Monotone-short solutions of the Tolman-Oppenheimer-Volkoff-de Sitter

equation
Cheng-Hsiung Hsu and Tetu Makino

Citation: Journal of Mathematical Physics 57, 092502 (2016); doi: 10.1063/1.4962724

View online: http://dx.doi.org/10.1063/1.4962724
View Table of Contents: http://scitation.aip.org/content/aip/journal/jmp/57/9?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

The non-linear asymptotic stability of De-Sitter: A perturbative approach
AIP Conf. Proc. 1458, 552 (2012); 10.1063/1.4734482

Schwarzschild and de Sitter solutions from the argument by Lenz and Sommerfeld
Am. J. Phys. 79, 662 (2011); 10.1119/1.3557070

General solutions of Einstein’s spherically symmetric gravitational equations with junction conditions
J. Math. Phys. 44, 5637 (2003); 10.1063/1.1621056

Erratum: “Symmetric tensor spherical harmonics on the N-sphere and their application to the de
Sitter group SO(N,1)” [J. Math. Phys. 28, 1553 (1987)]
J. Math. Phys. 43, 6385 (2002); 10.1063/1.1515382

Analysis of the cosmological Oppenheimer-Volkoff equations

J. Math. Phys. 41, 5582 (2000); 10.1063/1.533427

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 JOURNAL OF MATHEMATICAL PHYSICS 57, 092502 (2016)

Monotone-short solutions of the

Tolman-Oppenheimer-Volkoff-de Sitter equation
Cheng-Hsiung Hsu1,a) and Tetu Makino2,b)
1
Department of Mathematics, National Central University, Chung-Li 32001, Taiwan
2
Yamaguchi University, 11-45-1105, Sibanaka, Ube 755-0017, Japan
(Received 27 May 2016; accepted 31 August 2016; published online 20 September 2016)

It is known that spherically symmetric static solutions of the Einstein equations with
a positive cosmological constant for the energy-momentum tensor of a barotropic
perfect fluid are governed by the Tolman-Oppenheimer-Volkoff-de Sitter equation.
Some sufficient conditions for the existence of monotone-short solutions (with finite
radii) of the equation are given in this article. Then we show that the interior metric
can extend to the exterior Schwarzschild-de Sitter metric on the exterior vacuum
region with twice continuous differentiability. In addition, we investigate the analytic
property of the solutions at the vacuum boundary. Our result (Theorem 1) can be
considered as the de Sitter version of the result by Rendall and Schmidt [Classical
Quantum Gravity 8, 985-1000 (1991)]. Furthermore, one can see that there are
different properties of the solutions with those of the Tolman-Oppenheimer-Volkoff
equation (with zero cosmological constant) in certain situation. Published by AIP
Publishing. [http://dx.doi.org/10.1063/1.4962724]

I. INTRODUCTION
This paper is concerned with the existence of solutions of the Tolman-Oppenheimer-Volkoff-de
Sitter equation which have finite radii. To derive the equation, we consider a static and spherically
symmetric metric
ds2 = g µν dx µ dx ν = e2F(r )c2dt 2 − e2H (r )dr 2 − r 2(dθ 2 + sin2θdφ2)
which satisfies the Einstein-de Sitter equations
1 8πG
Rµν − g µν R − Λg µν = 4 Tµν ,
2 c
for the energy-momentum tensor of a perfect fluid
T µν = (c2 ρ + P)U µU µ − Pg µν .
Here Rµν is the Ricci tensor associated with the metric g µν dx µ dx ν and R = g α β Rα β is the scalar
curvature. c, G, and Λ are positive constants which represent the speed of light
(3.00 × 1010 cm/sec), gravitational constant (6.67 × 10−8cm3/g · s2), and cosmological constant,
respectively. Moreover, ρ is the mass density, P is the pressure, and U µ is the four-velocity. In
cosmology, the cosmological constant Λ is the value of the energy density of the vacuum of
space. For more descriptions about the Einstein-de Sitter equations, see the book of Landau and
Lifshitz [Ref. 11, Section 111].
If we choose the coefficients of the above metric by
2 2Gm(r) Λ 2
e2F(r ) = κ +e−2u(r )/c and e−2H (r ) = 1 − − r ,
c2r 3

a) E-mail: chhsu@math.ncu.edu.tw. Partially supported by the MOST and NCTS of Taiwan.

b) Email: makino@yamaguchi-u.ac.jp

0022-2488/2016/57(9)/092502/15/$30.00 57, 092502-1 Published by AIP Publishing.

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-2 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

then the Einstein-de Sitter equations are reduced to

4πr 3
c2Λ 3
dm dP G m+ P − r
= 4πr 2 ρ, = −(ρ + P/c2) c2 3 . (1)
dr dr 2Gm Λ

r2 1 − 2 − r2
cr 3
Here m(r 0) is the total mass inside radius r = r 0, u(r) and κ + are function and constant which will be
specified later. When Λ = 0, (1) turns out to be
4πr 3

dm dP G m+ P
= 4πr 2 ρ, = −(ρ + P/c2) c2 . (2)
dr dr 2Gm

r2 1 − 2
cr
17
System (2) was derived by Oppenheimer and Volkoff in 1939, and called the Tolman-Oppenheimer-
Volkoff equation. Therefore we shall call system (1) with Λ > 0 the Tolman-
Oppenheimer-Volkoff-de Sitter equation. Some mathematical progresses on the study of the Tolman-
Oppenheimer-Volkoff equation can be found in the literature.12,13 In this article, we will focus on the
solutions of system (1).
Historically speaking, the cosmological constant Λ of the above Einstein-de Sitter equations
was introduced by Einstein8 in 1917 and was discussed soon by de Sitter.20 Although it was intro-
duced for a static universe, it was not necessary for an expanding universe. So, later Einstein wrote
to Weyl the sentence
“If there is no quasi-static world, then away with the cosmological term!”
on May 23, 1923, and finally he rejected the cosmological term in Ref. 9, 1931. For the reason, it is
not necessary to explain the Hubble’s report on the redshifts of galaxies showing the expansion of
the universe (see Ref. 18 [Section 15e]). However, although the original motivation for introducing
the cosmological term disappeared, its status in the cosmological theories remained and it revived
with new meanings in the recent development of the theories and observations. For the details, see
the review5 which presents a pedagogical overview of cosmology in the presence of a cosmological
constant.
In the past years, the Tolman-Oppenheimer-Volkoff-de Sitter equation has already been system-
atically investigated from the physical point of view by Böhmer.2,3 We note that Böhmer assumed
that there exists a positive density ρb , so-called the “boundary density,” at which the pressure
P vanishes (i.e., P > 0 if and only if ρ > ρb ), and the cosmological constants Λ satisfying Λ <
4π ρb G/c2 are considered. Hence this situation is not treated in this article.
From the observation of state for neutron stars (see Refs. 14 and 21 [p.188]), the pressure and
density satisfy
 ζ  ζ
q4dq
P = Kc 5
and ρ = 3Kc 3
1 + q2q2dq, (3)
1 + q2

0 0

where K being a positive constant. Keeping relation (3) in mind, throughout this article, we assume
the fluid is barotropic, i.e., the pressure is a function of the density only and vice versa, and satisfies
the following assumption:
P(ρ) > 0, 0 < dP/dρ < c2 for ρ > 0, and P(ρ) → 0 as ρ → +0. Moreover, we as-
sume there are positive constants A, 1 < γ < 2 and an analytic function Ω(·) on a
neighborhood of [0, +∞) such that
Ω(0) = 1 and P(ρ) = Aργ Ω(Aργ−1/c2).
For the sake of notational conventions, we shall denote
2Gm Λ 2
κ(r, m) B 1 − − r , (4)
c2r 3
4πr 3
c2Λ 3
Q(r, m, P) B G m + 2 P − r , (5)
c 3

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-3 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

so that the second equation of (1) reads

dP Q(r, m, P)
= −(ρ + P/c2) 2 .
dr r κ(r, m)
Let us consider solutions of system (1) on the domain
D B {(r, m, P)| 0 < r, |m| < +∞, 0 < ρ, 0 < κ(r, m)}.
By the same proof of Ref. 14 [Proposition 1], it is known that system (1) has a solution germ
(m(r), P(r)) at r = +0, given the central density ρc > 0 with Pc B P(ρc ), such that
4π
m= ρc r 3 + O(r 5),
3

r2
P = Pc − (ρc + Pc /c2) 4πG(ρc + 3Pc /c2) − c2Λ + O(r 4), (6)
6
as r → +0. See also Ref. 12 [Section 2, pp. 57 and 58]. We are interested in the prolongation
of the solution germ to the right as long as possible in the domain D. Actually the prolonga-
tion is unique, since the right-hand sides of (1) are analytic functions of r, m, P in D, see Ref. 7
[Chap.1, Sec. V]. Especially, we want to provide a sufficient condition for which the prolongation
turns out to be “monotone-short” in the following sense.

Definition 1. A solution (m(r), P(r)), 0 < r < r +, of (1) is said to be monotone-short if r + < ∞,
κ + > 0, Q+ > 0, and
dP/dr < 0 for 0 < r < r + and P → 0 as r → r +,
where
2Gm+ Λ 2
κ + B lim κ(r, m(r)) = 1 − − r +,
r → r +−0 c2r + 3
c 2Λ 3
Q+ B lim Q(r, m(r), P(r)) = Gm+ − r ,
r → r +−0 3 +
with m+ B lim m(r).
r → r +−0

We are interested in solutions with monotone decreasing ρ(r). In view of (6), it is necessary that
the inequality
4πG
Λ < Λ∗ B (ρc + 3Pc /c2) (7)
c2
holds in order that dP/dr < 0 at least for 0 < r ≪ 1. In other words (7) should hold in order that the
prolongation be monotone-short. Discussion for the case of large Λ will be done in Sec. VI.
The rest of this paper is organized as follows. In Section II, we give the sufficient conditions
such that solution germ with (6) to the right turns out to be monotone-short. Section III concerns
with the monotonicity of the solution of the Tolman-Oppenheimer-Volkoff-de Sitter equation. In
Section IV, we extend the interior metric to the exterior Schwarzschild-de Sitter metric on the
vacuum region with twice continuous differentiability. In Section V, analytical properties of the vac-
uum boundary are investigated. Finally, we discuss about the case in which Λ is large in Section VI.

II. EXISTENCE OF MONOTONE-SHORT SOLUTIONS

To study the problem, we introduce the variable u by
 ρ
1
uB dP.
0 ρ + P/c2
Then it is easy to verify that
γ A γ−1 1 γ
u= ρ Ωu (Aργ−1/c2), ρ = A1u γ−1 Ω ρ (u/c2), and P = AAγ1 u γ−1 Ω P (u/c2),
γ−1

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-4 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

) 1
where A1 B (γ − 1)/(γ A) γ−1 , Ωu (·), Ω ρ (·), Ω P (·) are analytic functions on a neighborhood of
(

[0, +∞[ such that Ωu (0) = Ω ρ (0) = Ω P (0) = 1. In fact, the functions Ωu (·), Ω ρ (·), Ω P (·) only
depend on γ and the function Ω(·), which can be written by the form
 γ−1 ′
1 ζ Ω(ζ ) + γ ζ DΩ(ζ ) ′
′ ′
γ
−
Ωu (ζ) = dζ and Ω P (η) = Ω(ζ)Ωu (ζ) γ−1
ζ 0 1 + ζ Ω(ζ )
′ ′

with
γ−1 γ
ζ= ηΩ ρ (η), or equivalent to η = ζΩu (ζ).
γ γ−1
Let us fix a small positive number δΩ such that these functions Ωu (·), Ω ρ (·), Ω P (·) are defined and
analytic on a neighborhood of [−δΩ, +∞[. Using the variable u, we can write system (1) as
1
dm
= 4πr 2 A1u♯γ−1 Ω ρ (u/c2), (8)
dr
γ
4π 3 γ γ−1
c 2Λ 3
du G m + r AA u
1 ♯ Ω P (u/c 2
) − r
=− c2 3 ,
dr 2Gm Λ

r2 1 − 2 − r2
cr 3
where u♯ B max{u, 0}. According to the assumption 1 < γ < 2, which implies µ B 1/(γ − 1) > 1
and γ/(γ − 1) = µ + 1 > 2, we can see that the functions u → u♯µ , u♯µ+1 are of class C 1(R). Keeping
the fact in mind, we consider Equation (8) on the domain
Du B {(r, m,u)| 0 < r, |m| < ∞, −δΩ < u/c2 < +∞, κ > 0}.
Let us set
 ρc
dP γ A γ−1
uc B = ρc Ωu (Aργ−1
c /c ).
2
0 ρ + P/c 2 γ−1
We perform the homologous transformation of the variables
1
r = aR, m = a3b γ−1 · 4π A1 M, u = bU,
where a, b are positive parameters. Taking b = uc and a be the constant satisfying
2−γ
4πG A1a2b γ−1 = 1,
then system (8) turns out to
1
dM
= R2U♯γ−1 Ω ρ (αU), (9)
dR
γ
γ − 1 3 γ−1 1
M+ αR U♯ Ω P (αU) − βR3


dU 1 γ 3
=− 2 , (10)
dR R M 1
1 − 2α − α βR2


R 3
where U♯ B max{U, 0},
1
− γ−1 c2 − 1 c2
α B b/c2 = uc /c2, βBb λ= (uc ) γ−1 Λ with λ B Λ.
4πG A1 4πG A1
It follows that the domain of system (9) should be
DU B {(R, M,U)| 0 < R, |M | < ∞, − δΩ < U < 2, κ > 0},
where, of course,
M 1
κ = 1 − 2α − α βR2.
R 3

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-5 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

Let us concentrate ourselves to the conditions 0 ≤ α ≤ 1, 0 ≤ β ≤ 1, and consider a solution germ

(M(R),U(R)) at R = +0 which satisfies
R3
M(R) = Ω ρ (α) + O(R5), (11)
3
γ−1
R2
U(R) = 1 − Ω ρ (α) + 3 αΩ P (α) − β + O(R4).
γ 6
We first establish the continuous dependence of the solution (M(R),U(R)) on the parameters α and
β.

Proposition 1. There are positive numbers ϵ 0(< 1), R0 which depend upon only γ and the
function Ω(·) such that (M(R),U(R)) exists and satisfies dU/dR < 0 on 0 < R ≤ R0 provided that
0 ≤ α, β ≤ ϵ 0 and (M(R0),U(R0)) depends continuously on α, β ∈ [0, 1].

Proof. The proof is standard and sketched as follows.

First, we may assume that there is a positive δ > 0 such that
γ−1
Ω ρ (α) + 3 αΩ P (α) − β ≥ δ
γ
provided that 0 ≤ α, β ≤ ϵ 0(< 1) (this a fortiori guarantees (7) uniformly), by converting system (9)
to a system of integral equations under the condition (11) as
 R ′
3 1 Ω ρ (αU(R ))
q(R) = 3 U(R ′) γ−1 R ′2dR ′, (12)
R 0 Ω ρ (α)
γ
 R 1 Ω (α)q(R ′) + γ−1 αU(R ′) γ−1 Ω P (αU(R ′)) − 31 β
3 ρ γ
U(R) = 1 − R ′dR ′.
1 − 2α 1
Ω (α)q(R ′)R ′2 − 1 α βR ′2
0 3 ρ 3

Then, taking R0 sufficiently small uniformly on α, β, we see that the mapping (q,U) → (q̃, Ũ),
which is the right-hand side of the (12), is a contraction map from the space
F B {(q,U) ∈ C[0, R0]| 0 ≤ q ≤ Cq , 1/2 ≤ U ≤ 2}
into itself with respect to a suitable functional distance, where
 1 Ω ρ (αU) 
1/2 ≤ U ≤ 2, 0 ≤ α ≤ 1 .

Cq B max U γ−1
Ω ρ (α)
For the details, see Ref. 12 [pp. 57 and 58]. 
According to Proposition 1, we obtain the following result.

Theorem 1. Suppose that 6/5 < γ < 2. Then there exists a positive number ϵ 0(< 1) depending
upon only γ and the function Ω(·) such that if
4π γ − 1
γ−1
1 1
uc ≤ c2ϵ 0 and Λ ≤ G (uc ) γ−1 ϵ 0, (13)
c 2 γA
then the inequality (7) holds and the prolongation of the solution germ with (6) to the right turns out
to be monotone-short.

Proof. By Proposition 1, the right-hand side of system (9) depends continuously on α, β and
1
tends to (R2U♯γ−1 , −M/R2)T as α → 0, β → 0. Then the limiting system
dM dU M
= R2U♯µ , =− 2
dR dR R
is nothing but the Lane-Emden equation
1 d 2 dU

− R = U♯µ . (14)
R2 dR dR

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-6 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

Since 6/5 < γ < 2 (or 1 < µ < 5), the solution U = Ū(R) with Ū(0) = 1 of the Lane-Emden equa-
tion (14) is short, that is,
Ū > 0, dŪ/dR < 0 for 0 < R < ξ1 = ξ1(γ) and Ū(ξ1) = 0.
See Refs. 6 and 10. Of course we consider
dŪ
1 1

Ū(R) = R2 −
dR R=ξ 1 ξ1 R
harmonically on R ≥ ξ1. Thanks to Proposition 1, if ϵ 0 is sufficiently small,
α ≤ ϵ 0 and β ≤ ϵ 0, (15)
then U = U(R) exists and remains near to the orbit of U = Ū(R) on R0 ≤ R ≤ ξ1 + δ R , where δ R is
small enough such that
δΩ
− ≤ Ū(ξ1 + δ R ) < 0.
2
This is nothing but a direct application of Ref. 7 [Theorem 7.4]. Note that max{Ω P (η)| − δΩ ≤ η ≤
2} depends upon only γ and the function Ω(·), and −δΩ ≤ η = αU ≤ 2 provided that α ≤ ϵ 0 ≤ 1
and −δΩ < U < 2. Especially, if U(ξ1 + δ R ) < 0, then the radius R+ of U(R) should be found in the
interval ]0, ξ1 + δ R [. This completes the proof of Theorem 1, since condition (15) is nothing but
(13). 
The result of Theorem 1 can be considered as the de Sitter version of the result by A. D.
Rendall and G. B. Schmidt.19
Remark 1. (1) Note that (7) follows from (13) if ϵ 0 is sufficiently small, since
γ − 1
γ−1
1 1
ρ + 3P/c2 = u γ−1 Ω ρ+3P/c 2(u/c2),
γA
where
γ−1
Ω ρ+3P/c 2(η) B Ω ρ (η) + 3 ηΩ P (η)
γ
is a function depending upon only γ and the function Ω(·) and Ω ρ+3P/c 2(0) = 1 so that we can
assume that min{Ω ρ+3P/c 2(η)| − δΩ ≤ η ≤ 2} > ϵ 0.
(2) For the existence of uc satisfying (13), it is necessary that Λ enjoys
γ
2(2−γ) γ − 1
γ−1
1
Λ ≤ 4πc γ−1 G ϵ 0γ−1 .
γA
Thus one may ask whether the real value of the cosmological constant of our universe satisfies it or
not. But this question is not theoretical but experimental-observational and numerical. It is beyond
our ability to answer the question.
Furthermore, we note that even if γ ≤ 6/5, the solution with the central density ρc of the
Tolman-Oppenheimer-Volkoff equation, that is, (1) with Λ = 0, or (2), can be short, if ρc is
large and the function P(ρ) is very much different from the exact γ-law for large ρ. For a suffi-
cient condition for solutions to be short, see Ref. 14 [Proposition 3] or the proof of Ref. 12
[Theorem 1]. Therefore we can consider such a case, supposing that 1 < γ < 2 and the solution
(m, P) = (m0(r), P0(r)), 0 < r < r +0 , of (2) with the same central density ρc satisfies P0(r) → 0 as
r → r +0 − 0, with r +0 being finite. Then the associated (m,u) = (m0(r),u0(r)), 0 < r < r +0 , satisfies (8)
with Λ = 0, that is,
1
dm
= 4πr 2 A1u♯γ−1 Ω ρ (u/c2), (16)
dr
γ
4π 3 γ γ−1
+ 2


du G m 2
r AA 1 ♯ Ω P (u/c )
u
=− c .
dr 2Gm

r2 1 − 2
cr

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-7 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

In order to extend (m0(r),u0(r)) onto r ≥ r +0 , we put

m0(r) = m+0 (B m0(r +0 )),

c2 2Gm0
2Gm0


u0(r) = log 1 − 2 0+ − log 1 − 2 + ,
2 c r+ cr
for r ≥ r +0 . Then the extended (m0(r),u0(r)) satisfies (16) on 0 < r < +∞. Since the right-hand side
of (8) tends to that of (16) as Λ → 0, the solution of (8) under consideration exists and remains in a
neighborhood of the orbit (m0(r),u0(r)) on 0 < r ≤ r +0 + δr , provided that δr is a sufficiently small
positive number and Λ is also sufficiently small. Thus the following result follows immediately.

Theorem 2. Suppose that the solution of the Tolman-Oppenheimer-Volkoff equation (2) with
central density ρc is short. Then there exists a small positive number ϵ 1 such that if Λ ≤ ϵ 1, the
solution germ of the Tolman-Oppenheimer-Volkoff-de Sitter equation (1) with the central density ρc
has a monotone-short prolongation.

Here we point out that ϵ 1 may depend not only upon γ and the function Ω(·) but also upon
A, c, G, and ρc . In contrast with Theorem 1, we cannot specify the manner of its dependence.

III. MONOTONICITY
In this section, we study the monotonicity of the solution of the Tolman-Oppenheimer-Volkoff-
de Sitter equation (1).
When we studied the Tolman-Oppenheimer-Volkoff equation, that is, (1) with Λ = 0, or (2), we
see that if ]0,r +[,r + ≤ +∞, is the right maximal interval of existence of the solution in the domain
D, then
dP/dr < 0 for 0 < r < r + and P → 0 as r → r + − 0.
Note that proof is given in Refs. 12 and 14. In other words, we can say on the Tolman-Oppenheimer-
Volkoff equation that if the prolongation of the solution germ under consideration is short, it is
necessarily monotone-short. However it is not the case for the Tolman-Oppenheimer-Volkoff-de
Sitter equation with Λ > 0. Even if dP/dr < 0 for 0 < r ≪ 1 under the assumption (7), dP/dr may
turn out to be positive during the prolongation. Let us show it in the sequel.
In order to fix the idea, we suppose uc = 1 and put
c2
r = aR, m = a3 · 4π A1 M, 4πG A1a2 = 1, u = U, λ= Λ. (17)
4πG A1
Then system (1) is reduced to
dM
= R2U µ Ω ρ (U/c2), (18)
dR
dU 1 γ − 1 R3 µ+1 λ
2M λ
−1
=− 2 M+ U Ω P (U/c2) − R3 1 − 2 − 2 R2 . (19)
dR R γ c 2 3 c R 3c
The right-hand side of system (19) depends continuously upon the speed of light c and tends to
1 λ

T
R2U µ , − M − R3 as c → ∞.
R2 3
This non-relativistic limit equation can be written as
1 d 2 dU

− R = U µ − λ. (20)
R2 dR dR
In this situation, we assume that Λ depends upon c and c2Λ/(4πG A1) tends to λ. Since (20) is the
Lane-Emden equation when λ = 0, we shall call it “the Lane-Emden-de Sitter equation” supposing
that λ > 0.

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-8 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

Although we are supposing 1 < µ < +∞(⇔ 1 < γ < 2), we observe the limiting case µ = 1
(⇔ γ = 2). Then Equation (20) is linear and the solution U = Û(R) with Û(0) = 1 is given by
sin R
Û(R) = λ + (1 − λ)
R
explicitly. It follows that
1−λ 2
Û(R) = 1 − R + O(R4) as R → +0,
6
and condition (7) (i.e., Λ < Λ∗) reads λ < 1. Suppose that 1/2 ≤ λ < 1. Then we find that the
derivative
dÛ 1−λ sin R

= cos R −
dR R R
turns out to be positive for 3π/2 < R < 2π and so on. Thus Û(R) > 0 exists and oscillates on
0 < R < +∞ and converges to λ as R → +∞. Therefore, this explicit example tells us that if γ is
near to 2, c is sufficiently large, and c2Λ/(4πG A1) is near to a number λ in the interval [ 21 , 1[, then
the behavior of the solution under consideration may be similar, that is, the prolongation of the
solution germ with uc = 1 is not monotone.
Next, we spend few words concerning the conditions κ + > 0 and Q+ > 0 in Definition 1
for the monotone-short solutions of the Tolman-Oppenheimer-Volkoff-de Sitter equation. Let us
consider a solution (m, P) = (m(r), P(r)), 0 < r < r +, in D such that dP/dr < 0, which requires
Q(r, m(r), P(r)) > 0 for 0 < r < r +, and suppose P(r) → 0 as r → r + − 0, with r + being finite.
Note that when we are concerned with the Tolman-Oppenheimer-Volkoff equation (2) with
Λ = 0, the condition κ + > 0 follows automatically. The proof can be found in Ref. 14. Of course,
if Λ = 0, then Q+ = Gm+ > 0 a priori. However if Λ > 0, it seems that we cannot exclude the
possibility that κ + = 0 a priori. Generally speaking, since κ > 0 in D, we have κ + ≥ 0. Suppose
κ + = 0. Then
2Gm+ Λ
= 1 − r +2 ,
c2r + 3
and, since κ > 0 for r < r + and κ + = 0, we see
dκ  2G 2Gm 1 2
κ +′ B = lim − 4π ρ + 2 2 − Λr
dr r =r+−0 r → r+−0 c2

c r 3
2Gm+ 1 2 1
= − Λr + = (1 − Λr +2 ) ≤ 0.
c2 r +2 3 r+
On the other hand, we have
c2Λ 3 c2r +
Q+ = Gm+ − r = (1 − Λr +2 ) ≥ 0,
3 + 2
since Q > 0 for r < r +. Thus it should be the case that 1 − Λr +2 = 0 and Q+ = 0. In other words,
κ + = 0 requires Q+ = 0 and Λr +2 = 1. This is a very non-generic situation probably hard to occur, but
at the moment we have no reason to exclude this possibility.

IV. METRIC ON THE VACUUM REGION

Now we show that the interior metric of the solution can be connected with the Schwarzschild-
de Sitter metric on the exterior vacuum region in a smooth way.
Let (m(r), P(r)) be a fixed monotone-short solution of the Tolman-Oppenheimer-Volkoff-de
Sitter equation (1) for 0 < r < r +. Then it follows the metric
2 1 2
ds2 = κ +e−2u/c c2dt 2 − dr − r 2dω2
κ
on 0 ≤ r < r +, where dω2 = dθ 2 + sin2θdφ2. We plan to continue this metric to the exterior vacuum
domain r ≥ r +. According to the Birkhoff theorem,1 any spherically symmetric solution of the

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-9 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

vacuum field equations must be static and asymptotically flat. This means that we should take the
Schwarzschild-de Sitter metric
2Gm+ Λ
2Gm+ Λ
−1
ds2 = 1 − 2 − r 2 c2dt 2 − 1 − 2 − r 2 dr 2 − r 2dω2
cr 3 cr 3
on r ≥ r +. As the whole, we take
ds2 = g00c2dt 2 − g11dr 2 − r 2dω2,
where
2


 κ +e−2u(r )/c , if 0 ≤ r < r +
g00 =  ,

2Gm Λ
 1 − 2 + − r 2,

 if r + ≤ r < r E
 cr 3
2G m̃(r) Λ 2
−1
−g11 = 1 − − r if 0 ≤ r < r E ,
c2r 3
with
 m(r),
 if 0 ≤ r < r +
m̃(r) = 
 m+, .
 if r+ ≤ r < r E
Here the constants r E and r I (0 < r I < r E < +∞) are the values of r which are called the “cosmo-
logical horizon” and “black hole horizon,” respectively, that is, κ(r, m+) > 0 if and only if r I < r <
r E . In other words, we have
Λ
κ(r, m+) = (r − r I )(r E − r)(r + r I + r E ).
3r
See Ref. 4. But this situation is possible only if
√ c2
Λ< . (21)
3Gm+
If (21) does not hold, then κ(r, m+) ≤ 0 for all r > 0. However, since we assume κ + = κ(r +, m+) > 0,
condition (21) is supposed to hold and we have r I < r + < r E .
Now we discuss the regularity of this patched metric. First we obtain the following regularity of
u(r).
Proposition 2. The function u(r) is of class C 2 in a neighborhood of r + and
u(r) = B(r + − r)(1 + O(r + − r)) as r → r + − 0,
where B B Q+/r +2 κ +. Hence ρ(r) is of class C 1 and
(γ − 1)B
γ−1
1 1
ρ(r) = (r + − r) γ−1 (1 + O(r + − r)).
γA

Proof. Since u(r) satisfies Equation (8), whose right-hand side is a C 1-function of (r, m,u) near
(r +, m+, 0) thanks to µ = 1/(γ − 1) > 1. Recall that κ + > 0. Therefore, the continuous solution u(r)
turns out to be of class C 2 and satisfies
du  Q+
=r
= − 2 = −B.
dr r + −0
r+κ+
This completes the proof. 
Next, we consider the regularity of g00 and g11. Since
d  4πr 2 ρ(r), if r < r +
m̃(r) =


dr  0,
 if r + ≤ r < r E
is of class C 1, it implies that m̃(r) is of class C 2. Therefore g11 is twice continuously differentiable
across r = r +.

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-10 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

As for g00, since u vanishes at r = r + − 0, it is continuous thanks to the definition of κ +.

Elementary computations give
d 2κ + du  2Q+
g00 r =r+−0 = − 2 =

dr c dr r =r+−0 c2r +2
and
d 2Gm+ 2Λ
 2Q+
g00 r =r++0 = r r =r+ = 2 2 .

−
dr c2r 2 3 c r+
Therefore g00 is continuously differentiable. Moreover, we have
d2 4κ + du
2 2κ + d 2u

g = .

00 r =r +−0
− (22)
dr 2 c4 dr r =r+−0 c2 dr 2 r =r+−0
By differentiating the right-hand side of the equation for du/dr (the second equation of (8)) with a
tedious calculation, we have
d 2u  c2Λ 2Q+ 2(Q+)2
=r
= + 3 + 2 4 2. (23)
dr 2 r + −0 κ+ r+κ+ c r+κ+
Therefore, it follow from (22) and (23) that
d2 d2 4Q+
g = g00 r =r++0 = − 2 3 − 2Λ.
 
2 00 r =r + −0 2
dr dr c r+
Hence g00 is twice continuously differentiable across r = r +. Summing up, we have the following
result.
Theorem 3. Given a monotone-short solution of the Tolman-Oppenheimer-Volkoff-de Sitter
equation (1), we can extend the interior metric to the exterior Schwarzschild-de Sitter metric on the
vacuum region with twice continuous differentiability.

V. ANALYTICAL PROPERTY OF THE VACUUM BOUNDARY

Let us investigate the analytical property of a monotone-short solution (m(r), P(r)), 0 < r < r +,
of the Tolman-Oppenheimer-Volkoff-de Sitter equation (1). According to Proposition 2, we know
that the associated u(r) belongs to C 2([0,r +]) and
u(r) = B(r + − r)(1 + O(r + − r)) as r → r + − 0,
where B = Q+/r +2 κ +. To obtain the above high order terms of u(r) precisely, we first establish the
following lemma.

Lemma 1. Let µ > 1 and f α (x, x µ , y1, y2), α = 1, 2, be analytic functions of x, x µ , y1, y2 on a
neighborhood of (0, 0, 0, 0). Let ( y1(x), y2(x)), 0 ≤ x ≤ δ, be the solution of the problem
dyα
= f α (x, x µ , y1, y2), yα | x=0 = 0, α = 1, 2. (24)
dx
Then there are analytic functions ϕα of x, x µ on a neighborhood of (0, 0) such that yα (x) =
xϕα (x, x µ ) for 0 < x ≪ 1.
Proof. First we assume that µ is not an integer but a rational number, say, µ = q/p with
p, q ∈ N are relatively prime and p ≥ 2. Note that a function given by a convergent power series

ϕ(x) = c̃i j x i (x µ ) j , | c̃i j | ≤ M̃/δ̃ i+ j , (δ̃ < 1)
can be rewritten as
p−1

ϕ(x) = cl n x µn+l , |cl n | ≤ M/δ l for 0 ≤ n ≤ p − 1,
l n=0

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-11 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

where

cl n = { c̃i j | i + qJ = l, j = pJ + n, ∃J ∈ N},
M = M̃/(e δ̃ p ), and δ = δ̃/e. This rewriting is necessary, since not c̃i j ’s but cl n ’s can be uniquely
determined for the given function ϕ(x).
In fact, let us claim that µn cannot be an integer for n = 1, . . . , p − 1. Suppose the claim is false,
i.e., nq/p is an integer. We can assume q < p, by, if necessary, replacing q by q ′ B q − [q/p]p.
Since nq/p < n, we see that nq/p is either 1, . . . or n − 1, therefore q/p is either 1/n, . . . or
(n − 1)/n. Hence p is a divisor of n, a fortiori, p ≤ n, a contradiction to n ≤ p − 1. Hence the claim
holds. Therefore, if µn + l = µn ′ + l ′, n, n ′, l, l ′ ∈ N, 0 ≤ n, n ′ ≤ p − 1, it follows that n = n ′, l = l ′.
Thus we have a unique numbering (nk , l k )k ∈N of (n, l)’s such that µnk + l k < µnk+1 + l k+1. By
induction on k we can deduce cl k n k = 0 for ∀k from cl n x µn+l = 0 ∀x. This means the uniqueness

of the coefficients cl n in the above expansion of ϕ(x).
Anyway, we suppose
p−1
f α (x, x µ , y1, y2) =

alαnk1k2 x µn+l y1 1 y2 2,
k k

n=0

with
alαnk1k2 ≤ M/δ l+k1+k2, where 0 ≤ n ≤ p − 1
 

and put
p−1
 M  M 1 − xµp 1 1
x µn+l y1 1 y2 2 =
k k
F(x, y1, y2) = µ
.
δ l+k1+k2 n=0
1 − x/δ 1 − x 1 − y1/δ 1 − y2/δ
Then the problem
dY
= F(x,Y,Y ), Y | x=0 = 0
dx
has a solution of the form
p−1
µ

Y = M x(1 + [x, x ]1) = Cl n x µn+l , where 0 ≤ Cl n ≤ M ′/(δ ′)l .
l n=0

On the other hand, Equation (24) has a formal power series solution
p−1

yα = clαn x µn+l ,
l n=0

where the coefficients clαn ’s are determined by the recursive formula,

α α 1
c0n = 0, cl+1, n = bαl n ,
l + 1 + µn
bαL R =

alαnk1k2cl1′(1)n′(1) · · · cl1′(k1)n′(k1)cl2′′(1)n′′(1) · · · cl2′′(k2)n′′(k2).
Here the summation in the definition of bαL R is taken over
L = qJ + l + l ′(1) + · · · + l ′(k1) + l ′′(1) + · · · + l ′′(k2)
with l ′(1), . . . , l ′′(1), . . . , ≥ 1 and J ∈ N, and
pJ + R = n + n ′(1) + · · · + n ′(k1) + n ′′(1) + · · · + n ′′(k 2).
Then it can be shown inductively that |clαn | ≤ Cl n , which implies the convergence of the formal
power series solution.
A proof by a similar and easier majorant argument can be done when µ is an irrational number.
We omit the repetition. This completes the proof. 
Applying the result of Lemma 1, we can obtain the following result.

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-12 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

Theorem 4. Any monotone-short solution u(r), 0 < r < r +, of (1) enjoys the behavior at r =
r + − 0 such that
γ
u(r) = B(r + − r) 1 + [r + − r, (r + − r) γ−1 ]1 ,

therefore
(γ − 1)B
γ−1
1 1 γ
ρ(r) = (r + − r) γ−1 1 + [r + − r, (r + − r) γ−1 ]1 .


γA
Here [X1, X2]1 stands for a convergent double power series of the form

k k
a k 1k 2 X 1 1 X 2 2.
k 1+k 2 ≥1

Proof. Put µ = 1/(γ − 1), so that γ/(γ − 1) = µ + 1. First, we assume that µ is an integer. Then
proof for this case is easy. In fact, (m(r),u(r)) satisfies the system
dm
= 4πr 2 A1u µ Ω ρ (u/c2),
dr
4π 3 γ µ+1
c 2Λ 3
du G m + r AA 1 u Ω P (u/c 2
) − r
=− c2 3 (25)
dr 2Gm Λ

r2 1 − 2 − r2
cr 3
at least on 0 < r < r + and (m(r),u(r)) → (m+, 0) as r → r + − 0. Since µ is an integer, the right-hand
side of system (25) is analytic function of (r, m,u) in a neighborhood of (r +, m+, 0). This guaran-
tees that m(r),u(r) admit analytic prolongations beyond r = r + to the right. Hence the assertion of
the theorem holds. Of course this analytic prolongation is different from the C 2-prolongation as a
solution of (8), since u µ , u♯µ (= 0) for u < 0.
Next, we consider the case that µ is not an integer. Since u(r) is monotone decreasing on
0 < r < r +, it follow that u(r) has the inverse function r = r(u) defined on 0 < u < uc such that
r(u) → r + as u → +0. Then we have a solution (m,r) = (m(u),r(u)) of the system
dm 2Gm Λ

= −4πr 4 1 − 2 − r 2 · Q−1 · A1u µ Ω ρ (u/c2), (26a)
du cr 3
dr 2Gm Λ
= −r 2 1 − 2 − r 2 · Q−1,


(26b)
du cr 3
where
4π 3 γ µ+1
c 2Λ 3
Q=G m+ 2
r AA1 u Ω P (u/c2) − r .
c 3
Since (m(u),r(u)) → (m+,r +) and Q → Q+ > 0 as u → +0 and the right-hand sides of (26a) and
(26b) are analytic functions of u,u µ , m,r on a neighborhood of (0, 0, m+,r +), we can apply Lemma 1
to derive
m(u) = m+ + u[u,u µ ]0 and r(u) = r + + u[u,u µ ]0.

Here [·, ·]0 denotes a convergent double power series.

Since dm/du ∼ −Cu µ , with C = 4πr +4 (κ +/Q+)A1 and du/dr → −B, we have

m = m+ − Cu µ+1 +

m1nu µn+1 +

ml nu µn+l ,
n ≥2 n ≥0, l ≥2
1 
c1nu µn+1 +

r = r+ − u + cl nu µn+l .
B n ≥1 n ≥0, l ≥2

If m1n , 0 for some n ≥ 2, then dm/du would contain the term u µn with n ≥ 2. However, it is
impossible since the right-hand side of (26a) cannot contain such a term. Therefore, m1n = 0 for
∀n ≥ 2. If c1n , 0 for some n ≥ 1, then dr/du would contain the term u µn . However, it is still

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-13 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

impossible since the right-hand side of Equation (26b) cannot contain such a term. Hence c1n = 0
for ∀n ≥ 1. Thus we have

m = m+ − Cu µ+1 + ml nu µn+l ,
n ≥0, l ≥2
1 
r = r+ − u + cl nu µn+l .
B n ≥0, l ≥2

Moreover, we can show that ml n = cl n = 0 for 2 ≤ l ≤ n by induction on l. In fact, fix n ≥ 2, we

claim that m2n = c2n = 0. Since, otherwise, dm/du, the right-hand side of (26a), or dr/du, the
right-hand side of (26b), would contain the term u µn+1 with n ≥ 2, which is impossible. Now we
consider 3 ≤ l ≤ n. Assume that m n′,l ′ = cn′,l ′ = 0 for 2 ≤ l ′ ≤ n ′ with n ′ ≤ n, l ′ ≤ l − 1. If m nl , 0,
or, cnl , 0, then dm/du or dr/du would contain the term u µn+l−1, which is impossible by the
induction assumption. Therefore m nl = cnl = 0 for 2 ≤ l ≤ n. This implies that
1  1
r = r+ − u + cl nu(µ+1)n+l−n = r + − u(1 + [u,u µ+1]1).
B n ≥0, l ≥n+1, l ≥2
B

The inverse function u = u(r) then clearly enjoys an expansion of the form
u = B(r + − r) 1 + [r + − r, (r + − r) µ+1]1 .

This completes the proof. 

VI. DISCUSSION ABOUT LARGE Λ

Since we are interested in solutions with monotone decreasing ρ(r), we have been confined to
the case in which the inequality Λ < Λ∗ holds. In this section, we discuss about the case in which Λ
is large.

A. Λ = Λ∗
In this case, we have the constant density solution
(m(r), P(r)) = (4π ρc r 3/3, Pc ), 0<r <

3/L,
8πG
where L B 2 ρc + Λ. In this very special case, the metric is reduced to
c
L
−1
ds2 = c2dt 2 − 1 − r 2 dr 2 − r 2(dθ 2 + sin2θdφ2)
3
after a suitable change of the scale of t. The space √ at t = Const. is isometric to the half of the
compact 3-dimensional hypersphere with radius 3/L embedded in the 4-dimensional Euclidean
space R4 = {(ξ 1, ξ 2, ξ 3, ξ 4)} through

ξ 1 = r sin θ cos φ, ξ 2 = r sin θ sin φ, ξ 3 = r cos θ, ξ 4 = 3/L − r 2.
√
Thus the horizon r = 3/L − 0 is merely apparent. Of course, the condition Λ = Λ∗ is highly
unstable. This is nothing but the “Einstein’s steady state inverse (1917)” proposed in Ref. 8.

B. Λ > Λ∗
If Λ > Λ∗, then the solution germ satisfies dP/dr > 0 for 0 < r ≪ 1, that is, ρ(r) is monotone
increasing near the center. However, it is possible that dP/dr becomes negative and ρ(r) begins to
be decreasing when it is prolonged to the right. We show this fact in the sequel.
Let us recall that under normalization (17) considered in Section III, system (1) reduces to
Equations (18) and (19), which tends to the Lane-Emden-de Sitter equation,

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-14 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

1 d 2 dU

− R = U µ − λ,
R2 dR dR
as c → +∞. If µ = 1(⇔ γ = 2), the solution is
sin R
U = Û(R) = λ + (1 − λ) ,
R
which behaves as
1−λ 2
Û(R) = 1 − R + O(R4) as R → +0.
6
Now the condition Λ > Λ∗ reads λ > 1. Then dÛ/dR > 0 for 0 < R ≪ 1 but dÛ/dR becomes
negative and Û(R) oscillates and tends to the limit λ as R → +∞. Therefore, this explicit example
tells us that if γ is near to 2, c is sufficiently large, and c2Λ/(4πG A1) is near to λ > 1, then the
prolongation P(r) of the germ, which satisfies dP/dr > 0 as 0 < r ≪ 1, becomes decreasing.
Note that even if the density is increasing at the center, if the prolongation ρ(r) finally becomes
decreasing and hits zero as r approaches a finite r + with positive Q+, κ +, then maybe such a config-
uration should not be rejected as physically meaningless. The question whether such a behavior can
be observed or not, when Λ > Λ∗, remains as the task of study in the future.

√ c2
C. Λ≥ 3Gm+

Moreover, we have restricted our considerations to equilibrium configurations for which κ + =

κ(r +, m+) > 0. This a posteriori condition κ + > 0 guarantees that the inequality (21)
√ c2
Λ<
3Gm+
holds. Thus the inequality (21) is another restriction
√ on the magnitude of the admissible cosmolog-
ical constant Λ. However it is known that if Λ approaches c2/3Gm+ and (21) is corrupting, the
black hole horizon approaches the cosmological horizon on the metric in the vacuum region, or the
black hole singularity runs off to infinity. This situation is so-called “Nariai limit” (see Refs. 15 and
16), and it is a physically interesting case. But we have not yet done sufficient analysis about this
situation, and here we have a task reserved for the future, too.

ACKNOWLEDGMENTS
The authors would like to express their sincere thanks to the anonymous referee for reading the
manuscript carefully and giving helpful suggestions to improve the presentation, particularly, about
discussion for large Λ.
1 Birkhoff, G. D., Relativity and Modern Physics (Harvard University Press, Cambridge, Massachusetts, 1923).
2 Böhmer, C. G., “General relativistic static fluid solutions with cosmological constant,” Ph. D. Diplomarbeit, Max-Planck-
Institut für Gravitationsphysik (Albert-Einstein-Institut), Golm, 2002; e-print arXiv:gr-qc/0308057.
3 Böhmer, C. G., “Static perfect fluid balls with given equation of state and cosmological constant,” Ukr. J. Phys. 50,

1219–1225 (2005).
4 Brady, P. R., Chambers, C. M., Laarakkers, W. G., and Poisson, E., “Radiative falloff in Schwarzschild-de Sitter spacetime,”

Phys. Rev. D 60, 064003 (1999).

5 Carroll, S. M., “The cosmological constant,” Living Rev. Relativ. 4, 1 (2001).
6 Chandrasekhar, S., An Introduction to the Study of Stellar Structure (University of Chicago Press, 1939).
7 Coddington, E. A. and Levinson, N., Theory of Ordinary Differential Equations (McGraw-Hill, 1955).
8 Einstein, A., “Kosmologische Betrachungen zur allgemeinen relativitätstheorie,” Sitzungsber. Preuss. Akad. Wiss. VI,

142–152 (1917).
9 Einstein, A., “Zum kosmologischen Problem der allgemeinen relativitätstheorie,” Sitzungsber. Preuss. Akad. Wiss. XII,

235–237 (1931).
10 Joseph, D. D. and Lundgren, T. S., “Quasilinear Dirichlet problem driven by positive sources,” Arch. Ration. Mech. Anal.

49, 241–269 (1973).

11 Landau, L. D. and Lifshitz, E. M., The Classical Theory of Fields, 4th ed. (Pergamon Press, Oxford, 1975) [Teorija Polja

(Nauka, Moskva, 1973)].

12 Makino, T., “On spherically symmetric stellar models in general relativity,” J. Math. Kyoto Univ. 38(1), 55–69 (1998).

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58
 092502-15 C.-H. Hsu and T. Makino J. Math. Phys. 57, 092502 (2016)

13 Makino, T., “On the spiral structure of the (R,M)-diagram for a stellar model of the Tolman-Oppenheimer-Volkoff equation,”
Funkcial. Ekvac. 43, 471–489 (2000).
14 Makino, T., “On spherically symmetric solutions of the Einstein-Euler equations,” Kyoto J. Math. 56(2), 243–282 (2016).
15 Nariai, H., “On some static solutions of Einstein’s gravitational field equations in a spherically symmetric case,” Gen. Rel.

Gravit. 31(6), 951–961 (1999).

16 Nariai, H., “On a new cosmological solution of Einstein’s field equations of gravitation,” Gen. Rel. Gravit. 31(6), 963–971

(1999).
17 Oppenheimer, J. P. and Volkoff, G. M., “On massive neutron cores,” Phys. Rev. 55, 374–381 (1939).
18 Pais, A., Subtle is the Lord: The Science and the Life of Albert Einstein (Oxford University Press, 1982).
19 Rendall, A. D. and Schmidt, B. G., “Existence and properties of spherically symmetric static fluid bodies with a given

equation of state,” Classical Quantum Gravity 8, 985–1000 (1991).

20 de Sitter, W., “On the relativity of inertia. Remarks concerning Einstein’s latest hypothesis,” Proc. R. Acad. Amsterdam

XIX, 1217–1225 (1917).

21 Zeldovich, Y. B. and Novikov, I. D., Relativistic Astrophysics (University of Chicago Press, 1971), Vol. I.

Reuse of AIP Publishing content is subject to the terms: https://publishing.aip.org/authors/rights-and-permissions. Downloaded to IP: 146.83.173.12
On: Fri, 21 Oct 2016 20:29:58