SDWFS-MT-1: A SELF-OBSCURED LUMINOUS SUPERNOVA AT z 0.2

SDWFS-MT-1 (aka SN 2007va; Kozłowski et al. 2010b), where MT means mid-IR transient as a parallel notation to using OT for optical transient, is the most significantly variable source in the entire SDWFS field. It corresponds to the SDWFS source SDWFS J142623.24 + 353529.1. The [3.6] and [4.5] light curves were strongly correlated (r = 1) and the variability amplitude was 76 standard deviations from the mean for its average magnitude. A detailed definition of these variability criteria is given in Kozłowski et al. (2010a). The SDWFS survey consists of four epochs, taken 3.5 years, 6 months, and 1 month apart. The object is not detected in any of the four IRAC bands in the first epoch (also known as the IRAC Shallow Survey; Eisenhardt et al. 2004) on 2004 January 10–14 (JD' = JD − 24, 500, 00 ≈ 3016.5). The SDWFS single epoch 3σ detection limits of [3.6] = 19.67 mag, [4.5] = 18.73 mag, [5.8] = 16.33 mag, and [8.0] = 15.67 mag (4'' apertures corrected to 24''; Vega mag), set our upper limits on the object brightness in the first epoch. The observed part of the transient peaked in the second epoch taken 2007 August 8–13 (JD' ≈ 4322.5) at [3.6] = 15.93 mag and [4.5] = 15.61 mag (Figures 1 and 2). It was slightly fainter 6 months later in epoch 3 (2008 February 2–6, JD' ≈ 4500.5), with [3.6] = 16.09 mag and [4.5] = 15.76 mag. Then, in a matter of a month, the object faded by almost 2 mag to [3.6] = 17.99 mag and [4.5] = 17.72 mag in epoch 4 on 2008 March 6–10 (JD' ≈ 4533.5). We present all the available Spitzer data in Table 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

IRAC (Fazio et al. 2004) simultaneously observes the [3.6]/[5.8] or [4.5]/[8.0] bands, and in SDWFS the [4.5]/[8.0] observations were taken 40 s after the [3.6]/[5.8] ones. Each epoch consists of three 30 s exposures in each band, taken no more than 2 days apart, and the images for the whole mosaics were taken within 5 days. We confirmed that the transient was present in the individual artifact-corrected basic calibrated data (CBCD) frames corresponding to the position of the transient. The mosaic image for each epoch is then a combination of three to four of these frames, and there is no indication of problems in the SDWFS reductions (see Ashby et al. 2009).

We obtained a Keck/LRIS (Oke et al. 1995) spectrum of the transient's host galaxy on 2010 March 12, long after the transient was gone. The spectrum (shown in Figure 3) shows clear emission lines of Hα, Hβ, Hγ, Hδ, H, and Hζ, [O ii] at 3727 Å, [O iii] at 4363 Å, 4959 Å, and 5007 Å, and [Ne iii] at 3869 Å and 3968 Å. The redshift of the host galaxy is z = 0.1907, so the luminosity distance is 920 Mpc, the distance modulus is 39.82 mag, and the angular scale is 3.15 kpc arcsec⁻¹. Before measuring the line fluxes, we corrected the host galaxy spectrum for Galactic extinction (E(B − V) = 0.014 mag) using Schlegel et al. (1998). We do not detect the [N ii] 6584 Å line, which implies log([N ii]/Hα) < −1.5. Combining it with log([O iii] 5007 Å/Hβ) = 0.78, we conclude that the host is an irregular H ii galaxy (see, e.g., Terlevich et al. 1991; Kniazev et al. 2004). The lack of [N ii] and [Ne v] 3426 Å lines and the clear detection of the 4363 Å line rule out the presence of an AGN.

Zoom In Zoom Out Reset image size

We followed the prescription of Pilyugin & Thuan (2007) to derive the gas-phase oxygen abundance of the host. We measured the fluxes of the [O ii] 3727 Å, [O iii] 4363 Å, [O iii] 4959 Å, [O iii] 5007 Å,  and Hβ 4861 Å emission lines to find a metallicity of 12 + log(O/H) = 7.85 ± 0.13. We assumed an electron density of n_e = 100 cm⁻³, but the value does not change significantly with changes in the electron density (e.g., 12 + log(O/H) = 7.87 if we use n_e = 1000 cm⁻³ instead). Next, we verified this metallicity measurement using the prescriptions of Izotov et al. (2006) to get 12 + log(O/H) = 7.79. As a cross-check, we also used the oxygen-to-hydrogen flux ratio $R_{23}=({\rm [O\,\mathsc{ii}]+[O\,\mathsc{iii}]})/{\rm H}\beta =(F_{\lambda 3727}+F_{\lambda 4959}+F_{\lambda 5007})/F_{\lambda 4861}$ method of Pagel et al. (1979) and the Pettini & Pagel (2004) O3N2 method to derive oxygen abundances. The clear detection of the 4363 Å line and log([N ii]/Hα) < −1.5 strongly suggests that we should use the "lower" metallicity branch in the R₂₃ method. We use the relations from Skillman (1989) and Yin et al. (2007) with our measured log(R₂₃) = 0.94 to get oxygen abundances of 12 + log(O/H) = 7.78 and 7.80, respectively. The O3N2 method gives 12 + log(O/H) < 8.0. All these estimates are approximately 10% of the solar value (assuming 8.8 from Delahaye et al. (2010)).

In order to understand the origin of this mid-IR transient, we searched for any available UV and optical data for this area of the sky in the relevant time span.

The host galaxy was detected in the original NDWFS (Jannuzi & Dey 1999) with B_W = 23.98 ± 0.05 mag (λF_λ = 6.64 × 10⁴¹ erg s⁻¹), R = 22.78 ± 0.05 mag (1.10 × 10⁴² erg s⁻¹), and I = 22.58 ± 0.09 mag (8.39 × 10⁴¹ erg s⁻¹). It was unresolved in 12 (FWHM) seeing, which puts an upper limit on the size of the galaxy of 3.8 kpc. In archival Galaxy Evolution Explorer (GALEX; Morrissey et al. 2007) data, the host has NUV=24.30 ± 0.17 mag and no detection in FUV. The GALEX source is formally 087 from the SDWFS/NDWFS source, but there are no other blue sources within 20''. Both the colors of the host and its size are consistent with a blue, low-luminosity, low-metallicity galaxy. We used the template model of an irregular galaxy from Assef et al. (2010) to estimate the mid-IR magnitudes of the host. Fitting the GALEX NUV, and NDWFS B_W, R, and I-band magnitudes of the host galaxy, and assuming z = 0.1907, the expected apparent (absolute) IRAC magnitudes are [3.6] = 20.87(−18.56) mag, [4.5] = 20.69(−18.57) mag, [5.8] = 20.52(−20.02) mag, and [8.0] = 18.37(−21.45) mag. This is consistent with the failure to detect the host in the first SDWFS epoch. We estimate rest-frame host luminosities of M_B = −16.11 mag and M_R = −16.75 mag, making this an L/L 0.01 galaxy (Karachentsev 2005).

Although the field was monitored by at least five optical variability surveys, the peak of the transient occurred when the field was "behind" the Sun, so almost all the optical data correspond to the periods before epoch 2 and after epoch 4. The Boötes field was observed seven times between JD' = 3800 and 4620 with the 8k × 8k Mosaic CCD camera on the 2.4 m MDM Hiltner Telescope. The two most interesting observations took place 90 days before the second and 84 days after the fourth SDWFS epoch, bracketing the period of the transient. In these 180 s exposures, we detected no flux from the transient (or the host) at a 3σ level of R = 21.32 mag (4.2 × 10⁴² erg s⁻¹) and 21.15 mag (4.9 × 10⁴² erg s⁻¹), respectively.

The Catalina Sky Survey (CSS; Larson et al. 2003)/Catalina Real-Time Transient Survey (CRTS; Drake et al. 2009) provides light curves for the period prior to epoch 2 and overlapping epochs 3 and 4. Each epoch of four 30 s exposures on the 0.7 m Catalina Schmidt Telescope reaches an r-band magnitude limit of ~20 mag. These data provide the strongest limit on the optical-to-mid-IR flux ratios, where a stack of 12 images (3 epochs) gives a 3σ upper limit on the transient of r>20.93 mag, corresponding to 7.6 × 10⁴² erg s⁻¹ (large dark green symbol to the right in Figure 2), in the period between epochs 3 and 4.

The QUEST (Djorgovski et al. 2008) survey further limits the optical emission before epoch 2 and after epoch 4. There are 19 epochs each consisting of 1–4, 60–100 s exposures with the 1.2 m Samuel Oschin telescope in a wide red/IR filter that we calibrated to i band. The most interesting data were taken ~40 days prior epoch 2, where we stacked six adjacent QUEST epochs (JD' = 4276.71–4286.76) to obtain 3σ detection limit of i>20.70 mag (7.7 × 10⁴² erg s⁻¹, large orange symbol in Figure 2).

Similarly, by co-adding the high cadence RAPid Telescopes for Optical Response (RAPTOR; Woźniak et al. 2009) observations of the field, each of which is a 30 s exposure using the RAPTOR-P array of four 200 mm Canon telephoto lenses, we obtained 15 upper limits at various epochs between 2004 and 2008, before and after the transient (Figure 2). Calibrating the unfiltered observations to an R-band equivalent using Tycho 2 stars, we find no evidence for the transient, with C_R>18 mag (3σ).

Only the northern station of the All Sky Automated Survey (ASAS) data (Pojmański 1997) obtained data during the peak of the transient. Combining seven high-quality ASAS images during the transient (JD' = 4322.75–4525.11), we obtain a weak upper limit on the transient magnitude of V>16.3 mag (3σ), corresponding to 6.2 × 10⁴⁴ erg s⁻¹.

First, the transient cannot be explained by a Galactic star. The arguments against are as follows: (1) based on the Keck spectrum, we know that the source is spatially coincident with an irregular galaxy; (2) few stars can brighten by at least 5 mag (by more than 100 times) in the mid-IR and then stay at this level for 6 months before fading; and (3) most flaring stars do not flare in the mid-IR but in the UV. There are exceptions to (2) and (3), and these will generally be accreting systems where the disk can alter the spectral properties. However, the NDWFS field is a high latitude (b 67°) field with negligible Galactic extinction and unlikely to contain young (protostars) or massive (Be stars) stars with these characteristics, particularly if they must also have no optical counterpart in quiescence despite needing to be nearby sources because of the short path length out of the disk.

The second possibility is that we are really observing an AGN. First, we note that there are no signs of AGN activity in the host spectrum. The narrow lines in a galaxy are on relatively large scales and should show line ratios averaged over decades (or more) of AGN activity, so the lack of a spectrum during the transient does not invalidate using them to argue against an AGN. Moreover, the mid-IR colors at the peak of the transient, [3.6] − [4.5] = 0.3 mag and [5.8] − [8.0] = 0.0 mag (epoch 2), are inconsistent with that of an AGN (e.g., Stern et al. 2005; Assef et al. 2010; Kozłowski et al. 2010a). While the composite SED of an AGN and its host can have these colors, we know from the host flux limits in the first epoch that the SED at the peak should be dominated by the AGN, and a pure AGN at z 0.2 should have mid-IR colors closer to [3.6] − [4.5] 0.8 mag and [5.8] − [8.0] 1.1 mag (Assef et al. 2010). We also considered the possibility of a tidally disrupted star in the context of the super-Eddington accretion models of Loeb & Ulmer (1997). While these models have an inflated, relatively cool envelope around the black hole, the expected temperatures of T ≈ 7000(M_BH/10⁵ M_☉)^1/4 K for a black hole of mass M_BH are still too high. A black hole with low enough mass (M_BH ≈ 150 M_☉) to match the temperature (T ≈ 1350 K) would be unable to produce the observed luminosity. The Eddington luminosity for such a low mass is only L_Edd ≈ 2 × 10⁴⁰ erg s⁻¹.

The third possibility is that it is a γ-ray burst (GRB). We can rule it out as direct emission from a burst because the timescale (6 months) is too long and the SED is wrong for direct beamed emission (e.g., Sari et al. 1999). It cannot be indirect emission, where a dusty region absorbs and reradiates emission from the jet, because a relativistic jet expands past the scale of the photosphere (~10⁴ AU) too quickly (~60 rest-frame days) to produce the observed event duration.

Finally, this cannot be gravitational lensing of a background object by the host galaxy (see Wambsganss (2006) for a review). The timescales are such that it would have to be microlensing by the stars in the lens galaxy, but there are no known mid-IR sources that are sufficiently compact to show such a large, and apparently achromatic, degree of magnification. Moreover, almost any potential source would have to show a still stronger degree of optical magnification that would disagree even with our weak optical limits (see, e.g., Morgan et al. 2010).

Our working hypothesis is that SDWFS-MT-1 is an SN analogous to SN 2006gy, an extremely luminous Type IIn SN (Ofek et al. 2007; Smith et al. 2007, 2008b, 2010; Smith & McCray 2007; Miller et al. 2010). SN 2006gy peaked at M_V −22 mag, corresponding to a luminosity of L ~ 3 × 10⁴⁴ erg s⁻¹, and it radiated of order E_rad>2 × 10⁵¹ erg in total (Ofek et al. 2007; Miller et al. 2010). In a normal Type II SN, a very small fraction (1%) of the available energy is radiated. The initial energy from the shock is converted into kinetic energy by adiabatic expansion, leaving little to power the luminosity. Thus, an SN ejecting M_ej ~ 10 M_☉ at velocities of v_ej ~ 4000 km s⁻¹ contains enough kinetic energy, E_kin 2 × 10⁵¹ erg to power the transient, but it is only radiated on timescales of order 10³ years as the SN remnant develops. The scenario we present was proposed by Smith & McCray (2007) for SN 2006gy with further elaborations in Smith et al. (2008b) and Miller et al. (2010).

Very massive stars (50 M_☉) are known to undergo impulsive events ejecting significant fractions of the stellar envelope. The most famous example is the eruption of η Carinae in 1837–1857 (see Davidson & Humphreys 1997), but it also includes some of the so-called "supernova impostors" (e.g., Smith & Owocki 2006; Pastorello et al. 2007). The number and temporal distributions of these mass ejections relative to the final SN are unknown, but there are several cases where one seems to have occurred shortly before the final SN (e.g., Smith et al. 2007). Pair instability SNe can also produce a sequence of mass ejections just prior to the death of the star (see Woosley et al. 2007). Let us suppose that the progenitor of SDWFS-MT-1 was such a massive star and had two such eruptions in its final phases, roughly 300 and 5 years before the final explosion, each of which ejected of order M_shell 10 M_☉. For a typical ejection velocity of 200 km s⁻¹ (e.g., Smith & Owocki 2006), these shells would lie roughly 13000 and 200 AU from the star when it exploded.

The role of the first shell is to convert the kinetic energy of the explosion into radiation following Smith & McCray (2007). Roughly speaking, the collision between the ejected material and the first shell can extract fraction M_shell/(M_ej + M_shell) of the kinetic energy and convert it to radiation. The distance to the shell is tuned so that the shock crossing time is comparable to the Thomson photon diffusion time, thereby allowing the radiation to escape without significant adiabatic losses. This requires a total mass of approximately M = 4πR²m_pc/σ_Tv_ej = 2.6R²₁₀₀v⁻¹_ej,4000 M_☉, where R = 100R₁₀₀ AU and v_ej = 4000v_ej,4000 km s⁻¹. We can get up to 20 M_☉ either by making R 300 AU or by allowing for adiabatic losses. This then implies characteristic time, energy, luminosity, and temperature scales of t = R/v_ej = 50R₁₀₀/v_ej,4000 days, E = (1/4)Mv²_ej = 2 × 10⁵⁰R²₁₀₀v_ej,4000 erg, E/t = 5 × 10⁴³v²_ej,4000R₁₀₀ erg s⁻¹, and T = (L/4πR²σ)^1/4 = 14,000v^1/2_ej,4000R^−1/4₁₀₀ K, respectively, where we have used the estimate of the total mass and equally divided it between the shell and the SN ejecta. The basic energetics, but not the radiation transport, are largely confirmed by the simulations of van Marle et al. (2009). This set our choice of incident temperature in the DUSTY models. Thus, a shell of ejected material at R ~ 300v⁻¹_ej,4000 AU can produce the necessary energetics and luminosities but cannot match the observed timescales (too short) or temperatures (too high).

The role of the second shell is to match the time and temperature scales. Smith et al. (2008b) and Miller et al. (2010) invoke a second M_shell 10 M_☉ shell located at R 1 pc in order to explain the relatively bright, long-lived IR emission observed for SN 2006gy. Their shell is, however, optically thin to optical light (τ 0.02) because the total infrared emission is only a small fraction of the total. The optical depth of a shell is of order τ 0.1κ(M_shell/10 M_☉)(10⁴ AU/R)² for κ in cm² g⁻¹, which leads to an optically thin shell with τ_opt 0.08 for an R 1 pc shell radius and κ_opt 500 cm² g⁻¹ for a radiation temperature of order 10⁴ K (e.g., Semenov et al. 2003). This opacity, κ 500 cm² g⁻¹, corresponds to a dust-to-gas ratio of approximately 1%, which arguably may be high for a low-metallicity galaxy. Note, however, that SN 2006gy is probably also associated with a low-metallicity galaxy (see Section 3.4) and the shell producing its dust echo contains enough dust for our scenario.

To explain SDWFS-MT-1, we simply move the shell inward. At this point, however, we should improve on Equation (1) with a simple model for an optically thick dust echo. Suppose that the dust lies in an optically thick spherical shell of radius R and is exposed to luminosity L_p for a period of time t_p. Because all the dust is exposed to the same luminosity, the dust temperature is independent of position or time while irradiated. The observer, however, sees the luminosity pulse modulated by the light travel time 2R/c across the shell. In this simple model, the light curves are trapezoids with linear rises and falls, connected by a plateau, created by the changing fraction of the shell, illuminated by the outburst from the observer's perspective, but the dust temperature is constant. Models that can reproduce the drop in luminosity between epochs 3 and 4 have the pulse time t_p < 2R/c, with the simplest model having the luminosity plateau run from epoch 2 to epoch 3. The plateau duration is 2R/c − t_p = 180 days, while the rises and falls last t_p. The luminosity of the plateau is 2πRct_pσT⁴, which is smaller than L_p = 4πR²σT⁴ by the ratio 2t_p/Rc 0.2 between the light crossing time and the pulse length (if t_p>2R/c, the plateau luminosity is L_p). The factor of 3 drop in luminosity between epochs 3 and 4 is L₄/L₃ = 1 − t₃₄/t_p 1/3, where t₃₄ 33 days is the time between these epochs. Thus, we have t_p 50 days and R 19000 AU or 115 light days. With these time and distance scales, we come close to matching the epoch 2 and 3 luminosities given the observed temperatures, with an estimated luminosity of L₂ = L₃ = 5 × 10⁴³ erg s⁻¹. The true luminosity is L_p 2 × 10⁴⁴ erg s⁻¹, which is why the radius estimate is somewhat larger than the estimates of Equation (1). There is some tension between the plateau luminosity and the drop in luminosity between the last two epochs, but the overall agreement is good given the crudeness of the model both for the dust emission and the luminosity profiles.

The mass required to produce a shell with optical depth τ at this distance is M τ(κ/500 cm² g^-1)(R/19,000 AU)² M_☉, so there is little difficulty having an optically thick shell for masses of order 1–10 M_☉ even if the dust-to-gas ratio needs to be significantly reduced from ~1% because of the low metallicity of the galaxy. While the dust temperature is close to that for destroying dust (e.g., Dwek & Werner 1981), the situation is somewhat different from a normal SN because the (Thomson) optically thick inner shell protects the outer from any initial high-luminosity bursts of hard radiation. For the IR radiation, κ_IR 10 cm² g⁻¹, and the shell is optically thin to the mid-IR emission for reasonable shell masses, so the dust emission streams freely to the observer. Given the estimated pulse duration, t_p 50 days, and luminosity, L_p = 2 × 10⁴⁴ erg s⁻¹, we can estimate that the inner shell is at R 160 AU, the velocity must be v_ej 6300 km s⁻¹, and the mass involved must be M 4 M_☉, where the rapid decline pushes these values to small radii/masses and high velocities in order to keep t_p short. Given the crudeness of the model, it is remarkable how well all the observational features can be matched using relatively sensible physical parameters.

The nature of the host galaxy indirectly supports this hypothesis because both GRBs and most hyper-luminous Type IIn SNe are also associated with low-luminosity, low-metallicity dwarf galaxies. Figure 5, similar to Figure 1 in Stanek et al. (2006), shows the position of the host galaxy for SDWFS-MT-1 in the metallicity–luminosity plane in relation to ~126,000 star-forming SDSS DR4 galaxies (Tremonti et al. 2004; Adelman-McCarthy et al. 2006; Prieto et al. 2008). We also include the hosts of six local (z < 0.25) GRBs using updated estimates from Levesque et al. (2010) and including the host of GRB 100316D (Chornock et al. 2010; Starling et al. 2010). While SDWFS-MT-1 was not a GRB, its occurrence in a very low metallicity galaxy is striking, since such galaxies produce only a small fraction of massive stars in the local universe (e.g., see Figure 2 in Stanek et al. 2006). In Figure 5, we ignore small (20%) effects on metallicity measurements arising from the method being used to derive it (see Kewley & Ellison 2008).

Zoom In Zoom Out Reset image size

In fact, the hosts of most other hyper-luminous SNe also tend to be of very low luminosity (~0.01 L), suggestive of their low metallicity (unfortunately, oxygen abundances have not been measured for most of these hosts). SN 2005ap (Quimby et al. 2007) at z = 0.283 peaked at unfiltered magnitude of M = −22.7 mag and is associated with an M_R = −16.8 (M_B ≈ −16.1) mag dwarf galaxy. SN 2006tf (Quimby et al. 2007; Smith et al. 2008a) at z = 0.074 exploded in a host of low-luminosity M_r = −16.9 (M_B ≈ −16.2) mag. The over-luminous M_R = −21.3 mag Type Ic SN 2007bi exploded in a faint, M_B = −16.3 mag dwarf galaxy, with 12 + log(O/H) = 8.18 ± 0.17 (Gal-Yam et al. 2009; Young et al. 2010). SN 2008es at z = 0.213 (Miller et al. 2009; Gezari et al. 2009) reached M_V = −22.3 mag and exploded in a host which was not detected down to a faint luminosity of M_V> − 17.4 (M_B> − 17.1) mag. SN 2008fz at z = 0.133 peaked at M_V = −22.3 mag, but no host has been detected down to an upper limit of M_R> − 17 (M_B> − 16.6) mag (Drake et al. 2010). SN 2006gy (Quimby 2006; Smith et al. 2007; Ofek et al. 2007) is one of the two exceptions here, as it exploded in a fairly luminous S0 galaxy, NGC 1260 with M_B ≈ −20.3 mag. Another example is SN 2003ma (Rest et al. 2009) that released 4 × 10⁵¹ erg in an M_B ≈ −19.9 mag, 12 + log(O/H) ≈ 8.7 galaxy.

The general message is clear: GRBs and luminous SNe prefer low-metallicity, low-luminosity galaxies, so does SDWFS-MT-1.

We systematically searched for transients similar to SDWFS-MT-1 and SNe in general in the SDWFS data. We considered highly correlated [3.6] and [4.5] light curves (r>0.8) with flag =1 (photometry unaffected by the wings of bright stars), variability significances σ_[3.6]>1 and σ_[4.5]>1 and colors [3.6] − [4.5] < 0.4 to eliminate AGNs (see Kozłowski et al. 2010a). We selected as candidate objects that at their peak were brighter than [3.6] < 18.67 mag, corresponding to 1 mag brighter than the 3σ detection limit, and which changed in brightness by at least 1 mag between either epochs 1 and 2 or epochs 2 and 3/4. The latter two epochs are close enough together that an SN might not show a significant brightness change between them. We found six candidate objects, one of which is SDWFS-MT-1. Four are artifacts due to PSF structures (long spikes) from bright stars.

The remaining object, SDWFS J142557.64+330732.6 (hereafter SDWFS-MT-2), peaks in epoch 2 close to our selection limit at [3.6] = 18.52 ± 0.14 mag and [4.5] = 18.71 ± 0.33 mag. It lies 07 south–west of a potential host galaxy with GALEX NUV=24.02 ± 0.19 mag, NDWFS B_W = 23.77 ± 0.03 mag, R = 23.34 ± 0.04 mag, and I = 22.68 ± 0.06 mag that is not detected in SDWFS. In the deep images from NDWFS, the host galaxy appears to be an extended, possibly edge-on (length 30 in 12 seeing) disk, with a redder core and bluer outer regions. The transient lies in the bluer outer regions, although its position is only known to ~02 due to its faintness. We estimate a photometric redshift of z ≈ 1.0 for the host using the template models of Assef et al. (2010), corresponding to a comoving distance of 6.6 Gpc and a linear scale of 8.0 kpc arcsec⁻¹. This translates into the linear diameter of the host of ~24 kpc. The best-fit model for the SED of the host is a mixture of an AGN and a starburst template. The mid-IR color of the transient, [3.6] − [4.5] = 0.2 ± 0.5 mag, is ambiguous on the question of an SN or AGN origin for the transient, although the offset position would favor an SN. Given the redshift estimate, however, the implied luminosity is too high even for the brightest normal SN—for M_[3.6] ~ M_K ~ −20 mag (Mannucci et al. 2003), the peak magnitude of [3.6] ~ 24 mag would be far below our detection limits. A second event like SWFS-MT-1 would have to be even more extreme. Given the paucity of the data, interpreting this source as a false positive seems the most probable explanation.

For events with 6 month durations like SDWFS-MT-1, our survey time coverage is approximately ΔT = 2 years. SDWFS-MT-1 peaked in epoch 2 at fluxes of 0.12 mJy in [3.6] and 0.10 mJy in [4.5] at a luminosity distance of 920 Mpc. The peak flux limit for our systematic search is 9.6 μJy (5σ in [4.5]) which means we could have detected SDWFS-MT-1 at a luminosity distance of 3000 Mpc (920 × (100/9.6)^1/2) or about z = 0.52 ignoring K-corrections, which will initially aid detection because the SED peaks blueward of the IRAC bands. Thus, we could detect the source in a comoving volume of approximately V = 6 × 10⁶ Mpc³. Combining these factors, we estimate a rate for SDWFS-MT-1-like transients of r = (VΔT)⁻¹ ~ 8 × 10⁻⁸ Mpc⁻³ yr⁻¹ where we have not corrected the timescales for redshifting effects given the otherwise large uncertainties from the statistics of one event. This is approximately 10⁻³ of the Type II SNe rate (Horiuchi et al. 2009), ~0.1 of the hyper-luminous Type II SNe (Miller et al. 2009), and ~10 of the GRB rate (Nakar 2007).