Probing the Progenitors of Type Ia Supernovae using Circumstellar Material Interaction Signatures

This work aims to study different probes of Type Ia supernova progenitors that have been suggested to be linked to the presence of circumstellar material (CSM). In particular, we have investigated, for the first time, the link between narrow blueshifted NaID absorption profiles and the presence and strength of the broad high-velocity CaII near infrared triplet absorption features seen in Type Ia supernovae around maximum light. With the probes exploring different distances from the supernova; NaID>10$^{17}$cm, high-velocity CaII features<10$^{15}$cm. For this, we have used a new intermediate-resolution X-shooter spectral sample of 15 Type Ia supernovae. We do not identify a link between these two probes, implying either that, one (or both) is not physically related to the presence of CSM or that the occurrence of CSM at the distance explored by one probe is not linked to its presence at the distance probed by the other. However, the previously identified statistical excess in the presence of blueshifted (over redshifted) NaID absorption is confirmed in this sample at high significance and is found to be stronger in Type Ia supernovae hosted by late-type galaxies. This excess is difficult to explain as being from an interstellar-medium origin as has been suggested by some recent modelling, as such an origin is not expected to show a bias for blueshifted absorption. However, a circumstellar origin for these features also appears unsatisfactory based on our new results given the lack of link between the two probes of CSM investigated.


INTRODUCTION
Type Ia supernovae (SNe Ia) are the explosions of white dwarfs in binary systems, play a major role in our understanding of the expansion rate of the Universe and are essential for constraining the properties of dark energy to high precision (e.g., Scolnic et al. 2018). Despite this, there are still major gaps in our understanding of their underlying progenitor systems. While the observed rate of SNe Ia is measurable by sky surveys (e.g., Li et al. 2011;Frohmaier et al. 2019), the rate of the underlying progenitor channels is much more uncertain, with the likely possibility that there is more than one way to make a SN Ia (e.g., Maoz et al. 2014;Jha et al. 2019). Classical SN Ia models involve an exploding white dwarf that is close to the Chandrasekhar mass, which has accreted matter from a non-degenerate companion star and resulting in a runaway thermonuclear reaction (Whelan et al. 1973). Models involving the ex-plosion of sub-Chandrasekhar mass white dwarfs have also gained popularity and the explosion can occur through a violent or dynamical merger (Pakmor et al. 2010) or the triggering of a He-shell on the white dwarf surface, which results in a detonation of the core ('double-detonation' ;Taam 1980;Livne & Arnett 1995;Shen & Bildsten 2009;Fink et al. 2010).
One observational method that has been used to distinguish between potential SN Ia progenitor scenarios is the presence of circumstellar material (CSM), which is thought to be more likely in the case of a non-degenerate companion star (Hachisu et al. 1999a,b), although may be present in some merger scenarios (e.g., Raskin & Kasen 2013). Time-varying blueshifted Na D absorption features have been identified in some SNe Ia (Patat et al. 2007;Simon et al. 2009;Blondin et al. 2009Blondin et al. , 2017Stritzinger et al. 2010;Sternberg et al. 2014), with the interpretation that these were due to outflowing (hence blueshifted) CSM close to the SN (distances of ∼ 10 17 cm) and not interstellar material (ISM). Additionally, an excess of blueshifted compared to redshifted Na i D absorption features has been identified in larger SN Ia samples, consistent with a CSM origin in ∼20 % of SNe Ia (Sternberg et al. 2011;Maguire et al. 2013).
Potential links between the presence of blueshifted Na i D absorption and the properties of the SNe Ia themselves (Si velocity, colour at maximum) have been suggested (e.g., Foley et al. 2012;Maguire et al. 2013;Phillips et al. 2013). SNe Ia displaying blueshifted Na i D absorption are associated with redder B-V colours at maximum and in some cases also have higher Si velocities (but this trend is not seen in all studies e.g., Maguire et al. 2013). A link between SNe Ia showing high-velocity Si features and redder B-V colours at maximum was also previously suggested by Wang et al. (2009). SNe Ia with stronger Na D absorption in low-resolution spectra also have a preference for redshifted late-time emission line velocities (Förster et al. 2012). Phillips et al. (2013) found that SNe Ia displaying blueshifted Na D have stronger Na column densities compared to what would be expected from their colour-derived extinction. This was not the case for their non-blueshifted sample, where the extinction and Na column densities followed the Milky Way relation. This study concluded that the strong Na column densities for the blueshifted Na D sample was potentially evidence for outflowing CSM from these systems but there does not appear to be a direct relation; some SNe Ia that are identified as having time-varying Na D (thought to be a clear indicator of CSM interaction) do not follow this relation.
The ratio of total-to-selective extinction, R , measured in the Milky Way has a typical value of ∼ 3.1 (Fitzpatrick 1999), though a range of values are observed (Mörtsell 2013 found a value ∼3 with an uncertainty of around 10%, whereas Nataf et al. 2013 find a R value of 2.5 in the direction of the galactic core). Much lower values (R < 2) have been measured for some highly extincted SNe Ia (e.g., Amanullah et al. 2014;Burns et al. 2014), with these measurements confirmed through spectropolarimetry (Patat et al. 2015;Zelaya et al. 2017). However, whilst not measuring the parameter in the same way, larger cosmological samples of normal SNe Ia have identified a range of R , from R ≤ 2 (e.g., Astier et al. 2006;Conley et al. 2007), to values closer to 3 (e.g., Fink et al. 2010;Mandel et al. 2011). The cause of the low R values seen in some SNe Ia is debated. Some studies have suggested it could be due to multiple scattering in CSM around the SNe Ia (Wang 2005;Goobar 2008;Amanullah & Goobar 2011), due to enhanced Na abundances in the CSM (Phillips et al. 2013). Other studies have concluded that it is mainly produced in the ISM (Bulla et al. 2018a,b), where cloud-cloud collisions from the radiation pressure generated by the exploding SNe themselves may produce smaller dust grains (Hoang 2017;Giang et al. 2020).
High-velocity components in the Ca NIR triplet absorption feature are observed in nearly all SNe Ia spectra obtained around maximum light (Hatano et al. 1999;Gerardy et al. 2004;Mazzali et al. 2005;Childress et al. 2013a,b;Maguire et al. 2014), though the strengths of such components vary significantly between objects. The origin of these high-velocity components (with velocities a few thousand km s −1 greater than the photospheric components) is unclear but it has been proposed that the high-velocity components of the Ca NIR triplet could have contributions from CSM (at distances of ∼ 10 14 cm) (Gerardy et al. 2004;Mazzali et al. 2005;Quimby et al. 2006;Tanaka et al. 2008) or they may be intrinsic to the SN ejecta (Branch et al. 2006;Blondin et al. 2012). If a clear relation between blueshifted Na i D absorption and high velocity Ca ii absorption were to be established, it would be a strong indi-cation of local CSM interaction during the early phases of SNe Ia evolution.
The focus of this paper is to investigate the link between two potential probes of the presence of CSM around SNe Ia; the presence of blueshifted Na i D absorption features and the properties of Ca high-velocity features. We do this using a new sample of intermediate-resolution spectra of SNe Ia around maximum combined with the literature samples of Sternberg et al. (2011) and Maguire et al. (2013). Section 2 outlines the observations and reduction techniques employed in the construction of the dataset used in this work, while Section 3 details the analysis of the sample of spectra obtained, along with their complementary photometric observations and host galaxy measurements. Section 4 describes the main results and Section 5 discusses the implications of these results in the context of the progenitors of SNe Ia and previous studies. Section 6 presents a summary of our main findings. Throughout this paper, we assume a Hubble constant, H 0 of 70 km s −1 Mpc −1 and adopt a standard cosmology with Ω =0.27 and Ω Λ =0.73.

OBSERVATIONS AND DATA REDUCTION
In this section, we describe the new data sample used in this work of maximum light X-shooter spectra of 16 SNe Ia. In Section 2.1, we describe the reduction process and telluric correction. In Section 2.2, we discuss the estimates of the light curve properties (light curve width, colour at maximum), as well as the spectroscopic subclassification of each SN Ia in the sample. In Section 2.3, we discuss additional literature objects that are included to increase the sample size.

X-shooter spectroscopic observations
The primary dataset used in this work consists of maximum light (within approx. one week of peak) medium-resolution X-shooter spectroscopy of 16 SNe Ia with wavelength coverage of ∼3000 to 25000 Å across three arms (UVB, VIS, NIR) (Vernet et al. 2011). The SNe were chosen for observation with X-shooter in an unbiased manner, using the same criteria as Maguire et al. (2013) i.e., the target SNe Ia had already been spectroscopically classified and reported on the Transient Name Server (TNS) prior to maximum light to eliminate the risk of contamination from other types of transient with a redshift < 0.03 to enable a high signal to noise spectrum to be obtained. A summary of the sample, including their discovery information and properties of their host galaxies, is presented in Table 1.
The X-shooter spectra were reduced in a standard manner using the ESO Reflex reduction pipeline to produce flux-calibrated spectra (Modigliani et al. 2010;Freudling et al. 2013). Additional telluric corrections were applied to the VIS and NIR arm of the majority of the sample with the MOLECFIT software (Smette et al. 2015;Kausch et al. 2015). A telluric correction was not applied for 2 SNe Ia (SNe 2017hn and 2017bkc) for the NIR arm because of low signal-to-noise (S/N). SN 2017guh was consistent with the core of its host galaxy and suffered from significant host galaxy contamination and therefore is removed from further analysis. The spectra were corrected for Galactic extinction using the ( − ) values from Schlafly & Finkbeiner (2011) and the extinction law of Fitzpatrick (1999). The definition of the redshift of each SN is central to this analysis and is discussed further in Section 3.1.1. Table 1. Discovery and host galaxy information for the SN Ia sample. The discovery survey is that reported on the TNS. Two redshifts are listed, the heliocentric value from NED along with the final adopted redshift. The redshift calibration method for each adopted redshift is listed as either emission features (a mean value if there is more than one feature) seen in the X-shooter spectra or if no lines were seen, the value from NED is used.  (Sand et al. 2018). * No heliocentric host redshift is available on NED, with the host z taken from the original TNS classification report * * SN 2017cbv is notable for having a significant early excess in the bluer bands of its light curve (Hosseinzadeh et al. 2017). It was also investigated by Ferretti et al. (2017) exploring if it displayed any time variable Na D absorption features -none where detected. * * * SN 2017guh is excluded from further analysis due to strong host galaxy contamination.

Light curve properties and spectral classification
Light curves from the Asteroid Terrestrial-impact Last Alert System (ATLAS Tonry et al. 2018;Smith et al. 2020) sky survey were used to constrain the photometric properties of 9 events in the sample (as indicated in Table 2). Photometric information for these objects can be retrieved from the ATLAS forced photometry server (Shingles et al. 2021). A measure of the light curve width ('stretch') was estimated using the SiFTO light curve fitting code (Conley et al. 2008) for these events. The maximum light B-V colour was also estimated using SiFTO for two events (SN 2017ckq, SN 2017gah) with multi-band (ATLAS cyan and ATLAS orange) photometry around maximum light. For three objects in the sample, photometry from the literature was fit using SiFTO: SN 2017cbv (Hosseinzadeh et al. 2017 The 'stretch' values, along with available B-V colours at maximum, for the sample are presented in Table 2. The stretch range (∼0.82-1.44) is consistent with that found for the low-Palomar Transient Factory (PTF) SN Ia sample (Maguire et al. 2014).
Since light curve colour measurements were only available for five events, B-V colours around maximum were also estimated using the X-shooter spectra. These spectra were integrated through B and V band filters to obtain the colour at the phase of the spectrum. We have excluded SN 2016hvl from our spectral colour comparisons because its spectral phase (−7.8 ± 0.46 d) is significantly earlier than the rest of the sample of −4 to +5 d from peak.
The instrument response of X-shooter is relatively stable, with an estimated relative flux uncertainty of five per cent across the spectrum (Vernet et al. 2011). Since we are only interested here in the colour, the absolute flux calibration is not relevant. As a sanity check on the magnitude of this uncertainty, we have compared the photometric B-V values obtained at peak against the values measured from X-shooter spectra for four events (SN 2012cg, SN 2012ht, LSQ12dbr, SN 2013aj) from Maguire et al. (2013) and find agreement within the uncertainties for the four events. For the five events in our sample with both photometric and spectroscopic colour measurements, we have also compared these values and find that they are in agreement except for a slight discrepancy (∼1-) for SN 2017cbv, which we attribute to the earlier phase of its spectrum. To quantify the additional uncertainty that will be introduced to changes in phase, we have estimated the colour change within a phase range of ±4 d for the well-studied event SN 2017cbv (Hosseinzadeh et al. 2017) to be ∼0.1 mag, which is also consistent with the range for SN 2011fe and SN 2012cg (see Figure 2 of Hosseinzadeh et al. 2017). We have added this additional uncertainty due to phase in quadrature (see Table 2).
The SNe Ia in the sample were sub-classified using the SN IDentification SNID code (Blondin & Tonry 2007) applied to their X-shooter spectra. The bulk of the sample (11/15) were classified as 'normal' SNe Ia, three as 91T-like events, and one as a 91bg-like event (see Table 2). SN 2017fzw is classified as a 'normal' SN Ia based on the SNID comparison but has some unusual properties, as will be discussed further in Galbany et al. (in prep.).

Literature comparison sample
In order to increase the number of objects available for inclusion in the statistical analysis of this work, the 15 newly presented objects are supplemented by the inclusion of the combined dataset described in Maguire et al. (2013). This additional dataset consists of 16 SNe Ia observed with X-shooter and 16 SNe Ia from Sternberg et al. (2011) with similar intermediate-resolution spectra around maximum light. Our combined sample (referred to as the 'full' sample) consists of 47 SNe Ia. We note that the sample of Sternberg et al. (2011) has a weak selection effect, with six of the objects selected based on identification of strong narrow Na i D absorption features e.g., SN 2006X.

ANALYSIS
In this section, we describe the analysis performed to calibrate the spectra to the rest frame and determine the presence of blueshifted material (Section 3.1), the host galaxy morphologies and stellar mass measurements of the sample (Section 3.2), and the measurements of the broad SN absorption components of Si and Ca .

Identification of narrow absorption features
As the narrow features displayed by the Na i D and Ca H&K lines are at low velocities, displaying shifts of only a few hundred km s −1 from their expected rest positions (Patat et al. 2007), accurate redshift calibration is key in determining if any observed features are red or blueshifted. Any contaminating absorption features produced by unrelated interstellar material in the host galaxy would also be expected to be detected within a few hundred km s −1 of the SN, given the typical velocity distributions of material in galaxies. Such contamination is expected to produce an equal amount of both 'blue' and 'redshifted' absorption features (with respect to the SN rest-frame) with potential SN-generated features being exclusively blueshifted, due to their origin in out-flowing stellar material and the associated line of sight effects (but see Hoang 2017).

Redshift and zero-velocity calibration
To determine the accurate rest frame of each SN, measurements were made using host galaxy emission features identified in the two-dimensional spectral frames, close to the SN position. The host galaxy lines used were the H and N ii 6583 Å due to their proximity to the most important feature in this analysis, Na D. The lines used in the determination of the redshift for each event and the measured redshift is given in Table 1. For seven events in the sample, no host galaxy features could be identified in the Xshooter spectra. This was due to either the SN having a significant offset from its host and/or the host being an early-type galaxy that lacks strong emission features. This could potentially result in larger uncertainties due to the lack of an accurate rest frame correction at the SN position. However, for the seven events without identified host features, none showed absorption profiles for Na D -thus these cases are unaffected. The adopted redshift in Table 1 for these events is the heliocentric redshift retrieved from NED 1 .
Based on these redshifts, the 'position' of the zero-velocity for each spectrum can be obtained and the classification of the Na D and Ca H&K features into 'redshifted', 'blueshifted', 'red and blueshifted' or 'no absorption' with respect to the zero rest-frame position from the host galaxy lines can be made. This schema mirrors that used in Maguire et al. (2013), which was modified from that of Sternberg et al. (2011). We again note that those SNe calibrated using the host redshifts have increased uncertainties due to galactic rotation effects (though no SNe in this sample with observed Na D absorption was calibrated in this manner). Additionally if the material responsible for producing the narrow galaxy lines is located somewhat in front or behind the true SN position the calibration will also be affected though these offsets are expected to be small with the use of local lines from the 2D spectral frames. Figures 1 and 2 show the SNe with 'blueshifted' and 'red and blueshifted' Na D absorption profiles, Figure 3 shows an event (SN 2017cbv) identified with a redshifted Na D absorption profile (consistent with higher resolution spectra obtained with UVES Ferretti et al. 2017 studies of time-varying narrow features, (e.g., Patat et al. 2007) the Ca features are not seen to vary, which is suggested to be due to their higher ionisation potential compared to Na D. We find that the Ca classifications are generally in agreement with those of the Na D, with just two having different classifications, SN 2017fgc where the Ca H&K feature had blue and redshifted absorption while the Na D had no absorption and SN 2017gvp where the Ca H&K had blueshifted only absorption and the Na D had blue and red. A similar small number of inconsistencies between these features was also identified in Maguire et al. (2013). As in previous studies, we focus on the Na D features and only use the Ca H&K feature for confirming weak Na D features.

Continuum calibration and pseudo-equivalent widths
To make measurements of the pseudo-equivalent widths (pEQWs) of the narrow Na i D features, any continuum offset needs to be removed. This was done using a linear interpolation between uncontaminated regions on either side of the Na i D feature. The pEQWs of each feature were then measured following the method of Förster et al. (2012). Following previous works, the measurement of the stronger of the two lines making up the Na D doublet (Na i D 2 ) is used as the measure of the feature's overall strength.
To gain an understanding of overall absorption strength, as well as the strength of only the blueshifted absorption component, the measurement of the pEQW has been made in two velocity re- , the one object displaying redshifted Na i D absorption features. The method of redshift classification is also indicated below the spectra, with 'G' for a galactic redshift estimate from the literature. This classification is supported by the UVES observations described in (Ferretti et al. 2017) gions relative to the calibrated position of zero velocity. The overall strength of the feature measured in the region −200 -+200 km s −1 ('total pEQW') with the pEQW of the blueshifted component measured from −200 -0 km s −1 . These values are tabulated in Table  2. Some previous studies have used larger velocity ranges in the calculation of the pEQWs of Na D, such as Förster et al. (2012) who measured the features out to ± 300 km s −1 . However, the higher resolution of the X-shooter spectra allows a narrower region to be used.
Both these pEQW measurements have been made for the 15 new SNe Ia explored in this work, including those for which a Na D absorption feature could not be visually identified, to serve as a confirmation of the non-detection of these features. The regions selected for the continuum were varied randomly within 15 Å of their initially selected locations and the calculation repeated for a total of 5000 measurements to provide a measure of the uncertainty associated with the pEQW measurement. The pEQW values for the literature sample were taken from Maguire et al. (2013), which were obtained using an analogous method.

Host galaxy analysis
To search for correlations between the properties of SNe Ia and their environments, that could in turn be related to differing progenitor scenarios, we perform an analysis of the host properties of our SN sample. This work uses both visual morphological classification of the host galaxies of the SN Ia sample, along with photometric determinations of the total stellar mass of the hosts, where suitable data are available. We also calculate the projected distance between each SN and the nucleus of its host.

Host morphological classification
Host morphological classifications for the SN Ia sample were retrieved from NED and are sourced from the classification catalogues of Lauberts et al. (1981) and de Vaucouleurs et al. (1991). For two objects (SN 2017awz, SN 2017bkc), visual classification as spiral galaxies was performed as part of this work. These morphological classifications are given in Table 1    Morphological classifications for the hosts of the SNe included in the combined sample of objects from literature were adopted from their description in Sternberg et al. (2011) or Maguire et al. (2013), apart from four cases, where updated classifications were used. These objects are SN 2008C, which is updated from 'S0/a' to 'S' based on visual inspection of more recent Pan-STARRS 1 (PS1 Chambers et al. 2016) images, SN 2008fp is updated from 'Sa' to 'S0' (de Vaucouleurs et al. 1991), SN 2012fw is updated from 'S0/a' to 'S' based on the classification of Loveday (1996), and PTF12jgb is updated from unclassified to 'irregular' based on PS1 imaging.

Host stellar mass estimates
To estimate the stellar mass of the hosts of the SNe within this sample, we used photometric measurements of the host galaxies from the Sloan Digital Sky Survey (SDSS; Fukugita et al. 1996;Gunn et al. 1998Gunn et al. , 2006Blanton et al. 2017;Doi et al. 2010) and applied eqn. 8 of Taylor et al. (2011) as was used in Rigault et al. (2018). This equation provides a method of obtaining the stellar mass based on g and i band imaging and has an accuracy of ∼ 0.1 dex. The most appropriate magnitudes tabulated for SDSS for this analysis were found to be the Petrosian magnitudes (Taylor et al. 2011). Twenty one of the SNe Ia in our full sample have available photometry in SDSS Data Release 16 (Ahumada et al. 2020), and were obtained using the CasJobs data retrieval system. The photometry for the host of SN 2012ht was found to be unreliable due to potential blending with a coincident foreground star, and has been replaced by photometry, also in the SDSS filter system, from Mazzei et al. (2018). The estimated stellar masses using the method of Rigault et al. (2018) are given in Table 3. The mean of the log of the host stellar mass of this sample is determined to be 10.3, with a standard deviation of 0.7 dex. The log of the mean host stellar mass from the PTF sample of Pan et al. (2014) was 10.2 with a standard deviation of 0.8 dex, which is consistent with this sample within the uncertainties.

Projected distance from host nucleus
The projected distance between each SN and the centre of it's host has been determined assuming that SN and host are located at the same distance and are in the same plane i.e., inclination effects are not considered. The full results of these calculations are shown in Figure 7. It is expected that ISM effects will decrease with increasing galactocentric distance making it more likely that objects at large distances showing strong blueshifted Na D absorption are the result of CSM effects. The majority of the objects within the sample, regardless of Na D absorption classification are found at projected distances < 12.5 kpc, though two objects (LSQ12fxd and LSQ12gdj) with blueshifted Na i D absorption are observed at larger distances (26.5 and 17.5 kpc respectively). More generally, the means for each Na D absorption classification are found to be consistent (Blueshift: 6.5±7.2 kpc, Blue and redshift: 3.2±1.9 kpc, Non blueshift: 8.9±8.5 kpc, No absorption: 6.2±4.6 kpc) indicating no statistical relationship between Na D absorption classification and the projected distance between SN and host nucleus.

Line measurements of broad SN features
One of the main aims of this paper is to determine if there is a relation between the presence of blueshifted Na D features and the broad SN features of Si 6355 Å and the Ca NIR triplet seen in SNe Ia around maximum light. The typical identified broad features are shown for SN 2017fgc in Figure 6. In particular, we are interested in quantifying the presence of high-velocity components to these features with velocities 1500km s −1 higher than the photospheric contribution, which have been previously attributed to interaction with CSM (e.g., Mazzali et al. 2005). The identification of a relation would provide evidence of blueshifted Na D features being intrinsic to the SN system and likely not coming from the ISM.

Fitting procedure of the broad Si and Ca NIR features
The Si 6355 Å and Ca NIR triplet features were fit using Gaussian profiles broadly following the method of e.g., Childress et al. (2013b); Maguire et al. (2014). The spectra were rebinned to 1 Å resolution to increase the S/N and a sigma clipping was applied to remove any residual narrow skylines or other residual artefacts that were not corrected in the reduction process to prevent such features from influencing the fitting. The continuum was removed by selecting regions on either side of the feature and removing with a linear fit as described for the narrow Na D features. To investigate the presence of high-velocity features in the Si 6355 Å and Ca NIR triplet regions, we have fitted both features under two conditions: i) the presence of only photospheric-velocity components and ii) with the additional presence of high-velocity components.
The photospheric components of the Si feature were fit as a doublet with the relative strengths set assuming the optically thick regime (Childress et al. 2013b;Maguire et al. 2014), with the full width half-maximum (FWHM) and velocity offset constrained to be the same for both doublet features. The high-velocity components of the Si were set up in the same manner, with the high-velocity component values independent of the photospheric components. The same procedure was used for the photospheric and high-velocity components of the Ca NIR triplet, with three linked features for each component.
The weighted mean values of the relevant fit parameters (velocity, FWHM and pEQW) were determined by varying the position of the continuum selection region and refitting 5000 times, with the uncertainties given as the standard deviation of the successful fitting parameter values, combined in quadrature with the redshift and spectral resolution uncertainty for each object. In these fits, the absorption minima velocity of the fitted high-velocity feature must exceed that of the photospheric component by at least 1500 km s −1 . Based on the results of previous studies, we also make the assumption that the photospheric component is dominant and so the amplitude and width of the high-velocity component must be less than those of the photospheric component. Additionally the widths of both components must exceed 1000 km s −1 to disallow any fits where a narrow noise feature within spectrum has dominated the results. The Ca NIR triplet spectral region is less clean than the Si 6355 Å region, with a potential contribution from the O 8446 Å feature. We have performed tests including this potential contribution but do not obtain reliable results. Therefore, the O 8446 Å feature is not included in our fitting. To ensure consistency of results, the spectra forming the combined dataset of Maguire et al. (2013) and Sternberg et al. (2011) have also been fitted using this method where the necessary data are available. These results are presented in Tables A1 and A2

Determination of the presence of high-velocity components
The presence of a high-velocity Si feature is determined by the comparison of the fits with and without a high-velocity component using the Bayesian information criterion (BIC, Schwarz 1978), where the inclusion of a high-velocity component is strongly favoured if ΔB IC is > 10. Additionally, the fitting of a high-velocity feature must be successful in at least 500 iterations (at least 10 % of iterations) to prevent the comparisons being unduly weighted by a small number of outlying fits. The results of the fits for the Si feature are shown in Table 4, with example fits shown in Fig 8. Three SNe Ia (SNe 2017awz, 2017azw, 2017cbv) in the sample do not require the inclusion of high-velocity Si component and are well fit using a photospheric component alone. All remaining objects satisfy the requirements for a high-velocity component based on the BIC and on the number of successful fitting iterations. In a similar manner, the presence of a high-velocity Ca was also determined using the same constraints on BIC and number of successful fitting iterations (Table 5). Figure 9 showing example fits to several objects within the sample. A Ca feature was not observed in SN 2016hvl, likely due to its early phase (−8 d) and 91T-like classification. All of the remaining 14 objects satisfy the selection criteria and are thus identified as displaying high-velocity Ca absorption.

RESULTS
The following section first outlines the results of the analysis conducted of the broad SN features and the narrow Na D absorption features. It then describes the connection between the host galaxy properties and the Na D features, as well as quantifying the observed excess of blueshifted Na D absorption features. Following this, results of comparisons between the light curve properties and Si features to the classification of Na D features are described, before exploring similar comparisons with the properties of the Ca feature.

Broad SN features and the presence of high-velocity components
Fitting of the broad Si 6635 Å features found that three of the 15 objects (20 %) in the new sample were well described using a purely photospheric feature, with the rest being better fit with the inclusion of a high-velocity component ( Table 4). The mean photospheric components (velocity, pEQW, FHWM) of the subsample with and without a high-velocity are found to be consistent within the uncertainties. The fitting of the broad Ca NIR triplet absorption feature was found to heavily favour the inclusion of a   high-velocity feature, with all 14 objects for which it was possible to fit the Ca NIR triplet region found to be better described with the inclusion of a high-velocity component. The values measured for the fits are given in Table 5. The velocities measured for the Si and Ca photospheric components are found to be consistent within the uncertainties. However, values given in Table 5 show that the high-velocity components of the Si features are narrower and lower velocity than those measured for the Ca NIR triplet.

Na D equivalent widths
As outlined in Section 3.1.2, measurements of the strength for the Na i D features have been made, using the stronger of the two lines (Na i D 2 ) as a proxy for overall strength. Previous studies have suggested that those SNe Ia showing blueshifted Na i D absorption have a tendency to display stronger overall Na i D absorption features (higher pEQW) compared to the rest of the population (Phillips et al. 2013). When those SNe Ia showing 'no absorption' and the 'blue and redshifted' absorption are removed from the comparison in the full sample, the remaining objects (i.e. those showing 'only blueshifted' or 'non blueshifted' Na D absorption) are well described by a single pEQW distribution. Whilst the median values for the blueshifted and non-blueshifted groups are rather different (0.64 and 0.24 respec-tively), the corresponding mean values are found to be consistent at 0.46 ± 0.34 Å and 0.40 ± 0.39 Å respectively, additionally a K-S test indicates that the two groups are drawn from the same overall population (K-S statistic: 0.333, p-value: 0.657). We note that the non-blueshifted Na D sub-sample is small with only six objects but we do not identify any clear difference between them. This result is shown in Figure 10. As would be expected, this plot also shows that objects showing blue and redshifted Na D absorption tend to have stronger overall absorption features, which is unsurprising since these objects have multiple absorption components. The pE-QWs of the blueshifted components of objects with both blue and redshifted Na D absorption features are of comparable strength to those of objects showing purely blueshifted Na D absorption.

Host galaxy distribution and connection to narrow Na D absorption features
Previous studies have shown that SNe Ia with late-type host galaxies (spiral galaxies) typically have brighter absolute magnitudes and broader light curves (larger 'stretch' values), compared to those hosted by early-type galaxies (elliptical and lenticular galaxies), which have been seen to host intrinsically fainter and faster evolving events (Hamuy et al. 1995(Hamuy et al. , 1996(Hamuy et al. , 2000Riess et al. 1999). It has also been observed that SNe Ia occurring in late-type galaxies have a higher incidence of blueshifted Na i D absorption features compared to those in early-type galaxies which are more likely to show no Na i D absorption features (Sternberg et al. 2011;Maguire et al. 2013). The sample of new SNe Ia in this work confirms this previously observed trend with all six of the SNe with blueshifted Na i D absorption features associated with late-type spiral hosts, compared to just two of the SNe Ia showing no or non-blueshifted Na i D absorption features (left panel of Figure 11). The full combined sample of objects also shows a similar distribution (right panel of Fig 11). Figure 12 provides a comparison between the blueshifted Na i D EQW and the measured host stellar masses of the full sample. Given the reduced number of objects with available host mass estimates, no clear distinctions between the different sub-samples based on Na i D EQW can be identified, with the log mean host stellar masses of the sub-samples being consistent, 9.93 ± 0.59 M ('blue and redshifted'), 10.29 ± 0.13 M ('blueshifted only'), 10.36 ± 0.94 M ('no absorption'), with a mean of 10.31 ± 0.72 M for the full sample.

Observed excess of blueshifted Na D features
As previously described, for Na D features produced by absorbing material within the host galaxy of the SN but not related to the SN itself, it would be expected that an equal number of SNe would display unrelated blueshifted and redshifted Na D absorption features, as there is no physical reason why material along the same line of sight would have a location bias in front (producing blueshifted features) or behind (producing redshifted features) of a particular SN (but see Hoang 2017). However, previous studies (Sternberg et al. 2011;Maguire et al. 2013) have identified an excess of blueshifted compared to redshifted Na D components in SN Ia samples. Using The calculation used to determine the percentage excess of blueshifted Na i D features observed is detailed in Section 4.4.
our new sample of 15 SNe Ia, we have calculated the percentage excess of blueshifted Na D features (%EB) using, where B and R are the numbers of events showing exclusively blue and redshifted features respectively, N is the number of events with no observed Na D absorption features, and S represents the objects displaying a symmetric absorption profile around 0 km s −1 . This is similar to the method of Maguire et al. (2013) but for this calculation, any object that displays a combination of red and blueshifted features was excluded given that identifying the source of these features is problematic. Maguire et al. (2013) excluded such events from the counts of events showing red or blueshifted features but not from the total number of objects. This exclusion fully removes objects that may have narrow Na D absorption features from different sources (e.g., CSM and host galaxy ISM contamination), which could affect the statistical calculations. For the new sample of SNe Ia in this work, %EB is found to be 23 %, while the full sample has an %EB value of 24 %. A summary of the classifications of the objects included in the new and full samples is given Table 6. If the probability of a given SN displaying blue-or red-shifted Na i D features is taken to be equal, as would be expected through purely line-of-sight/host-galaxy contamination, the cumulative binomial probability of observing at least 14 events with blueshifted only Na i D features compared to five showing purely redshifted Na i D features is approximately 3.1 per cent.
It is expected that SNe in late-type galaxies show Na i D absorption features more commonly as a consequence of their more gas rich host environments relative to early-type galaxies (with the specifics dependent on the specific local environment of each SN) resulting in additional contamination from unrelated material on coincidental lines of sight to the SN. There is, however, no environmental process that should favour the generation of blue or redshifted features relative to the SN rest frame. To explore this further, %EB was also calculated separately for those objects with hosts classified as early-type (E and S0) and those with late-type (S), with those hosted by dwarf or irregular galaxies being excluded. Like the previous calculations, objects showing both red and blueshifted features are also excluded. A clear difference between the observed %EB in early-and late-type galaxies is seen (Table 7). For earlytype galaxies, the negative %EB (−8 per cent) indicates that fewer objects with blueshifted Na i D features have been observed than would be expected from an even split in distribution. This negative value is likely a result of the small sample size since there are no known processes that favour an overproduction of redshifted narrow Na i D absorption. Indeed this value is produced by the one detection of a purely redshifted feature in SN 2008fp, further highlighting that early-type galaxies with no absorption features (12 of 13 in this sample) appear to be the norm. In contrast, the %EB calculated for the SNe Ia in late-type galaxies shows a strong excess of observed blueshifted features (48 per cent). A corresponding increase in SNe Ia with redshifted Na D features is not seen in the late-type host galaxies, as would be expected from isolated ISM contribu- tions alone. The cumulative binomial probability of obtaining this distribution through random chance is approximately 1.5 per cent.

Light curve properties and Si velocities compared to Na D absorption shifts
In this section, we investigate the connection between light curve width and B-V colour at maximum and the Na D absorption classifications. The stretch distribution of this new sample alone and the full sample are shown in the left and right panels of Figure 13 weighted mean stretch values are measured to be 1.02 ± 0.11, 1.03 ± 0.08, 1.06 ± 0.05 and 0.86 ± 0.20 for blueshifted only, blue and redshifted, non-blueshifted, and no Na i D absorption, respectively. The average light curve stretch value for those SNe Ia showing no Na i D absorption features is smaller than the other classifications but they are all marginally consistent within the uncertainties. In agreement with Maguire et al. (2013), we do not identify any relation between the strength (as described by the pEQW) of the blueshifted Na i D absorption feature of a given SN and its light curve stretch.
The relationship between light curve width and Na i D absorption classification is associated with the host distribution of the SNe Ia within these samples, as those SNe with no Na i D absorption features largely occur in early-type galaxies, which as discussed in Section 4.3 are known to host less luminous and faster evolving SNe Ia. Two of the four SNe Ia with broad light curves (stretch values exceeding one) and showing no Na i D absorption (LSQ12dbr and PTF12jgb) occurred within irregular galaxies (that can have histories of recent star formation), with a third (SN 2017hn) hosted by an S0 galaxy, highlighting that the true relationship is more than a simple early/late-type host galaxy split.
Previous studies have identified a relation between B-V colour at maximum light and the strength of any blueshifted Na i D absorption features (Foley et al. 2012;Maguire et al. 2013). A similar investigation of the full sample has been conducted here (Figure 14) though we fit the those objects showing purely blueshifted Na D absorption separately from those showing both blue and redshifted Na D absorption. This analysis reveals the previously found result that SNe Ia with stronger blueshifted Na i D absorption features (higher pEQWs) display larger B-V values (i.e., redder colours) is present only for objects showing purely blushifted Na D absorption with no such trend observed for objects displaying blue and redshifted absorption. The fitted relation between B-V colour at maximum and the pEQW of the blueshifted Na i D 2 Na D for those objects classified as showing 'blueshifted' Na D absorption retrieves the correlation at 5.2-significance (provided by the ODR fitting routine). We discuss the possible origin of this correlation further in Section 5.3.
We also explored if there was a correlation between the measured velocity of the Si features and Na D pEQW. The full sample confirms the result of Maguire et al. (2013) that no such correla-tion is observed for either the photospheric or high velocity (where present) Si features.

Ca absorption properties compared to Na D absorption, light curve 'stretch' and SN colour
Previous studies (e.g., Mazzali et al. 2005;Childress et al. 2013a;Maguire et al. 2014), have shown that high-velocity components of the Ca NIR absorption features of SNe Ia obtained near the time of maximum light are very common. There are previous suggestions that they could be due to CSM (Gerardy et al. 2004;Mazzali et al. 2005;Quimby et al. 2006;Tanaka et al. 2008) or may be intrinsic to the SN itself (Branch et al. 2006;Blondin et al. 2012). Here we have found that 14 of the 15 objects for which we could obtain measurements display such features (93 per cent). The non-detection in one event (SN 2016hvl) is likely due to a combination of its early phase (∼ 8 d prior to maximum light) along with its classification as a 91T-like event, which typically show weaker absorption features from intermediate-mass elements. As the high-velocity Ca absorption features are so common, it is impossible to link the simple occurrence of such features to the properties of any Na i D absorption features, blueshifted or otherwise. Therefore, to examine any potential link between the strength of blueshifted Na i D absorption and high velocity Ca features, the detailed properties of the Ca features must be examined. Following the method of Childress et al. (2013a), the strength of the Ca high velocity features is characterised though the ratio of the pEQWs of the high-to photospheric-velocity components (Ca ii HVF ), with objects having a ratio exceeding one defined as those showing 'strong' high velocity Ca . In the interpretation that high-velocity Ca features come from the same CSM that may be responsible for the blueshifted narrow Na D features, then a connection between the Ca ii HVF around peak and the pEQW of the blueshifted narrow Na D features should be seen, with SNe Ia with the strongest high velocity Ca features also showing the highest blueshifted Na i D 2 pEQWs. We show this comparison for our full sample in the upper left panel of Figure 15  sample with no Na D absorption that display Ca ii HVF values about one, which is difficult to explain in scenario where both of these probes suggest interaction with CSM.
We have also investigated any potential link between the Ca ii HVF and the light curve stretch, B-V colour at maximum, and the stellar mass of the host galaxy for the sample. Previous studies (Childress et al. 2013a;Maguire et al. 2013) identified a relation between light curve width and Ca ii HVF ratio, with SNe Ia with high Ca ii HVF ratios also having broader light curves. We do not see this trend for our sample, with a significant fraction of SNe Ia with high Ca ii HVF ratios having narrow light curves (upper right panel of Figure 15). There is a preference for SNe Ia with blueshifted Na D to have broad light curves than the rest of the sample but as discussed in Section 4.3, this may be driven by the preference for SNe Ia with blueshifted Na D features for late-type host galaxies, which produce more SNe Ia with broad light curves. Similarly to Childress et al. (2013a), we do not observe a correlation between Ca ii HVF ratio and B-V colour at maximum (bottom left panel of Figure 15). Pan et al. (2014) identified that SNe Ia with high Ca ii HVF ratios are preferentially found in low-mass galaxies. Our sample size is small so we can not draw any strong conclusion from our data on this connection (bottom right panel of Figure 15).

DISCUSSION
In this paper, we have presented a new sample of 15 SNe Ia with medium-resolution X-shooter spectra, where the properties of narrow Na D features were studied. We have investigated the connection between the presence of narrow Na D features, broad SN spectral features (in particular high-velocity features), light curve properties, and the host galaxy stellar mass in a full sample of 47 SNe Ia (including literature events). Now, we further discuss the implications of these results for our understanding of SNe Ia and their progenitors.

Confirmation of an excess of blueshifted Na D in late-type hosts
The statistical analysis of Na D absorption shows a strong preference for displaying the feature blueshifted, with an excess of 23 per cent in the new sample and an overall 24 per cent excess in the full sample. These values are consistent with previous results of a lower limit of ∼20 per cent in spiral hosts showing an excess (Sternberg et al. 2011;Maguire et al. 2013). As was also seen in these previous studies, the SNe Ia displaying these blueshifted features are strongly concentrated in late-type host galaxies (Figs. 11 and 12). Late-type galaxies in general contain more gas and dust than earlytype galaxies (e.g., Galliano et al. 2018), thus it is expected that more SNe occurring within late-type galaxies would have spectra showing Na i D absorption features (both blue and redshifted relative to the zero velocity position of the SN). As such, a population of SNe Ia occurring within late-type galaxies would be expected to show more narrow absorption features than a population of SNe located in early-type galaxies, even if the properties of the underlying SNe were obtained from the same distribution. However, there is no reason for such a population to have a preference for blueshifted over redshifted Na i D features. We have found that the small sample of SNe Ia hosted by irregular or dwarf galaxies (four events) showed no Na D absorption features. Irregular galaxies tend to have young stellar populations and display star formation rates comparable to, though with a greater variance, than late-type galaxies, as well as having higher dust contents (Hunter 1997;Hunter et al. 2006). While the sample is small, the apparent difference in the Na D absorption behaviour of these objects warrants additional study to determine if it is significant and its origin. One difference between these populations would be the lower metallicities of irregular/dwarf galaxies compared to spiral galaxies (Grebel 2004), although why this would result in different Na D strengths is unclear.
We and Maguire et al. (2012) that SNe Ia with broader light curves are also more likely to display Na D than those with narrower light curves ( Figure 13). There is a well-known trend that late-type galaxies preferentially host brighter (broader light curve) SNe Ia (e.g., Sullivan et al. 2010) so the result that both the galaxy type and the SN light-curve width correlate with the presence of blueshifted Na D is unsurprising. However, it is unclear what is the driving force behind this correlation. Is it that brighter SNe Ia with broader light curves that occur in late-time hosts intrinsically show preferentially blueshifted Na D absorption or is this an age or metallicity effect? Do SNe Ia in spiral galaxies have a different progenitor population to those in early-type host galaxies?

Lack of a link between blueshifted Na D and high-velocity Ca features
In the previous section, we discussed how SNe Ia with blueshifted Na D have broader light curves but this may be driven by their preference for late-type host galaxies and their differing environments. One of the main aims of this study was to investigate, for the first time, if there is a connection between the presence of broad high-velocity Ca features in SNe Ia and the presence of narrow blueshifted Na D features. As previously discussed, high-velocity Ca features are ubiquitous in SNe Ia and we confirm this here finding that the vast majority of the sample require a high-velocity Ca component. We also identify high-velocity Si features in 12 SNe Ia in our sample, although these features are not clearly 'detached' from the photospheric component. The reason that high-velocity features are interesting to investigate in connection with blueshifted Na D features is that there are suggestions that the high-velocity Ca features may be at least partially due to ejecta-CSM interaction (e.g., Mazzali et al. 2005;Tanaka et al. 2006). Therefore, if a link between these quantities was identified, it would provide evidence that blueshifted Na D features are due to CSM also rather than contamination. We remind the reader that the two probes do explore different distances from the SN, with Ca originating at significantly shorter distances than Na D. However, when we examine the strength of the high-velocity Ca components (parameterised through the Ca ii HVF of Childress et al. 2013a) compared to the blueshifted Na i D 2 pEQW (15), we do not identify any clear trend between them. In particular, we also find that a number of SNe Ia with no Na D absorption at all have very high Ca ii HVF values, which is difficult to interpret in the context of a common CSM origin for both strong blueshifted Na D features and strong high-velocity Ca components, as with this interpretation of their origin a strong signature would be expected in both probes rather than singularly. We also find no correlation between the pEQW of the high-velocity Ca features and the pEQW of the blue-shifted Na D features. Since we do not identify a link between the Na D features and the strength of high-velocity Ca features, this suggests that either one (or both) of these measurements are unrelated to CSM. Alternatively, these features may explore different regions of CSM, with Na D absorption produced at much larger distances (10 16 -10 17 cm) from the SN (e.g., Patat et al. 2007) than the CSM potentially probed through strong high-velocity Ca absorption at < 10 15 cm (Gerardy et al. 2004;Mulligan & Wheeler 2018). An alternative scenario that has been suggested for the high-velocity Ca components is that they are due to a density or abundance enhancement in the SN ejecta (Branch et al. 2006;Blondin et al. 2012) and thus have no connection to the presence of CSM.

Can the excess of blueshifted Na D be explained by CSM?
We have identified a clear excess of blueshifted over redshifted Na D absorption features in both our new sample (33 per cent) and our full sample (24 per cent). Previous studies that have investigated these features have suggested that the most probable scenario to produce them is from CSM in the immediate vicinity of the SN (e.g., Sternberg et al. 2011;Maguire et al. 2012;Phillips et al. 2013). This conclusion was based on the original result of Patat et al. (2007), where time-varying Na D features were observed in SN 2006X and were suggested to be due to changes in the ionisation field of the CSM caused by the SN radiation. This material was estimated to be located at a radius of ∼ 10 16 cm from the SN. Other SNe Ia were subsequently identified with these time-varying Na D features and were also suggested to be due to CSM, most likely related to explosion scenarios involving non-degenerate companion stars (Simon et al. 2009;Blondin et al. 2009;Stritzinger et al. 2010;Sternberg et al. 2014). There are also models of doubledegenerate systems that could eject material several thousand years prior to the explosion that interacts with the ISM, and also could potentially produce similar Na D profiles (e.g., Raskin & Kasen 2013;Shen et al. 2013). However, it is likely that significant finetuning of double-degenerate models would be required to produce such material ejections in a large percentage of SNe Ia, as well as the host galaxy dependence of SNe Ia showing blueshifted Na D absorption features towards late-time host galaxies being less easy to explain with this scenario. Double degenerate scenarios not having a strong preference for a particular stellar population compared to single degenerate models that require the younger stellar populations observed in late type galaxies.
In this study, we have investigated a number of SN Ia properties directly to see if the presence of blueshifted Na D features are intrinsically linked to SNe Ia explosions. As discussed in Sections 5.1 and 5.2, there is a preference for the presence of Na D in late-type host galaxies, which also results in a correlation with light-curve stretch, although the driving force behind these correlations is not well understood. We also observe the presence of the previously identified relation that SNe Ia with redder B-V colours at maximum have have stronger Na D features, though we observe this trend only in objects showing purely blueshifted absorption, with objects showing both blue and redshifted absorption found to have consistent B-V colours regardless of the strength of their blueshifted Na D component. However, Phillips et al. (2013) showed that this is most likely driven by gas in the ISM and not due to CSM.
Related to the above, Phillips et al. (2013) investigated the connection between the Na D column densities seen in SNe Ia and extinction measured from the SN colours, finding that those with stronger Na D column densities, compared to the standard Milky Way-derived extinction relation, have preferentially blueshifted Na D features and hence may have an origin in the CSM. However, there was not a clear correlation with some SNe Ia with identified time-varying Na D not following the expected relation, suggesting that the situation is more complicated than previously thought and that strong blueshifted Na D features may be produced in some way other than through CSM (under the assumption that time-varying Na D are produced by CSM interaction).
The low R values seen in some SNe Ia have been attributed to multiple scatterings in CSM at typical radii (10 16 -10 19 cm) (Wang 2005;Patat et al. 2006;Goobar 2008;Amanullah & Goobar 2011), while some more recent studies have investigated the presence of blueshifted Na D features in the context of an origin in the ISM at 10 16 -10 20 cm (e.g Bulla et al. 2018a,b).
These ISM-origin models are based on cloud-cloud collisions produced by SN radiation interacting with clouds in the ISM in the vicinity of the SN at ∼ 10 18 cm (Hoang et al. 2019). This mechanism would reduce the dust grain sizes (that would explain the lower R ) and could accelerate the clouds towards the observer (that would explain the preference for blueshifted Na D features). This scenario, however, requires the IS cloud to be relatively close to the SN ( 1 pc, see Figure 6 of Giang et al. 2020) in order to explain the low Rv values and blueshifted Na D features observed already at early phases (pre-maximum). An acceleration of the blueshifted material with time would also be expected though has not been observed in SNe Ia, with blueshifted Na D features having been observed to evolve in intensity but not in velocity. Newer models (Hoang 2021) suggest that interaction with SN radiation is not needed at all and these small dust grains could be produced by the ISM radiation alone before the SN explosion, which would resolve the timescale issue above. There are, however, some remaining issues with this model, including the lack of observations for similar polarisation laws in the Milky Way as observed for SNe Ia (Cikota et al. 2018) and a lack of quantitative estimates of the velocities expected for the gas (and not the dust grains) within the ISM that would give rise to the blueshifted Na D features. These newer ISM only models also fail to explain why we see the clear excess in blueshifted over redshifted Na D features that is once again seen to be statistically significant in both our new and full samples of now 47 SNe Ia with the requisite intermediate-resolution spectroscopy around maximum light.
Additionally, there appears to be tentative evidence of a possible distinction in the absorption minimum velocities of the Na D features between the sub-classifications of absorption feature. With the minima of purely blueshifted absorption features, and the blueshifted components of those objects showing both blue and redshifted absorption, measured to be at higher relative velocities (Mean value: -51±62 km s −1 ) compared to the absorption minima of redshifted absorption features as shown in Figure 16 (Mean value: 22±12 km s −1 ). The small number of objects makes drawing firm conclusions based on this difficult but could potentially be explained by a combination of a symmetric population around 0 km s −1 produced by distant ISM with an additional population at higher, bluer relative velocities produced by CSM or close-by ISM.

CONCLUSIONS
This work has explored a new sample of moderate-resolution spectra of 15 SNe Ia obtained near maximum light to investigate the potential link between the properties of narrow Na D absorption features and high-velocity Ca NIR triplet absorption features, with the goal of constraining the progenitor scenarios of SNe Ia more firmly. Our main conclusions are: (i) we confirmed a %EB of 23 per cent blueshifted over redshifted Na D features in a new sample of maximum-light SN Ia intermediate-resolution spectra, in agreement with previous studies, (ii) we found that this excess was even stronger (48 per cent) in a full sample of 47 SNe Ia (including literature objects) when only spiral/late-type host galaxies were considered, (iii) we did not identify a trend between the presence or strength of high-velocity Ca features seen in the majority of SNe Ia around maximum light and the presence of narrow blueshifted Na D features, (iv) we observed that the previous identified correlation that stronger blueshifted Na D features are seen in SNe Ia displaying redder B-V colours at maximum light is present only for those objects displaying purely blueshifted absorption. No trend is observed for objects with blue and redshifted Na D absorption, (v) we have identified a tentative relation that the velocity distribution of the narrow blueshifted Na D features may extend to higher velocities than that of the non-blueshifted Na D sample (vi) we found that SNe Ia in irregular galaxies do not show Na D absorption at odds with their likely underlying stellar populations (although the sample size is very small) The lack of a link between the properties of the narrow Na D features and the high-velocity Ca components around maximum light in SNe Ia suggests either that one or both of these probes is not due to CSM as has previously been suggested or the presence of CSM around Type Ia SNe cannot be linked across the range of distances explored by the two probes. However, the confirmation of a clear excess of blueshifted over redshifted Na D absorption features in our new sample, and the full combined literature sample, cannot easily be explained through ISM interpretations of the Na D feature that do not involve an impact of the SN progenitor system or ejecta on these features. Further theoretical modelling is required to provide new scenarios to explain these results, most likely either in the context of close-in ISM or CSM. the contributions of the University of Hawaii Institute for Astronomy, the Queen's University Belfast, the Space Telescope Science Institute, the South African Astronomical Observatory (SAAO), and the Millennium Institute of Astrophysics (MAS), Chile.
Funding for the Sloan Digital Sky Survey IV has been provided by the Alfred P. Sloan Foundation, the U.S. Department of Energy Office of Science, and the Participating Institutions. SDSS acknowledges support and resources from the Center for High-Performance Computing at the University of Utah. The SDSS web site is www.sdss.org.
We thank the anonymous reviewer for their helpful comments and discussion.

DATA AVAILABILITY
The data underlying this article will be shared on reasonable request to the corresponding author.

APPENDIX A: LITERATURE OBJECTS: BROAD FEATURE FITTING RESULTS
For objects identified as not displaying a high velocity feature in their spectrum the 'Fitting Notes' column of the following tables details the first criterion each spectrum failed to meet. These criteria are described in more detail in Section 3.3.2 but are outlined here for clarity: (i) The fitting of a high velocity feature must be successful in at least 10% of iterations. (ii) The inclusion of a high velocity feature must be strongly favoured by the BIC indicated by a ΔB IC value >10. This paper has been typeset from a T E X/L A T E X file prepared by the author. Table A1. Summarised fitting results for the Si 6355 Å feature for the objects first described in Maguire et al. (2013). Means have been weighted using the relevant fitting uncertainties with the uncertainties of these means obtained as the standard deviation of the results. A positive value of ΔB IC indicates a preference for the inclusion of a high velocity feature based on statistical considerations alone. Mean -10250 ± 1120 6570 ± 840 69 ± 25.1 -12650 ± 1460 3810 ± 810 9.7 ± 10.5 0.034 ± 0.047 220 ± 160 * No ΔB IC is available for these objects as all fitting iterations when including a high velocity feature were considered to be unsuccessful (1) Values measured from a lower resolution, non X-shooter spectrum Roman numerals denote the first fitting criterion failed by each object not identified as displaying a high velocity feature, as described in Section 3.3.2 Table A4. Summarised fitting results for the Ca NIR triplet feature for the objects first described in Sternberg et al. (2011). Means have been weighted using the relevant fitting uncertainties with the uncertainties of these means obtained as the standard deviation of the results. A positive value of ΔB IC indicates a preference for the inclusion of a high velocity feature based on statistical considerations alone. -13500 ± 160 6060 ± 160 21.5 ± 0 -21650 ± 160 4890 ± 160 4.5 ± 0.0 0.045 ± 0.001 240 ± 10 (1) Mean -11330 ± 1820 6040 ± 450 28 ± 37.9 -19070 ± 2310 6290 ± 1560 11.5 ± 60.2 0.233 ± 0.322 290 ± 310 * No ΔB IC is available as all fitting iterations when including a high velocity feature were considered to be unsuccessful ** Residual pattern indicative of the presence of an additional feature (1) Values measured from a lower resolution, X-shooter spectrum Roman numerals denote the first fitting criterion failed by each object not identified as displaying a high velocity feature, as described in Section 3.3.2