CLUES TO THE NATURE OF SN 2009ip FROM PHOTOMETRIC AND SPECTROSCOPIC EVOLUTION TO LATE TIMES

Photometric monitoring of SN 2009ip began on 2012 September 22 with the spectral camera on Faulkes Telescope South (FTS, Siding Spring Observatory, Australia) in filters g'r'i'. The spectral camera is a 4096 × 4096 pixel Fairchild CCD with a 105 field of view. As Faulkes South monitored the rise of SN 2009ip, LCOGT's first three southern 1 m telescopes arrived on site at Cerro Tololo Inter-American Observatory in Chile. Photometric monitoring of SN 2009ip from CTIO began when the telescopes were operational, about one month later on UT 2012 October 21, with the first generation deployment camera: SBIG CCDs with a 15' field of view. In Figure 2, we show one of the first LCOGT r' images of SN 2009ip, and in Figure 3, we show the full LCOGT light curve, indicating the point at which the CTIO 1.0 m came online. We monitored SN 2009ip until the end of 2012 with as close to daily cadence as possible, in Johnson–Cousins filters UBVRI, Sloan filters g'r'i', and Pan-STARRS z-short. When SN 2009ip was again accessible in 2013 May, we monitored it with the 1.0 m network, including the new site at the South African Astronomical Observatory (SAAO), with near daily cadence in BVg'r'i'. The reduction and calibration of our photometry is discussed in detail in the Appendix. The photometry has been corrected for Milky Way extinction along the line-of-sight only where explicitly mentioned; otherwise, calibrated observed apparent magnitudes are presented. We present our photometry from the LCOGT 2.0 Faulkes Telescope and CTIO 1 m telescopes in Table 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Spectroscopic monitoring of SN 2009ip began on UT 2012 September 25 with the new FLOYDS robotic spectrograph on the 2.0 m FTS at Siding Spring Observatory in Australia. A twin spectrograph is located at the Faulkes Telescope North (FTN) on Haleakala, Hawaii. These robotic spectrographs have low resolution and were specifically designed for SNe. The design is a simple folded Schmidt camera, with a double-pass prism and reflective grating. FLOYDS uses a standard reflection grating (235 l mm⁻¹) as the primary disperser, with a cross-dispersed prism to image the first (5400–10000 Å) and second (3200–5700 Å) order light onto the CCD. This provides an overall wavelength coverage from 3200 Å to 10000 Å in a single exposure. Slit widths range from 09 to 60 but only the 20 slit was used in this work. The resolution in the red is R ∼ 400, or 16.5 Å per resolution element, and in the blue is R ∼ 690, or 8.25 Å per resolution element.

The LCOGT FLOYDS spectrographs, and the automatic pipeline with which these spectra were reduced, will be presented in D. J. Sand et al. (in preparation). FLOYDS-FTS achieved first light on 2012 May 7 and was still in commissioning when SN 2009ip was observed. The automated scheduler is always requested to observe at the paralactic angle and obtain a nearby standard star. At that time, the robotic target acquisition software was still being refined, and some of the spectra suffer more from flux miscalibration and residual fringing (e.g., 2012 October 4) than FLOYDS spectra being obtained and reduced in the present day. The telluric removal is done in the automatic pipeline but we still label the locations of tellurics in our plots so that incomplete removals are obvious. All FLOYDS spectra are calibrated using a sensitivity function derived from a standard star spectrum observed on the same night and then are flux calibrated with LCOGT photometry from the nearest epoch.

Spectroscopy of SN 2009ip was obtained on UT 2012 September 22 with the Goodman High Throughput Spectrograph (Clemens et al. 2004) on the 4 m SOAR Telescope at Cerro Pachon in Chile. This spectrograph has a 4 k × 4 k Fairfield imaging CCD. We used the long slit with a 084 width, the 600 lines mm⁻¹ grating, and the GG385 order-blocking filter in the "mid-" wavelength option for 4350–7020 Å coverage. These settings result in a read noise of 3.99 electrons, a gain of 2.06 electrons per analog-to-digital units (ADU), and resolution of R ∼ 1500 at 5500 Å. Three 600 s exposures of SN 2009ip at the paralactic angle were reduced with standard IRAF procedures, corrected for extinction, flux calibrated with standard star LTT2145, and median combined.

Spectroscopy of SN 2009ip was obtained nightly from UT 2012 October 11 to 2012 October 15 with the RC Spectrograph on the 4.0 m Mayall Telescope at KPNO. These spectra were also presented in RM14 but have been reduced and calibrated independently for this work. During the first two nights, lower resolution spectra were obtained but, because the spectra from each of these five nights are very similar, we will only present and discuss the higher resolution spectra obtained on the last three nights. For these, the higher resolution grating KPC-22B (632 lines mm⁻¹) was used in the first order, giving a spectral resolution of ∼2.7 Å and coverage of λ = 5500–8000 Å. All exposure times were 1200 s, with five, three, and four exposures taken on 2012 October 13, 14, and 15, respectively. All Mayall spectroscopy has been calibrated using a sensitivity function derived from a standard star spectrum observed on the same night and then flux calibrated to match the photometric magnitudes of the closest epoch to the time of the spectra's acquisition.

Spectroscopy of SN 2009ip was obtained on UT 2013 June 12 with the Gemini Multi-Object Spectrograph on the 8.0 m Gemini South Telescope at Cerro Pachon, Chile (PI: Howell, program GS-2013A-Q-62). We used a 10 longslit and an exposure time of 900 s for both the B600 and R400 gratings, with the OG515 order blocking filter on the red side. This gave spectral resolutions of 2.7 and 4.0 Å on the blue and red sides, respectively, and coverage over λ = 3500–9500 Å. These spectra were reduced and combined with standard procedures.

Far from a smooth procession, the light curve of SN 2009ip is more of a roller coaster ride. The photometric evolution is characterized by bumpy features that occur with greater strength in bluer bands, as shown in Figures 4 and 6. For ease of discussion, in Figure 4, we label five main "eras" of the light curve: rise, decline, bump, knee, and ankle. Peak brightness is reached on modified Julian date (MJD) = 56,206 (2012 October 6) but the decline does not begin in earnest until MJD = 56,215 (2012 October 15). The light curve shows the obvious "bump" in all bands starting on MJD ∼ 56,232 (2012 November 1), which is discussed further in Section 3.2. A more subtle bump follows, seen most clearly in bands g, B, and U on MJD = 56,248 (2012 November 17). The "knee" in the light curve starts on MJD = 56,264 (2012 December 1), after which the g and i bands drop more quickly than the r, until the "ankle" is reached on MJD = 56,278 (2012 December 16) and the light curve flattens out. These features have been discussed in other publications and are evidence of interaction with an inhomogeneous medium, such as shells of CSM previously released during the LBV phase (or disks of material; e.g., Levesque et al. 2014).

Zoom In Zoom Out Reset image size

Can we associate specific features in the 2012-B light curve with known past eruptions of SN 2009ip? If an explosion at time t_expl releases ejecta material with velocity v_expl, then the expected time of interaction, t_i, with material from a past eruption moving at velocity v_erupt since time t_erupt is given by

$$\begin{equation} t_{i} = t_{\rm expl} + \frac{ v_{\rm erupt} \times (t_{\rm expl}-t_{\rm erupt}) }{ v_{\rm expl} - v_{\rm erupt} }. \end{equation} \tag{ 1 }$$

Smith et al. (2010) find that the bulk outflow velocity during the 2009 event was v_erupt = 550 km s⁻¹. Here, we consider the physical scenario of Smith et al. (2014), that a core–collapse SN occurred at the start of the 2012-A event, after which the 2012-B event was powered by the interaction of fast-moving SN ejecta with material from past eruptions. In this scenario, the modified Julian date for the time of explosion is t_expl = 56,148, and we measure v_expl ∼ 4500 km s⁻¹ from the center of the blue side Hα absorption in our high-resolution spectrum in Section 4. Although multiple velocity components of explosion ejecta have been observed (e.g., RM14), we use v_expl = 4500 km s⁻¹ as representative of the bulk of the material. In Figure 7, black stars represent the magnitudes of past eruptions from Pastorello et al. (2013), plotted as a function of t_i, the predicted date at which eruption material would be encountered by the SN ejecta.

Zoom In Zoom Out Reset image size

Of course, we do not expect that all features will align exactly; the fastest-moving ejecta will reach the CSM first, and the distribution of ejecta velocities will smear out the timescale of interactions. For example, RM14 find three distinct velocity components in the blue side absorption features of Hα from high-resolution spectra of the 2012 event, and past eruptions did not all have precisely v_erupt = 550 km s⁻¹. Additionally, the material from each eruption has a unique mass and density, and the ejected material is likely to be asymmetrically distributed in the circumstellar environment (e.g., Levesque et al. 2014; Smith et al. 2014). However, in Figure 7, the pattern of past eruptions from 2011 and 2009 does generally agree with the characteristics of the 2012-B event. The temporal spacing of the small deviations in the light curve is about 10–20 days, similar to the expected spacing of interactions between SN ejecta and past material from the 2011 eruptions. Figure 7 also suggests that the abrupt and permanent change in the decline rate at the "ankle" may be related to material from the 2009 eruptions.

The opposing physical interpretation of SN 2009ip put forth by RM14 is that the 2012-A event was a precursor eruption to a stronger eruption at the start of 2012-B, which was subsequently powered by the immediate interaction material ejected in the 2012-A event. This scenario is also plausible under our current analysis. If we instead reflect the historical light curve around the start of the 2012-B event, as done in Figure 7 for the start of the 2012-A event, we find that material ejected at the start of the 2012-B event would encounter the slower moving material from 2011 at around the time of the "bump." Ultimately, we cannot rule out either scenario based solely on a comparison of the 2012-B light-curve morphology with past eruptions. We continue this discussion of CSM interactions as a probe of past mass loss in Section 4.1, where we use the Hα narrow line emission to constrain the amount of mass lost in the 2011 eruption.

MF13 fit splines to their Swift UV and UBVRI photometry at the time of the light-curve "bump," starting on MJD ∼ 56,232 (2012 November 1), and find that the "bump" occurs earlier in the redder filters. The "bump" itself is likely caused by explosion ejecta encountering an inhomogeneity in the CSM. MF13 point out that dust is not likely to cause this timing offset between UV and optical, suggesting instead that the delay is caused by the UV photons being more susceptible to scattering and therefore taking longer to emerge. The Swift UV data published by MF13 does not exhibit a bump so much as a brief plateau before the decline resumes a couple of days after the optical peak for the bump; this may bias the "bump" UV peak computation. While the optical and UV clearly behave differently, the difference in their UBVRI filters is ambiguous, as the UBV peak dates only differ by ∼1 day. The spline fit to RI filters appears to peak several days earlier but MF13 suggest that undersampling in their RI filters may be the true cause of this difference.

With our near daily cadence and multiband coverage at the time of the "bump," can we add to this discussion? In Figure 8, we fit our UBVRI and griz photometry with splines and find the bump reaches its peak at the same day in all optical filters. In addition, Figure 9 shows that the evolution in B − V and g − r colors show a brief blueward dip, coincident with the time of the "bump" in the optical. After this, the colors resume their steady redward progression seen over the bulk of the 2012-B event. The presence of dust would cause a bigger change in magnitude in the redder filters and so is unlikely to contribute to the "bump." This "bump" is likely caused by the ejecta material encountering a denser region of the CSM. This is supported by the spectroscopic evolution in the Hα line at this time: the decline rate of the flux in the narrow component flattens out at the time of the "bump," as discussed further in Section 4.2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In Figure 9(b), we show the evolution of SN 2009ip during the 2012-B event in four colors: U − B, B − V, g − r, and r − i. We also fit blackbody temperature and radius of the emitting region to the LCOGT griz photometry and plot the evolution in Figure 10. Under the interpretation of the 2012-B event as thermal radiation of the CSM that has been excited by the collision with fast ejecta, a blackbody is a useful and appropriate physical approximation to the spectrum.

Zoom In Zoom Out Reset image size

Together, these figures show that the steady redward progression of all colors is a result of blackbody cooling. The color evolution shows several blueward deviations that coincide with bumps in the light curve. The first, at MJD ∼ 56,235 (2012 November 4), coincides with the bump described in Section 3.2. The second, at MJD ∼ 56,248, is most prominent in U − B and coincides with the second bump, seen most clearly in the UBVRI light curve of Figure 6. For both events, the temperature evolution in Figure 10 shows a brief plateau in the cooling at these times, indicating a temporary injection of energy. In general, the color, temperature, and radius that we present agrees well with its interpretation as the explosive ejecta interacting with CSM previously released by the LBV. Material from the 2011 eruptions traveling at v_erupt = 550 km s⁻¹ since MJD = 55,700, 490 days later on MJD = 56,190, would be at distance ∼2.3 × 10¹⁵ cm. This agrees with the radius of the blackbody emitting region having a radius of ∼10¹⁵ cm, as seen in Figure 10.

After the "knee," starting at MJD ∼ 56,270, the redward color evolution in g − r flattens out, and in r − i, it takes a blueward turn. These observations agree with the published color evolution for SN 2009ip. MF13 present a flattening in all optical colors at this time, with blueward dips after MJD ∼ 56,270 for all colors except V − R, for which the dip starts 5–10 days later. Similarly, RM14 show some flattening in their B − V color evolution at this time but only their UV–optical colors show such a solid blueward turn. Unlike the blue deviations associated with small bumps in the light curve, these blueward turns are not related to an increase in temperature. Instead, at this time, the light-curve decay is steeper in the i and z bands, as we show in Figure 4. The slower decline in the r band at the "knee" is likely influenced by the flux in the Hα emission line, which as we present and discuss in Section 4.2, is simply declining linearly at this time. RM14 remark that at this time, the near-infrared is decaying more rapidly and there is a clear UV-excess over their blackbody fits, and we agree that, in general, SN 2009ip is not well fit by a blackbody at this late time. For this reason, although Figure 7 suggests that the temperature and radius evolution around the time of the "ankle" is related to interactions with material from the 2009 eruptions, we caution that this may not be the correct interpretation.

We integrate our blackbody fits from Section 3.3 and plot the evolution in bolometric luminosity (L_bol) during the main 2012-B event in Figure 11. In order to understand the relative sizes of inhomogeneities in the CSM, we measure the fraction of bolometric energy in the features. To do this, we extend the rate of decline during the 10 days prior to the "bump" (MJD = 56,220–56,230) forward in time and predict a L_bol ∼ 1.3 × 10⁴² erg cm⁻² s⁻¹ for MJD = 56,235. On this day, the "bump" feature has raised the bolometric luminosity to L_bol ∼ 2.5 × 10⁴² erg cm⁻² s⁻¹, doubling the total energy output. This additional injection of energy, L_bol ∼ 1.2 × 10⁴² erg cm⁻² s⁻¹, is ∼10% of the bolometric luminosity at peak. As discussed in Section 3.1, multiple qualities of the fast-moving ejecta and the CSM together produce these bumpy features, and we cannot derive direct constraints (i.e., it does not directly follow that inhomogeneities in the CSM are 10% of the total mass). What is clear is that the clumps in the CSM are large enough to indicate that episodic eruptive mass loss occurs in the late stages of massive star evolution. Alternative explanations include that these bolometric fluctuations are consistent with pulsations in a surviving A-type star (Martin et al. 2013), perhaps the mark of a binary system as presented in Kashi et al. (2013). Finally, we note that although these fluctuations appear to be unique for SN 2009ip, the relatively sparse data sampling in past events such as PTF10tel, SN 1998S, and SN 1999el shown in Figure 5 could hide such features.

Zoom In Zoom Out Reset image size

In Figure 11, we extend the bolometric light curve out to late times. Since the photometry is not very well fit by a blackbody at late times, we also use the method of converting a core–collapse SN's late-time V-band magnitude to bolometric luminosity presented in Hamuy (2003). The results are plotted as green points in Figure 11 and agree within the (rather large) uncertainties of L_bol derived from blackbody fits. In addition, we show how the integrated bolometric luminosity builds up over time (blue line), with most of the energy released within the first ∼20 days, and a final total energy output of ∼3.2× 10⁴⁹ erg.

Under the simplifying assumption that the bulk of the emission originates in the kinetic energy of the ejecta with a bulk velocity of v_expl = ∼4500 km s⁻¹, the mass ejected in the 2012 explosion is ∼0.1 M_☉. This is only a small fraction of the supposedly 50–80 M_☉ progenitor star (Smith et al. 2010) and suggests that an insufficient amount of the star was unbound in the explosion for it to be a true SN. However, NS14 point out that this is actually a minimum ejecta mass estimate because the CSM is not isotropic (e.g., Levesque et al. 2014), allowing much of the ejecta to escape, and the main brightening may be caused by only the fastest-moving ejecta. NS14 furthermore point out that 0.1 M_☉ of ejecta would be optically thin within days and that this is directly contradicted by the broad Hα emission component that is present until late in the 2012-B event, as we show in Figure 16. They argue that this indicates ejecta masses of at least 4–6 M_☉.

When SN 2009ip emerged from behind the Sun in 2013 May, we resumed our photometric monitoring program in BVgri using the LCOGT 1.0 m network. The calibrated photometry for gri is shown in Figure 12. Although SN 2009ip is steadily declining, the slope is not constant: on MJD ∼ 56,410, a steeper decay event began and lasted ∼20 days. This event is steeper in the g and i bands compared to the r-band filter, as was the "knee" in Figure 4. To emphasize this dip, in Figure 12, we make linear fits to the photometry for dates after MJD = 56,440 (solid lines) and extend them backward (dashed lines). The best-fit slopes in all bands after this dip are <0.005 mag day⁻¹, flatter than the typical decay of ⁵⁶Co, which is ∼0.01 mag day⁻¹. The combined evidence of features in the late-time light curve and a slope shallower than a ⁵⁶Co decay indicates the optical emission from SN 2009ip is still influenced by ongoing CSM interaction. This is not uncommon for SN IIn, as shown in Figure 5, and contributions from ⁵⁶Co decay from a SN cannot be excluded. As a test, if we include only dates 56390 < MJD < 56,430 in the linear fit, the slopes increase to 0.01, 0.009, and 0.019 mag day⁻¹ in the g, r, and i bands, respectively. With less complete late-time coverage, a late-time slope consistent with ⁵⁶Co decay could be derived. Without LCOGT's high-cadence observations at late times, we may well have drawn different conclusions about the late-time decline of SN 2009ip.

Zoom In Zoom Out Reset image size

When the late-time decline of SNe is powered by the decay of ⁵⁶Co to ⁵⁶Fe, which has an e-folding time of 111 days (or a 77 day half-life), the decline rate is ∼0.01 mag day⁻¹. Type IIn SNe are often seen to remain bright and decline slower than this at late times for a long time (e.g., Zhang et al. 2012), due to the heated CSM. This is exemplified by SN 1988Z and SN 1995G in Figure 5, and also by SN 2009ip in Figure 12. The only place in the SN 2009ip light curve where ⁵⁶Co decay could have been the dominant source of optical emission is the ∼100 day window between MJD = 56,290 and MJD = 56,390, when SN 2009ip was behind the Sun. During this time, the r-band magnitude declined from 17.5 to 18.6 mag, which is an average of 0.01 mag day⁻¹. We can only use this decline to estimate an upper limit on the potential mass of ⁵⁶Ni synthesized in the explosion. To do so, we use our bolometric luminosity, adopt the earliest possible explosion date (the start of the 2012-A event on MJD = 56,150), and use the expression for nickel mass from Hamuy (2003). We find M_Ni < 0.04 M_☉, consistent with previously reported estimates (e.g., MF13). However, for SN 2009ip, this estimate of potential ⁵⁶Ni mass is not very restrictive on the explosion mechanism.

Future monitoring will reveal whether the decline of SN 2009ip exhibits further features of interaction or continues to steadily decline. The faintest recorded historical magnitude since detection as a transient is m_V = 21.46 mag on 2009 November 24 (AP13), and in archival HST images, it was measured to be m_F606W = 21.8 mag (Smith et al. 2010). At the current rate of r-band decline, SN 2009ip will not pass this magnitude until 2016 mid-January. A complete disappearance would confirm that these events were powered by a terminal SN. Given that massive stars are frequently in binary systems with other massive stars, this may not happen (Sana et al. 2012). If half of the flux in the archival HST images actually came from a binary companion, the future magnitude of the remaining companion may be m_F606W = 22.6, easily detectable with HST.

Under the assumption that the 2012-B event is powered by fast-moving ejecta interacting with a shell of material from the 2011 eruption, as discussed in Section 3.1, we can constrain the amount of mass lost during that episode. Ofek et al. (2013b) provide a thorough examination of multi-wavelength mass loss rate indicators for SN 2009ip. For our calculation, we follow their Equation (4) for the mass of hydrogen, M_H, from the luminosity of the narrow component of Hα, L_Hα:

$$\begin{equation} M_{\rm H} = \frac{m_p \ L_{{\rm H}\alpha }}{h \ v_{\rm H} \ \alpha ^{{\rm eff}}_{\rm H} \ n}, \end{equation} \tag{ 2 }$$

where m_p is the proton mass, h is Planck's constant, and $\alpha ^{{\rm eff}}_{\rm H}$ is the recombination coefficient at 10,000 K, which is appropriate for our use as we find a blackbody temperature of ∼10, 000 K in Section 3.3. The term n is the particle density profile, which we express as

$$\begin{equation} n = \frac{M_{\rm H}/m_p}{V}, \end{equation} \tag{ 3 }$$

where the volume V is the shell occupied on MJD = 56,192 by material ejected at 550 km s⁻¹ for the 200 days starting on MJD = 55,700, which was the duration of the 2011 eruptions. At the peak of the 2012-B event, the flux in the narrow component of Hα was F_Hα ∼ 5 × 10⁻¹³ erg cm⁻² s⁻¹, which we convert to a luminosity using a distance of 25.2 Mpc as in Section 3. We find that M_H ∼ 0.1 M_☉ of material was released during the 2011 eruptions but note that this is an upper limit, as it relies on the assumption that all of the hydrogen was ionized by the 2012-B ejecta. The implied mass loss rate during the 2011 eruptions is ∼0.18 M_☉ yr⁻¹. As described in Ofek et al. (2013b), this upper limit is significantly higher than mass loss rates derived from, e.g., X-ray observations, but this discrepancy is important because it supports an asymmetric distribution of CSM (e.g., a disk, as in Levesque et al. 2014). For example, if the CSM is in a disk with ∼10% the volume of the assumed shell, then M_H ∼ 0.01 M_☉.

The Hα line exhibits three main components in varying strengths: narrow emission, broad emission, and blue side absorption. These components represent the physical evolution of the ejecta and CSM interactions powering the 2012-B event of SN 2009ip. In Figure 16, we show a time series of the Hα line from our LCOGT FLOYDS spectroscopic series, labeled with the date and the phase era (rise, decline, bump, knee, and ankle). The five spectra taken with the KPNO Mayall RC spectrograph between October 11–15 fit into the "decline" era after October 9. In the KPNO spectra, the Hα emission is dominated by a narrow Lorentzian profile from the CSM emission, and Figure 15 shows how there is effectively no evolution in the spectrum, just as the photometry is nearly constant at this time in Figure 4.

Zoom In Zoom Out Reset image size

To the Hα emission, we fit a single narrow Lorentz profile for dates up to October 20, after which this is combined with a broad Gaussian profile (after first subtracting the continuum). Blue-shifted absorption is clearly present on September 26 and after October 30 but exhibits multiple components so we do not fit the blue-shifted absorption. The broad Gaussian emission is fit to the red side only, with a peak wavelength restricted to within 5 Å of the narrow line peak. In Figure 16, the red line represents this fit to the narrow and broad emission only because we find that not including the absorption components has actually highlighted its shape and contribution to Hα. In the following two sections, we discuss the possible physical sources of the emission and absorption components in turn.

4.2.1. Narrow and Broad Emission Features

Over the 2012-B event, the width of the narrow component of Hα appears to increase from FWHM ∼ 700 to 1000 km s⁻¹, or ∼16 to 20 Å, but this is within uncertainty. As the continuum and the flux in the narrow component decline, the broad component re-emerges at the start of the "bump" era. The width of the best-fit broad component appears to increase from ∼10,600 km s⁻¹ on 2012 October 30 to ∼15,000 km s⁻¹ on 2012 November 13, encompassing the duration of the "bump," after which it appears to decrease back to ∼10,600 km s⁻¹.

The evolution of the integrated, continuum-subtracted flux in the fit narrow and broad components over the course of the 2012-B event are plotted in Figure 17. The flux in the narrow component rises and declines similar to the light curve, as it should, because the continuum emission is also dominated by CSM interaction. The decline of the flux in the narrow component exhibits a plateau at the time of the "bump," which is consistent with our interpretation of the "bump" as explosion ejecta interacting with an inhomogeneity in the CSM. Line emission as broad as that seen at post-"bump" eras of SN 2009ip is typically associated with fast expansion, and this leaves us with several questions: what is the physical cause of the broad emission? Why does its width and integrated flux appear to increase and decrease? How is it related to the light-curve "bump"?

Zoom In Zoom Out Reset image size

The coincidence of the "bump" with an increase in the relative contribution from the broad component seems difficult to resolve with the underlying cause of the "bump" being the result of ejecta encountering an inhomogeneity in the CSM, which is the source of narrow emission. MF13 point out that interaction with a clump of CSM can cause broader emission through photon scattering. If so, we might expect to see a similar increase in the relative contribution from the broad component at the time of the "knee" and/or "ankle"—but Figure 17 has only a few data points at those times, and they suggest a continuing decline. Given that the "knee" is a lower energy feature, and that our spectroscopic sampling at this phase is insufficient to see short timescale changes in the Hα broad component, we cannot confirm the photon scattering hypothesis.

NS14 interpret the 2012-A event as a true SN and the source of the broad component of Hα seen at the start of 2012-A and during the decline of 2012-B. With this model, the broad component is always there but during the 2012-B event, is swamped with continuum and narrow line emission from the subsequent interaction between SN ejecta and CSM. As the continuum luminosity declines, the broad component from the SN is again revealed, and it is simply a coincidence that it re-emerges at the same time as the ejecta encounters a CSM clump and creates the "bump." In this case, the width and luminosity of the broad component may only appear to rise and decline because the continuum is receding. NS14 further support this interpretation by showing that the Hα equivalent width, EW = |∫1 − (F_λ/F₀)dλ|, is constantly increasing, as is commonly seen for SNe (NS14). We also measure the EW of Hα for FLOYDS spectra, and find it increases from 100 to 1000 Å from the "rise" to the "ankle," in agreement with NS14. Note that a continually increasing EW is not at odds with the continually declining integrated line flux in Figure 17: this happens in the case where the continuum flux is declining faster than the line flux.

There remains the third interpretation that the broad emission is generated by another nonterminal eruption of the LBV star. This would explain both the apparent rise and decline of the broad component emission and the appearance of multiple new velocity components of Hα absorption presented by RM14 (discussed further in Section 4.2.2). In the scenario of continuing eruptions of an LBV-phase star, this would be third in a line of three eruptions spaced ∼40 days apart. Although the photometric evolution of the 2012-B event is quite distinct from the LBV phase, the 40 day spacing is similar to that seen during the 2011 eruptions in Figure 1. In this scenario, the light-curve "bump" would not represent a CSM inhomogeneity but newly ejected material interacting with the CSM. However, we are skeptical that the widths implied by our fits to the broad component are ∼twice that shown in the first spectrum from 2012 September 22 in Figure 13. How could there be a third powerful eruption without a correspondingly large increase in the continuum luminosity? We suspect that the simplest explanation is that the fit width and integrated flux of the broad component is compromised by the receding continuum during this era.

4.2.2. Absorption Features of the Ejecta Components

We obtained a late-time spectrum of SN 2009ip, shown in Figure 18, with the Gemini Multi-Object Spectrograph on Gemini South Telescope on 2013 June 12, when SN 2009ip was 249 days past the maximum brightness of the 2012-B event. At this time, it is similar to the spectrum taken on 2012 December 16 (the "ankle") in Figure 14, except that at these later times and in a higher resolution spectrum, the contributions from Fe and possibly O are more obvious and the He and individual Ca lines more distinguished. In Figure 19, we compare the late-time spectrum of SN 2009ip with spectra from an SN impostor, SNe IIn, and SNe IIP taken at similarly late phases, and also with LBV-phase spectra of SN 2009ip. These spectra were obtained from the Online Supernova Spectrum Archive (SUSPECT¹¹) and the Weizmann Interactive Supernova Data Repository (WISEREP¹²; Yaron & Gal-Yam 2012).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The late-phase spectrum of SN 2009ip is similar to SN 2008S, which was defined as an SN impostor by Smith et al. (2009). SN 2009ip has a flatter continuum flux and stronger helium lines but the Hα and calcium lines are equivalently strong. The major difference between SN 2009ip and SN impostor 2008S is the presence of a broad Hα component for SN 2009ip, indicating the high expansion velocities of a true SN, as discussed in NS14. Such broad emission is not seen for SN impostors.

At phases of ∼100 days, MF13 find the spectrum of SN 2009ip is well matched to SN 1998S, a peculiar SN IIn (e.g., Fassia et al. 2000). In Figure 19, we compare with the ∼300 day spectrum of SN 1998S (Pozzo et al. 2004). SN 1998S shows a strong, broad Ca ii emission, similar to SN 2009ip. The triple peak in the Hα line had emerged by day ∼230, as shown by Gerardy et al. (2000), who suggest the triple velocity components are caused by a ring-like disk structure in the CSM. This type of disk has also been suggested for SN 2009ip based on the Balmer decrement during the 2012-B event (Levesque et al. 2014). The fact that we do not see obvious multiple velocity components does not mean there is no disk, as orientation and relative velocities may differ from SN 1998S. In the inset of Figure 18, we look for asymmetry in the Hα line by reflecting it about the peak wavelength and comparing the blue and red sides. The excess on the red side could be attributed to asymmetry in the Hα emitting region, or as a depression in the blue side due to dust formation. Future confirmation of the CSM geometry will provide better constraints on the energy budget and the true explosion mechanism.

In Figure 19, we also compare our late-time Gemini spectrum with spectra for a typical SNe IIn, SN 1988Z at 115 days post-max (Turatto et al. 1993). At 1–4 months past peak, this object has similar spectral features to those shown by SN 2009ip in Figure 14, including asymmetry and blue-shifted absorption in the Hα feature. At these late times, their spectra are quite similar but SN 2009ip shows stronger, broader lines of He i and stronger and broader lines of Fe iii and Ca ii. At the chosen phase, SN IIn 1988Z still retains a broad component similar to that exhibited by SN 2009ip at earlier epochs.

For comparison with the next most similar kind of core–collapse SNe, we also show two late-time spectra from SNe IIP: SN 2004et at ∼250 days (Sahu et al. 2006) and SN 1990E at ∼250 days (Gómez & López 2000). Although similar to SN 2009ip in the presence of hydrogen and Ca ii, the SNe IIP show broader Balmer lines with P-Cygni profiles, stronger Ca ii emission, do not show the prominent He I, and have a weaker continuum in the blue end—a combination of a cool blackbody spectrum and possibly dust production. They also show prominent lines of forbidden oxygen at λ = 6300, 6346 Å, marked in light blue in Figure 19, which are not seen in the nebular epochs of SNe IIn 2009ip or 1988Z but are for 1998S. These lines of forbidden oxygen are also prominent in the Type Ic SNe associated with massive star progenitors (Gómez & López 2002). Although the late-time spectrum of SN 2009ip most resembles those of SNe IIn, to verify that a terminal SN explosion has occurred requires a positive identification of the forbidden lines common to the nebular phase spectra of core–collapse SNe. Even at late times, the true nature of SN 2009ip appears to be shrouded by CSM emission.

We also show a comparison to two spectra taken during the 2009 LBV-phase events of SN 2009ip: the thick line is 2009 September 25 and the thin line is 2009 October 22, both from Figure 5 of AP13. On September 25, SN 2009ip was just declining from its second detected outburst of the year, at R ∼ 19 mag. By October 22, it had been in a plateau at R ∼ 20 for 20 days, and the Hα emission had broadened slightly and lost its blue-shifted absorption feature. It is clear that SN 2009ip has not yet returned to its old quiescent self, as the Fe iii, He i, and Ca ii emission lines are all currently broader and stronger, and more similar to SN impostors and late-phase SN IIn. Only time will tell if SN 2009ip returns to its pre-2012, LBV-phase spectral signature.