Direct Analysis of Spectra of the Unusual Type Ib Supernova 2005bf

Among the hydrogen‐deficient Type I supernovae, the Type Ib (SN Ib) subclass is defined by the absence of strong Si ii and the presence of He i lines in the optical spectra. Type Ic supernovae (SNe Ic) are broadly similar to SNe Ib but lack conspicuous He i lines. SNe Ib and SNe Ic are generally thought to be core‐collapse supernovae whose progenitor stars lost most or all of their hydrogen and helium envelopes, respectively. (See Filippenko [1997] for a review of supernova spectral types.)

Supernova 2005bf was an exceptionally interesting Type Ib event (Anupama et al. 2005; Tominaga et al. 2005; Folatelli et al. 2006; hereafter A05, T05, and F06, respectively). The bolometric light curve reached an initial maximum about 15 days after explosion, but also a brighter second maximum (which hereafter we refer to as the maximum, without qualification) about 40 days after explosion. Such a double‐peaked light curve and such a long rise time to maximum light had not been observed previously. Both T05 and F06 invoked a double‐peaked radial distribution of ⁵⁶Ni to account for the light curve, although in other respects the models they discussed were quite different.

T05 and F06 attributed an absorption feature in very early spectra to Hα, forming at the high velocity of about 14,000 km s⁻¹, and F06 provided support for the identification by recognizing the presence of Ca ii and Fe ii features at the same velocity. A weaker absorption at the same wavelength in near–maximum‐light spectra was attributed by A05 to Hα and/or Si ii λ6355, while T05 favored Si ii, and F06 only remarked that by maximum light the early feature attributed to Hα had disappeared. In view of the unusual nature of SN 2005bf, and of our interest in the issue of whether SNe Ib and even SNe Ic eject some hydrogen during their explosions (Filippenko 1992; Branch et al. 2006, hereafter B06), we have carried out a direct analysis of selected spectra of SN 2005bf using a revised version (2.0) of the parameterized resonance‐scattering supernova synthetic spectrum code SYNOW.

Spectra of SN 2005bf are displayed and discussed in § 2, and our analysis of them is presented in § 3. In § 4, the significance of relationships between the spectra of SN 2005bf, the Type Ib SN 1999ex (Hamuy et al. 2002), and the Type Ic SN 1994I (Filippenko et al. 1995) is explored. The implications of our results are discussed in § 5.

Twelve spectra of SN 2005bf are shown in Figure 1. Following F06, epochs are with respect to the date of bolometric maximum light, 2005 May 9 (UT dates are used throughout this paper). Maximum in the B band occurred about 2.5 days earlier. The day −6, −4, and −2 spectra are from A05. The previously unpublished day +5 spectrum was obtained with the 3 m Shane reflector at Lick Observatory using the Kast spectrograph (Miller & Stone 1993); observations and reductions were similar to those of the day +2 spectrum, which was presented and described in F06. The other eight spectra are also from F06. Mild smoothing has been applied to some of the spectra. Because in this paper we are interested only in the spectral features, not in the shape of the underlying continuum, all observed and synthetic spectra are "flattened" by means of the local normalization prescription of Jeffery et al. (2007).

Zoom In Zoom Out Reset image size

From left to right, the dashed lines in Figure 1 correspond to He i λ5876 blueshifted by 7000 km s⁻¹, Hα λ6563 blueshifted by 15,000 km s⁻¹, and He i λλ6678 and 7065 blueshifted by 7000 km s⁻¹. He i absorptions blueshifted by about 7000 km s⁻¹ are clearly present at all epochs. (From careful measurements of the wavelength of the absorption minimum attributed to He i λ5876, T05 noticed that the blueshift increased slightly with time. This may have been due to the increasing strength of the feature; all else being equal, stronger features have higher blueshifts [Jeffery & Branch 1990].) The deep absorption attributed to high‐velocity Hα at early times becomes weaker between day −20 and day −6, but a feature persists at close to the same wavelength as late as day +5.

For the analysis presented in this paper, we have used a revised version (2.0) of SYNOW,³ which is described by Branch et al. (2007). New features employed here are (1) when using power‐law line optical depth profiles, different power‐law indices can be adopted for different ions; (2) a Gaussian line optical depth profile is now available, so that when detaching an ion from the photosphere, it is not necessary to introduce a discontinuity in the line optical depth profile; and (3) the output spectra can be flattened as in Jeffery et al. (2007). For this paper, the excitation temperature T_exc has been fixed at a nominal value of 7000 K. For ions that are not detached from the photosphere, a power‐law line optical depth distribution τ(v) = τ_p(v/v_phot)^-n is used, where v_phot and τ_p are the velocity and the line optical depth at the photosphere, respectively. We use n = 8 as the default value of the power‐law index, but other values are occasionally adopted to improve the fits. For ions that are detached from the photosphere, a Gaussian line optical depth distribution τ(v) = τ_g exp ^{-[(v - v_g)/σ_g]2} is used; i.e., the maximum line optical depth τ_g occurs at velocity v_g. We refer to ions that are undetached or only mildly detached as photospheric velocity (PV) ions, and those that are detached at high velocity (HV) relative to the photosphere as HV ions.

In this section, we cover some of the same ground as F06, but we provide all of the synthetic spectrum fitting parameters (unlike F06), and thanks to mild smoothing of the observed spectra and to the scale on which the fits are published, our fits can be examined more closely than those of F06.

3.1. Early Spectra

Figure 2 shows a close‐up view of the multiplet 42 region of Fe ii for the four early spectra. The vertical dashed lines correspond to λλ4924, 5018, and 5169, blueshifted by 13,000 km s⁻¹. Absorptions consistent with these transitions are clearly present in the day −32, −27, and −24 spectra. The vertical solid lines correspond to the same transitions blueshifted by 7000 km s⁻¹. The four early spectra have been compared with SYNOW spectra. Our interpretation of them is in good agreement with that of F06, although our results differ in detail because, for example, the version of SYNOW available to F06 did not have the Gaussian option.

Zoom In Zoom Out Reset image size

In Figure 3, the day −32 spectrum is compared with a synthetic spectrum that has v_phot = 8000 km s⁻¹ and includes lines of PV He i, O i, and Fe ii, as well as HV H i, Ca ii, and Fe ii. (The parameters of all synthetic spectra of this paper are mentioned in the text and/or listed in Table 1.) In the synthetic spectrum, PV He i is responsible for three features, and PV O i λ7773 is responsible for one. The main (but not exclusive) role of PV Fe ii is to produce the dip in the synthetic spectrum near 5040 Å. The influence of the HV ions is greater than that of the PV ions. The HV Ca ii infrared triplet produces the deep absorption near 8110 Å, HV Fe ii produces absorptions near 4930, 4790, 4710, and 4560 Å, and HV Hα produces the deep absorption near 6210 Å. The v_g values for HV Ca ii and H i are 17,000 km s⁻¹, while that of Fe ii is 15,000 km s⁻¹. We consider the HV Ca ii, Hα, and Fe ii identifications to be definite. Figure 4 is like Figure 3, except that HV Fe ii has been removed. We know of no way to restore the good fit of Figure 3 without using HV Fe ii.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In Figure 5, the day −27 spectrum is compared with a synthetic spectrum that is much like that of Figure 3 for day −32, but in Figure 5, v_phot = 7000 km s⁻¹ and PV Ca ii is introduced. (The difference between v_phot = 8000 km s⁻¹ for day −32 and v_phot = 7000 km s⁻¹ for day −27 is within our fitting uncertainties.) The HV and PV Ca ii parameters are chosen to produce a reasonable fit to the P Cygni profile extending from about 8000 to 8800 Å. The main changes between the synthetic spectra for day −32 and day −27 are that τ_p(He i) has increased by a factor of 1.5, τ_g(HV Ca ii) has decreased by a factor of 6.25, and PV Ca ii has been introduced with τ_p = 4. T05 used a different synthetic spectrum code, the Mazzali‐Lucy Monte Carlo code (Mazzali 2000), to fit a day −26 spectrum. They attributed the 6210 Å absorption to a blend of HV Hα and PV Si ii, and they did not explicitly⁴ introduce HV Ca ii and Fe ii features. (Their day −26 spectrum did not extend far enough to the red to show the HV Ca ii absorption at 8000 Å.)

Zoom In Zoom Out Reset image size

Between days −27 and −24, there are no major changes in our synthetic spectrum parameters (see Table 1). In Figure 6, the day −20 spectrum is compared with a synthetic spectrum that has v_phot = 8000 km s⁻¹. The major changes compared to day −27 are that for day −20, HV Fe ii is not used at all, τ_p(PV Ca ii) has increased by a factor of 12.5, and τ_p(PV Fe ii) has increased by a factor of 10.

Zoom In Zoom Out Reset image size

Significant spectral evolution occurred between days −32 and −20, especially between days −24 and −20. The day −32 spectrum is strongly influenced by HV features, but PV features also are present. By day −20, the spectrum is mainly PV, the only HV feature being Hα. From day −32 to day −20, τ_p(He i) increased by a factor of 2, suggestive of increasing nonthermal excitation (T05), while the fitting parameters for Hα varied only mildly.

3.2. Maximum Light

The five spectra of Figure 1 from day −6 to day +5 are rather similar, so we consider only day +2 here. In Figure 7, the day +2 spectrum is compared with a synthetic spectrum that has v_phot = 7000 km s⁻¹ and includes lines of PV He i, O i, Ca ii, and Fe ii, as well as HV H i. A05 noted that in the day −6 spectrum, the Fe ii absorptions were more blueshifted than the He i absorptions. We find the same for the day +2 spectrum: in the synthetic spectrum of Figure 7, Fe ii lines are mildly detached⁵ at 8000 km s⁻¹, while the He i lines are not.

Zoom In Zoom Out Reset image size

T05 found that in a day −5 spectrum, Si ii was satisfactory for the absorption that we attribute to Hα. Figure 8, which is like Figure 7, except that HV H i has been replaced by PV Si ii (τ_p = 1, n = 8), shows that for us, the Si ii absorption has its usual problem in SN Ib/c spectra: it is too blue to account for the observed absorption on its own. The synthetic spectrum of T05 has v_phot = 4600 km s⁻¹, compared to our value of 7000 km s⁻¹, and from their Figure 4, it appears that many of the synthetic absorptions are insufficiently blueshifted. This may account for their finding that Si ii is satisfactory.

Zoom In Zoom Out Reset image size

3.3. Postmaximum

The latest three spectra of Figure 1 are similar, so we consider only day +21. In Figure 9, the day +21 spectrum is compared with a synthetic spectrum that has v_phot = 5000 km s⁻¹ and includes the same lines as in Figure 7, except that HV H i is not used—the synthetic spectrum is entirely PV. The observed absorptions near 5460 and 5590 Å could be fitted reasonably well with PV Sc ii, but then Sc ii λ4247 would produce a deep unwanted absorption near 4100 Å. Part of the observed depression extending from 6290 to 6480 Å could be fit by introducing PV Hα, but the identification would not be convincing.

Zoom In Zoom Out Reset image size

F06 noted that the day −20 spectrum of SN 2005bf and a day +5 spectrum of the Type Ib SN 1999ex (Hamuy et al. 2002) were similar, in spite of the very different epochs with respect to maximum light. In Figure 10, the same two spectra are directly compared in a plot of linear flux, and with both spectra flattened (F06 presented a logarithm‐flux plot of unflattened spectra). Figure 10 shows that the two spectra are almost identical. They are so similar that it is not plausible that two different interpretations apply. Only minor tweaking of the synthetic spectrum for day −20 of SN 2005bf (Fig. 6), such as increasing the optical depth of HV Ca ii, would produce an equally good synthetic spectrum for day +5 of SN 1999ex. As discussed above, the day −20 spectrum of SN 2005bf is dominated by PV features but also includes HV Hα. The spectrum of SN 1999ex was interpreted in a similar way by B06, but the HV Hα identification was perhaps not entirely convincing. Figure 10 clinches the HV Hα identification in SN 1999ex.

Zoom In Zoom Out Reset image size

To account for the fact that based on very early spectra, SN 2005bf was initially classified as a Type Ic (Morell et al. 2005; Modjaz et al. 2005), F06 also noted that to some extent, the day −32 spectrum of SN 2005bf resembled a day −5 spectrum of the Type Ic SN 1994I (Filippenko et al. 1995). This leads us to ask whether the day −32 spectrum of SN 2005bf has something to teach us about how to interpret the near‐maximum spectra of SN 1994I. Figure 11 is like Figure 10, but for the day −32 spectrum of SN 2005bf and a day −4 spectrum of SN 1994I. The resemblance in Figure 11 is not as striking as in Figure 10, but it is sufficient to suggest that previous interpretations of the SN 1994I spectrum should be reconsidered. Millard et al. (1999), using the SYNOW code to interpret the day −4 spectrum of SN 1994I, did not consider the possibility that the spectrum was a composite HV and PV spectrum; therefore, in order to fit the Ca ii IR triplet feature, they had to use a high value of v_phot = 17,500 km s⁻¹. This led to difficulties in accounting for some of the other features. Millard et al. attributed the feature produced by He i λ5876 in SN 2005bf to Na i (with an imposed maximum velocity) in SN 1994I, and the feature produced by HV Hα in SN 2005bf as a blend of Si ii (with an imposed maximum velocity) and (detached) C ii in SN 1994I.

Zoom In Zoom Out Reset image size

Here we briefly consider a reinterpretation of the day −4 spectrum of SN 1994I. Using the same ions as for the day −32 spectrum of SN 2005bf, we varied the SYNOW input parameters and obtained the fit shown in Figure 12. The synthetic spectrum has v_phot = 12,000 km s⁻¹ (compared to 8000 km s⁻¹ for SN 2005bf), and the values of v_g for HV H i, HV Ca ii, and HV Fe ii are 23,000, 19,000, and 18,000 km s⁻¹, respectively (compared to 17,000, 17,000, and 15,000 km s⁻¹ for SN 2005bf). The fit is encouraging (although to fit the Ca ii infrared triplet, we have had to make Ca ii H and K too strong) and seems at least as plausible as that of Millard et al. (1999), as well as that of Sauer et al. (2006), who also attributed features in SN 1994I to Na i and to a blend of Si ii and C ii (and Ne i). A more thorough reconsideration of the early spectra of SNe Ic is deferred to a separate paper.

Zoom In Zoom Out Reset image size

B06 discussed the possible presence of HV Hα and PV He i in spectra of SN 1994I, as well as the possible presence of PV Hα as would be required by the tentative identification (Filippenko 1988, 1992) of centrally peaked Hα emission in SN 1994I and other SNe Ic. (Centrally peaked emission cannot be produced by highly detached hydrogen.) Figure 13 is like Figure 12, but with PV Hα included (with τ_p = 0.6, n = 8). PV Hα certainly does no harm; in fact, the fit is somewhat improved. It may be that PV Hα is actually in net emission in SN 1994I, as it is in SNe ii, rather than a resonance‐scattering profile as given by SYNOW; net emission in the synthetic spectrum would improve the fit near 6563 Å. The presence of PV Hα in SNe Ic remains a possibility.

Zoom In Zoom Out Reset image size

Our interpretation of the very early spectra, including the identification of HV Hα, is in good agreement with that of F06. In addition, we favor the Hα identification as late as day +2, and perhaps even day +5. Wang & Baade (2005) reported that an unpublished spectrum obtained on day −9 (presumably a very high signal‐to‐noise ratio spectrum, because spectropolarimetry was obtained) appeared to contain Hβ and Hγ absorptions. This increases our confidence in the Hα identification⁶ at day +2.

F06 discussed models for SN 2005bf. Although the numerical results they presented were based on spherical symmetry, they envisaged a grossly asymmetric explosion having many features in common with the supernovae that are associated with gamma‐ray bursts, for which the most popular model is the collapsar model (Woosley & Bloom 2006 and references therein). In SN 2005bf, according to F06, a collapsar launched relativistic jets that drove a small fraction of the ejected matter into high‐velocity (v≳14,000 km s⁻¹) bipolar flows that contained about 0.1 M_⊙ per pole of ⁵⁶Ni, were optically thick at early times, and produced the first light‐curve peak. When the bipolar ejecta became optically thin, the underlying low‐velocity ejecta, primarily equatorial and containing most of the mass and ⁵⁶Ni, powered the main light‐curve peak. However, it should be noted that the presence of PV features in the early spectra shows that the bipolar flows were not optically thick at the time of the first peak. Another issue is the distribution of hydrogen with respect to velocity. The progenitor star contained a surface layer of hydrogen. If the bipolar flows are of sufficiently wide opening angle, it may be that a pole‐on observer sees only HV hydrogen in absorption, not PV hydrogen. But an equator‐on observer would see PV hydrogen in absorption. Equator‐on is statistically more likely than pole‐on, yet it is not clear that any SN Ib/c has been seen to have PV, but not HV, hydrogen. Thus, the F06 interpretation of SN 2005bf requires that SN 2005bf–like events are quite uncommon; otherwise, SNe Ib/c with PV but not HV hydrogen in absorption should be found.

Like F06, T05 argued that the double‐peaked light curve required a double‐peaked radial distribution of ⁵⁶Ni, but T05 invoked jets that were not sufficiently energetic to reach the bottom of the helium layer, at v≃6000 km s⁻¹; these jets provided a small amount of ⁵⁶Ni at intermediate velocities (3900≲v≲5400 km s⁻¹) to power the first peak. Based on several perceived similarities, T05 suggested that SN 2005bf may have been a Cas A–like event. The T05 model is not necessarily highly asymmetric,⁷ so all of the hydrogen could be ejected at high velocity, and the issue of seeing SNe Ib/c with PV but not HV hydrogen does not necessarily arise.

In view of the almost perfect resemblance of the day +5 spectrum of SN 1999ex and the day −20 spectrum of SN 2005bf (Fig. 8), we regard the identification of Hα in SN 1999ex as definite. This supports the proposition that most SNe Ib eject some hydrogen (Deng et al. 2000; Branch et al. 2002; Elmhamdi et al. 2006). To our knowledge, there is no published clear explanation why hydrogen should generally not be entirely removed from the progenitors of supernovae that develop conspicuous He i lines.

The resemblance of the day +5 spectrum of SN 1999ex and the day −20 spectrum of SN 2005bf means that at these epochs, the conditions near the photosphere—composition, density structure, and temperature—were very similar. It is interesting to note that (1) for SN 2005bf, day −20 corresponds to about day +5 with respect to the initial maximum, and (2) these epochs correspond to similar times after explosion—20 days for SN 2005bf and 24 days for SN 1999ex (assuming a rise time for SN 1999ex of 19 days [Richardson et al. 2006]). But although the conditions at the photosphere were similar at these epochs, what was to follow—to be determined by what was still beneath the photosphere—would be very different.

Similarly, the times with respect to explosion for the day −4 spectrum of SN 1994I and the day −32 spectrum of SN 2005bf (Fig. 10) are not very different—about 8 days for SN 2005bf and 10 days for SN 1994I (assuming a rise time for SN 1994I of 14 days [Richardson et al. 2006]). The similarities between these two spectra raise our suspicion that SN 1994I ejected hydrogen. If so, then most or all other ordinary SNe Ic also do (see B06), and they are not explosions of bare carbon‐oxygen cores, as they usually are modeled. If SN 1994I did not eject hydrogen, then the spectroscopic coincidences between SNe Ib that do eject hydrogen and SNe Ic that do not eject hydrogen are even more striking than they appeared to B06.