CORE-COLLAPSE SUPERNOVAE AND HOST GALAXY STELLAR POPULATIONS

We consider images taken during the period from 3 months before to 12 months after discovery to be potentially contaminated by SN light. Arcavi et al. (2010) do not report the exact discovery dates of PTF SN, so, for these SN, the contamination window begins 3 months before the start and ends 12 months after the completion of the search period.

To assemble SDSS frames of each SN host, we first queried the SDSS SkyServer for any frames within 975 of the host galaxy center. If none was available without possible contamination from SN light (even with partial coverage not including from the field center), we instead assembled and constructed mosaics from potentially contaminated images. Such mosaics were used only to measure the deprojected offsets of SDSS spectroscopic fibers and the SN site in each host galaxy. The header of each SDSS image provides keywords that define a tangent plane projection (TAN) which maps coordinates on the sky to pixel coordinates. We discarded the small number of images retrieved from the SDSS server that lacked the keywords.

Our previous work has shown that SN Ic are more strongly associated with bright regions in their host galaxies' g'-band light than are SN Ib (Kelly et al. 2008), indicating that they have a distinct progenitor population, so we group them separately in this analysis. SN IIb and SN IIn subtypes are excluded from the "SN II" sample. A single SN with an associated long duration gamma-ray burst (LGRB), SN 2006aj, meets the sample criteria, but we consider it separately from SN Ic-BL discovered through their optical emission (which have no associated LGRB), as did Modjaz et al. (2008). Today's gamma-ray searches are not sensitive to normal SN explosions.

From a comprehensive set of spectra, we update the classification of SN 2005az. This SN was discovered approximately 17 days before maximum and spectroscopically classified 3 days after discovery as an SN Ic by Quimby et al. (2005a). The Nearby Supernova Factory (SNF), from a spectrum taken five days after discovery, suggested it was a Type Ib (Aldering et al. 2005). Spectral cross correlation using the Supernova Identification code (SNID; Blondin & Tonry 2007), applied to 24 spectra taken by the CfA Supernova Group from approximately 10 days before to 25 days after maximum, shows that it was a Type Ic explosion.

We update the spectroscopic types of 10 SN found in the Asiago catalog with reclassifications from CfA spectra (M. Modjaz et al. 2012, in preparation) using SNID. We also use new spectroscopic classifications for two SN from Sanders et al. (2012), based on revisions by the authors of the original IAU circulars. The SN in our sample that have new classifications from these papers have footnotes in Tables 7 and 8.

We exclude SN 2006jc, a peculiar SN Ib with narrow helium emission lines and an underlying broad-lined SN Ic spectrum (e.g., Foley et al. 2007; Pastorello et al. 2007), from our SN Ib statistical sample. The helium emission may reflect the collision of ejecta with a helium-rich circumstellar medium.

We group calcium-rich SN 2000ds, SN 2003dg, SN 2003dr, and SN 2005E separately from other SN (Ib+Ic) because of their potentially distinct progenitor population (Perets et al. 2010).

We exclude SN IIn imposters (e.g., Van Dyk et al. 2000; Maund et al. 2006), a group which includes SN 1997bs, SN 1999bw, SN 2000ch, and SN 2001ac.

While spectra taken over several epochs are necessary to observe the spectroscopic transition that defines SN IIb, such follow-up is not always available. Fortunately, the spectra of Type IIb SN similar to SN 1993J are sufficiently distinctive that cross correlation with spectroscopic templates (e.g., SNID) has been able to identify substantial numbers of explosions as Type IIb from a single spectrum. Although classifications based on a single spectrum may overlook examples of SN IIb, the Type IIb explosions they do identify should be reliable.

The Asiago catalog entries sometimes have less information than the IAU Circulars and published papers. For example, SN 1997dq and SN 1997ef were listed in the Asiago catalog (as of 2010 November) as "Type Ib/c" while Matheson et al. (2001) and Mazzali et al. (2004) identified them as SN Ic-BL. Motivated by these examples, we searched the circulars to see whether additional information was available. Despite making note of the presence or absence of helium more than 10 days after the explosion, some authors report a Type Ib/c classification. Authors may feel that an SN Ib/c classification was sufficiently precise while, in other cases, they may have wanted to emphasize conflicting spectroscopic characteristics. An example of the latter is SN 2003A which was classified as a Type Ib/c by Filippenko & Chornock (2003) who noted that "[w]eak He i absorption lines are visible, but the overall spectrum resembles that of Type Ic SN."

Classifications by the SNF (Aldering et al. 2002; Wood-Vasey et al. 2004) reported in circulars include an unusually high percentage of Type Ib/c. The high fraction of SN Ib/c reported by the SNF survey is hard for us to assess without being able to see the spectra or use impartial classification techniques. We have therefore excluded SN discovered by the SNF from our statistical sample.

We measure the host galaxy properties of SN discovered by both targeted surveys, which aristocratically discover almost all their SN in luminous galaxies, as well as galaxy-impartial surveys, which democratically scan swaths of sky without special attention to specific galaxies. Galaxy-impartial surveys generally employ larger telescopes (e.g., the SDSS 2.5 m and the PTF 1.2 m) than targeted surveys (e.g., the KAIT 0.76 m), have fainter limiting magnitudes, and image much greater numbers of low-mass galaxies. The SN harvested by galaxy-impartial surveys are found in host galaxies that are not apparently bright or nearby (and are not in bright galaxy catalogs). For example, in our sample, 34% (45/133) of galaxy-impartial SN but only 3.4% (13/186) of targeted SN have host galaxy masses smaller than 10⁹ M_☉.

We used the Discoverer column from the IAU classification⁷ to determine the provenance of each SN. There are relatively few galaxy-impartial discovery teams, because discovering substantial numbers of SN by impartially scanning the sky requires significant dedicated observing time and investment in data processing. Any SN whose discovery team we did not identify as part of a galaxy-impartial search effort, including amateur discoveries, was considered a targeted discovery.

Surveys that we considered galaxy-impartial are as follows: Catalina Real-Time Sky Survey and Siding Spring Survey (Djorgovski et al. 2011), La Sagra Sky Survey, PAN-STARRS (Kaiser et al. 2010), PTF (Law et al. 2009), ROTSE (Yost et al. 2006), ESSENCE (Miknaitis et al. 2007), Palomar-Quest (Djorgovski et al. 2008), SDSS-II (Sako et al. 2005), Supernova Legacy Survey (Astier et al. 2006), Supernova Cosmology Project (Perlmutter et al. 1999), Near Earth Asteroid Tracking Program (Pravdo et al. 1999), High-z Supernova Search (Riess et al. 1998), Experience de Recherche dObjets Sombres (Hardin et al. 2000), Great Observatories Origins Deep Survey (Dickinson et al. 2003), Deep Lens Survey (Wittman et al. 2002), and, except for discoveries in targets IC342, M33, M74, M81, NGC 6984, and NGC 7331, the Texas Supernova Search (Quimby et al. 2005b).

The control time t is the total time period during which an SN with a specific light curve, luminosity, and extinction by dust along the line of sight would have been detected by a survey's observations of a given galaxy (e.g., Cappellaro et al. 1999; Leaman et al. 2011). Extinction along the line of sight to potential explosion sites in each monitored galaxy as well as to the actual sites of SN explosions, however, is challenging to estimate. Therefore, control times are generally estimated for apparent luminosity functions and magnitudes uncorrected for extinction, instead of for dust-free luminosity functions and magnitudes with a galaxy-by-galaxy correction for obscuring dust. The expectation value of the number of discoveries of type T SN during a survey in the ith monitored galaxy is

$$\begin{equation} \big\langle N^T_i \big\rangle = r_i^T \times t_i^T, \end{equation} \tag{ 1 }$$

where r^T_i is the rate of and t^T_i is the control time for type T SN in the galaxy. The probability of detecting N^T_i type T SN in the ith galaxy imaged by a survey follows a Poisson distribution,

$$\begin{equation} P\big(N^T_i\big) = {\rm Pois}\big(r_i^T \times t_i^T\big). \end{equation} \tag{ 2 }$$

The ith galaxy observed by the survey has properties that include, for example, the stellar mass M_i. We are interested in making inferences about how the rates of the SN types may depend upon galaxy properties. The statistical approach we take in this paper is to look for differences among the distributions of host properties of the reported SN of each type S^T(M_i) (see Section 6),

$$\begin{equation} S^T(M_i) \approx {\rm Pois}\big(r_i^T \times t_i^T\big). \end{equation} \tag{ 3 }$$

Here, we cannot estimate control times because we lack information about the contributing surveys. We do not explicitly model the effect of different possible control times as well as other potential effects in our statistical analysis.

In the following sections, however, we find evidence that control times may be sufficiently similar that any differences do not dominate the results we find. We consider the published luminosity functions, place appropriate redshift upper limits on samples, and examine the redshift distributions of galaxy-impartial SN discoveries. Only differences among control times for SN species that correlate with galaxy properties will introduce bias into our statistical analysis.

The number of galaxies monitored by galaxy-impartial surveys grows with the volume within the limiting redshift (i.e., ∝z³). The number of galaxies monitored by targeted surveys, by contrast, likely does not increase as quickly with redshift, because generally the high-mass targets are selected from galaxy catalogs that have, for example, Malmquist bias.

Recent measurements have compared the mean luminosities of the core-collapse spectroscopic species, but any differences are not yet well constrained. Li et al. (2011b) (LOSS) and Drout et al. (2011) (Palomar 60'') measured the mean peak absolute magnitudes (before correction for host galaxy extinction) of core-collapse species, and these values are shown in Table 3. Drout et al. (2011) found that SN Ic-BL are intrinsically brighter explosions, on average, than normal Type Ic explosions, but the LOSS sample included too few examples to corroborate a difference. Although Li et al. (2011b) found some evidence that SN Ib and SN Ic have different mean intrinsic luminosities, Drout et al. (2011) did not find a similar indication.

Luminosity functions are broad, so the control time will not necessarily vary strongly with the mean SN luminosity. Li et al. (2011b) found, for example, that the SN (Ib+Ic) luminosity function has a standard deviation of 1.24 mag and that the SN II luminosity function has a standard deviation of 1.37 mag.

We exclude SN discovered at redshifts where only SN Ib, SN Ic, and SN II with brighter-than-average luminosities are detected. Targeted surveys generally employ smaller telescopes and have shallower limiting magnitudes, because their galaxy targets are nearby. We select LOSS and SDSS as the representative targeted and galaxy-impartial surveys, respectively, because they are responsible for the greatest numbers of discoveries in each category.

From luminosity functions and search limiting magnitudes, we estimate these upper redshift limits where detection efficiency falls below 50% as z = 0.023 and z = 0.08 for LOSS and SDSS-II, respectively. We use −16 as the mean absolute SN magnitude, because Li et al. (2011b) measured mean absolute magnitudes for SN (Ib+Ic), −16.09 ± 0.23 mag, and SN II, −16.05 ± 0.15 mag. For LOSS, which contributes 42% of our targeted sample, Leaman et al. (2011) report a median limiting magnitude of 18.8 ± 0.5, corresponding to a detection limit of z = 0.023 for SN (Ib+Ic) and SN II. LOSS survey observations are taken without a filter, and the total response function peaks in the R band (Li et al. 2011b).

Dilday et al. (2010) report a ∼21.5 mag r'-band detection limit for the SDSS-II survey, which accounts for 33% of the galaxy-impartial sample. For our sample of galaxy-impartial discoveries, the redshift upper limit is z = 0.008, corresponding to the SDSS-II detection limit. The PTF, accounting for 32% of galaxy-impartial SN, has a limiting R-band magnitude of ∼20.8 mag which corresponds to an upper detection limit of z = 0.056.

Varying the upper redshift limit for galaxy-impartial and targeted SN discoveries (e.g., from z = 0.023 to z = 0.02 or z = 0.06) does not alter the type-dependent trends we find. Table 4 presents the mean redshifts for each SN type.

From the information available in IAU circulars, we separated discoveries into those made by amateur astronomers, who generally use comparatively small telescopes, and by professional astronomers. Table 2 lists the numbers of SN in each sample that were amateur discoveries. We present significance values for several sample comparisons with and without amateur discoveries in Section 7.

We identified SDSS spectroscopic fibers that targeted the galaxy nucleus by visually inspecting images and fiber positions. Oxygen abundance varies primarily with offset from the galaxy center, and we also determined which fibers have host offsets within 3 kpc of the SN host offset. The numbers of fibers in each category are listed in Table 2.

Galaxy-impartial searches do not target specific galaxies, so their discovery rate may be expressed in terms of the rate of SN per unit volume,

$$\begin{equation} \big\langle N^T_{z_i} \big\rangle = r_{z_i}^T \times V_{z_i} \times t_{z_i}^T, \end{equation} \tag{ 4 }$$

where $r_{z_i}^T$ is the type T SN rate per unit volume, $V_{z_i}$ is the volume within the survey field of view, and $t_{z_i}^T$ is the control time for type T SN that explode in the z_i redshift bin (e.g., 0.01 < z < 0.015).

The power-law form of the Schechter luminosity function (Schechter 1976) cuts off exponentially at the characteristic absolute magnitude, M_*. The AGN and Galaxy Evolution Survey (AGES) found that M_* becomes ∼0.2 mag brighter between z = 0 and z = 0.2 (Cool et al. 2012). Moustakas et al. (2011) reported, also from AGES spectroscopy, that the mean gas-phase metallicity of galaxies decreases by only ∼0.02–0.03 dex from z = 0.1 to z = 0.2. Given this evidence for relatively modest changes in the galaxy population, we may reasonably expect the fractional representation of each SN type among SN that explode to change only modestly to z ≈ 0.2,

$$\begin{equation} \frac{r_{z_i}^A}{r_{z_i}^B} \approx \frac{r_{z_j}^A}{r_{z_j}^B}, \end{equation} \tag{ 5 }$$

where z_i and z_j are redshifts less than 0.2 and A and B are distinct spectroscopic classes of SN. The relationships in Equations (4) and (5) suggest that changes in the ratio between the numbers of detected SN of two species with increasing redshift depend primarily upon changes in the control times for the species,

$$\begin{equation} \frac{d}{dz} \frac{\langle N^A \rangle }{\langle N^B \rangle } \approx \frac{r^{A}}{r^{B}} \frac{d}{dz} \Bigg (\frac{t^A}{t^B} \Bigg). \end{equation} \tag{ 6 }$$

The left panel of Figure 1 plots the cumulative redshift distributions of core-collapse SN as well as Type Ia SN with z < 0.2 reported to the IAU by galaxy-impartial surveys. We extend the redshift upper limit beyond the z = 0.08 galaxy-impartial limit to compile a larger sample of SN discoveries. This plot suggests that surveys may have similar control times for SN II, SN IIb, SN IIn, SN Ib, SN Ic, and SN Ic-BL with increasing redshifts, although the numbers of some species are small. The Type Ia SN are, however, discovered at greater redshifts than the core-collapse species because of greater intrinsic brightness. This test may only be sensitive to strong differences among control times, however, and we expect that a more complete analysis based on detailed knowledge of each survey and improved constraints on SN luminosity functions will find differences among the control times for the core-collapse spectroscopic types.

Zoom In Zoom Out Reset image size

The control times for targeted surveys are likely to have similar behavior from z = 0 to z = 0.023 (where η = 0.5) as those for galaxy-impartial surveys from z = 0 to z = 0.08 (where η = 0.5). Targeted surveys generally have shallower limiting magnitudes, but their galaxy targets are also at smaller distance.

The redshift distributions of targeted SN, shown in the right panel of Figure 1, additionally depend on the set of galaxy search targets as well as type-dependent host galaxy preferences. A Malmquist effect in galaxy catalogs (e.g., the NGC) used to select targets (Leaman et al. 2011) means that more distant monitored galaxies are, on average, more luminous, and likely more metal-rich. The redshift distributions of the targeted SN samples in Figure 1 show greater differences than those for galaxy-impartial SN samples.

Here, we have only discussed possible systematic errors associated with differences between the luminosity functions of the core-collapse species and the redshift dependence of the galaxies monitored by SN searches. In Section 9, we list additional potential sources of systematic error including fiber targeting and spectroscopic classification.

The WCS provided in SDSS DR8 frame*.fits headers is a TAN approximation to the asTrans*.fits full astrometric solution and has subpixel accuracy (M. Blanton 2011, private communication), an improvement over the astrometric solution provided in DR7 fpC*.fit headers. We used SWarp (Bertin et al. 2002) to register and resample SDSS images of each host galaxy to a common pixel grid and co-added SDSS images in the u'g'r'i'z' bands. The SDSS DR8 frame*.fits images also feature an improved background subtraction (in comparison to DR7 fpC*.fit images). The DR8 background level is estimated from a spline fit across consecutive, adjacent frames from each drift scan, after masking objects in each image.

Host galaxy images were used to measure the color of the stellar population near the site of each SN, estimate each host's stellar mass, a proxy for chemical enrichment (e.g., Tremonti et al. 2004), and compute the deprojected host offsets of SN and SDSS spectroscopic fibers.

The SExtractor measurement MAG_AUTO, corresponding to the flux within 2.5 Kron (1980) radii, was used to estimate host galaxy stellar mass-to-light ratio (M/L) through fits with spectral energy distributions (SEDs) from PEGASE2 (Fioc & Rocca-Volmerange 1997, 1999) stellar population synthesis models using the appropriate SDSS instrumental response function. An estimate of the stellar mass was then computed: M = M/L × L where L is the galaxy's absolute luminosity. See Kelly et al. (2010) for a detailed description of the star formation histories used.

We then estimated the host galaxy's color near the SN location using two techniques. The first and more simple method was to extract the u'g'r'i'z' flux inside of a circular aperture with 300 pc radius centered on the SN location, after subtracting the median of the peripheral background regions. While this technique is straightforward, a small number of apertures centered at the sites of SN at large host offsets or found in low-luminosity hosts had low signal-to-noise ratio (S/N), especially in the u' and z' bands. The primary intent of the second method was to obtain higher S/N u'-band flux measurements near the sites of SN, in particular near the sites of the SN with faint hosts. To identify g'-band pixels with S/N > 1 associated with each host galaxy, we used SExtractor to generate a segmentation map of each image, which identifies the pixels associated with each object. We adjusted the SExtractor settings so that the segmentation map included only pixels with S/N > 1 (i.e., DETECT_THRESH = 1). The 20 pixels closest to the SN location contained in the g'-band segmentation maps define the aperture for measurements of u' − z' color and u'-band surface brightness.

The aperture generally consists of 20 pixels on the CCD array closest to the SN position (i.e., within a circle of radius ∼1''). Only for 20 of the 519 SN is the average distance between aperture pixels and the explosion site greater than 08. Excluding measurements where the average pixel is more than 08 from the explosion coordinates does not affect the distributions we plot in this paper. The median uncertainties of the measured u'g'r'i'z' magnitudes are 0.11, 0.06, 0.06, 0.06, and 0.13 mag, respectively. We correct for Galactic reddening using the Schlegel et al. (1998) dust maps.

The reported SN positions come from a variety of sources that may rely on catalogs (e.g., USNO B1, 2MASS, or GSC) with more modest astrometric accuracy than available for SDSS images. Repeated photometric follow-up, available for some SN, can also be useful for improving the accuracy of explosion coordinates (e.g., Hicken et al. 2012).

We use KCORRECT (Blanton & Roweis 2007) to estimate rest-frame u'g'r'i'z' magnitudes from the fluxes we measure. KCORRECT fits the input measured fluxes with a model for the rest-frame SED, and it uses this SED to estimate the rest-frame u'g'r'i'z' magnitudes for each galaxy. The estimated rest-frame u' − z' color, for example, therefore depends upon the full set of measured observer-frame u'g'r'i'z' magnitudes that inform the SED model. Consequently, even if u' and z' measurements are noisy, the estimated rest-frame u' − z' color may be informative, given constraints from less noisy measured g'r'i' fluxes.

To identify SDSS fibers that coincide with a host galaxy, we searched inside a catalog available online from an MPA-JHU Collaboration⁸ for fibers that fell within an aperture with radius (1.65/z)'' placed on the host center and with redshifts that agree with that of the SN. For an object in the Hubble flow, this angle corresponds to a physical distance of approximately 34 kpc. At z = 0.03, for example, this radius subtends a 55'' angle. If the g'-band SExtractor segmentation map ID at the fiber location was the same as the ID of the SN host galaxy, the fiber was considered a match to the galaxy after a visual check. The deprojected normalized offset of the fiber was then calculated by computing the offset at each pixel in the 3'' fiber aperture and averaging these offsets weighted by each pixel's g'-band counts.

Only the spectra classified by the SDSS pipeline as a galaxy spectrum (SPECTROTYPE = "GALAXY") enter our analysis, a restriction that excludes QSO and Type 1 active galactic nuclei (AGNs) whose continua have significant non-stellar contributions. The SDSS "galaxy" class, however, includes spectra with emission line strength ratios characteristic of Type 2 AGNs and low ionization nuclear emission regions (LINERs). The emission line patterns associated with AGN activity are significantly degenerate with variation in oxygen abundance, so AGN line ratios preclude metallicity measurements.

We use the classifications of fiber spectra as star forming, low S/N star forming, composite, AGNs, or low S/N AGNs made available by the MPA-JHU group following Brinchmann et al. (2004). That analysis uses each spectrum's position on the Baldwin et al. (1981) (hereafter BPT) diagram of [O iii] λ5007/Hβ and [N ii] λ6584/Hα line ratios.

From the fiber spectra (closest in deprojected offset to the SN sites), we estimate host galaxy reddening A_V using the Balmer decrement (Hα/Hβ), assuming the R_V = 3.1 Cardelli et al. (1989) extinction law. Following Osterbrock (1989), we assume a Case B recombination ratio of 2.85 when spectra are classified as star forming or low S/N star forming and a ratio of 3.1 when spectra are classified as composite, AGNs, or low S/N AGNs.

Our analysis uses both (1) abundances and specific SFR estimates available from the MPA-JHU Collaboration for SDSS fiber spectra and (2) abundances we compute using the Pettini & Pagel (2004) metallicity calibration. We only use galaxies with S/N > 3 Hβ, Hα, [N ii] λ6584, and [O iii] λ5007, as designated by the MPA-JHU analysis. For abundance measurements, we only analyze spectra classified as star forming. For specific SFR estimates, we use star forming, composite, and AGN spectra.

5.7.1. MPA-JHU Metallicity and Specific SFR

5.7.2. Pettini and Pagel Metallicity

Oxygen abundances measured from fibers centered on the host galaxy nucleus are, on average, only 0.01 dex (T04) greater than the abundance inferred from the stellar mass with the Tremonti et al. (2004) M–Z relation, with a scatter of 0.14 dex. If we instead select fibers closest in host offset to SN explosion sites, spectroscopic abundances are 0.053 dex (T04) less than abundances estimated from stellar masses with a scatter of 0.16 dex.

In the following sections, we test the null hypothesis that two samples are drawn from a single underlying distribution using the Kolmogorov–Smirnov (K-S) test. The K-S test statistic is defined as $D=\sup _x |F_1(x)-F_2(x)|$, the maximum difference between the samples' cumulative distribution functions, where $F_n(x)=({1}/{n})\sum _{i=1}^n I_{X_i\le x}$. The K-S distribution is the distribution of the test statistic D, given the null hypothesis that two distributions are identical. The p-value is the probability of observing a value of the test statistic, D, more extreme than the observed value of D given the null hypothesis that the two samples are drawn from a single underlying distribution. Low p-values (<5%) are significant evidence that the underlying distributions are distinct.

When two independent samples are drawn from the same distribution, there is, by definition, a 5% random chance of obtaining a p-value less than 5%. If we were to make, for example, 20 comparisons among samples drawn from identical distributions, one misleading p < 5% difference would occur by chance on average. The number of independent comparisons we make in this paper should therefore be taken into account when comparisons yield p-values of modest significance (p ≈ 5%). We note that the host properties we measure are correlated (e.g., host color and metallicity), so independent comparisons are fewer than the total number of comparisons.

As can be seen in Figure 2, SN IIb and SN Ic-BL erupt in exceptionally blue environments, while high u'-band surface brightness is typical of SN Ib and SN Ic sites. SN II sites show substantial overlap in color or surface brightness with the other classes. This plot shows u' − z' color versus u'-band surface brightness, measured inside an aperture consisting of the 20 pixels closest to the SN site with g'-band S/N > 1.

Zoom In Zoom Out Reset image size

The u' − z' color near the site of SN 2006aj, the SN–LGRB in our sample, was 0.88 mag.

Figure 3 helps to explain the exceptionally blue u' − z' color of SN IIb and SN Ic-BL sites and the high u'-band surface brightnesses near SN Ib and Ic sites. At one set of extremes, SN Ic-BL have generally low mass hosts, while SN IIb explode at larger offsets when they occur in galaxies of large masses. At another extreme, SN Ib and especially SN Ic more often occur inside the g'-band half-light radius of massive galaxies, sites expected to have redder color and high surface brightness.

Zoom In Zoom Out Reset image size

Host galaxy mass is a moderately precise (∼0.1 dex) proxy for chemical abundance (e.g., Tremonti et al. 2004) which does not suffer from the AGN selection effects. The hosts of SN (Ib+Ic), excluding SN Ic-BL, are more massive than SN II hosts.

The host stellar mass of SN 2006aj, the only SN–LGRB in our sample, was 8.0 × 10¹⁰ M_☉. We find with p = 12% that the SN IIn offset distribution is consistent with the SN II host offset distribution.

To probe the metallicities of the core-collapse hosts, we measure oxygen abundance from the fiber spectrum with deprojected offset most similar to the SN offset. Among the 311 host galaxies with spectroscopic fiber measurements, 139 have multiple SDSS fiber spectra. SDSS fiber spectra generally target the central regions of host galaxies (see Tables 1 and 2), with an average host offset in our sample of 0.45 × r_half-light. The low metallicities shown in Figure 4 for SN Ic-BL and SN IIb hosts and high metallicities for SN Ic hosts are consistent with the patterns we see among the species' colors near explosion sites, host offsets, and host masses.

Zoom In Zoom Out Reset image size

7.3.1. Every Abundance Measurement

7.3.2. When Fiber and SN Host Offsets Are Similar

SDSS spectra provide an estimate of the specific SFR (SFR M⁻¹_☉ yr⁻¹) within the aperture of the fiber, which generally targets the host galaxy within the g'-band half-light radius. As can be seen in Figure 5, there is a progression of increasing specific SFR from SN II to SN Ib to SN Ic host spectra. SN Ic-BL host spectra also have significantly greater specific SFR than SN II host spectra. These spectra yield significant evidence that the nuclei of SN (Ib+Ic) host galaxies have greater specific SFRs than those of SN II host galaxies (p = 10%).

Zoom In Zoom Out Reset image size

Strong central star formation among SN (Ib+Ic) hosts may overwhelm AGN-patterned emission and explain the relatively low AGN fraction among SN (Ib+Ic) host galaxies.

Although SN Ic hosts have stronger specific SFR within the half-light radius, the region where most SN Ic explode, the sites of SN Ic are not bluer than those of SN II (see Figure 2). Figure 6 shows that the high extinction of SN (Ib+Ic) host galaxies, measured from host spectra, may redden ongoing star formation in SN Ic host galaxies.

Zoom In Zoom Out Reset image size

The host galaxies of SN IIb have less extinction than SN (Ib+Ic) host galaxies. The average extinction difference between SN (Ib+Ic) and SN IIb hosts is A_V ≈ 0.5 mag, a u' − z' reddening of ∼0.6 mag. The approximately similar internal extinctions of SN II and SN IIb hosts, however, suggest that the stellar populations near SN IIb likely are intrinsically bluer than those near SN II sites.

SN (Ib+Ic) host reddening is consistent (p = 45%) with that estimated along the line of sight to 19 SN (Ib+Ic) from their light curve colors by Drout et al. (2011) using an empirical model of SN Ib/c photometric color evolution. Comparison between the Drout et al. (2011) sample and the SN II host A_V distribution yields p = 2.4%. Here, we plot only the Drout et al. (2011) Gold and Silver SN. There is a median A_V ≈ 1.2 mag extinction through SN (Ib+Ic) host fiber apertures (E(B − V) ≈ 0.4 mag).

SDSS spectroscopic fiber apertures have a fixed radius of 15. At increasing redshift, this aperture radius corresponds to larger physical scales: 0.3 kpc at z = 0.01; 0.6 kpc at z = 0.02; and 1.17 kpc at z = 0.04. For a sample of 11 NGC spiral galaxies, van Zee et al. (1998) found a mean radial abundance gradient of −0.052 dex kpc⁻¹. While metallicity gradients vary among galaxies and have some dependence on, for example, host galaxy morphology (e.g., Kewley et al. 2006) and the metallicity calibration (e.g., Moustakas et al. 2010), we use this as a representative value.

Within the targeted sample (z < 0.023), nuclear spectroscopic fibers extend at most 0.68 kpc away from the host center, corresponding to systematic shifts of order only ∼0.025 dex. Galaxy-impartial SN discoveries (to z < 0.08) account for a significant fraction of only the SN Ic-BL sample. The difference between the median abundances for SN Ic-BL and SN Ic hosts is ∼0.5 dex, substantially greater than an aperture effect may yield.

There may be variation among the classification practices of the different surveys that contribute to our samples. A concern is that surveys that monitor different host galaxy populations (e.g., galaxy-impartial and targeted) could have different classification practices, such as use of automated classification tools (e.g., SNID) or multi-epoch spectroscopic follow-up. For instance, the helium lines that identify SN Ib often emerge only after a couple of weeks (e.g., Li et al. 2011b).

The SDSS object detection algorithm mistakenly split many galaxies of large angular size into two or more components (see Figure 9 of Blanton et al. (2005)). The SDSS targeting algorithm then placed fibers on these false components, sometimes at significant offset from the true galaxy center. The error rate of these algorithms could depend on galaxy morphology (e.g., irregularity or an interacting neighbor), and we checked whether the offsets of fiber measurements depend on SN type. However, we found no evidence of strong variation with SN type.

SDSS fibers often target the local maxima of galaxy light distributions, including host nuclei and bright H ii regions. In our sample, fibers have mean offset of 0.45 × r_half-light, while matched fibers (where |r_env − r_fiber| < 3 kpc) have mean offset of 0.55 × r_half-light. Therefore, fiber sites are highly likely to be more metal-rich on average than they would be if SDSS fibers sampled galaxy light distributions more democratically. However, the fibers' offset distribution does not vary strongly with SN type.

The lifetimes of H ii regions may be shorter than the those of the progenitors (N. Smith 2010, private communication), and the signal at the SN site may be too weak. Programs that take host spectra at the location of the SN (e.g., Anderson et al. 2010; Modjaz et al. 2011; Leloudas et al. 2011) may only extract a metallicity estimate when there is sufficient nebular emission through the slit. Any such S/N requirement could possibly act as a type-dependent selection effect. The SDSS targets bright nuclei or H ii regions, moderating any such effect in our analysis.

There are strong connections among the type-dependent patterns in host galaxy photometry and spectroscopy.

• 1.  
  Host galaxies of SN Ic-BL in the sample generally have low mass and high specific SFR, helping to explain the blue colors at broad-lined SN Ic explosion sites. The SN IIb typically are found beyond the g'-band half-light radius in massive hosts, offering explanation for the blue colors of their sites. The SN Ic-BL and SN IIb host galaxy fiber spectra have lower abundances than SN Ic and SN Ib host galaxy fiber spectra closest in radius to the explosion site, respectively, although the SN IIb–SN Ib difference is significant for only one of two strong-line diagnostics.
• 2.  
  SN Ic often erupt at small offsets in massive galaxies with strong specific SFR, high oxygen abundance, and high extinction measured from fiber spectra. These fibers generally collect light from within the host's g'-band half-light radius. These explosion sites help to explain the high surface brightnesses near SN Ic explosion sites. High interstellar reddening helps to explain why the colors near SN Ic sites have colors similar to those of SN II, despite their hosts' high specific SFR.

The u' − z' color and u' surface brightness near SN explosion sites considerably separate SN IIb, SN Ic, and SN Ic (see Figure 2). SN Ic sites predominantly have high surface brightness, while SN IIb populate lower-surface brightness but extremely blue environments. SN Ib largely occupy the parameter space between the SN IIb and the SN Ic. By contrast, however, the explosion sites of SN II have no specific locus in the color–brightness plane.

These patterns suggest that the fraction of the overall stripped-envelope SN (Ib+Ic+IIb) to Type II SN may not vary strongly with environment. Perhaps the stars that may explode as one stripped-envelope species at one value of mass in one environment instead exploded as another stripped-envelope species in other environments where, for example, the chemistry is different.

The best-studied example of a Type IIb, SN 1993J, exploded at a distance of only 3.6 Mpc in M81 (Filippenko et al. 1993; Matheson et al. 2000), and archival imaging revealed that its progenitor was a K-type supergiant (Aldering et al. 1994). Hubble Space Telescope (HST) imaging after the SN disappeared found evidence for a B-type supergiant binary companion (Van Dyk et al. 2002; Maund et al. 2004). More recent studies of the sites of other SN IIb suggest, however, that a fraction of the SN IIb population may erupt from massive single stars (e.g., Crockett et al. 2008). Chevalier & Soderberg (2010) analyzed the radio emission, optical shock breakout, and nebular emission of a sample of SN IIb to constrain the extent of their progenitors' envelopes and the properties of their circumstellar material. They favor two progenitor populations: (1) extended progenitors (SN 1993J and SN 2001gd) with hydrogen envelope mass greater than ∼0.1 M_☉ and slow, dense winds; and (2) more compact and massive Wolf–Rayet progenitors (SN 1996cb, SN 2001ig, SN 2003bg, SN 2008ax, and SN 2008bo) with a less massive hydrogen envelope and lower density winds. PTF11eon/SN 2011dh, an SN IIb (Arcavi et al. 2011; Marion et al. 2011), was recently discovered by amateur astronomer Amadee Riou in M51, where pre-explosion HST imaging exists of the SN site. Analysis of the archival images finds evidence for a supergiant with T_eff ≈ 6000 K at or near the SN site (Van Dyk et al. 2011; Maund et al. 2011). Radio and X-ray observations (Soderberg et al. 2012) and the optical spectroscopic and photometric evolution (Arcavi et al. 2011; H. Marion et al. 2012, in preparation) both favor a compact progenitor, however, suggesting that this star may be a binary companion or not associated with the SN.

Our analysis finds three statistically significant, plausibly related patterns in the host environments of SN IIb: SN IIb environments are bluer than the environments of SN Ib, SN Ic, and SN II; their explosion sites may be more metal-poor than those of SN Ib or SN Ic (significant for one of two abundance diagnostics); and their host galaxy interstellar extinction is less than that of SN (Ib+Ic). These trends are statistically significant even when we analyze only the locations of targeted SN.

An unambiguous implication of the exceptionally blue colors of SN IIb environments is that the Type IIb progenitor population is distinct from that of Type Ib explosions. Lack of hydrogen features near maximum light in SN Ib spectra may reflect a more extensive loss of the progenitor's hydrogen envelope. Comparatively metal-poor SN IIb host galaxies suggest that metals may play an important role in achieving this loss of the outer envelope.

The SN IIb population may erupt from a combination of massive single stars and progenitors in close binary systems, so a possibility is that the blue colors of SN IIb environments indicate higher binary fractions. Although the current examples of each class are few, future efforts may be able to draw distinctions between the environments of the compact and extended SN IIb progenitors proposed by Chevalier & Soderberg (2010).

Li et al. (2011a) recently reported that the hosts of SN IIb detected by LOSS had greater K-band luminosities than SN II-P hosts (with p = 6.9%). Lower SN IIb progenitor metallicities are consistent with the PTF's diminished fraction of SN IIb and SN Ic-BL in "giant," presumably metal-rich galaxies, than in "dwarf" galaxies: 1 SN Ib, 3 SN IIb, 2 SN Ic-BL, and 9 SN II in "dwarf" galaxies, and 2 SN Ib, 2 SN IIb, 7 SN Ic, 1 SN Ic-BL, and 42 SNs II in "giant" galaxies (Arcavi et al. 2010).

Type Ic-BL are the SN that have been associated with coincident LGRB explosions (Galama et al. 1998; Matheson et al. 2003; Stanek et al. 2003; Hjorth et al. 2003; Malesani et al. 2004; Modjaz et al. 2006; Sanders et al. 2011; see Woosley & Bloom 2006 and Modjaz 2011 for reviews). Modjaz et al. (2008) showed that SN Ic-BL with associated LGRB prefer more metal-poor environments than do SN Ic-BL without an LGRB (but see Levesque et al. 2010).

We find that host galaxies of SN Ic-BL (without an associated LGRB) follow a significantly more metal-poor distribution than the hosts of normal SN Ic (or SN Ib) explosions, even when only galaxy-impartial discoveries are considered. The colors of SN Ic-BL local environments also follow a bluer distribution than those of SN Ic, further evidence for different progenitor populations. SN Ic-BL host galaxies have strong specific SFRs, similar to those of normal SN Ic.

Lower Type Ic-BL progenitor oxygen abundances may imply reduced rates of wind-driven mass loss, potentially enabling SN Ic-BL progenitor to retain greater angular momentum (e.g., Kudritzki 2002; Heger et al. 2003; Eldridge & Tout 2004; Vink & de Koter 2005). High angular momentum before the explosion may be important to the production of high velocity ejecta (Woosley et al. 1993; Thompson et al. 2004). Nonetheless, the means by which SN Ic-BL progenitors shed their outer envelopes, if not through their high metallicity, need explanation and may involve Roche lobe overflow (Podsiadlowski et al. 1992; Nomoto et al. 1995), stellar mergers (Podsiadlowski et al. 2010), or perhaps deep mixing.

Here, our measurements support a picture where both SN Ib and SN Ic have more metal-rich hosts on average than SN Ic-BL, consistent with the host galaxy magnitudes measured by Arcavi et al. (2010). It presents a contrast with the results of Modjaz et al. (2011) who recently measured the oxygen abundances at the sites of SN Ic-BL, SN (Ib+IIb), and SN Ic. There the SN Ic-BL distribution falls intermediate between those of SN (Ib+IIb) and SN Ic, although it is more similar to the comparatively metal-poor SN (Ib+IIb) distribution and neither comparison is statistically significant. These contrasting trends may relate to fact that Modjaz et al. (2011) constructed their samples for each SN type from approximately equal numbers of galaxy-impartial and targeted SN discoveries, or the inclusion of SN IIb (which we find inhabit metal-poor environments) with SN Ib. Modjaz et al. (2011) measurements were also taken at the explosion site, which may often differ significantly from the host abundance measured from SDSS fiber spectra (0.13 dex average disagreement with nuclear fiber measurements).

Svensson et al. (2010) found that host galaxies of LGRBs had smaller star masses than core-collapse SN hosts and had high surface brightness and more massive stellar populations. The only SN–LGRB that met our sample criteria, SN 2006aj, has low host stellar mass and comparatively blue u' − z' color near the explosion site.

In an earlier paper (Kelly et al. 2008), we showed that, while the positions of the other core-collapse SN follow the distribution of their hosts' light, Type Ic SN trace the brightest regions of their host galaxies in a pattern similar to that followed by LGRB (Fruchter et al. 2006). Possible explanations for this pattern include shorter lifetimes and higher masses of SN Ic progenitors (Raskin et al. 2008; Leloudas et al. 2010; Eldridge et al. 2011) as well as preference for metal-rich regions near the centers of hosts. Anderson & James (2008) showed, subsequently, that SN Ic also track their hosts' Hα emission more closely than SN II (their comparison with SN Ib lacked statistical significance).

The SDSS fiber spectra of core-collapse host galaxies, which generally sample inside of the g'-band half-light radius, reveal an increasing progression of specific SFR from SN II to SN Ib to SN Ic (and SN Ic-BL) hosts. This pattern persists when we study only the spectra from fibers targeting the host nucleus. SN Ic explode at comparatively small host offset, linking them to the strong star formation near their hosts' centers.

We find that the central star formation that yields SN Ic generally has high chemical abundance and extinction from interstellar dust. An SN Ic progenitor population tracking high metallicity would be expected to explode in massive galaxies with strong star formation in metal-rich gas near their centers, the pattern we observe.

The colors of SN Ib and SN Ic explosion sites may offer evidence that their progenitors are also younger and more massive than the progenitors of SN II. The distribution of the apparent u' − z' color at SN Ib and SN Ic sites is similar to that at SN II sites. However, we find that SN (Ib+Ic) host galaxies have higher interstellar extinction (ΔA_V ≈ 0.5 mag). This suggests that SN (Ib+Ic) sites have intrinsically bluer color than SN II sites, perhaps indicative of younger progenitor stellar populations.

SN Ib explosion sites have higher u'-band surface brightnesses than SN II sites, while SN Ib host galaxies generally have lower abundance than SN Ic in our sample. There is no statistically significant difference between the SN Ib and SN Ic host offset distributions in our sample (p = 60%), which may imply that host offset cannot, on its own, explain the uniquely strong association of SN Ic with bright host galaxy pixels (Kelly et al. 2008).

While analyses of pre- and post-explosion imaging have not yet identified a progenitor of an SN Ib or SN Ic, red supergiants (RSGs) have been found at the sites of SN II-P explosions (e.g., Barth et al. 1996; Van Dyk et al. 1999, 2003a, 2003b, 2012; Smartt et al. 2001, 2003, 2004; Li et al. 2005, 2007; Maund & Smartt 2005). Smartt et al. (2009) favor an 8.5–16.5 M_☉ mass range for SN II progenitors, although extinction along the line of sight to the progenitors is not well constrained (e.g., Walmswell & Eldridge 2012). Smith et al. (2011) note that Wolf–Rayet stars in binary systems, possible progenitors of SN Ib and SN Ic, are expected to be less luminous than single Wolf–Rayet stars. Brighter companions may outshine Wolf–Rayet progenitors, although mass-gaining companions may, in some cases, explode first (Podsiadlowski et al. 1992; Eldridge et al. 2011). Even for progenitors with close binary companions, metallicity and mass are expected to be important in determining the composition of the outer envelope, even though substantial mass loss may occur through Roche lobe overflow (Smith et al. 2011; Yoon et al. 2010; Eldridge et al. 2011).

Prantzos & Boissier (2003) and Boissier & Prantzos (2009) found that SN (Ib+Ib/c+Ic) hosts have greater absolute M_B luminosities than SN II hosts. Prieto et al. (2008) presented the first comparison between the oxygen abundances of SN (Ib+Ic) and SN II host galaxies with large sample sizes. Using the T04 metallicities available for the SDSS DR4 spectra, they found p = 5% evidence for a difference (the sample may not have been large enough to determine the effect of AGN contamination). Van den Bergh 1997, Tsvetkov et al. (2004), Hakobyan et al. (2009), Anderson & James (2009), and Leaman et al. (2011) have found that SN (Ib+Ib/c+Ic) occur preferentially toward galaxy centers, where oxygen abundances are generally higher. Habergham et al. (2010), examining 178 host galaxies for evidence of interaction, and Anderson et al. (2011), in a study of SN sites in Arp 299, have explored explanations for these patterns.

Modjaz et al. (2011) find a significant difference (∼0.2 dex on average) between the oxygen abundances at the sites of 12 SN Ic and a mixed sample of 16 SN (Ib+IIb) for one of three oxygen abundance calibrations (although see Anderson et al. 2010 and Leloudas et al. 2011). When only abundances measured at the SN site from these three studies are compared, a significant difference between SN Ib and SN Ic metallicities computed with the Pettini & Pagel (2004) diagnostic is evident (M. Modjaz, private communication; M. Modjaz, in preparation).

Among our set of host measurements, we find no statistically significant differences between the characteristics of SN IIn host environments and those of normal SN II. Anderson & James (2009), who have also studied SN IIn explosion sites, found no significant difference between the mean radial offsets of 12 SN IIn and 35 SN IIP from the host galaxy center.

Narrow line emission characterizes SN IIn spectra (Schlegel 1990) and is thought to be the result of the interaction of the ejecta with high density surrounding material. The existence of dense circumstellar material likely indicates strong pre-explosion mass loss (e.g., Chugai & Danziger 1994) and can increase the optical luminosity of the SN by thermalizing the emerging blast wave (e.g., Woosley et al. 2007; Smith et al. 2011; van Marle et al. 2010).

Luminous Blue Variable (LBV) stars (e.g., η Car), with their high mass loss rates (>10⁻⁴ M_☉ yr⁻¹), have been suggested as candidate progenitors, although standard stellar modeling positions the LBV period before an ultimate Wolf–Rayet phase (e.g., Langer 1993; Maeder et al. 2005). Dwarkadas (2011) has recently suggested that observations may only present a convincing case for an LBV progenitor in the case of SN 2005gl (Gal-Yam et al. 2007; Gal-Yam & Leonard 2009). Other means of potentially producing regions of high density circumstellar material include, for example, pulsation-driven superwinds from RSGs (Yoon & Cantiello 2010).