EVIDENCE FOR TYPE Ia SUPERNOVA DIVERSITY FROM ULTRAVIOLET OBSERVATIONS WITH THE HUBBLE SPACE TELESCOPE

Figure 2 shows the UV light curves of SNe 2004dt, 2004ef, and 2005M obtained with the HST ACS and the F220W, F250W, and F330W filters. Also plotted are the corresponding optical light curves (in BVI) from Ganeshalingam et al. (2010). As SN 2005cf was well observed and is judged to be a normal SN Ia, and its light curves have been published elsewhere (W09b), we use it as a template SN Ia for comparison study. To account for effects of the redshift on the observed SN flux, we applied K-corrections (Oke & Sandage 1968) to all of the light curves using the observed spectra (as shown in Figures 5 and 6) and the Hsiao et al. (2007) template.

Zoom In Zoom Out Reset image size

The light curves have also been corrected for reddening in the host galaxy, as well as in the Milky Way (MW). The total reddening toward an SN Ia can be derived from an empirical relation established between decline rate Δm₁₅(B) and color indexes for a low-reddening sample (e.g., W09b, and references therein). By comparing the observed color with that predicted by the Δm₁₅(B) versus (B_max − V_max) relation, we estimate the total reddening E(B − V)_total as 0.10 mag for SN 2004dt, 0.17 for SN 2004ef, 0.12 mag for SN 2005M, and 0.20 mag for SN 2005cf, with a typical error of ∼0.04 mag. The absorption in the MW was estimated by using the reddening maps of Schlegel et al. (1998) and the Cardelli et al. (1989) reddening law with R_V = 3.1, while extinction in the host galaxy was corrected applying an extinction law with R_V ≈ 2.3 (e.g., Riess et al. 1996; Reindl et al. 2005; Wang et al. 2006b; Kessler et al. 2009). Since all four SNe Ia have low host-galaxy reddening (e.g., E(B − V)_host ≲ 0.10 mag), a moderate change of the extinction-law coefficient will have limited effect on the UV light curves shown in Figure 2. According to the extinction-law coefficients recalculated by Brown et al. (2010), a change in R_V from 1.8 (LMC dust) to 3.1 (MW dust) will only result in a difference of ∼0.02 mag in F330W and ≲ 0.15 mag in F250W for an SN Ia with a host-galaxy reddening E(B − V)_host = 0.10 mag. Note, however, that the magnitude change in F220W could be larger due to a dramatic variation in the resultant extinction coefficient R_F220W. Table 3 lists the relevant parameters of these four SNe Ia.

One can see from Figure 2 that the UV light curves of our SNe Ia show a large dispersion, especially at shorter wavelengths, while the optical counterparts are generally similar except in the I band where there is more diversity. Of our sample, SN 2004ef appears relatively faint in all bands, consistent with the fact that it is a slightly faster decliner with Δm₁₅(B) ≈ 1.46 ± 0.06 mag. SN 2005M is a spectroscopically peculiar object like SN 1991T. This supernova has a slower decline rate (Δm₁₅(B) = 0.86 ± 0.05 mag), but it does not appear to be overluminous in either the UV or optical bands. However, our relative ignorance regarding the intrinsic color of these slow decliners might prevent us from deriving reliable reddenings for them. A notable feature in the plot is that the UV emission of SN 2004dt appears unusually strong around the B-band maximum and decreases rapidly after maximum light. Moreover, SN 2005M seems to peak a few days earlier than SN 2005cf in the F220W and F250W bands. Such a temporal shift seems to exist for SN 2004dt and SN 2004ef, although their light curves in UV are not well sampled around maximum brightness.

The color curves of our four SNe Ia, corrected for the MW and host-galaxy reddening, are presented in Figure 3. One can clearly see that their UV − V curves show remarkable differences despite having similar B − V curves. SN 2004dt is quite peculiar, being very blue compared with the other three objects. It is worthwhile to point out that the Swift UVOT has seen a number of events (7) that are also blue in the NUV–optical photometric colors (P. Milne 2012, private communication).

Zoom In Zoom Out Reset image size

Since SN 2004dt has Δm₁₅(B) ≈ 1.1 mag and a peak B − V color (∼0.0 mag) similar to that of SN 2005cf, it is of interest to perform a detailed comparison of their UV properties. Around maximum light, SN 2004dt is bluer than SN 2005cf by ∼0.8 mag in F250W − V and by ∼2.0 mag in F220W − V. About 2–3 weeks after maximum, the color difference becomes less significant as a result of the rapid decline of the post-maximum UV emission in SN 2004dt (see Figure 6 and the discussion in Section 3.2). In the first 15 days after the maximum, the decline of the UV light curve is measured to be ∼1.9 mag in F220W, ∼2.4 mag in F250W, and ∼2.5 mag in F330W for SN 2004dt. These are all larger than the corresponding values measured for SN 2005cf by ∼0.6–0.9 mag (see also Table 10 in W09b). A faster post-maximum decline usually corresponds to a shorter rise time for the light curves of SNe Ia in the optical bands, as evidenced by the fact that they can be better normalized through the stretch factor (Perlmutter et al. 1997; Goldhaber et al. 2001). For SN 2004dt, a shorter rise time in the UV is consistent with the high expansion velocity that could make the gamma-ray heating less effective. Moreover, a higher expansion velocity could extend the radius of the UV photosphere to higher velocities and hence result in a larger UV flux. The possible origin of the UV excess in SN 2004dt in the early phase is an interesting issue and is further discussed in Section 4.2.

As shown in Table 1, a total of 34 HST ACS PR200L prism spectra have been collected for SNe 2004dt, 2004ef, 2005M, and 2005cf. In Column 5 of Table 1, we list the total exposure time for each spectrum, resulting from a series of co-added CR-split exposures. Among the four SNe Ia observed with the HST ACS, SN 2005cf and SN 2004dt have better temporal sequences and higher S/N for their UV spectra (see Figures 5 and 6). Owing to very low spectral resolution at longer wavelengths, our analysis in this paper is restricted to the useful wavelength range 1800–3500 Å.

To evaluate the quality of the prism data, we compare the synthetic magnitudes derived from the UV spectra with those obtained by direct imaging observations for SN 2005cf and SN 2004dt (see Figure 4). In F330W, these two measurements match well with each other near maximum light, but the spectrophotometric magnitudes decline more slowly starting about one week after maximum. A similar trend in the decline rate exists in the F220W and F250W bands of SN 2004dt and in the F250W band of SN 2005cf. However, compared with the corresponding imaging observations, it is alarming that the flux of the prism spectra is too high in the F250W band (∼0.6 mag for SN 2004dt, ∼0.7 mag for SN 2005cf), and especially in the F220W band (∼1.0 mag for SN 2004dt, ∼2.0 mag for SN 2005cf). This suggests that the flux level of PR200L calibrated by the white dwarfs (which are very blue) may not be accurate for the flux calibration of the supernova.

Zoom In Zoom Out Reset image size

According to ACS/HRC Instrument science report ACS 2006-03 (Larsen et al. 2006), the flux calibration of the PR200L prism spectra based on white dwarfs is expected to be accurate to ∼5% over the wavelength range 1800–3000 Å. However, red objects usually suffer systematic errors in the UV flux calibration because the blue end of the spectrum can be highly contaminated by scattered light from the red end as a result of the "red pile-up" effect. Such an effect is probably a dominant factor accounting for the higher F250W and F220W fluxes, as indicated by the wavelength-dependent discrepancy between the spectrophotometry and observed photometry,⁶⁴ as well as for the slower post-maximum decline seen in all three UV bands for the SNe in our sample; SNe Ia are known to become progressively redder after maximum light, and the contamination of the red light becomes increasingly significant. However, the red pile-up effect is at present poorly quantified, and more test observations of objects covering a wide range of colors are clearly needed to better calibrate PR200L prism spectra.

Fortunately, we also have ACS HRC photometry taken nearly at the same time, making it possible to obtain accurate relative flux calibration of the PR200L prism spectra. Given the relatively smooth variations in the raw prism spectra, it is not difficult to model the continuum of the spectra with a simple polynomial function. The final calibrated spectra of SN 2005cf and SN 2004dt are shown in Figures 5 and 6; the spectrophotometry is consistent with the photometric results within ±0.05 mag. However, the observed increase in flux below about 2200 Å is almost certainly an artifact and should not be trusted.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The UV spectra of SN 2004ef have poor S/N; the object is relatively faint, being at a larger distance (∼120 Mpc). Only a single UV-prism spectrum was obtained of SN 2005M. The UV spectra of SN 2004ef and SN 2005M are shown in Figure 7.

Zoom In Zoom Out Reset image size

3.2.1. SN 2005cf

3.2.2. SN 2004dt

3.2.3. SN 2004ef and SN 2005M

3.2.4. Comparison of the UV Spectra

In Figure 8, we compare the near-maximum UV spectra of our four nearby SNe Ia with other mean low-redshift SN Ia spectra at similar phases. The spectra have been corrected for reddening in the MW. The mean low-redshift spectra come from several studies, including the mean spectrum constructed primarily with the HST STIS spectra (Cooke et al. 2011), the combined spectrum of a few nearby SNe Ia with archival HST and IUE data (Foley et al. 2012), and the spectral template established by Hsiao et al. (2007). As Cooke et al. (2011) did not apply any extinction correction in building their mean low-redshift spectrum, we apply a reddening correction to their spectrum, with E(B − V) = 0.10 mag and R_V = 3.1, to correct for the color difference. All of the spectra shown in the plot are normalized to approximately the same flux in the rest-frame wavelength range 4000–5500 Å. Note that the Hsiao et al. (2007) template spectra contain high-redshift SN Ia data from the Supernova Legacy Survey (Ellis et al. 2008); hence, they may suffer from evolutionary effects and may not represent the genuine spectrum of local SNe Ia in the UV region.

Zoom In Zoom Out Reset image size

The Foley et al. spectrum (mean phase t ≈ 0.4 days) and the Hsiao et al. t = −1 day spectrum agree fairly well within the uncertainties, and the STIS UV spectrum (mean phase t ≈ 1.5 days) also shows reasonable agreement given the difference in epoch. We see that there is a general shift of UV peak features to the red between t ≈ −1 day and t ≈ +1 day. Most notably, the large peak near 3100 Å at t ≈ −1 day shifts to ∼3150 Å at t ≈ +1 day. The underlying cause of the shift is probably just the recession in velocity of the UV photosphere with the expansion of the ejecta, but detailed modeling would be needed to verify this.

Of our sample, SN 2004dt appears to be unusually bright in the UV. Its UV emission is found to be stronger than the Foley et al. low-redshift mean spectrum for t ≈ −1 day by ≳ 75% (or at a confidence level ≳ 6σ–7σ) at wavelengths 2500–3500 Å. Differences are also present in the optical portion of the spectrum where the absorption features of IMEs are relatively strong and highly blueshifted (except for the S ii lines), although the integrated flux over this region does not show significant differences with respect to the templates. The ∼5000 Å absorption feature due to Fe ii and Fe iii blends appears to be rather weak, perhaps suggestive of a smaller amount of iron-peak elements in the outer ejecta of SN 2004dt.

The near-maximum spectrum of SN 2004ef displays prominent features of both IMEs and Fe-group elements, with the UV emission being weaker than the local composite spectrum by about 25%. Another notable aspect of Figure 8 is the deficiency of the UV emission in SN 2005M. This supernova exhibits the prominent Fe iii feature near 5000 Å in the earliest spectrum (Thomas 2005). SN 2005M has a smaller Δm₁₅(B); however, the UV flux emitted in the wavelength region 2500–3500 Å is found to be even lower than the mean value by ∼22%. One can see from Figure 8 that significant scatter is observed in these three events at wavelengths below ∼3500 Å. In this wavelength region, the continuum and the spectral features were proposed to be primarily shaped by heavy elements such as Fe and Co (Pauldrach et al. 1996). Thus, the large scatter in our sample might be related to variations of the abundance of Fe and Co in the outer layers of the exploding white dwarf (Höflich et al. 1998; Lentz et al. 2000; Sauer et al. 2008). On the other hand, the spectrum of SN 2005cf generally matches well the nearby comparison except for the Ca ii H&K feature, which exhibits the strongest difference in the optical region among our four events. The variations of Ca ii and Si ii absorptions at higher velocities suggest that additional factors, such as asphericity or different abundances in the progenitor white dwarf, affect the outermost layers (Tanaka et al. 2008).

Figure 9 shows a more detailed comparison of the UV–optical spectra of SN 2004dt and SN 2005cf at three different epochs (t ≈ −2 days, +5 days, and +14 days). The integrated fluxes of the spectra have been normalized over the 4000–5500 Å wavelength range. These two SNe Ia provide us good examples to examine the generic scatter in the UV region as they have quite similar photometric properties in the optical bands, such as the B − V color at maximum and the post-maximum decline rate Δm₁₅(B) (see Table 3). At t ≈ −2 days, the normalized flux obtained for SN 2004dt at 2500–3500 Å is much brighter than for SN 2005cf in the same wavelength region, with the flux ratio F_04dt/F_05cf ≈ 1.9. This flux ratio in the UV decreases quickly to ∼1.2 at t ≈ +5 days, and it becomes roughly 1.0 at t ≈ 2 weeks after maximum. Neither contamination by the background light from the host galaxy nor uncertainty in the reddening correction is likely to account for such a peculiar evolution of the UV flux for SN 2004dt; the influence of background emission would become more prominent as the object dims and SN 2004dt does not suffer significant reddening in the host galaxy (see Section 4). Such a fast decline of the UV flux is rarely seen in the existing sample of SNe Ia, possibly an indication of differences in the explosion physics or progenitor environment with respect to normal SNe Ia.

Zoom In Zoom Out Reset image size

The peak luminosity of SNe Ia in the optical bands has been extensively studied and found to correlate well with the width of the light curve around maximum light. Outliers have been identified as abnormal objects with different explosion mechanisms such as the overluminous objects SN 2003fk (Howell et al. 2006) and SN 2009dc (Silverman et al. 2011; Taubenberger et al. 2011) and the underluminous events SN 1991bg (Filippenko et al. 1992a; Leibundgut et al. 1993) and SN 2002cx (Li et al. 2003). Examining the relationship between the UV luminosity and Δm₁₅(B) may disclose a diversity, even for the so-called normal SNe Ia, since the UV emission is thought to be more sensitive to possible variations of the explosion physics or progenitor environment.

We have examined a sample of 20 SNe Ia with good photometry in the UV. Four among this sample are from the HST observations presented in this paper, and the rest are Swift objects published by Milne et al. (2010). The left panels of Figure 10 show the maximum absolute magnitudes of these objects in broadband U, B, and Swift uvw1/HST F250W versus Δm₁₅(B).⁶⁵ Following the analysis by Brown et al. (2010), we applied the red-tail corrections to the uvw1/F250W magnitudes to mitigate the effects of the optical photons on the UV flux. The U and uvw1/F250W magnitudes are also K-corrected for the large variation in the spectral flux at shorter wavelengths (but note that a UV/blue event like SN 2004dt may warrant a second spectral sequence for the sake of red-tail and K-corrections). The peak magnitudes in the uvw1/F250W filters are obtained by fitting the data with a polynomial or the template light curve of SN 2005cf. The parameters in the optical bands, such as the peak magnitudes and the B-band light-curve decline rate Δm₁₅(B), are estimated from the published light curves (see Table 3 and the references).

Zoom In Zoom Out Reset image size

After corrections for the extinction (assuming an extinction law with R_V = 2.3 for the SN host galaxy), linear fits to the subsample in Figure 10 with 0.8 mag < Δm₁₅(B) < 1.7 mag yield a root-mean-square scatter (i.e., σ values) of ∼0.2 mag in B and ∼0.4 mag in uvw1. Note that SN 2004dt and SN 2006X were excluded from the fit. The uvw1-band peak luminosity does show a correlation with Δm₁₅(B), but with steeper slopes and larger scatter than that in B and U. Given the fact that the dust obscuring the SN may have different origins (e.g., Wang 2005; Goobar 2008; W09a), the host-galaxy extinction corrections with R_V = 3.1 (MW dust) and R_V = 1.8 (LMC dust) are also applied to the absolute magnitudes. In both cases, the scatter also increases at shorter wavelengths. The relevant results of the best linear fit with different values of R_V to the M_max–Δm₁₅(B) relation are shown in Table 4.

The wavelength-dependent scatter can be driven in part by the uncertainty in the absorption corrections, as the UV photons are more scattered by the dust than the optical. Assuming a mean error ∼0.04 mag in E(B − V)_host (e.g., Phillips et al. 1999; Wang et al. 2006a), the residual scatter of the uvw1-band luminosity can be as large as ∼0.3 mag, which is still much larger than that found in B (∼0.15 mag). A notable feature in the left panel of Figure 10 is the UV excess seen in SN 2004dt and perhaps SN 2006X; they are found to be brighter than the corresponding mean value by ∼6.0σ and ∼3.0σ, respectively. The discrepancy of the observed scatter cannot be caused by the error in distance modulus, which applies equally to the UV and optical bands. This is demonstrated by the correlation between the peak colors and Δm₁₅(B), as shown in the right panels of Figure 10. The color correlation is distance independent, but it shows significantly large scatter in the UV. In particular, the uvw1/F250W − V colors of SN 2004dt and SN 2006X are found to be bluer than the mean value by ∼1.3 mag. This indicates that the uvw1/F250W filter is perhaps more a probe of SN Ia physics than a cosmological standard candle.

Strong emission in the UV is likely an intrinsic property of SN 2004dt rather than being due to improper corrections for the dust reddening, since we only applied a small reddening correction to this supernova, E(B − V)_host = 0.08 mag. Moreover, SN 2004dt still appears much brighter and bluer in the UV than the mean values defined by the other SNe Ia even if it is assumed to suffer little reddening. On the other hand, SN 2006X is heavily extinguished by dust having an abnormal extinction law, perhaps with R_V ≈ 1.5 (Wang et al. 2008a). Applying this extinction correction (which is an extrapolation from the results of Wang et al. 2008a) and the red-tail correction (which is an extrapolation from the results of Brown et al. 2010) to the UV magnitudes of SN 2006X would make it appear brighter than the other comparisons by ∼1.7 ± 0.6 mag in uvw1/F250W. We caution, however, that this result may suffer large uncertainties from the speculated extinction and red-tail corrections.

To examine how the relative flux in the UV evolves after the SN explosion, we compute the ratio of the UV emission (1800–3500 Å) to the optical (3500–9000 Å), F_UV/F_opt, for a few SNe Ia, as shown in Figure 11. The flux ratios obtained for SN 2005cf and SN 2005ke are overplotted. The dotted curve represents the Hsiao et al. (2007) template, and the dashed curve shows a combined template of SNe 1981B, 1989B, 1990N, 1992A, and 2001el (Stanishev et al. 2007). The flux ratios were calculated by integrating the flux-calibrated, UV–optical spectra except for SN 2005ke and SN 2006X. The integrated fluxes of these two SNe Ia are obtained approximately by the mean flux multiplied by the effective width of the passband. It is noteworthy that the flux ratio shown for SN 2006X might be just a lower limit, as we did not include the flux contribution below 2500 Å in the analysis because of the larger uncertainties in the extinction, K-corrections, and red-tail corrections at shorter wavelengths (e.g., Brown et al. 2010).

Zoom In Zoom Out Reset image size

We note that the F_UV/F_opt value of SN 2006X peaks at t ≈ 10 days before the B-band maximum, about 5 days earlier than for SN 2005cf and the templates. This feature is similarly observed in the fast-decliner SN 2005ke, which exhibited a possible signature of X-ray emission as well as evidence for a UV excess in the early nebular phase, perhaps suggestive of circumstellar interaction (Immler et al. 2006). Owing to the lack of earlier UV data, we cannot conclude whether such a feature is present in SN 2004dt. It is apparent that the flux ratio F_UV/F_opt exhibits a large scatter at comparable phases for the selected sample of SNe Ia. Consequently, the spread in bolometric corrections may affect significantly the inferring of the nickel masses from the light-curve peaks.

The overall properties of the SNe can be represented by their quasi-bolometric (UV/optical/IR) light curves shown in Figure 12. The near-infrared (NIR) photometry of SNe 2004ef and 2005M was taken from Contreras et al. (2010). The NIR emission of SN 2004dt was corrected on the basis of SN 2005cf (W09b). Similar corrections were applied to the comparison SNe Ia when the NIR observations were not available. Overall, the quasi-bolometric light curves of our SNe Ia are very similar in shape, with the exception of SN 2004dt, which shows a prominent "bump" feature. This "bump" is consistent with the secondary maximum visible in the I band (see Figure 2), being more prominent and occurring 10 days earlier than that of the other comparisons. This observed behavior suggests that SN 2004dt may have a larger, cooler iron core, or a higher progenitor metallicity according to the study of the physical relation between a supernova's NIR luminosity and its ionization state (Kasen 2006).

Zoom In Zoom Out Reset image size

The bolometric luminosity at maximum light of SN 2004dt is estimated to be L_max ≈ (1.7 ± 0.2) × 10⁴³ erg s⁻¹ with R_V = 2.3 and E(B − V)_host = 0.08 mag. This value is consistent with that obtained for SNe Ia with similar Δm₁₅ such as SN 2005cf [∼(1.5 ± 0.2) × 10⁴³ erg s⁻¹] and SN 2007af [∼(1.3 ± 0.2) × 10⁴³ erg s⁻¹] within 1σ–2σ errors. With the derived peak luminosity, we can estimate the ⁵⁶Ni mass ejected during the explosion (Arnett 1982). According to Stritzinger & Leibundgut (2005), the ⁵⁶Ni mass (M_Ni) can be written as a function of the bolometric luminosity at maximum and the rise time t_r:

$$\begin{eqnarray} L_{\rm max} &=& \big(6.45e^{-t_{r}/(8.8 \,{\rm days})} + 1.45e^{-t_{r}/(111.3 \,{\rm days})}\big)\nonumber\\ &&\times\left(\frac{M_{\rm Ni}}{M_{\odot }}\right)\times 10^{43} \,{\rm erg\ s^{-1}.} \emph {} \end{eqnarray} \tag{ 1 }$$

Taking the rise time t_r as ∼20 days for SN 2004dt and SN 2005cf, ∼18 days for SN 2004ef, and ∼24 days for SN 2005M (e.g., Ganeshalingam et al. 2011), the corresponding mass of the ejected ⁵⁶Ni is roughly estimated to be 0.9 ± 0.1 M_☉, 0.8 ± 0.1 M_☉, 0.4 ± 0.1 M_☉, and 0.7 ± 0.1 M_☉, respectively.

We see that the deduced M_Ni of the four SNe Ia shows significant scatter, and the value for SN 2004dt is slightly larger than the normal values. The net effect of a larger M_Ni would tend to delay the secondary maximum of the NIR light curves, inconsistent with what is seen in SN 2004dt. This inconsistency can be resolved if SN 2004dt has a shorter rise time, t_r ≈ 18.0 days. Given the observational evidence that longer t_r usually corresponds to slower post-maximum decline of the light curve (Ganeshalingam et al. 2011), SN 2004dt may rise to maximum at a faster pace. A shorter rise time is likely to be a common feature of the high-velocity SNe Ia (Zhang et al. 2010; Ganeshalingam et al. 2011). Therefore, the deduced nickel mass of SN 2004dt may have been overestimated to some extent. On the other hand, the implied M_Ni for SN 2005M seems quite normal and is lower than in slowly declining SNe Ia such as SN 1991T, indicating that the light-curve widths are not a single-parameter family of the ejected nickel mass.

The UV emission of SNe Ia is thought to originate predominantly from the outer layers of the ejecta because UV photons produced in deeper layers are subject to line blanketing by the wealth of bound–bound transitions associated with iron-group elements. Therefore, UV features are a promising probe to study the composition of the outer ejecta of SNe Ia.

Previous work has shown that the UV is particularly sensitive to the metal content of the outer ejecta (line-blocking effect) and their ionization (backwarming effect; Lentz et al. 2000; Sauer et al. 2008). Lentz et al. (2000) suggested a correlation between the emitted UV flux and the progenitor metallicity, with lower flux for models with higher metallicity. On the other hand, Sauer et al. (2008) proposed that the UV flux can become stronger at high metallicity in the outer layers due to an enhanced reverse-fluorescence scattering of photons from red to blue wavelengths and a change in the ionization fraction (backwarming effect). According to the model series for SN 2001eh and SN 2001ep (Sauer et al. 2008), as well as the prediction by Lentz et al. (2000), an abundance change of about ±2.0 dex (e.g., variations from 1/10 normal metallicity in the C+O layer to 10 times normal metallicity in the C+O layer) could lead to a change of up to ∼0.3 mag in F250W/uvw1. This is much smaller than that observed in SN 2004dt. Note that the above studies are based on the one-dimensional deflagration model W7 (Nomoto et al. 1984).

To illustrate the effect of different progenitor metallicity on the UV spectra in more detail, we also consider a sub-Chandrasekhar-mass detonation model proposed by Sim et al. (2010). In particular, we adopt the model of a 1.06 M_☉ CO white dwarf (model 1.06), which was found to give good agreement with the observed properties of normal SNe Ia in the optical. For this model, Sim et al. (2010) studied the effect of progenitor metallicity on the nucleosynthetic yields and, in turn, the synthetic observables. For that purpose they polluted the initial CO white dwarf with 7.5% ²²Ne (corresponding to a progenitor metallicity of ∼3 Z_☉). From this they find that the metallicity mainly affects the blue bands (specifically, they find the B-band peak magnitude to be ∼0.5 mag dimmer for the model with Z = 3 Z_☉ compared to the model with Z = 0). In Figure 13(a), we show the UV spectra of these models at maximum light, which were obtained using Monte Carlo radiative transfer code ARTIS (Kromer & Sim 2009). To get better coverage in progenitor metallicity, we added another model at Z = Z_☉ that was obtained in exactly the same manner as described by Sim et al. (2010). It is clearly seen that the UV flux increases significantly with decreasing progenitor metallicity. However, comparing the synthetic spectra to observed spectra of SN 2004dt, it is also obvious that the metallicity effect cannot be the dominant factor responsible for the unusual brightness of SN 2004dt in the UV, as it is not possible to enhance the UV flux farther than for the model with zero metallicity.

Zoom In Zoom Out Reset image size

SN 2004dt shows line-expansion velocities which are apparently larger than those of normal SNe Ia. To investigate if these higher ejecta velocities can explain the UV excess of SN 2004dt, we adopted the standard deflagration model W7 (Nomoto et al. 1984), which is known to reproduce the observed maximum-light spectra of SNe Ia very well. The t ≈ 0 day models shown in Figure 13(b) were calculated using the time-dependent PHOENIX code in local thermodynamic equilibrium (Jack et al. 2009, 2011). The velocities in W7 were increased by a uniform factor (20% or 40%). The densities were adjusted so that the total mass was conserved. Figure 13(b) clearly shows enhanced UV emission when the velocities are larger; the effects of line blending in the UV serve to increase the radius of the UV pseudophotosphere. Normalizing the spectral flux over the 4000–5500 Å region, we find that the UV flux emitted in the 2500–3500 Å region can be increased by about 40% if the expansion velocity v_exp is increased by 20% everywhere in W7. Note that increasing v_exp by 40% does not further enhance the UV flux, and the resulting flux increase in the 2500–3500 Å region drops to 34%. This indicates that an increase of the expansion velocity can lead to more UV flux, but it cannot reproduce the very large flux enhancement near 3000 Å in SN 2004dt.

Interestingly, SN 2004dt is the most highly polarized SN Ia ever observed. Across the Si ii line its polarization $P_{\rm Si\,\mathsc{ii}}$ reaches up to 1.6% at ∼1 week before maximum light (Wang et al. 2006a), indicating that its Si ii layers substantially depart from spherical symmetry. Among the comparison sample, SN 2006X also has a large degree of polarization in the early phase, with $P_{\rm Si\,\mathsc{ii}} \approx 1.1$% at 6 days before maximum light (Patat et al. 2009). Thus, it could well be that viewing-angle effects play a major role in the observed UV excess of SN 2004dt. Breaking of spherical symmetry in the explosion is also thought to be a critical factor responsible for the observed scatter among SNe Ia (e.g., Wang et al. 2007; Kasen et al. 2009; Maeda et al. 2010; Maund et al. 2010).

Kromer & Sim (2009) recently studied the effect of asymmetric ejecta on the light curves and spectra of SNe Ia using an ellipsoidal (prolate) toy model. They found that SNe observed along the equator-on axis are always brighter than those observed along the pole-on axis. This effect is strongest in the bluer bands because the photons at short wavelengths are more strongly trapped than photons in other bands and therefore tend to preferentially leak out along the equatorial plane where the photospheric velocity is smallest. Around maximum light, the difference ΔM = M_pole − M_equator measured in the U band could be ∼0.4 mag larger than that in the V band (Kromer & Sim 2009). After maximum light, this line-of-sight effect becomes weaker with time as the ejecta become optically thin at these wavelengths. Thus, the UV excess seen in SN 2004dt could also be due to a geometric effect.

We have described how there are several competing possible explanations for the enhanced UV flux in SN 2004dt. This is due to the complex nature of line blanketing in the barely optically thick, but highly scattering dominated, differentially expanding SN Ia atmosphere. This effect was described in part by Bongard et al. (2008), who showed that the complete spectrum is formed throughout the semitransparent atmosphere and that Fe iii lines produce features that are imprinted on the full spectrum, partially explaining the UV excess found in SN 2004dt. Indeed, the differing results of Höflich et al. (1998) and Sauer et al. (2008), compared with those of Lentz et al. (2000), are perhaps due to the complex nature of spectral formation, ionization, and radiative transfer effects.

While asymmetry could play some role, there seems to be enough variation in spherically symmetric models due to effects of varying ionization that is likely produced by a density profile that differs from that of W7. Recall that W7 has a density bump that occurs when the flame is quenched and momentum conservation causes material to pile up ahead of the dying flame. Also, the higher brightness and indications of a somewhat higher nickel mass in SN 2004dt will affect the ionization state of the iron-peak elements, which are almost certainly responsible in part for the observed UV excess. More UV observations of nearby SNe Ia are needed to thoroughly understand these effects, and whether they are due to asymmetries or other phenomena in the explosion.