1 Introduction

The transient nature of extragalactic type Ia supernovae (SN Ia) prevent studies from conclusively singling out unobserved progenitor configurations (Roelofs et al. 2008; Li et al. 2011a; Kilic et al. 2013). It remains fairly certain that the progenitor system of SN Ia comprises at least one compact C+O white dwarf (Chandrasekhar 1957; Nugent et al. 2011; Bloom et al. 2012). However, how the state of this primary star reaches a critical point of disruption continues to elude astronomers. This is particularly so given that less than ∼15 % of locally observed white dwarfs have a mass a few 0.1 M greater than a solar mass; very few systems near the formal Chandrasekhar-mass limit,Footnote 1 M Ch ≈1.38 M (Vennes 1999; Liebert et al. 2005; Napiwotzki et al. 2005, 2007; Parthasarathy et al. 2007).

Thus far observational constraints of SN Ia have been inconclusive in distinguishing between the following three separate theoretical considerations about possible progenitor scenarios. Along side perturbations in the critical mass limit or masses of the progenitors, e.g., from rotational support (Mueller and Eriguchi 1985; Yoon and Langer 2005; Chen and Li 2009; Hachisu et al. 2012; Tornambé and Piersanti 2013) or variances of white dwarf (WD) populations (van Kerkwijk et al. 2010; Dan et al. 2013), the primary WD may reach the critical point by accretion of material from a low-mass, radially-confined secondary star (Whelan and Iben 1973; Nomoto and Sugimoto 1977; Hayden et al. 2010a; Bianco et al. 2011; Bloom et al. 2012; Hachisu et al. 2012; Wheeler 2012; Mazzali et al. 2013; Chen et al. 2013), and/or through one of several white dwarf merger scenarios with a close binary companion (Webbink 1984; Iben and Tutukov 1984; Paczynski 1985; Thompson 2011; Wang et al. 2013b; Pakmor et al. 2013). In addition, the presence (or absence) of circumstellar material may not solely rule out particular progenitor systems as now both single- and double-degenerate systems are consistent with having polluted environments prior to the explosion (Shen et al. 2013; Phillips et al. 2013).

Meanwhile, and within the context of a well-observed spectroscopically normal SN 2011fe, recent detailed models and spectrum synthesis along with SN Ia rates studies, a strong case for merging binaries as the progenitors of normal SN Ia has surfaced (c.f., van Kerkwijk et al. 2010; Li et al. 2011b; Blondin et al. 2012; Chomiuk 2013; Dan et al. 2013; Moll et al. 2013; Maoz et al. 2013; Johansson et al. 2014). However, because no progenitor system has ever been connected to any SN Ia, most observational constraints and trends are difficult to robustly impose on a standard model picture for even a single progenitor channel; the SN Ia problem is yet to be confined for each SN Ia subtype.

As for restricting SN Ia subtypes to candidate progenitor systems: (i) observed “jumps” between mean properties of SN Ia subtypes signify potential differences of progenitors and/or explosion mechanisms, (ii) the dispersions of individual subtypes are thought to arise from various abundance, density, metallicity, and/or temperature enhancements of the original progenitor system’s post-explosion ejecta tomography, and (iii) “transitional-type” SN Ia complicate the already similar overlap of observed SN Ia properties (Nugent et al. 1995; Lentz et al. 2000; Benetti et al. 2005; Branch et al. 2009; Höflich et al. 2010; Wang et al. 2012, 2013c; Dessart et al. 2013a). Moreover, our physical understanding of all observed SN Ia subclasses remains based entirely on interpretations of idealized explosion models that are so far constrained and evaluated by “goodness of fit” comparisons to incomplete observations, particularly for SN Ia spectra at all epochs.

By default, spectra have been a limiting factor of supernova studies due to associated observational consequences, e.g., impromptu transient targets, variable intrinsic peak luminosities, a sparsity of complete datasets in wavelength and time, insufficient signal-to-noise ratios, and the ever-present obstacle of spectroscopic line blending (Payne-Gaposchkin and Whipple 1940). Subsequently, two frequently relied upon empirical quantifiers of SN Ia spectroscopic diversity have been the rate at which rest-frame 6100 Å absorption minima shift redward vis-à-vis projected Doppler velocities of the absorbing Si-rich material (Benetti et al. 2005; Wang et al. 2009a) and absorption strength measurements (a.k.a. pseudo equivalent widths; pEWs) of several lines of interest (see Branch et al. 2006; Hachinger et al. 2006; Silverman et al. 2012b; Blondin et al. 2012). Together these classification schemes more-or-less describe the same events by two interconnected parameter spaces (i.e. flux and expansion velocities, Branch et al. 2009; Foley and Kasen 2011; Blondin et al. 2012) that are dependent on a multi-dimensional array of physical properties. Naturally, the necessary next step for supernova studies alike is the development of prescriptions for the physical diagnosis of spectroscopic behaviors (see Sect. 2.2 and Kerzendorf and Sim 2014).

For those supernova events that have revealed the observed patterns of SN Ia properties, the majority are termed “Branch-normal” (Branch et al. 1993; Li et al. 2011b), while others further away from the norm are historically said to be “peculiar” (e.g., SN 1991T, 1991bg; see Filippenko 1997 and references therein). Although, many non-standard events have since obscured the boundaries between both normal and peculiar varieties of SN Ia, such as SN 1999aa (Garavini et al. 2004), 2000cx (Chornock et al. 2000; Li et al. 2001; Rudy et al. 2002), 2001ay (Krisciunas et al. 2011), 2002cx (Li et al. 2003), 2003fg (Howell et al. 2006; Jeffery et al. 2006), 2003hv (Leloudas et al. 2009; Mazzali et al. 2011), 2004dt (Wang et al. 2006; Altavilla et al. 2007), 2004eo (Pastorello et al. 2007a), 2005gj (Prieto et al. 2007), 2006bt (Foley et al. 2010b), 2007ax (Kasliwal et al. 2008), 2008ha (Foley et al. 2009, 2010a), 2009ig (Foley et al. 2012c; Marion et al. 2013), PTF10ops (Maguire et al. 2011), PTF11kx (Dilday et al. 2012; Silverman et al. 2013b), and 2012fr (Maund et al. 2013; Childress et al. 2013c).

The fact that certain subsets of normal SN Ia constitute a near homogenous group of intrinsically bright events has led to their use as standardizable distance indicators (Kowal 1968; Elias et al. 1985a; Branch and Tammann 1992; Riess et al. 1999; Perlmutter et al. 1999; Schmidt 2004; Mandel et al. 2011; Maeda et al. 2011; Sullivan et al. 2011a; Hicken et al. 2012). However, this same attribute of homogeneity remains the greatest challenge in the individual study of SN Ia given that the time-evolving spectrum of a supernova is unique unto itself from the earliest to the latest epochs.

Because SN Ia are invaluable tools for both cosmology and understanding progenitor populations, a multitude of large scale surveys, searches, and observing campaignsFootnote 2 are continually being carried out with regularly improved precision. Subsequently, this build-up of competing resources has also resulted in an ever growing number of new and important discoveries, with less than complete information for each. In fact, with so many papers published each year on various aspects of SN Ia, it can be difficult to keep track of new results and important developments, including the validity of past and present theoretical explosion simulations and their related observational interpretations (see Maoz et al. 2013 for the latest).

Here we compile some of the discussions on spectroscopic properties of SN Ia from the past decade of published works. In Sect. 2 we overview the most common means for studying SN Ia: light curves (Sect 2.1), spectra (Sect. 2.2), and detailed explosion models (Sect. 2.3). In particular, we overview how far the well-observed SN 2011fe has progressed the degree of confidence associated with reading highly blended SN Ia spectra. Issues of SN Ia diversity are discussed in Sect. 3. Next, in Sect. 4 we recall several SN Ia that have made up the bulk of recent advances in uncovering the extent of their properties and peculiarities (see also the Appendix for a guide of some recent events). Finally, in Sect. 5 we summarize and conclude with some observational lessons of SN 2011fe.

2 Common subfields of utility

2.1 Light curves

The interaction between the radiation field and the ejecta can be interpreted to zeroth order with the bolometric light curve. For SN Ia, the rise and fall of the light curve is said to be “powered” by 56Ni produced in the explosion (Colgate and McKee 1969; Arnett 1982; Khokhlov et al. 1993; Mazzali et al. 1998; Pinto and Eastman 2000a; Stritzinger and Leibundgut 2005). Additional sources are expected to contribute to the overall luminosity behavior at various epochs.Footnote 3

For example, Nomoto et al. (2003) has suggested that the variation of the carbon mass fraction in the C+O WD (C/O), or the variation of the initial WD mass, causes the diversity of SN Ia brightnesses (see Höflich et al. 2010). Similarly, Meng et al. (2011) argue that C/O and progenitor metallicity, Z, are intimately related for a fixed WD mass, and particularly for high metallicities given that it results in lower 3α burning rates plus an increased reduction of carbon via 12C(α,γ)16O. For Z>Z (∼0.02), Meng et al. (2011) find that both C/O and Z have an approximately equal influence on 56Ni production since, for a given WD mass, high progenitor metallicities (a greater abundance of species heavier than oxygen) and low C/O abundances (low carbon-rich fuel assuming a single-degenerate scenario) result in a low 56Ni yield and subsequently dimmer SN Ia. For near solar metallicities or less, the carbon mass fraction plays a dominant role in 56Ni production (Timmes et al. 2003). This then suggests that the average C/O ratio in the final state of the progenitor is an important physical cause, in addition to metallicity, for the observed width-luminosity relationship (WLRFootnote 4) of normal SN Ia light curves (Umeda et al. 1999a; Timmes et al. 2003; Nomoto et al. 2003; Bravo et al. 2010; Meng et al. 2011).

At the same time, the observed characteristics of SN Ia light curves and spectra can be fairly matched by adopting radial and/or axial shifts in the distribution of 56Ni, possibly due to a delayed- and/or pulsational-detonation-like explosion mechanism (see Khokhlov 1991b; Hoflich et al. 1995; Baron et al. 2008, 2012; Bravo et al. 2009; Maeda et al. 2010b; Dessart et al. 2013a) or a merger scenario (e.g., Dan et al. 2013; Moll et al. 2013). Central ignition densities are also expected to play a secondary role in the form of the WLR since they are dependent upon the accretion rate of H and/or He-rich material and cooling time (Röpke et al. 2005; Höflich et al. 2010; Meng et al. 2010; Krueger et al. 2010; Sim et al. 2013), in addition to the spin-down timescales for differentially rotating WDs (Hachisu et al. 2012; Tornambé and Piersanti 2013). Generally, discerning which of these factors dominate the spectrophotometric variation from one SN Ia to another remains a challenging task (Wang et al. 2012). As a result, astronomers are still mapping a broad range of SN Ia characteristics and trends (Sect. 3).

Meanwhile, cosmological parameters determined by SN Ia light curves depend on an accurate comparison of nearby and distant events.Footnote 5 For distant and therefore redshifted SN Ia, a “K-correction” converts an observed magnitude to that which would be observed in the rest frame in another bandpass filter, allowing for the comparison of SN Ia brightnesses at various redshifts (Hogg et al. 2002). Consequently, K-corrections require the spectral energy distribution (SED) of the SN Ia and depend on SN Ia broad-band colors and the diversity of spectroscopic features (Nugent et al. 2002). While some light curve fitters take a K-correction-less approach (e.g., Guy et al. 2005, 2007; Conley et al. 2008), an SED is still required. A spectral template time series dataset is usually used since there exists remarkable homogeneity in the observed optical spectra of “normal” SN Ia (e.g., Hsiao et al. 2007).

Unfortunately there do remain poorly understood differences regarding spectroscopic feature strengths and inferred expansion velocities for these and other types of thermonuclear supernovae (see Sects. 2.2 and 3). At best, the spectroscopic diversity of SN Ia has been determined to be multidimensional (Hatano et al. 2000; Benetti et al. 2005; Branch et al. 2009; Wang et al. 2009a). Verily, SN Ia diversity studies require numerous large spectroscopic datasets in order to subvert many complex challenges faced when interpreting the data and extracting both projected Doppler velocities and “feature strength” measurements. However, studies that seek to primarily utilize SN Ia broad band luminosities need only collect a handful of sporadically sampled spectra in order to type the supernova event as a bona fide SN Ia. We note that interests in precision cosmology conflict at this point with the study of SN Ia. This is primarily because obtaining UBVRI photometry for hundreds of events is cheaper than collecting complete spectroscopy for a lesser number of SN Ia at various redshifts.

Nevertheless, the brightness decline rate in the B-band during the first 15 rest-frame days post-maximum light, Δm 15(B), has proven useful for all SN Ia surveys. Phillips (1993) noted that Δm 15(B) is well correlated with the intrinsic luminosity, a.k.a. the width-luminosity relationship. Previously, Khokhlov et al. (1993) did predict the existence of a WLR given that the light curve shape is sensitive to the time-dependent state of the ejected material.

Kasen and Woosley (2007) recently utilized multi-dimensional time-dependent Monte Carlo radiative transfer calculations of Chandrasekhar-mass SN Ia models to access the physical relationship between the luminosity and light curve decline rate. They found that the WLR is largely a consequence of the radiative transfer inherent to SN Ia atmospheres, whereby the ionization evolution of iron redirects flux red ward and is hastened for dimmer and/or cooler SN Ia. Woosley et al. (2007) later explored the diversity of SN Ia light curves using a grid of 130 one-dimensional models. They concluded that a WLR is satisfied when SN Ia burn ∼1.1 M of material, with iron-group elements extending out to ∼8000 km s−1.

Broadly speaking, the shape of the WLR is fundamentally influenced by the ionization evolution of iron group elements (Kasen and Woosley 2007). However, since broad band luminosities are the sum of a supernova SED per wavelength interval, details of SN Ia diversity risk being “blurred out” for large samples of SN Ia. Therefore, decoding the spectra of all SN Ia subtypes, in addition to indirectly constraining detailed explosion models by the WLR, is of vital importance since variable signatures of iron-peak elements (IPEs) blend themselves within an SED typically populated by relatively strong features of overlapping signatures of intermediate-mass elements (IMEs).

2.2 Spectra

Supernova spectra detail information about the explosion and its local environment. To isolate and extract physical details (and determine their order of influence), several groups have invested greatly in advancing the computation of synthetic spectra for SN Ia, particularly during the early phases of homologous expansion (e.g., Mazzali and Lucy 1993; Hauschildt and Baron 1999; Kasen et al. 2002, 2006; Thomas et al. 2002, 2011a; Höflich et al. 2002; Branch 2004; Sauer et al. 2006; Jeffery and Mazzali 2007; Sim et al. 2010a; Hillier and Dessart 2012; Hoffmann et al. 2013; Pauldrach et al. 2013; Kerzendorf and Sim 2014). Although, even the basic facets of the supernova radiation environment serve as obstacles for timely computations of physically accurate, statistically representative, and robustly certain synthetic spectra (e.g., consequences of expansion).

It is the time-dependent interaction of the radiation field with the expanding material that complicates drawing conclusions about the explosion physics from the observations.Footnote 6 In a sense, there are two stages during which direct (and accessible) information about the progenitor system is driven away from being easily discernible within the post-explosion spectra: explosive nucleosynthesis and radiation transport.Footnote 7 That is to say, the ability to reproduce both the observed light curve and spectra, as well as the range of observed characteristics among SN Ia, is essential towards validating and/or restricting any explosion model for a given subtype.

Moreover, this assumes the sources of observed spectroscopic signatures in all varieties of SN Ia are known a priori, which is not necessarily the case given the immense volume of actively contributing atomic line transitions and continuum processes (Baron et al. 1995, 1996; Kasen et al. 2008; Bongard et al. 2008; Sauer et al. 2008). In fact, several features throughout the spectra have been either tentatively associated with a particular blend of atomic lines or identified with a multiple of conflicting suggestions (e.g., forbidden versus permitted lines at late or “nebular” transitional phases, see Bowers et al. 1997; Branch et al. 2005; Friesen et al. 2012; Dessart et al. 2013b). Meanwhile others are simply misidentified or unresolved due to the inherent high degeneracy of solutions and warrant improvements to the models for further study (e.g., Na I versus [Co III]; Dessart et al. 2013b).

For example, the debate over whether or not hydrogen and/or helium are detected in some early Ibc spectra has been difficult to navigate on account of the wavelength separation of observed weak features and the number of plausible interpretations (Deng et al. 2000; Branch et al. 2002; Anupama et al. 2005a; Elmhamdi et al. 2006; Parrent et al. 2007; Ketchum et al. 2008; Soderberg et al. 2008; James and Baron 2010; Benetti et al. 2011; Chornock et al. 2011; Dessart et al. 2012; Milisavljevic et al. 2013a, 2013b; Takaki et al. 2013). Historically, the term “conspicuous” has defined whether or not a supernova belongs to a particular spectroscopic class. By way of illustration, photographic spectrograms of type II events reveal conspicuous emission bands of hydrogen while type I events do not (Minkowski 1941). With the advent of CCD cameras in modern astronomy, it has been determined that 6300 Å absorption features (however weak) in the early spectra of some type Ibc supernovae are often no less conspicuous than 6100 Å Si II λ6355 absorption features in SN Ia spectra, where some 6300 Å features produced by SN Ibc may be due to Si II and/or higher velocity Hα (Filippenko 1988, 1992; Filippenko et al. 1990). That is, while SN Ibc are of the type I class, they do not necessarily lack hydrogen and/or helium within their outer-most layers of ejecta, hence the conservative definition of type I supernovae as “hydrogen/helium-poor” events.

This conundrum of which ion signatures construct each observed spectral feature rests proportionately on the signal-to-noise ratio (S/N) of the data. However, resolving this spectroscopic dilemma is primarily dependent on the wavelength and temporal coverage of the observations and traces back to the pioneering work of McLaughlin (1963) who studied spectra of the type Ib supernova, SN 1954A, in NGC 4214 (Wellmann 1955; Branch 1972; Blaylock et al. 2000; Casebeer et al. 2000). Contrary to previous interpretations that supernova spectra were the result of broad, overlapping emission features (Gaposchkin 1936; Humason 1936; Baade 1936; Walter and Strohmeier 1937; Minkowski 1939; Payne-Gaposchkin and Whipple 1940; Zwicky 1942; Baade et al. 1956), it was D. B. McLaughlin who first began to repeatedly entertain the idea that “absorption-like” features were presentFootnote 8 in regions that “lacked emission” (McLaughlin 1959, 1960, 1963).

The inherent difficulties in reading supernova spectra and the history of uncertain line identifications for both conspicuous and concealed absorption signatures are almost as old as the supernova field itself (Payne-Gaposchkin and Whipple 1940; Dessart et al. 2013b). Still, spectroscopic intuitions can only evolve as far as the data allow. Therefore it is both appropriate and informative to recall the progression of early discussions on the spectra of supernovae, during which spectroscopic designations of type I and type II were first introduced:

There appears to be a general opinion that the evidence concerning the spectrum of the most luminous nova of modern times was so contradictory that conclusions as to its spectra nature are impossible. This view is expressed, for example, by Miss Cannon: “With the testimony apparently so conflicting, it is difficult to form any conception of the class of this spectrum” (Gaposchkin 1936).

It also seems ill advised to conclude anything regarding the distribution of temperature in super-novae from the character of their visible spectra as long as a satisfactory explanation of some of the most important features of these spectra is completely lacking (Zwicky 1936).

The spectrum is not easy to interpret, as true boundaries of the wide emission lines are difficult to determine (Humason 1936).

Those [emission] bands with distinct maxima and a fairly sharp redward or violetward edge, excepting edges due to a drop in plate spectral sensitivity, may give an indication of expansion velocity (Popper 1937).

Instead of the typical pattern of broad, diffuse emissions dominated by a band about 4600 Å, it appeared like a continuum with a few deep and several shallow absorption-like minima. Two of the strongest “absorption lines,” when provisionally interpreted as λλ4026, 4472 He I, give velocities near −5000 km s−1 […] The author is grateful to N.U. Mayall and R. Minkowski for the use of spectrograms, and for helpful discussions. However, this does not imply agreement with the author’s interpretations (McLaughlin 1959).

It is hardly necessary to emphasize in detail the difficulties of establishing the correct interpretation of a spectrum which may reflect unusual chemical composition, whose features may represent emission, absorption, or both mixed, and whose details are too ill-defined to admit precise measures of wavelengths (Minkowski 1963).

Given that our general understanding of blended spectral lines remains in a continual state of improvement, the frequently recurrent part of “the supernova problem” is pairing observed features with select elements of the periodic table (Hummer 1976; Axelrod 1980; Jeffery and Branch 1990; Hatano et al. 1999b; Branch et al. 2000). In fact, it was not until nearly a half-century after Minkowski (1963), with the discovery and prompt spectroscopic follow-up of SN 2011fe (Fig. 1 and Sect. 4.1) that the loose self-similarity of SN Ia time series spectra from the perceived beginning of the event to near maximum light was roundly confirmed (Nugent et al. 2011, see also Garavini et al. 2005; Foley et al. 2012c; Silverman et al. 2012d; Childress et al. 2013c; Zheng et al. 2013).

Fig. 1
figure 1

Plotted is the SNFactory’s early epoch dataset of SN 2011fe presented by Pereira et al. (2013). We have normalized and over-plotted each spectrum at the 6100 Å P Cygni profile in order to show the relative locations of all ill-defined features as they evolve with the expansion of the ejecta. The quoted rise-time to maximum light (dashed black) is from Mazzali et al. (2013)

While SN 2011fe may not have revealed a direct confirmation on its progenitor system (Li et al. 2011a), daily spectroscopic records at optical wavelengths were finally achieved, establishing the most efficient approach for observing ill-defined features over time (Pereira et al. 2013). This is important given that UV to NIR line identifications of all observed complexes are highly time-dependent, are sensitive to most physically relevant effects, continuously vary between subtypes, and rely on minimal constraint for all observed events.Footnote 9

Even so, this rarely attainable observing strategy does not necessarily illuminate nor eliminate all degeneracies in spectral feature interpretations. However the advantage of complimentary high frequency follow-up observations is that the spectrum solution associated with any proposed explosion scenario can at least be consistently tested and constrained by the observed rapid changes over time (“abundance tomography” goals, e.g., Hauschildt and Baron 1999; Stehle et al. 2005; Sauer et al. 2006; Kasen et al. 2006; Hillier and Dessart 2012; Pauldrach et al. 2013). It then follows that hundreds of well-observed spectrophotometric datasets serve to carve out the characteristic information, f(λ;t), for each SN Ia between subtypes, in addition to establishing the perceived boundaries of the SN Ia diversity problem (see Fig. 11 of Blondin et al. 2012 for this concept at maximum light).

For supernovae in general, Fig. 1 also serves as a reminder that all relative strengths evolve continuously over time, where entire features are always red-shifting across wavelength (line velocity space) during the rise and fall in brightness. A corollary of this situation is that prescriptions for taking measurements of spectroscopic behaviors (whereby interpretations rely on a subjective “goodness of fit”) and robustly associating with any number of physical causes do not exist. Instead there are two primary means for interpreting SN Ia spectra and taking measurements of features for the purposes of extracting physical properties.

Indirect analysis assumes a detailed explosion model and is primarily tasked with assessing the accuracy and flaws of the model. Direct analysis seeks to manually measure via spectrum synthesis where one can either assume an initial post-explosion ejecta composition or give up abundance information altogether to assess the associated uncertainties and consequences of supernova line blending via purposeful high parameterizations. For the latter of these direct inference methods, the conclusions about spectroscopic interpretations—which are supported by remnants of inconsistencies throughout the literature—are summarized as follows.

For the most part, particularly at early epochs and as far as anyone can tell with current limiting datasets, the features in SN Ia spectra are due to IMEs and IPEs formed by resonance scattering of continuum and decay-chain photons, and have P Cygni-type profiles overall ((Hsiao et al. 2013; Pereira et al. 2013); see Fig. 2). Emission components peak at or near the rest wavelength and absorption components are blue-shifted according to the opacity profile of matter at and above the photospheric line forming region. The combination of these effects can often lead to “trumped” emission features Parrent et al. 2012, giving SN Ia spectra their familiar shapes.

Fig. 2
figure 2

Top: A schematic representation of how an assumed spherically sharp and embedded photosphere amounts to a pure line-resonance P Cygni profile under the conditions of Sobolev line transfer within a geometry of Absorbing, Emitting, and Occulted regions of material (Jeffery and Branch 1990; Branch et al. 2005). The approximate photospheric velocity, v phot , is inversely proportional to the blue ward shift of an unblended absorption minimum. Bottom: Application of the above P Cygni diagram to SN Ia spectra in terms of which species dominate and what other species are known to influence the temporal behavior (Bongard et al. 2008), each of which are constrainable from complete spectroscopic coverage. For each series of spectra, the black line in bold represents maximum light

Essentially all relevant atomic species (isotope plus ionization state) are present somewhere within the ejecta, each with its own 3-dimensional abundance profile. At optical wavelengths, conditions and abundance tomographies of the ejecta maintain the dominance of select singly−triply ionized subsets of C+O, IMEs, and IPEs Pskovskii 1969; Mustel 1971; Branch and Patchett 1973; Kirshner et al. 1973a. From shortly after the onset of the explosion to around the time of maximum light, the optical—NIR spectrum of a normal SN Ia consists of a continuum level with superimposed features that are primarily consistent with strong permitted lines of ions such as O I, Mg II, Si II, Si III, S II, Ca II, Fe II, Fe III, and trace signatures of C I and C II (Jeffery and Branch 1990). After the pre-maximum light phase, blends of Fe II (in addition to other IPEs) begin to dominate or influence the temporal behavior of many optical–NIR features over timescales from weeks to months (see (Hatano et al. 1999b) and references therein).

With the above mentioned approximated view of line formation in mind (Fig. 2), the real truth is that the time-dependent state of the ejecta and radiation field at all locations dictates how the material presence within the line forming regions will be imparted onto the spectral continuum, i.e. the radiation field and the matter are said to be “coupled.” With the additional condition of near-relativistic expansion velocities (∼0.1c), line identifications themselves can also be thought of as coupled to the abundance tomography of ejected material, which includes the projected Doppler velocities spanned by the recipe of absorbing material. Subsequently, while spectra can be used for constraining limits of some model parameters, it comes with a cost of certainty on account of natural uncertainties imparted by the large expansion velocities and associated expansion opacities.

As an exercise in this point, in Fig. 3 we have constructed an early epoch set of toy model line profiles that are representative of normal SN Ia line identification procedures (e.g., (Branch et al. 2006; Thomas et al. 2007; Bongard et al. 2008; Nugent et al. 2011; Parrent et al. 2012; Hsiao et al. 2013; Mazzali et al. 2013; Dessart et al. 2013a)) and over-plot them with an early optical–NIR spectrum (the observed outermost layers, sans UV) of SN 2011fe. We summarize the take away points of Fig. 3 as follows.

  • Even without considering weak contributions, at no place along the (UV–) optical–NIR spectrum is any observed feature removed from being due to less than 2 sources (more precisely, see also Branch et al. 2008). That is, under the basic assumptions of pure resonance line scattering and homologous expansion (Fig. 2), all features are complex blends of at least 2+ ions and are universally influenced by multiple regions of emitting and/or absorbing material (e.g., “high[-er] velocity” and “photospheric velocity” intervals of material, see also Branch et al. 2005; Parrent et al. 2011).

  • For supernovae, the components of the spectrum are most easily constrained via spectrum synthesis, and subsequently measurable (not the converse), when the bounds of wavelength coverage, λ a and λ b , are between ∼2000–3500 and 12000 Å, respectively. If λ b <7500–9500 Å, then the velocities and relative strengths of several physically relevant ions (e.g., C I, O I, Mg II, and Ca II) are said to be devoid of useful constraint and provide a null (or uncertain) measurement for every other overlapping spectral line signature (i.e. all features). That is, in order to viably “identify” and measure a single feature, the entire spectrum must be reproduced. While empirical measurements of certain absorption features are extremely useful for identifying trends in the observed behavior of SN Ia, these methods do not suffice to measure the truest underlying atomic recipe and its time-dependent behavior, much less the “strength” of contributing lines (e.g., multiple velocity components of Si II in SN 2012fr, Sect. 4.2.2). Specifically, empirical feature strength measurements at least require a proper modeling of the non-blackbody, IPE-dominated pseudo continuum level Bongard et al. 2008 or the use of standardized relative strength parameters (e.g., Marion et al. 2013).

  • Therefore, as in Fig. 2, employing stacked Doppler velocity scaled time series spectra provides useful and timely first-order comparative estimates for when (epoch) and where (projected Doppler velocity) contributing ions appear, disappear, and span as the photospheric region recedes inward over time.

Fig. 3
figure 3

SYN++ calculation comparisons to the early optical–NIR spectrum of SN 2011fe (Jeffery and Branch 1990; Branch et al. 2005). Calculations are based on an optical set of photospheric phase spectra (see (Bongard et al. 2008)) and are true-to-scale. Bands of color are intended to show overlap between lines under the simplified however informative assumption of permitted line scattering under homologous expansion. Some of the weaker lines have not been highlighted for clarity

We speak on this only to point out that even simple questions—particularly for homogeneous SN Ia—are awash in detection/non-detection ambiguities. However, it should be noted that a powerful exercise in testing uncertain line identifications and resolving complex blends can be done, in part, without the use of additional synthetic spectrum calculations. That is, by comparing a single observed spectrum to that of other well-observed SN Ia, where the analysis of the latter offers a greater context for interpretation than the single spectrum itself, one can deduce whether or not a “mystery” absorption feature is common to most SN Ia in general. On the other hand, if a matching absorption feature is not found, then one can infer the presence of either a newly identified, compositionally consistent ion or the unblended line of an already accounted for species (resulting from forbidden line emission, non-LTE effects, and/or when line strengths or expansion velocities differ between subtypes). Given also the intrinsic dispersion of expansion opacities between SN Ia, it is likely that an “unidentified” feature is that of a previously known ion at higher and/or lower velocities. It is this interplay between expansion opacities and blended absorption features that keep normal and some peculiar SN Ia within the description of a homogenous set of objects, however different they may appear.

In fact, when one compares the time series spectra of a broad sample of SN Ia subtypes, however blended, there is little room for degeneracy among plausible ion assignments (sans IPEs, e.g., Fe II versus Cr II during post-maximum phases). In other words, there exists a unique set of ions, common to most SN Ia atmospheres, that make up the resulting spectrum, where differences in subtype are associated with differences in temperature and/or the abundance tomography of the outermost layers (Bongard et al. 2008). The atomic species listed in Fig. 3 do not so much represent a complete account of the composition, or the “correct” answer, as they are consistent with the subsequent time evolution of the spectrum toward maximum light, and therefore serve to construct characteristic standards for direct comparative diversity assessments.

Said another way, it is the full time series dataset that enables the best initial spectrum solution hypothesis, which can be further tested and refined for the approximate measurement of SN Ia features Childress et al. 2013b. Therefore, this idea of a unique set of ions remains open since—with current limiting datasets—species with minimal constraint and competing line transfer processes can be ambiguously present,Footnote 10 even for data with an infinite S/N (i.e. sources with few strong lines, or lines predominately found blue ward of ∼6100 Å, e.g., C III, O III, Si IV, Fe I, Co II, Ni II). One can still circumvent these uncertainties of direct analysis by either using dense time series observations (e.g., (Branch et al. 2007a)) or by ruling out spurious inferred detections by including adjacent wavelength regions into the spectroscopic analysis (UV–optical–NIR; see Stritzinger et al. (2013)).

2.3 Models

A detailed account of SN Ia models is beyond the scope of our general review of SN Ia spectra (for the latest discussions, see Parrent et al. 2012). However, in order to understand the context by which observations are taken and synthetic comparisons made, here we only mention the surface layer of matters relating to observed spectra. For some additional recent modeling work, see Foley et al. 2012b; Hsiao et al. 2013; Mazzali et al. 2013, Wang and Han 2012; Nomoto et al. 2013; Hillebrandt et al. 2013; Calder et al. 2013; Maoz et al. 2013, Fryer and Diehl (2008), Bravo et al. (2009), Jordan et al. (2009), Kromer et al. (2010), Blondin et al. (2011), Hachisu et al. (2012), Jordan et al. (2012), Pakmor et al. (2013), Seitenzahl et al. (2013), Dan et al. (2013), and Kromer et al. (2013b).

Realistic models are not yet fully ready because of the complicated physical conditions in the binary stellar evolution that leads up to an expanding SN Ia atmosphere. For instance, the explosive conditions of the SN Ia problem take place over a large dynamic range of relevant length-scales (R WD ∼1 R and flame-thicknesses of ∼0.1 cm; Moll et al. (2013)), involve turbulent flames that are fundamentally multi-dimensional (Raskin et al. (2013), Timmes and Woosley 1992; Gamezo et al. 1999b; Khokhlov 1995, 2000; Reinecke et al. 2002a, 2002b; Gamezo et al. 2003), and consist of uncertainties in both the detonation velocity 2005 and certain nuclear reaction rates, especially 12C+12C (Seitenzahl et al. 2013, however see also (Domínguez and Khokhlov 2011)).

Most synthetic spectra are angle-averaged representations of higher-dimensional detailed models. Overall, the observed spectra of normal SN Ia have differed less amongst themselves than that of some detailed models compared to the data of normal SN Ia. This is not from a lack of efforts, but is simply telling of the inherent difficulty of the problem and limiting assumptions and interests of various calculations. Bravo et al. 2011 reviewed previous work done of N-dimensional SN Ia models and presented the first high-resolution 3D calculation of a SN Ia spectrum at maximum light. Their results are still in a state of infancy, however they represent the first step toward the ultimate goal of SN Ia modeling, i.e. to trace observed SN Ia properties and infer the details of the progenitor and its subsequent disruption by comparing 3D model spectra and light curves of 3D explosion simulations with the best observed temporal datasets.

Still, progress has been made in understanding general observed properties of SN Ia and their relation to predictions of simulated explosion models. For example, one-dimensional (1D) numerical models of SN Ia have been used in the past to test the possible explosion mechanisms such as subsonic flame or supersonic detonation models, as well as conjoined delayed-detonations (e.g., Bravo and Martínez-Pinedo 2012; Chen et al. 2013). The one-dimensional models disfavor the route of a pure thermonuclear detonation as the mechanism to explain most SN Ia events Kasen et al. (2008). Such a mechanism produces mostly 56Ni and almost none of the IMEs observed in the spectra of all SN Ia (e.g., Arnett 1968; Nomoto et al. 1984; Lentz et al. 2001a).

However, one-dimensional models have shown that a detonation can produce intermediate mass elements if it propagates through a Chandrasekhar-mass WD that has pre-expanded during an initial deflagration stage ((Hansen and Wheeler 1969; Arnett 1969; Axelrod 1980), Branch et al. 1982; Filippenko 1997; Gamezo et al. 1999b; Pastorello et al. 2007a; Khokhlov 1991a; Yamaoka et al. 1992; Khokhlov et al. 1993, 1997; Arnett and Livne 1994a). To their advantage, these deflagration-to-detonation transition (DDT) and pulsating delayed-detonation (PDD) models are able to reproduce the observed characteristics of SN Ia, however not without the use of an artificially-set transition density between stages of burning (1994b; Wheeler et al. 1995; Hoflich et al. 1995; Khokhlov 1991b, Hoflich et al. 1995; Lentz et al. 2001a; 2001b; Baron et al. 2008). Subsequently, a bulk of the efforts within the modeling community has been the pursuit of conditions or mechanisms which cause the burning front to naturally transition from a sub-sonic deflagration to a super-sonic detonation, e.g., gravitationally confined detonations Bravo et al. 2009, prompt detonations of merging WDs, a.k.a. “peri-mergers” Dessart et al. 2013a.

With the additional possibility that the effectively burned portion of the progenitor is enclosed or obscured by some body of circumstellar or envelope/disk of material (see (Jordan et al. 2009)), the intrinsically multi-dimensional nature of the explosion itself is also expected to manifest signatures of asymmetric plumes of burned material and pockets of unburned material within a spheroidal debris field of flexible asymmetries (see (Moll et al. 2013)). Add to this the degeneracy of SN Ia flux behaviors, i.e. colors are sensitive to dust/CSM extinction and intrinsic dispersions in the same direction Sternberg et al. 2011; Foley et al. 2012a; Förster et al. 2012; Scalzo et al. 2012; Raskin and Kasen 2013; Silverman et al. 2013d; Dan et al. 2013; Dessart et al. 2013a; Moll et al. 2013, whether large or small redshift-color dependencies Khokhlov 1995; Niemeyer and Hillebrandt 1995; Gamezo et al. 2004; Wang and Wheeler 2008; Patat et al. 2009; Kasen et al. 2009, and we find the true difficulty in constraining SN Ia models.

(Tripp and Branch 1999) recently presented and discussed the photometric and spectroscopic properties at maximum light of a sequence of 1D DDT explosion models, with ranges of synthesized 56Ni masses between 0.18 and 0.81 M. In addition to showing broad consistencies with the diverse array of observed SN Ia properties, the synthetic spectra of (Saha et al. 1999; Jha et al. 1999; Parodi et al. 2000; Wang et al. 2008a; Goobar 2008; Wang et al. 2009a; Foley and Kasen 2011; Mohlabeng and Ralston 2013) predict weaker absorption features of unburned oxygen (O I λ7774) at maximum light, in proportion to the amount of 56Ni produced. This is to be expected Blondin et al. (2013), however constraints on the remaining amount of unburned material, in addition to its temporal behavior, are more readily seen during the earliest epochs (within the outermost layers of ejecta) via C II λ6580 and O I λ7774 Blondin et al. (2013). Consequently, temporal spectrum calculations of detailed explosion models are needed for the purposes of understanding why the properties of SN Ia are most divergent well before maximum light (Hoflich et al. 1995).

Nucleosynthesis in two-dimensional (2D) delayed detonation models of SN Ia were explored by (Thomas et al. 2007; Parrent et al. 2011; Nugent et al. 2011). In particular, they focused on the distribution of species in an off-center DDT model and found the abundance tomography to be stratified, with an inner region of 56Ni surrounded by an off-center shell of electron-capture elements (e.g., Fe54, Ni58). Later, (Branch et al. 2006; Dessart et al. 2013a) investigated the late time emission profiles associated with this off-center inner-shell of material within several observed SN Ia and found a correlation between possible nebular-line Doppler shifts along the line-of-sight and the rate-of-decline of Si II velocities at earlier epochs. Their interpretation is to suggest that some SN Ia subtypes may represent two different hemispheres of the “same” SN Ia (LVG vs. HVG subtypes; see Sect. 3.2). Moreover, the findings of Maeda et al. (2010a) and Maeda et al. (2010b) remain largely consistent with the additional early and late time observations of the well-observed SN 2011fe Maeda et al. (2010b) and those of larger SN Ia samples Maund et al. (2010b). However, even the results of (Smith et al. 2011; McClelland et al. 2013) and others that rely on spectroscopic measurements at all epochs are not without reservation given that late time emission profiles are subject to more than line-shifts due to Doppler velocities and ionization balance (Blondin et al. 2012; Silverman et al. 2013a).

Maeda et al. (2010b) presented 14 3-dimensional (3D) high resolution Chandrasekhar-mass delayed-detonations that produce a range of 56Ni (depending on the location of ignition points) between ∼0.3 and 1.1 M. For this set of models, unburned carbon extends down to 4000 km s−1 while oxygen is not present below 10,000 km s−1. (Bongard et al. 2008; Friesen et al. 2012) conclude that if delayed-detonations are to viably produce normal SN Ia brightnesses, the region of ignition cannot be far off-center so as to avoid the over-production of 56Ni. As noted by Seitenzahl et al. (2013), these models warrant tests via spectrum synthesis given their 3D nature and possible predictive relations to the WLR, spectropolarimetry, and C+O “footprints” Seitenzahl et al. (2013).

Seitenzahl et al. (2013) recently compared synthetic light curves and spectra of a suite of DDT and PDD models. Based on comparisons to SN 2002bo and SN 2011fe, two SN Ia of different spectroscopic subtypes, and based on poor to moderate agreement between recent DDT models and observed SN Ia diversity (Howell et al. 2001; Baron et al. 2003; Thomas et al. 2007; Wang and Wheeler 2008), Dessart et al. (2013a) convincingly argue that these two SN Ia varieties (LVG vs. HVG, as above) are dissimilar enough to be explained by different explosion scenarios and/or progenitor systems (Blondin et al. 2011). For SN Ia in general, delineating spectroscopic diversity has been a difficult issue Dessart et al. (2013a), and has only recently been made clear with the belated release of decades-worth of unpublished data (Wang et al. 2013c).

3 Spectroscopic diversity of SN Ia

Observationally and particularly at optical wavelengths, SN Ia increase in brightness over ∼13 to 23 days before reaching maximum light (\(\overline{t}_{rise} = 17.38 \pm 0.17\); (Benetti et al. 2005; Branch et al. 2009)). However, it is not until ∼1 year later that the period of observation is said to be “complete.” From the time of the explosion our perspective as outside observers begins at the outermost layers if the SN Ia is caught early enough. In the approximate sense, this is because the line-forming region (the “photosphere”) recedes as the ejecta expand outward, which in turn means that the characteristic information for each explosion mechanism and progenitor channel is specified by the temporal spectrophotometric attributes of the “inner” and “outer” layers of freshly synthesized and remaining primordial material. In addition, because the expanding material cools as it expands, the net flux of photons samples different layers (of different states and distributions) over time. And since the density profile of the material roughly declines from the center outward, significant changes within the spectra for an individual SN Ia take place daily before or near maximum light, and weekly to monthly thereafter.

Documenting the breadth of temporal spectroscopic properties for each SN Ia is not only useful for theoretical purposes, but is also necessary for efficiently typing and estimating the epoch of newly found possible supernova candidates before they reach maximum light. Several supernova identification tools have been made that allow for fair estimates of both subtype and epoch (e.g., SNID; (Blondin et al. 2012; Silverman et al. 2012c), Gelato; Hayden et al. 2010b, Superfit; Blondin and Tonry 2007). In addition, the spectroscopic goodness-of-fit methods of Harutyunyan et al. 2008 allow one to find the “nearest neighbors” of any particular SN Ia within a sample of objects, enabling the study of so called “transitional subtype” SN Ia (those attributed with contrasting characteristics of two or more subtypes).

3.1 Data

One of the major limitations of spectroscopic studies has been data quality. For example, the signal-to-noise ratio, S/N, of a spectrum signifies the quality across wavelength and is usually moderate to high for high-z events. Similarly, and at least for low-z SN Ia, there should exist a quantity that specifies the density of spectra within a time series dataset. We suggest \(\mathcal{S}/\mathcal{N}\bullet (\mathcal{P})\) ≡ the number of continual follow-up spectra / the mean number of nights passed between exposures • (total number of spectra prior to maximum light). In Fig. 4 we apply this quantity to literature data.

Fig. 4
figure 4

Continual spectroscopic follow-up efficiencies for the most “well-observed” SN Ia at early phases (not counting multiple spectra per day). Some of the values reported may be slightly lower for instances of unpublished data. Dashed lines represent the upper-limit spectroscopic efficiencies and peak number of pre-maximum light spectra for one and two day follow-up cadences during the first 25 days post-explosion. See Sect. 3.1

An ideal dataset consisting of 25 spectra during the first 25 days post-explosion would yield \(\mathcal{S}/\mathcal{N}\bullet(\mathcal {P})=25\) (16) (e.g., SN 2011fe), whereas a dataset of spectra at days −12, −10, −7, −4, +0, +3, +8, +21, +48, +119 (a common occurrence) would be said to have \(\mathcal{S}/\mathcal {N}\bullet(\mathcal{P})=3.3\) (4) plus follow-up at days +21, +48, and +119. By including the total number of spectra prior to maximum light in parentheses, we are anticipating those cases where \(\mathcal{S}/\mathcal{N} = 1\), but with \(\mathcal{P} = 3\), e.g., a dataset with days −12, −9, and −6 observed. It may serve a purpose to also add second and third terms to this quantity that take into account the number of post-maximum light and late time spectra.

Regardless of moniker and definition, a quantity that specifies the density of spectra observed during the earliest epochs would aid in determining, quantitatively, which datasets are most valuable for various SN Ia diversity studies. Clearly such a high follow-up rate for slow-evolving events (e.g., SN 2009dc) or events caught at maximum light are not as imperative. However, when SN Ia are found and typed early, a high \(\mathcal{S}/\mathcal{N}\) ensures no loss of highly time sensitive information, e.g., when high velocity features and C+O signatures dissipate. Since most datasets are less than ideal for detailed temporal inspections of many events (by default), astronomers have instead relied upon comparative studies (Sect. 3.2); those that maximize sample sizes by prioritizing the most commonly available spectroscopic observables, e.g., line velocities of 6100 Å absorption minima near maximum light.

Another limitation of spectroscopic studies has been the localized release of all published data. The Online Supernova Spectrum Archive (SuSpect;Footnote 11 Howell et al. 2005) carried the weight of addressing data foraging during the past decade, collecting a total of 867 SN Ia spectra (1741 SN spectra in all). Many of these were either at the request of or donation to SuSpect, while some other spectra were digitized from original publications in addition to original photographic plates (Jeffery et al. (2007), Richardson et al. 2001). Prior to and concurrent with SuSpect, D. Jeffery managed a collection of SUpernova spectra PENDing further analysis (SUSPENDFootnote 12).

With the growing need for a manageable influx of data, the Weizmann Interactive Supernova Data Repository (WISeREP;Footnote 13 Casebeer et al. 1998) has since served as a replacement and ideal central data hub, and has increased the number of SN Ia spectra to 7661 (with 7933 publicly available SN spectra out of 13,334 in all). We encourage all groups to upload published data to WISeREP, whether or not made available elsewhere.

3.1.1 Samples

By far the largest data releases occurred during the past five years, and are available on WISeREP and their affiliated archives. 2000 and Yaron and Gal-Yam 2012 presented 2603 optical spectra (∼3700–7500 Å on average) of 462 nearby SN Ia (\(\tilde{z} = 0.02\); ∼85 Mpc) obtained by the Center for Astrophysics (CfA) SN group with the F.L. Whipple Observatory from 1993 to 2008. They note that, of the SN Ia with more than two spectra, 313 SN Ia have eight spectra on average. Matheson et al. (2008) and the Berkeley SuperNova Ia Program (BSNIP) presented 1298 optical spectra (∼3300–10,400 Å on average) of 582 low-redshift SN Ia (z<0.2; ∼800 Mpc) observed from 1989 to 2008. Their dataset includes spectra of nearly 90 spectroscopically peculiar SN Ia. Blondin et al. (2012) released 569 optical spectra of 93 low-redshift SN Ia (\(\tilde{z}\sim 0.04\); ∼170 Mpc) obtained by the Carnegie Supernova Project (CSP) between 2004 and 2009. Notably, 72 CSP SN Ia have spectra earlier than 5 days prior to maximum light, however only three SN Ia have spectra as early as day −12.

These samples provide a substantial improvement and crux by which to explore particular issues of SN Ia diversity. However, the remaining limitation is that our routine data collection efforts continue to yield several thousand SN Ia with few to several spectra by which to dissect and compare SN Ia atmospheres.

3.1.2 Comparisons of “well-observed” SN Ia

Given that both quantitative and qualitative spectrum comparisons are at the heart of SN Ia diversity studies, in Figs. 58, 913 we plot spectroscopic temporal snapshots for as many “well-observed” SN Ia as are currently available on WISeREP (Tables 1 and 2). Because the decline parameter, Δm 15(B), remains a useful parameter for probing differences of synthesized 56Ni mass, properties of the ejecta, limits of CSM interaction, etc., we have loosely ordered the spectra with increasing Δm 15(B) (top-down) based on average values found throughout the literature (Tables 3, 4, 5, 6) and M B (peak) considerations for cases that are reported as having the same Δm 15(B). The spectra have been normalized with respect to 6100 Å line profiles in order to amplify relative strengths of the remaining features (see caption of Fig. 5). We also denote the spectroscopic subtype for each object in color in order to show the overlap of these properties between particular SN Ia subclasses (see Sect. 3.2 and Munari et al. (2013)).

Fig. 5
figure 5

Early pre-maximum light, rest frame optical spectra of some of the most well-observed and often referenced SN Ia are plotted, loosely in order of increasing Δm 15(B) (top-down). Subtypes shown include bright SN 2006gz, 2009dc-like super-Chandrasekhar candidate (SCC; purple), high-ionization, shallow-silicon SN 1991T-like (SS; red), normal SN 1994D, 2005cf, 2011fe-like (CN; black), broad-lined SN 1984A, 2002bo-like (BL; green), and sub-luminous, low-ionization SN 1991bg, 2004eo-like (CL; blue) SN Ia. The horizontal dashed lines represent our normalization bounds that were applied to each spectrum. This ensures a fair comparison of all relevant spectroscopic features, sans continuum differences. For the SS SN Ia, in Figs. 5 and 6 only, we have normalized to the Fe III feature as indicated. For the purposes of this review, we have only included SN Ia that have received particular attention within the literature (see Sect. 4 and the Appendix). Many other time series observations can be found in Silverman et al. (2012a), Folatelli et al. (2013), and Matheson et al. (2008). The peculiar PTF09dav is shown in Fig. 8 for comparison, as it is not a prototypical SN Ia, however appearing similar to SN 1991bg-like events Silverman et al. (2012c)

Fig. 6
figure 6

1-Week pre-maximum light optical spectroscopic comparisons. See Fig. 5 caption

Table 1 References for Spectra in Figs. 513
Table 2 References for Spectra in Figs. 513
Table 3 References for M B (Peak) and Δm 15(B) plotted in Fig. 14: 1981–1992
Table 4 References for M B (Peak) and Δm 15(B) plotted in Fig. 14: 1994–1999
Table 5 References for M B (Peak) and Δm 15(B) plotted in Fig. 14: 2000–2005
Table 6 References for M B (Peak) and Δm 15(B) plotted in Fig. 14: 2006–2012

By inspection, the collected spectra show how altogether different and similar SN Ia (both odd and normal varieties) have come to be since nearly 32 years ago. With regard to the recent modeling of Pereira et al. (2013) and their accompanying synthetic spectra, we plot the spectra in Figs. 511 in the flux-representation of λ 2F λ for ease of future comparisons. These juxtapositions should reveal the severity of the SN Ia diversity problem as well as the future of promising studies and work that lie ahead.

3.2 Deciphering 21st century SN Ia subtypes

Observationally, the whole of SN Ia are hetero-, homogeneous events Foley et al. (2012b); some of the observed differences in their spectra are clear, while other suspected differences are small enough to fall below associable certainty. Because of this, observational studies have concentrated on quantitatively organizing a mapping between the most peculiar and normal events. In this section our aim is to review SN Ia subtypes. In all, three observational classification schemes will be discussed Silverman and Filippenko (2012), as well as the recent additions of so-called over- and sub-luminous events (see Munari et al. (2013); Childress et al. (2013c); Blondin et al. 2012 and references therein). For other relatively new and truly peculiar subclasses of supernova transients, we refer the reader to Blondin et al. (2013), (Oke and Searle 1974; Filippenko 1997) and references therein.

3.2.1 Benetti et al. ((Benetti et al. 2005; Branch et al. 2006; Wang et al. 2009a)) classification

Understanding the origin of the WLR is a key issue for understanding the diversity of SN Ia light curves and spectra, as well as their use as cosmological distance indicators. Brighter SN Ia with broader light curves tend to occur in late-type spiral galaxies, while dimmer, faster declining SN Ia are preferentially located in an older stellar population and thus the age and/or metallicity of the progenitor system may be relevant factors affecting SN Ia properties (Scalzo et al. 2012; see also Foley et al. 2013).

With this in mind, Silverman et al. 2013d studied the observational properties of 26 well-observed SN Ia (e.g., SN 1984A, 1991T, 1991bg, 1994D) with the intent of exploring SN Ia diversity. Based on the observed projected Doppler velocity evolution from the spectra,Footnote 14 in conjunction with characteristics of the light curve (M B , Δm 15), Kasliwal et al. (2012) considered three different groups of SN Ia: (1) “FAINT” SN 1991bg-likes, (2) “low velocity gradient” (LVG) SN 1991T/1994D-likes, and (3) “high velocity gradient” (HVG) SN 1984A-like events. The velocity gradient here is based on the time-evolution of 6100 (“6150”) Å absorption minima as inferred from Si II λ6355 line velocities. Overall, HVG SN Ia have higher mean expansion velocities than FAINT and LVG SN Ia, while LVG SN Ia are brighter than FAINT and HVG SN Ia on average 2005. Given an apparent separation of SN Ia subgroups from this sample of 26 objects, Hamuy et al. 1995; Howell 2001; Pan et al. 2013 considered it as evidence that LVG, HVG, and FAINT classifications signify three distinct kinds of SN Ia.

3.2.2 Branch et al. (Hicken et al. (2009a)) classification

Branch et al. (Benetti et al. (2005), (Blondin et al. 2012), Benetti et al. (2005), (Silverman et al. 2012b; Blondin et al. 2012), Benetti et al. (2005)) published a series of papers based on systematic, comprehensive, and comparative direct analysis of normal and peculiar SN Ia spectra at various epochs with the parameterized supernova synthetic spectrum code, SYNOW Footnote 15 2005. From the systematic analysis of 26 spectra of SN 1994D, 2006 infer a compositional structure that is radially stratified, overall. In addition, several features are consistent with being due to permitted lines well into the late post-maximum phases (∼120 days, see 2007b). Another highlight of this work is that, barring the usual short comings of the model, SYNOW is shown to provide a necessary consistency in the direct quantification of spectroscopic diversity 2008. Consequently, the SYNOW model has been useful for assessing the basic limits of a spectroscopic “goodness of fit” (Fig. 3), with room for clear and obvious improvements 2009.

In their second paper of the series on comparative direct analysis of SN Ia spectra, Thomas et al. 2011a studied the spectra of 24 SN Ia close to maximum light. Based on empirical pEW measurements of 5750, 6100 Å absorption features, in addition to spectroscopic modeling with SYNOW, (Fisher 2000; Branch et al. 2007a) organized SN Ia diversity by four spectroscopic patterns: (1) “Core-Normal” (CN) SN 1994D-likes, (2) “Broad-line” (BL), where one of the most extreme cases is SN 1984A, (3) “Cool” (CL) SN 1991bg-likes, and (4) “Shallow-Silicon” (SS) SN 1991T-likes. In this manner, a particular SN Ia is defined by its spectroscopic similarity to one or more SN Ia prototype via 5750, 6100 Å features. These spectroscopic subclasses also materialized from analysis of pre-maximum light spectra Branch et al. (2005).

The overlap between both Branch et al. 2008; Friesen et al. 2012 and (Branch et al. 2007a) classifications schemes comes by comparing Table 1 in (Friesen et al. 2012) to Table 1 of Branch et al. (2006), and it reveals the following SN Ia descriptors: HVG–BL, LVG–CN, LVG–SS, and FAINT–CL. This holds true throughout the subsequent literature Branch et al. (2006).

In contrast with (Branch et al. 2007b) who interpreted FAINT, LVG, and HVG to correspond to the “discrete grouping” of distinctly separate SN Ia origins among these subtypes, Benetti et al. (2005) found a continuous distribution of properties between the four subclasses defined above. We should point out that this classification scheme of Branch et al. (2006) is primarily tied to the notion that SN Ia spectroscopic diversity is related to the temperature sequence found by Benetti et al. (2005). That is, despite the contrast with Branch et al. (2006) (continuous versus discrete subgrouping of SN Ia), so far these classifications say more about the state of the ejecta than the various number of possible progenitor systems and/or explosion mechanisms (see also (Branch et al. 2009; Folatelli et al. 2012; Blondin et al. 2012; Silverman et al. 2012b)). Furthermore, the existence of “transitional” subtype events support this notion (e.g., SN 2004eo, 2006bt, 2009ig, 2001ay, and PTF10ops; see appendix).

Benetti et al. (2005) later analyzed a larger sample of SN Ia spectra. They found that SN 1991bg-likes are not a physically distinct subgroup Branch et al. (2006), and that there are probably many SN 1999aa-like events (A.5) that similarly may not constitute a physically distinct variety of SN Ia.

With regard to the fainter variety of SN Ia, Branch et al. (2006) made detailed comparative analysis of spectra of peculiar SN 1991bg-likes. They also studied the intermediates, such as SN 2004eo (A.23), and discussed the spectroscopic subgroup distribution of SN Ia. The CL SN Ia are dim, undergo a rapid decline in luminosity, and produce significantly less 56Ni than normal SN Ia. They also have an unusually deep and wide trough in their spectra around 4200 Å  suspected as due to Ti II Nugent et al. 1995, in addition to a relatively strong 5750 Å absorption (due to more than Si II λ5972; see Benetti et al. (2005)). Dessart et al. 2013a analyzed the spectra of SN 1991bg, 1997cn, 1999by, and 2005bl using SYNOW, and found this group of SN Ia to be fairly homogeneous, with many of the blue spectral features well fit by Fe II.

3.2.3 Wang et al. (Branch et al. (2009)) classification

Based on the maximum light expansion velocities inferred from Si II λ6355 absorption minimum line velocities, (Doull and Baron 2011) studied 158 SN Ia, separating them into two groups called “high velocity” (HV) and “normal velocity” (NV). This classification scheme is similar to those previous of Doull and Baron (2011) and (Filippenko et al. 1992b), where NV and HV SN Ia are akin to LVG–CN and HVG–BL SN Ia, respectively. That is, while the subtype notations differ among authors, memberships between these classification schemes are roughly equivalent (apart from outliers such as the HV-CN SN 2009ig, see Bongard et al. 2008).

Explicitly, Doull and Baron (2011) and 2009a subclassifications are based on empirically estimated mean expansion velocities near maximum light (±4 days; ±500–2000 km s−1) of 6100 Å features produced by an assumed single broad component of Si II. The notion of a single photospheric layer, much less a single-epoch snapshot, does not realistically account for the multilayered nature of spectrum formation Wang et al. (2009a), its subsequent evolution post-maximum light Benetti et al. (2005), and potential relations to line-of-sight considerations Branch et al. (2006). In the strictest sense of SN Ia sub-classification, “normal” refers to both of these subtypes since they differ foremost by a continuum of inferred mean expansion velocities and the extent of expansion opacities, simultaneously.

Furthermore, note from a sample of 13 LVG and 8 HVG SN Ia that Blondin et al. 2012 found \(10 \lesssim \dot{v}_{Si\ II}\) (km s−1 day−1) ≲67 (±7) and \(75 \lesssim \dot{v}_{Si\ II} \lesssim 125\) (±20) for each, respectively. Similarly, and from a sample of 14 LVG and 29 HVG SN Ia, Benetti et al. (2005) report that \(10\lesssim \dot{v}_{Si\ II}\lesssim 445\) (±50) and \(15 \lesssim \dot{v}_{Si\ II}\lesssim 290\) (±140) for LVG and HVG events, respectively. Additionally, the pEW measurements of 5750, 6100 Å absorption features (among others) are seen to share a common convergence in observed values Wang et al. (2009a). The continually consistent overlap between the measured properties for these two SN Ia “subtypes” implies that the notion of a characteristic separation value for \(\dot{v}_{Si\ II}\sim 70\) km s−1 day−1 (including the inferred maximum light separation velocity, v 0≳12,000 km s−1) is still devoid of any physical significance beyond overlapping bimodal distributions of LVG–CN and HVG–BL SN Ia properties (see Sect. 5.3 of (Bongard et al. 2008), Sect. 5.2 of (Patat et al. 1996; Scalzo et al. 2012), and (Maeda et al. 2010b; Blondin et al. 2011; Moll et al. 2013)). Rather, a continuum of empirically measured properties exists between the extremities of these two particular historically-based SN Ia classes (e.g., SN 1984A and 1994D). Given also the natural likelihood for a physical continuum between NV and HV subgroups, considerable care needs to be taken when concluding on underlying connections to progenitor systems from under-observed, early epoch snapshots of blended 6100 Å absorption minima.

Hence, the primary obstacle within SN Ia diversity studies has been that it is not yet clear if the expanse of all observed characteristics of each subtype has been fully charted. For the observed properties of normal SN Ia, it is at least true that \(\dot{v}_{Si\ II}\) resides between 10–445 km s−1 day−1, with a median value of ∼60–120 km s−1 day−1 Benetti et al. (2005), while the rise to peak B-band brightness ranges from 16.3 to 19 days Silverman et al. (2012b).

Recently, (Branch et al. 2006; Hachinger et al. 2006; Blondin et al. 2012; Silverman et al. 2012b) applied this NV and HV subgrouping to 123 “Branch normal” SN Ia with known positions within their host galaxies and report that HV SN Ia more often inhabit the central and brighter regions of their hosts than NV SN Ia. This appears to suggest that a supernova with “higher velocities at maximum light” is primarily a consequence of a progenitor with larger than solar metallicities, or that PDD/HVG SN Ia are primarily found within the galactic distribution of DDT/LVG SN Ia (c.f. Silverman et al. 2012b). This is seemingly in contrast to interpretations of Blondin et al. 2012 who propose, based on both early epoch and late time considerations, that LVG and HVG SN Ia are possibly one in the same event where the LVG-to-HVG transition is ascribed to an off-center ignition.

While it is true that increasing the C+O layer metallicity can affect the blueshift of the 6100 Å absorption feature—in addition to lower temperatures and increased UV line-blocking—this is not primarily responsible for the shift in 6100 Å absorption minima (Silverman et al. 2012a, (Benetti et al. 2005; Blondin et al. 2012; Silverman et al. 2012a)), where the dependence of this effect is not easily decoupled from changes in the temperature structure (Ganeshalingam et al. 2011; Mazzali et al. 2013). However, it is also worthwhile to point out that, while the early epoch spectra of SN 2011fe (a NV event) are consistent with a DDT-like composition with a sub-solar C+O layer metallicity (“W7+,” Wang et al. (2013c)) and a PDD-like composition Blondin et al. 2011, 2012; Dessart et al. 2013a, the outermost layers of SN 2010jn (a HV event; A.41) are practically void of unburned material and subsequently already overabundant in synthesized metals for progenitor metallicity to be well determined Maeda et al. (2010b). Therefore, discrepancies between NV and HV SN Ia must still be largely dependent on more than a single parameter, e.g. differences in explosion mechanisms Lentz et al. 2001a, where progenitor metallicity is likely to be only one of several factors influencing the dispersions of each subgroup 2001b.

It should be acknowledged again that metallicity-dependent aspects of stellar evolution are expected to contribute, in part, to the underlying variance of holistic SN Ia characteristics. However thus far, the seen discrepancies from metallicities share similarly uncertain degrees of influence as for asymmetry and line-of-sight considerations of ejecta-CSM interactions for a wide variety of SN Ia (Lentz et al. 2000). Similar to this route of interpretation for SN Ia subtypes are active galactic nuclei and the significance of the broad absorption line quasi-stellar objects (BALQSOs, see Mazzali et al. 2013).

3.2.4 Additional peculiar SN Ia subtypes

Spectroscopically akin to some luminous SS SN Ia are a growing group of events thought to be “twice as massive,” aka super-Chandrasekhar candidates (SCC, (Dessart et al. 2013a), (Hachinger et al. 2013); (Dessart et al. 2013a; Moll et al. 2013)). Little is known about this particular class of over-luminous events, which is partly due to there having been only a handful of events studied. Thus far, SCC SN Ia are associated with metal-poor environments (Lentz et al. 2000; Höflich et al. 2010; Wang et al. 2012). Spectroscopically, the differences that set these events apart from normal SN Ia are fairly weak Si II/Ca II signatures and strong C II absorption features relative to the strength of Si II lines. Most other features are comparable in relative strengths to those of normal SN Ia, if not muted by either top-lighting or effects of CSM interaction (Lentz et al. 2000; Kasen et al. 2003; Leloudas et al. 2013), and are less blended overall due to lower mean expansion velocities. In addition, there is little evidence to suggest that SCC SN Ia spectra consist of contributions from physically separate high velocity regions of material (≳4000 km s−1 above photospheric). This range of low expansion velocities (∼5000–18,000 km s−1), in conjunction with larger than normal C II absorption signatures, are difficult to explain with some M Ch explosion models (de Kool and Begelman 1995; Becker et al. 1997; Elvis 2000; Branch et al. 2002; Hamann and Sabra 2004; Casebeer et al. 2008; Leighly et al. 2009; Elvis 2012, however see also Howell et al. 2006; Jeffery et al. 2006; Hillebrandt et al. 2007; Hicken et al. 2007; Maeda et al. 2009; Chen and Li 2009; Yamanaka et al. 2009a for related discussions).

2013 recently searched the BSNIP and PTF datasets, in addition to the literature sample, and compiled a list of 16 strongly CSM interacting SN Ia (referred to as “Ia-CSM” events). These supernovae obtain their name from a conspicuous signature of narrow hydrogen emission atop a weaker hydrogen P Cygni profile that together are superimposed on a loosely identifiable SS-like SN Ia spectrum Scalzo et al. 2010; Tanaka et al. 2010; Yuan et al. 2010; Silverman et al. 2011; Taubenberger et al. 2011; Kamiya et al. 2012; Scalzo et al. 2012; Hachinger et al. 2012. Apart from exhibiting similar properties to the recent PTF11kx (Sect. 4.5.1) and SN 2005gj (Sect. 4.5.3), (Childress et al. 2011; Khan et al. 2011a) find that SN Ia-CSM have a range of peak absolute magnitudes (−21.3≤M R ≤−19), are a spectroscopically homogenous class, and all reside in late-type spiral and irregular host-galaxies.

As for peculiar sub-luminous events, (Branch et al. 2000; Leloudas et al. 2013) and Scalzo et al. 2012; Kamiya et al. 2012 discussed the heterogeneity of the SN 2002cx-like subclass of SN Ia. Consisting of around 25 members spectroscopically similar to SN 2002cx Hachisu et al. 2012; Dessart et al. 2013a; Moll et al. 2013, these new events generally have lower maximum light velocities spanning from 2000 to 8000 km s−1 and a range of peak luminosities that are typically lower than those of FAINT SN Ia (−14.2 to −18.9). In addition, this class of objects have “hot” temperature structures and—in contrast to SN Ia that follow the WLR—have low luminosities for their light curve shape. This suggests a distinct origin, such as a failed deflagration of a C+O white dwarf Silverman et al. (2013d) or double detonations of a sub-Chandrasekhar mass white dwarf with non-degenerate helium star companion (Aldering et al. 2006; Prieto et al. 2007; Leloudas et al. 2013). It is estimated that for every 100 SN Ia, there are 31\(\tiny\begin{array}{c}+17\\ -13\end{array}\) peculiar SN 2002cx-like objects in a given volume Silverman et al. (2013d).

3.2.5 SN Ia subtype summary

In Fig. 14 we plot average literature values of M B (peak), Δm 15(B), and \(\mathcal{V}_{\mathit{peak}}\)(Si II λ6355) versus one another for all known SN Ia subtypes. For M B (peak) versus Δm 15(B), the WLR is apparent. We have included the brightest SN 2002cx-likes (Fink et al. 2010; Sim et al. 2012; Wang et al. 2013b) for reference, as these events are suspected as having separate origins from the bulk of normal SN Ia (Foley et al. 2013). We have not included Ia-CSM events given that estimates of expansion velocities and luminosities, without detailed modeling, are obscured by CSM interaction. However, it suffices to say for Fig. 14 that Ia-CSM are nearest to SS and SCC SN Ia in both projected Doppler velocities and peak M R brightness Folatelli et al. (2012). At a separate end of these SN Ia diversity plane(s), \(\mathcal {V}_{peak}\)(Si II λ6355) versus Δm 15(B) further separates FAINT−CL SN Ia and peculiar events away from the pattern between SCC/SN 1991T-like over-luminous SN Ia and normal subtypes, where the former tend to be slow-decliners (i.e. typically brighter) with slower average velocities.

To summarize the full extent of SN Ia subtypes in terms of the qualitative luminosity and expansion velocity patterns, in Fig. 15 we have outlined how SN Ia relate to one another thus far (for quantitative assessments, see Blondin et al. (2012)). Broadly speaking, the red ward evolution of SN Ia features span low to high rates of decline for a large range of luminosities. Shallow Silicon and Super-Chandrasekhar Candidate SN Ia are by far the brightest, while Ia-CSM SN exhibit bright Hα emission features. These “brightest” SN Ia also show low to moderate expansion velocities and \(\dot{v}_{Si\ II}\). From BL to CN to SS/SSC SN Ia, mean peak absolute brightnesses scale up with an overall decrease in maximum light line velocities. Meanwhile, CL SN Ia fall between low velocity and high velocity gradients, but lean toward HVG SN Ia in terms of their photospheric velocity evolution. Comparatively, peculiar SN 2002cx-like and other sub-luminous events are by far the largest group of thermonuclear outliers.

Obtaining observations of SN Ia that lie outside the statistical norm is important for gauging the largest degree by which SN Ia properties diverge in nature. However, just as imperative for the cause remains filling the gaps of observed SN Ia properties (e.g., \(\dot{v}_{Si\ II}\), v neb , v C (t), v Ca (t), M B (peak), Δm 15(B-band), t rise , color evolution) with well-observed SN Ia. This is especially true for those SN Ia most similar to one another, aka “nearest neighbors” Pakmor et al. (2013), and transitional-type SN Ia.

3.3 Signatures of C+O progenitor material

If the primary star of most SN Ia is a C+O WD, and if the observed range of SN Ia properties is primarily due to variances in the ejected mass or abundances of material synthesized in the explosion (e.g., 56Ni), then this should also be reflected in the remaining amount of carbon and oxygen if M Ch is a constant parameter (see Blondin et al. (2012)). On the other hand, if one assumes that the progenitor system is the merger of two stars Folatelli et al. (2012) or a rapidly rotating WD Blondin et al. (2012)—both of which are effectively obscured by an amorphous region and/or disk of C+O material—then the properties of C and O absorption features will be sensitive to the interplay between ejecta and the remaining unburned envelope (see Pakmor et al. (2013)).

Oxygen absorption features (unburned plus burned ejecta) are often present as O I λ7774 in the pre-maximum spectra of SN Ia (Fig. 5). They may exhibit similar behavior to those seen in SN 2011fe (Sect. 4.1), however current datasets lack the proper temporal coverage of a large sample of events that would be necessary to confirm such claims. Still, comparisons of the blue-most wing in the earliest spectra of many SN Ia to that of SN 2009ig (Sect. 4.2.1), 2010jn (A.41), 2011fe (Sect. 4.1), 2012cg (A.43), and 2012fr (Sect. 4.2.2) may reveal some indication of HV O I if present and if caught early enough (e.g., SN 1994D; Blondin et al. (2012)).

Spectroscopic detections of carbon-rich material have been documented since the discovery of SN 1990N (see (Foley et al. 2013)) and have been primarily detected as singly ionized in the optical spectra of LVG–CN SN Ia (Hillebrandt et al. 2013). However, NIR spectra of some SN Ia subtypes have been suspected of harboring C I absorption features ((Silverman et al. 2013d), see also Blondin et al. 2012; Silverman et al. 2012b; Folatelli et al. 2013), while C III has been tentatively identified in “hotter” SS/SN 1991T-like SN Ia (Jeffery et al. 2007).

Observations of the over-luminous SCC SN 2003fg suggested the presence of a larger than normal C II λ6580 absorption signature Maeda et al. 2010a; Blondin et al. 2013; Dessart et al. 2013a. Later in 2006, with the detection of a conspicuous C II λ6580 absorption “notch” in the early epoch observations of the normal SN Ia, SN 2006D, (Webbink 1984; Iben and Tutukov 1984; Pakmor et al. 2013; Moll et al. 2013) reconsidered the question of whether or not spectroscopic signatures of carbon were a ubiquitous property of all or at least some SN Ia subtypes.

As follow-up investigations, (Hachisu et al. 2012) and Livio and Pringle 2011 presented studies of carbon features in SN Ia spectra, particularly those of C II λλ6580, 7234 (which are easier to confirm than λλ4267, 4745). However weak, conspicuous 6300 Å absorption features were reported in several SN Ia spectra obtained during the pre-maximum phase. It was shown that most of the objects that exhibit clear signatures are of the LVG—CN SN Ia subtype, while HVG—BL SN Ia may either be void of conspicuous signatures due to severe line blending, or lack carbon altogether, the latter of which is consistent with DDT models (e.g., Branch et al. 2005) and could also be partially due to increased progenitor metallicities Leibundgut et al. 1991; Jeffery et al. 1992; Branch et al. 2007b; Tanaka et al. 2008. This requires further study and spectrum synthesis from detailed models.

(Parrent et al. 2011) presented additional evidence of unburned carbon at photospheric velocities from observations of 5 SN Ia obtained by the Nearby Supernova Factory. Detections were based on the presence of relatively strong C II 6300 Å absorption signatures in multiple spectra of each SN, supported by automated fitting with the SYNAPPS code Höflich et al. 2002; Hsiao et al. 2013. They estimated that at least 22\(\tiny\begin{array}{c} +10\\-6\end{array}\) % of SN Ia exhibit spectroscopic C II signatures as late as day −5, i.e. carbon features, whether or not present in all SN Ia, are not often seen even as early as day −5.

Marion et al. 2006, 2009b later searched through the Carnegie Supernova Project (CSP) sample and found at least 30 % of the objects show an absorption feature that can be attributed to C II λ6580. (Hatano et al. 2002; Garavini et al. 2004; Chornock et al. 2006) searched for carbon in the BSNIP sample and found that ∼11 % of the SN Ia studied show carbon absorption features, while ∼25 % show some indication of weak 6300 Å absorption. From their sample, they find that if the spectra of SN Ia are obtained before day −5, then the detection percentage is higher than ∼30 %. Recently it has also been confirmed that “carbon-positive” SN Ia tend to have bluer near-UV colors than those without conspicuous C II λ6580 signatures (Howell et al. 2006).

Thomas et al. (2007) estimate the range of carbon masses in normal SN Ia ejecta to be (2–30)×10−3 M. For SN 2006D, Parrent et al. (2011) estimated 0.007 M of carbon between 10,000 and 14,000 km s−1 as a lower limit. Folatelli et al. (2012) also note that the most vigorous model of Hachinger et al. 2013 left behind 0.085 M of carbon in the same velocity interval. However, we are not aware of any subsequent spectrum synthesis for this particular model that details the state of an associated 6300 Å signature.

In the recent detailed study on SN Ia spectroscopic diversity, (Lentz et al. 2000; Meng et al. 2011; Milne et al. 2013) searched for signatures of C II λ6580 in a sample of 2603 spectra of 462 nearby SN Ia and found 23 additional “carbon-positive” SN Ia. Given that seven of the nine CN SN Ia reported by Thomas et al. (2011b) with spectra prior to day −10 clearly exhibit signatures of C II, and that ∼30–40 % of the SN Ia within their sample are of the CN subtype, it is likely that at least 30–40 % of all SN Ia leave behind some amount of carbon-rich material, spanning velocities between 8000–18,000 km s−1 (Thomas et al. 2011a).

Considering the volume-limited percentage of Branch normal SN Ia estimated by Folatelli et al. (2012), roughly 50 % or more are expected to contain detectable carbon-rich material in the outermost layers. If this is true, then it implies that explosion scenarios that do not naturally leave behind at least a detectable amount (pEWs ∼5–25 Å) of unprocessed carbon can only explain half of all SN Ia or less (sans considerations of Ia-CSM progenitors, subtype-ejecta hemisphere dualities, and effects of varying metallicities; see below).

Historically, time-evolving signatures of C II λλ6580, 7234 from the computed spectra of some detailed models have not revealed themselves to be consistent with the current interpretations of the observations. This could be due to an inadequate lower-extent of carbon within the models Silverman and Filippenko (2012) or the limits of the resolution for the computed spectra (Thomas et al. 2011b; Silverman and Filippenko 2012; Milne et al. 2013).

It should be noted that 6300 Å features are present in the non-LTE pre-maximum light spectra of Silverman and Filippenko (2012) who assessed metallicity effects on the spectrum for a pure deflagration model (see their Fig. 7). Overall, Thomas et al. 2007 find that an increase in C+O layer metallicities results in a decreased flux (primarily UV) in addition to a blue ward shift of absorption minima (primarily the Si II 6100 Å feature). While Thomas et al. 2007 did not discuss whether or not the weak 6300 Å absorption signatures are due to C II λ6580, it is likely the case given that an increase in C+O layer metallicities is responsible for the seen decrease in the strength of the 6300 Å feature. However, it should be emphasized that the “strength” of this supposed C II λ6580 feature appears to be a consequence of how Röpke et al. (2006) renormalized abundances for metallicity enhancements in the C+O layer. In other words, even though the preponderance of normal SN Ia with detectable C II λ6580 notches are of the NV subgroup, the fact that HV SN Ia are thus far void of 6300 Å notches does not imply robust consistency with the idea that nearest neighbor HV SN Ia properties are solely the result of a progenitor with relatively higher metallicities Blondin et al. (2012). Such a claim would need to be verified by exploring a grid of models with accompanying synthetic spectra.

Fig. 7
figure 7

Maximum light optical spectroscopic comparisons. See Fig. 5 caption

Fig. 8
figure 8

1-Week post-maximum light optical spectroscopic comparisons. See Fig. 5 caption

Fig. 9
figure 9

Two weeks post-maximum light optical spectroscopic comparisons. See Fig. 5 caption

Fig. 10
figure 10

1-Month post-maximum light optical spectroscopic comparisons. See Fig. 5 caption

Fig. 11
figure 11

2-Months post-maximum light optical spectroscopic comparisons. See Fig. 5 caption

Fig. 12
figure 12

100+ days post-maximum light optical spectroscopic comparisons. See Fig. 5 caption

Fig. 13
figure 13

Late-time optical spectroscopic comparisons. See Fig. 5 caption

Fig. 14
figure 14

Top: Peak absolute B-band magnitudes versus Δm 15(B) for most well-observed SN Ia found in the literature. Additional data (grey) taken from Narayan et al. (2011), Foley et al. (2013), and additional points discussed in (Li et al. 2003). Bottom: Expansion velocities at maximum light (± 3 days; via Si II λ6355 line velocities) versus Δm 15(B). All subtypes have been tagged in accordance with the same color-scheme as in Figs. 513. Included for reference are the brightest, peculiar SN 2002cx-likes (light blue circles). Outliers for each subtype have been labeled for clarity and reference. We also plot mean values for the SCC and CL subtypes (larger circles), and include the mean values for SS, CN, and BL SN Ia (large diamonds) as reported by (Foley et al. 2009; Jordan et al. 2012; Kromer et al. 2013a)

Fig. 15
figure 15

Subtype reference diagram. Dashed lines denote an open transitional boundary between adjacent spectroscopic subtypes

Additionally, carbon absorption features could signify an origin that is separate from explosion nucleosynthesis if most SN Ia are the result of a merger. For example, Blondin et al. (2012) recently presented angle-averaged synthetic spectra for a few “peri-merger” detonation scenarios. In particular, they find a causal connection between “normal” C II λ6580 signatures and the secondary star for both sub- and super-Chandrasekhar mass cases (c.f. (Parrent et al. 2011; Pereira et al. 2013; Cartier et al. 2013)). With constraints from UV spectra Li et al. (2011b) and high velocity features (Sect. 3.4), peri-megers can be used to explore the expanse of their spectroscopic influence within the broader picture of SN Ia diversity.

Coincident with understanding the relevance of remaining carbon-rich material is the additional goal of grasping the spectroscopic role of species that arise from carbon-burning below the outermost layers, e.g., magnesium (Thomas et al. 2007; Parrent et al. 2011; Blondin et al. 2012). While signatures of Mg II λλ4481, 7890 are frequently observed at optical wavelengths during the earliest phases prior to maximum light, these wavelength regions undergo severe line-blending compared to the NIR signatures of Mg II. Consequently, Mg II λ10927 has served as a better investment for measuring the lower regional extent and conditions during which a DDT is thought to have taken place (e.g., (Blondin et al. 2011); Lentz et al. (2000), Lentz et al. (2000), Lentz et al. (2000); Lentz et al. (2000), however see our Sect. 4.1).

3.4 High velocity (>16,000 km s−1) features

The spectra of many SN Ia have shown evidence for high-velocity absorption lines of the Ca II NIR triplet (IR3) in addition to an often concurrent signature of high-velocity Si II λ6355 ((Lentz et al. 2001b), Moll et al. (2013)). Most recently, high velocity features (HVFs) have also been seen in SN 2009ig Hicken et al. 2007; Zheng et al. 2013; Dessart et al. 2013a, SN 2012fr (Milne et al. 2013), and the SN 2000cx-like SN 2013bh (Wheeler et al. 1998). Overall, HVFs are more common before maximum light, display a rich diversity of behaviors Rudy et al. 2002, tend to be concurrent with polarization signatures Marion et al. 2003, and may be due to an intrinsically clumpy distribution of material 2006.

2009b showed that the Si II λ6355 line velocity decline rate, \(\dot{v}_{Si\ II}\), is correlated with the polarization of the same line at day −5, p Si II , and is consistent with an asymmetric distribution of IMEs. This interpretation is also complimentary with a previous finding that \(\dot{v}_{Si\ II}\) is correlated with v neb , the apparent Doppler line shift of [Fe II λ7155] emitted from the “core” during late times Hsiao et al. 2013. For the recent SN 2012fr, high velocity features of Ca II IR3 and Si II λ6355 at day −11 show concurrent polarization signatures that decline in strength during post-maximum light phases Mazzali et al. 2005a.

As for the origin of HVFs, they may be the result of abundance and/or density enhancements due to the presence of a circumstellar medium 2005b. If abundance enhancements are responsible, it could be explained by an overabundant, outer region of Si and Ca synthesized during a pre-explosion simmering phase (see (Foley et al. 2012c) and (Childress et al. 2013c)). On the other hand, the HVFs in LVG SN Ia spectra could indicate the presence of an opaque disk. For example, it is plausible that HVFs are due to magnetically induced merger outflows ((Silverman et al. 2013c), pending abundance calculations of a successful detonation), or interaction with a tidal tail and/or secondary star (e.g., (Childress et al. 2013b)).

Most recently, (Leonard et al. 2005; Tanaka et al. 2010) studied 58 low-z SN Ia (z<0.03) with well-sampled light curves and spectra near maximum light in order to access potential relationships between light curve decline rates and empirical relative strength measurements of Si II and Ca II HVFs. They find a consistent agreement with (Howell et al. 2001; Kasen et al. 2003; Thomas et al. 2004; Tanaka et al. 2006; Hole et al. 2010) in that the Ca II velocity profiles assume a variety of characteristics for a given Δm 15(B) solely because of the overlapping presence of HVFs. In addition, Maund et al. (2010b) show for their sample that the presence of HVFs is not strongly related to the overall intrinsic (BV) max colors. It is also seen that SN Ia with Δm 15(B)>1.4 continue to be void of conspicuous HVFs, while the strength of HVFs in normal SN Ia is generally larger for objects with broader light curves. Finally, and most importantly, the strength of HVFs at maximum light does not uniquely characterize HVF pre-maximum light behavior.

Notably, (Maeda et al. 2010b; Silverman et al. 2013a) find no correlation between nebular velocity and Δm 15(B), and for a given light-curve shape there is a large range of observed nebular velocities. Similarly (Maund et al. 2013) found no relation between the FWHM of late time 4700 Å iron emission features and Δm 15(B). This implies the peak brightness of these events do not translate toward uniquely specifying their late time characteristics, however the data do indicate a correlation between observed (BV) max and this particular measure of line-of-sight nebular velocities.

We should also note that while HVG SN Ia do not clearly come with HVFs in the same sense as for LVG SN Ia, the entire 6100 Å absorption feature for HVG SN Ia spans across velocity intervals for HVFs detected in LVG SN Ia. This makes it difficult to regard LVG and HVG subtypes as two separately distinct explosion scenarios. Instead we can only conclude that HVFs are a natural component of all normal SN Ia, whether conspicuously separate from a photospheric region or concealed as an extended region of absorbing material in the radial direction.

3.5 Empirical diversity diagnostics

The depth ratio between 5750 and 6100 Å absorption features, \(\mathcal{R}\)(“Si II”) (Gerardy et al. 2004; Quimby et al. 2006b), has been found to correlate with components of the WLR. In addition, Piro 2011 find a rich diversity of \(\mathcal{R}\)(Si II) pre-maximum evolution among LVG and HVG SN Ia.

As for some observables not directly related to the decline rate parameter, Zingale et al. 2011 studied a small sample of well observed SN Ia and found no apparent correlation between the blue-shift of the 6100 Å absorption feature at the time of maximum and Δm 15(B). Similarly, Ji et al. 2013 showed that \(\mathcal{R}\)(Si II) does not correlate well with v 10(Si II), the photospheric velocity derived from the Si II λ6355 Doppler line velocities 10 days after maximum light. This could arise from two or more explosion mechanisms, however Raskin and Kasen 2013; Moll et al. 2013 note that their interpretation is “rudimentary” on account of model uncertainties and the limited number of temporal datasets available at that time. In the future, it would be worthwhile to re-access these trends with the latest detailed modeling.

Childress et al. (2013b) made empirical measurements of spectroscopic feature pEWs, flux ratios, and projected Doppler velocities for 28 well-observed SN Ia, which include LVG, HVG, and FAINT subtypes. For normal LVG SN Ia they find similar observed maximum light velocities (via Si II λ6355; ∼9000–10,600 km s−1). Meanwhile the HVG SN Ia in their sample revealed a large spread of maximum light velocities (∼10,300–12,500 km s−1), regardless of the value of Δm 15(B). This overlap in maximum light velocities implies a natural continuum between LVG and HVG SN Ia (enabling unification through asymmetrical contexts; Maguire et al. (2012)). They also note that FAINT SN Ia tend to show slightly smaller velocities at B-band maximum for larger values of Δm 15(B), however no overreaching trend of maximum light expansion velocities from LVG to HVG to FAINT SN Ia was apparent from this particular sample of SN Ia.

Childress et al. (2013b) did find several flux ratios to correlate with Δm 15(B). In particular, they confirm that the flux ratio, \(\mathcal{R}\)(“S II λ5454, 5640”/“Si II λ5972”), is a fairly reliable spectroscopic luminosity indicator in addition to \(\mathcal{R}\)(Si II). Silverman et al. (2013a) conclude that these and other flux ratio comparisons are the result of changes in relative abundances across the three main SN Ia subtypes. In a follow-up investigation, Blondin et al. (2012) argue that the correlation with luminosity is a result of ionization balance, where dimmer objects tend to have a larger value of \(\mathcal{R}\)(Si II). (Nugent et al. 1995) later studied correlations between these and other flux ratios of SN Ia from the BSNIP sample and find evidence to suggest that CSM-associated events tend to have larger 6100 Å blue-shifts in addition to broader absorption features at the time of maximum light (see also Benetti et al. (2005)).

Patat et al. (1996) studied the \(\mathcal{R}\)(Si II) ratio and expansion velocities of intermediate-redshift supernovae. They find that the comparison of intermediate-redshift SN Ia spectra with high S/N spectra of nearby SN Ia do not reveal significant differences in the optical features and the expansion velocities derived from the Si II and Ca II lines that are within the range observed for nearby SN Ia. This agreement is also found in the color and decline of the light curve (see also Hatano et al. (2000)).

While the use of empirically determined single-parameter descriptions of SN Ia have proved to be useful in practice, they do not fully account for the observed diversity of SN Ia Hatano et al. (2000). With regard to SN Ia diversity, it should be reemphasized that special care needs to be taken with the implementation of flux ratios and pEWs. Detailed modeling is needed when attempting to draw connections between solitary characteristics of the observed spectrum and the underlying radiative environment, where a photon-ray’s route crosses many radiative contributions that form the spectrum’s various shapes, from UV to IR wavelengths. For example, the relied upon 5750, 6100 Å features used for \(\mathcal{R}\)(Si II) have been shown to be influenced by more than simply Si II, as well as from more locations (and therefore various temperatures) than a single region of line formation Hachinger et al. (2006). In fact, it is likely that a number of effects are at play, e.g., line blending and phase evolution effects. Furthermore, v 10(Si II) is a measure of the 6100 Å absorption minimum during a phase of intense line blending with no less than Fe II, which imparts a bewildering array of lines throughout the optical bands Maeda et al. 2010b; Maund et al. 2010b. Still, parameters such as \(\mathcal{R}\)(Si II) have served as useful tools for SN Ia diversity studies in that they often correlate with luminosity Hachinger et al. (2006) and are relatively accessible empirical measurements for large samples of under-observed SN Ia. A detailed study on the selection of global spectral indicators can be found in Hachinger et al. (2006).

3.6 The adjacent counterparts of optical wavelengths

3.6.1 Ultraviolet spectra

SN Ia are known as relatively “weak emitters” at UV wavelengths (< 3500 Å; Hachinger et al. (2009)). It has been shown that UV flux deficits are influenced by line-blanketing effects from IPEs within the outermost layers of ejecta Silverman et al. 2012b, overall higher expansion velocities Arsenijevic 2011; Foley et al. 2012a, progenitor metallicity Altavilla et al. (2009), viewing angle effects (e.g., Mohlabeng and Ralston 2013), or a combination of these (Hatano et al. 2000; Benetti et al. 2004; Pignata et al. 2004). Although, it is not certain which of these play the dominant role(s) in controlling UV flux behaviors among all SN Ia.

For SN 1990N and SN 1992A, two extensively studied SN Ia, pre-maximum light UV observations were made and presented by (Bongard et al. 2008) and (Baron et al. 1995, 1996), respectively. These observations revealed their expected sensitivity to the source temperature and opacity at UV wavelengths.

It was not until recently when a larger UV campaign of high S/N, multi-epoch spectroscopy of distant SN Ia was presented and compared to that of local SN Ia ((Bongard et al. 2006), see also Bailey et al. (2009)). Most notably, Panagia 2007 found a larger intrinsic dispersion of UV properties than could be accounted for by the span of effects seen in the latest models (e.g., metallicity of the progenitor, see (Sauer et al. 2008; Hachinger et al. 2013; Mazzali et al. 2013)).

As a follow-up investigation, (Foley and Kasen 2011; Wang et al. 2012) utilized and presented data from the STIS spectrograph onboard Hubble Space Telescope (HST) with the intent of studying near-UV, near-maximum light spectra (day −0.32 to day +4) of nearby SN Ia. Between a high-z and low-z sample, they find a noticeable difference between the mean UV spectrum of each, suggesting that the cause may be related to different metallicities between the statistical norm of each sample. Said another way, their UV observations suggest a plausible measure of two different populations of progenitors (or constituent scenarios) that could also be dependent on the metallically thereof, including potentially larger dependencies such as variable 56Ni mass and line blanketing due to enhanced burning within the outermost layers (Höflich et al. 1998; Lentz et al. 2000; Sim et al. 2010b). It should be noted, despite the phase selection criterion invoked by Blondin et al. 2011, it may not be enough to simply designate a phase range in order to avoid phase evolution effects (see Fig. 7 of (Moll et al. 2013)).

In order to confirm spectroscopic trends at UV wavelengths, a better method of selection will be necessary as the largest UV difference found by Leibundgut et al. (1991) and Kirshner et al. (1993) between the samples overlaps with the Si II, Ca II H&K absorption features (3600–3900 Å), i.e. a highly blended feature that is far too often a poorly understood SN Ia variable, both observationally (across subtype and phase) and theoretically, within the context of line formation at UV–NIR wavelengths Ellis et al. 2008. While it is true that different radiative processes dominate within different wavelength regions, there are a multitude of explanations for such a difference between the Si II–Ca II blend near 3700 Å. Furthermore, the STIS UV spectra do not offer a look at either the state of the 6100 Å absorption feature (is it completely photospheric?—the answer requires spectrum synthesis even for maximum light phases), nor is it clear if the same is true for Ca II in the NIR where high-velocity components thereof are most easily discernible Milne et al. 2013.

It is important to further reemphasize that the time-dependent behavior of the sum total of radiative processes that generate a spectrum from a potentially axially-asymmetric (and as of yet unknown) progenitor system and explosion mechanism are not well understood, much less easily decipherable with an only recently obtained continuous dataset for how the spectrum itself evolves over time at optical wavelengths.Footnote 16 Which is only to say, given the current lack of certain predictability between particular observational characteristics of SN Ia (e.g., spectroscopic phase transition times), time series observations at UV wavelengths would offer a beneficial route for the essential purposes of hand-selecting the ‘best’ spectrum comparisons in order to ensure a complete lack of phase evolution effects.

Recently, Höflich et al. 1998; Lentz et al. 2000 presented HST multi-epoch, UV observations of SN 2004dt, 2004ef, 2005M, and 2005cf. Based on comparisons to the results of Cooke et al. (2011) and (Marion et al. 2013), two studies that show a 0.3 magnitude span of UV flux for a change of two orders of magnitude in metallicity within the C+O layer of a pure deflagration model (W7; Cooke et al. (2011)), Childress et al. 2013b conclude that the UV excess for a HVG SN Ia, SN 2004dt (A.22), cannot be explained by metallicities or expansion velocities alone. Rather, the inclusion of asymmetry into a standard model picture of SN Ia should be a relevant part of their observed diversity (e.g., Cooke et al. (2011)).

More recently, Maguire et al. (2012) obtained 10 HST UV–NIR spectra of SN 2011fe, spanning −13.5 to +41 days relative to B-band maximum. They analyzed the data along side spectrum synthesis results from three explosion models, namely a ‘fast deflagration’ (W7), a low-energy delayed-detonation (WS15DD1; (Mazzali 2000; Kasen et al. 2003; Thomas et al. 2004; Foley 2012; Marion et al. 2013; Childress et al. 2013b)), and a third model treated as an intermediary between the outer-layer density profiles of the other two models (“W7+”). From the seen discrepancies between W7 and WS15DD1 during the early pre-maximum phase, in addition to optical flux excess for W7 and a mismatch in observed velocities for WS15DD1, (Lentz et al. 2000) conclude that their modified W7+ model is able to provide a better fit to the data because of the inclusion of a high velocity tail of low density material. In addition, and based on a spectroscopic rise time of ∼19 days, Pereira et al. 2013 infer a ∼1.4 day period of optical quiescence after the explosion (see Wang et al. (2012)).

3.6.2 Infrared light curves and spectra

By comparing absolute magnitudes at maximum of two dozen SN Ia, Lentz et al. (2000) argue that SN Ia can be best used as standard candles at NIR wavelengths (which was also suggested by Sauer et al. (2008), Nomoto et al. 1984), even without correction for optical light curve shape. Wang et al. (2012) later confirmed this to be the case from the analysis of 1087 near-IR (JHK) measurements of 21 SN Ia. Based on their data and data from the literature, they derive absolute magnitudes of 41 SN Ia in the H-band with rms scatter of 0.16 magnitudes. Kasen et al. 2009; Blondin et al. 2011 find a weak dependence of J-band luminosities on the decline rate from 9 NIR datasets, in addition to VJ corrected J-band magnitudes with a dispersion of 0.12 magnitudes. Mazzali et al. (2013) constructed a statistical model for SN Ia light curves across optical and NIR passbands and find that near-IR luminosities enable the most ideal use of SN Ia as standard candles, and are less sensitive to dust extinction as well. Iwamoto et al. 1999 analyzed the standardizability of SN Ia in the near-IR by investigating the correlation between observed peak near-IR absolute magnitude and post-maximum Δm 15(B). They confirm that there is a bimodal distribution in the near-IR absolute magnitudes of fast-declining SN Ia Mazzali et al. (2013) and suggest that applying a correction to SN Ia peak luminosities for decline rate is likely to be beneficial in the J and H bands, making SN Ia more precise distance indicators in the IR than at optical wavelengths Mazzali et al. (2013).

While optical spectra of SN Ia have received a great deal of attention in the recent past, infrared datasets (e.g., Piro and Nakar 2013; Chomiuk 2013) are either not obtained, or are not observed at the same epochs or rate as their optical counterparts. This has only recently begun to change. Thus far, the largest NIR datasets can be found in Krisciunas (2005) and Elias et al. 1985a. 1985b obtained NIR spectra (0.8–2.5 μm) of 12 normal SN Ia, with fairly early coverage. Later, Wood-Vasey et al. (2008) presented and studied a catalogue of NIR spectra (0.7–2.5 μm) of 41 additional SN Ia. In all, they report an absence of conspicuous signatures of hydrogen and helium in the spectra, and no indications of carbon via C I λ10693 (however, see our Sect. 4.1). For an extensive review on IR observations, we refer the reader to Folatelli et al. (2010).

3.7 Drawing conclusions about SN Ia diversity from SN Ia rates studies

It has long been perceived that a supernova’s local environment, rate of occurrence, and host galaxy properties (e.g., WD population) serve as powerful tools for uncovering solutions to SN Ia origins (Mandel et al. (2011); Kattner et al. (2012); (Krisciunas et al. 2009); (Barone-Nugent et al. 2012); Kirshner et al. 1973b; Meikle et al. 1996; Bowers et al. 1997; Rudy et al. 2002; Höflich et al. 2002; Marion et al. (2003), Marion et al. (2009b)). After all, a variety of systems, both standard and exotic scenarios—all unconfirmed—offer potential for explaining “oddball” SN Ia, as well as more normal events, at various distances (z; redshift) and associations with a particular host galaxy or WD population Marion et al. (2003).

Despite this broad extent of the progenitor problem, measurements of the total cosmic SN Ia rate, R SNIa (z), can be made to gauge the general underlying behavior of actively contributing systems Marion et al. (2009b). Further insight into how various progenitor populations impart their signature onto R SNIa (z) comes about by considering which scenarios lead to a “prompt” (or a “tardy”) stellar demise Phillips (2012). Whether or not mergers involve both a (“prompt”) helium-burning or (“tardy”) degenerate secondary star remains to be seen (Zwicky 1961 and references therein). Because brighter SN Ia prefer younger, metal-poor galaxies, and a linear relation exists between the SN Ia light curve shape and gas-phase metallicity, the principle finding has been that the rate of the universally prompt component is proportional to the star formation rate of the host galaxy, whereas the second delayed component’s rate is proportional to the stellar mass of the galaxy Hamuy et al. 1995. The SN Ia galaxy morphology study of van den Bergh et al. 2005 has since progressed this discussion of linking certain observed SN Ia properties with their individual environments. Overall, the trend of brighter/dimmer SN Ia found in younger/older hosts remains, however now with indications that a continuous distribution of select SN Ia subtypes exist in multiple host galaxy morphologies and projected distances within each host.

To understand the full form of R SNIa (z), taking into account the delay time distribution (DTD) for every candidate SN Ia system is necessary (see Mannucci 2005). Leaman et al. 2011 find that the DTD peaks prior to 2.2 Gyr and has a long tail out to ∼10 Gyr. They conclude that a DTD with a power-law t −1.2 starting at time t=400 Myr to a Hubble time can satisfy both constraints of observed cluster SN rates and iron-to-stellar mass ratios, implying that half to a majority of all SN Ia events occur within one Gyr of star formation (see also Li et al. 2011d).

In general, the DTD may be the result of binary mergers 2011c and/or a single-degenerate scenario (Yungelson and Livio 2000; Parthasarathy et al. 2007; Hicken et al. 2009a; Hachisu et al. 2012; Pakmor et al. 2013; Wang et al. 2013c; Pan et al. 2013; Kim et al. 2014), but with the consideration that evidence for delay times as short as 100 Myr have been inferred from SN remnants in the Magellanic Clouds (Maoz et al. 2012). From a recent comparison of low/high-z SN Ia rate measurements and DTDs of various binary population synthesis models, (Scannapieco and Bildsten 2005; Mannucci 2005) argue that single-degenerate systems are ruled out between 1.8<z<2.4. Overall, their results support the existence of a double-degenerate progenitor channel for SN Ia if the number of double-degenerate systems predicted by binary population synthesis models can be “aptly” increased Woods et al. 2011; Hillebrandt et al. 2013; Dan et al. 2013.

However, initial studies have primarily focussed upon deriving the DTD without taking into account the possible effects of stellar metallicity on the SN Ia rate in a given galaxy. Given that lower metallicity stars leave behind higher mass WD stars (Sullivan et al. 2006; Howell et al. 2007; Sullivan et al. 2010; Zhang 2011; Pan et al. 2013), Hicken et al. (2009a) and Bonaparte et al. 2013; Claeys et al. 2014 argue that the effects of metallicity may serve to significantly alter the SN Ia rate (see also Maoz et al. (2010)). In fact, models that include the effects of metallicity (e.g., Strolger et al. 2010; Meng et al. 2011) find similar consistencies with the observed R SNIa (z). Notably, recent spectroscopic studies do indicate a stronger preference of low-metallicity hosts for super-Chandrasekhar candidate SN Ia (Ruiter et al. 2009; Toonen et al. 2012; Nelemans et al. 2013), which may just as well be explained by low metallicity single-degenerate systems (Hachisu et al. 2008, 2012; Chen et al. 2013). While there are not enough close binary WD systems in our own galaxy that would result in SCC DD scenarios (Badenes et al. 2009; Maoz and Badenes 2010), sub-Chandrasekhar merging binaries may be able to account for discrepancies in the observed rate of SN Ia Graur et al. (2013).

Although, we wish to remind the reader that since spectrophotometry of SN Ia so far offer the best visual insight into these distant extragalactic events, and because there is no clear consensus on the origin of their observed spectrophotometric diversity, there is no clear certainty as to what distribution of progenitor scenarios connect with any kind of SN Ia since none have been observed prior to the explosion. Furthermore, whether or not brighter or dimmer SN Ia “tend to” correlate with any property of their hosts does not alleviate the discussion down to one or two progenitor systems (e.g., single- versus double-degenerate systems) since the most often used tool for probing SN Ia diversity over all distance scales, i.e. the “stretch” of a light curve, does not necessarily uniquely determine the spectroscopic subtype. Rather, such correlations reveal the degree of an underlying effect from samples of uncertain and unknown SN Ia subtype biases, i.e. dust extinction in star formation galaxies and progenitor ages also evolve along galaxy mass sequences (Maoz et al. 2010) and the redshift-color evolution of SN Ia remains an open issue (Umeda et al. 1999b; Timmes et al. 2003).

While it is important to consider the full redshift range over which various hierarchies of progenitor and subtype sequences may dominate over others, such studies are rarely able to incorporate spectroscopic diversity as input (a “serendipitous” counter-example being Kistler et al. (2013)). This is relevant given that the landscape of SN Ia spectroscopic diversity has not yet been seen to be void of line-of-sight discrepancies for all progenitor scenarios (particularly so for double degenerate detonations/mergers, e.g., Meng et al. (2011)). Ultimately, robust theories should be able to connect spectroscopic subtypes with individual or dual instances of particular progenitor systems, which requires detailed spectroscopic modeling.

Thus, the consensus as to how many progenitor channels contribute to SN Ia populations is still unclear. Broadly speaking, there are likely to be no less than two to three progenitor scenarios for normal SN Ia so long as single-degenerate systems remain viable Pan et al. 2013, if not restricted to explaining Ia-CSM SN alone (see Kistler et al. 2013). Given also a low observed frequency of massive white dwarfs and massive double-degenerate binaries near the critical mass limit with orbital periods short enough to merge within a Hubble time, some normal SN Ia are still perceived as originating from single-degenerate systems (Taubenberger et al. 2011; Childress et al. 2011). Meanwhile, some portion of events may also be the result of a core-degenerate merger (Hachisu et al. 2012), while some merger phenomena are possibly accelerated within triple systems (Parthasarathy et al. 2007). It likewise remains unclear whether or not some double-degenerate mergers predominately result in the production of a neutron star instead of a SN Ia ((Badenes and Maoz 2012; Kromer et al. 2013b), (Childress et al. 2013a); (Mohlabeng and Ralston 2013; Pan et al. 2013; Wang et al. 2013a)). At present, separately distinct origins for spectroscopically similar SN Ia cannot be ruled out by even one discovery of a progenitor system; the spectroscopic diversity is currently too great and too poorly understood to confirm without greater unanimity among explosion models and uniformity in data collection efforts.

4 Some recent SN Ia

During the past decade, several normal, interesting, and peculiar SN Ia have been discovered. For example, the recent SN 2009ig, 2011fe, and 2012fr are nearby SN Ia that were discovered extremely young with respect to the onset of the explosion Krughoff et al. 2011 and have been extensively studied at all wavelengths, yielding a clearer understanding of the time-dependent behavior of SN spectroscopic observations, in addition to a better context by which to compare. Below we briefly summarize some of the highlighted discoveries during the most recent decade, during which it has revealed a greater diversity of SN Ia than was previously known. In the appendix we provide a guide to the recent literature of other noteworthy SN Ia discoveries. We emphasize that these sections are not meant to replace reading the original publications, and are only summarized here as a navigation tool for the reader to investigate further.

4.1 SN 2011fe in M101

Thus far, the closest spectroscopically normal SN Ia in the past 25 years, SN 2011fe (PTF11kly), has provided a great amount of advances, including testing SN Ia distance measurement methods Shen et al. 2013; Pakmor et al. 2013; Raskin and Kasen 2013; Moll et al. 2013; Raskin et al. 2013. For example, the early spectroscopy of SN 2011fe showed a clear and certain time-evolving signature of high-velocity oxygen that varied on time scales of hours, indicating sizable overlap between C+O, Si, and Ca-rich material and newly synthesized IMEs within the outermost layers (Hachisu et al. 2012).

Han and Podsiadlowski 2006; Silverman et al. 2013d; Leloudas et al. 2013 carried out analysis of 18 spectra of SN 2011fe during its first month. Consequently, they were able to follow the evolution of C II λ6580 absorption features from near the onset of the explosion until they diminished after maximum light, providing strong evidence for overlapping regions of burned and unburned material between ejection velocities of at least 10,000 and 16,000 km s−1. At the same time, the evolution of a 7400 Å absorption feature experienced a declining Doppler-shift until 5 days post-maximum light, with O I λ7774 line velocities ranging 11,500 to 21,000 km s−1 (Parthasarathy et al. 2007). (Soker et al. 2013) concluded that incomplete burning (in addition to progenitor scenarios) is a relevant source of spectroscopic diversity among SN Ia (Thompson 2011; Dong et al. 2014).

Saio and Nomoto 1985 presented high quality spectrophotometric observations of SN 2011fe, which span from day −15 to day +97, and discussed comparisons to other observations made by 2004, Nomoto and Kondo 1991; Piersanti et al. 2003; Dan et al. 2013; Tauris et al. 2013, (Nugent et al. 2011; Foley et al. 2012c; Childress et al. 2013c), and (Matheson et al. 2012; Vinkó et al. 2012; Lee and Jang 2012). From an observed peak bolometric luminosity of 1.17±0.04×1043 erg s−1, t hey estimate SN 2011fe to have produced between ∼0.44±0.08–0.53±0.11 M of 56Ni.

By contrast, (Nugent et al. 2011) and Parrent et al. (2012) estimate 56Ni production for the normal SN 2005cf (A.26) to be ∼0.7 M. It is also interesting to note that SN 2011fe and the fast-declining SN 2004eo produced similar amount of radioactive nickel, however lower for SN 2004eo (∼0.4 M; (Nugent et al. 2011)). Parrent et al. (2012) also made comparisons between SN 2011fe, a SNFactory normal SN Ia (SNF20080514-002) and the broad-lined HV-CN SN 2009ig (Tanaka et al. 2008; Maeda et al. 2010a). Pereira et al. (2013) note similarities (sans the UV) and notable contrast with respect to high-velocity features, respectively.

Brown et al. (2012) calculated \(\dot{v}_{Si\ II}\) for SN 2011fe to be ∼60 (±3) km s−1 day−1, near the high end of low-velocity gradient SN Ia events (see Richmond and Smith (2012)). Given their high S/N, time series dataset, Vinkó et al. (2012) were also able to place tighter constraints on the velocities over which C II λ6580 is observed to be present in SN 2011fe. They conclude that C II is present down to at least as low as 8000 km s−1, which is 2000 km s−1 lower than that estimated by Munari et al. (2013), and is also ∼4000–6000 km s−1 (or more) lower than what is predicted by some past and presently favored SN Ia abundance models (e.g., W7; Pastorello et al. (2007a), and the delayed detonations of Wang et al. (2009b) and Mazzali et al. 2008).

Pereira et al. (2013) presented and discussed NIR time series spectra of SN 2011fe that span between day −15 and day +17. In particular, they report a detection of C I λ10693 on the blue-most side of a blended Mg II λ10927 absorption feature at roughly the same velocities and epochs as C II λ6580 found by (Foley et al. 2012c) and Pereira et al. (2013), which itself is blended on its blue-most side with Si II λ6355. While searches and studies of C I λ10693 are extremely useful for understanding the significance of C-rich material from normal to cooler sub-luminous SN Ia within the greater context of all C I, C II, C III, O I absorption features (C III for “hotter” SN 1991T-likes), blended C I λ10693 absorption shoulders are certainly no more (nor no less) useful for probing lower velocity boundaries than C II λ6580 absorption notches. This is especially true given that C I λ10693 absorption features are blended from the red-most side (lower velocities) by the neighboring Mg II line, which will only serve to obscure the lower velocity information of the C I profile for the non-extreme cases (e.g., SN 1999by, Pereira et al. (2013)).

Benetti et al. 2005; Blondin et al. 2012 used the observed temporal behavior, and later velocity-plateau, of Mg II λ10927 to estimate a lower extent of ∼11,200 km s−1 for carbon-burning products within SN 2011fe. Given that this in contrast to the refined lower extent of C II at ∼8000 km s−1 by Pereira et al. (2013), this could imply (i.e. assuming negligible temperature differences and/or non-LTE effects) that either some unburned material has been churned below the boundary of carbon-burning products via turbulent instabilities (Parrent et al. (2012), Nomoto et al. 1984, Höflich 2006) and/or the distribution of emitting and absorbing carbon-rich material is truly globally lopsided Röpke et al. 2012, and may indicate the remains of a degenerate secondary star Hsiao et al. (2013).

Detailed studies of this nearby, normal, and unreddenned SN 2011fe have given strong support for double-degenerate scenarios (assuming low environmental abundances of hydrogen) and have placed strong constraint on single-degenerate scenarios, i.e. MS and RG companion stars have been strongly constrained for SN 2011fe (see Parrent et al. (2011) and references therein, and also Pereira et al. (2013)). Höflich et al. 2002, Hsiao et al. (2013) and Pereira et al. (2013) confirm that the primary star was a compact star (R ≲ 0.1 R, c.f., Gamezo et al. 1999a). From the lack of evidence for an early shock outbreak 1999b, non-detections of radio and X-ray emissions 2004, non-detections of narrow Ca II H&K or Na D lines or pre-existing dust that could be associated with the event (Kasen et al. 2009; Maeda et al. 2010b; Blondin et al. 2011), and low upper-limits on hydrogen-rich gas (Moll et al. 2013), the paucity of evidence for an environment dusted in CSM from a non-degenerate secondary strongly supports the double degenerate scenario for SN 2011fe. Plus, this inferred ambient environment is consistent with that of recent merger simulations Shappee et al. 2013, and could signal an avenue of interpretation for signatures of carbon-rich material as well Hayden et al. 2010a; Bianco et al. 2011. Specifically, the remaining amount of carbon-rich material predicted by some explosion models may already be accounted for, and more so than would be required by the existence of low velocity detections of C I and C II. If this turns out to be the case, spectroscopic signatures of both C and O could tap into understanding (i) the sizes of merger C+O common envelopes, (ii) potential downward mixing effects between the envelope and the underlying ejecta, and/or (iii) test theories on possible asymmetries of C+O material within the post-explosion ejecta of the primary and secondary stars Nugent et al. (2011), which is expected to depend on the degree of coalescence Li et al. (2011a).

Of course, this all rests on the assumptions that (i) the surrounding environment of a single degenerate scenario just prior to the explosion ought to be contaminated with some amount of CSM, above which it would be detected Bloom et al. (2012), and (ii) the surrounding environment of a merger remains relatively “clean” Bloom et al. 2012; Piro and Nakar 2013; Chomiuk 2013. In this instance, and assuming similarly above that current DDT-like models roughly fit the outcome of the explosion, absorption signatures of C (+ HV O I) may point to super-massive single-degenerate progenitors with variable enclosed envelopes and/or disks of material (e.g., (Kasen 2010; Nakar and Sari 2012)) or sub-Chandrasekhar mass “peri-mergers” for resolve (see (Horesh et al. 2012; Chomiuk et al. 2012; Margutti et al. 2012) and references therein).

4.2 Other early discoveries

4.2.1 SN 2009ig in NGC 1015

(Patat et al. 2013; Johansson et al. 2013) obtained well-sampled, early UV and optical spectra of SN 2009ig as it was discovered 17 hr after the event (Lundqvist et al. 2013). SN 2009ig is found to be a normal SN Ia, rising to B-band maximum in ∼17.3 days. From the earliest spectra, (Dan et al. 2012) find Si II λ6355 line velocities around 23,000 km s−1, which is exceptionally high for such a spectroscopically normal SN Ia (see also (Branch et al. 2005; Dan et al. 2013; Moll et al. 2013)). SN 2009ig possess either an overall shallower density profile than other CN SN Ia, or a buildup of IMEs is present at high velocities.

(Livio and Pringle 2011) recently analyzed the photospheric to post-maximum light phase spectra of SN 2009ig, arguing for the presence of additional high-velocity absorption signatures from not only Si II, Ca II, but also Si III, S II and Fe II. Whether or not two separate but compositionally equal regions of line formation is a ubiquitous property of similar SN Ia remains to be seen. However, it should not be unlikely for primordial amounts of said atomic species to be present (in addition to singly-ionized silicon and calcium) on account of possible density and/or abundance enhancements within the outermost layers ((Moll et al. 2013; Raskin and Kasen 2013), (Justham 2011; Brown et al. 2012)). For example, simmering effects during convective phases prior to the explosion may be responsible for dredging up IMEs later seen as HVFs, which would give favorability to single-degenerate progenitor scenarios (see (Shen et al. 2013; Raskin and Kasen 2013)). Similarly, it is worthwhile to access the versatility of mergers in producing high-velocity features.

4.2.2 SN 2012fr in NGC 1365

Yoon and Langer 2004, 2005; Hachisu et al. 2012; Scalzo et al. 2012; Tornambé and Piersanti 2013; Dan et al. 2013 report on their time series spectroscopic observations of SN 2012fr Moll et al. 2013, complete with 65 spectra that cover between ∼15 days before and 40 days after it reached a peak B-band brightness of −19.3. In addition to the simultaneous spectropolarimetric observations of Foley et al. (2012c), the early to maximum light phase spectra of SN 2012fr reveal one of the clearest indications that SN Ia of similar type (e.g., SN 1994D, 2001el, 2009ig, 2011fe, and many others; Mazzali05a) tend to have two distinctly separate regions of Si-, Ca-based material that differ by a range of separation velocities (Kleiser et al. 2009; Navasardyan et al. 2009).

Foley et al. (2012c) and Blondin et al. 2012 discussed the various interpretations that have been presented in the past, however no firm conclusions on the origin of HVFs could be realized given the uncertainties of current explosion models. Despite this, the most recent advance toward understanding HVFs is the continual detection of polarization signatures due to the high-velocity Si II and Ca II absorption features, indicating a departure from a radially stratified, spherically symmetric geometry at some layer near or above the “photospheric region” of IMEs.

4.3 Super-Chandrasekhar candidate SN Ia

4.3.1 Over-luminous SN 2003fg (SNLS-03D3bb)

SN 2003fg was discovered as part of the Supernova Legacy Survey (SNLS); z=0.2440 Marion et al. (2013). Its peak absolute magnitude was estimated to be −19.94 in V-band, placing SN 2003fg completely outside the M V -distribution of normal low-z SN Ia (2.2 times brighter). Assuming Arnett’s rule, such a high luminosity corresponds to ∼1.3 M of 56Ni, which would be in conflict with SN 2003fg’s spectra since only ∼60 % of a Chandrasekhar pure detonation ends up as radioactive nickel (Thomas et al. 2004; Mazzali et al. 2005a, however see also 2005b). Given also the lower mean expansion velocities, this builds upon the picture of a super-Chandrasekhar mass progenitor for SN 2003fg and others like it Piro 2011; Zingale et al. 2011.

Childress et al. (2013c) proposed the formation of super-Chandrasekhar mass WD stars as a result of rapid rotation. (Klotz et al. 2012; Childress et al. 2012; Buil 2012) later reworked these models and found that the “prompt” detonation of a super-Chandrasekhar mass WD produces enough nickel, as well as a remainder of IMEs in the outer layers (in contrast to Maund et al. (2013)), to explain over-luminous SN Ia.

(Childress et al. 2013b) added to this model by taking into account processes of binary evolution. Namely, with the inclusion of mass-striping, optically thick winds of a differentially rotating primary star, Childress et al. (2013c) find three critical mass ranges that are each separated according to the spin-down time of the accreting WD. All three of these single-degenerate scenarios may explain a majority of events from sub-luminous to over-luminous SN Ia. So far no super-Chandrasekhar mass WD stars that would result in a SN Ia have been found in the sample of known WD stars in our Galaxy (Maund et al. (2013), see also (Howell et al. 2006)). However, this does not so much rule out super-Chandrasekhar mass models as it suggests that these systems are rare in the immediate vicinity within our own galaxy.

Steinmetz et al. 1992 proposed an alternative scenario involving only a Chandrasekhar-mass WD progenitor to explain the SN 2003fg event. They demonstrated that an off-center explosion of a Chandrasekhar-mass WD could explain the super-bright SN Ia. However, in this off-center explosion model it is not easy to account for the high Ni mass in the outer layers, in addition to the special viewing direction.

4.3.2 Over-luminous SN 2009dc in UGC 10064

Pfannes et al. 2010 presented early phase optical and NIR observations for SN 2009dc (Howell et al. 2006; Jeffery et al. 2006). From the peak V-band absolute magnitude they conclude that SN 2009dc belongs to the most luminous class of SN Ia (Δm 15(B)=0.65), and estimate the 56Ni mass to be 1.2 to 1.6 M. Based on the JHK photometry Yoon and Langer (2005) also find SN 2009dc had an unusually high NIR luminosity with enhanced fading after ∼ day +200 Pfannes et al. (2010). The spectra of SN 2009dc also show strong, long lasting 6300 Å absorption features (until ∼ two weeks post-maximum light) Based on the observed spectropolarimetric indicators, in combination with photometric and spectroscopic properties, Steinmetz et al. 1992 similarly conclude that the progenitor mass of SN 2009dc was of super-Chandrasekhar origin and that the explosion geometry was globally spherically symmetric, with a clumpy distribution of IMEs.

Hachisu et al. (2012) presented an analysis of 14 months of observations of SN 2009dc and estimate a rise-time of ∼23 days and Δm 15(B)=0.72. They find a lower limit of the peak bolometric luminosity ∼2.4×1043 erg s−1 and caution that the actual value is likely almost 40 % larger. Based on the high luminosity and low mean expansion velocities of SN 2009dc, Hachisu et al. (2012) derive a mass of more than 2 M for the white dwarf progenitor and a 56Ni mass of ∼1.4 to 1.7 M. Saffer et al. 1998 find the minimum 56Ni mass to be 1.8 M, assuming the smallest possible rise-time of 22 days, and the ejecta mass to be 2.8 M.

Kilic et al. 2012 compared photometric and spectroscopic observations of normal and SCC SN Ia at late epochs, including SN 2009dc, and find a large diversity of properties spanning through normal, SS, and SCC SN Ia. In particular the decline in the light curve “radioactive tail” for SCC SN Ia is larger than normal, along with weaker than normal [Fe III] emission in the nebular phase spectra. Hillebrandt et al. (2007) argue that the weak [Fe III] emission is indicative of an ejecta environment with higher than normal densities. Previously, Yamanaka et al. (2009a) carried out spectroscopic modeling for SN 2009dc and discussed the model alternatives, such as a 2 M rotating WD, a core-collapse SN, and a CSM interaction scenario. Overall, (Puckett et al. 2009; Harutyunyan et al. 2009; Marion et al. 2009a; Nicolas and Prosperi 2009) found the interaction scenario to be the most promising in that it does not require the progenitor to be super-massive. Yamanaka et al. (2009a) furthered this discussion in conjunction with their late time comparisons and conclude that the models of (Maeda et al. 2009; Silverman et al. 2011; Taubenberger et al. 2011) do not simultaneously match the peak brightness and decline of SN 2009dc (see also Tanaka et al. (2010)). Following the interaction scenario of Silverman et al. (2011), Silverman et al. (2011) propose a non-violent merger model that produces ∼1 M of 56Ni and is enshrouded by ∼0.6–0.7 M of C+O-rich material. In order to reconcile the low 56Ni production, Taubenberger et al. (2011) note that additional luminosity from interaction with CSM is required during the first two months post-explosion. Further support for CSM interaction comes from the observed suppression of the double peak in the I-band, which is thought to arise from a breaking of ejecta stratification in the outermost layers Taubenberger et al. (2013).

It is not yet clear if SN 2003fg, 2006gz (A.32), 2007if (A.34), and SN 2009dc are the result of a single super-Chandrasekhar mass WD star, given that even in our galaxy there is no observational evidence for the existence of such a system. Likewise, there is no direct observational evidence for the presence of very rapidly rotating massive WD stars, either single WDs or in binary systems as well. In fact, no double-degenerate close binary systems with a total mass amounting to super-Chandrasekhar mass configurations that can merge in Hubble-time have been found Taubenberger et al. (2013). Therefore, our current understanding of the origin of over-luminous SN Ia is limited, and more observations are needed. For example, progress has been made with the recent discovery of 24 merging WD systems via the extremely low mass Survey (see Hachinger et al. (2012) and references therein), however it is unclear if any are systems that would produce a normal SN Ia.

4.4 The peculiar SN 2002cx-like class of SN

4.4.1 SN 2002cx in CGCG-044-035

Hachinger et al. (2012) considered SN 2002cx as “the most peculiar known SN Ia” Taubenberger et al. (2013). They obtained photometric and spectroscopic observations which revealed it to be unique among all observed SN Ia. Hachinger et al. (2012) described SN 2002cx as having SN 1991T-like pre-maximum spectrum, a SN 1991bg-like luminosity, and expansion velocities roughly half those of normal SN Ia.

Photometrically, SN 2002cx has a broad peak in the R-band, a plateau phase in the I-band, and a slow late time decline. The BV color evolution are described as nearly normal, while the VR and VI colors are redder than normal. Spectra of SN 2002cx during early phases evolve rapidly and are dominated by lines from IMEs and IPEs, but the features are weak overall. In addition, emission lines are present around 7000 Å during post-maximum light phases, while the late time nebular spectrum shows narrow lines of iron and cobalt.

Yamanaka et al. 2013 presented late time spectroscopy of SN 2002cx, which includes spectra at 227 and 277 days post-maximum light. They considered it as a prototype of a new subclass of SN Ia. The spectra do not appear to be dominated by the forbidden emission lines of iron, which is not expected during the “nebular phase,” where instead they find a number of P Cygni profiles of Fe II at exceptionally low expansion velocities of ∼700 km s−1 Hachinger et al. (2012). A tentative identification of O I λ7774 is also reported for SN 2002cx, suggesting the presence of oxygen-rich material. Currently, it is difficult to explain all the observed photometric and spectroscopic properties of SN 2002cx using the standard SN Ia models (see Taubenberger et al. (2013)). However, the spectral characteristics of SN 2002cx support pure deflagration or failed-detonation models that leave behind a bound remnant instead of delayed detonations Taubenberger et al. (2013).

4.4.2 SN 2005hk in UGC 00272

(Kasen 2006; Kamiya et al. 2012) presented extensive multi-color photometry and optical spectroscopy of SN 2005hk (Parthasarathy et al. 2007). Kilic et al. 2012 also studied the spectrophotometric evolution SN 2005hk, covering pre-maximum phase to around 400 days after the event. These datasets reveal that SN 2005hk is nearly identical in its observed properties to SN 2002cx. Both supernovae exhibited high ionization SN 1991T-like pre-maximum light spectra but with low peak luminosities like that of SN 1991bg. The spectra reveal that SN 2005hk, like SN 2002cx, has expansion velocities that are roughly half those of typical SN Ia.

The R- and I-band light curves of both supernovae are also peculiar for not displaying the secondary maximum observed for normal SN Ia. Li et al. (2003) constructed a bolometric light curve from 15 days before to 60 days after B-band maximum. They conclude that the shape and exceptionally low peak luminosity of the bolometric light curve, low expansion velocities, and absence of a secondary maximum in the NIR light curves are in reasonable agreement with model calculations of a three-dimensional deflagration that produces 0.25 M of 56Ni. Note however that the low amount of continuum polarization observed for SN 2005hk (∼0.2 %–0.4 %) is far too similar to that of more normal SN Ia to serve as an explanation for the spectroscopic peculiarity of SN 2005hk, and possibly other SN 2002cx-like events (Wood-Vasey et al. 2002b).

4.4.3 Sub-luminous SN 2007qd

Li et al. (2003) obtained multi-band photometry and multi-epoch spectroscopy of SN 2007qd Jha et al. (2006a). Its observed properties place it broadly between those of the peculiar SN 2002cx and SN 2008ha (A.37). Optical photometry indicate a fast rise-time and a peak absolute B-band magnitude of −15.4. (Branch et al. 2004) carried out spectroscopy of SN 2007qd near maximum brightness and detect signatures of IMEs. They find the photospheric velocity to be 2800 km s−1 near maximum light, and note that this is ∼4000 and 7000 km s−1 less than that inferred for SN 2002cx and normal SN Ia, respectively. Foley et al. 2013 find that the peak luminosities of SN 2002cx-like objects are well correlated with their light curve stretch and photospheric velocities.

4.4.4 SN 2009ku

SN 2009ku was discovered by Pan-STARS-1 as a SN Ia belonging to the peculiar SN 2002cx class. (Jordan et al. 2012; Kromer et al. 2013a; Hillebrandt et al. 2013) studied SN 2009ku and find that while its multi-band light curves are similar to that of SN 2002cx, they are slightly broader and have a later rise to g-band maximum. Its peak brightness was found to be M V =−18.4 and the ejecta velocity at 18 days after maximum brightness was found to be ∼2000 km s−1. Spectroscopically, SN 2009ku is similar to SN 2008ha (A.37). Phillips et al. (2007) note that the high luminosity and low ejecta velocity for SN 2009ku is not in agreement with the trend seen for SN 2002cx class of SN Ia. The spectroscopic and photometric characteristics of SN 2009ku indicate that the SN 2002cx class of SN Ia are not homogeneous, and that the SN 2002cx class of events may have a significant dispersion in their progenitor population and/or explosion physics (see also (Quimby et al. 2005) for differences between this class and sub-luminous “calcium-rich” transients).

4.5 PTF11kx and the “Ia-CSM” Class of SN Ia

4.5.1 PTF11kx: a case for single-degenerate scenarios?

Sahu et al. (2008) studied the photometric and spectroscopic properties of another unique SN Ia event, PTF11kx. Using time series, high-resolution optical spectra, they find direct evidence supporting a single-degenerate progenitor system based on several narrow, temporal (∼65 km s−1) spectroscopic features of the hydrogen Balmer series, He I, Na I, Ti II, and Fe II. In addition, and for the first time, PTF11kx observations reveal strong, narrow, highly time-dependent Ca II absorption features that change from saturated absorption signatures to emission lines within ∼40 days.

Phillips et al. (2007) considered the details of these observations and concluded that the complex CSM environment that enshrouds PTF11kx is strongly indicative of mass loss or “outflows,” prior to the onset of the explosion of the progenitor system. Other SN Ia have exhibited narrow, temporal Na D lines before (e.g., SN 2006X, 2007le; see (Chornock et al. 2006; Maund et al. 2010a)), but none have been reported as having signatures of these particular ions, which are clearly present in the high-resolution spectra of PTF11kx. On the whole, and during the earliest epochs, McClelland et al. (2010) find that the underlying SN Ia spectroscopic component of PTF11kx most closely resembles that of SN 1991T (Bassett et al. 2007) and 1999aa McClelland et al. (2010).

As for the late time phases, McClelland et al. (2010) studied spectroscopic observations of PTF11kx from 124 to 680 days post-maximum light and find that its nebular phase spectra are markedly different from those of normal SN Ia. Specifically, the late time spectra of PTF11kx are void of the strong cobalt and iron emission features typically seen in other SN 1991T/1999aa-like and normal SN Ia events (e.g., Narayan et al. (2011)). For the most part, the late time spectra of PTF11kx are seen to be dominated by broad (FWHM ∼2000 km s−1) Hα emission and strong Ca II emission features that are superimposed onto a relatively blue, overly luminous continuum level that may be serving to wash out the underlying SN Ia spectroscopic information. Narayan et al. (2011) note that the Hα emission increases in strength for ∼1 yr before decreasing. In addition, from the absence of strong Hβ, He I, and O I emission, as well as a larger than normal late time luminosity, Kasliwal et al. 2012 conclude that PTF11kx indeed interacted with some form of CSM material; possibly of multiply thin shells, shocked into radiative modes of collisional excitation as the SN ejecta overtakes the slower-moving CSM. However, it should be noted that it is not yet clear if the CSM originates from a single-degenerate scenario or a H-rich layer of material that is ejected prior to a double-degenerate merger event (Dilday et al. (2012), see also Dilday et al. (2012)).

4.5.2 SN 2002ic

Simon et al. 2009; Patat et al. 2009, 2010, 2011; Sternberg et al. 2011 detected a large Hα emission in the spectra of SN 2002ic Dilday et al. (2012). Seven days before to 48 days after maximum light, the optical spectra of SN 2002ic exhibit normal SN Ia spectral features in addition to the strong Hα emission. The Hα emission line in the spectrum of SN 2002ic consists of a narrow component atop a broad component (FWHM of about 1800 km s−1). (Filippenko et al. 1992a; Gómez and López 1998) argue that the broad component arose from ejecta—CSM interaction. By day +48, they find that the spectrum is similar to that of SN 1990N. (Garavini et al. 2004) argue that the progenitor system contained a massive AGB star, associated with a few solar masses of hydrogen-rich CSM.

Silverman et al. (2013b) obtained the first high resolution, high S/N spectrum of SN 2002ic. The resolved Hα line has a P Cygni-type profile, indicating the presence of a dense, slow-moving outflow (about 100 km s−1). They also find a relatively large and unusual NIR excess and argue that this is the result of an infrared light-echo originating from the presence of CSM. They estimate the mass of CSM to be more than 0.3 M, produced by a progenitor mass loss rate greater than 10−4 M yr−1. For the progenitor, Ruiz-Lapuente and Lucy 1992; Salvo et al. 2001; Branch et al. 2003; Stehle et al. 2005; Kotak et al. 2005; McClelland et al. 2013; Silverman et al. 2013a favor a single-degenerate system with a post-AGB companion star.

Silverman et al. (2013b) obtained pre-maximum and late time photometry of SN 2002ic and find that a non-SN Ia component of the light curve becomes pronounced about 20 days post-explosion. They suggest the non-SN Ia component to be due to heating from a shock interaction between SN ejecta and CSM. Silverman et al. (2013b) also suggest that the progenitor system consisted of a WD and an AGB star in the protoplanetary nebula phase. Shen et al. 2013 and Soker et al. 2013 proposed that a nova shell ejected from a recurrent nova progenitor system, creating the evacuated region around the explosion center of SN 2002ic. They suggest that the periodic shell ejections due to nova explosions on a WD sweep up the slow wind from the binary companion, creating density variations and instabilities that lead to structure in the circumstellar medium. This type of phenomenon may occur in SN Ia with recurrent nova progenitors, however Hamuy et al. (2003) recently reported on an ongoing observational campaign of recurrent novae (RN) orbital period changes between eruptions. For at least two objects (CI Aquilae and U Scorpii), he finds that the RN lose mass, thus making RN unlikely progenitors for SN Ia.

Nearly one year after the explosion, (Wood-Vasey et al. 2002a) found that the supernova had become fainter overall, but Hα emission had brightened and broadened compared to earlier observations. From their spectropolarimetry observations, Hamuy et al. (2003) find that hydrogen-rich matter is asymmetrically distributed. Likewise, Hamuy et al. (2003) also found evidence of a hydrogen-rich asymmetric circumstellar medium. From their observations of SN 2002ic, Kotak et al. (2004) conclude that the event took place within a “dense, clumpy, disk-like” circumstellar medium. They suggest that the star responsible for SN 2002ic could either be a post-AGB star or WD companion (see also Kotak et al. (2004)).

4.5.3 SN 2005gj

Similar to SN 2002ic, Wood-Vasey et al. (2004) argue that SN 2005gj is a SN Ia in a massive circumstellar envelope (see also Wood-Vasey et al. (2004)), which is located in a low metallicity host galaxy with a significant amount of star formation. Their first spectrum of SN 2005gj shows a blue continuum level with broad and narrow Hα emission. Subsequent spectra reveal muted SN Ia features combined with broad and narrow Hγ, Hβ, Hα and He I λλ5876, 7065 in emission, where high resolution spectra reveal narrow P Cygni profiles. An inverted P Cygni profile for [O III] λ5007 was also detected, indicating top-lighting effects from CSM interaction Wood-Vasey et al. (2004). From their early photometry of SN 2005gj, Sokoloski et al. (2006) find that the interaction between the supernova ejecta and CSM was much weaker for SN 2002ic. Notably, both Schaefer (2011) and Wang et al. (2004) agree that a SN 1991T-like spectrum can account for many of the observed profiles with an assumed increase in continuum radiation from interaction with the hydrogen-rich material.

Wang et al. (2004) also find that the light curve and measured velocity of the unshocked CSM imply mass loss as recent as 1998. This is in contrast to SN 2002ic, for which an inner cavity in the circumstellar matter was inferred Deng et al. (2004). Furthermore, SN 1997cy, SN 2002ic, and SN 2005gj all exhibit large CSM interactions and are from low-luminosity hosts.

Consistent with this interpretation for CSM interactions is the recent report by Wang et al. (2004) that a NIR re-brightening, possibly due to emission from “warm” dust, took place at late times for both SN 2002ic and 2005gj. Notably, and in contrast to SN 2002ic, Hachisu et al. 1999; Han and Podsiadlowski 2006 find that the mid-IR luminosity of SN 2005gj increased to ∼ twice its early epoch brightness.

5 Summary and concluding remarks

Observations of a significant number of SN Ia during the last two decades have enabled us to document a larger expanse of their physical properties which is manifested through spectrophotometric diversity. While in general SN Ia have long been considered a homogeneous class, they do exhibit up to 3.5 mag variations in the peak luminosity, whereas “normal” SN Ia dispersions are ∼1 mag, and constitute several marginally distinct subtypes Aldering et al. (2006). Consequently, the use of normal SN Ia for cosmological purposes depends on empirical calibration methods (e.g., Prieto et al. 2007), where one of the most physically relevant methods is the use of the width-luminosity relation (Branch et al. 2000).

Understanding the physics and origin of the width-luminosity-relationship of SN Ia light curves is an important aspect in the modeling of SN Ia Aldering et al. (2006). Brighter SN Ia often have broad light curves that decline slowly after peak brightness. Slightly less bright or dimmer SN Ia have narrower and relatively rapidly declining light curves. In addition, several SN Ia do not follow the width-luminosity-relationship (e.g., SN 2001ay, 2004dt, 2010jn, SCC, CL and SN 2002cx-like SN Ia), which reinforces the notion that a significant number of physically relevant factors influence the diversity of SN Ia overall (see Aldering et al. (2006)).

Despite the ever increasing number of caught-early supernovae, our perspective on their general properties and individual peculiarities undergoes a continual convergence toward a set of predictive standards with which models must be seen to comply. The most recent observational example is that of SN 2012fr Prieto et al. (2007), a normal/low-velocity-gradient SN Ia that has been added to the growing list of similar SN Ia that exhibit stark evidence for a distinctly separate region of “high-velocity” material (>16,000 km s−1). While the origin of high velocity features in the spectra of SN Ia is not well understood, it is concurrent with polarization signatures in most cases which implies some amount of ejecta density asymmetries (e.g., Aldering et al. (2006)). Furthermore, since understanding the temporal behavior of high velocity Si II/Ca II depends on knowing the same for the photospheric component, studies that focus on velocity gradients and potential velocity-plateaus of the photospheric component could make clearer the significance of the physical separation between these two regions of material (see (Wood-Vasey et al. 2004)). However, it is at least certain that all viable models that encompass “normal” SN Ia conditions must account for the range of properties related to velocity evolution (see Fox and Filippenko (2013) and references therein), the occasionally observed however potentially under-detected signatures of C+O material at both low and high velocities (Fox and Filippenko (2013), (Blondin et al. 2012; Scalzo et al. 2012; Silverman et al. 2013d; Foley et al. 2013; Dessart et al. 2013a); Bailey et al. 2009), a high-velocity region of either clumps or an amorphous plumage of opaque Si-, Ca-based material (Phillips 1993; Phillips et al. 1999), and the supposed blue/red-shift of nebular lines emitted from the inner IPE-rich material (Khokhlov et al. 1993; Lentz et al. 2000; Timmes et al. 2003; Nomoto et al. 2003; Kasen and Woosley 2007; Kasen et al. 2009; Meng et al. 2011; Blondin et al. 2011).

For at least normal SN Ia, there remain two viable explosion channels (with a few sub- and super-M Ch sub-channels) regardless of the hierarchical dominance of each at various redshifts and/or ages of galactic constituents (c.f., Wang et al. 2012; Baron et al. 2012). Also, it may or may not be the case that some SN Ia are 2+ subtypes viewed upon from various lines of sight amidst variable CSM interaction (Maund et al. 2013; Childress et al. 2013c). However, with the current lack of complete observational coverage in wavelength, time, and mode (i.e. spectrophotometric and spectropolarimetric observations) for all SN Ia subtypes and “well-observed” events, there is a limit for how much constraint can be placed on many of the proposed explosion models and progenitor scenarios. That is, despite observational indications for and theoretical consistencies with the supposition of multiple progenitor channels, the observed diversity of SN Ia does not yet necessitate that each spectrophotometric subtype be from a distinctly separate explosive binary scenario than that of others within the SN Ia family of observed events; particularly so for normal SN Ia.

For the purposes of testing the multifaceted predictions of theoretical explosion models, time series spectroscopic observations of SN Ia serve to visualize the post-explosion material of an unknown progenitor system. For example during the summer of 2011 astronomers bore witness to SN 2011fe, the best observed normal type Ia supernova of the modern era. The prompt discovery and follow-up of this nearby event uniquely allowed for a more complete record of observed properties than all previous well-observed events. More specifically, the full range (in wavelength and time) of rapid spectroscopic changes was documented with continual day-to-day follow-up into the object’s post-maximum light phases and well beyond. However, the observational side of visualizing other SN Ia remains inefficient without the logistical coordination of many telescope networks (e.g., LCOGT; Kasen et al. 2003; Wang and Wheeler 2008; Smith et al. 2011; Maund et al. 2013), telescopes large enough to make nearly all SN “nearby” in terms of improved signal-to-noise ratios (e.g., The Thirty Meter Telescope, The Giant Magellan Telescope), or a space-based facility dedicated to the study of such time sensitive UV–optical–NIR transients.

Existing SN Ia surveys are currently acting toward optimizing a steady flow of discoveries, while other programs have produced a significant number of publicly available spectra Patat et al. 1996; Kasen et al. 2003; Tanaka et al. 2008; Foley et al. 2012c; Parrent et al. 2012; Scalzo et al. 2012; Childress et al. 2013c; Marion et al. 2013. However, for the longterm future we believe it is imperative to begin a discussion of a larger (digital) network of international collaboration by way of (data-) cooperative competition like that done for both The Large Hadron Collider Experiment and Fermi Lab’s Tevatron, with multiple competing experiments centered about mutual goals and mutual resources. Otherwise we feel the simultaneous collection of even very high quality temporal datasets by multiple groups will continue to create an inefficient pursuit of over-observing the most high profile event(s) of the year with a less than complete dataset.

Such observational pursuits require an increasingly focused effort toward observing bright and nearby events. For example, 206 supernovae were reported in 1999 and 67 were brighter than 18th magnitude while only three reached ∼13 magnitude.Footnote 17 By 2012 the number of found supernovae increased to 1045 while 78 were brighter than 16th magnitude and five brighter than 13th magnitude. This clearly indicates that supernovae caught early are more prevalent than ∼15 years ago and it is worthwhile for multiple groups to continually increase collaborative efforts for the brightest events. Essentially this could be accomplished without interfering with spectrum-limited high[er]-z surveys by considering a distance threshold (≲10–30 Mpc) as part of the public domain. Additionally, surveys that corroborate the immediate release of discoveries would further increase the number of well-observed events and could be supplemented and sustained with staggered observations given that there are two celestial hemispheres, unpredictable weather patterns, and caught-early opportunities nearly every week during active surveying.

In conclusion, to extract details of the spectroscopic behavior for all SN Ia subtypes, during all phases, larger samples of well-observed events are essential, beginning from as close to the onset of the explosion as possible (e.g., SN 1999ac, 2009ig, 2011fe, 2012cg, 2012fr), where SN Ia homogeneity diverges the most (see Blondin et al. 2012 for the most recent instance in SN 2013dy). Near-continuous temporal observations are most important for at least the first 1–2 months post-explosion and biweekly to monthly follow-up thereafter for ∼1 year. SN Ia spectra are far too complicated to do so otherwise. Even normal SN Ia deserve UV–optical–IR spectroscopic follow-up at a 1:1 to 2:1 ratio between days passed and spectrum taken, whenever possible, given that fine differences between normal SN Ia detail the variance in explosion mechanism parameters and initial conditions of their unobserved progenitor systems. It is through such observing campaigns that the true diversity to the underlying nature of SN Ia events will be better understood.