VOLUME

PHYSICAL REVIEW LETTERS

57, NUamER 20

17 NOvEMmR 1986

Equilibria of Self-Interacting

Boson Stars: Gravitational

Scalar Fields

Monica Colpi, t'l Stuart L. Shapiro, and Ira Wasserman

Center for Radiophysics and Space Research, Cornell Uni Uersity, Ithaca,

(Received 13 August 1986)

York I 4853

Spherically symmetric gravitational equilibria of self-interacting scalar fields p with interaction potential V(p) ,k p)" are determined. Surprisingly, the resulting configuration may differ markedly from
the noninteracting case even when A, &&1. Contrary to generally accepted astrophysical folklore, it is
found that the maximum masses of such boson stars may be comparable to the Chandrasekhar mass for
fermions of mass mf . -X, '~4mb
~

PACS numbers: 04.20Jb, 11.10.-z, 95.30.Sf

Recent speculations in particle physics and cosmology

have emphasized the potentially critical role played by
evolving scalar fields in the development of the early
Universe, ' and raised the possibility that various exotic
bosons comprise the cosmological missing mass. z3 In
view of this development, a detailed examination of the
gravitational equilibria of massive fields in asymptotical"boson stars" seems appropriate.
ly flat space-times
Earlier work has considered equilibrium configurations of noninteracting boson fields. 4 Such objects are
macroscopic quantum states, that are only prevented
from collapsing gravitationally by the Heisenberg uncertainty principle. This requiress that a state of characteristic size R has a typical boson momentum p 1/R
(here and throughout we set h 1 =c ). In a moderately
relativistic boson star one has p -rn, where m is the boson mass, so that R
I/m. Since hydrostatic equilibrium requires that the total mass M-R/G for relativistic
bound states, we find M -(Gm) ' M)~k/m for marginally relativistic boson stars. This mass is generally
much smaller than the Chandrasekhar
mass,
-M)l, k/m, characteristic of marginally relativistic
fermion st ras
In this paper we consider arbitrarily relativistic equihbria of self interacting c-omplex scalar fields. We as'
sume an interaction potential V(p) ,
where A, is
A, ( p (
a dimensionless coupling constant. We demonstrate explicitly that euen if A, ((1 the structure of the resulting
boson stars may differ radically from the A, 0 limit
To see why, first notice that the characteristic energy
density inside a A, 0 configuration is p-M)hkm in
the relativistic case. Since the energy density of noninbosons
teracting
is p m ( p ~, we
that
find
-Mp~, k inside relativistic X 0 boson stars. Next,
( p
consider the effect of increasing X, from zero. The importance of the interaction potential is measured by the
ratio V(p)/m ( p 2 of interaction energy to kinetic enerAt sufficiently
small X,, this ratio is just
gy.
A, M)t, k/m
. Thus, self-interactions may only be ignored if

s.

Moreover, families of gravitational equilibria may be

parametrized by the single dimensionless quantity
A

XM)(ak /4%m

for A))1 we find that M-A'I2M(~, ,k/

Mch, so that the masses of relatiuistic boson
stars may be comparable to those of their ferrnion coun
l.
terparts if
The starting point for our calculation is the scalarfield Lagrangean
Surprisingly,

m-X'

k-

J'-

lg""e,'-O,

&m'I

I' l&

(3)

I-',

tensor

which implies an energy-momentum

T," --'g" (~~,, +~,~,'.)

'&."&g"-~~-e;.+m'I e I'+ '&

~ I'j

-(4)

We explicitly ignore interactions of p with any other

fields. In particular, Eq. (3) assumes negligible coupling
to gauge fields, which is a good approximation for gauge
coupling constant e ((m/Mp~, k. Calculations for larger
values of e are currently under way.
We consider spherically symmetric, time-independent
solutions of Einstein's field equations

6" 8trGT"
coordinates

in Schwarzschild

ds2

(5)

B(r)dtz+A(r)dr2+r2d

Q, .

(6)

For such solutions to exist we require that

y(r, t)

@(r)e

((m

/M)tgngk

(6 7 x 10

GeV

)m

where @(r) is a real function. [We could equally well

consider "antiboson stars" with p(r, t) @(r)e' '. ] For
convenience we actually solve the (;) and (,') components of Eq. (5) coupled with the scalar wave equation

~-m'~-& e 'e-0,
I

which may be derived

1986 The American Physical Society

(8)

from Eq.

(5)

using Eq. (4) and

2485

pHYSI &L REVIEW LETTERS

NUMBER 20

VQLUME 57

i
tthee Bianchi
identities.
ties

Thee resultmg equuuations are

A'+1
+

A 4+
a+
+I cr+

x2

A2x

yEMBER 198
86

(0) a a 0) -0,' an d ~( )

()

fit

a')z
problem for &

a"
~, (r) have no no

(9

8'

ABx

x2

numerical

Acr

)' @/Mpi,g,

/,

If we ~rite
A

(4rrG ) 1/2C
by q.

denote d/dx,x, a
co rri an dAi gi

rimes

(x) [1 2AI(x)/x]

we may rewritete

At'(x)

',

E uation
e

hE.

(10

i,

that characteristic'c boson-

~max
m

= 0 22A'

o1 ~

'~Rane) /rii.
11y

derived
for A)&1 and is de

may bee seen

01 b tdiff
fl
ates
ill
11 and large

erivation

in our inin

from Fig. 3
t Ilues of A:
the different
values o

ra't""t" l'ngt .
rz d b lt

els are c
chara t
out to radii r-A

h
h

contributions s of the cr')'/2~

(9) are solved ssubject
j to the

'

s cr(x) drops smoo

h

8oona

e'

for severa diffent valu~ o

Our re
.
ellent
e en agreement with those of

trend mo
masses as a function o
2 is the relation

use

conditions

boundary

'a

'

nentia
ec
mass in the large-

g"

T..

th

Th'

'

A increases.
Figure 3 suggests an a lternative non dimensionaltzation

Ao

(9a')

w
9a ',) was
substituteted for Eq.. (9a)

seems reasonab e o

which h oldss asymptotica

below.

+
+I a ~
cr+

Following Ruffini an

). It

(10)

Eq.. (9a) as

-x'

it)

.
ing.

0,

(9c)
(4

shown

suits for

8'
+ 2+
+

(i.e. , reg"

(a')'

(9b)
g

ess at

.2

200

300

0~
FIG.. 1. Boson-star mass M as
a a function or a.
o, -(4m)
for A 0, 1, 10, 30, 100,

&/2

e.

boson-star
The so 1'd
I. curve
c
is the asymp

mas
nn
asses as a function
un

of A.

PHYSICAL REVIEW LETTERS

VoLUME 57, NUMBER 20

17 NovEnmER 1986

20

10

io (n'/B(o))
FIG. 3. Scalar
x for cz, 0. 1 but

field o as a function of dimensionless radius

A 0 and A 300. The dots are the solution
to Eqs. (12)-(15) with 02/8(0) from the exact cr, O. l,
A 300 model, scaled to A 300.

of Eqs. (9a'), (9b), and (9c) accurate at large A:

and At, At/A'/
Ignor. ing
o, oA'/, x, xA
terms O(A '), the scalar wave equation may be solved

',

to yield

algebraically

(0 /8 1)'

cr

which may be substituted

to the same accuracy,

4'

JKy

into the field equations to give,

(13)

y Py

FIG. 4. Boson-star mass vs 02/8(Q) for the limiting equilibria computed from Eqs. (12)-(l5).

Because of self-gravity, the ground state of the boson

field is not a zero-energy state. Moreover, the scalar
field, at large A, only varies on a relatively large length
scale A'/ m '&)m ', so that we can solve the scalar
wave equation locally, ignoring derivatives, to get Eq.
(12). This results in an effective equation of state, Eqs.
(15a) and 15(b), for the boson star, or on elimination of
fI2/8 and restoration of dimensional quantities

where pn

m4/4X and

+(pjpn)

and
1

x2

where prime denotes

8'~,

d/dx,

'/+)(

- '~-&(~2/g 1)2

(12)- (15)

and

'/

(15a)
(15b)

become essentially
exact for
A
At large A these equations may be used to generate approximate solutions. Figure 3 compares n(x)
based on Eqs. (12)- (15) with the exact o (x) for
A 300. As expected, the two calculations agree well
except at very large radii. Because A does not appear
explicitly in Eqs. (13)-(15), we can use these limiting
equations to determine the rescaled mass At~ M/
(A'~ M(~,qjrn) as a function of the single free parameter
/8(0). The results, shown in Fig. 4, imply the
peak value At
22 used in Eq. (11).

Equations

*=0.

((I+ ,' pipn)

I 1'.

'~'

(11) is

then equivalent to the statement that

for a fluid star with an equation of
state of the form of Eq. (16). It is straightforward to
show that the well-known theorems on stability of fluid
oo limiting scalar-field
stars may be applied to the A
equilibria satisfying Eqs. (12)-(15). In particular, M
denotes the boundary along the sequence between stable
and unstable equilibria as A
It seems reasonable
to suppose that the same stability criteria may be applied
for all (finite) A, although we have not proven this statement in general.
Generally accepted astrophysical folklore maintains
that boson stars, should they exist, must have negligibly
small masses. %'bile detailed studies have corroborated
this prejudice for bound states of noninteracting bosons,
the situation for self-interacting scalar fields may be very
different. In this paper we have shown that for an in'k p( much larger masses
teraction potential V(p) ,
will result provided that A=A/4x(Jm
1, an inequality

Equation

(14)

(16)

p -poF(p/po).

M~, -M$~,,q/p)

))

VOLUME 57, NUMBER 20

PHYSICAL REVIEW LETTERS

that may be satisfied even at A, &&1 for reasonable

scalar-boson masses. When A)&1 we have found that
the maximum boson-star mass is
M~gg

0.22A

M)[spy /m

-(0.10 Gev')~. ~'"/m',

which is comparable to the Chandrasekhar mass for fermions of mass m/1, 'l . It is conceivable that boson stars
with masses approaching
Eq. (18) could arise in the
course of gravitational
condensation of bosonic dark
matter in the early Universe. Equation (18) would imply masses possibly near, but below, the stellar mass
with m
1 GeV, but would
range for scalar neutrinos
to
require
ridiculously
masses,
appear
large
' Mo, for cosmologically relevant axions3
X,
M~,
(with m
10 eV) except for exceedingly tiny )I, .
It is a pleasure to acknowledge useful conversations
with J. Hartle, R. Ruffini, and M. Sasaki. This research
was supported in part by the National Science Foundation under Grant No. AST 84-15162 to Cornell University. One of us (I.W. ) acknowledges the hospitality of
the Harvard-Smithsonian
Center for Astrophysics,
where part of this work was carried out; I. %. is also an
Alfred P. Sloan Fellow. M. C. thanks the Astronomy
Department at Cornell University for kind hospitality
during the course of this research, and acknowledges
financial support from the Ministero della Pubblica Istruzione, Italy.

-10

' On

leave from Dipartimento

di Fisica, Universita

degli

17 NOVEMBER 1986

Studi di Milano, 20133 Milano, Italy.

tA. Linde, Phys. Lett. 1088, 389 (1982); A. Albrecht and
P. J. Steinhardt, Phys. Rev. Lett. 4$, 1220 (1982); A. Albrecht, P. J. Steinhardt, M. S. Turner, and F. Wilczek, Phys.
Rev. Lett. 4$, 1437 (1982). See also M. S. Turner, in
Proceedings of the Cargese School on Fundamental Physics
and Cosmology, edited by J. Adouze and J. Tran Thanh Van
(Editions Frontieres, Gif-Sur-Yvette, France, 1985), and references therein.
2J. S. Hagelin, G. L. Kane, and S. Raby, Nucl. Phys. 8241,

638 (1984).
3J. Preskill, M. Wise, and F. Wilczek, Phys. Lett. 120$, 127
(1983); L. Abbott and P. Sikivie, Phys. Lett. 1208, 133
(1983); M. Dine and W. Fischler, Phys. Lett. 1208, 137

(1983).
4R. Ruffini and S. Bonazzola, Phys. Rev. 1$7, 1767 (1969).
The calculations performed in this reference pertain to complex scalar fields, contrary to explicit statements made by the
authors.
5See, e.g. , E. Takasugi and M. Yoshimura, Z. Phys. C 26,
241 (1984).
sEquation (8) and its complex conjugate may also be derived
by considering variations of S~ Jd'xc with respect to p and
p, respectively.

7See B. K. Harrison, K. S. Thorne, M. Wakano, and J. A.

Wheeler, Gravitation
Theory and Gravitational
Collapse
(Univ. of Chicago Press, Chicago, 1965), Chaps. 3, 5, and 7,
and Appendix B; S. Chandrasekhar, Phys. Rev. Lett. 12, 114,
437 (1964). Just as for fluid stars, Eq. (14) may be derived by
consideration of first-order variations in the total mass, defined
in Eq. (13), at fixed boson number, with use of Eqs. (12) and
(15). Second-order variations yield the usual stability theory
for the analogous fluid stars. Extension of these results to general A, for which Eqs. (12) and (15) do not hold, does not appear to be straightforward.