Sie sind auf Seite 1von 22

Proper Formulation of Birkhoffs Theorem

R. E. Salvino
9 Thomson Lane, 15-06 Sky@Eleven
Singapore 927726
7 December 2014

Abstract
The conventional formulation of Birkhoffs theorem assumes that
the g22 metric function is both independent of the time-like coordinate and has a very specific radial dependence, g22 = r2 . To remove
these assumptions from the theorem without confusion, it is necessary to distinguish between the conventional formulation and a proper
formulation of the theorem which explicitly incorporates the assumptions. Proof of the proper formulation demonstrates that the time
dependent solution of the spherically symmetric vacuum field equations is the time independent solution if the metric is diagonal and
if and only if the g22 metric function is independent of the time-like
coordinate. Furthermore, the proper formulation of the theorem establishes the unique dependence of the g00 and the g11 metric functions
on the g22 metric function, but it does not establish the dependence of
the g22 metric function on the radial corrdinate r. Since g22|t = 0 is
a necessary and sufficient condition for the proper formulation of the
theorem, other time-dependent, spherically symmetric solutions of the
time dependent vacuum field equations exist as long as the g22 metric
function is time-dependent. In particular, the time dependent vacuum
field equations support gravitational wave solutions.
Keywords: Birkhoffs theorem, Combridge-Janne solution, time dependent gravitational field equations, gravitational waves

Introduction

Birkhoffs theorem [1,2] is a statement that the time-independent, spherically symmetric solution of the vacuum general relativistic field equations
is also the one and only solution of the time-dependent, spherically symmetric vacuum field equations. As an immediate consequence of the theorem,
it is clear that there is no need to investigate the time-dependent vacuum
equations at all since no other solutions exist. This means, in particular,
that it is not possible for a time-dependent spherical distribution of mass
to emit gravitational waves into the region exterior to the distribution since
the solution in that region is necessarily time-independent.
The conventional form of the theorem may be formally stated in the
following way:
Birkhoff s Theorem: Conventional Formulation. The solution of the time-dependent spherically symmetric vacuum field
equations is identical to the solution of the time-independent
spherically symmetric vacuum field equations. Furthermore, the
time-independent spherically symmetric solution is the textbook
or Hilbert version of the Schwarzschild solution.
Proofs of this conventional formulation of Birkhoffs theorem are predicated
on two unstated and unacknowledged assumptions about the g22 metric
function:
(A1) the g22 metric function is assumed to be independent of the time-like
coordinate, and
(A2) the g22 metric function is assumed to have a specified dependence on
the radial coordinate 1 , g22 = r2 .
From a logical point of view, these statements should be consequences of the
theorem, not assumptions upon which the theorem is based. This, however,
is not the case. In addition, the condition that the metric is diagonal, that
the g0k metric functions are zero for k = 1, 2, 3, is always assumed to be
satisfied. Since this diagonal condition is actually a necessary condition for
the theorem, it will be listed as an additional assumption:
1

A generalized Birkhoff theorem [3] addresses, to some extent, this second assumption
concerning the specified dependence on the radial coordinate. It does not, however, address
the assumption of time-independence.

(A3) the metric must be diagonalizable, that is, g0k = 0 for k = 1, 2, 3


All three statements, A1, A2, and A3, are essential to prove the conventional
form of the theorem; a proper formulation of the theorem requires these
statements to be incorporated into the theorem.
It should be clear that a diagonal metric is a necessary condition for
the theorem: if the metric can not be diagonalized, then it can not be
put into the form of the time-independent solution. It should also be clear
that a time-independent g22 function is also a necessary condition since the
time independent solution cannot have a time dependent g22 function. It
is shown below in Section 6 that a time-independent g22 metric function is
also a sufficient condition for the time-dependent solution of the vacuum field
equations to be identical to the time-independent solution once a diagonal
metric has been established. Both conditions, the diagonal property of the
metric and the time independence of the g22 function, are crucial for the
theorem to hold. In addition, the proper form of the theorem establishes
the unique dependence of the g00 and g11 metric functions on the time
independent g22 metric function but it makes no statement regarding the
functional dependence of the time independent g22 function on the radial
coordinate r.
The proper formulation of the theorem demonstrates that the conventional formulation of the theorem is false. That is, there are solutions to the
time dependent vacuum field equations that are not identical to the time independent solution. These solutions necessarily require a time dependent g22
metric function. An appeal to the source-free limit of the Robertson-Walker
(RW) metric [4, 5] or the McVittie metric [6] would provide a sufficient indication that the conventional formulation of Birkhoffs theorem is false.
Since the assumption of time independence for the g22 metric function was
never acknowledged, there was no perceived need to make a connection to
the source-free limit of the RW metric, the McVittie metric, or any other
metric.
To establish the proper formulation of Birkhoffs theorem, the time dependent, spherically symmetric vacuum field equations are developed ab initio. In Section 2, the form of the line element is established, including the
diagonality of the metric tensor. That line element is then used in Section 3
to provide the basic ingredients that are used for deriving the geodesic equations and the non-zero Christoffel symbols of the second kind. These results
are then used in Section 4 to derive the explicit forms of the time-dependent
Ricci tensor components and the time-dependent, spherically symmetric vacuum field equations. In Section 5, some basic symmetry properties of the
3

time-dependent Ricci tensor and field equations are examined. Section 6


provides the proof that the time-independence of the g22 metric function is
a sufficient condition for the theorem; Section 7 provides the proof that the
time-independence of the g22 metric function is also a necessary conditon for
the theorem. Some remarks on the solution presented in Section 7 are given
in Section 8, explicit gravitational wave solutions using linearized equations
are demonstrated in Section 9, and conclusions are presented in Section 10.

Time-dependent Metric Functions and g01 = g10 6= 0

The most general line element for a spherically symmetric system that
is neither static nor stationary has the form 2
2

ds2 = e c2 dt + 2 c T dt dr e dr2 R2 d2

(2.1)

d2 = d2 + sin2 d2

(2.2)

where the , T , , and R are functions of r and t only and d is the


differential solid angle. This line element is not static (g01 = g10 6= 0), is
not stationary (, T , , and R are not independent of the time coordinate
t), but is rotationally invariant ( and both leave the metric
unchanged). The distinguishing feature of equation (2.1) is contained in the
function R(r, t) which is treated as a function of r and t and not, as in the
standard approach, as identical to the independent radial coordinate r.
The metric may be diagonalized by a coordinate transformtaion in the
following way. First, the metric is rewritten as

ds2 = e c2

T
dt + 2 e dr
c

!2 

T 2

e 2 e
dr2 R2 d2
c

(2.3)

Now a new time marker is introduced either by


2

The material in this section follows a line of development that is outlined in ref [2].

dt = dt +

T
e dr
c2

(2.4)

or by

e/2 dt = e/2 dt +

T /2
e
dr
c2

(2.5)

depending on which form is integrable. In either case, the new time marker
provides a diagonalized metric with new metric functions that utilize the
new time marker defined by (2.4) or (2.5),
ds2 = e c2 dt2 e dr2 R2 d2
e = e

T
e
c2

(2.6)
(2.7)

Now the time-dependent problem may be approached in a simplified manner,


with a diagonalized metric, by using the coordinate set (ct, r, , ). Making
the conversion back to the original coordinate set (ct, r, , ), if necessary at
all, may be a much more difficult task. In any case, the analysis can now
begin with the diagonalized metric (2.6), the coordinate basis (ct, r, , )
using the new time marker t, and the new set of metric functions , , and
R.
The introduction of an integration factor, Eq. (2.4) or Eq. (2.5), to
obtain Eq. (2.6) may not be able guarantee integrability in all cases since
the integration factor is also subject to the field equations and it is not at
all obvious that the integration factor can fulfill both roles simultaneously.
However, a non-diagonal metric immediately provides a counterexample to
the conventional form of Birkhoffs theorem which requires the metric tensor
to be diagonal. Thus, it is clear that a diagonal metric is a necessary condition for the theorem to hold and, consequently, the focus is placed solely
on metrics that have diagonal form.

Geodesic Equations and Christoffel Symbols

It was shown in Section 2 that the most general line element for timedependent and spherically symmetric systems may be put in the diagonalized
form
ds2 = e (dx0 )2 e dr2 R2 d2

(3.1)

d2 = d2 + sin2 d2

(3.2)

where the , , and R are functions of r and t. The line element contains
three unknown metric functions (, , and R), is rotationally invariant, and
is static but not stationary (, , and R are functions of r and t).
Now to set up the Euler-Lagrange equations from this metric, it is convenient to define the function
2 R2 sin2 ()
F = e (x 0 )2 e (r)
2 R2 ()
2

(3.3)

where x 0 = dx0 /ds, r = dr/ds, = d/ds, and = d/ds. The EulerLagrange equations


d F
F
=0
ds x
x

(3.4)

then provide the equations for the geodesics. Using the vertical slash | to
denote standard partial differentiation, f|x f /x , the equations are

x
0 +


|t 1 |t 0 0
|t 2
1

x x + |r r x 0 + e (r)

c
2 c
2
c
R|t 2
R
+ e R |t sin2 ()
+ e R
()
2 = 0 (3.5)
c
c

|t 0
1
1
r + e |r (x 0 )2 +
x r + |r (r)
2
2
c
2
2 e RR|r sin2 ()
e RR|r ()
2 = 0 (3.6)
+ 2

R|t 0
R|r
x + 2
r sin cos ()
2=0
Rc
R

(3.7)

R|t 0
R|r
x + 2
r + 2 cot = 0
Rc
R

(3.8)

+ 2

for = 0, 1, 2, and 3, respectively. Comparing these equations with the


geodesic equations written in the form


x
+

x x = 0

(3.9)

then provides the determination of the Christoffel symbols of the second


kind. These results are summarized in Table 1. It should be noted that
these results revert to the time-independent functions for R(r, t) = R(r) [7]
and to the textbook functions for R(r, t) = r [4] as they must.

The Ricci Tensor and Field Equations


The Ricci tensor may be written as


R = ln g ||


+








ln g |

(4.1)

For spherically symmetric and time-dependent systems, the only non-zero


derivatives are the derivatives with respect to x0 = ct (for index = 0),
x1 = r (for index = 1), and x2 = (for index = 2).
7

Table 1: The non-zero Christoffel symbols of second kind obtained from the
geodesic equations, eqs. (3.5), (3.6), (3.7), and (3.8). The determinant of

the metric g = ||g || and ln g are also included in the table.




1


1
0



1


3


1

= 12 e |r

g = e+ R4 sin2


3

R|t
Rc

3
2


=

3
3

3
3
3

R|r
R

R|t
Rc


=


3

= 12 |r

= e RR|r sin2

R|r
R

2
2

= sin cos
3


3

= 12 |r

= 1c e RR|t

1
2c |t


=

e RR|r

1
|t
2c e

1
2
c e RR|t sin

1
2c |t

0

2

= cot

ln g = 12 ( + ) + 2 ln R + 12 ln(sin2 )

Computing the Ricci tensor components is a straightforward but tedious


and time-consuming task. The results of the computation are

R00


2|t |t |t 2|t R|t 4R|t|t 
1
= 2 |t|t +

+
2c
2
2
R
R


2

|r
|r |r
2|r R|r
e
|r|r +
(4.2)

+
2
2
2
R
R01 = R10 =

R11

R22

R33

2R|t|r
|r R|t |t R|r

Rc
Rc
Rc

(4.3)



2
|r
|r |r
2|r R|r
4R|r|r
1
=
+

+
2 |r|r
2
2
R
R

2|t |t |t 2|t R|t 
e

|t|t +

+
(4.4)
2c2
2
2
R



RR|t |t RR|t |t
e
2
= 2 RR|t|t + R|t
+
c
2
2




RR|r |r
RR|r |r

1 (4.5)
+ e
RR|r|r + R|r +
2
2



RR|t |t RR|t |t
e
2
2
= 2 sin RR|t|t + R|t
+
c
2
2




RR|r |r
RR|r |r
2

2
+ sin e
RR|r|r + R|r +

1 (4.6)
2
2

with all other R identically zero. As in the time-independent case, R33


is proportional to R22 , R33 = sin2 R22 , and so it does not provide an additional vacuum field equation. It is easily verified that the Ricci tensor
components reduce to the time-independent spherically symmetric components for R(r, t) = R(r) [7]. And, of course, the equations reduce to the
standard time-independent textbook equations for R(r, t) = r, R|r = 1 and
R|r|r = 0 [4].
9

The vacuum equations for the time-dependent case, R = 0, may be


conveniently written as



2
|r
|r |r
2|r R|r
1
+

+
2 |r|r
2
2
R

2|t |t |t 2|t R|t 4R|t|t 
e
=
|t|t +
(4.7)

+
2c2
2
2
R
R
2R|t|r
|r R|t |t R|r

=0
Rc
Rc
Rc

(4.8)



2
|r
|r |r
2|r R|r
4R|r|r
1

+
+
2 |r|r
2
2
R
R

2|t |t |t 2|t R|t 
e
|t|t +

+
(4.9)
=
2c2
2
2
R

(e



|r + |r
RR|r )|r 1 + (e RR|r )
2



|t + |t
1

= 2 (e RR|t )|t + (e RR|t )


(4.10)
c
2

which are the 00, 01, 11, and 22 equations, respectively. Since the R33
equation duplicates the R22 equation, this is a system of four equations in
three unknowns. Presumably at least one of these equations will duplicate
another so that the system is not overdetermined.

Symmetry Properties of the Ricci Tensor

Before proceeding with the proof of the proper formulation of Birkhoffs


theorem, it is worth noting some symmetry properties of the Ricci tensor
components. Interchanging the coordinates r and ct and the functions
and and comparing the results with the original Ricci tensor components,
Eqs. (4.2) - (4.5), shows that
10

R00 = R11

for r ct

and

(5.1)

R01 = R01

for r ct

and

(5.2)

R22 + 1 = (R22 + 1)

R33 + sin2 = (R33 + sin2 )

for

r ct

and

for r ct and

(5.3)

(5.4)

The interchange of the R00 and R11 components of the Ricci tensor reflects
the fact that interchanging r and ct corresponds to an interchange of the
0 (ct) and 1 (r) indices. Since R is symmetric in its indices, interchanging the 0 and 1 indices leaves the R01 = R10 components unchanged. The
behavior of the R22 and R33 components is ultimately due to the change
of signature of the metric under the interchange of the 0 and 1 indices, a
swapping of the time-like and radial space-like coordinates. For the interchange of , the metric functions g00 = e and g11 = e become
g00 e = g11 and g11 = e = g00 which introduces the change in
the signature of the metric. In the R00 and R11 components, these factors
appear in product form so that the interchange does not introduce an overall
sign change; in the R22 and R33 components, the factors e and e appear
separately and so a sign change is introduced. This can also be seen in
the = 0 and = 1 geodesic equations and the corresponding Christoffel
symbols (0, 2, 2 1, 2, 2 provides an example).
However, the interchange ct r and g00 = e g11 = e provides
both the interchange of the functions and and the requisite sign change
to keep the signature of the metric unchanged. Under this interchange, it is
found that both R22 and R33 are invariant

R22 = R22

for r ct

and g00 g11

(5.5)

R33 = R33

for r ct

and g00 g11

(5.6)

Thus, under the composite interchange x0 x1 and g00 g11 , the R00
and R11 components interchange, R00 R11 , while the other Ricci tensor components are unchanged. Since the vacuum field equations require
11

R = 0, this means the field equations are invariant under this composite
interchange.

R|t = 0: The Sufficient Condition

It will now be shown that imposing the condition R|t = 0 immediately


leads to the time-independent solution. This is actually an important restricted case of the full time-dependent field equations since the textbook
choice R(r, t) = r falls into this category. The field equations for R|t = 0 are




2
|r
2|t |t |t 
|r |r
2|r R|r
1
e

+
=
|t|t +
2 |r|r
2
2
R
2c2
2
2

|t R|r
=0
Rc

(6.1)

(6.2)



2
|r
|r |r
2|r R|r
4R|r|r
1
|r|r +

+
2
2
2
R
R

2|t |t |t 
e
=
|t|t +

(6.3)
2c2
2
2




RR|r |r
RR|r |r

1 =0
e
RR|r|r + R|r +
2
2

(6.4)

First, it should be noted that the condition R|t = 0 immediately reduces


the R22 = 0 equation, Eq. (6.4), to the time-independent equation. Second,
the form of R00 = 0 and R11 = 0 equations, Eq. (6.1) and Eq. (6.3),
show that the spatial derivative parts, although not separately equal to zero
as in the time-independent case, are equal to each other. These two facts
are sufficient to show that the time-dependent solutions for and are, to
within an overall scaling factor, the time-independent solutions.
Proceeding with the direct solution, the second equation requires either
R|r = 0 or |t = 0. Subtstituting R|r = 0 into Eq. (6.4), the R22 equation,
shows that R|r = 0 produces a contradiction and so is not a permissable
12

condition. Consequently, |t = 0. Thus, as a result of R|t = 0, the only possible time-dependence in the metric is contained in the g00 function . The
last two field equations for this R|t = 0 case are the time-independent R22
and R33 equations and, since the R01 equation requires to be independent
of t, the R00 and R11 equations reduce to the time-independent equations
as well,


2
|r
|r |r
2|r R|r
e
|r|r +
=0

+
2
2
2
R

(6.5)



2
|r
|r |r
2|r R|r
4R|r|r
1
+
=0

+
2 |r|r
2
2
R
R

(6.6)

From the initial four time-dependent field equations, the single constraint
R|t = 0 produces the three time-independent field equations with the additional R01 = 0 condition which requires |t = 0. The three time-independent
field equations are reducible to two independent equations [79] which produce the time-independent solution for in terms of R. Since R is independent of t by supposition, |t = 0 is satisfied automatically.
Although the metric function is, in principle, a function of r and t,
its t-dependence is not a factor in the field equations and so will remain
undetermined. This t-dependence must take the form of an overall scaling
factor on the g00 function which translates to an additive function of t to
the time-independent solution 0 (r), (r, t) = 0 (r) + f (t), so that all spatial derivatives contain the function 0 (r) only. Consequently, the single
restriction that R is a function of r only leads to the time-independent solution to the field equations with an arbitrary function of t multiplying the
time-independent g00 ,


f (t)
1
e
R


1

2
e = R|r 1
R

e = e0 (r)+f (t) =

(6.7)

(6.8)

R(r) = undetermined

(6.9)

f (t) = undetermined

(6.10)

13

The function f (t) may be asbsorbed into a new time coordinate defined by
dt0 = ef (t) dt. Alternatively, the far-field behavior may be used to show that,
as r ,


f
e 1
e (t)
r

(6.11)

which indicates the possibility of a time-dependent classical gravitational


potential. It has been previously shown [9] that c2 (1 g00 )/(2g00 g11 ) is the
classical potential, so that

VG (r) =

c2 (ef (t) 1)
Gm

2R02
RR02

(6.12)

The second term is the static gravitational potential that tends to the Newtonian form for R r in the asymptotically flat spacetime and identifies
as twice the geometric mass of the source, = 2mG = 2Gm/c2 . The
first term is a time-dependent non-Newtonian potential which tends to a
spatially-independent function of time in the asymptotically flat spacetime,
where R0 1 as R r. If no such time-dependent non-Newtonian term is
to be expected, then the requirement that f (t) = 0 follows. This, of course,
is consistent with simply absorbing the exponential function into the time coordinate. Thus, it has been established that R|t = 0 is a sufficient condition
that the time-independent solution is also the time-dependent solution of the
field equations. Furthermore, it was explicitly shown that the relevant timeindependent solution is the solution found independently by Combridge [10]
and Janne [11], denoted by the name Combridge-Janne solution [9], and not
the textbook or Hilbert version of the Schwarzschild solution. Note that for
R(r) < 2mG , t becomes space-like and must be identified as the radial coordinate while r becomes the time-like coordinate. Consequently R(r) 2mG
must hold since the g22 metric function was required to be independent of
the time-like coordinate in the derivation by supposition.

R|t = 0: The Necessary Condition

It now remains to show that R|t = 0 is a necessary condition for the


theorem. For this part, it is only necessary to present a time-dependent
14

solution of the field equations that is distinct from the Combridge-Janne


solution. As noted in Section 1, an appeal could be made to the source-free
limit of the Robertson-Walker metric [4, 5] or the McVittie metric [6] as
such a time-dependent solution. However, for the purpose of establishing
R|t = 0 as a necessary condition for the theorem, it may be of interest to
examine the symmetry properties presented in Section 5 and recognize that
a solution exists for the spatially homogeneous case, R|r = 0, so that the
g22 function is a function of t only, R(r, t) = R(t). It is straightforward to
verify that this solution has the form
ds2 = e c2 dt2 e dr2 R2 d2

e =

R|t2


c2 Ac

1
R


Ac

e =
1 eC0
R
R = R(t) = undetermined

(7.1)

(7.2)
(7.3)
(7.4)

where A and C0 are constants. The undetermined nature of R(t) is the


time-dependent analogue of the undetermined nature of R(r) in the timeindependent Combridge-Janne solution and is due to an insufficient number
of independent field equations [711].
The solution given by Eq. (7.1) - Eq. (7.4) is clearly distinguishable
from the time-independent Combridge-Janne solution and thus establishes
that R|t = 0 is a necessary condition for the time-dependent solution to
be identical to the time-independent solution of the vacuum field equations.
This completes the proof that, given a diagonal metric, R|t = 0 is a necessary
and sufficient condition that the solution of the time-dependent vacuum field
equations is given by the solution of the time-independent field equations.

Remarks On The R|r = 0 Solution

The purpose of the spatially uniform R|r = 0 solution is to present


a time-dependent vacuum solution that can not be reduced to the timeindependent Combridge-Janne solution. However, since this solution has
15

properties which relate it to wormhole solutions [9] that are presented as


time-independent solutions, it is worth making a few observations and comments on the relations between a wormhole solution and the R|r = 0 solution.
Three distinct cases need to be considered. First, if R(t) < Ac for all
t, then the signature of the metric is constant showing that this solution
does not undergo a transition to a time-independent metric for any value
of t. The solution remains time-dependent and spatially homogeneous for
all times: R(t) continues to evolve in time but is bounded from above by
Ac. This is the time-dependent analogue of the strictly time-independent
Combridge-Janne solution for which R(r) is bounded from below by a value
greater than or equal to 2mG .
On the other hand, if R(t) > Ac for all t, the signature of the metric then
shows that t is space-like and r is time-like. This means that the solution
is actually time-independent. This can most easily be seen by noting that
r = ct assumes the role of the radial coordinate and ct = r assumes the role
of the time-like coordinate. Due to the symmetry of the field equations, this
solution is identical to the time-independent Combridge-Janne solution for
which R(r) > 2mG for all r.
Finally, for the third case, if there is a time tH for which R(tH ) = Ac,
then the coordinates t and r swap roles for t > tH very much as they do in
the textbook wormhole solution for r < 2mG . Then, for t > tH the radial
coordinate is r = ct and the time coordinate is ct = r so that the metric
takes the form

ds2 = e c2 dt e dr2 R2 d2


Ac C0

e = 1
e
R

(8.1)
(8.2)

e =

R|r
(1

(8.3)

Ac
R)

R = R(r) = undetermined

(8.4)

This is just the Combridge-Janne form of the wormhole solution outside


r = 2mG with Ac = 2mG and C0 = 0. Although this solution is timeindependent for times t > tH , it is clearly time-dependent for t < tH .
This solution provides a means for understanding the physical origin of
the wormhole solution: the time independent state of the solution evolves in
16

time out of the spatially homogeneous solution. The metric originates at t =


0, say, as spatially homogeneous as described by Eq. (7.1) with R(0) defining
the original radial extent of spacetime. If R(0) = 0, then the behavior of the
g00 function depends on the behavior of R|t while the g11 function diverges.
As R(t) increases with t, the radial extent of spacetime increases until the
time tH for which R(tH ) = Ac. At this time, the signature of the metric
abruptly changes indicating that the time-like and radial coordinates swap
roles, ct = r and r = ct. Then, for all times t > tH the metric assumes the
form of the time independent generalized wormhole solution.
This time evolution can not be traced in a backward sense by a freely
infalling observer since the infall terminates at the event horizon. It must be
remembered that the point r = 0 does not correspond to the spatial origin
r = 0, it corresponds to the time origin t = 0. But since the solution inside
the event horizon is spatially homogeneous, all spatial points are equivalent
within the event horizon. Consequently, reaching the event horizon is equivalent to reaching the spatial origin. But the time evolution of the metric
within the event horizon is not retraced in a time-reversed sense.

Linearized Equations and Gravitational Waves

The solution presented in Section 7 establishes that g22|t = 0 is a necessary condition for the proper formulation of Birkhoffs theorem. Another
solution that accomplishes the same task can be obtained from the linearized
field equations. In addition, this solution to the linearized equations also
demonstrates that gravitational wave solutions exist for the time dependent
vacuum field equations.
It is simplest to provide a perturbation around flat Lorentzian spacetime.
In other words, weak field conditions are assumed such that e 1 + ,
e 1 + , and R = r + R. Thus, to first order in , , and R, the
metric is given by
ds2 (1 + )c2 dt2 (1 + )dr2 (r2 + 2rR)d2

(9.1)

and the field equations (4.7) - (4.10), to first order in , , and R, are


2|r
4R|t|t
1
|r|r +
= 2 |t|t +
r
c
r
17

(9.2)

2R|t|r |t = 0
2|r
4R|r|r
1
+
= 2 |t|t
r
r
c


|r |r
1
+ 2R|r + rR|r|r + r
= 2 rR|t|t
2
c
|r|r

(9.3)
(9.4)
(9.5)

The field equations can be written in a more suggestive form by noting that
for an arbitrary function of the magnitude of the radial coordinate f (r), the
Laplacian contains only the r-derivatives

2 f = f|r|r +

2f|r
r

(9.6)

Consequently, the field equations may then be written as




4R|t|t
1
= 2 |t|t +
c
r

(9.7)

2R|t|r |t = 0

(9.8)

2(|r + |r ) 4R|r|r
1
+
= 2 |t|t
r
r
c


|r |r

1
2 R 2 R|t|t +

=0
c
2
r
2

(9.9)
(9.10)

Clearly, Eq. (9.10) has the form of an inhomogeneous wave equation for R
with the other metric functions appearing to act as source functions.
The solutions of Eq. (9.7) - Eq. (9.10) are given by
g0
2
+ R
r2 c2 |t|t
g0
= 2R|r
r
R = undetermined

|r =

18

(9.11)
(9.12)
(9.13)

where g0 is a constant. The undetermined nature of R(r, t) continues the


trend established by the time-independent Combridge-Janne solution and
the time-dependent solution presented in Section 7. By noting that Eq.
(9.11) may be formally written as

g0
+ F (r, t)
r

with F|r =

2
R
c2 |t|t

(9.14)

the constant g0 may be identified as g0 = 2Gm/c2 where m is the mass of the


source of spacetime curvature. It should also be noted that setting R = 0
provides the textbook solution for the time-independent perturbation of flat
spacetime. It is clear in the above solution that the contribution from R
provides the time-dependence for the more general solution, although that
time-dependence is not determined by the field equations unless an auxiliary
condition is imposed.
By choosing the auxiliary condtion to be

|r |r =

2
r

(9.15)

the R22 equation (9.10) produces the homogeneous wave equation for R.
Eq. (9.15), which may be called the gravitational wave condition or gauge, is
fully consistent with the other equations for and . Direct substitution
of the solution for into Eq. (9.15) produces

|r =

g0
+ 22 R
r2

(9.16)

Comparing this with the previously derived expression for |r , eq. (9.11),
produces the homogeneous wave equation for R, consistent with Eq. (9.10)
with the gravitational wave condition imposed,


1
2
R 2 R|t|t = 0
c

(9.17)

This establishes the spherical wave solution for R for the time dependent
spherically symmetric vacuum field equations as a linear combination of
19



Rk = A(k)j0 (kr) + B(k)n0 (kr) ei(k)t

(9.18)

(k) = kc

(9.19)

where j0 (x) and n0 (x) are the zero-order spherical Bessel functions. Since
both j0 (kr) and n0 (kr) vanish as r , the boundary condition in the
asypmptotic region is satisfied. The boundary condition near the source,
for example as r 0, and the initial conditions will provide the necessary
information to completely specify the solution.
Note that if the gravitational wave condition, Eq. (9.15), is not imposed and some other condition is imposed instead, then a different type
of solution will be obtained for R(r, t) that may have nothing to do with
gravitational waves. In other words, gravitational waves are possible solutions to the linearized equations but they do not follow from the equations
necessarily. Nevertheless, this solution clearly demonstrates the the timedependent vacuum field equations do support gravitational wave solutions
as long as the g22 metric function is time-dependent and the gravitational
wave condition is imposed. Consequently, a radially collapsing, expanding,
or pulsating spherical distribution of mass can indeed emit gravitational
waves into the vacuum region outside of the mass distribution.

10

Conclusion

As stated in Section 1, it is necessary to distinguish between the conventional and proper formulations of Birkhoffs theorem. The conventional
formulation of Birkhoffs theorem states that the only solution of the timedependent spherically symmetric vacuum gravitational field equations is a
particular time-independent solution. The conventional formulation is false
since it is predicated on the assumptions that (1) the g22 function is independent of the time-like coordinate and (2) the g22 function has a particular
dependence on the radial coordinate r, g22 = r2 . It is the time-independent
assumption that is the primary reason that the conventional formulation is
invalid.
A proper formulation of Birkhoffs theorem must explicitly state the crucial but unacknowledged assumptions of the original theorem. This includes
the condition that the metric tensor must be diagonal. The critical nature
of the time-independent assumption, g22|t = 0, is demonstrated in Section 6
20

and Section 7: a time-independent g22 function is both a necessary and


sufficient condition that the solution of the time-dependent vacuum field
equations is the solution of the time-independent field equations. In addition, removing the unnecessary second coordinate dependence assumption, that the g22 function is given by g22 = r2 , shows that the relevant
time-independent solution is the Combridge-Janne solution. Consequently,
as outlined in Section 1, Birkhoffs theorem must be formulated in the following way:

Birkhoff s Theorem: Proper Formulation. The solution


of the time-dependent spherically symmetric vacuum field equations is identical to the solution of the time independent spherically symmetric vacuum field equations if the metric is diagonal
(g0k = 0 for k = 1, 2, 3) and if and only if the g22 metric function is independent of the time-like coordinate. Furthermore,
the time-independent spherically symmetric solution is the timeindependent Combridge-Janne solution.

An immediate and very important consequence of the proper formulation


of Birkhoffs theorem is that an entire class of uninvestigated solutions of
the time-dependent spherically symmetric vacuum field equations exists for
which R|t 6= 0. This may be stated with the force of a corollary to the proper
formulation of Birkhoffs theorem:

Corollary of Proper Formulation. Solutions of the timedependent spherically symmetric vacuum field equations exist
and are inequivalent to the time-independent solution if and only
if the g22 metric function is a function of the time-like coordinate.
The time dependent solutions presented in Section 7 and Section 9 provide
two distinct illustrations of this corollary. Of particular interest, the results
of Section 9 demonstrate the existence of spherical gravitational wave solutions of the vacuum equations, thereby permitting a collapsing, expanding,
or pulsating distribution of mass to emit such waves into the region exterior
to that mass distribution. Since this is not possible for systems for which
R|t = 0, this illustrates the crucial dependence of the gravitational wave
solution on the time dependence of the g22 metric function R(r, t).
21

References
[1] G.D. Birkhoff, Relativity and Modern Physics (Harvard University
Press, Cambridge MA, 1923).
[2] C.W. Misner, K.S. Thorne, and J.A. Wheeler, Gravitation (Freeman,
San Francisco, 1973).
[3] A.H. Abbassi, General Birkhoffs Theorem, arXiv:gr-qc/0103103v1
(2001).
[4] R. Adler, M. Bazin, and M. Schiffer, Introduction to General Relativity,
Second Edition (McGraw-Hill, New York, 1975).
[5] H.P. Roberson, Kinematics and World Structure, Astrophys J., 82,
284 (1935).
[6] G.C. McVittie, Mon. Not. R. Astron. Soc., 93, 325 (1933).
[7] R.E. Salvino and R.D. Puff, The Two Schwarzschild Solutions: A
Critica Appraisal, preprint (2013). Available on academia.edu.
[8] W. de Sitter, On Einsteins Theory of Gravitation, and Its Astronomical Consquences (First Paper), Month. Not. R. Astr. Soc., 76, 699
(1916). In particular, see sections 10 on pg. 711 and 11 on pg. 714.
[9] R.E. Salvino and R.D. Puff, The Combridge-Janne Solution and the
g22 Metric Function, preprint (2013). Available on academia.edu.
[10] J.T. Combridge, Phil. Mag., 45, 726 (1923).
[11] H.A. Jannes, Bull. Acad. R. Belg., 9, 484 (1923).

22

Das könnte Ihnen auch gefallen