Evidence for a Pulsar Wind Nebula in the Type Ib Peculiar Supernova SN 2012au

A low-resolution optical spectrum of SN 2012au was obtained with the 6.5 m Magellan telescope at Las Campanas Observatory on 2018 June 8. The Inamori Magellan Areal Camera and Spectrograph (IMACS; Dressler et al. 2011) was used with the f/4 camera in combination with the Mosaic3 array of eight thinned 2K × 4K × 15 μm E2V CCDs. A 300 line mm⁻¹ grating and a 09 long slit were used. Exposures of 2 × 3000 s were obtained and averaged. Resulting spectra have an effective wavelength range of 4520–10000 Å, with dispersion of 0.74 Å pixel⁻¹ and FWHM resolution of 5 Å (measured at 6000 Å). The spectrum has interruptions in coverage between detector gaps. The seeing was 06 and conditions were generally clear but not photometric.

Standard procedures to bias-correct, flat-field, and flux-calibrate the data were followed using IRAF. LA-Cosmic (van Dokkum 2001) was used to remove cosmic rays in individual images. Some cosmetic defects introduced from hot pixels and imperfect background subtraction have been manually removed. Spectrophotometric standards Feige 56 and LTT 3864 were observed and used for absolute flux calibration (Hamuy et al. 2002), which is believed to be accurate to within 30%. The spectrum has been corrected for the redshift z = 0.004483 of NGC 4790 (Theureau et al. 2007).

We obtained a 20 ks observation of SN 2012au with the Chandra X-ray Observatory (CXO) in combination with ACIS-S on 2018 August 2 through Director's Discretionary Time (PI: Patnaude; Observation ID 21660). Data have be reduced with the CIAO software v4.10 and corresponding calibration files. In a 2'' radius region centered on the SN, we detect a count rate of (3 ± 3.5) × 10⁻⁴ cts s⁻¹ (0.5–10 keV), consistent with background. Thus, no X-rays are detected from SN 2012au. The neutral hydrogen column density in the direction of SN 2012au is 3.8 × 10²⁰ cm⁻² (Kalberla et al. 2005). For an assumed non-thermal spectrum with index Γ = 2, we derive an unabsorbed flux limit of 2.8 × 10⁻¹⁵ erg s⁻¹ cm⁻², corresponding to luminosity L_X < 2 × 10³⁸ erg s⁻¹ (0.5–10 keV).

In Figure 1, we present our day 2270 spectrum of SN 2012au. Forbidden oxygen transitions [O i] λλ6300, 6364, [O ii] λλ7319, 7330, and [O iii] λλ4959, 5007 are clearly observed. The [S iii] λ9531 line is also detected but its doublet line [S iii] λ9069 that is intrinsically ≈1/3 the strength is not recovered from noise largely associated with the subtraction of telluric lines and fringing. Broad but weak emission centered around 7775 Å is also observed, which we identify with O i λ7774. Emission line fluxes and luminosities are listed in Table 1.

Zoom In Zoom Out Reset image size

All forbidden line profiles exhibit a clear asymmetry toward blueshifted wavelengths, peaking at −700 ± 50 km s⁻¹ (Figure 2). We measure the expansion velocity of [O iii], [O i], [O ii], and [S iii] using the half-width-zero-intensity shortward of 4959 Å, 6300 Å, 7325 Å, and 9531 Å, respectively. All measurements extend to approximately 2300 ± 100 km s⁻¹. The weakly detected O i 7774 has no clear peak in emission and does not appear to have the same emission line profile as the forbidden transitions.

Zoom In Zoom Out Reset image size

Excess flux is observed between 5100 and 5400 Å. This could potentially be the remnant of the plateau of emission observed in the day 321 spectrum (Milisavljevic et al. 2013; Figure 3). However, contaminating flux from the host galaxy cannot be ruled out. The fall-off in flux below 4800 Å is attributed to rapid loss of instrument sensitivity at these shorter wavelengths.

Zoom In Zoom Out Reset image size

No narrow (FWHM < 200 km s⁻¹) features are observed in the spectrum. Particular attention was made to possible emission in the region of Hα, which would be indicative of interaction with H-rich CSM. The 2D spectrum shows narrow emission lines of [O i] 6300, 6364, [N ii] 6548, 6583, and Hα extended spatially across the slit but nothing conspicuous and specific to the SN location (Figure 1, bottom). After careful subtraction of the local background we estimate an upper limit of Hα emission to be <4.1 × 10⁻¹⁷ erg s⁻¹ cm⁻².

The emission line intensity ratios of our spectra can be used with atomic rates from CHIANTI (Del Zanna et al. 2015) to constrain properties of the ejecta. The intensity ratio O i/[O ii] is a sensitive temperature diagnostic and the measured ratio of ≈0.27 indicates a temperature of T ≈ 5000 K. However, this should be viewed as an average temperature of the O⁺ zone. Because the line profile of O i does not closely follow those of the other lines (Section 3.1), some emission may originate from a different region of the ejecta and include cool gas that does not emit in the collisionally excited lines.

The [O iii]/[S iii] ratio reflects a higher temperature of the O⁺⁺/S⁺⁺ zone. If we assume that (S⁺⁺/S) = (O⁺⁺/O), then the measured ratio of [S iii]/[O iii] = 0.146 can be used in the expression

$$\begin{eqnarray*}&&[{\rm{S}}\,{\rm{III}}]/[{\rm{O}}\,{\rm{III}}]=5.98\times [{\rm{S}}/{\rm{O}}]\times \exp (12800/T)\end{eqnarray*}$$

to give

$$\begin{eqnarray*}&&{\rm{S}}/{\rm{O}}=0.024\times \exp (-12800/T).\end{eqnarray*}$$

The assumption of equal ionization fractions for S and O is plausible, but it potentially introduces an uncertainty of a factor of 2.

Notably, [S iii] λ9531 line emission is observed, but [S ii] λλ6717, 6731 is absent. This suggests electron densities above 10⁴ cm⁻³. Using the ratio of S/O from the equation above at 10⁴ K, assuming (S⁺/S) = (O⁺/O) in the singly ionized zone, and that $\mathrm{log}T$ is at least 3.7, then the ratio of [S ii]/[O ii] as a function of density can be determined. Estimating the upper limit to emission centered around the [S ii] λλ6716, 6731 lines to be ≈0.2 × I ([O ii]), we find that $\mathrm{log}T=3.8$ and $\mathrm{log}n\gtrsim 6.0$.

We thus conclude a sulfur to oxygen ratio ∼0.01 and a density of $\mathrm{log}n\gtrsim 6.0$. The high density is likely associated with significant clumping of ejecta, because a uniform sphere at that density with the radius R ∼ 4.3 × 10¹⁶ cm given by the time since explosion and observed expansion velocity would produce far more than the observed luminosity. The line profiles of all the forbidden oxygen lines are roughly the same (Figure 2), indicating that they are likely colocated, but with O⁺⁺ associated with a lower density, higher temperature zone.

Typically, late-time spectra of SNe I at stages reaching 1 yr are still powered by radioactive ⁵⁶Co in the ⁵⁶Ni → ⁵⁶Co → ⁵⁶Fedecay chain. However, radioactivity is not a plausible heating source for the oxygen-rich ejecta at t = 6.2 yr considering ≈0.3 M_⊙ of ⁵⁶Ni was produced in SN 2012au (Milisavljevic et al. 2013; Takaki et al. 2013) and ⁵⁶Co has a half-life of 77.3 days (Alburger et al. 1989). Furthermore, radioactivity is typically associated with neutral and singly ionized lines (see, e.g., Matheson et al. 2001), and yet our 6.2 yr spectrum exhibits a larger range of ionization levels.

SN–CSM interaction is the most common late-time emission mechanism for objects observed >3 yr post-explosion (see Milisavljevic et al. 2012; Chevalier & Fransson 2017, and references therein). Optical emission principally originates from a reverse shock that propagates upstream into outward expanding ejecta that gets heated and ionized. Such late-time detections all have clear signatures of interaction with an H-rich CSM including development of narrow emission lines and/or high-velocity H-rich ejecta (Milisavljevic & Fesen 2017; Figure 4). SNe Ib/c examples include SN 2001em (Chugai & Chevalier 2006), SN 2014C (Milisavljevic et al. 2015), and SN 2004dk (Mauerhan et al. 2018). A few SLSNe-I have exhibited hydrogen emission several hundred days after explosion (Yan et al. 2017). This contrasts with SN 2012au, which shows no spectral features indicative of SN–CSM interaction.

Zoom In Zoom Out Reset image size

If SN 2012au was indeed interacting with H-rich material, then H Balmer line emission would be expected. Certain H-rich CSM distributions could potentially inhibit optical emission at the time of observation, but such configurations are considered to be unlikely. It is also possible that SN 2012au is interacting with an H-poor environment. This scenario has been discussed in the context of SLSNe-I (Chatzopoulos & Wheeler 2012), but the detailed spectroscopic properties predicted for helium/carbon/oxygen-rich CSM interaction at these extremely late epochs is poorly explored.

Another late-time emission mechanism is pulsar interaction with expanding SN gas. A PWN is generated by the spin-down power of a central pulsar. In this scenario, photoionization of the inner regions of the expanding shell of ejecta can be the dominant source of optical line emission, especially at early times (<10 yr) when the expanding ejecta absorb much of the PWN-emitting ionizing radiation.

The Chevalier & Fransson (1992) model is based on the young Crab Nebula and has been widely used for pulsar and magnetar nebulae (e.g., Kasen & Bildsten 2010). The freely expanding SN density profile is approximated by an inner, flatter power law density profile surrounded by a steeper one. The inner profile has an index of m = 1 and the outer n = 9 (Chevalier & Soker 1989; Matzner & McKee 1999). For the SN 2012au parameters E_k ∼ 10⁵² erg and M_ej ≈ 4 M_⊙, the transition between these power laws occurs at 18,000 km s⁻¹. Thus the SN 2012au PWN is well within the inner, flatter region of the freely expanding gas. From Equation (2.11) of Chevalier & Fransson (1992), the velocity V of the PWN is

$$\begin{eqnarray}&&V=289{\dot{E}}_{39}^{0.25}{E}_{k51}^{0.25}{M}_{\mathrm{ej}1}^{-0.5}{t}_{\mathrm{yr}}^{0.25}\quad \mathrm{km}\,{{\rm{s}}}^{-1},\end{eqnarray} \tag{ 1 }$$

where ${\dot{E}}_{39}$ is in units of 10³⁹ erg s⁻¹, E_k51 in 10⁵¹ erg, M_ej1 in 10 M_⊙, and t_yr in years. With V = 2300 km s⁻¹ (assuming the emission is coming from close to the shell), E_k51 = 10, M_ej1 = 0.4, and t_yr = 6.2, we have $\dot{E}={10}^{40}$ erg s⁻¹. The Crab pulsar currently has $\dot{E}\approx 4\times {10}^{38}$ erg s⁻¹ (e.g., Condon & Ransom 2016) and extrapolating back with constant braking index gives an initial $\dot{E}\approx 4\times {10}^{39}$ erg s⁻¹, quite close to what is needed for SN 2012au. The Crab magnetic field is B ≈ 4 × 10¹² G (e.g., Condon & Ransom 2016), well below the magnetar field B > 10¹⁴ G. Hence, the pulsar in SN 2012au is potentially more Crab-like than magnetar-like.

The Chevalier & Fransson (1992) model includes line estimates for an O zone. It predicts an [O iii] luminosity at 1500 days to be 0.46 × 10³⁸ erg s⁻¹, which is close to the 1.4 ×10³⁸ erg s⁻¹ observed in SN 2012au. [O i] is weak because a thermal instability cools the gas, but somewhat different parameters could increase [O i]. [O ii] is weak because of low temperature and high density, but that could also change. The high energy of SN 2012au should result in a considerably lower density that the PWN is moving into than that in the Crab.

Chevalier & Fransson (1992) assumed that the swept up shell would be broken up by instabilities so the ionizing radiation from the PWN photoionizes freely expanding gas ahead of the shock front. In that case, there are different layers of ionized and neutral O in the O zone. However, the similar line profiles for [O i], [O ii], and [O iii] (Figure 2) do not support this picture.

A radiative shock wave driven into the freely expanding ejecta by the PWN is another possible excitation mechanism. However, the efficiency of this process is low, $L\sim 0.015\times \dot{E}$. Also, one would expect the characteristic boxy line profile for shell emission, which is not observed.

The most likely scenario seems to be photoionization of O zone gas that has been shocked by the high-pressure PWN and subjected to instabilities. In the Crab, the photoionization is of the H/He zone. Both SN 2012au and the Crab show fairly centrally peaked line profiles, as would be expected if Rayleigh–Taylor instabilities mix gas to the central region. Blondin & Chevalier (2017) performed a simulation of this process (see Figure 3 of that paper). Dense gas is in both Rayleigh–Taylor fingers and an outer shocked shell. Photoionization layers are anticipated, but on a small scale, so the line profiles of different ions will be similar. The compression in the PWN-driven shock also leads to the high density deduced from line diagnostics (Section 3.2).

A prediction of the above pulsar model is broadening of the emission lines with time because of the acceleration of the pulsar bubble. Following Chevalier & Fransson (1992), one would expect roughly steady pulsar power leading to R ∝ t^1.25 for the swept up shell or velocity V ∝ t^0.25. Specifically for SN 2012au, we anticipate

$$\begin{eqnarray*}&&\delta V/V=0.25\,\delta t/t=0.25(1\,\mathrm{year}/6.2\,\mathrm{years})=0.04\end{eqnarray*}$$

in one year, or a rate of increase in the emission line velocity width of ≈4% yr⁻¹.

In the above model, it is assumed that the initial pulsar spin-down timescale is much greater than the age. If the spin-down timescale is much less than the age, the pulsar input occurs early and the shell tends to a constant velocity; the shell comoves with the freely expanding gas and V = R/t. The kinetic energy of the shell depends on the mass, which can be found from the freely expanding density profile, and velocity V, yielding E_k = 2.5 × 10⁴⁸ erg. The energy of freely expanding gas is 1.3 × 10⁴⁸ erg, which implies that the initial rotational energy of the pulsar, E₀, was 1.3 × 10⁴⁸ erg, corresponding to a rotation period of 0.13 s. This is longer than the millisecond periods found in the magnetar models for SLSNe and is due to the V⁴ dependence of E_k (Equation (1)); line widths in the magnetar model approach ∼12,000 km s⁻¹ and remain constant.

We favor the long spin-down model because it implies a current pulsar power that is roughly consistent with that needed to produce the observed luminosity. A prediction of this model is increasing PWN velocities, as opposed to constant velocities in the short spin-down case. Deceleration is not expected in a PWN model.