Submitted to ApJ 2014-07-16, accepted 2015-01-12 A Preprint typeset using LTEX style emulateapj v. 03/07/07 CONNECTING GRBs AND ULIRGs: A SENSITIVE, UNBIASED SURVEY FOR RADIO EMISSION FROM GAMMA-RAY BURST HOST GALAXIES AT 0 <z< 2.5 D. A. Perley1,2,*, R. A. Perley3, J. Hjorth4, M. J. Micha lowski5, S. B. Cenko6,7, P. Jakobsson8, T. Kruhler¨ 9, A. J. Levan10, D. Malesani4, and N. R. Tanvir11 Submitted to ApJ 2014-07-16, accepted 2015-01-12 ABSTRACT Luminous infrared galaxies and submillimeter galaxies contribute significantly to stellar mass as- sembly and provide an important test of the connection between the gamma-ray burst rate and that of overall cosmic star-formation. We present sensitive 3 GHz radio observations using the Karl G. Jansky Very Large Array of 32 uniformly-selected GRB host galaxies spanning a redshift range from 0 <z < 2.5, providing the first fully dust- and sample-unbiased measurement of the fraction of GRBs originating from the Universe’s most bolometrically luminous galaxies. Four galaxies are de- −1 tected, with inferred radio star-formation rates ranging between 50–300 M⊙yr . Three of the four detections correspond to events consistent with being optically-obscured “dark” bursts. Our overall detection fraction implies that between 9–23% of GRBs between 0.5 <z < 2.5 occur in galaxies −1 −1 with S3GHz > 10µJy, corresponding to SFR > 50 M⊙yr at z ∼ 1 or > 250 M⊙yr at z ∼ 2. Similar galaxies contribute approximately 10–30% of all cosmic star-formation, so our results are consistent with a GRB rate which is not strongly biased with respect to the total star-formation rate of a galaxy. However, all four radio-detected hosts have stellar masses significantly lower than IR/submillimeter-selected field galaxies of similar luminosities. We suggest that the GRB rate may be suppressed in metal-rich environments but independently enhanced in intense starbursts, producing a strong efficiency dependence on mass but little net dependence on bulk galaxy star-formation rate. Subject headings: gamma-ray burst: general—galaxies: starburst—radio continuum: galaxies 1. INTRODUCTION 2012; Robertson & Ellis 2012; Salvaterra et al. 2013; One of the primary appeals of the study of long- Trenti et al. 2015). duration gamma-ray bursts (GRBs) lies in their Central to the utility of GRBs for this purpose is their potential to address broader questions of cosmology extreme luminosity at all electromagnetic wavelengths, and galaxy evolution. As the explosions of massive including wavelengths unaffected by absorption due to stars at cosmological distances, GRBs are intimately dust and gas. In particular, GRBs are first detected in hard X-rays and gamma-rays and are now routinely lo- connected with cosmic star-formation, and the evolution ′′ of the cosmic GRB rate and the changing demographics calized to ∼2 precision in soft X-rays using the Swift of their host galaxies with time should reflect overall X-ray Telescope (Burrows et al. 2005), meaning that— cosmological trends and inform our understanding of with sufficient follow-up effort—their host-galaxy popu- how galaxies and the stars within them form and evolve lation and redshift distribution can be constructed inde- over the Universe’s history (e.g., Hogg & Fruchter pendent of the effects of obscuration (e.g., Hjorth et al. 1999; Blain & Natarajan 2000; Porciani & Madau 2012; Jakobsson et al. 2012; Perley et al. 2013). 2001; Firmani et al. 2004; Natarajanet al. 2005; For this reason, one particular question in galaxy Kistler et al. 2008; Butler et al. 2010; Tanvir et al. evolution which GRBs showed significant promise to help resolve is the relative importance of extremely arXiv:1407.4456v3 [astro-ph.HE] 3 Feb 2015 1 Department of Astronomy, California Institute of Technology, luminous, dusty galaxies to cosmic star-formation MC 249-17, 1200 East California Blvd, Pasadena CA 91125, USA (e.g., Djorgovski et al. 2001; Ramirez-Ruiz et al. 2002; 2 Hubble Fellow Berger et al. 2003). The UV/optical tracers by which 3 National Radio Astronomy Observatory, P.O. Box O, Socorro, NM, 87801 galaxy and cosmic star-formation rates are normally es- 4 Dark Cosmology Centre, Niels Bohr Institute, Copenhagen, timated are significantly impacted by interstellar dust. Denmark Most galaxies are predominately optically thin and 5 Scottish Universities Physics Alliance, Institute for Astronomy, the effects of dust can be corrected for via measure- University of Edinburgh, Royal Observatory, Edinburgh, EH9 3HJ, UK ments of the UV spectral slope or Balmer decrement 6 NASA/Goddard Space Flight Center, Greenbelt, MD 20771 (Meurer et al. 1999; Smolˇci´cet al. 2009). However, the 7 Joint Space Science Institute, University of Maryland, College most bolometrically luminous galaxies (such as submil- Park, MD 20742 limeter galaxies [SMGs] and local ultra-luminous infrared 8 Centre for Astrophysics and Cosmology, Science Institute, Uni- versity of Iceland, Dunhagi 5, 107 Reykjav´ık, Iceland galaxies [ULIRGs]) contain sufficient dust so as to be 9 European Southern Observatory, Alonso de C´ordova 3107, Vi- optically thick at UV and optical wavelengths, caus- tacura, Casilla 19001, Santiago 19, Chile ing optical/UV-based tracers to inevitably underpredict 10 Department of Physics, University of Warwick, Coventry CV4 the true star-formation rates of these galaxies even after 7AL, UK 11 Department of Physics and Astronomy, University of Leices- dust correction (Blain et al. 2002; Goldader et al. 2002; ter, Leicester LE1 7RH, UK Chapman et al. 2005). Instead, the star-formation rates * e-mail: [email protected] . of these objects (and therefore their contribution to the 2 Perley et al. cosmic SFR) must be measured using long-wavelength have been either limited in scope (observing only small tracers such as far-IR or radio continuum, but these samples), limited in sensitivity (insensitive to even very methods have historically suffered from limited sensi- luminous galaxies beyond z > 1), or subject to uncertain tivity, uncertain calibration, uncertain dust tempera- selection biases (most commonly a bias in favor of tures, and AGN contamination. Estimates of the impor- GRBs with afterglow-based redshift determinations that tance of the most luminous and dusty galaxies to star- will systematically miss dust-obscured events, although formation have varied widely between different studies— other efforts have specifically targeted only the most in part due simply to varying definitions of what dis- heavily dust-obscured GRBs). tinguishes a “luminous” and dusty galaxy from an “or- In this paper, we present results of the first GRB host dinary” one—but have ranged from estimates of 50% galaxy survey that is simultaneously unbiased in regards or more (P´erez-Gonz´alez et al. 2005; Micha lowski et al. to target selection, sensitive enough to detect the long- 2010; Magnelli et al. 2013) down to only about 10% wavelength emission from luminous star-forming galax- (Rodighiero et al. 2011); see Casey et al. (2014) for a re- ies even out to z ∼ 2.5, and large enough to usefully view. statistically constrain the fraction of GRBs that origi- In principle, the contribution of dusty star-forming nate in such systems. Specifically, we survey a sample of galaxies (or any other galaxy population) to overall cos- 32 uniformly-selected GRB host galaxies (a factor of ∼ mic star-formation could be determined simply by mea- 2 larger than any previous long-wavelength survey) to a suring the fraction of GRBs hosted within such galaxies: limiting radio flux density of approximately 10µJy at 3 12 for example, if ULIRGs (galaxies with LIR > 10 L⊙) GHz (a factor of 2–3 deeper than any other radio host represent 40% of all star-formation at 1 < z < 2 survey)—detecting four in total and providing the first they should also produce 40% of all GRBs over the definitive measurement of the fraction of GRBs produced same redshift interval. However, this would strictly ap- by the Universe’s most luminous galaxies. Our sample ply only if GRBs represent an unbiased tracer of over- selection, observations, and analysis are discussed in §2. all star-formation. A large volume of evidence now We present our results and interpret our four detections suggests they do not: GRB host galaxies at z . as star-formation-associated emission from the GRB host 1.5 have lower masses, lower metallicities, bluer col- galaxies in §3. Further discussion of individual systems, ors, and more irregular morphologies compared to what including a detailed discussion of all detections and a few would be expected from an unbiased SFR-tracing pop- notable nondetections, is presented in §4 and §5. The ulation or when compared to the core-collapse super- overall statistical properties of our sample and its impli- nova host population (Fruchter et al. 2006; Modjaz et al. cations are discussed in §6, and conclusions are summa- 2008; Levesque et al. 2010; Graham & Fruchter 2013; rized in §7. Kelly et al. 2014; Vergani et al. 2014), even when includ- 2. SAMPLE AND OBSERVATIONS ing dust-obscured “dark” bursts (Perley et al. 2013, al- though c.f. Hunt et al. 2014). 2.1. Sample Selection Nevertheless, measuring the contribution of very lu- Given the heterogeneous nature of previous long- minous, dusty galaxies to the GRB rate remains im- wavelength efforts and the need to determine the frac- portant for understanding the overall link between tion of luminous hosts even among both obscured and GRBs and cosmic star-formation, and the influence unobscured GRBs (a few dusty and luminous galaxies of dust on our view of each. What causes GRBs have been reported hosting even the latter), the primary to favor certain environments over others is not well- consideration guiding the choice of targets was the need understood.