Optical Spectroscopy of Type Ia Supernovae by the Carnegie Supernova Projects I and II

We measured expansion velocities and pWs of selected features as described in Folatelli et al. (2013). Following the definitions in Garavini et al. (2007; see Figure 4 and Table 5 of Folatelli et al. 2013) we performed measurements of the following features: Ca ii H&K (pW1), Si ii λ4130 (pW2), Mg ii λ4481 (pW3; blended with Fe ii and dominated by Ti ii in the Branch CL class), Fe ii at ∼4600 Å (pW4, blended with S ii), the S ii "W" feature at ∼5400 Å (pW5), Si ii λ5972 (pW6), Si ii λ6355 (pW7), and the Ca ii "IR triplet" (pW8).

All measurements were made by means of the IRAF splot task from the onedspec package. Line expansion velocities were derived from Gaussian fitting of the minimum of each absorption trough. The pW measurements were obtained by direct integration between two defined pseudo-continuum positions. Error estimates were computed with splot setting the number of samples for error computation (the nerrsample parameter) to 100, the sigma0 parameter (uniform component of the uncertainty) to the rms flux measured in the nearby pseudo-continuum, and disregarding any Poissonian component of the uncertainty (i.e., setting the invgain parameter to zero). We adopted 1 Å as the minimum realistic uncertainty for pW measurements. For some critical features, such as Si ii λ5972, when unable to obtain a reliable measurement because the feature was very weak or not detected, we estimated upper limits of pW considering the signal-to-noise ratio in the spectral region where such a feature would be expected.

The measurements described above for our own spectra were analyzed in combination with similar measurements of spectra of the targets in Table 1, which are publicly available through WISeREP, specifically when our observations were not sufficiently close to maximum light to compute the desired spectroscopic parameters. The procedure followed is the same as that described in Folatelli et al. (2013) and is summarized as follows:

• 1.  
  In cases where several spectra encompassing maximum light were available to us, we fit low-order polynomials to our pW and expansion velocity measurements and used those functions to interpolate the values at maximum.
• 2.  
  When two spectra were available in the interval −4 to +4 days from maximum, we interpolated values at maximum from them.
• 3.  
  If only data before or after maximum were available, but one spectrum was obtained within 1 day of maximum light, an extrapolation was allowed.
• 4.  
  In the most frequent cases where only one spectrum was available within the range −4 to +4 days from maximum, we applied the slopes given in Tables 4 and 7 of Folatelli et al. (2013) to estimate values at maximum light, combining the errors estimated for our measurements with those coming from the assumed slopes.

This procedure allowed the determination of pWs and expansion velocities at maximum light for the selected features in 15 CSP I SNe Ia not included in our previous spectroscopic release (Folatelli et al. 2013), 113 SNe Ia observed by CSP II, and 27 historical SNe Ia. All targets for which we were able to derive spectroscopic parameters at maximum light are included, regardless of their being considered as normal or peculiar SNe Ia. The values of the pWs at maximum light are presented in Table 5. Measurements of expansion velocities at maximum light can be found in Table 6. Note that the SNe are sorted by Branch type (Branch et al. 2006) in both tables.

LSQ12ca. This SN Ia has the lowest value of Δm₁₅ in the CSP-II sample (0.618 ± 0.081 mag), although its s_BV (1.195 ± 0.097), while high, is comparable to that of other normal SNe Ia. From the spectrum available to us, obtained at phase 2.8 days past maximum, we derive normal values for its expansion velocities and pWs at maximum light. C ii absorption is probably present on the red side of the Si ii λ6355 line.

LSQ12gpw. There are three public spectra of this SN Ia from PESSTO, ⁴⁷ available in WISeREP: the one obtained at phase −1.5 days and the other two at phase −0.5 day. There is also one CSP spectrum at phase +5.8 days. All these spectra show an absorption line, redward of Si ii λ6355, well separated from the Si ii feature that, if identified as C ii λ6580, yields expansion velocities of approximately 7500, 7400, and 6200 km s⁻¹, for phases −1.5, −0.5, and +5.8 days, respectively.

LSQ12hno shows expansion velocities somewhat lower than usual (∼8000–9000 km s⁻¹) in our two spectra obtained at 2.4 and 1.5 days before maximum light, respectively.

LSQ12hnr. Two spectra of LSQ12hnr are available in WISeREP, both obtained by PESSTO, at phases +0.8 and +10.8 days, respectively. The classification report by Le Guillou et al. (2012) gave a possible redshift of z = 0.135 inferred from the SN Ia spectrum. No obvious host is detected in our follow-up images or in a deep, Very Large Telescope (VLT)-MUSE observation of the SN Ia site. However, an apparent cluster of galaxies is observed whose brightest member lies about 40'' west of the SN Ia location. We determined redshifts for the three brightest galaxies in that cluster (from a total of at least four), obtaining a weighted average of z = 0.1243 ± 0.0002, the value we therefore adopted for LSQ12hnr assuming this SN Ia occurred in a fainter member of the same cluster of galaxies.

OGLE-2013-SN-015. There is only one CSP spectrum of this SN Ia, which, albeit noisy, seems normal, and was obtained very close to maximum light (phase = +0.7 days). However, our photometric follow-up was poor for this target, and consequently, we decided not to consider this SN Ia in our analysis of SN properties at maximum light.

OGLE-2013-SN-123. There is only one spectrum available from WISeREP, obtained by the PESSTO Collaboration at maximum light, which shows clear evidence of host galaxy contamination, and therefore, the pWs measured from it are unreliable, as well as the spectral type determined via SNID. However, we obtained a spectrum of the host galaxy with the WFCCD instrument on the Las Campanas 2.5 m du Pont telescope in 2019 to determine its redshift. By scaling the host spectrum and subtracting it from the SN Ia observation so as to make the obvious stellar Na i D blend at ∼5892 Å absorption and the TiO feature at ∼7150 Å disappear, the other stellar features such as Ca ii H&K, the G band at ∼4300 Å, Hβ, and Mg ib λ5175 also mostly disappeared (see Figure 1). We, therefore, have used this host galaxy subtracted spectrum to derive the spectroscopic properties at maximum light for this SN Ia.

Zoom In Zoom Out Reset image size

ASASSN-15eb. The classification report by Childress et al. (2015) does not refer to any peculiarities; however, SNID yields some matches with 91T-like SNe at maximum light. According to our light curve, this spectrum corresponds to phase +4.5 days. The pWs are indeed small, but this is clearly caused by strong host contamination. Also, strong Galactic Na i D absorption is evident in that spectrum. The CSP spectrum published here, obtained at phase +11.0 days, also exhibits significant host galaxy contamination as well as strong Na i D absorption from the Milky Way. Both spectra show absorption minima on the red side of Si ii λ6355, which could be attributed to C ii.

For a limited number of CSP-II targets and historical SNe Ia, the available data span enough time to follow the evolution of the Si ii λ6355 expansion velocity to at least 20 days past maximum light. These observations are presented in the left and right panels of Figure 2, respectively, with the different Branch types indicated by the colors and shapes of the symbols. In general, albeit with less data here, the behavior observed in this figure is very similar to that of Figure 9 of Folatelli et al. (2013). In both panels, the shaded region represents the upper and lower 1σ dispersion of the Si ii λ6355 expansion velocity evolution for the whole CSP-I and II sample of Branch CN SNe with normal Wang classification subtypes (see Table 7 for details).

Zoom In Zoom Out Reset image size

From Figure 2 (left panel), we can see that SN 2012bl shows high Si ii λ6355 velocity, which persists 29 days past maximum light, although measurements of the minimum of Si ii λ6355 after 20 days post maximum are questionable due to blending with other features. SN 2012fr (Childress et al. 2013; Cain et al. 2018; Contreras et al. 2018) is an example of flat velocity evolution, in which the expansion velocity of Si ii λ6355 is almost constant over the period covered by our data (−5.0 to +17.8 days).

For all the SNe Ia presented in this paper with sufficient time coverage, the difference between the Si ii λ6355 velocity at maximum light and at 20 days past maximum, Δv₂₀(Si ii), was calculated using the same methodology described in Section 3.1.1 of Folatelli et al. (2013). These values are given in the last column of Table 6 and plotted in Figure 3 as a function of light-curve decline rate Δm₁₅ and color stretch s_BV, respectively, along with the CSP-I objects already presented in Folatelli et al. (2013). Figure 3 confirms the strong correlation ⁴⁸ observed by Folatelli et al. (2013) between Δv₂₀(Si ii) and Δm₁₅ for Branch SS, CN, and CL SNe, suggesting that these events form a single sequence. On the other hand, the lack of a correlation for the Branch BL events is consistent with previous hints that these objects may represent a distinct group of SNe Ia (e.g., see Wang et al. 2013; Burrow et al. 2020).

Zoom In Zoom Out Reset image size

In the analysis of spectroscopic parameters at maximum light, we consider separately the objects discovered by targeted and untargeted surveys. That is, SNe Ia in the CSP-I, CSP-II, and historical samples discovered by amateur astronomers or other targeted surveys (e.g., the Lick Observatory Supernova Survey or the Chilean Automatic Supernova Search) will be considered as one sample, while all the targets drawn from untargeted surveys such as the La Silla-QUEST (LSQ), the Palomar Transient Factory (PTF, iPTF), the Sloan Digital Sky Survey or the Catalina Real Time Transient Survey (CRTS), ASAS-SN, among others, will be considered as a second sample. The Calán/Tololo SNe are a special case since the photographic plates were taken in an untargeted fashion, but were searched for stellar objects that appeared near galaxies. We, therefore, have grouped these SNe with the targeted events. Most of the CSP-I SNe Ia belong to the targeted group and most of the CSP-II SNe Ia belong to the untargeted one. However, there are a few exceptions. CSP-II SNe Ia for which we derived spectroscopic parameters at maximum light, and that are included in the first group of targeted survey discoveries are PSN J13471211-2422171 and the SNe 2011iv, 2011jh, 2012E, 2012ah, 2012fr, 2012hd, 2012 hr, 2012ht, 2013aa, 2013fz, 2013gy, 2013hn, 2014I, 2014Z, 2014ao, 2014at, 2014dn, 2014eg, and 2015F. On the other hand, CSP-I targets for which we present spectroscopic parameters at maximum light that were discovered by untargeted searches are SNe 2007if, 2007ol, 2008bz, and 2008fr.

Figure 4 displays the Branch et al. (2006) diagram for the targeted (left) and untargeted (right) samples of CSP and historical SNe Ia. In defining the boundaries between the four classes (CN = "core normal," SS = "shallow silicon," BL = "broad lined," CL = "cool"), we follow the definitions adopted by Folatelli et al. (2013). The largest difference between these diagrams is for the BL SNe Ia, whose relative numbers are clearly different. To be precise, the BL SNe Ia represent 31% ± 5% of the targeted sample, but only 13% ± 4% of the untargeted events. This difference is explained by the fact that targeted searches are biased toward luminous galaxies, and high-velocity SNe Ia (approximately two-thirds of which belong to the BL class) are known to occur preferentially in luminous galaxies (Wang et al. 2013). Similarly to the BL events, CL SNe—those with relatively large values of pW6 (Si ii 5972)—represent 21% ± 4% of the targeted and 15% ± 4% of the untargeted samples, respectively. As for the CN SNe, they are more frequent in the untargeted sample, amounting to 50% ± 9% of the total, compared to 33% ± 6% of the targeted sample. The SS SNe, which correspond to the classical 1991T-like and 1999aa-like events, show a less significant difference, amounting to 22% ± 5% of the untargeted sample, while in the targeted sample, they represent 15% ± 4%. These percentages are illustrated for the targeted and untargeted samples, respectively, in Figure 5. For further comparison, we show in Figure 6 histograms of our pW6 and pW7 measurements at maximum light for the untargeted and targeted samples, separately for each of the Branch types.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Another way to look at these numbers is to consider ratios. In particular, how do the ratios of the numbers of SS, CL, and BL SNe Ia compare to the number of CN SNe Ia for the targeted and untargeted samples? In the case of the targeted SNe, N(SS)/N(CN) = 0.44, N(CL)/N(CN) = 0.64, and N(BL)/N(CN) = 0.93, while for the untargeted events we find N(SS)/N(CN) = 0.44, N(CL)/N(CN) = 0.31, and N(BL)/N(CN) = 0.27. These numbers imply that SS SNe Ia are equally common with respect to CN SNe in targeted versus untargeted surveys. On the other hand, CL SNe Ia are twice as common and BL SNe are more than three times more common with respect to CN SNe in targeted surveys compared to untargeted surveys. This likely reflects the fact that SS and CN SNe Ia occur in hosts over a large range of luminosity, whereas CL and BL SNe Ia prefer luminous hosts (e.g., see Neill et al. 2009; Wang et al. 2013; Pan 2020).

Note that Burrow et al. (2020) explored the Branch diagram through a cluster analysis instead of using predefined group boundaries as we have done in Figure 4. Comparing the Branch type classifications in Tables 1 and 2 for the 43 SNe Ia in common with the sample that Burrow et al. analyzed using a two-dimensional Gaussian mixture model (2D GMM), 60% are in agreement. Not surprisingly, the objects for which the classifications differ lie at or near the borders of the Branch groups.

Some studies have found correlations between SNe Ia ejecta velocities and their host properties (e.g., Pan et al. 2015; Pan 2020; Dettman et al. 2021). Considering that their results imply that high-velocity SNe Ia tend to prefer high stellar mass hosts, more frequent in targeted searches, we compared the Si ii λ6355 expansion velocities at maximum light derived for our targeted and untargeted samples as a whole, as well as for the different Branch types within each sample. While the expansion velocities tend to be higher for BL SNe Ia, and somewhat lower for SS SNe Ia, the differences between the targeted and untargeted samples are insignificant considering the uncertainties. Figure 7 presents the average velocities at maximum light for Si ii λ6355 as a function of the different Branch types, and for the whole targeted and untargeted samples, respectively.

Zoom In Zoom Out Reset image size

In their study of SNe Ia from CSP-I, Folatelli et al. (2013) searched for correlations between the different pW measurements at maximum light. In Figures 8–11, we reproduce the four strongest correlations that they found, plotting each separately for the targeted and untargeted samples. Folatelli et al. found that the correlations were tighter (Pearson correlation coefficients ρ > ∣0.75∣) if the high-velocity and fast-declining (v(Si ii(6355) <12,000 km s⁻¹ and Δm₁₅(B) < 1.7 mag) events were excluded from the fits, but in this paper, we do not make this distinction. In each of Figures 8–11, coefficients of determination (R²) and Pearson correlation coefficients (r) are shown near the top-left or bottom-right corners of each panel.

Zoom In Zoom Out Reset image size

As shown in the right panel of Figure 8, the strongest correlation for the untargeted sample is found between the pW values for Si ii λ4130 versus Si ii λ5972. In this plot, the CL SN Ia LSQ14ip appears as an outlier in an otherwise strong positive correlation. However, the left panel of Figure 8 corresponding to the targeted sample shows a large dispersion in the measurements for the CL SNe Ia, with LSQ14ip lying within this dispersion. LSQ14ip is an extremely cool SN Ia, very similar to SN 1986G (Phillips et al. 1987), which showed strong Ti ii absorption at maximum light. The Si ii λ4130 line is blended with the Ti ii absorption (e.g., see Ashall et al. 2016), making it difficult to measure an accurate pW value. This blending undoubtedly accounts for the large dispersion in the CL events observed in the targeted sample.

In Figure 9, we plot the pW values for the Si ii λ4130 and Si ii λ6355 absorptions. In this case, the correlation is tighter for the targeted sample. As might be expected, however, correlations between all three of the Si ii lines are similarly strong for both the targeted and untargeted samples.

Zoom In Zoom Out Reset image size

Figure 10 displays the pW values for the Ca ii IR triplet plotted versus the pW4 (Fe ii) measurements. The correlation is tighter for the SNe Ia discovered in untargeted searches. In Figure 11, the correlation between the pW8 (Ca ii IR) and pW7 (Si ii 6355) parameters is displayed. We see that the source of much of the dispersion, notably in the plot corresponding to the targeted searches, comes from the CL SNe Ia.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In order to provide a broader view of the possible correlations between the spectroscopic parameters under consideration, in Figure 12, we present correlation matrices for pairs of pW values and expansion velocities at maximum light for the objects in our targeted and untargeted samples, respectively.

Zoom In Zoom Out Reset image size

Here, we briefly review the strongest correlations between pW values at maximum light and the light-curve decline rate expressed by the typical Δm₁₅ parameter, or the more recently defined color stretch parameter, s_BV. Figure 13 shows the correlations between pW6 (Si ii 5972) and Δm₁₅. Hachinger et al. (2006) first pointed out that these two parameters correlate strongly. The relations are similarly tight for both the targeted and untargeted samples. The most discrepant measurements in the untargeted sample correspond to the CNs SN 2007ol and SN OGLE-2013-SN-123. As explained in Section 3.3, the PESSTO spectrum of the latter suffered from considerable host galaxy contamination. While our attempt to correct for this problem was largely successful, the error in the pW6 (Si ii 5972) measurement is likely underestimated due to uncertainties in the subtraction of the host galaxy spectrum. Accounting for this extra source of uncertainty would bring the error in pW6 up to ±7 Å. On the other hand, the spectrum of SN 2007ol looks good and does not show indications of host galaxy contamination.

Zoom In Zoom Out Reset image size

Figure 14 displays the relationship between pW6 (Si ii 5972) and s_BV. We might have expected an improvement in the correlation since the s_BV parameter does a better job of discriminating between light-curve shapes for fast-declining events. Nevertheless, the coefficients of determination and Pearson r coefficients are similar to those obtained using Δm₁₅.

Zoom In Zoom Out Reset image size

Folatelli et al. (2013) found that pW2 (Si ii 4130) also correlated strongly with the light-curve decline rate. In Figures 15 and 16, this parameter is plotted against Δm₁₅ and s_BV, respectively. Again, it is somewhat surprising to see that usage of s_BV does not significantly improve the tightness of the correlations. This could be related to the fact that s_BV seems not to work as well as Δm₁₅ for SS SNe Ia (C. Burns, private communication).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size