Discovery of the most luminous quasar of the last 9 Gyr

Abstract

We report the discovery of a bright ( $g = 14.5$ mag (AB), $K = 11.9$ mag (Vega)) quasar at redshift $z=0.83$ — the optically brightest (unbeamed) quasar at $z>0.4$ . SMSS J114447.77-430859.3, at a Galactic latitude of $b=+18.1^{\circ}$ , was identified by its optical colours from the SkyMapper Southern Survey (SMSS) during a search for symbiotic binary stars. Optical and near-infrared spectroscopy reveals broad Mg ii, H $\unicode{x03B2}$ , H $\unicode{x03B1}$ , and Pa $\unicode{x03B2}$ emission lines, from which we measure a black hole mass of $\log_{10}\! (M_{\mathrm{BH}}/\mathrm{M}_{\odot}) = 9.4 \pm 0.5$ . With its high luminosity, $L_{\mathrm{bol}} = (4.7\pm1.0)\times10^{47}\,\mathrm{erg\,s}^{-1}$ or $M_{i}(z=2) = -29.74$ mag (AB), we estimate an Eddington ratio of $\approx1.4$ . As the most luminous quasar known over the last ${\sim}$ 9 Gyr of cosmic history, having a luminosity $8\times$ greater than 3C 273, the source offers a range of potential follow-up opportunities.

1. Introduction

The observational study of quasars took off rapidly from the back-to-back-to-back papers of Nature’s 1963 March 16 issue, which featured the redshift determinations for 3C 273 and 3C 48 (Schmidt 1963; Oke 1963; Greenstein 1963). Barely two years later, the quasar 3C 9 became the first known object at a redshift greater than 2 (Schmidt 1965). But as exemplified by the 5-mag difference in the optical brightness between 3C 273 and 3C 9, the exploration towards higher redshifts became a push to fainter magnitudes.

Fortunately, the early recognition of radio-quiet quasars becoming prominent amongst blue, star-like objects beyond a magnitude of $V=14.5$ (Vega) provided an efficient means of identifying quasars from photometric techniques (Sandage 1965) and known quasars now number in the hundreds of thousands (see, e.g., the Million Quasar Catalogue, v7.5, hereafter Milliquas; Flesch 2021). Despite the proliferation of wide-area surveys across a range of wavelengths over the intervening fifty years, the search for bright quasars remains unfinished.

Here, we report on a spectroscopic investigation of a bright, blue, point-like source selected from the SkyMapper Southern Survey Data Release 2 (SMSS DR2; Wolf et al. 2018a; Onken et al. 2019), which demonstrates that SMSS J114447.77-430859.3 (SMSS DR2 object_ida 84280208; hereafter, J1144) is a $z=0.83$ quasar. Aside from one blazar object at $z=0.6$ (PKS 1424+240), this makes J1144 the optically brightest quasar known above a redshift of 0.4.

Spectroscopy of J1144 was first acquired during a search for symbiotic binaries, in which cool giant stars accrete onto smaller companions, using SMSS DR2 (Lucy 2021). While any known active galactic nucleus (AGN, as identified in SIMBAD; Wenger et al. 2000) was excluded, J1144 had only been identified as an AGN candidate by its near-IR and IR colours. It was identified as a candidate by Edelson & Malkan (2012) using the Two Micron All-Sky Survey (2MASS; Skrutskie et al. 2006) and the Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010; Mainzer et al. 2011). Similarly, Secrest et al. (2015) utilised the AllWISEb update to the IR dataset and selected J1144 as a quasar candidate from its IR colours alone. Notably, Shu et al. (2019) even estimated a photometric redshift for J1144 of $z=0.82$ from DR2 of the Gaia satellite mission (Gaia Collaboration et al. 2016; Gaia Collaboration 2018) and the unWISE (Schlafly, Meisner, & Green 2019) revision to the photometry from WISE.

However, the WISE colours of symbiotic stars sometimes fall in typical AGN selection regimes, which is why such sources were not excluded from the symbiotic star search. In fact, the flickering, accretion-powered symbiotic star EF Aql has an AGN-like $(W1-W2) \approx 0.9$ colour and was discovered via the UV-bright Quasar Survey (UVQS; Monroe et al. 2016; Margon et al. 2016; Zamanov et al. 2017). Lucy (2021) explored various SMSS selection mechanisms for different types of symbiotic stars, and J1144, being much bluer than most isolated cool giant stars, was caught by a $(u-g)$ / $(u-v)$ colour-only selection. Thus, along with 232 other symbiotic star candidates, optical spectroscopy of J1144 was obtained with the goal of confirming the presence of a cool giant star with emission lines, indicative of symbiotic binarity. J1144 was the only observed source to show AGN emission lines.

In Section 2, we describe the observations and data processing. In Section 3, we analyse the spectroscopic data and estimate the mass of the central black hole (BH). We summarise additional data available for J1144 in Section 4 and compare J1144 to other bright quasars in Section 5. Section 6 discusses the outlook to further study and utilisation of J1144, as well as the prospects for additional such discoveries in the future. Throughout the paper, we use Vega magnitudes for Gaia and IR data, and AB magnitudes for the SkyMapper passbands: uvgriz. We adopt a flat $\Lambda$ CDM cosmology with $\Omega_{\mathrm{m}}=0.3$ and a Hubble-Lemaître constant of $H_0=70\,\mathrm{km\,s}^{-1}\,\mathrm{Mpc}^{-1}$ .

2. Observations and data processing

The spectroscopic portion of the symbiotic star programme (described in Lucy 2021) principally used the South African Astronomical Observatory (SAAO) 1.9-m telescope and its SpUpNIC instrument (Spectrograph Upgrade: Newly Improved Cassegrain; Crause et al. 2019). Following the initial classification of J1144 as an AGN, additional optical spectroscopic data was obtained at higher spectral resolution with the Australian National University (ANU) 2.3-m telescope at Siding Spring Observatory (SSO) using the Wide Field Spectrograph (WiFeS; Dopita et al. 2007; Dopita 2010), and near-IR spectroscopy was obtained with the TripleSpec4.1 instrument (Schlawin et al. 2014) on the Southern Astrophysical Research (SOAR) 4.1-m telescope.

2.1 Optical spectroscopy with SAAO 1.9 m/SpUpNIC

SpUpNIC observations of J1144 were obtained on UT 2019 June 24 with the G7 grating. G7 is a low-resolution grating, which covered 3300–8930 Å with a resolving power $R\sim 500$ . A BG38 filter was manually inserted into the arc beam. The spectroscopic slit width was 2.24 arcsec in seeing of ${\sim}2^{\prime\prime}$ , and a spatial binning of 2 pixels was used. The exposure time was 1200 s, with the object at an airmass of 1.2.

Flux calibration was performed with observations of the spectrophotometric standard star, CD-32 9927 (Hamuy et al. 1994),c obtained on the same night. Data reduction, including bias subtraction and flat-fielding, used the standard tasks in the Image Reduction and Analysis Facility (IRAF; Tody 1986). A second pass at flux scaling was performed by processing the spectrophotometric standard in the same way as the science spectra and determining the residual correction needed to align the flux with the model values. The final signal-to-noise ratio (S/N) of the SpUpNIC spectrum was ${\sim}50$ per pixel.

2.2 Optical spectroscopy with ANU 2.3 m/WiFeS

WiFeS spectra were obtained on UT 2022 March 11 with two grating configurations. WiFeS is an integral field spectrograph and when used with a spatial binning of 2 pixels, it provides $1\times 1$ arcsec sampling over its $25\times 38$ arcsec field-of-view.

With the resolving power $R\sim 3000$ gratings, an exposure of 600 s was obtained, covering the wavelength range 3250–9550 Å across the two cameras of the spectrograph (the RT560 beamsplitter was used). For the high-resolution gratings ( $R\sim 7000$ ), an exposure time of 900 s was used. The B7000 grating covered 4180–5540 Å, while the I7000 grating covered 6810–9040 Å. All observations were obtained near an airmass of 1.2 with seeing of 1.8–2 arcsec in the i-band.

The spectrophotometric standard star, BD-12 2669 (Heap & Lindler 2010), was observed immediately after the J1144 spectra. The raw frames were reduced with the Python-based pipeline, PyWiFeS (Childress et al. 2014). We then extracted the spectra from the calibrated 3D data cubes using QFitsView,d selecting nearby source-free regions for sky subtraction. Variance and data-quality frames were extracted from the same regions. As with the SAAO data, a second iteration of flux scaling was performed by aligning the processed spectrophotometric standard spectrum to the model fluxes. The S/N in the final spectra ranged from 20-60 per pixel.

2.3 Near-IR spectroscopy with SOAR/TripleSpec4.1

We observed J1144 with TripleSpec4.1 on UT 2022 February 13 under NOIRLab programme 2022A-389756 (PI: X. Fan). TripleSpec4.1 utilises a fixed spectroscopic slit of $1.1\times28$ arcsec and produces cross-dispersed spectra that cover a simultaneous wavelength range from 0.95 to 2.47 microns, at a spectral resolution of ${\sim}3500$ .

The observations were performed in three consecutive ABBA patterns, with 40s exposures at each dither position. The detector was read out with 4-pair Fowler sampling (Fowler & Gatley 1990). The seeing was 1 arcsec in J-band. Observations of the A0V star, HIP 56984, were obtained immediately prior to J1144 to serve as a telluric and flux standard. The data were processed with the Spextool software package (Cushing, Vacca, & Rayner 2004; Cushing, Vacca, & Rayner 2014) written in the Interactive Data Language (IDLe), as modified for TripleSpec4.1.f Telluric correction was applied using the xtellcorr package (Vacca, Cushing, & Rayner 2003) in IDL. The final S/N was 50–100 per pixel.

Figure 1. Rest-frame spectrum of J1144, in units of $\mathrm{erg\,s}^{-1}$ Å^–1. Uncertainties are shown with grey errorbars, typically smaller than the thickness of the line. The dashed line shows a power-law continuum with slope $\unicode{x03B1}_{\lambda}$ =–1.56, which fits the spectrum well up to wavelengths of 7000 Å. The spectrum shown here has been corrected for Galactic reddening, but not for any internal reddening.

3. Spectroscopic analysis

We normalise the spectra by anchoring them to the photometric data which best overlap the cleanest wavelength regions of each spectrum. This involves the g, r, and i bands from SMSS DR3 and the 2MASS H-band for the near-IR spectrum. The bandpass details were retrieved from the Spanish Virtual Observatory (SVO) Filter Profile Serviceg (Rodrigo, Solano, & Bayo 2012; Rodrigo & Solano 2020). There is good agreement amongst the calibrations provided by the available photometric bands (typically better than 5%), consistent with the small levels of photometric variability discussed in Section 4.

We correct for Galactic reddening using the Fitzpatrick et al. (2019) extinction curve, as implemented in the dust_extinction Python package (Gordon 2021). We assume $R_V=3.1$ , and we take the $E(B-V)=0.123$ mag from Schlegel et al. (1998) and additionally apply the $\times0.86$ correction factor of Schlafly & Finkbeiner (2011).h The spectra are combined as the weighted mean on a new wavelength grid which sampled the spectra in pixels at $200\,\mathrm{km\,s}^{-1}$ spacing. The strong emission lines in the J1144 spectrum (Mg ii, H $\unicode{x03B2}$ , H $\unicode{x03B1}$ , and Pa $\unicode{x03B2}$ ) were used together (weighted mean) to provide a redshift estimate of $0.8314\pm0.0001$ . The observed spectrum is then transformed to rest-frame wavelengths and to luminosity in units of $\mathrm{erg\,s}^{-1}$ Å^–1, giving a velocity resolution of $109\, \mathrm{km\,s}^{-1}\,\mathrm{pixel}^{-1}$ and a final S/N between 100 and 250 per pixel. The combined spectrum is shown in Figure 1.

When fitting the spectrum, we separately consider the wavelength regimes around C iii, Mg ii, H $\unicode{x03B2}$ , H $\unicode{x03B1}$ , and Pa $\unicode{x03B2}$ . For each wavelength region, we first fit a combined power-law continuum and iron template. The best power-law slope for the combined UV/optical range is found to be $\unicode{x03B1}_{\lambda}=-1.56$ , although a flatter slope of –0.79 is a better fit for the near-IR, likely reflecting the contributions of hot dust in the region longward of $1\unicode{x03BC}$ m (cf. the WISE photometry in Figure 4). Because of the impact of the iron model on the remaining emission line profiles, we test the systematic effects of adopting various iron emission templates in the UV and optical portions of the spectrum. For the UV iron templates, we use those of Shen & Liu (2012),i Tsuzuki et al. (2006) and Mejia-Restrepo et al. (2016), while for the optical templates, we use those of Boroson & Green (1992; BG92, hereafter), Tsuzuki et al. (2006), Bruhweiler & Verner (2008), and Park et al. (2022). The velocity broadening of the template is a free parameter in each fit. On average, the best-fit iron full width at half-maximum (FWHM) was ${\sim}2500$ km s^–1. The systematic errors arising from the iron template choice are square-added to the statistical errors in the fit results below.

Figure 2. Emission line fits to C iii, Mg ii, H $\unicode{x03B2}$ , H $\unicode{x03B1}$ , and Pa $\unicode{x03B2}$ , as indicated in each panel. Black lines indicate the data. The model is plotted with progressively added elements: the power-law continuum (orange), then the pseudo-continuum from the broadened iron template (blue), then the emission line fits (red). The red dashed lines indicate the three Gaussian profiles used to fit each line. The particular fits shown here use the Shen & Liu (2012) and BG92 templates in the UV and optical, respectively.

For the subsequent steps of the spectral modelling, the emission lines are each fit with a sum of three Gaussian profiles. We estimate the statistical uncertainties on the fit parameters via a Monte Carlo approach, taking the RMS from 50 realisations in which the flux at each pixel is varied according to the error spectrum. Because of the weak [Oiii] and [Sii] emission and lack of evident narrow-line contribution to the Balmer lines, even in the original $R\sim7000$ spectrum of H $\unicode{x03B2}$ , no narrow-line subtraction is performed for the Balmer lines or the [Nii] lines near H $\unicode{x03B1}$ . Integrating $3\times$ the error spectrum over spectral windows of ${\pm}200\,\mathrm{km\,s}^{-1}$ around [Oiii] 5007 Å and [Sii] $6716+6731$ Å provides conservative upper limits of $5\times 10^{42}$ and $3\times 10^{42}\,\mathrm{erg\,s}^{-1}$ , respectively; however, the blending of H $\unicode{x03B1}$ with [Nii] precludes a similar upper limit estimate for the latter.

The emission line fits are shown in Figure 2. The C iii fit is poorly constrained, because of limited wavelength coverage, the lower spectral resolution of the SAAO data from which it is principally observed, and lack of de-blending with Aliii and Siiii. As a result, the parameters are omitted from Table 1. The features reward of H $\unicode{x03B2}$ are likely iron lines that have not been well modelled because of errors in the flux calibration at the long-wavelength limit of the optical spectra.

From the sum of the Gaussian fits, we determine the velocity shift (derived from the peak of the line profile), the integrated line luminosity, the FWHM, and the second moment of the line profile ( $\sigma_{\mathrm{line}}$ ). The emission line properties are summarised in Table 1. As was the case for the continuum luminosities, the emission line luminosity errors are dominated by the photometric calibration uncertainties.

3.1 Luminosity of J1144

From the power-law continuum fits, we determine the luminosities at several rest-frame wavelengths of interest:

$\log_{10}\! ( \lambda\,\mathrm{L}_{\lambda}(3000$ Å $)/\mathrm{erg\,s}^{-1} ) = 47.12 \pm 0.04$ ,

$\log_{10}\! ( \lambda\,\mathrm{L}_{\lambda}(5100$ Å $)/\mathrm{erg\,s}^{-1} ) = 46.94 \pm 0.04$ , and

$\log_{10}\! ( \lambda\,\mathrm{L}_{\lambda}(1\unicode{x03BC}\,\mathrm{m})/\mathrm{erg\,s}^{-1} )\,\,\,=\,46.67 \pm 0.04$ ,

where the errors are dominated by the 0.1 mag uncertainties in the flux calibration but do not incorporate the small additional factor of source variability (see Section 4). The luminosity at 3000 Å translates into an absolute magnitude of $M_{300\mathrm{nm}} = -28.70$ mag (AB). To aid comparison with higher-redshift samples, we extrapolate the best-fit continuum power-law (with $\unicode{x03B1}_{\lambda}=-1.56$ ) to shorter wavelengths and find $\log_{10}\! ( \lambda\,\mathrm{L}_{\lambda}(1450$ Å $)/\mathrm{erg\,s}^{-1} ) = 47.2$ or $M_{145\,\mathrm{nm}} = -28.36$ mag (AB). Using the prescription of Richards et al. (2006), we find $M_{i}(z=2) = -29.74$ mag (AB).

We adopt the bolometric correction (BC) for 3000 Å from Runnoe et al. (2012a, 2012b), although their BC values for 5100 Å (with spectral slope correction) or the BC values from Netzer (2019) yield similar results. The bolometric luminosity of J1144 is $(4.7\pm1.0)\times10^{47}\,\mathrm{erg\,s}^{-1}$ , where the uncertainty accounts for the variation arising from different BC assumptions. For the canonical radiative efficiency of 0.1 (e.g., Yu & Tremaine 2002), this equates to an accretion rate of ${\sim}80\,\mathrm{M}_{\odot}\,\mathrm{yr}^{-1}$ .

3.2 BH Mass of J1144

The wide wavelength coverage of our spectroscopic data provides access to several emission lines from the broad-line region (BLR), which can be used to estimate the mass of the central BH in J1144 through the virial relations. Anchored to the H $\unicode{x03B2}$ reverberation mapping results from the local AGN sample (see Peterson et al. 2004), the virial relations rely on a single epoch of line width and luminosity measurements to infer the velocity of the BLR gas around the BH, as well as the characteristic distance from the BH to the line-emitting gas in question (see, e.g., Cackett, Bentz, & Kara 2021). Each emission line may have its own velocity and distance, which should yield consistent mass estimates under the assumption that the gravity of the BH dominates the gas dynamics. However, we do note that J1144 represents an extrapolation of a factor of ${\sim}10$ in luminosity compared to the well measured H $\unicode{x03B2}$ reverberation mapping sample of Bentz et al. (2013).

We adopt the $M_{\mathrm{BH}}$ relation parameters indicated in Table 2 for a functional form of

(1) \begin{equation} M_{\mathrm{BH}} = 10^{A} \times \left(FWHM / 10^{3}\right)^B \times \left(L_{\mathrm{cont/line}} / 10^{44}\right)^{C}\end{equation}

for the emission line FWHM in units of $\mathrm{km\,s}^{-1}$ , and the luminosity of either continuum or emission line in units of $\mathrm{erg\,s}^{-1}$ . With the relations being primarily drawn from Le et al. (2020), the A parameters in Table 2 have been renormalised to that paper’s adopted virial factor of $f=1.12$ (from Woo et al. 2015) for these FWHM-based measurements.

With the emission line measurements of Table 1 and the continuum luminosities indicated above, we derive several complementary BH mass estimates. The $M_{\mathrm{BH}}$ values we estimate are presented in Table 3. For Mg ii, we do not apply the mass correction factors from Le et al. (2020) for the emission line shape (FWHM/ $\sigma_{\mathrm{line}}$ ratio) and spectral slope, which would increase that mass estimate by 0.3 dex and make it more discrepant with the other emission line results. The different emission lines produce BH mass estimates that span the range from $(1.9-3.8)\times 10^{9}\,\mathrm{M}_{\odot}$ . The high S/N of our spectroscopic data mean the statistical errors on the BH mass estimate are small compared to the systematic errors. Dalla Bontà et al. (2020) estimate the intrinsic scatter to be 0.371 dex for the best-measured FWHM-based virial relation (using H $\unicode{x03B2}$ and 5100 Å), and the systematic errors for the other estimates are likely to be larger. Thus, we take the mean value of our five measurements as the best estimate for the BH mass in J1144, $M_{\mathrm{BH}} = 2.6 \times 10^{9}\,\mathrm{M}_{\odot}$ , and conservatively adopt an uncertainty of 0.5 dex (cf. Vestergaard & Osmer 2009). Our measurements of the bolometric luminosity and BH mass yield an Eddington ratio of $\approx 1.4$ for J1144.

Table 1. Emission line fit results.

^a Measured with the emission line peak.

Table 2. Virial relations.

References: 1 - Le et al. (2020); 2 - Woo et al. (2015); 3 - Landt et al. (2013).

^a Rescaled from $f=1.4$ to $f=1.12$ .

In Table 4, we present a summary of the observed and derived properties of J1144.

Table 3. BH mass estimates.

Table 4. Summary of J1144 properties.

[1] From Schlegel et al. (1998).

3.3 Continuum Slope and Internal Reddening

Typical thin disk (Shakura & Sunyaev 1973) and slim disk (Abramowicz et al. 1988) models of BH accretion predict UV/optical continuum emission having a power-law slope of $\unicode{x03B1}_{\lambda} \approx -2.3$ ( $\unicode{x03B1}_{\nu} \approx +0.3$ ), though real quasars are rarely observed to be so blue (Xie et al. 2016). Taking such a spectral slope as the limiting case, we can assess the maximum amount of internal reddening that may be present in the emitted spectrum of J1144.

For a UV-flat reddening curve like that of (Gaskell & Benker 2007, GB07, hereafter), we find that a maximum intrinsic E(B-V) of 0.17 mag provides a reasonable fit to the data. In contrast, a UV-steep reddening curve – one lacking the strong bump at 2175 Å – such as the Gordon et al. (2003) model for the star-forming bar of the Small Magellanic Cloud, implies a maximum E(B-V) of 0.10 mag to avoid over-correcting C iii, but then under-corrects the spectrum near Mg ii. (Reddening curves retaining the 2175 Å bump perform even worse in the regime between C iii and Mg ii.) The GB07 reddening correction would lift the 3000 Å luminosity by 0.34 dex, which might be expected to increase each of the BH mass and the Eddington ratio by roughly half that margin. However, because the sources anchoring the virial relations have not been corrected for internal reddening, the appropriate adjustments for J1144 would be reduced in amplitude.

It is also worth highlighting that the power-law prescription for the rest-frame UV spectrum breaks down in accretion disk models at higher BH masses, as the high-energy turnover migrates to longer wavelengths (e.g., see Campitiello et al. 2018). For $M_{\mathrm{BH}} \sim 10^{9}\,\mathrm{M}_{\odot}$ , the departure from a power-law exceeds 0.1 mag between 2000 and 3000 Å, suggesting extreme care must be taken to disentangle the intrinsic spectral shape from any internal reddening. Observed-frame UV spectroscopy of J1144 should contribute to our ability to remove such degeneracies.

Finally, reddening will lead to an enhancement of the Balmer decrement, that is, the H $\unicode{x03B1}$ /H $\unicode{x03B2}$ flux ratio. Low-redshift quasars with blue spectral slopes (implying little intrinsic reddening) are often found to have Balmer decrements of 3.1 (Dong et al. 2008). Our observed Balmer decrement of 5.1 would imply E(B-V) $\approx 0.4$ mag, depending on the reddening curve adopted. However, Gaskell (2017) has argued that quasar Balmer decrements are consistent with an assumption of intrinsic Case B recombination and a flux ratio of 2.72, which would suggest even more reddening in J1144: E(B-V) $\approx0.5$ mag. In either case, such high reddening appears to be incompatible with the expected spectral slopes, suggesting an intrinsically higher Balmer decrement, which can arise from optical depth effects redistributing H $\unicode{x03B2}$ photons into H $\unicode{x03B1}$ and Pa $\unicode{x03B1}$ (Pottasch 1960; Netzer 1975).

4. Ancillary datasets

The unusually bright nature of J1144 raises the question of whether it has historically been fainter, contributing to its long-standing anonymity.

4.1 Optical data

On UT 1890 May 26, the 8-inch ‘Bache doublet’ telescope at ‘Mount Harvard’ near Chosica, Peru, obtained a 60-min exposure showing J1144. The photographic glass plate (b5269) has been digitised, and astrometrically and photometrically calibrated by the Digital Access to a Sky Century @ Harvard (DASCH) projectj (Laycock et al. 2008; Tang et al. 2013). The brightness of J1144 in the 1890 image is estimated to be $14.80\pm0.16$ mag (AB), calibrated to Pan-STARRS g-band using the Asteroid Terrestrial impact Last Alert System (ATLAS) All-Sky Stellar Reference Catalog (Tonry et al. 2018a). Similar plates available through DASCH from more recent epochs show J1144 with estimated g-band magnitudes between 13.9 and 15.1 mag (AB; omitting highly uncertain measurements or those close to the plate’s limiting depth). The heterogeneity of the data precludes a more detailed analysis, but the DASCH photographic archives indicate that the optical brightness of J1144 has not varied by more than a factor of ${\sim}2$ in the last 130 yr.

Considering data focused on the blue end of the optical spectrum, photographic glass plates taken by the UK Schmidt telescope at SSO on UT 1977 March 21, as part of the ESO/SERC Southern Sky Survey, measured $B_{J}=14.7\pm0.4$ mag (Vega), as cataloguedk by the SuperCOSMOS Sky Survey (Hambly et al. 2001a; Hambly, Irwin, & MacGillivray 2001b). The $B_{J}$ and SMSS g bandpasses are similar (e.g., see the $B_{J}$ throughput compared to the Sloan Digital Sky Survey $g_{\rm SDSS}$ in Richards et al. 2005) and the ( $B_{J}-g$ ) colours of 3673 quasars in the redshift range 0.6–1.0 from the 2dF and 6dF QSO Redshift Surveys (2QZ/6QZ; Croom et al. 2004) show a median of 0.10 mag with a scaled median absolute deviation (SMAD) of 0.38 mag. With an SMSS DR3 g-band magnitude of $14.534\pm0.014$ mag (AB), we conclude there has been no significant variation in the J1144 brightness at rest-frame ${\sim}2700$ Å compared to 45 yr ago.

On more recent timescales, the SMSS DR3 dataset incorporates images of J1144 acquired between 2015 February and 2018 June. Across that time window, each of the six SMSS filters shows a brightening of $\approx0.1$ mag, with 7–9 epochs per filter. Denser time sampling is available from the ATLAS telescopes (Tonry et al. 2018b; Smith et al. 2020), which have observed J1144 at a high rate (typically over 200 times per year) since December 2017.l Both the o (‘orange’; comprising 80% of the data) and c (‘cyan’) filters show a brightening of 0.2 mag relative to the earliest ATLAS epoch (2016 January), with a peak roughly in 2020 May. Together, these datasets provide a consistent picture of modest brightness variations over timescales of days to years.

To probe even shorter timescales, we examined the J1144 data available from two ${\sim}$ month-long visits by the Transiting Exoplanet Survey Satellite (TESS; Ricker et al. 2015), one beginning on UT 2019 March 26 (during the primary mission; Sector 10) and one beginning on 2021 April 02 (first extended mission; Sector 37). In the TESS Input Catalog (TIC; Paegert et al. 2022), J1144 has the designation TIC 61537875, but it was not selected for the Candidate Target List. As a result, photometry is only available from the Full Frame Image (FFI) dataset, which were acquired every 30 min in the primary mission and every 10 min in the first extended mission. The coarse spatial resolution of TESS results in J1144 being heavily blended with TIC 61537878, a star 30 arcsec away that is ${\sim}1$ mag brighter in TESS’s wide bandpass (600– $1000$ nm). As a result, precise photometry is difficult to obtain, but we note no significant fluctuations above 1% in the combined light curve of the quasar and star.

4.2 IR data

Turning to the IR, the NEOWISE 2022 Data Releasem (Mainzer et al. 2014) provides W1 and W2 photometry for J1144 at more than 250 epochs between UT 2014 January 09 and 2021 June 19. At rest-frame equivalents of 1.8 and 2.5 $\unicode{x03BC}$ m for W1 and W2, respectively, the WISE data is dominated by the hot dust near the quasar, rather than the quasar’s accretion disk that is probed at shorter wavelengths. In Figure 3, we show the WISE and ATLAS light curves, binned to 30-d median values.

Figure 3. ATLAS and WISE light curves for J1144 over the past 8 yr. Photometry was binned to 30-d median values for each bandpass. The ATLAS photometry is in AB magnitudes, while the the IR photometry is in Vega magnitudes and has been shifted vertically for convenience. Any increase in the optical brightness would take a decade to be reflected in the dust luminosity because of the large dust sublimation region around luminous quasars like J1144.

In contrast to the ATLAS photometry, J1144 exhibits a very slight fading in both IR bands over that time period, with an amplitude of ${\sim}0.05$ mag, comparable to the level of intraday photometric scatter. This uncorrelated behaviour can be understood in the context of the multi-year time lags for dust reverberation that would be expected from the large dust sublimation radius implied by the high luminosity of J1144. For the luminosity of J1144 ( $1.2 \times 10^{14}\,\mathrm{L}_{\odot}$ ), the predicted time lag for the IR response to optical variations would be 9.7 yr (Lyu, Rieke, & Smith 2019), slightly longer than the time span shown in Figure 3. Thus, the steadiness of the IR photometry since the start of WISE operations implies no significant and lengthy changes in the quasar luminosity over the past ${\sim}$ 20 yr.

As the recent increase in luminosity shown by the ATLAS light curves propagates into the dust surrounding the quasar, we could expect to see the IR similarly brighten within the next decade. However, the relative amplitude of IR variability in response to optical fluctuations is extremely broad, between factors of 0.1 and 10 for high-luminosity, high-time-lag sources (Lyu et al. 2019, cf. their Figure 14). Future monitoring by WISE and the Dynamic REd All-sky Monitoring Survey (DREAMS; Soon et al. 2020) may reveal the amplitude of the dust response in J1144.

4.3 X-ray data

In X-rays, there are no point-source counterparts to J1144 catalogued in the Second ROSAT all-sky X-ray Survey (2RXS; Boller et al. 2016). Cross-matching confirmed quasars from Milliquas, approximately 75% of the Gaia $G<16$ mag (Vega) quasars in the redshift range $z=0.7-0.9$ have 2RXS counterparts. For our extinction-corrected 2500 Å flux of $F_{\nu} = 3.8\times10^{-26}\,\mathrm{erg\,s}^{-1}\,\mathrm{cm}^{-2}\,\mathrm{Hz}^{-1}$ , we would expect a 2 keV X-ray flux of ${\sim}2\times10^{-30}\,\mathrm{erg\,s}^{-1}\,\mathrm{cm}^{-2}\,\mathrm{Hz}^{-1}$ (Bisogni et al. 2021), roughly a factor of 10 greater than the nominal 2RXS flux limit (Boller et al. 2016). Whether the non-detection is an indication of intrinsic X-ray weakness or absorption (local to the quasar or intervening) may be clarified by forthcoming data releases from SRG/eROSITA (the Spectrum-Roentgen-Gamma satellite’s extended ROentgen Survey with an Imaging Telescope Array; Predehl et al. 2021).

4.4 Radio data

The closest radio detection in DR1 of the Rapid ASKAP Continuum Surveyn (RACS; McConnell et al. 2020; Hale et al. 2021) is 48 arcsec from J1144, which at $z=0.83$ corresponds to more than 350 kpc projected distance. If we assume an association between J1144 and RACS-DR1 J114451.8-430920, then the flux density of $5.1\pm0.5$ mJy at 887.5 MHz implies a rest-frame 5 GHz luminosity density of $7.6\times10^{31}\,\mathrm{erg\,s}^{-1}\,\mathrm{Hz}^{-1}$ (for a spectral index of $\nu^{-0.6}$ ). For an optical (rest-frame 4400 Å) luminosity density of $1.4\times10^{32}\,\mathrm{erg\,s}^{-1}\,\mathrm{Hz}^{-1}$ (derived from the global power-law continuum shown in Figure 1), this gives an upper limit to the radio-loudness (using the definition of Kellermann et al. 1989) of $R < 0.54$ . Thus, even a putative associationo between the RACS DR1 source and J1144 leaves the quasar in the radio-quiet regime.

The absence of strong X-ray and radio emission, in conjunction with the low levels of UV-to-IR variability, make it unlikely for J1144 to have a relativistically beamed jet. Thus, we conclude that J1144 is not a blazar.

4.5 On the possibility of gravitational lensing

Given the high luminosity, it is natural to wonder if the source is gravitationally lensed (e.g., Fan et al. 2019). To check for a small-separation galaxy lens, we examine the corrected $B_{P}$ and $R_{P}$ flux excess, $C^{*}$ , from Gaia EDR3 (Riello et al. 2021), which makes a standardised comparison between the flux measured in the 0.35-arcsec-wide G-band photometric aperture with the integrated fluxes measured in the 3.5-arcsec-wide apertures for the $B_{P}$ and $R_{P}$ photometry (integrating along the wavelength dimension of the low-resolution spectra). With $C^{*} = 0.023$ , the variation in flux measured by the different extraction apertures for the Gaia photometry is within 2 $\sigma$ of the 0-value expected for point sources of the brightness of J1144 (where $\sigma = 0.012$ ). Amongst the small-separation gravitationally lensed quasarsp that are associated with a single Gaia EDR3 source, Q1208+101 has the smallest $C^{*}$ value at 0.168, a factor of 7 times larger than J1144. Similarly, the G-band variability proxy of Mowlavi et al. (2021) has a value of 0.015, suggesting that the Gaia small-aperture measurements across a range of scan directions have found peak-to-peak flux variations of ${\sim}0.05$ mag (cf. their Section 3).

While we conclude that there is no indication of gravitational lensing for J1144 in the existing Gaia data, the lack of microlensing-induced flux variations evident in the recent photometric sampling described above cannot fully exclude the existence of a lensing galaxy, as Mosquera & Kochanek (2011) found that roughly half of lensed quasars are likely to be in a ‘demagnified valley’ in any given 10-yr period. Thus, a high-spatial-resolution imaging study of J1144 would be of great interest.

5. Comparison with other bright quasars

As the single brightest quasar in the sky, 3C 273 is an important benchmark for luminous quasars, as its long observational history and low redshift have made it a forefront laboratory for exploring accretion processes (e.g., Courvoisier 1998; Gravity Collaboration et al. 2018). Moreover, the presence of the strong radio jet, with synchrotron emission that extends into the optical regime, has opened a window into the relatively rare class of radio-loud quasars (e.g., Bahcall et al. 1995; Jester et al. 2005; Uchiyama et al. 2006).

In Figure 4, we compare the rest-frame UV-to-IR spectral energy distribution (SED) of J1144 from recent data to the minimum and maximum luminosities of 3C 273 (including synchrotron flares), as observed over a 40-yr spanq (Türler et al. 1999; Soldi et al. 2008), as well as to SMSS J2157-3602, the most luminous known quasar in the Universe, at a redshift of $z=4.692$ (Wolf et al. 2018b; Onken et al. 2020). The J1144 SED is derived from a subset of the photometry presented in Table 4, namely, that of SMSS DR3 (u, v, g, r, i, z), 2MASS (J, H, K),r and AllWISE (W1, W2, W3, W4). The photometry in Figure 4 was corrected for Galactic extinction up to an observed-frame wavelength of 3 $\unicode{x03BC}$ m using the Fitzpatrick et al. (2019) extinction curve, a standard $R_{V}$ =3.1 Milky Way dust model, and the Schlegel et al. (1998) E(B-V) values of 0.123, 0.021, and 0.015 mag for J1144, 3C 273, and SMSS J2157, respectively, with the $\times0.86$ correction factor of Schlafly & Finkbeiner (2011). As with the spectroscopic calibration above, the bandpass details were retrieved from the SVO. The SMSS u- and v-band photometry were also corrected with the stellar-colour regression method of Huang et al. (2021), which makes them 0.088 and 0.069 mag fainter, respectively. (The corrections for g- and r-band are less than 0.01 mag and are therefore omitted.)

Figure 4. Rest-frame SED for J1144 (red circles) compared to the 40-yr range of luminosities of 3C 273 (grey shaded region) from Soldi et al. (2008), and to SMSS J2157 (blue stars), the most luminous known quasar. All three sources have been corrected for Galactic extinction. Single-epoch uncertainties for the J1144 photometry are smaller than the symbols. Three quasar templates from Lyu et al. (2017) are also shown: normal (solid), hot-dust-deficient (HDD; dotted), and warm-dust-deficient (WDD; dashed). The inset shows an expanded range in order to include the potential J1144 radio association from RACS DR1 as an upper limit (arrow) and to indicate the difference in long-wavelength slope from radio-loud quasars like 3C 273.

Across the entire observed UV/optical range, J1144 has an intrinsic luminosity that is roughly 8 times greater than the brightest observations of 3C 273, and only about 3 times less than the most luminous quasar known. Even with the occasional synchrotron flares elevating the peak luminosities of 3C 273 in the IR, the blazar has remained ${\sim}5\times$ less luminous than J1144. The inset in Figure 4 includes the maximum potential radio luminosity observed for J1144, on the assumption of the RACS DR1 detection being associated with the quasar, illustrating the dramatic difference in the long-wavelength SED of radio-quiet quasars compared to radio-loud sources such as 3C 273.

In Figure 4, we also show three SED templatess (Lyu, Rieke, & Shi 2017) exhibiting different mid-IR dust properties. The ‘Normal SED’ represents the typical broad-line quasar SED, while the hot-dust-deficient (HDD) and warm-dust-deficient (WDD) templates show reduced emission at shorter and longer mid-IR wavelengths, respectively. With the templates anchored to match the J1144 SED near 1 $\unicode{x03BC}$ m, the WISE photometry suggests that J1144 is an intermediate case and may be somewhat lacking in dust close to the quasar.

Figure 5. Bolometric luminosity of J1144 (large red point), 3C 273 (grey square), and SMSS J2157 (blue star), compared to sources from the SDSS DR14 quasar catalogue (DR14Q; black points; Rakshit et al. 2020) and Milliquas (MQ; green points), shown as a function of lookback time (bottom axis) and redshift (top axis). No known quasars are as luminous as J1144 in the last 9 Gyr, and J1144 is only a factor of 2 dimmer than the most luminous known quasar, SMSS J2157.

In addition, we note the similarity between the optical spectra of J1144 (Figures 1 and 2) and 3C 273 (Dietrich et al. 1999), with prominent Balmer lines and comparatively weak [Oiii] emission, suggesting a commonality in their central engines despite the differences in their radio properties. Compared to the sample of bright quasars analysed by BG92, J1144 is on the strong-Feii/weak-[Oiii] end of their ‘Eigenvector 1’ correlation, although the Feii equivalent widtht (EW) of ${\sim}40$ Å is typical of radio-quiet quasars. The ratio of Feii-to-H $\unicode{x03B2}$ EWs of 1.5 is at the high end of their distribution, but J1144 does not exhibit the enhanced blue H $\unicode{x03B2}$ asymmetry often seen for such sources.

In Figure 5, we compare the J1144 bolometric luminosity estimated in Section 3.1 to 3C 273, SMSS J2157, and a large number of quasars, drawn from either the SDSS DR14 quasar catalogue (DR14Q; Rakshit, Stalin, & Kotilainen 2020) or Milliquas, as a function of lookback time. In order to use a consistent bolometric correction, we estimate the continuum luminosities at 3000 Å for the literature sources. For 3C 273, we use the mean U-band flux from the 40-yr dataset of the ISDC. DR14Q values use the 3000 Å luminosity tabulated in the catalogue, or, at higher or lower redshifts, respectively, estimate $\log_{10}$ ( $\lambda\,L_{\lambda}(3000)$ ) from the 1350 Å luminosity as $4.887\,+\,0.89\,\log_{10}$ ( $\lambda\,L_{\lambda}(1350)$ ) or from the 5100 Å luminosity as $9.213\,+\,0.792\,\log_{10}$ ( $\lambda\,L_{\lambda}(5100)$ ), based on the best-fit relations from the DR14Q sources having both luminosities estimated. Sources with non-zero QUALITY_L3000 values were excluded. The varied literature sources compiled in Milliquas were cross-matched to the Gaia catalogue and an empirical scaling from the $R_{p}$ photometry to the 3000 Å luminosity as a function of redshift was applied.u The Milliquas sample has been restricted to sources with spectroscopic redshifts; cleaned of lensed sources, blazars, and a few spurious objects (including those with $>3\sigma$ parallax or proper motion estimates); and has omitted quasars from SDSS (to avoid duplication). Beyond $z=5.5$ , very few Milliquas sources have Gaia photometry.

As can be seen from Figure 5, J1144 is the most luminous known quasar out to $z=1.29$ , a lookback time of 8.7 Gyr, beyond which the quasar, HS 2154+2228 (Hagen, Engels, & Reimers 1999), is the first of a small sample of quasars found to be more luminous, up to the pinnacle of SMSS J2157. With an extinction-corrected i-band magnitude of 14.059 mag (AB), J1144 is nearly 1 full magnitude brighter than the SDSS cutoff at $i_{\mathrm{SDSS}}=15$ mag (AB) and nearly 1.5 mag brighter in $M_{i}(z=2)$ than any source actually found in the SDSS DR3 quasar luminosity function at redshifts below $z=0.9$ (Richards et al. 2006).

6. Discussion

The location of J1144 falls within a small gap in the GALEX All-Sky Imaging Survey (AIS; Martin et al. 2005), which explains why it did not appear in DR1 of the UVQS (Monroe et al. 2016). Utilising ${\sim}3000$ known quasars from Milliquas in the same redshift range as J1144, for which both GALEX and Gaia photometry exist, we use the $FUV-G$ and $NUV-G$ colours to predict J1144 to have (FUV, NUV) = (17, 16) mag (AB), with uncertainties of roughly 1 mag in each band. With $FUV=17$ mag (AB), J1144 would have been amongst the top 10% of the brightest discoveries by UVQS DR1, but at a redshift 0.2 higher than the rest.

The high luminosity of J1144 also implies a large size for its BLR. Extrapolating the radius–luminosity relation of Bentz et al. (2013) to the 5100 Å luminosity of J1144 suggests an H $\unicode{x03B2}$ -emitting size of ${\sim}1200$ light-days. With the additional time-dilation factor of ${\sim}2$ , a reverberation mapping campaign would be a long-term endeavour. However, the angular size of the BLR is expected to be in excess of 100 microarcsec. Since J1144 has a K-band magnitude of 11.9 mag (Vega), its BLR will be well within the reach of the upgraded GRAVITY+ instrumentv at ESO’s Very Large Telescopes. Thus, it may be possible to measure the Pa $\unicode{x03B2}$ dynamics in J1144, comparable to the Pa $\unicode{x03B1}$ measurement for 3C 273 (Gravity Collaboration et al. 2018).

Additional studies may make productive use of an exceptionally bright quasar like J1144 as a background source. For example, UV spectroscopy of J1144 may probe the Milky Way’s circumgalactic medium (Tumlinson, Peeples, & Werk 2017; Zheng et al. 2019; Bish et al. 2021).

Previous searches for quasars and other blue objects in the Southern hemisphere have usually not reached as close to the Galactic Plane as J1144, which lies at $b=+18.1^{\circ}$ . For example, the Edinburgh-Cape Blue Object Survey (Stobie et al. 1997; Kilkenny et al. 2016) was restricted to $|b|>30^{\circ}$ ; the Hamburg/ESO quasar survey (Wisotzki et al. 1996; Wisotzki et al. 2000) searched at $|b|>25^{\circ}$ ; and the Calán-Tololo Survey (Maza et al. 1988; Maza, Wischnjewsky, & Antezana 1996) observed to $|b|>20^{\circ}$ . Dedicated quasar searches closer to the Galactic Plane (e.g., Im et al. 2007; Fu et al. 2021) may produce samples of objects useful both in their own right and for studies of the gas and dust near the Galactic disk.

Moreover, the discovery power inherent in the recent generation of all-sky surveys like those of Gaia, WISE, and eROSITA motivate a fresh examination of what other bright quasars may have been missed in previous searches across the celestial sphere. A spectroscopic campaign underway at the ANU 2.3 m telescope has already identified ${\sim}80$ new, bright quasars (in addition to J1144), some with Galactic latitudes in excess of 60 deg. Thus, after 60 yr, it would appear we are finally approaching a complete census of bright quasars, with only the discovery of Changing Look Quasars (CLQ; e.g., LaMassa et al. 2015) from forthcoming surveys likely to add to the sample.