The Early Phases of Supernova 2020pni: Shock Ionization of the Nitrogen-enriched Circumstellar Material

We observed SN 2020pni with the Ultraviolet Optical Telescope (UVOT; Roming et al. 2005) on board the Neil Gehrels Swift Observatory (Gehrels et al. 2004) from 2020 July 16.8 until 2020 September 10.8 (δt = 1.0–56.9 days since first light). We performed aperture photometry using a 3''-radius circular region with uvotsource within HEAsoft v6.26, ¹⁵ following the standard guidelines from Brown et al. (2009). In order to remove contamination from the host galaxy, we employed images acquired at t ≈ 105 days after first light, assuming that the SN contribution is negligible. This is supported by visual inspection, in which we found no flux associated with SN 2020pni, although we cannot exclude some residuals given the relatively early phase of the SN, especially in the B and V bands. We subtracted the measured count rate at the location of the SN from the count rates in the SN images following the prescriptions of Brown et al. (2014). We detect UV emission from the earliest Swift epoch (δ t = 1.0 days; Figure 2) until δ t ≈ 32 days after first light. Subsequent nondetections in the u, w1, m2, and w2 bands indicate significant cooling of the photosphere.

Zoom In Zoom Out Reset image size

Additional griz-band imaging of SN 2020pni was obtained through the Young Supernova Experiment (YSE) sky survey (Jones et al. 2021) with the Pan-STARRS telescope (PS1; Kaiser et al. 2002) between 2020 July 17.3 and 2020 September 12.2 (δt = 1.5–58.4 days since first light). The YSE photometric pipeline is based on photpipe (Rest et al. 2005). Each image template was taken from stacked PS1 exposures, with most of the input data from the PS1 3π survey. All images and templates are resampled and astrometrically aligned to match a skycell in the PS1 sky tessellation. An image zero-point is determined by comparing point-spread function (PSF) photometry of the stars to updated stellar catalogs of PS1 observations (Chambers et al. 2017). The PS1 templates are convolved to match the nightly images, and the convolved templates are subtracted from the nightly images with HOTPANTS (Becker 2015). Finally, a flux-weighted centroid is found for each SN position, and PSF photometry is performed using "forced photometry": the centroid of the PSF is forced to be at the SN position. The nightly zero-point is applied to the photometry to determine the brightness of the SN for that epoch.

We obtained multiband NIR data for SN 2020pni on 2020 July 28 using the Multi-Object Spectrometer For Infra-Red Exploration (MOSFIRE; McLean et al. 2012) at Keck Observatory. We imaged the object using JHK_s filters. Standard flat-fielding has been applied, and the instrumental magnitudes were extracted through PSF photometry. We used the 2MASS catalog ¹⁶ (Skrutskie et al. 2006) for the flux calibration.

In addition to our own observations, we include g-band and r-band photometry from the ZTF public data stream (Bellm et al. 2019; Graham et al. 2019), which span from 2020 July 16.2 to 2020 September 2.2 (δt = 0.4–48.4 days since first light). All photometric observations are listed in Tables A1–A4.

The spectroscopic campaign of SN 2020pni started ∼1 day after discovery. Here we present the first 64 days of evolution. All of the spectra were reduced using standard techniques, which included correction for bias, overscan, and flat field. Spectra of comparison lamps and standard stars acquired during the same night and with the same instrumental setting have been used for the wavelength and flux calibrations, respectively. When possible, we further removed the telluric bands using standard stars. Given the various instruments employed, the data-reduction steps described above have been applied using several instrument-specific routines. Data from Keck using the LRIS, DEIMOS, and MOSFIRE instruments were processed with the PypeIt software package (Prochaska et al. 2020). We used standard IRAF ¹⁷ commands to extract the spectra from GMOS, Binospec, and ESI data. The SEDM spectrum was downloaded directly from the Transient Name Server ¹⁸ (TNS).

Spectra of SN 2020pni were obtained with the Kast spectrograph (Miller & Stone 1993) on the Shane 3 m telescope at Lick Observatory on 2020 July 18, 19, and 28, August 9 and 11, and September 7 (programs 2020A-S008 and 2020B-S001, PI Filippenko; program 2020A-S011, PI Foley). Observations were made with the 452/3306 and the 600/4310 grisms on the blue arm and the 300/7500 grating on the red arm, using the 20 slit aligned along the parallactic angle to minimize the effects of atmospheric dispersion (Filippenko 1982). Calibration observations (arc lamps, dome flats, and spectrophotometric standards) were performed on the same night. The data were reduced with standard IRAF/pyRAF and Python routines, including flat-fielding, determining the wavelength solution and small-scale wavelength corrections from night-sky lines, with flux calibration and telluric removal accomplished through the use of spectrophotometric standard stars. An additional spectrum was obtained on July 30 as part of the Lick Supernova Program (ToO) 2020A-S012 (PI Foley), using a different setup compared to the classical program (830/3460 grism, blue arm; 1200/5000 grating, red arm; 20 slit). The data were reduced in a similar manner.

We obtained spectra of SN 2020pni with the FLOYDS spectrograph on the Faulkes-N telescope at Haleakalā, Hawaii, as part of the Las Cumbres Observatory (LCO; Brown et al. 2013) network. The spectra were acquired on 2020 August 1, 4, 10, and 22 and September 7, as shown in Table A5. All spectra were obtained with the 16 slit under nearly photometric conditions, and the slit was aligned along the parallactic angle. Comparison-lamp and dome-flat exposures were obtained immediately before and after each observation. Following standard procedures in pyraf, we reduced all of these spectra using the FLOYDS pipeline (Valenti et al. 2014). ¹⁹ This included standard image reductions, aperture extraction, flux calibration, wavelength calibration, telluric removal, and order combination.

A spectrum was also taken with the Alhambra Faint Object Spectrograph and Camera (ALFOSC) on the Nordic Optical Telescope (NOT) at La Palma on 2020 July 25, using grism 4 and a 10 slit, aligned along the parallactic angle, and under clear observing conditions and good seeing. The spectrum was reduced with a custom pipeline running standard pyraf procedures.

We obtained an NIR spectrum of SN 2020pni in the YJHK bands on 2020 July 28 using MOSFIRE at Keck Observatory. The data were reduced using the MOSFIRE data-reduction pipeline, ²⁰ which performed flat-field correction, wavelength calibration using night-sky lines and arc-lamp observations, and spectral extraction. We then used xtellcor (Vacca et al. 2003) to perform flux calibration and telluric correction with observations of an A0V star HIP 71172.

A summary of all the telescopes, instruments, and configurations used for the spectroscopic observations of SN 2020pni is presented in Table A5. All of the spectra shown will be available at the Weizmann Interactive Supernova data REPository (WISeREP; ²¹ Yaron & Gal-Yam 2012).

We started observing SN 2020pni with Swift-XRT (Burrows et al. 2005) on 2020 July 16 until September 10 (δt ≈ 1.3–56 days since time of first light), for a total exposure time of 31.5 ks. We reduced the data following standard practice using HEASOFT v6.28 and the latest Swift calibration files. We find no evidence for X-ray emission at the location of the SN in the individual exposures and in the merged event file, for which we infer a 3σ count-rate upper limit of 8.0 × 10⁻⁴ counts s⁻¹ (0.3–10 keV). The neutral hydrogen column density in the direction of the SN is 1.40 × 10²⁰ cm⁻² (Kalberla et al. 2005). For a power-law spectrum F_ν ∝ ν⁻¹, the corresponding unabsorbed flux limit is F_x < 2.9 × 10⁻¹⁴ erg s⁻¹cm⁻², which is L_x < 1.9 × 10⁴⁰ erg s⁻¹ (0.3–10 keV). Individual segments of observations have a typical exposure time of ∼1.6 ks, which leads to flux limits F_x < 2.0 × 10⁻¹³ erg s⁻¹ cm⁻² (L_x < 1.2 ×10⁴¹ erg s⁻¹).

We emphasize that these limits are corrected only for the absorption component that originates in the Galaxy. However, the modeling of the optical spectra and early-time light curve (Section 4.3) indicates high densities in the local SN environment at distances <10¹⁵ cm, from which we estimate large intrinsic absorption corresponding to N_H−int ≳ 10²⁵ cm⁻² at the time of radiation breakout (assuming that a large fraction of the material is neutral). The lack of detected X-rays in SN 2020pni is thus most likely a consequence of the very large local absorption by the extended layer of CSM from which the H lines originate at early times. A later-time Swift-XRT observation was acquired on 2020 October 28 (δ t ≈ 105 days since time of first light, exposure time 3.5 ks), from which we derive L_x < 9 × 10⁴⁰ erg s⁻¹. The lack of detectable X-ray emission is consistent with the low-density, larger-scale environment inferred from the radio observations (Sections 3.4 and 4.3.3).

We observed SN 2020pni with the NSF's Karl G. Jansky Very Large Array (VLA) through our joint Fermi/VLA program SD1096/131096 (PI Margutti) on 2020 August 21.8 (δ t = 37.0 days after time of first light), 2020 November 21.5 (δ t = 128.7 days), and 2021 May 18.1 (δ t = 306.3 days). We carried out observations at a mean frequency of 10.0 GHz (X band) with a bandwidth of 4.096 GHz. The data were taken in standard phase-referencing mode, with 3C 286 as the bandpass and flux-density calibrator and 9C J1506+4239 and B3 1456+375 as the complex-gain calibrators. We calibrated the data using the VLA pipeline in the Common Astronomy Software Applications package (CASA; McMullin et al. 2007a) v5.6.2, with additional flagging. For imaging, we used Briggs weighting with a robust parameter of 1. No self-calibration was performed. The details of these observations are given in Table A7.

We find no evidence for radio emission at the SN location and infer a flux-density limit of F_ν < 19 μJy and F_ν < 12 μJy (3 × image rms) for the first and second epochs, respectively, corresponding to L_ν < 1.2 × 10²⁶ erg s⁻¹ Hz⁻¹ and L_ν < 0.8 ×10²⁶ erg s⁻¹ Hz⁻¹ at the distance of SN 2020pni. In the third epoch the VLA was in its D configuration and there was a significant contribution from the host galaxy at the location of SN 2020pni, but we found no evidence of a point source at the SN location. After fitting and subtracting the host emission in the image plane using PyBDSM (Python Blob Detection and Source Measurement; Mohan & Rafferty 2015), we find a 3σ upper limit of F_ν < 30 μJy (L_ν < 2 × 10²⁶ erg s⁻¹ Hz⁻¹).

The complete UV and optical light curves of SN 2020pni are presented in Figure 2. To estimate the time of escape of the first photons, we fit a three-parameter power-law function (e.g., m = a + bt^c ) to the early-time g and r data. From this, we infer a time of first light of MJD 59,045.8 ± 0.1. The error on the time is estimated based on the nonlinear least-squares fitting routine, and we point out that it likely underestimates the true (systematic) uncertainty of the measurement. We use this value throughout the paper. In order to estimate the peak absolute magnitudes, we fit a low-order polynomial to the SN 2020pni light curve. We obtain M_B = − 18.28 ± 0.10 mag at MJD 59,052.1 ± 0.2 and M_r =− 18.02 ± 0.20 mag at MJD 59,055.6 ± 0.2. Using the adopted time of first light, the rise time of SN 2020pni is t_r = 9.8 ± 0.3 days with respect to r-band maximum and t_B = 6.3 ± 0.2 days with respect to B-band maximum.

We then compare the r-band light curve of SN 2020pni to a sample of SNe II with and without shock-ionization features at early times. We include the sample of flash-spectroscopy objects from Khazov et al. (2016) and Bruch et al. (2021), in addition to the sample of "normal" SNe II from de Jaeger et al. (2019). We note that although SN 2013ej and SN 2014G were included in the latter collection, these objects showed flash-spectroscopy features in their early-time spectra (Valenti et al. 2014; Terreran et al. 2016), and they are therefore treated as such. Furthermore, we expand the shock-ionization sample with the addition of SN 1998S, SN 2016bkv, and SN 2017ahn (Leonard et al. 2000; Hosseinzadeh et al. 2018; Nakaoka et al. 2018; Tartaglia et al. 2021). The light curves of all the objects presented were retrieved from the OSC (https://sne.space; Guillochon et al. 2017), apart from those of Bruch et al. (2021), which were taken directly from their paper. For some objects the r band was not available, and we used the R band instead, without loss of generality.

We present the full comparison in the left panel of Figure 3. The great majority of flash-ionization objects populate the upper-end part of the plot, with SN 2020pni sitting right in the middle of the sample. Indeed, Khazov et al. (2016) claimed that the objects showing shock-ionization lines tended to be on average brighter than those that did not exhibit such features. On the other hand, Bruch et al. (2021) did not find the same trend from the analysis of their sample. The addition of the sample of "normal" SNe II from de Jaeger et al. (2019) seems to support the former study, although the presence of one extreme outlier complicates the scenario. In fact, SN 2016bkv (the lowest red line on the plot) is the most striking example that objects presenting shock-ionization features can indeed have luminosities well below the average of SNe II. There is some disagreement on the nature of this event, with claims for it to be an electron-capture SN rather than a core-collapse SN (Hosseinzadeh et al. 2018), but the presence of material around the progenitor star at the time of explosion should transcend the explosion mechanism responsible for the stellar demise.

Zoom In Zoom Out Reset image size

The apparently higher luminosity shown on average by flash objects could naturally be explained by the extra energy injection from the early interaction responsible for the shock ionization. However, selection criteria could also play a factor in the luminosity distribution. Brighter objects are typically of more interest to the community, and therefore a classification spectrum could be sought with more urgency. More low-luminosity SNe II could indeed exhibit flash-ionization features if observed sufficiently early. The size of the flash-spectroscopy sample is still quite small, preventing us from reaching a definitive conclusion on the matter.

We construct a pseudobolometric light curve of SN 2020pni using a combination of multicolor photometry from ZTF, PS1, and Swift observations. ²² For each epoch, luminosities are calculated through trapezoidal integration of SN flux in the ubvgri bands (3000–10000 Å). Uncertainties are estimated through a Monte Carlo simulation that includes 1000 realizations of the data. In time intervals without complete color information, we interpolate between light-curve data points using a low-order polynomial spline. The complete pseudobolometric light curve of SN 2020pni is presented in the right panel of Figure 3 for phases t < 60 days after explosion. We choose to estimate luminosities for SN 2020pni using a trapezoidal integration method rather than fitting a blackbody model so as to better compare SN 2020pni to other SNe II lacking UV or NIR photometric coverage.

We also construct a complete bolometric light curve of SN 2020pni by fitting the broadband photometry with a blackbody model that is dependent on radius and temperature. Each spectral energy distribution (SED) was generated from the combination of multicolor UV/optical photometry in the w2, m2, w1, u, b, v, g, r, and i bands (1500–9000 Å). In regions without complete color information, we extrapolate between light-curve data points using a low-order polynomial spline. This yields an initial radius of R_bo = (2.4 ± 0.14) × 10¹⁴ cm, as well as a peak temperature and luminosity of T = (2.5 ± 0.16) × 10⁴ K and L_bol = (2.7 ± 0.39) × 10⁴³ erg s⁻¹.

As shown in the left panel of Figure 3, we compare the r-band light-curve evolution of SN 2020pni to popular SNe II discovered within a few days of explosion, many of which have flash-ionized spectral features detected in their early-time spectra, such as SN 1998S (Leonard et al. 2000; Fassia et al. 2001; Shivvers et al. 2015), SN 2013fs (Yaron et al. 2017), SN 2014G (Terreran et al. 2016), and SN 2017ahn (Tartaglia et al. 2021). These objects were selected for reference to SN 2020pni because of their high-cadence optical sampling, complete pseudobolometric light curves, and distinct shock-ionization features in the early-time spectra that then disappeared at later phases. With respect to this sample, the peak r-band absolute magnitude of SN 2020pni is more luminous than that of SN 2013ej, SN 2013fs, and SN 2017ahn but less luminous than that of SN 1998S and SN 2014G. Its r-band rise time and post-maximum decline rate are most similar to those of SN 2013fs. The photometric evolution of SN 2020pni is comparable to that of SN 2017ahn out to t < 30 days post-peak, until the plateau in the light curve of SN 2017ahn falls off at ∼40 days.

In the right panel of Figure 3, we compare the pseudobolometric light-curve evolution of SNe II discovered within a few days of explosion to that of SN 2020pni. We find that SN 2020pni has a lower overall luminosity than SN 1998S and a pseudobolometric luminosity comparable to or higher than the other presented SNe. Like the r-band light curve, the overall pseudobolometric evolution of SN 2020pni is most comparable to that of SN 2013fs at phases t < 60 days after explosion. Out to phases t < 40 days post-explosion, its pseudobolometric rise time, light-curve decline rate, and peak luminosity are nearly identical to those of SN 2014G and SN 2017ahn.

The complete optical spectral evolution of SN 2020pni is presented in Figure 4, and a log of the presented spectra is reported in Table A6. We also include the ZTF classification spectrum that was taken 1 hr prior to our first spectrum (Bruch et al. 2020). We started our spectroscopic campaign less than 2 days after our estimated time of first light, and we kept monitoring SN 2020pni for ∼60 days, until constrained by the Sun. The early-time spectra show a blue continuum, with prominent narrow lines of H, He ii, C iv, N iii, and N iv. After 1 week, most of these features disappear, and only the H lines remain, which at this epoch are still narrow but with a P Cygni profile. At a phase of 17 days from first light, the spectrum seems completely featureless, while broad lines (v ≈ 10,000 km s⁻¹) begin to appear a few days later. More typical features of SNe II start to shape the spectrum (see, e.g., Gutiérrez et al. 2017), including H, Fe ii, Na i, Sc ii, Ba ii, Ti ii, and Ca ii. By the time SN 2020pni became Sun constrained, it was still in the photospheric phase, and no nebular lines were visible yet.

Zoom In Zoom Out Reset image size

At ∼12 days after first light, we acquired a Keck I+MOSFIRE NIR spectrum, in coordination with an optical spectrum with Shane+Kast (see Figure 5). Although the H Balmer series is clearly visible at this phase in the optical spectrum, we do not detect any Paschen or Brackett lines in the NIR. However, we identify a P Cygni profile of He i λ10830 and possibly He I λ20581.

Zoom In Zoom Out Reset image size

In Figure 6 we compare the early phases of SN 2020pni with those of other well-observed flash-spectroscopy events—SN 1998S (Leonard et al. 2000; Fassia et al. 2001; Shivvers et al. 2015), SN 2013fs (Yaron et al. 2017), SN 2014G (Terreran et al. 2016), and SN 2017ahn (Tartaglia et al. 2021). The spectra were first normalized by the underlying continuum ²³ and then scaled by an arbitrary factor for better display. A diversity in line intensities and timescales is clearly evident, and the presence of different elements with different ionization levels could be linked to variations in CSM composition as well. Prominent, highly ionized oxygen lines are present in the first spectrum of SN 2013fs, although no oxygen is detected in the first spectrum of SN 2020pni. These lines disappear very rapidly in SN 2013fs, so we cannot exclude that these lines were present in SN 2020pni at an earlier phase.

Zoom In Zoom Out Reset image size

The presence of a relatively prominent C iv λ5803 line in SN 2020pni, which is missing in the spectra of SN 2013fs, hints at a carbon-rich environment for SN 2020pni (see Figure 6). This carbon line seems to also be quite prominent in SN 2014G, and possibly detected in SN 1998S and SN 2017ahn. The spectroscopic similarities between SN 1998S and SN 2020pni are remarkable (see spectra at ∼4 days). Indeed, among the presented sample, SN 1998S is the only object other than SN 2020pni that still shows narrow lines beyond a week after explosion. SN 2013fs and SN 2014G present featureless spectra at ∼9 days after explosion, while SN 2020pni and SN 1998S show narrow H lines with P Cygni profiles, both with a low-velocity component in absorption (see also Section 4.2.1). This indicates that at this phase the photosphere is still in the slow-moving CSM, suggesting a more radially extended, high-density environment compared to that of SN 2013fs and SN 2014G.

We compare the later spectroscopic evolution of SN 2020pni with the same sample of objects. In Figure 7, we display spectra at ∼2 months after explosion. All of the objects present a very similar pattern, with the spectrum being dominated by P Cygni profiles of hydrogen, sodium, iron, and other metal lines like scandium and barium. Focusing on the P Cygni profile of Hα, SN 2017ahn and SN 1998S exhibit very shallow to no absorption, SN 2013fs shows a relatively deep trough, and SN 2020pni and SN 2014G are intermediate. We find a certain range of velocities of ejected material in the sample. The minimum of the Hα absorption feature of SN 2020pni sits at ∼6500 km s⁻¹, while that of SN 2014G is at ∼7600 km s⁻¹. Indeed, focusing on the position of the minimum of the absorption of the features in common, we notice that SN 2020pni has the slowest material. From the minimum of the Fe i λ5169 feature, we measure a velocity of ∼3700 km s⁻¹, which we can use as a proxy for the photospheric velocity (Hamuy et al. 2001). The slower material displayed by SN 2020pni manifests as a higher number of discernible features in the spectra of SN 2020pni, as a consequence of less severe line blending. This can be appreciated especially from the spectrum ∼64 days after explosion, where several Fe ii, Sc ii, and Ba ii multiplets can be identified (see Figure 7).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Focusing on the He i and Na i absorption blend around 5800 Å, it is clear from Figure 7 how this feature looks broad, and possibly with multiple components, in SN 2020pni, SN 2013fs, and SN 2014G. The same feature appears narrower in SN 1998S and SN 2013fs. It is tempting to attribute its reddest component to a resolved He i line. In SN 2020pni this identification would correspond to a velocity of the helium material of ∼ 5000 km s⁻¹, which is considerably slower than what is shown by the Hα minimum. Ba ii has a multiplet at 5854, 6142, and 6497 Å. Assuming then that the reddest absorption in the feature at 5800 Å is barium, we obtain a velocity of ∼ 4000 km s⁻¹, while from the feature at 6100 Å we measure ∼ 3000 km s⁻¹. Therefore, the Ba ii association also seems in conflict with other identifications of the same ion. It is likely that a combination of helium and barium is responsible for the broader feature at 5800 Å observed in some SNe II, such as SN 2020pni and SN 2013fs.

4.2.1. Line Evolution

At 6 days after first light, a very broad, although shallow, absorption starts to appear on the blue side of Hα (see Figure 8). This feature extends to ∼ 8000 km s⁻¹, while the emission component remains narrow. The broad component is clearly associated with the SN ejecta. After a few days, the absorption component develops a more pronounced dip. From the position of the minimum, we measure a velocity in close agreement with what is determined from the FWHM of the emission feature. The narrow Hα now has the full appearance of a P Cygni profile. The broad component is still present at these phases. This morphology, with a narrow P Cygni profile superimposed on a broader absorption, was also observed in SN 1998S (Leonard et al. 2000; Fassia et al. 2001; Shivvers et al. 2015; Dessart et al. 2016). At 14 days after first light, the emission component disappears completely, initially leaving only the narrow absorption, and then a featureless continuum. At 20 days a broad emission component starts to emerge with FWHM ≈ 8000 km s⁻¹, accompanied by a shallow absorption with a minimum at similar velocities, forming the classical P Cygni profile of typical evolved SNe II.

We now focus on the complex evolution of the Hα profile by studying the velocity evolution of each component described above. At t < 5 days, in order to reproduce both the narrow core and the broad wings of the pure emission feature, we used two Lorentzian components. Beyond this phase and until t < 13 days, we used single Lorentzian profiles to reproduce the narrow emission and absorption, while the broad absorption is well reproduced by a flat, boxy profile, with boundaries defined by a sigmoidal function. In Figure 9 we show an example of the modeling we performed during these phases. When the broad P Cygni profile is fully formed (t > 20 days), we instead used only two Gaussians to reproduce the line profile. We then used the FWHM as a proxy for the velocity of the emission component and the minimum of the absorption component as a proxy for the bulk velocity of the narrow absorption component. We also kept track of the middle point of the red-most sigmoidal boundary of the boxy profile (orange squares in Figure 9).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The evolution in time of the velocity of these features is shown in the bottom panel of Figure 9. At early phases, the narrow emission shows a clear increase in velocity, reaches a peak ∼4 days after first light, and then starts to decrease. We remark that this is not a spectral resolution effect, as the lines are fully resolved in all of the spectra (apart from the first epoch with the SEDM). One possible conclusion could be that the ejecta inside the CSM are accelerating the inner material (Moriya et al. 2011). However, the emission originates from the CSM in front of the shock (this is unshocked CSM). The shock then has yet to reach this part, so it could not be responsible for any acceleration at this phase.

The increasing velocity with time likely maps a velocity gradient of the CSM at larger radii from the explosion, hinting at complex CSM. Given that we see these narrow lines only during the early phases of evolution of SN 2020pni, it is fair to assume that this CSM was created by recent mass loss, and it is possible that the progenitor lost material having different velocities with time. In particular, we observe that the progenitor lost material with larger velocities at earlier times and then smaller velocities as it approached the explosion time. We point out, though, that this does not naturally reflect a variable mass-loss rate during this phase, as the velocity of the material is not necessarily linked to the amount of the material lost by the progenitor of SN 2020pni.

Finally, from Figure 9, one could also argue that the boxy profile extension evolves with time, as the orange square seems to increase in velocity from ∼8000 to ∼ 9500 km s⁻¹. However, given the modeling we performed, this parameter does not directly link to a specific physical property of the explosion. Although this parameter could be seen as a proxy for the maximum velocity of the ejecta, the signal-to-noise ratio (S/N) of the spectrum, the depth of the absorption function, the shallowness of the transition from the continuum to the floor of the boxy profile, and the fit to the continuum level all play a role in inferring this value. Therefore, the observed increase in velocity is probably not as significant as the figure might suggest, as the uncertainties are also likely underestimated. However, further studies of other objects showing this extended absorption are encouraged, and possibly a more meaningful physical quantity could be inferred from a larger sample.

4.3.1. CSM Properties at r ≤ 10¹⁵ cm from Early Light-curve Modeling

A number of observational features suggest that the shock's radiation is breaking out of a compact shell of dense CSM extending out to a radius R_w . Specifically, these include the fast rise to maximum optical light over a timescale of t_r ≈ 2 days, the bright and hot UV emission reaching a color temperature T ≥ 20,000 K (e.g., Section 4.1 and Figure 2), the rapid fading of shock-ionized spectral features, and the narrow P Cygni profiles (Figures 8 and 4; Section 4.3.2) by ∼15 days after first light. We expect R_w ≈ 10¹⁵ cm, comparable to the inferred best-fitting blackbody radius when the shock-ionized spectral features become subdominant (e.g., Section 4.1).

We employ the formalism by Chevalier & Irwin (2011; see Waxman & Katz 2017 for a recent review) to model the onset of the emission that breaks out from the thick shell of CSM under the reasonable assumption R_d < R_w , where R_d is the radius of the contact discontinuity at the diffusion time t_d (i.e., the radius where the diffusion of radiation becomes important). This assumption is motivated by the persistence of shock-ionized spectral features well beyond the time of bolometric peak. Indeed, the low-velocity P Cygni profiles of Hα are detectable until at least 15 days after first light. Under these circumstances, the escape of the radiation is delayed with respect to the onset of the explosion on a timescale ∼ t_d , which is set by the time necessary for the radiation to reach an optical depth τ_w ≈ c/v_sh, where v_sh is the shock velocity. Radiation is also released on the diffusion timescale, leading to a bolometric rise time t_rise ≈ t_d , which implies that the explosion started ∼ t_rise (i.e., at most a few days) before the estimated time of first light (Table 1). This result is consistent with the time of explosion estimated from the emergence of spectral features with v ≈ 8000 km s⁻¹ at t ≈ 8 days since first light and the measured radius of the photosphere at this time, which implies that the time of first light is delayed from the time of explosion by at most 1–2 days.

Following Chevalier & Irwin (2011) and using the solutions by Margutti et al. (2014), we find that the observed t_rise, radiated energy at breakout E_rad ≈ 0.5 × 10⁴⁹ erg, and breakout radius R_bo ≈ 2 × 10¹⁴ cm constrain the wind mass-loss rate to $\dot{M}\approx 0.01\,{M}_{\odot }\,{\mathrm{yr}}^{-1}$ for a wind velocity v_w = 200 km s⁻¹, similar to the FWHM of the narrow hydrogen component in the first optical spectrum. For these parameters the wind-shell mass is M_w ≈ 0.02 M_☉ enclosed within R_w ≈ 10¹⁵ cm. The wind mass within the breakout radius is M_w ( ≤ R_bo) ≈ 0.005 M_☉. In this model, after shock breakout, continued interaction with the wind material supports a luminosity ≳10⁴³ erg s⁻¹ for a few days, consistent with the observations. We note that these parameter values should be treated as order-of-magnitude estimates given the likely complexity of the SN environment and some simplifying assumptions inherent to our analytical modeling approach.

Zoom In Zoom Out Reset image size

4.3.2. CSM Properties at r ≤ 10¹⁵ cm from Spectral Modeling

4.3.3. Properties of the Larger-scale r > 10¹⁵ cm Environment from Radio Observations

We infer the density properties of the larger-scale environment at distances of (1–3) × 10¹⁶ cm using the radio nondetection in Section 3.4. In the context of synchrotron emission from the explosion's forward shock, and self-consistently accounting for both synchrotron self-absorption (SSA) and free–free absorption (FFA) (e.g., Chevalier 1998; Weiler et al. 2002), the radio nondetections at δt = 37.3–307 days enable constraints on the $\dot{M}$ versus v_shock parameter space shown in Figure 12. We followed the prescriptions from Chevalier (1998) to compute the SSA emission as a function of the radio-spectrum observables (see also the equations reported by Terreran et al. 2019), and we accounted for external FFA using the formalism for the optical depth to free–free radiation by Weiler et al. (2002). The shock velocity is self-consistently calculated using the self-similar solutions by Chevalier (1982). For this calculation, we have assumed a wind-like density profile of the CSM (ρ_CSM ∝ r⁻²) and a plasma temperature of ∼ 10⁴ K. We describe this process in further detail in Appendix A. We find that for a typical shock velocity of ∼0.1 c the lack of detectable radio emission is consistent with either a low-density medium, with density corresponding to $\dot{M}\lt {10}^{-5}\mbox{--}{10}^{-6}\,{M}_{\odot }\,{\mathrm{yr}}^{-1}$, or a higher-density medium, with $\dot{M}\gt 5\times {10}^{-4}\,{M}_{\odot }\,{\mathrm{yr}}^{-1}$, that would absorb the emission (Figure 12). These $\dot{M}$ values reported are for a wind velocity v_w = 200 km s⁻¹. The range of allowed mass-loss rates in the lower-density case is sensitive to the choice of shock microphysical parameter values _e and _B , which represent the fraction of post-shock thermal energy in relativistic electrons and magnetic fields, respectively.

In Figure 13 we summarize our inferences concerning the CSM that surrounded the progenitor of SN 2020pni at the time of explosion, similar to what Yaron et al. (2017) did for SN 2013fs. We marked in blue the densities we inferred from the CMFGEN modeling of the Keck II+DEIMOS optical spectrum taken ∼1.5 days after first light (see Section 4.3.2). The radio upper limits from the previous paragraph translate to an excluded region at a higher distance, marked in orange in the figure. The shock-ionization features are present in the spectra until ∼12 days after first light. The position of the shock at this phase (assuming a typical velocity of 0.1c) is marked in the figure with a vertical dashed magenta line. Considering the lack of narrow features after this epoch, it is fair to assume that the CSM was less dense beyond a radius of (2–4) × 10¹⁵ cm. The radio analysis suggested that both a high-density configuration and a low-density configuration were possible; however, the lack of narrow lines at later phases disfavors the high-density scenario. In Figure 13 we suggest a possible configuration of the CSM.

Zoom In Zoom Out Reset image size

In this section we compare the inferred physical properties of SN 2020pni with the growing population of flash-spectroscopy events. For this purpose we use the sample by Boian & Groh (2020), as they performed spectral modeling using the same code, CMFGEN. Boian & Groh (2020) used a grid of models for their analysis and three different surface abundance scenarios: solar-like, CNO-processed, and He-rich (see Boian & Groh 2020, for more details). In Figure 14 we show the inferred relations between the mass-loss rate $\dot{M}$, density factor $D\equiv \dot{M}/(4\pi {v}_{\mathrm{wind}}$), temperature T at the CSM inner boundary (where the optical depth to electron scattering is τ ≈ 10), and SN luminosity ${L}_{\mathrm{SN}}$. The mass-loss rate estimates rely on a measure of the wind velocity, which is not always possible to obtain, owing to the resolution of the classification spectra. Plotting the density factor instead of the mass loss has the advantage of not relying on an assumed wind velocity; however, it is less trivial to link this quantity with the general characteristics of the progenitor star. The electron temperature can be used as a proxy for the ionization level of the CSM, with higher temperatures indicating a higher ionization. SN 2020pni sits right in the middle of the distribution of all parameters. Indeed, the initial spectrum did not show features related to highly ionized ions such as O iv, O v, or N v the way SN 2013fs did (Figure 6). However, the temperature (and therefore the ionization level) is strongly influenced by the epoch at which the classification spectrum was acquired. SN 2013fs and PTF10gva are among the objects with the earliest spectra (t < 1 day), while the first spectrum of SN 2020pni was acquired later.

Zoom In Zoom Out Reset image size

In flash-ionization events, the ejecta lose kinetic energy as they are slowed down by the CSM. A denser medium allows for a more efficient conversion of kinetic energy into radiation, which can power a more luminous continuum and contribute to a more luminous SN at early phases. This energy conversion should follow D ∝ L^3/4, which is in rough agreement with what is shown in the right panel of Figure 14. Also, in this case SN 2020pni sits roughly in the middle of the distribution, with mass loss from the progenitor and therefore density parameter slightly lower than the average.

One conclusion by Boian & Groh (2020) was that the overall mass-loss estimates from the spectral modeling do not significantly differ from those inferred for SNe II (from optical observations), showing narrow lines for hundreds of days. This suggests that the difference between these more extreme interacting SNe and the shock-ionization events is not to be found in the density of the CSM surrounding the progenitor stars. Boian & Groh (2020) suggest, therefore, that the radial extension of the thick CSM could play a major role in shaping the evolution of the SN.

The persistence of the narrow lines in time t_line (i.e., for how long the shock-ionization lines are visible in the spectra) is an observable that was not taken into consideration by the analysis of Boian & Groh (2020). The hydrogen recombination timescale (for pure H composition) can be approximated by t_rec = (α n)⁻¹, where α ≈ (2 − 4) × 10⁻¹³ cm³ s⁻¹ for T = 10⁴ K (Osterbrock & Ferland 2006). Given the previously inferred density of ∼ 1.5 × 10⁸ particles cm⁻³, we find t_rec ≈ 2.2 ×10⁴ s, or 0.26 days. This is considerably smaller than t_line, implying the need for a source of ionizing photons that is active well beyond the time of shock breakout. Therefore, the CSM is likely kept ionized by the prolonged interaction of the SN ejecta with the inner boundary of the CSM. We can thus link the persistence of the narrow shock-ionization lines with some physical properties of the explosion itself: (i) an explosion that launches faster shocks would have the ejecta ram through the thick CSM earlier, and (ii) a radially less extended CSM would be engulfed at earlier times by the ejecta, and therefore the narrow lines would disappear sooner. This second point is particularly important, as a less extended CSM would have been created by the progenitor star at a time closer to the explosion. Hence, the persistence of these lines could be linked to the mass-loss history right before explosion.

We estimate the persistence of the narrow lines for all of the objects from the Boian & Groh (2020) sample that also had a published spectroscopic sequence beyond the classification spectrum. We identified the last spectrum showing narrow lines from the ionized CSM and the first spectrum in which the lines were absent, taking the midway point as the time of disappearance. For the majority of the targets, especially those with long-lived narrow lines, hydrogen is usually the only species left at the time of narrow-line disappearance, while the He and CNO lines disappear at earlier phases. These measurements are presented in Figure 15, where we look for correlations with the other physical properties studied above. The sample size does not allow us to come to any definitive conclusion, but at this time we do not see any clear correlations among the plotted quantities. However, an important conclusion to take is that the persistence of the narrow lines does not seem to depend on the composition of the CSM, considering that no clear distinction is evident among solar-like, CNO-processed, and He-rich events. Assuming that the CSM composition could be used as a proxy for the composition of the progenitor, this suggests that the composition probably does not play a major role in shaping the CSM.

Zoom In Zoom Out Reset image size

If we refer to point (ii) mentioned above, this conclusion could lead us to suggest that any physical mechanism responsible for the observed late-time mass loss must operate under different physical conditions and stellar progenitors. We caution that the sample studied here is quite limited. A significantly larger sample size than what is currently available might reveal correlations among the parameters of Figures 14 and 15 and has the potential to constrain the nature of the mass-loss mechanism at work.

So far we have shown that SN 2020pni was a particularly luminous (see Figure 3) hydrogen-rich SN. We found that the CSM surrounding the progenitor star was He-rich, suggesting that the progenitor had shed a large part of its envelope at the time of explosion, or alternatively that mixing was particularly high in the progenitor star. Assuming that the progenitor was a single star, the abundance of CN-processed material at the surface of the star, and therefore in the CSM immediately surrounding it, tends to increase with the ZAMS mass of the progenitor (Ekström et al. 2012). Therefore, the strong nitrogen and carbon lines observed at early phases would favor a stellar mass in the higher end of the progenitors of SNe II. A comparison with stellar evolution models computed with the Geneva code (Ekström et al. 2012) suggests an RSG or a yellow hypergiant progenitor star for SN 2020pni, with a ZAMS mass between 15 and 25 M_⊙ (Groh et al. 2013). These studies have also shown that luminous blue variables (LBVs) could be the potential progenitors of some of the more H-depleted SNe II. Both direct observations and models of stellar evolution revealed that LBVs are enriched in helium and nitrogen, while being depleted in carbon and oxygen (e.g., Meynet et al. 1994; Smith et al. 1994; Crowther 1997; Najarro et al. 1997), just as is shown by SN 2020pni. In addition, in Section 4.2.1 we demonstrated that the CSM surrounding the progenitor of SN 2020pni was not uniform. This is something that is also often observed in LBV nebulae (e.g., Smith 2006, 2014). Considering the further evolution of SN 2020pni as a relatively normal SN ii, we favor an RSG as its most probable progenitor. However, the fact that some characteristics of the CSM resemble LBV-like winds is remarkable. This once again highlights the lack of a full understanding of the mass-loss processes in RSG stars, especially during the last phases of their lives.

On a final note, binarity could have played a major role in shaping the CSM surrounding the progenitor of SN 2020pni and the other shock-ionization events. Considering the high percentage of progenitors of SNe II that are expected to be in binary systems (e.g., Zapartas et al. 2019), it is possible that the surface composition of the progenitor star could have been heavily modified by the interaction of a stellar companion or even by a merger event. However, one key aspect that emerges from Section 5.1 is that the mass loss observed in shock-ionization events appears to have been sustained for a very brief period of time (years), when compared to the whole lives of the stars themselves (millions of years). This necessarily means that the physical mechanism responsible for this mass loss is somehow linked to the actual end of the star's life. In other words, the mechanism "knows" that the end is coming. Interaction with a binary companion would not know about the advanced stage of evolution of the companion star. Considering that 30% of core-collapse SNe observed within 2 days since explosion show shock-ionization features (Bruch et al. 2021), the chances that so many objects exhibit enhanced mass loss induced by the binary interaction with a companion right before the death of the progenitor star are remarkably low. Therefore, although we cannot exclude the presence of a binary companion in the system of these stars, we disfavor the idea that the physical mechanism responsible for the appearance of shock-ionization features in young core-collapse SNe is directly related to the presence of a companion star.