DUST PRODUCTION AND PARTICLE ACCELERATION IN SUPERNOVA 1987A REVEALED WITH ALMA

Supernova (SN) 1987A in the Large Magellanic Cloud was the closest supernova explosion to Earth (50 kpc) observed since Kepler's SN1604AD, making it a unique target to study supernova and supernova remnant physics. Supernovae are thought to be one of the most important sources of dust in the universe. Stars synthesize heavy elements, and after they explode, rapid expansion and radiation cause the ejected gas to cool rapidly, allowing refractory elements to condense into dust grains. Dust masses have been measured in over 20 supernovae and supernova remnants (Gall et al. 2011); most contain 10⁻⁶–10⁻³ M_☉ of warm dust, far below predictions of 0.1–1.0 M_☉ (Todini & Ferrara 2001; Nozawa et al. 2003; Cherchneff & Dwek 2010). One of the few exceptions is SN 1987A—far infrared photometry (100–350 μm) with the Herschel Space Observatory in 2011 implied 0.4–0.7 M_☉ of dust at 20–26 K (Matsuura et al. 2011). Measured 24 years after the explosion, that dust mass is significantly higher than the 10⁻⁴ M_☉ reported from infrared measurements one to two years after the explosion (Wooden et al. 1993; Bouchet et al. 2004). While the Herschel result supports theoretical models of significant dust production, the discrepancy with prior results caused many to question whether all of the dust detected in 2011 was in fact produced in the supernova. SN 1987A's progenitor star could have created a massive dust shell during its red supergiant phase. Mid-infrared (MIR) photometry of SN 1987A in 2003 (Bouchet et al. 2004) and spectroscopy in 2005 (Dwek et al. 2010) only found 10⁻⁵ M_☉ of warm progenitor dust. However, observations of Galactic evolved stars with masses bracketing that of SN 1987A's progenitor reveal between 0.01 and 0.4 M_☉ of dust (the Egg Nebula, Jura et al. 2000 and Eta Carina, Gomez et al. 2010), so if only a minor warm component of the progenitor dust was detected in the MIR, a large cold progenitor dust mass detected for the first time by Herschel is plausible. Herschel had insufficient spatial resolution to distinguish between the ejecta, shocked progenitor wind, and nearby interstellar material.

Resolved observations of SN 1987A in the submillimeter regime are required to establish the location of the emission measured by Herschel. Multi-wavelength resolved images are required to determine its physical origin, and such images can be used to simultaneously study shock particle acceleration. Twenty-five years after the explosion, SN 1987A's blast wave had propagated to a radius of 7 × 10¹⁷ cm (09 on the sky), shocking material lost from the progenitor star (Crotts & Heathcote 2000). Electrons accelerated by the shocks produce a bright shell or torus of synchrotron radiation at centimeter wavelengths (Potter et al. 2009; Ng et al. 2013). The shocks are interacting with a dense equatorial ring tilted 43 degrees from the line-of-sight (Tziamtzis et al. 2011), dominating the emission at X-ray, optical to thermal infrared (<30 μm) wavelengths (Bouchet et al. 2004; Dwek et al. 2010). Interior to the shock and ring are the remnants of the star's metal-rich core, referred to hereafter as the inner ejecta. A supernova explosion can also leave behind a black hole or a neutron star. The observed neutrino burst implies that a neutron star must have at least temporarily formed in SN 1987A (McCray 1993), but it has yet to be detected. If the neutron star energizes a pulsar wind nebula (PWN), its emission would most likely have a spectrum with power-law spectral index of −0.3 to 0 (Gaensler & Slane 2006), flatter than synchrotron emission from the shock, perhaps detectable at millimeter wavelengths or even contributing to the Herschel far-infrared emission. Previous observations at 10 mm (Potter et al. 2009), 6.8 mm (Zanardo et al. 2013), and 3.2 mm (Lakićević et al. 2012b) tentatively detected such excess emission at the remnant's center, with large uncertainties from image deconvolution; high spatial resolution images from millimeter to submillimeter wavelengths are essential to disentangle the emission mechanisms.

The Atacama Large Millimeter/Submillimeter Array (ALMA) observations are executed according to the quality of the weather, and in general the observing requirements are more stringent the higher the frequency of the observation. SN 1987A was observed during 2012 multiple times with ∼20 antennas in configurations containing baselines between ∼20 m and ∼400 m. Bands 3,6,7, and 9 (2.8 mm, 1.4 mm, 870 μm, and 450 μm) were observed in approximately chronological order. Band 3: A002/X3c5ee0/X24b (April 5), A002/X3c7a84/X1c (April 6); Band 6: A002/X3c8f66/X352 (April 7), A002/X45f1dd/Xd13 (July 15), A002/X494155/X8be (August 10); Band 7: A002/X45e2af/X458 (July 14), A002/X4ae797/X776 (August 24); Band 9: A002/X4afc39/X8ce (August 25), A002/X4b29af/Xd21 (August 27), A002/X535168/X796 (November 5). Each data set was calibrated individually with Common Astronomy Software Applications (casa.nrao.edu). All data in a given wavelength band were then combined for imaging and deconvolution with the "clean" algorithm. Only the channels free of bright CO and SiO emission were used (212.56–213.59 GHz and 100.07–103.91 GHz in Bands 6 and 3). In synthesis imaging the weighting as a function of baseline length can be adjusted to achieve finer spatial resolution at the expense of somewhat higher noise per beam. Briggs weighting was used with "robust" parameters and resulting Gaussian restoring beams as noted in the figures. The Australia Telescope Compact Array (ATCA) images were deconvolved using the maximum entropy algorithm (Zanardo et al. 2013; Lakićević et al. 2012b).

In Figure 1, we present the first spatially resolved submillimeter continuum observations of SN 1987A, obtained with the ALMA.

Zoom In Zoom Out Reset image size

We can now separate emission from the torus from that of the inner ejecta at 450 μm, 870 μm, and 1.4 mm. Decomposition of the torus and inner ejecta was performed in the image plane. The torus was represented with the clean model (collection of point sources or clean components) from deconvolving the ATCA 6.8 mm image. The inner ejecta were represented by the ALMA Band 9 (450 μm) model. For each band, the image was decomposed into a clean model and residual. The torus or ejecta model, scaled in amplitude, was subtracted from the model, which was then restored with the appropriate Gaussian restoring beam, and the residual image added back in. This method results in minimal noise from the torus or ejecta observation being introduced, merely preserving the noise already present in each original image. The amplitude scaling is varied to minimize the residuals in the subtracted image. The torus and ejecta were also subtracted from each other in the Fourier plane, with consistent results. The torus and ejecta flux densities measured using these independent decompositions were combined to determine the flux densities and uncertainties listed in Table 1.

Table 1. Flux Densities

Component	ν	λ	Fν	Epoch	Telescope	Angular	Ref.
	(GHz)		(mJy)			Res.
Torus	36.2	8.3 mm	27 ± 6	2008	ATCA	03	Potter et al. (2009)
Torus	44	6.8 mm	40 ± 2	2011	ATCA	03	Zanardo et al. (2013)
Torus	90	3.2 mm	23.7 ± 2.6	2011	ATCA	07	Lakićević et al. (2012b)
Both	110	2.8 mm	27 ± 3	2012	ALMA	13	This Letter
Torus	215	1.4 mm	17 ± 3	2012	ALMA	07	This Letter
Ejecta	215	1.4 mm	<2	2012	ALMA	07	This Letter
Torus	345	870 μm	10 ± 1.5	2012	ALMA	05	This Letter
Ejecta	345	870 μm	5 ± 1	2012	ALMA	05	This Letter
Torus	680	440 μm	<7	2012	ALMA	03	This Letter
Ejecta	680	440 μm	50 ± 15	2012	ALMA	03	This Letter
Both	860	350 μm	54 ± 18	2010	Herschel	24''	Matsuura et al. (2011)
Both	860	350 μm	44 ± 7	2011	APEX	8''	Lakićević et al. (2012a)
Both	1200	250 μm	123 ± 13	2010	Herschel	18''	Matsuura et al. (2011)
Both	1900	160 μm	125 ± 42	2010	Herschel	95	Matsuura et al. (2011)
Both	3000	100 μm	54 ± 18	2010	Herschel	135	Matsuura et al. (2011)

Download table as:  ASCIITypeset image

We discuss the torus first since the physical emission mechanism, synchrotron emission from shock-accelerated particles, is less controversial than the inner ejecta emission mechanism. Images of the torus alone (central ejecta removed) are shown in Figure 2. The eastern side is brighter at all (sub)millimeter wavelengths, and the asymmetry decreases at shorter wavelengths. In contrast, Hα emission (Larsson et al. 2013) from shocks being driven into the equatorial ring, and soft X-ray emission (Helder et al. 2013) from hot plasma behind the shocks is now brighter in the west (Figure 1). All wavelengths shown have brightened with time, and the emission distribution has become geometrically flatter in the equatorial plane, as the shocks interact with the dense ring (Ng et al. 2013; Racusin et al. 2009). Asymmetry results from differences in how far that interaction has progressed (e.g., the X-ray asymmetry can be explained by faster shocks in the east; Zhekov et al. 2009).

Zoom In Zoom Out Reset image size

The integrated spectrum of the torus (Figure 3) is fit from 17 mm to 870 μm by a power law with a spectral index α = −0.8 ± 0.1 (F_ν∝ν^α) and no evident spectral break. This rules out significant synchrotron losses that would steepen the spectrum or contamination from free–free emission that would flatten it. The spectrum is quite steep (few high energy particles), compared to what would be produced by diffusive shock acceleration in the test particle limit, in a single shock at the observed ∼2000 km s⁻¹ expansion velocity (Helder et al. 2013; Ng et al. 2013). The shock structure is therefore likely being modified by pressure from the accelerated particles (Ellison et al. 2000) or by an amplified tangled magnetic field (Kirk et al. 1996).

Zoom In Zoom Out Reset image size

The torus with its simple spectrum can now be removed from the images to isolate the inner ejecta (Figure 4). Inner ejecta emission is well resolved at 450 μm, with a beam-deconvolved FWHM of 0.3 ± 003 by 0.16 ± 005, or 2.2 ± 0.2 × 10¹⁷ cm by 1.2 ± 0.4 × 10¹⁷ cm, corresponding to constant expansion velocities of 1350 ± 150 km s⁻¹ by 750 ± 250 km s⁻¹. It is marginally resolved at 870 μm, constrained by the observations to have FWHM<04, consistent with the 450 μm size.

Zoom In Zoom Out Reset image size

The spectrum of the inner ejecta rises steeply with frequency, inconsistent with PWN emission, but very well modeled by dust. The ALMA and Herschel photometry can be fit with 0.23 ± 0.05 M_☉ of amorphous carbon dust at 26 ± 3 K (Figure 3). This implies that that the carbon dust mass is much higher today than the 5 × 10⁻⁴ M_☉ measured two years after the explosion (Wooden et al. 1993; Ercolano et al. 2007), and that within uncertainties nearly all of the 0.23 ± 0.1 M_☉ of carbon (Thielemann et al. 1990; Woosley & Heger 2007) released in the explosion is now in dust. The dust mass depends on opacity, temperature, and optical depth but is quite robustly constrained by the data. We use amorphous carbon opacity κ = 10 and 3 cm² g⁻¹ at 450 and 870 μm, respectively (Rouleau & Martin 1991). Values in the literature range from 2 to 10 cm² g⁻¹ at 450 μm (e.g., Planck Collaboration et al. 2013; Jager et al. 1998); using a lower value would only raise the dust mass. Our analysis of CO emission observed with ALMA and Herschel finds >0.01 M_☉ of CO in SN 1987A (Kamenetzky et al. 2013). Within uncertainties, the carbon in dust and CO does not yet exceed the nucleosynthetic yield, but as future observations refine the CO and dust masses, strong constraints may be placed on nucleosynthesis, chemistry, dust coagulation, or all three.

The temperature is quite well constrained by these data, and theoretical models predict a similar temperature. Cooling by adiabatic expansion and radiation is offset by heating from ⁴⁴Ti decay and by external X-ray heating from the shocks. The current gas temperature in the absence of X-ray heating is modeled to be 20–100 K (Fransson et al. 2013). Models predict that significant X-rays do not yet penetrate inward as far as the 2 × 10¹⁷ cm radius of the ALMA 450 μm emission (Fransson et al. 2013). The compact and centrally peaked 450 μm emission supports this interpretation, since external heating would likely result in more limb-brightened or extended dust emission. Since the ionization fraction is below 1% (Larsson et al. 2013), less than ∼40% of the X-ray flux will go into heating (Xu & McCray 1991), and <5% of the observed total flux 4.7 × 10³⁶ erg s⁻¹ (Helder et al. 2013) is intercepted by the ejecta. Even if it did reach the dusty inner core, the energy deposition would be <10% of the heating from ⁴⁴Ti decay (Jerkstrand et al. 2011). The best-fit dust mass M_d scales approximately as $T_d^{-2}$, and M_d > 0.1 M_☉ for T_d < 50 K.

The emission is optically thin at 450 μm, and M_d is insensitive to unresolved clumpiness: If 0.23 M_☉ of dust uniformly filled a region the size of the ALMA emission, the peak surface density would be 0.02 g cm⁻², with an optical depth τ₄₅₀ <0.2. If the dust is clumpy, each clump remains optically thin unless the filling fraction is less than ∼2%; a filling fraction of 10%–20% was fitted to the infrared spectrum (Lucy et al. 1991; Ercolano et al. 2007) and consistent with the CO clump filling factor of 0.14 fitted to ALMA CO emission (Kamenetzky et al. 2013). Analytical formulae for radiative transfer in dense clumps (Városi & Dwek 1999) indicate that the effective optical depth of the ensemble of clumps remains low over a very wide range of clump filling fraction and number.

The data can also be fit by a combination of carbonaceous and silicate (Mg₂SiO₄) dust (Figure 3; dotted line), although the mass of carbonaceous dust cannot be significantly reduced because amorphous carbon has the highest submillimeter opacity κ among minerologies similar to the interstellar medium, and eliminating it would require dust masses of other compositions larger than the available metal mass (the best fit using pure silicate is 4 M_☉ at 21 K). Although these data do not strongly constrain the mass of silicate dust, it is unlikely that only carbon condensed in the SN 1987A, given the presence of large amounts of Mg and Si in the ejecta (0.065 ± 0.1 and 0.19 ± 0.1 M_☉, respectively) and the detection of SiO molecules at early epochs (Roche et al. 1991; Wooden et al. 1993). The absence of the 10 μm feature during days 615–775 puts a strong upper limit (15%) on the fraction of dust which is silicates, but does not exclude the presence of some silicate dust during that epoch, since the ejecta could have been optically thick at that wavelength (Ercolano et al. 2007). The silicate mass could be much larger now, similar to the much larger carbon mass.

Models and observations show that core collapse supernovae like SN 1987A are highly inhomogeneous, with instabilities and radioactive energy deposition mixing the initially chemically stratified stellar interior into clumps with different compositions, macroscopically mixed throughout the ejecta volume (McCray 1993; Jerkstrand et al. 2011). The distribution of emission from different atoms, molecules, and dust not only reveals the chemistry of their formation, but a snapshot of the early supernova interior. The extent of dust emission corresponds well to line emission from CO 2–1 and SiO 5–4 observed simultaneously with ALMA (Kamenetzky et al. 2013). The 1–2 × 10¹⁷ cm radial extent corresponds to expansion velocities of 750–1400 km s⁻¹, consistent with the 1250 km s⁻¹ half-width of the CO emission (Kamenetzky et al. 2013), and within the commonly used maximum core expansion velocity of ∼2000 km s⁻¹ (Jerkstrand et al. 2011). SiO is an important precursor to forming dust, so the relative distributions of dust and SiO in SN 1987A that ALMA will be able to measure with higher spatial resolution in the future will constrain formation theories.

In the last few years, the inner ejecta have developed a complex morphology in optical and infrared emission, including north–south elongation and a "hole" or fainter region in the center. The 450 μm emission peak is coincident with the hole, and although the optical emission is likely limb-brightened due to external X-ray heating (Larsson et al. 2011), the dust we detect does have significant optical opacity. Smooth dust would have τ_V >1000, and 100 dusty clumps filling 10% of the inner ejecta volume would have τ_V ∼1.7 (Városi & Dwek 1999). (About 100 clumps were predicted by collapse simulations (Hammer et al. 2010) and fitted to observed optical lines (Jerkstrand et al. 2011)).

We have used the powerful resolution of ALMA to show clearly that prodigious amounts of dust have formed in the cold inner ejecta of SN 1987A. Our data suggests that nearly all of the carbon has condensed into dust, so condensation must be efficient. A large mass of clumpy dust was tentatively detected at 1.3 mm four years after the explosion (Biermann et al. 1992)—if this is the same dust then it formed quite rapidly. If dust production in other supernovae resembles that in SN 1987A, then core-collapse supernovae might contribute as much dust to galaxies as asymptotic giant branch stars. In the absence of grain destruction, dust in high-redshift galaxies can be explained with only 0.1 M_☉ produced per Type II supernova (Dwek et al. 2007). SN 1987A has the largest measured dust mass, but a few other remnants contain almost 0.1 M_☉ (Cassiopeia A, Barlow et al. 2010; Crab Nebula, Gomez et al. 2012). However, dust formed in a supernova must survive passage through the reverse shock to be dispersed into the interstellar medium, and then also survive shock passages once dispersed (Jones & Nuth 2011). The reverse shock in SN 1987A located with spatially resolved Hα spectra (Michael et al. 2003; France et al. 2010) has a minimum radius in the equatorial plane of ∼4 × 10¹⁷ cm, significantly larger than the 450 μm emission (Figure 4). The new dust has not been significantly processed by the reverse shock. Models of dust destruction predict a wide range of survival fractions (Nozawa et al. 2007). If 0.23 M_☉ of dust is typically created in Type II supernovae, and 0.1 M_☉ passes through the reverse shock into the ISM, then Type II supernovae could dominate dust production in galaxies at all redshifts.

This Letter makes use of the following ALMA data: ADS/JAO.ALMA#2011.0.00273.S (PI: Indebetouw). ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada) and NSC and ASIAA (Taiwan), in cooperation with the Republic of Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO and NAOJ. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. M.M. was supported by NASA NAG5-12595 and NASA/ADAP NNX13AE36G.