2.1. Data reduction and supernova spectrum extraction

The WiFeS instrument is an image-slicing integral field spectrograph with a wide 25 arcsec × 38 ′′arcsec field of view. For AWSNAP, this frequently provided simultaneous integral field observations of SNe and their host galaxies. The WiFeS image slicer breaks the field of view into 25 ‘slitlets’ of width 1 arcsec, which then pass through a dichroic beamsplitter and volume phase holographic (VPH) gratings before arriving at 4k × 4k CCD detectors. AWSNAP observations were always conducted with a CCD binning of 2 in the vertical direction—this sets the vertical spatial scale of the detector to be 1 arcsec, yielding final integral field elements (or ‘spaxels’) of size 1 arcsec × 1 arcsec. Typically seeing at Siding Spring is roughly 2 arcsec (see Section 2.3).

The VPH gratings utilised by WiFeS provide a higher wavelength resolution than traditional glass gratings. The low- and high-resolution gratings provide resolutions of R = 3000 and R = 7000, respectively, yielding velocity resolutions of up to σ_v ~ 45 km s⁻¹ which is ideal for observing nebular emission lines from ionised regions in galaxies. For SNe, this can reveal narrow absorption features (see Section 4.2) from circumstellar material (CSM) which are typically smeared out by lower resolution spectrographs. AWSNAP observations were generally conducted with the lower resolution B3000 and R3000 gratings for the blue and red arms of the spectrograph, respectively, with the RT560 dichroic beamsplitter. Occasionally, the R7000 grating was deployed on the red arm to provide higher resolution observations of sodium absorption features. In Table 1, we provide the wavelength range, spectroscopic pixel size, and wavelength resolution (determined as the FWHM of calibration lamp emission lines) for the three gratings used for AWSNAP observations.

Table 1. Details of WiFeS gratings.

Data for AWSNAP observations were reduced with version 0.7.0 of the PyWiFeS pipeline (Childress et al. Reference Childress, Vogt, Nielsen and Sharp2014a). PyWiFeS performs standard image pre-processing such as overscan and bias subtraction, as well as cosmic ray rejection using a version of LACosmic (van Dokkum Reference van Dokkum2001) tailored for WiFeS data. The wavelength solution for WiFeS is derived using an optical model of the spectrograph which achieves an accuracy of 0.05 to 0.10Å (for R = 7000 and R = 3000, respectively) across the entire detector. Spectral flatfielding (i.e. correction of pixel-to-pixel quantum efficiency variations) is achieved with an internal quartz lamp, whilst spatial flat-fielding across the full instrument field of view is facilitated by twilight sky flats. Once the data has been pre-processed and flat-fielded, it is resampled onto a rectilinear three-dimensional (x, y, λ) grid (a ‘data cube’). We then flux calibrate the data cubes with the use of spectrophotometric standard stars (from, e.g. Oke Reference Oke1990; Bessell Reference Bessell1999; Stritzinger et al. Reference Stritzinger, Suntzeff, Hamuy, Challis, Demarco, Germany and Soderberg2005) and the Siding Spring Observatory extinction curve from Bessell (Reference Bessell1999), whilst telluric features are removed using observations of smooth spectrum stars.

To further facilitate data reduction for the AWSNAP data release, we developed an SQL database for WiFeS observations using the Python Django framework. We created modified versions of the PyWiFeS reduction scripts that allow the user to request that all data from a specific night using a specific grating be fully reduced. The reduction scripts then query the database for all science observations and calibration data from that night and perform the required reduction procedures. For nights where a full suite of calibrations was not available (e.g. due to guest observations on a non-AWSNAP night), calibration solutions from the closest (in time) AWSNAP night were automatically identified via the database and employed in the reduction. Both the wavelength solution and spatial flat-fielding of the instrument are incredibly stable on long (~year) timescales (see Section 2.2), thus validating the choice to use calibrations from different nights where necessary. This new database-driven data processing mode for PyWiFeS will be an important component of the upcoming effort to develop fully robotic queue-based observing capabilities for the ANU 2.3-m telescope.

The output of the PyWiFeS pipeline is a flux calibrated data cube that contains signal from the target supernova, the night sky, and occasionally the SN host galaxy. Extraction of the SN spectrum requires isolation of the spaxels containing supernova signal and subtraction of the underlying sky (and possibly galaxy) background. To achieve this, we constructed a custom Python-based GUI which allows the user to manually select spaxels containing the target SN (‘object’ spaxels) and spaxels used to determine the background (‘sky’ spaxels, which may contain some galaxy signal). The background spectrum is determined as the median spectrum across all ‘sky’ spaxels, and this median spectrum is subtracted from each ‘object’ spaxel. The final SN spectrum is then the sum of the sky-subtracted object spaxels, and the variance spectrum is the sum of the variance from all object spaxels (with no subtraction of sky variance).

This SN spectrum extraction technique produces excellent quality sky subtraction due to the robust spatial flat-fielding achieved with PyWiFeS. However, some obvious sky subtraction residual features are evident in the redder wavelengths of most R3000 spectra where the night sky exhibits sharp emission features from rotational transitions of atmospheric OH molecules. The intrinsic line width of these emission lines is below the resolution of the spectrograph, meaning the observed line width is that of the spectrograph—which in this case is only slightly larger than the detector pixel size. The natural wavelength solution of the spectrograph shifts spaxel-to-spaxel, so when all spaxels are resampled onto the same wavelength grid this means sky lines experience pixel-wise resampling that is not uniform across the full instrument field of view. Thus, when a median sky spectrum is subtracted from a specific spaxel, some residual features arise due to this resampling effect. A more robust technique for correcting this would be to model the intrinsic sky spectrum using the multiple samplings achieved across all spaxels and resample it to the wavelength solution of each spaxel before subtracting it (as was demonstrated for two-dimensional spectroscopic data by Kelson Reference Kelson2003). Such a technique is beyond the scope of the current data release, but is being prioritised for future AWSNAP releases.

2.2. Long-term behaviour of the WiFeS instrument

One key advantage of having a long-running observing programme is the ability to characterise the long-term behaviour of the WiFeS instrument. Below we analyse the stability of the wavelength solution and instrument throughput on multi-year timescales. WiFeS did undergo a major change in early 2013 when the detectors for both arms of the spectrograph were replaced with higher throughput E2V CCDs. Thus, we restrict our analysis to dates from late 2013 May until the end of the current data release in 2015 August.

We collected the wavelength solution fits for the B3000 (R3000) grating from 48 (45) distinct epochs following the WiFeS detector upgrade. In Figure 1, we plot the difference between the wavelength solution for each individual epoch and the mean wavelength solution across all epochs. Epochs are colour-coded from earliest (red) to latest (purple) to illustrate potential coherent long-term shifts in the wavelength solution. The top panels present the B3000 grating, whilst the bottom panels present the R3000 gratings. The left and middle columns represent the WiFeS slitlets at the top and middle of the detectors, respectively. In the right hand panels of this Figure, we show the average deviation of the wavelength solution for a given epoch from the global mean wavelength solution, as a function of epoch. This allows us to track any coherent evolution of the detector wavelength solution with time.

Figure 1. Evolution of the WiFeS wavelength solution over a 2-yr period spanning 2013 May to 2015 April. In the left two panels of each row, we plot the deviation of individual wavelength solutions from the mean solution (averaged over the full 2-yr period) for the top and middle slitlets of the instrument (left and middle columns, respectively) for the B3000 (top row) and R3000 (bottom row) gratings. Wavelength solution residuals are colour-coded by date from earliest (red) to latest (purple), with the pixel size (dashed black lines) and wavelength solution residual RMS (solid black lines) displayed for comparison. In the right panels of each row, we show the average deviation (across all wavelengths) from the mean wavelength solution as a function of epoch, to trace the temporal evolution of the instrument solution (points for individual epochs obey the same colour scheme as the left panels).

From these plots, we see clear demonstration of the remarkable stability of the WiFeS instrument. The RMS variation of wavelength solution is smaller than a single pixel for both gratings (i.e. both detectors) for nearly all wavelengths, with the exception of the lower throughput regions near the dichroic boundary. There is some evidence of a coherent shift of the blue detector wavelength solution over the 2-yr time period probed here, but this is still less than two pixels shift. For the red detector, we can say confidently that the wavelength solution has shifted by less than a pixel over a timescale of 2 yrs. This remarkable stability achieved (by design) by WiFeS means the use of the wavelength solution from an adjacent night yields negligible changes in wavelength, and thus small instrumental velocity shifts—highly suitable for supernova analyses.

We then collected illumination corrections (derived from twilight flat observations) from 46 (45) distinct epochs for the B3000 (R3000) grating. In Figure 2, we show the mean illumination correction (left panels) and RMS of the illumination correction (right panels—presented in fractional form, i.e. the RMS divided by the mean) for both the B3000 (top) and R3000 (bottom) gratings. We find the mean variation of the illumination correction to be 1.4 and 1.2% for the B3000 and R3000 gratings, respectively, for the full WiFeS field of view. The nature of the instrument optics are such that the outer regions of instrument field of view have slightly lower throughput, and thus a higher fractional RMS than the inner most region. For the innermost 8 arcsec × 8 arcsec region typically used for SN observation in AWSNAP, the RMS of the illumination correction is 0.8 and 0.3% for the blue and red detectors (i.e. B3000 and R3000 gratings). Thus, the throughput of the WiFeS instrument has remarkable spatial uniformity on long (multi-year) timescales.

Figure 2. The normalised mean (left column) and RMS (right column) of the illumination corrections for the B3000 (top row) and R3000 (bottom row) gratings. The RMS images are shown as fractional values of the mean illumination correction.

2.3. Observing conditions at siding spring observatory

In addition to the long-term behaviour of the WiFeS instrument, our observing programme allows us to monitor the observing conditions at Siding Spring Observatory. Below we briefly discuss the atmospheric throughput and seeing conditions experienced during AWSNAP observing.

We collected the flux calibration solutions from 59(54) epochs for the B3000 (R3000) grating from 2013 May to 2015 August, and these are plotted in Figure 3 after being normalised in the wavelength range with highest throughput (45 00–5 400 Å and 6 500–8 000 Å for B3000 and R3000, respectively). These curves represent the normalised throughput (in magnitudes) of the instrument and atmosphere as measured with spectrophotometric standard stars (typically from Oke Reference Oke1990; Bessell Reference Bessell1999; Stritzinger et al. Reference Stritzinger, Suntzeff, Hamuy, Challis, Demarco, Germany and Soderberg2005) whose flux has already been corrected using the nominal Siding Spring extinction curve from Bessell (Reference Bessell1999). The effects of the instrument throughput and atmospheric transmission are degenerate here, as WiFeS instrument throughput cannot be independently measured using local (i.e. terrestrial) calibration sources.

Figure 3. Flux calibration solutions for the B3000 (top) and R3000 (bottom) gratings. As in Figure 1, these are colour-coded by date from earliest (red) to latest (purple), with the mean flux calibration solution shown as the solid black line.

From the curves in Figure 3, we see that the total combined throughput of the instrument plus atmosphere is relatively stable. We calculated the RMS colour variation in the throughput curves and find variations of σ(U − B) = 0.09 mag for B3000 (note the V band runs into the wavelength range where the dichroic splits light between the blue and red channels) and σ(r − i) = 0.04 mag for R3000. We note these are calculated from the mean flux calibration solution for each night, where no attempt has been made to derive a unique extinction curve for a given night. These colour variations are relatively small, and comparable in size to the colour dispersion found when comparing spectrophotometry to imaging photometry for other SN spectroscopy samples (e.g. Silverman et al. Reference Silverman2012c; Blondin et al. Reference Blondin2012; Modjaz et al. Reference Modjaz2014).

We also measured the atmospheric seeing during our observing programme, using both guide star FWHM measurements recorded from the WiFeS guider camera and low airmass (secz ⩽ 1.5) WiFeS data cubes convolved with the B- and R-band filter curves (for the blue and red cameras, respectively). We plot the observed distribution of seeing values in Figure 4. The seeing distribution peaks slightly above 1.5 arcsec (for all measurements) with a tail predominantly filled to 2.5 arcsec, with little or no sub-arcsecond observations and a small number of observations with incredibly poor seeing (3.0 arcsec or greater).

Figure 4. Seeing measurements at Siding Spring Observatory as measured during AWSNAP observations. These include measurements from the WiFeS guider camera (green dash-dot histogram), and measurements of low airmass standard star WiFeS datacubes convolved with B-band (blue solid histogram—from the blue detector) and R-band (red dashed histogram—from the red detector) filter curves.

4.1. Spectral features in SNe Ia at maximum light

We begin by inspecting the spectra of SNe Ia close to the epoch of peak brightness. These maximum light spectra have a long history of providing key insights into the diversity of SN Ia explosions through the study of ‘spectral indicators’ (Nugent et al. Reference Nugent, Phillips, Baron, Branch and Hauschildt1995; Hatano et al. Reference Hatano, Branch, Lentz, Baron, Filippenko and Garnavich2000; Benetti et al. Reference Benetti2005; Bongard et al. Reference Bongard, Baron, Smadja, Branch and Hauschildt2006, Reference Bongard, Baron, Smadja, Branch and Hauschildt2008; Hachinger, Mazzali, & Benetti Reference Hachinger, Mazzali and Benetti2006; Hachinger et al. Reference Hachinger, Mazzali, Tanaka, Hillebrandt and Benetti2008; Branch et al. Reference Branch2006; Branch, Dang, & Baron Reference Branch, Dang and Baron2009; Bronder et al. Reference Bronder2008; Silverman, Kong, & Filippenko Reference Silverman, Kong and Filippenko2012b), as well as potential avenues for improving the cosmological standardisation of SNe Ia (Wang et al. Reference Wang2009; Bailey et al. Reference Bailey2009; Blondin, Mandel, & Kirshner Reference Blondin, Mandel and Kirshner2011; Silverman et al. Reference Silverman, Ganeshalingam, Li and Filippenko2012a). Thus, we actively targeted many SNe Ia (which had already been classified) on an epoch close to maximum light to obtain high-quality spectra facilitating such studies.

We isolated the sample SNe Ia with a spectrum within 5 d of estimated peak brightness. It is important to reiterate here that the phases for our SN Ia sample are extrapolated from the reported spectroscopic classification phase and date. It has been demonstrated previously (e.g. Blondin et al. Reference Blondin2012) that the phase determined for SNe Ia via spectroscopic matching codes such as SNID (Blondin & Tonry Reference Blondin and Tonry2007) typically has an uncertainty of at least 3–5 d. Thus, we may have included SN Ia spectra as much as 10 d removed from maximum light. Our analysis here is intended to be illustrative of the utility of our published spectra, and more detailed quantitative spectroscopic analyses should always be coupled to a robust photometric data set.

For the analysis that follows, we measure two key quantities of interest: the silicon absorption ratio R _Si, and the strength of high-velocity features (HVFs) R _HVF. For this work, we define R _Si as the ratio of the pseudo equivalent width (pEW) of the 5962 Å feature to the pEW of the 6355 Å feature. The pEW for each feature is measured by fitting a linear pseudo-continuum across narrow regions redward and blueward of the given feature (similar to methods employed in Silverman et al. Reference Silverman, Kong and Filippenko2012b), then integrating the flux in the normalised absorption feature.

R _Si is known to correlate strongly with the SN Ia light curve decline rate (Nugent et al. Reference Nugent, Phillips, Baron, Branch and Hauschildt1995). We quantified the relationship between our definition of R _Si (the pEW ratio) and light curve decline rate Δm ₁₅ by fitting a linear relationship between these quantities for a sample of 342 SNe Ia collated from the CfA, BSNIP, and CSP samples. The fit to these data is presented in Appendix A2, shown graphically in Figure A1 with a best fit trend given by Equation (A1). Below we will use this relation to display the corresponding range of Δm ₁₅ spanned by our observed values of R _Si.

We measure the HVF strength for our SN Ia sample using the same techniques employed in Childress et al. (Reference Childress, Filippenko, Ganeshalingam and Schmidt2014b), and similarly define R _HVF as the ratio of the pEW of the HVF to the pEW of the photospheric velocity feature (PVF). Briefly, we first normalise the Ca NIR feature using a linear pseudo-continuum fit. We then fit the Ca NIR feature with two Gaussians in velocity space with velocity centres, widths, and absorption depths fitted with a Python mpfit routine. Each component in velocity space corresponds to a triplet in wavelength space, and we set the absorption depth of each component equal to each other (i.e. the optically thick limit). As in Childress et al. (Reference Childress, Filippenko, Ganeshalingam and Schmidt2014b), we require the photospheric velocity component to have a velocity centre within 20% of the value derived for the Si 6355 Å feature, though in most cases the results are the same if this requirement is removed.

In Figure 10, we plot the measured HVF strength (R _HVF) versus the silicon absorption strength ratio R _Si. For illustrative purposes, we also show the scale of light curve decline rate Δm ₁₅ corresponding to the plotted range of R _Si using the above relation. A representative error bar for Δm ₁₅ from the above relation and a typical R _HVF error-bar are shown in the figure.

Figure 10. Strength of the high-velocity features (HVFs) in the Ca II NIR triplet—using the quantity R _HVF as defined by Childress et al. (Reference Childress, Filippenko, Ganeshalingam and Schmidt2014b)—plotted against the Si II absorption strength ratio R _Si defined by Nugent et al. (Reference Nugent, Phillips, Baron, Branch and Hauschildt1995). On the top axis, we show the rough equivalent light curve decline rate Δm ₁₅ values corresponding to the range of R _Si values. The crosshair in the upper right represents the characteristic errors in measurement of R _HVF (typically 20%, plotted here for the larger values of R _HVF) and the error in Δm ₁₅ when converted from R _Si.

Figure 10 demonstrates a tendency for SNe Ia with strong HVFs to also have low values of R _Si and thus broad light curves (low Δm ₁₅). This correlation of HVFs with SN Ia light curve width was first observed by Maguire et al. (Reference Maguire2012) in composite high-redshift SN Ia spectra, and subsequently confirmed by numerous studies of low-redshift SNe Ia (Childress et al. Reference Childress, Filippenko, Ganeshalingam and Schmidt2014b; Maguire et al. Reference Maguire2014; Pan et al. Reference Pan, Sullivan, Maguire, Gal-Yam, Hook, Howell, Nugent and Mazzali2015; Silverman et al. Reference Silverman, Vinkó, Marion, Wheeler, Barna, Szalai, Mulligan and Filippenko2015; Zhao et al. Reference Zhao2015). Our results support these studies with spectra alone.

4.2. Narrow sodium absorption features in SN Ia spectra

A great advantage of the higher resolution of WiFeS (compared to many spectrographs deployed for other SN spectroscopy surveys) is the ability to detect narrow absorption features, particularly the Na I doublet at λλ5890/5896 Å. This feature has a long history of being used to infer the presence of foreground dust in SNe Ia (Barbon et al. Reference Barbon, Benetti, Rosino, Cappellaro and Turatto1990; Turatto, Benetti, & Cappellaro Reference Turatto, Benetti, Cappellaro, Hillebrandt and Leibundgut2003; Poznanski et al. Reference Poznanski, Ganeshalingam, Silverman and Filippenko2011; Poznanski, Prochaska, & Bloom Reference Poznanski, Prochaska and Bloom2012; Phillips et al. Reference Phillips2013). Earlier works attempted to derive correlations between SN Ia colours and sodium absorption strength (i.e. the absorption equivalent width), but Poznanski et al. (Reference Poznanski, Ganeshalingam, Silverman and Filippenko2011) showed sodium to be a poor indicator of SN Ia reddening. Phillips et al. (Reference Phillips2013) refined this result by showing that SN Ia reddening exhibits a strong correlation with absorption strength of the diffuse interstellar bands (found exclusively in the interstellar medium) but some SNe Ia have an excess of sodium absorption that does not coincide with associated reddening of the SN. Thus, sodium features in SNe Ia remain instructive but should be considered with a measure of caution.

Recently, velocities of sodium absorption features in SN Ia spectra have been used as a diagnostic of CSM, including cases of some SNe Ia (Patat et al. Reference Patat2007; Simon et al. Reference Simon2009) where the sodium absorption features exhibit variability (though most do not—see Sternberg et al. Reference Sternberg2014), indicating SN-CSM interaction. More recently, a statistical analysis of the velocity distribution of sodium features in SNe Ia by Sternberg et al. (Reference Sternberg2011) found an excess of SNe Ia with blueshifted sodium features, indicating that a fraction of SNe Ia explode inside an expanding shell of material that was presumably shed by the SN Ia progenitor system prior to explosion. Maguire et al. (Reference Maguire2013) extended this work to show that the excess of SNe Ia with blueshifted sodium absorption was populated predominantly by the more luminous SNe Ia with slow declining light curves (i.e. high ‘stretch’).

We thus searched for the presence of sodium absorption features in our sample of SN Ia spectra. For most SNe Ia (13), this was done with the lower resolution (R3000) observations of the SN at maximum light. For six SNe Ia, we also had a higher resolution (R7000) observation of the SN at maximum light. In a few instances, this observation was triggered by the presence of strong reddening being reported in the SN classification announcement. For the other instances, we obtained both a low-resolution (R3000) and high-resolution (R7000) spectrum in the red whilst obtaining a longer exposure blue (B3000) spectrum during nights near full moon when the blue sky background was exceptionally high.

We fitted the sodium doublet absorption profile as follows. First, the spectrum was normalised to the local continuum by fitting a quadratic to the SN flux between 10–25 Å redward or blueward of the doublet centre (5893 Å). The absorption profile was fitted as two Gaussians with rest wavelengths set by the doublet wavelengths but with unknown (common) velocity shift and velocity width and (independent) absorption depths. This was done with an mpfit routine in Python, which accounts for variable covariances and returns the appropriate fit values and uncertainties. We show representative examples for fits to data from both gratings in Figure 11. The outcomes for our sodium profile fits are presented in Table A4 in Appendix A, and comprise six successful fits for targets with R7000 spectra, and 13 for R3000 spectra.

Figure 11. Sodium absorption fit examples for both the R7000 (top) and R3000 (bottom) gratings. Data (which have been normalised to the local continuum fit) are shown as blue diamonds whilst the best fit absorption profile is shown as the solid red curve. For reference, we also mark the continuum level (horizontal black line at value 1.00) and the rest wavelengths of the sodium doublet (vertical dotted gray lines).

The primary quantity we wanted to measure from the sodium absorption feature was its velocity with respect to the local standard of rest at the SN site. The default local rest velocity was initially set to be the systemic velocity of the SN host galaxy. In some cases, we detected clear nebular emission lines at the SN site present in the SN spectrum—for these cases, the local rest velocity (i.e. the rotational velocity of the host galaxy at the site of the SN) was measured from the Hα emission line.

In two cases (SN 2014ao and ASASSN-14jg), no local Hα emission was present in the SN spectrum, and the SN sodium velocity differed from the host systemic velocity by more than 100 km s⁻¹. To obtain the true local rest velocities for these two SNe Ia, we took advantage of the integral-field data provided by WiFeS, which allows us to measure host galaxy properties (such as velocity) over a broad field of view.

We illustrate this for SN 2014ao in Figure 12: The WiFeS field of view extends from the host galaxy core to the outer edge of its spiral arms in the direction of the SN. We were able to extract the velocity of an H ii region along the SN-host axis and found its velocity differed significantly from that of the host core (Δv = 174 ± 10 km s⁻¹). As galaxy velocity curves tend to flatten at large radii, we use the H ii region velocity as the local rest velocity of SN 2014ao—a value much closer to the measured sodium absorption velocity.

Figure 12. Determination of the local velocity for SN 2014ao in NGC 2615 with WiFeS. Left: SDSS (York et al. Reference York2000) gri colour composite—created with SWARP (Bertin et al. Reference Bertin, Mellier, Radovich, Missonnier, Didelon, Morin, Bohlender, Durand and Handley2002) and STIFF (Bertin Reference Bertin, Ballester, Egret and Lorente2012)—with the WiFeS field of view (red rectangle) and SN location (red dot) highlighted. Middle: Image of the SN 2014ao WiFeS data cube in the isolated wavelength range within ± 6 Å (i.e. ± 300 km s⁻¹) of the wavelength of Hα at the published redshift of NGC 2615, with host core (purple square) and nearby H ii region (brown square) highlighted—the SN is the bright object near the centre. Right: Extracted WiFeS spectra of the nearby H ii region (top) and host core (bottom) near the Hα+NII emission line group, with the expected location of those lines at the published redshift of NGC 2615 (z = 0.014083, Theureau et al. Reference Theureau, Bottinelli, Coudreau-Durand, Gouguenheim, Hallet, Loulergue, Paturel and Teerikorpi1998a) shown as the vertical gray lines.

With the final sodium absorption velocities for our sample, we can inspect the relationship between sodium velocity and other spectroscopic properties of our SN Ia sample. In Figure 13, we plot the silicon absorption ratio (R _Si), the HVF absorption strength (R _HVF), and the absorption strength (i.e. equivalent width) of sodium itself against the velocity of the sodium absorption feature. Based on results of Maguire et al. (Reference Maguire2013), we would expect SNe Ia with low R _Si and high R _HVF values (the high stretch, slow declining SNe Ia) to have a slight excess of blueshifted sodium absorption. Our sample size here is too small (and not well-selected) to make any robust statement about such preferences. However, we note that this analysis was not the explicit objective of our observations, but instead was a supplemental outcome facilitated by the nature of the WiFeS data.

Figure 13. Top: Silicon absorption ratio R _Si plotted against velocity centre of the narrow sodium absorption feature (as in Figure 10 we show the corresponding values of Δm ₁₅, though note the smaller range). Middle: HVF strength (R _HVF) plotted against sodium absorption velocity. Bottom: Absorption equivalent width of the combined D1+D2 sodium lines plotted against sodium absorption velocity. On the right axis of this panel, we use the relation of Poznanski et al. (Reference Poznanski, Prochaska and Bloom2012) to show the reddening values E(B − V) corresponding to the measured sodium equivalent widths if the absorption arises solely from the ISM—though this is unlikely to be true for all SNe Ia (Poznanski et al. Reference Poznanski, Ganeshalingam, Silverman and Filippenko2011; Phillips et al. Reference Phillips2013—see discussion in text). In all panels, higher resolution observations with the R7000 grating are displayed as green squares, whilst lower resolution R3000 observations are shown as blue circles.

Thus, WiFeS is an excellent instrument for measuring sodium absorption in SNe Ia, a key observable for investigating SN Ia progenitor systems. This is particularly true for observations taken with the R7000 grating, which provides sodium velocity uncertainties of order a few km s⁻¹, thus enabling a robust classification of the SN as being ‘blueshifted’ or ‘redshifted’. More importantly, perhaps, is the capability of WiFeS to observe a wide field of view around the SN. This enables a measurement of the local systemic velocity at the SN location, even in cases where emission from the host galaxy is weak at the SN location itself.

4.3. Spectroscopic evolution of SN 2012dn

The SN with the greatest number of spectroscopic epochs in AWSNAP is SN 2012dn, and we show its AWSNAP spectroscopic time series in Figure 14. SN 2012dn was a spectroscopically peculiar SN Ia whose photometric and spectroscopic evolution was studied extensively by Chakradhari et al. (Reference Chakradhari, Sahu, Srivastav and Anupama2014)—they found it to be similar to the ‘candidate super-Chandrasekhar’ SN Ia SN 2006gz (Hicken et al. Reference Hicken, Garnavich, Prieto, Blondin, DePoy, Kirshner and Parrent2007). For brevity, we will refer to this class of objects as ‘super-Chandra’, though we caution the reader that the origin of the extreme luminosity of these objects is still under debate (see, e.g. Taubenberger et al. Reference Taubenberger2011; Hachinger et al. Reference Hachinger, Mazzali, Taubenberger, Fink, Pakmor, Hillebrandt and Seitenzahl2012). We comment on only a few additional outcomes for SN 2012dn from the AWSNAP data, but refer readers to Chakradhari et al. (Reference Chakradhari, Sahu, Srivastav and Anupama2014) and Parrent et al. (Reference Parrent2016) for a thorough discussion of this interesting object.

Figure 14. AWSNAP time series of SN 2012dn, labelled by phase with respect to the date of maximum light (2012 July 24, as determined by Chakradhari et al. Reference Chakradhari, Sahu, Srivastav and Anupama2014). Note these observations come from the first semester of AWSNAP when observing time was allocated in multi-night blocks separated sometimes by a month or more.

We obtained a very high signal-to-noise spectrum of SN 2012dn at phase +91 d (with respect to the date of maximum light 2012 July 24, as determined by Chakradhari et al. Reference Chakradhari, Sahu, Srivastav and Anupama2014). At this epoch, the SN is beginning to enter the nebular phase when the ejecta become optically thin, revealing emission from the iron group elements (IGEs) near the centre of the SN. Spectra at these epochs provide an excellent diagnostic of the nucleosynthetic products of the SN explosion. In Figure 15, we present our +91 d spectrum of SN 2012dn compared to very late spectra of other candidate super-Chandra SNe Ia SN 2007if (Scalzo et al. Reference Scalzo2010; Yuan et al. Reference Yuan2010; Taubenberger et al. Reference Taubenberger2013) and SN 2009dc (Silverman et al. Reference Silverman, Ganeshalingam, Li, Filippenko, Miller and Poznanski2011; Taubenberger et al. Reference Taubenberger2011; Yamanaka et al. Reference Yamanaka2009; Tanaka et al. Reference Tanaka2010; Hachinger et al. Reference Hachinger, Mazzali, Taubenberger, Fink, Pakmor, Hillebrandt and Seitenzahl2012; Kamiya et al. Reference Kamiya, Tanaka, Nomoto, Blinnikov, Sorokina and Suzuki2012; Taubenberger et al. Reference Taubenberger2013), as well as the gold standard normal SN Ia SN 2011fe (Nugent et al. Reference Nugent2011; Li et al. Reference Li2011b; Parrent et al. Reference Parrent2012; Pereira et al. Reference Pereira2013).

Figure 15. SN 2012dn at its latest AWSNAP epoch (+91 d on 2012 October 23) compared to other candidate super-Chandra SNe Ia SN 2007if at +98 d (top panel, from Silverman et al. Reference Silverman, Ganeshalingam, Li, Filippenko, Miller and Poznanski2011) and SN 2009dc at +97 d (middle panel, from Taubenberger et al. Reference Taubenberger2011), as well as the normal SN 2011fe (bottom panel, from Pereira et al. Reference Pereira2013).

This comparison clearly reveals a strong spectroscopic similarity between SN 2012dn and the candidate super-Chandrasekhar SNe Ia, and a distinct dissimilarity with SN 2011fe. Perhaps, most prominent is the weaker Fe III line complex at ~ 4700 Å for the super-Chandra SNe Ia compared to SN 2011fe. This discrepancy is also evident in fully nebular spectra of super-Chandra SNe Ia at ~ 1 yr past maximum light, as discussed by Taubenberger et al. (Reference Taubenberger2013). This indicates that the ionisation state of the super-Chandra SNe Ia at these phases is different from normal SNe Ia—whether the diminished Fe iii emission arises from a higher or lower average ionisation state remains uncertain.

Additionally, the velocity profile of the emission features in the super-Chandra SNe Ia exhibits a marked difference to that of SN 2011fe (and other normal SNe Ia). The normal SN Ia profile appears very Gaussian (and indeed is generally well fit by a Gaussian profile—Childress et al. Reference Childress2015), whilst the super-Chandra velocity profile appears sharper at the centre. This is particularly evident for the line features at ~ 5 900 and ~ 6 300 Å, which at these epochs are dominated by Co iii. Further spectral modelling of this and other late-phase super-Chandra spectra may reveal important insights into the structure and composition of the super-Chandra ejecta.

Finally, the high resolution of WiFeS reveals narrow emission lines from the host galaxy of SN 2012dn. We isolated the narrow host galaxy lines using the longest exposure of SN 2012dn, the late-phase observation of 2012 October 23 (+91 d). We fit a simple linear ‘continuum’ near each emission line by fitting a line to the SN spectrum between 5 and 15 Å away from the line centre on both the blue and red sides of the line. We illustrate this technique and present the SN-subtracted emission features in Figure 16.

Figure 16. Extraction of host galaxy emission line flux (green) from late SN 2012dn spectrum (blue) using simple linear continuum fits (red).

In Table 3, we present the measured emission fluxes and errors for the major galaxy emission lines, all of which have been scaled by the observed flux in the Hβ line. By comparing the observed ratio of the Hα and Hβ lines to its expected value of 2.87 (the Balmer decrement Osterbrock & Ferland Reference Osterbrock and Ferland2006), we can determine the amount of reddening in the H ii regions giving rise to the host emission lines. We use the Cardelli, Clayton, & Mathis (Reference Cardelli, Clayton and Mathis1989) to find a host reddening of E(B − V) = 0.17 ± 0.10, a value remarkably similar to the SN reddening of E(B − V) = 0.18 (Milky Way plus host) determined by Chakradhari et al. (Reference Chakradhari, Sahu, Srivastav and Anupama2014). We correct the host emission line fluxes for the value E(B − V) = 0.17 and report the corrected values (which we also re-scale to the de-reddened Hβ flux) in Table 3.

Table 3. SN 2012dn local emission line fluxes.

^a Observer frame fluxes, scaled to Hβ.

^b De-reddened using Balmer decrement reddening of E(B − V) = 0.17 so that F(Hβ) = F(Hα)/2.87 with a CCM reddening law, and scaled to the de-reddened flux of Hβ.

With the de-reddened host galaxy emission line fluxes, we calculate a gas-phase metallicity at the site of SN 2012dn. The N2 method of Pettini & Pagel (Reference Pettini and Pagel2004, hereafter PP04) yields 12 + log(O/H) = 8.29 ± 0.04, whilst the O3N2 method of PP04 yields 12 + log(O/H) = 8.35 ± 0.03. The former value is remarkably close to the estimated site metallicity for SN 2006gz of 12 + log(O/H) = 8.26 calculated by Khan et al. (Reference Khan, Stanek, Stoll and Prieto2011) using the same metallicity method. If we convert the O3N2 value to the Tremonti et al. (Reference Tremonti2004) scale using the formulae of Kewley & Ellison (Reference Kewley and Ellison2008) as was done in Childress et al. (Reference Childress2011), we measure 12 + log(O/H) _T04 = 8.51 ± 0.04.

Comparing the above values to the solar oxygen abundance of 12 + log(O/H)_⊙ = 8.69 (Asplund et al. Reference Asplund, Grevesse, Sauval and Scott2009), we find the site metallicity for SN 2012dn is in the range 40–65% solar, depending on the chosen metallicity calibration. This sub-solar metallicity value is consistent with the previously reported trend for super-Chandra SNe Ia to prefer low metallicity environments (Taubenberger et al. Reference Taubenberger2011; Childress et al. Reference Childress2011; Khan et al. Reference Khan, Stanek, Stoll and Prieto2011).