The Time-domain Spectroscopic Survey: Target Selection for Repeat Spectroscopy

The SDSS uses the imaging data gathered by a dedicated 2.5 m wide-field telescope (Gunn et al. 2006), which collected light from a camera with 30 2k × 2k CCDs (Gunn et al. 1998) over five broad bands—ugriz (Fukugita et al. 1996; Doi et al. 2010)—in order to image 14,555 unique deg² of the sky. This area includes 7500 deg²in the North Galactic Cap (NGC) and 3100 deg² in the South Galactic Cap (SGC). The Eighth Data Release (DR8; Aihara et al. 2011) provides the full imaging data set and updated photometric calibrations. This catalog provides the magnitudes and astrometry used in constructing our target samples; all coordinates hereafter are J2000.

The Stripe 82 region of the SDSS (S82; $-60^\circ \lt \alpha \lt 50^\circ$ and $| \delta | \lt 1\buildrel{\circ}\over{.} 27$) covers ∼300 deg² and has been observed ∼60 times on average to search for transient and variable objects (Frieman et al. 2008; Abazajian et al. 2009). These multi-epoch data probe timescales ranging from 3 hr to 8 yr and provide well-sampled five-band light curves. The S82 variable and standard star catalogs (Ivezić et al. 2007) are used to train the TDSS SES selection in Morganson et al. (2015), and therefore the hypervariables selection (Section 3.4). The S82 data are used for variability selection in the RQS program (Section 4).

PS1 utilizes a 1.8 m telescope equipped with a 1.4-gigapixel camera. Over the course of 3.5 years of the 3π survey, up to four exposures per year in five bands, ${g}_{\mathrm{PS}1},{r}_{\mathrm{PS}1},{i}_{\mathrm{PS}1},{z}_{\mathrm{PS}1},{y}_{\mathrm{PS}1}$ were taken across the entire $\delta \gt -30^\circ$ sky (for full details, see Tonry et al. 2012; Metcalfe et al. 2013). Each nightly observation consists of a pair of exposures separated by 15 min to search for moving objects. For each exposure, the PS1 3π survey has a typical 5σ depth of 22.0 in the g-band (Inserra et al. 2013). The instrumentation, photometric system, and the PS1 surveys are described in Kaiser et al. (2010), Stubbs et al. (2010), and Magnier et al. (2013), respectively. A high-quality subset of PS1 data was first released in Processing Version 1 (PV1), and PV2 added data through a later date in the observing, as well as some previously missed earlier observations via better analysis failure handling (for an overall description of the PS1 database, see Flewelling et al. 2016).

Whereas variability selection formed the basis for the first epoch (SES) target selection described in Morganson et al. (2015), multi-epoch imaging data were also used to select known stars and quasars for the extremely variable (or "hypervariable"; Section 3.4) and RQS quasar (Section 4) samples. The SES selection uses a three-dimensional parameter space (magnitude, PS1-only variability, and SDSS–PS1 difference), designed to achieve a high-purity variable sample at a typical surface density on the sky of ∼10 deg⁻². As part of the SES selection, hypervariables were selected based on a single variability metric V that parameterizes variability in a two-dimensional space of two variability terms S₁ and S₂:

$$\begin{eqnarray}V & = & {(\mathrm{median}{(| {\mathrm{mag}}_{\mathrm{PS}1}-{\mathrm{mag}}_{\mathrm{SDSS}}| )}^{2}+4\mathrm{median}{({\mathrm{Var}}_{\mathrm{PS}1})}^{2})}^{1/2}\\ & = & {({S}_{1}^{2}+4{S}_{2}^{2})}^{1/2},\end{eqnarray} \tag{ 1 }$$

(Equation (11) of Morganson et al. 2015). Qualitatively, S₁ is the PS1–SDSS difference and represents long-term (multi-year) variability, and S₂ is the PS1-only variability characterizing short-term (days to a few years) variability. This non-standard variability measure was intended to combine short-term and long-term variability measures into one quantity. The training sets used to derive this quantity, as well as the detailed statistics of this selection, can be found in Morganson et al. (2015). Hypervariables with $V\gt 2$ mag were prioritized during the SES selection, whereas for the FES targets, a lower V threshold was used (see Section 3.4). As for the SES selection, an updated version of the "ubercalibrated" PS1 data (Schlafly et al. 2012), which include PV1 data up through 2013 July, was used to select FES hypervariables. For detailed definitions, see Morganson et al. (2015).

The RQS program targets known SDSS quasars based on median SDSS magnitude and highly significant variability (see Section 4). Here, the variability selection is based on the reduced ${\chi }^{2}$ of the light curve:

$$\begin{eqnarray}&&{\chi }_{\mathrm{pdf}}^{2}=\displaystyle \frac{1}{n-1}\displaystyle \sum _{i=1}^{n}{\left(\displaystyle \frac{{m}_{i}-\mu }{{\sigma }_{i}}\right)}^{2},\end{eqnarray} \tag{ 2 }$$

where n is the number of data points in a given filter, m_i, ..., m_n are the individual magnitudes, ${\sigma }_{i}$ is the error associated with m_i, and μ is the mean magnitude. The reduced ${\chi }_{\mathrm{pdf}}^{2}$ in both g and r bands is used to define the variability-selected subsamples (see Section 4 for detailed criteria). The ${\chi }_{\mathrm{pdf}}^{2}$ cuts remove noisy, sparse light curves where the variability is at low signal-to-noise ratio (S/N; e.g., see Figure 1), rather than removing quasars that do not intrinsically vary, since essentially all quasars should be variable in the absence of poor photometry or systematics (Butler & Bloom 2011).

Zoom In Zoom Out Reset image size

For the ${\chi }_{\mathrm{pdf}}^{2}$ calculation, all primary and secondary SDSS photometric observations are considered, along with PS1 PV2 data, matched to within $1^{\prime\prime}$ and without regard to morphology or data quality flags. The PS1 data include observations up through 2013 December, and the error inflations derived in Morganson et al. (2015) are applied. Point-spread function (PSF) magnitudes are adopted, as we are interested in the nuclear variability (e.g., in the case of resolved active galactic nuclei (AGNs)).Before the ${\chi }_{\mathrm{pdf}}^{2}$ calculation, the SDSS magnitudes are transformed to the PS1 system as described in Morganson et al. (2015). We also first remove deviant points, as defined by being $\gt 0.5$ mag from either the SDSS or PS1 running average. For this outlier rejection, we consider SDSS and PS1 data separately due to the gap in time between the two data sets. The outlier rejection is only applied to SDSS light curves with $n\geqslant 10$ points (i.e., Stripe 82 data), and to PS1 data with $n\geqslant 5$ points. To compute the running averages, we use a window of five points for SDSS data and three points for PS1 data. The outlier rejection affects 5% (10%) of the PS1 (Stripe 82) light curves. Among these, 12.5% (7%) of the PS1 (Stripe 82) epochs on average are removed as a result.²⁹

Figure 1 shows example light curves for four sources with S82 and PS1 photometry and with different values of ${\chi }_{\mathrm{pdf}}^{2}$.

All samples are constructed based on known spectroscopic classifications in SDSS. The BOSS spectrographs and their SDSS predecessors are described in detail by Smee et al. (2013). SDSS-III BOSS (Dawson et al. 2013) significantly expanded the coverage in the SGC (approximately $-2^\circ \lt \delta \lt 35^\circ$, $-30^\circ \lt \alpha \lt 30^\circ$, see Figure 2), and revisited the entire NGC area. Since the FES targets were planned before the start of SDSS-IV, they were restricted to SDSS-I/II/III observations. SDSS-IV eBOSS observations are planned to cover the entire SGC BOSS footprint and about half the NGC, targeting mostly quasars and galaxies (Blanton et al. 2017). Since the RQS targets were compiled in summer 2016, they also draw from newly confirmed quasars targeted as part of SDSS-IV (Myers et al. 2015; Palanque-Delabrouille et al. 2016). The ELG survey (Raichoor et al. 2017) footprint³⁰ covers most of S82, in particular the "Thin82" and "Thick82" regions outlined in Figure 2, totalling 620 deg² in the SGC. This area is also covered by eBOSS plates designed for luminous red galaxies (LRGs; Prakash et al. 2016) and quasar targets (including some that were observed after the SDSS-IV observations shown in Figure 2). The ELG footprint also includes 600 deg² in the NGC.

Zoom In Zoom Out Reset image size

The wavelength coverage of the SDSS (BOSS) spectrographs is 3800–9200 Å (3600–10400 Å), with a spectral resolution ranging from 1850 to 2200 (1560–2650). The SDSS and TDSS spectra presented in this work all have ${\lambda }_{\mathrm{eff}}=5400$Å, i.e., the plate holes were drilled to maximize the S/N at ${\lambda }_{\mathrm{eff}}$, and the BOSS spectra either have ${\lambda }_{\mathrm{eff}}=5400$ Å or ${\lambda }_{\mathrm{eff}}=4000$ Å (Dawson et al. 2013). To accurately compare spectra with differing ${\lambda }_{\mathrm{eff}}$, one must correct the spectra using the prescriptions given in Margala et al. (2016), Guo & Gu (2016), and Harris et al. (2016). Note that the Margala et al. (2016) corrections are applied in the DR14 release of spectra from the BOSS spectrographs (Abolfathi et al. 2017).

2.4.1. Quasar Catalogs and Temporal Baselines

As described by Richards et al. (2002), the bulk of SDSS quasar target candidates in SDSS-I/II were selected for spectroscopic observations based on their optical colors and magnitudes in the SDSS imaging data or their detection in the FIRST radio survey (Becker et al. 1995). Low-redshift, $z\lesssim 3$, quasar targets were selected based on their location in ugri-color space and the quasar candidates passing the ugri-color selection were selected to a flux limit of i = 19.1. High-redshift ($z\gtrsim 3$) objects were selected in griz-color space and are targeted to i = 20.2. Furthermore, if an unresolved, $i\leqslant 19.1$ SDSS object was matched to within $2^{\prime\prime}$ of a source in the FIRST catalog, it was included in the quasar selection. Additional quasars were also (inhomogeneously) discovered and cataloged in SDSS-I/II using X-ray, radio, and/or alternate odd-color information, and extending to fiber-magnitudes of about $m\lt 20.5$ (e.g., see Anderson et al. 2003).

Unless otherwise stated, we select quasars for repeat spectroscopy from one of the visually vetted quasar catalogs: the SDSS-I/II DR5/7 quasar catalogs (DR5Q, DR7Q; Schneider et al. 2007, 2010; Shen et al. 2011) or the DR12 quasar catalog (DR12Q, final quasar catalog of SDSS-III; Pâris et al. 2017). The RQS target selection considers confirmed SDSS-IV quasars from post-DR13 data (Myers et al. 2015; Palanque-Delabrouille et al. 2016), specifically the SpAll database version v5_9_1, which covers a region in the SGC (see Figure 2 and Table 2 for the SDSS-IV coverage at the time of target selection). The following SDSS-IV spectra are excluded from consideration.

• 1.  
  Those objects lacking primary (mode = 1) magnitudes in DR10 (these objects are faint and few in number);
• 2.  
  Those with OBJTYPE = SKY;
• 3.  
  Those objects with morphological TYPE = 0, according to DR10 photometry; and
• 4.  
  Those at redshift $z\geqslant 0.8$ with morphological TYPE = 3, since a resolved quasar should be at a lower redshift.

In order to determine the number of existing spectra in Section 4, we extract all spectroscopy within $2^{\prime\prime}$ from the SDSS DR12 SpecObjAll database. For the new SDSS-IV objects, we use the NSPEC field from SpAll-v5_9_1.

Figures 3 and 4 show the anticipated distribution of time lags between spectra for SDSS quasars. Figure 4 also shows the existing distribution of time lags and displays these distributions as a function of absolute magnitude M_i, where the M_i values are estimated from the apparent i magnitudes and distance modulus (no K-correction is applied). Note that these figures do not include the well-sampled quasar cadence from the SDSS Reverberation Mapping Program (Shen et al. 2015).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Magnetic activity is ubiquitous in stars at the transition between the M and L spectral types (ML dwarfs). In optical spectra, this activity is best identified with the Hα emission line, which traces chromospheric heating on these low-mass objects (e.g., Gizis et al. 2002; West et al. 2011). Hα is the optimal diagnostic in part because it is found in a relatively red portion of the spectrum, so is easier to observe for these very cool, red objects. Serendipitous and dedicated observations of Hα emission on multiple timescales have indicated Hα emission varies (sometimes dramatically; Hall 2002) on multiple timescales (e.g., Berger et al. 2009; Schmidt et al. 2015). Chromospheric heating covers only a small portion of the surface (<1% Schmidt et al. 2015), and those regions rotate in and out of view on timescales of hours to days (due to relatively rapid rotation; Reiners & Basri 2008) leading to Hα variability. On timescales of weeks to months, we expect the chromospheric emission regions to change in size due to underlying shifts in the magnetic field (similar to shifts in sunspots). Hα may also show variability over year- to decade-long timescales based on long-timescale magnetic field changes that are similar to the 11 year solar cycle. Analyses of Hα variability on ML dwarfs have so far been limited to ∼20 objects serendipitously observed by multiple groups, but the data indicate that 30%–50% of ML dwarfs exhibit significant variability over timescales that span months to years (Schmidt et al. 2015).

By comparing original spectra of ML dwarfs from the SDSS legacy survey with an additional spectrum from TDSS, we have a unique opportunity to monitor changes in Hα emission lines over timescales of 6–14 years. These observations can either be taken as indicators of the level of overall variability, or could be combined with data over shorter timescales to detect decadal magnetic cycles. The SDSS data for ML dwarfs also allow three-dimensional Galactic kinematics that can be leveraged to examine age and activity correlations among the multi-epoch observations.

To select the FES sample of ML dwarfs, we combined the West et al. (2011) M dwarf and Schmidt et al. (2010) L dwarf catalogs. We selected a subset of dwarfs from those catalogs with magnitudes between $17\lt i\lt 21$ and spectral types from M7 to L3. We also required an average S/N $\gt 3$ per 1.5 Å pixel in the continuum surrounding Hα (6530–6555 Å and 6575–6600 Å) so that the presence and strength of Hα emission can be reliably measured (e.g., West et al. 2008). The dwarfs in our initial sample are contained in the BOSS Ultracool Dwarfs catalog (S. J. Schmidt et al. 2017, in preparation), and we required that each of them have a photometric distance (based on SDSS photometry), proper motion (from SDSS-2MASS-WISE positions) and radial velocity (based on SDSS spectroscopy) from that catalog. We also restricted the sample to dwarfs within 300 pc of the Galactic plane.

The selection criteria resulted in a total of 3739 M7, 534 M8, 153 M9, and 23 L dwarfs. To reduce the sample to ∼1000 ML dwarfs, we binned the data by height above the Galactic plane and restricted each 25 pc wide bin to 60 dwarfs randomly drawn from each spectral type. The final target list included 1036 stars (583 M7, 283 M8, 147 M9, and 23 L dwarfs). Initial data from the ML dwarf FES sample included dwarfs that have no change in their activity level as well as those that have strong variability. The spectrum of a strongly active and variable M dwarf is shown in Figure 5.

Zoom In Zoom Out Reset image size

Recent studies have demonstrated that close binaries consisting of white dwarf-M dwarf pairs (WD-dM binaries) have a significant effect on the magnetic activity of their main sequence components (e.g., Morgan et al. 2012). The binary separations where increased magnetic activity is observed extend to ∼100 au. While many of the WD-dM are unresolved photometrically, the two components can be separated in low-resolution spectroscopy due to their vastly different spectral energy distributions. While there is evidence of increased magnetic activity in close pairs, there has been limited analysis of the variability of magnetic activity in binary systems. Variability studies can distinguish among possible causes of activity (e.g., irradiation, accretion, disk disruption, and spin-up). This program aims to re-observe ∼400 WD-dM binaries identified via their spectral energy distributions. By measuring the magnetic activity of the M dwarf via the Hα equivalent width (EW), the goals are to determine (i) the effect of binary separation on the variability of magnetic activity, (ii) the effect of rotation on stellar activity in close binaries, and (iii) the WD cooling age, spectral type, orbital parameters, metallicity, and Galactic height, and the corresponding effects on magnetic activity.

Starting with the WD-dM binary sample of Morgan et al. (2012), targets that met the following criteria were selected:

• 1.  
  within the magnitude range $17\lt i\lt 21$,
• 2.  
  clear Hα emission from visual inspection,³² and
• 3.  
  accurate proper motion measurements from the SDSS-USNOB proper motions table (MATCH = 1; PMRA and PMDEC $\ne \,0;$ and DIST22 $\gt 7^{\prime\prime}$, where DIST22 is the distance to the nearest neighbor with $g\lt 22$). This criterion is necessary for the fiber holes to be drilled in the correct locations.

The resulting sample contains 402 active WD-dM pairs that span several M dwarf spectral types. An example of a binary with variable Hα is shown in the bottom panel of Figure 5.

Carbon in stellar atmospheres—indeed, most of the carbon in the universe—is produced by the triple-α process of helium fusion (3 ⁴He$\,\to {}^{12}$C) in the interiors of red giant stars. Strong carbon molecular bands are historically expected to be seen only in asymptotic giant branch (AGB) stars that have experienced a third "dredge-up" (Iben & Renzini 1983). However, among stars showing such C₂ and CN molecular bands (C stars), the main sequence dwarf carbon stars (dCs) are numerically dominant in the Galaxy (Green et al. 1992). The accepted explanation for dCs is that they must all be in post-mass transfer binaries, where the former AGB star has since become a WD, leaving a carbon-enhanced dC primary. Indeed, a handful of "smoking gun" systems reveal evidence for this evolutionary scenario, having composite spectra with a hot DA WD component (Heber et al. 1993; Liebert et al. 1994; Green 2013; Si et al. 2014). While the connection has rarely been made in the literature, dC stars, having been rejuvenated by mass accretion, would likely be seen as blue stragglers if they were within a coeval stellar cluster. They are also probably the dwarf progenitors of the typically more luminous carbon-enhanced metal-poor (sgCH, CH) and perhaps barium (Ba ii) stars, which all show carbon and s-process enhancements (see the discussion and references in De Marco & Izzard 2017). The detection of WD companions, and the characterization of the orbital properties of dC stars, are therefore important for understanding the mass transfer processes that give rise to this fascinating family of stars.

However, to date, the only dC star with a measured binary orbit is the prototype dC G77-61 (Dahn et al. 1977; Dearborn et al. 1986), a single-line spectroscopic binary (245 day period and semi-amplitude 20 km s⁻¹), where the WD has cooled to ${T}_{\mathrm{eff}}\,\lt$ 6000 K. Since G77-61 represents the only known dC with a proven radial velocity (RV) orbit, the mass-transfer hypothesis for dCs remains to be confirmed, and can be investigated only through the properties of dCs to detect and characterize host binary systems. Models for dC formation in both the disk and halo (de Kool & Green 1995) predict a bimodal orbital period distribution, with a large peak at ∼a decade (for accretion of the AGB wind at a binary separation ∼10 au), and a smaller peak at ∼a year (for separations $\lesssim 1$ au) corresponding to systems that underwent a common envelope phase, where the companion was subsumed in the expanding atmosphere of the AGB star when it filled its Roche lobe. These models reproduce the better-studied distributions of CH and Ba ii giants, whose progenitors are almost certainly the dCs. The relic distribution of dC binary orbits should reveal the relative importance and efficiency of these types of accretion, which can substantially modify the dC, leaving it hotter and bluer (and perhaps more rapidly rotating) than expected for its age.

Green (2013) identified 1220 faint ($r\gtrsim 17\mbox{--}21$) C stars from SDSS spectra, ∼five times more than previously known, but also including a wider variety of dC properties than past techniques such as color or grism selection have netted. From those with significant proper motion measurements, they identified 730 definite dwarfs, including eight systems with clear DA WD companions. This data set represents the first significant sample of bona fide dCs appropriate for a population study.

The statistical analysis of large samples of sparsely sampled RV curves can be used to constrain the underlying properties (binary fraction and separation distribution) of the corresponding binary population (e.g., Maoz et al. 2012). The TDSS dwarf carbon star FES program will provide a second epoch of SDSS spectroscopy to measure RV variability for a large sample of dC stars, to produce first constraints on their binarity and the distribution of their orbital properties. The main aims of this program are to (i) test the binary evolution hypothesis for dC stars, (ii) constrain the distribution of orbital separations, and (iii) trace the chemistry and evolution of the oldest AGB stars. The strategy used in the program will: (i) measure the RV shift ${\rm{\Delta }}\mathrm{RV}$ for dC stars between SDSS and TDSS (5–18 years), and (ii) constrain the separation distributions and hence the mass transfer mode.

For the dC FES program, we selected all 730 SDSS C stars from Green (2013) that were listed as dwarfs with high probability based on either their measured proper motions, or because they were identified from their SDSS spectra as composite DA/dC spectroscopic binaries. We added another 99 dC stars found by Si et al. (2014), totalling 829 unique dC stars for repeat spectroscopy within the TDSS. An example of SDSS archival and TDSS spectra of a dC in our program is shown in Figure 6.

Zoom In Zoom Out Reset image size

This program targets the most highly photometrically variable stars/classical pulsators (defined as hypervariable stars), as well as hypervariable quasars in the TDSS. The spectroscopic variability for these objects can potentially reveal large structural changes in astrophysical sources, and is useful for finding rare, transient phenomena such as "changing-look quasars" (e.g., LaMassa et al. 2015; MacLeod et al. 2016; Ruan et al. 2016a; Runnoe et al. 2016). More importantly, this program is exploring unknown territory and therefore the scientific returns could be quite substantial.

During the variable target selection in the main TDSS SES program (Morganson et al. 2015), hypervariables are identified using a modified variability characterization that is designed to work in the extreme regions of variability space (see Section 2.3). The hypervariable targets for these FES programs all have previous spectra in the SDSS DR11 SpecObjAll table. Since the pipeline classifications were adopted here without further verification, a few targeted stars may actually be quasars, and vice versa.

For stars, defined as $i\lt 20$ point sources (uncorrected for Galactic extinction) with CLASS = STAR, the top 0.5% most significantly variable objects were selected, corresponding approximately to $V\gt 0.3$ mag (see Figure 7). These sources lie outside an approximately elliptical contour with SDSS-PS1 difference of 0.2 mag, a PS1-only variability of 0.15 mag, or some intermediate combination of the two (see Figure 5, Morganson et al. 2015). The SDSS images of these sources are visually examined to remove objects with close neighbors, nearby diffraction spikes or other imaging issues that could significantly affect photometry. The above criteria select 1150 stars (∼0.05 deg⁻²), which have the target flag TDSS_FES_HYPSTAR. Inspection of these targets' initial SDSS spectra suggest that this sample is rich in RR Lyrae variables and also includes M dwarfs, carbon stars and stars that are difficult to classify. For an example target, see Figure 8.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For quasars, defined as $i\lt 20$ point sources (uncorrected for Galactic extinction) with CLASS = QSO, the top 2% most significantly variable objects were selected, corresponding to approximately $V\gt 0.5$ mag (see Figure 7). These sources lie outside an approximately elliptical contour with SDSS-PS1 difference of 0.7 mag, a PS1-only variability of 0.25 mag, or some intermediate combination of the two. The SDSS images of these sources are also visually examined to remove objects with close neighbors, nearby diffraction spikes or other imaging issues that could significantly affect photometry. The above criteria select 1555 quasars (∼0.05 deg⁻²), which have the target flag TDSS_FES_HYPQSO. Inspection of these targets' initial SDSS spectra suggest that this sample is rich in BAL quasars and blazars, but otherwise contains a wide range of quasar types (we leave a detailed census to a later publication). For example spectra, see Figure 8.

This FES program will build upon recent systematic, sample-based studies of BAL variability (e.g., Barlow 1993; Lundgren et al. 2007; Filiz Ak et al. 2012, 2013, 2014; Vivek et al. 2014, and references therein) by re-observing ∼3000 BAL quasars from the SDSS and BOSS. About 2/3 of the sample was selected from Gibson et al. (2009) and has already been mostly observed as part of a BOSS ancillary proposal (see Filiz Ak et al. 2013) and probes rest-frame timescales of ≈4–7 years. The TDSS is obtaining a third spectroscopic epoch for this subsample, typically spanning an additional 1–3 years in the rest frame beyond the most recent BOSS observations. The TDSS data yield improved measurements of the dependence of BAL EW variability upon rest-frame timescale, enabling a test of the extent to which long-term variability trends found in the SDSS-I/II versus BOSS data persist. A third epoch also allows for the possibility of detecting BAL acceleration or re-emergence/disappearance. The long timescales sampled by this project are highly beneficial since velocity shifts associated with BAL acceleration/deceleration accumulate over time; the first results on this project's BAL acceleration are presented in Grier et al. (2016). Constraints upon BAL disappearance and emergence provide key insights into the lifetime of BALs. Furthermore, BAL re-emergence events at the same velocity argue strongly against models where the variability is due to gas motions, instead favoring models where ionization changes play a key role. The first results on BAL re-emergence/disappearance are presented in McGraw et al. (2017). Finally, these observations further characterize the coordinated EW variations of BAL quasars with multiple troughs; these coordinated variations constrain models for BAL variability (e.g., Filiz Ak et al. 2012, 2013).

Figure 9 shows two examples of a variable BAL quasars observed in the TDSS. The selection recipes used to obtain these targets are detailed below in a step-by-step manner.

Zoom In Zoom Out Reset image size

3.5.1. Main BAL Sample

3.5.2. DR12 Objects

Supermassive black hole binaries (SBHBs) are throught to be a common consequence of the merger of two massive galaxies. According to the evolutionary scenario described by Begelman et al. (1980), sometime after the merger of the parent galaxies, the two black holes form a bound binary whose separation decays first by dynamical friction, then by scattering of stars, and finally by the emission of gravitational radiation.³⁵ The slowest stage in this evolutionary scheme is thought to correspond to an orbital separation of $0.01\,\mathrm{pc}\lesssim a\lesssim 1\$ pc. Thus, observational efforts have focused on finding SBHBs at these orbital separations using RV variations of the BELs (by analogy with double-lined or single-lined spectroscopic binary stars; e.g., Gaskell 1983, 1996). So far, direct observational evidence for SBHBs with two active black holes via this method has been elusive (e.g., Eracleous et al. 1997; Liu et al. 2016). Recent surveys have concentrated on candidate SBHBs with one active black hole and have utilized the large samples of quasar spectra available in the SDSS archive (Tsalmantza et al. 2011; Eracleous et al. 2012; Ju et al. 2013; Shen et al. 2013; Liu et al. 2014; Runnoe et al. 2017). The general strategy of these surveys is to select quasars whose broad Balmer or Mg ii lines are offset from the frame defined by the narrow lines by $\sim 1000\ \mathrm{km}\ {{\rm{s}}}^{-1}$ or more and/or search for systematic RV variations between the first-epoch spectra and spectra taken several years later.

This program is a continuation of the work of Ju et al. (2013) who studied the broad Mg ii emission lines of $0.36\lt z\lt 2$ quasars with multiple SDSS observations. The spectra from this program can be used to detect velocity shifts in SBHBs with separations of $\sim 0.1$ pc and orbital periods of ∼100 years, assuming that the black holes have masses of order ${10}^{9}\ {M}_{\odot }$.

From the sample of all quasars in DR7Q with multiple SDSS spectra of the Mg ii line, Ju et al. (2013) identified seven robust SBHB candidates along with 57 more candidates that were less secure, for a total of 64 targets. The program is designed to obtain a third-epoch spectrum for all candidates, with highest priority given to the seven robust candidates, in order to search for monotonic velocity shifts relative to first epoch. The first results from this program were reported in Wang et al. (2017), in which the authors rule out a binary model for the bulk of candidates by comparing the variations in the velocity shifts over 1–2 years and 10 years. They also find that $\lesssim 1$% of active SMBHs reside in binaries with ∼0.1 pc separations observed in the TDSS. The example shown in the bottom panel of Figure 9 is a candidate from Wang et al. (2017) with a prominent line shift.

Broad Balmer lines with double peaks, twin shoulders, or flat tops can be found in about 15% of radio-loud AGNs at $z\lt 0.4$ (Eracleous & Halpern 1994, 2003) and in about 3% of AGNs at $z\lt 0.33$ in the SDSS (Strateva et al. 2003), depending on radio-loudness and possibly Eddington ratio. Although a number of ideas have been discussed in the literature for the origin of these line profiles, a physical model attributing the emission to the outer parts of the accretion disk is the most successful in explaining the Balmer line profiles and other properties of these objects (see the discussion in Eracleous & Halpern 1994, 2003; Eracleous et al. 2009, and references therein). Thus, we refer to these objects as disk-like emitters hereafter. Previous long-term monitoring of disk-like emitters has sampled about two dozen objects over 20 years (e.g., Sergeev et al. 2000, 2017; Storchi-Bergmann et al. 2003; Gezari et al. 2007; Flohic 2008; Lewis et al. 2010; Popović et al. 2011, 2014, and references therein) at $z\lt 0.4$, most of which are radio loud.

This FES program expands the scope of past monitoring efforts by re-observing for at least one more epoch a much larger number of disk-like emitters drawn from the SDSS. This selection method leads to a much wider variety of objects than those targeted by previous campaigns, namely more luminous objects, objects with higher Eddington ratios, and radio-quiet objects. This program also targets objects at $z\sim 0.6$, which are even more luminous than those at $z\lt 0.4$.

Included in the target list are 1251 objects from DR7Q distributed over ∼6300 deg² (i.e., 0.2 deg⁻²). The targets comprise "classic" disk-like emitters (at $z\lt 0.33$ taken from Strateva et al. 2003) and higher-redshift analogs ($z\sim 0.6;$ from Luo et al. 2013), as well as additional objects identified by Shen et al. (2011). A total of 220 objects are "classic" disk-like emitters (objects whose Balmer profiles can easily be modeled by a rotating accretion disk; e.g., Eracleous et al. 2009) while the remaining objects have very asymmetric Balmer profiles that can plausibly be attributed to a perturbed disk (for example one with a prominent spiral) or to an SBHB (see Section 3.6). The magnitudes of the targets are $i\lt 18.9$. The TDSS spectra will cover Hα and Hβ for the $z\lt 0.4$ objects, and Hβ and Mg ii for the $z\sim 0.6$ objects. The time baseline will be $\gt 10$ years for most objects. The 1251 targets of this program include 28 objects identified as promising sub-pc binary SMBH candidates with observed Hβ line shifts between two epochs in SDSS-I/II from Shen et al. (2013).

By combining existing SDSS spectra and those collected during the TDSS, this program aims to address the following scientific goals. First, the observations will empirically characterize the variability of the BEL profiles, i.e., determine what property of the profiles is varying (e.g., width, asymmetry, shift, relative strengths and velocities of the peaks or shoulders), as well as the magnitude and timescale of the variations. Second, the data will be compared to a wide array of models of disk perturbations, including warps, self-gravitating clumps, and spiral or other waves. Third, this program aims to determine whether the variations represent systematic drifts of the line profiles and evaluate whether these changes are consistent with RV shifts due to orbital motion in an SBHB.

An example of disk-like emitter variability seen in one of the targets of this program is shown in the top panel of Figure 10.

Zoom In Zoom Out Reset image size

This program will yield second (or third) epoch spectra of bright, low-redshift ($z\lt 0.8$) SDSS quasars with existing high-S/N spectra (requiring that the median S/N per spectral pixel across the full SDSS spectral range is $\gt 23$). The combination of old and new spectra will be used to study the general broad-line variability of quasars, including line shape changes and line centroid shifts, on multi-year timescales. The scientific goals are similar to those of the previous program (see Section 3.7). In addition to furthering our understanding of the dynamics of the gas in the broad-line region, the data from this program will be important for two more applications: (i) a comparison of the variability properties of typical quasar BELs to the variability properties of disk-like emission lines (see Section 3.7), and (ii) selection of SBHB candidates via velocity shifts.

The focus of this program is quasars in DR7Q at $z\lt 0.8$. Thus, the spectra will include the Hβ line, as well as the narrow [O iii] doublet that will provide a reliable redshift and a velocity reference (e.g., Hewett & Wild 2010). Included in this sample are 1486 quasars with a median S/N $\gt 23$ per pixel.

For an example of a quasar targeted in this program, see Figure 10. This program is also producing serendipitous discoveries, for example the changing-look quasar from Runnoe et al. (2016) was identified from NQHISN spectra.