Obtaining spectroscopic observations of the progenitors of core-collapse supernovae is often unfeasible, due to an inherent lack of knowledge as to what stars will go supernova and when they will explode. In this Letter, we present photometric and spectroscopic observations of the progenitor activity of SN 2023fyq before the He-rich progenitor explodes as a Type Ibn supernova. The progenitor of SN 2023fyq shows an exponential rise in flux prior to core-collapse. Complex He I emission line features are observed in the progenitor spectra, with a P-Cygni like profile, as well as an evolving broad base with velocities on the order of 10,000  km s${}^{-1}$. The luminosity and evolution of SN 2023fyq is consistent with a Type Ibn, reaching a peak $r$-band magnitude of $-18.8$ mag, although there is some uncertainty in the distance to the host, NGC 4388, located in the Virgo cluster. We present additional evidence of asymmetric He-rich material being present prior to the explosion of SN 2023fyq, as well as after, suggesting this material has survived the ejecta-CSM interaction. Broad [O I], C I, and the Ca II triplet lines are observed at late phases, confirming that SN 2023fyq was a genuine supernova, rather than a non-terminal interacting transient. SN 2023fyq provides insight into the final moments of a massive star’s life, highlighting that the progenitor is likely highly unstable before core-collapse.

Spectroscopic observations of progenitor activity 100 days before a Type Ibn supernova

1 Introduction

Massive stars ($\gtrsim$ 8 – 10 M${}_{\odot}$) will eventually end their lives as core-collapse supernovae (CCSNe; Woosley & Weaver, 1995; Heger et al., 2003; Janka, 2012; Crowther, 2012). With the increasing capabilities of photometric transient surveys (Bellm, 2014; Chambers et al., 2016; Tonry et al., 2018) a growing sample of supernova (SN) progenitors have been observed experiencing outbursts in the weeks – years prior to their ultimate core-collapse (Foley et al., 2007; Pastorello et al., 2007; Mattila et al., 2008; Ofek et al., 2013; Fraser et al., 2013; Margutti et al., 2014; Ofek et al., 2014; Strotjohann et al., 2021; Fransson et al., 2022; Jacobson-Galán et al., 2022; Smith et al., 2022; Hiramatsu et al., 2023). These precursor events are difficult to explain, and are a relatively new area of stellar astrophysics (Dessart et al., 2010; Quataert & Shiode, 2012; Leung et al., 2021; Strotjohann et al., 2021; Tsang et al., 2022), however are often invoked to explain complex circumstellar material (CSM) shortly before collapse (Matsumoto & Metzger, 2022; Hiramatsu et al., 2023). Precursor activity is likely accompanied by violent mass ejections by a currently unclear mechanism. Potential scenarios include binary interaction (Yoon et al., 2010; Smith, 2011; Kashi et al., 2013; Yoon et al., 2017; Zapartas et al., 2021; Tsuna et al., 2024), outbursts from Luminous Blue Variables (LBVs; Humphreys & Davidson, 1994; Smith et al., 2010; Kilpatrick et al., 2018; Smith et al., 2022), and instabilities towards the end of a massive stars life (Woosley et al., 2007; Smith & Arnett, 2014; Müller et al., 2016; Fuller & Ro, 2018; Leung et al., 2021; Wu & Fuller, 2022).

Figure 1: Composite $giy$ color image of the host galaxy of SN 2023fyq from the Pan-STARRS survey, with the transient environment given in the top right inset. An $r$-band image of SN 2023fyq from the NOT/ALFOSC from 2023-08-16 is given in the top left inset.

This Letter presents spectra (photometric) observations of the progenitor of a Type Ibn SN several months (years) before core-collapse, as well as SN 2023fyq itself. The precursor activity of SN 2023fyq was discovered by the Zwicky Transient Facility (ZTF; Graham et al., 2019; Bellm et al., 2019; Masci et al., 2019; Dekany et al., 2020) at R.A. = 12:25:45.874, Dec. = $+$12:39:48.87 (J2000) on 2023-04-17 at $g=19.51$ mag (AB) (De, 2023). Spectroscopic observations of the progenitor were obtained as part of the Census of the Local Universe (CLU) survey (De et al., 2020) as it satisfied the selection criteria; low-luminosity transient in a nearby galaxy. SN 2023fyq was later classified by the Nordic optical telescope Unbiased Transient Survey 2 (NUTS2) as a Type Ib-pec SN (Valerin et al., 2023) on 2023-07-25, around 5 days pre-peak. SN 2023fyq occurred around 10${}^{\prime\prime}$ from the core of the active spiral galaxy NGC 4388 (Damas-Segovia et al., 2016) in the Virgo Cluster, as presented in Fig. 1. For consistency with the literature, we assume a distance of 17.2 Mpc to NGC 4388 (Lianou et al., 2019; Böhringer et al., 1997), and a distance modulus ($\mu$) of 31.2 mag, giving SN 2023fyq a peak $r$-band magnitude of $-18.8$ mag. However the distance is uncertain and may be as high as 25 Mpc (e.g. Ekholm et al., 2000), meaning SN 2023fyq may be $\sim 1.5$ mag more luminous. We correct for foreground Milky Way (MW) extinction using ${\rm R_{V}}=3.1$, ${\rm E(B-V)_{MW}}$ = 0.029 mag (Schlafly & Finkbeiner, 2011), with the extinction law given by Cardelli et al. (1989). Measurements of host emission lines indicates substantial host attenuation, as inferred from the Balmer (H$\alpha$/H$\beta$) decrement, of ${\rm E(B-V)_{host}\approx 0.4\pm 0.1}$ mag (Osterbrock, 1989). We correct for this extinction in a similar fashion to Milky Way extinction mentioned above using an ${\rm R_{V}}=2$. We adopt a redshift of $z=0.0084$ and the rest-frame phase is reported with respect to the $r$-band peak magnitude on 2023-07-30 ($t_{peak}$ = 60155.1).

2 Observations

2.1 Photometric observations

Science-ready images taken as part of the ZTF survey were obtained in $gri$ from the NASA/IPAC Infrared Science Archive (IPAC) service. Template subtracted photometry was performed using the AutoPhOT pipeline (Brennan & Fraser, 2022). Photometry from the ATLAS forced-photometry server (Tonry et al., 2018; Smith et al., 2020; Shingles et al., 2021) were obtained in $c$ and $o$ bands. Several epochs of optical photometry were obtained with the Alhambra Faint Object Spectrograph and Camera (ALFOSC) on the 2.56 m Nordic Optical Telescope (NOT) in $gri$, the Liverpool Telescope (LT) with the optical imaging component of Infrared-Optical imager: Optical (IO:O) in $griz$, and the Spectral Energy Distribution Machine (SEDM; Blagorodnova et al., 2018; Rigault et al., 2019) on the Palomar 60-inch (P60) telescope in $gri$. We also obtained $gri$ images from a 32-inch Ritchey-Chretien telescope (RC32) at Post Observatory in Mayhill, New Mexico. UV photometry were acquired using the Ultra-violet Optical Telescope (UVOT) onboard the Neil Gehrels Swift Observatory (Gehrels et al., 2004; Roming et al., 2005). We reduced the images using the Swift HEAsoft V. 6.26.1. toolset, as detailed in Irani et al. (2023). A UV template image was constructed using archival observations prior to MJD 59100 and after the SN declined to the host level. We then removed the local host-galaxy contribution by subtracting the SN site flux from the fluxes of the individual epochs.

2.2 Spectroscopic observations

Follow-up spectra were obtained using NOT/ALFOSC, the Low Resolution Imaging Spectrometer (LRIS; Oke et al., 1995) on the 10m Keck telescope, the Double Spectrograph (DBSP) on the Palomar 200-inch telescope (P200), the Kast spectrograph at the Lick Observatory, the Gemini Multi-Object Spectrographs (GMOS) mounted on the Gemini North 8m telescope (Hook et al., 2004), and P60/SEDM (Blagorodnova et al., 2018). Spectra were reduced in a standard manner, using LPipe (Perley, 2019), DBSP_DRP (Mandigo-Stoba et al., 2022) and PypeIt (Prochaska et al., 2020a, b, 2020), for Keck/LRIS, P200/DBSP, and NOT/ALFOSC, pysedm (Rigault et al., 2019; Kim et al., 2022) for SEDM spectra, the UCSC spectral pipeline(Siebert et al., 2019) for the Kast/Lick spectrum, and the DRAGONS (Labrie et al., 2019) for our GEMINI/GMOS spectrum. Observations using the P60, P200,and Keck telescopes were coordinated using the FRITZ data platform (van der Walt et al., 2019; Coughlin et al., 2023).

NGC 4388 was observed with the Multi Unit Spectroscopic Explorer (MUSE) at the 8.2 m ESO Very Large Telescope on 12 March 2022 (ID. 108.229J, PI. Venturi). Observations were reduced using the MUSE pipeline v2.8.7 and the ESO workflow engine ESOReflex (Freudling et al., 2013; Weilbacher et al., 2020). We extracted all spectra with the MUSE Python Data Analysis Framework (MPDAF) version 3.6 (Bacon et al., 2016). The spectrum of the SN site was extracted with a circular aperture. The aperture radius of $0.5^{\prime\prime}$ translates to a physical scale of 89.5 pc at the projected distance of SN 2023fyq, comparable to the largest giant molecular clouds in the Milky Way.

Figure 2: Spectral observations of SN 2023fyq and its progenitor. Each spectrum has been normalised and offset for clarity, as well as corrected for redshift and extinction. We mark regions of telluric contamination using the $\oplus$ symbol and grey vertical bands.

Spectroscopic observations for the progenitor activity and SN 2023fyq are presented in Fig. 2.

3 Results and Discussion

3.1 The photometric evolution of SN 2023fyq and its progenitor activity

Figure 3: Multiband light curve of SN 2023fyq, with each band offset for clarity. Vertical grey bars denote epochs of spectral observations. Absolute magnitudes, based on a distance of 17.2 Mpc, are shown on the right Y-axis, uncorrected for Milky Way or host extinction. Note the broken X-axis, highlighting the long-lived progenitor activity.

Figure 3 gives the photometric evolution of SN 2023fyq, highlighting the pre-SN evolution which lasts around 5 years, as well as the SN explosion itself. Figure 4 provides the bolometric evolution constructed from the host-subtracted photometry presented in Fig. 3, and calibrated using the extinction and distance given in Sect. 1.

Figure 4: The blackbody luminosity of SN 2023fyq and its progenitor are depicted in the figure. Star markers indicate blackbody fits to the synthetic photometry from spectra presented in Fig. 2. Circle markers are blackbody fits to epochs of photometry which contain at least $u$-band. Similarly, blackbody radius and temperature are provided in the middle and lower panels, respectively.

The blackbody luminosity (${\rm L_{BB}}$) of the progenitor is well-fitted by an accelerating rise (i.e., exponential), followed by a fast rise after the progenitor has likely exploded. The blackbody radius (${\rm R_{BB}}$) of the progenitor is seen to increase linearly at approximately 200  km s${}^{-1}$, peaking at ${\rm 1.75\times 10^{14}\leavevmode\nobreak\ cm}$, after which a sharp increase is observed. This radius ($\sim$ 2400 ${\rm R_{\odot}}$) is unlikely to represent a photosphere of the progenitor star, but perhaps signifies an extended/inflated envelope, due to the final stages of a binary merger event (Ivanova et al., 2013; Irani et al., 2023), or a confined CSM, as this value is consistent with the shock breakout radius reported for SN 2023ixf (Zimmerman et al., 2023). Assuming this second increase is a result of core-collapse, we extrapolated backwards to ${\rm R_{BB}\approx 0cm}$ and find an explosion time (${t_{\text{exp}}}$) of -13.9 days. Due to our sparse photometry around the explosion epoch, we are uncertain whether SN 2023fyq follows the evolution depicted by the analytical fits in Fig. 4, or rather a multi-component rise (e.g., SN 2021qqp, SN 2009ip; Hiramatsu et al., 2023). We adopt a rise time for SN 2023fyq of approximately 14 days, which is consistent with what is seen in other Type Ibn SNe (Hosseinzadeh et al., 2017).

From SN 2023fyq, we measure a max luminosity of ${\rm 4.2\times 10^{43}\leavevmode\nobreak\ erg\leavevmode\nobreak\ s^{-1}}$, ${\rm L(t=t_{peak})\approx 2.2\times 10^{41}\leavevmode\nobreak\ erg\leavevmode\nobreak\ s^{-1}}$, and a total radiated energy of ${\rm 6.0\times 10^{49}\leavevmode\nobreak\ erg}$. These value is consistent with predicted activity from binary interactions (Tsuna et al., 2024), and similar to progenitor activity in other SNe (Strotjohann et al., 2021). We measure the radiated energy of the pre-SN evolution as ${\rm L(t<t_{exp})\approx 6.6\times 10^{47}\leavevmode\nobreak\ erg}$. Due to the lack of UV and IR coverage of the progenitor and SN 2023fyq, these values are taken as lower limits. After the SN has occurred, a dramatic increase in ${\rm L_{BB}}$ and ${\rm T_{BB}}$ is observed, with temperatures reaching $\sim 30\leavevmode\nobreak\ \text{kK}$. This brightness increase is mainly seen in the bluer bands (see Fig. 3), and is consistent with shock breakout (e.g., SN 2008D; Chevalier & Fransson, 2008; Modjaz et al., 2009).

Assuming an efficient conversion of kinetic energy to radiation, we get ${\rm M_{CSM}\approx\frac{2E_{rad}}{\epsilon v^{2}}\approx 0.05}$ M${}_{\odot}$, assuming 50% energy conversion, and a velocity of 1700  km s${}^{-1}$ (taken from the observed P-Cygni minimum in pre-SN spectra). This material would be very optically thick at $10^{14}$ cm, which is consistent with the observed thermalised spectra. For SN 2023fyq, a CSM breakout with $\sim$0.05 M${}_{\odot}$ would last a few days, with a peak luminosity of a few $10^{43}$ erg s${}^{-1}$ followed by shock emergence and cooling, consistent with Fig. 4. In this scenario (Khatami & Kasen, 2023, e.g. Edge-breakout; Heavy CSM scenerio ), we expect ${\rm M_{ejecta}>M_{CSM}}$, and given the higher mass and lower velocity of the shocked CSM material, the cooling emission diffuses out on a longer timescale (Piro, 2015). We estimate the diffusion time (taken to be ${\rm t_{d}\approx\sqrt{\frac{\kappa M_{CSM}}{vc}}}$; Khatami & Kasen, 2019) as $\sim$12 days, assuming material expanding at 12000  km s${}^{-1}$ from Fig. 4, roughly consistent with the adopted explosion time. Assuming our final $R_{BB}$ measurement before the adopted ${\rm t_{exp}}$ indicates the radius of this CSM material, we find ${\rm Log(R/R_{\odot})}\approx 3.4$, which overall agrees with models from Khatami & Kasen (2023).

We measure the progenitor to have ${\rm\log(T_{eff}/K)}\approx 4.1\pm 0.1$ and ${\rm\log(L/L_{\odot})}\approx 7.6\pm 0.4$ from host-subtracted photometry. These values are likely super-Eddington and are reminiscent of the Giant Eruption of Eta Carina, which, possibly through multi-star interactions, ejected $\sim$10 M${}_{\odot}$ of material but did not unbind the star (Smith, 2011; Prieto et al., 2014). This suggests a multi star scenario, although without detailed late-time stellar evolutionary modeling, the mechanism powering the pre-SN evolution is unclear.

3.2 Spectroscopic evolution

Figure 5: Evolution of the He I $\lambda$5876 emission profile. Pre-SN spectra show complex P-Cygni like profiles. Soon after the SN has occurred, an additional absorption component is seen in the $-$6d Lick/KAST spectrum at around $-$5000  km s${}^{-1}$. At late times, the P-Cygni absorption is no longer observed, but rather a double-peaked emission profile consisting of host emission and redshifted He I emission at similar velocities as seen in the $-$93d Keck/LRIS spectrum. Spectra marked with an asterisk have been re-binned for clarity, with the raw spectrum given in grey. We manually mask out telluric lines and artefacts from the sky subtractions.

Observations of CCSN progenitors are rare, with the only previous examples, to our knowledge, being SN 2009ip (Smith et al., 2010; Foley et al., 2011; Pastorello et al., 2013) and SN 2015bh (Elias-Rosa et al., 2016; Thöne et al., 2017). In agreement with stellar evolutionary theory, the progenitor of the Type Ibn SN 2023fyq display features expected for He-rich stars (Crowther, 2007; Yoon et al., 2010). Figure 5 shows the evolution of He I $\lambda$5876, during the pre- and post- SN epochs, as well as the post-peak appearance. Similar to the H$\alpha$ profile seen for SN 2009ip in 2009 (Pastorello et al., 2013), SN 2023fyq shows a complex P-Cygni like profile for He I $\lambda$5876 (and similarly for other He I lines) with an absorption centered at $-1700$  km s${}^{-1}$ and extending out to $-2500$  km s${}^{-1}$. Perhaps similarly to H$\alpha$ in late time spectra of SN 2009ip (Fraser et al., 2015), the He I in SN 2023fyq shows a redshifted emission peak, centered at +1400  km s${}^{-1}$, which is present during the slow pre-SN rise, as well as post explosion. The velocities of these profiles do no evolve significantly during the pre-SN stage, although this may be a result of poor S/N.

A spectrum of the explosion site from VLT/MUSE taken $\sim$2 years prior to SN 2023fyq is given in Fig. 2. We investigate if any potential flux from the progenitor activity is present in this spectrum by comparing it to a similar spectrum extracted 2.4${}^{\prime\prime}$. We note a tentative broad excess at the position of He I $\lambda\lambda$ 6678,7065, possibly due to progenitor emission.

Due to the sodium laser guide star, the wavelength range covering He I $\lambda$5876 is masked, preventing any investigation of potential host contamination of He I $\lambda$5876 for SN 2023fyq. We suspect this redshifted component is originating from He I $\lambda$5876 as it is seen in several He I lines in the pre-SN spectra, as well as a similar emission component at similar velocities seen in the post-SN spectra. This would mean that the asymmetric material responsible for this emission was not destroyed in the SN explosion. SN ejecta interacting with asymmetric CSM has been used to explain irregular emission line profiles (e.g. Leloudas et al., 2015; Andrews & Smith, 2018; Pursiainen et al., 2022; Hosseinzadeh et al., 2022; Smith et al., 2023) and bumpy light curves (Nyholm et al., 2017; Woosley, 2018; Wang et al., 2022), and SN 2023fyq provides the first clear spectroscopic evidence of asymmetric structure prior to core-collapse.

A broad component is observed for He I $\lambda$5876 profile in the $-37$d and $-17$d spectra, with wings extending to $\pm 13,000$  km s${}^{-1}$ and a full width half maximum (FWHM) around 14,000  km s${}^{-1}$. Such velocities are normally associated with SN ejecta, although have also been observed in non-terminal eruptions (Pastorello et al., 2013; Smith et al., 2018). To investigate if this broad component evolves, we measure the flux in a small wavelength bin around 5800 Å in the continuum-subtracted flux-calibrated spectra. An order of magnitude increase in flux from $-103$ to $-10$ days is observed. This confirms that the broad component is rapidly evolving in strength prior to the explosion of SN 2023fyq rather than not being detectable in earlier spectra due to low signal-to-noise.

While these features may represent fast moving material, an alternative explanation could be electron scattering. Some mechanism(s) cause the progenitor to be surrounded by a dense CSM (such as acoustic waves or a merger; Smith, 2011; Yoon et al., 2017; Fuller & Ro, 2018; Chevalier, 2012; Fransson et al., 2022; Tsuna et al., 2024), and may lead to shock dissipation and emission of radiation in the optically thick CSM. We model the $-17$ d spectrum using the electron scattering model from Brennan et al. (2023), assuming ${\rm V_{\text{shock}}=1700}$  km s${}^{-1}$, ${\rho\propto r^{-2}}$, ${\rm\tau_{e}\approx 18}$, ${\rm T_{e}=2\times 10^{4}}$ K, and ${\rm R_{2}=1.75\times 10^{14}}$ cm. As shown in Fig. 5, the model reproduces the wings of He I $\lambda$5876 well, suggesting that under certain conditions, material moving at $>10^{4}$  km s${}^{-1}$ may not be required to produce the broad features during the pre-SN phase. Assuming a random walk, the diffusion time is $t_{d}=\frac{R^{2}}{\lambda_{\text{mfp}}c}\approx 4$ days. Our models do not produce the P-Cygni-like profile or offset emission and are poorly fit in the red wing, highlighting the complex geometry likely responsible for these features.

The post-SN He I $\lambda$5876 appears double peaked with similar shaped profiles also seen in the ${{Ca\textsc{II}}\leavevmode\nobreak\ \lambda\lambda 8498,8542,8662}$ near-infrared triplet. Double peaked emission lines have been observed in Type Ibn SNe (e.g. SN 2018bcc Karamehmetoglu et al., 2021), as well as Type IIn SNe (Brennan et al., 2022a; Hiramatsu et al., 2023), and are attributed to asymmetric SN ejecta or CSM or both (Andrews et al., 2019; Pursiainen et al., 2022).

A long standing question for SN 2009ip-like transients is whether the transient is indeed a SN or rather a “SN impostor” (Van Dyk et al., 2000; Mauerhan et al., 2013; Fraser et al., 2013; Brennan et al., 2022b; Smith et al., 2022). As shown in Figs. 2 and 5, SN 2023fyq shows broad emission from ${\rm[{O\textsc{I}}]\leavevmode\nobreak\ \lambda\lambda 6300,6364}$ at nebular phases, showing that SN 2023fyq is a genuine CCSN (Jerkstrand et al., 2014; Kuncarayakti et al., 2015), whose progenitor underwent some form of instabilities shortly before core-collapse. Late time followup of SN 2023fyq make reveal significant reddening due to possible dust formation. We note the appearance of O I and C I emission lines, meaning the environment of SN 2023fyq is conducive for amorphous and graphite dust grains.

Events like SN 2023fyq and SN 2009ip suggest that certain massive stars experience eruptive activity preceding core-collapse. However, several nearby transients lack notable eruptive activity (Jacobson-Galán et al., 2022; Ransome et al., 2023; Dong et al., 2023), implying a diverse mechanisms for pre-supernova emissions in the final stages before core-collapse.

4 Conclusion

In this Letter, we present an unprecedented dataset of progenitor activity of a Type Ibn SN almost 150 days before core-collapse occurs. Pre-SN spectra reveal a complex evolving He I profile. These observations of SN 2023fyq and the final moments of the progenitor, highlight that the progenitors to CCSNe can undergo some extreme instabilities shortly before their final demise.

Progenitor analysis typically occurs after the star has been destroyed, by searching through archival images, and measuring the photometric properties of the assumed progenitor. Although this area of transient astronomy is in its infancy, the repercussions of detecting precursor activity are immense, highlighting that the progenitor is not in a equilibrium state, and may not be represented well by standard stellar evolutionary models. SN 2023fyq and similar transients, highlight that the pre-SN appearance of the progenitor is non-trivial, and without careful consideration, may produce misleading results in supernova progenitor studies.

Data availability

The spectroscopic and photometric data underlying this article are available in the Weizmann Interactive Supernova Data Repository (WISeREP; Yaron & Gal-Yam, 2012).