A low-energy explosion yields the underluminous Type IIP SN 2020cxd

We present observations and analysis of SN 2020cxd, a Low luminous (LL), long-lived Type IIP SN. This object was a clear outlier in the magnitude-limited SN sample recently presented by the ZTF Bright Transient Survey. We demonstrate that SN 2020cxd is an additional member of the group of LL SNe, and discuss the rarity of LL SNe in the context of the ZTF survey, and how further studies of these faintest members of the CC SN family might help understand the underlying initial mass function for stars that explode.


Introduction
Stars more massive than about 8 M end their lives with the collapse of their iron core. Type II supernovae (SNe) are the most common among these core-collapse (CC) explosions, and are characterized by the presence of hydrogen in their spectra. Type II SNe constitute a diverse class, with light curves (LCs) showing different decline rates across a continuum, and a large range in A&A proofs: manuscript no. astroph cally ∼ 0.005 M (Spiro et al. 2014, their fig. 13). LL SNe II often also display slow expansion velocities suggesting low explosion energies.
It is now well established that the progenitors of many SNe II are red supergiant (RSG) stars (Smartt et al. 2009) and it is suspected that the LL SNe IIP originate from stars with relatively low zero-age main-sequence (ZAMS) masses (∼ 8 − 10 M , Pumo et al. 2016;O'Neill et al. 2021). However, other studies have suggested the possibility that the progenitors are more massive RSGs with large amounts of fallback material (Zampieri et al. 2003), or electron capture supernova (ECSN) explosions of super-asymptotic giant branch (SAGB) stars (Hiramatsu et al. 2021). There are several methods available to derive information about the exploded star, including identifying SN progenitors in archive images (e.g., Smartt et al. 2009), light curve modeling (e.g., Arnett & Fu 1989), and late-time nebular spectral modeling (e.g., Jerkstrand et al. 2012Jerkstrand et al. , 2018Dessart et al. 2013;Lisakov et al. 2016).
For a typical Initial Mass Function (IMF), over 40% of the potential CC SN progenitors reside in the 8 − 12 M range and 25% in the 8 − 10 M range (Sukhbold et al. 2016). The fact that the LL SNe are rarely detected and under-explored is of course due to the many observational biases challenging the discovery and follow-up of such faint transients. Most of them have been discovered in targeted searches of relatively nearby galaxies. However, with the ongoing revolution in transient science, with systematic non-targeted surveys uncovering thousands of transients, this is beginning to change. In this paper we use data from one of the ongoing surveys, the Zwicky Transient Facility (ZTF; Bellm et al. 2019;Graham et al. 2019). In particular, Fremling et al. (2020) introduced the ZTF Bright Transient Survey (BTS), which provides a large and purely magnitudelimited sample of extragalactic transients in the northern sky, suitable for detailed statistical and demographic analysis. Perley et al. (2020a) presented early results of the BTS, which is almost spectroscopically complete down to a target magnitude of 18.5. In particular, they presented a CC SN luminosity function, in which a significant fraction of the CC SNe is very dim. They state that this is in agreement with works arguing that the "SN rate problem" (Horiuchi et al. 2011) can be resolved using galaxy-untargeted surveys and including the faint end of the SN luminosity function.
Similar efforts to construct an unbiased view of the Type II SN luminosity function comes from the parallel ZTF volumelimited survey, the ZTF Census of the Local Universe (CLU) experiment, which extends the classification threshold to m 20 mag for transients occurring in known galaxies within D < 200 Mpc (De et al. 2020). These results also imply that the lowest luminosity events are more common than previously appreciated (Tzanidakis et al. in prep.). The ultimate aim of such large efforts is to be able to connect SN rates with star-formation rates and to couple the stellar evolution IMF to the known SN populations.
Out of the 171 Type II SNe presented by Perley et al. (2020a) one object clearly stands out as both longer duration and considerably fainter than the rest of the population, SN 2020cxd (a.k.a. ZTF20aapchqy, Perley et al. 2020a, their fig. 7). The authors declare that this is "almost certainly the explosion of a massive star". In this paper we take the opportunity to present SN 2020cxd in more detail. We argue that this transient is not only interesting as a single object given the rarity of LL SNe in the literature, but that this event is important in context with the overall population of SNe II provided by the strict and clear criteria of the highly complete BTS sample.
The paper is structured as follows. In Sect. 2, we outline the observations and corresponding data reductions, including Sect. 2.1 where we present the discovery and classification of SN 2020cxd. The ground-based optical SN imaging observations and data reductions are presented in Sect. 2.2, and in Sect. 2.3 we describe the Swift observations. A search for pre-explosion outbursts is done in Sect. 2.4, the optical spectroscopic follow-up campaign is presented in Sect. 2.5, and a discussion of the host galaxy is provided in Sect. 2.6. An analysis and discussion of the results is given in Sect. 3 and we provide a summary in Sect. 4. Some of the more technical aspects of the analysis are presented in the Appendices.

Observations and Data reduction
2.1. Discovery and classification SN 2020cxd was first discovered with the Palomar Schmidt 48-inch (P48) Samuel Oschin telescope on February 19, 2020 (JD discovery = 2458899.0306), as part of the ZTF survey. The first ZTF detection was made in the r band, with a host-subtracted magnitude of 17.69 ± 0.05 mag, at the J2000 coordinates α = 17 h 26 m 29.26 s , δ = +71 • 05 38.6 . The first g-band detection came in 31 minutes later at 17.63 ± 0.04. An on-duty astronomer (JS) immediately triggered follow-up observations, and several telescopes started observing shortly thereafter. The discovery was reported to the Transient Name Server (TNS 1 ) by Nordin et al. (2020) with a note that the last non-detection was three days earlier on February 16, with a global limit of 20.4 in the g band. We can therefore constrain the epoch of explosion for this SN with good precision. In this paper we will adopt an explosion date of JD explosion = 2458897.5301 with an uncertainty of ±1.5 days, as given by the epoch half way between discovery and last non-detection 2 .
SN 2020cxd is positioned in the nearby spiral galaxy NGC 6395 which has a measured redshift of z = 0.003883. According to the NASA/IPAC Extragalactic Database (NED) 3 catalog, the peculiar motion corrected distance for a standard cosmology (Ω M = 0.3, Ω Λ = 0.7, and h = 0.7) is 23 Mpc, whereas the most recent Tully-Fisher measurement reported on the same site is 20 Mpc. In this paper we will adopt a distance modulus of 31.7±0.3 mag (22 ± 3 Mpc). We also obtain the amount of Galactic extinction along the line of sight using the NED extinction tool 4 (based on the dust map of Schlafly & Finkbeiner 2011) and adopt E(B − V) = 0.035 mag. The SN together with the host galaxy and the field of view is shown in Fig. 1. SN 2020cxd was classified as a Type II SN (Perley et al. 2020b;Perley 2020) based on a spectrum obtained on February 20, 2020, 17 hours past discovery, with the Liverpool telescope (LT) equipped with the SPectrograph for the Rapid Acquisition of Transients (SPRAT). That spectrum revealed a blue continuum with hydrogen P-Cygni features (broad Hα and Hβ). The classification was consolidated with a spectrum taken 4 hours thereafter with the Palomar 60-inch telescope (P60; Cenko et al. 2006) equipped with the Spectral Energy Distribution Machine (SEDM; Blagorodnova et al. 2018). Fig. 1 SN 2020cxd in the nearby galaxy NGC 6395. The g-band image subtraction is shown in the top panels (subtraction to the right), with the SN image observed on 2020 April 24, 64.9 days after the first ZTF detection (to the left). The bottom panel shows a gri-colour composite image of the host galaxy and its environment. It was composed of ZTF g−, r− and i−band pre-explosion images of the field.

Optical photometry
Following the discovery, we obtained regular follow-up photometry during the 100+ days long plateau phase in g, r and i bands with the ZTF camera (Dekany et al. 2020) on the P48. Early LT photometry in ugriz was also obtained at one epoch to measure the colours, and a campaign with the Swift observatory was launched (Sect. 2.3). Later on, after the drop from the plateau, we also obtained a few epochs of photometry in gri with the SEDM on the P60, with the LT telescope and with the Nordic Optical telescope (NOT) using the Alhambra Faint Object Spectrograph (ALFOSC). The photometric magnitudes of SN 2020cxd are listed in Table 2.
The LCs from the P48 come from the ZTF pipeline (Masci et al. 2019). Photometry from the P60 were produced with the image-subtraction pipeline described in Fremling et al. (2016), with template images from the Sloan Digital Sky Survey (SDSS; Ahn et al. 2014). This pipeline produces point spread function (PSF) magnitudes, calibrated against SDSS stars in the field. For the late NOT images, template subtraction was performed with hotpants 5 , using archival SDSS images. The magnitudes of the transient were then measured using SNOoPY 6 and calibrated against SDSS stars in the field. All magnitudes are reported in the AB system. The reddening corrections are applied using the Cardelli et al. (1989) extinction law with R V = 3.1. We do not correct for host galaxy extinction, since there is no sign of narrow Na i d absorption lines in our spectra. This is basically an assumption, and we discuss some of the implications of this towards the end of the paper. The gri LCs are shown in Fig. 2.
We used a Gaussian Process (GP) algorithm 7 to interpolate the photometric measurements and found 8 that the peak happened at m peak r = 17.14±0.01 after t rise r = 109.50±1.60 rest frame days past explosion. In the i band, the photometric behavior followed the same trend and reached a maximum at m peak i = 17.00 after t rise i = 113.46 rest frame days. The g-band light curve actually peaks at the first detection epoch. It slowly declines before it flattens out on a plateau which reached a maximum at m g = 18.08 after 115.26 days. Thereafter the SN declined fast, and soon became too faint for detection with the P48. We then activated the LT and NOT telescopes. The r-and i-band observations on the tail suggest a linear decline. We performed a linear fit after ∼ 150 days, and found that the best fit slopes for the r and i band are (8.86 ± 1.33) × 10 −3 and (9.99 ± 0.79) × 10 −3 mag days −1 , which are consistent with the radioactive cobalt decay slope (λ Co = 1/111.3 × 2.5/ln(10) ∼ 9.76 × 10 −3 mag days −1 ). For the g-band images on the tail, we found no apparent flux in the LT and NOT residual images down to 2.5 sigma (upper limits are also presented in Table 2).

Swift-observations: UVOT photometry
A series of ultraviolet (UV) and optical photometry observations were obtained with the UV/Optical Telescope onboard the Neil Gehrels Swift observatory (UVOT; Gehrels et al. 2004;Roming et al. 2005) between February 20 and June 2, 2020. Our first Swift/UVOT observation was performed on February 20, 2020, (JD = 2458899.8624), 0.83 days past discovery (2.3 days past estimated explosion). The SN was detected in all bands. The brightness in the UVOT filters was measured with UVOTspecific tools in HEAsoft 9 .
Source counts were extracted from the images using a circular region with a radius of 3 . The background was estimated using a circular region with a radius of 39 . The count rates 9 https://heasarc.gsfc.nasa.gov/docs/software/heasoft/ version 6.26.1.
were obtained from the images using the Swift tool uvotsource and converted to magnitudes using the UVOT photometric zero points (Breeveld et al. 2011) and the new calibration data from September 2020 10 . We obtained a final UVOT epoch in August 2020, after the SN had faded, to remove the host contribution from the transient light curve. A log of the Swift observations is provided in Table 3. This includes six epochs in total, and these epochs are also indicated in Fig. 2. Such a dense and early UV coverage is very rare for LL SNe, and is important for constructing the bolometric light curve needed to model the SN in Sect. 3.3, see also Sect. A.

Pre-discovery imaging
Since the SN occurred in a very nearby galaxy we decided to search for pre-explosion activity of the progenitor star. For this purpose we downloaded IPAC difference images 11 , performed forced photometry at the SN position (Yao et al. 2019), and applied quality cuts and searched for pre-explosion detections as described by Strotjohann et al. (2021). ZTF started to monitor the position of SN 2020cxd 2.5 years before the discovery and we searched for pre-explosion outbursts in 999 observations collected on 307 different nights that were obtained in the g, r and i bands. No outbursts are detected above the 5σ threshold when searching unbinned or binned (1 to 90day long bins) LCs. Figure 3 shows the absolute magnitude LC in 7-day-long bins, where filled data points are significant at the 5σ level while arrows indicate 5σ upper limits. The g-band (rband) observations are available in 69 (68) out of 132 weeks before the SN discovery and we can exclude flares brighter than an absolute magnitude of −10 in 46 weeks (41 weeks), i.e. 34% (31%) of the time.

Optical spectroscopy
Spectroscopic follow-up was conducted with SEDM mounted on the P60 and with the LT SPRAT spectrograph on La Palma. Further spectra were obtained with the NOT using ALFOSC as well as one epoch with Gemini North and GMOS and a final spectrum with Keck and LRIS. A log of the 9 obtained spectra is provided in Table 4. SEDM spectra were reduced using the pipeline described by Rigault et al. (2019) and the LT and NOT spectra were reduced using standard pipelines. The spectra were finally absolute calibrated against the r-band photometry using the GP interpolated magnitudes and then corrected for Milky Way (MW) extinction. All spectral data and corresponding information is made available via WISeREP 12 (Yaron & Gal-Yam 2012). We present the sequence of spectra in Fig. 4, and the epochs of spectroscopy are also highlighted in Fig. 2. Overall, we managed to acquire spectra over the full relevant range of the SN evolution, but it should also be noted that the photospheric part of the evolution is mainly covered by robotic telescopes providing limited resolution and signal.

Host galaxy
NGC 6395 is a nearby grand Scd spiral galaxy in the Draco constellation, which has not been reported to host any SN before. Simbad reports an AB magnitude of 13 in the r band, but there is no SDSS spectrum of this galaxy and we have not been able to find any estimates of the metallicity of NGC 6395. We therefore made use of some of the spectroscopic observations of the SN to also measure some of the stronger emission lines from the nearby regions as a way to probe the metallicity of the gas close to the site of the explosion. The late Keck spectrum was taken with the slit oriented through the galaxy, and we extract the spectrum of an H II region located 8 north of the SN to probe the gas-phase metallicity there. We discuss the results in Sect. 3.4.

Light curves
The g-, r-and i-band LCs of our SN are displayed in Fig. 2. In the figure we have also included the most restrictive upper limits as triangles (2.5σ), while the arrows on top of the figure illustrate epochs of spectroscopy. The GP interpolation (at early phases) and linear fits (during nebular phases) are also shown.
It is clear that SN 2020cxd must have risen very fast in the first few days. The rise is 2.57 magnitudes (r band) in 3 days or less, with the slope thus larger than 0.85 mag per day. This fast rise of SN 2020cxd is comparable to that of the LL SN 2005cs 12 https://wiserep.weizmann.ac.il (0.90 mag/day in the R band, see Fig. 5 and Pastorello et al. 2009, their fig. 4).
After that initial rise follows a plateau phase of 100+ days, which firmly establishes SN 2020cxd as a Type IIP SN. The plateau is, however, not completely flat. In the g band there is an initial decline which levels off and then very slightly rises until the end of this phase. The r-band LC shows an initial undulation, but is then slowly rising by about 0.5 mag in 100 days, and the i band is following in that rise. The LC is well sampled with 58 observations (r band) over the plateau that allow us to discern such unusual LC structures.
The drop from the plateau is sharp and fast. The g-band LC declines by 4.98 mag in only 28 days. The drop in the r and i bands are slightly shallower, making SN 2020cxd redder at the very late-time phases, but the better sampled r-band LC also sharply plummets by 1.9 mag in only 6 days.
When comparing to the large compilation of Type II LCs from Anderson et al. (2014) and using their nomenclature 13 , we find that the Optically Thick Phase duration (OPTd) of SN 2020cxd is 118.3±3.0 d 14 , which is one of the longest plateau lengths compared to the Anderson sample, and the decline rate during plateau phase (s 2 ) is −0.73 mag per 100 days, which is also very rare 15 . Most of the SNe IIP have instead a clearly positive s 2 slope, i.e the plateau is slowly declining in luminosity, and SN 2020cxd is thus an unusually well sampled example of a brightening plateau.
In Fig. 5 we show the r/R-, g/V-and i/I-band LCs in absolute magnitudes together with the LCs of several other LL SNe II (see Table 1 for detailed information). Since most of the LL SNe II in the literature were observed in the Johnson photometric system, we compare LCs between r/R, g/V, and i/I, correspondingly. They show similarity with the canonical Type II plateau SN 1999em (green circles). The magnitudes in Fig. 5 are in the AB system 16 and have been corrected for distance modulus, MW extinction and host extinction if any, and are plotted versus rest frame days past estimated explosion epoch (see Table 1).
The figure demonstrates that SN 2020cxd is a clear member of this population of LL SNe, and in many respects a rather typical representative example for this class. The better photometric sampling compared to most other objects allows to clearly see the on-plateau evolution. The well determined explosion epoch and the sharp drop from the plateau also allows a more precise measure of the plateau duration (OPTd).
The colour evolution of SN 2020cxd and the other LL SNe is shown in Fig. 6. We plot (g − r)/(V − R) in the upper panel and (r − i)/(R − I) in the lower panel, both corrected for MW extinction. In doing this, no interpolation was used. Given the excellent LC sampling we used only data where the pass-band magnitudes were closer in time than 0.1 days. Comparison is made with the 13 With SN 2020cxd in the g band, whereas their sample was in V band. 14 As done by Anderson et al. (2014), we fit g-band data using a χ 2 minimizing procedure with a composed function of Gaussian, Fermi Dirac and straight line, following Olivares E. et al. (2010), from the estimated time of explosion. 15 As shown in Anderson et al. (2014, their fig. 2), the mean of the s 2 distribution is 1.27 mag per 100 days with a variance of 0.93 mag. The s 2 slope of SN 2020cxd is 2.2 σ off the distribution. However, there are a few SNe in their sample with similarly negative s 2 . We have been made aware, however, that such a negative plateau slope might be more common among the LL SNe II, see for example SN 1999br (Pastorello et al. 2004), SN 2009N (Takáts et al. 2014) and SNe 2013K and 2013am (Tomasella et al. 2018). 16 The Vega/AB magnitude conversion follows Blanton & Roweis (2007).   Table 4. The epoch (rest frame days since explosion) of the spectra is provided to the right. The spectra are shifted in flux by a constant for clarity. Late time spectra are scaled for comparison. The scaling factors are also listed to the right. Some spectra are also binned (shown as black lines), with a window size of 5 Å for LT, and 10 Å for NOT/Gemini/Keck, and their original spectra are shown in shaded grey. colour evolution for other LL SNe, which are known to typically be redder than ordinary SNe II. Compared to this sample, our SN is similar, which implies that the amount of additional host extinction is indeed likely small compared to the rest of this population, in accordance with our assumption of negligible host extinction. Our SN shows normal g − r colors at early phases, and evolves towards the redder part of the sample population after about a month. The r − i stays in the lower part of the LL SN II color space over the entire plateau phase.

Spectra
The log of the spectroscopic observations is provided in Table 4 and the sequence of spectra is shown in Fig. 4. Overall, the spectra of SN 2020cxd are typical for a Type II SN. In this section, we compare the spectra of SN 2020cxd with spectra from other LL Type IIP SNe at different phases.
Spectral comparisons between SN 2020cxd and other LL SNe IIP at photospheric phases are provided in Fig. 7. All spectra have been corrected for redshift and galactic reddening. Within ∼10 days after explosion, all of the comparison SNe show very Fig. 5 Absolute magnitudes of SN 2020cxd (black stars) together with the LCs of other LL SNe IIP, as well as for the normal Type II SN 1999em (green circles). We highlight SN 2013am (orange circles) which is the brightest within the LL SN II sample, and SNe 1999eu (purple circles) and 1999br (blue circles) which are the faintest. We also use a larger marker for SN 2018hwm (red hexagons), which was also monitored by ZTF in gri photometry. The other SNe are displayed with smaller and fainter markers. The LCs are plotted versus rest frame days past explosion. The radioactive cobalt decay rate is shown as a black line in the r-and i-band panels for comparison. The photometric data were obtained via the Open Supernova Catalog. Data reference for SN 1999em is Leonard et al. (2002), and for the LL SNe II these are listed in Table 1. blue continuum, with dominant line features of Balmer lines and Na I D 17 .
As SN 2020cxd enters the plateau phase, the spectrum becomes redder as the photospheric temperature decreases. At +93 days, the Ca ii near-infrared (NIR) triplet of SN 2020cxd is well developed, and we can also discern several metal lines, e.g. Fe ii, Sc ii, and Ba ii (Fig. 7). 17 Could also be blended with He i λ5876, (see e.g., Pastorello et al. 2006, their fig. 5).
From ∼120 days, SN 2020cxd enters the optically thin phase, and the luminosity decreases rapidly. The spectra displayed in Fig. 4 are all absolute calibrated against r-band photometry as interpolated from the LC, and the last three nebular spectra have been scaled by factors of 2, 10, and 15 respectively, for easier comparison. Figure 8 shows the scaled late-time spectra compared to other LL SNe II at similar phases. From 160 days, the spectra are dominated by Hα, [Ca ii] λλ7291, 7323 and the Ca ii NIR triplet.  Table 1. Fig. 7 Sequence of early photospheric spectra for SN 2020cxd, compared with spectra of other LL Type IIP SNe in grey. The epoch of the spectrum is provided to the right. Some identified lines (H, Na I D and Ca NIR triplet) are marked as vertical dashed lines. We also mark several metal lines as short vertical lines, which might be present in the last P60 spectrum. During the photospheric phase, we measured the velocities for SN 2020cxd using iraf 18 to fit a Gaussian to the minimum of the absorption lines of the corresponding P-Cygni profiles. Velocities and their uncertainties were estimated by a random sampling on the Gaussian fits to the minimum, by shifting the anchor continuum points a 1000 times within a region of ±5Å, using astropy.modeling.Gaussian1D for the iteration. Fits with bad chi-squared values were excluded.
The difference between the minimum of the best-fit Gaussian and the line location was translated to an expansion velocity. In the late nebular phase, we instead estimated the velocities from the emission line Full Width at Half Maximum (FWHM), which is the width measured at half level between the continuum and the peak of the line, and corrected for the instrumental resolutions obtained from the sky lines. These velocities are displayed in Fig. 9, where we also compare to other LL Type IIP SNe. The velocities for the comparison sample are taken from Pastorello et al. (2004) and Spiro et al. (2014). The time evolution of the velocities measured for Hα matches well with those of other LL Type IIP SNe at similar epochs, but also extend to earlier phases. The velocities of all these LL SNe, including SN 2020cxd, are very low, clearly lower than for the normal Type IIP SN 1999em. We notice that the Hα velocity of SN 2020cxd became very low around 250 days after the explosion, i.e. we measure an intrinsic FWHM of 478 km s −1 , in the last Keck spectrum (Fig. 8). This is among the narrowest lines ever measured for a SN IIP.
An important method for diagnosing the progenitor mass is through late-time spectral modeling. At the optically thin phase we are observing deeper into the progenitor structure (Jerkstrand et al. 2012). Figure 10 shows our last Keck spectrum taken during the optically thin phase, compared to modelling. The Keck spectrum is absolute calibrated against r-band interpolated photometry. In order to estimate the progenitor mass of SN 2020cxd, we have collected spectral synthesis models of SNe II (with 12, 15 and 19 M ZAMS masses) from the literature, i.e. from  Dessart et al. (2013). The models are selected to have similar phases to the SN observation, and are shifted to the distance of SN 2020cxd, as well as scaled by the nickel mass. As discussed, e.g. in Jerkstrand et al. (2014) and Müller-Bravo et al. (2020), the luminosity of some lines, like [O i] λ6300, 6364, scales relatively linearly with the ejected nickel mass, thus it is reasonable to scale the model fluxes by nickel mass for the comparison. To fit the models to the data, we implement a simple Monte Carlo approach that randomly scaled the models while χ 2 values were calculated to quantify the comparisons 19 . In Fig. 10  larger (by ∼ 30 − 40%). The Jerkstrand models do not reproduce the Ca ii NIR triplet, while the Lisakov 12 M model does a good job reproducing it, better than in the Dessart 15 M model. The Lisakov model over-predict the [Ca ii] λλ7291, 7323 by a factor of 2, whereas the other models agree better. In terms of χ 2 values, the scaled Dessart 15 M model gives the best match, slightly better than the other three models. Overall, we conclude that the progenitor of SN 2020cxd had a ZAMS mass less than 19 M . However, to properly distinguish between 12 M and 15 M models, a more detailed comparison is needed. We note that for lower-mass progenitors, a smaller line ratio of [O i]/[Ca ii] is also expected due to the lower O-core mass (Maeda et al. 2007;Jerkstrand et al. 2012;Fang & Maeda 2018). Using our absolute calibrated Keck spectrum, we measure the fluxes for [O i] λλ6300, 6364 and [Ca ii] λλ7291, 7323, and estimate a ratio of 0.82, which favours a lower mass progenitor.

Bolometric lightcurve
In order to estimate the total radiative output, we attempted to construct a bolometric LC. We perform Blackbody (BB) fitting every 2 days with the GP interpolated gri data of SN 2020cxd during its photospheric phase, i.e. up to ∼ 120 rest frame days. More details about the diluted BB fitting and the UV/IR contributions to the bolometric LC are discussed in Appendix A.
At nebular phases, the BB fits are not applicable and we use the bolometric correction (BC) approach to estimate the bolometric magnitudes of SN 2020cxd. In this work, we use the BC of SN 2018hwm to estimate the bolometric LC of SN 2020cxd. More information on this approach is provided in Appendix B.
The best fit bolometric luminosity LC of SN 2020cxd is thus composed of the early part from BB fitting and the tail inferred using the BC, and is shown as red stars in Fig. 11, compared to other LL Type IIP SNe discussed throughout the paper. As shown, SN 2020cxd is similar to the other LL SNe II, which are fainter than the famous SN 1987A. The slope of the LC tail matches well with the Cobalt decay rate. They were obtained via WISeREP. The models are scaled by nickel mass ratio. All model spectra are shifted to the SN distance by the inverse square of the distance. Their rest frame phases and scaling factors are shown in the legend.
We can now calculate the bolometric peak luminosity of SN 2020cxd, which is L max bol = 1.91 × 10 41 erg s −1 at 103 rest frame days and a total radiated energy over the first 200 rest frame days of E rad = 1.52 × 10 48 erg.
From the BB approximation we also obtain the temperature and the evolution of the BB radius. The obtained temperatures and radii are compared to those of other LL SNe in Fig. 12. SN 2020cxd follows a similar evolution in these parameters as the other LL SNe II in this sample, with both temperature and radius at the lower bound of the distribution. This is sensitive to the amount of host extinction and allowing for additional extinction would increase the temperature.
The ejected 56 Ni mass can be inferred by measuring the luminosity tail, which is powered by the decay of radioactive elements, i.e. 56 Co, synthesized in the explosion. Using, L = 1.45 × 10 43 exp(− t τ Co )( M Ni M ) erg s −1 from Nadyozhin (2003) implies that we would require 0.002 ± 0.001 M of 56 Ni to account for the tail luminosity (> 140 days). We note that the uncertainty in distance modulus affects this estimate with a system-atic uncertainty of 30%, but an additional host extinction of E(B − V) host = 0.25 mag would increase this by a factor two.
In order to estimate progenitor and explosion parameters from the photometry, we make use of a Monte Carlo code that was recently presented by Jäger et al. (2020), which fits semianalytic models to quasi-bolometric LCs (see Appendix C for more details). We employ the MCMC code to fit the semianalytic models to SN 2020cxd, the acceptable fits are displayed as light solid black lines in Fig. C.1. After marginalization, we have estimates with confidence intervals (1σ) for each of these parameters, i.e. SN 2020cxd has R 0 = 1.3 +0.4 −0.1 × 10 13 cm (∼ 187 R ), M e j = 9.5 +1.3 −1.0 M , E kin = 4.3 +0.9 −0.9 × 10 50 erg, v exp = 2747 +714 −235 km s −1 , E th = 1.5 +0.3 −0.5 × 10 50 erg, for its progenitor radius, ejecta mass, kinetic energy, expansion velocity, and thermal energy respectively. The nickel mass was simultaneously estimated as 0.003 ± 0.002 M . Reguitti et al. (2021) estimated the ejected mass of SN 2018hwm to 8 M . By assuming the mass of the compact remnant and a typical mass loss during the pre-SN stage, they estimated that the initial progenitor mass (zero age main sequence, Fig. 11 Bolometric luminosities of several LL SNe IIP and the famous SN 1987A. The luminosity of SN 2020cxd (shown as the red stars) was calculated after accounting for MW extinction and distance. The slope of the luminosity tail matches well with the radioactive decay of 56 Co.
ZAMS mass) of SN 2018hwm was in the range of 9.4−10.9 M . For SN 2020cxd, the ejected mass is estimated to be ∼ 9.5 M . With the same assumptions, its ZAMS mass was then between 10 and 14 M .
The initial bolometric decline (first 20 days) somewhat mimics those suggested to be powered by CSM in some hydrodynamical models (Morozova et al. 2017), but no evidence was found for flash spectroscopy features (e.g., Bruch et al. 2021) in our very early spectra (< 3 days past explosion). Overall, there is no evidence for CSM interaction from narrow spectral lines, UV brightness or from the shape of the LC. These consideration could disfavour a SAGB star as a potential progenitor star, since these are thought to have high mass-loss rates (compared to those of RSG stars of similar initial mass, Hiramatsu et al. 2021).
There has been extensive studies on the duration of the LC plateau for Type II SNe in the literature, exploring for example the effect of radioactivity on this phase. Although well established that the luminosity in this period derives from the diffusion as the photosphere recedes, it has also been suggested that the duration of the plateau is affected by radioactive input (Popov 1993;Kasen & Woosley 2009;Nakar et al. 2016). The most important parameter for the plateau duration in the investigation of Pumo et al. (2016) was the amount of stripping of the ejecta, where extensive stripping led to shorter plateau lengths. In this context, the very long duration in SN 2020cxd might suggest that the ejecta must have remained relatively intact. Kasen & Woosley (2009) derived scaling relations for the properties of SNe IIP, and conclude that the plateau duration is correlated with the explosion energy and the progenitor mass (where nickel in the ejecta tends to extend the plateau). However, given the low plateau luminosities of SN 2020cxd and SN 2018hwm these scaling relations would suggest a too low initial mass; this discrepancy was explained for SN 2018hwn by the extremely low explosion energy (Reguitti et al. 2021) 20 .
In Fig. 13 we compare the estimated radioactive 56 Ni mass of SN 2020cxd to those estimated for LL SNe II in the literature. As shown, the 56 Ni mass of SN 2020cxd is significantly lower than for normal CC SNe, and in the range of the LL SN class. We also examine the correlation between the V band (g band for our case) absolute magnitude and the ejected nickel mass. As shown, the linear relation of Hamuy (2003) can be confirmed also for the LL Type II SNe, extending the relation to the lower nickel mass region. There is evidence both from the photospheric velocities (Fig. 9), the very narrow nebular emission lines (Fig. 8) and the LC modeling that the explosion has very low energy. A priori, there need not be any correlation between the explosion energy, the luminosity at the plateau (Fig. 13) and the ejected mass of radioactive nickel, but this SN adds to the evidence that such correlations exist. Explosion models of neutrino-driven Fe CC indeed predict that in this ZAMS mass regime, where the iron core mass scales with ZAMS mass, lower explosion energies and thus ejecta velocities are expected from lower ZAMS mass stars, which also eject less mass (Barker et al. 2021). Fig. 12 Temperature and radius of several LL SNe IIP, including SN 2020cxd (shown as black stars). Upper panel: Evolution of their photospheric temperature estimated by fitting a black-body to the SED in the optically thick phase. Bottom panel: Evolution of their photospheric radius based on the same fits. We have followed Jäger et al. (2020) in the methodology, and the included SNe are also from that paper.

Host galaxy properties
The host galaxy of SN 2020cxd, NGC 6395, is classified as a Scd galaxy in NED, and reported to have a stellar mass log M * = 9.29 ± 0.10 M by Leroy et al. (2019). This is close to the center of the distribution of galaxy masses for SNe II hosts measured by Schulze et al. (2020) for the Palomar Transient Factory (PTF) sample (median log M * = 9.65 M ). Also many of the other LL SNe have been discovered in large spirals (Spiro et al. 2014).
No metallicity measurement for NGC 6495 is reported in the literature. As described in Sect. 2.6, we therefore extracted the spectrum of a nearby H II region from the day 242 Keck spectrum in order to obtain a metallicity estimate. We measure the fluxes of a range of emission lines (Table 5) and use the flux ratios to calculate several common strong-line metallicity indicators using the pyMCZ code (Bianco et al. 2016). The Balmer decrement indicates moderate extinction, E(B− V) = 0.36 ± 0.02 mag towards this H II region. We measure a slightly sub-solar metallicity in most indicators, for example we can adopt the N2 scale of the Pettini & Pagel (2004) calibration using the flux ratio between [N ii] λ6583 and Hα. We found that 12+log(O/H) = 8.50 ± 0.15. We also employed another metallicity diagnostic of Dopita et al. (2016)

Summary and Conclusions
We have presented SN 2020cxd, a LL SN IIP discovered by the ZTF which has one of the longest plateaus, the slowest ejecta and the coolest photospheres among such objects. The LC is wellsampled with a well determined explosion epoch and shows an initial sharp rise, and then a negative slope (s 2 ) during the plateau phase which peaks after about 100 days. The long-duration (118 days, OPTd) plateau length ends with a very sharp drop to a low-luminosity radioactive tail, implying 0.003 M of 56 Ni being ejected in the explosion. We find no evidence for a precursor at the site of the explosion, which is measured to have slightly subsolar metallicity.
We compare the nebular spectra to spectral synthesis models to constrain the progenitor mass through the [O i] λ6300, 6364 lines, and find relatively good agreements with progenitors of 12 and 15 M , excluding more massive progenitor scenarios. This argues against a high mass (25 − 40 M ) RSG (as a Fe CC SN) with large amounts of material from the Ni-rich region falling back onto the newly formed degenerate core (Zampieri et al. 2003).
We constructed the bolometric LC of SN 2020cxd, and estimated a total radiated energy of E rad = 9.71 × 10 49 erg over the first 200 days. The semi-analytic modelling suggests that SN 2020cxd originates from a progenitor which ejected about 9.5 M of ejecta. Although we can not conclusively determine the explosion mechanism (Electron capture or Fe CC), or the nature of the progenitor, all evidence are consistent with a Fe CC of a progenitor with a ZAMS mass on the order of 12 M .  Table 1), green squares are the SNe IIP from Hamuy (2003). SN 2020cxd is shown as a red data point. The grey dashed vertical lines represent the average nickel mass for different types of CC SNe from Anderson (2019). We perform linear fits on M V and logM Ni for the Hamuy 2003 sample and for the LL SN IIP sample. The best fits are shown as the green and black dashed lines, and give consistent relations.
This presentation provides an observational account of SN 2020cxd, which deserved a detailed investigation given its unique position in the larger systematic BTS survey. We have argued that this SN is most likely a result from a CC in a low ZAMS mass star. This is in accordance with the notion that the low-luminosity tail of the SN luminosity function is underexplored, and that understanding the SN progenitor IMF require further studies of large well-controlled samples in tandem with detailed studies of individual examples to connect them to the stellar progenitors. Table 1. Properties of the sample of LL SNe IIP. The sample was composed by Spiro et al. (2014) andMüller-Bravo et al. (2020), and was extended with 3 recent objects, i.e. SN-NGC 6412, SN 2018hwm andSN 2020cxd. t Reguitti et al. (2021); (21) this paper.