SN Iax environments

Investigating the diversity of supernovae type Iax: A MUSE and NOT spectroscopic study of their environments


SN 2002cx-like Type Ia supernovae (also known as SNe Iax) represent one of the most numerous peculiar SN classes. They differ from normal SNe Ia by having fainter peak magnitudes, faster decline rates and lower photospheric velocities, displaying a wide diversity in these properties. We present both integral-field and long-slit visual-wavelength spectroscopy of the host galaxies and explosion sites of SNe Iax to provide constraints on their progenitor formation scenarios. The SN Iax explosion site metallicity distribution is similar to that of core-collapse (CC) SNe and metal-poor compared to normal SNe Ia. Fainter members, speculated to form distinctly from brighter SN Iax, are found at a range of metallicities, extending to very metal-poor environments. Although the SN Iax explosion sites’ ages and star-formation rates are comparatively older and less intense than the distribution of star forming regions across their host galaxies, we confirm the presence of young stellar populations (SP) at explosion environments for most SNe Iax, expanded here to a larger sample. Ages of the young SP (several  to  yrs) are consistent with predictions for young thermonuclear and electron-capture SN progenitors. The lack of extremely young SP at the explosion sites disfavours very massive progenitors such as Wolf-Rayet explosions with significant fall-back. We find weak ionised gas in the only SN Iax host without obvious signs of star-formation. The source of the ionisation remains ambiguous but appears unlikely to be mainly due to young, massive stars.

supernovae: general

paragraph \crefformatequationEq. #2#1#3

1 Introduction

The relative homogeneity of Type Ia supernovae (SN Ia) has allowed them to serve as precise extragalactic distance indicators after the application of empirically derived relations for light curve shape (Phillips, 1993) and color (Tripp, 1998). The similarity among the vast majority of SNe Ia suggests a commonality among their progenitor stars and their explosion physics. SNe Ia are generally thought to be the complete disruption of Chandrasekhar-mass white dwarfs (WDs) which undergo a thermonuclear runaway process of explosive carbon and oxygen burning. Despite a large population of ‘normal’ SNe Ia, increasing numbers of apparently similar events, but with spectroscopic and light curve peculiarities have been discovered. It is not clear whether such under- (e.g. Filippenko et al., 1992a; Leibundgut et al., 1993; Turatto et al., 1996) and over-luminous (Ruiz-Lapuente et al., 1992; Phillips et al., 1992; Filippenko et al., 1992b; Howell et al., 2006) events represent extreme extensions of the SN Ia progenitor model or are distinct in their progenitor stars and explosion mechanisms. One particular event, SN 2002cx, also appeared to fall into the category of SN Ia, following the traditional SN classification scheme (Filippenko, 1997), but garnered the title of the ‘most peculiar known’ SN Ia (Li et al., 2003). With early spectral similarity to the over-luminous SN 1991T-like events – consisting of conspicuous high-ionization spectral features associated with \ionFeii and \ionFeiii – but being relatively faint at peak ( mag), SN 2002cx prompted the classification of other ‘SN 2002cx-like events’ (e.g., Jha et al., 2006; Narayan et al., 2011). A comprehensive study of all objects designated as SN 2002cx-like events (hereafter SNe Iax) was presented by Foley et al. (2013). The sample of has since expanded and a recent review by Jha (2017) discusses the current status of SN Iax observations and theory.

The primary distinctions of SNe Iax from normal SNe Ia are fainter peak magnitudes, lower ejecta velocities, no secondary near-infrared maxima (with light curves peaking in optical before near-infrared, unlike normal SNe Ia), and late time spectra that do not become fully nebular (see Jha et al., 2006; Foley et al., 2013; Foley et al., 2016, for more in depth discussion). However even within the events classified as members of this group there appears a large degree of diversity. The low ejecta velocities of SNe Iax ( cf.  km s for normal SNe Ia around peak) implies that their explosions are of significantly lower energies, and their faint peak magnitudes indicate production of smaller amounts of radioactive Ni ( cf.  M for normal SN Ia, e.g. Stritzinger et al. 2006; Childress et al. 2015). One of the leading scenarios to explain SNe Iax is that they are deflagrations of Chandrasekhar mass WDs (e.g. Branch et al., 2004; Phillips et al., 2007; Jordan et al., 2012; Kromer et al., 2013) that have accreted helium-rich material from a companion star (Foley et al., 2009; Liu et al., 2015b). Due to the extension of SNe Iax to low explosion energies and inferred ejecta masses, there is indirect evidence that at least some fraction of these SNe could leave bound remnants.

Support for the helium-accreting WD scenario has come from direct progenitor constraints and environmental constraints. McCully et al. (2014a) present the detection of a blue source at the location of SN 2012Z that is similar to the Galactic helium nova V445 Pup and limits for SN 2014dt provided by Foley et al. (2015) are also consistent with a similar progenitor system. Liu et al. (2015c) suggest a hybrid CONe WD + He-star as the most favourable progenitor system for SN 2012Z. The delay time to explosion for WD+He-star systems has been investigated by Wang et al. (2013, 2014) and Liu et al. (2015b) who found a timescale of 100 Myr, in good agreement with observational constraints on their ages. Lyman et al. (2013) found that SN Iax host galaxy types closely match one drawn from the star-formation (SF) rate density of the local universe and the environments of SNe Iax trace the SF of their hosts at a similar level to core-collapse SNe II with expected delay times of tens-hundred Myr. This statistical estimate of their environment ages was backed up through analyses of the local stellar populations (SP) of SNe 2008ha (Foley et al., 2014) and 2012Z (McCully et al., 2014a), which found environments harbouring young SP with ages of 10–80 Myr.

Consistent with their young environments, it has also been suggested that some SNe Iax may be due to the collapse of massive stars. One particular example that has drawn attention is SN 2008ha. This was an extremely weak SN, even in comparison to other SN Iax, producing 0.003 M of Ni and possibly ejecting only a few tenths M of material (Valenti et al., 2009; Foley et al., 2009), although it is not alone in this regard (e.g., SN 2010ae; Stritzinger et al., 2014). Massive star models explored to explain these very low-energy and -luminosity events include the collapse of very massive Wolf-Rayet (WR) stars with significant fall-back onto a nascent black hole (Valenti et al., 2009; Moriya et al., 2010) and stripped-envelope electron-capture (EC) SNe (Pumo et al., 2009; Moriya & Eldridge, 2016). The scenario of very low-luminosity SNe accompanying massive stellar collapse may be appealing to explain apparently SN-less gamma ray bursts (GRBs, e.g. Fynbo et al., 2006; Della Valle et al., 2006; Michałowski et al., 2016), if indeed they are related to other long-duration GRBs (LGRBs), and thus a result of the collapse of massive stars (see, e.g., Gehrels et al., 2006; Gal-Yam et al., 2006).

Adding further diversity to the sample we note that for one event, SN 2008ge, no evidence for a young SP at the explosion site, or within the S0 host galaxy, was found by Foley et al. (2010a). Furthermore, for two events, SNe 2004cs and 2007J, helium (and potentially hydrogen) was detected in their spectra. Doubt as to their validity as SNe Iax members has been raised by White et al. (2015) who classify them instead as core-collapse Type IIb SNe, although this was refuted by Foley et al. (2016).

It is apparent that the group of SNe that have been labelled SN Iax are a heterogeneous group, and multiple progenitor systems and explosion mechanisms could be responsible (Foley et al., 2013; Jha, 2017). The sample has now grown to a size where statistical studies can elucidate further information on the progenitor systems, beyond characterising individual examples. Environmental diagnostics have played a central role in the understanding of other well-known transient types (see Anderson et al., 2015, and references therein) and, in particular, the Multi-Unit Spectroscopic Explorer (MUSE; Bacon et al., 2010) instrument mounted at the Very Large Telescope (VLT) is revolutionising the way such studies are performed. MUSE is an integral field spectrograph (IFS) providing spatial and spectral information at the transient explosion site and across the host galaxy system in a single pointing (e.g., Galbany et al., 2016a; Prieto et al., 2016; Krühler et al., 2017; Izzo et al., 2017).

In this paper we perform spectroscopic environmental measurements for a large number of SNe Iax using both IFS and long-slit data from VLT/MUSE and Nordic Optical Telescope/Andalucia Faint Object Spectrograph and Camera (NOT/ALFOSC), thereby providing additional constraints on the nature of the progenitor systems and investigating whether their wide-ranging explosion characteristics are matched by large spreads in environmental measures. In \crefsec:obs we present our observations. Our methods for the MUSE-observed sample are given in \crefsec:musemethods and in \crefsec:notmethods we highlight similarities and differences for our NOT-observed sample’s analysis. Results are presented and discussed in \crefsec:results,sec:discussion, respectively. Throughout the paper we use dex as the solar abundance (Asplund et al., 2009) and adopt the cosmological parameters  km s Mpc (Riess et al., 2016) and .

2 Observations and Data reduction

The current number of SN Iax discovered is around 50 events (Jha, 2017). Our sample consists of 37 SNe Iax host galaxies observed with VLT/MUSE or NOT/ALFOSC. The hosts span a redshift range of 0.004 to 0.08. Membership of our sample as SN Iax up to SN 2012Z is presented in Foley et al. (2013). Additionally we add SN 2013dh (Jha et al., 2013), SN 2013en (Liu et al., 2015a), SN 2013gr (Hsiao et al., 2013), SN 2014ey3 (Carnegie Supernova Project, in prep), SN 2014ck (Tomasella et al., 2016), SN 2014dt (Ochner et al., 2014), SN 2015H (Magee et al., 2016), PS 15aic (Pan et al., 2015), PS 15csd (Harmanen et al., 2015), SN 2015ce (Balam & Graham, 2015) and SN 2016ado (Tonry et al., 2016). Here we detail the observations taken for our sample on the two instruments.

2.1 Vlt/muse

Observations of the host galaxies were carried out in a MUSE programme running between September 2015 and March 2016 apart from the host of SN 2002bp, which was observed as part of an earlier programme (as this galaxy has also hosted other SNe) in April 2015. The time lag between the SN and our observations is more than two years for all but two events, SNe 2014dt and 2015H. In both these cases it is possible to identify broad forbidden line emission arising from the SN ejecta (see Foley et al., 2016 for an overview of SN Iax late-time spectra). These are analysed further in \crefapp:14dt15H where the late time SN spectra and any impact on our environmental analyses are presented. In essence, we do not consider the presence of faint underlying signal from these hydrogen-poor, non-interacting events to significantly impact our emission-line based environmental analyses.

The strategies for sky-subtraction with MUSE depended on the angular size of each host galaxy. For galaxies that did not cover the field of view of MUSE, blank sky spaxels within the on-source exposures could be used to create the sky-spectrum to subtract. Otherwise, we included two off-source shorter exposures of nearby blank sky interspersed among the on-source exposures. Four on-source exposures, rotated 90 degrees from each other, were used in each case to account for detector artefacts. The details of the exposures taken are given in \creftab:musesample. The seeing values were determined by the FWHM of a point source in the (spectrally) flattened data cubes, in the absence of a suitable source we used the observatory’s estimate of the conditions provided in the headers. All data were reduced and combined using version 1.6.2 of the ESO MUSE pipeline via reflex (Freudling et al., 2013) with default parameters. The ESO MUSE pipeline employs an empirical method for the removal of the sky signal, however this can leave significant residuals, particularly in the redder part of the spectrum where sky lines are prominent. To combat this, Zurich Atmospheric Purge (ZAP4; Soto et al. 2016) has been developed. We additionally applied this method to our already reduced data cubes to correct these residuals. Where we took off-source sky exposures, we applied ZAP to the reduced off-source blank-sky exposures, before applying the results to the on-source combined data cubes. For those hosts that did not fill the field of view of MUSE, the correction to be applied by ZAP was calculated using blank regions of the on-source data cubes. Although some residuals were still present, they were significantly reduced compared to the standard sky-subtraction method.

SN name Host galaxy Date Obs. Exp. time Seeing
(arcsec) (s) (arcsec)
1991bj IC 344 0.018 14.4 Sept 2015 b 1.7
2002bp UGC 6332 0.021 36.9 Apr 2015 1.2
2002cx CGCG 044-035 0.024 20.8 Jan 2016 b 1.7
2004cs UGC 11001 0.014 38.7 Mar 2016 1.1
2005P NGC 5468 0.009 78.9 Jan 2016 1.0
2005hk UGC 272 0.013 43.4 Oct 2015 1.4
2008ae IC 577 0.030 15.8 Nov 2015 b 0.9
2008ge NGC 1527 0.004 111.5 Sept 2015 0.6
2008ha UGC 12682 0.005 40.5 Nov 2015 1.4
2009J IC 2160 0.016 61.3 Sept 2015 2.0
2010ae ESO 162-017 0.004 61.3 Sept 2015 b 1.8
2010el NGC 1566 0.005 249.6 Sept 2015 1.5
2011ce NGC 6708 0.009 34.5 Sept 2015 0.9
2012Z NGC 1309 0.007 65.7 Oct 2015 0.5
2013gr ESO 114-007 0.007 54.6 Sept 2015 1.8
2014dt NGC 4303 0.005 193.7 Jan 2016 1.9
2014ey CGCG 048-099 0.032 19.8 Feb/Mar 2016a b 0.7
2015H NGC 3464 0.012 77.1 Dec 2015 1.1
  • The best five exposures from two incomplete observing attempts were combined.

  • Sky subtraction was done using blank regions of on-source exposures.

Table 1: Observations of SN Iax host galaxies with VLT/MUSE.

2.2 Not/alfosc

We observed 21 host galaxies of SNe Iax at the NOT using ALFOSC over two campaigns during March, September 2016 and April 2017. The data were reduced largely following the same procedures described in Taddia et al. (2013, 2015b), which we additionally describe here.

We obtained long-exposure ( 1800 s), long-slit spectra of the \ionHii regions within the host galaxies, by placing the slit at the SN position and choosing a position angle such that the galaxy centre or other bright \ionHii regions near the SN location were captured. In most cases the slit contained a few \ionHii regions bright enough for our analysis. For all but SN 2008A the slit was positioned to go through the host galaxy nucleus. The instrumental setup chosen was the same as adopted for the study presented in Taddia et al. (2013, 2015a), i.e. ALFOSC with grism #4 (wide wavelength range 35009000 Å) and a 1 arcsec-wide slit, resulting in a FWHM (full width half maximum) spectral resolution of 16–17 Å. Details of the observations and exposure times adopted for each host galaxy observation are listed in \creftab:notsample.

First, the 2-D (dimensional) spectra were bias subtracted and flat-field corrected in a standard way. When available, multiple exposures were then median-combined to remove any spikes produced by cosmic rays. Otherwise, we removed them using the L.A.Cosmic removal algorithm (van Dokkum, 2001).

We extracted the trace of the brightest object in the 2-D spectrum (either the galaxy nucleus, a bright star, or a \ionHii region with a bright continuum) and fitted with a low-order polynomial. The precision of this trace was checked by plotting it over the 2-D spectrum. We then shifted the same trace in the spatial direction to match the position of each \ionHii region visible in the 2-D spectrum, and then extracted a 1-D spectrum for each \ionHii region. The extraction regions were chosen by looking at the H flux profile. The extracted spectra were wavelength and flux calibrated using an arc-lamp spectrum and a spectrophotometric standard star, observed the same night or (in March 2016) during the same week, respectively. Following Stanishev (2007), we removed the second order contamination, which characterises the spectra obtained with grism #4, from each spectrum. In this study, we included all the spectra showing at least H and [N ii6583 emission lines.

SN name Host galaxy Date Obs. Exp. time Seeing
(arcsec) (s) (arcsec)
1999ax SDSS J140358.27+155101.2 0.023a Mar 2016 0.8
2002cx CGCG 044-035 0.024 20.8 Mar 2016 1.1
2003gq NGC 7407 0.021 59.9 Sep 2016 0.8
2004gw CGCG 283-003 0.017 38.8 Mar 2016 1.1
2005cc NGC 5383 0.008 94.9 Mar 2016 1.2
2006hn UGC 6154 0.017 29.3 Mar 2016 1.7
2007J UGC 1778 0.017 36.1 Sep 2016 1.3
2007qd SDSS J020932.73-005959.8 0.043 25.1 Sep 2016 1.0
2008A NGC 634 0.016 62.7 Sep 2016 0.7
2009ku APMUKS(BJ) B032747.73-281526.1 0.079 Sep 2016 1.7
2011ay NGC 2315 0.021 40.5 Mar 2016 1.0
PS1-12bwh CGCG 205-021 0.023 23.8 Apr 2017 1.1
2013dh NGC 5936 0.013 43.4 Mar 2016 1.6
2013en UGC 11369 0.012 34.5 Mar 2016 1.5
2014ck UGC 12182 0.005 38.8 Mar 2016 1.7
2014ek UGC 12850 0.023 28.7 Sep 2016 1.0
2015H NGC 3464 0.012 77.1 Mar 2016 1.3
2015ce UGC 12156 0.017 28.7 Sep 2016 1.0
PS 15aic SDSS J133047.95+380645.0 0.056 22.3 Mar 2016 1.2
PS 15csd SDSS J020455.52+184815.0 0.044b Sep 2016 0.7
2016ado SDSS J020305.81-035024.5 0.043 Sep 2016 1.3
  • An adjacent galaxy is present at (as measured from H in the NOT spectrum).

  • Redshift from H emission of host galaxy in PESSTO classification spectrum.

Table 2: Observations of SN Iax host galaxies with NOT/ALFOSC.

3 MUSE data analysis and methods

To analyse the data cubes, we used ifuanal5, a package developed in Python to perform spaxel binning, stellar continuum and emission line fitting of IFU data following work done by Stanishev et al. (2012) and Galbany et al. (2014). Further documentation is available for the package but we also detail the relevant analysis procedures in the following subsections.

The effects of Galactic reddening were removed for each cube using the extinction maps of Schlafly & Finkbeiner (2011) and adopting a Cardelli et al. (1989) reddening law before correcting to rest-frame using redshifts for each host given by the NASA/IPAC Extragalactic Database (NED)6.

Figure 1: An example of the binning and fitting routines performed on each data cube described in \crefsec:binning,sec:contfitting,sec:emisfitting, as shown for NGC 6708 (the host of SN 2011ce). Top: The spectral axis has been convolved with the transmission profile of the appropriate filter to produce an -band (left) and H (middle) image of the MUSE cube; North is up, East is left. The bins that were produced by the binning algorithm (\crefsec:hiibinning) are shown (arbitrarily coloured based on their H brightness) overlaid by the location of the nucleus and the SN explosion site, given by black and ‘star’ symbols, respectively (right). Bottom: The continuum fitting of an example spectrum (left). Black shows the observed spectrum and orange the starlight fit. Regions around strong emission lines were masked and these are shown as vertical shaded regions. Red elements in the residual plot are those flagged as bad pixels or clipped by starlight. The continuum fits were subtracted from the spectra before emission line fitting (right). Orange dashed line shows the multiple gaussian fit over the observed residual emission spectrum – line identifications for the main lines used in our analysis are shown.

3.1 Spaxel binning

As the spaxels in MUSE data are typically smaller than the seeing and typical sizes of objects of interest (i.e. \ionHii regions), we utilised binning methods in order to combine the signal of adjacent physically-related spaxels, as well as improve the SNR for faint regions. For every galaxy except NGC 1527 (the host of SN 2008ge) we adopted a binning method aimed at segmenting \ionHii regions. For NGC 1527 we used the Voronoi binning method based on achieving a target SNR of the bins in the continuum. Additional to these algorithmically created bins, we manually added two custom bins, a fibre-like 2 arcsec bin on the host nucleus and a bin at the explosion site of the SN using the same radius limit as for our \ionHii region binning (\crefsec:hiibinning). For any cases where it was evident that foreground stars were affecting the binning algorithm, these were removed using a circular mask before repeating the binning. A single weighted-average spectrum was used to represent each bin, which was forwarded for stellar continuum and emission line fitting. An overview of the binning performed for each data cube is given in \creftab:bins.

\ionHii region binning

In order to separate \ionHii regions into bins we use a method based on the Hiiexplorer7 algorithm (Sánchez et al., 2012; Galbany et al., 2016a). A narrow 30 Å filter was simulated in the data cubes, centred on the H emission line, and from this we subtracted a continuum level determined by interpolation of the flux in two simulated neighbouring continuum filters. From the H map a region of blank sky was used to determine the background median and standard deviation. Seeds of potential bins were found as all peaks in the flux distribution that were above the background median. The H map was smoothed with a narrow gaussian before the peak detection algorithm was run to avoid finding spurious peaks due to noise fluctuations. Seeds were ordered in descending H flux and, starting with the brightest seed, every pixel in the H map that satisfied the following conditions was added to the candidate bin:

  1. within 250 pc (or 1/2 FWHM of the MUSE cube if this was larger) of the seed pixel;

  2. a flux at least 10 per cent of the seed pixel flux;

  3. above the background median.

If the pixels satisfying these conditions were a contiguous group of 8 or more then the bin was stored, otherwise the bin was rejected. The next seed (that had not already been assigned a bin) was then used. This process was repeated until all seeds were used.

The nature of the Hiiexplorer algorithm means that bins in bright areas of the galaxy are grown to a similar size (i.e. the maximum radius specified). Our limit is conservative in comparison to observed sizes of \ionHii regions, however our choice was driven by the approximate mode of \ionHii region sizes (e.g. Oey et al., 2003). Thus, although we will crop any giant regions that can have diameters up to 1 kpc, our limit means the majority of bins are representative of \ionHii region sizes. As shown in the top panels of \creffig:museanalysis, the algorithm captures each of the individual regions above our flux limit, although some of the diffuse emission and the outer extents of very large complexes may be missed.

The typical minimum flux in the H narrow band image we consider as a bin seed was a few  erg sec cm Å. The number of bins found by the algorithm for each galaxy is shown in \creftab:bins.

Continuum binning

NGC 1527 is a smooth-profiled, early-type S0 galaxy (Foley et al., 2010a) and as such we chose another method of binning the spaxels. Specifically, we adopted the Voronoi binning method of Cappellari & Copin (2003), designed to attain a minimum signal-to-noise in the continuum ( Å) for each bin whilst maximising spatial resolution. Individual spaxels were required to have a signal-to-noise ratio (SNR) of 3 in the continuum window and these were binned by the algorithm to a target of SNR = 40 per bin. In practice, individual spaxels in the central  kpc of the galaxy had SNR  40 and so these were not binned. Even with the reasonably high SNR target, bin sizes only reached radii of  pc in the outskirts of the field of view.

Host nucleus bins

For each MUSE cube where the central nucleus of the host galaxy was in the field of view we simulated a representative survey fibre by creating a 2 arcsec diameter bin centred on the nucleus. The results pertaining to these bins will be referred to as “nucleus” values.

SN explosion site bins

We manually added a bin at the location of the SN explosion site in each host. The explosion site bins were circular with a diameter equal to that used for the \ionHii region binning (\crefsec:hiibinning). The location was found via astrometric alignment of a SN discovery image or, if no image could be obtained, we used offsets from the host centre given in discovery IAU circulars (the method used for each is given in \creftab:bins). Where a suitable image was found, the data cube was convolved with a filter transmission matching the SN image filter to create a MUSE image. The MUSE field of view is comparatively small at  arcmin and so, generally, only a small number of common sources with which to tie the astrometric alignment were present. We therefore first registered the SN and MUSE images with their WCS information before fitting a transformation that allowed a shift only.

Two of our MUSE images detected their respective SN at late times - SN 2014dt and SN 2015H, and we used these as a test of our alignment procedures. With 3 sources to tie the fit, our transformed location of SN 2015H was 0.6 pixels (0.12 arcsec) offset from the centroid location of the SN in the MUSE image itself. For SN 2014dt the MUSE image is quite poor seeing and we are limited to using diffuse \ionHii regions as common sources; nevertheless we find our transformed location to be only 1 pixel offset from the centroid of the SN. When compared to the typical sizes of the explosion site bins, uncertainties in the alignment of pixel do not significantly impact our results.

SN name Binning Survey fibre Explosion site
algorithma bin? astrometryb
1991bj \ionHii region 31 Y NTT/EMMI (R)
2002bp \ionHii region 67 Y IAUC offsets
2002cx \ionHii region 17 Y IAUC offsets
2004cs \ionHii region 90 Y IAUC offsets
2005P \ionHii region 127 N Swope/SITe3 ()
2005hk \ionHii region 87 Y VLT/FORS1 (R)
2008ae \ionHii region 58 Y Swope/SITe3 ()
2008ge continuum 10237 Y Gemini/GMOS-S ()
2008ha \ionHii region 50 Y Swope/SITe3 ()
2009J \ionHii region 61 Y Swope/SITe3 ()
2010ae \ionHii region 39 Y NTT/EFOSC (R)
2010el \ionHii region 88 Y NTT/EFOSC (R)
2011ce \ionHii region 182 Y NTT/EFOSC (R)
2012Z \ionHii region 311 Y Swope/SITe3 ()
2013gr \ionHii region 48 Y NTT/EFOSC (V)
2014dt \ionHii region 109 N SOAR/Goodman (clear)
2014ey \ionHii region 95 Y Swope/SITe3 ()
2015H \ionHii region 148 Y NTT/EFOSC (V)
  • The \ionHii region and continuum binning are described in \crefsec:hiibinning,sec:contbinning, respectively.

  • If an image of the SN was used to astrometrically align the MUSE cube, the Telescope/instrument (filter) is given. Otherwise offsets from the host given in the discovery IAU Circular were used.

Table 3: Details of spatial binning for the data cubes.

3.2 Stellar continuum fitting

Although our analysis of the underlying stellar continuum is restricted as our regions of interest are emission dominated, and generally without sufficient signal in the continuum to permit a robust analysis from the fitting of absorption indices, we nevertheless must account for the continuum in order to determine accurate emission line results. In order to fit the stellar continuum we used the spectral synthesis package starlight (Cid Fernandes et al., 2005). A minimisation of the (model data) residuals is performed by summing contributions of single stellar populations (SSPs) after masking regions dominated by emission lines and assuming a dust screen following the same reddening law we assume for Galactic extinction. SSP base models were taken from Bruzual & Charlot (2003)8 for the MILES spectral library (Sánchez-Blázquez et al., 2006) using a Chabrier (2003) stellar initial mass function (IMF) over the range 0.1–100 M. The components for the base models comprised 16 ages from 1 Myr to 13 Gyr for each of 4 metallicities (Z ). We note that tests with somewhat different base model sets and choice of IMF did not significantly affect continuum subtraction and so do not affect the emission line results on which our analyses rely. Our procedure largely mirrors that performed on MUSE data of supernova hosts elsewhere (e.g., Galbany et al., 2016a).

3.3 Emission line fitting

Emission line fitting was performed on the spectra after subtracting the best-fit continuum, as determined in \crefsec:contfitting. Prior to fitting, the continuum-subtracted spectra were median filtered with a width of 120 Å – this width was chosen in order to remove any broad residuals left by the imperfect continuum fitting, whilst leaving the narrow emission lines unaffected. The emission line model was a composite model of single gaussians initially centred at the wavelengths of H, [O iii] 4959,5007, H, [N ii6548,6583 and [S ii] 6716,6731 (the two exceptions to this set up, SNe 2002bp and 2008ge, are detailed later). The fitting was performed via the weighted Levenberg-Marquardt least-squares algorithm. Such fitting is however susceptible to finding local minima solutions that are dependent on the initial guesses for the gaussian parameters. In order to circumvent the dependency on initial guesses, and to restrict to physically plausible solutions we added the following caveats and steps to our fitting routine:

  1. The velocity offset of the gaussian means were restricted to within 500 km s relative to the galaxy’s rest-frame (i.e., after accounting for the redshift of the galaxy nucleus). This comfortably encloses the limits of rotational velocities seen in late-type galaxies (Sofue & Rubin, 2001) whilst preventing misidentification of lines.

  2. The velocity offset of the Balmer and forbidden lines were each tied, i.e a single velocity offset for H and H was fit and a separate offset was fit for the other lines.

  3. The standard deviations of the gaussians were restricted to  km s and pairs of lines for [O iii], [N ii] and [S ii] had tied values for the standard deviation (in km s).

  4. In order to semi-brute force locate the global minimum, a list of several initial guesses for each of the means, standard deviations and amplitudes to be fit were made. A grid was then formed of all possible combinations of these parameters to use as initial guesses for the Levenberg-Marquardt fitter, which found the local minimum for each initial guess. The residual between the data and model was used as an estimator of the goodness of fit for each grid position and used to identify the best overall fit.

Where the fitting process failed, we manually inspected the emission line spectrum. In most cases, low signal-to-noise or non-detection of the emission features was the cause and we did not consider these bins in our further analysis. Another source of failure was the nuclear bins of some hosts that harbour an active galactic nucleus (AGN). AGN can have a variety of spectral morphologies including broad emission line components. Their presence prevents much of our emission line analyses of the host nuclei bins since metallicity and age indicators are calibrated based on ionising radiation from recently-formed young stars, these cases were identified as described in \crefsec:ionising.

Observations of UGC 6332, the host of SN 2002bp, were taken under relatively poor sky conditions, and its redshift means that the [S ii] lies in a region of telluric absorption. We attempted to correct this with molecfit (Smette et al., 2015) using the galaxy nucleus (as there are no bright stars in the MUSE field of view for this host). However we were not able to clean the spectrum sufficiently to recover the line satisfactorily. Instead we opted to set the flux of [S ii] equal to  [S ii] for each bin. This was chosen as the relation provided a good fit to the vast majority of all other bins in our MUSE sample. We assigned a factor of two uncertainty in the estimated flux, which encompasses the range of intensity ratios for this doublet for typical \ionHii regions (Osterbrock & Ferland, 2006).

As mentioned, the host of SN 2008ge is an S0 galaxy with the consequential expectation of weak or no emission lines from SF. We found after a initial run with the full emission line list detailed above, that the fitting routine was failing as it relies on finding a fit simultaneously for each line. Thus, although inspection of certain bin spectra showed conspicuous emission in stronger lines (e.g. H, [N ii]), the fits were failing due to non-detections of the weaker lines. We opted to fit this observation for H, H and Nii 6548,6583 only, which allows us to still estimate E() and determine N2 metallicities for the small number of bins where these were detected. This host galaxy is discussed further in \crefsec:08ge.

The fitted gaussian parameters were used to determine the signal-to-noise ratio, flux, equivalent width (EW), FWHM and velocity offset of the lines. Uncertainties on these quantities were found by propagating the statistical uncertainties of the fitted parameters and accounting for photon noise in the manner of Gomes et al. (2016b, and references therein). The H and H fluxes were used to determine the Balmer decrement compared to the expected ratio of 2.86 (assuming Case B recombination,  K and  cm, Osterbrock & Ferland 2006) and give the reddening of the gas component, E(), again adopting a Cardelli et al. (1989) reddening law. This was used to correct the flux and EW values for the effects of dust extinction. Although assumptions for the physical parameters of the gas are inherent to this correction, they are representative of observed \ionHii regions. Furthermore, our metallicity measurements, for example, rely on emission lines nearby in wavelength and as such are largely insensitive to reddening.

Luminosity measurements have their statistical uncertainties quoted, however the systematic uncertainty due to the distance of the hosts is a dominating source in most cases. For example, a 500 km s peculiar galaxy velocity at the median redshift of our sample ( = 0.0165) equates to an uncertainty of dex in (H).

Ionising source

We created a BPT diagram (Baldwin et al., 1981) for each host, in order to discard bins where the emission lines are not driven by the radiation from young, hot stars, such as regions around AGN, and thus where measures of line strengths and ratios are not appropriate tracers of metallicity, SFR etc. We used the redshift-dependent classification criterion for regular \ionHii regions being powered by SF of Kewley et al. (2013). The theoretically-determined limit on the ionisation ratios from pure SF is also used (Kewley et al., 2001). We analyse any region below the theoretical SF limit as the region is still consistent with being driven by SF, although there may be smaller contributions from other sources of ionisation. Above this limit we consider the ionising source of the bin to no longer be dominated by SF and do not include these in our results.

An example for the host of SN 2009J, which harbours an AGN and strongly star-forming regions, is shown in \creffig:bptmap. Those bins that were found to be above the theoretical SF line (shown in this plot) were discarded from further analysis as they are powered primarily via other sources of ionisation.

Figure 2: An example BPT map and diagram shown for IC 2160 (the host of SN 2009J). Left: Heatmap classification of the ionising source (SF: star-formation, CP: composite, AGN: active galactic nucleus) of each bin determined using the relations of Kewley et al. (2001, 2013, see text), overlaid on a H map of the galaxy; North is up, East is left. The location of the host nucleus and the SN explosion site are indicated by the black and ‘star’ symbols, respectively. Right: The classification limits for each ionising source shown for [N ii]/H and [O iii]/H line ratios with the location of each bin in this parameter space.


Many emission line metallicity abundance indicators are present in the literature, with significant systematic differences in results between methods, in particular based on whether the method is empirical or theoretically motivated (e.g. Kewley & Ellison, 2008). The wavelength range of MUSE for low-redshift galaxies covers many emission lines used in various strong emission line methods for determining the gas-phase metallicity, although it does not extend sufficiently into the blue to cover the most commonly used temperature-sensitive lines – a thorough discussion of metallicity determinations with a view to MUSE data of low redshift galaxies is given in Krühler et al. (2017). We concentrate here on determining metallicities on the scale of relative oxygen abundance and use the theoretically-motivated calibration based on photo-ionization models presented by Dopita et al. (2016, hereafter D16). This indicator uses the [S ii] lines as well as H and [N ii] and is thus well-suited to MUSE observations of \ionHii regions in the local Universe. The relation is given as:




This is a good description of the theoretical results over a wide range of metallicities from very sub- to super-solar. Although this indicator is robust to changing ionization parameter (D16), it was noted by Krühler et al. (2017) that there exists an apparent  dex systematic offset to lower abundances with this indicator compared to -based values.

We also present values based on the O3N2 () calibration of Marino et al. (2013, hereafter M13)9, primarily to facilitate comparison with literature values for the environments of other transient types. There is a 1 uncertainty of 0.18 dex associated to measurements with this indicator due to the observed spread of -based abundances about this relation. The indicator is given over the range dex. Where literature values were presented in the O3N2 calibration of Pettini & Pagel (2004), we convert these to M13. We adopt a binning regime that means our effective spatial scale is similar to the size of \ionHii regions, and we are therefore not as sensitive to significant variations of the ionisation parameter that can occur within individual regions on small spatial scales. For ionisation parameter-dependent abundance indicators, such as O3N2, this creates strong gradients within individual regions (Krühler et al., 2017).

Although we correct our emission line fluxes for reddening (E() estimated by the Balmer decrement), we note that since the line ratios used in the above metallicity indicators are relatively nearby in wavelength, we are not strongly affected by uncertain reddening values.

We determined metallicity gradients for each host with a linear fit to the bin metallicities and deprojected galactocentric distances. Deprojected distances were found following the method of Andersen & Bershady (2013) to determine the position angle and inclination from the velocity map of the H line and then normalised by the (25th -band mag arcsec surface brightness radius) value of the host. values were taken from NED. For SN 2008ge the velocity map used was that of the stellar continuum as determined by starlight. For SNe where we do not directly detect emission lines at the location of the SN we use these metallicity gradients to estimate the metallicity of the region local to the SN. The uncertainty on this estimate is taken as the rms of the observed metallicity values about the linear fit.

An example of a metallicity map and gradient is shown in \creffig:d16map and stamps for all hosts (except SN 2008ge) are given in \crefapp:zstamps.

Figure 3: A gas-phase metallicity map for NGC 1309 (the host of SN 2012Z). Metallicities are given on the oxygen abundance scale using the relations of D16. Left: A heatmap of the metallicity of each bin overlaid on a H map of the galaxy; North is up, East is left. The adopted locations of the nucleus and the SN explosion site are indicated by the black and ‘star’ symbols, respectively. Right: A cumulative distribution of the metallicity in the galaxy weighted by the SFR – i.e. – of each bin (top) and the deprojected metallicity gradient of the galaxy (bottom). The SN explosion site metallicity is highlighted in these panels also.

SF rates

We use Kennicutt (1998) to convert our extinction-corrected (H) values (determined as in \crefsec:emisfitting) into SFRs with the relation:



Ages for the youngest SP in galactic regions and at SN explosion sites are often made based on the EW of (primarily Balmer) emission lines in comparison to theoretical predictions. Such mappings between EW values and ages are subject to significant uncertainty due to potentially unaccounted-for effects (stellar multiplicity, SN feedback) and the large spread in physical characteristics of nebular gas in star forming regions (electron density, physical size). With these caveats in mind, it is nevertheless widely accepted that the EW of nebular emission lines should have some inverse relation with the age of the youngest SP for an instantaneous or rapidly-declining SF history for the region. We used results from the Binary Population and Spectral Synthesis code, bpass (e.g. Eldridge & Stanway, 2009, 2012), and processed the stellar continua with cloudy (Ferland et al., 1998). This was done with bpass version 2.1 (Stanway et al., 2016, Eldridge & Stanway, in prep) in the manner described in Stanway et al. (2014) for a fiducial nebula gas model of 10 cm hydrogen gas density in a spherical distribution with inner radius of 10 pc. From this we obtained the evolution of EW measurements for H with age for instantaneous SF episodes, which is shown in \creffig:bpassew. These were compared with our measured EW values at the explosion sites to provide estimates of the age of the youngest SP at these locations. We note here the strong effect of including binary stellar evolution for these calculations, which act to strengthen the emission relative to the continuum, beginning after several Myr. A binary fraction would result in evolution somewhere between the two cases shown in \creffig:bpassew. A pertinent study of the effect of including binary stellar evolution when determining age estimates of CCSN-hosting regions has found that regions are generally older than previously thought, thus revising progenitor initial mass estimates down (Xiao et al., 2017, Xiao et al. in prep)

Figure 4: The evolution of EW(H) for an instantaneous SF episode calculated with bpass and cloudy (see text). The IMF used has a power-law slope of over the range 0.5 to 100 M. We stress here that this is for a single fiducial gas model of assumed size and density and that the physical properties of the gas distribution can affect results significantly (see Stanway et al., 2014, Eldridge & Stanway, in prep). Different metallicities are represented by colours with binaries and single populations as solid and dashed lines, respectively.

Pixel statistics

The location of transients within the light distribution of their hosts has been used to infer further constraints on the nature of the progenitor systems (see Anderson et al., 2015, for a review). We use the Normalised Cumulative Rank (NCR) method presented in James & Anderson (2006): briefly, pixel values for an image of a transient host are sorted, cumulatively summed and then normalised by the total sum of the values. The location of the transient’s explosion site in this cumulative sum provides the fraction of light in the host at a level lower than the intensity at the explosion site. When using an appropriate SF tracer, such as H, the fraction of SF in the host below the level of the explosion site can be inferred; a distribution of H NCR values describes the association to SF for a sample of transients.

The NCR analysis has been presented for an initial sample of SNe Iax in Lyman et al. (2013). We used the same H maps constructed from our MUSE data cubes that were used to create our spaxel bins (\crefsec:hiibinning) in order to calculate NCR values for a larger sample of SN Iax. The maps were binned 33 to circumvent astrometric uncertainties before applying the NCR method. Where we had an existing NCR value from Lyman et al. (2013) and our MUSE observation did not cover the extent of the host, we use the pre-existing value.

We further introduce a metallicity cumulative rank in \crefsec:reszrank, which we use to assess the presence of metallicity bias in the production of SNe Iax within their host galaxies.

4 NOT data analysis and methods

Figure 5: An example of the analysis performed on our NOT observations for UGC 1778 (the host of SN 2007J). Top left: The open filter acquisition image. Overlaid are the 25th -band magnitude elliptic contour (black solid line), the position of SN 2007J (red star) and the centre of the galaxy (red circle). The slit position is shown, and a color code is used to show the D16 metallicity measurements at the position of each \ionHii region that we inspected. Top right: The flux at H along the slit, with the location of the SN and host nucleus marked. The D16 metallicity measurements are shown at the corresponding positions in the top sub-panel. Bottom: The metallicity gradient of UGC 1778. The linear fit on our D16 measurements is shown with a solid red line. The interpolated metallicity at the SN distance is marked with a empty red square and its uncertainty corresponds to the fit error. The positions of SN and nucleus are marked with vertical dotted lines. The solar metallicity and the LMC metallicity are indicated with horizontal dotted lines.

We here detail the analysis of our NOT observations, highlighting differences from the MUSE analysis. Further details of the analysis methods used can be found in Taddia et al. (2013, 2015a). Prior to analysis, NOT spectra were corrected for Galactic extinction and dereddened in the same manner as the MUSE data cubes. The main differences from the MUSE analysis were the identification of \ionHii regions versus spaxel binning routines and the fitting of emission lines due to a lower spectral resolution. Stellar continuum fitting, identification of the ionising source and our metallicity, SFR and age measurements followed those described in \crefsec:musemethods.

Intrinsic (deprojected) galactocentric offsets of SN explosion sites and other \ionHii regions were calculated and normalised to following the method of Hakobyan et al. (2009); Hakobyan et al. (2012). Host galaxy position angles and inclinations were obtained from NED, supplemented by data from HYPERLEDA10, since we were not able to determine a velocity map as was done for the MUSE data (\crefsec:methmetallicity). Again, these were normalised by the of the hosts. These deprojected distances were used to determine metallicity gradients for the host galaxies where more than two \ionHii regions were extracted. For SNe where we could not extract a \ionHii region underlying the explosion site we used these gradients to estimate the metallicity of the SN. Statistical uncertainties on these interpolated or extrapolated estimates were determined based on the uncertainties of our linear gradient fit (which generally dominate over the rms of \ionHii regions about the relation, whereas the opposite is true for MUSE, \crefsec:methmetallicity).

Since we only have long-slit spectra for these targets, we do not perform our pixel statistics, as done for the MUSE sample (\crefsec:pixelstat).

For two observations, those of UGC 12182 (SN 2014ck) and SDSS J020305.81-035024.5 (SN 2016ado) we were not able to perform our main emission line-based analysis at any location within the slit due to poor observing conditions. We remove these two events from our sample when discussing our spectroscopic results as they provide no meaningful measurements or limits (however we include the host-normalised offset of SN 2014ck, ).

For the hosts of SN 1999ax and PS 15aic we were only able to extract emission at the host nuclei.11 As we have no additional spatial information to be gained from the NOT long-slit spectra, we opted to use existing SDSS spectra which offer improved spectral resolution and depth on which to perform our analysis.

For SN 2009ku and PS 15csd we take the host nucleus measurements as those of the SN explosion site owing to the small apparent size of the hosts and the very small offset of the SNe.

Within our NOT-observed sample we have two duplicates of MUSE-observed events: SNe 2002cx and 2015H. We found very good agreement (within 1) between our results obtained for the explosion site of SN 2002cx and host nuclei of each (the explosion site of SN 2015H was not detected in emission lines in the NOT data) for the two sets of data. We present our results for these SNe based on the MUSE data results since these are subject to smaller statistical uncertainties owing to greater signal-to-noise of the emission lines and increased spectral resolution, and also allowed for more robust determinations of the metallicity gradients as more \ionHii regions could be analysed.

4.1 Hii region identification

We performed an inspection of each acquisition image as well the H line flux trace in order to identify \ionHii regions from which to extract spectra (i.e. peaks in the flux trace). An example of this procedure is shown in the top of \creffig:notanalysis, extraction widths were ad hoc in order to maximise the signal to noise ratio.

For our observation of SN 2008A the slit did not cover the host nucleus and thus we do not have a measurement of the nucleus for this host.

4.2 Emission line fitting

Emission line fitting followed broadly the same methodology as for the MUSE spectra (\crefsec:emisfitting), however, the lower resolution of the NOT spectra meant we had to use a different fitting routine as nearby lines are blended. Firstly, the H and [N ii] 6548,6583 lines were simultaneously fit with the gaussians of fixed width (determined by the spectral resolution), and fixed centroid offsets using the known wavelengths of the lines. Furthermore, the flux of [N ii] 6548 was fixed as one third that of [N ii] 6583 (Osterbrock & Ferland, 2006) – this assumption was required to allow for a proper fit of this faint line, although it is not used in our metallicity determinations, to remove its contamination from H. Similarly, the [S ii] 6716,6731 doublet was fitted with two gaussians of fixed width and separation. We also attempted fits to H, [O iii] 5007 for each extracted spectrum, however, due to poorer sensitivity in the blue part of the spectrum we were were not able to measure these in several cases.

\arraybackslashGalaxy SN Offseta (H) EW(H) NCR (SN site)b (Host nucleus)c
\arraybackslash [] [ erg s] [Å] D16 M13 D16 M13
MUSE sample
\arraybackslashIC 344 1991bj 1.383 39.400.01 98.93.4 0.29 8.260.01 8.370.01 8.550.01 8.510.01
\arraybackslashUGC 6332 2002bp 0.920 37.130.36 1.81.5 0.00 8.770.25 AGN
\arraybackslashCGCG 044-035 2002cx 0.765 39.190.01 45.81.9 0.33 8.250.02 8.360.01 8.520.01 8.500.01
\arraybackslashUGC 11001 2004cs 0.519 39.230.01 78.62.2 0.72 8.420.02 8.430.01 8.680.01 8.560.02
\arraybackslashNGC 5468 2005P 0.443 37.500.02 9.20.9 0.06d 8.070.13 8.360.06 not in FOV
\arraybackslashUGC 272 2005hk 1.089 37.250.04 20.95.1 0.00 7.850.28 8.340.09 8.390.02 8.450.01
\arraybackslashIC 577 2008ae 0.881 39.060.01 58.65.4 0.07 8.660.03 8.500.02 8.990.04 8.530.01
\arraybackslashNGC 1527 2008ge 0.055 0.00
\arraybackslashUGC 12682 2008ha 0.263 37.910.01 24.91.0 0.41d 7.910.04 8.220.01 7.990.02 8.190.00
\arraybackslashIC 2160 2009J 0.551 38.810.01 34.31.7 0.00d 8.510.03 8.410.02 AGN
\arraybackslashESO 162-017 2010ae 0.070 39.240.01 39.70.8 0.61 8.260.01 8.320.00 8.280.01 8.310.00
\arraybackslashNGC 1566 2010el 0.106 38.660.01 20.40.6 0.03 8.850.02 8.510.02 AGN
\arraybackslashNGC 6708 2011ce 0.136 39.940.01 86.41.9 0.59 8.760.01 8.590.01 8.880.01 8.600.02
\arraybackslashNGC 1309 2012Z 0.766 37.800.01 36.11.9 0.00d 8.260.03 8.410.01 AGN
\arraybackslashESO 114-007 2013gr 0.266 37.780.01 31.01.5 0.06 7.840.05 8.270.01 7.960.03 8.270.01
\arraybackslashNGC 4303 2014dt 0.218 38.780.01 33.00.6 0.04 8.850.02 8.520.02 not in FOV
\arraybackslashCGCG 048-099 2014ey 0.616 38.870.01 47.83.6 0.22 8.360.04 8.430.02 8.670.01 8.570.02
\arraybackslashNGC 3464 2015H 0.426 37.970.02 7.60.4 0.01 8.660.07 8.500.10 8.970.05 8.580.01
NOT sample
\arraybackslashSDSS J140358.27+155101.2 1999ax 7.780.10 8.230.02
\arraybackslashNGC 7407 2003gq 0.133 39.570.01 23.50.9 8.510.07 8.330.02 AGN
\arraybackslashCGCG 283-003 2004gw 0.568 9.230.24 8.540.02
\arraybackslashNGC 5383 2005cc 0.061 38.810.01 28.70.4 8.800.07 8.500.02 8.860.08 8.510.02
\arraybackslashUGC 6154 2006hn 0.498 38.980.01 18.40.6 8.810.17 8.500.02 8.950.05 8.480.01
\arraybackslashUGC 1778 2007J 0.514 39.350.01 116.99.4 8.410.04 8.380.02 8.840.04 8.450.01
\arraybackslashSDSS J020932.73-005959.8 2007qd 0.531 8.540.03 8.460.01
\arraybackslashNGC 634 2008A 0.789 not in FOV
\arraybackslashAPMUKS(BJ) B032747.73-281526.1 2009kue 40.410.01 47.860.82 8.430.06 8.370.01 8.430.06 8.370.01
\arraybackslashNGC 2315 2011ay 0.468 8.690.01
\arraybackslashCGCG 205-021 PS1-12bwh 0.318 8.980.07 8.560.01
\arraybackslashNGC 5936 2013dh 0.219 39.140.01 34.33.7 8.770.05 8.580.01 9.180.03 8.630.01
\arraybackslashUGC 11369 2013en 0.480 39.320.01 20.70.7 8.480.09 8.460.03 9.050.02 8.630.01
\arraybackslashUGC 12850 2014ek 0.248 39.830.01 34.80.4 8.500.02 AGN
\arraybackslashUGC 12156 2015ce 0.394 8.920.09 8.630.01
\arraybackslashSDSS J133047.95+380645.0 PS 15aic 0.666 39.500.02 2.30.1 8.720.02
\arraybackslashSDSS J020455.52+184815.0 PS 15csde 39.550.02 25.11.0 8.110.17 8.290.03 8.110.17 8.290.03
  • The N2 relation of M13 was used instead of O3N2.

  • Helium detected in SN spectra.

  • The deprojected offset of the explosion site normalised by the value of the host.

  • The abundance measured in the scales of D16 and M13 for the SN explosion site. Uncertainties are statistical only: M13 have a systematic uncertainty of 0.18 dex; D16 do not have well quantified systematics but appear lower than -based estimates (Krühler et al., 2017).

  • As for but for the host nucleus. ‘AGN’ means the host was deemed to host an active galactic nucleus, ‘not in FOV’ means the host nucleus was not captured in the IFU or slit.

  • NCR value taken from Lyman et al. (2013).

  • Due to the SN being overlaid on the nucleus of an unresolved host at very small offset, we use the host nucleus results as those of the SN site also.

Table 4: Results for SN explosion sites and host nuclei.
Galaxy SN name D16 M13
MUSE sample
IC 344 1991bj
UGC 6332 2002bp
CGCG 044-035 2002cx
UGC 11001 2004cs
NGC 5468 2005P
UGC 272 2005hk
IC 577 2008ae
UGC 12682 2008ha
IC 2160 2009J
ESO 162-017 2010ae
NGC 1566 2010el
NGC 6708 2011ce
NGC 1309 2012Z
ESO 114-007 2013gr
NGC 4303 2014dt
CGCG 048-099 2014ey
NGC 3464 2015H
NOT sample
NGC 7407 SN 2003gq
CGCG 283-003 SN 2004gw
NGC 5383 SN 2005cc
UGC 6154 SN 2006hn
UGC 1778 SN 2007J
SDSS J020932.73-005959.8 SN 2007qd a a
NGC 2315 SN 2011ay a
CGCG 205-021 PS1-12bwh a
NGC 5936 SN 2013dh
UGC 11369 SN 2013en
UGC 12850 SN 2014ek
UGC 12156 SN 2015ce
  • The N2 relation of M13 was used instead of O3N2.

  • Only two \ionHii regions were used to determine the gradient. We adopt an uncertainty of 0.15 dex on .

Table 5: Metallicity gradients for SN Iax host galaxies and SN explosion site metallicities. The metallicity gradients for SN Iax host galaxies are parameterised as , where is the deprojected offset normalised by the host , in the scales of D16 and M13. are estimates of based on these metallicity gradients at the offset of the SN, and are our directly measured values for the \ionHii underlying the SN position.

5 Results

We present our main results of the SNe explosion sites and host nuclei in \creftab:results. We indicate those observations where the host nucleus was not captured in the FOV or where the ionising source was deemed to be powered by an AGN (\crefsec:ionising,fig:bptmap). Individual line fluxes for the explosion sites (where they could be measured) are given in \creftab:fluxes.

5.1 Metallicity

Gradients and estimating explosion site metallicities

In some cases we were not able to directly measure emission lines at the explosion site of the SN. As discussed in \crefsec:notmethods, in these cases we resort to estimating the metallicity based on the observed metallicity gradient of the galaxy as has been done previously for other studies of SN environments.

Here we utilise the power of the MUSE data, in combination with the results of NOT to assess the reliability of this method in the extreme regimes of many and few \ionHii regions from which to determine the gradient. In \creftab:zgrad we present our determined metallicity gradients alongside the estimated (based on the gradient) and measured metallicities of the SN explosion sites, which are plotted in \creffig:Zgrad.12 The estimates agree very well with the measured values – the one outlier in the figure is SN 2005P, although this is still only at the level and interestingly only in the D16 indicator. We crucially also see that even with a low number of regions from which to fit a metallicity gradient (in the case of NOT data), the measured and estimated values agree well, offering further support of results derived using this method. The associated uncertainties (determined from the residuals of the measured \ionHii region metallicities about the gradient or based on the uncertainties of the linear fit; \crefsec:musemethods,sec:notmethods), are generally larger than the statistical uncertainties on the measured values, but are still mainly dominated by the systematic uncertainties on the calibrations. We note that these findings corroborate those of Galbany et al. (2016b) who investigated a number of local metallicity estimators in IFS data, finding that an interpolation of the metallicity gradient is a robust estimate.

For some observations we were not able to either measure the metallicity at the SN explosion site or determine a metallicity gradient. In the cases of SNe 1999ax, 2009ku, PS 15aic and PS 15csd we were only able to measure the host nucleus metallicities, and for SN 2008A we could only measure a single bright \ionHii region slightly offset from the host nucleus13. SN 2009ku and PS 15csd lie on top of their respective hosts which are of small apparent size and so we adopt the host nucleus metallicities as those of the SN (\creftab:results). Our determined metallicity gradients cover a large range, making any estimates for SNe 1999ax, 2008A or PS 15csd based on an average gradient of limited use as the associated error would be very large (Galbany et al. 2016b also caution against this methodology). We are limited to supposing their metallicities are less than or equal to their respective host nucleus metallicities (for SN 2008A our sole \ionHii region metallicity is at an offset of / = 0.168 with a metallicity of dex, dex). This does not introduce the requirement for especially low metallicities (compared to the rest of the sample) for these events.

Given the very good agreement between our direct explosion site metallicities and those estimated based on the hosts’ gradients, we supplement our explosion site metallicity values with gradient estimates where appropriate, and present and discuss this enlarged sample in our results.

Figure 6: A comparison between the metallicity determined directly at the location of SN explosion sites () and that which would be estimated based on the metallicity gradient alone (). Metallicities are determined using the calibrations of D16 and M13 in the top and bottom panels, respectively. Empty markers in the lower panel indicate the N2 relation was used (versus O3N2 for filled markers). Uncertainties on the measured metallicities are statistical only. The gradient uncertainties are either the uncertainty on the gradient linear fit, or the root mean square of \ionHii region metallicity residuals about the gradient (see text). Orange dashed lines indicate the one to one relation for each.


In \creffig:Zall the metallicity distribution of SNe Iax explosion sites is shown alongside that of the host nuclei and all \ionHii regions extracted in our MUSE sample14. The cumulative weight of each \ionHii region in this plot is given by its H luminosity, i.e. SFR, in order to show the cumulative distribution of SFR with metallicity in SN Iax hosts. Although we only have global metallicity determinations for the MUSE half of our sample, the main discriminating factor between this and the NOT sample is only declination of the source, and so we expect no bias in terms of host properties. Additionally, a few of our MUSE observations do not cover the entire hosts, but they do cover areas around the same location and offset as the SNe, and cover reasonable fractions of their respective hosts. As a check, we also plot in \creffig:Zall the metallicity distribution of SF in our MUSE sample only including observations that cover approximately all the host galaxy (i.e., we exclude observations of SNe 2005P, 2010el, 2012Z, 2013gr and 2014dt; see \creffig:zstamps). The distribution for this sub-sample of our MUSE galaxies is almost identical to the full MUSE sample (as was found in our other environmental measure also) and so the partially covered hosts do not introduce significant biases into the distributions.

We find that the metallicity of SNe Iax sites are at lower metallicities than where typical SF is occurring in the hosts (in each metallicity indicator shown the KS test gives that the two are drawn from the same distribution). As expected, given the SN explosion sites are offset from their host nuclei and we find mostly negative radial metallicity gradients in our hosts, there is an offset between the host nuclei and explosion site metallicities.

Figure 7: The metallicity distributions of SNe Iax explosion sites, their host nuclei, and all detected Hii regions in the MUSE sample. Metallicities are determined using the calibrations of D16 and M13 (O3N2) in the top and bottom panels, respectively. When appropriate, SN explosion site metallicity estimates based on the gradients of the hosts are included (see \crefsec:zgrads). Dotted lines are the distributions after adding/subtracting statistical uncertainties to/from all values and are somewhat extreme limits for the true distribution. The cumulative sum of SN explosion sites and host nuclei are unweighted (i.e. a step in these histograms equals one SN or host nucleus, as appropriate). The Hii regions have been weighted by their H luminosity such that this histogram shows the cumulative fraction of all detected SF with metallicity across the MUSE sample. The thin orange dashed line shows the sub sample of MUSE \ionHii regions for which the observations covered the whole host and appears almost indistinguishable from the full \ionHii region sample (see text).

Metallicity ranking of explosion sites

In order to further assess the indications of \crefsec:reszdist, we have calculated SFR-weighted metallicity NCR (NCR) values for the explosion sites of out MUSE sample of SNe Iax. The NCR is calculated in a similar manner to the traditional NCR value of SNe (\crefsec:pixelstat, see also James & Anderson, 2006; Anderson et al., 2012), but the cumulative distribution of the SFR-weighted metallicity in the galaxy is used to determine the rank of each SN. The NCR is thus the fraction of stars being formed at or below the metallicity of the explosion site in that host. This may be seen visually in the top right panel of \creffig:d16map, where the SN in question has NCR = 0.003. In \creffig:ZNCR we show the distributions of NCR for D16 and M13. In the case of metallicity-unbiased formation of the progenitors, a uniform distribution (indicated by dashed lines) should be recovered, i.e. the SNe have no dependence on where in the metallicity distribution of SF they form, and so sample it uniformly. We see for both distributions the SN explosion sites are systematically shifted to lower values than the uniform distribution, indicating the SNe are preferentially exploding in metal poor regions, and not unbiasedly tracing the SF in their hosts. For M13, although the distribution is systematically below the uniform distribution, the difference is more marginal than for D16.

Figure 8: The SFR-weighted metallicity NCR (NCR) values of SNe Iax explosion sites in two metallicity indicators, as observed with MUSE. In the case of metallicity-unbiased production of the progenitors, the SNe should create a uniform distribution in NCR (thick dashed lines). Uncertainties on the distributions were found by creating many realisations of the cumulative sums (based on the uncertainties of the values) for each SN host and recalculating each explosion site’s rank. Dotted lines represent the lower and upper distributions of the 95 per cent confidence interval of the ranks.

5.2 Ages and SFR

We show the EW(H) and L(H) measurements for our explosion site bins in \creffig:HaEWall,fig:Halumall, respectively. As was done for the metallicities, we also show the cumulative distribution of SF in these measurements for our MUSE-observed hosts. The regions of SF at the locations of SNe Iax are occurring at significantly lower EW and luminosities than the overall SF in the hosts (KS test for each). Adopting a H luminosity–SFR relation (Kennicutt, 1998), the H luminosities imply that the explosions sites are typically of lower SFR than that of the population of SF regions in the hosts, ranging from M yr.

As discussed in \crefsec:methages, ages for the young stellar component of star forming regions can be estimated from EW measurements of emission lines but are subject to sources of uncertainty. The bulk of our explosion site EW(H) measurements are tens, up to  Å. We show in \creffig:bpassew (see \crefsec:methages) the evolution of EW(H) with age for bpass models for an instantaneous SF episode at a range of metallicities. As the consensus is that the vast majority of (particularly massive) stars are in some form of binaries (e.g. Sana et al., 2012) we compare results with the binary population models.

Our observed range of EW(H) suggests young SP components of several  to  yr at the locations of SNe Iax. We again note the inherent uncertainties in selecting a single fiducial model for the nebular gas properties, however these values indicate that there are moderately young SP components at the location of the majority of SNe Iax explosion sites.15 Notably, although our EW values are actually somewhat lower limits of the true young SP EW (as there will be an existing underlying, older SP that contributes to the continuum but not emission lines), they are not exceptionally high. This would seem to disfavour very young, and therefore very massive, SPs at their locations with ages of several Myr, since initially very high EW values tend to drop off quickly and largely independent of model differences and reasonable gas parameters.

As an additional check for the presence of very young SPs at the location of SNe Iax we also attempted fitting for He i in our emission fitting routine (\crefsec:emisfitting) for the MUSE sample.16 He i in emission is present only for ages up to a few Myr (e.g. González Delgado et al., 1999). We did not detect this line at any SN Iax explosion site. Within our MUSE data we found detections of He i within 3 kpc of the SN for SNe 2008ha, 2009J, 2010ae, 2012Z and 2013gr. Assuming a 5 Myr age for the young SP (also consistent with EW(H) found in these regions), the average velocities of the SN Iax progenitors would have to have been in excess of 234, 128, 269, 580, 159 km s, respectively, in order to have originated from these regions.

Two events with higher EW(H) measurements at their explosion sites are debated members of the SN Iax sample as they displayed helium and perhaps hydrogen in their spectra. The removal of these events (SNe 2004cs and 2007J) would lean the typical ages of the young SPs at the locations of the rest of the sample slightly higher (\crefsec:hetero).

Figure 9: As for \creffig:Zall but here showing EW(H) of SN Iax explosion sites.
Figure 10: As for \creffig:Zall but here showing L(H) of SN Iax explosion sites. The top axis indicates the corresponding SFR (Kennicutt, 1998).

5.3 Offsets

Offsets of SNe Iax and \ionHii regions are shown for kpc and host-normalised values in \creffig:Offsetall. SNe Iax appear to trace a similar offset distribution as the overall SF of their hosts when accounting for the varying sizes of the host galaxies, although shifted systematically to slightly larger offsets. Since the outer regions of late type galaxies are likely to be more metal poor and less intensely star forming, this may be a contributing factor to the difference we observe in our other measurements.

Figure 11: As for \creffig:Zall but here showing the deprojected galactocentric offsets of SN Iax explosion sites. The offsets are given in kpc and normalised to the host galaxies’ values in the top and bottom panels, respectively.

6 Discussion

6.1 Correlation between environment and SN properties

Tying the diversity of progenitor systems to the observed properties of SNe is an open issue amongst all SN types. It may be expected that some imprint of the nature of the progenitor (which can be determined via direct detection or, as here, inferred through environmental analysis) is evident in the SN light curve and spectra.

To investigate this for SNe Iax we plot SN properties, where they have been measured in the literature, against our environmental measures in \creffig:snprops. The light curve peak () and decline rate () in - or -band, as well as an estimate of the photospheric velocity around peak light () are shown versus the D16 metallicity, EW(H) and the host-normalised offsets. Values and references for SN properties are provided in \creftab:litsn. In order to expand our comparison sample of SN properties we include preliminary analyses for SNe 2010el, 2013gr and 2014ey (Stritzinger et al., in preparation)17 – these objects are to be the subject of more detailed studies in preparation, but our values here are representative.

SN name Filter Reference
2002cx 0.540.06 -17.640.15 5600 Foley et al. (2013)
2003gq 0.710.10 -17.370.15 5200 Foley et al. (2013)
2004cs 1.110.07 -16.550.45 Magee et al. (2016)
2005cc 0.650.01 -17.130.15 5000 Foley et al. (2013)
2005hk 0.700.02 -18.070.25 6000 Stritzinger et al. (2015)
2007qd 2800 Foley et al. (2013)
2008A 0.510.01 -18.230.15 6400 Foley et al. (2013)
2008ae 0.710.13 -17.760.16 6100 Foley et al. (2013)
2008ha 0.970.02 -14.410.15 3700 Foley et al. (2010b)
2009J 0.790.05 -15.470.22 2200 Foley et al. (2013)
2009ku 0.250.03 -18.700.15 3300 Foley et al. (2013)
2010ae 1.010.03 -14.590.81 5500 Stritzinger et al. (2014)
2010el 1.080.03 -15.430.50 3000 This work
2011ay 0.440.14 -18.430.19 5600 Foley et al. (2013)
2012Z 0.660.02 -18.600.09 8000 Stritzinger et al. (2015)
PS1-12bwh 0.600.05 -17.690.24 5700 Magee et al. (2017)
2013gr 0.990.20 -15.170.50 5300 This work
2014ck 0.580.05 -17.290.15 3000 Tomasella et al. (2016)
2014dt 4100 Foley (2015)
2014ey 0.620.02 -18.130.10 5000 This work
2015H 0.690.04 -17.270.07 5500 Magee et al. (2016)
PS 15csd 1.060.06 -17.750.06 Magee et al. (2016)
  • Based on a similarity to SN 2002cx post-peak.

Table 6: Values for literature SN properties.

From these plots we see that brighter ( mag) members appear to cover almost the full range of the metallicity, EW and offset distributions as found for the full sample, with large spreads in each parameter. Although we have fewer fainter members with peak absolute peak magnitude determinations, we find none at large galactocentric offsets. This trend with galactocentric offset cannot be attributed to observational biases as these work in the opposite direction (very faint transients are more difficult to detect in the brighter central regions of galaxies) and so may actually be more pronounced than is shown. The faint members appear to cluster at a very small range in EW(H) of 25–40 Å, perhaps indicating more strict age constraints for these members, assuming they arise from the local young stellar population. Their metallicities appear diverse although most appear quite metal-poor.18

Evidence has been presented for a relation between and for SNe Iax, but any relation is certainly less tight than that seen for normal SNe Ia and with notable outliers (e.g. McClelland et al., 2010; Narayan et al., 2011; Foley et al., 2013). The slowly fading ( mag) members of our sample occupy a wide range of environments, with the faster declining members generally occupying more restricted ranges, analogous to the trends seen for the bright and faint members. In particular we observe distinct clustering in metallicity with – slowly declining members are systematically more metal-rich than the faster decliners (barring the fast declining, metal-rich SN 2010el). A composite figure showing the and parameter space for SNe Iax now coded by the explosion site metallicity is shown in \creffig:mm15. SN 2004cs (based on unfiltered imaging) was a fast declining member although somewhat brighter than the other fast-decliners in this plot, and shows a significantly larger EW(H) than the other fast decliners (or any of the sample for which we have light-curve information). This was one of two proposed SN Iax members that showed He features, and is further discussed in \crefsec:hetero.

We do not observe any strong clustering or correlated behaviour between our environmental measures and the estimates of photospheric velocity near peak. We caution here, however, that these values have been determined using a variety of methods (spectral synthesis, line measurements based on differing elements) at slightly varying epochs around maximum light, and there thus may be some systematics within the sample. We assign 1000 km s uncertainties when plotting, which are likely to be overestimates for individual measurements but account better for potential systematics arising from differing methods.

Figure 12: The properties of SNe Iax and their environments. and values are given in the - (red squares), -band (red circles) or unfiltered (for SN 2004cs, empty squares). values are estimates of the photospheric velocity.
Figure 13: - and -band versus for SNe Iax (e.g. Narayan et al., 2011; Foley et al., 2013; White et al., 2015; Tomasella et al., 2016) in our sample, now colour coded by their environmentally-derived metallicities in the indicator of D16 where possible. SN 2004cs is highlighted as the classification of this event is debated (\crefsec:hetero).

6.2 SNe Iax environments in the context of other transients

Despite sharing some similarities to SNe Ia in the general spectral classification sense, the host galaxies of SN Iax are almost exclusively (barring SN 2008ge) late type and star-forming (e.g. Perets et al., 2010; Foley et al., 2013) and the transients appear to be associated with regions of ongoing SF (this work; Lyman et al., 2013). This is the case for CCSNe and thus a comparison between our results here and those of other SN types in the literature can inform on similarities or differences in the progenitor environments.

The construction of our comparison SN samples do not constitute unbiased, blindly-targeted events by any means. However, the same holds for the SN Iax sample, which were discovered over a variety of surveys. Our comparison is thus limited to an initial, indicative comparison until such time as a reasonable sample of homogeneously discovered SNe Iax exists.


As we are working with literature samples, there are almost no abundance measurements in the D16 scale owing to its recent inclusion in the literature. We therefore collect available values based on the O3N2 indicator (these were generally presented in the calibrations of Pettini & Pagel 2004 and have been translated to those of M13). We used only values based at or nearby the SN explosion site, or based on determined gradient values (i.e. we exclude those where the metallicity was determined based on the host nucleus but the SN was at a significant offset and not covered by the slit/fibre).

We take values from the analyses of Anderson et al. (2010); Modjaz et al. (2011); Leloudas et al. (2011); Sanders et al. (2012); Kuncarayakti et al. (2013a, b); Stoll et al. (2013) and Galbany et al. (2016a). In the case of SNe II (including both II-L and II-P) we take values from Anderson et al. (2016, which includes the results of ) and Galbany et al. (2016a) directly in M13, and values from Stoll et al. (2013). On the values of Anderson et al. (2016) we impose a distance cut of 3 kpc on the \ionHii region—SN distance. We present SNe Ib and Ic separately, however uncertain classifications mean there is likely to be some cross-contamination. Where appropriate we use the updated classifications of Modjaz et al. (2014); Shivvers et al. (2017) and count still uncertain ‘Ib/c’ designations with half weight in each cumulative distribution.

We plot the explosion-site metallicity distributions for SNe Iax and our comparison samples in \creffig:Zcompare. There is no significant difference between SNe Iax explosion-site metallicities and those of SNe Ib, Ic, II or IIb and the median value for SNe Iax (8.41 dex) is close to that of the others (Ib: 8.42 dex, Ic: 8.49 dex, II: 8.47 dex, IIb: 8.39 dex). The distribution for SNe Iax covers broadly the same range as these other SN types, perhaps not extending as metal-rich or -poor as SNe Ib, Ic or II, however there are only a very small number of events in these samples on the extreme edges of their distributions. Compared to the SNe Ia distribution (local metallicity measurements explosion sites from IFS data taken from Galbany et al., 2016a), median = 8.53, we see SNe Iax are metal-poor (KS test Ia vs Iax metallicities ).

Although the overall SN Iax distribution appears to follow broadly the distributions of other well-known SN types, barring SNe Ia, it is overall quite different from the distribution of low redshift LGRB explosion sites, which are in the range dex (e.g. Modjaz et al., 2008; Sanders et al., 2012). Similar very low metallicities have also been found for the the vast majority of super-luminous supernova hosts (e.g. Lunnan et al. 2014; Leloudas et al. 2015).

Figure 14: The explosion site metallicity distributions of SNe Iax compared to literature samples of other principal SN types. Metallicities are based on the O3N2 indicator of M13.


-normalised offset distributions for SNe Ib, Ic, II and IIb were taken from Taddia et al. (2013). SNe Ia offsets were taken from the catalogue of Hakobyan et al. (2016) and calculated following the prescription described therein. The SNe Ia host-normalisation are based on -band measurements as opposed to -band measurements used for the other samples. This may introduce a systematic offset in the values, however, given the considerable filter overlap and the modest variation of apparent galaxy sizes around these wavelength ranges (Vulcani et al., 2014), the effect is likely to be small. SNe Ia in the catalogue are those that exploded in disk galaxies (types S0-Sm).

Similarly to the metallicities, we see in \creffig:offsetcompare that the offsets of SNe Iax closely match other SNe types, in particular SNe Ia and Ib. Offsets do not appear to be a strong factor in distinguishing SNe Iax, whereas their host galaxy type distribution proves more discriminatory against that seen for SNe Ia (Perets et al., 2010; Lyman et al., 2013).

Figure 15: The explosion site galactocentric offset distributions of SNe Iax compared to literature samples of other well-known SN types. Offsets have been normalised by their hosts’ respective values.

Pixel Statistics

An analysis of the NCR (\crefsec:pixelstat) distribution of SNe Iax was presented by Lyman et al. (2013), who found the association of SNe Iax to SF in their hosts was at a similar level to that of SNe IIP and inferred similar progenitor ages (tens of Myr). As we are able to create H maps from our MUSE data cubes we present an updated comparison figure with this extended combined sample of SNe Iax in \creffig:ncr. Other SN sample data are taken from Anderson & James (2008); Anderson et al. (2012, 2015) and, following Lyman et al. (2013), we have corrected the SN Ia sample to account for the fact that only late-type galaxy hosts were used but 27 per cent of SNe Ia explode in early-type hosts (Li et al., 2011), with no appreciable ongoing SF. For SNe 2005P, 2008ha, 2009J and 2012Z we use values from Lyman et al. (2013) as H imaging presented there covers the full spatial extent of the hosts. The NCR values for these determined in Lyman et al. (2013) and here are [0.06, 0.41, 0.00, 0.00] and [0.01, 0.33, 0.06, 0.03], respectively.

Following Lyman et al. (2013), we confirm for a larger sample that SNe Iax display a level of association to the SF of their hosts most similar to that of SNe II. Their distribution is formally discrepant, based on the KS test, with the SNe Ia and Ic distributions at (considering only SNe Ia in late-type hosts and excluding SN 2008ge from the SN Iax sample, the SN Ia vs Iax discrepancy is ).19 Although there are few H NCR values for LGRBs (due to their typically much larger distances), studies have shown these to be very strongly associated to the brightest SF regions of their hosts, at a level exceeding SNe Ic (Fruchter et al., 2006; Svensson et al., 2010; Lyman et al., 2017). Assuming the same would hold for SF as traced by H, this infers drastically different environments of SNe Iax compared to LGRBs.

We note that although we detected H at the locations of all our SNe Iax in the MUSE sample, we have some where NCR = 0.00. In order to facilitate comparison to literature samples, which were performed using narrow band imaging, we construct H narrow band images (filter width 30Å) and subtract the neighbouring continuum within our data cubes (\crefsec:hiibinning). As such, very faint sources of emission may become dominated by shot noise or inaccuracies in continuum subtraction during the H map construction, leaving the pixel within the noise floor of the image (defined as NCR=0, see James & Anderson, 2006; Anderson & James, 2008, for more thorough discussion). Additionally, the spatial extend of our spaxel bins is larger than the binning employed for the NCR calculations in order to replicate the method used in the literature.

Figure 16: The explosion site Normalised Cumulative Rank (NCR) distributions of SNe Iax compared to literature samples of other well-known SN types. NCR values were calculated using H maps from our MUSE data cubes and supplemented with values presented in Lyman et al. (2013).

6.3 Ionised gas in the early-type host of SN 2008ge

Although there are likely to be contributions from different progenitor channels to such a diverse class of objects (see \crefsec:hetero), one SN in the Iax sample has been marked out by its unique environment and host. SN 2008ge was hosted by NGC 1527, an early, weakly-barred S0 galaxy (de Vaucouleurs et al., 1991), in contrast to the strongly star-forming, late-type hosts of other known Iax (Perets et al., 2010; Lyman et al., 2013). Motivated by this, Foley et al. (2010a) investigated the host galaxy and explosion site for signs of a young SP. They concluded from a smooth galaxy profile, non-detection of far-infrared or HI 21cm emission in the host, and the lack of narrow nebular emission lines in their host galaxy spectrum that there is no significant ongoing SF (and thus young SP) in NGC 1527.

Our MUSE observations provide deep optical spectra, spatially resolved, across a reasonable fraction of the central regions of the host galaxy. After an initial pass with our emission line fitting as presented in \crefsec:musemethods, we reran the emission line fitting this time fitting only for H and [N ii6548,6583 in order to determine a subset of the bins with evidence for ionised emission lines. We then attempted to fit H and [O iii5007 for these bins. In accordance with Foley et al. (2010a), we found no significant emission lines at the location or in the close vicinity of the SN explosion site (either within the Voronoi bin directly underlying the explosion site or by extracting circular apertures of varying sizes). However, within our MUSE data we detected an ionised gas stream arcing from the NE of the nucleus with evidence for a further region of emission to the SE of the host nucleus. As NGC 1527 is particularly nearby ( 17 Mpc) the MUSE observations only cover the central 2–3 kpc around the nucleus. We have summed the H flux captured by the MUSE FOV and determine (L(H) = 37.6 erg s and note that this is not corrected for the effects of reddening as we were unable to detect H in many of the bins. For the case that this emission is driven by ongoing SF, the relation of Kennicutt (1998) suggests the detected emission equates to (SFR) = -3.48 M yr. This is consistent with the limit that Foley et al. (2010a) provide for NGC 1527 of (SFR) M yr. The metallicity (for SF-driven ionisation) appears to be around solar to slightly sub-solar based on the N2 indicator (M13). The EW(H) of the bins are Å.

Morphologically the main stream appears somewhat coherent, in a thin parabolic stream about the nucleus. The emission looks different to the more extended, and generally more symmetric, spiral arms seen in some early-type galaxies (ETG; see Gomes et al., 2016a, and references therein). The nature of the ionising sources in ETGs is debated and there appears to be contributions from a variety of phenomena, with different mechanisms dominating in different galaxies (Goudfrooij, 1999; Sarzi et al., 2006, 2010, e.g.). Unfortunately, we are limited in our analysis for the ionising source based on emission line flux ratios as we are close to our detection limits and thus flux ratios are not well constrained. Within \creffig:08ge we plot the BPT diagram for bins where all lines were detected at SNR > 1. Although the position of each bin is rather uncertain within this diagram, there appears to be a general clustering of the bins around or above the maximal star-formation relation of Kewley et al. (2001), indicative that ongoing SF may not be the dominant process driving this emission. (Those bins that are more consistent with being ionised solely due young, massive stars are those SE of the nucleus, not in the main arc). With a comparison to the study of Sarzi et al. (2010), the stronger line ratios and morphology may be more reminiscent of emission due to shocks in the galaxy, although these are not expected to be dominating sources of ionisation in ETGs. Alternatively it may be due to diffuse ionising SPs, such as post-asymptotic giant branch (pAGB) stars.

The ionising flux from pAGB stars can lead to line ratios similar to those seen for low-ionization emission-line regions (LIERs Binette et al., 1994; Sarzi et al., 2006), and the bulk of the bins in \creffig:08ge are located close to LIER areas of the parameter space (e.g. and EW(HÅ, Cid Fernandes et al., 2011). One would, however, expect a pAGB population to be more or less pervasive across the galaxy and follow the distribution of stellar mass. This results in the correlation between ionised line flux and stellar continuum flux seen in ETGs with pAGB ionising sources (Sarzi et al., 2010). A more structured ionised region (as seen here) could therefore be due to variations in gas column density within the galaxy. The ionising output and subsequent EW(H) predictions from various models for pAGB star contributions predict roughly constant low values from to  years (Cid Fernandes et al., 2011), meaning such values do not offer strong constraints on the relative age of the underlying population.

Figure 17: The central ionised gas component of NGC 1527 (the host of SN 2008ge). Left: A heatmap of the metallicity (using the N2 relation of M13) for each bin where both H and [N ii6583 has SNR , overlaid on an inverted -band image of the galaxy; North is up, East is left. The adopted location of the nucleus and the SN explosion site are indicated by the black and ‘star’ symbols, respectively. Left, inset: An inverted, continuum-subtracted [O iii] image of the central region of NGC 1527. Each Voronoi bin has had its best fitting stellar continuum model (\crefsec:contfitting) subtracted. The residual ionised gas component is evident Top right: A BPT diagram for those bins with detected emission. The format is the same as shown in \creffig:bptmap. Bottom right: The observed spectrum (black) with the starlight fit to the continuum (orange, dashed) and the residual emission (purple) around H and [N ii]. This is shown for the SN explosion site, where we do not detect any significant emission lines, and in one of the bins NE of the galaxy nucleus, from which we were able to measure a metallicity. Wavelengths are in the rest-frame of the nucleus of NGC 1527.

6.4 Heterogeneity of the class and contaminants

As is the case for many astrophysical transients, in particular for peculiar and relatively rare objects, membership to a particular class (or even designating a single class) can be contentious. Often the underlying continua of the events properties are at odds with defining distinct regions of parameter space to assign one class or another. This is exemplified in the case of SNe Iax where multiple progenitor channels and explosion mechanisms may be contributing to empirically similar transients.

Two examples in the sample of Foley et al. (2013) showed evidence for helium in their spectra: SNe 2004cs and 2007J. Helium can remain hidden in SN spectra as it is difficult to ionise and so it is possible that helium is present in the ejecta of other members, however at the time of writing these two members remain the only posited SNe Iax to show detections. One potential confusion, as highlighted by Foley et al. (2013), is with SN 2005E-like events (Perets et al., 2010, also known as Ca-rich transients/SNe) – these similarly faint-and-fast events display helium features in their spectra before quickly evolving to unusually Ca-dominated nebular spectra (Kasliwal et al., 2012). Instead White et al. (2015) favour the original classifications of SNe 2004cs and 2007J as CCSNe. In addition to helium they also find evidence for hydrogen in their spectra, prompting classifications of SNe IIb. However, a reanalysis by Foley et al. (2016) argued against the presence of hydrogen in SN 2007J, citing an identification as [Feii] of the same feature in SN 2002cx. Furthermore, they find inconsistencies between the light curve of SN 2004cs and other known SNe IIb. As we are discussing only 2 objects, statistical inferences from environmental measures are weak. This is further exacerbated by the fact that our measures overlap significantly for SNe Iax and SNe IIb (\crefsec:sncompare). For completeness we note EW values at the helium-SNe Iax explosion sites are the among the highest (1st and 4th) of our sample, pointing to the presence of younger SPs at their explosion sites than typical SNe Iax. Their offsets () and metallicities (O3N2 ) are typical of the samples of both SNe Iax and IIb. Their exclusion would revise our estimates of the typical ages of the young SP at SN Iax explosion sites slightly higher, although the quantitative effect is likely to be dwarfed by the inherent uncertainties present in determining ages from nebular gas spectra (Stanway et al., 2014, \crefsec:methages), and would not significantly affect our discussion elsewhere.

SN 1999ax was classified as a “somewhat peculiar” SN Ia by Gal-Yam et al. (2008) at . On the basis of SDSS spectra of a nearby potential host showing it to be at , Foley et al. (2013) re-analysed its spectrum and classified it as a SN Iax based on similarity to SN 2002cx post-peak. We note our NOT spectroscopy of the explosion site also covered another nearby object in SDSS that is at a similar (projected) offset from SN 1999ax as the galaxy at (with the SN being located between the two). We find this galaxy is at based on the H line in the NOT spectrum. As the SN is located directly between the two potential hosts it is difficult to distinguish the likely host from environments alone. The velocities of the features in the spectrum appear low compared to normal SN Ia, and a consideration of the SN at still poses peculiarities to a typical SN Ia classification, as such we retain the SN Iax classification making the lower redshift galaxy the probable host.

6.5 Implications for progenitor models

Pure deflagrations (e.g. Branch et al., 2004; Jha et al., 2006; Phillips et al., 2007; Jordan et al., 2012; Kromer et al., 2013; Fink et al., 2014; Magee et al., 2016), pulsational-delayed detonations (Stritzinger et al., 2015) and helium-ignition (Wang et al., 2014) of WDs have been presented as progenitor models to explain SNe Iax. Although too weak to account for normal SNe Ia, they broadly agree with the low Ni masses and kinetic energies displayed by SNe Iax. There are currently limited predictions as to the progenitor environments for these based on population studies, with the main constraints in the form of delay-time distributions and age constraints.

Any double-degenerate channel (either merger or accretion between two WDs), would require circumstances to explain the young environments since such progenitor systems would be expected to be prevalent in old SPs and early-type galaxies, inconsistent with the locations of SNe Iax as a whole. As such, single-degenerate formation channels are generally favoured for SN Iax.

Meng & Podsiadlowski (2014) found that delay times as low as 30 Myr were possible for massive hybrid CONe WDs accreting material until the Chandrasekhar mass, and would likely produce lower-luminosity events (cf. normal SNe Ia) making them an attractive possibility for SNe Iax. Wang et al. (2013) explore detonation of CO WDs via ignition in a helium envelope that is accreted from a He companion. For the case of a non-degenerate companion their delay times to explosion are  yrs, consistent with the ages of young SPs at the location of most SNe Iax. If the companion is a He WD the delay time can be significantly extended (up to the Hubble time), potentially providing an explanation for SNe Iax in old populations (e.g., SN 2008ge) or those with no detected signs of SF at the explosion site and in the outskirts of their hosts (e.g., SN 2008A). We note that these are for solar metallicity populations, which is metal-rich compared to the SN Iax population (\creffig:Zall). Similarly, Liu et al. (2015b) find that the WD + He star channel best reproduces the young ages of SN Iax explosion sites, but note an extended delay time exists for main sequence or red giant companions. Our age constraints on SNe Iax seem to be in good agreement with predictions of young explosion scenarios of single-degenerate thermonuclear progenitor models.

Faint and fast ‘SN 2008ha-like events’

The faintest, lowest-energy members of SNe Iax could represent a distinct population of events, with the most famous example being SN 2008ha (see Foley et al., 2009, for further discussion of models in light of the physical properties of the SNe). These very faint examples are difficult to produce with models proposed for other SNe Iax (Fink et al., 2014), perhaps indicating a single model is not able to explain the full diversity of all events labelled as such. For the purposes of this section we simply refer to faint ( mag) and (/or) fast evolving ( mag) members of the sample, ignoring a detailed spectroscopic distinction.

Valenti et al. (2009) suggested SN 2008ha and, by extension, other SNe Iax were of core-collapse origin. Our environmental analyses indicate that the fainter members arise from low metallicity regions, quite centrally located on their host galaxies (\creffig:snprops). From an environmental viewpoint, their distributions are largely consistent with those seen for well-known CCSN types. The question of whether they are due to lower mass,  M, stars stripped of their hydrogen envelopes or very massive WR stars, probable progenitors of LGRBs, experiencing significant fall-back, was posed by Valenti et al. (2009). Our analysis of the ages of the youngest SPs here suggests ages of order tens of Myr for their explosion sites (\crefsec:resultsages). This is in agreement with resolved SP studies of explosion sites of nearby SNe Iax (Foley et al., 2014; McCully et al., 2014a) and the comparative association of SNe Iax to H emission in their host galaxies at the level of SNe II (\crefsec:pixelstatres, Lyman et al., 2013). For WR stars, however, we may expect much younger ages,  Myr (for stars with initial masses of  M). Consequently we would expect larger EW values at the explosion sites (e.g. see the low redshift LGRB environment studies of Thöne et al., 2008; Christensen et al., 2008; Krühler et al., 2017; Izzo et al., 2017), and an association to ongoing star formation of their hosts at a level much higher than is seen for SNe Iax (\crefsec:pixelstatres, see also Kangas et al. 2017).

A search for He i in emission (indicative of stellar populations of only a few Myr) in the hosts of our MUSE sample found 5 SNe Iax with such young regions within 3 kpc. This include 4 of the 5 events with peak luminosities known to be  mag plus SN 2012Z (the missing low-luminosity event is SN 2010el). The host galaxies of these fainter members are typically more irregular and exhibit signatures of extremely young stellar populations within them, with EW(H) reaching hundreds of Å. In order to establish a causal link between the SNe and their nearest very young regions, progenitor velocities of hundreds of km s must be invoked (\crefsec:resultsages). To produce such high mass and velocity runaways requires very rare dynamical ejections or unfeasibly large ejection velocities after a binary companion goes SN (Eldridge et al., 2011, see also discussion in Krühler et al. 2017). Our results would therefore disfavour the very massive progenitor scenario (for the SN Iax sample as a whole and the SN 2008ha subset). Moriya et al. (2010) concur, finding that their 25 and 40 M fall-back models cannot reproduce the observations of SN 2008ha, but find better agreement with a 13 M model.

Deflagrations of hybrid CONe have been proposed to explain the weak SN 2008ha by Kromer et al. (2015). Although occurring in Chandrasekhar-mass WDs, the deflagration is not propagated into the outer ONe layer, which results instead in lower energy release and consequently low ejecta masses (cf. CO WD explosions), leaving behind a bound remnant. The ejecta mass in the model is below estimates for SN 2008ha, resulting in too quickly rising and fading light curves, and also affecting the spectral comparison. The delay time to explosion for CONe WDs with helium donor stars has been estimated at few  to  yrs by Wang et al. (2014); Kromer et al. (2015), albeit at solar metallicity which may not be representative of the typical environments of SN 2008ha-like events. This is in agreement with this study and Foley et al. (2014).

Moriya & Eldridge (2016) present an initial study of the light curves and rates for binary evolution channels of stripped-envelope electron-capture (SE-EC) SNe, which may have relation to the low-luminosity CCSN scenario of Valenti et al. (2009). They find broad agreement with SN 2008ha-like events in terms of peak magnitude and their rapidly-evolving light curves, and that the rates of SE-ECSNe are significantly enhanced at low-metallicities. For our 5 events peaking at  mag we determined a wide spread of metallicities (see also Foley et al., 2009; Stritzinger et al., 2014), including very low metallicity events and the quite metal-rich SN 2010el. In the case of the rapidly declining events (note that these are not all overlapping with the faint events, e.g., PS 15csd; Magee et al. 2016), we similarly find low metallicities for all but SN 2010el. SE-ECSNe progenitors are expected to come from the lower end of the mass range of massive SNe progenitors  M and we therefore expect to see evidence for young, but not exceptionally young SPs at their explosion sites – this is borne out in the relatively modest EW(H) values we find of 25–40 Å. We therefore confirm here the preference for faint and/or fast subsets of SNe Iax (SN 2008ha-like events) to be preferentially in lower-metallicity environments with signatures of relatively young SPs at their explosion sites, consistent with current predictions for SE-ECSNe by Moriya & Eldridge (2016, see also ). In this scenario the proposed donor or remnant of SN 2008ha (Foley et al., 2014) would be the companion star that may have stripped the envelope of the SN progenitor. The spectral predictions for SE-ECSNe that are needed to provide a full comparison to SN 2008ha-like events are currently lacking.

7 Conclusions

We have presented a spectroscopic survey of the explosion sites and host galaxies of SNe Iax and fitted emission line regions in our VLT/MUSE IFS and NOT/ALFOSC long-slit observations.

We determined deprojected metallicity gradients for all hosts where possible and found that estimates of the explosion sites’ metallicities based on these gradients agreed well with our direct measurements. With this our explosion site metallicity sample was supplemented with gradient estimates where we could not measure the location directly. The majority of SN Iax explosion sites were found to be sub-solar ( dex in the scale of D16), and generally metal poor compared to the distribution of SF in their hosts (\crefsec:resz).

SNe Iax explosion sites appeared to be less intensely star-forming and somewhat older than the typical SF region of the hosts (\crefsec:resultsages), although they do follow a similar host-normalised offset distribution (\crefsec:resultsoffsets). Through comparison with a fiducial gas model and the results of the SP synthesis code bpass we estimate the ages of the young SPs at the explosion sites of SNe Iax to be several to years old (although note the caveats on assigning quantitative values in \crefsec:methages). The relatively young ages at the explosion sites are confirmed through a similarity in their association to the ongoing SF of their host galaxies to that seen for SNe II (\crefsec:pixelstatres), which are expected to have typical ages of tens of Myr, extending up to years (Zapartas et al., 2017).

Assessing our environmental measures in terms of SN properties, we find that the brighter, more slowly fading objects define our range of metallicities, EW(H) and offsets by covering the entire ranges (\crefsec:sncompare). Faint and faster events appear to occupy more restricted ranges of EW values and host offsets although we are limited to only a few objects for which we can investigate. Their metallicities extend across a wide range including some of the most metal-poor and -rich in the whole sample. We find no correlations between the velocity of the SNe and our environmental parameters. The only two designated SNe Iax to display helium (SNe 2004cs, 2007J) have among the highest EW(H) values of the sample, indicate the presence of comparatively younger SPs. Their membership as SNe Iax has been debated (White et al., 2015; Foley et al., 2016). With just two events we are limited in searching for statistical difference between these and the rest of the SNe Iax and their environments appear typical of both SNe Iax and IIb (the suggested typing of White et al. 2015).

When compared to other SN types, SNe Iax as a whole display a similar metallicity distribution to that seen for SNe Ib/c and II, with similar median values and covering a broadly similar range in metallicity. The host-normalised offset distribution of SNe Iax follows closely that seen for SNe Ia in disk galaxies and SNe Ib. Using pixel statistics, SNe Iax trace the ongoing SF of their hosts at a level most similar to SN II and significantly more so than SNe Ia, but less than SNe Ic (and LGRBs).

For the S0 host of SN 2008ge, NGC 1527, we discover a stream and clumps of ionised gas not seen in previous studies. A limited analysis of emission line ratios would disfavour a SF-driven ionising flux, with an evolved SP or shocks more likely to be powering it. We find no evidence of ionised gas at, or nearby, the location of SN 2008ge and confirm its environment is old compared to the rest of SNe Iax. The ionised component suggests there may have been SF in NGC 1527 relatively recently and thus Gyrs old progenitors are not required.

The young explosion sites we confirm here remain in good agreement with the expectations for progenitors consisting of CO/CONe WD + He stars and moderately massive CCSNe with fall-back or SE-ECSNe. The lack of features associated with very young environments disfavour the presence of very massive stars at the explosion site. We would therefore consider the model of massive WR stars suffering fall-back on to a black hole upon collapse as unlikely progenitors (without having to invoke implausibly large runaway velocities), ruling out faint SNe Iax as a way to explain local, apparently SN-less LGRBs.


We thank Ryan Foley for providing the late time spectra of SN 2014dt. David Balam is thanked for providing the spectrum of PSN J22412689+3917220 (now SN 2015ce) for its registration on the Transient Name Server. Eric Hsiao and the iPTF collaboration are thanked for their assistance registering PTF 14ans (now SN 2014ey) onto the Transient Name Server. Lise Christensen is thanked for comments on a draft of the manuscript. JDL gratefully acknowledges support from the UK Science and Technology Facilities Council (grant IDs ST/L000733/1 and ST/P000495/1). MDS gratefully acknowledges support provided by the Danish Agency for Science and Technology and Innovation realized through a Sapere Aude Level 2 grant, the VILLUM FONDEN (research grant 3261), and the Instrument Centre for Danish Astrophysics (IDA). Based on observations collected at the European Organisation for Astronomical Research in the Southern Hemisphere under ESO programmes 095.D0091, 096.D-0263 and 099.D-0022. Based on observations made with the Nordic Optical Telescope (proposal numbers 52-004 and P3-005; PI: Stritzinger), operated by the Nordic Optical Telescope Scientific Association at the Observatorio del Roque de los Muchachos, La Palma, Spain, of the Instituto de Astrofisica de Canarias. This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. We acknowledge the usage of the HyperLeda database ( This research made use of Astropy, a community-developed core Python package for Astronomy (Astropy Collaboration et al., 2013).

Appendix A The late-time spectra of SNe 2015H and 2014dt

Figure 18: The late-time spectrum of SN 2014dt extracted from the MUSE data cube at an epoch of +450 days after maximum, centred around prominent forbidden lines. Also shown are the +270 and +410 days spectra of Foley et al. (2016), we confirm a general lack of evolution in the strength or widths of the forbidden lines between the +270d spectra, although the [Ni ii7378 flux is slightly depressed. Our spectrum shows very little evolution from the +410d spectrum, with only perhaps a slight weakening of the 7670Å feature.

Our MUSE observations included a significant time lag in order to analyse the underlying galaxy light at the explosions sites. Nevertheless, for SNe 2014dt and 2015H we were able to detect prominent forbidden emission that is associated with SNe Iax at late phases. Although a full analysis of the SN signal is beyond the scope of this work we present their spectra for completeness and compare to the comprehensive analysis of late-time SN Iax spectra sample by Foley et al. (2016). For this we extracted flux in spaxels centred on the SN emission within a radius equal to the seeing of the image.

Our SN 2014dt spectrum is at a phase of approximately +450 rest-frame days post maximum, which makes it the latest spectrum of a SN Iax to our knowledge. In \creffig:14dtlate we compare our spectrum in a region of strong forbidden line emission to those of Foley et al. (2016), which were taken at +270 and +410d. As those authors noted based on their +410 days spectrum (to which our +450 days spectrum is almost identical), there is a general lack of evolution between these epochs in line strength or widths. Our SN 2014dt observations were taken under challenging sky conditions and the regions of interest are affected by residual sky emission not removed by the MUSE data reduction (\crefsec:museobs) – for the sake of presentation of \creffig:14dtlate we have manually subtracted the signal from a very faint region of the data cube.

The early spectra and light curve of SN 2015H were analysed by Magee et al. (2016) and our MUSE observation adds a late time spectrum at an epoch of +291 days past maximum. SNe Iax display a wide range of morphologies in their spectra around these wavelengths at similar epochs, as shown in Figure 8 of Foley et al. (2016) and we note a very strong similarity between SN 2002cx and SN 2015H. As with SN 2014dt we present a region of strong forbidden line emission in \creffig:15Hlate where we also plot the best-fitting 10 parameter model of Foley et al. (2016) in order to compare more quantitatively with SN 2002cx. As with SN 2002cx, we see that [Ni ii7378 emission is dominated by its broad component (FWHM  km s, cf. 7870 km s for SN 2002cx), with the narrow components of [Fe ii7155 and [Ca ii7291,7324 being more prominent, albeit somewhat narrower (FWHM  km s cf. 1430 km s for SN 2002cx).

We additionally note that we do not expect emission from these SNe that would affect our main environmental analysis, in the absence of circumstellar interaction, of which there appears to be no evidence. When running these SN spaxel bins as part of our main fitting procedure, starlight masked over the SNe features (such as those shown in \creffig:14dtlate,fig:15Hlate) as part of a sigma clipping routine. We found that manually masking large regions of the input spectra that may be affected by SN emission made very little difference to our emission line results. Late in the manuscript’s preparation we were able to check the effect of the SN emission for SN 2015H as MUSE re-observed the SN location. Although the observations were aborted due to scheduling restrictions,  s exposures were taken. We measured metallicities at the SN location on the first exposure (as the second was of poorer quality). Our values of and dex agree excellently with the values from our original observations of and dex for D16 and M13 (N2), respectively.

With these results we highlight the power of MUSE as an extremely efficient optical spectrograph to obtain nebular phase spectra of SNe, whilst obtaining spectral information at the explosion site and across the host for free, providing legacy value to the data for a wealth of galactic studies not possible with traditional long slit spectroscopy.

Figure 19: As for \creffig:14dtlate but here showing SN 2015H at an epoch of +291 days. The 10 component model, following the methodology of Foley et al. (2016) is shown. The morphology and line kinematics of the spectrum are similar to those of SN 2002cx at a similar epoch.

Appendix B Explosion site line fluxes

In \creftab:fluxes we provide individual flux measurements for the SN Iax explosion sites. These values have been corrected for Galactic reddening but not local reddening, i.e., before our Balmer decrement correction.

SN H H [O iii] [N ii] [S ii]
[ erg s]
MUSE sample
1991bj 21.59 0.278 0.100 0.309 0.064 0.184 0.224 0.159
2002bp 0.15 1.136
2002cx 9.05 0.301 0.113 0.355 0.066 0.188 0.231 0.177
2004cs 27.00 0.290 0.078 0.218 0.083 0.254 0.231 0.172
2005P 1.73 0.252 0.264 0.044 0.163 0.301 0.202
2005hk 0.51 0.290 0.248 0.059 0.112 0.313 0.204
2008ae 3.78 0.287 0.060 0.168 0.118 0.392 0.235 0.166
2008ha 32.13 0.439 0.376 0.990 0.026 0.079 0.167 0.119
2009J 7.72 0.284 0.124 0.339 0.102 0.323 0.258 0.185
2010ae 378.23 0.280 0.169 0.493 0.058 0.177 0.216 0.150
2010el 24.57 0.202 0.070 0.169 0.202 0.595 0.263 0.181
2011ce 249.84 0.245 0.020 0.052 0.127 0.378 0.179 0.129
2012Z 4.81 0.316 0.110 0.267 0.086 0.231 0.300 0.207
2013gr 4.08 0.310 0.195 0.547 0.035 0.106 0.300 0.202
2014dt 42.74 0.234 0.045 0.144 0.168 0.539 0.237 0.158
2014ey 1.75 0.261 0.085 0.203 0.078 0.258 0.271 0.197
2015H 2.95 0.144 0.084 0.072 0.411 0.277 0.143
NOT sample
2003gq 38.99 0.158 0.126 0.132
2005cc 55.52 0.135 0.302 0.204
2006hn 15.69 0.309 0.170 0.398 0.282
2007J 38.60 0.288 0.288 0.200 0.295
2009ku 17.82 0.282 0.363 0.219 0.316
2013dh 37.82 0.324 0.069 0.372 0.282
2013en 44.03 0.295 0.251 0.331
2014ek 60.88 0.209 0.295
PS 15csd 8.56 0.295 0.525 0.112 0.363
  • SN overlaid on a small, unresolved host.

Table 7: Emission line fluxes at the explosion sites of SNe Iax, corrected for Galactic extinction. Line strengths are given relative to the H flux, H). For the NOT sample [S ii] are given as the sum of the lines .

Appendix C Metallicity maps

In \creffig:zstamps we show stamps of D16 metallicity maps for the MUSE sample of SNe Iax hosts, excluding SN 2008ge, which is discussed and shown in \crefsec:08ge).

Figure 20: The metallicity maps of SNe Iax host galaxies using the indicator of D16 (see \crefsec:methmetallicity). Host nuclei and the SN explosion site positions are given by black and ‘star’ symbols, respectively. North is up, East is left. Note that PTF 14ans = SN 2014ey


  1. pubyear: 2017
  2. pagerange: Investigating the diversity of supernovae type Iax: A MUSE and NOT spectroscopic study of their environmentsC
  8. Using the 2016 update at
  9. Where we did not reliably detect [O iii] 5007 in our cubes we use the N2 relation of M13. These cases are highlighted in tables of results.
  11. For SN 1999ax we additionally detect another nearby source of \ionHii at an apparently higher redshift, this is discussed in \crefsec:hetero.
  12. We exclude NGC 1527, the host of SN 2008ge, as even if the ionised gas we detect is driven by SF (see \crefsec:08ge), it is very irregular in morphology, and perhaps not intrinsic to the luminous S0 host. Thus a metallicity gradient is not likely to be an appropriate description.
  13. In agreement with our findings based on the NOT spectroscopy here, Lyman et al. (2013) show a H image of the SN location with no underlying detected emission and mainly faint, diffuse emission throughout the host. McCully et al. (2014b) also present HST broadband observations of the environment, showing it to be in the outskirts of its host, in a fairly low density environment.
  14. We do not include the emission from the host of SN 2008ge, see \crefsec:08ge.
  15. Although we were not able to measure emission lines at the locations of some NOT-observed examples, these data were comparatively shallower than the MUSE data, where signatures of ongoing SF were found for all but one explosion site (SN 2008ge). This is demonstrated by our duplicate observations of SN 2015H: in the NOT data we were not able to extract an emission line spectrum at the explosion site but we could in the MUSE data.
  16. Our comparatively shallower NOT data with lower spectral resolution is not conducive to providing meaningful detections of this faint line, which is close to other, much stronger features.
  17. For the three SNe we find  mag,  mag and  km s, respectively. For SN 2010el we assign an uncertainty of 0.5 mag on the peak owing to the quite uncertain distance to NGC 1566.
  18. For another member, SN 2007qd, McClelland et al. (2010) find a faint peak magnitude ( mag), although this is uncertain since the light curve does not have full coverage (Foley et al., 2013). We note for this event we only have a poorly constrained (2 regions) metallicity gradient from which to estimate the D16 abundance as 8.60.
  19. Anderson-Darling tests provided very similar levels of significance, with slightly higher significance for the uncorrected SNe Ia vs Iax ().


  1. Andersen D. R., Bershady M. A., 2013, \apj, 768, 41
  2. Anderson J. P., James P. A., 2008, \mnras, 390, 1527
  3. Anderson J. P., Covarrubias R. A., James P. A., Hamuy M., Habergham S. M., 2010, \mnras, 407, 2660
  4. Anderson J. P., Habergham S. M., James P. A., Hamuy M., 2012, \mnras, 424, 1372
  5. Anderson J. P., James P. A., Habergham S. M., Galbany L., Kuncarayakti H., 2015, \pasa, 32, e019
  6. Anderson J. P., et al., 2016, \aap, 589, A110
  7. Asplund M., Grevesse N., Sauval A. J., Scott P., 2009, \araa, 47, 481
  8. Astropy Collaboration et al., 2013, \aap, 558, A33
  9. Bacon R., et al., 2010, in Ground-based and Airborne Instrumentation for Astronomy III. p. 773508, doi:10.1117/12.856027
  10. Balam D. D., Graham M. L., 2015, The Astronomer’s Telegram, 7931
  11. Baldwin J. A., Phillips M. M., Terlevich R., 1981, \pasp, 93, 5
  12. Binette L., Magris C. G., Stasińska G., Bruzual A. G., 1994, \aap, 292, 13
  13. Branch D., Baron E., Thomas R. C., Kasen D., Li W., Filippenko A. V., 2004, \pasp, 116, 903
  14. Bruzual G., Charlot S., 2003, \mnras, 344, 1000
  15. Cappellari M., Copin Y., 2003, \mnras, 342, 345
  16. Cardelli J. A., Clayton G. C., Mathis J. S., 1989, \apj, 345, 245
  17. Chabrier G., 2003, \pasp, 115, 763
  18. Childress M. J., et al., 2015, \mnras, 454, 3816
  19. Christensen L., Vreeswijk P. M., Sollerman J., Thöne C. C., Le Floc’h E., Wiersema K., 2008, \aap, 490, 45
  20. Cid Fernandes R., Mateus A., Sodré L., Stasińska G., Gomes J. M., 2005, \mnras, 358, 363
  21. Cid Fernandes R., Stasińska G., Mateus A., Vale Asari N., 2011, \mnras, 413, 1687
  22. Della Valle M., et al., 2006, \nat, 444, 1050
  23. Dopita M. A., Kewley L. J., Sutherland R. S., Nicholls D. C., 2016, \apss, 361, 61
  24. Eldridge J. J., Stanway E. R., 2009, \mnras, 400, 1019
  25. Eldridge J. J., Stanway E. R., 2012, \mnras, 419, 479
  26. Eldridge J. J., Langer N., Tout C. A., 2011, \mnras, 414, 3501
  27. Ferland G. J., Korista K. T., Verner D. A., Ferguson J. W., Kingdon J. B., Verner E. M., 1998, \pasp, 110, 761
  28. Filippenko A. V., 1997, \araa, 35, 309
  29. Filippenko A. V., et al., 1992a, \aj, 104, 1543
  30. Filippenko A. V., et al., 1992b, \apjl, 384, L15
  31. Fink M., et al., 2014, \mnras, 438, 1762
  32. Foley R. J., 2015, \mnras, 452, 2463
  33. Foley R. J., et al., 2009, \aj, 138, 376
  34. Foley R. J., et al., 2010a, \aj, 140, 1321
  35. Foley R. J., Brown P. J., Rest A., Challis P. J., Kirshner R. P., Wood-Vasey W. M., 2010b, \apjl, 708, L61
  36. Foley R. J., et al., 2013, \apj, 767, 57
  37. Foley R. J., McCully C., Jha S. W., Bildsten L., Fong W.-f., Narayan G., Rest A., Stritzinger M. D., 2014, \apj, 792, 29
  38. Foley R. J., Van Dyk S. D., Jha S. W., Clubb K. I., Filippenko A. V., Mauerhan J. C., Miller A. A., Smith N., 2015, \apjl, 798, L37
  39. Foley R. J., Jha S. W., Pan Y.-C., Zheng W. K., Bildsten L., Filippenko A. V., Kasen D., 2016, \mnras, 461, 433
  40. Freudling W., Romaniello M., Bramich D. M., Ballester P., Forchi V., García-Dabló C. E., Moehler S., Neeser M. J., 2013, \aap, 559, A96
  41. Fruchter A. S., et al., 2006, \nat, 441, 463
  42. Fynbo J. P. U., et al., 2006, \nat, 444, 1047
  43. Gal-Yam A., et al., 2006, \nat, 444, 1053
  44. Gal-Yam A., Maoz D., Guhathakurta P., Filippenko A. V., 2008, \apj, 680, 550
  45. Galbany L., et al., 2014, \aap, 572, A38
  46. Galbany L., et al., 2016a, \mnras, 455, 4087
  47. Galbany L., et al., 2016b, \aap, 591, A48
  48. Gehrels N., et al., 2006, \nat, 444, 1044
  49. Gomes J. M., et al., 2016a, \aap, 585, A92
  50. Gomes J. M., et al., 2016b, \aap, 588, A68
  51. González Delgado R. M., Leitherer C., Heckman T. M., 1999, \apjs, 125, 489
  52. Goudfrooij P., 1999, in Carral P., Cepa J., eds, Astronomical Society of the Pacific Conference Series Vol. 163, Star Formation in Early Type Galaxies. p. 55 (arXiv:astro-ph/9809057)
  53. Hakobyan A. A., Mamon G. A., Petrosian A. R., Kunth D., Turatto M., 2009, \aap, 508, 1259
  54. Hakobyan A. A., Adibekyan V. Z., Aramyan L. S., Petrosian A. R., Gomes J. M., Mamon G. A., Kunth D., Turatto M., 2012, \aap, 544, A81
  55. Hakobyan A. A., et al., 2016, \mnras, 456, 2848
  56. Harmanen J., et al., 2015, The Astronomer’s Telegram, 8264
  57. Howell D. A., et al., 2006, \nat, 443, 308
  58. Hsiao E. Y., et al., 2013, The Astronomer’s Telegram, 5612
  59. Izzo L., et al., 2017, preprint, (arXiv:1704.05509)
  60. James P. A., Anderson J. P., 2006, \aap, 453, 57
  61. Jha S. W., 2017, preprint, (arXiv:1707.01110)
  62. Jha S., Branch D., Chornock R., Foley R. J., Li W., Swift B. J., Casebeer D., Filippenko A. V., 2006, \aj, 132, 189
  63. Jha S. W., et al., 2013, The Astronomer’s Telegram, 5143
  64. Jordan IV G. C., Perets H. B., Fisher R. T., van Rossum D. R., 2012, \apjl, 761, L23
  65. Kangas T., et al., 2017, \aap, 597, A92
  66. Kasliwal M. M., et al., 2012, \apj, 755, 161
  67. Kennicutt Jr. R. C., 1998, \araa, 36, 189
  68. Kewley L. J., Ellison S. L., 2008, \apj, 681, 1183
  69. Kewley L. J., Dopita M. A., Sutherland R. S., Heisler C. A., Trevena J., 2001, \apj, 556, 121
  70. Kewley L. J., Maier C., Yabe K., Ohta K., Akiyama M., Dopita M. A., Yuan T., 2013, \apjl, 774, L10
  71. Kromer M., et al., 2013, \mnras, 429, 2287
  72. Kromer M., et al., 2015, \mnras, 450, 3045
  73. Krühler T., Kuncarayakti H., Schady P., Anderson J. P., Galbany L., Gensior J., 2017, preprint, (arXiv:1702.05430)
  74. Kuncarayakti H., et al., 2013a, \aj, 146, 30
  75. Kuncarayakti H., et al., 2013b, \aj, 146, 31
  76. Leibundgut B., et al., 1993, \aj, 105, 301
  77. Leloudas G., et al., 2011, \aap, 530, A95
  78. Leloudas G., et al., 2015, \mnras, 449, 917
  79. Li W., et al., 2003, \pasp, 115, 453
  80. Li W., Chornock R., Leaman J., Filippenko A. V., Poznanski D., Wang X., Ganeshalingam M., Mannucci F., 2011, \mnras, 412, 1473
  81. Liu Z.-W., et al., 2015a, \mnras, 452, 838
  82. Liu Z.-W., Moriya T. J., Stancliffe R. J., Wang B., 2015b, \aap, 574, A12
  83. Liu Z.-W., Stancliffe R. J., Abate C., Wang B., 2015c, \apj, 808, 138
  84. Lunnan R., et al., 2014, \apj, 787, 138
  85. Lyman J. D., James P. A., Perets H. B., Anderson J. P., Gal-Yam A., Mazzali P., Percival S. M., 2013, \mnras, 434, 527
  86. Lyman J. D., et al., 2017, \mnras, 467, 1795
  87. Magee M. R., et al., 2016, \aap, 589, A89
  88. Magee M. R., et al., 2017, \aap, 601, A62
  89. Marino R. A., et al., 2013, \aap, 559, A114
  90. McClelland C. M., et al., 2010, \apj, 720, 704
  91. McCully C., et al., 2014a, \nat, 512, 54
  92. McCully C., et al., 2014b, \apj, 786, 134
  93. Meng X., Podsiadlowski P., 2014, \apjl, 789, L45
  94. Michałowski M. J., et al., 2016, preprint, (arXiv:1610.06928)
  95. Modjaz M., et al., 2008, \aj, 135, 1136
  96. Modjaz M., Kewley L., Bloom J. S., Filippenko A. V., Perley D., Silverman J. M., 2011, \apjl, 731, L4+
  97. Modjaz M., et al., 2014, \aj, 147, 99
  98. Moriya T. J., Eldridge J. J., 2016, \mnras, 461, 2155
  99. Moriya T., Tominaga N., Tanaka M., Nomoto K., Sauer D. N., Mazzali P. A., Maeda K., Suzuki T., 2010, \apj, 719, 1445
  100. Narayan G., et al., 2011, \apjl, 731, L11
  101. Ochner P., Tomasella L., Benetti S., Cappellaro E., Elias-Rosa N., Pastorello A., Turatto M., 2014, The Astronomer’s Telegram, 6648
  102. Oey M. S., Parker J. S., Mikles V. J., Zhang X., 2003, \aj, 126, 2317
  103. Osterbrock D. E., Ferland G. J., 2006, Astrophysics of gaseous nebulae and active galactic nuclei
  104. Pan Y.-C., et al., 2015, The Astronomer’s Telegram, 7534
  105. Perets H. B., et al., 2010, \nat, 465, 322
  106. Pettini M., Pagel B. E. J., 2004, \mnras, 348, L59
  107. Phillips M. M., 1993, \apjl, 413, L105
  108. Phillips M. M., Wells L. A., Suntzeff N. B., Hamuy M., Leibundgut B., Kirshner R. P., Foltz C. B., 1992, \aj, 103, 1632
  109. Phillips M. M., et al., 2007, \pasp, 119, 360
  110. Prieto J. L., et al., 2016, preprint, (arXiv:1609.00013)
  111. Pumo M. L., et al., 2009, \apjl, 705, L138
  112. Riess A. G., et al., 2016, \apj, 826, 56
  113. Ruiz-Lapuente P., Cappellaro E., Turatto M., Gouiffes C., Danziger I. J., della Valle M., Lucy L. B., 1992, \apjl, 387, L33
  114. Sana H., et al., 2012, Science, 337, 444
  115. Sánchez-Blázquez P., et al., 2006, \mnras, 371, 703
  116. Sánchez S. F., et al., 2012, \aap, 546, A2
  117. Sanders N. E., et al., 2012, \apj, 758, 132
  118. Sarzi M., et al., 2006, \mnras, 366, 1151
  119. Sarzi M., et al., 2010, \mnras, 402, 2187
  120. Schlafly E. F., Finkbeiner D. P., 2011, \apj, 737, 103
  121. Shivvers I., et al., 2017, \pasp, 129, 054201
  122. Smette A., et al., 2015, \aap, 576, A77
  123. Sofue Y., Rubin V., 2001, \araa, 39, 137
  124. Soto K. T., Lilly S. J., Bacon R., Richard J., Conseil S., 2016, \mnras, 458, 3210
  125. Stanishev V., 2007, Astronomische Nachrichten, 328, 948
  126. Stanishev V., Rodrigues M., Mourão A., Flores H., 2012, \aap, 545, A58
  127. Stanway E. R., Eldridge J. J., Greis S. M. L., Davies L. J. M., Wilkins S. M., Bremer M. N., 2014, \mnras, 444, 3466
  128. Stanway E. R., Eldridge J. J., Becker G. D., 2016, \mnras, 456, 485
  129. Stoll R., Prieto J. L., Stanek K. Z., Pogge R. W., 2013, \apj, 773, 12
  130. Stritzinger M., Mazzali P. A., Sollerman J., Benetti S., 2006, \aap, 460, 793
  131. Stritzinger M. D., et al., 2014, \aap, 561, A146
  132. Stritzinger M. D., et al., 2015, \aap, 573, A2
  133. Svensson K. M., Levan A. J., Tanvir N. R., Fruchter A. S., Strolger L.-G., 2010, \mnras, 405, 57
  134. Taddia F., et al., 2013, \aap, 558, A143
  135. Taddia F., et al., 2015a, \aap, 574, A60
  136. Taddia F., et al., 2015b, \aap, 580, A131
  137. Thöne C. C., et al., 2008, \apj, 676, 1151
  138. Tomasella L., et al., 2016, \mnras, 459, 1018
  139. Tonry J., Denneau L., Stalder B., Heinze A., Sherstyuk A., Rest A., Smith K. W., Smartt S. J., 2016, The Astronomer’s Telegram, 8680
  140. Tripp R., 1998, \aap, 331, 815
  141. Turatto M., Benetti S., Cappellaro E., Danziger I. J., Della Valle M., Gouiffes C., Mazzali P. A., Patat F., 1996, \mnras, 283, 1
  142. Valenti S., et al., 2009, \nat, 459, 674
  143. Vulcani B., et al., 2014, \mnras, 441, 1340
  144. Wang B., Justham S., Han Z., 2013, \aap, 559, A94
  145. Wang B., Meng X., Liu D.-D., Liu Z.-W., Han Z., 2014, \apjl, 794, L28
  146. White C. J., et al., 2015, \apj, 799, 52
  147. Xiao L., Eldridge J. J., Stanway E., Galbany L., 2017, preprint, (arXiv:1705.03606)
  148. Zapartas E., et al., 2017, \aap, 601, A29
  149. de Vaucouleurs G., de Vaucouleurs A., Corwin Jr. H. G., Buta R. J., Paturel G., Fouqué P., 1991, Third Reference Catalogue of Bright Galaxies. Volume I: Explanations and references. Volume II: Data for galaxies between 0 and 12. Volume III: Data for galaxies between 12 and 24.. Springer
  150. van Dokkum P. G., 2001, \pasp, 113, 1420