SPT0346-52: NEGLIGIBLE AGN ACTIVITY IN A COMPACT, HYPER-STARBURST GALAXY AT z = 5.7

SPT0346-52 was observed with the Advanced CCD Imaging Spectrometer (ACIS; Garmire et al. 2003) on board Chandra on 2015 July 29. The source was placed at the aim point of the back-illuminated ACIS-S3 chip. The data were taken in Very Faint mode and were initially processed by the Chandra X-ray Center (CXC) using software version 10.4.1 and CalDB version 4.6.8.

We reprocessed the data with the Chandra Interactive Analysis of Observations (CIAO; version 4.7) tool chandra_repro. All of the bad pixels were removed and the standard grade (0, 2, 3, 4, 6), status, and good-time filters were applied. The net exposure time for the observation is 49.52 ks. We performed energy filtering on events into three Chandra bands: the soft (SB; 0.5–2.0 keV), hard (HB; 2–8 keV), and full (FB; 0.5–8.0 keV) bands. The soft and hard bands probe rest-frame energies 3.3–13.3 keV and 13.3–53.2 keV for z = 5.656, respectively. No detectable X-ray emission is expected from the foreground lens. Existing optical and radio data show no evidence for an AGN in the lens. Moreover, the foreground lens is an elliptical galaxy (based upon the light profile fitting by Ma et al. 2015), and thus should also have negligible X-ray emission from star formation (more than three orders of magnitude below the detection threshold).

We matched the source to the position of ALMA (Figure 2) and used a source extraction radius of 175 enclosing all of the lensed images, which is ∼1.3 times the 90% encircled-energy aperture radius (at 03 off-axis angle). The aperture was chosen such that it is large enough to enclose the ALMA contours without including too much background. The background counts were estimated by placing 78 circular apertures with the same size at random positions in the field. We only detected 3.19 net source counts in the full band (0.73 in the SB and 2.47 in the HB). Due to the low count level, we utilize the tool aprates in the srcflux script in CIAO to place a proper upper limit on the X-ray flux. This tool employs Bayesian statistics to compute the background-marginalized posterior probability distribution for source counts/flux. The posterior distribution can be used to determine flux value and confidence intervals or upper limits. The resultant FB flux is consistent with a non-detection with a 3σ upper limit of 6.0 × 10⁻¹⁵ erg cm⁻² s⁻¹. We derive the rest-frame 0.5–8 keV apparent luminosity L_{0.5−8 keV} (without absorption correction) using the following equation:

$$\begin{eqnarray}&&{L}_{0.5\mbox{--}8\mathrm{keV}}\,=\,4\pi {d}_{{\rm{L}}}^{2}{f}_{0.5\mbox{--}8\mathrm{keV}}{(1+z)}^{{{\rm{\Gamma }}}_{\mathrm{eff}}-2},\end{eqnarray} \tag{ 1 }$$

where d_L is the luminosity distance at z = 5.656 and Γ is the effective power-law photon index. In principle, the photon index can be derived from the hardness ratio, which is the ratio of the photon count rates in the HB and the SB. However, we cannot derive a reliable hardness ratio based on the upper limits in both bands. Instead, Γ_eff is fixed to 1.4 following Xue et al. (2011) and Wang et al. (2013b).

Zoom In Zoom Out Reset image size

We then derive the rest-frame 0.5–8 keV absorption-corrected luminosity L_{0.5−8 keV,unabs} by replacing Γ_eff with intrinsic photon index Γ_int and f_{0.5−8 keV} with unabsorbed flux f_{0.5−8 keV,unabs} in Equation (1). We assume Γ_int = 1.8, a typical value for AGN. The unabsorbed flux is estimated using the tool modelflux within the srcflux script. We run simulations adopting Sherpa (Freeman et al. 2001) models xspowerlaw × xszphabs × xsphabs with fixed Γ = 1.8 for the power-law model and hydrogen column density N_H for the (intrinsic and Galactic) absorption models. We scale the measured 3σ upper limit on the flux by a fractional correction for absorption based on the typical intrinsic N_H (2.3 × 10²³ cm⁻²) from the ALESS SMG sample (Wang et al. 2013b). The absorption-corrected 3σ upper limit is f_{0.5−8 keV,unabs} < 7.6 × 10⁻¹⁵ erg cm⁻² s⁻¹. Since SPT0346-52 is gravitationally lensed, we further correct the X-ray flux and luminosity for lensing magnification assuming there is no differential magnification between the FIR (i.e., ALMA) and the X-ray emission (Hezaveh et al. 2012). The magnification-corrected FB flux and luminosity upper limits are listed in Table 2.

SPT0346-52 was observed with ATCA for 3960 s at 5.5 and 9.0 GHz and 4068 s at 2.1 GHz on 2012 January 25 in the 6A array configuration using the CABB in the 1M-0.5k mode. The data was reduced in the same manner as in Aravena et al. (2013). The resultant synthesized beam sizes are 77 × 52 at 2.1 GHz, 33 × 22 at 5.5 GHz, and 21 × 13 at 9.0 GHz. The continuum was not detected in any band and we place 3σ upper limits (i.e., 3× rms noise values calculated within a 1' region around the source position) of 0.213 mJy at 2.1 GHz, 0.114 mJy at 5.5 GHz, and 0.138 mJy at 9.0 GHz on the radio emission from the source.

We place SPT0346-52 on the L_FIR–L_X plane (Figure 3) in the context of X-ray quasars and starburst galaxies. There, we compare it with other SMGs that are identified as SMG–AGN or SMG-starbursts to determine if it is star formation dominated or AGN dominated. As shown in Figure 3, the starburst galaxies (squares labeled with "S") from the literature occupy different locations than AGN-dominated galaxies (squares labeled with "A") and quasars (dashed line and the gray region). The well-studied unobscured quasars of Elvis et al. (1994) provide the fiducial X-ray to FIR luminosity ratios (the median ratio is L_X/L_FIR = 0.05 and the gray region indicates the standard deviation) for AGN-dominated sources. The luminosity ratio for pure starburst galaxies is about two orders of magnitude lower.

Zoom In Zoom Out Reset image size

For SMGs, which may involve the co-evolution of SMBHs and the host galaxies, a heterogeneous population has been observed. The dividing line between starbursts and AGN is typically taken to be L_X/L_FIR ∼ 0.004 (Alexander et al. 2005). The extensively studied ALESS SMGs from Wang et al. (2013b) are consistent with this notion. The X-ray to FIR ratio upper limit for SPT0346-52 is 0.0038, slightly leftward of the dividing line, which indicates that it is consistent with being starburst dominated in the X-ray. One caveat is that if the absorption for SPT0346-52 is exceptionally high, the X-ray upper limit could be weakened. For example, if N_H is 4 times higher (the highest N_H for ALESS SMGs) than what we adopted, the upper limit will be 1.3 times higher, which would move the limit slightly to the right of the dividing line between starbursts and AGN in Figure 3. For this reason, we also cannot exclude Compton-thick AGN by L_FIR–L_X alone (e.g., Murphy et al. 2009). The second potential caveat is AGN X-ray variability. Using the relation between X-ray luminosity and variability found by Lanzuisi et al. (2014), the fractional variability of SPT0346-52 is expected to be 30%. A third potential caveat is the possibility of differential magnification between the star formation region and the AGN (e.g., Hezaveh et al. 2012). We have assumed a constant magnification of μ = 5.6. It could be that the star formation region is more magnified than the AGN. The maximum difference in magnification between components of this galaxy found by Spilker et al. (2015) is Δμ ∼ 2, which would move the red upper limit in Figure 3 by the same amount. However, without a robust X-ray detection and resolved X-ray image of the system, it is impossible to say any more.

If we adopt L_X/L_FIR = 0.05 for typical quasars and take the ratio (L_X/L_FIR)_SPT0346‐52/(L_X/L_FIR)_quasars following Alexander et al. (2005), the AGN fractional contribution to the FIR luminosity is estimated to be at most 8%. We note that an AGN in SPT0346-52 would be fainter than the quasars studied by Elvis et al. (1994), and that Seyfert 1 galaxies tend to be relatively more X-ray luminous than quasars. Using the extensive Seyfert observations of Rush et al. (1996b) would shift the gray band in Figure 3 (and AGN/starburst boundary) to the right by a factor of a few. This would mean that the FIR contribution of any AGN in SPT0346-52 would be even smaller than our upper limit of 8%. Although we cannot completely rule out the presence of an AGN in the system, it is certainly the star formation that is dominating the FIR emission, which is confirmed by fitting SEDs including an AGN component in Section 3.2.

We previously performed SED fitting on SPT0346-52 with CIGALE FORTRAN assuming no AGN contribution in the IR (Ma et al. 2015), which is consistent with the NIR photometric upper limits and FIR detections. Now we employ CIGALE PYTHON (Roehlly et al. 2014), which includes up-to-date star formation history models, stellar population synthesis models, IR re-emission models, and AGN templates, to constrain the potential AGN contribution in the IR. The photometric data points used in the SED fitting are listed in Table 1. We adopt a Chabrier (2003) IMF and the Bruzual & Charlot (2003) stellar population synthesis models. For star formation history, we assume a delayed-τ model, which rises at early ages and then declines exponentially. This form of star formation history is generally expected for especially high-redshift galaxies (Pacifici et al. 2013; Simha et al. 2014; da Cunha et al. 2015). We use a combination of Dale et al. (2014) IR models accounting for the dust emission from the stellar component and Fritz et al. (2006) AGN templates. The Fritz et al. (2006) AGN models take into account two emission components associated with the AGN: a power law from the central source and the thermal and scattering dust-torus emission. The relative normalization of these components is handled through a parameter, fracAGN, which is the fractional contribution of the AGN to the total IR luminosity (${L}_{\mathrm{IR},\mathrm{total}}\,=\,{L}_{\mathrm{Starburst}}+{L}_{\mathrm{AGN}}$). The model curve is also extended to radio wavelengths. The extension relies on the well-established FIR-to-radio correlation for star-forming galaxies. CIGALE does not include synchrotron emission from AGN.

A pure starburst SED remains the best-fit SED (i.e., with the minimal reduced χ²), constrained by all of the photometric detections and upper limits from NIR to radio wavelengths (Figure 4). Parameters are analyzed under a Bayesian approach generating a posterior probability distribution function. The SED fitting failed to tightly constrain fracAGN given the lack of mid-IR photometric points (which differentiate between AGN and star-forming galaxies) and the loose constraint from the NIR. The posterior probability distribution suggests that the AGN component contributes at most ∼20% to the total IR luminosity, with the highest probability being assigned to a contribution of 0%–5%. This is consistent with the estimation from the L_X/L_FIR ratio in Section 3.1. The model SED with the maximum 20% AGN fraction is also plotted in Figure 4, which is inconsistent with the 100 μm Herschel/PACS upper limit.

Zoom In Zoom Out Reset image size

The ATCA radio continuum upper limits at 2.1 and 5.5 GHz are consistent with the FIR-to-radio correlation for star-forming galaxies. The radio part of the SED for radio-loud AGN would be at least a factor of ∼2 higher (e.g., Rush et al. 1996a, 1996b; Morić et al. 2010).

We have provided evidence that SPT0346-52 is star formation dominated in the IR. Since our SFR is mainly constrained by the FIR photometric data, L_IR being almost all from star formation suggests that the inferred SFR reflects the true SFR. We examine whether this observed high SFR can be physically explained in the framework of "maximum" starbursts (Elmegreen 1999) where a substantial fraction of available gas is consumed to make stars. Following Tacconi et al. (2006), the maximum ("Schmidt-law") SFR can be written as

$$\begin{eqnarray}&&{\mathrm{SFR}}_{\max }\,=\,\displaystyle \frac{\epsilon {f}_{g}{M}_{\mathrm{tot}}}{{t}_{\mathrm{dyn}}}\,=\,630\left(\displaystyle \frac{\epsilon }{0.1}\right)\left(\displaystyle \frac{{f}_{g}}{0.4}\right){\left(\displaystyle \frac{{v}_{c}}{400}\right)}^{3}({M}_{\odot }\,{\mathrm{yr}}^{-1}),\end{eqnarray} \tag{ 2 }$$

where Elmegreen (1999) defines as the star formation efficiency, f_g is the gas fraction, M_tot is the total (dynamical) mass of the system, t_dyn is the dynamical time, and v_c is the circular velocity in km s⁻¹. We use the definition in Tacconi et al. (2006), v_c = 0.67 × v_FWHM = 410 km s⁻¹ (v_FWHM = 613 ± 30 km s⁻¹ measured from the low-J CO(2-1) line; Aravena et al. 2016). Assuming a gas fraction of 0.3–0.8, we utilize the observed SFR of 3600 ± 300 M_☉ yr⁻¹ to¹⁹ derive , which turns out to be in the range of 0.3–0.7. This range is higher than the typical star formation efficiency (0.15–0.2) observed in DSFGs (Tacconi et al. 2006).

We further examine the SFR surface density ${{\rm{\Sigma }}}_{{\mathrm{SFR}}_{\max }}$ expected from this framework;

$$\begin{eqnarray}\begin{array}{rcl}{{\rm{\Sigma }}}_{{\mathrm{SFR}}_{\max }} & = & \displaystyle \frac{{\mathrm{SFR}}_{\max }}{\pi {R}^{2}}\\ & = & 43\left(\displaystyle \frac{\epsilon }{0.1}\right)\left(\displaystyle \frac{{f}_{g}}{0.4}\right)\left(\displaystyle \frac{{{\rm{\Sigma }}}_{\mathrm{tot}}}{5000}\right){\left(\displaystyle \frac{R}{2}\right)}^{-1}({M}_{\odot }\,{\mathrm{yr}}^{-1}\,{\mathrm{kpc}}^{-2}),\end{array}\end{eqnarray} \tag{ 3 }$$

where Σ_tot is the total (dynamical) mass density within radius R, in units of M_☉ pc⁻². For SPT0346-52, R = 1.8 ± 0.2 kpc and M_tot = (1.5 ± 0.2) × 10¹¹ M_☉, derived in a spatially resolved CO imaging study by Spilker et al. (2015). The observed Σ_SFR = 1540 ± 130 M_☉ yr⁻¹ kpc⁻², in²⁰  which the dust emission (i.e., the stellar emission reprocessed by dust) is distributed in a compact area with an effective radius of 0.61 ± 0.03 kpc, surpasses the theoretical ${{\rm{\Sigma }}}_{{\mathrm{SFR}}_{\max }}$ by at least a factor of ∼2. It is one of the highest star-formation densities of any known galaxy in the universe (Rujopakarn et al. 2011; Diamond-Stanic et al. 2012), although the nuclei of Arp 220 have Σ_SFR ∼ 10⁴ M_☉ yr⁻¹ kpc⁻² (Barcos-Muñoz et al. 2015). Figure 6 shows how SPT0346-52 compares in SFR surface density to other starburst galaxies (black) and starbursts found in quasar host galaxies (green) at z > 3 in the literature. SPT0346-52 has an order of magnitude higher SFR than most other sources that lie within a factor of a few in star-formation surface density. SPT0346-52 stands out as the most extreme source at high redshift.

If the starburst is supported by radiation pressure on dust grains in a disk, ${L}_{\mathrm{IR}}$ is consistent with the Eddington-limited luminosity, but the SFR surface density is above the ∼1000 M_☉ yr⁻¹ kpc⁻² theoretical limit by Thompson et al. (2005). The Eddington limit depends upon opacity, and the observed SFR for SPT0346-52 lies in between the SFR_max,thin for the optically thin limit and SFR_max,thick for the optically thick limit from Younger et al. (2008). The observed brightness temperature (∼50 K) is comparable to the dust temperature, which suggests that the gas may be approaching the optically thick regime (Murray et al. 2005; Younger et al. 2008). Thus, the star formation activity in this source, while very vigorous, may still be sub-Eddington. It is natural to suspect that the radiation pressure would drive outflows of cold dusty gas, which have been commonly observed from starbursting galaxies including ultra-luminous IR galaxies (e.g., Martin 2005; Spoon et al. 2013; Veilleux et al. 2013).

Figure 5 demonstrates the allowed and f_gas in the theoretical framework of "maximum starbursts." This observed extremely high Σ_SFR may be explained by an especially high star-formation efficiency ( > 0.4) relative to what has been observed in DSFGs. The gas fraction is constrained to be at least 40%. So far we have only detected a handful of DSFGs at z > 5, only a few hundred million years from the Big Bang (Capak et al. 2011; Combes et al. 2012; Walter et al. 2012; Riechers et al. 2013; Strandet et al. 2016). These sources are expected to harbor larger gas reservoirs available for star formation and be able to sustain a more elevated star-formation efficiency than typical DSFGs (Béthermin et al. 2015). Gas fractions of SPT DSFGs derived using low-J CO observations are in the range of 0.3–0.8 (Aravena et al. 2016). Bothwell et al. (2016) found that SPT DSFGs (SPT0346-52 is not in this sample) on average have very high gas fractions (f_gas ∼ 0.6–0.8) based on atomic carbon observations. The very high gas fractions could raise the Σ_SFR without invoking extremely high star-formation efficiency.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Emission from SPT0346-52 has proved to be beyond the reach of our existing HST and Spitzer data (Ma et al. 2015), which has prevented a detailed characterization of the established stellar mass. Thus, SPT0346-52 would be an ideal object for follow-up observations with JWST.

In this paper we presented a pilot X-ray observation of a single extreme star-forming galaxy at high redshift with Chandra. Studying a sample of such vigorously star-forming galaxies in the early universe will help us constrain the formation and co-evolution of the massive galaxies and SMBHs.