THE AFTERGLOW AND ULIRG HOST GALAXY OF THE DARK SHORT GRB 120804A

Short-duration gamma-ray bursts (GRBs) occur in a wide range of environments that include elliptical and star-forming galaxies in the field and in clusters (e.g., Berger 2011 and references therein). These galaxies have redshifts of z ≈ 0.1 to 1 (Berger et al. 2007; Rowlinson et al. 2010), star formation rates of SFR 0.01 to ≈40 M_☉ yr⁻¹ (Berger 2009; Perley et al. 2012), and stellar masses of M_* ≈ 10⁹–4 × 10¹¹ M_☉ (Leibler & Berger 2010). These properties are suggestive of a progenitor population that tracks both stellar mass and star formation activity (though with a significant delay of ~0.3 Gyr; Leibler & Berger 2010), in agreement with the popular compact object coalescence model (e.g., Eichler et al. 1989; Narayan et al. 1992).

In a similar vein, short GRBs also exhibit a range of explosion properties, with isotropic-equivalent energies of E_{γ, iso} ~ E_{K, iso} ~ 10⁴⁹–10⁵² erg (Berger 2007; Nakar 2007; Nysewander et al. 2009), jet opening angles of θ_j ≈ few to 20 deg in a few cases (Burrows et al. 2006; Grupe et al. 2006; Soderberg et al. 2006; Watson et al. 2006; Fong et al. 2012), and circumburst densities of n 1 cm⁻³ (Berger et al. 2005; Soderberg et al. 2006; Fong et al. 2012). The distribution of opening angles is of particular importance since it impacts the true energy scale and event rate. To date, all measurements or limits on θ_j have relied on X-ray observations thanks to the relative brightness and high detection fraction of the afterglow in the X-ray band compared to the optical and radio bands (Nysewander et al. 2009; Berger 2010).

Here we present the optical discovery and subarcsecond localization of the optical and X-ray afterglow of the short GRB 120804A, as well as optical, near-IR, and radio detections of its host galaxy. The afterglow data constrain the burst properties (E_{K, iso}, θ_j, n). The host galaxy observations identify it as an ultraluminous infrared galaxy (ULIRG) at a photometric redshift of z ≈ 1.3, making GRB 120804A one of the most distant short bursts known to date. This is the first ULIRG host in the short GRB sample, exceeding even the luminous infrared galaxy (LIRG) likely host of GRB 100206A (Perley et al. 2012). We present the afterglow and host galaxy observations in Section 2, extract the explosion properties in Section 3, and determine the host galaxy photometric redshift and properties in Section 4. Throughout the paper we report magnitudes in the AB system (unless otherwise noted), use a Galactic extinction value of E(B − V) ≈ 0.204 mag (Schlafly & Finkbeiner 2011), and employ the standard cosmological parameters: H₀ = 71 km s⁻¹ Mpc⁻¹, Ω_Λ = 0.73, and Ω_M = 0.27.

GRB 120804A was discovered with the Swift Burst Alert Telescope (BAT) on 2012 August 4 at 00:55:47.8 UT (Baumgartner et al. 2012) and was also detected with Konus-WIND (Sakamoto et al. 2012). The burst duration is T₉₀ = 0.81 ± 0.08 s (15–350 keV) with a fluence of F_γ = (8.8  ±  0.5) × 10⁻⁷ erg cm⁻² (15–150 keV) and (1.45  ±  0.30) × 10⁻⁶ erg cm⁻² (15–1000 keV). The 16 ms peak flux is (6.0 ± 2.7) × 10⁻⁶ erg cm⁻² s⁻¹ (15–1000 keV). A joint analysis of the BAT and Konus-WIND data indicates a peak energy of E_p = 135⁺⁶⁶₋₂₉ keV (Sakamoto et al. 2012). The spectral lags are 16  ±  12 ms (15–25 to 50–100 keV) and −5 ± 6 ms (25–50 to 100–350 keV).

Recently, Bromberg et al. (2012) suggested, based on prompt-emission properties alone, that the 50% probability dividing line for Swift long and short GRBs is about 0.8 s, similar to the duration of GRB 120804A. At the redshift of z ≈ 1.3, which we infer in Section 4, the resulting rest-frame values of E_p (≈310 keV) and the spectral lag (≈9 ms) along with the isotropic-equivalent γ-ray energy (≈6 × 10⁵¹ erg) and the peak luminosity (≈3 × 10⁵² erg s⁻¹) are generally intermediate between the correlations exhibited by short and long GRBs (e.g., Figures 3 and 4 of Zhang et al. 2009). However, the combination of lag and duration most closely matches the short-burst distribution. In the absence of a specific analysis of GRB 120804A along the lines of Bromberg et al. (2012) that can argue for a long-burst origin, we consider GRB 120804A to be in the short-duration category.

Swift/X-ray Telescope (XRT) observations commenced about 78 s after the burst and led to the identification of a fading source, located at R.A. = 15^h35^m4751, decl. = −28°46'569 with an uncertainty of 14 radius (90% containment, UVOT-enhanced; Osborne et al. 2012). Observations with the UV/Optical Telescope (UVOT) began about 97 s after the burst, but no counterpart was detected to a 3σ limit of 21.4 mag in the white filter (at δt ≈ 97–247 s; Chester & Lien 2012). Optical and near-IR observations with GROND starting at δt ≈ 1.5 hr also led to non-detections with 22 mag (griz) and 20.6 mag (the J band; Sudilovsky et al. 2012).

2.1. X-Ray Observations

2.2. Optical Afterglow Discovery and Relative X-Ray Astrometry

We obtained two epochs of i-band imaging with the Gemini Multi-Object Spectrograph (GMOS; Hook et al. 2004) on the Gemini-North 8 m telescope on 2012 August 4.27 and 7.30 UT (δt ≈ 0.23 days and ≈3.26 days, respectively). The observations consisted of 1980 s and 2880 s, respectively, in 065 seeing. We process the data using the gemini package in IRAF and perform photometry using a zero point of 28.46 ± 0.10 mag measured on the nights of August 2–8 UT. We perform digital image subtraction of the two epochs with the ISIS package (Alard 2000) and recover a fading source with m_i = 26.2 ± 0.2 AB mag (with an additional systematic uncertainty of ±0.1 mag due to the zero-point uncertainty), which we consider to be the optical afterglow of GRB 120804A. Corrected for Galactic extinction, the resulting flux density is F_ν = 0.17 ± 0.05 ; μJy at δt ≈ 0.23 days. We note that for a typical power-law decline rate in the optical band, F_ν∝t⁻¹, the expected flux density at δt ≈ 3.26 days is less than 1/10th of the flux density at δt ≈ 0.23 days, and we can therefore safely approximate the flux density in the second epoch as zero. Images of the two epochs and the resulting subtraction are shown in Figure 1.

Zoom In Zoom Out Reset image size

We determine the absolute position of the afterglow by astrometrically matching the images to the 2MASS reference frame using 45 common sources. The resulting astrometric uncertainty is 015 in each coordinate. The afterglow position in the residual image is R.A. = 15^h35^m47479, decl. = −28°46'5617 (J2000), with a centroid uncertainty of about 005 in each coordinate.

To locate the Chandra X-ray afterglow on the optical images, we perform differential astrometry using four common sources. We find a relative offset between the two coordinate frames (Chandra to Gemini) of δR.A. = −0.06 ± 021 and δdecl. = -0.01 ± 017, leading to a refined X-ray afterglow position of R.A. = 15^h35^m47478, decl. = −28°46'5630 with an uncertainty of about 035 radius that takes into account a centroid uncertainty of about 006; see Figure 1). The Chandra position is in excellent agreement with the optical afterglow position.

2.3. Optical/Near-IR Host Galaxy Observations

2.4. Radio Observations

We observed GRB 120804A with the Karl G. Jansky Very Large Array (JVLA) starting on 2012 August 4.97 UT (δt ≈ 0.93 days) at a mean frequency of 5.8 GHz. We utilized the WIDAR correlator (Perley et al. 2011) with a bandwidth of about 1 GHz in each sideband, centered at 4.9 and 6.7 GHz. All observations were undertaken in the B configuration, utilizing 3C 286 for bandpass and flux calibration, with interleaved observations of J1522−2730 for gain calibration. We calibrate and analyze the data using standard procedures in the Astronomical Image Processing System (AIPS; Greisen 2003) and list the resulting flux density measurements in Table 2.

Table 1. Optical and Near-IR Observations of the Host and Nearby Galaxy

Filter	λobs	Host Galaxy		Nearby Galaxy
		AB Magnitudea	Fνa	AB Magnitudea	Fνa
	(μm)		(μJy)		(μJy)
r	0.630	25.9 ± 0.2b	0.16 ± 0.04b	26.0 ± 0.2	0.15 ± 0.03
i	0.779	24.8 ± 0.15	0.44 ± 0.07	24.4 ± 0.1	0.63 ± 0.06
Y	1.021	23.7 ± 0.3	1.20 ± 0.40	23.0 ± 0.2	2.30 ± 0.46
J	1.235	23.0 ± 0.2	2.19 ± 0.44	22.4 ± 0.15	3.98 ± 0.60
Ks	2.146	22.0 ± 0.1	5.75 ± 0.58	21.8 ± 0.1	6.85 ± 0.69

Notes. ^aThese magnitudes and flux densities are based on a 08 radius aperture and have been corrected for Galactic extinction of E(B − V) ≈ 0.204 mag (Schlafly & Finkbeiner 2011). ^bThis value has been corrected for an expected afterglow contribution of ≈0.03 μJy.

Download table as:  ASCIITypeset image

Table 2. JVLA Observations

Date	δt	On-source Time	Frequency	Fν
(UT)	(days)	(minutes)	(GHz)	(μJy)
2012 Aug 4.97	0.93	80	4.9	44 ± 8
			6.7	25 ± 7
2012 Aug 5.99	1.95	80	4.9	40 ± 8
			6.7	26 ± 7
2012 Aug 8.02	3.98	60	4.9	45 ± 9
			6.7	25 ± 8
2012 Aug 14.94	10.90	55	4.9	45 ± 10
			6.7	24 ± 9
2012 Sep 11.01	37.97	35	4.9	40a
			6.7	36a
GRB 100206A
2012 Sep 10.27	946	35	5.8	100b

Notes. ^aLimits are 3σ. ^bLimits are 5σ.

Download table as:  ASCIITypeset image

In the first four epochs (δt ≈ 0.9–10.9 days) we detect a single unresolved source coincident with the optical and X-ray afterglow, as well as the host galaxy: R.A. = 15^h35^m47485 (±0.008), decl. = −28°46'5644 (±0.20). However, the source remains constant in brightness, with flux densities of F_ν(4.9 GHz) = 43 ± 4 μJy and F_ν(6.7 GHz) = 25 ± 4 μJy. The steady brightness and optically thin spectrum, F_ν∝ν^{−1.7 ± 0.8}, are unprecedented for GRB radio afterglows at early time (Granot & Sari 2002) but are instead suggestive of emission from the host galaxy. Indeed, the centroid of the radio source has a smaller offset relative to the optical host galaxy centroid (010) than to the optical afterglow position (028), although we note that even the latter corresponds to about 1σ. The probability of chance coincidence for a source of this brightness within ~1'' of the optical/near-IR host galaxy position is P_cc ≈ 2 × 10⁻⁴, based on the number counts of faint 5 GHz sources (Fomalont et al. 1991). This indicates that the radio source is the host galaxy of GRB 120804A. The observation on 2012 September 11 UT was obtained during a JVLA re-configuration leading to poorer noise characteristics in the resulting image. The 3σ limits are consistent with the earlier detections of the steady source.

We use the lack of variability to place a 3σ upper limit on the radio afterglow brightness of F_ν(5.8 GHz) 20μJy.

In addition, we obtained JVLA observations of the LIRG host galaxy of GRB 100206A (Perley et al. 2012) on 2012 September 10.27 UT to determine whether its large total star formation rate inferred from optical IR data (~40 M_☉ yr⁻¹) produces radio emission. The observing setup and data analysis follow the procedure described above. We observed 3C 48 for flux and bandpass calibration and interleaved observations of J0238+1636 for gain calibration. There is a bright, contaminating source in the field (27 mJy at 4.9 GHz and 9 mJy at 6.7 GHz), located about 45 from the position of GRB 100206A. We therefore image the field utilizing self-calibration techniques on this bright source, leading to a non-detection of radio emission with a conservative 5σ limit of F_ν(5.8 GHz)   80 μJy. The synthesized beam size of 17 × 08 is well matched to the angular size of the host galaxy (Perley et al. 2012).

We model the X-ray data, optical detection, and radio upper limits using the standard afterglow synchrotron model (Granot & Sari 2002). We follow the standard assumptions of synchrotron emission from a power-law distribution of electrons (N(γ)∝γ^−p for γ ≥ γ_m) with constant fractions of the post-shock energy density imparted to the electrons (_e) and magnetic fields (_B). The additional free parameters of the model are the isotropic-equivalent blast-wave kinetic energy (E_{K, iso}) and the circumburst density (n for a constant density medium).

We note three key observational facts to guide the afterglow modeling. First, the X-ray light curve is best fit with a single decline rate of α_X = −0.93 ± 0.06, which, coupled with the spectral index of β_X ≈ −0.9 ± 0.1 (Section 2.1), indicates that the synchrotron cooling frequency is ν_c ~ ν_X and that p ≈ 2.2. Second, the peak of the X-ray afterglow is ≈20 μJy at δt ≈ 200 s. This suggests that the radio light curve will eventually reach a similar peak flux density if the synchrotron peak frequency is ν_m ≈ ν_X at δt ≈ 200 s (i.e., F_{ν, m} ≈ 20 μJy at 200 s). Such a peak flux density is consistent with our inferred radio upper limits. On the other hand, if ν_m is located well below the X-ray band at δt ≈ 200 s, this would imply that F_{ν, m} 20 μJy, thereby violating the radio limits at later time. Finally, the optical and X-ray flux densities at δt ≈ 5.5 hr are comparable, indicating that the optical to X-ray spectral index is β_OX = 0.03 ± 0.06, compared to an expected slope of ≈0.6 (for p = 2.2 and ν_c ~ ν_X). This shallow slope indicates that GRB 120804A can be classified as a "dark" burst, following the definition of Jakobsson et al. (2004).

To guide the reader, we use the constraints listed above along with the synchrotron emission equations in Granot & Sari (2002) to find the following rough constraints on the burst parameters:

$$\begin{equation} \nu _m(200\,{\rm s})\approx \nu _X\Rightarrow E_{K,{\rm iso},52}^{1/2}\, \bar{\epsilon }_e^2\, \epsilon _B^{1/2}\approx 1.1\times 10^{-3}, \end{equation} \tag{ 1 }$$

$$\begin{equation} \nu _c(200\,{\rm s})\approx \nu _X\Rightarrow E_{K,{\rm iso},52}^{1/2}\, n_0\, \epsilon _B^{3/2}\approx 6.9\times 10^{-5}, \end{equation} \tag{ 2 }$$

$$\begin{equation} F_{\nu ,m}(200\,{\rm s})\approx 20\,{\rm \mu Jy}\Rightarrow E_{K,{\rm iso},52}\, n_0^{1/2}\, \epsilon _B^{1/2}\approx 3.0\times 10^{-3}. \end{equation} \tag{ 3 }$$

Combining these equations with the constraint that _e, _B 1/3 indicates that E_{K, iso} ~ 3 × 10⁵¹ erg and n₀ ~ 10⁻³ cm⁻³.

Using the full model from Granot & Sari (2002), we find that the entire data set can be fit with the following parameters (using z = 1.3; see Section 4): E_{K, iso} ≈ 8 × 10⁵¹ erg, n ≈ 10⁻³ cm⁻³, _e ≈ 0.3, _B ≈ 0.1, and p ≈ 2.2, in good agreement with the basic constraints discussed above. The inferred blast-wave energy is comparable to the isotropic-equivalent γ-ray energy, E_{γ, iso} ≈ 6 × 10⁵¹ erg. To explain the suppressed optical emission, we also require A^host_V ≈ 2.5 mag (at z = 1.3). The resulting light curves are shown in Figure 2. Models with higher values of E_{K, iso} and/or n violate the radio limits (e.g., dotted line in Figure 2).

Zoom In Zoom Out Reset image size

In addition, the lack of a break in the X-ray light curve to at least ≈46 days (Burrows et al. 2012) places a lower bound on the jet collimation angle, with θ_j 11°, where we use the values of E_{K, iso} and n inferred above. This indicates a beaming correction factor of f⁻¹_b ≡ [1 − cos(θ_j)]⁻¹ 55 and hence a beaming-corrected energy scale of E_γ + E_K ≈ 2.6 × 10⁵⁰–1.4 × 10⁵² erg, with the upper bound set by isotropy.

Since, in the afterglow model above, the X-ray light curve sets the overall flux density scale, an alternative explanation for the radio non-detections and the low optical flux density is that the X-ray emission is dominated by a different emission mechanism. One possibility is contribution from inverse Compton emission (Sari & Esin 2001), but this requires a large density of n 10² cm⁻³. Another possibility is emission from a newly born rapidly spinning magnetar (e.g., Zhang & Mészáros 2001), but the expected evolution in this case is a relatively constant brightness for a duration similar to the spin-down timescale, followed by F_ν∝t⁻² at later time (for a typical braking index of 3). This is quite distinct from the observed single power law of F_ν∝t⁻¹ at δt 200 s.

The observed optical/near-IR spectral energy distribution (SED) of the host galaxy exhibits a red color of i − K = 2.8 ± 0.25 mag (=4.3 Vega mag); the r − K color is 3.9 ± 0.3 mag (=5.6 Vega mag). There is also noticeable steepening between the i- and Y-band filters, with i − Y = 1.1 ± 0.45 mag. These properties are indicative of a Balmer/4000 Å break at z ≈ 1.3 and either an evolved or dusty stellar population. Using the Maraston (2005) evolutionary stellar population synthesis models to fit the riYJK_s-band data, we infer a photometric redshift of z = 1.3^+0.3_−0.2 (1σ uncertainties); see Figure 3.

Taken in conjunction with the radio detection, the SED is reminiscent of ULIRGs (L_FIR 10¹² L_☉). To investigate this possibility, we compare the host galaxy fluxes to the SED⁹ of the local ULIRG Arp 220. As shown in Figure 4, at z = 1.3 a simple scaling of Arp 220 provides a remarkable fit to the data. An elliptical galaxy template (2 Gyr old population; Polletta et al. 2007) at z = 1.3 provides a reasonable fit in the optical/near-IR (although it underestimates the observed r- and K_s-band fluxes) but cannot explain the radio emission (Figure 4).

At z = 1.3, the rest-frame B-band absolute magnitude is M_B ≈ −20.2 mag, or L_B ≈ 0.2 L* in comparison to the B-band luminosity function at z ~ 1.3–2 (Ilbert et al. 2005). The rest-frame K-band absolute magnitude inferred from the model is M_K ≈ −22.1 mag (−24.0 Vega mag, or L_K ≈ 0.4 L*; Caputi et al. 2006b). This luminosity corresponds to a stellar mass of M_* ≈ 5 × 10¹⁰ M_☉ (using a characteristic mass-to-light ratio of M_*/L_K ≈ 0.3; Drory et al. 2004). The infrared bolometric luminosity scaled using the SED of Arp 220 is L_FIR ≈ 10¹² L_☉.

The unobscured star formation rate is inferred from the observed r band (λ₀ ≈ 2700 Å) to be ≈1 M_☉ yr⁻¹. However, from the observed radio emission we estimate the total star formation rate to be much larger (Yun & Carilli 2002):

$$\begin{equation} {\rm SFR}\approx \frac{F_{\rm \nu ,\mu Jy}\,d_{L,{\rm Gpc}}^2/(1+z)}{25\nu _{\rm 0,GHz}^{-0.75}+0.7\nu _{\rm 0,GHz}^{-0.1}}\approx 300\, M_{\odot} \,{\rm yr}^{-1}, \end{equation} \tag{ 4 }$$

where ν₀ = (1 + z)ν_obs is the rest frequency and d_L is the luminosity distance. Assuming that the host stellar mass has been assembled with this star formation rate gives a characteristic age of about 0.15 Gyr. Using the limit on radio emission from GRB 100206A in Equation (4), we find SFR 50 M_☉ yr⁻¹.

The large star formation rate in the host of GRB 120804A is also expected to produce X-ray emission, with L_X ≈ 7 × 10³⁹ × SFR erg s⁻¹ (Watson et al. 2004; Vattakunnel et al. 2012). For the values inferred above we find an expected luminosity of L_X ≈ 2 × 10⁴² erg s⁻¹, corresponding to a flux of F_X ≈ 4.5 × 10⁻¹⁶ erg s⁻¹ cm⁻². This is about 55 times lower than the afterglow flux measured with XMM-Newton at δt ≈ 19 days and hence consistent with a star formation origin for the radio emission.

An alternative interpretation of the radio detection is emission from an active galactic nucleus (AGN). Matching the standard template of a radio-quiet AGN (Shang et al. 2011) to the observed radio flux density (using z = 1.3) leads to a substantial overestimate of the optical/near-IR brightness and a much bluer color (Figure 4). Similarly, the expected X-ray flux in this scenario is F_X ≈ 6.4 × 10⁻¹⁴ erg cm⁻² s⁻¹, several times brighter than the XMM-Newton detection of the afterglow. Thus, we can rule out a radio-quiet AGN origin. If we instead use the standard radio-loud AGN template (Shang et al. 2011) matched to the radio flux density, we find that the expected optical/near-IR and X-ray fluxes are ~1–2 orders of magnitude fainter than measured (Figure 4). Indeed, the observed lower bound on the ratio of radio to X-ray luminosity, νL_{ν, rad}/L_X 2 × 10⁻⁴, along with an upper bound of L_X 5 × 10⁴⁴ erg s⁻¹ are in good agreement with samples of low-luminosity AGNs (e.g., Terashima & Wilson 2003). In this scenario, the optical/near-IR emission is instead dominated by a stellar component, from either an old population or a reddened young population with a modest star formation rate of a few M_☉ yr⁻¹.

For the radio luminosity and stellar mass of the host galaxy, the fraction of galaxies with radio-loud AGNs is about 2 × 10⁻³ (e.g., Heckman & Kauffmann 2002). Since the presence of a putative AGN should be unrelated to the GRB, this result indicates that the AGN hypothesis has a low likelihood. Similarly, while both a ULIRG and an AGN origin can explain the observed radio emission, we note that the former also offers a natural explanation for the inferred extinction and fits the broadband host SED with a single component. We therefore consider a ULIRG as the more likely explanation for the host galaxy of GRB 120804A.

The offset between the optical afterglow position and host galaxy centroid is 027 ± 015, corresponding to 2.2 ± 1.2 kpc at z ≈ 1.3. This is a relatively small offset for short GRBs (with a median of ≈5 kpc; Fong et al. 2010; Berger 2010), though not unprecedented. Still, in the context of a ULIRG origin, the small offset is consistent with the inferred afterglow extinction.

Finally, we determine a photometric redshift for the galaxy located ≈14 from the host galaxy position and find z = 1.25 ± 0.10, consistent with that of the host galaxy (Figures 3 and 4). Thus, it appears likely that these galaxies, with a projected separation of about 11 kpc, are interacting or merging. This is not unexpected since at least some ULIRG activity is triggered by galaxy mergers (e.g., Sanders & Mirabel 1996).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We presented the discovery and subarcsecond localization of the optical and X-ray afterglow of the short GRB 120804A. A comparison of the observed fluxes points to substantial rest-frame extinction of A^host_V ≈ 2.5 mag, commensurate with the large neutral hydrogen column density, N_{H, int} ≈ 2 × 10²² cm⁻³. In conjunction with deep radio limits, we infer an energy of E_{γ, iso} ≈ E_{K, iso} ≈ 7 × 10⁵¹ erg and a low circumburst density of n ~ 10⁻³ cm⁻³. The lack of a break in the X-ray afterglow at ≈46 days leads to θ_j 11°, in line with existing measurements of short GRB jets (Fong et al. 2012). The prompt-emission properties of GRB 120804A, combined with its low circumburst density, point to a genuine short-duration origin.

We also detect the host galaxy of GRB 120804A, which exhibits red optical/near-IR colors and radio emission that are well matched by the SED of a ULIRG (Arp 220) at z ≈ 1.3. The inferred total star formation rate is ≈300 M_☉ yr⁻¹. A low-luminosity radio-loud AGN in an elliptical galaxy cannot be definitively ruled out, but the ULIRG interpretation, combined with the small projected offset of 2.2 ± 1.2 kpc, more naturally explains the inferred afterglow extinction. The host galaxy is part of an interacting/merging system, which is not unexpected for ULIRGs. Observations with the Atacama Large Millimeter/submillimeter Array (ALMA) will be able to robustly distinguish the two scenarios. At 240 GHz (ALMA band 6), the expected flux densities are about 0.3 mJy and 1 μJy, for the ULIRG¹⁰ and AGN cases, respectively. The former can be detected with high significance in a short observation.

The host galaxy of the short GRB 100206A at z = 0.407 is also a luminous infrared galaxy, L_IR ≈ 4 × 10¹¹L_☉, with SFR ≈ 30–40 M_☉ yr⁻¹ inferred from SED fitting of data at ~0.3–10 μm (Perley et al. 2012). Here we find SFR 60 M_☉ yr⁻¹ from radio observations. Of the ≈25 short GRBs with robust host galaxy associations (Fong et al. 2013), the hosts of GRBs 120804A and 100206A are the only galaxies with clear LIRG/ULIRG properties. At z ~ 1, the population of ULIRGs and bright LIRGs accounts for ~25% of the total star formation rate density (Le Floc'h et al. 2005; Caputi et al. 2007), but their contribution to the stellar mass density is small, ~1% to few percent (e.g., Caputi et al. 2006a). Thus, the fraction of ~5%–10% in the short GRB host population appears to be intermediate between the star formation and mass weighted fractions. This is expected for a progenitor population that tracks both star formation and stellar mass, consistent with previous findings for short GRBs (Leibler & Berger 2010).

We thank Karina Caputi, Ranga Chary, Francesca Civano, Martin Elvis, and Jane Rigby for useful discussions about ULIRGs and AGNs. We also thank Brian Metzger for information on the magnetar scenario. The Berger GRB group at Harvard is supported by the National Science Foundation under Grant AST-1107973 and by NASA/Swift AO7 grant NNX12AD69G. E.B. acknowledges partial support of this research while in residence at the Kavli Institute for Theoretical Physics under National Science Foundation Grant PHY11-25915. The Dark Cosmology Centre is funded by the Danish National Research Foundation. V.M. acknowledges funding through contract ASI-INAF I/004/11/0. Observations were obtained with the JVLA (program 12A-394) operated by the National Radio Astronomy Observatory, a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. The paper includes data gathered with the 6.5 m Magellan Telescopes located at Las Campanas Observatory, Chile; with ESO Telescopes at the La Silla Paranal Observatory under program ID 089.D-0450; with the Gemini Observatory, which is operated by the Association of Universities for Research in Astronomy, Inc., under a cooperative agreement with the NSF on behalf of the Gemini partnership: the National Science Foundation (United States), the Science and Technology Facilities Council (United Kingdom), the National Research Council (Canada), CONICYT (Chile), the Australian Research Council (Australia), Ministério da Ciência, Tecnologia e Inovação (Brazil) and Ministerio de Ciencia, Tecnología e Innovación Productiva (Argentina). This work is based in part on observations obtained with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and the USA (NASA).

Facilities: Swift (XRT) - Swift Gamma-Ray Burst Mission, CXO (ACIS-S) - Chandra X-ray Observatory satellite, XMM - Newton X-Ray Multimirror Mission satellite, Gemini:Gillett (GMOS) - Gillett Gemini North Telescope, Magellan:Baade (FourStar) - Magellan I Walter Baade Telescope, VLT:Yepun (HAWK-I) - Very Large Telescope (Yepun), EVLA - Expanded Very Large Array