Disappearance of a massive star in the Andromeda Galaxy due to formation of a black hole

TL;DR

The paper investigates the disappearance of a massive star in M31 (M31-2014-DS1) as evidence for black hole formation via a failed supernova. By combining NEOWISE mid-infrared monitoring with archival and follow-up optical/near-infrared data, the authors characterize a dust-enshrouded remnant that fades over ~1000 days, with no accompanying bright optical supernova. They construct two progenitor scenarios using MESA: a hydrogen-poor envelope that yields minimal mass ejection ($\sim 10^{-1}\,M_\odot$) and a hydrogen-rich envelope that ejects more mass ($\sim 0.3\,M_\odot$), both leading to fallback onto a stellar-mass BH; the early luminosity can reach a super-Eddington plateau (near $(0.3$–$0.5)\,L_{\rm Edd}$) before declining as accretion falls below the Eddington rate. The results are consistent with a failed SN for M31-2014-DS1, and they draw parallels with a prior event in NGC 6946-BH1, reinforcing the interpretation that some massive stars die without bright SNe while forming BHs, with dust obscuration concealing X-ray emission. The work provides empirical constraints on fallback physics, dust formation, and the rates of failed SNe in nearby galaxies, advancing our understanding of BH formation channels in massive stars.

Abstract

When a massive star reaches the end of its lifetime, its core collapses and releases neutrinos that drive a shock into the outer layers (stellar envelope). A sufficiently strong shock ejects the envelope, producing a supernova. If the shock fails to eject it, the envelope is predicted to fall back onto the collapsing core, producing a stellar-mass black hole (BH) and causing the star to disappear. We report observations of M31-2014-DS1, a hydrogen-depleted supergiant in the Andromeda Galaxy. In 2014 it brightened in the mid-infrared. From 2017 to 2022 it faded by factors of $\gtrsim10^4$ in optical light, becoming undetectable, and $\gtrsim10$ in total light. We interpret these observations, and those of a previous event in NGC 6946, as evidence for failed supernovae forming stellar-mass BHs.

Figures (17)

• Figure 1: Location and disappearance of M31-2014-DS1. (A) Optical color composite image of the discovery field taken from the Panoramic Survey Telescope and Rapid Response System (PanSTARRS/PS1 survey; suppmat). The yellow dashed square indicates the region shown in panels B-D (where white indicates brighter pixels) and the yellow cross-hair marks the position of the star. (B) NEOWISEMainzer2014 MIR image taken in 2017, (C) NEOWISE image in 2010 and (D) the difference between them. The other panels show zoomed-in images of the star (as indicated by the scale, with black indicating brighter pixels) taken in the labeled years: (E-H) optical HST images; (I) near-infrared HST image; (J) near-infrared Keck image.
• Figure 2: Brightness of M31-2014-DS1 as a function of time. The measured flux in millijansky (mJy, left axis) and equivalent luminosity at the distance of the Andromeda Galaxy (right axis), both on logarithmic scales, are plotted as a function of time in modified Julian date (MJD, lower axis) and Gregorian year (upper axis). Archival data are from the Palomar Transient Factory (PTF), NEOWISE, PS1, Gaia and Zwicky Transient Facility (ZTF) surveys; also shown are follow-up photometric data from the MMT Observatory suppmat. Error bars are $1\sigma$ confidence (smaller than the symbol size for the NEOWISE and pre-2014 data); hollow symbols with downwards arrows are $5\sigma$ upper limits. For the Gaia photometry, we show the raw measurements as light symbols, while the dark symbols are the averages within $45$ d of the closest epoch of NEOWISE data. The luminosity is monochromatic $\lambda\,F_\lambda$, where $\lambda$ is the wavelength and $F_\lambda$ is the flux density, scaled to a wavelength of $2$ µ m. The MIR data are shown on a linear scale in Figure \ref{['fig:linearlc']}, and all photometric data are provided (with references provided in suppmat) in the online data repository.
• Figure 3: Spectral energy distribution and physical properties of the progenitor. (A) The ultraviolet to MIR SED of M31-2014-DS1 (solid and hollow circles, in units of the monochromatic luminosity $\lambda\,L_{\lambda}$, where $L_\lambda$ is the luminosity density) observed with HST and SST from 2005-2012. Lines are the best-fitting DUSTY model (parameters are listed in Table \ref{['tab:dusty_fits']}). Total flux (black solid line), dust emission (orange dot-dashed), dust-scattered stellar emission (red-dashed) and dust-attenuated stellar emission (blue dotted). Error bars are $1\sigma$ confidence (smaller than the symbol sizes), and downward arrows are $5\sigma$ upper limits. (B) The observed luminosity and effective temperature of M31-2014-DS1 (green star, from panel A) compared to theoretical single star evolutionary tracks at different initial masses (colored dashed lines) Choi2016. The observed progenitor of NGC 6946-BH1 (brown plus symbol) and those of known hydrogen-rich SNe Smartt2015 (type indicated in legend) are shown. Error bars are $1\sigma$ confidence. Colored circles are our stellar evolution model for the progenitor of M31-2014-DS1 suppmat, with the residual hydrogen envelope mass indicated by the color bar. The model at the time of core collapse is indicated for M31-2014-DS1 (blue star) and for NGC 6946-BH1 (red plus symbol).
• Figure 4: Evolution of the stellar and dust properties of M31-2014-DS1. Temporal evolution of the dust shell and stellar photosphere inferred from the SED model fitting: (A) the dust temperature ($T_d$; black, left axis) and inner shell radius ($R_in$; red, right axis) and (B) the effective temperature (black, left axis) and inferred stellar radius (red, right axis). These are plotted as functions of time since the start of the MIR brightening in 2014 (MJD 56674.19; Figure \ref{['fig:optirlc']} and suppmat). The shaded regions are the corresponding model parameters for the progenitor from 2005-2012. Other parameters are shown in Figure \ref{['fig:dust_param_evol']}. The SED of the remnant of M31-2014-DS1 (C) in $2022$ to $2023$ [photometry in black circles and spectrum in gray lines (shown on a linear scale in Figure \ref{['fig:nires']})]. The best-fitting DUSTY model with line styles as in Figure \ref{['fig:prog']}, and parameters are listed in Table \ref{['tab:dusty_fits']}. The corresponding progenitor photometry (gray circles) and SED model (dashed grey line) are shown. Error bars are $1\sigma$ confidence and downward arrows are $5\sigma$ upper limits.
• Figure 5: Constraints on the mass ejection and stellar envelope fallback in M31-2014-DS1. (A) The luminosity and duration of transients powered by energy injection into hydrogen-rich envelopes, for ejecta with different velocities ($v_{\rm ej}$) and kinetic energies ($E_{\rm ej}$). The diamonds, circles, and squares indicate ejecta with velocities of $\approx 1\times$, $\approx 5\times$, and $\approx 25\times$ the progenitor's escape velocity respectively, with colors indicating ejecta mass ($m_{\rm ej}$; right color bar). The cross-hatched area shows the phase space ruled out by the optical photometry suppmat. Also shown are the luminosity and transient duration for the H-poor (stars) and H-rich (plus symbols) models as a function of ejecta kinetic energy (bottom color bar). (B) The bolometric fading of M31-2014-DS1 compared to models of mass fallback rates for different explosion energies. The hydrogen-poor and hydrogen-rich progenitor models both assume $5$% accretion radiative efficiency. For comparison, the dotted pink line shows $100$% accretion in the H-poor case and the dashed orange line shows an example H-rich case. Black solid circles are parameters inferred from the SED models, and hollow circles are estimates using a bolometric correction to MIR luminosity suppmat. Error bars are $1\sigma$ confidence. The gray shaded region shows the progenitor luminosity, and the black dashed line shows the Eddington luminosity ($L_{\rm Edd}$) for a $5$ M$_\odot$ BH. The right axis shows the corresponding total mass fallback rate ($\dot{M}_{\rm fall, total}$) from the models.