Draft version September 6, 2023

Typeset using LATEX twocolumn style in AASTeX63

Collapse of rotating massive stars leading to black hole formation and energetic supernovae
Sho Fujibayashi,1 Yuichiro Sekiguchi,2, 3 Masaru Shibata,1, 2 and Shinya Wanajo1
1 Max-Planck-Institut für Gravitationsphysik (Albert-Einstein-Institut), Am Mühlenberg 1, D-14476 Potsdam-Golm, Germany
2 Center for Gravitational Physics and Quantum Information, Yukawa Institute for Theoretical Physics, Kyoto University, Kyoto,
606-8502, Japan
3 Department of Physics, Toho University, Funabashi, Chiba 274-8510, Japan
arXiv:2212.03958v2 [astro-ph.HE] 5 Sep 2023

(Received; Revised September 6, 2023; Accepted)

ABSTRACT
We explore a possible explosion scenario resulting from core collapses of rotating massive stars that
leave a black hole by performing radiation-viscous-hydrodynamics simulations in numerical relativity.
We take moderately and rapidly rotating compact pre-collapse stellar models with the zero-age-main-
sequence mass of 9M⊙ and 20M⊙ based on stellar-evolution calculations as the initial conditions. We
find that the viscous heating in the disk formed around the central black hole powers an outflow.
The moderately rotating models predict a small ejecta mass of order 0.1M⊙ and explosion energy of
≲ 1051 erg. Due to the small ejecta mass, these models may predict a short-timescale transient with
the rise time 3–5 d. It can lead to a bright (∼ 1044 erg/s) transient like superluminous supernovae in
the presence of a dense massive circum-stellar medium. For hypothetically rapidly rotating models
that have a high mass infall rate onto the disk, the explosion energy is ≳ 3 × 1051 erg, which is
comparable to or larger than that of typical stripped-envelope supernovae, indicating that a fraction of
such supernovae may be explosions powered by black-hole accretion disks. The explosion energy is still
increasing at the end of the simulations with a rate of > 1050 erg/s, and thus, it may reach ∼ 1052 erg.
The nucleosynthesis calculation shows that the mass of 56 Ni amounts to ≳ 0.1M⊙ , which, together
with the high explosion energy, may satisfy the required amount for broad-lined type Ic supernovae.
Irrespective of the models, the lowest value of the electron fraction of the ejecta is ≳ 0.4; thus, the
synthesis of heavy r-process elements is not found in our models.

Keywords: stars: neutron; general–hydrodynamics–neutrinos–relativistic processes

1. INTRODUCTION If the collapsing star has a very compact core, the

At the final evolution stage of massive stars, their iron proto-neutron star is likely to collapse into a black hole
cores become gravitationally unstable and collapse. Af- with no successful shock revival due to the strong ram
ter the core bounce due to the formation of a proto- pressure by the matter infall (e.g., O’Connor & Ott
neutron star, a shock wave is formed on its surface and 2011; but see Burrows et al. 2019). Even in this case,
propagates outward. The shock wave then stalls pri- there is a possibility of the explosion if a massive disk
marily because of the photodissociation of heavy (iron- is formed around the black hole due to the rotation of
group) nuclei in the infalling matter swept by the shock. the progenitor star. In the disk, the magnetorotational
In the standard neutrino-driven delayed-explosion sce- instability (MRI) could amplify the magnetic field, de-
nario, the shock is revived by the heating of neutrinos veloping a turbulent state inside the disk (e.g., Balbus &
emitted by the proto-neutron star (e.g., Janka et al. Hawley 1991). The turbulent motion then induces an ef-
2012). fective viscosity in the disk, which governs the evolution
of the disk through the angular momentum transport

Corresponding author: Sho Fujibayashi

[email protected]
2 Fujibayashi et al.

and heating. The viscous heating rate is estimated by associated with long-duration gamma-ray bursts (e.g.,
MacFadyen & Woosley 1999, Pruet et al. 2003, Na-
Lvis ∼ νΩ2 Mdisk
gataki et al. 2007, Surman et al. 2006, and Hayakawa
  2
αvis cs & Maeda 2018), for which the promising central en-
≈ 7 × 1051 erg/s
0.03 109 cm/s gine of the gamma-ray bursts is likely to be a spinning
 1/2  −3/2   black hole penetrated by a strong magnetic field. In this
MBH rdisk Mdisk
× , (1) scenario, however, we additionally need the supernova
10M⊙ 107 cm 0.1M⊙
component. Eisenberg et al. (2022) suggested, based
where MBH is the mass of the black hole, Mdisk and on their simulations, that the observationally inferred
rdisk are the mass and typical radius of the disk, cs velocity distribution of the supernova-component is not
is the sound speed, and Ω is the local angular veloc- likely reproduced only by the relativistic jet. This indi-
ity. Equation (1) is a radially integrated form of viscous cates that there has to be another energy source to drive
heating rate per unit area. See, e.g., Frank et al. (2002). the supernova component in addition to the relativistic
Here, we assumed a Keplerian rotation and the Shakura- jet accounting for the gamma-ray burst. Kohri et al.
Sunyaev-type alpha-viscosity model for the kinetic vis- (2005) applied a disk explosion scenario to normal su-
cous coefficient ν = αvis cs H (Shakura & Sunyaev 1973) pernovae by analytically solving a stationary neutrino-
with the disk scale height H = cs /Ω. αvis is the so- cooled accretion-disk model and indicated that an en-
called alpha parameter, which is likely to be of order ergetic outflow could be driven from the collapse of
10−2 in the presence of MRI turbulence (e.g., Balbus rotating stars when the accretion flow is advection-
& Hawley 1998; Hawley et al. 2013; Suzuki & Inutsuka dominated.1
2014; Shi et al. 2016; Kiuchi et al. 2018; Held & Ma- Second, it has been speculated that the matter in the
matsashvili 2022). Accretion disks formed around black disk outflow could be neutron-rich, and thus, the out-
holes in the collapsar scenario (MacFadyen & Woosley flow may be a site for r-process nucleosynthesis (Surman
1999) often become a neutrino-dominated accretion disk et al. 2006; Pruet et al. 2003; Kohri et al. 2005). Siegel
(e.g., Kohri & Mineshige 2002; Di Matteo et al. 2002) in et al. (2019) suggested, based on their magnetohydrody-
which internal energy generated by the viscous heating namics simulations with an approximate neutrino treat-
is released primarily by the neutrino emission. However, ment in a fixed black-hole spacetime, that neutron-rich
the latest studies for the systems of a black hole and a matter may be ejected from the disk cooled by neutrinos
compact accretion disk have shown that in the late stage and the heavy nuclei up to third peak of the r-process el-
of the viscous evolution, the neutrino cooling rate drops ements may be synthesized. In a similar setup but with
due to the viscous angular momentum transport and Monte-Carlo neutrino transfer, however, Miller et al.
subsequent expansion of the disk (Fernández & Metzger (2020) pointed out that the electron fraction (Ye ) of the
2013; Just et al. 2015; Fujibayashi et al. 2020a,b; Just ejecta is higher than 0.3, and thus, nuclei only up to the
et al. 2022b). In such a stage, the viscous heating can second peak of r-process elements are synthesized. Just
be used for launching a strong outflow from the disk in et al. (2022a) performed viscous hydrodynamics simula-
a viscous timescale estimated by tions in Newtonian gravity with general relativistic cor-
2
 −1  −2 rections incorporating moment-based neutrino radiation
rdisk αvis cs
tvis ∼ ≈ 4s transfer for the collapse of a rotating massive star and
ν 0.03 109 cm/s showed that the outflow from the disk formed around
1/2  1/2
the black hole has the electron fraction higher than 0.4.

MBH rdisk
× . (2)
10M⊙ 107 cm Therefore, the speculation in this field has not converged
yet. Moreover, no fully general relativistic work, which
The order of magnitude of the explosion energy gener-
self-consistently takes into account the self gravity of the
ated by the viscous heating is estimated by ∼ Lvis tvis ,
collapsing star and the formed black hole in general rel-
which is comparable to or even larger than that of typi-
ativistic manner, has been carried out. Obviously more
cal supernovae (∼ 1051 erg) for plausible values of MBH ,
detailed studies are required.
Mdisk , and rdisk . This motivates us to explore a scenario
of the explosion from a massive accretion disk around a
spinning black hole formed during the rotating stellar 1 Note that it is still possible to achieve successful neutrino-driven
core collapse. explosion in proto-neutron star phases even for collapses for com-
pact progenitor stellar cores (see a series of work Obergaulinger
The sub-relativistic outflow from the disk is of im- & Aloy 2020; Aloy & Obergaulinger 2021; Obergaulinger & Aloy
portance for several aspects. First, it can be an essen- 2021, 2022 and also Fujibayashi et al. 2021).
tial energy source to power a supernova-like explosion
 Collapsar leaving BH-disk 3

Third, recent high-cadence transient surveys have with the fourth-order Lagrange interpolation is imple-
shown that there is a variety of optical transients that mented.
are not canonical supernovae. Those with timescales The neutrino radiation transfer equations are approx-
of a few days, which are much shorter than that of imately solved using a leakage scheme together with the
normal supernovae (> 10 d), are such examples (e.g., truncated moment formalism (Fujibayashi et al. 2017;
Drout et al. 2014, Prentice et al. 2018, and Tampo et al. see also Sekiguchi 2010). In this formulation, the neu-
2020). Despite intensive photometric and spectral obser- trino field is split into two components; trapped and
vations, the progenitors of the transients different from free-streaming neutrinos. The trapped neutrinos are as-
the canonical supernovae are still not clear. There are sumed to be tightly coupled with the fluid and have the
several scenarios in which a collapse of a massive star same local temperature and velocity as those of the fluid.
leading to black-hole formation plays a central role (e.g., This component is treated as a part of the fluid and con-
Margutti et al. 2019, Perley et al. 2019). However, the tributes to the internal energy and pressure. It becomes
previous studies are limited only to the ones based on the free-streaming component with the generation rate
simplified models (see Piran et al. 2019 and Gottlieb controlled by the local diffusion rate of neutrinos. The
et al. 2022b for a recent simulation-based model). Thus, free-streaming neutrinos are assumed to obey radiation
it is important to provide predictions based on reliable transfer equations, which are solved by a truncated mo-
numerical simulations for interpreting the observation ment formalism with the M1 closure relation (Thorne
and confirming the origins of the mysterious transients. 1981; Shibata et al. 2011). Following our previous work
Motivated by these current situations, in this pa- (Fujibayashi et al. 2020a,b,c), we solve the equations for
per, we explore the long-term evolution of the collapse the frequency-integrated energy and momentum density
of rotating massive stars by fully general relativistic for the three neutrino radiation fields (electron, electron
radiation-viscous-hydrodynamics simulations with an anti-, and other neutrinos).
approximate neutrino transfer. The viscous hydrodynamics equations are solved using
This paper is organized as follows. In § 2, our method the formulation described in Shibata et al. (2017), in
of the simulations is briefly described. We also intro- which the energy-momentum tensor is written as
duce the pre-collapse stellar models which we employ 0
Tµν = ρhuµ uν + P gµν + νρhτµν , (3)
from the stellar evolution calculations. Then, in § 3, the
results of our numerical-relativity simulations and the where ρ := mu nb is the rest-mass density, h = c2 + ε +
nucleosynthesis calculations are presented. We discuss P/ρ is the specific enthalpy with the specific internal
the possible optical transients and implications to broad- energy ε and pressure P , ν is the kinematic viscous co-
lined type Ic supernovae and gamma-ray bursts based on efficient, and gµν is the spacetime metric. The viscous
0 0 µ
our results in § 4. We also discuss the possible produc- tensor τµν is a symmetric tensor that satisfies τµν u =0
tion of light r-process nuclei and effects on the optical (Israel & Stewart 1979), and it is determined by the fol-
transient. § 5 is devoted to a summary. Throughout this lowing equation
paper, G, c, and kB denote the gravitational constant, 0
Lu τµν 0
= −ζ(τµν − σµν ), (4)
speed of light, and Boltzmann’s constant, respectively.
where Lu denotes the Lie derivative with respect to uµ
2. METHOD and ζ is a coefficient, for which we set ζ −1 = O(10µs).
Assuming the form of the shear tensor as
2.1. Numerical code
Numerical-relativity simulations are performed with σµν = ∇µ uν + ∇ν uµ = Lu (gµν + uµ uν ), (5)
our latest axisymmetric neutrino-radiation viscous- where ∇µ is the covariant derivative with respect to gµν ,
hydrodynamics code. The detail of the code is described 0
we obtain the evolution equation for τµν := τµν −ζ(gµν +
in Fujibayashi et al. (2017, 2020c). In this code, Ein- uµ uν ) as
stein’s equation is solved in the original version of the 0
Lu τµν = −ζτµν . (6)
Baumgarte-Shapiro-Shibata-Nakamura formalism (Shi-
bata & Nakamura 1995; Baumgarte & Shapiro 1998) We only need to solve the spatial part of τµν because
0 µ
with a constraint propagation prescription to make of the presence of the condition τµν u = 0. The spatial
the constraint violation to propagate outward (Hilditch part obeys the following evolution equation (in Carte-
et al. 2013). A dynamical gauge condition described in sian coordinates):
Fujibayashi et al. (2017) is employed. To impose the √ √
∂t (ρut −gτij ) + ∂k (ρuk −gτij )
axisymmetry for the geometrical variables, the so-called √ √ 0
cartoon method (Alcubierre et al. 2001; Shibata 2000) + ρut −g(τik ∂j v k + τjk ∂i v k ) = −ρ −gζτij . (7)
4 Fujibayashi et al.

Table 1. Model description. The columns provide from left to right: model name, progenitor model name, mass of the
progenitor star, angular velocity profile, equation of state, compactness just prior to the collapse, innermost grid spacing, the
values of δ and N , and the location of the outer boundaries along each axis. The last column shows a model for the black
hole-disk system.

Model Progenitor Progenitor star Ω profile EOS ξ2.5 ∆x0 (m) δ N L (cm)
AD09x1 AD09 MZAMS = 9M⊙ Original ×1 DD2 0.68 175 0.01 975 1.0 × 1010
AD20x1 AD20 MZAMS = 20M⊙ Original ×1 DD2 0.66 175 0.01 975 1.0 × 1010
AD20x2 AD20 Original ×2
BHdisk – 10M⊙ BH-3M⊙ disk – DD2 – 220 0.01 801 3.6 × 109

The effective viscosity in the disk is believed to arise as AD09x1

a result of the turbulence induced by magnetohydrodyn- 1018 AD20x1
AD20x2
mical instabilities such as MRI (Balbus & Hawley 1991; jISCO
Balbus & Hawley 1998) and Kelvin-Helmholtz instabil- 1017

j (cm2/s)
ity (e.g., Obergaulinger et al. 2010). Following Shakura
& Sunyaev (1973), we define the viscous coefficient by 1016
ν = cs ℓtur , (8)
1015
where ℓtur is the mixing length scale (or the largest eddy
size) in the turbulence. In the alpha disk model, ℓtur is
1014
written as αvis H (Shakura & Sunyaev 1973). In this 0 2 4 6 8 10 12 14 16 18
study, we assume that ℓtur is proportional to the size of Mr /M
the black hole as
Figure 1. Specific angular momentum (solid curves)
2GMBH and that for innermost stable circular orbits (dashed curves)
ℓtur = 0.03 × , (9)
c2 along the equatorial direction as functions of enclosed mass
where the black-hole mass, MBH , continuously increases for each model.
with time due to the matter accretion. Thus, we
employ a time-varying form of ℓtur . By the above The grid structure is the same as in the 2D sim-
definition, the mixing length scale becomes ℓtur ≈ ulations recently performed with the same code (Fu-
0.9 km(MBH /10M⊙ ). Since the disk scale height H is jibayashi et al. 2023), in which the cylindrical coordi-
larger than 2GMBH /c2 for most parts of the disk around nates (R, z) are employed. In the inner cylindrical re-
the black hole, Equation (9) implies that we assume gion of R ≤ 100∆x0 and z ≤ 100∆x0 , a uniform grid
αvis ≤ 0.03, i.e., a conservative value of αvis . Even for with the grid spacing of ∆x0 is prepared, while in the
such a conservative value of αvis , we will find a signifi- outer region, a non-uniform grid with an increase rate
cant effect in § 3. This spatially constant mixing length of the grid spacing of 1 + δ is prepared. The value of δ,
scale, Eq. (9), may lead to a smaller kinetic viscous co- grid number N for each axis, and location of the outer
efficient than that employed in Just et al. (2022a) (see boundaries along each axis (denoted by L) are listed in
also Just et al. 2022b). Table 1. We assume the plane symmetry with respect
We note that the viscosity is incorporated just prior to to the z = 0 plane (the equatorial plane).
the formation of the disk around the black hole. Specifi-
cally, we first perform simulations without viscosity until 2.2. Models
the disks are formed. We then go back to the time slice One of the theoretically accepted central engines of
prior to the disk formation and rerun the simulation long-duration gamma-ray bursts is the system of a spin-
with viscosity. We check that there are any significant ning black hole with a surrounding accretion disk (e.g.,
differences of the density and angular velocity profiles Woosley 1993). For the formation of a massive accre-
for infalling matter between the results of the simula- tion disk around the black hole, the progenitor star
tions with and without viscosity.2 has to be rapidly rotating. The progenitor models
of Aguilera-Dena et al. (2020) may be promising for
2 such a scenario. We employ, from their work, two of
The infalling matter at the formation of the disk is that from
carbon-oxygen-neon layer of the star, which does not have signif- the rotating stars with zero-age-main-sequence (ZAMS)
icant differential rotation. In this sense, turning on the viscosity masses of 9 and 20M⊙ (hereafter AD09 and AD20, re-
does not have a significant effect during the collapse. spectively). Because of nearly chemically homogeneous
 Collapsar leaving BH-disk 5

evolution, the pre-collapse stars have very massive cores 400GMBH /c2 ≈ 5900 km, respectively. The constant
that are very compact. The compactnesses at the radius entropy per baryon of s = 7kB is simply assumed.
with its enclosed mass 2.5M⊙ (referred to as r2.5M⊙ ), In this paper, we employ a tabulated equation of state
ξ2.5 := 2.5/(r2.5M⊙ /1000 km), are 0.68 and 0.66 for AD09 (EOS) referred to as DD2 (Banik et al. 2014). We ex-
and AD20, respectively. This suggests that they are likely tended the table down to low-density (ρ ≈ 0.17 g/cm3 )
to form a black hole with no neutrino-driven explosion and low-temperature (kB T = 10−3 MeV); see Hayashi
if it is non-rotating (see, e.g., O’Connor & Ott 2011 but et al. (2022) for the procedure.
see Burrows et al. 2019.) To evolve black holes in a good precision, the radius
It should be noted that the angular momentum trans- of the apparent horizon has to be resolved well (for the
port via the convection, circulation, and magnetohy- dependence of the black-hole mass and spin on the grid
drodynamical interactions is taken into account only in resolution, see, e.g., Fujibayashi et al. 2020a). Employ-
an approximate way in current stellar evolution stud- ing a stiff EOS like DD2, which predicts a larg maxi-
ies. The original rotation profile is applied for models mum mass of neutron star, is advantageous to numeri-
AD09x1 and AD20x1, while for model AD20x2, the origi- cally resolve black holes in a good accuracy (for a given
nal angular velocity is doubled to investigate a possible computational resource), because the black-hole mass
case in which the rotation of the star is even faster. (i.e., the radius of the apparent horizon) at its forma-
Figure 1 shows the specific angular momentum along tion is larger for a stiffer EOS. Thus, for the present
the equatorial direction as a function of enclosed mass study, the DD2 EOS would be a better choice to save
for the models employed in this paper. The specific an- computational resources.
gular momentum at the innermost stable circular orbit
for the black hole with the enclosed mass and angular 2.3. Diagnostic
momentum, jISCO , is also plotted in the dashed curves. We define the mass infall rate to the central region by
For each progenitor model, the enclosed mass at which
√
Z
the solid and dashed curves cross approximately gives Ṁfall = −gρuk dsk , (10)
the black-hole mass at which the infalling matter starts r=rin
forming an accretion disk around the black hole. This
where dsi = r2 δir d cos θdϕ is the surface element of a
figure indicates that disks are likely to be formed when
sphere, g = det(gµν ), and rin is chosen in the following
the black holes grow to 4.7, 4.9, and 7.8M⊙ for models
manner: In the early phase, it is the largest value among
AD20x2, AD09x1, and AD20x1, respectively. In § 3, we
the radii of the surface of a standing accretion shock
will confirm that this prediction is approximately cor-
formed as a result of the core bounce; after this shock
rect.
disappears due to the matter infall from the outer region
At the time of the disk formation, the mass infalling
which induces the collapse of a proto-neutron star to a
rates for model AD20x2 are higher than those for AD09x1
black hole, the maximum radius of the apparent horizon
and AD20x1. This makes a difference in the evolution
rAH is used for rin ; after the formation of the disk, we
process of the disk and the onset timing of the outflow
again choose the largest value among the radii of the
from the disk as found in § 3.3 and § 3.2.
surface of a standing accretion shock, which is formed
In several previous studies, the mass ejection and
along the disk surface. In the late phase, we also define
nucleosynthesis calculation were carried out by simply
the mass accretion rate to the black hole by
modeling collapsar remnants as a black hole-disk sys-
tem (Siegel et al. 2019; Miller et al. 2020). The evolu- √
Z
tion process of the black hole-disk system as a result of ṀBH = −gρuk dsk . (11)
r=rAH
stellar collapse may be different from that of an isolated
disk around a black hole. To clarify the possible differ- The definitions of the unbound matter and explosion
ences, we also perform a simulation for a system of a energy are the same as those in Fujibayashi et al. (2021).
spinning black hole and a massive disk (model BHdisk) We first define the specific binding energy and binding
as in our previous studies (Fujibayashi et al. 2020a,b). energy flux density of the matter as
The initial condition is constructed using a code of Shi-
Tt t
bata (2007). The masses of the black hole and disk are ebind := − − (1 + εmin ), (12)
10 and 3M⊙ , respectively, and the dimensionless spin ρut
i
of the black hole is ≈ 0.6. The inner and outer edges fbind := −Tt i − ρui (1 + εmin ), (13)
of the disk are chosen to be 4GMBH /c2 ≈ 59 km and
where Tt t and Tt i are the time-time and time-space com-
ponents of the energy-momentum tensor and εmin is the
6 Fujibayashi et al.

Table 2. Main results. The columns provide from left to right: model name, post-bounce time of black hole formation and
explosion, ejecta mass, explosion energy, mass of the matter with maximum temperature higher than 5 GK (1 GK = 109 K),
mass of 56 Ni, estimated peak bolometric luminosity, and rise time of bolometric light curve. For Mej and Eexp , there are two
columns with different extraction radius. The values in parentheses for columns of Mej and Eexp are, respectively, the mass
and binding energy of stellar matter above the extraction radius. Those for columns of Lpeak and trise are values for the case in
which a half of the mass and binding energy above the extraction radius are considered.

Model tBH texp Mej Eexp M>5 GK MNi Lpeak trise

(s) (s) (M⊙ ) (1051 erg) (M⊙ ) (M⊙ ) (1042 erg/s) (days)
rext = 1 × 109 cm 2 × 109 cm 1 × 109 cm 2 × 109 cm
AD09x1 0.87 13.2 0.08 (1.6) 0.12 (1.5) 0.57 (−0.21) 0.53 (−0.19) 0.04 0.01 0.50 ( 0.28) 3.3 (13.1)
AD20x1 0.92 20.9 0.22 (6.0) 0.25 (5.6) 1.8 (−1.4) 1.3 (−1.4) 0.14 0.06 2.6 ( 0.83) 4.4 (27.8)
AD20x2 2.33 15.2 0.96 (7.9) 1.3 (6.6) 3.5 (−1.7) 3.1 (−1.6) 0.63 0.15 4.1 ( 2.0) 10.9 (26.8)
BHdisk 0.0 12.2 > 0.088 (–) > 0.3 (–) > 0.088 > 0.037 – –
Values with “>” denote that they are still increasing at the end of the simulation.

minimum specific internal energy of the employed EOS per baryon should be released into the internal energy,
table (≈ −0.0013c2 for DD2). We then define the condi- which can be an important energy source of the explo-
tion for unbound matter as ebind > 0. We find however sion of the star (if the internal energy is not carried
that the binding energy of the outer layer of the star in away by the neutrino emission). However, this effect is
our computational domain is not accurately calculated absent in our simulation because of the assumption of
due to a numerical error accumulated in long-term simu- nuclear statistical equilibrium (NSE) in constructing the
lations. Therefore, we decide to calculate the diagnostic EOSs, for which the low-temperature matter in the com-
explosion energy as well as the mass of the unbound putational domain is assumed to be composed mostly
matter as of 56 Ni. In addition, 56 Ni is photo-dessociated into
Z
√ lighter nuclei if the temperature of the matter increases
Eexp := ebind ρut −gd3 x to ≳ 7 GK, consuming more internal energy than in the
ebind >0,r<rext
Z tZ dissociation of 12 C or 16 O. These temperature condi-
+ k
fbind dsk dt, (14) tions are found when the shock formed near the disk
ebind >0,r=rext around the black hole has propagated sufficiently far out
√ within the star and, hence, an excessively large energy
Z
Mej := ρut −gd3 x of 8.6 MeV/nucleon, instead of ≈ 8.0 MeV/nucleon, is
ebind >0,r<rext
Z tZ consumed. Thus, the explosion energy in our present
+ ρukbind dsk dt, (15) simulations is underestimated. This point will be dis-
ebind >0,r=rext cussed in § 3.6.
where dsk is the surface element of the sphere with the We also note that the inclusion of εmin in Eqs. (12)
radius rext . These quantities are defined by the volume and (13) is important to estimate the ejecta mass and
integral for the matter of the positive binding energy explosion energy in a physically correct way. As men-
inside an extraction radius rext plus the time integral tioned above, the low-temperature matter in the outer
for the components of the positive binding energy flux region of the star is assumed to be composed of iron
at the radius. group nuclei, which have smaller rest masses per baryon
A part of the stellar matter is located outside the ex- than the atomic mass unit (mu ; the mass of 12 C divided
traction radius and even outside the computational do- by 12). Due to the low-temperature, such matter has a
main. Its mass and binding energy can contribute to low specific internal energy, which leads to ε < 0 for the
the ejecta mass and explosion energy. To estimate their matter. If we do not include εmin in Eqs. (12) and (13),
contribution, we first obtain the enclosed mass at the such matter with ε < 0 is not recognized as ejecta, even
extraction radius when the shock wave reaches the ex- though it gains sufficient energy to be unbound after
traction radius. We then estimate the binding energy being swept by the shock wave.
of the matter above the radius with the same enclosed
mass in the pre-collapse profile. We compare the explo- 2.4. Tracer-particle method
sion energies with different extraction radii in § 3.4. To perform the nucleosynthesis calculation, we apply
We note that the matter in the progenitor stars is our post-process tracer-particle method for the results
composed mainly of 4 He, 12 C, 16 O, and 20 Ne. When of our simulations. The method is the same as that in
16
O burns into 56 Ni, the rest-mass energy of 0.67 MeV Fujibayashi et al. (2023). The tracer particles are dis-
 Collapsar leaving BH-disk 7

1.0 traction radius. The mass of each particle is deter-

AD09x1
AD20x1
mined based on the mass flux at the extraction radius
AD20x2 √
0.8 Ṁfall as ∆m = rext 2 ∆Ωρur −g∆tset , where ∆Ω is the solid
Ṁfall, ṀBH (M /s)

ṀBH
angle element.
0.6
3. RESULT
0.4
3.1. Evolution after bounce to disk formation
0.2 For all the collapse models, a proto-neutron star is
initially formed after the core bounce together with the
0.0 standing accretion shock formation. Subsequently, due
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
tpb (s)
to the matter accretion from the outer region, the proto-
neutron star collapses to a black hole. The black-hole
1054
AD09x1 BHdisk formation time is tpb = 0.9, 0.9, and 2.3 s for models
AD20x1 Lν
1053 AD20x2 Viscous heating AD09x1, AD20x1, and AD20x2, respectively (denoted by
tBH in Table 2). Here tpb denotes the time after the
1052 core bounce. For model AD20x2, the formation of the
Lν (erg/s)

black hole is delayed due to the significant centrifugal-

1051
effect associated with the rapid rotation. These results
1050 illustrate that the formation process of the black hole
(and subsequent evolution process of the black hole and
1049 disk) depends on the profiles of the density and specific
angular momentum of the progenitor stars.
1048
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 For seconds after the black-hole formation, the in-
tpb (s) falling matter does not have enough angular momen-
tum to form disks, and thus, it is swallowed simply by
AD09x1

10−1 AD20x1 the black hole. A geometrically thin disk starts form-
AD20x2
BHdisk ing at tpb ≈ 7.0, 10.5, and 5.0 s at which the black-hole
10−2 mass is MBH /M⊙ ≈ 5.0, 7.0, and 4.5 for models AD09x1,
Lν /ṀBHc2

AD20x1, and AD20x2, respectively (see, e.g., panel (a) of

10−3 Fig. 3 for model AD20x1). The above black-hole mass
is consistent with the ones inferred from the distribu-
10−4
tions of density and specific angular momentum of the
10−5 progenitor stars (see § 2.2).
The disks at their formation are geometrically thin be-
10−6 cause of the lower pressure inside the disk than the ram
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
tpb (s) pressure of the infalling matter (Sekiguchi & Shibata
2011). The disks then become geometrically thick when
Figure 2. Top: Mass infall rate across the surface of r = rin the pressure in the outer part of the disk and ram pres-
(solid curves) and mass accretion rate onto the black hole sure of the infalling matter become comparable. The
(dashed curves). Circles on each curve denote the time at
vertical expansion of the disk occurs at tpb ≈ 13, 21,
which the viscosity is turned on. Middle: Total neutrino lu-
minosity (solid curves) and total viscous heating rate (dashed and 10 s for models AD09x1, AD20x1, and AD20x2, re-
curves) inside the shock wave. Bottom: Neutrino cooling effi- spectively (see panel (b) of Fig. 3 and panels (a) and
ciency defined by the total neutrino luminosity divided by the (d) of Fig. 4). In the presence of the viscosity, the disk
mass accretion rate onto the black hole. For model BHdisk, expansion occurs at time slightly later than that in the
tpb = 0 corresponds to the beginning of the simulation. corresponding non-viscous model. The reason for this is
that the viscous angular momentum transport acceler-
tributed for 128 polar angles in the range of θ = [0 : π/2] ates the accretion of the disk matter onto the black hole
on the arc with the radius of rext = 2 × 104 km. The and the increase of the pressure in the disk is delayed.
particles are continuously set with the time interval of After the vertical expansion of the disk, a shock sur-
∆tset := rext ∆θ/⟨v r ⟩, where ∆θ = (π/2)/128 and ⟨v r ⟩ face between the disk and infalling matter expands (see,
is the average radial velocity of the ejecta at the ex- e.g., panel (b) of Fig. 3). The geometrical cross section
of the shock surface becomes large, which enhances the
8 Fujibayashi et al.

Figure 3. Snapshots at tpb = 20.0 (a), 21.0 (b), 21.4 (c), and 28.7 s (d) for model AD20x1. The panels (a)–(d) show
the snapshots just prior to the formation of a geometrically thick disk, at the onset of the outflow, at the expanding phase,
and the final snapshot, respectively. Each panel has four sub-panels of rest-mass density (top-left), entropy per baryon (top-
right), temperature (bottom-left), and electron fraction (bottom-right). The black solid circle for the first panel shows the
region inside the apparent horizon (for other panels, it is too small to be seen because the plotted region is much wider than
GMBH /c2 ). Note that the regions of the plots are different for each snapshot. An animation for this model is available at
https://www2.yukawa.kyoto-u.ac.jp/~sho.fujibayashi/share/AD20x1_multiscale.mp4.

dissipation of the kinetic energy of the infalling matter 3.2. Models AD09x1 and AD20x1: Models of lower
at the shock more efficiently (Sekiguchi & Shibata 2011). infalling rate at disk formation
As a consequence, the neutrino luminosity increases dur- For models AD09x1 and AD20x1, the rate of mass sup-
ing this phase. The neutrino luminosity depends on the ply to the disk after the vertical expansion of the disk,
rate of mass supply to the disk (see Fig. 2). For a rapidly Ṁfall − ṀBH , is small (≲ 0.1M⊙ /s; see the top panel
rotating model AD20x2, the neutrino luminosity is higher of Fig. 2). As a result, the disk temperature cannot be
because the mass-infall rate is higher. high enough for the efficient cooling by neutrino emis-
After the vertical disk expansion, the models with the sion. Thus, the viscous heating dominates over the neu-
original rotational profiles AD09x1 and AD20x1, and a trino cooling in the entire phases after the disk expan-
rapidly rotating model AD20x2 show different evolution sion (the middle panel of Fig. 2); a neutrino-dominated
process in terms of mass infall rate and neutrino cooling accretion disk is not formed in these models. The dom-
efficiency. We describe the evolution processes in the inance of the viscous heating leads to an early launch of
following subsections separately. the outflow (several hundreds of milliseconds) after the
 Collapsar leaving BH-disk 9

Figure 4. The same as Fig. 3 but for models AD09x1 (top panels) and AD20x2 (bottom panels). For each model, the panels
from left to right show the snapshots at the formation of a geometrically thick disk, after the onset of the outflow, and the final
snapshots, respectively.

disk expansion for these models. The outflow is launched negligible compared with the kinetic energy of electrons
mainly toward the equatorial direction (see panel (c) of ∼ kB T (see Just et al. 2022b, Arcones et al. 2010, and
Fig. 3 and (b) of Fig. 4). The final snapshots (see panel Beloborodov 2003). The result of these models are sim-
(d) of Fig. 3 and (c) of Fig. 4) also show the deformed ilar to those found in Just et al. (2022a) (although our
profile of shock surfaces and the outflow from the disk prescription of the viscous hydrodynamics is different
toward the equatorial direction. from that of their study).3 Because of the high electron
The top panel of Fig. 5 shows the mass histogram as fraction of the ejecta, we do not expect an r-process
a function of the electron fraction of the ejecta. For nucleosynthesis (see § 4.2).
these models, the value of Ye is not low (at lowest 0.47; The middle panel of Fig. 5 shows the mass histogram
see the top panel of Fig. 5) because of their low disk as a function of entropy per baryon, s/kB . These models
density and low neutrino cooling efficiency throughout have large values of entropy per baryon s/kB ≈ 30–50
the disk evolution. This leads to weak electron degen- and short expansion timescales r/v r ∼ 50 ms than a
eracy of the disk matter and thus keeps higher values rapidly rotating model AD20x2 (see § 3.3). This result
of Ye . Some components have the electron fraction even can be explained by the following three reasons. First,
higher than 0.5. This reflects the fact that the positron for moderately rotating models, the typical radius of the
capture proceeds in a shorter timescale than the elec- disk is smaller at the onset of the outflow because of the
tron capture under the condition of low electron degen- shorter time of the onset of the outflow after the disk for-
eracy and mildly high temperature kB T ≳ 1 MeV. In mation. This requires a larger heating efficiency for the
this condition, the positron capture n + e+ → p + νe is
energetically more preferred than the electron capture 3 We note that our prescription of the viscosity may lead to smaller
because the mass difference between a free neutron plus kinetic viscous coefficient than that in Just et al. (2022a).
an electron and a free proton (mn + me − mp )c2 is not
10 Fujibayashi et al.

100 3.3. Model AD20x2: Models of higher infalling rate at

AD09x1
AD20x1 disk formation
AD20x2
10−1 BHdisk This rapidly rotating model has appreciable differ-
∆M/Mej,Tmax>5 GK

ences from models AD09x1 and AD20x1 in the follow-

ing two aspects: Specific angular momentum is much
10−2
larger, and as a result, the mass infalling rate after the
disk formation is much higher. In particular, the latter
10−3 difference significantly modifies the evolution process of
the disk from the model AD20x1.
10−4
For this model, a high-mass infall phase continues
0.35 0.40 0.45 0.50 0.55 0.60 prior to the disk outflow for a long timescale of ∼ 10 s.
Ye (T = 5 GK)
The reason for this is that during such a phase, the
100 ram pressure from the infalling matter is strong and the
AD09x1
AD20x1 temperature in the shocked region is high enough for
AD20x2
10−1 BHdisk
enhancing the neutrino cooling, and hence, the thermal
∆M/Mej,Tmax>5 GK

pressure generated by the viscous heating cannot be high

enough for launching the disk outflow.
10−2 As a consequence, model AD20x2 exhibits a long-term
quasi-steady phase of the disk in which the viscous
10−3 heating and neutrino cooling rates are comparable for
tpb ≈ 10–16 s. In this quasi-steady phase, the neutrino
cooling efficiency, defined by the neutrino luminosity Lν
10−4 divided by ṀBH c2 , is several to ten per cent (see the
101 102 103
s/kB (T = 5 GK) bottom panel of Fig. 2), which is sufficiently high for
100 neutrinos to carry away the energy generated by viscous
AD09x1
AD20x1
heating. That is, a neutrino-dominated accretion flow
AD20x2 (NDAF) is established in this phase.
10−1 BHdisk
After the onset of the outflow from the disk, an ex-
∆M/Mej,Tmax>5 GK

panding shock with a slightly oblate shape is formed (see

10−2 the panel (e) of Fig. 4). During the shock expansion, the
matter infall onto the black hole and disk still continues
in the polar and equatorial directions because the out-
10−3 flow from the disk is launched along the surface of the
geometrically thick disk (z ≈ 0.5R). The outflow to-
10−4 wards the polar and equatorial directions is suppressed
10−2 10−1 100 101
r/v r (s; T = 5 GK)
for different reasons. Because the outflow is launched
mainly from the surface of the inner side of the disk,
Figure 5. Mass histograms as a function of the electron the outflow is prohibited in the equatorial direction due
fraction (top), entropy per baryon (middle), and expansion to the presence of the dense outer disk (bound matter).
timescale (r/v r ; bottom) at T = 5 GK of the tracer particles Near the polar axis, the infalling matter that passed
that experience temperature higher than 5 GK. through the expanding shock surface converges toward
the polar region. As a result, the ram pressure near the
disk matter to be unbound and for the outflow matter to polar region is enhanced and becomes larger than that
have the higher velocity. Second, since the outflow sets in the other direction. This prevents the outflow toward
in before the disk settles into a quasi-steady phase, the the polar direction. The final snapshot for this model
shear of the poloidal velocity field in addition to that of also shows that the outflow from the disk expands in a
Keplerian motion leads to more efficient viscous heating. diagonal direction (see panel (f) and (i) of Fig. 4; the
Third, the entropy generated by the viscous heating is matter with higher entropy, s/kB ≳ 30, is the outflow
not efficiently lost by the neutrino emission because of component launched from the disk).
the lower neutrino cooling efficiency. These properties Although a neutron-rich region with Ye < 0.2 is
of moderately rotating models produce the ejecta with present in the inner mid-plane region of the disk reflect-
a higher entropy and a shorter timescale. ing the high density there (see panel (e) of Fig. 4), the
 Collapsar leaving BH-disk 11

lowest value of Ye of the ejecta is 0.40 for this model. The AD09x1( 1.6, 1.4)
reason for this high value is that the outflow is launched 1.5 AD20x1( 6.0, 5.6)
AD20x2( 7.9, 6.6)
from the disk surface region, in which the electron de- BHdisk( 0.0, 0.0)
rext = 2 × 109 cm
generacy is not as high as that in the disk mid-plane rext = 1 × 109 cm

Mej (M )
and the value of Ye is close to 0.5. The component with 1.0
Ye > 0.5 is present because of the same reason as for
models AD09 and AD20x1 (see § 3.2). From the ejecta
with Ye ≳ 0.5 and with T ≳ 5 GK, a substantial amount 0.5
of 56 Ni can be synthesized. We discuss this topic in § 3.6.
The typical value of s/kB is found to be 10–20 for
this rapidly rotating model, which is lower than those 0.0
0 1 2 3 4 5 6 7 8 9
for models AD09x1 and AD20x1 (see the middle panel of t − texp (s)
Fig. 5). This is because the outflow is developed well 4
AD09x1(−0.2, −0.2)
(∼ 10 s) after the formation of quasi-steady disks in a AD20x1(−1.4, −1.4)

milder manner for the rapidly rotating model than for AD20x2(−1.7, −1.5)
3 BHdisk(0.0, 0.0)
the moderately rotating models. This trend is also illus- rext = 2 × 109 cm

Eexp (1051 erg)

rext = 1 × 109 cm
trated in mass histogram of the expansion timescale (see
2
the bottom panel of Fig. 5). The expansion timescale for
the rapidly rotating model is typically ≳ 0.1 s, which is
longer than for the moderately rotating models. The 1
entropy and expansion-timescale distributions for the
rapidly rotating model are similar to those for model 0
BHdisk, indicating that the formation of a dense disk in
0 1 2 3 4 5 6 7 8 9
this model is a key to determining the properties of the t − texp (s)
ejecta.
The disk mass is ≈ 0.2M⊙ for model AD20x2 and the Figure 6. Time evolution of ejecta mass (top) and ex-
cooling efficiency is low at the termination of the sim- plosion energy (bottom) as functions of post-explosion time
ulation. The electron fraction for the bulk of the disk t − texp . The dashed and solid curves are the values calcu-
lated with rext = 1 × 109 cm and 2 × 109 cm, respectively.
matter is frozen to be Ye = 0.4–0.5, because of a long
The values in the legend denote the mass (in units of solar
weak interaction (electron/positron capture) timescale. mass) and binding energy (in units of 1051 erg) above the ex-
Therefore, the electron fraction of the matter expected traction radius when the shock wave reaches these radii (see
to be ejected in a longer timescale is likely to be 0.4–0.5 § 2.3).
(see also § 3.5).
As these results illustrate, the evolution of the disk, Figure 6 shows the mass of the ejecta and the diag-
the timing for the onset of the explosion, and the elec- nostic explosion energy as functions of time after the
tron fraction of the ejecta depend strongly on the dis- explosion, t − texp . Here, the time of the explosion is de-
tribution of the specific angular momentum of the pro- fined as the time at which the explosion energy reaches
genitors and the resulting mass infall rate on the disk. 1 × 1050 erg.
For the models with relatively low angular momentum For all the viscous-hydrodynamics simulations, we
(AD09x1 and AD20x1), the explosion occurs in a short find an explosion by the energy injection from the disk
timescale after the formation of a geometrically thick around the black hole. For models AD09x1 and AD20x1,
disk, at which the ram pressure of the infalling matter for which the angular momentum of the progenitor stars
is low enough for launching an outflow. In this case, the is relatively low, the formation of the disk is delayed, and
electron fraction of the ejecta resulting from the explo- the onset of the explosion occurs in a late phase (see Ta-
sion cannot be very low. By contrast, for the model with ble 2). As a consequence, only a small amount of mass
relatively high angular momentum (AD20x2), the explo- remains outside the black hole at the explosion. The
sion takes place at ∼ 10 s after the formation of the disk, small energy budget at the formation of the disk leads to
and in this case, the electron fraction of the ejecta can a slow increase in the explosion energy. For these mod-
be low with Ye < 0.45, and thus, light trans-iron nuclei els, the explosion energies amount to ≈ (0.5–1)×1051 erg
can be synthesized (see § 4.2). at the termination of the simulations, which are com-
parable or slightly smaller than the canonical value for
3.4. Ejecta mass and explosion energy core-collapse supernovae. The ejecta masses for these
12 Fujibayashi et al.

models are 0.1–0.3M⊙ at the termination of the simu- 3.5. Comparison of the collapse models with model
lations, which are an order of magnitude smaller than BHdisk
that for canonical supernovae (but see below). For model BHdisk, which consists of a 3M⊙ disk
For the rapidly rotating model AD20x2, after the around a 10M⊙ black hole with the dimensionless spin
launch of the outflow, the explosion energy increases of 0.6, the disk matter accretes onto the black hole in
beyond 1 × 1051 erg with a timescale of ∼ 1 s. For this a quasi-steady manner in the first 5 seconds. The out-
model, the explosion energy at the termination of the flow is then launched at t ≈ 5 s because the neutrino
simulation reaches ≳ 3 × 1051 erg. At the termination cooling rate has dropped far below the viscous heating
of the simulations, disks with 0.2M⊙ is present around rate (see the middle panel of Fig. 2). The increase rate
the black holes for this model. In addition, matter infall in the explosion energy for model BHdisk is much lower
still continues around the central region. Therefore, an than for the collapsing star models (see bottom panel
energy budget to provide more explosion energy is still of Fig. 6). The primarily reason for this is that, for
present. The result suggests that if the massive progen- the collapsing stars, a velocity shear is present not only
itors are rapidly rotating, a high-energy supernova-like in the accretion disk with the nearly Keplerian motion
explosion could occur from the disk outflow triggered by but also on the surface of the disk resulting from the
viscous heating. We discuss this point in §4.1. infalling matter onto the disk, which is absent for model
For all the collapse models, appreciable stellar matter BHdisk. The strong shear on the disk surface signifi-
is still present above the extraction radius. This can cantly enhances the viscous heating rate, which results
contribute additionally to the ejecta mass and explosion in the higher increase rate of the explosion energy.
energy. The masses outside the extraction radius are The origin of the velocity shear associated with the in-
larger than those of the ejecta, and thus, they may have falling matter is different from the Keplerian motion of
an impact on the ejecta velocity. The terminal aver- the disk. Here, for the latter we suppose that the MRI
age velocity of the ejecta depends on the amount of the turbulence (Balbus & Hawley 1991) is the origin of the
stellar matter that becomes ejecta in the later phase. effective viscosity. For the surface region of the disk,
The mass and binding energy of the matter outside on the other hand, we suppose that the shear region on
the extraction radii rext = 1 × 109 cm and 2 × 109 cm the disk surface should induce the Kelvin-Helmholtz in-
are listed in Table 2 (see the values in parentheses for stability. In such regions, magnetic fields are supposed
columns of Mej and Eexp ; also see Fig. 6). At the ter- to be enhanced significantly leading to the development
mination of the simulations, the absolute values of the of turbulence and dissipating the kinetic energy of the
binding energy are below the explosion energies for mod- infalling matter (e.g., Zhang et al. 2009; Obergaulinger
els AD09x1, and AD20x2. Thus, the explosion is likely et al. 2010; Rembiasz et al. 2016; Viganò et al. 2020).
successful for these models. For model AD20x1, on the Thus, it is natural to consider that the effective viscosity
other hand, the value is slightly above the explosion en- on such a region is also high.4 However, to clarify this
ergy for rext = 2 × 109 cm. Thus, it is not clear whether process in the first-principle way, magnetohydrodynam-
this model results in successful explosion. Note however ics simulation is necessary. In the future work, we plan
that the bound matter outside the extraction radius at to perform this to confirm that our assumption is indeed
the termination of the simulation has a sufficiently large correct.
angular momentum to circularize around the black hole For model BHdisk, we stopped the simulation at t ≈
(see Fig. 1). Thus, a possible increase may be expected 21 s. For this model, the mass ejection still continues
in the explosion energy for the later phase if such a mat- with the ejection rate slightly higher than the mass ac-
ter falls into the central region to power an additional cretion rate onto the black hole. This suggests that
outflow. We plan to perform a longer-term simulation a large fraction of the disk matter will be eventually
in our future work to address the possibility of the ex- ejected from the system. At the termination of the sim-
plosion for these models. ulation, the total ejecta mass is only ∼ 0.1M⊙ . However,
As already mentioned in § 2.3, the composition of the the disk mass is still ≈ 2M⊙ . Extrapolating the mass
computational domain is assumed to be that in NSE in ejection rate at the final time ∼ 10−2 M⊙ /s, we infer
our simulation. In reality, the outer region of the star
is composed mainly of 12 C, 16 O, and 20 Ne, and hence,
4
the nuclear burning of them in the stellar mantle swept For this case, the kinetic viscosity is likely to be proportional to
the infall velocity, not to the sound velocity, as ν = ℓtur vinfall ,
by the shock wave can provide an additional energy. In
where vinfall is comparable to or larger than cs , and thus, our
§ 3.6, we will investigate this more quantitatively. present treatment for the viscous coefficient may be conservative
if ℓtur /H = O(10−2 ).
 Collapsar leaving BH-disk 13

that the mass ejection continues for more than 100 s for gest that massive and rapidly rotating stars leading to a
this model. black hole and massive disk are candidates for the pro-
The mass ejection for model BHdisk sets in after the genitors of type Ic-BL supernovae.
neutrino cooling efficiency of the bulk of the disk drops We also note that the numerical simulation for this
(cf. the middle panel of Fig. 2). The ejecta for model model underestimates the total ejecta mass because we
BHdisk has a low-Ye component down to Ye ≈ 0.4. stopped the simulations at a time when the ejecta mass
The value is determined by the electron fraction in elec- and explosion energy are still increasing. Our results
tron/positron capture equilibrium when the timescale of here indicate the lower bound for the 56 Ni mass.
these reactions becomes comparable to that of the vis- It is important to note that, as can be found in Ta-
cous expansion (see Fujibayashi et al. 2020c; Just et al. ble 2, the produced 56 Ni mass fraction relative to the
2022b). Thus, the mass ejection mechanism for rapidly mass of ejecta exceeding 5 GK varies from 24% (AD20x2)
rotating model AD20x2 is qualitatively similar to that to 43% (AD20x1) depending on the electron fraction, en-
for model BHdisk. tropy, and expansion timescale of the outflowing matter
for each model. The conditions of Ye ≳ 0.5, low entropy,
3.6. Production of 56
Ni and slow expansion are favored for the efficient produc-
tion of 56 Ni. Therefore, the mass of ejecta with ≥ 5 GK
Using the time evolution of the temperature and den-
(as frequently used in the literature) only serves as a
sity along tracer particles, the post-process nucleosyn-
loose upper limit for the produced amount of 56 Ni.
thesis calculations are performed with the nuclear reac-
Suppose that 56 Ni is synthesized from 16 O, the rest-
tion network code rNET (Wanajo et al. 2018). The initial
mass energy released into the internal energy due to the
composition of the nucleosynthesis calculation depends
nuclear burning is 1.6×1049 , 7.5×1049 , and 1.9×1050 erg
on the thermal history of the tracer particles. If the
for models AD09x1, AD20x1, and AD20x2, respectively.
maximum temperature along a particle is higher than
These contributions are several to tens percents of the
10 GK, we start the nucleosynthesis calculation at the
explosion energy estimated in § 2.3 and can have notable
time that the temperature decreases to T = 10 GK with
effects. Especially for model AD20x1, the explosion en-
the mass fraction of free protons and nucleons, Ye and
ergy plus binding energy above the extraction radius
1 − Ye , respectively. For the tracer particles with the
rext = 2 × 109 cm becomes negative but the value is
maximum temperature lower than 10 GK, we start the
comparable to the energy generated by nuclear burning.
nucleosynthesis calculation at t = 0 of the simulation
Thus, to clarify whether such a marginal model explodes
with the composition depending on the position of the
successfully, feedback of the nuclear reaction has to be
particle in the progenitor star (mostly consisting of 16 O
taken into account in hydrodynamics simulations (see
and 20 Ne).
Bollig et al. 2020 and Navó et al. 2022 for a recent at-
The resulting mass of 56 Ni is listed in Table 2. For
tempt).
models AD09x1 and AD20x1, the 56 Ni masses are smaller
than 0.1M⊙ (0.01 and 0.06M⊙ , respectively), because 4. DISCUSSION
of their smaller ejecta masses. Thus for moderately ro-
tating progenitor models, the 56 Ni mass is likely to be 4.1. Optical transients
comparable to or smaller than that for an ordinary su- Using the ejecta properties obtained in the present
pernova. These models predict the presence of moder- study, we analytically calculate bolometric luminosity
ately bright, but rapidly varying optical transients as models for photons following Arnett (1982). The ther-
found in § 4.1. malization efficiency for gamma-rays is estimated follow-
It is found that for model AD20x2 the mass of 56 Ni ing Colgate et al. (1997) with the optical depth for non-
amounts to 0.15M⊙ , reflecting the large mass of the thermal gamma-rays κγ = 0.03 cm2 /g. For the optical
ejecta that experiences high temperature ≳ 5 GK. Thus, depth of thermal photons, we simply set κ = 0.1 cm2 /g.
the mass of 56 Ni found for this model could be high We note that the “Arnett” model tends to infer a larger
56
enough for explaining high-energy supernovae such as Ni mass by a factor of a few than that inferred by the
broad-lined type Ic (type Ic-BL) supernovae, considering “radioactive tail” luminosity of a supernova, the latter
that the 56 Ni mass inferred with the so-called “Arnett’s being less ambiguous (e.g., Meza & Anderson 2020, Af-
rule” (Arnett 1982) is possibly overestimated (e.g., Meza sariardchi et al. 2021, and Rodrı́guez et al. 2022). This
& Anderson 2020, suggesting that the 56 Ni masses in- indicates that the luminosity predicted by the Arnett
ferred from the radioactive tail luminosity for nearby model for a given 56 Ni mass may be underestimated
two type Ic-BL supernovae, SN2009bb and SN2016coi, by a factor of a few (see Dessart et al. 2015, 2016 and
are 0.08 and 0.10M⊙ , respectively). These results sug- Khatami & Kasen 2019). We also note that for the
14 Fujibayashi et al.

κ = 0.1 cm2 /g AD09x1 κρR = 1

increased ejecta mass and decreased explosion energy,
AD20x1
AD20x2
Ib
Ic
peak luminosity and timescale for all the light curves
1043 +0.5(M, Ebind )r>rext Ic-BL become smaller and longer, respectively.
For rapidly rotating model AD20x2, the timescale of
Lbol (erg/s)

the light curve is longer than those typically found for

1042
type Ib/c supernovae. Karamehmetoglu et al. (2022)
reported such long-timescale type Ib/c supernovae re-
1041 cently. The long-timescale supernovae are reported as
the explosions of massive (MZAMS ≳ 25M⊙ ) stars with
the explosion energies comparable to typical type Ib/c
1040 supernovae, but they are infered to have a larger amount
−20 0 20 40 60 80
t − tpeak (day) (≳ 0.1M⊙ ) of 56 Ni mass. These facts indicate that our
rapidly rotating model can explain such a subclass of
Figure 7. Bolometric light curve models. The time ori-
type Ib/c supernovae.
gin is chosen to be the peak time for each curve. The solid
curves denote the light curves with ejecta mass and explosion For the moderately rotating models AD09x1 and
energy extracted for rext = 2 × 109 cm (see Table 2 for the AD20x1, on the other hand, the timescales become ≈ 13
values). The dashed curves denote the ones with the assump- and 28 d, respectively, comparable to or longer than that
tion that a half of the mass and binding energy outside the of typical type Ib/c supernovae. The peak luminosity
extraction radius contribute to the ejecta properties. The is, however, about 10 times dimmer than that of typical
point shows the time at which the ejecta becomes optically type Ib/c supernovae. This indicates that rotating mas-
thin, τ = κρR = 1, to thermal photons for each model. The
sive stars exploded by outflows from black hole-disk sys-
shaded regions denote the templates of the bolometric light
curves with standard deviations for type Ib, Ic, and Ic-BL
tems may produce a variety of transients depending on
supernovae taken from Lyman et al. (2016). the rotation profile of the progenitors and the presence
of the stellar envelope, although the explosion mecha-
nism is qualitatively universal.
rapidly rotating model AD20x2, light trans-iron elements
As found in the comparison of the solid and dashed
could be synthesized in the ejecta (see § 4.2), and hence,
curves in Fig. 7, the features of the bolometric light
the opacity for optical wavelengths may be higher than
curves depend on the possible contribution of the mat-
0.1 cm2 /g. For more quantitative study, we obviously
ter outside the extraction radius. The quantitative pre-
need a radiation transfer simulation for photons taking
diction of the optical transients requires us to perform
into account a realistic opacity table.
simulations for entire stars until the outer layer of the
Figure 7 shows the bolometric light curves for all the
star is swept by the shock wave.
models investigated in this paper. For deriving the
Another possible astrophysical transients can be pow-
solid curves of Fig. 7, we take into account only the
ered by the interaction of the ejecta with a circum-stellar
ejecta mass and explosion energy extracted at rext =
medium that can result from the strong mass loss of their
2×109 cm. For models AD09x1 and AD20x1, the luminos-
progenitor prior to the stellar core collapse. The progen-
ity evolves rapidly. The rise times, defined by the time
itor models provided by Aguilera-Dena et al. (2018) are
until the maximum luminosity is reached, are trise ≈ 3.3
likely to be surrounded by a dense, massive (∼ 0.1–1M⊙
and 4.4 d, respectively, for these models (see Table 2).
within ∼ 1015 cm) circum-stellar medium at the core col-
Such fast transients may be discovered in the future
lapse. Since the ejecta mass is an order of 0.1–1M⊙ for
high-cadence transient surveys. On the other hand, for
models AD09x1 and AD20x1, the ejecta will be signifi-
model AD20x2, trise ≈ 10 d, which is consistent with that
cantly decelerated in the circum-stellar medium, releas-
of type Ib/c supernovae (see, e.g., Taddia et al. 2018).
ing a substantial fraction of its kinetic energy ≲ 1051 erg.
This indicates that there could be a possible subclass
The optical depth of the circum-stellar medium is esti-
of type Ib/c supernovae driven by the disk outflow in
mated as
the black hole-forming core collapses of rotating mas-  
sive stars. 3κMCSM κ
τCSM ∼ ≈ 170
To investigate the possible effects of the stellar mat- 4πRCSM 2 0.35 cm2 /g
ter outside the extraction radius, we also calculate the 
MCSM

RCSM
−2
light curve for each model assuming that a half of the × . (16)
1M⊙ 1015 cm
mass and binding energy of the matter for r > rext =
2 × 109 cm contribute to the ejecta mass and explosion Here, the opacity for photons is assumed to be dom-
energy (see the dashed curves in Fig. 7). Because of the inated by the Thomson scattering of fully ionized
 Collapsar leaving BH-disk 15

medium with Ye = 0.875 (i.e., hydrogen and helium 100

AD09x1
with mass fractions 0.75 and 0.25, respectively). Since −1
AD20x1
AD20x2
10
the circum-stellar medium is optically thick, the released BHdisk

energy diffuses out from the circum-stellar medium with 10−2

the diffusion time. The luminosity is then estimated as
10−3
ϵEexp RCSM

X(A)
L∼ ·
tdiff v∞ tdiff 10−4
 −1
κ
≈ 5.5 × 1043 erg/s 10−5
0.35 cm2 /g
  2  
ϵ v∞ RCSM 10−6
× , (17)
0.1 1 × 109 cm/s 1015 cm
10−7
p 0 20 40 60 80 100
where v∞ = 2Eexp /(Mej + MCSM ) is the terminal ve- A
locity of the ejecta plus circum-stellar medium, ϵ is the
radiation efficiency, and Figure 8. Isobaric mass fraction for our models. Nuclei at
s A = 12 and 16 are, respectively, predominantly unprocessed
12
κ(Mej + MCSM ) C and 16 O in the progenitor stars.
tdiff ∼
4πv∞ c
 1/2 to be synthesized for a rapidly rotating model AD20x2
κ (as well as BHdisk). It is known that in slightly neutron
≈ 26 d
0.35 cm2 /g rich (Ye ≳ 0.4) ejecta, such trans-iron nuclei are synthe-
3/4  −1/4
sized predominantly in quasi-nuclear statistical equilib-

Mej + MCSM Eexp
× , (18) rium (QSE, Meyer et al. 1998; Wanajo et al. 2018) un-
2M⊙ 1051 erg
der an alpha-rich condition. Since the neutron-richness
is the diffusion time of the expanding ejecta plus circum- is not very high, heavy r-process elements with A > 100
stellar medium (see, e.g., Matsumoto & Metzger 2022 for are not synthesized for any of the present models.
similar expression). The second factor of the first line The first peak nuclei of r-process, especially Zr and Y
in Eq. (17) is the contribution of the adiabatic cooling. (synthesized in QSE here), are known to have opacities
This can naturally lead to an optical transient like su- higher than those of iron-group elements (Kawaguchi
perluminous supernovae (see, e.g., Moriya et al. 2018 for et al. 2021). Therefore, if such elements are appreciably
a review). We note, however, that the properties of the synthesized, the resulting optical transients may have
transient depend not only on the mass and radius but longer timescales than those without such elements. In
also on the density profile of the circum-stellar medium addition, the peak luminosity will be lower and the spec-
(see, e.g., Chevalier & Irwin 2011 and Suzuki et al. 2020 trum could be redder. To quantify the light curve and
for the curcum-stellar medium like a stationary wind). spectrum, a radiation transfer simulation is needed in
future work.
4.2. Possible synthesis of light trans-iron nuclei
Figure 8 shows isobaric mass fractions obtained by 4.3. Implications for gamma-ray bursts
the nucleosynthesis calculations for our models. We find Figure 9 shows the masses (top) and dimensionless
prominent peaks at alpha-nuclei (with A multiple of 4) spins (bottom) of the black holes for our models as func-
and A = 56. The peaks at A = 12 and 16 reflect the tions of post-bounce time. Here, the mass and dimen-
initial composition of 12 C and 16 O, respectively, in pro- sionless spin of the black holes are estimated from the
genitor stars. The bulk of these nuclei remains unpro- equatorial and polar circumference radii of the apparent
cessed owing to relatively low temperature achieved. On horizon (e.g., Shibata 2016).
the other hand, the nuclei at peaks of A = 20–40 (cor- For models AD09x1 and AD20x1, in which we use the
responding to elements Ne, Mg, Si, S, Ar, and Ca) are original angular momentum distribution of the stellar
synthesized from 12 C and 16 O in the ejecta that expe- evolution simulations, the mass accretion and thus the
rience higher temperature but lower than that required spin up of the black hole are suppressed because of the
for achieving NSE. The peak at A = 56 corresponds to late-time disk formation and the quicker launch of the
56
Ni, which is synthesized predominantly in NSE with outflow from the disk. The dimensionless spins for these
Ye ≳ 0.5. Interestingly, a certain amount of nuclei heav- models are ∼ 0.4 and 0.7 respectively. For these models,
ier than the iron group (A ≳ 60), up to A ≈ 90, is found the Blandford-Znajek mechanism (Blandford & Znajek
16 Fujibayashi et al.

11 it is also found that the rapidly rotating models show a

10 large average velocity of the ejecta ≳ 2×109 cm/s, which
9
8 is necessary for the broad-line features for these super-
MBH/M

7 novae. In addition, the large explosion energy ≳ 1052 erg

6
5 AD09x1
AD20x1
observationally inferred from type Ic-BL supernovae is
4
3 AD20x2 likely achieved if we consider longer evolution of the sys-
1.0 tem than simulated in this study. Therefore, the rapidly
0.8 rotating models in this study may reasonably represent
0.6 the supernovae accompanying long-duration gamma-ray
χBH

0.4
bursts.

0.2
4.4. Possible effects of relativistic jet
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
tpb (s) We briefly discuss possible effects of a relativistic
jet that may be launched in the polar direction by
Figure 9. Masses (top) and dimensionless spins (bottom) some mechanisms, accompanied with the formation of
of black holes for our models as functions of post-bounce a rapidly rotating black hole. Here, we suppose that the
time.
jet is launched during the phase of a high mass accretion
rate onto the black hole.
1977), which is one of the most promising mechanisms
If a jet is powerful enough, it may partly prevent the
to power gamma-ray bursts (e.g., Gottlieb et al. 2022a),
infall of the stellar matter (Tominaga 2009). If so, the
may provide only moderately large Poynting luminosity
ram pressure of the infalling matter decreases, and as
for launching intense electromagnetic waves because of
a result, the outflow from the disk surface is launched
its strong dependence on the black-hole spin. Moreover,
earlier. The decrease of the matter infall also prevents
because of the low density of the disk for these models,
the matter supply to the disk, and thus, the total energy
strong magnetic fields may not be sustained in the vicin-
for the outflow may be reduced.
ity of the black hole. This indicates that more rapidly
Another possible effect of the relativistic jet is that
rotating progenitors than those of AD09 and AD20 would
the energy injection by the jet can be a source of 56 Ni
be preferred for generating powerful relativistic jets.
production (see, e.g., Tominaga et al. 2007, Barnes et al.
For the rapidly rotating model AD20x2, the mass ac-
2018, and Leung et al. 2023). If the jet luminosity is high
cretion from the disk onto the black hole continues for
enough (≳ 1053 erg/s; Tominaga et al. 2007), a signifi-
a long timescale (≳ 20 s) because of the presence of the
cant amount of 56 Ni (≳ 0.1M⊙ ) may be synthesized and
long-term high-mass-accretion rate phase. This leads to
the optical transient may become more luminous. The
a rapidly spinning black hole, the dimensionless spin of
energy injection by the jet may also modify the mor-
which is ∼ 0.8 at the termination of the simulations.
phology of the ejecta, which can affect the features of
During the long-term evolution of the black hole, the
the optical transient.
accretion disk is likely to be in a turbulent state associ-
ated with magnetohydrodynamical instabitlities such as 5. SUMMARY
the MRI (Balbus & Hawley 1991) in a realistic situation.
By this, the magnetic-field strength should be enhanced, In this paper, we studied the explosion in the rotating
and the magnetic flux penetrating the black hole is in- massive-star collapse leading to a black hole and a mas-
creased as a result of the mass accretion (and magnetic sive disk in fully general relativistic radiation-viscous-
flux accretion) onto the black hole. Such a highly spin- hydrodynamics simulations with an approximate neu-
ning black hole penetrated by a strong magnetic field trino radiation transfer, employing evolved stars with a
could be a promising central engine for relativistic jet compact core (Aguilera-Dena et al. 2020) as the initial
via the Blandford-Znajek process (e.g., Christie et al. conditions. We adopt the original or doubled angular
(2019); Hayashi et al. (2022) for a related topic). velocity for models AD09 and AD20. For all the models
It is known that at least a fraction of long-duration investigated in this paper, we found the formation of an
gamma-ray bursts are accompanied by type Ic-BL super- accretion disk after a proto-neutron star collapsed into
novae (e.g., Cano et al. 2017). As found in § 3.6, rapidly a black hole although the time at the onset of the disk
rotating models synthesize a large amount of 56 Ni (> formation depends strongly on the rotational profile of
0.1M⊙ ), which is consistent with required amount to ex- the progenitor stars. The evolution after the disk for-
plain the light curves of type Ic-BL supernovae. In § 3.4, mation was also found to depend strongly on the degree
of rotation of the progenitor stars.
 Collapsar leaving BH-disk 17

For moderately rotating models, AD09x1 and AD20x1, there might be a possible subclass of bright stripped-
for which the original angular momentum profiles ob- envelope supernovae driven by the outflow from a mas-
tained in stellar evolution calculations are employed, sive disk around a rapidly spinning black hole formed
the rest-mass density and the neutrino cooling efficiency from the collapse of a massive rotating star. The pos-
were already low at the disk formation. For these mod- sible existence of high-opacity trans-iron elements (such
els, the viscous heating efficiency is always higher than as Y and Zr) may lead to a longer timescale and redder
the neutrino cooling one after the disk formation, and transient.
thus, no NDAF phase is established. As a result, the Depending on the effect of the mass in the outer layer
outflow is launched before a massive disk is formed, and of the star, the resulting optical transient can have a
the explosion occurs at several hundreds of millisecond very short timescale (a few days) or that comparable to
after the disk formation. Because the disk starts forming normal supernovae. If dense and massive circum-stellar
at a late stage of the stellar collapse (at tpb > 10 s), the media are present as predicted in Aguilera-Dena et al.
mass of the envelop is relatively small, and hence, the en- (2018), a very bright (∼ 1044 erg/s) transient with a
ergy budget is small. As a consequence, the ejecta mass timescale of months is expected due to the interaction
and explosion energy are relatively small as ∼ 0.1M⊙ of ejecta with the circum-stellar medium.
and ≈ (0.5–1) ×1051 erg, respectively. The electron frac- To more rigorously predict observational features
tion of the ejecta is always higher than 0.47, and thus, an (photometric luminosity and spectra) of optical tran-
r-process nucleosynthesis cannot proceed in the ejecta sients based on our present results, we need to perform
for these models. a photon-radiation transfer simulation. The inclusion
For the model with a rapidly rotating progenitor of high-opacity trans-iron elements, which could exist
AD20x2, the mass infall rate to the disk and the black in the ejecta, may drastically change the observational
hole are high (Ṁfall ≳ 0.3M⊙ /s). As a result, the disk feature, which will be investigated in our future work.
settles to an NDAF phase and evolves quasi-steadily We employed viscous hydrodynamics to incorporate
prior to the onset of the outflow, which starts ∼ 10 s angular momentum transport and dissipation of kinetic
after the disk formation. The outflow is launched from energy to internal energy in the region in which a veloc-
the surface of the disk after the mass infall rate decreases ity shear or differential rotation is present. This enables
(i.e., the ram pressure of the infalling matter drops). At us to approximately capture the effective viscosity in-
the launch of the outflow, the neutrino cooling efficiency duced by the magnetohydrodynamical turbulence. How-
around the disk surface is lower than that deeper in the ever, it is obviously necessary to perform first-principle
disk. The ejecta mass and explosion energy amount to magnetohydrodynamics simulations in order to strictly
≳ 1M⊙ and ≳ 3 × 1051 erg, respectively. The electron explore the effects of the angular momentum trans-
fraction of the ejecta is at lowest ≈ 0.4, which is still port and (effectively) viscous dissipation. Thus, three-
not sufficient for an r-process nucleosynthesis. Indeed, dimensional radiation-magnetohydrodynamics simula-
the nucleosynthesis calculation shows that heavy nuclei tion is necessary in future work.
are synthesized at most up to A ≈ 100. However, there A missing but potentially important ingredient of the
is low-electron fraction (Ye < 0.2) matter deep inside scenario presented in this work is the possible existence
the disk in which the rest-mass density is high enough of a relativistic jet. The disk evolution may be affected
to enhance the electron degeneracy. If such components by this because the history of mass supply is modified by
were ejected by a very efficient mass ejection process the feedback of the jet. The relativistic jet, if powerful
with a shorter timescale, e.g., by magnetohydrodynam- enough, can also synthesize a significant amount of 56 Ni,
ics processes, r-process elements might be synthesized. which makes the optical transient more luminous. It
Magnetohydrodynamics simulation for this problem is can also modify the ejecta morphology, and as a result,
one of our future issues. may affect the features of the optical transient. These
For moderately rotating models, the synthesized 56 Ni possible effects are also the issues to be investigated in
mass is less than 0.1M⊙ , and the luminosity of the our future work.
supernova-like explosion is inferred to be comparable
to those of the ordinary supernovae. By contrast, for
the rapidly rotating model AD20x2, the synthesized 56 Ni
mass is larger than 0.1M⊙ , and hence, a luminous
supernova-like explosion may be expected. The bolo-
metric light curve for this model is suitable for a light-
curve model of type Ib/c supernovae. This suggests that
18 Fujibayashi et al.
1.0 1054
T20 T20 T20 Viscous heating
1018 jISCO Ṁfall Lν
ṀBH 1053
0.8

Ṁfall, ṀBH (M /s)

1017 1052

Lν (erg/s)
j (cm2/s)

0.6
1051
1016
0.4
1050
1015 0.2
1049

14
10 0.0 1048
0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24
Mr /M tpb (s) tpb (s)
(a) (b) (c)
100
T20 T20 T20

10 −1 1.50

10−1 1.25

∆M/Mej,Tmax>5 GK
10−2
1.00
Lν /ṀBHc2

Mej (M )
10−3 10−2 0.75
10−4 0.50
10−3
10−5 0.25

0.00
10−6 10−4
0 2 4 6 8 10 12 14 16 18 20 22 24 0.35 0.40 0.45 0.50 0.55 0.60 0 1 2 3 4 5 6 7 8
tpb (s) Ye (T = 5 GK) t − texp (s)
(d) (e) (f)
100
T20

10−1
T20 κ = 0.1 cm2 /g T20 Ib
+0.5(M, Ebind )r>rext Ic
10−2
κρR = 1 Ic-BL
3 1043
10−3
Eexp (1051 erg)

X(A)
Lbol (erg/s)

2 1042 10−4

10−5
1 1041
10−6

0
1040 10−7
0 1 2 3 4 5 6 7 8 −20 0 20 40 60 80 0 20 40 60 80 100
t − texp (s) t − tpeak (day) A
(g) (i) (h)
Figure 10. The same figures as Figs. 1, 2, 5–8 but for model T20.

ACKNOWLEDGMENTS
We thank Kunihito Ioka, Keiichi Maeda (at Kyoto
University), and Takashi Moriya for helpful discussions
and Koh Takahashi and David Aguilera-Dena for pro-
viding their stellar evolution models. Numerical compu-
tation was performed on Sakura, Cobra, and Raven clus-
ters at Max Planck Computing and Data Facility. This
work was in part supported by Grant-in-Aid for Scien-
tific Research (grant Nos. 20H00158 and 23H04900) of
Japanese MEXT/JSPS.

APPENDIX

A. A HIGH MASS SUPPLY RATE CASE

We here present an additional result for a model in which we assume a hypothetically rapid rotation. The motivation
to consider this additional model is (i) to explore the case for which a disk is formed in a short timescale after the
formation of a black hole and (ii) to understand the evolution process of the system in which the matter infalling rate
onto the disk is very high.
For these purposes, we employ the star evolved from a helium star with the initial mass of MHe = 20M⊙ , which is
taken from Takahashi et al. (2018) (hereafter referred to as model T20). This progenitor star also has a very compact
 Collapsar leaving BH-disk 19

Figure 11. Density (top-left), neutrino cooling rate density (bottom-left), cooling efficiency (top-right), and viscous heating
rate density (bottom-right) for model T20 at tbp ≈ 11.6 (left panel) and 15.4 s (right panel), respectively.

core with ξ2.5 = 0.74. Because the progenitor is non-rotating, we add an artificial rotation profile (although this is
not self-consistent to the stellar model but we attempt to mimic the profile of more realistic models of Aguilera-Dena
et al. 2020) as

1
Ω(r) = , (A1)
1/Ω0 + 1/ΩrK =r0 (r)

where Ω0 = 0.1 rad/s and

p
Gm(r) · r0
ΩrK =r0 (r) = (A2)
r2

is the angular velocity for which the fluid element settles into a Keplerian motion with the cylindrical radius of r0 .
Note that the shellular rotation is assumed and the initial angular velocity is given as a function only of the spherical
radius r. In this study, we set r0 = 1000 km. This implies that the maximum radius at which the infalling matter is
circularized is 1000 km. The panel (a) of Fig. 10 shows the specific angular momentum profile of this model. From
the comparison between the profile and jISCO , the disk is expected to be formed when the black hole mass becomes
≈ 4M⊙ . This mass is smaller than those for the other models shown in this paper. In this model, we intentionally
prepare a hypothetically rapidly rotating progenitor star to investigate the effect of earlier disk formation around a
formed black hole on the subsequent evolution of the system. A soft EOS, SFHo (Steiner et al. 2013), is employed to
accelerate the black hole formation. The grid structure is similar to those of the other models, but with ∆x0 = 150 m,
δ = 0.01, and N = 991, which results in L = 1.0 × 1010 cm.
As in model AD20x2, for this model, the disk evolves in a quasi-steady manner for ≈ 10 s after the vertical expansion
of the disk. Thus, the evolution process of the disk is similar to that of model AD20x2; the explosion also sets in at
∼ 10 s after the vertical expansion of the disk. In the quasi-steady phase, however, the disk mass increases with a
rate Ṁdisk = Ṁfall − ṀBH ≈ 0.2M⊙ /s by the matter infall (see panel (b) of Fig. 10), which is much higher than other
models. This high-mass infall rate stems from the earlier formation of the disk and characterizes this model.
The panel (c) of Fig. 10 shows that the neutrino luminosity is slightly higher than the viscous heating rate. This situ-
ation is possible because of the presence of an additional heating by the infalling matter, which releases its gravitational
20 Fujibayashi et al.

binding energy to settle into a nearly Keplerian rotation, with the rate of

GMBH Ṁdisk
Linfall ∼
2rdisk
   −1
51 MBH Ṁdisk rdisk
≈ 3 × 10 erg/s , (A3)
6M⊙ 0.2M⊙ 500 km

where the infalling matter is assumed to settle into a Keplerian motion at a radius rdisk . Interestingly, the neutrino
luminosity is approximately constant in time for the quasi-steady NDAF phase, 4 s ≲ tpb ≲ 14 s.
After the infalling rate Ṁfall starts decreasing at tpb ≳ 14 s, the viscosity-driven outflow is launched because the
viscous heating plus shock heating dominate over the neutrino cooling on the surface region of the disk. Figure 11
shows the rest-mass density, local neutrino cooling efficiency, neutrino cooling rate, and viscous heating rate before
(tpb = 11.6 s: left) and after (tpb = 15.4 s: right) the onset of the outflow. Here, the local neutrino cooling efficiency
+ +
is defined by qν− /(qvis + qν+ ), where qν− and qν+ are neutrino cooling and heating rate densities and qvis is the viscous
heating rate density (see Fujibayashi et al. 2018 for its definition). The balance between the neutrino cooling and
the viscous heating is established in the mid-plane region (white regions) at both phases. For a given cylindrical
radius, this region extends over z ≲ H, where H := cs /Ω is the disk scale height evaluated at z = 0. The scale
height is found to be 0.2–0.3R, where R is the cylindrical radius. In this region, the timescale of weak interaction,
≈ 0.03 s(kB T /2 MeV)−5 , is much shorter than the timescale of radial motion, r/v r ≈ 1–10 s. This indicates that the
region may be well modeled by the NDAF solution. On the surface region of the disk, which extends over z ≳ H, a
velocity shear between the disk matter and the infalling matter is clearly present and produces a high viscous heating
rate (see the bottom-right plot in each panel). In addition, it is found that the neutrino cooling is not efficient on the
disk surface region, z ∼ 2H ≈ 0.5R, because of the low temperature there. This results in a launch of the outflow
from the surface region of the disk once the ram pressure of the infalling matter becomes weak.
For model T20, the lowest value of Ye for the ejecta is ≈ 0.46 (see panel (e) of Fig. 10), which is higher than that for
model AD20x2, although the disk is more massive and Ye at the mid-plane is lower for model T20. This is because, as
mentioned above, the outflow is launched mainly from the surface region, which is the only place the neutrino cooling
is not efficient (see Fig. 11). This results in the ejection of the matter with a weak electron degeneracy, and thus, with
a high-Ye .
For model T20, a disk with mass of ≈ 2M⊙ still remains at the termination of our simulations and the bulk of
the disk is still cooled by neutrino emission efficiently. When the cooling efficiency drops, a stronger outflow may be
launched (e.g., Fernández & Metzger 2013; Just et al. 2015; Fujibayashi et al. 2020a; Just et al. 2022b). Suppose that
10% of the disk matter becomes ejecta eventually, the matter with mass of O(0.1)M⊙ becomes the ejecta in a later
phase. Since the outflow matter experiences a condition with temperature higher than 5 GK, a non-negligible amount
of 56 Ni may be synthesized in the later-time ejecta of this model. Also, the explosion energy at the termination of the
simulation reaches ≳ 3 × 1051 erg, and is still increasing (see panel (g) of Fig. 10). Since the timescale of the energy
increase appears long, we stopped the simulation by t = 30 s. Thus, the ejecta mass and explosion energy which
we describe below give the lower limit. Therefore, this model would represent more energetic transients than typical
supernovae such as type Ic-BL (see also explosion energy for the similar scenario found in Just et al. 2022a)
For this model, the mass of 56 Ni is found to be ≈ 0.56M⊙ , reflecting the large mass of the ejecta that experiences
high temperature ≳ 5 GK. This amount of 56 Ni is high enough for explaining a typical 56 Ni mass inferred from the
observations of type Ic-BL supernovae ≈ 0.4M⊙ (see, e.g., Cano et al. 2017 and Anderson 2019). These results suggest
that very massive and rapidly rotating stars leading to black hole plus massive disk formation are candidates for the
progenitors of type Ic-BL supernovae. We also note that, since we stopped the simulation at a time when the ejecta
mass and explosion energy are still increasing, it is the lower bound for the 56 Ni mass.

REFERENCES
Afsariardchi, N., Drout, M. R., Khatami, D. K., et al. 2021, Aguilera-Dena, D. R., Langer, N., Moriya, T. J., &
ApJ, 918, 89, doi: 10.3847/1538-4357/ac0aeb
Schootemeijer, A. 2018, ApJ, 858, 115,
Aguilera-Dena, D. R., Langer, N., Antoniadis, J., & Müller,
doi: 10.3847/1538-4357/aabfc1
B. 2020, ApJ, 901, 114, doi: 10.3847/1538-4357/abb138
 Collapsar leaving BH-disk 21

Alcubierre, M., Brügmann, B., Holz, D., et al. 2001, Fernández, R., & Metzger, B. D. 2013, MNRAS, 435, 502,
International Journal of Modern Physics D, 10, 273, doi: 10.1093/mnras/stt1312
doi: 10.1142/S0218271801000834 Frank, J., King, A., & Raine, D. J. 2002, Accretion Power
Aloy, M. Á., & Obergaulinger, M. 2021, MNRAS, 500, in Astrophysics: Third Edition
4365, doi: 10.1093/mnras/staa3273 Fujibayashi, S., Kiuchi, K., Nishimura, N., Sekiguchi, Y., &
Anderson, J. P. 2019, A&A, 628, A7, Shibata, M. 2018, ApJ, 860, 64,
doi: 10.1051/0004-6361/201935027 doi: 10.3847/1538-4357/aabafd
Arcones, A., Martı́nez-Pinedo, G., Roberts, L. F., & Fujibayashi, S., Kiuchi, K., Wanajo, S., et al. 2023, ApJ,
Woosley, S. E. 2010, A&A, 522, A25, 942, 39, doi: 10.3847/1538-4357/ac9ce0
doi: 10.1051/0004-6361/201014276
Fujibayashi, S., Sekiguchi, Y., Kiuchi, K., & Shibata, M.
Arnett, W. D. 1982, ApJ, 253, 785, doi: 10.1086/159681
2017, ApJ, 846, 114, doi: 10.3847/1538-4357/aa8039
Balbus, S. A., & Hawley, J. F. 1991, ApJ, 376, 214,
Fujibayashi, S., Shibata, M., Wanajo, S., et al. 2020a,
doi: 10.1086/170270
PhRvD, 101, 083029, doi: 10.1103/PhysRevD.101.083029
Balbus, S. A., & Hawley, J. F. 1998, Rev. Mod. Phys., 70,
—. 2020b, PhRvD, 102, 123014,
1, doi: 10.1103/RevModPhys.70.1
doi: 10.1103/PhysRevD.102.123014
Banik, S., Hempel, M., & Bandyopadhyay, D. 2014, ApJS,
214, 22, doi: 10.1088/0067-0049/214/2/22 Fujibayashi, S., Takahashi, K., Sekiguchi, Y., & Shibata, M.
Barnes, J., Duffell, P. C., Liu, Y., et al. 2018, ApJ, 860, 38, 2021, ApJ, 919, 80, doi: 10.3847/1538-4357/ac10cb
doi: 10.3847/1538-4357/aabf84 Fujibayashi, S., Wanajo, S., Kiuchi, K., et al. 2020c, ApJ,
Baumgarte, T. W., & Shapiro, S. L. 1998, ApJ, 504, 431, 901, 122, doi: 10.3847/1538-4357/abafc2
doi: 10.1086/306067 Gottlieb, O., Lalakos, A., Bromberg, O., Liska, M., &
Beloborodov, A. M. 2003, ApJ, 588, 931, Tchekhovskoy, A. 2022a, MNRAS, 510, 4962,
doi: 10.1086/374217 doi: 10.1093/mnras/stab3784
Blandford, R. D., & Znajek, R. L. 1977, MNRAS, 179, 433, Gottlieb, O., Tchekhovskoy, A., & Margutti, R. 2022b,
doi: 10.1093/mnras/179.3.433 MNRAS, 513, 3810, doi: 10.1093/mnras/stac910
Bollig, R., Yadav, N., Kresse, D., et al. 2020, arXiv e-prints, Hawley, J. F., Richers, S. A., Guan, X., & Krolik, J. H.
arXiv:2010.10506. https://arxiv.org/abs/2010.10506 2013, Astrophys. J., 772, 102,
Burrows, A., Radice, D., & Vartanyan, D. 2019, MNRAS, doi: 10.1088/0004-637X/772/2/102
485, 3153, doi: 10.1093/mnras/stz543 Hayakawa, T., & Maeda, K. 2018, ApJ, 854, 43,
Cano, Z., Wang, S.-Q., Dai, Z.-G., & Wu, X.-F. 2017, doi: 10.3847/1538-4357/aaa76c
Advances in Astronomy, 2017, 8929054, Hayashi, K., Fujibayashi, S., Kiuchi, K., et al. 2022,
doi: 10.1155/2017/8929054 PhRvD, 106, 023008, doi: 10.1103/PhysRevD.106.023008
Chevalier, R. A., & Irwin, C. M. 2011, ApJL, 729, L6,
Held, L. E., & Mamatsashvili, G. 2022, Mon. Not. Roy.
doi: 10.1088/2041-8205/729/1/L6
Astron. Soc., 517, 2309, doi: 10.1093/mnras/stac2656
Christie, I. M., Lalakos, A., Tchekhovskoy, A., et al. 2019,
Hilditch, D., Bernuzzi, S., Thierfelder, M., et al. 2013,
MNRAS, 490, 4811, doi: 10.1093/mnras/stz2552
PhRvD, 88, 084057, doi: 10.1103/PhysRevD.88.084057
Colgate, S. A., Fryer, C. L., & Hand, K. P. 1997, in NATO
Israel, W., & Stewart, J. M. 1979, Annals of Physics, 118,
Advanced Study Institute (ASI) Series C, Vol. 486,
341, doi: 10.1016/0003-4916(79)90130-1
Thermonuclear Supernovae, ed. P. Ruiz-Lapuente,
R. Canal, & J. Isern, 273, Janka, H.-T., Hanke, F., Hüdepohl, L., et al. 2012, Progress
doi: 10.1007/978-94-011-5710-0_19 of Theoretical and Experimental Physics, 2012, 01A309,
Dessart, L., Hillier, D. J., Woosley, S., et al. 2015, MNRAS, doi: 10.1093/ptep/pts067
453, 2189, doi: 10.1093/mnras/stv1747 Just, O., Aloy, M. A., Obergaulinger, M., & Nagataki, S.
—. 2016, MNRAS, 458, 1618, doi: 10.1093/mnras/stw418 2022a, ApJL, 934, L30, doi: 10.3847/2041-8213/ac83a1
Di Matteo, T., Perna, R., & Narayan, R. 2002, Astrophys. Just, O., Bauswein, A., Ardevol Pulpillo, R., Goriely, S., &
J., 579, 706, doi: 10.1086/342832 Janka, H. T. 2015, MNRAS, 448, 541,
Drout, M. R., Chornock, R., Soderberg, A. M., et al. 2014, doi: 10.1093/mnras/stv009
ApJ, 794, 23, doi: 10.1088/0004-637X/794/1/23 Just, O., Goriely, S., Janka, H. T., Nagataki, S., &
Eisenberg, M., Gottlieb, O., & Nakar, E. 2022, MNRAS, Bauswein, A. 2022b, MNRAS, 509, 1377,
517, 582, doi: 10.1093/mnras/stac2184 doi: 10.1093/mnras/stab2861
22 Fujibayashi et al.

Karamehmetoglu, E., Sollerman, J., Taddia, F., et al. 2022, Prentice, S. J., Maguire, K., Smartt, S. J., et al. 2018,
arXiv e-prints, arXiv:2210.09402. ApJL, 865, L3, doi: 10.3847/2041-8213/aadd90
https://arxiv.org/abs/2210.09402 Pruet, J., Woosley, S. E., & Hoffman, R. D. 2003, ApJ,
Kawaguchi, K., Fujibayashi, S., Shibata, M., Tanaka, M., & 586, 1254, doi: 10.1086/367957
Wanajo, S. 2021, ApJ, 913, 100, Rembiasz, T., Obergaulinger, M., Cerdá-Durán, P., Müller,
doi: 10.3847/1538-4357/abf3bc E., & Aloy, M. A. 2016, MNRAS, 456, 3782,
Khatami, D. K., & Kasen, D. N. 2019, ApJ, 878, 56, doi: 10.1093/mnras/stv2917
doi: 10.3847/1538-4357/ab1f09 Rodrı́guez, Ó., Maoz, D., & Nakar, E. 2022, arXiv e-prints,
Kiuchi, K., Kyutoku, K., Sekiguchi, Y., & Shibata, M. 2018, arXiv:2209.05552. https://arxiv.org/abs/2209.05552
PhRvD, 97, 124039, doi: 10.1103/PhysRevD.97.124039 Sekiguchi, Y. 2010, Progress of Theoretical Physics, 124,
Kohri, K., & Mineshige, S. 2002, Astrophys. J., 577, 311, 331, doi: 10.1143/PTP.124.331
doi: 10.1086/342166 Sekiguchi, Y., & Shibata, M. 2011, ApJ, 737, 6,
Kohri, K., Narayan, R., & Piran, T. 2005, ApJ, 629, 341, doi: 10.1088/0004-637X/737/1/6
doi: 10.1086/431354 Shakura, N. I., & Sunyaev, R. A. 1973, A&A, 500, 33
Leung, S.-C., Nomoto, K., & Suzuki, T. 2023, arXiv Shi, J.-M., Stone, J. M., & Huang, C. X. 2016, Mon. Not.
e-prints, arXiv:2304.14935, Roy. Astron. Soc., 456, 2273,
doi: 10.48550/arXiv.2304.14935 doi: 10.1093/mnras/stv2815
Lyman, J. D., Bersier, D., James, P. A., et al. 2016, Shibata, M. 2000, Progress of Theoretical Physics, 104, 325,
MNRAS, 457, 328, doi: 10.1093/mnras/stv2983 doi: 10.1143/PTP.104.325
MacFadyen, A. I., & Woosley, S. E. 1999, ApJ, 524, 262, Shibata, M. 2007, Phys. Rev. D, 76, 064035,
doi: 10.1086/307790 doi: 10.1103/PhysRevD.76.064035
Margutti, R., Metzger, B. D., Chornock, R., et al. 2019, Shibata, M. 2016, Numerical Relativity, doi: 10.1142/9692
ApJ, 872, 18, doi: 10.3847/1538-4357/aafa01 Shibata, M., Kiuchi, K., Sekiguchi, Y., & Suwa, Y. 2011,
Matsumoto, T., & Metzger, B. D. 2022, ApJ, 936, 114, Progress of Theoretical Physics, 125, 1255,
doi: 10.3847/1538-4357/ac892c doi: 10.1143/PTP.125.1255
Meyer, B. S., Krishnan, T. D., & Clayton, D. D. 1998, ApJ, Shibata, M., Kiuchi, K., & Sekiguchi, Y.-i. 2017, PhRvD,
498, 808, doi: 10.1086/305562 95, 083005, doi: 10.1103/PhysRevD.95.083005
Meza, N., & Anderson, J. P. 2020, A&A, 641, A177, Shibata, M., & Nakamura, T. 1995, PhRvD, 52, 5428,
doi: 10.1051/0004-6361/201937113 doi: 10.1103/PhysRevD.52.5428
Miller, J. M., Sprouse, T. M., Fryer, C. L., et al. 2020, ApJ, Siegel, D. M., Barnes, J., & Metzger, B. D. 2019, Nature,
902, 66, doi: 10.3847/1538-4357/abb4e3 569, 241, doi: 10.1038/s41586-019-1136-0
Moriya, T. J., Sorokina, E. I., & Chevalier, R. A. 2018, Steiner, A. W., Hempel, M., & Fischer, T. 2013, ApJ, 774,
SSRv, 214, 59, doi: 10.1007/s11214-018-0493-6 17, doi: 10.1088/0004-637X/774/1/17
Nagataki, S., Takahashi, R., Mizuta, A., & Takiwaki, T. Surman, R., McLaughlin, G. C., & Hix, W. R. 2006, ApJ,
2007, ApJ, 659, 512, doi: 10.1086/512057 643, 1057, doi: 10.1086/501116
Navó, G., Reichert, M., Obergaulinger, M., & Arcones, A. Suzuki, A., Moriya, T. J., & Takiwaki, T. 2020, ApJ, 899,
2022, arXiv e-prints, arXiv:2210.11848, 56, doi: 10.3847/1538-4357/aba0ba
doi: 10.48550/arXiv.2210.11848 Suzuki, T. K., & Inutsuka, S.-i. 2014, Astrophys. J., 784,
Obergaulinger, M., & Aloy, M. Á. 2020, MNRAS, 492, 121, doi: 10.1088/0004-637X/784/2/121
4613, doi: 10.1093/mnras/staa096 Taddia, F., Stritzinger, M. D., Bersten, M., et al. 2018,
—. 2021, MNRAS, 503, 4942, doi: 10.1093/mnras/stab295 A&A, 609, A136, doi: 10.1051/0004-6361/201730844
—. 2022, MNRAS, 512, 2489, doi: 10.1093/mnras/stac613 Takahashi, K., Yoshida, T., & Umeda, H. 2018, The
Obergaulinger, M., Aloy, M. A., & Müller, E. 2010, A&A, Astrophysical Journal, 857, 111,
515, A30, doi: 10.1051/0004-6361/200913386 doi: 10.3847/1538-4357/aab95f
O’Connor, E., & Ott, C. D. 2011, ApJ, 730, 70, Tampo, Y., Tanaka, M., Maeda, K., et al. 2020, ApJ, 894,
doi: 10.1088/0004-637X/730/2/70 27, doi: 10.3847/1538-4357/ab7ccc
Perley, D. A., Mazzali, P. A., Yan, L., et al. 2019, MNRAS, Thorne, K. S. 1981, MNRAS, 194, 439,
484, 1031, doi: 10.1093/mnras/sty3420 doi: 10.1093/mnras/194.2.439
Piran, T., Nakar, E., Mazzali, P., & Pian, E. 2019, ApJL, Tominaga, N. 2009, ApJ, 690, 526,
871, L25, doi: 10.3847/2041-8213/aaffce doi: 10.1088/0004-637X/690/1/526
 Collapsar leaving BH-disk 23

Tominaga, N., Maeda, K., Umeda, H., et al. 2007, ApJL, Woosley, S. E. 1993, ApJ, 405, 273, doi: 10.1086/172359
657, L77, doi: 10.1086/513193
Viganò, D., Aguilera-Miret, R., Carrasco, F., Miñano, B., & Zhang, W., MacFadyen, A., & Wang, P. 2009, ApJL, 692,
Palenzuela, C. 2020, PhRvD, 101, 123019,
doi: 10.1103/PhysRevD.101.123019 L40, doi: 10.1088/0004-637X/692/1/L40
Wanajo, S., Müller, B., Janka, H.-T., & Heger, A. 2018,
ApJ, 852, 40, doi: 10.3847/1538-4357/aa9d97