1195

Progress of Theoretical Physics, Vol. 49, No. 4, April 1973

Hydrodynamic Calculations of Spherical Gravitational Collapse

in the Scalar-Tensor Theory of Gravity

Takuya MATSUDA and Hidekazu NARIAI*

Department of Aeronautical Engineering, Kyoto University, Kyoto

*Research Institute for Theoretical Physics, Hiroshima University
Takehara, Hiroshima-ken

(Received November 30, 1972)

Hydrodynamic equations for spherical gravitational collapse in the scalar-tensor theory

of gravity are approximated by finite-difference equations. The dynamical motion of a gaseous
sphere is calculated numerically on the assumption that the sphere consists of a perfect gas
without energy flow and, therefore, its total mass is conserved. In order to avoid the difficulty
of ~atching of the metric and scalar fields at .the surface of the gaseous sphere, the sphere
is divided into two parts, i.e., a central core and an extended tenuous atmosphere. In the
collapsing core, scalar waves are generated around its central region at the final stage, but
their effect is not so large as to deviate various physical quantitiEs appreciably from those to
be obtained in the general relativistic treatment, except in the inner-most region of self-closure.

§ l. Introduction and summary

Brans and Dicke 1l have proposed a theory of gravitation called the scalar-
tensor theory, according to which gravitational phenomena should be described
by a combination of the metric tensor and a scalar field ¢> whose reciprocal is
in proportion to the Newtonian gravitation constant G, in contrast with Einstein's
general theory of relativity. Relevance of a theory of gravity should be decided
by its experimental or observational test, but not by our taste or philosophy. It
is therefore important to investjgate how a prediction of the Brans-Dicke theory
would depart from that in general relativity. As regards the solar gravitational
field, their predictions for the deflection of a light ray and the peripherion advance
of the Mercury are somewhat different from Einstein's. There are, however,
many unknown factors to settle the problem, such as an inner rotation of the
sun and the nature of inter-planetary plasma. Accordingly it is difficult to com-
pare directly their respective predictions with observational data.
The departure of the tw~ theories will become important only when the gen-
eral relativistic effect itself becomes dominant. Salmona2l has examined the struc-
ture of a neutron star in the Brans-Dicke theory and shown that the effect of the
inertial scalar field is small. Matsuda 3l has reexamined the problem, but his con-
clusion has been similar to Salmona's except for a slight difference in numerical
results. As the gravitational field of a collapsing object is stronger than that of
a neutron star, the phenomenon of gravitational collapse will be suitable to find

Downloaded from https://academic.oup.com/ptp/article-abstract/49/4/1195/1932304

by guest
on 28 July 2018
 1196 T. Matsuda and H. Nariai

out some characteri stic feature of the Brans-Dic ke theory of gravity to be com-
pared with Einstein's .
As regards the gravitatio nal collapse occurring in the process of galaxy for-
mation, there is a numerical analysis by Nariai and Fujimoto 4l for the dynamical
behavior of a rotating gaseous ellipsoid with uniform density in the Friedman n
universe, according to which the ellipsoid becomes free from the cosmic expan-
sion and starts to collapse _when the density contrast relative to the cosmologi cal
backgroun d attains the values of 5.4 to 5.6, provided that its angular momentum
is moderate. Dynamica l equations for such an ellipsoid in the Brans-Dic ke uni-
verse have been derived by Nariai,"l but their numerical analysis is not yet per-
formed.
On the other hand, Thorne and Dykla6 l have analyzed qualitative ly the prob-
lem of gravitatio nal collapse in the scalar-ten sor theory and concluded that the
gravitatio nal collapse produces black holes identical with those in general relativity.
Their argument is mainly based on the points: a) in the spherical gravitatio nal
collapse of an uncharged star, the scala,! field ¢ approache s a constant value when
t---HXJ, b) as a result, the field equations and their solutions must agree with those
in general relativity. However, Matsuda7l has pointed. out that their argument
is not always true if the scalar field is taken to be asymptoti cally constant. An.
other reason is that the Brans type-I solution for a sphericall y symmetric vacuum
region does not approach the usual Schwarzsc hild solution, but the truncated
Schwarzsc hild solution having a naked singularity . 7l This means that the gravi-
tational collapse beyond a non-singu lar event horizon may not occur in the scalar-
tensor theory, while the optical appearanc e of a collapsing star is not so different
from its general relativisti c counterpa rt because of the fact that the metrics in
the region R>Ru=2 GM/c2 (M is the total mass) are similar in both theories. 8l
The aim of this paper is to study the gravitatio nal collapse of a gaseous
sphere by solving numerical ly the full field equations in the scalar~tensor theory,
in contrast to Thorne and Dykla's 6l method relying on a power series expansion
of the field equations with respect to 1/w, where w is the Dicke coupling con-
stant.
The gravitatio nal collapse of -a gaseous sphere has been investigat ed by many
authors in terms of general relativity. The hydrodyna mic equations have b~en
formulated by Misner and Sharp,9 l and numerical calculatio ns of the gravitation-al
collapse of various astronomi cal objects have been performed by May and
White/ 0a),IObJ Voropinov et al., 11 l Schwarz/2l Matsuda and Sato. 13l It has been
shown that gaseous spheres with masses larger th~n a certain critical value can-
not settle into hydrostati c equilibriu m states but become black holes.
The hydrodyna mic equations requisite for our numerical analysis have recently
been formulate d by Nariai, 14l where we must solve an inhomoge neous wave equa-
tion for the scalar field as well as the field equations for th~e metric tensor. We
have encounter ed with many difficulties in numerical calculation s. a) As a scalar

Downloaded from https://academic.oup.com/ptp/article-abstract/49/4/1195/1932304

by guest
on 28 July 2018
 Hydrodynamic Calculations of Spherical Gravitational Collapse 1197

wave generated in the interior region propagates with the light velocity, finer
time meshes for dealing with the finite-difference scheme are needed than in the
general relativistic treatment, and therefore much computer time is needed. b)
Unless we employ -a suitable difference scheme, which will be described in the
Appendix, the wave equation for ¢ leads to the occurrence of oscillations of two
zones in wave-length and large amplitudes. c) As the Birkoff theorem does not
hold in the scalar-tensor theory of gravity, we cannot settle the time-dependent
vacuum metric surrounding the collapsing object even in the case of spherical
symmetry. Therefore we cannot settle the boundary conditions of the metric and
scalar fields at the surfac~ of the object. [In Ref. 14), the Brans type-I metric
was prop'osed to be adopted .as the vacuum metric, but this was shown to be
insufficient in the course of our numerical analysis.] To avoid this difficulty, we
must introduce a tenuous envelope surrounding the main bulk of the object such
that any signal caused at the surface of the envelope cannot reach the bulk within
the time considered. Then the evolution of bulk is not affected by the manner
specifying the boundary conditions.
In § 2 the hydrodynamic equations in the scalar-tensor theory are summarized.
Our main concern is to clarify how the characters of gravitational collapse in
that theory deviate from those in general relativity. Accordingly we assume a
simple equation of state for gas and a simple configuration for gaseous sphere,
so that the model considered does not necessarily correspond to a real astronomi-
-cal object.
In § 3 the numerical results are presented, which shows that the deviation
is small if the relativi~tic effect itself is small and the bounce phenomenon occurs
in the central region. In the case of a continued collapse, oscillations of the
scalar field develop in the central region, but they cannot halt the collapsing.
Therefore, characteristic nature of the collapse resembles qualitatively that in
general relativity. The present method is, for numerical difficulties, incompetent
to see the asymptotic (t~oo) state of the metric and scalar fields, so that we
cannot say anything about the final state of the collapsing object, e.g., the validity
of Thorne and 'Dykla's conjecture.

§ 2. Hydrodynamic equations
In this section, we briefly summarize the hydrodynamic equations in terms
of the scalar-tensor theory of gravity. We consider an ideal fluid and neglect
heat transfer, pair creations and magnetic fields. Assuming a spherical symmetry
leads to the metric
(1)
where v and A are functions of t and r; r is a radial coordinate defined by the
rest mass of matter included in a sphere with radius r, i.e.,
(2)

Downloaded from https://academic.oup.com/ptp/article-abstract/49/4/1195/1932304

by guest
on 28 July 2018
 1198 T. Matsuda and H. Nariai

where p is the rest-mass density. The coordinate system which moves with the
fluid leads to the energy-mo mentum tensor:

(3)
where P is the pressure, s is the total density defined by s = p (1 + ejc 2) and e is
the internal energy per gram. We introduce comoving derivative s

(4)
and define the following quantities similar to those defined originally by Misner
and Sharp :7 >

U=DeR , u=tDe¢ , (5)

T=DrR , r=tDr¢ (6)
and

U=U+u R, T=T+rR , (7)

Gm:=tc 2 Re~"{1 + (Ujc)2-T2}; .(8)
where ¢=In (G¢) (¢ is the inertial scalar field).
On inserting above relations into the Brans-Dic ke field equations, we obtain
the following s.et of equations :14>

Deu= _ _l_DrP+ 4nGe-~'> (s- 3P) +c2 (Dr+ 2T)t

pw 3+2w c2 R
-u (D-:__fl + 2[}) +u 2 { 1- (3+2w)
r R L~}, (10)
, r.
D 1 ln(pR 2)=-Djp +u {1-(3+2 w) r;}. (11)

D 1e= -PDe(l_) , (12)

. p.

Dr (V /2) = - - 1-DrP, (13)

c2 pw

D 1m = - 4nP ( ~) 2 e-N2U , (14)

DerJ = - { (u 2 + c2r 2) U- 2c2urT} R 2e~"12 , (15)

where m and rJ are defined by

Gm =e~"1'{Gm +H3 + 2w) rJ}, (16)

Downloaded from https://academic.oup.com/ptp/article-abstract/49/4/1195/1932304

by guest
on 28 July 2018
 Hydrodynamic Calculations of Spherical Gravitational Collapse 1199

and w is the relativistic enthalpy defined by

w= e+Pc-2 1+!:... +~. (17)
p c2 pc2
Note that the mass function m is conserved at the surface of a star, but not m.
As we are concerned only with the dynamical behaviour of a collapsing object
in the scalar-tensor theory of gravity, in comparison with that in general relativity,
we assume the following simple form as the equation of state:
P=frpe. (18)
These equations can be solved numerically, using a finite-difference method
similar to the one used in the case of a general relativistic collapse. 18l Shocks
are treated by means of an artificial viscosity similar to that of Richtmyer and
von Neumann.
The boundary conditions used are
R = U = (] = m = 11 = r = 0 at r= 0, (19)
P=v=r=O, ¢=ln[(4+2w)l(3+2w)] at r=rb, (20)
where rb ( = const) is the radial coordinate specifying the surface of the collaps-
ing object. It should be noted that true boundary conditions for rand¢ at r=rb
are not known to us, unless we solve the time-dependent vacuum metric surround-
ing the object. As was discussed in § 1, we avoid this difficulty by assuming a
tenuous extended envelope surrounding the main bulk of the object, which we
call a core, and removing the surface rb so far that any signal caused at rb can-
not reach the core surface re within the time considered. Therefore any choices
of boundary conditions at rb do not affect the evolution of the core. We assume
Rb=Re+l.2c·t1 , where t 1. is a free-fall time of the core; t 1 = (Gp 0) - 112 , where Po
1s the central ,density and rb is determined by Rb.
Initial conditions are assumed to be
U=(J =u=r=IJ=O, ¢=In[ (4+2w)l(3+2w)]. (21)
The internal energy e is assumed to be constant, e0 , in the core and decreases
as e = e0 (Rei R)2 in the envelope, where Re is the radius of the core. As the
density distribution, we adopt the Emden function of index 0.1 in the core and
p = Pe (Rei RY in the envelope, where Pe is the density of the outermost shell of
the core. The choice of that index is made, in part, because for a core with
higher index there appears a central region of extreme curvature necessitating a
finer zoning; and, in part, because the assumption of uniform density (index 0)
leads to an unfavourable gap of physical quantities at the core surface. The
mass function m is determined by Eqs. (6), (7), (8), (19) and

Dr(Gm) =4nGp(1+elc2 )R 2T+rGm+ 3 + 2 w {(u 2 +c 2 r 2 )T-2ur(J}R 2e~,

2
(22)

Downloaded from https://academic.oup.com/ptp/article-abstract/49/4/1195/1932304

by guest
on 28 July 2018
 1200 T. Matsuda and H. Nariai

and m is determined by Eqs. (16) and (21). If we put cf;=u=r=O identically,

the above equations are reduced to the hydrodynam ic equations in general rela-
tivity.9l

§ 3. · Numerical results
Numerical calculations have been performed for various initial configuratio ns.
When the mass of a gaseous sphere is so small that its contracting motion is
halted, it has been found that there is no noticeable difference between the nu"
merical results obtained by using the Brans-Dicke theory and the general relativity.
Therefore, in this section, we present an example of continued collapse, in which
we may expect noticeable differences.
Initial model of the gaseovs sphere is assumed as R. = 3 ·109 em, p0 = 10 6 g/ cm 8,
e0 = 1015 erg/g. .As was noted before, this model does not correspond to a real
astronomica l object. The Dicke. coupling constant ()) is assumed to be 10, because
a larger ()) does not make noticeable difference and a smallar ()) has been rejected
by recent observations .ul
In Figs. 1 through 7, spatial distribution s of p, R, m, U, rR, ¢ and e"l2 r,
are represented in the case of the Brans-Dicke hydrodynam ics from the initial
rest state to the final state where the calculation is stopped by numerical difficulty.
The abscissa is the zone number of a space mesh. In these calculations , rather
coarse zoning such as 36 zones for the core and 10 zones for the atmosphere is
employed, partly because the hydrodynam ic calculations in the Brans-Dicke theory
consume computer time much longer than that in g~neral relativity and partly
because we are concerned only with the qualitative nature of gravitationa l collapse
but not precise numerical results.
Fig. 1. log p versus zone number J at various
17 time during the collapse from rest of a spheri-
cal gaseous object. The center of the object
is J =1 which is not expressed in the figure,
0 0 sec the surface of the core J=36, and the sur-
300 0.1895 sec
face of the atmosphere J=46. Initially, den-
600 0.21 21 sec
sity distribution is expressed by the Emden
900 0.2134 sec
function of index 0.1 in the core and by P"'
1200 0.2137 sec
R-5 in the atmosphere, the central density
is p0 =108 g/cms, and internal energy is as-
sumed to be homogeneous in the core, eo=
1Ql5 erg/g, and e-..-R-2 in the envelope. Initial
core radius is R 0 ( =3 ·109 em) and atmospheric
II
radius Rb is determined Rb=R 0 +1.2c(GPo)-112 •
The core baryon mass is r 0 ( = 6.97 ·1036 g)
10
and the atmospheric baryon mass is rb ( =
9.06 ·1036 g). The numbers denoted in the
figure are numbers N of time mesh; N=300
corresponds to t=0.1895 sec, N=600 to t=
0.2121 sec, N=900 to t=0.2134 sec, N=1000
7 2L-..J--'---'--1-'-o-'--'--'--'--'---'--'--'--'--'--'-..l..-lW to t=0.2135 sec, N=llOO to t=0.2136 sec,
20 30 36
Zone N=1200 to t=0.2137sec.

Downloaded from https://academic.oup.com/ptp/article-abstract/49/4/1195/1932304

by guest
on 28 July 2018
 Hydrodynamic Calculations of Spherical Gravitational Collapse 1201

E
u

0.

-"'
0

"'0

Zone

Fig. 2. log R versus zone number at various

Zone
time for the same model as in Fig. ~-
-0.01 r-'r-lr-1--r--r--r--r-,-,--r-,--,-,...,...,--r-,
Fig. 3. m versus zone number for the same
model as in Fig. 1.

0.2 I 35 sec
0.21 36 sec
0.21 37 sec

36
Zone
...
0::
No
Fig. 4.
Fig. 4. U versus zone number for the same 0
model as in Fig. 1.
Fig. 5. r versus zone number (upper figure) and
102r R versus zone number (lower figure). Note 900
that the departure of the Brans-Dicke hydrody-
namics from the general relativistic one can be
-1.02
estimated by the ratio lrR/rl; the larger lrR/ 4 6 8 10 12 14 16 18
rj, the larger departure. Therefore, in there- Zone
gion of r~o. the largest departure exists. Fig. 5.

Downloaded from https://academic.oup.com/ptp/article-abstract/49/4/1195/1932304

by guest
on 28 July 2018
 1202 T. Matsuda and H. Nariai

2 4 6 8 10 12 14 16 18
Zone

Zane
Fig. 6. ¢(=:ln(G¢)) versus zone number. Note
that scalar waves are generated in the inner re- Fig. 7. e•l 2 versus 'zone number.
gion of the object at the later stage of collapse.

These figures show the character-istic aspects of relativistic collapse with

pressure, such as the proceeding of a central condensation, Fig. 1, the increase
of mass in the central region, Fig. 3, and the occurrence of T ~o (or self-closure)
at some intermediate shell, Figs. 2 and 5; compare these figures with those .of
general relativistic collapse. 10a),lOb),lS) Our calculation was stopped by a difficulty
of increasing numerical errors. around inner zones.
The' growing of scalar waves in the central region is shown in the lower
figure of Fig. 5 and in Fig. 6. The departure of the Brans-Dicke hydrodynamics
from the Einstein's can be estimated by the values of iuR/UI and irR/TI. If
these values are small, the departure can be neglected, except for the fact that
the gravitational coupling constant ¢- 1 become smaller than G for positive (J). In
our case, the value of iuRfUI is found to be smaller than 0.1 throughout the
interior region, although it is not shown in these figures. As can be seen in
Fig. 5, IrRjTl is small except in the region of r "-'0. This means that the effect
of the scalar field on the dynamics can be neglected in the gaseous sphere except
for the region of "-'0. r
In the region of T ,..__,o, however, the departure of the Brans-Dicke hydrody-
namics from the Einstein's is so large that the approximation6 l to the Brans-Dicke
theory, constructed by the method of a power series expansion with respect to
1/w, breaks down. In spite of this, the present method is, for numerical difficulty,
incompetent to see the asymptotic (t~oo) state of the metric and scalar fields,
so that we cannot say any conclusive statement about the final state of the col-
lapsing object, e.g., the validity of Thorne and Dyklii's conjecture or not. Only

Downloaded from https://academic.oup.com/ptp/article-abstract/49/4/1195/1932304

by guest
on 28 July 2018
 Hydrodynamic Calculations of Spherical Gravitational Collapse 1203

some conjectures can be given. The asymptotic metric may have a naked singu-
larity: (a) The Brans type-I solution for a static spherically symmetric vacuum
region approaches the truncated Schwarzshild solution having a naked singularity
when ¢ approaches a constant value ;7l (b) True singularity will appear first in the
neighborhood of the region of T=O, which will become an event horizmi, and
there is no other event horizon. (We conjecture that even in general relativity
metrics with a singularity covered by a noncsingular event horizon, such as
Schwarzshild metric and Kerr metric, are exceptional ones and general metric
may have a naked singularity. Though this idea is not currently accepted, the
discovery of the rotating W eyl metric by Tomimatsu and Sato/6l which has a
naked singularity outside an event horizon, has strengthen the idea.) If we
define a black h()le by a mass covered by non-singular event horizons and without
naked singularity, it is conjectured that there are scarcely black holes in the
universe.
The optical appearance of a collapsing star for a distant observer in the Brans-
Dicke theory is not so different from that in general relativity, because the metrics
outside the event horizon are not so different. The only remarkable difference
for a distant observer is the emission of scalar waves, while gravitational waves
are not emitted from a spherical collapsing object in general relativity.

Acknowledgements
One of the authors (T.M.) wishes to thank Professor T. Sakurai for his
encouragements, and Mr. A. Tomimatsu for his valuable discussions. Numerical
calculations were performed by F ACOM 230-60 at the Data Processing Center
of Kyoto University.

Appendix
--Technique of solving a wave equation numerically--
The technique of solving hydrodynamic equations by a difference scheme IS
discussed by many authors (see, for example, Refs. lOb), 12) and 13)). Here
we discuss the method of solving Eq. (10), which governs the behavior of scalar
field rfo and is originated from
D¢= SrcT
(3 + 2w) c4
Although hydrodynamic equations are of the hyperbolic type, too, they are rather
stable when pressure is negligible. The collapsing problem is this case. However,
Eq. (10) is unstable and scalar waves with two zones in wavelength develop
unless some, suitable procedure to damp out waves with short wavelength is em-
ployed. We are not interested in scalar waves with short wavelength but those
with wavelength comparable with the radius of the collapsing object.

Downloaded from https://academic.oup.com/ptp/article-abstract/49/4/1195/1932304

by guest
on 28 July 2018
 1204 T. Matsuda and H. Nariai

Let us consider the following equation :

Bt = f(. Y; y I ' y " ). '

f}y

where prime denotes spatial derivative. A natural difference scheme is

y"+lf2=y"+if"·.dt '
y"+l = y" + ]"+1/2 C51-"+lta ... ) Jt .

This scheme is known to be unstable. To get stable solutions, we use the fol-
lowing scheme
y"+l=y"+ f"· .dt'
y"+ 1= y" + [ (1- a)f'" + a]"+ 1(y"+ 1· · ·)] .dt, a>t.
In the present calculations, we adopt a= 1. This difference scheme is a kind of
low path filter, which damps out oscillations of high frequencies. This scheme
is used recently in the numerical weather prediction.

References

1) C. Brans and R. H. Dicke, Phys. Rev. 124 (1961), 925.

2) A. Salmona, Phys. Rev. 154 (1967), 1218.
3) T. Matsuda, Prog. Theor. Phys. 48 (1972), 341.
4) H. Nariai and M. Fujimoto, Prog. Theor. Phys. 47 (1972), 105.
5) H. Nariai, Prog. Theor. Phys. 47 (1972), 118.
6) K. S. Thorne and J. ]. Dykla, "Black Holes in the Dicke-Brans-Jordan Theory of Gravity",
Preprint OAP-237 (1971).
7) T. Matsuda, Prog. Theor. Phys. 47 (1972), 738.
8) H. Nariai, Prog. Theor. Phys. 42 (1969), 742; 43 (1970), 334.
9) C. W. Misner and D. H. Sharp, Phys. Rev. 136 (1964), B571; Phys. Letters 15 (1965), 279.
lOa) M. M. May and R.N. White, Phys. Rev. 141 (1966), Bl232.
lOb) M. M. May and R. N. White, Methods in Computational Physics Vol. 7 (Academic Press,
New York and London, 1967), p. 219.
11) A. I. Voropinov, V. L. Zaguskin and M. A. Podurets, Soviet Astron.-AJ 11 (1967), 383.
12) R. A. Schwarz, Ann. of Phys. 43 (1967), 42~
13) T. Matsuda and H. Sato, Prog. Theor. Phys. 41 (1969), 1021.
14) H. Nariai, Prog. Theor. Phys. 47 (1972), 832,
15) K. S. Thorne, C. M. Will and W. T. Ni, Proceedings of the Conference on Experimental
Tests of Gravitational Theories (NASA JPL technical memorandum, 1971).
16) A. Tomimatsu and H. Sato, Phys. Rev. Letters 29 (1972), 1344.

Downloaded from https://academic.oup.com/ptp/article-abstract/49/4/1195/1932304

by guest
on 28 July 2018