HIGH DUST DEPLETION IN TWO INTERVENING QUASAR ABSORPTION LINE SYSTEMS WITH THE 2175 Å EXTINCTION BUMP AT z ∼ 1.4^*

Interstellar dust grains play an important role in the evolution of galaxies, star formation, and planet formation. The size distribution and composition of dust grains are mainly inferred from the observed extinction curve and the re-radiated infrared emission. So far, the best-studied interstellar extinction curves are the ones for the Milky Way (MW; Savage & Mathis 1979), the Large and Small Magellanic Clouds (LMC and SMC; Fitzpatrick 1989). The most conspicuous difference between them is the sequential increment in the strength of the 2175 Å extinction bump (Stecher 1965) from SMC to LMC to MW. The feature is rarely seen in other local galaxies (Keel & White 2001).

Wang et al. (2004) composed a comprehensive review on the detection of the 2175 Å absorption bump in quasar absorption line (QAL) systems (see also references therein) and reported three bumps at high redshift z ∼ 1.4. The 2175 Å absorber in several cases (Cohen et al. 1999; Wucknitz et al. 2003) were confirmed to be damped Lyα absorbers (DLAs). Recently, several new detections have been reported. Srianand et al. (2008) found a 2175 Å extinction feature in two Mg ii systems at z ∼ 1.3 and detected associated 21 cm absorption, which usually traces cold gas content. Noterdaeme et al. (2009) presented a detection of carbon monoxide molecules (CO) at z = 1.6408 toward a red quasar and a pronounced 2175 Å bump at the same redshift. A super-strong 2175 Å absorption galaxy at z = 0.8839 toward the quasar SDSS J100713.68+285348.4 was reported by Zhou et al. (2010). Besides the quasar absorption approach to search for the 2175 Å dust extinction feature in high-redshift galaxies, the analysis of gamma-ray burst (GRB) afterglow spectra has also revealed several positive detections from intervening absorbers and from gas in the GRB host galaxy (e.g., Ellison et al. 2006; Elíasdóttir et al. 2009; Prochaska et al. 2009).

The physical conditions of absorbers, such as ionization state, electron density, temperature, gas kinetics, metallicity, and dust depletion, can be measured from the relative strength and the profile of metal absorption lines (e.g., Pettini et al. 1994; Prochaska & Wolfe 1998; Jenkins & Tripp 2001; Prochaska et al. 2002). The measurements of metal absorption lines associated with very rare 2175 Å quasar absorbers can definitely help us to understand the physical characteristics giving rise to the dust extinction features at high redshift (i.e., z > 1.0). In this paper, we analyze the dust depletion levels of the two 2175 Å absorbers in lines of sight toward the quasars SDSS⁶ J012147.73+002718.7 (hereafter J0121+0027; z = 2.2241) and SDSS J145907.19+002401.2 (hereafter J1459+0024; z = 3.0124).

Dust depletion in MW has been widely studied over ∼240 sight lines in the past several decades with the ultraviolet (UV) space telescopes (e.g., Jenkins 2009 and references therein). Significant deviation of depletion levels in different Galactic environments is found. Generally speaking, heavy elements are more depleted onto dust grains in cold gas clouds than in warm and halo gas clouds (e.g., Sembach & Savage 1996; Welty et al. 1997, 1999). Comparing the dust depletion patterns of the 2175 Å absorbers with the Galactic clouds can provide important clues on the general astrophysical characteristics of them. The dust depletion for the two absorbers in this work is very similar to that measured in cold clouds of MW, which are characterized by cold, dense, and primarily molecular gas. Dust depletion is also usually used to identify the dust content in DLAs (e.g., Pettini et al. 1994, 1997, 1999; Lu et al. 1996).

This paper is organized as follows. In Section 2 we describe the observations, Section 3 presents the analysis of UV extinction bumps, in Section 4 we present the measurements of elemental column densities in the two absorption line systems, in Section 5 we explore the dust depletion patterns, discussions are presented in Section 6, and future work and conclusions are given in Section 7.

The Sloan Digital Sky Survey (SDSS) spectra of these two quasars are remarkable with depressed flux around ∼5500 Å. They were initially uncovered by Patrick B. Hall when he carried out a visual inspection of the SDSS spectra of the DR1 quasars with z ⩾ 1.6 (Schneider et al. 2003) to identify unusual quasar subtypes (i.e., primarily broad absorption line quasars, nitrogen-strong quasars, and dust reddened quasars). The features were then interpreted as 2175 Å extinction bumps associated with two strong intervening Mg ii absorbers at z∼ 1.4 in lines of sight toward the quasars by Wang et al. (2004).⁷ The point-spread function magnitudes measured from the SDSS images are u = 20.68 ± 0.06, g = 19.09 ± 0.01, r = 18.46 ± 0.01, i = 17.93 ± 0.01, and z = 17.65 ± 0.02 for J1459+0024; u = 20.44 ± 0.06, g = 19.97 ± 0.02, r = 19.38 ± 0.01, i = 19.03 ± 0.02, and z = 18.70 ± 0.04 for J0121+0027, respectively. The SDSS spectra cover ∼3800–9200 Å with a spectral resolution R ∼ 2000 (Stoughton et al. 2002).

We performed follow-up spectroscopic observations of the quasars at higher spectral resolution to explore the dust depletion in these two 2175 Å absorbers. The medium-resolution (R ∼ 10, 000), medium signal-to-noise ratio (S/N ⩾ 10) quasar spectra were obtained with the 10 m Keck telescope using the Echellette Spectrographs and Imager (ESI; Sheinis et al. 2002). The entire spectra cover the full optical range from 3900 Å to 1.1 μm, recorded by a single 2K × 4K Lincoln Labs CCD with 15 μm pixels. The quasars were observed on 2003 December 20 (UT) with total exposure times of 900 s and 2700 s for J1459+0024 and J0121+0027 under good conditions but variable seeing of approximately full width at half-maximum (FWHM) = 08. We implemented the 05 slit for J1459+0024 and the 075 slit for J0121+0027 which yields an FWHM resolution of 33 km s⁻¹ and 44 km s⁻¹, respectively. An additional 1800 s exposure for J1459+0024 was obtained on 2004 February 18 (UT). The data were reduced and calibrated using the ESIRedux software package (v1.0) developed by JXP (see Prochaska et al. 2003 for details). Then the two reduced spectra of J1459+0024 were combined. The data were normalized by fitting a series of polynomials to absorption-free regions of the quasar spectrum.

Jiang et al. (2010) used a parameterized extinction curve (FM parameterization; Fitzpatrick & Massa 1990) constituted by a linear component and a Drude component to describe the optical/UV extinction curve in the rest frame of the quasar absorber. The linear component is used to model the underlying extinction,⁸ while the Drude component is used to model the possible 2175 Å extinction bump. The parameterized extinction curve is written as

$$\begin{equation} A(\lambda)=c_1+c_2x+c_3D(x,x_0,\gamma),\vspace*{-3pt} \end{equation} \tag{ 1 }$$

where x = λ⁻¹ and D(x, x₀, γ) is a Drude profile, which is expressed as

$$\begin{equation} D(x,x_0,\gamma)={\frac{x^2}{\big(x^2-x_0^2\big)^2+x^2\gamma ^2},}\vspace*{-3pt} \end{equation} \tag{ 2 }$$

where x₀ and γ are the peak position and FWHM of the Drude profile, respectively. Our aim is to unveil the 2175 Å absorption feature associated with absorption line systems on quasar spectra. We do not try to derive the absolute extinction curve. Our derived curve is a relative extinction curve without being normalized by E(B − V). We cannot measure the conventional extinction parameters A_V, E(B − V), and R_V from it. But all the features of the 2175 Å absorption bump are preserved. The strength of the bump is measured by the area of the bump A_bump = πc₃/(2γ), which can be interpreted as rescaling the integrated apparent optical depth of bump absorption ($A_{\lambda }={\frac{2.5}{{\rm ln} 10}}\tau _{\lambda }$). We fit the SDSS spectra of J1459+0024 and J0121+0027 by reddening the SDSS composite quasar spectrum (Vanden Berk et al. 2001) with parameterized extinction curves at the redshifts z_abs = 1.3947 for the former and z_abs = 1.3888 for the latter, respectively. The redshifts are those of the strongest component in the absorption line profiles, which are measured on our Keck/ESI spectra, for the two 2175 Å absorbers. Since the 2175 Å extinction bump is a very broad feature whose FWHM is usually greater than 300 Å (Fitzpatrick & Massa 2007), the small differences between these redshifts and the redshifts of the centroid of absorption line profiles could not affect the results of spectrum fitting in this work. To focus on fitting the continuum of quasar spectrum, the regions with strong emission lines and known strong absorption lines are masked. Several intervening Lyα absorption lines are present blueward of the Lyα emission lines on the spectrum of J1459+0024 with z > 2.6. We remove them iteratively by clipping the outliers beyond 4σ error. The results of spectrum fitting are listed in Table 1. The best models are overplotted with SDSS data in Figures 1(a) and 2(a). To emphasize the requirement of a absorption bump on the extinction curve, we also overplot the reddened composite quasar spectrum by using the linear component of best model (green lines in Figures 1(a) and 2(a)) only. The required bump of J1459+0024 is so broad that its wings exceed the wavelength coverage of the SDSS spectrum. Thus, the entire observed spectrum of J1459+0024 is completely off from the green model in Figure 1(a).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The intrinsic variation of quasar spectra can mimic an extinction bump in some cases (Pitman et al. 2000). We gauge the significance of the extinction bumps in this work with the simulation technique developed by Jiang et al. (2010). The simulation technique begins with the selection of a control sample of SDSS quasar spectra at a similar redshift to the quasar of interest. Then we fit each of them by reddening the composite quasar spectrum with a parameterized extinction curve at the redshift of the absorber of interest. The parameters x₀ and γ in the parameterized extinction curve are fixed to the best values fitting the 2175 Å extinction bump of interest. The distribution of bump strengths is expected to be Gaussian by assuming random fluctuation of continuum on each spectrum in the control sample. If the bump strength of the absorber is far away from this distribution, then the bump has statistical significance. For each quasar, we select the spectra in the SDSS DR7 database classified as quasars (specClass = QSO or HIZ_QSO) with redshift in the range of z_emi − 0.05 < z < z_emi + 0.05 and S/N SN_I ⩾ 6,⁹ as its control sample. The resultant control sample of J1459+0024 is constituted by 1002 SDSS quasars around z = 3.0124. The resultant control sample of J0121+0027 is constituted by 2351 SDSS quasars around z = 2.2241. The extracted distributions of bump strengths are presented in Figures 1(c) and 2(c). Both of the bumps in J1459+0024 and J0121+0027 are statistically significant at a confidence level of >5σ. The width of the bump strength distribution for J1459+0024 is much greater than that for J0121+0027. That is because the bump on the spectrum of J1459+0024 (γ = 2.68 μm⁻¹) is much broader than that of J0121+0027 (γ = 0.80 μm⁻¹). For the spectrum without a real extinction feature, the fitted bump strength is an integration of random fluctuations over the wavelength range covered by the bump of interest. In principle, a longer integration makes the resultant distribution of bump strengths broader. Therefore, a broad bump has to be stronger (with larger bump area) than a narrow one in order to be significantly distinguished from random fluctuations of quasar spectra. In theory, the mean value of the bump strength distribution should be zero. But, the mean value of the bump strength distribution is about 1.0 for J1459+0024 (see Figure 1(c)). The significant non-zero mean bump strength is mainly caused by the low quality of the SDSS quasar composite spectrum in far-UV band of its rest frame. There are only about 150 low (or moderate) observed quasar spectra being used to create the composite spectrum in far-UV band (Vanden Berk et al. 2001).¹⁰ Thus, it is very likely that the composite quasar spectrum cannot represent the average properties of quasar spectra in far-UV band completely.

We also perform continuum fitting by reddening the SDSS composite spectrum with different types of average extinction curves, namely that from the SMC, the LMC Supershell (LMC2), the LMC (Gordon et al. 2003), and the MW (Fitzpatrick & Massa 2007). The 2175 Å extinction bump is absent in the SMC extinction curve and it is strongest in the LMC and the MW extinction curves. The strength of the bump in the LMC2 extinction curve is medium. These extinction curves are given in the FM parameterization. The two free parameters in the continuum fitting are E(B − V) and a normalization scale. The results of spectrum fitting are presented in Table 2. We show the models using different extinction curves for the two reddened quasar spectra in Figure 3. The spectrum of J0121+0027 can be fitted by the average LMC (χ²_ν = 1.14) and the average MW (χ²_ν = 1.10) extinction curve models. But the spectrum of J1459+0024 cannot be fitted by any model (χ²_ν = 3.88 in the best case). That is because the extinction bump on J1459+0027 is much broader (γ = 2.68 ± 0.07 μm⁻¹) than that of the average LMC (γ = 0.934 ± 0.016 μm⁻¹) extinction curve or the average MW (γ = 0.922 μm⁻¹) extinction curve.

Zoom In Zoom Out Reset image size

We compare the two extracted 2175 Å absorption bumps at z∼ 1.4 with the bumps observed in the MW (see Figure 4; Fitzpatrick & Massa 2007). The peak position and strength of bumps are within the range of Galactic values. However, the FWHM of bump in J1459+0024 is larger than the broadest Galactic bump, which is measured in the line of sight toward the star HD 29647. The direct comparison of Optical/UV extinction curves of J1459+0024 and HD 29647 is plotted in Figure 4(c) and the comparison of J0121+0027 and HD 164816 (the latter has a similar bump with the former) is in Figure 4(d). Since the extracted extinction curves for quasar absorbers are relative extinction curves, we cannot measure the absolute visual extinction A_V on them. We add an arbitrary normalization on the extracted curves while plotting them with Galactic curves.

Zoom In Zoom Out Reset image size

The total number of absorption components is not fully resolved because of our limited resolution and the intrinsic line-blending. For the line of sight toward J0121+0027, three Mg ii absorption systems are seen at z = 1.3877, 1.3888, and 1.3907, which match the approximate redshift of the 2175 Å absorption bump (z∼ 1.39; Wang et al. 2004) detected in the SDSS spectrum. Since the first two strong systems are only separated by ∼140 km s⁻¹ (corresponding to z = 1.3888), they are considered as two associated components in one absorption system at z = 1.3888 in this work. The third one could be another component, but it is neglected in our analysis because it is too weak to affect the measurements. For the line of sight toward J1459+0024, one strong Mg ii absorption system is seen at z = 1.3947. We identified numerous absorption lines associated with the absorption system, consistent with the 2175 Å absorption bump feature redshift (z∼ 1.39) measured in the SDSS spectrum. The metal absorption lines of interest are plotted in velocity space in Figures 5 and 6.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We carry out the Apparent Optical Depth Method (AODM; Savage & Sembach 1991) to measure the column densities of low-ionization absorption lines in those two 2175 Å absorption systems. In general, the column densities are measured by integrating the optical depth over the velocity range of the spectra covering all of the absorption in the detected transitions. The uncertainties for detections are reported as 1σ errors. The upper limits are measured by integrating the optical depth over the same velocity range as an unsaturated line (i.e., Fe ii λ2374) and are reported as 3σ statistical limits. For the saturated lines, the column densities derived from the AODM are reported as lower limits. We also measured the column densities of transitions in red and blue components of the absorption system toward J0121+0027 separately. The boundary of these two components is chosen at the point of −70 km s⁻¹ compared to redshift z = 1.3888 (the thin blue dot line in Figure 6). All ionic column densities are presented in Table 3.

The column density measurements of ions from absorption lines only represent the elements in the gas phase. If one assumes that ionization corrections are small for what is expected to be a primarily neutral gas, the observed relative gas-phase abundances reflect the underlying nucleosynthetic abundance pattern modified by the differential depletion of the elements. Frequently, the relative abundance, [Fe/Zn] = log(N(Fe)/N(Zn))−log(N(Fe)/N(Zn))_☉,¹¹ is used to infer the dust depletion level, even when N(H i) is not obtainable, by assuming that Zn, which is assumed to be nearly undepleted due to its low condensation temperature (Meyer & Roth 1990; Pettini et al. 1994), tracks the Fe Peak closely in nucleosynthesis.

The detected transitions of zinc are Zn ii λ2026 and Zn ii λ2062 on our Keck spectra. These two absorption lines are weakly blended with Mg i λ2026 and Cr ii λ2062 lines, respectively. Since the oscillator strength of transition Cr ii λ2062 is fairly weak and other transitions of Cr ii are only marginally detected in both of the absorption systems, we think the absorption line of Zn ii λ2062 is not contaminated by blending and take the column density of zinc measured on it to derive dust depletion patterns. For other species, the weighted mean values are adopted if there are two or more transitions detected. The depletion patterns are listed in Table 4.

We compare the dust depletion patterns in the two 2175 Å absorption systems with the depletion patterns observed in SMC ISM, LMC ISM, and cold Galactic disk clouds (Welty et al. 1997, 1999) in Figure 7. The dust depletion levels increase sequentially from SMC, LMC, to MW. It is clear that the depletion patterns of both absorption systems closely resemble that in cold Galactic clouds, especially from the best measured [Fe/Zn] values. The strength of the 2175 Å extinction bumps increases in the same manner as dust depletion from SMC, LMC, to MW. It is very likely that heavy dust depletion is required to give rise to 2175 Å absorption. See more discussions in Section 6.

Zoom In Zoom Out Reset image size

The depletion levels, [Fe/Zn] ≈−1.5 and [Si/Zn] <−0.67, in both of absorbers are among the highest dust depletion ever measured for QAL systems. The mean value of [Fe/Zn] > − 0.6 was measured by York et al. (2006) with a large sample constituted by 809 Mg ii absorption systems at 1 ⩽ z < 2. A comparable depletion measurement with ours has been reported by Petitjean et al. (2002) in a DLA bearing molecular hydrogen with [Fe/Zn] = −1.59 at z = 1.973 toward the quasar Q0013-004. In addition, Noterdaeme et al. (2009) measured a similar high dust depletion level ([Fe/Zn] = −1.47 and [Si/Zn] = −1.07) of a 2175 Å absorber at z = 1.64 toward the quasar SDSS J160457.50+220300.5 with VLT/UVES. We compare the dust depletion level in this work with a combined DLA/sub-DLA sample, for which high-resolution spectroscopic data are available (Keck/ESI and Keck/HIRES data from Prochaska et al. 2007; VLT/UVES data from Noterdaeme et al. 2008), in Figure 8. We ignore censored data in the sample when plotting the figure. To present the dust depletion of two 2175 Å absorbers in the same figure, we assume the column density of neutral hydrogen N(H i) = 10²¹ cm⁻¹ in both of them. This assumption is arbitrary and is for illustration purposes only. DLAs bearing H₂ are marked with black filled circles. It is clear that DLAs showing high dust depletion level, [Fe/X]¹²<−0.4, have a high H₂ detection rate (e.g., Ge & Bechtold 1997; Ge et al. 2001; Ledoux et al. 2003; Cui et al. 2005; Noterdaeme et al. 2008). It is reasonable to assume, therefore, that the gas in these two 2175 Å absorbers bears molecules.

Zoom In Zoom Out Reset image size

The dust depletion patterns ([X/Zn]) in the SMC ISM are very similar to that measured in Galactic halo clouds, while the depletion in the LMC ISM is very similar to that in Galactic warm disk clouds (see Table 4). Since the overall metallicity and dust-to-gas ratio are very different in SMC, LMC, and MW (e.g., Pei 1992; Welty et al. 1997), the similarities probably indicate that dust depletion can hardly be dependent on these two physical parameters. In the MW, the most severe depletions are found in lines of sight with the largest mean densities and largest fraction of H₂; the least severe depletions are observed for halo clouds and for high-velocity clouds in the Galactic disk (e.g., Jenkins et al. 1984; Savage & Sembach 1996; Fitzpatrick 1996; Trapero et al. 1996; Jenkins & Wallerstein 1996; Welty et al. 1997). Such a dependence of dust depletion on environment probably has to do with the modification or destruction of dust grains by supernova shocks. In cold dense molecular clouds, large dust grains can form substantially by coagulation of finer grains, while in warmer environments, large grains are usually destroyed by energetic processes and some depleted elements return to the gas phase.

The fact that the strength of 2175 Å extinction bumps becomes weaker with decreasing dust depletion levels from MW, LMC, to SMC may indicate, that the destruction of 2175 Å absorption carriers happens simultaneously with the destruction of large grains. The large polycyclic aromatic hydrocarbon (PAH) molecules, which have strong π → π* absorption in the 2000–2500 Å region, are proposed to be the carrier of 2175 Å absorption (Li & Draine 2001). These large molecules can be easily destroyed by photon-thermo dissociation, coulomb explosion, and/or X-ray destruction (e.g., Voit 1992 and references therein).

We search for the literatures and find 15 lines of sight in MW having both measurements of UV absorption bump and metal abundance available (see Table 5). Dust depletion is measured with the abundance ratio [Fe/H] by assuming that the underlying abundances are solar values. The measurements of Galactic 2175 Å extinction bump are taken from Fitzpatrick & Massa (2007). Note that the area of the bump defined in FM07 is different from that defined in this work. Since the extinction curve has been normalized by E(B − V) in FM07, A_bump = E(B − V)×A*_bump, where A*_bump is the area defined in FM07. We plot the relative bump strength ($\frac{A_{\rm bump}}{A_V}$) versus [Fe/H] in Figure 9(a). We calculate Spearman's ρ, a non-parametric measure of statistical dependence between two variables, for the relative strength of the bump and [Fe/H]. The resultant Spearman's ρ = −0.46 suggests a tentative anti-correlation between them with statistical confidence level of 90%. The fairly large scatter of bump strength could be caused by the varying size of PAH molecules from one sightline to another. Draine (2003) interpreted that the observed variations in FWHM (and small variations in peak position) of the Galactic 2175 Å extinction bump profile resulted from differences in the PAH mix. This differences could also affect the oscillator strength per molecule and lead to the deviation of bump strengths. Note that the two 2175 Å extinction bumps in this work have very different widths and strengths.

Zoom In Zoom Out Reset image size

In Figure 9(c), we confirm the anti-correlation between f (H₂) and [Fe/H] (Spearman's ρ = −0.81) and interpret it as the simultaneous dissociation of molecular hydrogen with destruction of large dust grain. In addition, we find a tentative correlation between dust-to-gas ratio and [Fe/H] with Spearman's ρ = +0.54 (Figure 9(d)). It is consistent with the grain destruction scenario. Assuming constant density and spherical dust grains, the total cross section of finer grains would be greater than the total cross section of large grains even if some materials of dust grains are evaporated during the destruction processes. We summarize the discussions as a molecule and large grain destruction scenario: (1) molecules (e.g., PAH and H₂) and large dust grains form substantially in cool and dense environment; (2) molecules and large grains would be destroyed simultaneously when the environment is heated (e.g., heated by supernova shock waves; irradiated by extreme UV photons); and (3) the destruction of large dust grains releases some depleted elements into the gas phase; the dissociation of H₂ molecules decreases the f (H₂), while the dissociation of PAH molecules diminishes the strength of 2175 Å absorption bump.

In this work, we measure the dust depletion levels for the two 2175 Å absorbers. More observations in other wavelength bands will definitely help us to unveil the astrophysical conditions giving rise to 2175 Å absorption. For the two absorbers at z ∼ 1.4 in this work, the HST/COS¹³ spectrometer covers the important H i and H₂ transitions and can be used to determine the gas metallicity and f (H₂). We are planning to propose Keck/HIRES observations to catch the C i and other transitions blueward of the ESI coverage to the atmospheric cutoff. Atomic carbon is an excellent tracer of cold, dense gas because of its low ionization potential (below 1 Ryd). The ratios of C i and C ii transitions will provide us a powerful tool to probe the physical environment of the absorbers.

By using the parameterized extinction curve fitting technique, we are searching for more 2175 Å absorbers in SDSS quasar spectra database. Our preliminary result shows 18 detections of 2175 Å extinction bump associated with dusty strong Mg ii quasar absorbers in SDSS DR3.

In this paper, we report on follow-up moderate resolution spectroscopy of the 2175 Å absorption systems at z∼ 1.4 in lines of sight toward two quasars J0121+0027 and J1459+0024. The column densities of heavy elements are measured by low-ionization absorption lines with AODM. We derived the dust depletion patterns of both absorption systems and found that they closely resemble that of cold diffuse disk clouds in MW. The values, [Fe/Zn] ≈−1.5 and [Si/Zn]<−0.67, are among the highest dust depletion measured for QAL systems. We conclude that heavy dust depletion (i.e., a characteristic of cold dense clouds in MW) is required to produce a pronounced 2175 Å extinction bump.

This work was partially supported by NSF with grant NSF AST-0451407, AST-0451408, and AST-0705139 and a China NSF grant (NSF-10973012). P.J acknowledges support from China Scholarship Council. This research has also been partially supported by the CAS/SAFEA International Partnership Program for Creative Research Teams. J.X.P. is partially supported by an NSF CAREER grant (AST-0548180) and by NSF grant AST-0908910.

The authors recognize and acknowledge the very significant cultural role and reverence that the summit of Mauna Kea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain.

Funding for the SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, the Max Planck Society, and the Higher Education Funding Council for England. The SDSS Web site is http://www.sdss.org/.

The SDSS is managed by the Astrophysical Research Consortium for the Participating Institutions. The Participating Institutions are the American Museum of Natural History, Astrophysical Institute Potsdam, University of Basel, University of Cambridge, Case Western Reserve University, University of Chicago, Drexel University, Fermilab, the Institute for Advanced Study, the Japan Participation Group, Johns Hopkins University, the Joint Institute for Nuclear Astrophysics, the Kavli Institute for Particle Astrophysics and Cosmology, the Korean Scientist Group, the Chinese Academy of Sciences (LAMOST), Los Alamos National Laboratory, the Max-Planck-Institute for Astronomy (MPIA), the Max-Planck-Institute for Astrophysics (MPA), New Mexico State University, Ohio State University, University of Pittsburgh, University of Portsmouth, Princeton University, the United States Naval Observatory, and the University of Washington.