THE DESTRUCTION OF THE CIRCUMSTELLAR RING OF SN 1987A

Supernova 1987A is unique: we can study in real time and with high spatial resolution physical processes that we can only infer for other supernovae. Emission from SN 1987A included a flash of ultraviolet radiation in the first hours, thermal diffusion of radioactive energy input for weeks, and prompt reprocessing of radioactive power for decades (e.g., McCray 1993). However, now, conversion of the supernovaʼs kinetic energy to heat through shock interactions dominates the radiative energy budget (Larsson et al. 2011, hereafter L11).

The ring system around SN 1987A was created by mass loss from the progenitor star in the 20,000 years before the explosion (Crotts & Heathcote 1991). The rings may have been created as a result of mass ejections from a rapidly rotating single star (Chita et al. 2008) or a merger of two massive stars (Morris & Podsiadlowski 2009; Akashi et al. 2015). Observations of the interaction of the ejecta with the circumstellar medium can provide clues to the mass loss history and thereby the nature of the progenitor.

A blast wave, driven by the supernova ejecta, initially with a velocity of ≳35,000 km s⁻¹ (Staveley-Smith et al. 1993), slowed down to ∼4000 km s⁻¹ after interacting with the H ii region created by the B3 Ia progenitor star in the swept-up red supergiant wind (Chevalier & Dwarkadas 1995; Dewey et al. 2012; Potter et al. 2014). The first interaction with the circumstellar ring occurred in 1995. This event was manifested by the appearance of the first hotspot (Sonneborn et al. 1998; Lawrence et al. 2000), where the blast wave entered a relatively dense clump of matter and suddenly slowed to velocities of a few 100 km s⁻¹. Additional hotspots erupted in dense clumps of gas over the next few years, gradually involving the whole ring. Emission at optical (Gröningsson et al. 2008a), X-ray (Maggi et al. 2012; Helder et al. 2013), and radio (Ng et al. 2013) wavelengths have since then undergone a steep rise.

We have followed the evolution of these shocks from the very beginning through regular monitoring with the Hubble Space Telescope (HST) and the Very Large Telescope (VLT). In this Letter we report observations of the ring system over the last ∼20 years, and show that the ring interaction has entered a new phase involving the final destruction of the ring but also interactions with material outside of this.

In order to make light curves of the ring we use all HST imaging observations in the R band (F675W and F625W) and B band (F439W, F435W and F438W) obtained since 1994. Details of all observations up to 2009 December are included in L11. Additional observations are given in Table 1. All images were processed in a standard way in order to remove cosmic rays and drizzled to combine dithered exposures (L11). Figure 1 shows a time series of the combined "R" and "B" band images, illustrating the evolution of the supernova during the last 20 years.

Zoom In Zoom Out Reset image size

The light curve of the ring was measured using an aperture in the form of an elliptical annulus, with semi-major axis 05 (11) for the inner (outer) boundary and an axial ratio of 0.78. Corrections were applied as described in L11 in order to account for differences between the different filters/instruments as well as the loss of charge transfer efficiency in the late WFPC2 images. The resulting light curves are shown in the top panel of Figure 2.

Zoom In Zoom Out Reset image size

From the images we can also determine the positions and fluxes of the hotspots. To define the hotspots, we used an ACS image taken in the F625W filter on 2006 December 6, which provides the best spatial resolution close to the time when the ring was brightest and when all hot spots are present (Figure 1). A two-dimensional Gaussian was fitted around each local maximum in the ring, resulting in the identification of 28 hotspots (see Figure 3). The distance of each hotspot from the center of the ring was measured by fitting a one-dimensional Gaussian to the intensity along a radial ray at the angular location of the spot in question. In images taken prior to the onset of a hotspot, this method returns the radial location of the ring. Figure 3 shows the time evolution of the spot positions, given as a distance to the mean position of the ring before the appearance of the spots.

Zoom In Zoom Out Reset image size

The velocities for each spot were determined from a linear fit to the positions after the impact, seen as a sudden decrease in the radius. For some of the spots there are indications of a decreasing velocity after ∼8000 days. Given the errors in the measurements, we have, however, limited the analysis to a linear least square fit from the time of the impact to the last observation. The resulting velocities range between 180 and 950 km s⁻¹ with a mean at ∼540 km s⁻¹. We estimate the errors in the velocities to be ∼50 km s⁻¹.

The flux from each spot was determined from the FWHM and peak of the Gaussian. The result for each epoch is shown in the middle panel in Figure 3, corrected for differences in filters and instruments in the same way as the total light curve.

To complement the imaging, high resolution spectra were obtained at regular intervals with VLT/UVES between 2000 December and 2013 December. These spectra allow us to separate the contribution from shocked and unshocked gas. Details of observations are given in Gröningsson et al. (2008b) for observations up to 2007 and later in Table 1. Reductions are similar, using the updated UVES pipeline. The 08 slit was centered at the supernova and oriented at a P.A. of 30°.

We estimate the accuracy of the fluxes from the UVES spectra by comparing them to those from HST imaging in the R and B bands, taking the slit width and seeing into account. We find that the UVES fluxes are within 20% of the HST fluxes, and that they also follow the same trend in time (Figure 2). The errors in the fluxes for most of the lines are dominated by this systematic error. However, for Hα at the last epochs the separation between the increasingly faint narrow component from the much brighter shocked component becomes increasingly difficult. This may introduce a systematic underestimate of the flux from the unshocked spots by up to ∼50% for the last observations, resulting in a similar increase in these fluxes. The Hα light curve would then be marginally flatter, but the general strong decrease should not be affected by this.

The light curves of the ring in the upper panel of Figure 2 are characterized by a slowly decreasing part up to day ∼5000, followed by a fast rise, which reached its peak at day ∼8000, after which they have decayed. The B band is dominated by emission lines from Hγ, Hδ and [S ii] λ4069, while the R band is dominated by Hα, [N ii] λλ6548, 6583, and [O i] λλ6300, 6364. The initial decay is a result of the recombination and cooling after the initial ionization from the shock breakout (Lundqvist & Fransson 1996). The strong increase in the flux began as the outward shock interacted with the whole ring, producing the hotspots, which brightened with time. This phase ended ∼8000 days after the explosion and the hotspots are now fading in our observations that extend to 9975 days.

The turn-on of the hot spots occurred between 5000–6000 days at the same time as the emission peaks moved inward (Figure 3). These dense "protrusions" may have recombined at early time due to their high density and only became visible when hit by the shock. The flux from the hotspots peaked earliest on the east side of the ring.

The high-resolution spectroscopy from the UVES observations allows us to separate the contribution to the light curve from the unshocked ring (emitting narrow lines with FWHM ∼10 km s⁻¹; Gröningsson et al. 2008b) and the shocked hotspots (emitting lines with typical velocities ∼300 km s⁻¹). In the middle panel of Figure 2 we show the light curves of three different lines from the shocked gas (Hα, [Fe xiv] λ5303, and [O iii] λ5007), which all show a similar evolution as the HST light curves. In the lower panel we show the fluxes of the strongest narrow lines from the unshocked gas (Hα, [N ii] λ6583, and [O iii] λ5007). Up to ∼7000 days the unshocked [O iii] and Hα lines increase in flux, caused by pre-ionization by the soft X-rays. After this epoch all three lines, however, decline more steeply than the lines from the shocked gas.

From the UVES observations of Hα we also find that the shocks are radiative up to at least ∼700 km s⁻¹, and possibly up to ∼1000 km s⁻¹, depending on the angle of the shocks relative to the line of sight (Gröningsson et al. 2008a; K. Migotto et al. 2015, in preparation). Using the cooling time from Gröningsson et al. (2008a) this indicates a density of up to $\sim 6\times {{10}^{4}}\;{\rm c}{{{\rm m}}^{-3}}$ in the clumps. This can be compared to the velocities derived from the proper motions in Section 2, where the average was found to be ∼540 km s⁻¹. This directly reflects the propagation of the shocks through the clumps and and concomitant acceleration of the post-shock gas. Although the radial and proper motion velocities are in perpendicular directions, the 43° inclination of the ring results in a similar correction, assuming that the expansion is radial from the explosion center. These two methods therefore result in similar shock velocities.

The most interesting recent changes are shown in Figure 4, with the emergence of new, faint spots, as well as diffuse emission, outside the ring. These new spots are most clearly seen in the final image from 2014, but with hindsight, some of these are faintly visible in the 2013 image. The new hotspots are a factor of 10–20 fainter than the old hotspots. The new spots, as well as diffuse emission outside the ring, can also be seen in the narrow-band images in Figure 4. It is interesting that most of the new spots are in the same region as where the faintest hotspots in the ring are present. The diffuse emission is likely to be a combination of Hα from the reverse shock region (France et al. 2015) and gas producing line emission outside the ring. The latter may be a result of photoionization of the wind from the red supergiant stage of the progenitor by X-rays from the shock.

Zoom In Zoom Out Reset image size

To interpret these observations we consider the ram pressure acting on the clumps, given by $P\propto {{\rho }_{{\rm ejecta}}}{{V}^{2}}\propto {{V}^{-n+2}}{{t}^{-3}}\propto {{t}^{n-5}}$, where n is the power-law index of the ejecta density profile, ${{\rho }_{{\rm ejecta}}}\propto {{V}^{-n}}$, V is the ejecta velocity at the reverse shock, assumed to be at a constant radius, and t is the time since explosion. The pressure will increase with time as long as the reverse shock is in the steep part of the ejecta density profile. This pressure drives transmitted shocks into clumps of gas in the circumstellar ring.

If the density of a clump is sufficiently high, radiative cooling will cause the clump to collapse to an even higher density. The result will be an optically emitting hotspot (Pun et al. 2002). If instead, the density of a clump falls below the threshold density for radiative cooling, the gas behind the transmitted shock will radiate X-rays and very little optical radiation. This threshold density is a sensitive function of the velocity, ${{V}_{{\rm clump}}}$, of the shock entering the clump. Thus, at any given time, the requirement for radiative cooling acts as a high contrast filter, limiting optical radiation to only the denser clumps. This may explain the still increasing soft X-ray emission (Maggi et al. 2012; Helder et al. 2013), in contrast to the optical. Dewey et al. (2012) find from the modeling of the X-ray observations a filling factor of ∼30% for the dense clumps.

If the radiative shocks can be characterized by one velocity the luminosity will depend on time as $L\approx {{\rho }_{{\rm clump}}}V_{{\rm clump}}^{3}A\propto \rho _{{\rm clump}}^{-1/2}{{t}^{(3n-15)/2}}A\propto {{t}^{5.4}}A$, where ${{\rho }_{{\rm clump}}}$ is the density and A is the cross section of the clumps and we have assumed n ≈ 8.6 (Shigeyama & Nomoto 1990). Explosion models show that the reverse shock is still likely to be in the steep density part of the ejecta (Dewey et al. 2012; Fransson et al. 2013). Further, the velocity for which the shocks are radiative is not likely to decrease because the density is increasing as long as $n\gt 3$. The most likely reason for a decrease in the luminosity is therefore that the area of the radiative shocks is decreasing. Alternatively, or in addition, as the driving pressure increases, the transmitted shock velocity may exceed the threshold for radiative cooling. Likewise, the threshold density for a clump to accommodate a radiative shock will increase with time. In either case it will lead to the clumps getting dissolved.

From the fading and acceleration of the hotspots we conclude that the clumps are destroyed by the blast wave, in agreement with hydrodynamic simulations (Borkowski et al. 1997; Pun et al. 2002). The fast decay of the narrow lines from the unshocked gas can be explained as a result of the non-radiative shocks, which traverse lower density gas, replacing the narrow line emission with soft X-ray emission, in combination with a decreasing emission measure of the pre-shock gas in the clumps.

The range of densities in the unshocked ring has been estimated to be $1\times {{10}^{2}}\;{\rm c}{{{\rm m}}^{-3}}$ up to $3\times {{10}^{4}}\;{\rm c}{{{\rm m}}^{-3}}$ (Mattila et al. 2010), while the shock speed from the X-ray imaging is ∼1820 km s⁻¹ (Maggi et al. 2012; Helder et al. 2013), coming from shocks propagating in the low-density component of the gas. Assuming that the interaction with the ring started on day 5600 (Helder et al. 2013), the blast wave has expanded by $\sim 7\times {{10}^{16}}$ cm up to day 9975, or ∼10 % of the radius of the ring. This is of the same order as the distance of the new hotspots outside the ring. It is therefore conceivable that the new hotspots are a result of either the interaction of the blast wave with very dense clumps embedded in gas with density $\sim {{10}^{3}}\;{\rm c}{{{\rm m}}^{-3}}$ outside the ring, and close to the equatorial plane, or pre-ionization by the X-rays. The fact that the main part of the ring has faded most in the south–east, where the majority of the new spots are seen supports the former interpretation.

The timescale for the complete destruction of the ring depends on the mass, geometry, and density distribution in the clumps (Pun et al. 2002). The mass estimate for the ring, $5.8\times {{10}^{-2}}$ ${{M}_{\odot }}$ (Mattila et al. 2010), only refers to the mass ionized by the initial shock breakout, and the mass of the hotspots and gas outside of the ring could be considerably larger. However, extrapolating the light curves in Figure 2 shows that the ring should fade away between ∼2020 (based on the individual lines) and ∼2030 (based on the HST light curves). The former estimate is likely to be more reliable, as the lines isolate the emission from the shocked ring component. The hotspots will be gradually destroyed by instabilities and conduction by the hot surrounding gas (Borkowski et al. 1997; Pun et al. 2002). The shocked gas which has cooled and increased its density by a factor 100–1000 may fragment and survive as overdense "nodules" in the hot gas, unless heat conduction destroys them (Poludnenko et al. 2004; Raga et al. 2007; Silvia et al. 2010). This depends sensitively on the magnetic field and its orientation. Heat conduction or X-ray irradiation may then provide some energy input to the surviving fragments, but at a low level. The cloud crushing time is $\sim 2{{r}_{{\rm clump}}}/{{V}_{{\rm clump}}}$ (e.g., Poludnenko et al. 2004). With a decay timescale of ∼2000 days and shock velocity of ∼500 km s⁻¹ this corresponds to a clump diameter of $\sim 0.9\times {{10}^{16}}$ cm. For the individual spots in Section 2 we find an FWHM of 1.5 pixels, consistent with being unresolved with the ACS. This corresponds to a diameter $\lt 0\buildrel{\prime\prime}\over{.} 04$, or $\lt 3\times {{10}^{16}}$ cm. We do not find any evidence of a change in the lateral direction within ±2°.

The structure outside the inner ring is complex with a possible hour-glass shape (Sugerman et al. 2005; France et al. 2015), which will now be possible to probe. As the shock progresses beyond the circumstellar ring, it will trace the history of mass loss from the supernovaʼs progenitor, revealing the distribution of gas that is now unseen, and providing useful information to discriminate among different models for the progenitor of SN 1987A.

The lower general density will result in a mainly adiabatic shock wave, except for clumps with densities comparable to the inner ring. However, the new hotspots we found probably only represent a small fraction of the mass outside the ring. We expect SN 1987A will become more thoroughly X-ray dominated as the youngest supernova remnant evolves.

We are grateful to the referee for detailed comments. This work was supported by the Swedish Research Council and the Swedish National Space Board, NASA grant NNX12AF90G, NSF grant AST-1109801. Support for the HST observing program was provided by NASA through a grant from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc. Partially based on observations collected at the European Southern Observatory, Chile (ESO Programmes 080.D-0727, 082.D-0273, 086.D-0713, 088.D-0638, 090.D-645, 092.D-0119, 094.D-0505).