NEOWISE-R Caught the Luminous SN 2023ixf in Messier 101

NEOWISE observed the SN site pre-explosion, as part of routine operations, 152 times between 2013 December 18.26 and 2022 December 18.99 UT (MJD 56644.2618 and 59931.9958, respectively). The first pre-SN pair of single exposures occurred 3438.49 d prior to explosion, while the last was 150.75 d pre-SN. The progenitor candidate was not detected in any of these observations (e.g., Hiramatsu et al., 2023; Jencson et al., 2023; Soraisam et al., 2023; Van Dyk et al., 2024). See also Section A of this paper. The lack of detection tends to rule out luminous eruptions or outbursts from the progenitor candidate during that time period prior to explosion, as discussed in the above studies. (The star was also not detected by WISE and the post-cryogenic NEOWISE prior to reactivation, between 2009 and 2011; see Van Dyk et al. 2024.)

The SN itself was detected during the nineteenth sky coverage since the start of the Reactivation Mission, from day 3.631 through 10.901 (2023 May 22.38 through May 29.65). We note that the gap of several days in the early-time data is caused by a “Moon toggle” procedure, in which the spacecraft pointing skips ahead to avoid the Moon, and then slews back to observe the “skipped” area of sky (see Wright et al. 2010 for a description); in our case, this was advantageous, as we were able to sample the SN’s evolution about a week after the first detections. All of these data are publicly available via the NASA/IPAC Infrared Science Archive (IRSA; https://irsa.ipac.caltech.edu/); see Figure 1. The SN was then captured again at late times during the twenty-first sky coverage, from day 211.736 through 213.351 (2023 December 16.49 through 18.10), and then again from day 370.875 through 372.476 (2024 May 23.63 through 25.23). Those data were still pre-release at the time of this writing and were also obtained via IRSA. Given the separation of the SN from the general environs of the neighboring giant H ii region NGC 5461, the detections are relatively clean. The SN detections are listed in Table 1. The quantities W1mpro and W2mpro are profile-fit photometry magnitudes at 3.4 and 4.6 $\mu$m (bands W1 and W2), respectively, in the Vega system¹¹1See the NEOWISE Data Release Explanatory Supplement, https://wise2.ipac.caltech.edu/docs/release/neowise/expsup/., without any further special processing or additional background subtraction applied. The resulting light curves are shown in Figure 2.

Rather than select every pair of observed NEOWISE W1 and W2 data points to analyze, for illustrative purposes we have chosen to consider just two sets at early times, the very first one from MJD 60086.381 (day 3.631) and from 60093.586 (day 10.836). These adequately represent the two periods of early-time sampling of the light curves in these bands.

In order to put the NEOWISE data in context with the overall SED at these two epochs, we accumulated published light-curve data at UV, optical, and near-IR wavelengths corresponding to (or bracketing) the epochs. The data sources were then an amalgam of Swift UVW2, UVM2, UVW1 from Zimmerman et al. (2024), SDSS $ugriz$ and Johnson $BVJHK_{s}$ from Teja et al. (2023), Johnson $UBV$, and SDSS $griz$ from Hiramatsu et al. (2023), and Johnson $JHK$ from Yamanaka et al. (2023).

Since the NEOWISE measurements are in the Vega system, the entirety of the dataset presented by Teja et al. (2023), which is in the AB system, had to be converted to Vega magnitudes. The SDSS magnitudes from Hiramatsu et al. (2023) also required a similar conversion. No conversion was needed for the Yamanaka et al. (2023) $JHK$ photometry. For day 3.631 the available UV-optical data were quasi-contemporaneous with the NEOWISE points; however, $JHK$ required a linear interpolation between two bracketing epochs (the earlier epoch was at day 3.4, very close in time to NEOWISE). For day 10.836 none of the complementary data were contemporaneous, so we were forced to interpolate between bracketing epochs at all wavelengths.

The resulting SED is shown in Figure 3. In addition to the UV-optical photometric points we included an FTN-FLOYDS-N spectrum of the SN obtained by Bostroem et al. (2023) on 2023 May 22, which we downloaded from WISeREP⁴⁴4https://www.wiserep.org (Yaron & Gal-Yam, 2012). Both the spectrum and the photometry were first reddening-corrected. This spectrum was further renormalized to the (dereddened) SN $V$-band brightness. As can be seen in the figure, the overall agreement is reasonable between the spectrum and the photometric points across the common wavelength range.

We then attempted to fit a single, simple blackbody to the SED at day 3.631. We found that a hot, $\sim 26,630$ K blackbody provides a good fit to the optical data, although a clear excess exists in the UV relative to this fit. The fit implies that the SN luminosity at that epoch was $\gtrsim 4.1\times 10^{43}$ erg s^-1; this is a lower limit, since there is clearly additional luminosity in the UV. This is consistent with the evidence for strong, early-time interaction of the SN shock with dense CSM (e.g., Bostroem et al., 2023; Jacobson-Galán et al., 2023; Teja et al., 2023; Martinez et al., 2024; Zimmerman et al., 2024); specifically, interaction in SNe II can strengthen the continuum flux and boost emission lines in the UV simultaneously (Dessart & Hillier, 2022). The blackbody radius is then $R_{\rm hot}\approx 3.4\times 10^{14}$ cm. The early-phase photometric and spectroscopic UV-optical SN observations provided evidence for strong interaction of the SN shock with a dense, confined ($<2\times 10^{15}$ cm) CSM (e.g., Jacobson-Galán et al., 2023; Smith et al., 2023; Bostroem et al., 2023; Teja et al., 2023). Zimmerman et al. (2024) further refined the extent of the dense CSM to $R_{\rm CSM}\approx 2\times 10^{14}$ cm and concluded that it actually delayed shock breakout (SBO) from hours after explosion to $\sim 3$ d. In other words, the very first NEOWISE data were likely obtained within just hours after SBO, and the inferred $R_{\rm hot}$ is consistent with the shock having already overrun the confinements of the dense CSM. In fact, following Zimmerman et al. (2024) and assuming a SN expansion velocity $v_{\rm exp}=8,000$ km s^-1, on day 3.631 the shock would have been at $\sim 2.5\times 10^{14}$ cm, roughly consistent with $R_{\rm hot}$ (if $v_{\rm exp}$ instead had been a somewhat higher, $\sim 11,000$ km s^-1, the two radii would be in better agreement).

Particularly fascinating here is that an obvious excess in flux, relative to the hot blackbody, can be seen in Figure 3 at $\gtrsim 1.5$ $\mu$m. We found that we could account for this IR excess with an additional much cooler blackbody, at $\sim 1620$ K. Including this extra blackbody provides a reasonable fit to $JHK_{s}$ and W1, although it does not quite fit W2 as well. This additional source of IR emission is of a comparatively far smaller luminosity, $\sim 3.2\times 10^{40}$ erg s^-1, than the SN shock (it accounts for $\lesssim 0.1$% of the total emission). The corresponding blackbody radius is $R_{\rm IR}\approx 2.6\times 10^{15}$ cm. Kilpatrick et al. (2023) and Van Dyk et al. (2024), for example, inferred that the RSG progenitor candidate was surrounded by a dusty shell with an inner radius of $R_{\rm in}\approx(0.5$–$1.0)\times 10^{15}$ cm. The assumption Van Dyk et al. (2024) made in their modeling of the star was that the shell extended out to $1000\ \times R_{\rm in}$, with the dust density declining $\propto r^{-2}$. Furthermore, $\sim 1620$ K is roughly the estimated evaporation temperature of $\sim 0.01$–0.1 $\mu$m-sized silicate-dominated dust in SN environments (e.g., Gall et al., 2014). (Note that Van Dyk et al. 2024 were able to fit the reddening-corrected SED of the pre-explosion dust shell with a simple $\sim 1761$ K blackbody.) Thus, we speculate that the IR excess was emanating from the dusty CSM shell, with the SN shock still within $R_{\rm in}$.

This analysis including the NEOWISE data lends credence to the overall picture of the progenitor star, inferred via the modeling of the observed SED of the star. The estimated luminosity from the optically-thin dust was still about two orders-of-magnitude larger than the luminosity of the progenitor candidate ($\sim 9\times 10^{4}\ L_{\odot}$, or $\sim 3.5\times 10^{38}$ erg s^-1; Van Dyk et al. 2024), which implies that the dust shell was likely heated by and was reprocessing the UV emission from the interaction of the SN shock with the dense inner CSM. We note that much of the CSM dust was likely destroyed immediately after explosion by high-energy (extreme UV to $\gamma$-ray) photons from the blast, through grain sublimation, vaporization, and extreme grain charging effects (Jones, 2004).

We built a similar SED from the available UV/optical/near-IR/NEOWISE photometry from day 10.836. Again, we added to this a dereddened and renormalized FTN-FLOYDS-N SN spectrum from WISEReP from 2023 May 29 (Bostroem et al. 2023). The resulting SED is shown in Figure 4. We attempted to fit a warm, $\sim 9050$ K blackbody to the data; however, as can be seen in the figure, while this fit is very good in the IR, it diverges significantly for wavelengths $<1\ \mu$m. If we consider the radiation models for SNe IIP by Dessart et al. (2013; D13), specifically, the m15mlt1 model (with an initial progenitor mass of $15\ M_{\odot}$ and radius $R_{\star}=1107\ R_{\odot}$) computed at day 11, it provides a remarkably good comparison, from the UV through to the mid-IR, with the observed data. This implies that what we were seeing at that epoch, including with NEOWISE, was SN ejecta-dominated emission. If we take into account the blackbody fit, the inferred luminosity is $\gtrsim 5.8\times 10^{42}$ erg s^-1, which is a lower limit, since a significant amount of luminosity is still emerging from the SN at $<1\ \mu$m. In fact, if we integrate the D13 model (for instance), we obtain a luminosity of $\sim 1.5\times 10^{43}$ erg s^-1. The blackbody radius is $R_{\rm warm}\approx 1.1\times 10^{15}$ cm. Once again, assuming $v_{\rm exp}=8,000$ km s^-1, the shock radius would have been $\sim 7.5\times 10^{14}$ cm, roughly consistent with $R_{\rm warm}$; similar to day 3.631, a somewhat higher $v_{\rm exp}=11,000$ km s^-1 would result in better agreement between the two radius estimates.

The blackbody radii and temperatures that we have calculated for both days 3.6 and 10.8 are in a good agreement with the results of Singh et al. (2024). Note, however, that we have assumed a spherical and homogeneous medium, while early-phase spectropolarimetric data implied the presence of an aspherical dense CSM and a clumpy, low-density extended CSM around SN 2023ixf (Vasylyev et al. 2023; Singh et al. 2024). Nevertheless, our conclusions also fit with that of Vasylyev et al. (2023), in that they concluded that the expanding SN ejecta just emerged from the dense CSM region around Day 3.5.

We computed the averages of the various measurements obtained by NEOWISE at late times in the SN’s evolution, between MJD 60294.5 and 60296.1. SN 2023ixf was already well on the radioactively-powered exponential tail by this point. The averages are $11.895\pm 0.023$ mag and $10.509\pm 0.020$ mag in W1 and W2, respectively, with reference to day 213.

Thanks to the active worldwide follow-up observations of SN 2023ixf, we can also build the late-time optical-NIR SED for the SN. We assembled the unpublished optical $BVgriz$ photometry from Konkoly Observatory (from 2023 December 18.5, day 214; J. Vinkó et al. 2024, in preparation) and Lick Observatory $BVRI$ data (interpolated to the same date), as well as a single Lick optical spectrum (from 2023 December 12, day 208; W. Zheng et al. 2024, in preparation); see Section 2. We show the combined optical SED in Figure 5. Since we are unaware of any available near-IR data close to this late epoch, we included spectra and fluxes from models of SN IIP explosions for comparison: The day 207 s15p2 model spectrum courtesy of Luc Dessart, and the interpolated and distance-scaled day +193 and +242 model fluxes for the Type IIP SN 2012aw from Pejcha & Prieto (2015). (We compared the Fe ii line velocities for SN 2023ixf from Zheng et al. 2024, to that of SN 2012aw from Bose et al. 2013 and found a very good match even at late phases; we also adopted the SN 2012aw distance of 9.9 Mpc from Bose et al. 2013.) Both the model spectrum and the set of optical model fluxes appear to compare well with our late-time measured optical data; thus, we find it to be a reasonable approach to use the near-IR components of these models as references during the analysis of the late-time optical-IR SED.

As can be seen in Figure 5, the NEOWISE 3.4 and 4.6 $\mu$m data show a clear excess relative to the model fluxes. We fit a single blackbody to the optical-near-IR part of the SED and found a warm component with temperature $\sim 4650$ K and corresponding radius $\sim 5.6\times 10^{14}$ cm. While SN ejecta are not expected to be realistically represented by a blackbody at late epochs, we can then use the result of this fit to provide a characterization of the colder, longer-wavelength excess that is apparent.

As we noted in Section 3.1, we assume that the majority of pre-existing dust grains, located in the dense, confined circumstellar shell, have been evaporated at early times after explosion; therefore, these can no longer be responsible for the excess. Thus, we consider two possible alternate scenarios: (i) At day 213, newly-formed dust existed either in the inner (unshocked) ejecta or in the contact discontinuity between the forward and reverse shocks (the CDS); or (ii) radiation was being emitted by more distant pre-existing dust grains in the SN environment, heated collisionally or radiatively by the (forward) SN shock. However, from only the two mid-IR data points, a detailed investigation of the origin and properties of the assumed late-time dust is challenging to carry out — we hope that pending JWST observations of the SN will provide greater insight into these two scenarios. Furthermore, we note that the large 4.6 $\mu$m flux excess at day 213 can also be explained by the emergence of the 1–0 vibrational band of carbon monoxide (CO) at 4.65 $\mu$m, as seen in some SNe IIP with observed mid-IR spectra at a similar age (e.g., Kotak et al. 2005; 2006; Szalai et al. 2011; Szalai & Vinkó 2013), and also predicted by modeling of exploding RSGs by McLeod et al. (2024). The presence of a late-time mid-IR excess is in agreement with the results from Singh et al. (2024); they found that the flattening in the $K_{s}$-band light curve and the attenuation of the red-edge of the H$\alpha$ line profile $\sim$125 d after explosion is indicative of the onset of molecular CO and, hence, dust formation in SN 2023ixf. This possibility should be also taken into account during any analysis.

Thus, we also excluded the W2 flux and fit a two-component blackbody model to the remainder of the optical-IR SED, holding the parameters from the hot component fixed. Since a fit to only one mid-IR point would not provide any physically relevant information, we applied temperature constraints and assumed two possible scenarios for the “warm” dust component: (i) First, we assumed the highest theoretical dust temperature possible for amorphous carbon dust, $T_{\rm IR}=2600$ K (see, e.g., Gall et al. 2014); and, (ii) second, we assumed a “typical” temperature of “warm” dust ($T_{\rm IR}=700$ K) seen in SNe IIP at a similar age (see, e.g., Kotak et al. 2009; Szalai et al. 2011; Szalai & Vinkó 2013). Using these two assumptions, we found a blackbody radius of $R_{\rm IR}\sim 1.5\times 10^{15}$ cm and $\sim 1.6\times 10^{16}$ cm, respectively, for the two assumptions; see Figure 6).

Extrapolating the Fe ii velocities (W. Zheng et al. 2024, in preparation) to day 213, we estimated $\sim 1400$–1500 km s^-1 for the ejecta velocity, which results in (2.6–2.8) $\times 10^{15}$ cm for the ejecta radius. This implies that, in the case of very high-temperature dust (the first scenario above), these grains possibly could be within the ejecta. However, assuming a more realistic dust temperature (the second scenario), the dust would more likely be outside the ejecta. As Van Dyk et al. (2024) concluded, the dusty shell of the progenitor likely had an inner radius of $\sim 10^{15}$ cm, with an $r^{-2}$ density distribution extending outward. Thus, the pre-explosion shell itself could easily have extended out to $\sim 1.6\times 10^{16}$ cm (well beyond the confined volume); the dust we infer here could have been located within the CSM, being either pre-existing or newly-formed in the CDS.

We present in Figure 7 an optical-IR SED comprised of data obtained in the day 370–372 interval. Since at that range of epochs the shape of the measured optical SED appears to differ significantly from Type IIP atmospheric models by either Dessart et al. (2023) or Pejcha & Prieto (2015), we are unable to directly estimate the amount of IR excess in the manner that we did for the day 211–213 SED. Nevertheless, clear excesses at both 3.3 and 4.6 $\mu$m are obvious.

We have considered all of the NEOWISE data for SN 2023ixf, in the context of the mid-IR emission from other SNe, particularly SNe IIP. To achieve this, we compared SN 2023ixf with the sample of SNe from Szalai et al. (2019), who presented photometric data obtained by Spitzer in the IR Array Camera (IRAC; Fazio et al. 2004) bands at 3.6 and 4.5 $\mu$m during both the cryogenic and Warm missions; see Figure 8. While the bandpasses for Spitzer IRAC differ slightly from those of NEOWISE W1 and W2 channels, we can draw some basic conclusions.

First, as noted above, SN 2023ixf has become the best-observed SN IIP in the mid-IR during the first several days after explosion. Second, SN 2023ixf is one of the most luminous SNe IIP in the mid-IR ever seen. This statement is especially striking, considering the 4.5/4.6 $\mu$m photometric evolution of SNe IIP (Figure 8; right panel) — at day 213, SN 2023ixf is more luminous than at early times and than any other SNe IIP detected so far at late times. We can speculate on why this is the case. The early-time mid-IR luminosity is consistent with the overall excess observed at other wavelengths and can be explained by the shock-CSM interaction. Among the Spitzer sample, most of the SNe were either of low luminosity (e.g., SN 2004dj) or were otherwise normal (e.g., SN 2007od, SN 2011ja); only SN 2004et was observed to be somewhat extraordinary (e.g., Maguire et al. 2010; see also Shahbandeh et al. 2023). The late-time mid-IR excess for SN 2023ixf could be due to post-explosion dust formation, as we have mentioned above. However, we cannot say much more about this, based on the NEOWISE data alone; observations with JWST will likely provide significantly more insight on this.