An imperfectly passive nature: Bright sub-millimeter emission from dust-obscured star formation in the z=3.717 "passive" system, ZF20115

TL;DR

Deep multi-wavelength observations reveal that the high-redshift galaxy ZF20115, previously thought to be quiescent, hosts an obscured starburst, emphasizing the importance of sub-millimeter data in classifying galaxy activity.

Contribution

This study demonstrates that dust-obscured star formation can mimic post-starburst features, challenging previous classifications of high-redshift galaxies as quiescent.

Findings

A luminous starburst is located close to the galaxy's optical emission.

Obscured star formation significantly affects stellar mass estimates.

Balmer features are not exclusive to post-starburst galaxies in dusty environments.

Abstract

The identification of high-redshift massive galaxies with old stellar populations may pose challenges to some models of galaxy formation. However, to securely classify a galaxy as quiescent, it is necessary to exclude significant ongoing star formation, something that can be challenging to achieve at high redshift. In this letter, we analyse deep ALMA/870um and SCUBA-2/450um imaging of the claimed "post-starburst" galaxy ZF-20115 at z=3.717 that exhibits a strong Balmer break and absorption lines. The far-infrared imaging reveals a luminous starburst located 0.4+/-0.1 arcsec (~3kpc in projection) from the position of the rest-frame ultra-violet/optical emission, with an obscured star-formation rate of 100 Mo/yr. This star-forming component is undetected in the rest-frame ultraviolet but contributes significantly to the lower angular resolution photometry at restframe wavelengths >3500A, significantly complicating the determination of a reliable stellar mass. Importantly, in the presence of dust obscuration, strong Balmer features are not a unique signature of a post-starburst galaxy and are indeed frequently observed in infrared-luminous galaxies. We conclude that the ZF20015 system does not pose a challenge to current models of galaxy formation and that deep sub-/millimeter observations are a prerequisite for any claims of quiescence. The multi-wavelength observations of ZF20115 unveil a complex system with an intricate and spatially-varying star-formation history. ZF20115 demonstrates that understanding high-redshift obscured starbursts will only be possible with multi-wavelength studies that include high-resolution observations, available with the JWST, at mid-infrared wavelengths.

Figures

Full text

1. Introduction

In the local Universe, the most massive galaxies are giant spheroidal galaxies that formed the bulk of their stellar populations in a burst of star formation at $z$ \mathrel{\raise 1.50696pt\hbox{\scriptstyle>}\kern-6.00006pt\lower 1.72218pt\hbox{{\scriptstyle\sim}}} 2 (e.g. Nelan et al. 2005). Identifying the progenitors of these galaxies, in either a high-redshift passive or starburst phase, has become a major focus of galaxy formation surveys (e.g. Simpson et al. 2014; Straatman et al. 2014).

One route to isolate passive galaxies at high redshifts is to search for galaxies that have extremely red colors (e.g. $H$–4.5 $\mu$m $>$ 4). These colors may arise due to the presence of a redshifted Balmer (3646 $\mathrm{\AA}$) or 4000 $\mathrm{\AA}$ break in the spectral energy distribution (SED) of the source. However, the degeneracy between stellar age, redshift and dust extinction means that proposed “passive” samples, selected on apparent colors, suffer high (\mathrel{\raise 1.50696pt\hbox{\scriptstyle>}\kern-6.00006pt\lower 1.72218pt\hbox{{\scriptstyle\sim}}} 80 %) contamination from dusty interlopers at low– and high–redshift (Smail et al., 2002b; Toft et al., 2005; Boone et al., 2011; Caputi et al., 2012). Indeed, an early attempt to identify a $z$ $>$ 3 passive galaxy was presented by Mobasher et al. (2005), who claimed the detection of a bright post-starburst galaxy at $z$ $\sim$ 6.5. Dunlop et al. (2007) subsequently showed that allowing for extreme values of dust obscuration in the SED model yields a high likelihood that the source is a $z$ $\sim$ 2 dusty starburst.

Despite these challenges a number of studies have continued to claim the detection of large numbers of passive galaxies at high redshift from wide-field near-infrared imaging (Marchesini et al. 2010; Nayyeri et al. 2014). The most massive of these are thought to have stellar masses of \mathrel{\raise 1.50696pt\hbox{\scriptstyle>}\kern-6.00006pt\lower 1.72218pt\hbox{{\scriptstyle\sim}}} 10 11 $\mathrel{\rm M_{\odot}}$ at z\mathrel{\raise 1.50696pt\hbox{\scriptstyle>}\kern-6.00006pt\lower 1.72218pt\hbox{{\scriptstyle\sim}}} 3 that, if correct, may pose challenges for models of galaxy formation. However a convincing spectroscopic confirmation of a truly massive, quiescent galaxy at z\mathrel{\raise 1.50696pt\hbox{\scriptstyle>}\kern-6.00006pt\lower 1.72218pt\hbox{{\scriptstyle\sim}}} 3 has yet to be presented.

Recently, Glazebrook et al. (2017) presented deep near-infrared spectroscopy of ZF 20115, a purported “passive” galaxy at $z$ = 3.717. ZF 20115 was selected from the ZFOURGE survey based on the presence of a strong Balmer break identified in near-infrared photometry (Straatman et al., 2014). Near-infrared spectroscopy then confirmed Balmer absorption lines with high equivalent width (EW) that were suggested to show that ZF 20115 is a “post-starburst” galaxy with a stellar age of 0.2–1 Gyr. The Balmer lines, combined with fits to the broad-band photometry, led Glazebrook et al. (2017) to conclude in favour of an age in the range 0.5–1 Gyr, corresponding to a formation redshift $z_{\mathrm{form}}\sim$ 5–8. Combined with the estimated stellar mass of 1.5–1.8 $\times$ 1011 $\mathrel{\rm M_{\odot}}$ this indicates a rapid conversion of baryons into stars at high redshift, as expected from studies of sub-millimeter galaxies (SMGs; e.g. Lilly et al. 1999; Smail et al. 2002a).

In this letter we analyse sub-millimeter observations of ZF 20115 with SCUBA-2 and ALMA which identify an intense, obscured starburst within 0.4 $\pm$ 0.1 ′′ of the restframe ultraviolet component (ZF 20115–UV). Throughout we adopt a $\Lambda$CDM cosmology with $H_{\rm 0}$ = 70 km s*-1* Mpc*-1*, $\Omega_{\Lambda}$ = 0.7, and $\Omega_{\rm m}$ = 0.3 and a Chabrier initial mass function (Chabrier, 2003).

2. Observations

The galaxy ZF 20115, located in the CANDELS region of the COSMOS field, was selected based on its rest-frame optical color in the ZFOURGE survey and we use this photometric catalogue in our analysis (Straatman et al., 2016).

ZF 20115 was observed in ALMA Cycle 2 at 870 $\mu$m for 1.4 min as part of program 2013.1.01292.S. The ALMA observations employed 39 antennae yielding a synthesized beam of 1.1*′′* $\times$ 0.6*′′, reach a depth of $\sigma_{870}$ = 0.2 mJy beam-1* and reveal a significant (6.9 $\sigma$) source (ZF 20115–FIR) within 0.4 $\pm$ 0.1*′′* of ZF 20115–UV. We use CASA / imfit to model the emission and determine that ZF 20115–FIR is unresolved with a total flux density of 1.4 $\pm$ 0.2 mJy beam*-1*.

The central area of the CANDELS / COSMOS is being mapped at 450 and 850 $\mu$m by the SCUBA-2 Ultra Deep Imaging EAO Survey (STUDIES: PI: W.-H. Wang), a large program at the James Clerk Maxwell Telescope (JCMT). We make use of the first STUDIES release, which reaches a depth of $\sigma_{450}$ $\sim$ 1 mJy in the vicinity of the ZF 20115 111Including archival data from Geach et al. (2017).. We detect ZF 20115–FIR at a 3 $\sigma$ significance level in the 450 $\mu$m imaging, identifying a $S_{450}$ = 3.1 $\pm$ 1.0 mJy source within the expected 1– $\sigma$ positional uncertainty (2.1*′′*; Ivison et al. 2007). ZF 20115–FIR is detected in the STUDIES SCUBA-2 850 $\mu$m imaging with a flux density of $S_{850}$ = 1.49 $\pm$ 0.15 mJy, consistent with the ALMA detection.

The CANDELS / COSMOS region was imaged with Herschel / PACS and SPIRE as part of the Herschel–CANDELS survey and Herschel Multi-tiered Extragalactic Survey (HerMES; Oliver et al. 2012). Straatman et al. (2014) present the PACS photometry for ZF 20115, with stated 1–$\sigma$ uncertainties of 0.4 mJy at 100 and 160 $\mu$m. At 160 $\mu$m the claimed depth is below the measured confusion limit of PACS but it is not possible to verify this as the reduced Herschel–CANDELS imaging is not publicly available and the data reduction is not detailed in the literature. The deblended SPIRE imaging reaches 1–$\sigma$ depths of 3.1, 3.5 and 4.1 mJy at 250, 350, and 500 $\mu$m (Swinbank et al., 2014). ZF 20115 is not detected in the 100–500 $\mu$m imaging. Finally, ZF 20115 is not detected in the available 1.4 and 3 GHz imaging (Schinnerer et al., 2010; Smolcic et al., 2017), consistent with lying at a redshift of $z$ \mathrel{\raise 1.50696pt\hbox{\scriptstyle>}\kern-6.00006pt\lower 1.72218pt\hbox{{\scriptstyle\sim}}} 2.5.

We align the astrometry of the wide-field imaging presented here by correcting for the mean offset between the Spitzer IRAC / 3.6 $\mu$m image and the relevant image. We cannot apply this technique to the single ALMA pointing but we note that the overall ALMA astrometry has been found to be in agreement with the solution for the CANDELS / COSMOS imaging (Schreiber et al., 2017).

3. Analysis

In Figure 1 we show the archival ALMA 870 $\mu$m data for ZF 20115, contoured on the HST / CANDELS imaging. We first consider that the observed 870 $\mu$m emission (ZF 20115–FIR) is located 0.4 $\pm$ 0.1*′′* ($\sim$ 3 kpc in projection) from the detected rest-frame $UV$ emission. At the redshift of ZF 20115–UV ($z$ = 3.717) the near-infrared wavebands (\mathrel{\raise 1.50696pt\hbox{\scriptstyle<}\kern-6.00006pt\lower 1.72218pt\hbox{{\scriptstyle\sim}}} 2.0 $\mu$m) trace rest-frame ultraviolet emission and thus it is unsurprising to find positional offsets between obscured and unobscured components. Indeed, the measured offset between ZF 20115–UV and ZF 20115–FIR is consistent with Chen et al. (2015), who determine an intrinsic offset between the HST- and ALMA-traced emission in SMGs of 0.55*′′* (1 $\sigma$), evidence for structured dust obscuration in these sources (Chapman et al., 2004; Hodge et al., 2015).

To test whether ZF 20115–UV and ZF 20115–FIR are physically associated we calculate the probability that they arise due to the chance alignment of two sources on the sky. First, we create mock catalogues with source surface densities that satisfy the ZFOURGE color-criterion for a massive, quiescent galaxy (Straatman et al., 2014), and sub-millimeter sources at $S_{870}>$ 1 mJy (Dunlop et al., 2017). We do not account for any bias due to gravitational lensing but comment that this is unlikely given that ZF 20115 is located at $z$ = 3.717. From these mock catalogues we determine a probability of $\sim$ 1.5 $\times$ 10*-4* that ZF 20115–UV and ZF 20115–FIR are a chance alignment, in agreement with the corrected Poissonian probability (Downes et al. 1986). Given this low probability, and the similarity with the properties of high-redshift dusty starbursts, we conclude that ZF 20115 is a composite dusty starburst.

3.1. Properties of the starburst

To determine the far-infrared properties of ZF 20115–FIR we fit the far-infrared photometry with a single-temperature, optically-thin, modified blackbody (mBB) function. The 100 $\mu$m photometry (restframe $\sim$ 20 $\mu$m) is dominated by Polycylic Aromatic Hydrocarbon features and is not included in our SED fitting. The dust emissivity is fixed at $\beta=$ 1.8 and the best-fit parameters and associated uncertainties are determined using a Monte-Carlo Markov Chain (MCMC) approach (Simpson et al., 2016), accounting for the effect of the Cosmic Microwave Background following da Cunha et al. (2013).

Our SED fitting shows that ZF 20115–FIR has a best-fit dust temperature of $T_{\mathrm{d}}$ = 30.4${}^{+1.4}_{-12.0}$ K and a far-infrared luminosity (8–1000 $\mu$m) of $L_{\mathrm{FIR}}$ = 7.7${}^{+1.0}_{-5.2}$ $\times$ 1011 $\mathrel{\rm L_{\odot}}$. A single temperature mBB is known to underestimate the total far-infrared luminosity by $\sim$ 20 % (Swinbank et al. 2014) and correcting for this we determine that ZF 20115 has a total far-infrared luminosity $L_{\mathrm{FIR}}$ = 9.2${}^{+1.2}_{-6.2}$ $\times$ 1011 $\mathrel{\rm L_{\odot}}$ and an obscured star-formation rate (SFR) of 100${}^{+15}_{-70}$ $\mathrel{\rm M_{\odot}\,yr^{-1}}$ (Kennicutt, 1998).

4. Discussion

Interest in ZF 20115 (e.g. Rong et al. 2017) has arisen due to the claim that its stellar population is massive and quiescent at $z$ = 3.717 (Glazebrook et al., 2017). The presence of an obscured starburst, blended or co-located, with the optical-to-near-infrared emission requires a re-assessment of the properties of ZF 20115.

4.1. Origin of Balmer Absorption Lines

Near-infrared spectroscopy of ZF 20115 identified Balmer absorption lines with a combined EW of 38 $\pm$ 6 $\mathrm{\AA}$ (summing H$\beta$, H$\gamma$, and H$\delta$; Glazebrook et al. 2017). This requires that the detectable stellar continuum around 4000–5000$\mathrm{\AA}$ is dominated by A-stars with lifetimes of $\sim$ 0.1–1 Gyr and is often taken as an indicator of a “post-starburst” population. However, Balmer absorption features are not unique to post-starburst galaxies; they appear frequently in dusty starburst galaxies as well (Poggianti & Wu, 2000).

Poggianti & Wu (2000) present detections of the H $\delta$ absorption line for a sample of IRAS–selected galaxies ($L_{\mathrm{FIR}}$ $>$ 5 $\times$ 1011 $\mathrel{\rm L_{\odot}}$) at $z$ \mathrel{\raise 1.50696pt\hbox{\scriptstyle<}\kern-6.00006pt\lower 1.72218pt\hbox{{\scriptstyle\sim}}} 0.05. This far-infared-bright sample is dominated by star-forming galaxies, and over 60 % of the sources show evidence of on-going interactions or mergers. In Figure 2 we show the SED of Mrk 331 (Brown et al., 2014) from the sample of Poggianti & Wu (2000), scaled in luminosity to match ZF 20115. Mrk 331 is a late-stage merger and exhibits H$\delta$ absorption with EW = 4.1$\mathrm{\AA}$ and a strong break at \mathrel{\raise 1.50696pt\hbox{\scriptstyle<}\kern-6.00006pt\lower 1.72218pt\hbox{{\scriptstyle\sim}}} 4000 $\mathrm{\AA}$. We note that few local examples reach the high EW Balmer absorption observed from ZF 20115 but comment that this may be a reflection of the low survey volumes and decreased activity at $z$ $\sim$ 0. The SED of Mrk 331 is in qualitative agreement with that of ZF 20115, illustrating the challenge of deriving accurate star-formation histories for high-redshift “red” galaxies from optical-to-near-infrared photometry alone.

The high EW Balmer absorption features that appear common in starburst galaxies can be explained by age-dependant dust obscuration (Poggianti & Wu, 2000). This can be considered in terms of a single isolated galaxy, or triggered star-formation in an ongoing merger. Given the presence of two components with a projected offset of 3 kpc we suggest that the ZF 20115 system is an ongoing merger between two gas-rich progenitors. In this scenario the centroid of the observable stellar emission likely traces one component of the merger that underwent a burst of star formation, triggered by an earlier interaction.

Hopkins et al. (2013) present numerical simulations of the merger of gas-rich disks at high-redshift, which demonstrate that star-formation is triggered in the individual disks $\sim$ 0.5 Gyr before coalescence, and the triggering of a nuclear starburst. The Balmer absorption in the spectrum of ZF 20115 is likely the consequence of a similar merger event, with the strength of the lines indicating a minimum age for this initial star-formation episode of $\sim$ 100 Myr (González Delgado et al., 1999; Glazebrook et al., 2017).

4.2. Stellar mass of ZF 20115

The detection by ALMA of a strongly star-forming, obscured component within 0.4 $\pm$ 0.1*′′* of the claimed quiescent component ZF 20115–UV significantly complicates the analysis of this system. To some extent the issue about whether the two components are part of a single “galaxy” or are distinct components within a larger halo, is one of semantics. What is unarguable is that the presence of this luminous, star-forming component will result in significant contamination of the flux measurements in longer wavelength bands, which are critical for estimating the stellar mass of the system. At the redshift of ZF 20115, the stellar emission from the obscured starburst peaks in the 2*′′*–resolution Spitzer / IRAC imaging (rest-frame 0.8–1.7 $\mu$m), where the two components are strongly blended (Figure 1). At these wavelengths, the mass-to-light ratio of a stellar population varies by an order of magnitude for ages of 50–1000 Myr (Hainline et al., 2011) and so it is crucial that we take the contribution from ZF 20115–UV into account.

We use two methods to estimate the likely contribution from ZF 20115–FIR to the integrated photometry. First, we deblend the 3.6 / 4.5 $\mu$m emission using three 2-D Gaussian components representing ZF 20115–UV, ZF 20115–FIR, and the northern optically-bright galaxy (Figure 1). Using an MCMC approach we determine that ZF20115–FIR contributes $\sim$ 40-–50$\%$ of the integrated flux density at 3.6 and 4.5 $\mu$m, consistent with the two components being indistinguishable in the IRAC imaging. Second, we estimate the contribution by considering the spectroscopically-identified, ALMA-located–SMGs from Danielson et al. (2017). We fit the templates for these SMGs to the far-infrared photometry of ZF 20115–FIR and in Figure 2 show those SEDs that are near-infrared–bright and have far-infrared properties matching ZF 20115–FIR. We create an average SED from these best-fit templates and estimate that the obscured starburst contributes 20–75 % (median 33 $\pm$ 4 %) of the measured photometry of ZF 20115 at observed 1.5–8.0 $\mu$m, broadly consistent with our estimates from deblending the IRAC imaging.

To estimate the mass of the stellar component associated with ZF 20115–UV we subtract the average starburst SED from the global photometry. Using magphys (da Cunha et al., 2015) to model the residual photometry we estimate that this component has a stellar mass of $\sim$ 0.8 $\times$ 1011 $\mathrel{\rm M_{\odot}}$, less than half of that estimated by Glazebrook et al. (2017). We cannot determine an accurate star-formation history for this component but note that a rapid formation at high redshift ($z$ \mathrel{\raise 1.50696pt\hbox{\scriptstyle>}\kern-6.00006pt\lower 1.72218pt\hbox{{\scriptstyle\sim}}} 5) is no longer required and that the implied SFR and duration are consistent with the population of obscured starbursts that are known to reside at $z$ \mathrel{\raise 1.50696pt\hbox{\scriptstyle>}\kern-6.00006pt\lower 1.72218pt\hbox{{\scriptstyle\sim}}} 4 (e.g. Riechers et al. 2013; Strandet et al. 2016; Simpson et al. 2016). Using the stellar mass estimates presented by Danielson et al. (2017) we estimate that ZF 20115–FIR has a mass of $\sim$ 0.4 $\times$ 1011 $\mathrel{\rm M_{\odot}}$. At the current SFR and assuming a constant star-formation history this stellar component could form in $\sim$ 350 Myr, equivalent to a formation redshift of $z$ = 3.7–4.6.

The apparent quiescent nature of ZF 20115 led Glazebrook et al. (2017) to suggest that this source, and the parent sample of proposed quiescent galaxies, may be in conflict with current models of galaxy formation (e.g. Steinhardt et al. 2016). The contamination of the stellar mass estimates for ZF 20115–UV has implications for the conclusions presented by Glazebrook et al. (2017) and we revisit those here.

First, we adopt the halo mass of $\sim$ 3 $\times$ 1012 $\mathrel{\rm M_{\odot}}$ for ZF 20115 estimated by Glazebrook et al. (2017). Assuming a cosmic baryon fraction of 16% (Planck Collaboration et al., 2016), we estimate that the stellar mass for ZF 20115–UV corresponds to a baryon conversion efficiency of $\sim$ 15 –40 $\%$, at a formation redshift of $z_{\mathrm{form}}$ = 4–5. Genel et al. (2014) demonstrate that comparable halos in the Illustris simulation have a conversion efficiency of $\sim$10 $\%$ at $z$ $\sim$4, consistent with observational studies of $UV$–selected galaxies (Finkelstein et al. 2015). The conversion of baryons into stars appears to be progressing more efficiently in ZF 20115–UV, although it remains well below the theoretical limit. Indeed, simulations of galaxy formation do predict the presence of $z$ $\sim$ 4 star-forming systems with stellar masses of $\sim$ 1011 $\mathrel{\rm M_{\odot}}$ (Wellons et al. 2015; Rodriguez-Gomez et al. 2016), at a space density in agreement with observational studies (e.g. Mortlock et al. 2017). Thus, while ZF 20115 remains an interesting system it does not appear to be in conflict with our understanding of galaxy formation.

5. Conclusions

In this letter we have analysed ALMA and SCUBA-2 detections of far-infrared emission from ZF 20115, a proposed massive, quiescent galaxy at $z$ = 3.717. The far-infrared imaging locates an obscured starburst only 0.4 $\pm$ 0.1*′′* from the unobscured ZF 20115–UV. The far-infrared luminous starburst appears associated with ZF 20115–UV bringing into doubt the claim that this is a quiescent system. Indeed, we show that the Balmer absorption features exhibited by ZF 20115 are not a unique tracer of a “post-starburst” galaxy, such features appear frequently in local IRAS–selected starbursts (Poggianti & Wu, 2000).

ZF 20115–UV and ZF 20115–FIR are separated by $\sim$ 3 kpc on the sky, a strong indication that they reside within the same halo at $z$ = 3.717. Indeed, HST studies of ALMA-identified SMGs identify similar offsets between the obscured and unobscured components of comparably luminous galaxies at these redshifts. What is unavoidable is that the presence of a starburst within 0.4 $\pm$ 0.1*′′* of ZF 20115–UV means that an allowance must be made for blending in the low-spatial resolution SED of this system, particularly at long wavelengths.

Including a dusty starburst component in the SED fitting of ZF 20115 indicates that $\sim$ 30–50 $\%$ of the total rest-frame 1–2 $\mu$m emission arises from the on-going starburst. Correcting for this contribution and refitting the SED we determine that ZF 20115–UV has a stellar mass of $\sim$ 0.8 $\times$ 1011 $\mathrel{\rm M_{\odot}}$, although we stress that the uncertainties on this measurement are large. The corrected stellar mass is $\sim$ 50 $\%$ lower than previously published values and as we show this removes the tension between ZF 20115 and models of galaxy formation.

The multi-wavelength imaging and spectroscopy of the ZF 20115 system provides key insights into the properties of obscured starbursts at high redshifts. ZF 20115 comprises a blend of obscured and unobscured components that host intense but, spatially-variable, star-formation. The strength of the Balmer absorption features suggest that the system has a complex star-formation history and was undergoing a star-formation event at least 100 Myrs prior to the current episode. This can be explained by adopting a merger-model for the system but is challenging to understand if secular processes dominate the formation mechanism of high-redshift starbursts. What is clear from our analysis is that to untangle this complex system, and thus truly understand high-redshift starbursts, we require the high-resolution imaging at mid-infrared wavelengths that will only be available with the James Webb Space Telescope.

Acknowledgements

The authors thank Nick Scoville for sharing his catalog of ALMA observations and Alice Shapley for helpful comments. JMS acknowledges an EACOA fellowship. IRS acknowledges support from DUSTYGAL 321334, a RS/Wolfson Merit Award and STFC (ST/L00075X/1). DR acknowledges support under grant number AST-1614213.

This paper makes use of data taken with SCUBA-2 / JCMT, and ALMA (ADS/JAO.ALMA$\#$2013.1.01292.S). The JCMT is operated by the EAO on behalf of The National Astronomical Observatory of Japan, Academia Sinica Institute of Astronomy and Astrophysics, the Korea Astronomy and Space Science Institute, the National Astronomical Observatories of China and the Chinese Academy of Sciences, with additional support from the Science and Technology Facilities Council of the United Kingdom and participating universities in the United Kingdom and Canada.