Early observations of the nearby type Ia supernova SN 2015F

We used several instruments to obtain photometry of SN 2015F, the main characteristics of which are summarized in Table 1. Data obtained with the European Southern Observatory (ESO) Faint Object Spectrograph and Camera (v2) (EFOSC2) on the New Technology Telescope (NTT) were reduced with the PESSTO pipeline described in detail in Smartt et al. (2015), and the Las Cumbres Observatory Global Telescope Network (LCOGT) images were reduced using the Obsevatory Reduction and Acquisition Control Data Reduction pipeline (ORAC-DR Jenness & Economou, 2015). The reduction steps of the data obtained with the PROMPT telescopes are described in Pignata et al. (2011).

A optical photometric sequence of stars around the SN was calibrated in the $UBV$ and $gri$ bandpasses against standard stars from the Landolt (1992) and Smith et al. (2002) catalogues, respectively. The magnitudes of the optical photometric sequence around SN 2015F (see Fig. 1) are presented in Table 8 in the Appendix. Point spread function (PSF) fitting photometry was performed on the SN frames using daophot (Stetson, 1987), and calibrated using the photometric sequence around the SN. The optical photometry of SN 2015F is in Table 9.

The NIR data were obtained with the Son of Isaac (SofI) camera on the NTT, and were reduced using our own IRAF²²2IRAF is distributed by the National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation. scripts to create a clean sky image, which we subtract from the science images. Pixel-to-pixel variations were removed by dividing the science images by a flat field image. We used scamp (Bertin, 2006) to obtain an astrometric solution, and swarp (Bertin et al., 2002) to combine the dithered images into a single image. The NIR photometric sequence was calibrated by observing Persson et al. (1998) standard fields close in time and in airmass to the SN observations. The photometry of the NIR photometric sequence and of SN 2015F are presented in Tables 11 and 12, respectively.

The $UBV$, $gri$ and $JHK_{s}$ light curves of SN 2015F are shown in Fig. 2. Fitting a polynomial to the $B$-band light curve of SN 2015F, we measure a decline rate of $\Delta m_{15}(B)=1.35\pm 0.03$ mag, and we estimate the epoch of maximum light as MJD $57106.45\pm 0.02$ d (2015 March 25.4). Using the SiFTO light curve fitter (Conley et al., 2008), we determine a stretch of $s=0.906\pm 0.005$. A summary of the epoch of peak brightness, the peak magnitude, and the decline rate of the light curve ($\Delta m_{15}$) obtained from fitting a polynomial to the light curves of SN 2015F in each filter is presented in Table 2.

In Table 3 we summarize the values of the host galaxy reddening along the line-of-sight to SN 2015F ($E(B-V)_{\mathrm{host}}$), calculated from the optical colours (Phillips et al., 1999), and $V-$NIR colours (Krisciunas et al., 2004) of SN 2015F. The $E(B-V)_{\mathrm{host}}$ values derived using different methods are in good agreement with each other, and the weighted average is $E(B-V)_{\mathrm{host}}=0.085\pm 0.019$ mag; the uncertainty corresponds to the standard deviation from the weighted average.

We obtained optical spectra of SN 2015F with EFOSC2/NTT, with the Robert Stobie Spectrograph (RSS Kobulnicky et al., 2003) on the Southern African Large Telescope (SALT), with WiFeS (Dopita et al., 2010) on the Australian National University (ANU) 2.3-m Telescope, and with the FLOYDS spectrograph on the 2-m Faulkes Telescope South (FTS) at the Siding Spring Observatory. The EFOSC2 spectra were reduced as described in Smartt et al. (2015), and the FLOYDS data were reduced with the floydsspec pipeline³³3https://www.authorea.com/users/598/articles/6566. The RSS is a long slit spectrograph, the spectra of SN 2015F were reduced using the pysalt pipeline which is described in detail in Crawford et al. (2010). WiFeS is an integral field unit spectrograph, one dimensional spectra were reduced and extracted from the data cube using a PSF-weighted fit using pywifes pipeline (Childress et al., 2014a). In Table 5 we summarize the optical spectroscopic observations of SN 2015F. We correct all our spectra for Galactic extinction using a Cardelli et al. (1989) extinction law with $R_{V}=3.1$, and convert to the rest-frame using the recession velocity of the host galaxy. All reduced PESSTO data will be available from the ESO Science Archive Facility in Spectroscopic Survey Data Release 3 (SSDR3), due for submission in late 2016. Details will be posted on the PESSTO website⁴⁴4http://www.pessto.org. All our spectra are available from the WISeREP archive⁵⁵5http://wiserep.weizmann.ac.il/ (Yaron & Gal-Yam, 2012).

In Fig. 5 we present the Si ii $\lambda 6355$ velocity ($v_{\mathrm{Si}}$) measured from the minimum of the $\lambda 6355$ Å absorption feature observed near 6150 Å. In the figure, we compare with several low-velocity gradient (LVG) and high-velocity gradient (HVG) SNe Ia⁶⁶6Benetti et al. (2005) separated SNe Ia into three groups: High-velocity gradient (HVG) events, consisting of objects with a velocity gradient of Si ii $\lambda 6355$ $\dot{v}_{\mathrm{Si}}\geq 70$ km s^-1 d^-1 and $\Delta m_{15}(B)\leq 1.5$, low-velocity gradient (LVG) events, consisting of objects with $\dot{v}_{\mathrm{Si}}\leq 70$ km s^-1 d^-1 and $\Delta m_{15}(B)\leq 1.5$, and FAINT events with $\Delta m_{15}(B)\geq 1.5$.. We estimate $v_{\mathrm{Si}}$ at maximum light ($v^{0}_{\mathrm{Si}}$) and $\dot{v}_{\mathrm{Si}}$, by fitting a first degree polynomial to the $v_{\mathrm{Si}}$ measurements over $-7$ to $+30$ d, and interpolating to obtain $v^{0}_{\mathrm{Si}}=10400$ km s^-1 and $\dot{v}_{\mathrm{Si}}=50$ km s^-1 d^-1. These values place SN 2015F in the LVG group.

At phases prior to $-10$ d, SN 2015F presents a slower $v_{\mathrm{Si}}$ than SN 2011fe (see Fig. 5). The difference in $v_{\mathrm{Si}}$ between these two SNe at $-14$ d is $\sim$1000 km s^-1. After $-10$ d, SN 2015F shows a $v_{\mathrm{Si}}$ evolution similar to SN 2011fe and SN 2004eo.

We next compare in detail the spectra of SN 2015F to other SNe Ia from the literature. We focus our analysis at phases earlier than $-10$ d, and also make a comparison to sub-luminous SN 1991bg-like SNe Ia.

The Ca ii high velocity (HV) features are strong in the very early spectra of SN 2015F, dominating the Ca ii features until a week before maximum light. After these phases the photospheric component becomes dominant, and the HV component disappears around maximum light. In Fig. 8 we show the Ca ii H&K and the NIR triplet in velocity space for the $-14$ d to $-11$ d spectra. For Ca ii H&K, the expansion velocity is calculated with respect to the average wavelength of the H&K lines ($3951$ Å), while for the NIR triplet the velocity is with respect to the strongest line ($8542$ Å).

Following Childress et al. (2013); Childress et al. (2014b), Maguire et al. (2014) and Silverman et al. (2015), we fit Gaussian profiles to the Ca ii lines, defining a pseudo-continuum on either side of the profile, and using it to normalize the spectrum. To model the Ca ii NIR feature, we used both HV and photospheric components each composed of three Gaussian profiles with fixed relative positions and a common width. We allowed the relative strength of the three lines to vary, but to be the same for both components. The best-fitting parameters are listed in Table 6 and are shown in Fig. 8. The relative strengths are similar across the different epochs, and close to the theoretical values expected from atomic physics.

The Ca ii H&K region is more complex due to additional photospheric Si ii at $\lambda 3856$, coupled with blanketing from iron-peak elements that makes the definition of the pseudo-continuum less reliable. Following Maguire et al. (2014) and Silverman et al. (2015), we modelled each pair of Ca ii H&K lines using two Gaussians where the relative positions and their widths were fixed, and the relative strength of the two Gaussians fixed to unity. We consider the HV and photospheric Ca ii components, the Si ii $\lambda 3856$ line, and a weak feature at $\sim$3600 Å, possibly caused by the blending of several lines of iron-peak elements (the feature can be seen at $\simeq$27000 km s^-1 in Fig. 8). We fixed the position of the Si ii $\lambda 3856$ line to be within 5 per cent of the photospheric velocity of the Ca ii NIR line, and the velocity of the Ca ii $H\&K$ HV and photospheric components to be within 10 per cent of the values from the Ca ii NIR lines.

The Si ii feature is required in our fits from $-11$ d. The weak feature at $\sim$3600 Å is blended with the Ca ii H&K line at early stages, but around maximum light the feature becomes more detached, becoming an independent feature on the blue side of the H&K profile around peak brightness. The extension of the Ca ii material reaches a velocity of $\sim$29000 km s^-1 in the very outermost layers of SN 2015F.

In Fig. 10, we present the spectral sequence of SN 2015F around the C ii $\lambda 6580$ and $\lambda 7234$ lines in velocity space, compared to SN 2011fe. The C ii $\lambda 6580$ photospheric absorption is detected until $-4$ d at velocities similar to Si ii (see Fig. 9), although the measurement is difficult at phases prior to $-10$ d as the feature does not have a well defined minimum; see Fig. 10). We also note that around maximum light, the Si ii $\lambda 6355$ line becomes stronger while the C ii lines become weaker; thus P-Cygni emission from Si ii may affect the C ii feature by moving the minimum to redder wavelengths (lower velocities).

Fig. 10 also presents a spectral sequence showing the evolution of the $\sim 6800$ Å absorption feature. One interpretation for this feature is that it is HV C ii, which we discuss in Section 5.2.

Our favoured explanation of the $\sim 6800$ Å spectral feature in SN 2015F is photospheric Al ii (see Section 4), expanding at a velocity of $\sim 13000$ km s^-1 (Fig. 9). The Al ii material has to be confined in a relatively narrow range of velocity, as the $\sim 6800$ Å feature does not appear to evolve in velocity over 16 days (Fig. 10). However, we caution that the feature is quite weak and is affected by telluric absorption; a definitive statement about the velocity evolution is difficult to make.

Aluminium in SNe Ia has not been commonly reported in the literature. To our knowledge, the only previous claim was in the peculiar ‘.Ia’ (Bildsten et al., 2007) candiate SN 2010X (Kasliwal et al., 2010). ²⁷Al  is the only stable aluminium isotope, which according to nucleosynthesis calculations is $\sim 10^{3}$ more abundant than the radioactive ²⁶Al  isotope (Iwamoto et al., 1999; Seitenzahl et al., 2013). However, the expected mass fraction of ²⁷Al  in SNe Ia is relatively low, only $10^{-3}$ to $10^{-2}$ times the total mass of ²⁸Si  (Iwamoto et al., 1999; Seitenzahl et al., 2013), the latter being the most abundant silicon isotope in SNe Ia.

Given this low predicted abundance of Al, strong Al ii features seem unexpected. The yield of ²⁷Al obtained from the W7 (Nomoto et al., 1984) nucleosyntheis models of Iwamoto et al. (1999), and the three-dimensional N100 delayed-detonation models of Seitenzahl et al. (2013), predict a strong dependence of the abundance of ²⁷Al  on metallicity. A change from zero to solar metallicity in the progenitor white dwarf produces an increase of an order of magnitude in the yield of ²⁷Al by mass. As a comparison, the abundances of ¹²C and ²⁸Si remain essentially flat as function of progenitor metallicity. Thus a relatively metal-rich progenitor may help to explain the presence of Al in SN 2015F.

In Section 3.2 we noted the common presence of the $\sim 6800$ Å feature in SN 1991bg-like SNe. This class appears to explode in more massive, higher metallicity galaxies; we also note the non-detection of this feature in SN 2011fe which seems to be the result of a sub-solar metallicity progenitor (see Mazzali et al., 2014). As SN 1991bg-like SNe exhibit lower photospheric temperatures than normal SNe Ia, in principle, the presence of the Al ii lines could be explained by a temperature effect and not as a metallicity effect. However, in the case of SN 2015F a temperature effect can be ruled out by the simultaneous detection of the $\sim 6800$ Å feature with Si iii lines, which are a signature of a hot SN ejecta, and are strong at -6 d and -4 d (see Section 3.2.3).

A second explanation for the $\sim 6800$ Å feature is high-velocity (HV) C ii $\lambda 7234$; photospheric C ii is clearly detected. This suggests that the outermost layers ($\gtrapprox$18000 km s^-1) of SN 2015F are mostly unburned, consistent with the Mazzali et al. (2014) model for SN 2011fe, in which the outermost layers of the SN ejecta (>$19400$ km s^-1) are unburned, and are composed mainly of carbon. The fact that SN 2015F has a faster decline rate than SN 2011fe, and is thus a dimmer/cooler event, suggests a less efficient burning, and perhaps an even larger amount of unburned material in the outer layers than in SN 2011fe.

In recent delayed-detonation models (Seitenzahl et al., 2013), the outermost layers of the ejecta ($v_{exp}$ >$20000$  km s^-1) are mostly composed of carbon and oxygen, and this may explain any HV C ii. Nevertheless, we do not see a correspondingly strong HV O i line in SN 2015F. In Seitenzahl et al. (2013), carbon could also be present down to about $10000$ km s^-1, which may explain photospheric C ii in SN 2015F (but see also Mazzali et al., 2014).

Under the assumption that the 6800 Å absorption feature corresponds to HV C ii $\lambda 7234$, we show its velocity evolution in Figs. 9 and 10. The first spectrum of SN 2015F has lower signal-to-noise ratio implying a larger uncertainty in the minimum of the feature, located at $\simeq$16900 km s^-1 ($6828$ Å). The feature evolves getting weaker and moving to redder wavelengths with time. We measured the minimum of this absorption feature at phases $<-6$ d (note the measurement is affected by telluric on the red side; see Fig. 10), and we show its expansion velocity in Fig. 9. The feature appears confined to a narrow range in velocity space from $\simeq$18700 km s^-1 to $\simeq$17000 km s^-1, but is persistent, and is still present in the spectrum at $+2$ d. The feature is also observed in SN 2007af at slightly higher velocities until $-4$ d, and then disappears.

The possibility that the HV features of C ii and Ca ii are produced close in velocity space may suggest a common origin for the HV material (see Fig. 9). HV Ca ii features exhibit a plateau in their velocity evolution between $-10$ and $-4$ d (Fig. 9). At the same phase, the velocity measured for the possible HV C ii $\lambda 7234$ feature is similar, but slightly lower.

Using syn++ to model the spectra of SN 2015F (see Section 4), we have identified lines of Ti ii, V ii, Cr ii, and Fe ii expanding at $\simeq$14800 km s^-1. This implies a non-negligible amount of iron-group elements in the region between $15000$ to $20000$ km s^-1of the SN ejecta. Hatano et al. (1999) reported strong Fe ii absorptions in the $-12$ d spectrum of SN 1994D at $\sim$4300 and $\sim$4700 Å, and included in their synow model a HV Fe ii component extending from $22000$ km s^-1 to $29000$ km s^-1, to reproduce these features.

Strong and broad lines of iron-peak elements at such high-expansion velocities are generally unexpected in SNe Ia, since the pre-expansion suffered by the layers at higher velocities than $10000$ to $13000$  km s^-1 will decrease the density too much to burn the material to iron-peak elements. Only in the case of a rapid transition in the burning speed front, from a sub-sonic deflagration to a super-sonic detonation, would the flame burn the outermost layers ($v_{\mathrm{exp}}$ >$13000$  km s^-1) to iron-peak elements, yielding mainly radioactive ⁵⁶Ni and not enough IMEs to reproduce the characteristic spectral features of normal SNe Ia. The decay of ⁵⁶Ni mixed in the outermost layers can then heat the ejecta, producing strong lines of doubly-ionized species such as Fe iii and Si iii as in the brightest SNe. In the earliest spectra of SN 2015F ($<-12$ d), this is not observed; by contrast, SN 2015F exhibits a spectrum dominated mainly by singly-ionized species (Fe ii, Si ii, S ii, Ca ii), consistent with a normal or relatively low ejecta temperature. At later phases, from about $-11$ d, SN 2015F begins to exhibit Si iii $\lambda 4560$ and Si iii $\lambda 5540$ lines, now suggesting heating from the decay of radioactive material mixed in the outer layers of the SN ejecta.

The absorption features produced by iron-group elements in the very early spectra of SNe Ia could be explained by iron-peak elements synthesized during the SN explosion and mixed to the outermost layers of the SN ejecta, or as absorptions of iron-peak elements present in the white dwarf surface at the moment of the explosion (see Lentz et al., 2000) – or a combination of both. Recent three-dimensional delayed detonation models predict that freshly synthesized iron-peak elements are located mainly at intermediate velocities ($\sim$3000 to 10000 km s^-1; Seitenzahl et al., 2013), the latter not corresponding with the observations of SN 2015F. However, Seitenzahl et al. (2013) remark that models with a strong (turbulent) deflagration phase, which are rather symmetric under rotation on large scales, exhibit strong inhomogeneities in the burning products on small scales. This may explain pockets of iron-peak material, observed at high-velocity, mixed in the outermost layers as in the case of SN 2015F.

To disentangle the metallicity of the progenitor from the fraction of freshly synthesized iron-group elements mixed to the outermost layers would require a detailed modelling using the abundance tomography technique (Stehle et al., 2005; Mazzali et al., 2005; Tanaka et al., 2011; Hachinger et al., 2013; Mazzali et al., 2014). This is beyond the scope of this paper, and will be the subject of a future article.