Relativistic Roche problem for stars in precessing orbits around a spinning black hole

Matteo Stockinger1 and Masaru Shibata1, 2

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
Tidal disruptions of stars on the equatorial plane orbiting Kerr black holes have been widely
studied. However thus far, there have been fewer studies of stars in inclined precessing orbits
around a Kerr black hole. In this paper, by using tensor virial equations, we show the presence
of possible resonances in these systems for typical physical parameters of black hole-neutron star
binaries in close orbits or of a white dwarf/an ordinary star orbiting a supermassive black hole. This
arXiv:2403.06834v1 [astro-ph.HE] 11 Mar 2024

suggests the presence of a new instability before the tidal disruption limit is encountered in such
systems.

I. INTRODUCTION equations adapted to our system, by which we derive

The classical Roche problem studies the equilibrium
and stability of a self-gravitating fluid in a binary. Its
main result is the presence of a critical orbital distance,
the so-called Roche limit, below which there is no stable
configuration, i.e., mass shedding and tidal disruption of
the star are induced [1, 2]. Extensions of these results to
eccentric orbits have also been the focus of some studies
[3, 4]. Furthermore, the concept of the Roche limit has
been extended to general relativistic scenarios, such as
when the binary companion is a black hole (BH) [5–8].
The previous studies for this focused on the star in cir-
cular equatorial orbits around the BH, and to our knowl-
edge, the only exception is [9].
In this paper, we study the case of a star in circular or-
bits on the non-equatorial plane around a spinning BH.
To study this, we assume that the star is incompress-
ible and utilize the tensor virial formalism developed by
Chandrasekhar [2]. This formalism has generally been
used to study the assumed ellipsoidal shape of a self-
gravitating fluid and provided a number of knowledge on
the equilibrium and stability of the stars.
After reviewing the previous results, we show the pres-
ence of a resonance when the orbit has an inclination with
respect to the equatorial plane. To compute the impact
of the inclination on the shape of the star, we use a per-
turbative expansion assuming that the inclination angle
is small: We perform this expansion up to the second or-
der in the angle of the inclined orbit ε = θ−π/2 and show
that the resulting equations take a linear form, which is
usually solvable. When this is not the case, we posit a
new kind of resonance for such systems. We show that
the resonances can play an important role in the evolu-
tion of BH-neutron star (NS) binaries in close orbits or of
a white dwarf/an ordinary star orbiting a supermassive
BH.
This paper is organized as follows: In Sec. II, we
present the working assumptions and the tensor virial
∗ matteo.stockinger@aei.mpg.de
the equilibrium solution of a star on an equatorial orbit
around a BH. In Sec. III, we present our method for the
perturbative expansion, which we apply up to the sec-
ond order in ε. In Sec. IV, we show the presence of new
resonances in these systems, using the equations derived
in Sec. III. Finally, we discuss the implications of our
findings in Sec. V. Throughout this paper, we use the
geometrical units of c = G = 1 where c and G are the
speed of light and gravitational constant, respectively.
II. FRAMEWORK OF THE SYSTEM
A. System
The purpose of this paper is to study the tidal effect
on a star orbiting a Kerr BH. We assume that the star
is modeled by an incompressible fluid of mass M and
constant density ρ. This star is assumed to have an
ellipsoidal shape determined by its principal semi-axes
(a1 , a2 , a3 ). We assume that it orbits a spinning BH
of mass m and of spin parameter a. When referencing
its position relative to the BH, we will use the Boyder-
Lindquist coordinates in which the line element is written
as
∆
ds2 = − (dt − a sin2 θdϕ)2 +
Σ
sin2 θ 2 Σ
[(r + a2 )dφ − adt]2 + dr2 + Σdθ2 , (1)
Σ ∆
where ∆ = r2 − 2mr + a2 and Σ = r2 + a2 cos2 θ. In the
following we assume that M ≪ m and do not consider the
self-gravitating effect of the stars on the orbital motion
around the BH. That is, we assume that the star has a
geodesic orbit around the BH.
To study the tidal effect on the stars by the tidal force
from the BH, we prepare Fermi-normal coordinates on
the geodesics [10]. The tidal force to the star by the
BH is evaluated on the local inertial frame on the Fermi-
normal coordinates. This allows us to treat the problem
in the similar way to Newtonian hydrodynamics.
 2
h
The geodesics on the Kerr spacetime are described by C23 = −3 − ar cos θ(S + T )Ia
the four constants of motion, E: specific energy, L: spe-
i√ ST
cific angular momentum for the BH spin direction, K:
+(a2 cos2 θS − r2 T )Ib sin Ψ, (11)
the Carter constant [11], and the mass of the star. The KΣ2
geodesic equations are written as (e.g., [12]) where
2 2 2 2 2 mr 2
dt [(r + a ) − ∆a sin θ]E − 2mraL Ia = (r − 3a2 cos2 θ), (12)
= (, 2) Σ3
dτ ∆Σ
ma cos θ
(3r2 − a2 cos2 θ),
 2
dr 2 Ib = (13)
Σ2 E(r2 + a2 ) − aL − ∆(r2 + K) Σ3

=
dτ
S = r2 +K, T = K −a2 cos2 θ, and Ψ is a time-dependent
:= R(r), (3)
 2 angle which obeys the following equation [10]
dθ (aE sin2 θ − L)2 √ 
Σ2 = K − a2 cos2 θ − (4) dΨ K E(r2 + a2 ) − aL L − aE sin2 θ

dτ sin2 θ = +a .
    dτ Σ r2 + K K − a2 cos2 θ
dφ 1 2mraE 2mr L
= + 1− , (5) (14)
dτ ∆ Σ Σ sin2 θ We note that higher-order corrections in Rstar /r0 for the
tidal tensors are found in [15]. One important aspect of
where τ is an affine parameter of the geodesics. In this
the tidal tensor is that C23 and C13 are non-zero only
paper, we will assume that the star has a circular orbit
in the presence of the BH spin and for θ ̸= π/2. This
with a fixed value of r = r0 ; R = 0 = dR/dr = 0 at
implies that for a star in precessing orbits around a Kerr
r = r0 . These relations give us the two relations among
BH, a qualitatively new tidal force, which is absent in
E, L, and K.
Newtonian gravity, is exerted.
In this paper, we consider precessing orbits with re-
The expression of the tidal tensor can be simplified by
spect to the equatorial plane, i.e., θ ̸= π/2. For a given
changing the frame we are working on. Indeed in a rotat-
maximum value of θ for the orbit, we then have an addi- ⃗ along the 3-axis with magnitude
tional relation among E, L, and K, and thus, these quan- ing frame of rotation Ω
tities are written as a function of r0 (see, e.g., [13, 14]). Ω = dΨ/dτ , the tidal tensor becomes [10]
In the inertial frame defined above, we refer to the  ST (r2 − a2 cos2 θ) 
axes of the corresponding orthonormal basis as 1, 2, and C̃11 = 1 − 3 Ia
KΣ2
3. We assume that the axis 3 points in the same direction ST
as the orbital angular momentum of the fluid or in other +6ar cos θ Ib , (15)
KΣ2
words, the direction of (∂/∂θ)µ . Note that our 2 and 3
axes agree with the 3 and 2 axes of [10], and that we C̃22 = Ia , (16)
changed the direction of our 2-axis. Assuming that the  r2 T 2 − a2 cos2 θS 2 
stellar radius, Rstar , is much smaller than the curvature C̃33 = 1 + 3 Ia
KΣ2
radius of the BH, r0 , one can write the tidal tensor as [10] ST
−6ar cos θ Ib , (17)
KΣ2
 ST (r2 − a2 cos2 θ) 
C11 = 1−3 2
cos2 Ψ Ia C̃12 = 0, (18)
KΣ h
ST C̃13 = 3 − ar cos θ(S + T )Ia
+6ar cos θ cos2 ΨIb , (6)
KΣ2 i √ST
2 2 2
 ST (r2 − a2 cos2 θ) 2
 +(a cos θS − r T )Ib , (19)
C22 = 1−3 sin Ψ Ia KΣ2
KΣ2
ST C̃23 = 0. (20)
+6ar cos θ sin2 ΨIb , (7)
KΣ2 In the following, all the analyses will be carried out in
 r2 T 2 − a2 cos2 θS 2  this frame.
C33 = 1+3 Ia
KΣ2
ST
−6ar cos θ I ,
2 b
(8) B. Tensor virial equations
h KΣ
C12 = −3 (a2 cos2 θ − r2 )Ia
As already stated in Sec. I, we employ the tensor virial
i ST equations [2, 16] to study the equilibrium state of stars
+2ar cos θIb cos Ψ sin Ψ, (9) orbiting a Kerr BH. The hydrodynamics equation for the
KΣ2
h fluid on the local inertial frame is written as
C13 = 3 − ar cos θ(S + T )Ia
3
i √ST ρ
dUi
=−
∂P
−ρ
∂ϕ
−ρ
X
Cil Xl , (21)
2 2 2
+(a cos θS − r T )Ib cos Ψ, (10) dτ ∂Xi ∂Xi
KΣ2 l=1
 3

where Xi = (X1 , X2 , X3 ) denote the coordinates in the tegral variables:

local inertial frame, Ui is the velocity field of the star, P Z
the pressure, and ϕ the Newtonian potential of the fluid, Iij = ρxi xj d3 x, (27)
which obeys the Poisson equation as V
Z 3 X
X 3
△ ϕ = 4πρ, (22) 3
2Tij = ρui uj d x = Qik Qjl Ikl , (28)
V k=1 l=1
where △ denotes the Laplacian in the flat space. We Z 3
X
assume that the self-gravity of the star is not so strong Jij = ρui xj d3 x = Qik Ikj , (29)
that the equation of motion is written as in Newtonian V k=1
manner. Z
As seen above, it is preferable to place ourselves in a Mij = − ρxi ∂j ϕd3 x, (30)
⃗ as this simplifies the expres- V
rotating frame of rotation Ω, Z
sion of the tidal tensor. In this case, we have Π = P d3 x. (31)
V

dui ∂P ∂ϕ
ρ =− −ρ + ρ(Ω2 xi − δi3 Ω2 x3 ) Then, Eq. (26) is written as
dτ ∂xi ∂xi
3 3 3
d
Jij = 2Tij + Πδij + Mij + Ω2 Iij − δi3 Ω2 I3j
X X X
+ 2ρ ϵil3 ul Ω + ρ ϵil3 xl Ω̇ − ρ C̃il xl , (23)
dτ
l=1 l=1 l=1 3 3 3
X X X
where xi and ui denote the coordinates and the velocity + 2Ω ϵil3 Jlj + ϵil3 Ω̇Ilj − C̃il Ilj . (32)
l=1 l=1 l=1
field of the star in the rotating frame, δij is the Kronecker
delta, ϵijk is the completely antisymmetric tensor in 3 From this, we can deduce the symmetrized version
dimension, and Ω̇ = dΩ/dτ .
We then assume that the pressure is given by 1 d2
Iij = 2Tij + Πδij + Mij
3
! 2 dτ 2
x2i
 
X
2 1
P = pc 1 − , (24) +Ω Iij − (δi3 I3j + δj3 I3i )
a2
i=1 i
2
3
X
where pc is the central pressure and ai (i = 1–3) denote +Ω (ϵil3 Jlj + ϵjl3 Jli )
the axial length of the star for the corresponding direc- l=1
tions. This setting enables us to assume that the velocity 3
1 X
depends linearly on the position as + Ω̇ (ϵil3 Ilj + ϵjl3 Ili )
2
l=1
3
X 3
ui = Qik xk , (25) 1 X
− (C̃il Ilj + C̃jl Ili ), (33)
k=1 2
l=1

where Qik is a matrix to be defined. and the antisymmetrized version of the tensor virial re-
We now follow the method developed by Chan- lation [16],
drasekhar to derive the second-rank tensor virial tensor
equations [2]. First, we multiply xj to Eq. (23) and in- d
tegrate over the volume V of the entire star. Then we (Jij − Jji ) = −Ω2 (δi3 I3j − δj3 I3i )
dτ
obtain X3
Z Z Z +2Ω (ϵil3 Jlj − ϵjl3 Jli )
dui 3 ∂P 3 ∂ϕ
ρ xj d x = − xj d x − ρ xj d3 x l=1
V dτ V ∂xi V ∂xi 3
Z X
+ ρ(Ω2 xi − δi3 Ω2 x3 )xj d3 x +Ω̇ (ϵil3 Ilj − ϵjl3 Ili )
V l=1
3
X Z 3
X Z 3
X
+ 2Ω ϵil3 ul xj d3 x + ϵil3 Ω̇ xl xj d3 x − (C̃il Ilj − C̃jl Ili ). (34)
l=1 V l=1 V l=1
3 Z
X
− ρC̃il xl xj d3 x. (26) Equation (34) is interpreted as the evolution equation of
l=1 V the angular momentum. Equation (32) will serve as our
basis for the exploration of the influence of the tidal force
To rewrite this equation, we introduce the following in- on stars orbiting a Kerr BH.
 4

C. Equatorial plane case In this paper, we pay attention to the cases of f = 0 and
1. For f = 0, the fluid has no internal motion in the
When the star orbits a BH on its equatorial plane, the rotating frame. In the inertial frame, the fluid’s inter-
equations become analytically solvable. Although this nal motion comes only from the rotation. On the other
case was already studied in [8], we will here restate the hand, for f = 1, the fluid has a certain vorticity in the
main results, as they are the zeroth-order solutions for rotating frame, while in the inertial frame, the fluid does
the main problem in this paper. not have any circulation. We refer to this case as ir-
Assuming that the star is in equilibrium, Eq. (33) is rotational Roche-Riemann ellipsoid (IRRE) [2]. It has
written as been shown that this velocity configuration is typically
  (approximately) satisfied for binary neutron stars and
1 BH-NS binaries in the late inspiraling stage [17, 18].
0 = 2Tij + Πδij + Mij + Ω2 Iij − (δi3 I3j + δj3 I3i )
2 In the case of the equatorial orbits, Eq. (35) becomes
3 3
X 1X
+Ω (ϵil3 Jlj + ϵjl3 Jli ) − (C̃ik Ikj + C̃jk Iki ), pc ∆0
2 0 = Λ̃2 − 2α2 Ω̃Λ̃ − 2A1 + 2 + 3Ω̃2 , (46)
l=1 k=1 πρ2 a21 P0
(35) pc
0 = α22 Λ̃2 − 2α2 Ω̃Λ̃ − 2α22 A2 + 2 2 2 , (47)
πρ a1
while Eq. (34) is trivially satisfied because Ω̇ = 0 and  
off-diagonal components of C̃ij vanish: In this case, the pc ∆0
0 = −2α32 A3 + 2 2 2 − α32 Ω̃2 3 − 2 , (48)
non-zero components of the tidal tensor are πρ a1 P0
 ∆0  where we defined dimensionless quantities, α2 := a2 /a1 ,
C̃11 = Ω20 1 − 3 , (36)
P0 α3 := a3 /a1 , Ω̃ := Ω0 /(πρ)1/2 , and Λ̃ := Λ/(πρ)1/2 .
C̃22 = Ω20 , (37) After fixing f and eliminating pc /(πρ2 a21 ), we can de-
 ∆0  termine α2 and Ω̃ as functions of α3 . We plot the the
C̃33 = Ω20 − 2 + 3 , (38) relations between Ω̃ and α3 for f = 0 and 1 in Fig. 1,
P0
which agree with those in [8].
where Ω20 = m/r03 [6], ∆0 = r02 − 2mr0 + a2 , and P0 = One notices that Ω̃ reaches a maximum value Ω̃crit cor-
1/2
r02 − 3mr0 + 2am1/2 r0 . Like in the standard Roche responding to α3,crit . Above this value, there is no possi-
problem, we can assume that the principal axes of the ble equilibrium configuration of the star. This gives the
ellipsoid coincide with the axes of our frame. Thus, we relativistic Roche limit, already computed in [5, 8]. As
have [2] in [8], we find that at the innermost stable circular orbit
1 (ISCO) Ω̃2crit ≈ 0.06364 for f = 1, and Ω̃2crit ≈ 0.06640
Iij = M a2i δij , (39) for f = 0 [5]. As in the classical case, one can expect that
5
the star is unstable against mass shedding/tidal disrup-
Mij = −2πρAi Iij , (40)
Z ∞ tion for α3 < α3,crit and stable when α3 > α3,crit . Thus,
du for a given value of ρ, Ω has a maximum value, i.e., the
Aj = a1 a2 a3 , (41)
0 D(a2j + u) orbital separation r0 has a minimum value, and for the
smaller orbital separation, the star should start the tidal
where disruption process.
q
D= (a21 + u)(a22 + u)(a23 + u). (42)

As we assume that the internal motion is characterized III. STARS IN SLIGHTLY PRECESSING
by a vorticity around the 3-axis, we set ORBITS

a1 a−1
 
0 2 0 A. Method for the expansion
Q = Λ −a−11 a2 0 0 , (43)
0 0 0
We now explore the cases that the star does not lie
where Λ is the magnitude of the vorticity. The circulation on the equatorial plane. For this, we perform perturba-
of the star in the inertial frame is then written as tive calculations using ε = θ − π/2 as an infinitesimal
parameter. The zeroth-order solution is determined by
a2 + a22
 
CΓ = πa1 a2 2Ω0 − 1 Λ the equations for the equatorial orbits reviewed in the
a1 a2 previous section. We then need to consider a small per-
= 2πa1 a2 Ω0 (1 − f ), (44) turbation away from this configuration for 0 < ε ≪ 1.
To obtain the perturbed configuration, we assume that
where every fluid element is displaced as xi → xi + ξi where
a21 + a22 Λ ξi is an infinitesimal displacement. For this, we as-
f= . (45) sume the form of ξi =
P3
ξ x , which can satisfy
2a1 a2 Ω0 j=1 ij j
 5

We consider a spherical orbit with always the same

radius r0 but different values of θ. Following [14], to
study the orbit of the star, we fix r0 and C and deduce the
other parameters from R = 0 = dR/dr at r = r0 . Thus,
E, L, and K depend on the inclination angle. These
changes occur at the second order in ε0 (see appendix).
The tidal tensor components can be decomposed using
a Taylor expansion around θ = π/2 :
∞
(k)
X
C̃ij = C̃ij εk . (54)
k=0

The components only contain even-order contributions,

if its indices are even on the component 3, and odd-order
contributions, if its indices are odd on the component 3.
Likewise, we write Ω and Ω̇ as a Taylor expansion,
which containr only the terms of even power in ε. We
FIG. 1. Ω̃2 as a function of arccos(α3 )/(π/2) at the ISCO for m
f = 0 and f = 1. We note that ∆0 /P0 = 4/3 at the ISCO
note Ω0 = at the zeroth order of Ω.
r03
irrespective of the dimensionless spin parameter, a/m, and We do the same for the displacement coefficients ξij .
hence, the curves are universal irrespective of a/m. By considering the symmetry x3 → −x3 , one can con-
strain the expression of the displacement. We thus have
that if the indices of the displacement are odd on the com-
the perturbation equations for the incompressible fluid
ponent 3, then the series expansion only contains terms
self-consistently (see below). Furthermore, to solve this
of odd power in ε, and likewise for indices even on compo-
system in such an assumption, it is sufficient to consider
nent 3, the series expansion only contains terms of even
the modified second-rank tensor virial equations [2] to-
power in ε.
gether with the incompressible condition, which leads to
Following [2], the integrals following a displacement ξi
ξ11 + ξ22 + ξ33 = 0. (49) of the fluid are given as
3 
The displacement will depend on time, as the orbital X dVi;k dVj;k
2δTij = Qjk + Qik
value of θ does too. Thus, as a first step, we have to deter- dτ dτ
k=1
mine the time-dependent orbital motion. To do this we 
analyze the geodesic equation for θ, (4), which is rewrit- − ((Q2 )jk Vi;k + (Q2 )il Vj;k ) , (55)
ten for ε = θ − π/2 ≪ 1 as
 2 δIij = Vi;j + Vj;i , (56)
4 dε
= C − ε2 a2 (1 − E02 ) + L20 , 3
 
r (50) X
dτ δMij = −2πρBij Vij + πρa2i δij Ail Vll , (57)
2 k=1
where C = K − (L − aE) , which vanishes for equatorial Z
orbits, and we took into account the terms at O(ε2 ). Note δΠ = δP d3 x, (58)
that E0 and L0 denote E and L for equatorial circular V
orbits [12] while for C we need the second-order quantities
d2 Vi;j
Z
d
in ε. Then, for r = r0 , we obtain δ ρui xj d3 x =
dτ V dτ 2
ε(τ ) = ε0 cos(ωθ τ ), (51) 3  
X dVj;k dVi;k
+ Qik − Qjk ,(59)
where dτ dτ
p k=1
L20 + a2 (1 − E02 ) Z
dVi;j X
3
ωθ = δ ρui xj d3 x = + [Qik Vj;k − Qjk Vi;k ](60)
,
r02 dτ
V k=1
√
s
r02 − 4a mr0 + 3a2 where
= Ω0 , (52)
P0 Z
√ Vi;j = ρξi xj d3 x = ξij Ijj , (61)
C V
ε0 = . (53) Z ∞
ωθ udu
√ Bij = a1 a2 a3 , (62)
0 D(ai + u)(a2j + u)
2
Here ε0 and C are first-order parameters which we will Z ∞
use for the perturbative expansion. It is worthy to note du
Aij = a1 a2 a3 . (63)
at this stage that cos θ(τ ) = −ε(τ ) at the first order. 0 D(ai + u)(a2j + u)
2
 6

The resulting equations can be divided in two types as nation of the orbit.
shown in the following subsections. In the following, we
only consider equations which contain non-zero contribu-
tions of lower-order solutions of the fluid. This allows us B. First-order calculations
to focus only on the displacement induced by the incli-
To solve the first-order equations for the displacement,
we select only the contribution of order ε0 in the equa-
tion. If we start from Eq. (32), this gives us

3 3 3
d2 X dVi;k X X
2
Vi;j − 2 Qjk =− ((Q2 )jl Vi;l + (Q2 )il Vj;l ) − 2πρBij Vij + πρa2i δij Ail Vll
dτ dτ
k=1 l=1 l=1
3 3
! 3 3
X d X X (0)
X (1)
+ Ω20 Vij − δi3 Ω20 V3j + δij δΠ + 2 ϵil3 Ω0 Vl;j + (Qlk Vj;k − Qjk Vl;k ) − C̃ik Vkj − C̃ik Ikj . (64)
dτ
l=1 k=1 k=1 k=1

(1)
As before, we can write the corresponding symmetrized (−ωθ2 − Ω20 + C̃22 − C̃33 )V2;3
and antisymmetrized equations. We only need to con- + (ωθ2 − Ω20 + 2Ω0 Λα2−1 + C̃22 − C̃33 )V3;2
(1)
(72)
sider 4 of these equations at first order. Indeed only the
(1) (1)
equations for the indices (i, j) = (1, 3), (3, 1), (2, 3), (3, 2) − 2Ω0 ωθ V1;3 + 2Λωθ α2 V3;1 = 0.
contain the non-zero contribution from the inertia tensor.
Since the first-order perturbative quantities should These equations can be recasted as a matrix equation
vary periodically in time with the angular velocity ωθ , 3
(1) (1) (1)
X
we use the ansatz for the variables as Mik Vk = Ci , (73)
(1) k=1
V1;3 = V1;3 cos(ωθ τ ), (65)
(1)
where
V3;1 = V3;1 cos(ωθ τ ), (66) (1) (1) (1) (1) (1)
Vi = (V1;3 , V3;1 , V2;3 , V3;2 ), (74)
(1)
V2;3 = V2;3 sin(ωθ τ ), (67)
(1) (1)
V3;2 = V3;2 sin(ωθ τ ). (68) (1) M C̃13 2
Ci =− (a3 + a21 , a23 − a21 , 0, 0), (75)
5
Then the symmetrized equations become
(1) (1)
(−ωθ2 + 4πρB13 − Ω20 + C̃11 + C̃33 )V1;3 C̃13 = C̃13 cos(ωθ τ )
p
+ (−ωθ2 + 4πρB13 − Ω20 − 2Λ2 m r02 + K0 (r02 + 5K0 )
= 3aε0 cos(ωθ τ ) √ , (76)
(1) (69) K0 r 6
+ 2Ω0 Λα2 + C̃11 + C̃33 )V3;1
2 and
(1) (1) (1) a + a21
− 2Ω0 ωθ V2;3 − 2Λωθ α2−1 V3;2 = −M C̃13 3 , (1)
5 M11 = − ωθ2 + 4πρB13 − Ω20 + C̃11 + C̃33 ,
(1)
M12 = − ωθ2 + 4πρB13 − Ω20 − 2Λ2
(1)
(−ωθ2 + 4πρB23 − Ω20
+ C̃22 + C̃33 )V2;3 (77)
+ 2Ω0 Λα2 + C̃11 + C̃33 ,
+ (−ωθ2 + 4πρB23 − Ω20 − 2Λ2 (1)
(70) M13 = − 2Ω0 ωθ ,
(1)
+ 2Ω0 Λα2−1 + C̃22 + C̃33 )V3;2 (1)
M14 = − 2Λωθ α2−1 ,
(1) (1)
− 2Ω0 ωθ V1;3 − 2Λωθ α2 V3;1 = 0,
(1)
while the antisymmetrized equations are M21 = − 2Ω0 ωθ ,
(1)
(1) M22 = − 2Λωθ α2 ,
(−ωθ2 − Ω20 + C̃11 − C̃33 )V1;3
(1)
(1) M23 = − ωθ2 + 4πρB23 − Ω20 + C̃22 + C̃33 , (78)
+ (ωθ2 − Ω20 + 2Ω0 Λα2 + C̃11 − C̃33 )V3;1 (71) (1)
2 M24 = − ωθ2 + 4πρB23 − Ω20 − 2Λ2
(1) (1) (1) a − a21
− 2Ω0 ωθ V2;3 + 2Λωθ α2−1 V3;2 = −M C̃13 3 , + 2Ω0 Λα2−1 + C̃22 + C̃33 ,
5
 7
(1) (−2πρ(B13 − B12 ) + 2Ω0 Λα2 )V1;1
M31 = − ωθ2 − Ω20 + C̃11 − C̃33 ,
(1)
 d2
M32 =ωθ2 − Ω20 + 2Ω0 Λα2 + C̃11 − C̃33 , + + 2πρ(3B22 − B23 ) − 2Ω20 − 2Λ2 + 2Ω0 Λα2−1
(1)
(79) dτ 2
M33 = − 2Ω0 ωθ ,   d2 
(1) + 2C̃22 V2;2 + − 2 − 2πρ(3B33 − B23 ) − 2C̃33 V3;3
M34 =2Λωθ α2−1 , dτ
d d
+ 2Ω0 V1;2 + 2Λα2 V2;1
dτ dτ
(2) (2)
= −C̃22 I22 + C̃33 I33 + C̃13 (V3;1 + V1;3 )
(1)
M41 = − 2Ω0 ωθ , + 4Ω(2) (Ω0 − Λα2−1 )I22 ,
(1)
M42 =2Λωθ α2 , (83)
(1)
(80)
M43 = − ωθ2 − Ω20 + C̃22 − C̃33 ,
(1)
M44 =ωθ2 − Ω20 + 2Ω0 Λα2−1 + C̃22 − C̃33 .
d d
(2Λα2 + 2Ω0 ) V1;1 − (2Ω0 + 2Λα2−1 ) V2;2
(1) (1) dτ dτ
Note that all the components of Mij and Ci are writ-  d2 
2 2
ten in the zeroth-order quantities, and hence, the solu- + 2
+ 4πρB12 − 2Ω0 − 2Λ + C̃11 + C̃22 V1;2
(1) dτ (84)
tion for Vk is obtained straightforwardly by inverting  d2 
the matrix equation (see Sec. IV). + + 4πρB12 − 2Ω20 − 2Λ2 + C̃11 + C̃22 V2;1
dτ 2
= −C̃13 (V3;2 + V2;3 ) + Ω̇(I22 − I11 ),

C. Second-order calculations
d d
(2Λα2 − 2Ω0 ) V1;1 − (2Ω0 − 2Λα2−1 ) V2;2
To find the second-order equations, we need to take dτ dτ
 d2
into account second-order corrections to Ω and Ω̇ for the

+ + C̃11 − C̃ 22 V1;2
equations. We denote the former as Ω = Ω0 + Ω(2) + dτ 2 (85)
O(ε40 ). The second-order corrections to Ω and Ω̇ are given  d2 
in the appendix. + − 2 + C̃11 − C̃22 V2;1
dτ
For the second-order perturbation, we only need to = −C̃13 (V3;2 + V2;3 ) + Ω̇(I22 + I11 ),
consider the even terms on the component 3. We then
have 5 equations. By adding Eq. (49), we obtain 6 equa-
tions for 6 variables, V1;1 , V2;2 , V3;3 , V1;2 , V2;1 , and δΠ.
Eliminating δΠ from the equations, and using the re- 1 1
V1;1 + V2;2 + 2 V3;3 = 0. (86)
lations α22 α3

3 3
X X There is an additional subtlety in this second-order
− 2B11 V11 + a21 A1l Vll + 2B33 V33 − a23 A3l Vll equations since the right hand side of the equations has
l=1 l=1
two different time dependence because of ε2 (t) = ε20 [1 +
= −(3B11 −B13 )V11 +(B23 −B12 )V22 +(3B33 −B13 )V33 , cos(2ωθ τ )]/2. That is, there is a contribution from the
(81) constant terms and from the terms of angular velocity
2ωθ .
we obtain We must treat the different time dependence sepa-
rately. We thus separate the terms of different frequen-
 d2 cies and make the ansatz
+ 2πρ(3B11 − B13 ) − 2Ω20 − 2Λ2 + 2Ω0 Λα2
dτ 2 
(2,0) (2,2)
+ 2C̃11 V1;1 + (−2πρ(B23 − B12 ) + 2Ω0 Λα2−1 )V2;2 V1;1 = V1;1 + V1;1 cos(2ωθ τ ),
(2,0) (2,2)
 d2  V2;2 = V2;2 + V2;2 cos(2ωθ τ ),
+ − 2 − 2πρ(3B33 − B13 ) − 2C̃33 V3;3 (2,0) (2,2)
dτ V3;3 = V3;3 + V3;3 cos(2ωθ τ ), (87)
d d
−2Λα2−1 V1;2 − 2Ω0 V2;1 V1;2 =
(2,0)
V1;2 +
(2,2)
V1;2 sin(2ωθ τ ),
dτ dτ
(2) (2) (2,0) (2,2)
= −C̃11 I11 + C̃33 I33 + 4Ω(2) (Ω0 − Λα2 )I11 , (82) V2;1 = V2;1 + V2;1 sin(2ωθ τ ).
 8
(2,2)
We then separate the source terms in the same way: M41 =(−4Λα2 + 4Ω0 )ωθ ,
(2,2)
M42 =(−4Λα2−1 + 4Ω0 )ωθ ,
(2) (2,0) (2,2)
C̃11 = C̃11 + C̃11 cos(2ωθ τ ), (2,2)
M43 =0, (93)
(2) (2,0) (2,2)
C̃22 = C̃22 + C̃22 cos(2ωθ τ ), (2,2)
M44 = − 4ωθ2 + C̃11 − C̃22 ,
(2) (2,0) (2,2)
C̃33 = C̃33 + C̃33 cos(2ωθ τ ), (2,2)
M45 =4ωθ2 + C̃11 − C̃22 ,
(2) (2,0) (2,2)
Ω =Ω +Ω cos(2ωθ τ ),
(2)
Ω̇ = Ω̇ sin(2ωθ τ ), (2,2)
M51 =1,
(1) (1) (1)
C̃13 (V3;1 + V1;3 ) (2,2) 1
C̃13 (V3;1 + V1;3 ) = [1 + cos(2ωθ τ )], M52 = 2,
2 α2
(1) (1) (1)
C̃13 (V3;2 + V2;3 ) (2,2) 1 (94)
C̃13 (V3;2 + V2;3 ) = sin(2ωθ τ ). M53 = 2,
2 α3
(88) (2,2)
M54 =0,
(2,2)
For the complete expression of these terms, we refer the M55 =0.
leader to the appendix.
(2,2) (2,2) (2,2) (2,2) (2,2)
As before, we write the corresponding equations using V (2,2) = (V1;1 , V2;2 , V3;3 , V1;2 , V2;1 ), and
the matrix notation
(2,2) (2,2) (2,2)
C1 = − C̃11 I11 + C̃33 I33 + 4Ω(2,2) (Ω0 − Λα2 )I11 ,
3 (2,2) (2,2) (2,2) (1) (1) (1)
(2,2) (2,2) (2,2) C2 = − C̃22 I22 + C̃33 I33 + C̃13 (V3;1 + V1;3 )
X
Mik Vk = Ci (89)
k=1 + 4Ω(2,2) (Ω0 − Λα2−1 )I22 ,
(1) (1) (1)
where (2,2) C̃13 (V3;2 + V2;3 )
C3 =− + Ω̇(2) (I22 − I11 ),
2
(2,2)
M11 = − 4ωθ2 + 2πρ(3B11 − B13 ) (2,2)
(1) (1) (1)
C̃13 (V3;2 + V2;3 )
C4 =− + Ω̇(I22 + I11 ),
− 2Ω20 − 2Λ2 + 2Ω0 Λα2 + 2C̃11 , 2
(2,2)
M12
(2,2)
= − 2πρ(B23 − B12 ) + 2Ω0 Λα2−1 , C5 =0.
(2,2)
(90) (95)
M13 =4ωθ2 − 2πρ(3B33 − B13 ) − 2C̃33 ,
(2,2) The differential equations for the constant term must
M14 = − 4Λωθ α2−1 , be considered with more care. If we proceed as before, we
(2,2) 3
M15 = − 4Ω0 ωθ , X (2,0) (2,0) (2,0)
obtain again linear equations Mik Vk = Ci .
k=1
(2,0) (2,2)
(2,2)
The matrix of this equation Mik is the same as Mik
M21 = − 2πρ(B13 − B12 ) + 2Ω0 Λα2 , when we set ωθ = 0. However, this matrix is never in-
(2,2)
M22 = − 4ωθ2 + 2πρ(3B22 − B23 ) versible as its third and fourth line are colinear.
However, in this case the term in the right hand side,
− 2Ω20 − 2Λ2 + 2Ω0 Λα2−1 + 2C̃22 , (2,0)
Ci , is also simpler because C3
(2,0)
= C4
(2,0)
= 0 (see
(2,2)
(91)
M23 =4ωθ2 − 2πρ(3B33 − B23 ) − 2C̃33 , Eq. (88)). This would imply that we have 4 equations for
(2,2)
5 variables. We have thus a certain freedom for setting
M24 =4Ω0 ωθ , (2,0) (2,0)
V1;2 and V2;1 . We could set them both to 0.
(2,2) By exploring further, we also find that one can assume
M25 =4Λωθ α2 ,
(2,0) (2,0)
V1;2 = w1 τ , V2;1 = w2 τ . The differential equations
only impose w1 + w2 = 0. It thus seems that there is a
(2,2) certain freedom on the choice of the angular momentum
M31 =(−4Λα2 − 4Ω0 )ωθ , at this order. While we could fix all these variables to 0 as
(2,2)
M32 =(4Λα2−1 + 4Ω0 )ωθ , a simplifying assumption, we could also enforce the value
(2,2) f , as defined in Eq. (45), at this order by comparing the
M33 =0, corresponding vorticity of this second-order correction to
(2,2) the second-order correction of Ω. We may then set
M34 = − 4ωθ2 + 4πρB12 − 2Ω20 − 2Λ2 + C̃11 + C̃22 ,
(2,2)
M35 = − 4ωθ2 + 4πρB12 − 2Ω20 − 2Λ2 + C̃11 + C̃22 , 2a21 a22 Ω(2,0) f
w1 = . (96)
(92) a21 + a22
 9

D. Higher orders

Even for higher orders, the form of the equations stays

essentially the same (the matrix form is identical). In-
deed, as before, we have linear equations for higher-order
displacements.
For higher orders, however, there are three main differ-
ences from the previous analyses. The first comes from
the right hand side of the matrix equations. It then
reflects the contribution of all the lower-order displace-
ments and the higher-order contributions of the physical
parameters of the system. The second comes from the
different frequencies which one has to consider: Specifi-
cally, the frequency is higher. Finally, higher-order con-
tributions to the geodesic motion have to be taken into
account.

E. Elliptic orbits

The method described in the previous subsections can

be extended to other types of orbits, like slightly ellip-
tic orbits. For this, we write the orbital separation as
r(τ ) = r0 [1 + ε(τ )] and assume that ε(τ ) is an infinitesi-
mal quantity.
As before, we need to determine the time-dependence
FIG. 2. Ω̃2 as functions of r0 /m for the first- and second-
of ε. Using the definition of ε and writing E, L and K
order resonances, along with the tidal disruption limit Ω̃2crit
by the geodesic constants of a circular geodesic of radius
as well as the expected values of Ω̃2 for typical values of BH-
r0 , we get from Eq. (3) as NS binaries and supermassive BH-white dwarf/ordinary star
 2 binaries. For both plots, we assumed a/m = 0.8. The upper
3(L − aE)2
 
dε
= (δr)2 − ε2 Ω20 1 − , (97) and lower panels show the results for f = 0 and f = 1.
dτ r02
where δr is given by the perturbed values of the geodesic
constants, E and L, and K = (L−aE)2 for the equatorial dependence of the displacement. These are justified as
orbits. We then obtain long as the matrix equations are solvable. This hinges on
the inversability of the matrices considered.
ε(τ ) = δr cos(ωe τ ), (98)
However, during our research, we found that these ma-
where trices could have determinant 0 for particular equilibrium
s r states. Furthermore, the terms in the right hand side
3(L − aE)2 ∆0 were not in the image of the matrices. Thus the lin-
ωe = Ω0 1− = Ω0 4−3 . (99) ear equations were not solvable for the particular cases
r02 P0
in which the determinant vanishes. This would indicate
In this setting, the tidal tensor throughout the orbit is that the ansatz used in these cases are inappropriate.
rewritten in the perturbative form. One has then a resonance. To solve the equations, a bet-
We can also consider more general orbits, which are ter ansatz would be: ξ⃗ = (ξ⃗0 + ξ⃗1 τ )eiωθ τ . In this case,
neither circular nor equatorial. In this case, when we are the displacements would eventually not be infinitesimal.
at order n, we need to consider all frequencies lωθ + mωe Thus our perturbative expansion breaks down at such
for all |l| + |m| = n, with l, m ∈ Z. This is a topic beyond points.
the scope of our present work. These resonances can come either from the first- or
second-order equation. For f = 0, we find a resonance
for both the first and second-order displacements. For
IV. RESONANCES f = 1, we find it only for the second-order equations.
Near the resonant angular frequencies, ω ≈ ωθ or 2ωθ , the
A. Determination of resonances amplitude of the displacement scales as |ω−kωθ |−1 where
k = 1 or 2, indicating that a significant deformation from
An important aspect of the linear equations described the ellipsoidal shape happens not only at the resonant
above comes from the ansatz we made about the time angular frequency but also around these frequencies, i.e.,

FIG. 3. The displacement vector ξi for the first-order pertur-
bation on the surface of the star represented by the red ellipse
for ωθ τ = 0 (upper panel) and ωθ τ = π/2 (lower panel). The
axes X, Y , and Z correspond to the axes 1, 2, and 3 of the
problem in units of a1 , respectively. For all the plots, the
vectors are plotted for a/m = 0.8, f = 0 and at a distance
r0 = 20 m with Ω̃2 ≈ 0.03459. In this case, α2 ≈ 0.886 and
α3 ≈ 0.858.
at ω that satisfies |ω − kωθ | ≪ kωθ . This indicates that
for a star approaching the resonance, the degree of the
displacement is significantly enhanced.
In order to assess the relevancy of these resonances, we
plot the corresponding values of Ω̃2 of these resonances,
along with the values of Ω̃2 that are modeled by a neu-
tron star, a white dwarf, and an ordinary star orbiting a
variety of Kerr BHs. We estimate Ω̃2 using typical radius
Rstar and mass M for these objects. We then have Ω̃2 as
a function of r0 as
 3
4 m Rstar
Ω̃2 = . (100)
3M r0
In Fig. 2, we plot the values of Ω̃2 for the resonances
of ω = ωθ and 2ωθ as functions of r0 /m together with
the Roche (f = 0) and Roche-Riemann (f = 1) lim-
its. We also plot the curves along hypothetical sequences
of inspiraling BH-NS binaries in close orbits and white
10
FIG. 4. The displacement vector ξi for a second-order pertur-
bation on the surface of the star represented by the red ellipse
for Ωθ τ = 0 (upper panel) and ωθ τ = π/2 (lower panel). The
axes X, Y , and Z correspond to the axes 1, 2, and 3 of the
problem in units of a1 , respectively. For all the plots, the
vectors are plotted for a/m = 0.8, f = 0 and at a distance
r0 = 20 m with Ω̃2 ≈ 0.06758. In this case, α2 ≈ 0.734 and
α3 ≈ 0.694.
dwarf/ordinary stars orbiting supermassive BHs. We
note that for the ω = 0 mode in the second-order dis-
placement, the resonance appears near the Roche/Roche-
Riemann limits. This corresponds to the onset of dy-
namical instability for the ellipsoid, similarly as in the
classical Roche problem [2]. This figure shows that these
resonances may destabilize the objects before reaching
the Roche and Roche-Riemann limits. While we show
figures only for one value of a/m(= 0.8), we observe the
same behavior for any non-zero value of a/m.
These resonances seem to be similar to the instabilities
found for a star in elliptic orbits [4] (note that these are
instabilities in purely Newtonian gravity). They appear
when ωe is equal to one of the oscillatory modes of the
star on the equatorial plane. This then creates the res-
onance discussed before. In the case of a star orbiting a
Kerr BH, the possible resonances are more numerous as
a result of the more complicated nature of the geodesics
due to the relativistic effects.
In our present context, the displacement associated
with an oscillatory mode gets excited by the variation

of the θ-parameter of the orbit. It will then grow linearly
and eventually, it will not be infinitesimal anymore. Fig-
ures 3 and 4 show the displacement vectors associated
with the excited modes, assuming a/m = 0.8. Again the
spin parameter does not have an important impact on
the qualitative aspect of the resonant displacement.
From Fig. 3, we can see that the resonance in the first-
order displacement will mostly imply a precession of the
3-axis. As such, this mode is quite reminiscent of the
transverse-shear mode of the MacLaurin spheroids dis-
cussed in [2]. By the non-linear amplification of this
mode, matter may be ejected from the surface to the
outside of the star, resulting possibly in the modification
of the precessing orbit.
From the second-order displacement shown in Fig. 4,
we observe that it will involve mostly a quadrupole dis-
placement in the 1-2-plane. This mode reminds us of
toroidal modes also discussed in [2].
The resonance in the first- and second-order dis-
placements is always encountered slightly outside the
Roche/Roche-Riemann limits. For BH-NS binaries in
precessing orbits, thus, the resonant deformation may be
excited just prior to the tidal disruption and the tidal
disruption may be assisted by the prior resonance if the
relativistic effect on the self-gravity of the star, which we
do not take into account in the present study, does not
qualitatively change the property of the resonance.
B. Analytical resolution of the first-order
resonance
The resonance in the first-order displacement with
f = 0 can be determined analytically. If we introduce
λ such that ωθ2 = (1 + λ)Ω20 where λ is a purely relativis-
tic correction (see Eq. (52)), we have C̃11 = (−2 − λ)Ω20 ,
C̃22 = Ω20 , and C̃33 = (1 + λ)Ω20 . Then,
(1)
det(Mij ) = − 16πρB23 (1 + λ)Ω40
h i
× − 4πρB13 λ + (3 + λ)(1 + λ)Ω20 ,
(101)
and thus, we find that the first-order resonance occurs
when
4B13 λ
Ω̃2 = . (102)
(3 + λ)(1 + λ)
We find that the relativistic correction is the key for this
resonance.
V. DISCUSSION
The main result of this paper is the discovery of possi-
ble resonances of stars with precessing orbits around Kerr
BHs. In particular, we found that these resonances could
be relevant for systems such as white dwarfs/ordinary
11
stars orbiting a supermassive BH and BH-NS binaries in
close orbits.
One interpretation of our results is that as an orbiting
star approaches gradually a Kerr BH, e.g., by the radia-
tion reaction of gravitational waves, it will approach one
of these resonances before encountering the tidal disrup-
tion limit. The closer it gets to it, the more its displace-
ment associated with it will grow. At some point before
the resonance, the displacement will not be infinitesimal
anymore. Our equations are then not valid.
As the first-order resonance has only been found for
f = 0, it will probably not be relevant for BH-NS binaries
for which the NS is likely to be nearly irrotational [17, 18]
in such systems. If we consider white dwarfs/ordinary
stars orbiting supermassive BH, we expect that the cor-
responding star may be corotational because the viscous
angular momentum redistribution can have a timescale
shorter than the orbital evolution one. For these systems,
the star may encounter the first-order resonance before
tidal disruption. If the system has an elliptic orbit, we
expect also resonances of frequency ωe . However, if the
object stays on the equatorial plane, the modification
to the orbit will only give contributions to the diagonal
components of the tidal tensor. As such, the resonances
should correspond to the toroidal mode. If the orbit is
neither equatorial nor spherical, the overall picture gets
more complicated. According to our perturbation analy-
sis, it is likely that there are more possible resonances of
frequency ωθ +ωe and ωθ −ωe , and thus, these resonances
may be excited as well.
To our knowledge, these results have never been re-
ported before. A major counterpoint to our analysis is its
apparent reliance on some questionable assumptions, like
the incompressibility hypothesis. However, as the gen-
eral mechanism for the resonance holds true, we expect
resonances even with more realistic equations of state.
Indeed, whatever the equation of state is, we have char-
acteristic modes for our star. We may then observe a res-
onance between these modes and the oscillation present
in the tidal tensor. Thus, while the equations can change,
the resulting phenomenon should not qualitatively. Of
course, this explanation is just tentative and should be
confirmed by a more serious analysis of the question.
It is possible that this resonance might provide an ex-
planation as to why certain objects end up in a circular
equatorial orbit around a Kerr BH. To justify this, one
would have to inquire about the ultimate fate of the star.
As our equations are not valid near the resonance, one
would have to change our method. One could use dif-
ferent analytical tools or hydrodynamics simulations to
investigate this issue. We leave this open to future work.
ACKNOWLEDGMENTS
We thank the member of Computational Relativis-
tic Astrophysics group at the Max Planck Institute for
Gravitational Physics for discussion. This work was
 12

in part supported by Grant-in-Aid for Scientific Re-
search (grant Nos. 20H00158 and 23H04900) of Japanese
MEXT/JSPS.
compute the perturbed expressions for the energy E, an-
gular momentum around the z-axis L, and the Carter
constant K for the corresponding orbit.
To do this we will rely on the formula given in [14].
Assuming that C is a small quantity, we expand them as

Appendix A: Formula for second-order equations in E = E0 + E1 C + O(C 2 ), (A1)

ε L = L0 + L1 C + O(C 2 ), (A2)
K = K0 + K1 C + O(C 2 ), (A3)
In order to compute all the second-order quantities
necessary for the corresponding equations, we need to where

√
a a − mr
E1 = √ , (A4)
2r02 P0
√ √ 5/2 1/2
a3 m + a mr02 + r0 (−3m + r0 ) + a2 r0 (−m + r0 )
L1 = √ , (A5)
2 mP0 r05
√ √
r0 (−a + mr0 )(a2 + a mr0 + (−3m + r0 ))
r
K1 = 1− . (A6)
m P0

Then, we have

√ √
(2) 2 2 m(r02 − 3mr0 + 2a mr0 )
Ω = a cos θ 7/2 √
r0 ( mr0 − a)2
√
K0 E1 (r02 + a2 ) − aL1 K1 (E0 (r02 + a2 ) − aL0 ) a(L1 − aE1 ) aK1 (L0 − aE0 ) (A7)
  
+ C 2K1 Ω0 + 2 − + −
r0 r02 + K0 (r02 + K0 )2 K0 K02
= a2 cos2 θΩa + CΩC .

Therefore (2) 6M 2
C22 = − a cos2 θ, (A12)
r05
a2 ε20
Ω(2,0) = Ωa + CΩC ,
2 (A8)
(2,2) a2 ε20
Ω = Ωa .
2
Likewise, we have
√ √
2 m(r02 − 3mr0 + 2a mr0 )
Ω̇ = −a θ̇ sin 2θ 7/2 √
+ O(C 2 ),
r0 ( mr0 − a)2
(A9)
and thus
√ √
m(r02 − 3mr0 + 2a mr0 )
Ω̇(2) = −a2 ε20 ωθ 7/2 √
. (A10)
r0 ( mr0 − a)2
Finally, the second-order contributions to the tidal ten-
sor are given as
15K02 + 14r02 K0 + r04 15K02 + 12r02 K0 + r04
     
(2) 3M K1 C (2) 3M K1 C
C11 = 3 − 2 + a2 cos2 θ , C33 = − a2
cos 2
θ .
r0 r0 r04 K0 r03 r02 r04 K0
(A11) (A13)
 13

Thus
a2 ε20 15K02 + 14r02 K0 + r04
  
(2,0) 3M K1 C
C11 = − + ,
r03 r02 2 r04 K0
(2,2) 3M a2 ε20 15K02 + 14r02 K0 + r04
C11 = ,
2r03 r04 K0
(2,0) 3M a2 ε20
C22 =− ,
r05
(2,2) 3M a2 ε20
C22 =− ,
r5
0
a2 ε20 15K02 + 12r02 K0 + r04
 
(2,0) 3M K1 C
C33 = 3 − ,
r0 r02 2 r04 K0
(2,2) 3M a2 ε20 15K02 + 12r02 K0 + r04
C33 = .
2r03 r04 K0
(A14)

[1] M. L. Aizenman, The equilibrium and the stability of
the roche-riemann ellipsoids, The Astrophysical Journal
153, 511 (1968).
[2] S. Chandrasekhar, Ellipsoidal figures of equilibrium
(1969).
[3] A. Nduka, The Roche Problem in an Eccentric Orbit,
Astrophys. J. 170, 131 (1971).
[4] J. Vidal and D. Cébron, Inviscid instabilities in rotat-
ing ellipsoids on eccentric kepler orbits, Journal of Fluid
Mechanics 833, 469–511 (2017).
[5] L. G. Fishbone, The Relativistic Roche Problem, Astro-
phys. J. 175, L155 (1972).
[6] L. G. Fishbone, The relativistic Roche problem. II. Sta-
bility theory., Astrophys. J. 195, 499 (1975).
[7] B. Mashhoon, On tidal phenomena in a strong gravita-
tional field, The Astrophysical Journal 197, 705 (1975).
[8] M. Shibata, Relativistic Roche-Riemann Problems
around a Black Hole, Progress of Theoretical Physics
96, 917 (1996), https://academic.oup.com/ptp/article-
pdf/96/5/917/5226638/96-5-917.pdf.
[9] P. Banerjee, S. Paul, R. Shaikh, and T. Sarkar, Tidal ef-
fects away from the equatorial plane in kerr backgrounds,
Physics Letters B 795, 29 (2019).
[10] J.-A. Marck, Solution to the equations of parallel trans-
port in kerr geometry; tidal tensor, Proceedings of the
Royal Society of London. Series A, Mathematical and
Physical Sciences 385, 431 (1983).
[11] B. Carter, Global structure of the Kerr family of gravi-
tational fields, Phys. Rev. 174, 1559 (1968).
[12] J. M. Bardeen, W. H. Press, and S. A. Teukolsky, Ro-
tating black holes: Locally nonrotating frames, energy
extraction, and scalar synchrotron radiation, Astrophys.
J. 178, 347 (1972).
[13] D. C. Wilkins, Bound geodesics in the kerr metric, Phys-
ical Review D 5, 814 – 822 (1972).
[14] E. Teo, Spherical orbits around a kerr black hole, Gen-
eral Relativity and Gravitation 53, 10.1007/s10714-020-
02782-z (2021).
[15] M. Ishii, M. Shibata, and Y. Mino, Black hole tidal prob-
lem in the Fermi normal coordinates, Phys. Rev. D 71,
044017 (2005), arXiv:gr-qc/0501084.
[16] E. N. Parker, Tensor Virial Equations, Physical Review
96, 1686 (1954).
[17] C. S. Kochanek, Coalescing Binary Neutron Stars, As-
trophys. J. 398, 234 (1992).
[18] L. Bildsten and C. Cutler, Tidal interactions of inspi-
raling compact binaries, Astrophysical Journal, Part 1
(ISSN 0004-637X), vol. 400, no. 1, p. 175-180. 400, 175
(1992).