Placing High-redshift Quasars in Perspective: A Catalog of Spectroscopic Properties from the Gemini Near Infrared Spectrograph–Distant Quasar Survey

A persistent problem in extragalactic astrophysics is understanding how supermassive black holes (SMBHs) and their host galaxies coevolve over cosmic time (e.g., Di Matteo et al. 2008; Merloni et al. 2010; Bromm & Yoshida 2011; Heckman & Best 2014). This problem touches upon several aspects of galaxy evolution, including the SMBH mass (M_BH), which correlates with properties of the host galaxy, such as the bulge mass and stellar velocity dispersion (e.g., Ferrarese & Merritt 2000; Gebhardt et al. 2000; Woo et al. 2010; Kormendy & Ho 2013; McConnell & Ma 2013; Reines & Volonteri 2015), the accretion rate, which probes the accretion flow and efficiency of the accretion process, (e.g., Croton et al. 2006; Hopkins & Quataert 2010; Blaes 2014), and the kinematics of material outflowing from the vicinity of the SMBH, which may affect star formation in the host galaxy (e.g., Hopkins & Elvis 2010; Maiolino et al. 2012; Carniani et al. 2018). For nearby (z ≲ 1) active galactic nuclei or quasars, most of the parameters required for exploring these topics can be most reliably estimated using optical diagnostics, namely the broad Hβ λ4861 and narrow [O iii]  λλ4959, 5007 emission lines. However, at z ≳ 1, which includes the epoch of peak quasar activity (from z = 1–3), these diagnostic emission lines are redshifted beyond λ_obs ∼ 1 μm, firmly into the near-infrared (NIR) regime. Since the vast majority of large spectroscopic quasar surveys have been limited to λ_obs ≲ 1 μm, investigations of large samples of quasars at z ≳ 1 are usually forced to use spectroscopic proxies for Hβ and [O iii]. Using indirect proxies can lead not only to inaccurate redshifts (e.g., Gaskell 1982; Hewett & Wild 2010; Denney et al. 2016a; Shen et al. 2016; Dix et al. 2020), but also to systematically biased and imprecise estimates of fundamental parameters such as M_BH and the accretion rate (e.g., Shen & Liu 2012; Trakhtenbrot & Netzer 2012; Denney et al. 2016b).

NIR spectra have been obtained for a few hundred quasars at z ≳ 1, but these spectra constitute a heterogeneous collection of relatively small samples (≈10–100 sources) that span wide ranges of source-selection criteria, instrument properties, spectral band and resolution, and signal-to-noise ratio (S/N) (e.g., McIntosh et al. 1999; Shemmer et al. 2004; Sulentic et al. 2004; Netzer et al. 2007; Trakhtenbrot et al. 2011; Shen & Liu 2012; Zuo et al. 2015; López et al. 2016; Mejía-Restrepo et al. 2016; Shen 2016; Coatman et al. 2017). Thus, the current NIR spectroscopic inventory for high-redshift quasars is biased in a multitude of selection criteria, and none of these mini-surveys are capable of providing a coherent picture of SMBH growth across cosmic time.

To mitigate the various systematic biases present in the current NIR spectroscopic inventory, we have obtained NIR spectra of 272 quasars at high redshift using the Gemini Near-Infrared Spectrograph (GNIRS; Elias et al. 2006), at the Gemini North Observatory, with a Gemini Large and Long Program.¹⁹ By utilizing spectroscopy in the ∼0.8–2.5 μm band of a uniform, flux-limited sample of optically selected quasars at 1.5≲ z ≲ 3.5, our Distant Quasar Survey (GNIRS-DQS) was designed to produce spectra that, at a minimum, encompass the essential Hβ and [O iii] region in each source while having sufficient S/N in the NIR band to obtain meaningful measurements of this region. This survey assembles the largest uniform sample of z ≳ 1 quasars with rest-frame optical spectroscopic coverage. The spectral inventory presented in this catalog will allow development of single-epoch prescriptions, as opposed to C iv reverberation mapping, for rest-frame ultraviolet (UV) analogs of key properties such as M_BH and accretion rate, along with revised redshifts based primarily on emission lines in the rest-frame optical band.

This paper describes the GNIRS observations and structure of the catalog; subsequent investigations will present the scientific analyses enabled by this catalog. Section 2 describes the target selection, and Section 3 describes the GNIRS observations and the spectroscopic data processing. Section 4 presents the catalog of basic spectral properties, along with a smaller catalog of additional features that can be measured reliably in some of the spectra. Section 5 summarizes the main properties of our catalog as well as comments on its future applications. Throughout this paper we adopt a flat ΛCDM cosmology with ${{\rm{\Omega }}}_{{\rm{\Lambda }}}=1-{{\rm{\Omega }}}_{0}=0.7$ and H₀ = 70 km s⁻¹ Mpc⁻¹ (Spergel et al. 2007).

The GNIRS-DQS targets were selected from the spectroscopic quasar catalog of the Sloan Digital Sky Survey (SDSS; York et al. 2000), primarily from SDSS Data Release 12 (Pâris et al. 2017) and supplemented by SDSS Data Release 14 (DR14; Pâris et al. 2018). Sources were selected to lie in three narrow redshift intervals, 1.55 ≲ z ≲ 1.65, 2.10 ≲ z ≲ 2.40, and 3.20 ≲ z ≲ 3.50, in order to cover the Hβ+[O iii] emission complex, and in order of decreasing NIR brightness, down to m_i ∼ 19.0, a limit at which the SDSS is close to complete in each of those redshift intervals (Richards et al. 2002). Figure 1 displays the luminosity–redshift distribution of GNIRS-DQS sources with respect to sources from the SDSS DR14 catalog. For the redshift distributions in the selected intervals, shown in Figure 2 along with their respective magnitude distributions, the Hβ+[O iii] emission complex reaches the highest S/N in the centers of the J, H, and K bands, respectively. The selected redshift intervals also ensure coverage of sufficient continuum emission and Fe ii line emission flanking the Hβ+[O iii] complex, enabling accurate fitting of these features. We visually inspected the SDSS spectrum of each candidate and removed sources having spurious redshifts, instrumental artifacts, and other anomalies. The combined SDSS-GNIRS spectroscopic coverage of each source includes, at a minimum, the C iv λ1549, Mg ii λλ2796, 2803, Hβ, and [O iii]  emission lines; the Hα λ6563 emission line is present in all sources at 1.55 ≲ z ≲ 2.50, representing ∼87% of our sample. We note that the 2.10 ≲ z ≲ 2.40 redshift bin comprises ∼67% of our entire sample, given that this redshift bin is three times wider than that of the lower redshift bin, and sources in this bin are brighter than the sources in the higher redshift bin.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In summary, the GNIRS-DQS sources constitute an optically selected, NIR flux-limited sample of quasars, spanning wide ranges in rest-frame UV spectral properties, including broad absorption line (BAL) and non-radio-quiet quasars²⁰  (comprising ∼30%²¹ and ∼12% of the sample, respectively; Pâris et al. 2018). Figure 3 shows the radio-loudness distribution of the GNIRS-DQS sources. The GNIRS-DQS sample is broadly representative of the general quasar population of luminous, high-redshift quasars during the epoch of most intense quasar activity (e.g., Hewett et al. 1993; Hasinger et al. 2005; Richards et al. 2006).

Zoom In Zoom Out Reset image size

The observations were designed to yield data of roughly comparable quality, in terms of both S/N and spectral resolution, to the respective SDSS spectra at λ_obs ∼ 5000 Å. The GNIRS spectra were thus required to have a ratio of ∼40 between the mean flux density and the standard deviation of that flux density in a rest-frame wavelength interval spanning 100 Å around λ_rest = 5100 Å, and a spectral resolution of R ∼ 1100 across the entire GNIRS band. These requirements enable accurate measurements of redshift based on [O iii]  line peaks, with the high S/N contributing to reducing the uncertainties below the spectral resolution limit, ∼300 km s⁻¹ (Shen et al. 2016). As explained below in Section 4, we determine that, on average, our spectra produce uncertainties on the measured line peak of [O iii] λ5007 of order ∼50 km s⁻¹, stemming from pixel-to-wavelength calibration and our fitting procedures.

All spectra were obtained in queue-observing mode with GNIRS configured to use the Short Blue camera (015 pixel⁻¹), the 32 lines mm⁻¹ grating in cross-dispersed mode, and the 045 wide slit. This configuration covers the observed-frame ∼0.8–2.5 μm band in each source, simultaneously, in six spectral orders with overlapping spectral coverage. Our observing strategy utilized an ABBA method of slit nodding to enable sky subtraction. Exposure times ranged from ∼10 to 40 minutes for each object, with an additional 15 minutes of overhead per source. Each observation included calibration exposures, and either one or two ABBA sequences depending on source brightness. We also observed a telluric standard star either immediately before or after the observation in a spectral range of B8 V to A4 V, with 8200 K ≲ T_eff ≲ 13,000 K, and typically within ≈10°–15° from each quasar.

The observation log of the original 272 sources appears in Table 1. Column (1) is the SDSS designation of the quasar. Column (2) provides the most reliable reported redshift estimate from SDSS (Pâris et al. 2018, Table A1, column 9 "Z"). Columns (3), (4), and (5) list the respective J, H, and K magnitudes of each quasar from the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006). Columns (6) and (7) give the observation date and semester, respectively. Column (8) is the net science exposure time, Column (9) provides comments, if any, concerning the observation, Column (10) provides a flag for whether or not the quasar is a BAL quasar (as defined in Pâris et al. 2018), and Column (11) provides a flag for whether or not the quasar is considered non-radio-quiet (see footnote 20).

We classify an acceptable observing night for this survey based on our programs' approved observing conditions including no more than 50% cloud cover and 85% image quality,²² however some objects were observed under worse conditions, and are noted as such in Table 1 (which brings the total number of lines in that table to 284). Additionally, 12 sources were observed over two observing sessions. These additional observations are recorded separately and immediately follow the initial observation in Table 1. For these objects, all available observations were utilized in the reduction process.

Our data-processing procedure generally follows the XDGNIRS pipeline developed by the Gemini Observatory (Mason et al. 2015 ²³ : see also Shen et al. 2019a) with the Gemini package in PyRAF.²⁴ Following standard image cleaning for artifacts and other observational anomalies, we pair-subtract the images to remove the bulk of the background noise by directly combining the sky-subtracted object exposures. Quartz lamps and IR lamps were used to create flat fields to correct pixel-by-pixel variation across the detector. The flat-fielded images were corrected for optical distortions. Several objects required replacement flat fields due to pixel shifting of dead pixels in the detector into the GNIRS spectra directly (marked accordingly with a corresponding comment in Table 1), which produced a notable increase in the uncertainty of spectroscopic measurements for these objects, particularly in the bluer bands. On average, the increased flux uncertainty from these spectra is on the order of ∼3%. At this stage, of the 272 sources observed, 47 observations did not yield a meaningful spectrum due to bad weather, instrument artifacts, or other technical difficulties (Note 4 in Column 9 of Table 1), leaving the final sample at 226 sources.

Wavelength calibration was performed using two argon lamp exposures in order to assign wavelength values to the observed pixels. The uncertainties associated with this wavelength calibration are not larger than 0.5 Å rms, corresponding to ≲10 km s⁻¹ at ∼15000 Å.

Spectra of the telluric standards were processed in a similar fashion, followed by a careful removal of the stars' intrinsic hydrogen absorption lines. This process was performed by fitting Lorentzian profiles to the hydrogen absorption lines, and interpolating across these features to connect the continuum on each side of the line. Following the line cancellation, the quasar spectra were divided by the corrected stellar spectra. The corrected spectra were multiplied by an artificial blackbody curve with a temperature corresponding to the telluric standard star, which yielded a cleaned, observed-frame quasar spectrum. Each quasar spectrum was flux calibrated by comparing local flux densities to the J, H, and K 2MASS magnitudes from Table 1 and using the magnitude-to-flux conversion factors from Table A.2 of Bessell et al. (1998). For the final spectra, we masked any noise present from cosmic rays, regions of high levels of atmospheric absorption, and band gap interference.

We chose this method as opposed to flux calibrating via the telluric standards to avoid any differences in atmospheric conditions between observations of the object and the telluric standard. This preference was also motivated by our use of a relatively narrow slit in order to prioritize spectral resolution at the cost of potentially larger slit losses in the observations. Although the 2MASS and Gemini observations are separated by several years in the quasars' rest frames, the cross-calibrations are subject to minimal uncertainties since ∼88% of our sources are luminous radio-quiet quasars at high redshift. Such sources typically show UV–optical flux variations on the order of ≲10% over such timescales (e.g., Vanden Berk et al. 2004; Kaspi et al. 2007; MacLeod et al. 2012). In fact, the effects stemming from the differences in airmass between the quasars and their respective telluric standard stars, as well as the slit losses, are typically larger than the expected intrinsic quasar variability.

In order to further test the reliability of our flux calibration, we compared the flux densities in overlapping continuum regions, λ_obs ∼ 8000–10000 Å, between our GNIRS spectra and those of the respective SDSS spectra; this test was feasible for ∼90% of our sources that have both high-quality GNIRS and SDSS spectral data where we can obtain meaningful comparisons that avoid reductions in quality that can occur in this region for both surveys. We found that the flux densities in the SDSS spectra are, on average, smaller than the GNIRS flux densities by ∼40% (μ = −0.155), with a 1σ scatter of ∼60% (σ = 0.2013) (see Figure 4, where μ and σ are the logarithms of the mean and standard deviation, respectively). Therefore, the flux densities when directly comparing both spectral sets are consistent at the 1σ level, despite the presence of this systematic offset. This systematic offset should be taken into account when comparing fluxes between SDSS and GNIRS spectra, however, it does not affect the emission-line measurements presented in this survey. This scatter may include discrepancies such as those due to intrinsic quasar variability, fiber light loss in SDSS spectra, and differences in airmass between quasars and their respective standard star observations. Examples of prominent emission lines in final, flux-calibrated spectra appear in Figure 5.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The final GNIRS quasar spectra were fit by using multiple localized linear continua, explained in Section 4.1, constrained by no less than six narrow (∼200 Å wide, rest-frame) line-free regions, and performing Gaussian fits to the emission lines. The Fe ii and Fe iii emission complexes were modeled via empirical templates from Boroson & Green (1992) and Vestergaard & Wilkes (2001) for the rest-frame optical and UV bands, respectively. These templates were scaled and broadened by convolving a Gaussian with an FWHM value that was free to vary between 1300 and 10,000 km s⁻¹. Given that the Fe ii, Fe iii, and Hβ lines likely originate from different physical regions (e.g., Barth et al. 2013), we kept the FWHM of the iron templates as a free parameter. The FWHM values selected to broaden each template were determined using a least-squares analysis on each fitted region.

For the [O iii]  lines, the widths of each line were restricted to be identical to each other, and their flux ratios were kept constant at I₅₀₀₇/I₄₉₅₉ = 3 (e.g., Storey & Zeippen 2000, and references therein); additionally, the rest-frame wavelength difference between the λ5007 and λ4959 lines was kept constant, which proved adequate for the fits of each object.

We fit two Gaussians to each broad emission-line profile to accommodate possible asymmetry present in the profile due to, e.g., absorption, or outflows. We note that the two Gaussians fit per broad emission line is adopted for fitting purposes only, and they do not represent physically distinct regions. Fitting the line profiles with more complex models was not warranted given the quality of our GNIRS spectra. The constraints on the Gaussian profiles for each emission line were that the peak wavelengths can differ from their known rest-frame values by up to ±1500 km s⁻¹, on initial assessment (see, e.g., Richards et al. 2011, Figure 5) with a max flux value ranging from zero to a value calculated to be twice the maximum value of the emission line. Visual inspection yielded some exceptions beyond an offset of ±1500 km s⁻¹, whereupon manual fitting was performed to compensate for the larger velocity offset.

4.1. Continuum Fitting

4.2. Mg ii

4.3. Hβ

The Hβ region, for most of our objects, provided reliable measurements given the survey was designed with this region in mind. However, in ≲2% of our objects, the Hβ emission line was adjacent to the edge of the observing band, resulting in larger uncertainties when fitting the Fe ii  emission complex. This misalignment of Hβ stems from selecting our sample using UV-based redshifts, based primarily on the peak wavelength of the C iv emission line, which suffer from systematic biases due to outflows that can be as large as ≈5000 km s⁻¹ (Dix et al. 2020; B. M. Matthews et al. 2021, in preparation). This misalignment also results in reduced coverage of the Fe ii  blends for these objects. Despite this complication, we were able to adequately fit two Gaussians to each of the Hβ emission lines.

By design, our survey targeted highly luminous quasars, biased toward having higher L/L_Edd values (see Figure 1 in Netzer 2003), which typically also tend to have relatively strong Fe ii  emission. As a result, we relied on the broad iron bumps on either side of the Hβ line, rest frame ∼4450–4750 and 5100–5400 Å (Vanden Berk et al. 2001), as our primary region for fitting the Fe ii  complex. While reasonable in most cases, these fits are likely affected by He ii λ4686 emission-line contamination, however the He ii emission line is unresolvable in this sample due to uncertainties from a variety of factors (see Section 4.5). On average, the corresponding Fe ii  EW values in those sources is ∼20 Å. Additionally, ∼5% of our objects differed from the well-known trends of "Eigenvector 1" (Boroson & Green 1992), having a blend of strong [O iii]  and Fe ii  emission, resulting in their spectra exhibiting "shelves" on the red side of the Hβ profile. These features required a more careful fitting, and we did not see any evidence of [O iii]  outflows directly contributing to this emission complex. An example of a shelf-like fit is presented in Figure 6.

Zoom In Zoom Out Reset image size

Figure 7 shows the distribution of [O iii]  EWs in the GNIRS-DQS sample. As explained in Section 4.6, for those objects that do not have detectable [O iii]  emission, we must use the Mg ii line to determine systemic redshifts (z_sys); for those objects that lack both [O iii]  and Mg ii, we must utilize the Hβ line for that purpose, which is present in every GNIRS-DQS spectrum.

Zoom In Zoom Out Reset image size

4.4. Hα

4.5. Uncertainties in Spectral Measurements

4.6. The Catalog

We present a catalog of spectroscopic properties obtained from NIR observations of a uniform, flux-limited sample of 226 SDSS quasars at 1.5 ≲ z ≲ 3.5, which is the largest uniform inventory for such sources to date. The catalog includes basic spectral properties of Mg ii, Hβ, [O iii], Fe ii, and Hα emission lines, as well as Hδ, Hγ, [O ii], and [Ne iii] emission lines for a subset of the sample. A spectral resolution of R ∼ 1100 was achieved for this data set, which is roughly comparable to the value of the corresponding SDSS spectra. These measurements provide a database to comprehensively analyze and investigate rest-frame UV–optical spectral properties for high-redshift, high-luminosity quasars in a manner consistent with studies of low-redshift quasars.

In particular, the catalog will enable future work on robust calibrations of UV-based proxies to systemic redshifts and black hole masses in distant quasars. Such prescriptions are becoming increasingly more important as millions of quasar optical spectra will be obtained in the near future by, e.g., the Dark Energy Spectroscopic Instrument (DESI; Levi et al. 2013; DESI Collaboration et al. 2016) and the 4 m Multi-Object Spectroscopic Telescope (de Jong et al. 2012), where reliable estimates of z_sys and M_BH will be crucial to extract the science value from these surveys. In forthcoming papers we will present, among other facets, redshift calibrations via indicative emission lines such as [O iii] (B. M. Matthews et al. 2021, in preparation), SMBH estimates using the Hβ and Mg ii profiles measured in this survey (C. Dix et al. 2021, in preparation), and correlations among UV–optical emission lines (e.g., Boroson & Green 1992; Wills et al. 1999; Shen & Liu 2012).

In the future, we should continue to push the redshift barrier for the Hβ and [O iii]  emission lines, as current investigations have been confined to z ≲ 3.5, in order to gain an increased understanding of the coevolution of SMBHs and their host galaxies, along with more reliable redshifts. However, at redshifts higher than z ∼ 3.5, these observations cannot be obtained via ground-based telescopes. Future studies in this respect could include a two-pronged approach using small calibration surveys. The first survey, for example, can use higher resolution instruments such as Gemini's Spectrograph and Camera for Observations of Rapid Phenomena in the Infrared and Optical (SCORPIO; Robberto et al. 2018) which will better measure weak emission-line profiles and obtain more accurate measurements of the prominent emission lines. This information will reinforce the measurements of this survey and allow for more confident applications to much higher redshifts, even beyond z > 6. The second survey would be a select sample of a few dozen highly luminous z > 3.5 objects using space-based observations from the James Webb Space Telescope (Gardner et al. 2006) for optimal spectral quality, with the possibility for a contemporaneous SCORPIO survey to obtain measurements of lines such as C iv from the ground.

This work is supported by National Science Foundation grants AST-1815281 (B.M.M., O.S., C.D.), AST-1815645 (M.S.B., A.D.M.), AST-1516784 (W.N.B.), and AST-1715579 (Y.S.). A.D.M. was also supported, in part, by the Director, Office of Science, Office of High Energy Physics of the U.S. Department of Energy under Award No. DE-SC0019022. Y.S. acknowledges support from an Alfred P. Sloan Research Fellowship. We thank an anonymous referee for thoughtful and valuable comments that helped improve this manuscript. 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. We thank Jin Wu and the Gemini Observatory staff for helpful discussions, and providing assistance with data reduction for this project. This work was enabled by observations made from the Gemini North telescope, located within the Maunakea Science Reserve and adjacent to the summit of Maunakea. We are grateful for the privilege of observing the universe from a place that is unique in both its astronomical quality and its cultural significance.