Detections of [O iii] 88 μm in two quasars in the reionization epoch

ABSTRACT

We report on detections with the Atacama Large Millimeter/sub-millimeter Array of the far-infrared [O iii] 88 μm line and the underlying dust continuum in two quasars in the reionization epoch, J205406.48−000514.8 (hereafter J2054−0005) at z = 6.0391 ± 0.0002 and J231038.88+185519.7 (hereafter J2310+1855) at z = 6.0035 ± 0.0007. The [O iii] luminosities of J2054−0005 and J2310+1855 are L_{[O iii]} = (6.8 ± 0.6) × 10⁹ and |$(2.4 \pm 0.6) \times 10^{9}\,$|L_⊙, corresponding to ∼0.05% and 0.01% of the total infrared luminosity, L_TIR, respectively. Combining these [O iii] luminosities with [C ii] 158 μm luminosities in the literature, we find that J2054−0005 and J2310+1855 have [O iii]-to-[C ii] luminosity ratios of 2.1 ± 0.4 and 0.3 ± 0.1, respectively, the latter of which is the lowest among objects so far reported at z > 6. Combining [O iii] observations at z ≈ 6–9 from our study and the literature, we identify the [O iii] line deficit: objects with a larger L_TIR (total infrared luminosity) have lower L_{[O iii]}-to-L_TIR ratios. Furthermore, we also find that the anti-correlation is shifted toward a higher L_TIR value when compared with the local [O iii] line deficit.

1 Introduction

Quasars are powered by supermassive black holes (SMBHs) with a mass of |${\sim } 10^{8-10}\,$|M_⊙ (e.g., De Rosa et al. 2014; Wu et al. 2015). From wide-area surveys, ∼100 quasars have been discovered at z > 6 (e.g., Fan et al. 2003; Jiang et al. 2016; Mazzucchelli et al. 2017; Matsuoka et al. 2018) and up to z = 7.54 (Bañados et al. 2018). How SMBHs have accreted within |${\sim } 1 \,$|Gyr of the Big Bang is one of the most important questions in modern astronomy (Valiante et al. 2017).

In the local Universe, there is a tight correlation between the central black hole mass and the bulge mass (Häring & Rix 2004; Kormendy & Ho 2013). Given the coevolution of SMBHs and their host galaxies, an understanding the host galaxy properties at the earliest Universe is crucial. Rest-frame far-infrared (FIR) dust continuum observations show that high-z quasar host galaxies have star formation rates (SFRs) of ∼50–2700 M_⊙ yr⁻¹ and large dust masses ≈10⁷–|$10^{9}\,$|M_⊙ (e.g., Wang et al. 2008; Venemans et al. 2018). Carbon monoxide (CO) line observations reveal a large amount of gas mass in the host galaxies (⁠|${\sim } 10^{10}\,$|M_⊙; e.g., Wang et al. 2010; Venemans et al. 2017a; Feruglio et al. 2018). The FIR fine structure line of [C ii] 158 μm is widely used to obtain the precise redshift and the dynamical mass (e.g., Maiolino et al. 2005; Wang et al. 2013; Venemans et al. 2017b; Decarli et al. 2018; Izumi et al. 2018).

Combinations of multiple FIR fine structure lines are useful to obtain the physical properties of the interstellar medium (ISM) such as the gas-phase metallicity, the electron density, and the ionization parameter (e.g., Nagao et al. 2011; Pereira-Santaella et al. 2017). Among the FIR lines, the [O iii] 88.356 μm line (ν_rest = 3393.006244 GHz) would be a good next target after [C ii] because it is the second most commonly observed line in normal star-forming galaxies at z > 6 (e.g., Inoue et al. 2016; Tamura et al. 2019). Indeed, recent Atacama Large Millimeter/sub-millimeter Array (ALMA) observations demonstrate that [O iii] is detectable even at z = 9.11 (Hashimoto et al. 2018).

In this paper we report on the results of our ALMA band 8 observations targeting [O iii] in two quasars at z ≈ 6, J205406.48−000514.8 (hereafter J2054−0005) and J231038.88+185519.7 (J2310+1855). With our observations (section 2), we successfully detect [O iii] and the underlying dust continuum (section 3). In conjunction with [C ii] measurements in the literature, we discuss their [O iii]-to-[C ii] line luminosity ratios (section 4). Throughout this paper we adopt a flat ΛCDM cosmology (⁠|$\Omega _{\small {m}} = 0.272$|⁠, |$\Omega _{\small {\Lambda}} = 0.728$|⁠, and |$H_{\small {0}} = 70.4$| km s⁻¹ Mpc⁻¹; Komatsu et al. 2011). The solar luminosity, L_⊙, is 3.839 × 10³³ erg s⁻¹, and k_B represents the Boltzmann constant.

2 Our sample and ALMA band 8 data

2.1 Sample

At the time of writing our proposal, 2017 April, there were 13 quasars with [C ii] detections at z ≥ 6.0.¹ We excluded one object with too high a declination for ALMA observations. We then omitted four objects with the redshift at which [O iii] emission is strongly affected by atmospheric absorption.² To secure the [O iii] line detection within reasonable ALMA integration times, we selected objects with (i) a bright total infrared luminosity, L_TIR, and (ii) a relatively lower z among the candidates. Finally, this leaves us with two objects, J2054−0005 and J2310+1855, which have a very bright total infrared luminosity, log(L_TIR|$/$|L_⊙) ≈13, at z = 6.0. In fact, J2310+1855 has the brightest infrared, [C ii], and CO(6–5) luminosities at z ≥ 6.0 (Decarli et al. 2018; Venemans et al. 2018; Feruglio et al. 2018).

These objects were originally discovered by the Sloan Digital Sky Survey data (Jiang et al. 2008, 2016). J2054−0005 (J2310+1855) has a UV absolute magnitude of M₁₄₅₀ = −26.1 (−27.8) and a bolometric luminosity of |$2.8\times 10^{13}\,$|L_⊙ (⁠|$9.3\times 10^{13}\,$|L_⊙)—see Wang et al. (2013). The BH mass in J2054−0005 (J2310+1855) is estimated to be |$0.9^{+1.6}_{-0.6}\times 10^{9}\,$|M_⊙ (⁠|$2.3^{+5.1}_{-1.8}\times 10^{9}\,$|M_⊙) under the assumption of Eddington-limited mass accretion (Wang et al. 2013; Willott et al. 2015). The [C ii] redshift value of J2054−0005 (J2310+1855) is z = 6.0391 ± 0.0002 (6.0031 ± 0.0002)—see Wang et al. (2013).

2.2 Observations and data

We performed observations of [O iii] with ALMA band 8 from 2018 March through July (ID 2017.1.01195.S, PI: T. Hashimoto). In J2054−0005 (J2310+1855), 43 antennas with baseline lengths of 15–785 m (15–360 m) were used, and the total on-source exposure time was 127 min (176 min). Four spectral windows (SPWs) with a bandwidth of 1.875 GHz were used in frequency division mode. Two slightly overlapping SPWs (0 & 1) were used to target [O iii], covering the frequency range of 480.71–483.68 GHz (483.19–486.31 GHz) for J2054−0005 (J2310+1855). The other two SPWs (2 & 3) were used to observe the continuum, covering 492.21–495.96 GHz (494.70–498.45 GHz) for J2054−0005 (J2310+1855). The quasar J1924−2914 (J2258−2758) was used for bandpass and flux calibrations, and the quasar J2101+0341 (J2253+1608) was used for phase calibrations.The data were reduced and calibrated using the CASA (Common Astronomy Software Applications) pipeline, version 5.1.1-5. We produced images and cubes with the CLEAN task using the natural weighting. To create a pure dust continuum image, we collapsed all off-line channels. To create a pure line image, we subtracted the continuum using the off-line channels in the line cube with the CASA task uvcontsub. For J2310+1855, we could not obtain the data product in SPW 1 due to very strong atmospheric absorption.³

With the CASA task imstat, we estimate the rms level of the continuum image of J2054−0005 (J2310+1855) to be 67|$\, \mu$|Jy beam⁻¹ (106|$\, \mu$|Jy beam⁻¹). The spatial resolution of the continuum image is |${0{^{\prime \prime }_{.}}38} \times {0{^{\prime \prime }_{.}}34}$| (⁠|${0{^{\prime \prime }_{.}}69} \times {0{^{\prime \prime }_{.}}60}$|⁠) in full width at half maximum (FWHM) with a beam position angle, PA, of 69° (−61°). The typical rms level of the line cube is 0.6 mJy beam⁻¹ (0.8 mJy beam⁻¹) per 30 km s⁻¹ bin.

3 Results

3.1 Dust continuum

Our data probe dust continuum emission at the rest-frame wavelength, λ_rest, of |${\sim } 87\,$|μm. The top left and bottom left panels of figure 1 show dust continuum images of J2054−0005 and J2310+1855, respectively. Our measurements are summarized in table 1.

Fig. 1.

Dust continuum image at |${\sim } 87\,$|μm (left), the [O iii] 88 μm line image (middle), and the continuum-subtracted [O iii] spectrum (right). In the left and middle panels, the ellipse at the lower left corner indicates the synthesized beam size of ALMA, and the scale bar is shown at the upper left corner. Negative and positive contours are shown by the white dashed and black solid lines, respectively. The white crosses show the optical position in the SDSS z-band image whose astrometry is calibrated by GAIA astrometry (see subsection 3.2). The size of the white cross corresponds to the |$1 \, \sigma$| positional uncertainty of the optical counterpart. We do not see significant (⁠|$\gt\!\! 3 \, \sigma$|⁠) positional offsets between the ALMA and optical images. In the right panel, the continuum-subtracted [O iii] spectrum is extracted from the region with |$\gt\!\! 3 \, \sigma$| detections in the velocity-integrated intensity images. The black solid line is the best-fit Gaussian for the [O iii] line, while the black dashed lines show the |$\pm 1 \, \sigma$| noise levels. (Top) J2054−0005: Dust continuum contours drawn at (−2, 2, 4, 8, 16, 32, 64, 100) × σ, where |$\sigma = 67\, \mu$|Jy beam⁻¹. The [O iii] line contours are drawn at (−2, 2, 4, 8, 12, 16) × σ, where σ = 94 mJy beam⁻¹ km s⁻¹. (Bottom) J2310+1855: Dust continuum contours at |${\sim } 87\, \mu$|m drawn at (−2, 2, 4, 8, 16, 32, 64, 128, 200) × σ, where |$\sigma = 106\, \mu$|Jy beam⁻¹. The [O iii] line contours are drawn at (−2, 2, 4, 6) × σ, where σ = 149 mJy beam⁻¹ km s⁻¹. (Color online)

Table 1.

Summary of observational results.

	J2054−0005	J2310+1855
|$z_{\rm [O\, {\small {III}}]}$|	6.0391 ± 0.0002	6.0035 ± 0.0007
FWHM([O iii]) [km s−1]	282 ± 17	333 ± 72
[O iii] integrated flux [Jy km s−1]	3.79 ± 0.34	1.38 ± 0.34
[O iii] luminosity [|$10^{9}\,$|L⊙]	6.79 ± 0.61	2.44 ± 0.61
S  ν, 87 [mJy]	10.35 ± 0.15	24.89 ± 0.21
Dust deconvolved size* [arcsec2]	(0.23 ± 0.01) × (0.15 ± 0.02)	(0.31 ± 0.01) × (0.22 ± 0.02)
[O iii] deconvolved size* [arcsec2]	(0.49 ± 0.07) × (0.45 ± 0.06)	(0.44 ± 0.27) × (0.38 ± 0.13)
ALMA dust position (ICRS)	|${20^{\rm h}54^{\rm m}06{^{\rm s}_{.}}503}, -00^{\circ }{05^{\prime}14{^{\prime \prime }_{.}}43}$|	|${23^{\rm h}10^{\rm m}38{^{\rm s}_{.}}902}, +18^{\circ }{55^{\prime}19{^{\prime \prime }_{.}}83}$|
ALMA [O iii] position (ICRS)	|${20^{\rm h}54^{\rm m}06{^{\rm s}_{.}}503}, -00^{\circ }{05^{\prime}14{^{\prime \prime }_{.}}48}$|	|${23^{\rm h}10^{\rm m}38{^{\rm s}_{.}}900}, +18^{\circ }{55^{\prime}19{^{\prime \prime }_{.}}80}$|
SDSS optical position† (ICRS)	|${20^{\rm h}54^{\rm m}06{^{\rm s}_{.}}486}, -00^{\circ }{05^{\prime}14{^{\prime \prime }_{.}}50}$|	|${23^{\rm h}10^{\rm m}38{^{\rm s}_{.}}882}, +18^{\circ }{55^{\prime}19{^{\prime \prime }_{.}}61}$|
T  d [K]	50 ± 2	37 ± 1
βd	1.8 ± 0.1	2.2 ± 0.1
L  TIR  ‡ [|$10^{13}\,$|L⊙]	|$1.3^{+0.2}_{-0.2}$|	|$1.9^{+0.2}_{-0.1}$|
SFRIR§ [M⊙ yr−1]	|$1897^{+265}_{-216}$|	|$2873^{+294}_{-232}$|

*The values represent major- and minor-axis FWHM values of a two-dimensional (2D) Gaussian profile.

^†The SDSS optical position recalibrated with the GAIA astrometry (see subsection 3.3).

^‡The total luminosity, L_TIR, is estimated by integrating the modified blackbody radiation at 8–|$1000\,$|μm.

^§The SFR value is obtained following Kennicutt and Evans (2012) under the assumption of the Kroupa initial mass function (IMF; Kroupa 2001) in the range 0.1–100 |$M_{\odot}$|⁠. The value should be treated as an upper limit on the obscured SFR (see subsection 3.4).

• J2054−0005: To estimate the flux density and the beam deconvolved size of the dust continuum, we apply the CASA task imfit assuming a 2D Gaussian profile for the specific intensity. We estimate the continuum flux density to be S_ν,87μm = 10.35 ± 0.15 mJy. The beam deconvolved size is (0.23 ± 0.01) × (0.15 ± 0.02) arcsec², corresponding to (1.34 ± 0.06) × (0.88 ± 0.13) kpc² at z = 6.0391, with PA = 177° ± 7°.

• J2310+1855: The continuum flux density is 24.89 ± 0.21 mJy. The beam deconvolved size is (0.31 ± 0.01) × (0.22 ± 0.02) arcsec², corresponding to (1.81 ± 0.06) × (1.28 ± 0.13) kpc² at z = 6.0391, with PA = 154° ± 8°.

These deconvolved sizes and position-angle values are consistent with those obtained by Wang et al. (2013) using ALMA band 6 data within 1–|$2 \, \sigma$| uncertainties.⁴

3.2 [O iii] 88 μm

[O iii] is detected in both quasars. Our measurements are summarized in table 1. The middle panels of figure 1 show a velocity-integrated intensity image between 481.7–482.6 GHz (484.1–485.0 GHz) for J2054−0005 (J2310+1855). The peak intensity is 1.67 ± 0.10 Jy km s⁻¹ beam⁻¹ (0.94 ± 0.15 Jy km s⁻¹ beam⁻¹). We perform photometry on the image with the CASA task imfit assuming a 2D Gaussian profile for the line intensity.

For J2054−0005, the total line flux is estimated to be 3.79 ± 0.34 Jy km s⁻¹. The beam-deconvolved size is (0.49 ± 0.07) × (0.45 ± 0.06) arcsec², corresponding to (2.87 ± 0.41) × (2.63 ± 0.35) kpc² at z = 6.0391, with PA = 75° ± 82°. Likewise, for J2310+1855, the total line flux is estimated to be 1.38 ± 0.34 Jy km s⁻¹. The beam-deconvolved size is (0.44 ± 0.27) × (0.38 ± 0.13) arcsec², corresponding to (2.57 ± 1.58) × (2.22 ± 0.76) kpc² at z = 6.0035, with PA = 70° ± 97°. We note that both quasars have an [O iii] emitting region size of ∼2–3 kpc (FWHM), which is significantly larger than the continuum emitting region size of ∼1 kpc (FWHM).

The top (bottom) right panel of figure 1 shows the continuum-subtracted spectrum of J2054−0005 (J2310+1855) extracted from the [O iii] region with |$\gt 3 \, \sigma$| detections in the velocity-integrated intensity image. We obtain an [O iii] redshift of 6.0391 ± 0.0002 (6.0035 ± 0.0007) and an FWHM value of 282 ± 17 km s⁻¹ (333 ± 72 km s⁻¹). Based on a combination of the flux and redshift values, we obtain [O iii] luminosities of (6.79 ± 0.61) × 10⁹ and |$(2.44 \pm 0.61) \times 10^{9}\,$|L_⊙ for J2054−0005 and J2310+1855, respectively.

To investigate a possible broad velocity component in the [O iii] line, as found in a z = 6.4 quasar in [C ii] (Maiolino et al. 2012), we extract two additional spectra from the [O iii] regions with |$\gt\!\! 1 \, \sigma$| and |$\gt\!\! 2 \, \sigma$| detections in the velocity-integrated intensity images. We do not find any broad velocity component in the spectra.

We compare our [O iii] measurements with [C ii] measurements presented by Wang et al. (2013). For J2054−0005, the [O iii] emitting region size, (0.49 ± 0.07) × (0.45 ± 0.06) arcsec², is consistent with the [C ii] emitting region, (0.35 ± 0.04) × (0.32 ± 0.05) arcsec², within |${\sim } 2 \, \sigma$| uncertainties. Likewise, the [O iii] line FWHM, 282 ± 17 km s⁻¹, is consistent with that of [C ii], 243 ± 10 km s⁻¹, within |${\sim } 2 \, \sigma$| uncertainties. For J2310+1855, the [O iii] emitting region size, (0.44 ± 0.27) × (0.38 ± 0.13) arcsec², is consistent with that of [C ii], (0.56 ± 0.03) × (0.39 ± 0.04) arcsec², within |$1 \, \sigma$| uncertainties. Likewise, the [O iii] line FWHM value, 333 ± 72 km s⁻¹, is consistent with that of [C ii], 393 ± 21 km s⁻¹, within |$1 \, \sigma$| uncertainties. The case of J2054−0005 might reveal that the size of the [O iii] emitting region and the line FWHM are larger than those of [C ii]. If this is the case, [O iii] and [C ii] lines may trace different regions of the quasar. However, because these differences are only marginal (⁠|${\sim } 2 \, \sigma$|⁠), we do not attempt to discuss this further.

3.3 Astrometry

For J2054−0005 (J2310+1855), we find that the spatial positions of the dust continuum and [O iii] are consistent within a 48 (72) mas uncertainty. Hereafter, we use the dust continuum positions due to their high-significance detections. Based on the IRAF task imexam, J2054−0005 is at |$(\alpha , \delta ) = ({20^{\rm h}54^{\rm m}06{^{\rm s}_{.}}503}, -00^{\circ }{05^{\prime}14{^{\prime \prime }_{.}}43})$|⁠, and J2310+1855 is at |$(\alpha , \delta ) = ({23^{\rm h}10^{\rm m}38{^{\rm s}_{.}}902}, +18^{\circ }{55^{\prime}19{^{\prime \prime }_{.}}83})$| in the International Celestial Reference System (ICRS), on which ALMA relies (table 1).

We also compare the positions of the ALMA and the SDSS optical images (cf. Shao et al. 2019; Wang et al. 2019). To do so, we first recalibrate the astrometry of the SDSS z-band image (Eisenstein et al. 2011) where the quasars are detected. Using nearby bright stars whose positions are accurately measured in the GAIA second data release (DR2) catalog in the ICRS frame (Gaia Collaboration 2016, 2018), we performed IRAF tasks ccmap and ccsetwcs to recalibrate the astrometry of the SDSS image. The astrometry uncertainty (i.e., systematic uncertainty) is estimated to be ∼120 and 100 mas around J2054−0005 and J2310+1855, respectively, based on a comparison of bright star positions in the GAIA catalog and the SDSS image with recalibrated astrometry. In addition, we estimate the positional uncertainty arising from the IRAF task imexam (i.e., measurement uncertainty) to be ∼350 (40) mas for J2054−0005 (J2310+1855).⁵ The relatively large uncertainty for J2054−0005 is due to the low signal-to-noise ratio in the z-band image. We regard 470 and 140 mas as the final positional uncertainties for J2054−0005 and J2310+1855, respectively.

In the left panels of figure 1, white crosses indicate the optical positions in the SDSS z-band image with recalibrated astrometry. The optical position of J2054−0005 is |$(\alpha , \delta ) = ({20^{\rm h}54^{\rm m}06{^{\rm s}_{.}}486}, -00^{\circ }{05^{\prime}14{^{\prime \prime }_{.}}50})$|⁠, ∼250 mas offset from the ALMA position. Likewise, the optical position of J2310+1855 is |$(\alpha , \delta ) = ({23^{\rm h}10^{\rm m}38{^{\rm s}_{.}}882}, +18^{\circ }{55^{\prime}19{^{\prime \prime }_{.}}61})$|⁠, ∼350 mas offset from the ALMA position. Given the positional uncertainties, we do not conclude that there is a significant (⁠|${>}3 \, \sigma$|⁠) offset between the two images (cf. Shao et al. 2019; Wang et al. 2019). ALMA higher angular resolution data and deeper and higher angular resolution optical images would be useful for further investigating whether there is a possible spatial offset between the two images.

3.4 Tight constraints on the dust temperature and the infrared luminosity

Previous studies often assume that far-IR (FIR) dust continuum emission of quasars at |$\lambda _{\rm rest} \gtrsim 50\,$|μm is mainly powered by star formation activity, with negligible contribution from active galactic nuclei (AGNs; e.g., Leipski et al. 2013). Assuming that FIR dust continuum emission is described as optically thin modified blackbody radiation, |$I_{\rm \nu } \propto \nu ^{3+\beta _{\rm d}}/[\exp (h\nu /kT_{\rm d}) - 1]$|⁠, we constrain the single dust temperature, T_d, and the dust emissivity index, β_d, of the two quasar host galaxies, taking cosmic microwave background (CMB) effects into account (da Cunha et al. 2013)—see table 1.

For J2054−0005 we use four flux density measurements of 12.0 ± 4.9 mJy, 10.35 ± 0.15 mJy, 2.98 ± 0.05 mJy, and 2.38 ± 0.53 mJy obtained with Herschel 350 μm data (Leipski et al. 2013), our ALMA 488 GHz, ALMA 262 GHz data (Wang et al. 2013), and MAMBO 250 GHz data (Wang et al. 2008), respectively. These data sample λ_rest ≈ 50–200 μm. By fitting modified blackbody models corrected for the CMB effects to the photometry data, we obtain T_d = 50 ± 2 K and β_d = 1.8 ± 0.1 based on the χ² statistics. The best-fit model is shown in the left panel of figure 2. Integrating the modified blackbody radiation over 8–1000|$\, \mu$|m, we obtain the total infrared luminosity to be L_TIR|$= 1.3^{+0.2}_{-0.2} \times 10^{13}\,$|L_⊙. Following Kennicutt and Evans (2012) under the assumption of the Kroupa IMF (Kroupa 2001) in the range of 0.1–100 M_⊙, we obtain the IR-based star formation rate SFR_IR ≈ 1900 M_⊙ yr⁻¹. Note that our ALMA band 8 data are useful for constraining T_d because the data probe the wavelengths close to the peak of the dust spectral energy distribution (SED).

Fig. 2.

FIR dust SED of J2054−0005 (left) and J2310+1855 (right). In each panel, black squares denote the measurements, with error bars typically smaller than the symbols, while the best-fit data are shown with yellow crosses. The red line corresponds to the best-fit SED. See the text for details of the data used in the fit. (Color online)

Likewise, for J2310+1855 we use five flux density measurements of 24.89 ± 0.21 mJy, 8.91 ± 0.08 mJy, 8.29 ± 0.63 mJy, 0.40 ± 0.05 mJy, and 0.41 ± 0.03 mJy obtained with our ALMA 488 GHz, ALMA 262 GHz data (Wang et al. 2013), MAMBO 250 GHz data, MAMBO 99 GHz data (Wang et al. 2008), and ALMA 91.5 GHz (Feruglio et al. 2018), respectively. These data sample λ_rest ≈ 90–500 μm. We obtain T_d = 37 ± 1 K, β_d = 2.2 ± 0.1, L_TIR|$=1.9^{+0.2}_{-0.1} \times 10^{13}\,$|L_⊙, and SFR|$_{\rm IR} \approx 2900\,$|M_⊙ yr⁻¹ in the same way as for J2054−0005. We use these T_d and SFR_IR values to interpret our results in section 4.

These T_d and β_d values are within the ranges obtained in a mean SED of six quasar host galaxies at z = 1.8–6.4, T_d = 47 ± 3 K and β_d = 1.6 ± 0.1 (Beelen et al. 2006), and in a mean SED of seven quasar host galaxies at z ≈ 4–5, T_d = 41 ± 5 K and β_d = 1.95 ± 0.3 (Priddey & McMahon 2001). Nevertheless, our results demonstrate the variety of dust properties on an individual basis.

In the discussion above we have assumed that the dust continuum emission is purely powered by star formation activity. However, recent studies have decomposed the FIR SED of local AGNs and quasars into components heated by star formation and AGN activity (e.g., Symeonidis et al. 2016; Ichikawa et al. 2019); these studies show that powerful AGN activity can actually dominate the dust continuum emission up to |$\lambda _{\rm rest} \approx 90\,$|μm or longer wavelengths (see also Schneider et al. 2015; Symeonidis 2017). Because the two z ≈ 6 quasars studied here are very luminous, a significant fraction of L_TIR could be powered by AGN activity.⁶ In particular, J2054−0005 has a very compact dust continuum emitting region (FWHM ≈1 kpc), implying that its high T_d could be largely due to heating by the quasar. Therefore, our values should be treated as upper limits on the dust-obscured SFRs.

3.5 Luminosity ratios

3.5.1 [O iii]-line deficit and its redshift evolution

It is widely known that the FIR line luminosity-to-L_TIR ratio anti-correlates with L_TIR, the so-called FIR line deficit, particularly in the high L_TIR regime (L_TIR|$\gtrsim 10^{11}\,$|L_⊙; e.g., Malhotra et al. 1997; Stacey et al. 2010; Graciá-Carpio et al. 2011; Díaz-Santos et al. 2013, 2017; Herrera-Camus et al. 2018a, 2018b). The line deficit was first identified in [C ii]. A number of hypotheses have been proposed to explain the [C ii] line deficit in extreme objects such as (ultra-)luminous infrared galaxies, (U)LIRGs, and luminous quasars, although no consensus has yet been reached (e.g., Kaufman et al. 1999; Malhotra et al. 2001; Luhman et al. 2003; Abel et al. 2009; Graciá-Carpio et al. 2011; Langer & Pineda 2015; Muñoz & Oh 2016; Díaz-Santos et al. 2017; Lagache et al. 2018; Herrera-Camus et al. 2018a; Rybak et al. 2019).

Focusing on [O iii], based on a compiled sample of local dwarf and spiral galaxies with high dynamic ranges in metallicity and L_TIR, Cormier et al. (2015) showed that the |$L_{\rm [O\, {\small {III}}]}$|-to-L_TIR ratio anti-correlates with L_TIR (see also, e.g., Malhotra et al. 1997; Graciá-Carpio et al. 2011; Díaz-Santos et al. 2017; Herrera-Camus et al. 2018a, 2018b). The local galaxies have |$L_{\rm [O\, {\small {III}}]}$|-to-L_TIR ratios ranging from ∼10⁻⁵ to ∼10⁻². Recently, Tamura et al. (2019) investigated the relation at higher-z based on a compiled sample of z ≈ 7–9 galaxies, showing that at least high-z galaxies with dust continuum detections follow a similar relation as in the local Universe. Our two quasars are useful for further investigating the trend at the reionization epoch because of their high L_TIR values. The luminosity ratios are log(⁠|$L_{\rm [O\, {\small {III}}]}/$|L_TIR) = −3.3 ± 0.1 and −4.0 ± 0.1 for J2054−0005 and J2310+1855, respectively.

In the left panel of figure 3 we plot the two quasars along with eight objects at z > 7 (see the caption for details) and lower-z objects. The latter include various populations of local galaxies taken from Herschel Dwarf Galaxy Survey (DGS; Cormier et al. 2015) and SHINING (Herrera-Camus et al. 2018a) samples, local spirals (Brauher et al. 2008), and lensed sub-millimeter galaxies (SMGs) at z ≈ 1–4 (Zhang et al. 2018). We confirm a trend that high-z objects follow a similar relation to the local Universe. Following the explanations for the [C ii] line deficit, we propose two possible explanations for the [O iii] line deficit. First, the collisional de-excitation of [O iii] may significantly reduce |$L_{\rm [O\, {\small {III}}]}$| (and hence the luminosity ratio) in the high electron density environment (n_e > n_crit ≈ 500 cm⁻³). Secondly, the strong AGN radiation can contribute to L_TIR (e.g., Symeonidis et al. 2016) while ionizing oxygen higher than O²⁺ (e.g., Spinoglio et al. 2015). Detailed modeling is needed to conclude the origin of the [O iii]-line deficit, which we leave for future studies.

Fig. 3.

(Left) |$L_{\rm [O\, {\small {III}}]}$|-to-L_TIR ratio plotted against L_TIR, where the luminosities are corrected for magnification, if any. Orange hexagons show the two quasars. The other eight filled symbols represent z ≈ 7–9 objects compiled by Tamura et al. (2019): SPT0311−58 E/W at z = 6.90 (triangles, Marrone et al. 2018), BDF-3299 at z = 7.11 (pentagon with two arrows, Carniani et al. 2017), B14-65666 at z = 7.15 (filled diamond, Hashimoto et al. 2019), SXDF-NB1006-2 at z = 7.21 (circle with two arrows, Inoue et al. 2016), MACS0416_Y1 at z = 8.31 (five-pointed star, Tamura et al. 2019), A2744_YD4 at z = 8.38 (square, Laporte et al. 2017), and MACS1149-JD1 at z = 9.11 (triangle with two arrows, Hashimoto et al. 2018). The open symbols show lower-z galaxies including the Herschel DGS (open circles, Cormier et al. 2015) and SHINING samples (thin open squares, Herrera-Camus et al. 2018a), the median of local spirals (thick open square, Brauher et al. 2008), and z ≈ 2–4 dusty star-forming galaxies with spectroscopic redshifts (open triangles, Zhang et al. 2018). For the z > 6 objects, except for the two quasars and SPT0311−58 E/W, we have assumed L_TIR = 50 K and β_d = 1.6 for consistency. The blue-to-green color code shown for MACS0416_Y1, SXDF-NB1006-2, MACS1149-JD1, B14-65666, and local dwarfs indicates the best-fitting oxygen abundances. (Right) [O iii]-to-[C ii] luminosity ratio at z > 6. The sample includes the two quasars (two red five-pointed stars), SPT0311−58 E/W (triangles), B14-65666 (circle), BDF-3299 (“+” symbol), and SXDF-NB1006-2 (“−” symbol with an upward arrow). Note that the definition of L_bol is different in quasars and star-forming galaxies: L_bol for quasars indicates the quasar/AGN bolometric luminosity, taken from the literature. On the other hand, L_bol for other galaxies without AGN activity indicates the galaxy's bolometric luminosity. These are obtained from Hashimoto et al. (2019) as the sum of the UV luminosity and L_TIR, where we assume L_TIR = 50 K and β_d = 1.6 except for SPT0311−58 E/W. For the two Lyα emitters with L_TIR upper limits, the lower limit corresponds to the UV luminosity, while the upper limit denotes the summation of the UV luminosity and the |$3 \, \sigma$|L_TIR upper limits. (Color online)

In the left panel of figure 3 we find that the relation between |$L_{\rm [O\, {\small {III}}]}$|-to-L_TIR and L_TIR at high z is shifted toward higher L_TIR values (or higher |$L_{\rm [O\, {\small {III}}]}$|-to-L_TIR values). Such a shift toward higher L_TIR values is also found in the relation between |$L_{\rm [C\, \small {II}]}$|-to-L_TIR and L_TIR (e.g., Stacey et al. 2010). In the case of [C ii], to understand the origin of this shift in high-z galaxies, Narayanan and Krumholz (2017) coupled analytical models for the structure of giant molecular clouds in galaxies with chemical equilibrium networks and radiative transfer models. The authors proposed that the shift could arise if high-z galaxies have larger gas masses at a given SFR (i.e.. L_TIR). In this case, the gas surface density of an individual molecular cloud (i.e., the star formation efficiency, SFE = SFR/M_mol, defined as the SFR per molecular gas mass) becomes lower, which in turn reduces the ability for self-shielding of molecular clouds against ionizing photons, leading to a brighter [C ii] luminosity. Although it is not clear whether the same discussion is applicable for [O iii], there is growing evidence that the molecular gas properties are the key to regulating the general trend in the FIR line deficit; Herrera-Camus et al. (2018a) show that the discrepancy between low- and high-z objects becomes less prominent if one plots the L_{[C ii]}-to-L_TIR ratio against the IR surface brightness (Σ_IR), the latter of which is related to the gas surface density or the SFE. Such an analysis for [O iii] would be interesting in future by the use of larger samples in the local and high-z Universe.

3.5.2 [O iii]-to-[C ii] luminosity ratio

We next turn our attention to the [O iii]-to-[C ii] luminosity ratio. Based on a compiled sample of five galaxies at z ≳ 7 with [O iii] and [C ii] observations (Inoue et al. 2016; Carniani et al. 2017; Marrone et al. 2018), Hashimoto et al. (2019) have demonstrated a trend that the [O iii]-to-[C ii] line luminosity ratio becomes small if a galaxy has a large galaxy bolometric luminosity. Their sample includes two Lyα emitters, one Lyman break galaxy, and two SMGs. J2054−0005 and J2310+1855 provide an invaluable opportunity to investigate the line luminosity ratio in quasars at z > 6 for the first time.

For J2054−0005 (J2310+1855), combining our [O iii] luminosity and the [C ii] luminosity of |$3.3 \pm 0.5\times 10^{9}\,$|L_⊙ (⁠|$8.7 \pm 1.4\times 10^{9}\,$|L_⊙) in Wang et al. (2013), we obtain a line luminosity ratio of 2.1 ± 0.4 (0.3 ± 0.1). In the right panel of figure 3, red star symbols show the line luminosity ratio of the two quasars plotted against the quasar/AGN bolometric luminosity. Just for comparison, the black and blue symbols show the line luminosity ratios in star-forming galaxies without AGN activity at z > 6 plotted against the galaxy bolometric luminosity. Notably, J2310+1855 has the lowest [O iii]-to-[C ii] ratio so far reported among objects at z > 6.

4 Discussion and summary

To interpret [O iii] in quasars, we need to separate the [O iii] contributions from star formation and AGN activity. For the latter, [O iii] can arise from the narrow-line region (NLR) of AGNs because the NLR has a relatively small electron density of 100–300 cm⁻³ (e.g., Bennert et al. 2006; Kakkad et al. 2018), which is smaller than the critical density of [O iii] 88 μm (≈500 cm⁻³). Indeed, the sizes of the [O iii] emitting regions of the two quasars (∼2–3 kpc in FWHM) are reasonable for the size of stellar disks or extended NLRs. Ideally, one can separate the contribution based on spatially resolved diagnostics such as Baldwin–Philips–Terlevich diagrams (Baldwin et al. 1981), as performed in local Seyfert and LINER galaxies (Kakkad et al. 2018). However, it is difficult to separate the contribution based on the [O iii]-to-[C ii] line ratio alone because it is insensitive to the presence of AGNs; Herrera-Camus et al. (2018a) demonstrated that both star formation and AGN activity can reproduce the [O iii]-to-[C ii] luminosity ratio of ∼0.1–2.0 (see their figure 11).

Thus we focus on the fact that [O iii] and [C ii] have consistent redshifts, emitting region sizes, and line FWHMs within uncertainties (subsection 3.2). These results would imply that [O iii] arises from star-forming regions as traced by [C ii] (see a similar argument by Walter et al. 2018), although we cannot rule out the possibility that both [O iii] and [C ii] are partly affected by the NLRs.

Next we examine whether [O iii] luminosity-based SFRs (SFR|$_{\rm O\, {\small {III}}}$|⁠) are comparable to SFR_IR values (table 1). It is expected that the two SFR values are consistent with each other if [O iii] is mainly powered by star-forming activity. For the conversion of the [O iii] luminosity to the SFR we use the empirical relation in the local Universe, which assumes the Kroupa IMF in the range of 0.1–100 M_⊙ (De Looze et al. 2014). The authors presented different empirical relations for, e.g., metal-poor dwarf galaxies, starburst, the composite of star formation and AGNs, ULIRGs, and the entire sample. Because the two quasars presented in this study have L_TIR|$\approx 10^{13}\,$|L_⊙, the empirical relation for ULIRGs would be the most suitable among the relations in De Looze et al. (2014). Although the typical uncertainty of the relation is a factor of 2.5 (see table 3 in De Looze et al. 2014), it is unclear whether the local relation can be applied to luminous quasars at z = 6 as presented in this study. Therefore, the actual uncertainty would be larger than a factor of 2.5. On the other hand, as discussed in subsection 3.4, the SFR values estimated from L_TIR are also highly uncertain because it is possible that AGN significantly contaminates the FIR dust continuum emission (e.g., Symeonidis et al. 2016). With these in mind, we find that SFR|$_{\rm O\, {\small {III}}}$| in J2054−0005 is about five times larger than SFR_IR, whereas the two SFR values in J2310+1855 are consistent within a factor of two. Given the large systematic uncertainties in the two SFR values, there is no strong evidence that supports AGN contamination of [O iii] in the two objects in this study.⁷

A possible way to disentangle the [O iii] contributions from star formation and AGN activity is to investigate a spatially resolved map of the [O i] 63 μm-to-[C ii] line ratio defining the AGN-dominated region. This is because [O i] is significantly enhanced in the presence of AGNs due to the fact that [O i] becomes a more efficient coolant than [C ii] in dense and warmer gas (Herrera-Camus et al. 2018a). Combining this map with high-angular-resolution [O iii] data, future studies can infer the [O iii] flux fraction of each component.

Although the origin of [O iii] emission is not clear given the current data, we try to interpret the [O iii]-to-[C ii] line luminosity ratio of the two quasars. In the local Universe, based on a compiled sample of star-forming galaxies and AGN-dominated galaxies, Herrera-Camus et al. (2018a) have statistically demonstrated that [O iii] becomes stronger than [C ii] if galaxies have a higher dust temperature (see also Díaz-Santos et al. 2017). The two quasars seem to be consistent with the trend in the sense that J2054−0005 (J2310+1855) has high (low) dust temperature, T_d = 50 ± 2 K (37 ± 1 K). An interpretation of this result is that J2054−0005 has a harder UV stellar + AGN radiation field than J2310+1855.⁸ Assuming the same dust covering fraction and dust grain size distribution, a harder UV radiation field leads to higher T_d. The harder UV radiation field also naturally enhances [O iii] (ionization potential ∼35 eV) against [C ii] (ionization potential ∼11 eV) if we assume a constant C|$/$|O abundance ratio. This hypothesis can be tested with the line luminosity ratio of [N ii] 205 μm against [O iii], which is a good tracer of the UV radiation hardness (Ferkinhoff et al. 2010). Alternatively, the weak [O iii] in J2310+1855 may be due to its high electron density that causes collisional de-excitation. This can be investigated using the line ratio of [O iii] 88 μm to [O iii] 52 μm, which is sensitive to the electron density because of their different critical densities (Pereira-Santaella et al. 2017). Our results highlight the potential use of [O iii] (and the underlying continuum) as a useful tracer of the ISM in quasar host galaxies.

Acknowledgements

We thank an anonymous referee for valuable comments that have greatly improved the paper. We are grateful to Toru Nagao, Nobunari Kashikawa, Yoshiki Matsuoka, Kohei Ichikawa, and Takuma Izumi for discussions. This paper makes use of the following ALMA data: ADS/JAO.ALMA#2017.1.01195.S. ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada), NSC and ASIAA (Taiwan), and KASI (Republic of Korea), in cooperation with the Republic of Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO, and NAOJ.

Funding for SDSS-III has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, and the U.S. Department of Energy Office of Science. The SDSS-III web site is http://www.sdss3.org/. SDSS-III is managed by the Astrophysical Research Consortium for the Participating Institutions of the SDSS-III Collaboration including the University of Arizona, the Brazilian Participation Group, Brookhaven National Laboratory, Carnegie Mellon University, University of Florida, the French Participation Group, the German Participation Group, Harvard University, the Instituto de Astrofisica de Canarias, the Michigan State/Notre Dame/JINA Participation Group, Johns Hopkins University, Lawrence Berkeley National Laboratory, Max Planck Institute for Astrophysics, Max Planck Institute for Extraterrestrial Physics, New Mexico State University, New York University, Ohio State University, Pennsylvania State University, University of Portsmouth, Princeton University, the Spanish Participation Group, the University of Tokyo, University of Utah, Vanderbilt University, University of Virginia, University of Washington, and Yale University.

This work has made use of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/gaia), processed by the Gaia Data Processing and Analysis Consortium (DPAC, https://www.cosmos.esa.int/web/gaia/dpac/consortium). Funding for DPAC has been provided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement.

T.H. and A.K.I. appreciate support from NAOJ ALMA Scientific Research Grant Number 2016-01A. We are also grateful to KAKENHI grants 26287034 and 17H01114 (K.M. and A.K.I.), and 17H06130 (Y.T.). This work was partly supported by a Grant-in-Aid for Scientific Research 19J01620 (T.H.).

We thank Rodrigo Herrera-Camus for kindly providing us with their SHINING sample data, used in the left panel of figure 3.

Footnotes

1

J1148+5251 (Maiolino et al. 2005, 2012), J112+0641 (Venemans et al. 2012), J2348−3054, J0109−3047, J0305−3150 (Venemans et al. 2016), J2310+1855, J1319+0950, J2054−0005 (Wang et al. 2013), J0100+2802 (Wang et al. 2016), P036+03 (Bañados et al. 2015), J0210−0456 (Willott et al. 2013), J0055+0146, and J2229+1457 (Willott et al. 2015).

2

J1148+5251 is excluded from the candidates because of its high declination for ALMA observations. Four objects, J0305−3150, J1319+0950, P036+03, and J2229+1457, are additionally omitted because their [O iii] frequencies are strongly affected by atmospheric absorption.

3

According to the QA2 Report, the ALMA staff tried to keep part of the data by changing the parameters of the pipeline, but it did not work.

4

Wang et al. (2013) used the CASA task imfit to obtain the beam-deconvolved sizes of [C ii] and dust continuum emitting regions in the same manner as this study.

5

We performed the IRAF task imfit with five cursor positions, the peak flux pixel and the four adjacent pixels, to obtain the position. We then adopt the standard deviation of the results as the 1 σ uncertainty due to the fitting.

6

Recently, Shao et al. (2019) performed detailed multi-wavelength SED analyses for J2310+1855. Figure 4 in their study shows that star formation activity may be a dominant source for the FIR dust continuum emission in J2310+1855.

7

Walter et al. (2018) also compared the two SFR values to examine the origin of [O iii]. Using the empirical relation for high-z (z > 0.5) galaxies, the authors argued that the two SFRs are in good agreement. However, the [O iii] empirical relation for high-z galaxies was constructed with only three objects (see table 3 in De Looze et al. 2014). Thus, their comparison is also subject to uncertainties as described in this study.

8

In the case of J2054−0005, AGN activity may significantly contribute to the radiation field because the object has a very compact dust emitting region.