A Modification of Einstein-Schrödinger Theory That Contains Both General Relativity and Electrodynamics

Gen Relativ Gravit (2008) 40:17451769
DOI 10.1007/s10714-007-0572-6
RESEARCH ARTICLE
A modication of EinsteinSchrdinger theory
that contains both general relativity
and electrodynamics
J. A. Shifflett
Received: 20 September 2007 / Accepted: 3 December 2007 / Published online: 9 January 2008
Springer Science+Business Media, LLC 2007
Abstract We modify the EinsteinSchrdinger theory to include a cosmological
constant
z
which multiplies the symmetric metric, and we show how the theory can
be easily coupled to additional elds. The cosmological constant
z
is assumed to
be nearly cancelled by Schrdingers cosmological constant
b
which multiplies the
nonsymmetric fundamental tensor, such that the total =
z
+
b
matches mea-
surement. The resulting theory becomes exactly EinsteinMaxwell theory in the limit
as |
z
| . For |
z
| 1/(Planck length)
2
the eld equations match the ordinary
Einstein and Maxwell equations except for extra terms which are <10
16
of the usual
terms for worst-case eld strengths and rates-of-change accessible to measurement.
Additional elds can be included in the Lagrangian, and these elds may couple to the
symmetric metric andthe electromagnetic vector potential, just as inEinsteinMaxwell
theory. The ordinary Lorentz force equation is obtained by taking the divergence of
the Einstein equations when sources are included. The EinsteinInfeldHoffmann
(EIH) equations of motion match the equations of motion for EinsteinMaxwell the-
ory to Newtonian/Coulombian order, which proves the existence of a Lorentz force
without requiring sources. This xes a problem of the original EinsteinSchrdinger
theory, which failed to predict a Lorentz force. An exact charged solution matches
the ReissnerNordstrm solution except for additional terms which are 10
66
of the
usual terms for worst-case radii accessible to measurement. An exact electromagnetic
plane-wave solution is identical to its counterpart in EinsteinMaxwell theory.
Keywords EinsteinSchrodinger theory EinsteinStraus theory
Cosmological constant
J. A. Shifflett (B)
Department of Physics, Washington University,
1 Brookings Drive, St Louis, MO 63130, USA
e-mail: shifflet@hbar.wustl.edu
123
1746 J. A. Shifflett
1 Introduction
The EinsteinSchrdinger theory is a generalization of vacuum general relativity
which allows non-symmetric elds. The theory without a cosmological constant was
rst proposed by Einstein and Straus [15]. Schrdinger later showed that it could
be derived from a very simple Lagrangian density if a cosmological constant was
included [68]. Einstein and Schrdinger suspected that the theory might include
electrodynamics, but no Lorentz force was found [9, 10] when using the Einstein
InfeldHoffmann (EIH) method [11, 12]. Here we show that a simple modication
of the EinsteinSchrdinger theory closely approximates EinsteinMaxwell theory,
and the Lorentz force does result from the EIH method, and in fact the ordinary
Lorentz force equation results when sources are included. The modication is the
addition of a second cosmological term
z
g
, where g
is the symmetric metric.

We assume this term nearly cancels Schrdingers bare cosmological term
b
N
,
where N
is the nonsymmetric fundamental tensor. The total cosmological con-

stant =
b
+
z
can then match cosmological measurements of the accelerating
universe. Our theory is related to one in [13], but it is roughly the electromagnetic
dual of that theory, and it allows coupling to additional elds (sources), and it allows
= 0.
The origin of our
z
is unknown. One possibility is that
z
could arise from vac-
uumuctuations, an idea that has been discussed by many authors [1417]. Zero-point
uctuations are essential to both QED and the Standard-Model, and are the cause of
the Casimir force [15] and other effects. Another possibility is that
z
arises dynam-
ically, related to the minimum of a potential of some additional eld in the theory.
Speculation about the origin of this second cosmological constant is beyond the scope
of this paper. Our main goal here is to demonstrate that the theory closely approximates
EinsteinMaxwell theory.
Like EinsteinMaxwell theory, our theory can be coupled to additional elds using
a symmetric metric g
and vector potential A
, and it is invariant under a U(1)

gauge transformation. The theory does not enlarge the invariance group. When cou-
pled to the Standard Model, the combined Lagrangian is invariant under the usual
U(1) SU(2) SU(3) gauge group. The usual U(1) gauge term F
is incorpo-
rated together with the geometry, and is not explicitly in the Lagrangian. Whether this
is a step backwards from EinsteinMaxwell theory coupled to the Standard Model,
or whether the SU(2) and SU(3) gauge terms could also be incorporated using non-
Abelian elds as in [18, 19], or by using higher spacetime dimensions, is speculation
beyond the scope of this paper.
This paper is organized as follows. In Sect. 2 we discuss the Lagrangian density. In
Sects. 35 we derive the eld equations and quantify howclosely they approximate the
eld equations of EinsteinMaxwell theory. In Sect. 6 we derive the ordinary Lorentz
force equation by taking the divergence of the Einstein equations when sources are
included. In Sect. 7 we derive the Lorentz force using the EIH method, which requires
no sources in the Lagrangian. In Sect. 8 we give an exact charged solution and show
that it closely approximates the ReissnerNordstrm solution. In Sect. 9 we give an
exact electromagnetic plane-wave solution which is identical to its counterpart in
EinsteinMaxwell theory.
123
A modication of EinsteinSchrdinger theory 1747
2 The Lagrangian density
EinsteinMaxwell theory can be derived from a Palatini Lagrangian density,
L(
, g
, A
) =
1
16
g
_
g
( ) +(n2)
b
_
+
1
4
gA
[,]
g
A
[,]
+L
m
(u
, , g
, A
, . . .). (1)
Here
b
is a bare cosmological constant. The L
m
termcouples the metric g
and elec-
tromagnetic potential A
to additional elds, such as a hydrodynamic velocity vector

u
, spin-1/2 wavefunction , or perhaps the other elds of the Standard Model. The
original EinsteinSchrdinger theory allows a nonsymmetric N
and

in place
of the symmetric g
and
, and excludes the

gA
[,]
g
A
[,]
term. Our
-renormalized EinsteinSchrdinger theory introduces an additional cosmologi-
cal term
g
z
,
L(
, N
) =
1
16
N
_
N
) +(n2)
b
_
1
16
g (n2)
z
+L
m
(u
, , g
, A
, . . .), (2)
where
b
z
so that the total matches astronomical measurements [20],
=
b
+
z
10
56
cm
2
, (3)
and the physical metric and electromagnetic potential are dened to be
g g
=

NN
()
, A
[]
/[(n1)
_
2
b
]. (4)
Equation (4) denes g
unambiguously because

g =[det(
g g
)]
1/(n2)
.
Here and throughout this paper we use geometrized units with c = G = 1, the
symbols ( ) and [ ] around indices indicate symmetrization and antisymmetrization,
g =det(g
), N =det(N
), and N
is the inverse of N
such that N
.
The dimension is assumed to be n = 4, but n is retained in the equations to
show how easily the theory can be generalized. The L
m
term is not to include a
gA
[,]
g
A
[,]
part but may contain the rest of the Standard Model. In (2),
R
) is a form of Hermitianized Ricci tensor [1],

R
) =

((),)
+

()

[]
[]
/(n1). (5)
This tensor reduces to the ordinary Ricci tensor when
[]
=0 and
[,]
=0, as
occurs in ordinary general relativity.
123
It is helpful to decompose

into a new connection

, and A
from (4),
+(
)
_
2
b
, (6)
where

+(
[]

[]
)/(n1). (7)
By contracting (7) on the right and left we see that

has the symmetry
()
=

, (8)
so it has only n
3
n independent components. Substituting (6) into (5) gives
R
) = R
(

) +2A
[,]
_
2
b
. (9)
Using (9), the Lagrangian density (2) can be written in terms of

and A
,
L(
, N
) =
1
16
N
_
N
(

R
+2A
[,]
_
2
b
) +(n2)
b
_
1
16
g (n2)
z
+L
m
(u
, , g
, A
, . . .). (10)
Here

R
=R
(

), and from (8, 5) we have
(,)
+

. (11)
From (6, 8),

and A
fully parameterize

and can be treated as independent

variables. It is simpler to calculate the eld equations by setting L/

= 0 and
L/ A
= 0 instead of setting L/
= 0, so we will follow this method.

To do quantitative comparisons of this theory to EinsteinMaxwell theory we will
need to use some value for
z
. One possibility is that
z
results from zero-point
uctuations [1417], in which case using (3) we get
b

z
C
z
4
c
l
2
P
10
66
cm
2
, (12)
c
= (cutoff frequency)1/l
P
, (13)
C
z
=
1
2
_
fermion boson
spin states spin states

_

60
2
(14)
where l
P
=(Planck length)=1.610
33
cm. We will also consider the limit
c
,
|
z
| ,
b
as in QED, and we will prove that
lim
|
z
|
_
-renormalized
EinsteinSchrdinger theory
_
=
_
EinsteinMaxwell
Theory
_
. (15)
123
The Hermitianized Ricci tensor (5) has the following invariance properties
R
T
) = R
) (T = transpose), (16)
R
,]
) = R
) for an arbitrary (x
). (17)
From (16, 17), the Lagrangians (2, 10) are invariant under charge conjugation,
QQ, A
, N
, N
,
(18)
and also under an electromagnetic gauge transformation
e
i
, A

h
Q

,
,

+
2
h
Q

,]
_
2
b
,
(19)
assuming that L
m
is invariant. With
b
> 0,
z
< 0 as in (12) then

, N
and N
are all Hermitian,

R
and R
) are Hermitian from (16), and g
, A
and L are real from (4, 2, 10).

In this theory the metric (4) is used for measuring spacetime intervals, for calcu-
lating geodesics, and for raising and lowering of indices. The covariant derivative ;
is always done using the Christoffel connection formed from g
=
1
2
g
(g
,
+ g
,
g
,
). (20)
We will see that taking the divergence of the Einstein equations using (20, 4) gives
the ordinary Lorentz force equation. The electromagnetic eld is dened in terms of
the potential (4)
F
= A
,
A
,
. (21)
However, we will also dene another eld f

g f

=

NN
[]
1/2
b
/
2 i. (22)
Then from (4), g
and f

2 i
1/2
b
are parts of a total eld,
(
N/
g )N
= g
+ f

2 i
1/2
b
. (23)
We will see that the eld equations require f
to a very high precision. The

definitions (4) of g
and A
in terms of the fundamental elds N
may seem
unnatural from an empirical viewpoint. On the other hand, our Lagrangian density (2)
seems simpler than (1) of EinsteinMaxwell theory, it contains fewer elds, and these
123
elds have no symmetry restrictions. However, these are all very subjective consider-
ations. It is much more important that our theory closely matches EinsteinMaxwell
theory, and hence measurement.
Note that there are many nonsymmetric generalizations of the Ricci tensor besides
the HermitianizedRicci tensor R
) from(5) andthe ordinaryRicci tensor R
).
For example, we could form any weighted average of R
), R
), R
T
)
and R
T
), and then add any linear combination of the tensors

[,]
,

[|,|]
,
[]
[]
,

[]
[]
, and

[]
[]
. All of these generalized Ricci tensors would be
linear in

,
, quadratic in

, and would reduce to the ordinary Ricci tensor when
[]
= 0 and
[,]
= 0 as occurs in ordinary general relativity. Even if we limit
the tensor to only four terms, there are still eight possibilities. We assert that invari-
ance properties like (16, 17) are the most sensible way to choose among the different
alternatives, not criteria such as the number of terms in the expression.
Finally, let us discuss some notation issues. We use the symbol
for the Chris-

toffel connection (20) whereas Einstein and Schrdinger used it for our

and

respectively. We use the symbol g
for the symmetric metric (4) whereas Einstein

and Schrdinger used it for our N
, the nonsymmetric fundamental tensor. Also, to

represent the inverse of N
we use N
instead of the more conventional N
,
because this latter notation would be ambiguous when using g
to raise indices.
While our notation differs from previous literature on the EinsteinSchrdinger the-
ory, this change is required by our explicit metric definition, and it is necessary to be
consistent with the much larger body of literature on EinsteinMaxwell theory.
3 The Einstein equations
To set L/(
NN
)= 0 we need some initial results. Using (4) and the identities

det(s M)= s
n
det(M), det(M
1
)= 1/det(M) gives
N = (det(
NN
))
1/(n2)
, (24)
g = (det(
g g
))
1/(n2)
= (det(
NN
()
))
1/(n2)
. (25)
Using (24, 25, 4) and the identity (det(M
))/ M
= M
1
det(M
) gives
N
(
NN
)
=
N
(n2)
,
g
(
NN
)
=
g
(n2)
. (26)
Setting L/(
NN
)= 0 using (10, 26) gives the eld equations,
+2A
[,]
_
2
b
+
b
N
+
z
g
= 8S
, (27)
where S
and the energy-momentum tensor T
are dened by
S
2
L
m
(
NN
)
= 2
L
m
(
gg
)
, (28)
123
T
1
2
g
, S
= T
1
(n 2)
g
. (29)
The second equality in (28) results because L
m
in (2) contains only the metric
g g
NN
()
from (4), and not

NN
[]
. Taking the symmetric and
antisymmetric parts of (27) and using (21) gives
R
()
+
b
N
()
+
z
g
= 8
_
T

1
(n 2)
g
_
, (30)
N
[]
= F
2 i
1/2
b

R
[]
1
b
. (31)
Also from the curl of (31) we get
R
[,]
+
b
N
[,]
= 0. (32)
To put (30) into a form which looks more like the ordinary Einstein equations, we
need some preliminary results. The definitions (4, 22) of g
and f
can be inverted
exactly to give N
in terms of g
and f
. An expansion in powers of
1
b
will
better serve our purposes, and is derived in Appendix A,
N
()
=g
2
_
f

1
2(n2)
g
f

f
1
b
+( f
4
)
2
b
(33)
N
[]
= f
2 i
1/2
b
+( f
3
)
3/2
b
(34)
Here the notation ( f
3
) and ( f
4
) is for terms like f
and f
.
Let us consider the size of these higher order terms relative to the leading order term
for worst-case elds accessible to measurement. In geometrized units an elementary
charge has
Q
e
= e
_
G
c
4
=
_
e
2
hc
G
h
c
3
=

l
P
= 1.38 10
34
cm (35)
where = e
2
/
hc is the ne structure constant and l

P
=
_
G
h/c
3
is the Planck
length. If we assume that charged particles retain f
1
0
Q/r
2
down to the smallest
radii probed by high energy particle physics experiments (10
17
cm) we have from
(35, 12),
| f
1
0
|
2
/
b
(Q
e
/(10
17
)
2
)
2
/
b
10
66
. (36)
Here | f
1
0
| is assumed to be in some standard spherical or cartesian coordinate system.
If an equation has a tensor term which can be neglected in one coordinate system, it
can be neglected in any coordinate system, so it is only necessary to prove it in one
coordinate system. The elds at 10
17
cm from an elementary charge would be larger
than near any macroscopic charged object, and would also be larger than the strongest
123
plane-wave elds. Therefore the higher order terms in (3334) must be <10
66
of the
leading order terms, so they will be completely negligible for most purposes.
In Sect. 5 we will calculate the connection equations resulting from L/

=0.
Solving these equations gives (53, 54, 57, 59), which can be abbreviated as
()
=
+O(
1
b
),

[]
= O(
1/2
b
), (37)
= G
+O(
1
b
),

R
[]
= O(
1/2
b
), (38)
where
is the Christoffel connection (20),

R
=R
(

), R
=R
( ) and
=

R
()

1
2
g
, G
= R

1
2
g
R. (39)
In (38) the notation O(
1
b
) and O(
1/2
b
) indicates terms like f

;
f

;
1
b
and
f
[,];
1/2
b
.
From the antisymmetric part of the eld equations (31) and (34, 38) we get
f
= F
+O(
1
b
). (40)
So f
and F
only differ by terms with

b
in the denominator, and the two become
identical in the limit as
b
. Combining (30) with its contraction, and substituting
(39, 33, 3) gives the Einstein equations
= 8T

b
_
N
()
1
2
g
_
+
z
_
n
2
1
_
g
, (41)
= 8T
+2
_
f
1
4
g
f

f
_
+
_
n
2
1
_
g
+( f
4
)
1
b
(42)
From (28, 29) we see that T
will be the same as in ordinary general relativity, for

example when we include classical hydrodynamics or spin-1/2 elds as in [21, 22].
Therefore from (40, 38), equation (42) differs from the ordinary Einstein equations
only by terms with
b
in the denominator, and it becomes identical to the ordinary
Einstein equations in the limit as
b
(with an observationally valid total ). In
Sect. 5 we will examine how close the approximation is for
b
from (12).
4 Maxwells equations
Setting L/ A
= 0 and using (10, 22) gives

0 =
2 i
1/2
b
2
g
(
NN
[]
)
,
4j
=
(
g f

)
,
g
4j
, (43)
123
where
j
=
1
g
_
L
m
A
_
L
m
A
,
_
,
_
. (44)
From (43, 21) we get Maxwells equations,
f

;
= 4j
, (45)
F
[,]
= 0. (46)
where f
= F
+O(
1
b
) from(40). From(2, 44) we see that j
will be the same as

in ordinary general relativity, for example when we include classical hydrodynamics
or spin-1/2 elds as in [21, 22]. From (40), we see that equations (45, 46) differ from
the ordinary Maxwell equations only by terms with
b
in the denominator, and these
equations become identical to the ordinary Maxwell equations in the limit as
b
.
In Sect. 5 we will examine how close the approximation is for
b
from (12).
Because L
m
in (2) couples to additional elds only through g
and A
, any equa-
tions associated with additional elds will be the same as in ordinary general relativity.
For example in the spin-1/2 case, setting L/

=0 will give the ordinary Dirac equa-
tion in curved space as in [21, 22]. It would be interesting to investigate what results
if one includes f
, N
or

in L
m
, although there does not appear to be any
empirical reason for doing so. A continuity equation follows from (45) regardless of
the type of source,
j
;
=
1
4
f

;[;]
= 0. (47)
Note that the covariant derivative in (45, 47) is done using the Christoffel connection
(20) formed from the symmetric metric (4).
5 The connection equations
Setting L/

= 0 with a Lagrange mulitiplier term
[]
to enforce the sym-
metry (8), and using (10, 43) gives
(
NN
)
,
+

NN
NN
NN
=
8
2 i
(n1)
1/2
b
g j
[
. (48)
These are the connection equations, analogous to g
;
=0 in the symmetric case.
Note that we can also derive Amperes law (43) by antisymmetrizing and contracting
these equations. From the definition of matrix inverse, N
=(1/N) N/ N
and
123
N
we get the identity

(
N )
,
=

N
N
N
,
=
N
2
N
N
,
=
N
2
N
,
N
. (49)
Contracting (48) with N
using (8, 49), and dividing this by (n2) gives,

(
N )
,

N =
8
2 i
(n1)(n2)
1/2
b
g j
N
[]
. (50)
Multiplying (48) by N
and using (50) gives

N
,
=
8
2 i
(n1)
1/2
b
N
_
N
[
N
]
+
N
[]
N
(n2)
_
j
. (51)
Equation (51) together with (30, 32, 8) are often used to dene the Einstein
Schrdinger theory, particularly when T
=0, j
=0.
Equations (48) or (51) can be solved exactly [23, 24], similar to the way g
;
=0
can be solved to get the Christoffel connection. An expansion in powers of
1
b
will
better serve our purposes, and such an expansion is derived in Appendix E of [21],
and is also stated without derivation in [25],
, (52)
()
=2
_
f

(
f
)
;
+ f

f
(;)
+
1
4(n2)
(( f

f
)
,
2( f

f
)
,(
)
)
+
4
(n2)
j
_
f

+
2
(n1)
f
(
)
__
1
b
+( f
4
)
2
b
. . . , (53)
[]
=
_
1
2
( f
;
+ f

;
f

;
)+
8
(n1)
j
[
]
_
2 i
1/2
b
+( f
3
)
3/2
b
. . .
(54)
In (52),
is the Christoffel connection (20). The notation ( f

3
) and ( f
4
) refers to
terms like f

f

[;]
and f

f

(;)
. As in (33, 34), we see from (36)
that the higher order terms in (5354) must be <10
66
of the leading order terms, so
they will be completely negligible for most purposes.
Extracting
of (52) from the Hermitianized Ricci tensor (11) gives,

R
()
(

) = R
( )+
();
(;)
()
()
[]
[]
+
()
, (55)
R
[]
(

) =
[];
()
[]
[]
()
+
[]
. (56)
123
Substituting (5254, 45) into (55), and using (39) gives
(

G
) =
_
2 f

(
f
)
;;
+2 f

f
(; );
f

;
f

;
+ f

;
f
;
+
1
2
f

;
f

;
g
f

f
;;

1
4
g
( f

f
)
,
3
4
g
f
[;]
f
[
;
]
+8j
f
(;)

32
2
(n1)
j
+
16
2
(n1)
g
+( f
4
)
_
1
b
(57)
From (42) we can dene an effective energy momentum tensor

T
which applies
when G
is used in the Einstein equations and L

m
=0,
8

T
= 2
_
f
1
4
g
f

f
_
(

G
). (58)
Substituting (54, 45) into (56) gives
R
[]
=
_
3
2
f
[,];
+2 f

;[;]
2 f

;[;]
8(n2)
(n1)
j
[,]
_
2 i
1/2
b

(59)
As we have noted in Sects. 3 and 4, the
b
in the denominator of (57, 59) causes
our Einstein and Maxwell equations (42, 45, 46) to become the ordinary Einstein and
Maxwell equations in the limit as
c
, |
z
| ,
b
, and it also causes
the relation f
from (40) to become exact in this limit. Let us examine how

close these approximations are when
b
10
66
cm
2
as in (12).
We will start with the Einstein equations (42). Let us consider worst-case values
of the O(
1
b
) terms in (57) and compare these to the ordinary electromagnetic term
in (58). If we assume that charged particles retain f
1
0
Q/r
2
down to the smallest
radii probed by high energy particle physics experiments (10
17
cm) we have,
| f
1
0;1
/f
1
0
|
2
/
b
4/
b
(10
17
)
2
10
32
, (60)
| f
1
0;1;1
/f
1
0
|/
b
6/
b
(10
17
)
2
10
32
. (61)
So for electric monopole elds, terms like f

;
f

;
1
b
and f

f
(; );
1
b
in
(57) must be <10
32
of the ordinary electromagnetic term in (58). And regarding j
as a substitute for (1/4) f

;
from (45), the same is true for the j
terms. For an
electromagnetic plane-wave in a at background space
A
= A
sin(k
),
= 1, k
= k
= 0, (62)
f
= 2A
[,]
= 2A
[
k
]
cos(k
), j
= 0. (63)
123
Here A is the magnitude, k
is the wavevector, and
is the polarization. Substituting

(62, 63) into (57), all of the terms vanish for a at background space. Also, for the
highest energy gamma rays known in nature (10
20
eV, 10
34
Hz) we have from (12),
| f
1
0;1
/f
1
0
|
2
/
b
(E/
hc)
2
/
b
10
16
, (64)
| f
1
0;1;1
/f
1
0
|/
b
(E/
hc)
2
/
b
10
16
. (65)
So for electromagnetic plane-wave elds, even if some of the terms in (57) were
non-zero because of spatial curvatures, they must still be <10
16
of the ordinary
electromagnetic term in (58). Therefore even for the most extreme worst-case elds
accessible to measurement, the extra terms in the Einstein equations (42) must all be
<10
16
of the ordinary electromagnetic term.
Now let us look at the approximation f
from (40), and Maxwells equa-

tions (45, 46). From the covariant derivative commutation rule, the cyclic identity
2R
[]
= R
, the definition of the Weyl tensor C
, and the Einstein equa-

tions R
= g
+( f
2
) . . . from (42) we get
2 f

;[;]
=
1
2
f

C
+
(n2)
(n1)
f
+( f
3
) (66)
Substituting (34) into the eld equations (31) and using (59, 66) we get
f
= F
+
_
[,]
+ f

C
+
2(n 2)
(n 1)
f
+
8(n 2)
(n 1)
j
[,]
+( f
3
)
_
1
b
(67)
where
= (LeviCivita tensor), C
= (Weyl tensor), and
=
1
4
f
[,]
, f
[,]
=
2
3

. (68)
The
[,]
1
b
term in (67) is divergenceless so that it has no effect on Amperes
law (45). The f
/
b
term is 10
122
of f
from (3, 12). The ( f

3
)
1
b
term is
<10
66
of f
from (36). The largest observable values of the Weyl tensor might
be expected to occur near the Schwarzschild radius, r
s
= 2Gm/c
2
, of black holes,
where it takes on values around r
s
/r
3
. The largest value of r
s
/r
3
would occur near
the lightest black holes, which would be of about one solar mass, where from (12),
C
0101
b
r
2
s
=
1
b
_
c
2
2Gm
_
2
10
77
. (69)
And regarding j
as a substitute for (1/4) f

;
from (45), the j
[,]
1
b
term is
<10
32
of f
from (61). Therefore, the last four terms in (67) must all be <10
32
123
of f
. Consequently, even for the most extreme worst-case elds accessible to mea-
surement, the extra terms in Maxwells equations (45, 46) must be < 10
32
of the
ordinary terms.
The divergenceless term
[,]
1
b
of (67) should also be expected to be
<10
32
of f
from (60, 61, 68). However, we need to consider the possibility where
changes extremely rapidly. Taking the curl of (67), the F
and j
[,]
terms drop
out, and we get something similar to the Proca equation [26, 27],
=
_
; ;
+
1
2
( f

C
[
)
,]
+
(3n7)
(n1)
+( f
3
)
_
1
2
b
(70)
Here the constraint
;
= 0 results from the definition (68) and we are using a
(1, 1, 1, 1) metric signature. Equation (70) suggests that
Proca-wave solu-
tions might exist in this theory. Assuming that the magnitude of C
is roughly
proportional to
for such waves, and assuming that f
goes according to (67) with

F
=0, the extra terms in (70) could perhaps be neglected in the weak eld approx-
imation. Using (70) and
b
z
=C
z
4
c
l
2
P
from (12), such Proca-wave solutions
would have an extremely high minimum frequency
Proca
=
_
2
b

_
2C
z

2
c
l
P
10
43
rad/s, (71)
where the cutoff frequency
c
and C
z
come from (13, 14).
There are several points to make about (70, 71). (1) A particle associated with a
eld would have mass

h
Proca
, which is much greater than could be produced by
particle accelerators, and so it would presumably not conict with high energy physics
experiments. (2) We have recently shown that sin[ krt ] Proca-wave solutions do not
exist in the theory, using an asymptotically at NewmanPenrose 1/r expansion sim-
ilar to [28, 29]. However, it is still possible that wave-packet solutions could exist. (3)
Substituting the k =0 at space Proca-wave solution
=(0, 1, 0, 0)sin[
Proca
t ] and
F
=0 into(67, 58, 57), andassuminga at backgroundspace gives

T
00
=2/
b
<0.
This suggests that Proca-wave solutions might have negative energy, but because
sin[ kr t ] solutions do not exist, and because of the other approximations used,
this calculation is extremely uncertain. (4) With a cutoff frequency
c
1/l
P
from
(13) we have
Proca
>
c
from (71, 13, 14), so Proca-waves would presumably be cut
off. More precisely, (71) says that Proca-waves would be cut off if
c
>1/(l
P
2C
z
).
Whether
c
is caused by a discreteness, uncertainty or foaminess of spacetime near the
Planck length [3034], or by some other effect, the same
c
which cuts off
z
in (12)
should also cut off very high frequency electromagnetic and gravitational waves, and
Proca-waves. (5) If wave-packet Proca-wave solutions do exist, and they have negative
energy, it is possible that
could function as a kind of built-in PauliVillars eld.

Pauli-Villars regularization in quantum electrodynamics requires a negative energy
Proca eld with a mass

h
Proca
that goes to innity as
c
, as we have from
(71). (6) As mentioned initially, it might be more correct to take the limit of this
theory as
c
, |
z
| ,
b
, as in quantum electrodynamics. In this
limit (70, 71) require that
0 or
Proca
, and the theory becomes exactly
EinsteinMaxwell theory as in (15). (7) Finally, we should emphasize that Proca-wave
123
solutions are only a possibility suggested by equation (70). Their existence and their
possible interpretation are just speculation at this point. We are continuing to pursue
these questions.
6 The Lorentz force equation
A generalized contracted Bianchi identity for this theory can be derived using only
the connection equations (48) and the symmetry (8) of

,
(
NN
NN
)
,

NN
R
,
= 0. (72)
This identity can also be written in terms of g
, f

and

G
from (4, 22, 39),
;
=
_
3
2
f

R
[,]
+ f

;

R
[]
_
2 i
1/2
b
. (73)
The identity was originally derived [3, 7] assuming j
=0 in (48). The derivation for

j
=0 was rst done [25] by applying an innitesimal coordinate transformation to

an invariant integral, and it is also done in Appendix B of [21] using a much different
direct computation method. Clearly (72, 73) are generalizations of the ordinary con-
tracted Bianchi identity 2(
g R
)
,
g g
R
,
= 0 or G
;
= 0, which is
also valid in this theory.
Another useful identity [13] can be derived using only the definitions (4, 22)
_
N
(
)
1
2
_
;
=
_
3
2
f

N
[,]
+ f

;
N
[]
_
2 i
1/2
b
. (74)
The ordinary Lorentz force equation results from taking the divergence of the
Einstein equations (41) using (73, 45, 31, 74, 21)
T
;
=
1
8
_

G
;
+
b
_
N
(
)
1
2
_
;
_
= F
. (75)
Note that the covariant derivatives in (73, 74, 75) are all done using the Christoffel
connection (20) formed from the symmetric metric (4).
7 The EinsteinInfeldHoffmann equations of motion
For EinsteinMaxwell theory, the EIH method allows the equations of motion to be
derived directly from the electro-vac eld equations. For neutral particles the method
has been veried to PostNewtonian order [11], and in fact it was the method rst
used to derive the PostNewtonian equations of motion [35]. For charged particles the
method has been veried to PostCoulombian order [12, 36, 37], meaning that it gives
the same result as the Darwin Lagrangian [27]. In Sect. 6 we derived the exact Lorentz
force equation for this theory by including source terms in the Lagrangian. Here we
123
derive the Lorentz force using the EIHmethod because it requires no source terms, and
also to showdefinitely that the well known negative result of [9, 10] for the unmodied
EinsteinSchrdinger theory does not apply to the present theory. We will only cover
the bare essentials of the EIH method which are necessary to derive the Lorentz force.
We will also only calculate the equations of motion to Newtonian/Coulombian order,
because this is the order where the Lorentz force rst appears.
The EIHmethod assumes the slowmotion approximation, meaning that v/c1.
The elds are expanded in the form [11, 12, 36, 37],
g
/2, (76)
00
=
2
00
2
+
4
00
4
(77)
0k
=
3
0k
3
+
5
0k
5
(78)
i k
=
4
i k
4
(79)
A
0
=
2
A
0
2
+
4
A
0
4
(80)
A
k
=
3
A
k
3
+
5
A
k
5
(81)
f
0k
=
2
f
0k
2
+
4
f
0k
4
(82)
f
i k
=
3
f
i k
3
+
5
f
i k
5
(83)
where v/c is the expansion parameter, the order of each term is indicated with
a left subscript [9],
= diag(1, 1, 1, 1), and Latin indices run from 1 to 3.

The eld
(often called

h
in other contexts) is used instead of g
only because
it simplies the calculations. Because v/c, when the expansions are substituted
into the Einstein and Maxwell equations, a time derivative counts the same as one
higher order in . The general procedure is to substitute the expansions, and solve the
resulting eld equations order by order in , continuing to higher orders until a desired
level of accuracy is achieved. At each order in , one of the
l
terms and one of

the
l
f
terms will be unknowns, and the equations will involve known results from
previous orders because of the nonlinearity of the Einstein equations.
The expansions (7783) use only alternate powers of essentially because the
Einstein and Maxwell equations are second order differential equations [35], although
for higher powers of , all terms must be included to predict radiation [12, 36, 37].
Because v/c, the expansions have the magnetic components A
k
and f
i k
due
to motion at one order higher in than the electric components A
0
and f
0i
. As in
[12, 36, 37], f
0k
and f
i k
have even and odd powers of respectively. This is the oppo-
site of [9, 10] because we are assuming a direct definition of the electromagnetic eld
(22, 34, 67, 21) instead of the dual definition f

=
N
[]
/2 assumed in [9, 10].
The eld equations are assumed to be of the standard form
G
= 8T
where G
= R

1
2
g
, (84)
or R
= 8S
where S
= T

1
2
g
. (85)
123
However, with the EIH method we solve a sort of quasi-Einstein equations,
0 =

G
8

T
, (86)
where
= R

1
2
,

T
= S

1
2
. (87)
Here the use of
instead of g
is not an approximation because (85) implies (86)

whether

G
and

T
are dened with
or g
. Note that the references use many

different notations in (86): instead of

G
others use
/2+
/2+
or
[LS: ] and instead of 8

T
others use 2S
or [RS: ].
The equations of motion result as a condition that the eld equations (86) have
acceptable solutions. In the language of the EIH method, acceptable solutions are
those that contain only pole terms and no dipole terms, and this can be viewed
as a requirement that the solutions should resemble ReissnerNordstrm solutions
asymptotically. To express the condition of solvability we must consider the integral
of the eld equations (86) over 2D surfaces S surrounding each singularity,
l
C
=
1
2
S
_
(
l

G
k
8
l

T
k
)n
k
dS. (88)
Here n
k
is the surface normal and l is the order in . Assuming that the divergence
of the Einstein equations (84) vanishes, and that (86) has been solved to all previous
orders, it can be shown [11] that in the current order
(
l

G
k
8
l

T
k
)
|k
=0. (89)
Here and throughout this section | represents ordinary derivative [11]. FromGreens
theorem, (89) implies that
l
C
in (88) will be independent of surface size and shape

[11]. The condition for the existence of an acceptable solution for
4
i k
is simply
4
C
i
= 0, (90)
and these are also our three O(
4
) equations of motion [11]. The C
0
component of
(88) causes no constraint on the motion [11] so we only need to calculate

G
i k
and

T
i k
.
At this point let us introduce a Lemma from [11] which is derived from Stokess
theorem. This Lemma states that
S
_
F
( )kl|l
n
k
dS = 0 if F
( )kl
= F
( )lk
, (91)
where F
( )kl
is any antisymmetric function of the coordinates, n
k
is the surface nor-
mal, and S is any closed 2D surface which may surround a singularity. The equation
123
4
C
i
=0 is a condition for the existence of a solution for
4
i k
because
4
i k
is found by
solving the O(
4
) eld equations (86), and
4
C
i
is the integral (88) of these equations.
However, because of the Lemma (91), it happens that the
4
i k
terms in
4

G
i k
integrate
to zero in (88), so that
4
C
i
is actually independent of
4
i k
. In fact it is a general rule
that C
i
for one order can be calculated using only results from previous orders [11],
and this is a crucial aspect of the EIH method. Therefore, the calculation of the O(
4
)
equations of motion (90) does not involve the calculation of
4
i k
, and we will see
below that it also does not involve the calculation of
3
f
i k
or
4
f
0k
.
The
4

G
i k
contribution to (88) is derived in [11]. For two particles with masses m
1
,
m
2
and positions
i
1
,
i
2
, the O(
4
) term from the integral over the rst particle is
G
4
C
i
=
1
2
1
_
4

G
i k
n
k
dS = 4
_
m
1
i
1
m
1
m
2
i
1
_
1
r
_
_
, (92)
where
r =
_
(
s
1

s
2
)(
s
1

s
2
)
. (93)
If there is no other contribution to (88), then (90) requires that

G
4
C
i
=0 in (92), and the
particle acceleration will be proportional to a (m
1
m
2
/r) Newtonian gravitational
force. These are the EIH equations of motion for vacuum general relativity to O(
4
),
or Newtonian order.
Because our effective energy momentum tensor (58) is quadratic in f
, and the
expansions (77-83) begin with
2
terms, the O(
2
) O(
3
) calculations leading to
(92) are unaffected by the addition of the electromagnetic terms to the vacuum eld
equations. However, the 8
4

T
i k
contribution to (88) will add to the
4

G
i k
contribution.
To calculate this contribution, we will assume that our singularities in f
are simple
moving Coulomb potentials, and that
=0, =0. Then from (67, 8283) we see

that
2
F
0k
=
2
f
0k
, and from inspection of the extra terms in our Maxwell equations
(45, 46, 67) and Proca equation (70), we see that these equations are both solved to
O(
3
). Because (58) is quadratic in f
, we see from (8283) that only

2
f
0k
can
affect the O(
4
) equations of motion. Including only
2
f
0k
, our f
is then a sum of
two Coulomb potentials with charges Q
1
, Q
2
and positions
i
1
,
i
2
of the form
2
A
= (
2
, 0, 0, 0),
2
f
0k
= 2
2
A
[k|0]
=
2
|k
, (94)
2
=
1
+
2
,
1
= Q
1
/r
1
,
2
= Q
2
/r
2
, (95)
r
a
=
_
(x
s
s
a
)(x
s
s
a
)
, a = 1, . . . , 2 . (96)
Because (58) is quadratic in both f
and g
, and the expansions (77-83) start at
2
in both of these quantities, no gravitational-electromagnetic interactions will occur
at O(
4
). This allows us to replace covariant derivatives with ordinary derivatives,
and g
with
in (58). This also allows us to replace

T
from (86, 87) with (58).

123
Keeping only O(
4
) terms when (94) is substituted, the spacial part of (58) gives,
8
4

T
sm
= 2
_
f
s
0
f
0m

1
2
sm
f
r0
f
0r
_
+
_
2 f
a0
f
0(s|m)|a
+ f
0
s|a
f
0m|
a
+ f
0
a|s
f
a
0|m
1
2

sm
( f
r0
f
0r
)
|
a
|a
_
1
b
.
(97)
Note that
2
from (95) obeys Gausss law,
|a|a
= 0. (98)
Substituting (94) into (97) and using (98) gives
8
4

T
sm
= 2
_
|s
|m
+
1
2
sm
|r
|r
_
+
_
2
|a
|s|m|a

|s|a
|m|a
+
|a|s
|a|m
+
1
2
sm
(
|r
|r
)
|a|a
_
1
b
(99)
= 2
_
|s
|m
+
1
2
sm
|r
|r
_
2(
|[s
|a]|m
+
|r
|r|[s
a]m
)
|a
1
b
. (100)
From (91), the second group of terms in (100) integrates to zero in (88), so it can have
no effect on the equations of motion. The rst group of terms in (100) is what one gets
with EinsteinMaxwell theory [12, 36, 37], so at this stage we have effectively proven
that the theory predicts a Lorentz force.
For completeness we will nish the derivation. First, we see from (100, 98) that
4

T
sm|s
= 0. This is to be expected because of (89), and it means that the 8
4

T
sm
contribution to the surface integral (88) will be independent of surface size and shape.
This also means that only 1/distance
2
terms such as
sm
/r
2
or x
s
x
m
/r
4
can contribute
to (88). The integral over a term with any other distance-dependence would depend
on the surface radius, and therefore we know beforehand that it must vanish or cancel
with other similar terms [11]. Now,
|i
=
1
|i
+
2
|i
from (95). Because
1
|i
and
2
|i
both
go as 1/distance
2
, but are in different locations, it is clear from(100) that contributions
can only come from cross terms between the two. Including only these terms gives,
8
4

T
c
sm
= 2(
1
|s
2
|m
+
2
|s
1
|m
+
sm
1
|r
2
|r
). (101)
Some integrals we will need can be found in [11]. With = 1/
x
s
x
s
we have,
1
4
0
_

|m
n
m
dS = 1,
1
4
0
_

|a
n
m
dS =
1
3
am
. (102)
123
Using (101, 102, 95) and integrating over the rst particle we get,
1
2
1
_
_
8

T
sm
_
n
m
dS =
1
2
1
_
2(
1
|s
2
|m
+
2
|s
1
|m
+
sm
1
|r
2
|r
)n
m
dS (103)
=4Q
1
2
|s
(
1
)
_
1
3
1+
1
3
_
= 4Q
1
2
|s
(
1
). (104)
Using (90, 88, 104, 92, 95) we get
0 =
4
C
i
= 4
_
m
1
i
1
m
1
m
2
i
1
_
1
r
_
_
4Q
1
2
|i
(
1
) (105)
= 4
_
m
1
i
1
m
1
m
2
i
1
_
1
r
_
+ Q
1
Q
2
i
1
_
1
r
_
_
, (106)
where
r =
_
(
s
1

s
2
)(
s
1

s
2
)
. (107)
These are the EIH equations of motion for this theory to O(
4
), or Newtonian/
Coulombian order. These equations clearly exhibit the Lorentz force, and in fact they
match the O(
4
) equations of motion of EinsteinMaxwell theory.
8 An exact electric monopole solution
Here we give an exact charged solution for this theory which closely approximates
the ReissnerNordstrm solution [38, 39] of EinsteinMaxwell theory. A MAPLE
program [40] which checks the solution and the derivation [41] are available. The
solution is
ds
2
= cadt
2
1
ca
dr
2
cr
2
d
2
cr
2
sin
2
d
2
, (108)
f
10
=
Q
cr
2
,
N = r
2
sin ,
g = cr
2
sin , (109)
F
01
= A
0
=
Q
r
2
_
1 +
4M
b
r
3

4
3
b
+2
_
c 1
Q
2
b
r
4
__
1

b
__
, (110)
a = 1
2M
r

r
2
3
+
Q
2

V
r
2
_
1

b
_
, (111)
123
where (
) means /r, and c and

V are very close to one for ordinary radii,
c =
_
1
2Q
2
b
r
4
= 1
Q
2
b
r
4

(2i )!
[i !]
2
4
i
(2i 1)
_
2Q
2
b
r
4
_
i
, (112)
V =
r
b
Q
2
_ _
r
2
c dr
r
3
3
_
= 1+
Q
2
10
b
r
4
+
(2i )!
i !(i +1)! 4
i
(4i +1)
_
2Q
2
b
r
4
_
i
, (113)
and the nonzero connections are
1
00
=
aa
c
2
2

4a
2
Q
2
b
r
5
,

0
10
=

0
01
=
a
2a
,

1
11
=
a
2a
,
2
12
=

2
21
=

3
13
=

3
31
=
1
r
, (114)
1
22
= ar,

1
33
= ar sin
2
,

3
23
=

3
32
= cot ,

2
33
= sin cos ,
2
02
=

2
20
=

3
03
=

3
30
=
a
2 i Q
b
r
3
,

1
10
=

1
01
=
2a
2 i Q
b
r
3
.
The solution matches the ReissnerNordstrm solution except for terms which are
negligible for ordinary radii. To see this, rst recall that /
b
10
122
from (3, 12),
so the terms are all extremely tiny. Ignoring the terms and keeping only the
O(
1
b
) terms in (110, 111, 112, 113) gives
F
01
=
Q
r
2
_
1 +
4M
b
r
3

4Q
2
b
r
4
_
+O(
2
b
), (115)
A
0
=
Q
r
_
1 +
M
b
r
3

4Q
2
5
b
r
4
_
+O(
2
b
), (116)
a = 1
2M
r
+
Q
2
r
2
_
1 +
Q
2
10
b
r
4
_
+O(
2
b
), (117)
c = 1
Q
2
b
r
4
+O(
2
b
). (118)
For the smallest radii probed by high-energy particle physics we get from (36),
Q
2
b
r
4
10
66
. (119)
The worst-case value of M/
b
r
3
might be near the Schwarzschild radius r
s
of black
holes where r =r
s
=2M and M/
b
r
3
=1/2
b
r
2
s
. This value will be largest for the
lightest black holes, and the lightest black hole that we can expect to observe would
be of about one solar mass, where we have
M
b
r
3

1
2
b
r
2
s
=
1
2
b
_
c
2
2Gm
_
2
10
77
. (120)
123
From (119, 120, 3, 12) we see that our electric monopole solution (108-111) has a
fractional difference from the ReissnerNordstrm solution [38, 39] of at most 10
66
for worst-case radii accessible to measurement. Clearly our solution does not have
the deciencies of the Papapetrou solution [42, 43] in the original theory, and it is
almost certainly indistinguishable from the ReissnerNordstrm solution experimen-
tally. Also, when this solution is expressed in NewmanPenrose tetrad form, it can
be shown to be of Petrov Type-D [24]. And of course the solution reduces to the
Schwarzschild solution for Q= 0. And from (115-118) we see that the solution goes
to the ReissnerNordstrm solution exactly in the limit as
b
.
The only significant difference between our electric monopole solution and the
ReissnerNordstrmsolution occurs on the Planck scale. From(108, 112), the surface
area of the solution is [44],
_
surface
area
_
=
_
0
d
2
_
0
d
= 4r
2
c = 4r
2
_
1
2Q
2
b
r
4
. (121)
The origin of the solution is where the surface area vanishes, so in our coordinates the
origin is not at r = 0 but rather at
r
0
=
_
Q(2/
b
)
1/4
. (122)
From (35, 12) we have r
0
l
P
10
33
cm for an elementary charge, and r
0
2M for
any realistic astrophysical black hole. For Q/M<1 the behavior at the origin is hidden
behind an event horizon nearly identical to that of the ReissnerNordstrm solution.
For Q/M>1 where there is no event horizon, the behavior at the origin differs mark-
edly from the simple naked singularity of the ReissnerNordstrm solution. For the
ReissnerNordstrm solution all of the relevant elds have singularities at the origin,
with g
00
Q
2
/r
2
, A
0
= Q/r, F
01
= Q/r
2
, R
00
2Q
4
/r
6
and R
11
2/r
2
. For our
solution the metric has a less severe singularity at the origin, with g
11
r/
r r
0
.
Also, the elds N
, N
N, A
g f

,
g f
gg
gg
, and
the functions a and

V all have nite nonzero values and derivatives at the origin,
because it can be shown that

V(r
0
)=
2 [ (1/4)]
2
/6
2/3 = 1.08137. The elds

F
and
g

R
are also nite and nonzero at the origin, so if we use the tensor
density form of the eld equations (41, 45), there is no ambiguity as to whether the
eld equations are satised at this location.
9 An exact electromagnetic plane-wave solution
Here we give an exact electromagnetic plane-wave solution for this theory which
is identical to the electromagnetic plane-wave solution in EinsteinMaxwell theory,
usually called the BaldwinJeffery solution [4548]. We will not do a full derivation,
but a MAPLE program [40] which checks the solution is available. We present the
solution in the form of a pp-wave solution [47], and a gravitational wave compo-
nent is included for generality. The solution is expressed in terms of null coordinates
123
x, y, u = (t z)/
2, v = (t + z)/
2,
g
1 0 0 0
0 1 0 0
0 0 H 1
0 0 1 0
g f

=
0 0 0

f
x
0 0 0

f
y
0 0 0 0

f
x

f
y
0 0
, (123)
f
= 2A
[,]
= 2A
,[
k
]
=
0 0

f
x
0
0 0

f
y
0
f
x

f
y
0 0
0 0 0 0
g =

N = 1 (124)
where
k
= (0, 0, 1, 0), A
= (0, 0, A, 0), A =
2(x

f
x
+ y

f
y
), (125)
H = 2

H + A
2
(126)
= 2(h
+
x
2
+h
xy h
+
y
2
) +2(

f
2
x
+

f
2
y
)(x
2
+ y
2
), (127)
H = h
+
x
2
+h
xy h
+
y
2
+(y

f
x
x

f
y
)
2
. (128)
and the nonzero connections are
1
33
=
1
2
H
x
,

2
33
=
1
2
H
y
,

4
33
=
1
2
H
u

2
b
(

f
2
x
+

f
2
y
)
u
,
4
13
=
1
2
H
x

2i

f
x
u
,

4
31
=
1
2
H
x
+
2i

f
x
u
, (129)
4
23
=
1
2
H
y

2i

f
y
u
,

4
32
=
1
2
H
y
+
2i

f
y
u
.
Here h
+
(u), h
(u) characterize the gravitational wave component,

f
x
(u),

f
y
(u) char-
acterize the electromagnetic wave component, and all of these are arbitrary functions
of the coordinate u = (t z)/
2.
The solution above has been discussed extensively in the literature on Einstein
Maxwell theory [4548] so we will not interpret it further. It is the same solution
which forms the incoming waves for the BellSzekeres colliding plane-wave solution
[48], although the full BellSzekeres solution does not satisfy our theory because the
electromagnetic eld is not null after the collision.
10 Conclusions
The EinsteinSchrdinger theory is modied to include a cosmological constant
z
which multiplies the symmetric metric. This is assumed to be nearly cancelled by
Schrdingers bare cosmological constant
b
which multiplies the nonsymmetric
fundamental tensor, such that the total cosmological constant =
b
+
z
matches
123
measurement. The resulting theory closely approximates EinsteinMaxwell theory for
|
z
| 1/(Planck length)
2
, and it becomes exactly EinsteinMaxwell theory in the
limit as |
z
| .
Acknowledgments I am grateful to Clifford Will for discussions and for helpful comments on drafts of
this manuscript. This work was supported in part by the National Science Foundation under grant PHY 03-
53180.
Appendix A: Solution for N
in terms of g
and f
Here we invert the definitions (4, 22) of g
and f
to obtain (33, 34), the approxi-

mation of N
in terms of g
and f
. First let us dene the notation
f

= f

2 i
1/2
b
. (130)
We assume that |

f

| 1 for all components of the unitless eld

f

, and nd a
solution in the form of a power series expansion in

f

. Lowering an index on (23)

gives
(
N/
g )N
. (131)
Let us consider the tensor

f

=

f

g
. Because g
is symmetric and

f

is anti-
symmetric, it is clear that

f

=0. Also because

f
is symmetric it is clear that
= 0. In matrix language therefore tr(

f ) = 0, tr(

f
3
) = 0, and in fact
tr(

f
p
) =0 for any odd p. Using the well known formula det(e
M
) = exp (tr(M)) and
the power series ln(1x) = x x
2
/2 x
3
/3 x
4
/4 we then get [49],
ln(det(I

f )) = tr(ln(I

f )) =

f

/2 +(

f
4
) (132)
Here the notation (

f
4
) refers to terms like

f

. Taking ln(det()) on
both sides of (131) using (132) and the identities det(s M) = s
n
det(M), det(M
1
) =
1/det(M) gives
ln
_
g
_
=
1
(n 2)
ln
_
N
(n/21)
g
(n/21)
_
=
1
2(n 2)
+(

f
4
) (133)
Taking e
x
on both sides of (133) and using e
x
= 1 + x + x
2
/2 gives
g
= 1
1
2(n2)
+(

f
4
) (134)
123
Using the power series (1x)
1
=1 +x +x
2
+x
3
, or multiplying (131) term by
term, we can calculate the inverse of (131) to get [49]
(
g/
N )N
+

f

+

f

+

f

+(

f
4
) (135)
N
= (
N/
g )(g
+

f
+

f
+

f
+(

f
4
) ). (136)
Here the notation (

f
4
) refers to terms like

f
. Since

f
is sym-
metric and

f
is antisymmetric, we obtain from (136, 134, 130) the nal

result (33, 34).
References
1. Einstein, A., Straus, E.G.: Ann. Math. 47, 731 (1946)
2. Einstein, A.: Rev. Mod. Phys. 20, 35 (1948)
3. Einstein, A.: Can. J. Math. 2, 120 (1949)
4. Einstein, A., Kaufman, B.: Ann. Math. 62, 128 (1955)
5. Einstein, A.: The Meaning of Relativity, 5th edn revised. Princeton University Press, Princeton (1956)
6. Schrdinger, E.: Proc. R. Ir. Acad. 51A, 163 (1947)
7. Schrdinger, E.: Proc. R. Ir. Acad. 52A, 1 (1948)
8. Schrdinger, E.: SpaceTime Structure, pp. 93, 108, 112. Cambridge Press, London (1950)
9. Callaway, J.: Phys. Rev. 92, 1567 (1953)
10. Infeld, L.: Acta Phys. Pol. X, 284 (1950)
11. Einstein, A., Infeld, L.: Can. J. Math. 1, 209 (1949)
12. Wallace, P.R.: Am. J. Math. 63, 729 (1941)
13. Kur suno glu, B.: Phys. Rev. 88, 1369 (1952)
14. Zeldovich, Ya.B.: Sov. Phys. Uspekhi 11, 381 (1968)
15. Sahni, V., Starobinsky, A.: Int. J. Mod. Phys. D9, 373 (2000) [arXiv:astro-ph/9904398]
16. Peskin, M.E., Schroeder, D.V.: An Introduction to Quantum Field Theory, pp. 790791. Westview
Press, Boulder (1995)
17. Carroll, S.M., Press, W.H., Turner, E.L.: Annu. Rev. Astron. Astrophys. 30, 499 (1992)
18. Borchsenius, K.: Phys. Rev. D 13, 2707 (1976)
19. Borchsenius, K.: Nuovo Cim. 46A, 403 (1978)
20. Astier, P., et.al.: Astro. Astrophys. (2005) [arXiv:astro-ph/0510447]
21. Shifflett, J.A.: (unpublished) [arXiv:gr-qc/0411016]
22. Birrell, N.D., Davies, P.C.: Quantum Fields in Curved Space. pp. 8188. Cambridge University Press,
Cambridge (1982)
23. Tonnelat, M.A.: C. R. Acad. Sci. 239, 231 (1954)
25. Antoci, S.: Gen. Rel. Grav. 23, 47 (1991) [arXiv:gr-qc/0108052]
26. Proca, A.: J. Phys. Radium 7, 347 (1936)
27. Jackson, J.D.: Classical Electrodynamics, 3rd edn. p. 596, 600. Wiley, London (1999)
28. Persides, S., Ioannides, I.: Prog. Theor. Phys. 58, 829 (1977)
29. Persides, S.: Prog. Theor. Phys. 320, 349 (1970)
30. Sakharov, A.D.: Sov. Phys. Dokl. 12, 1040 (1968)
31. Garay, L.L.: Int. J. Mod. Phys. A10, 145 (1995)
32. Padmanabhan, T.: Gen. Rel. Grav. 17, 215 (1985)
33. Padmanabhan, T.: Class Quantum Grav. 19, 3551 (2002)
34. Smolin, L.: Sci. Am. 290, 1 (2004)
35. Einstein, A., Infeld, L., Hoffmann, B.: Ann. Math. 39, 65 (1937)
36. Wallace, P.R.: On the Relativistic equations of motion in electromagnetic theory. Doctoral Thesis,
University of Toronto (1940)
37. Gorbatenko, M.V.: Theor. Math. Phys. 142, 138 (2005)
38. Reissner, H.: Ann. Phys. (Leipzig) 50, 106 (1916)
123
39. Nordstrm, A.: Proc. Kon. Ned. Akad. Wet. 20, 1238 (1918)
40. Shifflett, J.A.: http://www.artsci.wustl.edu/~jashif/MAPLEtxt.html/
42. Papapetrou, A.: Proc. R. Ir. Acad. 52, 69 (1948)
43. Takeno, H., Ikeda, M., Abe, S.: Prog. Theor. Phys. VI, 837 (1951)
44. Chandrasekhar, S.: The Mathematical Theory of Black Holes, p. 317. Oxford University Press,
New York (1992)
45. Baldwin, O.R, Jeffery, G.B: Proc. R. Soc. A 111, 95 (1926)
46. Misner, C.W., Thorne, S., Wheeler, J.A.: Gravitation, pp. 961962. W. H. Freeman and Company,
San Francisco (1973)
47. Stephani, H., Kramer, D., MacCallum, M.A.H., Hoenselaers, C., Herlt, E.: Exact Solutions to Einsteins
Field Equations, 2nd edn, p. 385. Cambridge University Press, Cambridge (2003)
48. Grifths, J.B.: Colliding Plane Waves in General Relativity, pp. 1821, 137145. Oxford University
Press, Oxford (1991)
49. Deif, A.S.: Advanced Matrix Theory For Scientists and Engineers, p. 20, 153, 183, 171. Wiley,
New York (1982)
123

A Modification of Einstein-Schrödinger Theory That Contains Both General Relativity and Electrodynamics

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

A Modification of Einstein-Schrödinger Theory That Contains Both General Relativity and Electrodynamics

Caricato da

Copyright:

Formati disponibili

Gen Relativ Gravit (2008) 40:17451769

is the symmetric metric.

is the nonsymmetric fundamental tensor. The total cosmological con-

and vector potential A

, and it is invariant under a U(1)

to additional elds, such as a hydrodynamic velocity vector

, and excludes the

) is a form of Hermitianized Ricci tensor [1],

into a new connection

has the symmetry

and can be treated as independent

= 0, so we will follow this method.

spin states spin states

are all Hermitian,

) are Hermitian from (16), and g

and L are real from (4, 2, 10).

to a very high precision. The

in terms of the fundamental elds N

) from(5) andthe ordinaryRicci tensor R

, and would reduce to the ordinary Ricci tensor when

for the Chris-

respectively. We use the symbol g

for the symmetric metric (4) whereas Einstein

, the nonsymmetric fundamental tensor. Also, to

instead of the more conventional N

)= 0 we need some initial results. Using (4) and the identities

)= 0 using (10, 26) gives the eld equations,

and the energy-momentum tensor T

hc is the ne structure constant and l

is the Christoffel connection (20),

only differ by terms with

will be the same as in ordinary general relativity, for

= 0 and using (10, 22) gives

will be the same as

= 0 with a Lagrange mulitiplier term

we get the identity

using (8, 49), and dividing this by (n2) gives,

and using (50) gives

is the Christoffel connection (20). The notation ( f

of (52) from the Hermitianized Ricci tensor (11) gives,

is used in the Einstein equations and L

from (40) to become exact in this limit. Let us examine how

as a substitute for (1/4) f

is the wavevector, and

is the polarization. Substituting

from (40), and Maxwells equa-

, the definition of the Weyl tensor C

, and the Einstein equa-

= (Weyl tensor), and

from (3, 12). The ( f

as a substitute for (1/4) f

changes extremely rapidly. Taking the curl of (67), the F

for such waves, and assuming that f

goes according to (67) with

eld would have mass

=0 into(67, 58, 57), andassuminga at backgroundspace gives

could function as a kind of built-in PauliVillars eld.

from (4, 22, 39),

=0 in (48). The derivation for

=0 was rst done [25] by applying an innitesimal coordinate transformation to

= diag(1, 1, 1, 1), and Latin indices run from 1 to 3.

in other contexts) is used instead of g

terms and one of