CLOSE COMPANIONS TO TWO HIGH-REDSHIFT QUASARS^*

J0256+0019 was confirmed as a z = 4.8 quasar using longslit spectroscopic observations obtained with the MMT Red Channel spectrograph on 2011 October 1. The Red Channel observations were performed with a 1''×180'' slit and the low dispersion 270 mm⁻¹ grating, delivering a resolution of R ∼ 640 over the range 5700 Å to 9700 Å. Three 10 m exposures were combined to produce the final spectrum. Data were processed in a standard fashion; details are given in McGreer et al. (2013).

We immediately noticed extended Lyα emission in the two-dimensional (2D) spectrum. Although the seeing during the spectroscopic observations was ∼08, the emission in the vicinity of the quasar's Lyα line extends to roughly 2''. The Lyα detection was purely serendipitous: only after the observations did we notice extended emission in the same location in the stacked i-band image from Stripe 82 (Figure 1). Although we applied no morphological criteria in selecting quasar candidates from Stripe 82, J0256+0019 would have likely been targeted anyway, as the i-band detection for the quasar has SExtractor CLASS_STAR = 0.95 (where 1 represents a stellar profile); the excess i-band emission is quite weak relative to the quasar.

Zoom In Zoom Out Reset image size

We obtained near-IR imaging in the J and H bands with the MMT Smithsonian Widefield Infrared Camera (SWIRC; Brown et al. 2008) on 2011 October 15. The seeing was 07 and conditions were variable and non-photometric. J0256+0019 was observed in the J and H bands using dithered integrations for a total exposure time of 20.7 m and 16.3 m, respectively. The images were shifted and stacked, and then registered to an astrometric solution obtained by matching UKIDSS stars within the field. Flux calibration was obtained from observations of faint UKIRT standard stars from Leggett et al. (2006) and checked against UKIDSS matches. The quasar is clearly detected in both images, with fluxes of J = 22.11 ± 0.14 and H = 21.43 ± 0.10 measured from 18 apertures (Figure 2). The companion object is not detected to 3σ limits of J = 22.8 and H = 23.2 (both AB).

Zoom In Zoom Out Reset image size

Additional imaging and spectroscopic observations were obtained with the LBT Multi-Object Double Spectrograph (MODS1) instrument on 2012 September 21. MODS1 is an optical imager/spectrograph mounted on the LBT 2×8.4 m telescope (Pogge et al. 2006). We obtained both direct imaging and grating spectroscopy of J0256+0019. During the observations conditions were non-photometric with mostly clear skies and good seeing, improving from 08 to 06.

2.2.1. Imaging

In acquisition mode the MODS1 red CCD employs a 1024×1024 array with a pixel scale of 0123 and a FOV of 21. Immediately following a series of longslit spectroscopic observations (see below), a single 600 s exposure was obtained with the SDSS i filter. Guiding was continuous throughout both observations.

The i-band image was reduced in a standard fashion. A bias correction was applied from a median of bias images. Sky flats were obtained at twilight and used to provide the flat field. The image was masked of regions affected by scattered light from a nearby bright star and from obstruction by the guide probe camera, leaving roughly 2/3 of the field usable. Astrometric registration and flux calibration were obtained by matching stellar objects from the Stripe 82 coadded imaging. Object detection was performed with SExtractor (Bertin & Arnouts 1996) using standard parameters.

We measure a separation between the quasar and the companion object of 18 from the MODS1 i-band image, corresponding to a projected physical separation of 12.1 kpc for our adopted cosmology. The quasar has i = 22.08 ± 0.02, somewhat fainter than measured from the Stripe 82 coadded imaging (i = 21.73 ± 0.02). This is likely due to blending, although variability could also contribute to the flux difference. The two objects are well resolved in the MODS1 image (Figure 3). According to the SExtractor star/galaxy separation, the quasar is consistent with a point source (CLASS_STAR = 0.97) while the companion is extended (CLASS_STAR = 0.17). The companion has i = 23.60 ± 0.07 and measured size of FWHM = 10 (compared to 057 for the quasar). The companion is not detected in the Stripe 82 coadded z-band image to a 3σ limit of z = 23.5; so i − z ≲ 0.1 for this object.

Zoom In Zoom Out Reset image size

2.2.2. Spectroscopy

In grating spectroscopic mode the MODS1 Red channel utilizes a 8 k × 3 k array. We used the single grating mode with the G670L grating and a 1'' longslit, providing a spectral resolution of R ∼ 1400 and wavelength coverage from 5800 Å to 1 μm at a dispersion of 0.8 Å pixel⁻¹. Three exposures of 1200s each were obtained.

The position angle between the quasar and the companion is 312 from the MODS1 imaging. We estimated a P.A. of 40° from the Stripe 82 coadded imaging and placed the MODS longslit at this angle for the spectroscopic observations. Although the slit angle was offset by ∼9° from the true P.A. between the two objects, the quasar was also slightly off-center relative to the slit, while the companion galaxy was relatively well-centered.

The images were first processed with the MODS CCD reduction utilities (modsTools v0.3) to obtain bias and flat-field corrections. Transformations between image pixels and wavelength were determined from polynomial fits to arc calibration images. A 2D sky model was fit to each image using b-splines (Kelson 2003) and then subtracted. Cosmic rays were identified and masked during the construction of the sky model. The individual exposures were combined with inverse variance weighting to produce the final 2D spectrum.

Figure 4 displays a section of the sky-subtracted and combined 2D spectrum obtained with MODS1. Two components are clearly visible at ∼7045 Å and are well separated. The lower component is the quasar J0256+0019 included in our Stripe 82 survey. The only obvious feature in the galaxy spectrum (upper component) is prominent Lyα emission.

Zoom In Zoom Out Reset image size

One-dimensional (1D) spectra of both components were obtained using optimal extraction (Horne 1986) while allowing for multiple, blended object profiles (Hynes 2002). We used the direct image obtained immediately following the spectroscopic observations to obtain the profile weights for extraction. We fit a Moffat profile to the quasar and an exponential profile to the galaxy, then collapsed the profile fits along the slit axis. The results are similar to those obtained with a simple boxcar extraction, but with lower noise as expected.

Flux calibration is provided by observations of the spectrophotometric standard star GD71. This calibration is affected by the non-photometric conditions and slit losses. We obtain rough corrections for the slit losses using the direct imaging obtained immediately after the spectroscopic observations, by calculating the ratio of the synthetic fluxes obtained from integrating the spectra to the total fluxes measured from the i-band imaging. We adopt these corrections (0.65 for the quasar and 0.60 for the galaxy), but note that systematic effects may introduce additional (unaccounted for) uncertainty into these values.

Figure 5 displays the extracted spectra of both components, with slit loss corrections applied. The quasar spectrum has typical emission and absorption features for a high-redshift quasar. The galaxy spectrum is dominated by Lyα emission. Although it is not immediately apparent from Figure 5, we also detect continuum flux from the companion galaxy. Figure 6 presents a collapsed version of the 2D spectrum, binning pixels to enhance the S/N until the weak continuum of the companion is apparent (${\rm S/N} \sim 0.5\ {\rm pix{\rm el}}^{-1}$). The non-zero continuum level is also evident in the extracted 1D spectrum.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Based on the MMT spectrum alone we considered the possibility that the companion represents a secondary image of a single source quasar due to gravitational lensing, and is visible only in Lyα emission as this was the strongest feature in the quasar spectrum. However, the quasar's C iv line is sufficiently strong that it should have been detected from the companion galaxy if its spectrum were simply scaled from the quasar's spectrum. This conclusion is even stronger based on the higher S/N LBT observations: the flux ratios of the two spectra at the peak of the Lyα emission are ∼2:1, compared to a ratio at the C iv peak of ≳10:1 and a continuum ratio of ∼2.5:1. Furthermore, the companion galaxy is resolved in the LBT imaging, and no candidate lens galaxy is apparent. Finally, the detailed Lyα profiles show clear differences (see Figure 7). Thus we reject the lensing hypothesis for this object.

Zoom In Zoom Out Reset image size

Our main results are the detection of strong Lyα emission as well as rest-frame UV continuum from the companion galaxy. By fitting a power law to the spectrum redward of Lyα, we obtain f₁₅₀₀ = (4.7 ± 0.2) × 10⁻¹⁹erg s⁻¹ cm⁻²Å⁻¹ (after applying the slit loss correction). This flux density corresponds to a continuum luminosity of M₁₅₀₀ = −22.7, or ≈5.5L* at this redshift (Bouwens et al. 2014). Although this value is high compared to local non-active galaxies, objects with even higher UV luminosities have been found at z ∼ 3 (Bian et al. 2012) and at z ∼ 7 (Bowler et al. 2014).

We measure a power-law slope of β_λ = −1.1 ± 0.3 ($f_\lambda \propto \lambda ^{\beta _\lambda }$) by fitting the UV continuum. This value is quite uncertain given the low S/N and the limited wavelength coverage. We fit the power-law continuum of the quasar and obtain a slope α_ν = −0.5 ($f_\nu \propto \nu ^{\alpha _\nu }$)⁹. This agrees well with typical UV slopes measured for quasars (e.g., Vanden Berk et al. 2001) and suggests that the relative flux calibration is reliable. A slope β_λ = −1.1 is fairly red, but consistent with measurements of galaxies at z ∼ 5, especially given that the spectral slope becomes redder at higher luminosities (Bouwens et al. 2011).

We measure a redshift of z = 4.789 ± 0.005 from a single Gaussian fit to the O i emission line of the quasar. This is a low ionization transition which normally has smaller systematic redshift offsets than lines such as C iv (e.g., Shen et al. 2011); however, we adopt a conservatively large uncertainty on the redshift to account for a possible shift of the O i peak relative to the systemic velocity. We do not attempt to define a separate redshift for the companion galaxy, as the only detected line is Lyα, from which it is notoriously difficult to obtain a reliable redshift; however, the Lyα redshifts for the quasar and galaxy agree to <200 km s⁻¹, suggesting that the true separation of the two objects is not much larger than the projected separation of 12 kpc.

Figure 7 shows the Lyα profiles of both the quasar and companion galaxy. The companion galaxy has strong Lyα emission, with f_Lyα = 2.4 × 10⁻¹⁶erg s⁻¹cm⁻²Å⁻¹ and EW₀ ∼ 100 Å. The equivalent width is not atypical when compared to high-redshift Lyα emitters (LAEs) selected from narrow-band surveys (e.g., Kashikawa et al. 2011); however, the line strength tends to decrease with continuum luminosity in these objects; thus the galaxy stands out both for its bright continuum and the equivalent width of the Lyα emission line. The line luminosity is ∼6 × 10⁴³ erg s⁻¹; if the Lyα emission were entirely due to star formation, this would correspond to SFR(Lyα) ∼60 M_☉ yr⁻¹ using the Kennicutt (1998) relation for Hα and assuming Case B recombination. This is likely to be an underestimate due to intergalactic medium attenuation and dust extinction, both of which would dampen the observed Lyα line flux.

On the other hand, the Lyα line profile is unusual when compared to LAEs. From Figure 7 it is evident that the red wing of the line profile extends over a range |Δv| ∼ 1000  km s⁻¹. A fit to the line with a single Gaussian results in FWHM ∼700  km s⁻¹; a fit with a double Gaussian has a broad wing with FWHM ∼1000  km s⁻¹. Ohyama et al. (2004) observed a similar feature in the spectrum of one of the Lyα companions to BR 1202-0725. They attributed the broad linewidth to outflows of neutral hydrogen driven by star formation. Given the high estimated star formation rate (SFR) from Lyα (roughly four times that of BR 1202-0725 Lyα-1) this scenario is plausible.

In Table 1 we provide values obtained from Gaussian fitting of the emission lines in both spectra. For the quasar Lyα line we restrict the fitting to the red wing of the line in order to mask the strong absorption features in the blue wing. Because of the difficulty in fitting the Lyα profile we do not provide formal uncertainties; similarly, the C iv line is in a region with substantial night sky line residuals and is thus difficult to fit reliably.

The Lyα emission from the quasar is unusually narrow. We measure a FWHM of ∼750 km s⁻¹ from a single Gaussian fit to the line. This is the narrowest line of the 35 z ∼ 5 quasars from our MMT observations from which we can measure the Lyα line (the median is ∼2000 km s⁻¹ with a standard deviation of ∼500 km s⁻¹; only one other object has FWHM <1000 km s⁻¹). The Lyα line alone would qualify this object as a Type II quasar candidate according to the criteria of Alexandroff et al. (2013). On the other hand, the C iv line is broader (FWHM ∼ 2000 km s⁻¹) and the UV continuum is strong (M₁₄₅₀ ≈ −24), indicating at best a moderate amount of nuclear extinction. It is evident from Figure 7 that the Lyα line has multiple strong absorption features; this likely explains the narrowness of the line. We re-fit the line using only the red wing and obtain FWHM ∼1400 km s⁻¹. Although we do not attempt to model the Lyα absorption in detail, the implication is that J0256+0019 has a high covering fraction of neutral gas along the line of sight when compared to quasars at a similar redshift. A virial estimate for the BH mass from the C iv line width using the relation of Vestergaard & Peterson (2006) gives a mass of ≈1.2 × 10⁸ M_☉ and an Eddington ratio of λ ≈ 0.9, indicating that the quasar is in a phase of rapid growth.

No rest-frame UV emission lines other than Lyα are detected from the companion galaxy; thus it is unlikely to be an active galactic nucleus (AGN). Figure 8 shows the C iv emission region of both the quasar and the companion. We obtain an upper limit on the C iv emission from the companion galaxy by using the redshift obtained from the quasar O i line to set the wavelength and a fiducial linewidth of σ_v = 200 km s⁻¹. This limit is EW₀ < 6 Å (1σ), whereas nearly all of the Type II quasar candidates in Alexandroff et al. (2013) have EW₀(C iv)  > 10 Å. Additionally, no N v emission is detected from the galaxy, to a limit of EW₀(N v) < 0.2 Å. These properties are similar to the Lyα-1 companion galaxy of BR 1202-0725, which Williams et al. (2014) found to have ratios of C iv flux to Lyα flux  <  0.033 and N iv/Lyα  <  0.019, compared to >0.2 and ≳ 0.1, respectively, for typical Seyfert galaxies and obscured quasars. These ratios are ≈0.05 and <0.001 for the J0256+0019 companion, suggesting that AGN photoionization plays at best a minor role.

Zoom In Zoom Out Reset image size

We have been performing a systematic search for evidence of gravitational lensing among z ∼ 6 quasars through an HST SNAP program (#12184, PI: X. Fan). This program includes nearly all of the known z ∼ 6 quasars and utilizes WFC3/IR F105W imaging. Full results of the SNAP study will be presented in a forthcoming work (I. D. McGreer et al., in preparation). During the SNAP program J0050+3445 was observed with two 180 s exposures. It was immediately apparent that the resulting image was inconsistent with a single point-spread function (PSF) detection, as there was excess flux detected at ∼1'' from the quasar position. However, the shallow, single-band SNAP observations did not constrain the nature of the excess emission, which could be attributed to (1) a foreground interloper, (2) a high-redshift galaxy, or (3) a secondary lensed image of the quasar.

3.1.1. HST Cycle 19 Imaging

To further explore the nature of the excess flux in the SNAP imaging we obtained additional HST imaging of J0050+3445 in Cycle 19 (#12493, PI: I. D. McGreer). These observations consisted of two components: (1) three orbits of ACS/WFC imaging with the F775 bandpass, and (2) two orbits of WFC3/IR imaging with the F105W bandpass. This filter combination isolates objects at the redshift of the quasar by taking advantage of the strong spectral break feature introduced by the nearly saturated Lyα forest absorption at z = 6.25. A single broadband color can provide strong evidence that a detected object is at high redshift if the filters straddle the Lyman break. Figure 9 displays the filter choices for the HST observations with spectral energy distributions (SEDs) of a high-redshift quasar and galaxy for reference. Neither of our filters includes Lyα emission at the redshift of the quasar.

Zoom In Zoom Out Reset image size

The ACS/WFC imaging occurred on 2012 June 13. The three orbits were divided into two exposures, the first with four dither positions and the second with two dither positions. Images were processed using AstroDrizzle. First, the standard pipeline-reduced images were combined into two drizzled images, which were then aligned with the tweakshifts routine. After propagating the calculated offsets to the input images, they were combined into a single drizzled image. The full integration time in the final mosaic is 8072 s. We used pixfrac = 0.8 and a scale of 003 for the output pixel grid (Koekemoer et al. 2011).

The WFC3/IR imaging occurred on 2012 June 8. A similar dither pattern was employed. Although cosmic rays in individual images can be rejected by the continuous sampling of the detector array, we found some residual hot pixels in the individual images. We thus used tweakreg and AstroDrizzle in a single iteration with cosmic ray rejection enabled to produce a final drizzled image combining all six individual images, with a total integration time of 5018 s. We used pixfrac = 0.8 and a scale of 006 for the final pixel grid.

We found small (<1'') offsets between the final ACS and WFC3 mosaics when compared to SDSS imaging over the same area. We thus aligned the ACS image to the SDSS DR9 (Ahn et al. 2012) i-band image of the field using tweakshifts, and then aligned the WFC3 image to the corrected ACS image. The final astrometric accuracy is <01 when compared to SDSS.

Figure 10 presents the HST imaging obtained for J0050+3445. Within the central ∼1'' the ACS image contains only a single faint detection corresponding to the z = 6.25 quasar, while the WFC3 image includes both the bright quasar PSF component and an additional extended component (a galaxy), consistent with the results from the shallow SNAP imaging.

Zoom In Zoom Out Reset image size

We first construct PSF models in order to decompose the images. For both the ACS and WFC3 images we generate empirical PSF models from stars within the field using IRAF DAOPHOT tasks. Analysis of the ACS image is somewhat insensitive to the accuracy of the PSF model, as the quasar is relatively weak and the galaxy lies well beyond the extent of the quasar's light profile. On the other hand, the quasar is much brighter in the WFC3 image, and positive flux from the wings of the PSF extends to the position of the galaxy. We selected eight isolated stars with high S/N detections to construct the WFC3 PSF model. We also experimented with TinyTim PSF models, but found they were not an improvement over the empirical PSF (cf. Mechtley et al. 2012).

We next use GALFIT to model the quasar and galaxy in the HST images. The ACS image is fit with a single PSF component to represent the quasar. The model for the WFC3 image includes both a PSF component for the quasar and an exponential disk model for the galaxy. The exponential disk profile was selected empirically; we fix the axis ratio and position angle to values obtained from visually matching the galaxy profile in order to reduce the fit degeneracies. Finally, we find that the PSF subtraction is improved by masking the core of the quasar PSF during the fit. We mask the central 3×3 pixels (018). The results of the GALFIT fitting are given in Table 2.

Table 2. Properties of J0050+3445

	QSO	Gal
Δα	0.00	+045
Δδ	0.00	−074
i775	24.032 ± 0.008	>26.9
Y105	20.325 ± 0.003	25.06 ± 0.23
rs	⋅⋅⋅	009 ± 003
a/b	⋅⋅⋅	0.4
P.A.	⋅⋅⋅	635
M1500	−26.64a	−21.76
SFR(UV)	⋅⋅⋅	27 M☉ yr−1

Notes. Measured quantities are from GALFIT. The axis ratio (a/b) and position angle (P.A.) were fixed during fitting. P.A. is measured east of north. ^aFrom Willott et al. (2010b).

Download table as:  ASCIITypeset image

The neighboring galaxy is not formally detected in the ACS image. We obtain an upper limit for its flux by first subtracting the PSF component and then fitting the residual image with a model for the galaxy. The parameters for the galaxy model are fixed to the values obtained from fitting the WFC3 image. We then increase the flux in the model galaxy until the fit results in Δ(χ²) = 1 compared to the PSF-only fit. We adopt this value as the upper limit for the i₇₇₅ flux of the galaxy given in Table 2. The limit obtained from this method, i₇₇₅ > 26.9, is bright compared to the average depth of the ACS image. This results from a slight positive fluctuation at the position of the galaxy in the ACS image (∼  + 1σ pixel⁻¹ for several neighboring pixels). However, this fluctuation appears to be associated with cosmic ray hits that appear in two out of the six ACS frames. We insert fake galaxies at random locations around the quasar position at the same radius of the neighboring galaxy and then repeat the process for deriving an upper limit. We find that in general the limiting flux is i₇₇₅ > 28. Assuming that the positive flux is simply a noise fluctuation, the lower limit on the color is i₇₇₅ − Y₁₀₅ ≳ 3 (1σ), consistent with a z = 6.25 galaxy. However, we adopt the more conservative limit obtained at the galaxy's position in the image, yielding a color i₇₇₅ − Y₁₀₅ > 1.8.

The original goal of the SNAP program was to search for gravitational lenses, and one of the primary goals of the subsequent imaging was to test the lensing hypothesis for J0050+3445. As one of most luminous quasars in the CFHQS, with M₁₄₅₀ = −26.6, and given the excess emission detected in the SNAP imaging, J0050+3445 was a prime candidate for a gravitationally lensed quasar. However, the Cycle 19 HST imaging rules out the lensing hypothesis, at least at ≳ 01 scales, and thus suggests the measured luminosity is intrinsic. First, the secondary component is extended. While this could be a lens galaxy, no additional lensed quasar images are detected. The limit obtained from the ACS imaging is even more stringent, with higher resolution and easier image decomposition. The quasar is detected at ∼200σ, and the detection limit for point sources is i₇₇₅ ∼ 29, so flux ratios ≲ 60: 1 can be excluded at ≳ 005 separations. These constraints effectively rule out all of the parameter space expected for strong lensing of high-redshift quasars (Turner et al. 1984; Comerford et al. 2002; Richards et al. 2004); even the inclusion of ellipticity or shear to the lens model will generally result in strong lensing with multiple images that should be apparent with HST's resolution and depth (Keeton et al. 2005).

The neighboring galaxy is also unlikely to be a foreground interloper or even a high-redshift galaxy unassociated with the quasar. The non-detection in the ACS imaging indicates an extremely red i−Y color, akin to Lyman break galaxies (LBGs) at z > 6. To examine the significance of the i−Y color, we compare to the CANDELS GOODS-South compilation of Guo et al. (2013), which includes ACS-i₇₇₅ imaging from GOODS (Giavalisco et al. 2004) and WFC3-Y₁₀₅ imaging from CANDELS (Grogin et al. 2011). We examine the distribution of i₇₇₅ − Y₁₀₅ colors for objects within the GOODS-S area with coverage in both filters, amounting to an area of ∼70 arcmin². We make conservative cuts of Y₁₀₅ < 25 and i₇₇₅ − Y₁₀₅ > 2.0, finding a density of ∼1 arcmin⁻². At this density, the probability of finding a galaxy similar to the J0050+3445 companion within 1'' of the quasar by chance is ∼8 × 10⁻⁴. This is strong evidence that the neighboring galaxy is indeed associated with the quasar. The projected separation between the two objects is 5.0 kpc if they are at the same redshift.

The Lyα line falls outside of both the F775W and F105W bandpasses (Figure 9). The detected emission is most likely dominated by UV continuum, while the presence of strong Lyα emission—as in the case of J0256+0019—is unconstrained by our observations. The galaxy is resolved in the WFC3 image, with a fitted disk scale length of 009, or 0.5 kpc at z = 6.25. The derived UV luminosity is M₁₅₀₀ = −21.8, which is ≈5L* at this redshift (e.g., McLure et al. 2013; Schenker et al. 2013), and is comparable to the brightest galaxies found in ground-based surveys at this redshift (Bowler et al. 2012).

A possible explanation for the high UV luminosity of the J0050+3445 companion galaxy is that it is powered by an AGN. At lower redshift, the fraction of galaxies hosting an AGN increases steeply with the host luminosity (e.g., Juneau et al. 2013). In fact, the derived UV luminosity of the J0050+3445 companion is comparable to that of Seyfert galaxies. The possibility of an AGN is intriguing, as it would imply two massive BHs at z = 6.25 separated by <10 kpc and potentially merging in a brief timescale. However, the currently available data do not provide any indicators to assess the presence of a low-luminosity AGN in the companion galaxy, other than to note that its morphology does not appear to be dominated by a central point source. Obtaining ground-based spectroscopy of the companion would be challenging, given its small separation from the nearby luminous quasar and its faintness.

Our two-band imaging was designed to provide information on the excess flux detected <1'' from the quasar, but also allows color selection of neighboring galaxies over the entire field. We thus search for associated galaxies up to ∼2' from the quasar. The total area of overlap between the ACS and WFC3 imaging is 4.9 arcmin². We construct matched catalogs by first aligning the ACS image to the WFC3 image with AstroDrizzle  and then executing SExtractor in dual-image mode with the WFC3 image for object detection.

We use simulations of the Lyα forest (see McGreer et al. 2013) and a simple power-law model for the UV continuum from galaxies with a range −2.5 < β_λ < −1 to estimate the range of galaxy colors expected at z = 6.25¹⁰. From this approach the typical color is i₇₇₅ − Y₁₀₅ ≈ 3.1, with i₇₇₅ − Y₁₀₅ > 2.8 for all simulated galaxies. The quasar has i₇₇₅ − Y₁₀₅ = 3.7. We estimate a 3σ detection limit in the ACS image of i₇₇₅ = 28.2 in a 024 aperture, though the same limit is somewhat weaker using the elliptical apertures defined for the WFC3 detections (i.e., SExtractor MAG_AUTO), for which we obtain i₇₇₅ ≈ 27.8. We thus survey for associated galaxies by searching for i₇₇₅-dropouts with Y₁₀₅ < 25.0; more specifically, we require i₇₇₅ − Y₁₀₅ > 2.8 to the detection limit of i₇₇₅ ≈ 28.

We identify four candidates from the SExtractor catalogs with these criteria (excluding the J0050+3445 quasar and companion galaxy). Of those objects, three are near the selection limit (Y₁₀₅ ≈ 25) and visual inspection of the ACS images shows they are weakly detected in the optical band. We do not consider them to be good candidates for z ∼ 6.25 galaxies. The final object has Y₁₀₅ = 24.24 ± 0.05 and is also weakly detected in the ACS image. Using a circular 1'' diameter aperture instead of the MAG_AUTO photometry, which roughly matches the size of the object in the WFC3 image, we obtain i₇₇₅ − Y₁₀₅ ≈ 1.9. We also do not consider this object to be a good candidate for an associated galaxy. From visual inspection we identify one object 115 from the quasar that appears to be a true optical dropout; however, with Y₁₀₅ = 25.4 it is too faint for the color to be conclusive (i₇₇₅ − Y₁₀₅ ≳ 2.6).

We also broaden our search to all optical dropouts to a depth of Y₁₀₅ = 27. The ACS non-detections do not provide strong constraints on the color for these objects. However, we can compare the number counts of the faint WFC3 detections to those from the GOODS-S catalogs from CANDELS described in Guo et al. (2013). We restrict the GOODS-S area to regions covered by both ACS/F775W and WFC3/F105W with integration times in each band greater than the integration times in our observations. We then consider any GOODS-S object with i₇₇₅ > 28 a "dropout," as it would be below the detection limit of our ACS imaging. We find no excess of such objects in our imaging after comparing to the number density obtained from the GOODS-S data (see Figure 11).

Zoom In Zoom Out Reset image size

The fact that we see no excess of associated galaxies on large scales around the quasar J0050+3445 is consistent with previous HST results that did not find overdensities to be ubiquitous among high-redshift quasars (e.g., Kim et al. 2009). However, using F105W for detection restricts us to objects with bright rest-frame UV continuum; we are not sensitive to obscured sources or weak continuum sources with strong Lyα emission as the line falls outside our bandpass. On the other hand, a depth of Y₁₀₅ = 27 corresponds to ≈L* at this redshift (McLure et al. 2013; Schenker et al. 2013); thus a strong overdensity of massive, star-forming galaxies should have been detectable with our observations.

The spectra of both quasars include strong Lyα absorption features near the quasar redshift (Figures 7 and 12). Such features are not uncommon in high-redshift quasar spectra, and may be indicative of a large reservoir of cold neutral gas surrounding the quasar. Further evidence for a high covering fraction of neutral gas in J0256+0019 comes from its relatively weak and narrow Lyα line. Such halos may give rise to fluorescent Lyα emission (Haiman & Rees 2001) powered by the quasar's ionizing continuum. Could the companions be dense clouds reprocessing the ionizing flux from the quasars, rather than self-luminous systems?

Zoom In Zoom Out Reset image size

We consider the contribution of fluorescent Lyα emission to the observed spectrum of the J0256+0019 companion galaxy by determining whether the incident ionizing flux from the quasar onto the galaxy is sufficient to power its strong Lyα flux. We adopt the projected separation between the two objects and a radius of 3 kpc for the companion (roughly its measured extent from the LBT image). We further assume that the quasar UV emission is isotropic and has a typical UV power-law continuum (Shull et al. 2012). The incident ionizing photon flux on the galaxy is then ∼4 × 10⁵⁴ erg s⁻¹. If every photon resulted in a Lyα photon, the resulting Lyα luminosity would be ∼6 × 10⁴³ erg s⁻¹, which is roughly equal to the measured luminosity. In other words, the observed strong Lyα emission could be due entirely to fluorescence if all of the quasar's ionizing photons incident on the galaxy are re-emitted as Lyα photons by the galaxy, whereas only ∼60% of ionizing photons should result in Lyα photons even for a single optically thick cloud (Gould & Weinberg 1996).

The J0256+0019 companion galaxy also has strong continuum emission, with νL_ν ≈ 10⁴⁵ erg s⁻¹ at rest-frame 1500 Å, so its spectrum cannot be purely fluorescent. Similarly, for J0050+3445 the observed emission arises only from continuum redward of Lyα. The lack of other emission lines in the spectrum of the J0256+0019 companion further demonstrates that neutral gas within its ISM is not being ionized by the nearby quasar (e.g., Gunn 1971; Filippenko 1985; see also the case of Hanny's Voorwerp, a cloud of gas illuminated by a (now dormant) quasar that has strong high ionization lines (Keel et al. 2012)). However, it is intriguing to consider that, at least in the case of J0256+0019, some of the strong Lyα emission may arise from fluorescence in a neutral halo of metal-poor gas surrounding the galaxy. This interpretation is consistent with the larger SFR inferred from the Lyα emission than the UV continuum (about a factor of four, see Table 1).

The companion galaxies may themselves be AGNs. Indeed, in merger-driven models for high-redshift quasar triggering (e.g., Li et al. 2007) it is expected that the progenitor galaxies each carry their own massive BH, although it may not be likely for both to be active at the same time. We argued in Section 2.3 that the lack of AGN lines in J0256+0019 argues against an AGN in this object. The constraints on AGN activity in the J0050+3445 companion are weaker, although we noted in Section 3.3 that its observed emission is not nuclear-dominated.

While we cannot conclusively rule out AGN activity in either system, the current observations are consistent with both galaxies being powered by internal star formation. We next discuss the implications of finding two unusually bright galaxies in the vicinity of high-redshift quasars.

The presence of bright galaxies within ∼10 kpc of two luminous, high-redshift quasars is highly suggestive of ongoing major mergers in both systems. Far-IR measurements of C ii line widths of z ∼ 6 quasars have shown that their dynamical masses are roughly ∼10^10–11 M_☉ (Wang et al. 2013). C ii dynamical masses may underestimate the full mass of the stellar bulge if the line emission is mainly concentrated in the center of the host galaxy; however, CO observations further indicate masses >10¹⁰M_☉ in molecular gas alone (Wang et al. 2010). This result is consistent with an expectation that the hosts of high-redshift quasars are massive, with the archetypal case being J1148+5251 at z = 6.4 (Walter et al. 2003).

Next, we consider the galaxy companions. According to the z = 5.7 Lyα luminosity function from Kashikawa et al. (2011), the J0256+0019 companion galaxy is at ∼6 L*, and thus on the steeply falling bright end of the luminosity function. It can be compared to Himiko, an extremely bright LAE discovered in the Subaru Deep Field (Ouchi et al. 2009). The J0256+0019 companion galaxy has many features in common with Himiko: it is resolved in ground-based imaging, and has a comparable Lyα luminosity (if not somewhat greater), equivalent width, and UV continuum luminosity. Using multiwavelength SED fitting, Ouchi et al. (2009) inferred a stellar mass of ∼4 × 10¹⁰ M_☉ for Himiko.

We can compare J0050+3445 to field LBGs selected at a similar redshift. Curtis-Lake et al. (2013) examined luminous (L > 1.2 L*) galaxies with spectroscopic redshifts 5.5 < z < 6.5 and found that the stellar masses lie in the range 10⁹ < M_* < 10¹⁰ M_☉, with typical stellar population ages of 50–200 Myr. At ≈5L*, the companion to J0050+3445 is comparable to the most extreme objects in the Curtis-Lake et al. (2013) sample, and thus likely lies at the higher mass end of this distribution. Similarly, the stellar mass-UV luminosity relation of Stark et al. (2009) derived at z ∼ 4 would predict a mass of M_* ∼ 10¹⁰ M_☉ at this luminosity; this relation shows little or no evolution to higher redshift (e.g., Curtis-Lake et al. 2013).

Thus, comparison to results from high-redshift surveys implies that both the quasar host galaxies and the companion galaxies have masses >10¹⁰M_☉. The small projected separations between the quasars and companion galaxies (≲ 10 kpc) in both cases place the quasar host and companion within the virial radius of a 10¹¹–10¹² M_☉ halo (the mass range of a halo likely to host a luminous quasar at this redshift), such that their eventual coalescence seems inevitable. A likely conclusion, then, is that in both cases we are witnessing extreme major merger events, potentially fueling luminous quasar and powerful starburst activity simultaneously.

Both of the systems discussed here were discovered serendipitously and it is not straightforward to draw conclusions about how common or rare they may be. However, we can ask whether similar systems would have been detected in the course of our surveys.

J0256+0019 was discovered in slit spectroscopic observations. We observed 55 quasars at z ∼ 5 with the 180'' longslit on MMT Red Channel during the course of our Stripe 82 survey (McGreer et al. 2013). Roughly half were observed with a 1'' slit, the other half with a 15 slit, so we use an average slit width of 125. If we consider neighbors within 20 kpc (∼3''), the fraction of area within that radius of the quasar subtended by the slit is ∼13% (ignoring slit losses for miscentered objects). Thus we would have detected 1 in ∼7.5 galaxy neighbors in our quasar survey. Experimenting with a single exposure of J0256+0019, if the Lyα emission were a factor of ∼2 fainter, it would have still been noticeable during the data reduction. From this crude analysis we conclude that less than 1 in ∼55/7.5 ≈ 7 quasars in our z = 5 survey have a companion with a Lyα flux within a factor of two of the J0256+0019 companion.

It is somewhat easier to draw statistical conclusions from the J0050+3445 observations. A total of 29 z ∼ 6 quasars were observed during the HST SNAP survey. Brighter quasars (z_AB < 20.5) had a total exposure time of 300 s, while the fainter objects were exposed for 1200 s. J0050+3445 was in the bright sample and its companion galaxy was detected near the limit of the imaging; the faint sample could reach galaxies ∼1.6 times fainter. Thus we could have detected companion galaxies to a limit of ≈5L* for the 23 bright objects observed, and ≈2L* for the 6 faint objects. One other bright quasar has excess emission in the SNAP imaging and is scheduled for further observations in Cycle 21; as part of the search for gravitational lenses we can rule out any other companions at separations <3''. Thus the incidence of UV-bright galaxies associated with quasars at z ∼ 6 is ≲ 2/29 for ≳ 5L* galaxies and <1/6 for 2 ≲ L ≲ 5L* galaxies.

At least within the context of our observations, bright companions to high-redshift quasars are uncommon. We conclude with some final thoughts related to the incidence of such systems.