Fibre Bundles, Connections, General Relativity, and Einstein-Cartan Theory 1110.1018v1

a
r
X
i
v
:
1
1
1
0
.
1
0
1
8
v
1

[
g
r
-
q
c
]

5

O
c
t

2
0
1
1
FIBRE BUNDLES, CONNECTIONS, GENERAL RELATIVITY,
AND EINSTEIN-CARTAN THEORY
M. Socolovsky
Instituto de Ciencias Nucleares, Universidad Nacional Autonoma de Mexico
Circuito Exterior, Ciudad Universitaria, 04510, Mexico D. F., Mexico
Introduction
The main purpose of this article is to present in the most natural way, that is, in the context of the
theory of vector and principal bundles and connections in them, fundamental geometrical concepts related to
general relativity (GR) and one of its extensions, the Einstein-Cartan theory (EC). Central concepts are the
curvature tensor R
, the torsion tensor T
and the non-metricity tensor Q
= D
as properties
of connections in a Riemannian or pseudo-Riemannian manifold, with metric g
and ane connection
.
(D
is the covariant derivative with respect to
.) GR has to do with a metric symmetric connection,

the Levi-Civita connection, that only allows for R
; EC theory involves a metric but not necessarily

symmetric connection, that allows also for T
; while the theory of Weylian manifolds involves a non

necessarily metric (Q
= 0) and non necessarily symmetric (T
= 0) connection. (In units of length [L],

[R
] = [L]
2
, [T
] = [Q
] = [
] = [L]
1
, while [g
] = [L]
0
.)
One of the most beautiful equations of Physics is the equality to zero of the Einstein tensor, that is, the
Einsteins equations in vacuum:
G
= 0,
where
G
= R

1
2
g
R,
with R = g
= g
= g
. G
= 0 is equivalent to Ricci atness:

R
= 0.
This however does not imply vanishing curvature; so, in GR, empty space-time can be curved. Instead, in
EC theory, torsion must be zero in vacuum.
It is remarkable that G
appears naturally when the Bianchi equations for the Levi-Civita connection
are expressed in terms of the Ricci tensor R
and the scalar curvature R. Then, G
is a purely geometric
object.
The use of the tetrads (e
c
) formalism along with their duals, the coframes or anholonomic coordinates
(e
a
), allows us to discover how GR and EC theory have an internal or gauge symmetry (Utiyama, 1956),
implemented by a connection that takes values in the Lie algebra of the Lorentz group L
4
: the spin connection
ab
. The variations of the Einstein-Hilbert action of pure gravity or gravity coupled to Dirac matter with
respect to
ab
and e
c
lead, respectively, to the Cartan and Einstein equations, the former involving torsion
and the spin of matter, and the latter involving curvature and the energy-momentum of matter.
Later, through a shift of the e
c
s one nds the translation gauge potential B
a
; together, B
a
and
bc
,
dene a Poincare connection, extending the symmetry group of GR and EC theory to the semidirect product
P
4
D, where P
4
= T
4
L
4
is the Poincare group, with T
4
the translation group, and D the group of general
coordinate transformations.
1
Finally, in the last section, we discuss the problem of dening a gauge invariant eld strength for the
Maxwell eld coupled to gravity, and the subsisting problem of the U(1)-gauge dependence of torsion in the
solution of the Cartan equation.
CONTENT
1. Connections in smooth real vector bundles
2. Linear connection in a dierentiable manifold M
3. Total covariant derivative of a section in
4. Local expressions for the connection (covariant derivative) and the total covariant derivative
5. Local expressions for the covariant derivatives of vector elds and 1-dierential forms in a manifold
6. Example: Trivial connection in R
n
7. Transformation of the connection coecients in a manifold (local concept). Provisional denition of
tensors in a manifold. Tensors in arbitrary vector bundles.
8. Directional covariant derivative and parallel transport of tensors; geodesics.
9. Curvature and torsion of a connection in
10. Geometric interpretation of curvature and torsion
11. Exterior covariant derivative and curvature 2-form
12. Bianchi equation and Bianchi identities
13. The Levi-Civita connection
14. Physics 1: Equivalence principle in GR
15. Covariant components of the curvature tensor
16. Ricci tensor for the Levi-Civita connection
17. Physics 2: (Local) Einstein equations in empty space-time
18. Ricci (or curvature) scalar (with Levi-Civita connection)
19. Einstein tensor
20. Physics 2
: (Local) Einstein equations in empty space-time

21. Examples in m = D = 1, 2, 3. Generalization to m 4 and Weyl tensor
22. For the Levi-Civita connection, G
= 0
23. Physics 3: (Local) Einstein equations in the presence of matter
2
24. Tensor bundles as associated bundles to the bundle of frames of M
n
, F
M
n
25. Vertical bundle of a principal bre bundle
26. Soldering form on F
M
n
27. Linear connection in a manifold M
n
on F
M
n
28. Tetrads and spin connection
29. Curvature and torsion in terms of spin connection and tetrads. Cartan structure equations;
Bianchi identities
30. Spin connection in non-coordinate basis
31. Locally inertial coordinates
32. Einstein-Cartan equations
33. Lorentz gauge invariance of Einstein and Einstein-Cartan theories
34. Poincare gauge invariance of Einstein and Einstein-Cartan theories
35. Torsion and gauge invariance
Acknowledgements
References
Appendix A
Appendix B
Appendix C
Appendix D
Appendix E
Appendix F
1. Connections in smooth real vector bundles
Let : R
m
E

M
n
be a smooth m dimensional real vector bundle over M
n
M, a dierentiable
manifold of dimension n. Let (TM) denote the sections of the tangent bundle of M and (E) denote the
sections of E. E is an m + n dimensional dierentiable manifold; this can be easily shown from the local
triviality condition.
A connection in is a function
: (TM) (E) (E),
(X, s) (X, s)
X
s
3
which has the following properties:
i)
X+X
s =
X
s +
X
s
ii)
fX
s = f
X
s, where f C
(M, R): smooth real valued functions on M

iii)
X
(s +s
) =
X
s +
X
s
iv)
X
(fs) = X(f)s +f
X
s (Leibnitz rule)
(TM) and (E) are innite dimensional vector spaces over R, but modules over C
(M, R) as a ring,
with dimensions n and m respectively. The Leibnitz rule shows that is not C
(M, R)-linear in the second

entry. As will be shown below, this will be reected in the fact that under a change of local coordinates, the
set of connection coecients (Christoel symbols) is not a tensor.
The value of the connection at (X, s) is called the covariant (or invariant) derivative of s in the direction
of X.
X
s : M E, x
X
s(x) = (x, (
X
s)
x
), with (
X
s)
x
E
x
: the bre in E over x; E
x
is a real m
dimensional vector space.
Notice that we can dene the operator
X
: (E) (E),
X
(s) =
X
s
X
Lin
R
((E)) and obeys the Leibnitz rule.
One summarizes these concepts in the following diagram:
R
m
|
E
s
X
s
R
n
TM
M
M
X
E =
xM
{x} E
x

xM
E
x
.
Note: M is a dierentiable manifold; as such is a topological space. This global structure is dened in
M prior to any connection on .
2. Linear connection in a dierentiable manifold M
A linear connection on M is a connection in its tangent bundle. With E = TM we have:
: (TM) (TM) (TM),
(X, Y ) (X, Y )
X
Y
with
i
)
X+X
(Y ) =
X
Y +
X
Y
ii
)
fX
(Y ) = f
X
Y
4
iii
)
X
(Y +Y
) =
X
Y +
X
Y
iv
)
X
(fY ) = X(f)Y +f
X
Y
We shall denote by Conn() the set of connections in the vector (or principal, where appropiate) bundle
.
Again,
X
Lin
R
((TM)) and obeys the Leibnitz rule.
3. Total covariant derivative of a section in
Let (T
M E) be the set of dierential 1-forms in M with values in E. T
M E is a vector bundle
on M:
R
nm
T
M E M, with T
M E =
xM
T
x
M E
x
.
The section s (T
M E) is dened by
s : (TM) (E), X s(X) :=
X
s.
As for any dierential form on M, s is C
(M, R)-linear i.e. s(fX) = fs(X); however,

(fs) = sdf +fs.
In fact, (fs)(X) =
X
(fs) = X(f)s +f
X
s = fs(X) +sX(f) = fs(X) +sdf(X) = (fs +sdf)(X).
s is called the total covariant derivative of the section s. In detail,
s : M T
M E, x (s)(x) = (x, (s)

x
), (s)
x
=
x
v
x
: T
x
M E
x
, X
x
(s)
x
(X
x
) =
x
(X
x
)v
x
=
x
v
x
where
x
T
x
M, v
x
E
x
and
x
R.
For a linear connection on M,
Y : (TM) (TM), X (Y )(X) =
X
Y.
4. Local expressions for
X
s and s
Let (U
,
i
)
J, i{1,...,m}
be a basis of local sections of E i.e.
i
: U
=
1
(U
), x
i
(x) = (x,
ix
) {x} E
x
, with
i
= Id
U
, and such that if s (E
), then s =
m
i=1
s
i
i
with
s
i
C
(U
, R). Let

x
be a local coordinate basis of (TU
) i.e. if X (TU) then X = X
with X
(U
, R). (If the domains of the local trivializations of the bundle E do not coincide with the
domains of the atlas U of the manifold M, one can always consider their intersections.) Then, locally,
X
s =
X
(s
i
i
) = X
(s
i
i
) = X
((
s
i
)
i
+s
i
i
);
and since
i
(E
) then
i
:=
j
i
j
for a unique set of n m
2
functions
j
i
: U
R, called the Christoel symbols of the connection in the

atlas U. For a linear vector bundle, m = 1, and then there are only n symbols:
j
i

.
5
We then write
X
s = X
((
s
i
)
i
+s
i
j
i
j
) = X
s
j
+s
i
j
i
)
j
= X
j
i
+
j
i
)s
i
j
= X
D
j
i
s
i
j
where we have dened the local covariant derivative operator
D
j
i
:=
j
i
+
j
i
.
(Units: [
j
i
] = [x
]
1
; in natural units [
j
i
] = [mass] if [x
] = [length].)
We can also write
X
s = X
s
j
;
j
with D
s
j
s
j
;
= D
j
i
s
i
= s
j
,
+
j
i
s
i
and s
j
,
=
s
j
.
Notice that the ordinary derivative term s
j
,
is due to the Leibnitz rule.

Locally, we can write
s = dx
s.
In fact, s(X) = (dx
s)X = dx
(X)
s = dx
(X
s = X
dx
s = X
s =
X
s =
X
s. Then
s = dx
j
i
+
j
i
)s
i
j
= dx
s
j
j
+
j
i
s
i
j
) = dx
s
j
j
+dx
j
i
s
i
j
= ds
j
j
+
j
i
s
i
j
where
j
i
is an mm matrix of 1-forms on U
given by
j
i
= dx
j
i
.
We can write
j
i
s
i
j
=
j
i
s
i

j
= (s)
j
j
and then
s = ds
j
j
+ (s)
j
j
= (ds
j
+ (s)
j
)
j
= ((d + )s)
j
j
i.e.
s = (d + )s (T
U E).
Then, locally,
= d + .
Let s
i
;
(x)=0 i.e.
s
i
,
(x) +
i
j
(x)s
j
(x) = 0.
Multiplying by dx
|
x
we obtain
s
i
,
(x)dx
|
x
=
i
j
(x)s
j
(x)dx
|
x
T
x
U.
The r.h.s.
i
j
(x)s
j
(x)dx
|
x
(
||
s)
i
|
x
,
is called the innitesimal parallel transport (transfer) (Schroedinger, 1950) of the section s
i
by the connection
along the 1-form dx
i
|
x
(see section 8), and we see that for a covariantly constant section, it coincides with
the dierential of s
i
at x, ds
i
|
x
. The transfer of the section, proportional to dx
|
x
and to the section itself,
just follows the values of s
i
along dx
(when s
i
;
= 0). When s
i
;
= 0, (
||
s)
i
|
x
still is the parallel transfer of
s
i
through dx
, but it fails to follow the value of the section. (A more detailed discussion can be found in
Cheng, 2010.)
6
5. Local expression for
X
Y
When E = TM,
i
=
and
for a unique set of n

3
smooth functions
: U
R. (
is not necessarily symmetric or antisymmetric

in and .) Then,
X
Y = X
, D
,
D
,
X
Y = (
X
Y )
,
(
X
Y )
= X
)Y
.
The quantity
D
;
D
is the covariant derivative of the local vector eld Y
in the direction of

x
. Then,
(
X
Y )
= X
.
If V
is a local vector eld and A
is a local dierential 1-form in M

n
, then = V
is a scalar
(0-rank tensor) i.e.
= or V
= V
under x
. The covariant derivative of a scalar is

naturally dened as
;
=
,
and the Leibnitz rule is assumed for the covariant derivative of the product of arbitrary tensors T and S:
(TS)
;
= T
;
S +TS
;
. Then,
;
= (V

A
)
,
= (
)A
+V
= V
;
A
+V
A
;
and so
V
A
;
= V
,
A
+V
A
,
V
;
A
= V
,
A
+V
A
,
V
,
A
= V
A
,
= V
A
;
where
A
;
= A
,

is the covariant derivative of the local 1-form A
in the direction of

x
.
6. Example: the trivial connection in R
n
Consider M = R
n
; then with X, Y (TM) we dene
0
X
Y := X(Y
= X
(Y
.
Additivity in both X and Y , and C
(M, R)-linearity in X are trivial; nally,

0
X
(fY ) = X
(fY
=
X(f)Y +fX(Y
= X(f)Y +f
0
X
Y .
In particular,
) =
0
=
0
= 0
and therefore
= 0.
7
Note. We can compare with the Lie derivative L
X
Y :
L
X
Y = [X, Y ] = X(Y
Y (X
= [X, Y ]
=
0
X
Y
0
Y
X;
then
0
X
Y
0
Y
X [X, Y ] = 0.
As we shall see later, this means that
0
is torsion free.
For
0
Y we have:
0
Y = d(Y
.
In fact, (
0
Y )(X) = dx
(Y
(X) = dx
(X)
(Y
= X(x
(Y
= X
(Y
= X(Y
=
0
X
(Y ).
7. Transformation of
under a change of local coordinates (charts) in M

From the denition of the
-functions and using
=

x
=

x
one obtains:
x
(

x
) =
x
(
x
) =
x
(

x
(
x
)

x
+
x
(
x
))
=
x
(
x
2
x
+
x
) =
x
2
x
+
x
=

2
x
+
x
which implies, by the linear independence of the coordinate tangent elds

x
=

2
x
+
x
;
nally, multiplying by the inverse transformation
x
and using
x
one obtains
=
x
+
x
2
x
.
The second term on the r.h.s., which comes from the fact that obeys the Leibnitz rule, shows that the
functions do not transform as tensors.

Provisional denition of a tensor (previous to the existence of any metric in the manifold)
In a chart U
(x
) of M
n
, a tensor with r contravariant and s covariant indices is a set of
n
r+s
functions on U
with values in R,
{T
1...r
1...s
,
k
,
l
= 1, . . . , n, k = 1, . . . , r, l = 1, . . . , s}
such that in a chart U
(x
) which overlaps with U
(x
) becomes the set

T
1...r
1...s
=
x
1
x
1
. . .
x
r
x
r
x
1
x
1
. . .
x
s
x
s
T
1...r
1...s
8
for x
= x
(x) in the overlap U
.
Remark: The concept of covariant and contravariant indices has sense only if there exists a metric in
the manifold.
An r-contravariant and s-covariant tensor can be considered a C
(M, R)-multilinear map from the

tensor product of s factors of (TM) and r factors of (T
M) with values in C
(M, R). On a chart U
,
T = T
1...r
1...s
dx
1
. . . dx
s

x
1
. . .

x
r
: (
s
((TU
)) (
r
(T
)) C
(U
, R),
V
1
1
x
1
. . . V
s
s
x
s
A
11
dx
1
. . . A
rr
dx
r
T
1...r
1...s
dx
1
(V
1
1
x
1
) . . . dx
s
(V
s
s
x
s
(A
11
dx
1
) . . .

x
r
(A
rr
dx
r
) = T
1...r
1...s
V
1
1

1
1
. . . V
s
s

s
s
A
11
1
1
. . . A
rr
r
r
= T
1...r
1...s
V
1
1
. . . V
s
s
A
11
. . . A
rr
.
Well call
r
s
(M) to the C
(M, R)-module of r-contravariant and s-covariant tensors on M

n
. For
example,
1
0
(M) = (TM): vector elds on M;
0
1
(M) = (T
M): dierential 1-forms on M. In general,
r
s
(M) = (T
r
s
M): sections of the bundle of r(s) contravariant (covariant) tensors on M, with T
1
0
M = TM
and T
0
1
M = T
M.
Locally, given a tensor T
1...r
1...s
and a connection
in the manifold M
n
, the covariant derivative of
T
1...r
1...s
in the direction of

x
is given by
D
T
1...r
1...s
T
1...r
1...s;
=

x
T
1...r
1...s
+
1
1
T
12...r
1...s
+. . . +
r
r
T
1...r1r
1...s

1
1
T
1...r
12...s
. . .
s
s
T
1...r
1...s1s
.
It can be veried that T
1...r
1...s;
is an r-contravariant and s + 1-covariant tensor.
Remark: Notice that while the operators
X
send tensors (or sections in general) of a given order to
tensors of the same order, for both covariant and contravariant indices, the operators D
map (r, s)-tensors

into (r, s + 1)-tensors.
Tensors in arbitrary vector bundles
In U
M
n
consider the change of local coordinates and sections:

x
=

x
and
k
= f
1
j
k
j
with x
= x
, x
= x
,
k
=
k
, and
j
=
j
; , = 1, . . . , n; j, k = 1, . . . , m; at each x U
, f
and f
1
take values in GL
m
(R) with
j
= f
k
j
k
. We study the transformation of the Christoel symbols
i
j
of a connection in : R
m
E

M
n
:
i
=
j
i
j
=
x
i
= x
f
1
j
i
j
=
x
f
1
j
i
j
=
x
((

x
f
1
j
i
)
j
+f
1
j
i

x
j
)
=
x
(

x
f
1
l
i
+f
1
j
i
l
j
)
l
) =
j
i
f
1
l
j
l
i.e.
j
i
f
1
l
j
=
x
(

x
f
1
l
i
+f
1
j
i
l
j
);
9
multiplying by f
r
l
and using f
1
l
j
f
r
l
=
r
j
we obtain
r
i
=
x
f
1
j
i
f
r
l
l
j
+
x
f
r
l
(f
1
l
i
).
The homogeneous part in the connection coecients gives the general law for the tensorial transformation
of an object with r(v) contravariant or upper (covariant or lower) internal indices, r, v = 1, . . . , m, and s(t)
contravariant (covariant) space-time (external) indices, s, t = 1, . . . , n:
T
1...sa1...ar
1...tb1...bv
=
x
1
x
1

x
s
x
s
x
1
x
1

x
t
x
t
f
a1
c1
f
ar
cr
f
1
d1
b1
f
1
dv
bv
T
1...sc1...cr
1...td1...dv
.
For example,
T
a
=
x
f
a
b
T
b
, T
ab
=
x
f
1
c
a
f
1
d
b
T
cd
.
8. Directional covariant derivative and parallel transport of tensors; geodesics
If c : (a, b) M
n
, c(), with (a, b) an open interval in R, is a smooth path in M
n
locally given by
c() = (x
1
(), . . . , x
n
()), then the covariant derivative of T
1...r
1...s
along c is the tensor dened by
(
DT
d
)
1...r
1...s
:=
dx
d
D
T
1...r
1...s
.
We have then dened the covariant derivative operator along the path c through
D
d
=
dx
d
D
.
In detail,
(
DT
d
)
1...r
1...s
=
dx
d
(

x
T
1...r
1...s
+
1
1
T
12...r
1...s
+. . .
s
s
T
1...r
1...s1s
)
=
d
d
T
1...r
1...s
+
dx
d

1
1
T
12...r
1...s
+. . .
dx
d

s
s
T
1...r
1...s1s
.
Symbolically,
DT
d
=
d
d
T + c
where denotes the contractions.

For a vector eld V
,
(
DV
d
)
=
d
d
V
+
dx
i.e.
(
DV
d
)|
= (
dV

()
d
+
dx
()
d

()V

())

x
where the dependence of is through c, and for a dierential 1-form A
,
(
DA
d
)
=
dA
d

dx
i.e.
(
DA
d
)|
= (
dA
()
d

dx
()
d

()A
())dx
.
10
If, in particular, V
=
dx
d
: the tangent vector to the curve c at , i.e. V = c, then
(
D c
d
)
=
d
2
x
d
2
+
dx
dx
d
.
Symbolically
DV
d
= (
d
d
+ c)V i.e.
D
d
=
d
d
+ c for vectors, and
DA
d
= (
d
d
c)A i.e.
D
d
=
d
d
c for
1-forms.
A tensor T is said to be parallel transported by the connection from c(
0
) to c(
1
) along the smooth
curve c in M
n
(a <
0
<
1
< b), if
DT
d
= 0 for all [
0
,
1
]
i.e. if
dT()
d
= c
()T
() + c
()T
().
This is a system of n
r+s
ordinary dierential equations of rst order (ODE-1). By general theorems on
ODE-1, if T
0

r
s
(c(
0
)) then there exists and is unique a parallel transported tensor T() along c, in
particular at c(
1
), such that T(
0
) = T
0
.
The parallel transport of T depends on i.e. on and on the path c. There exists a vector space
isomorphism
P
c
:
r
s
(c(
0
))
r
s
(c(
1
)), T
0
P
c
(T
0
) = T
1
with (P
c
)
1
= P
c
1
where c
1
() = c(
1
+
0
).
The equations of parallel transport for vector elds and dierential 1-forms are
(
DV
d
)
= 0
dV
d
=
dx
and
(
DA
d
)
= 0
dA
d
=
dx
respectively.
In particular, a curve c is a geodesic in M
n
with respect to the connection , if its tangent vector c is
parallel transported along c:
(
D c
d
)
= 0
d
2
x
d
2
=
dx
d
dx
d
.
Symbolically,
D c
d
= 0 c = c
2
.
In more detail,
x
() +
() x
() x
() = 0,
which is a system of n ordinary dierential equations of second order (ODE-2) for c() (x
()). Given
(x
0
, v
x0
) TM, there always exists a unique solution to this system of equations in an interval (
0
,
0
+),
> 0 with the initial conditions c(
0
) = x
0
and c(
0
) = v
x0
.
The geodesic equation is invariant under the change a +b, with a, b R, a = 0. (See also section
13.)
11
Notice the arbitrariness of v
x0
at x
0
, and the fact that the whole geodesic is determined in the interval
(
0
,
0
+) (globally) from the initial data.
For the trivial connection in R
n
(section 6),
0
= 0 and then the solutions of the geodesic equation

are straight lines:
x
() = 0 x
() = a
+b
.
9. Curvature and torsion of a connection
Let be a connection on : R
m
E

M
n
. The curvature of is dened as follows:
R : (TM) (TM) (E) (E), (X, Y, s) R(X, Y, s) := ([
X
,
Y
]
[X,Y ]
)(s)
i.e.
R(X, Y, s) =
X
(
Y
(s))
Y
(
X
(s))
[X,Y ]
(s).
Clearly, R(X, Y, s) = R(Y, X, s).
Well show that R is C
(M, R)-linear in its three entries. This will have as a consequence that the set
of local components of R behaves as a tensor.
i) R(fX, Y, s) =
fX
Y
s
Y
fX
s
[fX,Y ]
s = f
X
Y
s
Y
(f
X
s)
f[X,Y ]Y (f)X
s =
f
X
Y
s Y (f)
X
s f
Y
X
s
f[X,Y ]
s +
Y (f)X
s = f
X
Y
s Y (f)
X
s f
Y
X
s f
[X,Y ]
s +
Y (f)
X
s = f(
X
Y

Y
X

[X,Y ]
)(s) = fR(X, Y, s);
ii) R(X, fY, s) = R(fY, X, s) = fR(Y, X, s) = fR(X, Y, s);
iii) R(X, Y, fs) =
X
Y
(fs)
Y
X
(fs)
[X,Y ]
(fs) =
X
(Y (f)s+f
Y
s)
Y
(X(f)s+f
X
s)
[X, Y ](f)sf
[X,Y ]
s = X(Y (f))s+Y (f)
X
s+X(f)
Y
s+f
X
Y
sY (X(f))sX(f)
Y
sY (f)
X
s
f
Y
X
s X(Y (f))s +Y (X(f))s f
[X,Y ]
s = f(
X
Y

Y
X

[X,Y ]
)(s) = fR(X, Y, s).
Locally (in a common chart for and M),
R(
,
j
) = [
](
j
)
[,]
(
j
), but [
] = 0 and
0
s =
0X
s = 0
X
s = 0, then
R(
,
j
) = [
](
j
) =
(
j
))
(
j
)) =
(
i
j
i
)
(
i
j
i
)
=
(
i
j
)
i
+
i
j
l
i
(
i
j
)
i
i
j
l
i
l
= R
k
j
k
with
R
k
j
=
k
j

k
j
+
k
i
i
j
k
i
i
j
.
Then
R(X, Y, s) = X
s
j
R(
,
j
) = X
s
j
R
i
j
i
.
For a linear connection in a manifold,
R
,
with R(X, Y, Z) = X
R(
) = X
.
In particular, R(
, Z) = [
](Z) = Z
or R(
) = [
](
) = R
.
Dening
R : (TM) (TM) End

C
M
((TM)),

R(X, Y ) : (TM) (TM),

R(X, Y )(Z) := R(X, Y, Z)
12
one obtains, in particular,
R(
) = [
],
which is the usual expression of curvature in terms of a commutator of local covariant derivatives.
If in R
we contract with we obtain the antisymmetric tensor

S
= S
,
and if we contract with we obtain the tensor
R
.
In general, R
is non symmetric, not even when
. In this case, however, its antisymmetric part

is the half of S
:
R
{}
=
1
2
(R
) =
1
2
S
.
But the denition of S
does not require a symmetric connection.

In GR, where is the Levi-Civita connection (section 13) uniquely determined by the metric in a
pseudo-riemannian (lorentzian) manifold, it is usual to denote
R
= R
with R R : (TM) (TM) (TM) (TM), (X, Y, Z) R(X, Y, Z) = ([

X
,
Y
]
[X,Y ]
)(Z) =
X
. This denition holds for any connection, like the Weyl connection (non-metric symmetric),
or that corresponding to the Einstein-Cartan theory (metric non-symmetric).
Clearly, R
= R
. Since R(
) = R
= R
, then
< dx
, R(
) >=< dx
, R
>=< dx
> R
= R
where < , > denotes the 1-form-vector contraction, which is independent of the metric.
(R : (TU) (TU) (TU) (TU), (
) R(
).)
For a symmetric connection,
(see section 13),

R
. ()
The torsion T of a linear connection on a manifold M
n
is dened as follows:
T : (TM) (TM) (TM), (X, Y ) T(X, Y ) :
X
Y
Y
X [X, Y ].
It holds:
i) T(X, Y ) = T(Y, X)
ii) T
0
= 0 for the trivial connection
0
in R
n
.
iii) T is C
(M, R)-linear in X and Y i.e. T(fX, Y ) = T(X, fY ) = fT(X, Y ).

13
iv) Locally, in a chart U
(x
), J,
T(X, Y ) = (T(X, Y ))
= ((
X
Y )
(
Y
X)
[X, Y ]
)

x
= X
2T
with
T
=
1
2
(
) = T
[]
.
If X
i.e. X =
=

x
, and Y
i.e. Y =
=

x
, then
T(

x
,

x
) = (
)

x
= 2T
i.e. T
=
1
2
(T(

x
,

x
))
.
A straightforward calculation leads to:
[D
, D
]V
= D
(D
) D
(D
) = (V
;
)
;
(V

;
)
;
= R
2T
;
.
If is a scalar, then [D
, D
]() = D
) D
) = D
(
;
) D
(
;
) =
;
+
;
= (
i.e.
[D
, D
]() = 2T
.
So, [D
, D
]() = 0 if T
= 0.
Then, with = V
and using the Leibnitz rule, for a covariant vector (1-form) one obtains
[D
, D
]W
= R
2T
W
;
.
The generalization for a tensor T
1...r
1...s
is
[D
, D
](T
1...r
1...s
) = R
1
T
2...r
1...s
+ +R
r
T
1...r1
1...s
R
1
T
1...r
2...s
R
s
T
1...r
1...s1
2T
T
1...r
1...s;
.
For a symmetric connection,
and therefore
T
= 0
in all charts. I.e. a symmetric connection (like the Levi-Civita connection (section 13)) is torsion free. From
the transformation of the
s, it is clear that T
is a tensor. In particular one has

[D
, D
]V
= R
.
The modied torsion tensor is dened as
= T
1
n 1
(
)
where T
= T
is the torsion vector (in fact, it is a 1-form).

T
is traceless i.e.

T
= 0.
10. Geometric interpretation of curvature and torsion
Curvature
14
Consider the innitesimal parallelogram pqrs in M
n
with coordinates x
, x
, x
, x
respectively, with |
|, |
| 1. Let c and c
, not necessarily part of geodesics, be curves which join p with

r through q and s respectively, and (locally ) be an arbitrary connection in M
n
. Let V
p
T
p
M; its
variation from p to q through c is obtained from the formula of the covariant derivative of a vector along
a curve (section 8), which implies DV
= d(
DV
d
)
= dV
+ dx
; if the transport is parallel then

DV
= 0 i.e. dV
= dx
. So through c, dV
|
c
= V
q
|
c
V
p
and with dx
one obtains
V
q
|
c
= V
(p)V
p
.
Then,
V
r
|
c
= V
q
|
c
(q)V
q
|
c
= V
(p)V
(q)(V

p

(p)V
p
)
= V
(p)V
(p) +
(p)
)(V

p

(p)V

p
)
= V
p
V
(p)
(p)
(p)
+V
(p)
= V
p
V
(p)
(p)
p
(
(p)
(p)
(p))
,
where we have neglected terms of order higher than two in
s and
s; to obtain V
r
|
c
we simply change
which is equivalent to the change :
V
r
|
c
= V
p
V
(p)
(p)
p
(
(p)
(p)
(p))
.
Then the dierence of the two parallel transports is
V
r
|
c
V
r
|
c
= (
(p)
(p) +
(p)
(p)
(p)
(p))V
= R
(p)V
(p)V
p
A
where A
is the innitesimal area
enclosed by the curves c and c
. Clearly,
V
r
|
c
= V
r
|
c
if and only if R
(p) = 0.
Then, the curvature tensor measures the dierence between the parallel transport of a vector through the
paths c and c
, where c (c
) is a loop.
V
r
|
c
V
r
|
c
amounts to a rotation, since norms of vectors do not change by parallel transport induced
by metric connections (appendix B); then one says that curvature is the rotational part of the connection.
When parallel transport is independent of the path, that is, for a vanishing curvature, the connection
is said to be integrable (or at).
Torsion
As before, consider the points p, q and s with coordinates x
, x
and x
, respectively. Consider
the innitesimal vectors at p,
|
p
and
|
p
(
p
=
p
=
); regarded as innitesimal displacements

(translations) in M
n
, they respectively dene the points q and s. Make the parallel transport of
|
p
along
p
: we obtain the vector at s, V
s
=
(p)
p
; so the total displacement vector from p to r is
p
+
(p)
p
;
similarly, making the parallel displacement of

x
|
p
along
one obtains the vector at q, V
q
=
(p)
p
; and the total displacement vector from p to r is
p
+
(p)
p
.
The dierence between the two vectors is
(p) +
(p) =
p
(
(p)
(p)) = 2T
(p)
p
.
15
So, the torsion measures the failure of the closure of the parallelogram made of the innitesimal displacement
vectors and their parallel transports.
Since the last expression is a translation, one says that torsion is the translational part of the connection.
11. Exterior covariant derivative and curvature 2-form
Up to here we have considered the vector bundle : R
m
E

M
n
. Now we shall consider the vector
bundle
k
whose sections are the k-dierential forms s on M
n
with values in E:
R
n!
k!(nk)!
m
k
T
M E

k
M
with
k
T
ME =
xM

k
T
x
ME
x
; clearly,
0
= . s (
k
T
ME) with (s)(x) = (x, (s)

x
),
( s)
x

k
T
x
M E
x
; if, as before, {
i
}
m
i=1
is a basis of sections of E in U
M, x U
, and {x
}
n
=1
are local coordinates on U
, then
{dx
1
|
x
. . . dx
|
x
ix
, i = 1, . . . , m, n
k
> . . . >
1
1}
is a basis of
k
T
x
M E
x
. So,
( s)
x
=
m
i=1
n
k
>...>11
t
i
1...
k
dx
1
|
x
. . . dx

ix
,
t
i
1...
k
R.
We dene the set of total exterior covariant derivative operators
{d
0
, d
1
, . . . , d
n1
}, d
k
: (
k
T
M E) (
k+1
T
M E), k = 0, . . . , n 1,
as the R-linear extension of
d
k
( s) = (d
k
) s + s
with
(d
k
) s + s : M
k+1
T
M E, x (x, (d
k
)
x
s
x
+
x
(s)
x
).
For k = 0, d
0
= : (E) (T
M E).
Let us study the composition d
1
d
0
. If s (E), then
d
1
d
0
(s) = d
1
(s) R(s) (
2
T
M E),
i.e.
R(s) : (TM)(TM) (E), (X, Y ) R(s)(X, Y ) : M E, x R(s)(X, Y )(x) = (x, R(s)
x
(X
x
, Y
x
)).
R(s), also denoted by
2
(s), is called the second total covariant derivative of the section s. (In general,
d
k+1
d
k
= 0; compare with d
k+1
d
k
= 0 in De Rham theory.)
Locally,
R(s) =
2
(s) = d
1
((s)) = d
1
(dx
(s)) = d
1
(dx
((
s
j
)
j
+
j
i
s
i
j
))
= d
1
(dx
) ((
s
j
)
j
+
j
i
s
i
j
) +dx
((s
j
)
j
+
j
i
s
i
j
)
= dx
((
s
j
x
j
) +(
j
i
s
i
j
)) = dx
(d(
s
j
x
)
j
+
j
i

s
i
x
j
+d(
j
i
s
i
)
j
+
j
k
k
i
s
i
j
)
16
= dx
(dx
(
s
j
x
)
j
+dx
(
j
i
s
i
)
j
+dx
j
i
s
i
x
j
+dx
j
k
k
i
s
i
j
)
= ds
i

j
i

j
+dx
dx
(
j
i
s
i
)
j
+
k
i

j
k
s
i
j
= ds
i

j
i

j
+dx
dx
(
j
i
)s
i
j
+dx
dx
j
i
s
i
x

j
+
k
i

j
k
s
i
j
= (dx
d(
j
i
)s
i
+
k
i

j
k
s
i
)
j
= (d(
j
i
) dx
s
i
+s
i
k
i

j
k
)
j
= (d
j
i
+
j
k

k
i
)s
i
j
i.e.
R(s) =
2
(s) = R
i
j
s
j
i
where (R
i
j
) is the mm matrix with entries in the dierential 2-forms on U M i.e. (R
i
j
) (
2
T
U)(m)
(or R
i
j

2
DR
(U)) given by
R
i
j
= d
i
j
+
i
k

k
j
.
(Symbolically, R = d + = (d + ).)
From the last expression for R
i
j
it is clear that R(s) is closely related to the curvature tensor R. Let us
see this relation:
d
i
j
= d(
i
j
dx
) =
i
j
dx
dx
=
1
2
(
i
j
i
j
)dx
dx
i
k

k
j
=
i
k
dx

k
j
dx
=
i
k
k
j
dx
dx
=
1
2
(
i
k
k
j

i
k
k
j
)dx
dx
,
then
R
i
j
=
1
2
R
i
j
dx
dx
.
Also,
R(s)(X, Y ) = R
i
j
s
j
i
(X, Y ) =
1
2
R
i
j
dx
dx
(X, Y )s
j
i
=
1
2
R
i
j
1
2
(dx
dx
dx
dx
)(X, Y )s
j
i
=
1
2
R
i
j
1
2
(X
)s
j
i
=
1
2
R
i
j
X
s
j
i
i.e.
R(X, Y, s) = 2R(s)(X, Y ).
The matrix (R
i
j
) is called the curvature 2-form matrix.
12. Bianchi identities
For each pair i, j in {1, . . . , m}, R
i
j
is a local 2-form. Its exterior derivative is a local 3-form. Since
d
2
= 0, we have dR
i
j
= d(
i
k
d
k
j
) = d
i
k

k
j

i
k
d
k
j
, then dR
i
j
+
i
k
d
k
j
d
i
k

k
j
= 0 or
dR
i
j
+
i
k
(R
i
j

k
i

i
j
) (R
i
k
i
l

l
k
)
k
j
= 0. I.e.
dR
i
j
+
i
k
R
k
j
R
i
k

k
j
= 0.
These are the Bianchi equations. Simbolically, dR+ RR = 0.
The l.h.s. is an m m matrix of 3-forms. So, in terms of the curvature form and the connection
coecients one has m
2
equations. However, when the R
i
j
s are written in terms of the
i
j
s one obtains m
2
identities: d(d
i
j
+
i
k

k
j
) +(
i
k
(d
k
j
+
k
l

l
j
) (d
i
k
+
i
l

l
k
)
k
j
= (d
i
k
)
k
j

i
k
d
k
j
+
i
k
d
k
j
+
i
k

k
l

l
j
(d
i
k
)
k
j

i
l

l
k

k
j
= 0 identically.
17
From the explicit expression of R
i
j
and
k
l
in terms of local coordinates, in the Bianchi equations
one has: dR
i
j
=
1
2
d(R
i
j
dx
dx
) =
1
2
R
i
j,
dx
dx
dx
,
i
k
R
k
j
=
i
k
dx
1
2
R
k
j
dx
dx
=
1
2
i
k
R
k
j
dx
dx
dx
, R
i
k

k
j
=
1
2
R
i
k
dx
dx

k
j
dx
=
1
2
R
i
k
k
j
dx
dx
dx
; then
1
2
(R
i
j,
+
i
k
R
k
j
R
i
k
k
j
)dx
dx
dx
= 0
i.e.
R
i
j,
+
i
k
R
k
j
R
i
k
k
j
= 0.
For a linear connection in M
n
, with i, j, k = , , ,
R
,
+
= 0.
In GR, R
= R
, then the Bianchi equations are:

R
,
+
= 0.
Note 1. In section 22, well see that in the case E = TM and the connection is that of Levi-Civita
(section 13), when the Bianchi equations are written in terms of the Ricci tensor (see section 16) and the
scalar curvature (section 18), we obtain the vanishing of the covariant divergence of the Einsteins tensor
(see section 19): G
= 0.
Note 2. In electromagnetism, dF = 0 in terms of the curvature tensor F (eld strength), amounts to
the homogeneous Maxwell equations. Instead, if F = dA is used, we obtain an identity.
Remark: Up to here, all the results have been independent of the existence of a metric g
in the
manifold M
n
i.e. of a non degenerate symmetric scalar product at each tangent space T
x
M
n
. This metric
is introduced in the next section.
13. The Levi-Civita connection
In Appendix A we shall prove the Fundamental Theorem of Riemannian or Pseudo-Riemannian Ge-
ometry, which states that in a riemannian or pseudo-riemannian manifold (M
n
, g
) there exists a unique

symmetric and metric linear connection, the Levi-Civita connection, given by
=
1
2
g
) = g
with
=
1
2
(
).
Then g
= g
.
It holds:
i) D
= g
;
= 0 (and also D
= g
; = 0). A consequence of this is that for any smooth path

c : (a, b) M
n
the metric tensor g
is parallel transported along c:

(
Dg
d
)
=
dx
d
D
= 0,
18
where
dx
d
= c
. Then it follows that the scalar product of two parallel transported vectors along c by the
Levi-Civita connection is also parallel transported i.e. covariantly constant:
D
d
(g
) = (
Dg
d
)
+ g
(
DV
d
)
+g
(
DW
d
)
= 0.
In particular, if V = W = c, then g
( c, c) || c||
2
remains constant by parallel transport; if
|| c||
2
> 0,=0,<0 the geodesic is respectively called timelike, null or lightlike, and spacelike. Since
= g
,
then
;
= 0.
ii)
D
= D
(g
) = g
i.e. the covariant derivative commutes with the raising or lowering of the indices.
It can be shown that if c : (a, b) M
n
is a smooth path that extremizes the proper time (or path
length) =
_
c
df
1/2
, with f = g
dx
d
dx
d
, then c is a geodesic of the Levi-Civita connection. Also, a
change of parameter = () preserves the form of the geodesic equation if and only if is an
ane transformation, i.e.
= a +b,
where a, b R and a = 0. For an arbitrary transformation one obtains
d
2
x
d
2
+
dx
d
dx
d
=
d
2
d
2
(
d
d
)
2
dx
d
.
This means that the derivation of the geodesic equation in section 8, forces the parameter to be an ane
parameter i.e. a parameter linearly related, up to an additive constant, to the proper time .
It is important to mention that the fact that the connection coecients
depend on the metric

function g
, is the usual argument in the literature for denying to G.R. the character of a gauge theory of
gravity. More on this below.
14. Physics 1: Equivalence principle in GR
Massive free point particles move along timelike geodesics. Massless free point particles move along
lightlike geodesics; in this case can not be the proper time since (d)
2
= 0, (Dirac, 1975).
15. Covariant components of the curvature tensor
Starting from the expression () in section 9, a long but straightforward calculation leads to the result,
for the Levi-Civita connection,
R
= g
=
1
2
(
) +g
).
Clearly, R
is a covariant 4-rank tensor. It has n

4
components (e.g. if n = 4, 4
4
= 256).
Algebraic properties of R
i) R
=
1
2
(
) +g
) = R
(symmetry under the interchange between the rst pair of indices with the second pair of indices i.e.
R
AB
= R
BA
with A = and B = ).
19
ii) R
= R
, R
= R
, then R
= R
which can be summarized as

R
= R
= R
.
iii) R
+R
+R
= 0 (cyclicity).
If one denes
A
:= R
+R
+R
,
it can be proved that it is totally antisymmetric in its four indices. Since A
= 0, this imposes
_
m
4
_
=
m!
4!(m4)!
conditions on R
. (Number of ways one can take four distinct elements among m; obviously it
must be m 4.)
Let us determine the number of algebraically independent components of R
. Let S
ab
and A
ab
be respectively a symmetric and antisymmetric tensor in m dimensions. The corresponding number of
independent components are N(S
ab
; m) =
m(m+1)
2
and N(A
ab
; m) =
m(m1)
2
. So, we have
m 1 2 3 4 5 6 . . . 10 . . .
N(S
ab
; m) 1 3 6 10 15 21 . . . 55 . . .
N(A
ab
; m) 0 1 3 6 10 15 . . . 45 . . .
Notice that
N(A
ab
;m)
N(S
ab
;m)
1
as m . If we write R
R
AB
, since under or R
is antisymmetric, each index A or B contributes with

m(m1)
2
independent components; but now one has
a two-index symmetric matrix R
AB
with A, B {1, . . . ,
m(m1)
2
}, which gives
1
2
m(m1)
2
(
m(m1)
2
+ 1) =
1
8
m(m1)(m(m1) +2) =
1
8
m(m1)(m
2
m+2) independent components for R
. But iii) and then

the antisymmetry of A
imposes
m!
4!(m4)!
=
m(m1)(m2)(m3)
4!
conditions. Then,
N(R
; m) =
1
8
m(m1)(m
2
m+ 2)
1
4!
m(m1)(m2)(m3) =
m
2
(m
2
1)
12
.
So, we have
m 1 2 3 4 5 6 . . . 10 11 . . . 26 . . .
N(R
; m) 0 1 6 20 50 105 . . . 825 1210 . . . 38025 . . .

16. Ricci tensor for the Levi-Civita connection
Dene the covariant 2-tensor
R
:= g
.
We contracted indices 1 and 3; contracting 1-2 and 3-4 gives zero; contracting 1-4, 2-3 and 2-4 gives R
:
g
= g
= R
, g
= g
= R
, g
= g
= R
. So, up to
a sign, the Ricci tensor is uniquely dened from R
and g
.
R
is symmetric: R
= g
= g
= g
= R
. Then,
N(R
; m) =
m(m+ 1)
2
.
We have
m 1 2 3 4 5 6 . . . 10 11 . . .
N(R
; m) 1 3 6 10 15 21 . . . 55 66 . . .
We can write
R
= g
= R
=< dx
, R(
) >=< dx
, R(
) > .
20
17. Physics 2: Einsteins hypotesis: (Local) equations for empty (vacuum, without
matter, vanishing cosmological constant , only g
) space-time
In all charts of the manifold i.e. in all reference frames:
R
= 0,
, = 1, . . . , m or 0, . . . , m 1. R
(U
; R), (Dirac, 1975). These are equations at each chart or

reference system (U
, x
) of M
m
. The number of algebraically independent components of R
equals the
number of independent components of g
. R
= 0 does not imply R
= 0. In other words, in GR
empty space-time can be curved.
18. Ricci (or curvature) scalar (with Levi-Civita connection)
R := g
.
R C
(U
; R).
Then, in empty space-time,
R = 0.
19. The Einstein tensor is dened by
G
:= R

1
2
g
R.
G
(U
; R). Also, it is symmetric (10 algebraically independent components in 4 dimensions).

Proposition (Mathematics)
G
= 0 R
= 0
for m = D = 2, D = d + 1 (space-time dimension).
Proof.
) G
= 0 R
=
1
2
g
R R
= R =
1
2
g
R =
D
2
R (
D
2
1)R = 0. If
D
2
= 1 i.e. D = 2, then
R = 0 and then R
= G
= 0. In particular this holds for D = 3 + 1 = 4.

) R
= 0 R
= R = 0 G
= 0. This holds for all Ds=ms.

20. Physics 2: (Local) Einstein equations in empty space-time
G
= 0
for all space-time dimensions D = m = d + 1. (For D = 4 these are ten equations.)
21. Examples in m = D = 1, 2, 3. Generalization to m 4 and Weyl tensor
D = 1. N(R
; 1) = 0; then R
1111
(or R
0000
)=0. This reects the fact that the curvature tensor
represents intrinsic properties of the space in question, and not how the space (in this case a line, straight
or curved) is embedded in a higher dimensional space. Let g
11
(x) g(x) be the metric tensor in D = 1.
21
If x x
= x
(x) then g
(x
) = (
dx
dx
)
2
g(x). Choose x
(x) =
_
x
dy
_
|g(y)|; then (
dx
dx
)
2
=
1
|g(x)|
and so
g
(x
) =
g(x)
|g(x)|
which equals +1 (-1) if g(x) > 0 (< 0). From the constancy of g(x), (x) =
1
2g(x)
d
dx
g(x) = 0.
D = 2. N(R
; 2) = 1. By antisymmetry in and , the only possibilities for a non-vanishing

R
are R
1010
, R
1001
, R
0110
and R
0101
. We choose R
0101
and it is easily veried that the unique R
which satises the algebraic properties of section 15 and gives R

0110
= R
1010
= R
1001
= R
0101
is
R
= (g
)
R
0101
g
,
with g = det
_
g
00
g
01
g
10
g
11
_
= g
00
g
11
g
01
g
10
= g
00
g
11
g
2
01
. In the presence of matter, Einsteins equations
are
G
= const. T
where T
is the energy-momentum tensor of matter. Then, for D = 2 = 1 + 1,

T
= 0.
This means that in D = 2, the unique solutions to Einsteins equations are those corresponding to the
vacua i.e. T
= 0.
D = 3. Since N(R
; 3) = N(R
; 3) = 3, the curvature tensor can be expressed in terms of g
and
R
; the most general form satisfying the symmetry properties of section 15 is

R
= A(g
) +B(g
+g
)
with A and B numerical constants. From the denition of the Ricci tensor, R
= g
= (2A +
BR)g
+BR
, where R is the Ricci scalar; then B = 1 and A =

1
2
R. Therefore,
R
=
R
2
(g
) +g
+g
.
D = m 4. In all these cases
N(R
; m) N(R
; m) := N(C
; m) =
m(m + 1)(m+ 2)(m3)
12
> 0
where C
, the Weyl tensor, has the same algebraic properties as R
but cant be obtained from g
and R
. One writes
R
= C(g
) +D(g
+g
) +C
with all traces of C
vanishing i.e.
C
= 0.
Then
R
= g
= C(mg
) +D(mR
2R
+g
R) = g
((m1)C +DR) + (m2)DR
.
Then D =
1
m2
and C =
R
(m1)(m2)
. Therefore,
R
=
R
(m1)(m2)
(g
) +
1
m2
(g
+g
) +C
.
22
The Weyl tensor in terms of curvature and metric is given by
C
= R
+
2R
(m1)(m2)
g
[
g
]
2
m2
(g
[
R
]
g
[
R
]
).
One has the table
m 4 5 6
N(C
; m) 10 35 84
Clearly C
= 0 for m = 2, 3.
22. For the Levi-Civita connection, G
= 0
From the general expression for the covariant derivative of a tensor T
1...r
1...r
in section 7, the relation
between the covariant and the ordinary derivatives of R
is given by
R
,
= R
.
Then, the last form of the Bianchi equations in section 12 becomes
R
;
+R
;
+ R
;
= 0.
Since the covariant derivative commutes with the lowering of indices, contracting with g
we obtain
R
;
+R
;
+R
;
= 0.
Contracting and with g
,
R
;
+R
;
+R
;
= 0,
and we obtain, in terms of the Ricci and curvature tensors,
R
;
+R
;
+R
;
= 0.
Finally, contracting and with g
, we obtain, in terms of the Ricci tensor and the scalar curvature,

(R
1
2
R);
= 0
i.e.
G
= 0.
So, the Einstein tensor, apparently arbitrarily dened in section 19, appears naturally when the Bianchi
equations for the Levi-Civita connection are expressed in terms of the Ricci tensor and the scalar curvature,
and some contractions are done. G
is, then, a purely geometrical object.

23. Physics 3: (Local) Einstein equations in the presence of matter
G
=
8G
c
4
T
,
(Einstein, 1956), where G is the Newton gravitational constant, c is the velocity of light in the vacuum,
and T
is the energy momentum tensor of matter: all other elds than g
. Clearly, T
is symmetric
(T
= T
) and covariantly conserved:

T
; = 0.
23
Units: [G
] = [R
] = [R
] = [L]
2
,
[G]
[c
4
]
=
[t]
2
[M][L]
=
1
[force]
; then [T
] =
[M]
[L][t]
= [energy density],
where [L], [M], and [t] denote the units of length, mass, and time, respectively.
24. Tensor bundles as associated bundles to the bundle of frames of M
n
T
r
s
M
n
is the total space of the (n + n
r+s
)-dimensional real vector bundle of r-contravariant and s-
covariant tensors on M
n
, with bre R
n
r+s
= {
i1...ir
j1...js
R, i
k
, j
l
{1, . . . , n}, k = 1, . . . , r, l = 1, . . . , s}
{
}. The bundle of frames of M

n
, F
M
n, is the principal bundle with structure group GL
n
(R) (the bre of
the bundle) on M
n
(the base space), and with total space FM
n
consisting of the set of all ordered basis of
the tangent space at each point of M
n
, namely
FM
n
=
xM
n
{r
x
(v
1x
, . . . , v
nx
), {v
kx
}
n
k=1
: basis of T
x
M
n
} =
xM
n{x} {(v
1x
, . . . , v
nx
)}
xM
n
(FM
n
)
x
,
where (FM
n
)
x
is the bre over x, with dim
R
(FM
n
)
x
= n
2
. The bundles of orthogonal frames, Lorentz
frames, restricted Lorentz frames, etc. of M
n
, are obtained by reducing the group GL
n
(R) respectively to
O(n), O(n1, 1), SO
0
(n1, 1), etc. If x U
U, then v
kx
=
n
=1
v
k
(x)

x
|
x
; also, dim
R
FM
n
= n+n
2
.
The n + n
2
local coordinates on F
U
is the set (x
, X
) with x
(x, r
x
) = x
(x) and X
(x, r
x
) = v
x
,
, , {1, . . . , n}.
One has:
n 1 2 3 4 5 . . . 10 . . .
dim
R
FM
n
2 6 12 20 30 . . . 110 . . .
The bundle structure of F
M
n is represented by
F
M
n : GL
n
(R) FM
n
F
M
n
where
F
is the projection
F
(x, (v
1x
, . . . , v
nx
)) = x and GL
n
(R) FM
n
represents the right action of
GL
n
(R) on FM
n
given by
FM
n
GL
n
(R)

FM
n
, ((v
1x
, . . . , v
nx
), a) (v
1x
a
1
1
+ +v
nx
a
n
1
, . . . , v
1x
a
1
n
+ +v
nx
a
n
n
)
(v
1x
, . . . , v
nx
)a.
The left action of GL
n
(R) on R
n
r+s
, given by
GL
n
(R) R
n
r+s
R
n
r+s
, (a,
) (a
)
i1...ir
j1...js
= a
i1
k1
. . . a
ir
kr
a
1
l1
j1
. . . a
1
ls
js
k1...kr
l1...ls
,
induces the associated bundle FM
n
GLn(R)
R
n
r+s
which turns out to be isomorphic (through , see below)
to T
r
s
M
n
. One has the following commutative diagram:
FM
n
GLn(R)
R
n
r+s
T
r
s
M
n
F

r
s
M
n
Id
M
n
M
n
where:
([((v
1x
, . . . , v
nx
),
)]) =
n
i
k
,j
l
=1
i1...ir
j1...js
v
i1x
. . . v
irx
w
j1
x
. . . w
js
x
,
with [((v
1x
, . . . , v
nx
),
)] = {((v
1x
, . . . , v
nx
)a, a
1
)}
aGLn(R)
; {w
1
x
, . . . , w
n
x
} is the dual basis of
24
{v
1x
, . . . , v
nx
} i.e. w
i
x
(v
jx
) =
i
j
; and
F
([((v
1x
, . . . , v
nx
),
)]) =
F
(x, (v
1x
, . . . , v
nx
)) = x. Using the
fact that the dual basis vectors w
j
x
transform with a
1
, it is easily veried that is well dened i.e. it is
independent of the representative element of the class [((v
1x
, . . . , v
nx
),
)].
A section of the bundle of frames of M
n
i.e. a smooth function s : M
n
FM
n
with
F
s = Id
M
n,
trivializes F
M
n and therefore all the tensor bundles associated with it (in particular the tangent bundle
of M
n
). The same occurs for any of the reductions of F
M
n (bundle of Lorentz frames, restricted Lorentz
frames, etc.).
25. Vertical bundle of a principal bre bundle
Let be a principal bre bundle (p.f.b.), = (P
r+s
, B
s
, , G
r
, , U) : G
r
P
r+s

B
s
, where
B
s
B (base space) and P
r+s
P (total space) are dierentiable manifolds of dimensions s and r + s
respectively, G
r
G is an r-dimensional Lie group with right action on P, P G P, (p, g) pg, and U
is a system of local trivializations
1
(U
)

U
G with
1

= .
For each p P there exists a canonical vector space isomorphism
p
between G = Lie(G) : the Lie
algebra of G, and V
p
= T
p
P
(p)
: the tangent space to the bre over (p) at p, the vertical space at p:
p
: G V
p
, A
p
(A) A
p
,
with
A
p
: C
(P, R) R, f
d
dt
f(pe
tA
)|
t=0
.
We used the fact that T
p
P
(p)
T
p
P; if A
i
, i = 1, . . . , r is a basis of G, then
p
(A
i
) is a basis of V
p
; in
general, neither A
i
nor
p
(A
i
) are canonical basis.
Given p, p
P, since
p
: G V
p
and
p
: G V
p
are isomorphisms, there is a canonical vector space
isomorphism (absolute teleparallelism) between V
p
and V
p
, for all p, p
P:
V
p
p

1
p
V
p
.
Remark: This result, namely, the existence of
p
at each p P, is independent of any connection.
This implies the triviality of the vertical bundle V
of the p.f.b. :
V
: R
r
V
2r+s
V
P,
with V
2r+s
=
pP
V
p
=
pP
{p} V
p
and
V
(p, v
p
) = p.
In fact, V
admits r independent global sections

i
: P V
2r+s
,
i
(p) = (p,
p
(A
i
)); then there is the
following vector bundle isomorphism:
R
r
R
r
| |
V
2r+s

P R
r
V

1
P
Id
P
with (p, v
p
) = (p,
r
i=1
p
(A
i
)) = (p, (
1
, . . . ,
r
)). is not canonical since it depends on the basis A
i
of G.
25
Let be a connection on , i.e. (T
P G) with : P T
P G, p (p) = (p,
p
),
p
: T
p
P G, v
p

p
(v
p
) =
1
p
(ver(v
p
)). Since ker(
p
) = H
s
p
H
p
: the horizontal vector space at p,
then
p
is |H
p
| = 1. However,
p
|
Vp
: V
p
G,
p
|
Vp
(v
p
) =
1
p
(v
p
) is a vector space isomorphism i.e.
p
|
Vp
=
1
p
.
In other words, if is a connection on , then at each p P, gives an inverse of
p
. Therefore, for the
isomorphism between V
p
and V
p
, one has
V
p
p
|V
p
1
p|Vp
V
p
.
In particular, we are interested in the case P = F
M
n, the frame bundle of a dierentiable manifold M
n
,
where (P, B, , G, , U) = (FM
n
, M
n
,
F
, GL
n
(R), , U); p = (x, r
x
) FM
n
, and r
x
= (v
1x
, . . . , v
nx
). Its
vertical bundle is isomorphic to the product bundle FM
n
R
n
2
:
R
n
2
R
n
2
| |
V
F
M
n
c
FM
n
R
n
2
V

1
FM
n
Id
FM
n
where
c
is the canonical isomorphism determined by the canonical basis of gl
n
(R) = R(n) given by the n
2
matrices (A
ij
)
kl
=
ik
jl
. The similar result holds for the reductions of GL
n
(R) to O(n), SO
0
(n1, 1), etc.
mentioned in section 24.
In particular, for the case n = 4 and G = SO
0
(3, 1), with dim
R
(SO
0
(3, 1)) = dim
R
(so(3, 1)) =
dim
R
(o(3, 1)) = 6, case relevant in GR, FM
4
is the bundle of Lorentz frames F
L
M
4
, and we have the
vector bundle isomorphism
R
6
R
6
| |
V
F
L
M
4
L
c
F
L
M
4
R
6
V
| |
1
F
L
M
4
Id
F
L
M
4
with dim
R
(F
L
M
4
) = 4 + 6 = 10 and dim
R
(V
F
L
M
4
) = 16. In this case, the canonical basis of so(3, 1) (or of
o(3, 1)) is the set of matrices
{
_
_
_
0 0 0 0
0 0 0 0
0 0 0 1
0 0 1 0
_
_
_,
_
_
_
0 0 0 0
0 0 0 1
0 0 0 0
0 1 0 0
_
_
_,
_
_
_
0 0 0 0
0 0 1 0
0 1 0 0
0 0 0 0
_
_
_,
_
_
_
0 1 0 0
1 0 0 0
0 0 0 0
0 0 0 0
_
_
_,
_
_
_
0 0 1 0
0 0 0 0
1 0 0 0
0 0 0 0
_
_
_,
_
_
_
0 0 0 1
0 0 0 0
0 0 0 0
1 0 0 0
_
_
_}, respectively {l
23
a
1
, l
31
a
2
, l
12
a
3
, l
01
b
1
, l
02
b
2
, l
03
b
3
}, where the rst three
matrices generate rotations around the axis x, y, and z, and the second three matrices generate boosts along
the same axis, respectively. The derivation of the canonical basis is as follows: one starts from the denition
of the Lorentz transformations :
L
:=
T
L
, with
L
= (
00
,
11
,
22
,
33
) = (+1, 1, 1, 1)
(or (1, +1, +1, +1)) and
ab
= 0 if a = b; if () is a smooth path through the identity (0) = I,
the corresponding tangent vector at I,

(0) = L, obeys the equation L
T
= L. The generators a
i
and b
i
obey [a
i
, a
j
] =
ijk
a
k
, [b
i
, b
j
] =
ijk
a
k
, [a
i
, b
j
] =
ijk
b
k
. If l =
3
i=1
(
i
b
i
+
i
a
i
) and l
=
26
3
i=1
(
i
b
i
+
i
a
i
) are in o(3, 1) (
s,
s R), then is Lie bracket [l, l
] = l
3
i=1
(
i
b
i
+
i
a
i
) o(3, 1)
with
1
=
2
3
+
2
2
,
2
=
3
1
+
3
3
,
3
=
1
2
+
1
1
,
1
=
2
3
+
2
3
+
3
2
,
2
=
3
1
+
3
1
+
1
3
,
3
=
1
2
+
1
2
+
2
1
.
26. Soldering form on F
M
n
Given the dierentiable manifold M
n
, the soldering or canonical form on F
M
n is the R
n
-valued
dierential 1-form on FM
n
i.e. (T
FM
n
R
n
) dened as follows:
: FM
n
T
FM
n
R
n
, (x, r
x
) ((x, r
x
)) = ((x, r
x
),
(x,rx)
),
(x,rx)
: T
(x,rx)
FM
n
R
n
, v
(x,rx)

(x,rx)
(v
(x,rx)
) = r
1
x

F
(x,rx)
(v
(x,rx)
)
i.e.
(x,rx)
= r
1
x
d
F
|
(x,rx)
,
where
F
is the projection in the bundle F
M
n (section 24) and r
x
is the vector space isomorphism
r
x
: R
n
T
x
M, (
1
, . . . ,
n
) r
x
(
1
, . . . ,
n
) =
n
i=1
i
v
ix
with inverse
r
1
x
(
n
i=1
i
v
ix
) = (
1
, . . . ,
n
). We have the commutative diagram
R
n
Id
R
n
(x,rx)
r
1
x
T
(x,rx)
FM
n
dF |
(x,rx)
T
x
M
n
Notice that dim
R
(T
(x,rx)
FM
n
) = n+n
2
. Since d
F
|
(x,rx)
is onto,
(x,rx)
is a vector space epimorphism, with
ker(
(x,rx)
) = V
(x,rx)
, the vertical space of the bundle FM
n
at (x, r
x
), with dim
R
ker(
(x,rx)
) = n
2
. The
existence of is independent of any connection. Also, it is clearly a global section of the bundle T
FM
n
R
n
.
(e
)
j
=
j
, , j = 1, . . . , n is the canonical basis of R
n
, then
(x,rx)
=
n
=1
(x,rx)
e
where
(x,rx)
T
(x,rx)
FM
n
.
In local coordinates on F
U
,
=
n
=1
(X
1
)
dx
where (X
1
)
(x, r
x
) = (X
(x, r
x
))
1
. [In fact, if v
(x,rx)
T
(x,rx)
FU
, then v
(x,rx)
=
n
=1
|
(x,rx)
+
n
,=1
|
(x,rx)
with d
F
|
(x,rx)
(v
(x,rx)
) =
n
=1
|
x
T
x
U
; then
(x,rx)
(v
(x,rx)
)
= r
1
x
d
F
|
(x,rx)
(v
(x,rx)
) = r
1
x
(
n
=1
|
x
) =
n
=1
r
1
x
(

x
|
x
) =
n
,=1
(X
(x, r
x
))
1
e
;
on the other hand,
(x,rx)
(v
(x,rx)
) = (
n
=1
(x,rx)
e
)(
n
=1
|
(x,rx)
+
n
,=1
|
(x,rx)
)
= (
n
,=1
(X
1
)
(x, r
x
)dx
|
(x,rx)
e
)(
|
(x,rx)
+
n
,=1
|
(x,rx)
)
=
n
,,=1
(X
(x, r
x
))
1
n
,=1
(X
(x, r
x
))
1
.]
27
Thus, a local section on F
M
n, s
: U
FU
, x s
(x) = (x, r
x
), gives rise to a set of n local
dierential 1-forms
on FU
.
If is a connection on F
M
n, and H
(x,rx)
is the horizontal space at (x, r
x
), then
(x,rx)
|
H
(x,rx)
= r
1
x
d
F
|
(x,rx)
|
H(x,rx)
is a vector space isomorphism, since d
F
|
(x,rx)
is a canonical isomorphism between H
(x,rx)
and T
x
M
n
:
R
n
Id
R
n
(x,rx)
|
H
(x,rx)
r
1
x
H
(x,rx)
dF |
(x,rx)
T
x
M
n
We emphasize that
(x,rx)
|
H
(x,rx)
depends on both the frame at x (r
x
) and the connection .
Any connection on the frame bundle F
M
n, together with the canonical soldering form , trivializes
the tangent bundle of F
M
n. This fact is known as absolute parallelism. The canonical bundle isomorphism
(only depending on ) is given through the following diagram:
R
n+n
2
R
n+n
2
| |
(TFM
n
)
2(n+n
2
)
c
(FM
n
)
n+n
2
R
n+n
2
F

1
(FM
n
)
n+n
2 Id
(FM
n
)
n+n
2
with
c
(((x, r
x
), v
(x,rx)
)) = ((x, r
x
), (
(x,rx)
|
H
(x,rx)

(x,rx)
|
V
(x,rx)
)(hor(v
(x,rx)
), ver(v
(x,rx)
))),
where v
(x,rx)
T
(x,rx)
FM
n
and gl
n
(R) = R(n)
= R
n
2
.
Absolute parallelism in the bundle of Lorentz frames F
L
M
n is given by the diagram
R
n(n+1)
2
R
n(n+1)
2
| |
(TF
L
M
n
)
n(n+1)

L
c
(F
L
M
n
)
n(n+1)
2
R
n(n+1)
2
L

1
(F
L
M
n
)
n(n+1)
2
Id
(F
L
M
n
)
n(n+1)
2
with
L
c
(((x, e
x
), v
(x,ex)
)) = ((x, e
x
), (
(x,ex)
|
H
(x,ex)

L
(x,ex)
|
V
(x,ex)
)(hor(v
(x,ex)
), ver(v
(x,ex)
))),
e
x
= (e
1x
, . . . , e
nx
), and H
(x,ex)
= ker(
L
(x,ex)
).
In particular, for the n = 4 case:
R
10
R
10
| |
(TF
L
M
4
)
20

L
c
(F
L
M
4
)
10
R
10
L

1
(F
L
M
4
)
10
Id
(F
L
M
4
)
10
.
28
27. Linear connection in a manifold M
n
on F
M
n
A gl
n
(R)-valued linear connection on M
n
is locally given by
U
=
dx
where (E
with , , , = 1, . . . , n is the canonical basis of R(n) = gl

n
(R) = Lie(GL
n
(R)),
and
are the Christoel symbols (section 4).

On FU, the connection
FU
such that
U
=
(
FU
) with : U FU the local section given by
x
(x
) = (x
)) is given by
FU
= (X
1
)
(dX
dx
) E
.
(Kobayashi and Nomizu, 1963; pp 140-143)) Clearly,
U
(T
U gl
n
(R)) and
FU
(T
FU gl
n
(R)).
Real-valued connection 1-forms
U
and
FU
are dened by
U
=
U
and
FU
=
FU
.
The horizontal lift of a local vector eld

x
by the connection in F
M
n is then given by
(

x
=

x
.
In fact,
FU
((

x
) = (X
1
)
(dX
dx
)(

x
)
= (X
1
)
dX
(

X
) +
dx
(

x
))
= (X
1
)
) = (X
1
)
) = 0.
28. Tetrads and spin connection
1. At each chart U M
n
we can take as a basis of (TU) the local vector elds (Vielbeine)
e
a
= e
a
, a = 1, . . . , n
with r
x
= (e
1x
, . . . , e
nx
) (FU)
x
. Since the n n matrices (e
a
(x)) GL
n
(R), there exist the inverse
vector elds e
1
a
e
a
= e
a
dx
: 1-forms with e
a
= (e
a
)
1
GL
n
(R) and
e
a
b
=
b
a
and e
a
a
=
.
Then e
a
e
a
= e
a
e
a
i.e.
= e
a
e
a
.
In general, the e
a
s are called non-coordinate basis and the e

a
s anholonomic coordinates. For n = 4, the
Vierbeine e
a
are called tetrads.
2. While [
] = 0, the Vielbeine have non-vanishing Lie brackets. In fact, applying the commutator
[e
a
, e
b
] to a function f C
(U, R), one easily obtains

[e
a
, e
b
] =
c
ab
e
c
with
c
ab
= e
c
(e
a
e
b
) e
b
e
a
)) =
c
ba
.
29
3. For a local vector eld, V = V
= V
a
e
a
= V
a
e
a
, with
V
a
= e
a
V
, V
= e
a
V
a
.
4. At each x M
n
, the bre of the co-frame bundle of M
n
,
F
M
n : GL
n
(R) (FM
n
)

M
n
,
is the set
(FM
n
)
x
= {ordered basis of T
x
M} = {(f
1
x
, . . . , f
n
x
), f
a
x
= f
a
(x)dx
|
x
}.
Again, (f
a
(x)) GL
n
(R) and, locally, (f
a
)
1
= f
a
with
f
a
b
=
b
a
and f
a
a
=
.
Also,
f
a
= f
a
dx
and dx
= f
a
f
a
.
From the duality relation dx
) =
we obtain
= f
a
f
a
(e
b
e
b
) = f
a
b
f
a
(e
b
); imposing the duality
relation between f
s and e
s,
f
a
(e
b
) =
a
b
one obtains
f
d
= e
d
and f
a
= e
a
.
Then,
f
a
= e
a
dx
and dx
= e
a
f
a
.
(Another usual notation for f
a
is
a
.)
5. Given an (r, s)-tensor in M
n
,
T = T
1...r
1...s

1
. . .
r
dx
1
. . . dx
s
,
we obtain
T = T
1...r
1...s
e
1
a1
e
a1
. . . e
r
ar
e
ar
f
b1
1
f
b1
. . . f
bs
s
f
bs
= T
a1...ar
b1...bs
e
a1
. . . e
ar
f
b1
. . . f
bs
with
T
a1...ar
b1...bs
= e
1
a1
. . . e
r
ar
f
b1
1
. . . f
bs
as
T
1...r
1...s
.
For example,
T = T
dx
= T
a
e
a
dx
= T
a
e
a
dx
= T
f
b
f
b
= T
f
b
= T
a
e
a
f
b
f
b
= T
a
b
e
a
f
b
.
6. Let g = g
dx
dx
be a metric in M
n
. g has a signature, given by the diagonal metric
ab
, equal
to
ab
in the euclidean case ( =
E
), or with -1s and +1s in the general riemannian case; the lorentzian
case, relevant for GR, has =
L
(section 25.). g
being a symmetric matrix, at any point x M

n
it
can be diagonalized to
ab
. The metric and its signature distinguish the subset of Vielbeine which obey the
following orthonormality condition:
g(e
a
, e
b
) =
ab
.
In detail,
g
dx
dx
(e
a
, e
b
) = g
e
a
e
b
dx
)dx
) = g
e
a
e
b
= g
e
a
e
b
30
i.e.
g
e
a
e
b
=
ab
(a).
This relation can be easily inverted, namely, g
e
a
e
b
a
e
b
= g
= g
i.e.
ab
e
a
e
b
= g
or g =
ab
e
a
e
b
. (b)
The unique solution of g
( x) =
ab
e
a
( x)e
b
( x) =
is e
a
( x) =
a
. Such a frame is called inertial at x.

Notice that formula (b) is fundamental: since it holds in all charts in M
n
, we have, up to topological
obstructions, trivialized the metric everywhere i.e. globally, at the expense of the x-dependence of the
coframes e
a
.
It is usual the rough statement that the duals of the Vielbeine are the square roots of the metric. In
particular, for the lorentzian n = 4 case, det(g
) = (det(e
a
))
2
. Also, equation (a) allows to interprete
the n n matrices e
c
as the matrices which diagonalize the metric g
to the Lorentz metric

ab
. (b) says
that the e
a
s are more fundamental than the metric.
7. Equation (b) in the last subsection appears naturally when describing spinor elds in curved space-
times. If
(x) are the Dirac matrices in M

n
, then
{
(x),
(x)} = 2g
(x)I.
The solution
(x) = E
a
(x)
a
()
with
a
the at Dirac matrices obeying {
a
,
b
} = 2
ab
I, leads to
ab
E
a
(x)E
b
(x) = g
which says that the E
a
s are the duals of the Vielbeine e
a
. It can be proved that the solution () is

unique (ORaifeartaigh, 1997).
8. It is clear that through (b) in subsection 6., the n
2
quantities involved by a Vielbein determine
uniquely a metric g
; however, the set of

n(n+1)
2
components of a metric determines a Vielbein only up to
an equivalence relation i.e. it determines a class of Vielbeine, whose elements are related by a group G of
n(n1)
2
elements.
Let e
a
= h
a
c
e
; then
ab
= g
h
a
c
e
h
b
d
e
= g
h
a
c
h
b
d
=
cd
h
a
c
h
b
d
= h
a
c
cd
h
b
d
i.e.
= hh
T
.
So, if =
L
then h L
n
= O(1, n 1); if =
E
then h O
n
= O(n); etc.
In the following we shall restrict to the case of orthonormal frames in the sense dened in 6., so one has
a principal G-bundle over M
n
:
G
n(n1)
2
F
G
M
n
,
with G = L
n
or O
n
, and F
G
FM
n
. The interest in GR is for G = L
4
and one has the lorentzian bundle
L
4
F
L4
M
4
31
and the bundle reduction
L
4
GL
4
(R)

F
L4
FM
4
L4

F
M
4
Id
M
4
where is the inclusion.
We want to emphasize that the bundle is trivial, that is, F
L4
= M
4
L
4
, if M
4
is contractible i.e. if it
is of the same homotopy type as a point.
At each U M one has a local Lorentz group L
U
which, at x U, takes the value L(x). Since the
Vielbeine are natural basis of (TU), and the metric necessarily has a signature
ab
, the principal L
n
-bundles
over M
n
are also natural. (We emphasize that here, we do not use the word natural in its technical sense,
but in a coloquial sense.) We also notice the natural appearance of a new group at each x M
n
, besides
the group of general coordinate transformations in the intersections of open sets: the Lorentz group.
9. In section 4., the covariant derivative of a local section
i
of an arbitrary vector bundle E was dened
by
i
=
j
i
j
. Let
i
= e
a
; then
e
a
=
(e
a
) = (
e
a
+e
a
= (
e
a
+e
a
.
We now dene the spin connection coecients
b
a
through
e
a
:=
b
a
e
b
and
e
a
:=
a
b
e
b
.
(The
b
a
s are, in the present case, nothing but the
j
i
s.) Then,
b
a
e
b
= (
e
a
+e
a
; from the
linear independence of the coordinate basis, and multiplying by e
c
we obtain
c
a
= e
e
a
+e
c
e
a
(c)
or
c
ab
= e
b
e
,a
c
+e
a
e
b
(c
)
with
c
ab
= e
ab
,
ab
= g
ab
, e
,a
c
=
a
e
c
,
a
= e
a
= e
a
.
Multiplying by e
c
a
we obtain the inverse relation:
= e
a
a
+e
c
c
a
(d)
or
=
c
da
e
c
a
e
d
e
a,d
e
a
d
(d
)
with e
a,d
=
d
e
a
.
Multiplying (c) by e
c
and (d) by e
b
we obtain, respectively,
e
a
e
a
c
a
e
c
= 0 (e)
and
b
+
b
a
e
a
= 0. (f)
The covariant derivative of tensors with upper and lower internal (Lorentz) and external (space-time)
indices is dened by:
D
T
1...ra1...at
1...sb1...bu
=
T
1...at
1...bu
+
1
1
T
12...at
1...bu
+. . . +
at
ct
T
1...at1ct
1...bu
1
1
T
1...at
12...bu
. . .
cu
bu
T
1...at
1...bu1cu
.
32
With this denition, the equations (e) and (f) read
D
e
a
= 0 and D
b
= 0
respectively. (We also denote D
T
1...ra1...at
1...sb1...bu
= T
1...ra1...at
1...sb1...bu ;
.)
From the Lorentz transformations of the tetrads e
a
and their inverses, e
a
= e
c
h
1
c
a
we obtain:
e
e
a
= e
r
h
1
r
c
(h
a
d
e
) = e
r
h
1
r
c
(h
a
d
)e
+e
r
h
1
r
c
h
a
d
(h
a
d
)h
1
d
c
+h
a
d
(e
)h
1
r
c
,
and
e
c
e
a
= e
r
h
1
r
c
h
a
d
e
= h
a
d
(e
r
e
)h
1
r
c
.
Then
c
a
= h
a
d
(e
+e
r
e
)h
1
r
c
+
(h
a
d
)h
1
d
c
= h
a
d
r
d
h
1
r
c
+
(h
a
d
)h
1
d
c
.
I.e.
= h
h
1
+ (dh)h
1
or, equivalently,
= h
1
h h
1
dh.
So, the 1-form
a
b
:=
a
b
dx
is not a 1-1 L-tensor, since its transformation has an inhomogeneous term.

Notice that from (c), has the structure
= e
1
( + )e,
and from (d), the structure of is
= e( +)e
1
.
It can be easily shown that for a metric connection , for which
D
= 0,
the spin connection with lower Lorentz indices
bc
=
a
c
ab
is antisymmetric in these indices. In fact, using (e) and (f),
0 = g
;
= (
ab
e
a
e
b
)
;
=
ab;
e
a
e
b
+
ab
e
a
;
e
b
+
ab
e
a
e
b
;
=
ab;
e
a
e
b
= (
c
a
cb
c
b
ac
)e
a
e
b
= (
ba
+
ab
)e
a
e
b
,
then
0 = (
ba
+
ab
)e
a
e
b
e
c
e
d
= (
ba
+
ab
)
a
c
b
d
= (
dc
+
cd
)
i.e.
dc
=
cd
.
Thus, we see that it is the condition of metric compatibility which reduces the Lie algebra of the gauge group
from gl
n
(R) to o(1, n1) (or o(n)), where the 1-form
bc
=
bc
dx
takes values. The reduced gauge group

33
can be O(1, n 1) (or O(n)) or, for the case n = 4, SL(2, C) if h
a
c
L
+
= SO
0
(1, 3) at each x M
4
(Randono, 2010).
Up to here the content of this section does not depend on the symmetry properties of
in its lower
indices. If in particular the Levi-Civita connection of section 13. is inserted in (b), using (c) (valid for any
connection with local coecients
), we obtain the spin connection coecients

c
a
in terms of the
Vielbeine, their derivatives, and the Lorentz metric
ab
:
c
a
= e
e
a
+
1
2
e
c
e
a
fd
e
f
e
d
hk
(
(e
h
e
k
) +
(e
h
e
k
)
(e
h
e
k
)).
So, we have the result analogous to the dependence of the Levi-Civita connection on the metric: the depen-
dence of the spin connection on the tetrads.
10. Explicitly, on each chart (U, x
) in M, the (metric) spin connection with values in so(3, 1), is

constructed as follows:
=
1
2
ab
dx
l
ab
(T
U so(3, 1)), x (x) = (x,

x
) with
x
=
1
2
ab
(x)dx
|
x
l
ab

T
x
U so(3, 1)
i.e.
x
: T
x
U so(3, 1), v
x

x
(v
x
) =
1
2
ab
(x)dx
|
x
(v
x
)l
ab
=
1
2
ab
(x)v
x
l
ab
=
1
2
ab
(x)l
ab
with
ab
(x) =
ba
(x) :=
ab
(x)v
x
and
x
(v
x
) =
01
l
01
+. . .
31
l
31
.
For later use, consider the connected component of the Poincare group P
4
, the semidirect sum of the of
the translation group T
4
and the connected component of the Lorentz group L
4
:
P
4
= T
4
SO
0
(3, 1), (a
)(a, ) = (a
a,
)
with Lie algebra
p
4
= R
4
so(3, 1), (
, l
)(
, l) = (l
, [l
, l]) = (
, l)(
, l
).
(P
4
is a subgroup of the ane group A
4
; for arbitrary n, A
n
= R
n
GL
n
(R) with Lie algebra a
n
=
R
n
R(n).)
Also, on each chart (U, x
n
) in M, the tetrad (1-form) e
a
with values in Lie(T
4
) = R
4
i.e. e
a
(T
U R
4
) is constructed as follows:
e
a
= e
a
dx
: U T
U R
4
, e
a
(x) = (x, e
a
x
), e
a
x
= e
a
(x)dx
|
x
x
U R
4
,
e
a
x
: T
x
U R
4
, e
a
(x)dx
|
x
(v
x
)
=
x
with
x
= e
a
(x)v
x
R.
11. The Lorentz bundle L
4
F
L4
M
4
in subsection 8. extends the symmetry group of GR, the
group of general coordinate transformations of M
4
, D, to the semidirect sum
G
GR
= L
4
D,
with composition law given by
(h
, g
)(h, g) = (h
(g
hg
1
), g
g).
34
In fact, it is easy to verify that D has a left action on L
4
at each bre F
L4x
of the bundle, given by the
commutative diagram
F
L4x
h
F
L4x
g g
F
L4x
h
F
L4x
which denes
h
= ghg
1
L
g
(h)
(conjugation action), with g(x, (e
1
(x), . . . , e
4
(x))) = g(e
a
(x)) = g(e
a
(x)

x
|
x
) = e
(x)

x
|
x
, e
(x) =
x
e
a
(x). The action is left since h
= g
1
= g
(ghg
1
)g
1
= (g
g)h(g
g)
1
i.e. L
g
g
(h) =
L
g
L
g
(h). (See the extension to G
GR
= P
4
D in section 34.)
29. Curvature and torsion in terms of spin connection and tetrads. Cartan structure
equations; Bianchi identities
In what follows we shall designate by
k
(L
r
s
) the real vector space of k dierential forms on M with
values in the (r, s)-Lorentz tensors.
Given the Vielbeine e
a
and the spin connection
a
b
on the chart (U, x
) on M, we have the dierential

forms
e
a
= e
a
dx

1
(L
1
) and
a
b
=
a
b
dx
.
(
a
b
/
1
(L
1
1
) since
a
b
is not an L
1
1
-tensor, but a connection on the Lorentz bundle L
4
F
L4
M
4
.)
Then we have the 2-forms
T
a
= de
a
+
a
b
e
b
=
1
2
T
a
dx
dx

2
(L
1
), ()
with
T
a
a
+
a
b
e
a
b
e
b
= T
a
,
and
R
a
b
= d
a
b
+
a
c

c
b
=
1
2
R
a
b
dx
dx

2
(L
1
1
), ()
with
R
a
b
=
a
b
a
b
+
a
c
c
b
a
c
c
b
.
() and () are known as the Cartan structure equations. As we shall show below, T
a
and R
a
b
are,
respectively, the torsion and curvature 2-forms of section 9.
For de
a
one has
de
a
= d(e
a
dx
) =
a
dx
dx
=
a
dx
dx

a
with
= (de
a
)
=
1
2
(e
,
a
e
,
a
) =
a
.
Also,
a
bc
= (de
a
)
bc
= e
b
e
c
=
1
2
e
a
(e
c
e
b
e
b
e
c
) =
1
2
e
a
(e
b,c
e
c,b
) =
a
cb
.
Comparing with
a
bc
of 28.2, we have
a
bc
=
1
2
a
bc
and therefore
[e
b
, e
c
] = 2
a
bc
e
a
= 2(de
a
)
bc
e
a
.
35
So
a
bc
also measures the non-commutativity of the Vielbeine. By the Jacobi identity,
f
ad
d
bc
+
f
bd
d
ca
+
f
cd
d
ab
= 0.
It is easy to show that for a metric connection, the curvature tensor with lower Lorentz indices
R
ab
=
ad
R
d
b
is antisymmetric i.e.
R
ab
= R
ba
.
In fact, R
ab
=
ad
(d
d
b
+
d
c

c
b
) =
ad
d
d
b
+
ad
d
c

c
b
= d
ab
+
ac

c
b
= (d
ba
+
ca

c
b
) =
(d
ba
c
b
ca
) = (d
ba
cb
c
a
), while R
ba
=
bc
R
c
a
=
bc
(d
c
a
+
c
d
d
a
) = d
ba
+
bd
d
a
=
d
ba
db

d
a
.
Symbolically we write
T = de + e, R = d + .
We notice that torsion is related to the tetrads as curvature is related to the spin connection.
On the other hand, while curvature involves only , torsion involves both e and (not only e). This
is related to the fact that the Poincare group is the semidirect (not direct) sum of R
4
(translations) and
SO
0
(3, 1) (spacetime rotations).
A manifold equipped with a metric g
and a connection
compatible with the metric but with non-

vanishing torsion, is called an Einstein-Cartan manifold. The metric induces the Levi-Civita connection,
(
LC
)
(section 13) with
= (
LC
)
+ contortion tensor.
In the Einstein-Cartan (E-C) theory of gravity, the 1-forms
{e
a
,
ab
}
are called gauge or gravitational potentials, respectively translational and rotational, while the 2-forms
{T
a
, R
a
b
}
are called gauge or gravitational eld strengths, respectively translational and rotational. (See, however,
section 34.) At a point, it is always possible to set e
pt
= 1 and
pt
= 0, i.e. respectively e
a
=
a
(16
conditions) and
ab
= 0 (24 conditions). (Hehl, 1985; Hartley, 1995.) The total number of conditions,
40, coincides with that for making zero the Christoel symbols in the case of the Levi-Civita connection
(|{(
LC
)
}| = 40).
Comment. Together with the comments in section 10, we have the following relations:
curvature spin connection spacetime rotations,
torsion tetrads spacetime translations.
On the other hand, from Noether theorem, we have the relations:
spacetime rotations angular momentum,
spacetime translations energy momentum.
36
Naively, one should then expect the following relations:
curvature angular momentum,
torsion energy momentum.
However, in Einstein-Cartan theory, based on a non-symmetric metric connection, the sources of cur-
vature and torsion are respectively energy momentum and spin angular momentum i.e.
curvature energy momentum,
torsion spin angular momentum.
This crossing of relations is due to holonomy theorems (Trautman, 1973).
These facts can be better understood as follows: In (special) relativistic eld theory (r.f.t.), elds belong
to irreducible representations of the Poincare group P
4
, which are characterized by two parameters: mass
and spin. Invariance under translations (T
4
) and rotations (L
4
) respectively leads, by Noether theorem, to
the conservation of energy-momentum (T
) and angular momentum: orbital + intrinsic (spin, with density

S
). On the other hand, dierential geometry (d.g.), through holonomy theorems, relates curvature (R
)
with the Lorentz group and torsion (T
) with translations (section 10). Finally, Einstein (E) equations

make energy-momentum the source of curvature, while Cartan (C) equations makes spin the source of torsion.
This is summarized in the following diagram:
L
4
d.g. r.f.t.
R
E C
T

e
T
r.f.t. d.g.
T
4
In a formulation of the Einstein-Cartan theory based on tetrads and spin connection, the Einstein equations
are obtained by variation with respect to the tetrads (
e
), related to translations, and the Cartan equations
by variation with respect to the spin connection (
), related to rotations. (See section 32.)

Locally, as dierential 2-forms with values in so(3, 1) and Lie(T
4
) = R
4
, R and T
a
are respectively
given as follows:
R =
1
2
R
ab
dx
dx
l
ab
(
2
U so(3, 1)), R
ab
=
ad
R
d
b
, R(x) = (x, R
x
),
R
x
=
1
2
R
ab
(x)dx
|
x
dx
|
x
l
ab

2
x
U so(3, 1), R
x
: T
x
U T
x
U so(3, 1), R
x
(v
x
, w
x
)
=
1
4
R
ab
(x)(v
x
w
x
v
x
w
x
)l
ab
;
T
a
= T
a
dx
dx
(
2
U R
4
), T
a
(x) = (x, T
a
x
), T
a
x
= T
a
(x)dx
|
x
dx
|
x
,
T
a
x
: T
x
U T
x
U R
4
, T
a
x
(v
x
, w
x
) =
1
2
T
a
(x)(v
x
w
x
v
x
w
x
)
.
From the denition of T, we have, since d
2
= 0, dT = d e de = d e (T e) =
d e T+ e, i.e. dT+ T = (d + ) e, that is
dT+ T = R e
3
(L
1
). ()
37
In Lorentz components,
dT
a
+
a
b
T
b
= R
a
b
e
b
. (
)
For R one has dR = d d = (R ) (R ) = R R, i.e.
dR+ RR = 0
3
(L
1
1
). ()
In components,
dR
a
b
+
a
c
R
c
b
R
a
c

c
b
= 0. (
)
() and () (or (
) and (
)) are the so called Bianchi identities. (Compare () with the corresponding

equation in section 12.)
Dening the covariant exterior derivative operator acting on Lorentz tensors-valued dierential forms
D
= d +
we have the equations
T = D
e, D
T = R e, D
R = R .
Though is not a Lorentz tensor, one has R = D
.
It is easy to verify that T
a
is nothing but twice the torsion tensor of section 9.:
e
a
T
a
= e
a
((de
a
)
+(
a
b
e
b
)
) = e
a
a
+
a
b
e
a
b
e
b
) = (e
a
a
+e
a
a
b
)
(e
a
a
+e
a
a
b
) =
= 2T
.
A similar calculation leads to
e
a
b
R
a
b
= e
a
b
(d
a
b
+
a
c
c
b
)
= e
a
b
(
a
b
a
b
+
a
c
c
b
a
c
c
b
) = R
.
For the Ricci tensor and the Ricci scalar of sections 16 and 18 respectively (but now not restricted to
the Levi-Civita connection) we have
R
= R
= e
a
b
R
a
b
and R = R
= e
a
b
R
a
b
g
= e
a
e
b
R
ab
.
We summarize the above formulae in the following table:
1
(L
1
)
1
(L
1
1
)
2
(L
1
)
2
(L
1
1
)
3
(L
1
)
3
(L
1
1
) Components
e e
a

a
b
T = de + e T
a
= de
a
+
a
b
e
b
R = d + R
a
b
= d
a
b
+
a
c

c
b
dT+ T = R e dT
a
+
a
b
T
b
= R
a
b
e
b
dR+ R = R dR
a
b
+
a
c
R
c
b
= R
a
c

c
b
30. Spin connection in non-coordinate basis
The Christoel symbols for a metric connection with torsion is given in Appendix B:
= (
LC
)
+K
38
where the contortion tensor depends on the metric and the torsion, while (
LC
)
only depends on the metric

and its derivatives. Contracting with e
a
s and e
b
s one has
e
a
e
b
c
(
LC
)
+e
a
e
b
c
K
= e
a
e
b
,
and using (c
) in 28.9 we obtain
c
ab
+e
b
e
,a
c
= e
a
e
b
c
(
LC
)
+K
c
ab
.
Using the expressions for g
, etc. in
LC
in terms of e
s and their derivatives, a straightforward

calculation leads to
dab
=
abd
+K
dab
(U
4
space) ()
where
abd
=
dab
+
abd
bda
=
adb
are the Ricci rotation coecients, with |{
abd
}| =
n
2
(n1)
2
, (24 for n = 4), and
X
abc
=
ad
X
d
bc
, X = , , , K; X
d
cb
= X
d
bc
, X = , , .
If T
a
= 0, then
dab
=
abd
. (V
4
space)
We emphasize that the
abd
s come from the metric, the Vielbeine and their inverses and derivatives. So,
the parallel transport and concomitant rotations of vectors by () has two sources: metric (g) and torsion
(): from
ab
=
ab
+K
ab
,
we have
(
||
A)
a
|
x
=
ab
(x)A
b
(x)dx
|
x
= (
(g)
||
A)
a
|
x
+ (
()
||
A)|
x
with
(
(g)
||
A)
a
|
x
=
ab
(x)A
b
(x)dx
|
x
and
(
()
||
A)
a
|
x
= K
ab
(x)A
b
(x)dx
|
x
.
(A
a
= e
a
, A
b
=
bc
A
c
.)
31. Locally inertial coordinates
Let (M
n
, g, ) be a U
n
-space (see Appendix B), x M
n
and (U, = (x
)) a chart on M
n
with x U
and x
(x) = 0, = 0, . . . , n 1. Let (U
= (x
)) be an intersecting chart with x
(x) = 0 and
x
= x
1
2
()
(x)x
,
where () means symmetrization. The antisymmetric part
[]
= T
= T
(torsion) does not contribute

to the change of coordinates.
The condition of metricity at x,
0 = g
;
(x) = g
,
(x)
(x)g
(x)
(x)g
(x),
which, being a tensor also holds in the chart U
, the tensor transformation formula of g
(section 7), and

the diagonalization of g
to
at x (uniquely determined up to a Lorentz transformation (section 28.8)),

lead to the equations:
g
(x
) =
+ (
(x) +
(x))x
+O(x
2
) =
+

x
(x)x
+O(x
2
), (a)
39
and
(
LC
)
(x) =
1
2
(g
)(x) +
(g
)(x)
(g
)(x)). (b)
So,
T
(x) = 0
(g
)(x) = 0 and (
LC
)
(x) = 0,
i.e. the vanishing of the torsion at x is a sucient condition for having a local inertial system at x.
However, the condition is not necessary: in fact,
(p) +
(p) = T
(p) +T
(p) = 0
implies that T
is also antisymmetric in its second and third indices, and then it is totally antisymmetric,
since T
= T
= T
= T
.
A calculation gives:
n = 2:
T
0
01
= T
1
01
= 0
n = 3:
T
0
01
= T
0
02
= T
1
01
= T
1
12
= T
2
02
= T
2
12
= 0,
T
0
12
= T
2
10
= T
1
02
n = 4:
T
0
01
= T
0
02
= T
0
03
= T
1
01
= T
1
12
= T
1
13
= T
2
02
= T
2
12
= T
2
32
= T
3
03
= T
3
13
= T
3
23
= 0,
T
0
12
= T
2
10
= T
1
02
,
T
0
13
= T
3
10
= T
1
03
,
T
0
23
= T
3
20
= T
2
03
,
T
1
23
= T
2
31
= T
3
12
In each case, the number of independent but not necessarily zero components of the torsion tensor coincides
with the number of independent components of the totally antisymmetric torsion tensor with covariant
indices, number which results from the condition that the denition of geodesics as world-lines of particles
(parallel transported velocities, section 8) to coincide with their denition as extremals of length. This last
fact can be seen as follows:
As world-lines, geodesics are dened in section 8, the equation being
d
2
x
d
2
+
()
dx
d
dx
d
= 0 ()
where only the symmetric part of
contributes:
()
= (
LC
)
(T
+T
) = (
LC
)
(T
+T
) = (
LC
)
2T
(
)
with g
= g
= T
; notice that the covariant form of the torsion tensor, T
, is anti-
symmetric in the rst two indices: T
= T
. (With this denition of T
, the covariant form of the

contortion tensor is
K
= g
= T
+T
,
40
which is antisymmetric in the last two indices i.e. K
= K
.)
On the other hand, the equation of geodesics dened as extremals of arc-length:
0 =
_
ds =
_
(g
dx
dx
)
1
2
,
turns out to be (Carroll, 2004, pp.106-109)
d
2
x
d
2
+ (
LC
)
dx
d
dx
d
= 0. ()
Then, for the denitions () and () to coincide, T
(
)
must vanish i.e. T
= T
=
g
= T
i.e. T
must be 1-3 antisymmetric; but this implies that T
is also 2-3
antisymmetric: T
= T
= T
= T
. Since, by denition, T
is antisymmetric in the rst

two indices, it then turns out to be totally antisymmetric; in n dimensions, its number of independent
components is
_
n
3
_
=
n(n1)(n2)
6
N. Some values are:
n 2 3 4
N 0 1 4
The set of allowed non-vanishing components of the torsion tensor still leads to physical (geometrical)
eects in the sense of section 10. The non-closure of a parallelogram with innitesimal sides
and
is
measured by the vector
= 2T
= T
)
. For n = 4 its components are:
0
= T
0
12
(
1
1
) +T
0
13
(
1
1
) +T
0
23
(
2
2
),
1
= T
1
23
(
2
2
) +T
1
02
(
0
0
) +T
1
03
(
0
0
),
2
= T
2
31
(
3
3
) +T
2
10
(
1
1
) +T
2
03
(
0
0
),
3
= T
3
10
(
1
1
) +T
3
20
(
2
2
) +T
3
12
(
1
1
),
which can be distinct from zero.
In summary, the necessary and sucient condition for erecting a locally inertial coordinate system at a
point x in a U
4
-space, is that the symmetric part of the contortion tensor vanish up to terms of order (x
)
2
,
where x
(x) = 0. (Socolovsky, 2010.)

In the above sense, the weak equivalence principle, which only refers to the free motion of point-like
and therefore classical particles, still holds in a U
4
-space (U
4
= (M
4
, g, ) with a metric connection). (A
similar result was recently found by Fabbri (Fabbri, 2011).)
32. Einstein-Cartan equations
(We owe this derivation to L. Fabbri, 2010.)
A. Pure gravitational case (vacuum)
We start from the curvature 2-form R
a
b
with components R
a
b
given in section 29; we dene the Ricci
tensor
R
b
:= R
a
b
e
a
41
and the Ricci scalar
R = R
b
bc
e
c
.
Then
R =
bc
R
a
b
e
a
e
c
= R(
a
b
, e
c
).
The gravitational action is given by the Einstein-Hilbert lagrangian density eR,
S
G
=
_
d
4
x eR
with e =
_
detg
.
A.1. Variation with respect to the spin connection: =
(Cartan equations)
Varying R
a
b
w.r.t. , using (
c
d
) =
(
c
d
), and adding and subtracting (
a
b
)
one obtains
R
a
b
= D
(
a
b
) D
(
a
b
) + 2
a
b
T
where
D
(
a
b
) =
(
a
b
)
c
b
(
a
c
) +
a
c
(
c
b
)
(
a
b
)
and
D
(
a
b
) =
(
a
b
)
c
b
(
a
c
) +
a
c
(
c
b
)
(
a
b
)
are covariant derivatives since
c
d
is a tensor (
c
d
is a connection, not a tensor, but the dierence of two
connections is a tensor). Then,
S
G
=
_
d
4
x ee
a
bc
e
c
R
a
b
=
_
d
4
x e(D
((
a
b
)e
a
bc
e
c
) D
((
a
b
)e
a
bc
e
c
)
+2e
a
bc
e
c
a
b
T
) =
_
d
4
x e(D
+ 2(
ac
)T
ac
)
with V
the 4-vector given by

V
= (
ac
)(e
a
e
c
e
a
e
c
),
where we have used the Leibnitz rule for D
, D
e
a
= 0, D
bc
= 0, and raised Lorentz indices with
ab
.
For D
one has
D
= (D
LC
)
+K
+ (
LC
)
+K
,
then
D
= (D
LC
)
+K
= e
1
(eV
)
,
2T
where we have used (

LC
)
(appendix C), appendix B, and the denition of the torsion 1-form

T
= T
(section 9). Neglecting the surface term

_
d
4
x (eV
)
,
one obtains
0 =
S
G
=
_
d
4
x e(T
ac
T
ac
) =
_
d
4
x e(
ac
)(e
c
T
a
+e
a
T
c
+T
ac
)
with T
a
= T
e
a
, which, due to the arbitrariness of
ac
leads to the Cartan equations for torsion (in
vacuum):
T
ac
+e
a
T
c
e
c
T
a
= 0. (i)
Multiplying (i) by e
a
e
c
one obtains the Cartan equations in local coordinates:
T
= 0. (ii)
42
Proposition 1: Torsion vanishes.
Proof. Taking the trace in (ii) leads to T
+ T
4T
= 2T
= 0; then T
= 0 and from (ii)

again,
T
= 0. qed
Note: The above result holds in n dimensions for n = 2: the trace gives (2 n)T
= 0. For n = 2, T
is arbitrary with independent components T

0
01
= T
1
and T
1
01
= T
0
. Also, notice that (i) (or (ii)) is not a
dierential equation, but an algebraic one; this is the mathematical expression of the fact that in E-C theory
torsion does not propagate.
Proposition 2: Let
T
= S
(iii)
with S
= S
, and a constant. ((iii) corresponds to a non-vacuum case and will be used in part B.)
Then,
T
= (S
1
2 n
(
)) (iv)
where n = 2 is the dimension of the manifold and S
= S
. In particular, for n = 4,
T
= (S
+
1
2
(
)). (v)
Proof. Again taking the trace in (iii), (2 n)T
= S
, then T
=
1
2n
S
and T
is (iv). qed
For n = 2, the unique solution of (iii) is S
= 0: in fact,
= 2 and then S
= 0; so S
0
= S
1
01
= 0 and
S
1
= S
0
10
= 0.
A.2. Variation with respect to the tetrads: =
e
( Einstein equations)
Again from S
G
,
e
S
G
=
_
d
4
x ((R)e +Re) =
_
d
4
x (R
a
b
bc
e((e
a
)e
c
+e
a
e
c
) Ree
d
e
d
)
=
_
d
4
x (2R
a
Re
a
)ee
a
= 0
where we used e = ee
d
e
d
(appendix C), and from the arbitrariness of e

a
, we obtain the Einstein

equations for curvature (in vacuum):
R
a
1
2
Re
a
= 0 (vi)
or
G
a
= 0 (vii)
with
G
a
= R
a
1
2
Re
a
. (viii)
(R
a
=
ac
R
c
.) Since in vacuum R = 0 (section 18), (vi) amounts to
R
a
= 0. (ix)
Of course, multiplying (vi) by e
a
we obtain Einstein equations in local coordinates (section 20).

43
In summary, for the pure gravitational case, Einstein theory=Einstein-Cartan theory; this is a conse-
quence of the form of the Einstein-Hilbert action S
G
. From the form of R, gravity has been expressed as an
interacting gauge theory (see section 33) between the spin connection
a
b
and the coframes eld e
a
; both
a
b
and e
a
are pure geometric elds, which live in the frame and coframe bundles L
4
F
L4
M
4
and
L
4
F
L4
M
4
respectively.
B. Minimal coupling to Dirac elds
The Dirac-Einstein action is given by
S
DE
= k
_
d
4
x eL
DE
= k
_
d
4
x e (
i
2
(
a
(D
a
) (

D
a

)
a
) m
)
where
D
a
= (e
a
i
4
abc
bc
) = e
a
i
4
bc
bc
) = e
a
and
D
a

= e
a

+
i
4
abc

bc
= e
a

+
i
4
bc

bc
) = e
a
are the covariant derivatives of the Dirac eld and its conjugate

=
0
with respect to the spin
connection, which give the minimal coupling between fermions and gravity; they are obtained through the
replacement
d
a
D
a
i.e. e
a
= e
a
D
a
which amounts to the comma goes to semicolon rule for tensors but here adapted to spinor elds.
bc
=
i
2
[
b
,
c
], and the
a
s are the usual numerical (constant) Dirac gamma matrices satisfying {
a
,
b
} = 2
ab
I,
0
=
0
and
j
=
j
. k = 16
G
c
4
(16 in natural units). Then the action is
S
DE
= k
_
d
4
x e (
i
2
+
1
8
bc
[
b
,
c
])
i
2
(

1
8
bc

[
b
,
c
])
)
where
= e
a
a
=
(x).
B.1. Variation with respect to the spin connection: =
S
DE
=
k
8
_
d
4
x e

{
,
bc
}
bc
=
k
2
_
d
4
x e S
bc
bc
with S
bc
= e
a
S
abc
, where
S
abc
=
1
4
{
a
,
bc
}
is the spin density tensor of the Dirac eld. S
abc
is totally antisymmetric and therefore in 4 dimensions it
has 4 independent components: S
012
, S
123
, S
230
and S
301
.
Combining this result with the corresponding variation for the pure gravitational eld (part A), we
obtain
0 =
(S
G
+S
DE
) =
_
d
4
x e
ac
(T
ac
+e
a
T
c
e
c
T
a
+
k
2
S
ac
)
and therefore
T
ac
+e
a
T
c
e
c
T
a
=
k
2
S
ac
,
the Cartan equation. Multiplying by e
a
e
c
one obtains
T
=
k
2
S
44
with
S
=
1
4
}.
From (iii) in part A, with =
k
2
, we obtain the torsion in terms of the spin tensor:
T
=
8G
c
4
(S
+
1
2
(
))
with S
= S
. In natural units, G = c = h = 1 and so T
= 8(S
+
1
2
(
)).
B.2. Variation with respect to the tetrads: =
e
From S
DE
and using appendix C for e, one obtains
e
S
DE
= k
_
e (
i
2
(

a
(D
) (

D

)
a
) e
a
L
DE
)e
a
.
For the Dirac elds which obey the equations of motion
S
DE
=
S
DE
= 0
i.e.
i
a
(

D
a

) +m
= i
a
D
a
m = 0
the Dirac-Einstein lagrangian vanishes i.e. L
DE
|
eq. mot.
= 0. Then, combining this result with the corre-
sponding variation for the pure gravitational eld (part A),
0 =
e
(S
G
+S
DE
) =
_
d
4
x e (2R
a
Re
a
+k
i
2
(
a
(D
) (

D

)
a
))e
a
,
and from the arbitrariness of e
a
,
R
a
1
2
Re
a
=
k
2
T
a
()
with
T
a
=
i
2
(
a
(D
) (

D

)
a
)
the energy-momentum tensor of the Dirac eld. Multiplying () by e
a
one obtains
R
1
2
R
=
k
2
T
or R
1
2
Rg
=
k
2
T
,
the Einstein equation.
Note: For L
DE
one has
L
DE
= e
a
T
a
i.e. T
a
couples to the tetrad. On the other hand,

T
a
=
a
+
bc
S
abc
where
=
i
2
(

)
a
)
is the canonical energy-momentum tensor of the Dirac eld. Then,
L
DE
= e
a
+e
a
bc
S
abc
m
= e
a
+
abc
S
abc
m
.
45
So,
a
couples to the tetrad while spin couples to the spin connection; moreover, since S
abc
is totally
antisymmetric, the Dirac eld only interacts with the totally antisymmetric part of the connection.
33. Lorentz gauge invariance of Einstein and Einstein-Cartan theories
Under Lorentz transformations h
a
b
(x) in the tangent, cotangent and, in general, tensor spaces at each
point x of the manifold M
4
(for simplicity we restrict the discussion to four dimensions), the frames e
a
(or
e
a
), coframes e
a
(or e
a
) and spin connection
a
b
(or
a
b
) transform as indicated in sections 28.8 and 28.9.
The volume element d
4
x e in S
G
(or S
DE
) is invariant; in fact,
g
(x) =
ab
e
a
(x)e
b
(x) =
ab
e
c
h
1
c
a
e
d
h
1
d
b
= e
c
e
d
h
1
c
a
ab
h
1
d
b
= e
c
e
cd
= g
(x)
and then det(g
(x)) = det(g
(x)) i.e. e
(x) = e(x). Then d

4
x e = d
4
x
since x
= x
.
A. Pure gravitational case (vacuum)
Though
a
b
=
a
ab
dx
is a connection and transforms as
c
a
= h
c
d
r
d
h
1
r
c
+ (dh
a
d
)h
1
d
c
,
its curvature R
a
b
is a Lorentz tensor:
R
a
b
= h
b
d
h
1
c
a
R
c
d

2
(L
1
1
)
(see table in section 29); then the Ricci scalar is gauge invariant:
R = R
a
b
e
a
bc
e
c
= h
b
d
R
c
d
h
1
c
a
h
a
f
e
bg
h
g
l
e
l
= R
c
d
f
c
e
dl
e
l
= R
c
d
e
dl
e
l
= R
.
(An explicit proof of the gauge invariance of R is given in Appendix D.)
Remark: In this section, the tetrads e
a
(or their duals e
a
) are not gauge potentials; only the spin
connection
a
b
is a gauge potential, related to the Lorentz group and therefore to space-time rotations (at
each point of the manifold). So, if only the Lorentz group is gauged, the metric g =
ab
e
a
e
b
(see eq. (b)
in section 28.6) does not come from the connection.
B. Minimal coupling to Dirac eld
See ORaifeartaigh (1997), Ch. 5,
_
3, pp 115-116.
34. Poincare gauge invariance of Einstein and Einstein-Cartan theories
We need the concepts of ane structures: spaces, bundles and connections (
_
3, chapter III, Kobayashi
and Nomizu, 1963).
The general linear group GL
n
(R) in n real dimensions acts from the left on the vector space R
n
by
simple matrix multiplication: (g, ) g, which is a linear operation.
The ane space
A
n
= {
_
1
_
, R
n
}
is acted by the ane group in n real dimensions
GA
n
(R) = {
_
g
0 1
_
, g GL
n
(R), R
n
}
46
as follows:
GA
n
(R) A
n
A
n
, (
_
g
0 1
_
,
_
1
_
)
_
g +
1
_
.
Then, one has the following diagram of short exact sequences (s.e.s.s) of groups and group homomorphisms:
0 R
n

GA
n
(R)
GL
n
(R) 0
Id
0 R
n
|
P
n
|
L
n
0
with () =
_
I
n

0 1
_
and (
_
g
0 1
_
) = g. is 1-1, is onto, and ker() = Im() = {
_
I
n

0 1
_
, R
n
}.
We have also restricted and (respectively | and |) to the connected components of the Poincare (P
n
)
and Lorentz (L
n
) groups in n dimensions. Both s.e.s.s split, i.e. there exists the group homomorphism
: GL
n
(R) GA
n
(R), g (g) =
_
g 0
0 1
_
and its restriction | to L
n
, such that = Id
GLn(R)
and
| | = Id
Ln
. So
GA
n
(R) = R
n
GL
n
(R), P
n
= R
n
L
n
with composition law
(
, g
)(, g) = (
+g
, g
g).
As a consequence, the factorization of an element of GA
n
(R) (P
n
) in terms of elements of R
n
and GL
n
(R)
(L
n
) is unique:
_
g
0 1
_
= ()(g) (or |()|(g)). The dimensions of GA
n
(R), GL
n
(R), P
n
and L
n
are,
respectively, n +n
2
, n
2
,
n(n+1)
2
and
n(n1)
2
(20, 16, 10 and 6 for n = 4).
The above s.e.s.s pass to s.e.s.s of the corresponding Lie algebras:
0 R
n

ga
n
(R)

gl
n
(R) 0
Id
0 R
n
|
p
n
|
l
n
0
with gl
n
(R) = R(n), ga
n
(R) = R
n
gl
n
(R) with Lie product
(
, R
)(, R) = (R
, [R
, R]),
where [R
, R] is the Lie product in gl

n
(R) and [
, ] = 0 in R
n
, () = (, 0), (, R) = R, and (R) =
(0, R). , and (and their corresponding restrictions |, | and |) are Lie algebra homomorphisms, with
= Id
gln(R)
and | | = Id
ln
. The s.e.s.s split only at the level of vector spaces i.e. if (, R) ga
n
(R),
then (, R) = () + (R), but (, R) = () (R).
Let us denote:
M
n
: n-dimensional dierentiable manifold
F
M
n: frame bundle of M
n
: GL
n
FM
n
F
M
n
A
M
n: ane frame bundle of M
n
: GA
n
AM
n
A
M
n
GL
n
: general linear group in n dimensions (section 24), dim
R
GL
n
= n
2
47
GA
n
: general ane group in n dimensions, dim
R
GA
n
= n +n
2
F
L
M
n: bundle of Lorentz frames of M
n
: L
n
F
L
M
n
F |
M
n
(section 25)
F
P
M
n: bundle of Poincare frames of M
n
: P
n
A
P
M
n
A|
M
n
One has the following diagram of bundle homomorphisms:
AM
n
GA
n
A
P
M
n
P
n
||
||
F
L
M
n
L
n
FM
n
GL
n
A

A
|
F
|
F
AM
n

A
P
M
n
|
F
L
M
n

FM
n
A

A
|
F
|
F
M
n
Id
M
n
Id
M
n
Id
M
n
where is the bundle homomorphism
AM
n
GA
n
FM
n
GL
n
A

F
AM
n
FM
n
A

F
M
n
Id
M
n
between the bundle of ane frames and the bundle of linear frames over M
n
, with
(x, (v
x
, r
x
)) = (x, r
x
), (x, r
x
) = (x, (0
x
, r
x
)), 0 T
x
M
n
,
AM
n
=
xM
n({x} AM
n
x
), AM
n
x
= {(v
x
, r
x
), v
x
A
x
M
n
, r
x
F
x
},
where A
x
M
n
is the tangent space at x considered as an ane space (Appendix E); and
A
((x, (v
x
, r
x
)), (, g)) = (x, (v
x
+r
x
, r
x
g))
is the action of GA
n
on AM
n
, with r
x
=
n
a=1
e
ax
a
. |
F
, . . . , |
A
, . . . , |, . . ., etc. are restrictions; in
particular |(a) = (a) and |(e) = (e).
A general ane connection (g.a.c.) on M
n
is a connection in the bundle of ane frames A
M
n. If
A
is
the 1-form of the connection, then
A
(T
AM
n
ga
n
)
i.e.
A
: AM
n
T
AM
n
ga
n
, (v
x
, r
x
) ((v
x
, r
x
),
A(vx,rx)
),
A(vx,rx)
: T
(vx,rx)
AM
n
ga
n
,
V
(vx,rx)

A(vx,rx)
(V
(vx,rx)
) = (, R) R R
n
gl
n
(R).
Obviously,
A
obeys the usual axioms of connections.
48
From the smoothness of , the pull-back
(
A
) is a ga
n
-valued 1-form on FM
n
:
(
A
) =
F
,
where
F
is a connection on FM
n
, and is an R
n
-valued 1-form. There is a 1-1 correspondence between
g.a.c.s on AM
n
and pairs (
F
, ) on FM
n
:
{
A
}
g.a.c.
{(
F
, )}.
A
is an ane connection (a.c.) on M
n
if is the soldering (canonical) form on FM
n
(section 26).
Then, if
A
is an a.c. on AM
n
,
(
A
) =
F
where
F
is a connection on FM
n
. There is then a 1-1 correspondence
{
A
}
a.c.
{
F
},
since is xed. Also, if
A
is the curvature of
A
, then
(
A
) = D
F

F
.
But D
F
= T
F
: the torsion of the connection
F
on FM
n
: in fact, from section 26,
= (X
1
)
dx
, then
a
= e
= e
a
(X
1
)
dx
= (X
1
)
a
dx
= e
a
dx
= e
a
, so D
F
a
= d
a
+
a
Fb
e
b
= T
a
F
. Therefore,
(
A
) = T
F

F
.
The facts that A
P
M
n
is a subbundle of AM
n
and F
L
M
n
is a subbundle of FM
n
, with structure groups
and Lie algebras the corresponding subgroups and sub-Lie algebras, and the existence of the restrictions
| : A
P
M
n
F
L
M
n
and | : F
L
M
n
A
P
M
n
, allow us to obtain similar conclusions for the relations
between ane connections on the Poincare bundle and linear connections on the Lorentz bundle:
There is a 1-1 correspondence between ane Poincare connections
P
on F
P
M
n
and Lorentz connections
on F
L
M
n
:
{
P
} {
L
}
with
|
(
P
) =
L
L
where
L
=
FM
n|
F
L
M
n is the canonical form on F
L
M
n
. Also,
|
(
P
) = D
L
L
= T
L
L
.
So, there is a 1-1 correspondence between curvatures of ane connections on F
P
M
n
and torsion and curva-
ture pairs on F
L
M
n
:
{
P
} {(T
L
,
L
)}.
For pure gravity governed by the Einstein-Hilbert action, T
L
= 0, as it was shown in section 32.
The Poincare gauge invariance of G.R. and E-C theory has been discussed by several authors (Hayashi
and Shirafuji, 1980; Ali et al, 2009; Gronwald and Hehl, 1996; Hehl, 1998). To explicitly prove it, we have
to consider as gauge transformations both the Lorentz part, already studied in the previous section, and the
translational part. This has to be done using the bundle of Poincare frames F
4
: P
4
A
P
M
4
P
M
4
(
P
=
A
|), (DOlivo and Socolovsky, 2011). The action of P
4
over on A
P
M
4
is given by
P
: A
P
M
4
P
4
A
P
M
4
, (
P
=
A
|),
P
((x, (v
x
, r
x
)), (, h)) (x, (v
x
, r
x
))(, h) = (x, (v
x
+r
x
, r
x
h))
49
= (x, (v
x
, r
x
)), where r
x
= (e
ax
), a = 1, 2, 3, 4, is a Lorentz frame, h L
4
, and R
4
= R
1,3
is a Poincare
gauge translation. For a pure translation, h = I
L
i.e. h
a
b
=
a
b
and therefore
(x, (v
x
, r
x
))(, I
L
) = (x, (v
x
+r
x
, r
x
I
L
)) = (x, (v
x
+r
x
, r
x
))
i.e.
r
x
= r
x
.
Therefore e
ax
= e
ax
, a = 1, 2, 3, 4, and then, from (c) or (c
) in section 28.9,
a
b
=
a
b
since
= (
LC
)
+K
remains unchanged (in the case of pure gravity K
= 0), and so the coordinate

Ricci scalar R of section 32.A is also a gauge scalar and S
G
is invariant.
The Poincare bundle extends the symmetry group of GR and E-C theory to the semidirect sum
G
GR
= P
4
D
cf. 28.11, with composition law
((
, h
), g
)((, h), g) = ((
, h
)(g
(, h)g
1
), g
g).
The left action of D on P
4
is given by the commutative diagram
A
P
M
4
(,h)
A
P
M
4
g g
A
P
M
4
(
,h
)
A
P
M
4
with
g : A
P
M
4
A
P
M
4
, (x, (v
|
x
, (e
ax
|
x
))) (x, (v
|
x
, (e
ax
))),
where v
x
=
x
|
x
v
x
and e
ax
=
x
|
x
e
ax
.
In section 29, following Hehl (Hehl, 1985; Hehl et al, 1976; Hammond, 2002), we called e
a
s the
translational gravitational gauge potentials. This is not, however, strictly correct, since the e
a
s or their
duals, the tetrad elds e
a
= e
a
, are not connections, but tensors in both their world () and internal
(a) indices (Hayashi, 1977; Leston, 2008; Leston and Socolovsky, 2011). The translational potentials B
a
(in
fact their inverses B
a
) are given by the 1-form elds locally dened as follows (Hayashi and Nakano, 1967;
Aldrovandi and Pereira, 2007):
B
a
= e
v
a
x
x
or B
a
= e
a
dv
a
x
,
where v
x
=
3
a=0
v
a
x
e
ax
A
x
M
4
; the v
a
x
s are here considered the coordinates of the tangent space at x.
The transformation properties of the B
a
s are the following:
Internal Lorentz:
B
a
= h
a
b
B
(h
b
a
)v
b
x
or B
a
= h
b
a
B
b
(dh
b
a
)v
b
x
.
Proof: B
a
= e
v
a
x
x
, with e
a
= h
b
a
e
b
and v
a
x
= h
b
a
v
b
x
, then B
a
= h
b
a
e
(h
b
a
v
b
x
)
x
.
50
General coordinate transformations:
B
a
=
x
a
.
Proof: e
a
=
x
a
i.e. e
a
is a 1-form, and

x
=
x
.
Internal translations:
B
a
= B
a
or B
a
= B
a
d
a
.
Proof: B
a
= e
v
a
x
x
= e
v
a
x
x
= B
(v
a
x
v
a
x
)
x
= B

a
x
.
Then, B = B
dx
= B
a
dx
b
a
, where b
a
, a = 0, 1, 2, 3, is the canonical basis of R
4
, is the connection
1-form corresponding to the translations.
Remark: The local (x
) dependence of the internal Lorentz and translational transformations is a

consequence of the general denition of a gauge transformation in bre bundle theory (Appendix F).
In terms of the B
a
elds and the spin connection, the Ricci scalar in section 32 is given by
R = (
v
a
x
x
v
b
x
x
+
v
a
x
x
b
+
v
b
x
x
a
+B
a
B
b
)(
ab

ab
+
ac
c
b

ac
c
b
). (41)
If one intends to use this Lagrangian density as describing a (B
a
,
bc
) (or (e
a
,
bc
)) interaction (Randono,
2010), then immediately faces the problem that the B
a
(or e
a
) does not have a free part (in particular a
kinematical part), since all its powers are multiplied by s or s. So an interpretation in terms of elds
interaction seems dicult, and may be, impossible.
35. Torsion and gauge invariance
It is well known the problem of the violation, in the presence of torsion, of the local gauge invariance
of theories like Maxwell and Yang-Mills due to the straightforward application of the minimal coupling
procedure to introduce the interaction with the gauge elds: the comma goes to semicolon rule. In fact,
as we shall show below, the prohibition of this procedure should only be applied to the denition of the
eld strengths F, as emphasized by Hammond (Hammond, 2002). For simplicity of the presentation we shall
restrict to the abelian case.
In special relativity, for the eld strength in terms of the gauge potential one has F = dA = d(A
dx
) =
(
)dx
dx
=
1
2
(
)dx
dx
= F
dx
dx
, which is clearly gauge independent: F(A) =

F(A +d). Replacing
by D
one obtains
F
= D
= (
) (
) = F
[]
A
= F
2T
. ()
When torsion vanishes, f
= F
i.e. A
;
A
;
= A
,
A
,
; when T
= 0, f
= F
and moreover,
f
is not gauge invariant: if A
, then f
with
f

g.tr.
(f
) = 2T
. ()
The equations () and () show that the acquired dependence on torsion of the classical electric and magnetic
elds, also depends on the chosen gauge (by the presence of ), what is inadmissible.
At this point, we criticize the solution given by some authors (Hehl et al, 1976; de Sabbata, 1997),
which consists in the assertion that torsion does not couple to the gauge eld. This statement would have
sense if also the Levi-Civita part of the connection would not couple, since both
LC
and torsion come
51
together in the sum =
LC
+K (appendix B), where torsion is the antisymmetric part of the contortion
K. However,
LC
does couple. Moreover, de Sabbata (de Sabbata, 1997) shows that at the microscopic
quantum level photons couple to torsion in a gauge invariant way through virtual pairs e
+
e
creation. There
is no reason to expect that in the transit to the classical limit the coupling should disappear; though, as we
shall see below, partly due to the absence of the intermediate fermion eld, gauge invariance breaks down.
A partial solution to this problem has been given by Benn, Dereli and Tucker (Benn et al, 1980), leaving,
as we show below, F = dA with F
= dA
= F if A
= A +d in a completely natural way.

Let A = A
dx
= A
a
e
a
be the connection 1-form of the electromagnetic eld, with A
a
= e
a
and
e
a
= e
a
dx
. If
a
b
=
a
b
dx
is the spin connection with

cb
=
ca
a
b
=
bc
, then the exterior covariant
derivative of A
a
with respect to
a
b
is given by
DA
a
= dA
a
b
a
A
b
( )
with dA
a
= dx
A
a
and
b
a
A
b
= A
b
b
a
dx
. ( ) gives the minimal coupling of the electromagnetic

connection with the space-time connection, i.e. dA
a
m.c.
DA
a
.
Exterior multiplication with e
a
gives
DA
a
e
a
= (da
a
b
a
A
b
) e
a
= (dA
a
) e
a
A
b
b
a
e
a
,
and using the expression for torsion T
a
= de
a
+
a
b
e
b
(section 29), we obtain
DA
a
e
a
= (dA
a
) e
a
A
b
(T
b
de
b
) = (dA
a
) e
a
A
b
T
b
+A
b
de
b
i.e.
DA
a
e
a
+A
b
T
b
= (dA
a
) e
a
+ A
a
de
a
= d(A
a
e
a
) = dA.
Then,
F = dA = DA
a
e
a
+A
b
T
b
. ( )
F is closed,
dF = d
2
A = 0
and, most important, U(1)-gauge invariant:
A A
= A +d =F F
= F +d
2
= F.
We notice however that DA
a
is not U(1)-gauge invariant: in fact, with d
a
= e
a
,
DA
a
(DA
a
)
= D(A
a
+d
a
) = d(A
a
+d
a
)
b
a
(A
b
+d
b
) = DA
a
+d(d
a
)
b
a
d
b
= DA
a
+ (
b
a
b
a
)d
b
i.e.
(DA
a
)
= DA
a
+D
b
a
d
b
with
D
b
a
=
b
a
d
b
a
.
Nevertheless, even with a gauge invariant eld strength, due to the non gauge invariance of the spin
density tensor of the electromagnetic eld, the solution of the Cartan equation gives a U(1)-gauge dependent
torsion, which points to a dicult (if not impossible) to cure illness of the EC theory. In fact, the Maxwell-
Einstein action describing the interaction between the electromagnetic eld and gravity is given by
S
ME
= l
_
d
4
x eL
ME
= l
_
d
4
x e(
1
4
F
)
52
with l =
G
c
4
. From the r.h.s. of ( ) and using ( ), we obtain the expression for F
:
(dA
a

b
a
) e
a
= ((
A
a
)e
b
a
A
b
e
a
)dx
dx
,
and
A
b
T
b
= A
b
T
b
dx
dx
,
then
F = ((
A
a
b
a
A
b
)e
a
+A
b
T
b
)dx
dx
and so
F
= 2((
[
A
a
b
[a
A
b
)e
]
a
+A
b
T
b
).
The Cartan equation results from the variation with respect to the spin connection of the total action
S
G
+S
ME
:
0 =
S
G
+
S
ME
.
The rst term was obtained in section 32.A.1.; for the second term, using the expression for T
b
in section
29,
1
4
(F
) =
1
2
F
= F
(
1
2
((
A
a
b
a
A
b
)e
a
(
A
a
b
a
A
b
)e
a
)
+A
b
(
b
+
b
a
e
b
a
e
a
))
= F
(
1
2
(
b
a
)A
b
e
a
+
1
2
(
b
a
)A
b
e
a
+A
b
((
b
a
)e
a
(
b
a
)e
a
))
= F
(
b
a
)A
b
e
a
2F
A
b
(
b
a
)e
a
= F
A
b
e
a
(
b
a
) = F
A
b
e
ba
= F
A
[b
e
a]
ba
where we used the antisymmetry
ba
=
ab
. Then
0 = e
a
T
b
+e
b
T
a
+T
ba
lF
A
[b
e
a]
.
Multiplying by e
b
e
a
we obtain
T
= lF
[
A
]
=
l
2
(F
) =
l
2
S
where
S
= F
is the canonical spin density tensor of the electromagnetic eld obtained from the gauge invariant Lagrangian
density L
ME
=
1
4
F
through the Noether theorem (Bogoliubov and Shirkov, 1980). S
is antisym-
metric in its lower indices but it is not gauge invariant:
g.tr.
(S
) = 2F
if
g.tr.
(A
) =
. In contradistinction with the density of energy-momentum T
of any matter eld, which

always can be made gauge invariant (and symmetric), there is no known way to construct a gauge invariant
spin density tensor for the electromagnetic eld. However, at least in the special relativistic classical and
quantum eld theory context, after space integration of S
0
, all the results for the conserved spin angular

momentum tensor of the electromagnetic eld are physical (light polarization, helicity states, etc.), and
therefore gauge independent. This means that S
is not directly observable, and then the same could be

concluded for the non gauge invariant torsion tensor produced by the electromagnetic spin.
By the same method of section 32 applied to the Dirac eld, for the torsion one obtains
T
=
l
2
(S
+
1
2
(
))
53
with S
= S
= F
.
Acknowledgements
This work was partially supported by the projects PAPIIT IN 118609, 113607-2, and 101711-2, DGAPA-
UNAM, Mexico. The author thanks for the hospitality at the IAFE-UBA-CONICET, Argentina.
References
Aldrovandi, R. and Pereira, J. G. (2007). An Introduction to Teleparallel Gravity, Instituto de Fsica
Te orica, UNESP, Sao Paulo, Brazil.
Ali, S. A., Cafaro, C., Capozziello, S., and Corda, Ch. (2009). On the Poincare Gauge Theory of
Gravitation, International Journal of Theoretical Physics 48, 3426-3448; arXiv: gr-qc/0907.0934.
Benn, I. M., Dereli, T., and Tucker, R. W. (1980). Gauge Field Interactions in Spaces with Arbitrary
Torsion, Physics Letters 96B, 100-104.
Bogoliubov, N. N. and Shirkov, D. V. (1980). Introduction to the Theory of Quantized Fields, Wiley,
50-51.
Carroll, S. (2004). Spacetime and Geometry. An Introduction to General Relativity, Addison Wesley,
San Francisco.
Cartan, E. (1922). Sur une generalization de la notion de courbure de Riemann et les espaces ` a torsion,
Comptes Rendus Acad. Sci., Paris, 174, 593-595.
Cheng, T.-P. (2010).Relativity, Gravitation and Cosmology, 2nd. ed., Oxford University Press; pp.
307-308.
de Sabbata, V. (1997). Evidence for torsion in gravity?, in Spin Gravity: Is it possible to give an
experimental basis to torsion?, International School of Cosmology and Gravitation, XV Course, Erice, Italy,
eds. P. G. Bergmann et al, World Scientic, 52-85.
Dirac, P. A. M. (1975). General Theory of Relativity, Princeton University Press, Princeton, New Jersey,
1996; pp. 15, 25.
DOlivo, G. and Socolovsky, M. (2011). Poincare gauge invariance of general relativity and Einstein-
Cartan theory; arXiv: gr-qc/1104.1657.
Einstein, A. (1956). The Meaning of Relativity, Chapman and Hall, p. 96.
Fabbri, L. (2010). Private communication.
Fabbri, L. (2011). On the Principle of Equivalence, in Einstein and Hilbert: Dark Matter, editor V. V.
Dvoeglazov, Nova Publishers; arXiv: gr-qc/0905.2541.
Gronwald, F., and Hehl, F. W. (1996). On the Gauge Aspects of Gravity, in Proceedings of the In-
ternational School of Cosmology and Gravitation: 14th Course: Quantum Gravity, Erice, Italy, 1995; P. G.
Bergmann et al eds., World Scientic, Singapore; arXiv: gr-qc/9602013.
Hammond, R. T. (2002). Torsion gravity, Reports on Progress in Physics 65, 599-649.
54
Hartley, D. (1995). Normal frames for non-Riemannian connections, Classical and Quantum Gravity
12, L103-L105.
Hayashi, K. (1977). The gauge theory of the translation group and underlying geometry, Physics Letters
69B, 441-444.
Hayashi, K. and Nakano, T. (1967). Extended Translation Invariance and Associated Gauge Fields,
Progress of Theoretical Physics 38, 491-507.
Hayashi, K., and Shirafuji, T. (1980). Gravity from Poincare Gauge Theory of the Fundamental Parti-
cles. I, Progress of Theoretical Physics 64, 866-882.
Hehl, F. W. (1985). On the Kinematics of the Torsion of Space-Time, Foundations of Physics 15,
451-471.
Hehl, F. W. (1998). Alternative Gravitational Theories in Four Dimensions, in Proceedings 8th M.
Grossmann Meeting, ed. T. Piran, World Scientic, Singapore; arXiv: gr-qc/9712096.
Hehl, F. W., von der Heyde, P., Kerlick, G. D., and Nester, J. M. (1976). General relativity with spin
and torsion: Foundations and prospects, Review of Modern Physics 48, 393-416.
Kobayashi, S. and Nomizu, K. (1963). Foundations of Dierential Geometry, Volume I, J. Wiley, New
York.
Leston, M. (2008). Simetras Innito-Dimensionales en Teoras de Gravedad, Tesis doctoral, FCEN-
UBA, Argentina.
Leston, M., and Socolovsky, M. (2011). The Status of Gravity as a Gauge Theory, in Einstein and
Hilbert: Dark Matter, ed. V. V. Dvoeglazov, Nova Scie. Pub, pp. 141-156.
ORaifearteigh, L. (1997). The Dawning of Gauge Theory, Princeton University Press, Princeton.
Randono, A. 2010. Gauge Gravity; a forward-looking introduction, arXiv: gr-qc/1010.5822.
Schroedinger, E. (1950). Space-Time Structure, Cambridge University Press, Cambridge.
Socolovsky, M. (2010). Locally inertial coordinates with totally antisymmetric torsion, arXiv: gr-
qc/1009.3979.
Trautman, A. (1973). On the structure of Einstein-Cartan equations, in Dierential Geometry, Symp.
Mathematics, Vol. 12, Academic Press, London, 139-162.
Utiyama, R. (1956). Invariant Theoretical Interpretation of Interaction, Physical Review 101, 1597-
1607.
Weyl, H. (1918). Gravitation und Elektrizit at, Sitzungsber. d. Preuss., Akad. d. Wissensch., 465.
(English translation in ORaifearteigh (1997), 24-37.)
Appendix A
Fundamental theorem of riemannian geometry
55
Denition: Let (M, g, ) be a riemannian manifold with a linear connection . and g are compatible
or, equivalently, is a metric connection in M, if for any smooth path c : (a, b) M, c(), and any
pair of parallel vector elds V and V
along c (i.e.
DV
d
=
DV
d
= 0), then g(V, V
) < V, V
> is constant
along c.
Theorem: In any riemannian manifold (M, g) there exists a unique symmetric linear connection , i.e.
a connection in the tangent bundle of M, which is compatible with the metric.
Proof:
Let X, Y, Z (TM), and let be a metric and symmetric linear connection in M. Then:
<
X
Y, Z >= X < Y, Z > < Y,
X
Z >= X < Y, Z >= X < Y, Z > < Y,
Z
X + [X, Z] >
= X < Y, Z > < Y,
Z
X > < Y, [X, Z] >;
Z < Y, X >=<
Z
Y, X > + < Y,
Z
, X > implies < Y,
Z
X >= Z < Y, X > <
Z
Y, X >
= Z < Y, X > <
Y
Z + [Z, Y ], X >; then:
<
X
Y, Z >= X < Y, Z > Z < Y, X > + <
Y
Z + [Z, Y ], X > < Y, [X, Z] >;
Y < Z, X >=<
Y
Z, X > + < Z,
Y
X > implies <
Y
Z, X >= Y < Z, X > < Z,
Y
X >
= Y < Z, X > < Z,
X
Y + [Y, X] >= Y < Z, X > <
X
Y, Z > < Z, [Y, X] >;
then:
<
X
Y, Z >= X < Y, Z > Z < Y, X > +Y < Z, X > <
X
Y, Z > < Z, [Y, X] >
+ < [Z, Y ], X > < [X, Z], Y >
and therefore
<
X
Y, Z >=
1
2
(X < Y, Z > Z < X, Y > +Y < Z, X > < X, [Y, Z] > + < Y, [Z, X]+ < Z, [X, Y ])
=
1
2
(X < Y, Z > < X, [Y, Z] > +Y < Z, X > + < Y, [Z, X] > Z < X, Y > + < Z, [X, Y ]),
which gives an explicit expression for <
X
Y, Z > in terms of X, Y, Z and < , >= g.
In local coordinates, we choose X =
i
, Y =
j
, Z =
k
; then [
i
,
j
] = 0 and therefore
<
i
j
,
k
>=<
l
ij
l
,
k
>=
l
ij
<
l
,
k
>=
l
ij
g(
l
,
k
) =
l
ij
g
lk
=
1
2
(
i
<
j
,
k
> +
j
<
k
,
i
>
k
<
i
,
j
>) =
1
2
(
i
g
jk
+
j
g
ki
k
g
ij
); multiplying by g
mk
= (g
1
)
mk
,
l
ij
g
mk
g
lk
=
l
ij
m
l
=
m
ij
and therefore
m
ij
=
1
2
g
mk
(
i
g
jk
+
j
g
ik
k
g
ij
). (qed)
Clearly,
m
ij
=
m
ji
.
Remark: The theorem is also valid in pseudo-riemannian geometry; in particular for lorentzian manifolds.
56
Note: In m dimensions, the number of independent components of the metric tensor and the Levi-Civita
connection are N(g
; m) =
m(m+1)
2
and N(
; m) =
m
2
(m+1)
2
.
Appendix B
General form of the local version of a non (necessarily) metric and non (necessarily)
symmetric connection
Given a linear connection in a manifold M
n
(L
n
-space), with local Christoel symbols
, if in addition
the manifold is riemannian or pseudo-riemannian (M
n
, g) (V
n
-space), one has a (L
n
, g)-space. The non-
metricity tensor is dened as minus the covariant derivative of the metric:
Q
= D
.
(Q
= Q
.) Using section 7, by cyclic permutation of indices, one obtains
= (
LC
)
+K
+
1
2
g
(Q
+Q
)
where (
LC
)
is the Levi-Civita connection of section 13, and

K
= T
+g
+g
is the contortion tensor, with

K
= (K
A
)
+ (K
S
)
,
(K
A
)
= T
= (K
A
)
,
(K
S
)
= (K
S
)
= g
(T
+T
).
A metric connection is one in which Q
= 0 i.e. the connection is compatible with the metric, but non

necessarily symmetric:
=
LC
+contortion.
(U
n
-space.) In particular, scalar products and lengths of vectors are constant in parallel transport. (In fact,
||V ||
2
,
= (g
)
,
= g
,
V
+2g
,
= 2V
(V
,
+
) = 2V
;
= 0.) A physical case
corresponds to the Einstein-Cartan theory of gravity. (Cartan, 1922.)
A symmetric connection is one in which torsion vanishes i.e. T
= 0:
=
LC
+non metricity.
A particular case is the Weyl connection (1918):
non metricity = Q
= g
with A = A
dx
a 1-form. (In natural units, [A
] = [mass] if [x
] = [length].)
Appendix C
1. If a = (a
ij
) is an invertible matrix with a
ij
= a
ij
(x), then
(det a) = a
ij,
(det a)a
ij
, where
a
ij
= (a
1
)
ij
. In particular, for the metric tensor,
g
,
= g
,
gg
57
where g = detg
. Then, for the Levi-Civita connection,

(
LC
)
=
1
2
g
) =
1
2
g
g
,
=
1
2
g
1
(gg
g
,
) =
1
2g
g
,
=
1
2(g)
(g)
,
=
1
(g)
1
2
(g)
1
2
=
.
2. From 1.,
2
g
g
,
; then,
= (
)x
2
g
g
,
x
2
g
. Using g
ab
e
a
e
b
(section 28.6), we obtain g
= 2e
d
d
and so e = ee
d
d
with e =
. From e
d
f
=
f
d
, e = ee
d
e
d
.
Appendix D
Lorentz gauge invariance of the Ricci scalar
(This proof is due to G. DOlivo.)
The Ricci scalar is given by
R =
bd
e
a
e
d
a
b
a
b
+
a
c
c
b
a
c
c
b
)
bd
e
a
e
d
((3) (4) + (1) (2)).

Under the transformation
a
c
= h
c
l
r
l
h
1
r
a
+ (h
c
l
)h
1
l
a
we have:
(1) = (a) + (b) + (c) + (d) with
(a) = h
c
l
r
l
h
1
r
a
h
b
g
s
g
h
1
s
c
, (b) = h
c
l
r
l
h
1
r
a
(
h
b
g
)h
1
g
c
,
(c) = h
b
g
s
g
h
1
s
c
(
h
c
l
)h
1
l
a
, (d) = (
h
c
l
)h
1
l
a
(
h
b
g
)h
1
g
c
;
(2) = (e) + (f) + (g) + (h) with
(e) = h
c
g
s
g
h
1
s
a
h
b
l
r
l
h
1
r
c
, (f) = h
c
g
s
g
h
1
s
a
(
h
b
l
)h
1
l
c
,
(g) = h
b
l
r
l
h
1
r
c
(
h
c
g
)h
1
g
a
, (h) = (
h
c
l
)h
1
l
a
(
h
b
g
)h
1
g
c
;
(3) = [1] + [2] + [3] + [4] with
[1] = h
b
n
h
1
t
a
(
t
n
), [2] =
t
n
(h
b
n
h
1
t
a
), [3] = (
h
b
n
)h
1
n
a
, [4] = (
h
b
n
)(
h
1
n
a
);
and (4) = [5] + [6] + [7] + [8] with
[5] = h
b
l
h
1
s
a
(
s
l
), [6] =
s
l
(h
b
l
h
1
s
a
), [7] = (
h
b
l
)h
1
l
a
, [8] = (
h
b
l
)(
h
1
l
a
).
Now,
[3] [7] = (
h
b
n
)h
1
n
a
(
h
b
l
)h
1
l
a
= 0,
(b) + (c) =
r
l
h
1
r
a
h
b
l
s
g
h
b
g
h
1
s
a
,
(f) + (g) =
s
g
h
1
s
a
h
b
g
r
l
h
b
l
h
1
r
a
;
58
so
((b) + (c)) ((f) + (g)) =
r
l
(h
1
r
a
h
b
l
)
s
g
(h
1
s
a
h
b
g
);
also,
[2] [6] =
s
g
(h
b
g
h
1
s
a
)
r
l
(h
b
l
h
1
r
a
);
then
((b) + (c)) ((f) + (g)) + ([2] [6]) = 0.
Also,
[4] [8] = (
h
b
l
)(
h
1
l
a
) (
h
b
l
)(
h
1
l
a
)
and
(d) (h) = (
h
1
l
a
)(
h
b
l
) (
h
1
l
a
)(
h
b
l
);
so
([4] [8]) + ((d) (h)) = 0.
Finally,
[1] [5] + (a) (e) = h
b
l
h
1
s
a
(
s
l
s
l
+
s
r
r
l
s
r
r
l
).
Therefore,
R =
bd
e
a
e
d
h
b
l
h
1
s
a
(
s
l
s
l
+
s
r
r
l
s
r
r
l
) =
lt
e
s
e
t
s
l
s
l
+
s
r
r
l
s
r
r
l
)
= R
. (qed)
Appendix E
Ane spaces
An ane space is a triple (V, , A) where V is a vector space, A is a set, and is a free and transitive
left action of V as an additive group on A:
: V A A, (v, a) v +a,
with
0 +a = a and (v
1
+v
2
) +a = v
1
+ (v
2
+a), for all a A and all v
1
, v
2
V.
Then, given a, a
A, there exists a unique v V such that a
= v+a. Also, if v
0
is xed in V ,
v0
: A A,
v0
(a) = (v
0
, a) is a bijection.
Example. A = V : The vector space itself is considered as the set on which V acts. In particular, when
V = T
x
M
n
and A = T
x
M
n
, the tangent space is called ane tangent space and denoted by A
x
M
n
. The
points a of A
x
M
n
are the tangent vectors at x.
59
It is clear that to dene an action of GA
n
(R) on A
x
M
n
, we need a frame at x i.e. we have to consider
the bundle of ane frames AM
n
so that (v
x
, r
x
)(, g) = (v
x
+r
x
, r
x
g).
Appendix F
Gauge transformations in G-bundles
A gauge transformation or vertical automorphism of a (smooth) principal G-bundle : G P

B is
a dieomorphism : P P such that the following diagram commutes:
P G
IdG
P G

P

P

B
IdB
B
( is the action of G on P.) That is:
= ( Id
G
) i.e. (pg) = (p)g
and
= i.e. ((p)) = (p).
So, (p) = ph with h G.
The set of gauge transformations of , G(), is called the gauge group of the bundle.
Local form of
A local trivialization of is given by the commutative diagram
P
U
U
U G

1
U
IdU
U
i.e.
1

U
= , where P
U
=
1
(U),
U
is a dieomorphism, U is an open subset of B, and
1
(p, g) = p.
U
denes the local section of ,
U
: U P
U
,
U
(b) =
1
U
(b, e) where e is the identity in G. Then
there exists the smooth function
U
: U G, b
U
(b)
which determines for all p P
U
. In fact, let p =
U
(b); then
(
U
(b)) =
U
(b)g P
b
=
1
({b}),
and so
U
(b) = g
with g unique since acts freely on P and transitively on bers. If p
P
b
, then p
=
U
(b)h and
(p
) = (
U
(b)h) = (
U
(b))h = (
U
(b)
U
(b))h =
U
(b)(
U
(b)h). This holds for all b U.
e-mail: socolovs@nucleares.unam.mx
60

Fibre Bundles, Connections, General Relativity, and Einstein-Cartan Theory 1110.1018v1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fibre Bundles, Connections, General Relativity, and Einstein-Cartan Theory 1110.1018v1

Uploaded by

Copyright:

Available Formats

a

, the torsion tensor T

and the non-metricity tensor Q

and ane connection

is the covariant derivative with respect to

.) GR has to do with a metric symmetric connection,

; EC theory involves a metric but not necessarily

; while the theory of Weylian manifolds involves a non

= 0) and non necessarily symmetric (T

= 0) connection. (In units of length [L],

= 0 is equivalent to Ricci atness:

and the scalar curvature R. Then, G

: (Local) Einstein equations in empty space-time

(M, R): smooth real valued functions on M

(M, R)-linear in the second

M E) be the set of dierential 1-forms in M with values in E. T

(M, R)-linear i.e. s(fX) = fs(X); however,

M E, x (s)(x) = (x, (s)

be a local coordinate basis of (TU

) i.e. if X (TU) then X = X

R, called the Christoel symbols of the connection in the

is due to the Leibnitz rule.

for a unique set of n

is not necessarily symmetric or antisymmetric

is the covariant derivative of the local vector eld Y

is a local vector eld and A

is a local dierential 1-form in M

. The covariant derivative of a scalar is

is the covariant derivative of the local 1-form A

(M, R)-linearity in X are trivial; nally,

under a change of local coordinates (charts) in M

-functions and using

which implies, by the linear independence of the coordinate tangent elds

functions do not transform as tensors.

) which overlaps with U

) becomes the set

(x) in the overlap U

(M, R)-multilinear map from the

(M, R). On a chart U

(M, R)-module of r-contravariant and s-covariant tensors on M

M): dierential 1-forms on M. In general,

map (r, s)-tensors

where denotes the contractions.

where the dependence of is through c, and for a dierential 1-form A

= 0 and then the solutions of the geodesic equation

R : (TM) (TM) End

we contract with we obtain the antisymmetric tensor

is non symmetric, not even when

. In this case, however, its antisymmetric part

does not require a symmetric connection.

with R R : (TM) (TM) (TM) (TM), (X, Y, Z) R(X, Y, Z) = ([

(see section 13),

(M, R)-linear in X and Y i.e. T(fX, Y ) = T(X, fY ) = fT(X, Y ).

is a tensor. In particular one has

is the torsion vector (in fact, it is a 1-form).

, not necessarily part of geodesics, be curves which join p with

; if the transport is parallel then

is the innitesimal area

enclosed by the curves c and c

); regarded as innitesimal displacements

one obtains the vector at q, V

ME) with (s)(x) = (x, (s)

, then the Bianchi equations are: