View metadata, citation and similar papers at core.ac.uk brought to you by CORE
provided by CERN Document Server
Cosmological PPN Formalism and
Non-Machian Gravitational Theories
1
G. Dautcourt
received
1. INTRODUCTION
There app ears to b e little agreement what Mach's principle actually is - this was my
chief impression from a recent conference on Mach's principle at Tubingen (Germany).
Should we relate it to the requirement that the inertial mass of a b o dy is determined by
the remaining b o dies in the Universe? Or should we follow one of Barb ours suggestions
(Barb our 1995), p erhaps that theories must b e formulated in terms of relative quantities
to b ecome Machian, since we observe relative quantities (p ositions, velo cities) only? Mach
himself seems to b e of little help in this resp ect since for almost every formulation of a
Machian principle one may nd an appropriate citation in his publications.
Another item from Mach's General Store (Brill 1995) is the rather vague statement
that distant parts of the Universe should have some in uence on the form of lo cal laws
of physics. In a non-Machian gravitational theory a lo cal exp eriment should therefore b e
independent of the cosmological environment. In particular, no elds of cosmological origin
should have an in uence on the lo cal motion of matter, "lo cal" b eing taken to denote a
suciently small space region. The purp ose of this note is to show that - within a p ost-
Newtonian approximation - metric theories of gravitymay b e divided according to the
requirementof time-dependent potentials for a consistent lo cal description of cosmological
mo dels. This kind of dep endence on cosmological b oundary conditions is used to de ne
Machian and non-Machian theories.
To deal with a fairly general class of metric gravitational theories, the Parametrized
Post-Newtonian (PPN-) formalism by Kenneth Nordtvedt and Cli ord M. Will will b e
used. The notation is taken from Misner et al. (1973), otherwise I follow Will's standard
b o ok (Will 1993).
The PPN formalism was clearly not designed to deal with cosmology. On the other
hand, one has to face the fact that a Newtonian cosmology was develop ed decades ago
(see, e.g., Heckmann and Schuecking 1955, 1956, 1959, Trautman 1966), with results very
similar to those of General Relativity. The relation of this Newtonian cosmology to the
Friedman mo dels remained unclear, however. One exp ects that a Newtonian cosmology
should emerge as a rst approximation of the general-relativistic theory. Then one may ask
for a p ost-Newtonian cosmology as a second-order approximation. Notice the Newtonian
1
Humb oldt-Universitat zu Berlin, Institut fur Physik, D-10115 Berlin, Germany 1
cosmology as discussed here is based on a Lorentzian manifold, and hence quite di erent
from the Newton-Cartan formulation of a transition to Newton's gravity (for the latter
see, e.g., Trautman 1956, Ehlers 1981, Lottermoser 1988, Dautcourt 1990, and references
therein).
Somewhat surprisingly, the PPN formalism as it stands is already able to some extent
to deal with cosmological problems, as will b e shown subsequently. The keys are (i) to use
di erential relations to avoid the Minkowskian b oundary conditions at spatial in nityin
the usual integral formulation of the PPN formalism, and (ii) to take the ratio distance
(measured from the center of a selected spatial region) to Hubble distance c=H as the
0
expansion parameter. This leads to a simple lo cal description of cosmological mo dels,
which should b e valid for smal l values of the expansion parameter.
Section 2 summarizes basic asp ects of Newtonian cosmology. In Section 3 PPN rela-
tions as well as the divergence freedom of the matter tensor are used to construct lo cally
isotropic cosmological solutions b eyond the Newtonian approximation.
Restricting to homogeneous mo dels in Section 4, an inconsistency is found in generic
metric theories (with the exception of the non-Machian class) at the p ost-Newtonian level.
We suggest that the discrepancy might b e caused by the neglection of time-dep endent
comp onents in the Newtonian and p ost-Newtonian p otentials at the origin of the PPN
co ordinate system.
Section 5 illuminates the problem from a di erent p oint of view: Starting from a co-
moving cosmological co ordinate system, in which the cosmological uid with a divergence-
free matter tensor is homogeneous and expands isotropically, a transformation into the
standard PPN co ordinate system pro duces generically time-dep endent lo cal p otentials.
Again the p otentials are absent in a small subset of theories, formed by a non-Machian
class of gravitational theories.
2. NEWTONIAN COSMOLOGY
The PPN formalism assumes the existence of a co ordinate system where the metric tensor
can b e written as
2 6
g = 1+ 2U 2 U +4 A+O( ); (1)
00
7 1
5
g = V W + O ( ); (2)
0i 1 i 2 i
2 2
4
g = (1 + 2 U )+O( ): (3)
ik ik
The space-time functions U; ; A;V ;W are de ned in terms of the matter density,
i i
the pressure p and velo city comp onents v (x;t) of the uid:
i
Z
0 0 0
U (x;t)= dx(x;t)= j x x j; (4)
Z
0 0 0 0
dx(x;t)(x ;t)= j x x j; (5) (x;t)=
Z
0 0 0 0
V (x;t)= dx(x;t)v (x ;t)= j x x j; (6)
i i 2
Z
0 0 0 0 i 0i 0 3
W (x;t)= dx(x;t)[(x x )v (x ;t)](x x )= j x x j ; (7)
i
Z
0 0 0 0 2 0 3
A(x;t)= dx(x;t)[(x x )v (x ;t)] = j x x j ; (8)
where
3
2
(x;t)= v + U + p= : (9)
1 2 4 0
2
As it stands, the PPN formalism app ears not to b e applicable to cosmology: If a
homogeneous matter density (t)isintro duced, the Newtonian p otential U as calculated
from (4) diverges - this is the well-known Seeliger-Neumann paradoxon, seen to b e present
also in a higher-order correction to the Newtonian p otential. A simple waytoovercome this
dicultywas prop osed by Heckmann and Schucking (1959), see also Trautman (1966): In
agreement with the equivalence principle, the Newtonian p otential and its rst derivatives
are considered as quantities sub ject to changes induced by co ordinate transformations
beyond the Galileo group. The second derivatives only, which are determined by the
generating mass density, should havea physical meaning. Thus the function
2
2
U = G(t)r ; (10)
3
was taken as the p otential in Newtonian cosmology. It is evidently a solution of the Poisson
equation for U ,
U = 4; (11)
generated by homogeneous matter density. The divergence of U for large r can b e consid-
ered as co ordinate singularity,now caused by the use of a lo cal (non-cosmological) co or-
dinate system. The singularityvanishes, if a transformation to cosmological (comoving)
co ordinates is p erformed, see b elow Eq.(19).
The p ost-Newtonian approximation as applied to ob jects in the Solar system uses
p
two expansion parameters = j U j and v=c, where U is the maximal value of the
m m
Newtonian p otential and v is the velo city of b o dies moving under the in uence of gravity.
For the b o okkeeping of the di erent approximations b oth parameters are usually considered
as of equal order. This is justi ed also in the case considered here: The cosmological
p
p otential (10) gives r 2 G=3 r=L, where L is a typical Hubble distance. The
0
radial velo city of galaxies is of equal order as is evident from Hubbles redshift-distance
relation. As dimensionless expansion parameter one may therefore take the ratio distance
to Hubble distance.
To Newtonian approximation the metric tensor is given by
4
g = 1+ 2U +o( ); (12)
00
3
g = o( ); (13)
0i
2
g = + o( ); (14)
ik ik
and the dynamical equations reduce in the case of dust matter, which will b e assumed for
simplicity throughout this note, to
i i
@ =@ t + @ (v )=@ x =0; (15)
(_v + v v )=U ; (16)
i i;k k ;i 3
where the p otential U is given by the solution (10) of the Poisson equation (11). For an
isotropic expansion with a scale factor R(t)
i i i
_
v = x f (t)=x R=R; (17)
0
(15) and (16) together with the p otential (10) lead to matter conservation and to the
Friedman equation:
3 2
_
R = const; R =2GM=R + const: (18)
This similaritybetween Newtonian and general-relativistic cosmology has sometimes b een
taken as evidence that Newtonian gravity mighthave a wide range of applicability in cos-
mology - in the hop e to discard the mathematical apparatus of Einstein's theory in favour
of the much simpler apparatus of Newton's theory. This is to some degree misleading,
and the validity of the Friedman equation within Newtonian cosmology is easily under-
k
stood: One can intro duce comoving cosmological co ordinates ; instead of the lo cal
k
PPN-co ordinates x ;t used so far by requiring
(i) that the space comp onents of the matter 4-velo cityvanish and that
(ii)g ^ = 1; g^ = 0 in cosmological co ordinates.
00 0i
The transformation
2 2 k k
_
= t r R=(2R)+o(t ); =x=R + o(r) (19)
leads to
00 4
g^ = 1+o( ); (20)
0k 3
g^ = o( ); (21)
ik 2 2
g^ = =R( ) + o( ); (22)
ik
and satis es (i) and (ii) to Newtonian order. It is now easy to see why the Newtonian scale
factor R(t)must b e a solution of the Friedman equation: Consider the Friedman equation
for the correct general-relativistic scale factor S ( ). Writing
2 2
_ _ _
S ( )= S(t r R=(2R)) S (t) Sr R=(2R) (23)
2
in this equation and expanding, terms (r=L) cancel, and one is left with the Friedman
equation for the Newtonian scale factor R(t) S (t).
Thus it app ears that the applicability of the Newtonian approximations (12) - (18)
with the p otential (10) is con ned to a neighbourhood of the observer, corresponding to
distances smal l compared to the Hubble distance. As a consequence, the matter density
i
in lo cal co ordinates is a pure time function only on the observer world line x = 0 and
changes with r further out just to ensure homogeneity or a Cop ernican principle. Indeed,
any scalar dep ending only on cosmic time in cosmological co ordinates, dep ends in a lo cal
PPN co ordinate system also on spatial co ordinates (see Eq. (19)).
Manyyears ago, Callan, Dicke and Peebles (1962) have presented similar ideas con-
cerning the relation b etween Newtonian mechanics and cosmological expansion. 4
3. POST-NEWTONIAN EQUATIONS
Beyond the Newtonian approximation we use instead of (5),(6) the corresp onding di er-
ential formulations (G = 1 subsequently,we also write = + , assuming dust):
1 1 2 2
= 4 ; (24)
V = 4v : (25)
k k
Di erential relations also hold for the remaining functions. W may b e written
i
Z
1 @
0 0 0 0
W = dx (x ;t)(x x )v j
i
0
i 0
@x j x x
Z
@
k k 0 0 0 0 0
= x V + dx (x ;t)x v = j x x j;
;i
i
@x
or after applying the Laplacian
W = 4v 2V : (26)
i i k;ki
Finally we list some further relationships (see Will 1993, chapter 4.1) already in di erential
form:
V = U ; (27)
k;k ;0
= 2U; (28)
= V W ; (29)
;0k k k
= A : (30)
;00 1
The di erential formulation of the PPN formalism used in this article is based on the
(not all indep endent) relations (24) - (30) and (11). We write down expansions in terms
of p owers of r similar to (10) for the p ost-Newtonian p otentials V ;W ; ;A as well as for
i i
the uid density and velo city, and lo ok for homogeneous and isotropic purely expanding
solutions. The matter density and velo city for isotropically moving dust can b e expanded
as
2
= + r ; (31)
0 2
k k 2
v = x (f + f r ); (32)
0 2
where ; ;f ;f are time functions. Using this p ower law expansion, a straightforward
0 2 0 2
calculation yields
1 2
2 4
r r ; (33) U =
0 2
3 5
1
4 2 2 2
)r ; (34) +2 ( 3 f =
2 1 0
0 0
15
2
2 k
V = f r x ; (35)
k 0 0
5
W = V ; (36)
k k
4 1
2 4
(3f + )r : (37) A =
0 0
0
5 3 5
Up to p ost-Newtonian order, the metric tensor can now b e written:
2 4
g = 1+br + cr ; (38)
00
2 i
g = mr x ; (39)
0i
2
g = (1 + gr ): (40)
ik ik
The time functions b; c; m; g are determined by the distribution and motion of matter:
4
; (41) b =
0
3
2 4 1
2 2 2
c = + (6 3 10 ) f (4 +3); (42)
2 2 0 1
0 0
5 45 5
1
m = f (7 ); (43)
0 0 1 2
5
4
g = : (44)
0
3
From the conservation law of the matter tensor, T = 0, one obtains in Newtonian
;
approximation, in agreement with (18),
4
2
_
f = f ; (45)
0 0
0
3
_ = 3f : (46)
0 0 0
The p ost-Newtonian terms lead to equations for the next co ecients f and :
2 2
2 4 4
2 2 2
_
f = 4f f + k f + k ; (47)
2 0 2 1 0 2 2
0 0
15 45 5
2
2
f (2 3 ); (48) _ = 5f 5f +
0 2 0 2 2 0
0
3
with
k = 12 9 + 42 6 20 +5;
1 1 1 2
k = 20 +12 6 + 21 3 20 :
2 2 1 2
The system (45) - (48) can b e integrated fairly easily: From (45), (46) one obtains a
second-order linear di erential equation for f alone, is determined algebraically up on
0 0
solution of the f equation. Similarly, if a solution of (45),(46) is known, (47),(48) can
0
b e transformed into a linear second-order di erential equation for f , leaving an algebraic
2
equation for .
2
4. HOMOGENEITY REQUIREMENTS
The mo dels obtained in this way represent - in a p ost-Newtonian approximation -
a large class of lo cally isotropic, but in general inhomogeneous expanding matter dis-
tributions in the neighb ourho o d of the observer. Restrictions for homogeneitymaybe 6
intro duced by several metho ds. Usually one relies on groups of motions. A more intu-
itiveway is to imp ose homogeneity with a Cop ernican principle. The origin of the PPN
co ordinate system may b e asso ciated with a freely falling observer P . The worldline of a
second observer P at rest with regard to the cosmological substratum in his neighb ourho o d
i i
is given by x (t), with the three functions x (t) as solutions of the di erential equations
0 0
i i 2
x_ = x (f + f r ): The co ordinate transformation leading from the P -system to the P -
0 2
0 0 0
system is a generalization of the Chandrasekhar-Contop oulos (1967) transformation, since
the second observer moves with acceleration. A Cop ernican principle should state the
equivalence of all these systems. More sp eci c, in the lo cal PPN system of any comoving
observer the expansion co ecients of matter density and cosmic velo city (Eqns. (31), (32))
should have the same dep endence on the corresp onding co ordinate time. Wemay set up
a Cop ernican principle for the matter density and velo cityaswell as for the metric or
derived quantities such as the expansion rate of the matter congruence. It is not obvious
whether all these principles lead to the same conclusions.
For purp ose of this note we can avoid the construction of a generalized Chandrasekhar-
Contop oulos transformation by refering to Eq.(19): Homogeneity for scalars can b e
2
de ned by a dep endence on the cosmological time only, hence ( )=(t r f =2)
0
2
_
(t) r f (t)=2, and the lo cal expansion co ecients and are related by =
0 0 2 2
_