SN 2005ip: A LUMINOUS TYPE IIn SUPERNOVA EMERGING FROM A DENSE CIRCUMSTELLAR MEDIUM AS REVEALED BY X-RAY OBSERVATIONS

Type IIn supernovae (SNe IIn) are a class of supernovae (SNe) that show prominent narrow H emission lines in their optical spectra (Schlegel 1990). There is a general consensus that these SNe occur in dense circumstellar media (CSMs) created by pre-SN mass losses, and that the narrow H lines are produced by photoionization of the dense winds irradiated by X-rays from the region behind the forward shock. However, the detailed nature of SNe IIn is still unclear because well-studied objects are sparse (See Taddia et al. 2013; Kiewe et al. 2012 for recent reviews).

In particular, their progenitors have not yet been established. One of the most important quantities to constrain the progenitor is a mass-loss rate, which varies from ∼10⁻⁶–10⁻⁵ M_☉ yr⁻¹ for red supergiants (RSG), ∼10⁻⁵–10⁻⁴ M_☉ yr⁻¹ for stripped Wolf–Rayet stars, and to ∼10⁻⁴–10 M_☉ yr⁻¹ for luminous blue variable (LBV) stars (Kiewe et al. 2012). We can infer mass-loss rates of progenitor stars by measuring Hα and X-rays from SNe, as they arise due to interactions between the CSM and ejecta. So far, mass-loss rates have been inferred for several SNe IIn. The values scatter within 10⁻⁴–10 M_☉ yr⁻¹, which is consistent with massive eruptions seen for LBV stars (Kiewe et al. 2012; Taddia et al. 2013). Indeed, possible detections of LBV-like progenitors in pre-SN images were reported for SNe 2005gl and 2009ip (Gal-Yam & Leonard 2009; Mauerhan et al. 2013; Pastorello et al. 2013).

SN 2005ip, discovered on 2005 November 5.163 in the nearby galaxy NGC 2906 (Boles et al. 2005), was initially designated a normal Type II SN, based on an optical spectrum taken one day after discovery (Modjaz et al. 2005). Later, however, Smith et al. (2009b) found clear narrow Hα emission even on day 1 as well as an unusually rich forest of narrow coronal lines at later times and categorized SN 2005ip as SNe IIn. Fox et al. (2009) reported strong IR emission over more than 900 days post-discovery. Subsequent studies on IR photometric and spectroscopic observations showed two dust components: one presumably formed in the ejecta or cool dense shell, and the other, preexisting dust in the surrounding CSM (Fox et al. 2010). X-ray emission from SN 2005ip was detected on 2007 November with Swift (Immler & Pooley 2007). For an assumed foreground column density of N_H = 3.7× 10²⁰ cm⁻² and a thermal spectrum with a plasma temperature of kT = 1 keV, an absorption-corrected (or intrinsic) X-ray luminosity was calculated to be 1.6 ± 0.3 × 10⁴⁰ erg s⁻¹ (Immler & Pooley 2007).

Previous optical and IR observations have led to different estimates of mass-loss rates for the progenitor of SN 2005ip. From the Hα intensity together with the CS wind speed (V_w = 100 km s⁻¹), Smith et al. (2009b) inferred that $\dot{M} =$ (2–4)× 10⁻⁴ M_☉ yr⁻¹ and argued that the progenitor is a RSG star having eruptive mass ejections like VY Canis Majoris (Smith et al. 2009a). By contrast, a large amount of warm dust contained in the CS shell led Fox et al. (2010) to suggest a very high mass-loss rate of 0.075–0.38 M_☉ yr⁻¹. Such a mass-loss rate exclusively favors a LBV-like progenitor rather than a RSG. Most recently, Moriya et al. (2013) estimated that $\dot{M} =$ (1.2–1.4)× 10⁻³ M_☉ yr⁻¹, based on analytical modeling of the bolometric light curve. This result further supports a LBV-like progenitor.

In this article, we report on archival X-ray monitoring observations of SN 2005ip from ∼1 yr to ∼6 yr post-discovery, acquired by Swift and Chandra. X-ray emission has been detected in all the observations, exhibiting clear spectral softening with time. The spectral evolution can be best interpreted by a significant decline of absorption due to the external CSM. The X-ray spectra enable us to measure a correct X-ray luminosity as well as an absorbing column density, allowing for a new estimate of the mass-loss rate, independent of the other previous ones. Throughout this article, we use a distance to SN 2005ip of 35 Mpc (Stritzinger et al. 2012) and adopt 2005 October 27.2 (nine days prior to the discovery) as the time of explosion (Smith et al. 2009b).

We analyze several archival Swift and Chandra data for SN 2005ip using the HEAsoft⁹ and CIAO¹⁰ packages. Detailed information of these observations is summarized in Table 1. We combine the latest five Swift observations taken in 2012 March into one data file owing to the poor photon statistics for individual data sets. As a result, we have five observation epochs, i.e., 2007 February (Swift), 2008 June (Chandra), 2008 November (Swift), 2009 November (Swift), and 2012 March (Swift).

We extract X-ray spectra from a circular region with a radius of 30'' (3'') for Swift (Chandra) observations and subtract background from the surrounding annular region. The background levels to the source are quite small: ∼0.3% (Chandra) and ∼3% (Swift). Each spectrum is grouped into bins, so that each bin contains at least five counts. We have checked that even if we set the number of the minimum grouping count to be unity, we obtain the same fit results shown below. The background-subtracted spectra are shown in Figure 1. Clearly, the spectra are softening with time, as is evident in hardness ratios summarized in Table 1.

Zoom In Zoom Out Reset image size

For spectral modeling, we apply an absorbed (TBabs; Wilms et al. 2000) thermal emission model in ionization equilibrium, i.e., the apec model (Smith et al. 2001) in XSPEC (Arnaud 1996). We allow the absorbing H column density (N_H), the temperature (kT), and normalizations to be free, while we treat the temperature to be common among all the five spectra in order to better constrain the parameters. Elemental abundances in both TBabs and apec models are set to the solar values (Lodders 2003), which are consistent with the metallicity of the host galaxy NGC 2906 (Stritzinger et al. 2012). To find the best-fit model, we use maximum likelihood statistics for a Poisson distribution, the so-called c-statistics (Cash 1979), which minimizes the C value defined as C = −2∑_i(M_i − D_i + D_i(lnD_i − lnM_i)) with M_i and D_i being the model-predicted and observed counts in each spectral bin i, respectively. This method is suitable for low-counts spectra that do not allow for the χ² test. Errors on the parameters can be estimated similarly to the χ² method, based on the fact that ΔC = C − C_min can be used similarly to Δχ² (Cash 1979), where C_min is the minimum C value obtained at the best-fit. As can be seen in Figure 1, the best-fit models represent the data well. The best-fit parameters are summarized in Table 2 (No. 1). We find a clear temporal decrease in N_H. This result is fairly robust from a statistical point of view, as is evident from confidence contours of kT versus N_H in Figure 2, where thick and thin lines correspond to the first and the final epochs, respectively. Judging from the high temperature (≳6.7 keV), most of the X-ray emission likely arises from a hot plasma in the CSM heated by the forward shock.

Zoom In Zoom Out Reset image size

Table 2. Spectral Properties from the One-component Model

Parameter		2007–02	2008–06	2008–11	2009–11	2012–03
NH (1022 cm−2)	1	5.38$^{+3.02}_{-2.18}$	3.12$^{+1.44}_{-0.87}$	1.35$^{+0.89}_{-0.51}$	0.86$^{+1.04}_{-0.68}$	0.21(< 0.5)
	2	5.35$^{+3.09}_{-2.16}$	3.11$^{+1.36}_{-0.87}$	1.33$^{+0.88}_{-0.49}$	0.85$^{+1.05}_{-0.66}$	0.21(< 0.41)
	3	5.33$^{+2.10}_{-1.55}$	3.08$^{+0.69}_{-0.61}$	1.33$^{+0.45}_{-0.38}$	0.86$^{+0.55}_{-0.47}$	0.22(< 0.46)
	4	1.62$^{+0.25}_{-0.22}$	Linked to 2007–02
kT (keV)	1	68.6(> 6.7)	Linked to 2007–02
	2	81.5(> 8.5)	Linked to 2007–02 with assumed evolution of t−0.2
	3	172.6(> 109.6)	Linked to 2007–02 with assumed evolution of t−0.53
	4	135.0(> 52.6)	100.0(> 26.3)	87.8(> 26.3)	7.1$^{+39.7}_{-3.2}$	3.0$^{+1.5}_{-0.9}$
Observed flux (10−13 erg s−1 cm−2)	1	5.69$^{+1.67}_{-1.40}$	6.00$^{+1.03}_{-0.93}$	8.11$^{+1.42}_{-1.27}$	7.80$^{+1.79}_{-1.56}$	4.46$^{+1.00}_{-0.87}$
	2	5.73$^{+1.68}_{-1.41}$	6.0$^{+1.0}_{-0.9}$	8.17$^{+1.43}_{-1.28}$	7.84$^{+1.80}_{-1.56}$	4.48$^{+1.0}_{-0.87}$
	3	5.73$^{+1.68}_{-1.41}$	6.04$^{+1.04}_{-0.93}$	8.17$^{+1.43}_{-1.28}$	7.80$^{+1.79}_{-1.56}$	4.40$^{+0.98}_{-0.86}$
	4	3.91$^{+1.14}_{-0.96}$	4.95$^{+0.85}_{-0.77}$	8.55$^{+1.50}_{-1.34}$	6.91$^{+1.59}_{-1.38}$	3.46$^{+0.79}_{-0.69}$
Absorption-corrected flux (10−13 erg s−1 cm−2)	1	9.65$^{+2.83}_{-2.37}$	8.94$^{+1.54}_{-1.38}$	10.86$^{+1.90}_{-1.71}$	9.91$^{+2.28}_{-1.98}$	5.03$^{+1.12}_{-0.98}$
	2	9.61$^{+2.82}_{-2.36}$	8.93$^{+1.54}_{-1.38}$	10.85$^{+1.90}_{-1.71}$	9.91$^{+2.28}_{-1.97}$	5.04$^{+1.12}_{-0.98}$
	3	9.61$^{+2.82}_{-2.36}$	8.93$^{+1.54}_{-1.38}$	10.85$^{+1.90}_{-1.70}$	9.91$^{+2.28}_{-1.98}$	5.01$^{+1.12}_{-0.98}$
	4	5.34$^{+1.55}_{-1.31}$	6.63$^{+1.14}_{-1.02}$	11.65$^{+2.04}_{-1.83}$	11.20$^{+2.57}_{-2.23}$	7.47$^{+1.71}_{-1.49}$
La (1041 erg s−1)	1	1.43$^{+0.42}_{-0.35}$	1.32$^{+0.23}_{-0.20}$	1.61$^{+0.28}_{-0.25}$	1.47$^{+0.34}_{-0.29}$	0.74$^{+0.17}_{-0.15}$
	2	1.42$^{+0.42}_{-0.35}$	1.32$^{+0.23}_{-0.20}$	1.60$^{+0.28}_{-0.25}$	1.47$^{+0.34}_{-0.29}$	0.75$^{+0.17}_{-0.15}$
	3	1.42$^{+0.42}_{-0.35}$	1.32$^{+0.23}_{-0.20}$	1.60$^{+0.28}_{-0.25}$	1.47$^{+0.34}_{-0.29}$	0.74$^{+0.17}_{-0.14}$
	4	0.79$^{+0.23}_{-0.19}$	0.98$^{+0.17}_{-0.15}$	1.72$^{+0.30}_{-0.27}$	1.66$^{+0.38}_{-0.33}$	1.10$^{+0.25}_{-0.22}$
C/dof	1	56.1/62
	2	56.1/62
	3	56.0/62
	4	93.1/62

Note. ^aCalculated in 0.2–10 keV at a distance of 35 Mpc.

Download table as:  ASCIITypeset image

We also fit the data with the same model in different ways that take account of temperature evolution. First, we assume that the temperature evolves as T∝t^−0.2, as the temperature is proportional to the square of the shock speed, which decreases as t^−0.1 in a self-similar solution (e.g., Chevalier 1982); the SN radius, R, is proportional to t^{(n − 3)/(n − s)}, where n and s are density profiles in the ejecta and the CSM, respectively (ρ_ej∝r⁻ⁿ and ρ_csm∝r^−s), so that the shock speed, dR/dt, is proportional to t^{(n − 3)/(n − s) − 1}, resulting in dR/dt ∼ t^−0.1 for typical density profiles of core-collapse SNe, n ∼ 10 and s ∼ 2 (e.g., Chevalier & Fransson 2003). The fit results listed in Table 2 (No. 2) are found to be consistent (except for the temperature) with those of the No. 1 case. Second, we take into account effects of temperature non-equilibration between electrons and ions due to slow equilibration through Coulomb heating. The temperature nonequilibration may be applicable for SN 2005ip, in which X-rays are likely due mainly to an adiabatic forward shock (e.g., Fransson et al. 1996). To this end, we follow Nymark et al. (2009) who showed that the electron temperature is proportional to $V_{\rm sh}^{-4/3} t^{-2/3}$ for Coulomb heating (Equation (10) in Nymark et al. 2009). Using the relation V_sh∝t^−0.1, we obtain T_e∝t^−0.53. By fitting the five spectra with this temperature evolution taken account, we obtain the results shown in Table 2 (No. 3), which are consistent (except for the temperature) with the other cases. Finally, we allow kT to vary freely among the five spectra, while we let N_H be a common free parameter. This strategy fails to fit the data (C value of 93.1; the fit results are presented in Table 2 No. 4), indicating that the spectral softening is not simply due to cooling. Therefore, we are confident that the spectral softening is mainly due to evolution in the absorption.

While non-detection of X-ray emission from reverse-shocked ejecta is not so peculiar for SNe IIn (cf. Chandra et al. 2012a, 2012b), for completeness we also apply an absorbed two-component apec model, in which the two components are supposed to represent the reverse-shocked ejecta and the forward-shocked CSM (e.g., Fransson et al. 1996; but see also Nymark et al. (2006, 2009) who presented a more detailed modeling especially focusing on X-ray emission from the ejecta by coupling a spectral code to a hydrodynamical code). Since it is difficult to determine a number of spectral-fit parameters with the relatively poor photon statistics, we initially link temperatures and normalizations of the low-temperature (ejecta) component among the five spectra, according to a theoretical expectation by Fransson et al. (1996). The best-fit parameters are summarized in Table 3 (No. 1). The two-component model gives a slightly better fit than the one-component model: ΔC ∼ −4.5 for two extra free parameters (i.e., kT and normalization of the ejecta component). The slight improvement in the fit is not strongly significant, yielding a confidence level of 92% based on the F-test; we assume that the C values can be treated as χ² values and performed F-test using the fit-statistics listed in Table 3 (No. 1). It should be noted that, as can be seen from the best-fit parameters listed in Tables 2 and 3, the basic best-fit parameters, i.e., N_H, kT, and normalization of the high-temperature component, derived by the one- and two-component models, roughly agree with each other.

Table 3. Spectral Properties from the Two-component Model

Parameter	2007–02	2008–06	2008–11	2009–11	2012–03
NH (1022 cm−2) 1	7.24$^{+3.27}_{-2.19}$	5.06$^{+1.29}_{-1.64}$	3.14$^{+1.04}_{-1.34}$	2.78$^{+1.15}_{-1.05}$	2.32$^{+0.91}_{-1.29}$
2	2.94$^{+0.79}_{-0.52}$	Linked to 2007–02
3	3.26$^{+0.98}_{-0.90}$	Linked to 2007–02
kTl (keV) 1	0.22$^{+0.70}_{-0.07}$	Linked to 2007–02
2	0.22$^{+0.07}_{-0.07}$	Linked to 2007–02
3	0.30$^{+0.15}_{-0.12}$	Linked to 2007–02 with assumed evolution of t−0.2
kTh (keV) 1	9.19(> 4.63)	Linked to 2007–02
2	88(> 36)	Linked to 2007–02
3	102.3(> 9.0)	Linked to 2007–02 with assumed evolution of t−0.2
$L_l ^{\rm a}$ (1041 erg s−1) 1	35.0$^{+10.4}_{-9.3}$	Linked to 2007–02
2	0(<5.2)	0(<5.8)	35.5$^{+20.7}_{-16.6}$	49.3$^{+34.7}_{-27.0}$	74.9$^{+26.9}_{-22.6}$
3	0(<1.46)	0(<2.63)	19.6$^{+11.3}_{-9.2}$	35.5$^{+24.1}_{-19.0}$	82.3$^{+29.7}_{-24.9}$
$L_h ^{\rm a}$ (1041 erg s−1) 1	1.73$^{+0.52}_{-0.43}$	1.58$^{+0.29}_{-0.26}$	1.90$^{+0.38}_{-0.33}$	1.81$^{+0.48}_{-0.41}$	0.91$^{+0.26}_{-0.22}$
2	1.04$^{+0.30}_{-0.25}$	1.29$^{+0.22}_{-0.20}$	1.84$^{+0.38}_{-0.34}$	1.80$^{+0.52}_{-0.45}$	0.80$^{+0.29}_{-0.23}$
3	1.04$^{+0.30}_{-0.25}$	1.29$^{+0.22}_{-0.20}$	1.82$^{+0.38}_{-0.34}$	1.76$^{+0.53}_{-0.45}$	0.80$^{+0.29}_{-0.24}$
C/dof 1	51.6/60
2	58.4/60
3	58.0/60

Note. ^aCalculated in 0.2–10 keV at a distance of 35 Mpc.

Download table as:  ASCIITypeset image

Next, to see if the spectral softening can be explained by the increasing low-temperature component rather than the N_H evolution, we fit the data with the two-component model in a different fitting strategy, in which we link the N_H parameter among the five epochs while we let the normalization of the low-temperature component vary freely for individual epochs. The best-fit parameters are listed in Table 3 (No. 2), from which we find that the model gives a reasonable but slightly worse fit than previous fittings. Moreover, the intrinsic X-ray luminosity for the low-temperature component is derived to be ∼10⁴³ erg s⁻¹ at the final epoch (∼6 yr after explosion). This may be unrealistically bright especially for reverse-shock emission; it is two orders of magnitude brighter than ever recorded at this late phase (Dwarkadas 2012). The intrinsic X-ray luminosity for the cooler component would be even greater than the value we obtained. This is because there is almost always a radiative cooling shell between a reverse shock and a forward shock, and the cooling shell will provide an extra absorption of the soft X-rays coming from the reverse shock. We have also checked that temperature evolution of T∝t^−0.2 for both of the two components (as is expected in a reasonable self-similar solution described above) does not affect the fit result (see the results No. 3 in Table 3). Accordingly, in the following discussion, we basically adopt the results from the one-component model in Table 2.

The archival X-ray data of SN 2005ip revealed clear spectral softening with time. Our spectral analysis showed that the spectral softening is best explained by declining evolution of the absorbing column density; N_H ∼ 5 × 10²² cm⁻² at the first epoch (∼1 yr after explosion) gradually goes down to be consistent with the Galactic absorption, N_H ∼ 4 × 10²⁰ cm⁻², at the final epoch (∼6 yr after explosion). On the other hand, we find that the intrinsic X-ray luminosity stays constant until the final epoch when it drops by a factor of ∼2 (see Table 2). These results suggest that the SN was initially embedded in a dense CSM, but the forward shock has penetrated the dense CSM region at a certain time between 2009 November (fourth epoch) and 2012 March (final epoch).

SN 2005ip is the second example in which we see possible evolution in absorption by the external CSM after SN 2010jl (Chandra et al. 2012a). By contrast, there are several SNe that do not show significant N_H evolution. These include SNe 2006jd (Chandra et al. 2012b), 1987A (Dewey et al. 2012), 1998S (Pooley et al. 2002), and 1999em (Pooley et al. 2002). SN 1993J would be intermediate between the two types, i.e., evolving absorption and constant absorption; Chandra et al. (2009) noted possible N_H evolution that may be caused by a cool shell between the forward shock and the reverse shock. A simple speculation to explain the difference between the two types is that the CSM is formed as a disk-like shell due to equatorial mass ejections. For this geometry, CSM absorption strongly depends on the inclination angle of the disk in the sense that strong absorption and its declining evolution is expected for an edge-on view (which would be responsible for SNe 2005ip and 2010jl), while little absorption for a face-on view (for others). An alternative scenario is that the class of SNe without evolution in absorption has somewhat peculiar CSM geometries (while the SNe class with strongly evolving absorption has either a relatively spherically symmetric CSM or an edge-on disk). For example, Chandra et al. (2012b) proposed for SN 2006jd that a low-column-density region is present in one direction (i.e., a cylindrical shape toward the observer) while strong interactions are taking place over much of the rest of the solid angle. In order to constrain the CSM structures, it is essential to increase SNe samples that allow for absorption measurements with Chandra, XMM-Newton, and Swift.

We estimate mass-loss rates of the progenitor star of SN 2005ip in two different ways. One is based on the column density, which is a robust measure in X-ray spectroscopy. We compare the measured column density with a theoretical calculation, i.e., Equation (4.1) in Fransson et al. (1996). As shown in Figure 3, the data are well represented by either $\dot{M} = 1.5\times 10^{-2}$ M_☉ yr⁻¹ and s = 2 (solid) or $\dot{M} = 4\,M_\odot$ yr⁻¹ and s = 3 (dotted) with reasonable wind and shock speeds (V_w =100 km s⁻¹ and V_sh = 14, 000 km s⁻¹; Smith et al. 2009b). Here, we should not put much weight on the last-epoch data, since the forward shock has likely penetrated the dense CSM region (resulting in a change in the s value) between the fourth epoch and the final epoch. From an astrophysical point of view, the latter case is not realistic; $\dot{M} = 4\,M_\odot$ yr⁻¹ is too high and s = 3 is too steep especially for Type IIn SNe, which are embedded in usually dense environments. Therefore, we favor the former case with the mass-loss rate of ∼0.015 M_☉ yr⁻¹ and s = 2.

Zoom In Zoom Out Reset image size

The other way is based on the X-ray luminosity, as is often used to estimate wind parameters in conjunction with Equation (3.11) in Fransson et al. (1996), which describes X-ray emission from the shocked CSM. The much larger absorption than that (N_H = 3.7 × 10²⁰ cm⁻²) assumed by Immler & Pooley (2007) boosts up the intrinsic 0.2–10 keV luminosity to be ∼1.5×10⁴¹ erg s⁻¹, which is by an order of magnitude higher than the original value of 1.6 ± 0.3 × 10⁴⁰ erg s⁻¹ reported by Immler & Pooley (2007). The revised luminosity places SN 2005ip as one of the brightest X-ray SNe among others being SNe 2001em (Pooley & Lewin 2004), 2005kd (Immler et al. 2007; Pooley et al. 2007; Dwarkadas 2012), 2006jd (Chandra et al. 2012b) and 2010jl (Chandra et al. 2012a). By integrating the equation from 0.2 keV to 10 keV, it can be rewritten as 2.77×10³⁶ (1−s)⁻¹ ($\dot{M}_{-4}$/V_w2)² (V_sh/10⁴ km s⁻¹)^{3 − 2s} (t/11.57 days)^{3 − 2s} erg s⁻¹ for T = 6 × 10⁸ K, where $\dot{M}_{-4} = (\dot{M} / 10^{-4}\, M_\odot$ yr⁻¹) and V_w2 = (V_w/100 km s⁻¹). The intrinsic luminosity during the plateau phase, first–fourth epochs when we can reasonably assume that s = 1.5 from 3 − 2s = 0, gives a mass-loss rate of $\dot{M} = (2\hbox{--}2.5)\times 10^{-2} V_{w2}$ M_☉ yr⁻¹, which is in agreement with the lower limit of the N_H-based $\dot{M}$ estimated above. Therefore, our X-ray-based mass–loss rates are significantly higher than those inferred from the Hα luminosity ((2–4) × 10⁻⁴ M_☉ yr⁻¹; Smith et al. 2009b) and the bolometric light curve ((1.2–1.4)× 10⁻³ M_☉ yr⁻¹; Moriya et al. 2013), but are close to the estimate from IR observations ((7.5–38)× 10⁻² M_☉ yr⁻¹; Fox et al. 2010). Our result supports eruptive mass ejections as seen in LBV stars.

Interestingly, the evolution in the X-ray luminosity, which is constant for the initial ∼4 yr and drops between the fourth epoch and the final epoch, is similar to that of Hα emission (Stritzinger et al. 2012). This similarity confirms a general picture that late-time Hα emission arises by reprocessing of X-ray emission. The decreasing X-ray and Hα luminosities at the final epoch lead us to suggest that the forward shock broke out a dense CSM region (i.e., entered into a steeper density-profile region) at some point between the fourth epoch and the final epoch. The distance that the forward shock (V_sh = 14, 000 km s⁻¹; Stritzinger et al. 2012) traveled for ∼5 yr, a time span for the forward shock to be within the dense CSM region, is calculated to be ∼2.2 × 10¹⁷ cm. This would be the spatial extent of the erupted material. Dividing the distance by the wind speed (V_w ∼ 100 km s⁻¹; Smith et al. 2009b), we can estimate the duration of the eruption to be ∼700 yr (t/5 yr) (V_sh/14,000 km s⁻¹) (V_w/100 km s⁻¹)⁻¹. Then, the total mass ejected in this eruptive period is calculated to be ∼14 (t/700 yr) ($\dot{M}$/0.02 M_☉ yr⁻¹) M_☉, indicating that the progenitor is quite massive (≳ 25 M_☉).

As mentioned above, the constant evolution in the X-ray luminosity (the plateau phase) requires that s = 1.5 in the framework of general SN evolutionary models (e.g., Fransson et al. 1996). At a glance, this appears to conflict with the N_H-based s-value ranging from ∼2 to ∼3. However, we should note that such SN evolutionary models assume spherically symmetric CSM structures, which may not be valid for SN 2005ip. For example, the CSM geometry may be a disk-like shell, as discussed earlier in this section. In this case, the constant X-ray luminosity is expected for a density profile of s = 2 instead of s = 1.5 (Stritzinger et al. 2012). On the other hand, evolution in N_H would be similar to that expected for a spherically symmetric CSM, if the disk is edge on. Therefore, a disk-like CSM structure with s = 2 would explain evolution in both N_H and luminosity simultaneously without invoking exotic conditions.

There still remains the possibility that the CSM structure of SN 2005ip is spherically symmetric. In this case, evolution in N_H and the luminosity leads to discrepant density profiles, which seems to be a real problem. By noting that absorbing material decreases more rapidly than what is expected from the luminosity evolution, one may think of a possibility that the external cool CSM is fully ionized by strong X-ray emission behind the forward shock. We thus calculate the ionization parameter, ξ = L/nr². For the X-ray luminosity of L =1.5× 10⁴¹ erg s⁻¹, a lower limit density of n = 3 × 10⁵ cm⁻³ (Smith et al. 2009b), and a radius of r = 10¹⁷ cm, which is a distance for a shock of V_sh = 10⁴ km s⁻¹ to propagate during ∼1000 days, we obtain that ξ ≲ 50. This is not sufficient to fully ionize C, N, and O, which are main absorbers of X-rays (cf. Chandra et al. 2012b). Therefore, photoionization is unable to explain the discrepancy in the derived density profile. The other possibility is to introduce clumpiness in the CSM wind, in which shocked clumps sustain the high X-ray luminosity (Chugai & Danziger 1994), while a covering fraction of dense clumps is so small that they do not noticeably affect dynamics of the system and the CSM absorption. In this model, a flatter distribution of clumps than that of a global CSM density profile can qualitatively explain both the constant intrinsic luminosity and the rapidly declining N_H. Given that the presence of small clumps is strongly suggested by the late-time development of intermediate-width component of Hα (Smith et al. 2009b), this scenario may be at work. The clumps will produce significant soft X-ray emission, which is not required in the one-component model but is seen in the two-component model (cf. Table 3). Therefore, the clump scenario cannot be excluded at the expense of a remarkably strong X-ray emission from the reverse shock.

X-ray spectra from SN 2005ip showed strong spectral softening with time, which we interpret to be due to declining in absorption by the external CSM. With the absorption taken into account, SN 2005ip turned out to be one of the brightest SNe IIn in the X-ray regime. Both the column density and the high luminosity require a mass-loss rate of at least ∼1 × 10⁻²V_w2 M_☉ yr⁻¹. Such a high mass-loss rate suggests eruptive mass ejections by a LBV-like progenitor. Based on evolution in the X-ray luminosity and N_H, we speculate that the CSM structure of SN 2005ip is a disk-like equatorial shell. Alternatively, a relatively spherical wind with small clumps could be another possibility at the expense of a remarkably bright reverse shock X-ray emission.

We are grateful to Drs. Jonathan Mackey and Takashi Moriya for kindly pointing out a correction to Figure 3 prior to publication. S.K. is supported by the Special Postdoctoral Researchers Program in RIKEN. The work by K.M. and T.N. is supported by World Premier International Research Center Initiative (WPI Initiative), MEXT, Japan. S.K. and K.M. are supported by Grant-in-Aid for Scientific Research for Young Scientists (25800119, 23740141). T.N. has been supported by the Grant-in-Aid for Scientific Research of the Japan Society for the Promotion of Science (22684004, 23224004).