et al. 2001; Baldry et al. 2004), color (Baade 1958; Tinsley 1978), star formation rates (SFRs), and gas fractions in the population of present-day galaxies. In color–magnitude space, this bimodality is seen as a red sequence and a blue cloud (Strateva et al. 2001; Baldry et al. 2004). The red sequence includes early-type galaxies (ETGs), which are relatively gas-poor, have redder optical colors, ellipsoidal morphologies, and typically a quenched star formation. Blue cloud galaxies, on the other hand, are usually late-type galaxies (LTGs), which are gas-rich, have bluer colors, flattened disk morphologies, and are actively forming stars.

From z = 1 → 0, the total mass of blue cloud galaxies has remained roughly constant (Noeske et al. 2007), while the red sequence has doubled in mass (Bell et al. 2012). This suggests that once its star formation is quenched, a blue cloud galaxy migrates to the red sequence (Harker et al. 2006). The reverse migration (from red sequence to blue cloud) does not commonly occur (Young et al. 2014), except in gas-rich mergers or other such extreme events (Kannappan et al. 2009). Few galaxies are seen in the intermediate "green valley" space of color–magnitude diagrams. It has been suggested that this lack is due to the rapid timescales of the transition from blue cloud to red sequence (Faber et al. 2007). While there is a correlation between galaxy morphology and SFR, it does not necessarily imply causation. Analyses with the Sloan Digital Sky Survey (SDSS), the Galaxy Evolution Explorer (GALEX; Martin et al. 2005), and the Galaxy Zoo (Lintott et al. 2008) data have shown that while morphological ETGs do transition rapidly through the green valley, LTGs do not; instead, they retain their morphologies as specific SFRs decline very slowly (Schawinski et al. 2014).

These differing rates of transition suggest that there may be several transitional states within the green valley. It has been proposed that star-forming galaxies could transition as a result of the cosmic supply of gas being shut off (e.g., at a critical halo mass), and depletion of the remaining gas (via secular or external means) over many billions of years (e.g., Larson et al. 1980; Schawinski et al. 2014; Peng et al. 2015; Lilly & Carollo 2016). ETGs, on the other hand, would require their gas reservoir to be exhausted on very short timescales after their morphology has transformed from that of an LTG (e.g., via a major or minor merger). Other processes have also been invoked to explain the transition of galaxies to the red sequence: ram pressure stripping and/or strangulation following the infall of a galaxy into a massive cluster potential (Gunn & Gott 1972; Chung et al. 2009); tidal disruption and harassment in group interactions (Bitsakis et al. 2014); morphological quenching (Martig et al. 2009, 2013); and active galactic nucleus (AGN) feedback (Di Matteo et al. 2005; Hopkins et al. 2008; Fabian 2012; Kormendy & Ho 2013).

Identifying galaxies in the midst of the transition, as star formation is being quenched, has not been so straightforward. Traditionally, post-starburst galaxies (also known as E + A or K + A because of the prevalence of A stars in an elliptical galaxy-like spectrum, which contains K stars; Dressler & Gunn 1983; Quintero et al. 2004) have been identified based on the presence of strong Balmer absorption lines, which select for intermediate-age A stars, and the absence of strong star formation lines, such as Hα and [O ii]λ3727 (e.g., Goto 2007). These criteria are able to select recently quenched galaxies, but they present an incomplete set because they miss objects with line emission from AGN (Cales et al. 2013), objects with substantial emission from post-asymptotic giant branch stars (post-AGB; Yan et al. 2006), and shocks (Rich et al. 2011; Alatalo et al. 2016a), as all of these phenomena can produce non-negligible [O ii] and Hα emission.

Identifying post-transition galaxies solely photometrically has the potential to revolutionize our understanding of galaxy evolution, removing the necessity of the observationally expensive follow-up spectroscopy. Notably, post-starburst galaxies exhibit near-uniform optical and near-infrared (NIR) properties (Kriek et al. 2010; Melnick & De Propris 2014). Post-starbursts also exhibit mid-IR colors that place them into a distinct section of color space (Ko et al. 2013; Yesuf et al. 2014; Alatalo et al. 2017). While mid-IR colors are effective at low redshift, they become increasingly difficult to observe at high redshift. In contrast, rest-frame UV becomes observable in the optical, making UV observations an excellent tool for identifying and characterizing transitioning galaxies, even at high redshift. While rest-frame UV is capable of identifying post-starbursts, directly predicting UV emission in these systems has remained challenging. Melnick & De Propris (2014) find that stellar population synthesis (SPS) models reproduce the spectral energy distributions (SEDs) of post-starburst galaxies well at optical and near-IR wavelengths, they systematically overpredict the observed UV fluxes of these galaxies. The authors attribute the discrepancy in UV to a combination of inadequate modeling of the synthetic SEDs, and non-uniform distribution of dust leading to an underestimation of the reddening of the intermediate-age populations.

Kaviraj et al. (2007) reconstructed the star formation histories of 38 post-starburst galaxies using UV and optical data from GALEX and SDSS, respectively, and found that the burst of star formation that dominates the post-starburst signature usually takes place within a Gyr, consistent with the presence of A stars. They found that short timescales for the burst (0.01–0.2 Gyr) and high stellar mass fractions (20%–60%) suggested high SFRs during the burst, resulting in a tight positive correlation with galaxy mass. The SFRs are comparable to those found in luminous and ultra-luminous infrared galaxies (LIRGs and ULIRGs, respectively) at low redshift, suggesting that low-redshift massive LIRGs may be the progenitors of massive post-starburst galaxies. The authors were also able to follow the quenching evolution of each modeled post-starburst galaxy, confirming that both the optical and UV colors evolve from blue to red. They also demonstrated that the interrelation between the optical and UV provide further constraints capable of identifying and understanding the evolving galaxy population. The ultraviolet is a wavelength regime that is able to pinpoint galaxies that are currently undergoing that transition. Wild et al. (2014) utilized principle component analysis to differentiate between galaxy types at z = 0.9–1.2 based on their SEDs. This analysis was able to directly pinpoint post-starburst galaxies, with the lynchpin wavelength coverage appearing in the UV.

Star formation is far from the only source of UV emission present in galaxies. Radiative shock waves induced by supersonic turbulence in gas reservoirs provide an additional source of emission especially relevant for transitioning galaxies. High-velocity shocks can ionize the gas and emit highly excited UV emission lines (Allen et al. 1998). Evidence of shocks is frequently seen in merging systems (Rich et al. 2011, 2014); galaxies in the outskirts of groups (Appleton et al. 2013) and clusters (Braglia et al. 2009); and AGN-dominated galaxies (Ogle et al. 2010; Villar Martín et al. 2014; Lanz et al. 2016), all processes that may be associated with the quenching of star formation and the transition of blue cloud galaxies into the red sequence. While the emission-line spectra will depend strongly on the physical and ionization structure of the shock, the total luminosity of the shock will effectively only depend on the gas density and shock velocity. Because shocks are characterized by highly ionized regions of high electron temperature, their spectra can contain several collisionally excited UV lines (Allen et al. 2008), capable of contaminating broadband measurements.

Traditionally, discriminating between different excitation mechanisms has relied on emission-line ratios, usually of optical wavelengths. Most notably, ratios of [N ii]λ6583/Hα versus [O iii]λ5007/Hβ, [S ii]λλ6717, 31/Hα versus [O iii] λ5007/Hβ, and [O i]λ6300/Hα versus [O iii] λ5007/Hβ have been used to distinguish between star-forming galaxies, composite AGN/H ii galaxies, Seyferts and low-ionization nuclear emission-line regions (LINERs; Heckman 1986), and objects excited by shockwave heating, in line diagnostic diagrams (sometimes referred to as "BPT/VO87 diagrams"; Baldwin et al. 1981; Veilleux & Osterbrock 1987; Kewley et al. 2006). Allen et al. (1998) showed that UV emission lines also provide a useful way to discern the presence of shocks in galaxies, particularly N iii λ991, N iii]λ1750, C iii λ977, C iv λ1549, C iii] λ1909, C ii] λ2326, and He ii λ1640.

The Shocked POststarburst Galaxies Survey (SPOGS; Alatalo et al. 2016a)¹⁷ seeks to identify galaxies in the process of transformation overlooked by classical selection criteria, with nebular emission lines excited by shocks rather than star formation. The SPOGS selection criteria are formally defined in Section 2 of Alatalo et al. (2016a). Briefly, the catalog selects galaxies from the SDSS DR7 (Abazajian et al. 2009) with z < 0.2, using the Oh–Sarzi–Schawinski–Yi (OSSY) absorption and emission-line catalog (Oh et al. 2011) to determine line strengths. Following continuum and emission-line signal-to-noise (S/N) cuts to ensure robust detections of spectral lines, a criterion is used to select strong Balmer absorption lines for a stellar population with a recent burst of star formation (Falkenberg et al. 2009). Specifically, we use a threshold of EW(Hδ) > 5 Å, consistent with the Balmer post-starburst selection criteria of Goto (2007) and Falkenberg et al. (2009). Next, shock boundaries are defined based on grids of shock models generated from MAPPINGS III (Dopita & Sutherland 1995) from the following optical emission-line ratios: [N ii]/Hα, [S ii]/Hα, [O i]/Hα, and [O iii]/Hβ (see Alatalo et al. 2016a for details). The remaining galaxies are subclassified based on their line diagnostic ratios as either Seyferts, LINERS, composites, star-forming, or ambiguous according to the classification lines of Kewley et al. (2006). To limit contamination, we exclude galaxies that are classified as star-forming or composite in all of the three line diagnostic diagrams. The final SPOGS catalog contains the 1067 objects (which we refer to as the SPOGs sample) that meet the aforementioned criteria.

In Section 2, we explain our methods for obtaining GALEX UV and SDSS optical photometry for our SPOGs sample, as well as our comparison samples. In Section 3, we present the UV properties of SPOGs. In Section 4, we discuss and provide interpretations for the large scatter in UV photometry and blue UV colors the SPOGs. In Section 5, we provide a summary. The cosmological parameters H₀ = 70 km s⁻¹ Mpc⁻¹, Ω_M = 0.3 and Ω_Λ = 0.7 are assumed throughout (Spergel et al. 2007).

GALEX has two UV filters: the far-ultraviolet (FUV; 1350–1750 Å) centered at 1530 Å and the near-ultraviolet (NUV; 1750–2750 Å) centered at 2310 Å. We obtained GALEX photometry from the Catalog Archive Server Jobs System (CasJobs),¹⁸ which provides access to the GALEX Data Release 6 (GR6) object catalogs (Bianchi et al. 2014). We used the "fuv_mag" and "nuv_mag" magnitudes, which correspond to the flux within elliptical "Kron" apertures (with semimajor axis scaled to 2.5 times the first moment of each source's radial profile, as first suggested by Kron 1980). We also visually inspected intensity maps in both FUV and NUV, to ensure only robust detections (i.e., undeniably present in the image) were included in the analysis. Although most detections were only discernible point sources, GALEX did successfully resolve a few objects, 12 of which are shown in Figure 1. Our selection and inspection resulted in 444 SPOGs detected in both bands, 365 detected in NUV-only, and 24 detected only in FUV, for a total of 833 SPOGs with UV detections. The other 234 SPOGs were not observed by GALEX in either band. We used the same SDSS Data Release 9 (Ahn et al. 2012) photometry (u, g, r, i, z bands) to be consistent with Alatalo et al. (2016a).

Zoom In Zoom Out Reset image size

2.1. Comparison Samples

2.2. Photometric Corrections

For all of the comparison samples, SDSS magnitudes are taken from DR9, the same photometry used for SPOGs. Both GALEX and SDSS magnitudes are corrected for Galactic extinction using the Schlafly & Finkbeiner (2011) extinction map provided by the NASA/IPAC Extragalactic Database (NED), assuming a Fitzpatrick (1999) reddening law, R_V = A_V/E(B − V) = 3.1, and GALEX extinction laws: A_NUV/E(B − V) = 8.741 and A_FUV/E(B − V) = 8.376 from Wyder et al. (2005). Corrections for intrinsic extinction are done in two ways. For the AGN and interacting samples, which do not have E(B − V) values from the OSSY catalog (Oh et al. 2011), we apply the analytic formulae of Cho & Park (2009), using the SDSS isophotal axis ratios (a/b) of the 25 mag arcsec⁻² isophote in the i-band, and concentration indices (c = R₉₀/R₅₀). Specifically, we use Equation (18) in Cho & Park (2009) to calculate the extinction-corrected r-band absolute magnitude

$$\begin{eqnarray*}&&{M}_{r,0}=-20.77\\ &&\quad +\,\displaystyle \frac{-1+\sqrt{1\,+\,4{\rm{\Delta }}({M}_{r,\mathrm{obs}}+20.77+4.93{\rm{\Delta }})}}{2{\rm{\Delta }}},\end{eqnarray*}$$

where

$$\begin{eqnarray*}{\rm{\Delta }} & \equiv & 1.06\times 0.223\\ & & \times [1.35{(c-2.48)}^{2}-1.14]{\mathrm{log}}_{10}(a/b)\leqslant 0.\end{eqnarray*}$$

We calculate the intrinsic extinction in the r-band as ${A}_{r,\mathrm{in}}\,={M}_{r,0}-{M}_{r,\mathrm{obs}}$.

For the remaining samples, both SDSS and GALEX magnitudes are corrected for intrinsic extinction using the stellar E(B − V) values from the OSSY catalog (Oh et al. 2011) and SDSS extinction values from Schlafly & Finkbeiner (2011). Oh et al. (2011) calculate their reddening values using Sarzi et al.'s (2006) gandalf code to match the stellar continuum and nebular emission of each galaxy to various templates. Their models include two reddening components: one which represents dust diffusion in the entire spectrum, and another which only takes into account nebular emission. We used the former for our calculations.

All photometric magnitudes are k-corrected using the "k-corrections calculator" Python script¹⁹ (Chilingarian et al. 2010; Chilingarian & Zolotukhin 2012). To check our k-corrections, we plotted k-corrected NUV–r and FUV–NUV colors against redshift and found no apparent dependence on redshift for either color, suggesting the corrections are indeed sufficiently robust (Figure 2).

Zoom In Zoom Out Reset image size

Table 1.  NUV–r and FUV–NUV Colors of All of Our Samples

Sample	NUV–r	FUV–NUV
SPOGs	2.18 ± 1.15	0.75 ± 0.55
Quiescent	4.54 ± 1.00	0.74 ± 0.58
Star-forming	2.01 ± 0.68	0.64 ± 0.34
E+A	3.51 ± 0.65	1.17 ± 0.47
AGN	2.48 ± 1.24	0.67 ± 0.40
Interacting	2.03 ± 1.66	0.86 ± 0.47

Note. For each sample, the mean color is given as well as the standard deviation.

Download table as:  ASCIITypeset image

2.3. Sample Coverage

To determine how representative the set of 833 SPOGs with GALEX UV observations are compared to the full sample of 1067, we performed multiple checks by plotting UV SPOGs against the full sample in optical color–magnitude diagrams (Figure 3), optical magnitude-redshift space (Figure 4), and emission-line ratio diagnostic diagrams (Figure 5). We find that the UV SPOGs populate a similar parameter space to the full SPOGs sample. In optical color–magnitude space (Figure 3), we see that SPOGs with both FUV and NUV photometry are slightly bluer than the full SPOGs sample, which could bias our discussion to galaxies at lower redshift with more signatures of star formation, potentially biasing our conclusions. The sample becomes more representative when SPOGs with only NUV photometry are included. Based on Anderson–Darling (AD) tests (Anderson & Darling 1952), the distributions of u–r colors for NUV-only SPOGs and FUV+NUV SPOGs are distinct (p < 10⁻⁴). Similarly, the distribution of NUV–r colors is also distinct between the two samples (p < 10⁻⁵). In order to check for possible biases in the UV properties of the optically bluer sample of FUV+NUV SPOGs, we tested subsamples of FUV+NUV SPOGs that had statistically similar distributions of u–r colors as the full SPOGs sample (using Anderson–Darling tests and p > 0.05). Using three of these distinct subsamples of 100 FUV+NUV SPOGs, we found no significant biases in UV colors or emission-line ratios for any of these subsamples when compared to the full SPOGs sample. We performed a similar test for NUV–r colors, where we matched the NUV–r distribution of FUV+NUV SPOGS to the full sample of SPOGs with NUV detection and found no significant biases in UV colors or emission-line ratios.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In magnitude-redshift space (Figure 4), FUV+NUV SPOGs have a similar coverage to the full sample of SPOGs, especially when including SPOGs with only NUV photometry as well. In emission-line ratio space (Figure 5), we notice the same trend, that SPOGs with UV photometry occupy a similar space to the full SPOGs sample. These checks indicate that the UV properties of the GALEX -detected subsample is representative of the full sample of SPOGs.

3.1. Color–magnitude Space

In Figure 6, we show the distribution of SPOGs in NUV–r versus M_i and FUV–NUV versus M_i space compared to galaxies that are star-forming, quiescent, E + A, AGN hosts, and interacting (the mean NUV-r and FUV-NUV colors of all of our samples are shown in Table 1). In NUV–r, SPOGs occupy a region similar to the star-forming sample (and distinct from quiescents). The E + A, AGN, and interacting samples occupy distinct regions of NUV–r space, and all share some overlap with SPOGs, suggesting that SPOGs are a heterogeneous group comprised of galaxies with different mechanisms for UV emission. In FUV–NUV (bottom panel of Figure 6), we find a similar trend, with SPOGs displaying a larger range in FUV–NUV colors compared to the star-forming sample. Once again the E + A, AGN, and interacting samples all share a similar color distribution as SPOGs, but seem to occupy distinct subregions. These plots suggest that UV emission in SPOGs may be an amalgam of the mechanisms that the comparison samples represent in different proportions. Future multiwavelength and high-resolution morphological analyses will clarify which individual SPOGs exhibit similar UV emission mechanisms to each comparison group.

Zoom In Zoom Out Reset image size

Figure 7 shows the cumulative distribution of FUV–NUV colors for SPOGs compared to each of the comparison samples. We used AD tests to compare the shapes of these 1D UV color distributions. For samples in the full range of masses, all samples, except the AGN sample, have statistically distinct color distributions from SPOGs with greater than 99% significance. The AGN sample is distinct from SPOGs at the 78% significance level. When restricting the comparison to a specific range of masses (−23 < M_i < −21), all samples, except the interacting sample, have statistically distinct color distributions from SPOGs with greater than 99% significance. The interacting sample is distinct from SPOGs at the 85% significance level. This test shows that the differences in UV colors are not attributable to differences in stellar mass distribution.

Zoom In Zoom Out Reset image size

3.2. Mixture Models

To further test which of the comparison samples contribute to the FUV–NUV colors of SPOGs and in what fractions, we used the method of Gaussian mixture models (GMMs)²⁰ . GMM models some underlying distribution whose true probability density function is unknown as the sum multiple Gaussian distributions (Ivezic et al. 2014). Here, we first modeled each of our comparison samples as a mixture of Gaussian components, then we modeled our SPOGs distribution as a mixture of our comparison samples. Specifically, we only compared SPOGs to SFs, Qs, and AGN because to zeroth-order, these constitute a base set of galaxy types. Interacting galaxies constitute a heterogeneous set composed of different kinds of galaxies, so we did not include them in our mixture models in order to keep our results about the composition of SPOGs simple and clear.

We approximated the distribution of FUV–NUV colors of the SFs, Qs, and AGN comparison samples as mixtures of Gaussian distributions. To fit GMMs to each of our samples, we used an expectation-maximization algorithm (Dempster et al. 1977), which iteratively adjusts the parameters of each Gaussian component until a maximum likelihood is reached for a given sample. This was done for GMMs with different number of components, and the Bayes Information Criterion (BIC) was used to select the optimal number of components for each comparison sample. The SF and AGN samples were modeled as two-component Gaussian mixtures, while the sample of Qs required a mixture of four Gaussians. We also modeled the SPOGs distribution as a three-component Gaussian mixture, and E + A as a single component Gaussian, which we used to compare to our mixtures of components. In Figure 8, we show the smoothed shapes of the FUV–NUV distributions of SFs, Qs, AGN, and SPOGs. We note that no single comparison distribution has a shape identical to that of the SPOGs distribution. As a result, we attempt to mix different fractions of SFs, Qs, and AGN to match the SPOGs distribution.

Zoom In Zoom Out Reset image size

We created mixture models with every two-component combination of these three base distributions (SFs+Qs, SFs+AGN, and Qs+AGN). For each mixture model, we varied only the weights of the two components, constrained by the fact that the sum of the weights had to be unity. A grid search method was used to search the parameter space (weights had to be between 0 and 1) for the likelihood of different mixture weights. At each weight, a likelihood was calculated as the sum of the probability density function of the GMM over all data points we were trying to fit. The grid size was of 1000 values.

We perform this grid search for both SPOGs and E + A galaxies and show the results in Figure 9. These plots show the potential contribution of each component to SPOGs and E+As. We observe that the SPOGs distribution is best reproduced with a large fraction of star-forming galaxies (>60%), and a small fraction of AGN galaxies (<30%). E + A galaxies require a large fraction of quiescent galaxies (>90%) and a smaller fraction of star-forming (<20%) galaxies. The differences in the contributions of the quiescent and star-forming populations in the models for SPOGs and E + A galaxies supports the idea that E + A galaxies are at a later stage of galaxy evolution (i.e., more quiesced star formation) than SPOGs.

Zoom In Zoom Out Reset image size

4.1. SPOG FUV–NUV Colors

4.2. Heterogeneity in the UV Colors of SPOGs

In this paper we investigated the UV properties of SPOGs compared with quiescent, star-forming, E + A galaxies, AGN-dominated galaxies, and interacting galaxies. We found that:

• 1.  
  SPOGs show a larger scatter in their FUV–NUV and NUV–r colors compared to most of the other samples, although quiescent, star-forming, E + A galaxies, AGN, and interacting galaxies all occupy overlapping UV-color parameter space similar to SPOGs. This result suggests that SPOGs are a heterogeneous group, possibly including several of these classes of objects.
• 2.  
  SPOGs exhibit FUV–NUV colors that are consistent with star-forming galaxies, and are much bluer than E + A galaxies, suggesting that they are at an earlier stage of their transition from blue to red.
• 3.  
  We used GMMs to attempt to measure the relative contribution of quiescent, star-forming, and AGN hosts to the UV colors of SPOGs, and find that they require a >60% contribution of star-forming, and a small fraction of AGN population (<30%).
• 4.  
  We ran GMMs on E + A galaxies and found that their UV colors required a much smaller contribution of star-forming, supporting both the optical and infrared picture that SPOGs are earlier in their transition than E + A galaxies.

Analyzing the full SEDs of SPOGs will allow us to obtain a better understanding of stellar populations, masses, star formation histories, and radiation sources. Obtaining UV spectral data of these objects will also be necessary to fully understand the UV emission. Ultimately, understanding the UV emission in SPOGs will allow us to better place them within the context of galaxy evolution.

F.A. thanks Mark Seibert and Michael A. Strauss for their guidance and comments on this work; Andy Skemer for introducing some of the statistical techniques that permitted the analysis; and the anonymous referee for the informative and insightful report and suggestions.

The following software and programming languages made this research possible: PYTHON (v2.7); ASTROPY (v2.0; Astropy Collaboration et al. 2013; The Astropy Collaboration et al. 2018), a community-developed core PYTHON package for Astronomy; Scikit-learn (v0.19; Pedregosa et al. 2011); SEABORN (v0.8.1;); and Tool for OPerations on Catalogs And Tables (TOPCAT, v4.3; Taylor 2005).

Support for this project was made possible by the collaboration between Princeton University and the Observatories of the Carnegie Institution. Support for KA is provided by NASA through Hubble Fellowship grant #HST-HF2-51352.001 awarded by the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., for NASA, under contract NAS5-26555. Support for A.M.M. is provided by NASA through Hubble Fellowship grant #HST-HF2-51377 awarded by the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., for NASA, under contract NAS5-26555. T.B. would like to acknowledge support from the CONACyT Research Fellowships program. K.N. acknowledges support from the grant associated with Spitzer proposal 11086. J.F.-B. acknowledges support from grant AYA2016-77237-C3-1-P from the Spanish Ministry of Economy and Competitiveness (MINECO). This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration.

GALEX (Galaxy Evolution Explorer) is a NASA Small Explorer, launched in 2003 April. We gratefully acknowledge NASA's support for construction, operation, and science analysis for the GALEX mission, developed in cooperation with the Centre National d'Etudes Spatiales of France and the Korean Ministry of Science and Technology. Funding for SDSS-III has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, and the U.S. Department of Energy Office of Science. The SDSS-III web site is http://www.sdss3.org/.

Facility: GALEX - Galaxy Evolution Explorer satellite, Sloan.