Lorentz Connections and Gravitation

J. G. Pereira
Instituto de Fı́sica Teórica, UNESP-Univ Estadual Paulista
Caixa Postal 70532-2, 01156-970 São Paulo, Brazil

Abstract. The different roles played by Lorentz connections in general relativity and in telepar-
allel gravity are reviewed. Some of the consequences of this difference are discussed.

1 Introduction
arXiv:1210.0379v1 [gr-qc] 1 Oct 2012

A key point of gravitation is that the metric tensor defines neither curvature nor torsion by
itself [1]. As a matter of fact, curvature and torsion require a connection to be defined, and
many different connections, with different curvature and torsion, can be defined on the very
same metric spacetime [2]. How can we determine the relevant connection for gravitation? This
is a fundamental question, which has more than one answer. For example, when constructing
general relativity, Einstein chose the zero–torsion Levi–Civita, or Christoffel connection, which
is a connection completely specified by the ten components of the metric tensor. In this theory,
therefore, torsion is chosen to vanish from the very beginning, and the gravitational field is
represented by curvature.
A second possibility would be to choose a zero–curvature Lorentz connection not related
to gravitation, but to inertial effects only. The gravitational theory that emerges from this
choice is teleparallel gravity, a gauge theory for the translation group, in which curvature is
assumed to vanish from the very beginning. In this theory, the gravitational field turns out to
be represented by a translational gauge potential, which appears as the non–trivial part of the
tetrad field and gives rise to a non–vanishing torsion, the field strength of the theory. One may
wonder why a gauge theory for the translation group, and not for any other group related to
spacetime. The answer is related to the source of gravitation: energy and momentum. From
Noether’s theorem, an instrumental piece of gauge theories [3], we know that the energy–
momentum tensor is conserved provided the source lagrangian is invariant under spacetime
translations. If gravity is to be described by a gauge theory with energy–momentum as source,
therefore, it must be a gauge theory for the translation group [4].
Although equivalent to general relativity, teleparallel gravity provides a new insight into
gravitation. The purpose of these lectures is to explore some of these insights, as well as discuss
how this approach could help to answer some old questions permeating general relativity, like
for example the energy localizability of the gravitational field and the problem of quantum
gravity.

2 Linear Frames and Tetrads

Spacetime is the common arena on which the four presently known fundamental interactions
manifest themselves. Electromagnetic, weak and strong interactions are described by gauge
theories involving transformations taking place in internal spaces, by themselves unrelated to
spacetime. The basic setting of gauge theories are the principal bundles, in which a copy of
the corresponding gauge group is attached at each point of spacetime — the base space of the

1
bundle. Gravitation, on the other hand, is deeply linked to the very structure of spacetime. The
geometrical setting of gravitation is the tangent bundle, a natural construction always present
in any differentiable manifold: at each point of spacetime there is a tangent space attached to
it — the fiber of the bundle, which is seen as a vector space. In what follows we are going to
use the Greek alphabet (µ, ν, ρ, · · · = 0, 1, 2, 3) to denote indices related to spacetime, and the
first letters of the Latin alphabet (a, b, c, · · · = 0, 1, 2, 3) to denote indices related to the tangent
space, a Minkowski spacetime whose Lorentz metric, in cartesian coordinates, is assumed to
have the form
ηab = diag(+1, −1, −1, −1). (1)
A general spacetime is a 4-dimensional differential manifold, indicated R3,1 , whose tangent
space is, at each point, a Minkowski spacetime. Spacetime coordinates will be denoted by
{xµ }, whereas tangent space coordinates will be denoted by {xa }. Such coordinate systems
determine, on their domains of definition, local bases for vector fields, formed by the sets of
gradients
{∂µ } ≡ {∂/∂xµ } and {∂a } ≡ {∂/∂xa }, (2)
as well as bases {dxµ } and {dxa } for covector fields, or differentials. These bases are dual, in
the sense that
dxµ (∂ν ) = δνµ and dxa (∂b ) = δba . (3)
On the respective domains of definition, any vector or covector can be expressed in terms
of these bases, which can furthermore be extended by direct product to constitute bases for
general tensor fields of any order.

2.1 Trivial Frames

Trivial frames, or trivial tetrads, will be denoted by

{ea } and {ea }. (4)

The above mentioned coordinate bases

{ea } = {∂a } and {ea } = {dxa } (5)

are very particular cases, whose name stems from their relationship to a coordinate system.
Any other set of four linearly independent fields {ea } will form another basis, and will have a
dual {ea } whose members are such that

ea (eb ) = δba . (6)

Notice that, on a general manifold, vector fields are (like coordinate systems) only locally
defined — and linear frames, as sets of four such fields, are only defined on restricted domains.
These frame fields are the general linear bases on the spacetime differentiable manifold
R 3,1 . The whole set of such bases, under conditions making of it also a differentiable manifold,
constitutes the bundle of linear frames. A frame field provides, at each point p ∈ R 3,1 , a basis
for the vectors on the tangent space Tp R 3,1 . Of course, on the common domains they are
defined, each member of a given base can be written in terms of the members of any other.
For example,
ea = ea µ ∂µ and ea = ea µ dxµ , (7)

2
and conversely,
∂µ = ea µ ea and dxµ = ea µ ea . (8)
On account of the orthogonality conditions (6), the frame components satisfy

ea µ ea ν = δµν and ea µ eb µ = δba . (9)

Notice that these frames, with their bundles, are constitutive parts of spacetime: they are
automatically present as soon as spacetime is taken to be a differentiable manifold.
A general linear base {ea } satisfies the commutation relation

[ea , eb ] = f c ab ec , (10)

with f c ab the so–called structure coefficients, or coefficients of anholonomy, or still the an-
holonomy of frame {ea }. As a simple computation shows, they are defined by

f c ab = ea µ eb ν (∂ν ec µ − ∂µ ec ν ). (11)

A preferred class is that of inertial frames, denoted e′a , those for which

f ′a cd = 0. (12)

Base {e′a } is then said to holonomic. Of course, all coordinate bases are holonomic. This is not
a local property, in the sense that it is valid everywhere for frames belonging to this inertial
class.
Consider now the Minkowski spacetime metric, which in cartesian coordinates has the form

ηµν = diag(+1, −1, −1, −1). (13)

In any other coordinates, ηµν will be a function of the spacetime coordinates. The linear frame

ea = ea µ ∂µ , (14)

provides a relation between the tangent–space metric ηab and the spacetime metric ηµν . Such
relation is given by
ηab = ηµν ea µ eb ν . (15)
Using the orthogonality conditions (9), the inverse relation is found to be

ηµν = ηab ea µ eb ν . (16)

Independently of whether ea is holonomic or not, or equivalently, whether they are inertial or

not, they always relate the tangent Minkowski space to a Minkowski spacetime. These are the
frames appearing in special relativity, and are usually called trivial frames — or trivial tetrads.

2.2 Nontrivial Frames

Nontrivial frames, or tetrads, will be denoted by

{ha } and {ha }. (17)

They are defined as linear frames whose coefficient of anholonomy is related to both inertial
effects and gravitation. Let us consider a general pseudo–riemannian spacetime with metric
components gµν in the some dual holonomic basis {dxµ }. The tetrad field

ha = ha µ ∂µ and ha = ha µ dxµ , (18)

3
is a linear basis that relates gµν to the tangent–space metric ηab through the relation

ηab = gµν ha µ hb ν . (19)

The components of the dual basis members ha = ha ν dxν satisfy

ha µ ha ν = δµν and ha µ hb µ = δba , (20)

so that Eq. (19) has the inverse

gµν = ηab ha µ hb ν . (21)
We see from these relations that
√
h = det(ha µ ) = −g , (22)

with g = det(gµν ).
A tetrad basis {ha } satisfies the commutation relation

[ha , hb ] = f c ab hc , (23)

with f c ab the structure coefficients, or coefficients of anholonomy, of frame {ha }. The basic
difference in relation to the linear bases {ea } is that now the f c ab represent both inertia and
gravitation. As before, the structure coefficients are given by

f c ab = ha µ hb ν (∂ν hc µ − ∂µ hc ν ). (24)

Although nontrivial tetrads are, by definition, anholonomic due to the presence of gravitation,
it is still possible that locally, f c ab = 0. In this case, dha = 0, which means that ha is locally a
closed differential form. In fact, if this holds at a point p, then there is a neighborhood around
p on which functions (coordinates) xa exist such that

ha = dxa .

We say that a closed differential form is always locally integrable, or exact. This is the case
of locally inertial frames, which are always holonomic. In these frames, inertial effects locally
compensate gravitation.

3 Lorentz Connections
A Lorentz connection Aµ , frequently referred to also as spin connection, is a 1-form assuming
values in the Lie algebra of the Lorentz group,
1
Aµ = 2 Aab µ Sab , (25)

with Sab a given representation of the Lorentz generators. As these generators are antisym-
metric in the algebraic indices, Aab µ must be equally antisymmetric in order to be lorentzian.
This connection defines the Fock–Ivanenko covariant derivative [5, 6]

D µ = ∂µ − 2i Aab µ Sab , (26)

whose second part acts only on the algebraic, or tangent space indices. For a scalar field φ, for
example, the generators are
Sab = 0. (27)

4
For a Dirac spinor ψ, they are given by [7]
i
Sab = 4 [γa , γb ] , (28)

with γa the Dirac matrices. A Lorentz vector field φc , on the other hand, is acted upon by the
vector representation of the Lorentz generators, matrices Sab with entries [8]

(Sab )c d = i (ηbd δac − ηad δbc ) . (29)

The Fock–Ivanenko derivative is, in this case,

D µ φc = ∂µ φc + Ac dµ φd , (30)

and so on for any other fundamental field.

On account of the soldered character of the tangent bundle, a tetrad field relates tangent
space (or internal) tensors with spacetime (or external) tensors. For example, if φa is an
internal, or Lorentz vector, then
φρ = ha ρ φa (31)
will be a spacetime vector. Conversely, we can write

φa = ha ρ φρ . (32)

On the other hand, due to its non–tensorial character, a connection will acquire a vacuum,
or non–homogeneous term, under the same operation. For example, to each spin connection
Aa bµ , there is a corresponding general linear connection Γρ νµ , given by

Γρ νµ = ha ρ ∂µ ha ν + ha ρ Aa bµ hb ν ≡ ha ρ D µ ha ν , (33)

where D µ is the covariant derivative (30), in which the generators act on internal (or tangent
space) indices only. The inverse relation is, consequently,

Aa bµ = ha ν ∂µ hb ν + ha ν Γν ρµ hb ρ ≡ ha ν ∇µ hb ν , (34)

where ∇µ is the standard covariant derivative in the connection Γν ρµ , which acts on external
indices only. For a spacetime vector φν , for example, it is given by

∇µ φν = ∂µ φν + Γν ρµ φρ . (35)

Using relations (31) and (32), it is easy to verify that [9]

D µ φd = hd ρ ∇µ φρ . (36)

Equations (33) and (34) are simply different ways of expressing the property that the total
covariant derivative of the tetrad — that is, a covariant derivative with connection terms for
both internal and external indices — vanishes identically:

∂µ ha ν − Γρ νµ ha ρ + Aa bµ hb ν = 0. (37)

5
3.1 Curvature and Torsion
Curvature and torsion require a Lorentz connection to be defined [1]. Given a Lorentz connec-
tion Aa bµ , the corresponding curvature is a 2-form assuming values in the Lie algebra of the
Lorentz group,
Rνµ = 12 Rab νµ Sab . (38)
Torsion is also a 2-form, but assuming values in the Lie algebra of the translation group,

Tνµ = T a νµ Pa , (39)

with Pa = ∂a the translation generators. The curvature and torsion components are given,
respectively, by
Ra bνµ = ∂ν Aa bµ − ∂µ Aa bν + Aa eν Ae bµ − Aa eµ Ae bν (40)
and
T a νµ = ∂ν ha µ − ∂µ ha ν + Aa eν he µ − Aa eµ he ν . (41)
Through contraction with tetrads, these tensors can be written in spacetime–indexed forms:

Rρ λνµ = ha ρ hb λ Ra bνµ , (42)

and
T ρ νµ = ha ρ T a νµ . (43)
Using relation (34), their components are found to be

Rρ λνµ = ∂ν Γρ λµ − ∂µ Γρ λν + Γρ ην Γη λµ − Γρ ηµ Γη λν (44)

and
T ρ νµ = Γρ µν − Γρ νµ . (45)
Since the spin connection Aa bν is a four–vector in the last index, it satisfies

Aa bc = Aa bν hc ν . (46)

It can thus be verified that, in the anholonomic basis {ha }, the curvature and torsion compo-
nents are given respectively by

Ra bcd = hc (Aa bd ) − hd (Aa bc ) + Aa ec Ae bd − Aa ed Ae bc − f e cd Aa be (47)

and
T a bc = Aa cb − Aa bc − f a bc , (48)
where, we recall, hc = hc µ ∂µ . Use of (48) for three different combinations of indices gives

Aa bc = 21 (fb a c + Tb a c + fc a b + Tc a b − f a bc − T a bc ). (49)

This expression can be rewritten in the form

◦
Aa bc = Aa bc + K a bc , (50)

where ◦
Aa bc = 1
2 (fb a c + fc a b − f a bc ) (51)

6
is the usual expression of the general relativity spin connection in terms of the coefficients of
anholonomy, and
K a bc = 12 (Tb a c + Tc a b − T a bc ) (52)
is the contortion tensor.
Equation (50) is actually the content of a theorem, which states that any Lorentz connection
can be decomposed into the spin connection of general relativity plus the contortion tensor
[10]. The corresponding expression in terms of the spacetime–indexed linear connection reads
◦
Γρ µν = Γρ µν + K ρ µν , (53)

where ◦
σ 1
Γ µν = 2 gσρ (∂µ gρν + ∂ν gρµ − ∂ρ gµν ) (54)
is the zero–torsion Christoffel, or Levi–Civita connection, and

K ρ µν = 1
2 (Tν ρ µ + Tµ ρ ν − T ρ µν ) (55)

is the spacetime–indexed contortion tensor.

3.2 Behavior Under Lorentz Transformations

A local Lorentz transformation is fundamentally a transformation of the tangent space coor-
dinates xa :
x′a = Λa b (x) xb . (56)
Under such a transformation, the tetrad transforms according to

h′a = Λa b (x) hb . (57)

At each point of a riemannian spacetime, Eq. (21) only determines the tetrad up to transfor-
mations of the six–parameter Lorentz group in the tangent space indices. This means that
there exists actually an infinity of tetrads ha µ , each one relating the spacetime metric gµν to
the tangent space metric ηcd by Eqs. (19) and (21). In fact, any other Lorentz–rotated tetrad
{h′a } will also relate the same metrics

gµν = ηcd h′c µ h′d ν . (58)

Under a local Lorentz transformation Λa b (x), the spin connection undergoes the transformation

A′a bµ = Λa c (x) Ac dµ Λb d (x) + Λa c (x) ∂µ Λb c (x). (59)

Of course, both curvature Ra bνµ and torsion T a νµ transform covariantly:

R′a bνµ = Λa c (x) Λb d (x) Rc dνµ and T ′a νµ = Λa b (x) T b νµ . (60)

3.3 Purely Inertial Lorentz Connection

In special relativity, Lorentz connections represent only inertial effects present in a given frame.
In the class of inertial frames, where these effects are absent, the Lorentz connection vanishes
identically. Since this is the class most used in field theory, Lorentz connections do not routinely
show up in relativistic physics. Of course, as long as physics is frame independent, it can be

7
described in any class of frames. For the sake of simplicity, however, one always uses the class
of inertial frames when dealing with non–gravitational physics.
To see how an inertial Lorentz connection shows up, let us denote by ea µ a generic frame
in Minkowski spacetime. The class of inertial (or holonomic) frames, defined by all frames for
which f ′c ab = 0, will be denoted by e′a µ . In a general coordinate system, the frames belonging
to this class have the holonomic form
e′a µ = ∂µ x′a , (61)
with x′a a spacetime–dependent Lorentz vector: x′a = x′a (xµ ). The spacetime metric
′
ηµν = e′a µ e′b ν ηab (62)
still represents the Minkowski metric, but in a general coordinate system. In the specific case
of cartesian coordinates, the inertial frame assumes the form
e′a µ = δµa (63)
′ is that given by Eq. (13). Under a local Lorentz transformation,
and the spacetime metric ηµν
the holonomic frame (61) transforms according to
ea µ = Λa b (x) e′b µ . (64)
As a simple computation shows, it has the explicit form
• •
ea µ = ∂µ xa + Aa bµ xb ≡ D µ xa , (65)
where •
Aa bµ = Λa e (x) ∂µ Λb e (x) (66)
is a Lorentz connection that represents the inertial effects present in the new frame ea µ . As
can be seen from Eq. (59), it is just the connection obtained from a Lorentz transformation of
•
the vanishing spin connection A′e dµ = 0:
• •
Aa bµ = Λa e (x) A′e dµ Λb d (x) + Λa e (x) ∂µ Λb e (x). (67)
Starting from an inertial frame, different classes of frames are obtained by performing local
(point–dependent) Lorentz transformations Λa b (xµ ). Inside each class, the infinitely many
frames are related through global (point–independent) Lorentz transformations, Λa b = con-
stant.
The inertial connection (66) is sometimes referred to as the Ricci coefficient of rotation [11].
Due to its presence, the transformed frame ea µ is no longer holonomic. In fact, its coefficient
of anholonomy is given by   • •
f c ab = − Ac ab − Ac ba , (68)
• •
where we have used the identity Aa bc = Aa bµ ec µ . The inverse relation is
•
a 1
A bc = 2 (fb a c + fc a b − f a bc ) . (69)
•
Of course, as a purely inertial connection, Aa bµ has vanishing curvature and torsion:
• • • • • • •
a
R bνµ ≡ ∂ν Aa bµ − ∂µ Aa bν + Aa eν Ae bµ − Aa eµ Ae bν = 0 (70)
and • • •
a
T νµ ≡ ∂ν ea µ − ∂µ ea ν + Aa eν ee µ − Aa eµ ee ν = 0. (71)

8
3.4 Equation of Motion of Free Particles
As a concrete example, let us consider the equation of motion of a free particle. In the class
of inertial frames e′a µ , such particle is described by the equation of motion
du′a
= 0, (72)
dσ
with u′a the particle four–velocity, and

dσ 2 = ηµν dxµ dxν (73)

the quadratic Minkowski invariant interval. In a anholonomic frame ea µ , related to e′a µ by the
local Lorentz transformation (64), the equation of motion assumes the manifestly covariant
form
dua •
+ Aa bµ ub uµ = 0, (74)
dσ
where
ua = Λa b (x) u′b (75)
is the Lorentz transformed four–velocity, with

uµ = ua ea µ (76)

the spacetime–indexed four–velocity, which has the usual holonomic form

dxµ
uµ = . (77)
dσ
Observe that the inertial forces coming from the frame non–inertiality are represented by the
inertial connection of the left–hand side, which is of course non–covariant.

4 General Relativity
General Relativity conceives the gravitational interaction as a change in the geometry of space-
time itself. Specifically, as a change from the Lorentz metric ηµν of Minkowski space into a
riemannian metric gµν . This new metric plays the role of basic field, and is in principe de-
fined everywhere. Derivatives compatible with this overall presence of the same metric must
preserve it, must parallel–transport it everywhere. Of all such Lorentz connections preserving
gµν , the most natural choice from the point of view of universality is to pick up the Christoffel,
or Levi–Civita connection
◦
Γσ µν = 1
2 gσρ (∂µ gρν + ∂ν gρµ − ∂ρ gµν ) , (78)

which is a connection determined solely by the ten components of the metric tensor gµν . It is
the only metric–preserving connection with vanishing torsion, a magnitude which is then found
not to play any role in the general–relativistic description of the gravitational interaction. The
corresponding spin connection is
◦ ◦
a
A bµ = ha ν ∂µ hb ν + ha ν Γν ρµ hb ρ . (79)

Its Riemann curvature,

◦ ◦ ◦ ◦ ◦ ◦ ◦
Ra bνµ = ∂ν Aa bµ − ∂µ Aa bν + Aa eν Ae bµ − Aa eµ Ae bν , (80)

9
represents the fundamental field of the theory: gravitation is present whenever at least one of
its components is non–vanishing.
The field equation governing the dynamics of general relativity is Einstein equation
◦ ◦
Ra ν − 21 R ha ν = k Θa ν , (81)

where k = 8πG/c4 ,
◦ ◦ ◦ ◦
a
R ν = Rρa ρν and R = ha ν Ra ν (82)
are, respectively, the Ricci and the scalar curvature, and
1 δLs
Θa ν = − √ (83)
−g δha ν
is the symmetric source energy–momentum tensor modified by the presence of gravitation,with
Ls the source field lagrangian. This equation can be obtained from the lagrangian
◦
L = L + Ls , (84)

where
1 √
◦ ◦
L=−−g R (85)
2k
is the Einstein–Hilbert lagrangian of general relativity.

4.1 How Does General Relativity Describe Gravitation?

The curvature of the Levi–Civita connection gives rise to a geometric description of the grav-
itational interaction. To understand what such a geometrical description does mean, let us
consider the motion of a (spinless) particle in a gravitational field. In Minkowski spacetime,
such a particle obeys the equation
dua
= 0, (86)
dσ
with dσ the Minkowski invariant interval (73). To obtain the equation valid in the presence
of gravitation, a rule turns up, which is reminiscent of the gauge prescription: the minimal
coupling prescription. According to this prescription, all ordinary derivatives must be replaced
by covariant derivatives. In the specific case of general relativity, the free equation of motion
(86) becomes the usual geodesic equation
dua ◦
+ Aa bν ub uν = 0, (87)
ds
where
ds2 = gµν dxµ dxν (88)
is the riemannian spacetime quadratic interval. This equation describes the motion of a test
particle in the presence of a gravitational field. It says essentially that the four–acceleration
of the particle vanishes:
◦
aa = 0. (89)
This means that in General Relativity there is no concept of gravitational force. In this theory,
the gravitational interaction is geometrized: the presence of gravitation produces a curvature
in spacetime, and the gravitational interaction is described by letting particles to follow freely
the spacetime curvature.

10
5 Teleparallel Gravity
Teleparallel gravity corresponds to a gauge theory for the translation group. Accordingly, the
gravitational field is represented by a translational gauge potential B a µ , a 1-form assuming
values in the Lie algebra of the translation group:

Bµ = B a µ Pa . (90)

It appears as the non–trivial part of the tetrad,

ha µ = ea µ + B a µ , (91)

with • •
ea µ ≡ D µ xa = ∂µ xa + Aa bµ xb (92)
the trivial (non–gravitational) tetrad. Under a gauge translation

δxa = εa , (93)

the potential B a µ transforms according to

•
δB a µ = − D µ εa . (94)

The tetrad is consequently gauge invariant:

δha µ = 0. (95)

The field strength of teleparallel gravity is a 2-form assuming values in the Lie algebra of
the translation group. In a general Lorentz frame its components are given by
• • •
T a µν = ∂µ B a ν − ∂ν B a µ + Aa bµ B b ν − Aa bν B b µ , (96)

or equivalently
• • •
a
T µν = Dµ B aν − Dν B aµ . (97)
Since • • • •
a a
Dµ D ν x − Dν D µ x = 0, (98)
it can be rewritten in the form
• • •
a
T µν = D µ ha ν − D ν ha µ . (99)

We see in this way that the field strength is nothing else, but torsion. On account of the gauge
invariance of the tetrad, the field strength is also invariant under gauge transformations:
• •
T
′a
µν = T a µν . (100)

This is an expected result. In fact, considering that the generators of the adjoint representation
are the coefficients of structure of the group taken as matrices, and considering that these
coefficients vanish for abelian groups, fields belonging to the adjoint representations of abelian
gauge theories will always be gauge invariant.

11
5.1 Teleparallel Lorentz Connection
The fundamental Lorentz connection of teleparallel gravity is the purely inertial connection
(66). This means that in this theory Lorentz connections keep the special–relativistic role
of representing inertial effects only. Of course, as a purely inertial connection, its curvature
vanishes identically:
• • • • • • •
a
R bµν = ∂µ Aa bν − ∂ν Aa bµ + Aa eµ Ae bν − Aa eν Ae bµ = 0. (101)

However, for a tetrad involving a non–trivial translational gauge potential B a µ , that is, for
•
B a µ 6= D µ εa , (102)

torsion will be non–vanishing:

• • •
T a µν = ∂µ ha ν − ∂ν ha µ + Aa eµ he ν − Aa eν he µ 6= 0. (103)

In teleparallel gravity, therefore, gravitation is represented by torsion, not by curvature. This

◦
is at variance with General Relativity, whose spin connection Aa bµ has vanishing torsion
◦ ◦ ◦
a
T µν = ∂µ ha ν − ∂ν ha µ + Aa eµ he ν − Aa eν he µ = 0, (104)

but non–vanishing curvature

◦ ◦ ◦ ◦ ◦ ◦ ◦
a
R bµν = ∂µ Aa bν − ∂ν Aa bµ + Aa eµ Ae bν − Aa eν Ae bµ 6= 0. (105)

The spacetime–indexed linear connection corresponding to the inertial spin connection (66)
is • • •
ρ
Γ νµ = ha ρ ∂µ ha ν + ha ρ Aa bµ hb ν ≡ ha ρ Dµ ha ν . (106)
This is the so–called Weitzenböck connection. Its definition is equivalent to the identity
• •
∂µ ha ν + Aa bµ hb ν − Γ ρ νµ ha ρ = 0. (107)
•
In the class of frames in which the spin connection Aa bµ vanishes, it reduces to
•
∂µ ha ν − Γ ρ νµ ha ρ = 0, (108)

which is the absolute, or distant parallelism condition, from where teleparallel gravity got its
name. We notice finally that, for the specific case of the Weitzenböck connection, identity (53)
assumes the form • ◦ •
ρ ρ ρ
Γ µν = Γ µν + K µν , (109)
where • • • • 
K ρ µν = 1
2 T µ ρ ν + T ν ρ µ − T ρ µν (110)

is the contortion of the Weitzenböck torsion

• • •
T ρ νµ = Γ ρ µν − Γ ρ νµ . (111)

12
5.2 Teleparallel Lagrangian
The lagrangian density of teleparallel gravity is [12]
• h • •
L= T ρµν S ρµν , (112)
4k
where • • • • •
ρµν
S = − S ρνµ = K µνρ − g ρν T σµ σ + gρµ T σν σ (113)
is the so–called superpotential, with
• • • • 
K ν ρµ = 1
2 T ρ ν µ + T µ ν ρ − T ν ρµ (114)

the contortion tensor of the teleparallel torsion. In terms of contortion it assumes the form
• h  • µνρ • •
µρ
•
ν

L= K K ρνµ − K µ K ρν . (115)
2k
•
Substituting K ρµν , we find
• h  1 • ρ • µν 1
• • • • 
L= T µν T ρ + T ρ µν T νµ ρ − T ρ µρ T νµ ν . (116)
2k 4 2

The first term corresponds to the usual lagrangian of internal gauge theories. The existence
of the other two terms is related to the soldered character of the bundle. In fact, the presence
of a tetrad field allows internal and external indices to be treated on the same footing, and
consequently new contractions turn out to be possible. In terms of algebraic–indexed torsion,
the teleparallel lagrangian assumes the form
• h  1 • a • bc 1 • a • cb •
a
•
cb

L = T bc T a + 2 T bc T a − T ba T c . (117)
2k 4
Notice that torsion is a Lorentz tensor — it transforms covariantly under local Lorentz trans-
formations. It then follows that each term of this lagrangian is local Lorentz invariant, and
consequently the whole lagrangian is also invariant independently of the numerical value of the
coefficients.

5.3 Equivalence with Einstein–Hilbert

As we have already discussed, the curvature of the Weitzenböck connection vanishes identically:
• • • • • • •
Rρ λνµ ≡ ∂ν Γ ρ λµ − ∂µ Γ ρ λν + Γ ρ ην Γ η λµ − Γ ρ ηµ Γ η λν = 0. (118)

Substituting the relation

• ◦ •
ρ
Γ µν = Γρ µν + K ρ µν , (119)
we find
• ◦ •
ρ
R θµν ≡ Rρ θµν + Qρ θµν = 0, (120)
◦
where Rρ θµν is the curvature of the Levi–Civita connection, and
• • • • • • •
Qρ θµν = ∂µ K ρ θν − ∂ν K ρ θµ + Γ ρ σµ K σ θν − Γ ρ σν K σ θµ
• • • • • • • •
− Γ σ θµ K ρ σν + Γ σ θν K ρ σµ + K ρ σν K σ θµ − K ρ σµ K σ θν (121)

13
is a tensor written in terms of the Weitzenböck connection only. By taking appropriate con-
tractions, the scalar version of identity (120) is found to be
◦ • • • • •  2  • 
− R = Q ≡ K µνρ K ρνµ − K µρ µ K ν ρν + ∂µ h T νµ ν . (122)
h
Comparing with the teleparallel lagrangian (115), we see that
• ◦ h • 
L = L − ∂µ T νµ ν , (123)
k
where
h ◦
◦
L=−
R (124)
2k
is the Einstein–Hilbert lagrangian of general relativity. Up to a divergence, therefore, the
lagrangian of teleparallel gravity is equivalent to the lagrangian of general relativity.
To understand the presence of a divergence term between the two lagrangians, let us recall
that the Einstein–Hilbert lagrangian (124) depends on the metric, as well as on the first and
second derivatives of the metric. Equivalently, in the context of the tetrad formalism, we can
say that it depends on the tetrad, as well as on the first and second derivatives of the tetrad
field. The terms containing second derivatives, however, reduce to a divergence term [13]. In
consequence, it is possible to rewrite the Einstein–Hilbert lagrangian in a form stating this
aspect explicitly:
◦ ◦ √ µ
L = L1 + ∂µ ( −g w ), (125)
◦
where L1 is a lagrangian that depends solely on the tetrad and on its first derivatives, and wµ is
a four–vector. On the other hand, the teleparallel lagrangian (117) depends only on the tetrad
and on its first derivative. The divergence in the equivalence relation (123) is then necessary
to remove the terms containing second derivatives of the tetrad from the Einstein–Hilbert
lagrangian.

5.4 Field Equations

Consider now the lagrangian
•
L = L + Ls , (126)
with Ls the lagrangian of a general source field. Variation with respect to the gauge potential
B a ρ — or equivalently, with respect to the tetrad field ha µ — yields the teleparallel version
of the gravitational field equation [4]
• •
∂σ (hS a ρσ ) − k hJ a ρ = k h Θa ρ . (127)

In this equation,
•
•
ρσ ∂L • • •
hS a ≡−k a
= K ρσ a − ha σ T νρ ν + ha ρ T νσ ν (128)
∂(∂σ h ρ )

is the superpotential, whereas the term

•
• ∂L 1 • • ha ρ • 1 • c • ρσ
hJ a ρ ≡ − a = ha µ S c νρ T c νµ − L + A aσ S c (129)
∂h ρ k h k

14
stands for the gauge current, which in this case represents the Noether energy–momentum
density of gravitation itself [14]. Finally,
 
ρ δLs ∂Ls ∂Ls
h Θa = − a ≡ − − ∂µ (130)
δh ρ ∂ha ρ ∂µ ∂ha ρ

is the source energy–momentum tensor. Due to the anti–symmetry of the superpotential in the
last two indices, the total — that is, gravitational plus source — energy–momentum density
is conserved in the ordinary sense:
•
∂ρ hJ a ρ + h Θa ρ = 0.


(131)

The left–hand side of the gravitational field equation (127) depends on the Weitzenböck
connection only. Using the identity (119), through a lengthy but straightforward calculation,
it can be rewritten in terms of the Levi–Civita connection only:
• • ◦ ◦

∂σ h S a ρσ − k hJ a ρ = h Ra ρ − 1
ha ρ R .


2 (132)

We see from this expression that, as expected due to the equivalence between the corresponding
lagrangians, the teleparallel field equation (127) is equivalent to Einstein’s field equation
◦ ◦
Ra ρ − 1
2 ha ρ R = k Θa ρ . (133)

Observe that the energy–momentum tensor appears as the source in both theories: as the
source of curvature in general relativity, and as the source of torsion in teleparallel gravity.
This shows that, according to teleparallel gravity, curvature and torsion are related to the same
degrees of freedom of the gravitational field.

5.5 How Does Teleparallel Gravity Describe Gravitation?

Let us take the geodesic equation of general relativity:
dua ◦
+ Aa bν ub uν = 0. (134)
ds
Substituting the identity
◦ • •
a
A bν = Aa bν − K a bν , (135)
we obtain
dua • •
+ Aa bν ub uν = K a bν ub uν . (136)
ds
This is the teleparallel equation of motion of a particle of mass m in a gravitational field
— as seen from a general Lorentz frame. It is a force equation, with contortion playing
the role of gravitational force. The inertial forces coming from the frame non–inertiality are
represented by the connection of the left-hand side, which is non-covariant by its very nature.
In teleparallel gravity, therefore, whereas the gravitational effects are described by a covariant
force, the non–inertial effects of the frame remain geometrized in the sense of general relativity,
and are represented by an inertia–related Lorentz connection. Notice that in the geodesic
equation (134), both inertial and gravitational effects are described by the connection term of
the left–hand side.

15
 Considering that the teleparallel force equation and the geodesic equation of general relativ-
ity are formally the same, the teleparallel description of the gravitational interaction is found to
be equivalent to the description of general relativity. There are conceptual differences, though.
In general relativity, a theory fundamentally based on the weak equivalence principle, curva-
ture is used to geometrize the gravitational interaction. The gravitational interaction in this
case is described by letting (spinless) particles to follow the curvature of spacetime. Geometry
replaces the concept of force, and the trajectories are determined, not by force equations, but
by geodesics. Teleparallel gravity, on the other hand, attributes gravitation to torsion. Tor-
sion, however, accounts for gravitation not by geometrizing the interaction, but by acting as a
force. In consequence, there are no geodesics in teleparallel gravity, only force equations quite
analogous to the Lorentz force equation of electrodynamics [15]. This is actually an expected
result because, like electrodynamics, teleparallel gravity is a gauge theory.

6 Final Remarks
Although equivalent to general relativity, teleparallel gravity presents several distinctive fea-
tures and achievements in relation to general relativity. For example, according to the geomet-
ric description of general relativity, which makes use of the torsionless Levi–Civita connection,
there is a widespread belief that gravity produces a curvature in spacetime. In consequence, the
Universe as a whole should be curved. However, the advent of teleparallel gravity breaks this
paradigm. In fact, it becomes a matter of convention to describe the gravitational interaction
in terms of curvature or in terms of torsion. This means that the attribution of curvature to
spacetime is not an absolute, but a model–dependent statement. Here, we will discuss two ad-
ditional points: the possibility of separating inertial effects from gravitation, and the existence
of a true gravitational variable in the usual sense of classical field theory.

6.1 Separating Inertial Effects from Gravitation

Let us consider again the tetrad field
•
ha µ = D µ xa + B a µ . (137)
Whereas the first term on the right–hand side is purely inertial, the second is purely gravita-
tional. This means that both inertia and gravitation are included in ha µ . As a consequence,
the coefficient of anholonomy of ha ,
f c ab = ha µ hb ν (∂ν hc µ − ∂µ hc ν ), (138)
will also represent both inertia and gravitation. Of course, the same is true for the spin
connection of general relativity,
◦
Aa bc = 12 (fb a c + fc a b − f a bc ). (139)
In a local frame in which inertial effects exactly compensate gravitation, that connection van-
ishes,
◦ .
Aa bc = 0, (140)
and gravitation becomes locally undetectable. On the other hand, considering that the telepar-
allel spin connection represents inertial effects only, the identity
◦ • •
Aa bc = Aa bc − K a bc (141)

16
 ◦
corresponds actually to a separation of Aa bc into inertial and gravitational parts [16]. In fact,
in the local frame in which (140) holds, the identity (141) becomes
• . •
Aa bc = K a bc . (142)

This expression shows explicitly that, in such a local frame, inertial effects (left–hand side)
exactly compensate gravitation (right–hand side).
◦
It is interesting to remark that, although the inertial part of Aa bc does not contribute
to some physical quantities, like curvature and torsion, it does contribute to others. An ex-
ample is the energy–momentum density of gravitation, whose expression in general relativity
always include, in addition to the energy–momentum density of gravity itself, also the energy–
momentum density of inertial effects, which is non–tensorial by its very nature. This is the
reason why in general relativity this density always shows up as a pseudotensor.∗ Further-
more, owing to its odd asymptotic behavior, the contribution of the inertial effects often yields
unphysical (divergent or trivial) results for the total energy and momentum of a gravitational
system. As a consequence, it is in general necessary to make use of a regularizing process to
eliminate the spurious contribution coming from those inertial effects [26]. Due to the pos-
sibility of separating inertial effects from gravitation, in teleparallel gravity it is possible to
write down a purely gravitational energy–momentum density which, as for any other field, is
a true tensor. The existence of such a tensorial density allows one to compute unequivocally
the energy and momentum of any gravitational system without necessity of a regularization
process [27].

6.2 A Genuine Gravitational Connection

Due to the fact that spin connection of general relativity involves both gravitation and inertial
effects, it is always possible to find a local frame in which inertial effects exactly compensate
gravitation, in such a way that the connection vanishes at a point:
◦ .
Aa bc = 0. (143)

Since we know there is gravitation at that point, such connection is not a genuine gravitational
variable in the usual sense of field theory. Notice, in particular, that any approach to quantum
gravity based on this connection will necessarily include a quantization of the inertial forces —
whatever that may come to mean. Considering furthermore the divergent asymptotic behavior
of the inertial effects, such approach will likely face additional difficulties.
◦
Notice furthermore that the connection behavior of Aa bc under local Lorentz transfor-
mations is due to its inertial content, not to gravitation itself. This can be seen from the
decomposition (141): whereas the first term on the right–hand side represents its inertial,
non–covariant part, the second represents its gravitational part, which is a tensor. This means
that it is not a genuine gravitational connection — its gravitational content is covariant —
but just an inertial connection. One should not expect, therefore, any dynamical effect coming
from a “gaugefication” of the Lorentz group. In this sense, local Lorentz transformations are
similar to diffeomorphism, another symmetry empty of dynamical meaning. As a matter of
fact, these two kind of transformations are used indistinctly in the metric formulation of general
relativity, leading sometimes to the somewhat strange concept of “locally inertial coordinate
∗
A sample of different pseudotensors can be found, for example, in Refs. [17, 18, 19, 20, 21, 22, 23, 24, 25].

17
system”. This concept makes sense only if local Lorentz transformations between frames are
considered on an equal footing with general coordinate transformations. Of course, this can
be done as both transformations are empty of dynamical meaning.
In teleparallel gravity, on the other hand, the gravitational field is not represented by
Lorentz connections, but by a translational–valued gauge potential B a µ , the non–trivial part
of the tetrad field. In this theory, Lorentz connections keep their special relativistic role,
representing inertial effects only. Considering that the translational gauge potential represents
gravitation only, to the exclusion of inertial effects, it cannot be made to vanish in a point
through a choice of an appropriate frame. It is, for this reason, a true field variable in the
usual sense of classical field theory. It is, furthermore, a genuine gravitational connection, and
consequently the natural field–variable to be quantized in any approach to quantum gravity [4].

Acknowledgments
The author would like to thank R. Aldrovandi for useful discussions. He would like to thank
also FAPESP, CAPES and CNPq for partial financial support.

References
[1] S. Kobayashi and K. Nomizu, Foundations of Differential Geometry, 2nd edition, Wiley–
Intersciense, New York, 1996.

[2] R. Aldrovandi and J. G. Pereira, An Introduction to Geometrical Physics, World Scientific,

Singapore, 1995.

[3] N. P. Konopleva and V. N. Popov, Gauge Fields, Harwood, New York, 1980.

[4] R. Aldrovandi and J. G. Pereira, Teleparallel Gravity: An Introduction, Springer, Dor-

drecht, 2012.

[5] V. A. Fock and D. Ivanenko, Z. Phys. 54, 798 (1929).

[6] V. A. Fock, Z. Phys. 57, 261 (1929).

[7] P. A. M. Dirac, in Planck Festscrift, edited by B. Kockel, W. Macke and A. Papapetrou,

Deutscher Verlag der Wissenschaften, Berlin, 1958.

[8] P. Ramond, Field Theory: A Modern Primer, 2nd edition, Addison–Wesley, Redwood,
1989.

[9] T. W. B. Kibble, J. Math. Phys. 2, 212 (1961).

[10] W. Greub, S. Halperin and R. Vanstone, Connections, Curvature, and Cohomology: Lie
Groups, Principal Bundles, and Characteristic Classes, Academic Press, New York, 1973.

[11] C. W. Misner, K. S. Thorne and J. A. Wheeler, Gravitation, Freeman, San Francisco,

1973.

[12] J. W. Maluf, J. Math. Phys. 35, 335 (1994).

[13] L. D. Landau and E. M. Lifshitz, The Classical Theory of Fields, Pergamon, Oxford, 1975.

18
[14] V. C. de Andrade, L. C. T. Guillen and J. G. Pereira, Phys. Rev. Lett. 84, 4533 (2000),
arXiv:gr-qc/0003100.

[15] V. C. de Andrade and J. G. Pereira, Phys. Rev. D 56, 4689 (1997), arXiv:gr-qc/9703059.

[16] R. Aldrovandi, L. C. T. Guillen, J. G. Pereira and K. H. Vu, Bringing Together Gravity and
the Quanta, in Albert Einstein Century International Conference, edited by J.-M Alimi
and A. Füzfa, AIP Conference Proceedings 861, American Institute of Physics, New York,
2006, arXiv:gr-qc/0603122.

[17] R. C. Tolman, Relativity, Thermodynamics and Cosmology, Oxford University Press, Ox-
ford, 1934.

[18] A. Papapetrou, Proc. Roy. Irish Acad. A 52, 11 (1948).

[19] P. G. Bergmann and R. Thompson, Phys. Rev. 89, 400 (1953).

[20] C. Møller, Ann. Phys. (NY) 4, 347 (1958).

[21] L. B. Szabados, Class. Quantum Grav. 9, 2521 (1992).

[22] J. M. Aguirregabiria, A. Chamorro and K. S. Virbhadra, Gen. Rel. Grav. 28, 1393 (1996),
arXiv:gr-qc/9501002.

[23] J. W. Maluf, Gen. Rel. Grav. 30, 413 (1998), arXiv:gr-qc/9710124.

[24] S. Deser, J. S. Franklin and D. Seminara, Class. Quantum Grav. 16, 2815 (1999)
arXiv:gr-qc/9905021.

[25] S. V. Babak and L. P. Grishchuk, Phys. Rev. D 61, 024038 (2000), arXiv:gr-qc/9907027.

[26] J. W. Maluf, M. V. O. Veiga and J. F. da Rocha–Neto, Gen. Rel. Grav. 39, 227 (2007),
arXiv:gr-qc/0507122.

[27] T. Gribl Lucas, Yu. N. Obukhov and J. G. Pereira, Phys. Rev. D 80, 064043 (2009),
arXiv:gr-qc/0909.2418.

19