You are on page 1of 52

Home Search Collections Journals About Contact us My IOPscience

Torsion gravity

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2002 Rep. Prog. Phys. 65 599

(http://iopscience.iop.org/0034-4885/65/5/201)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 200.145.112.194
This content was downloaded on 16/03/2016 at 16:18

Please note that terms and conditions apply.


INSTITUTE OF PHYSICS PUBLISHING REPORTS ON PROGRESS IN PHYSICS
Rep. Prog. Phys. 65 (2002) 599–649 PII: S0034-4885(02)12647-9

Torsion gravity
Richard T Hammond

Department of Physics, North Dakota State University, Fargo, ND 58105, USA

E-mail: rich.hammond@ndsu.nodak.edu

Received 21 November 2001, in final form 14 February 2002


Published 27 March 2002
Online at stacks.iop.org/RoPP/65/599

Abstract

Theoretical and experimental research on general relativity with torsion is reviewed. An


introductory section establishes definitions and notation, introduces tetrads, the anholonomic
formulation and the Dirac equation in curved space with torsion. After that, gauge theories
of gravitation are introduced, starting with local Poincaré gauge theory, in which the torsion
arises as a translational gauge field strength, and other gauge approaches are described. Torsion
that is derived from a potential, including a scalar, vector, and tensor potential is discussed,
with emphasis on the antisymmetric tensor of the string theory kind. Teleparallel theories are
described, conformal invariance is discussed and a brief section on the equation of motion
is presented. Experiments that have searched for, or bounded, torsion are described and the
possible physical manifestations are broken down into the broad areas of quantum effects,
laboratory scale phenomena and large-scale tests. Finally, a discussion of the relationship
between string theory and torsion is presented.
(Some figures in this article are in colour only in the electronic version)

0034-4885/02/050599+51$90.00 © 2002 IOP Publishing Ltd Printed in the UK 599


600 R T Hammond

Contents

Page
1. Introduction 601
1.1. Overview 601
1.2. Notation and definitions 603
1.3. Background 604
1.4. To propagate or not to propagate 605
1.5. Tetrads and anholonomies 607
1.6. The Dirac equation 610
1.7. Other reviews 612
2. Gauge theories of gravitation with torsion 612
2.1. Reasons for a gauge theory of gravity 612
2.2. The beginning: local Poincaré gauge theory 612
2.3. Other gauge theories 615
3. The potential of torsion 617
3.1. Torsion from a scalar, vector potential 617
3.2. The tensor potential 619
3.3. The early days of the antisymmetric field 626
4. Teleparallelism 628
5. Conformal invariance 630
6. Equation of motion 630
7. Physical manifestations and experiments 632
7.1. General remarks 632
7.2. Quantum effects 633
7.3. Laboratory tests 636
7.4. Large-scale tests 638
7.5. Cosmological tests 640
7.6. Exact solutions 640
8. String theory and torsion 641
9. Closing thought 645
Acknowledgments 645
References 645
Torsion gravity 601

Nature is a murky gem with many facets, and opposite the shining side that reflects our
own prejudices are darker sides we have not yet penetrated . . .

1. Introduction

1.1. Overview
The concept of the gauge covariant derivative has played an integral role in transforming how
we view the natural world: many major advances over the last century in understanding nature
at the most fundamental level are built upon this foundation. However, the compact, finite-
dimensional representation groups SU (n), which are so successful in giving us the standard
view of elementary particles and their interactions, were not in the mix when the foundation
was poured. Einstein used the concept of the covariant derivative to transform his beautiful
physical ideas into hard, mathematical equations, which, when solved, exposed the reality of
the beauty Einstein saw.
Consider the covariant derivative of an arbitrary vector Aσ ,
∇µ Aσ = Aσ,µ +
µν σ Aν (1)
where Aσ,µ ≡ ∂Aσ /∂x µ and the indices run from 0 to 3. This covariant derivative transforms
as a tensor under the group of diffeomorphisms, or general coordinate transformations. The
σ
tensorial nature of this derivative is achieved through the affine connection
µν , which is a
64 component object (in four dimensions) with no symmetry properties in its indices. The
covariant derivative, along with the metric tensor are the heart and soul of general relativity,
and one may view the covariant derivative as the mechanism of transforming Minkowski
kinematics into a dynamic theory of gravitation. With the understanding that it operates on a
vector Aν (and sums over ν) we may write this derivative as
δνσ ∇µ = δνσ ∂µ +
µν σ (2)
and compare it to the gauge covariant derivative of electromagnetism
Dµ = ∂µ + ieφµ (3)
or the gauge covariant derivative of the colour symmetry describing QCD,
i
Dµ = ∂µ + gλα Aαµ . (4)
2
The similarities are striking, but the differences run deep. One such difference is that the
gauge groups associated with the elementary particles, the SU (n), are compact and therefore
have finite-dimensional representations, while the translation group is not compact. Moreover,
while the electromagnetic φµ and gluon field Aα µ are potentials in the covariant derivative, the
affine connection of Einstein’s original 1915 theory is a field strength, which is derived from
a potential (this point will be discussed more fully below).
It is often helpful to interpret the affine connection in geometric terms. For example, as is
well known, in order to compute the derivative of a vector it must be evaluated at two different
spacetime points, and it is therefore necessary to transport the displaced vector back to its
original position for comparison. If the vector is parallel transported along the infinitesimal
dx µ , then we assume that the change due to this transport is given by

µν σ Aν dx µ , (5)
602 R T Hammond

Figure 1. The infinitesimal vector ζ µ is parallel transported along χ µ , and vice versa. The
non-closure is proportional to the torsion.

which leads to the definition (1). In fact, one may define the curvature tensor as the result
of parallel transporting a vector Aν around a closed path ξ µ ,

Aσ = 21 Aν Rβµν σ ξ µ dx β (6)

which gives the definition of the curvature tensor.


However, more things than this can be extracted from the use of geometrical inventions.
Consider transporting the infinitesimal vector ζ α along χ α , and compare that to transporting
the infinitesimal vector χ α along ζ α . These are shown in figure 1.
Suppose we define the vector Aα = ζ α + χ α −
µνα ζ µ χ ν , which represents the sum of the
two lower-right vectors in figure 1, and define the vector B α = χ α + ζ α −
µνα χ µ ζ ν , which is
the sum of the two vectors in the upper left. Their difference is C α , and we see that
C α = 2Sµν α ζ µ χ ν (7)
where
Sµν α ≡
[µν] α = 21 (
µν α −
νµ α ) (8)
α α
which is defined as the torsion tensor. We see that the vectors A and B do not form a
parallelogram, and it is sometimes said that parallelograms do not exist in spacetime with
torsion, which is an overstatement, but makes a point. The non-closure of the parallelogram
is proportional to the torsion, as the figure, with (7), shows.
I stated above that the affine connection and the metric tensor are the heart and soul of
gravitation, and it is natural therefore, to seek a connection between the two. Mathematically
speaking, they are independent objects, but when building a physical theory we sometimes seek
to find a link between them. There are generally two procedures: one, based on variational
principles, considers independent variations of the metric tensor and the affine connection
(or some quantity related to the affine connection). In this case the resulting field equations
provide links, if any, that may arise. The other procedure is to make use of the nonmetricity
tensor Qµασ which is defined by
∇µ g ασ ≡ Qµασ (9)
where it is assumed that the metric tensor is symmetric throughout this paper (modern
investigations impose severe limitations on non-symmetric metric theories [42]). Using
gµν g µσ = δνσ , one also finds ∇µ gασ = −Qµασ , and using (9), one finds the relation

ασν = {ασ
ν
} + Sασν + S νασ + S νσ α + 21 (Qασν + Qσ αν − Qνασ ) (10)
where the Christoffel symbol, which is symmetric in its lower indices, is defined by
ν
{ασ } = 21 g µν (gαµ,σ + gσ µ,α − gασ,µ ). (11)
Torsion gravity 603

In 1915 it was not only reasonable but extraordinarily insightful to set Sασν to zero, and to
make Qασν vanish as well, but modern day physics should not be shackled by rusty assumptions
mired in the murky dawn of relativity. Physics is an experimental science and only experiments
will tell us if these objects are real or ethereal. Of course, a theoretical motor describing these
quantities must be up and running before credence can be given to physical observations, and
many theorists have been oiling the gears for decades.
A pessimist might be discouraged by (10), and worry about the 24 unknown functions
introduced by the torsion and the 40 unknown guests invited by the nonmetricity, but
the optimist is excited, and looks hopefully at the richness provided by these new fields.
Many physicists are realists, falling somewhere in between, trying to balance simplicity and
solvability against the possibility of developing a complete unified theory, trying to simplify
the equations in order to extract physical manifestations and measurable predictions while
not ignoring, or missing, fundamental new windows that allow us to peer into the soul of the
Universe.
In the following I will try and describe some of the various theories that tap the richness
provided by (10), paying most attention to torsion. I will concentrate on the development over
the last 10 years or so, using a thin cloth to conceal my attempts to sew torsion into the fabric
of gravity, where it adds strength and beauty to one of our most beloved views of nature.

1.2. Notation and definitions


At the risk of slowing the pace, this section contains some basic definitions and sets the notation,
hopefully making the ensuing material more transparent. Most of the notation coincides with
Schouten [178]. A geometric n-dimensional manifold is called an Xn (for a more detailed
definition of Xn see [178]). If Qµνσ = 0 the connection is called metric. If the connection
is metric and gµν is real Xn is a Un , and if furthermore the connection is symmetric then it is
called a Vn , which is also called Riemannian geometry.
From (6) we obtain the Riemann–Christoffel tensor, or curvature tensor,
φ
Rβµν σ =
µν,β σ −
βν,µ σ +
βφσ
µνφ −
µφσ
βν (12)
and contracting gives the Ricci tensor,
Rµν = Rσ µνσ . (13)
This is not symmetric, and in fact the antisymmetric part is
R [µν] = (∇σ + 2Sσ )(S µνσ + S νg µσ − S µg νσ ) (14)
ν νσ
where S = S σ is called the torsion trace or the torsion vector.
The Einstein tensor in spacetime with torsion is defined by
Gµν = R µν − 21 g µν R (15)
and has the same antisymmetric part as the Ricci tensor, and the curvature scalar is defined by
R = Rσσ . However, in geometry with nonmetricity Rβµνσ = −Rβµσ ν , and there is also the
segmental curvature tensor defined by Rµνσσ . The Bianchi identities are given by
∇ν Gµν = 2S µαβ Rβα − Sαβγ R µγβα . (16)
Although the definition of covariant differentiation is given by (1), since the connection
is not symmetric one may equally well choose
∇˜ µ Aσ = Aσ,µ +
˜ νµσ Aν . (17)
This definition leads to an alternate curvature tensor
φ
Rβµν σ =
˜ µν,β
σ

˜ βν,µ
σ
+
˜ βφσ
˜ µνφ −
˜ µφσ
˜ βν . (18)
604 R T Hammond

This opens the door to many new invariants that may be formed. Einstein solved the
problem, in his unified theory, by adopting the Hermitian principle, yet even in the case of
a symmetric metric tensor this ambiguity survives, and the notion of transposition invariance
was implemented [72]. In this case one considers a tensor of the form
Pβµν σ = 21 (Rβµν σ + R̃βµν σ ) (19)
which is invariant under the transposition
µνσ →
˜ µνσ . More recently this has been
reconsidered in the case of a non-symmetric metric tensor [63]. The problem was also discussed
in its relation to the physical interpretation of the geometry [21]. Nevertheless, for the present
we will adhere to the definition (1).
An equivalent definition of the curvature tensor is given by the commutation of covariant
differentiation,
∇[µ ∇ν] Aσ = − 21 Rµνσ φ Aφ − Sµνφ Aσ . (20)
The contortion (not contorsion!) tensor is defined as
Kασν = −(Sασν + S νασ + S νσ α ) (21)
1
and the modified torsion tensor is
T αβσ = S αβσ + S β g ασ − S α g βσ . (22)
Although the covariant derivative was defined in (1), sometimes the covariant derivative
using only the Christoffel symbol is useful, designated by
Aσ;µ = Aσ,µ + {µν
σ
}Aν . (23)
The star derivative saves some writing at times and is defined by

∇ µ Aσ = Aσ;µ + 2Sµ Aσ . (24)
The curvature of tensor of U4 spacetime can be written in terms of the Riemann curvature
tensor of V4 spacetime, o Rσ µνα , plus everything else according to
Rσ µν α = o Rσ µν α − Kµν ;σ α + Kσ ν ;µ α + Kσ γ α Kµν γ − Kµγ α Kσ ν γ (25)
Rµν = o Rµν − Kµν ;σ σ + Kσ ν ;µ σ + Kσ γ σ Kµν γ − Kµγ σ Kσ ν γ (26)
and
R = o R + 4(S σ;σ − S σ Sσ ) − Kαβγ K γ αβ (27)
or
R = o R + 4(S σ;σ − S σ Sσ ) + Sαβγ S αβγ − 2Sαβγ S βγ α . (28)
An important special case of this corresponds to totally antisymmetric torsion, and yields,
R = o R − Sµνσ S µνσ . (29)

1.3. Background
Although Einstein initially assumed that torsion vanishes, he later used it in his fernparallelism
(teleparallel) theory in 1928. Cartan, who had introduced torsion in 1922 and elucidated its
geometrical significance and mathematical foundation in [31], wrote to Einstein about his work,
and the series of letters and interchanges has been documented in [32]. A substantial portion
of the Einstein–Cartan ‘collaboration’ went into developing a teleparallel theory (which is
described in more detail below), and much of Einstein’s later work focused more on the
1 Schouten uses Tασ ν in place of Kασν , and some authors use Tασ ν in place of Sασν .
Torsion gravity 605

antisymmetric metric tensor than torsion. Schrödinger picked up the thread and developed a
unified theory of gravity and electromagnetism where torsion was related to the electromagnetic
potential [179]. One consequence of his theory was that photons acquired a non-zero mass,
which spurred Schrödinger on to finding the best (at that time) experimental upper bound on
the photon mass [180]. Despite this effort, as these early decades eventually succumbed to
growing political unrest and the horror of war, many physicists lost interest and fell quiet, and
the word torsion echoed from distant cliffs, heard by only the most devout listeners.
The birth of local gauge theory in the 1950s breathed new life into torsion. With the
first attempts by Utiyama [210] paving some ground, and Sciama [181] emphasizing torsion
as being related to spin, Kibble [119] showed how to describe gravity with torsion as a local
gauge theory of the Poincaré group, and by 1976 Hehl et al [91] formulated a complete
gravitation theory with torsion as resulting from local Poincaré gauge invariance. With this
seminal paper, torsion flared back to life like a nova. The beauty and success of global Lorentz
invariance, which was Einstein’s easel in his kinematical view of nature, was generalized with
modern gauge principles to form a compelling new picture. The two Casimir invariants of the
Poincaré group, P 2 and L2 (the square of the translation operator and Pauli–Lubański spin
operator) found perfect interpretations in a theory of gravity with torsion: generalizing the
notion that mass curves space, now we also have spin giving rise to torsion.
Hopes were high, back then. It was anticipated that having formulated gravitation as a
gauge theory the charm and success of the quantization of the SU (n) theories might rub off
on gravity, yielding a quantizable theory. Disappointment grew as morning’s light dispelled
the dream, and excitement faded further as early predictions indicated torsion forces were too
week to measure, and, in some formulations, torsion did not even propagate into vacuum. Like
the nova burning its last remnants of hydrogen, torsion dimmed quickly from sight. Torsion
was relegated to become a mathematical curiosity, a force too feeble to make its presence felt.
Indeed, like the gravitational field that persists long after the collapse of a black hole,
dark thoughts propagated long after torsion’s flare. The misconception that torsion does not
propagate (exist in vacuum like other fields) grabbed the imagination of too many, and its
putative weakness formed thick iron bars, imprisoning torsion in a jail that few minds, and even
less funding, would reach (especially in the United States). This situation reminds me of the
argument concerning Euclidian geometry versus Riemannian geometry. A post-it calculation
evaluating 1 − 2 mG/r on the Earth’s surface yields a departure from Euclidian geometry of
one part in a billion, prompting Euclid’s advocate to claim it is too small to measure and should
be ignored. Riemannn’s advocate takes umbridge, challenging his adversary to jump out the
window, warning the effect is a matter of life and death. So, although torsion may be small in
one context, it may be of paramount importance in another.

1.4. To propagate or not to propagate


Although torsion is a field that in general propagates, it is useful to see a special case in which
it does not. Consider the variational principle
δ(Ig + Im + Is ) = 0 (30)
where the geometrical part is

1 √
Ig = −gR d4 x (31)
2k
where k = 8πG/c4 and we blithely define sources according to


δIm = 2 d4 x −gT µν δgµν
1
(32)
606 R T Hammond

and


δIs = 1
2
d4 x −gµµν σ δSµν σ . (33)

Suppose the nonmetricity is zero but torsion is not. Then in order to determine the 24 unknown
functions in Sµνσ let us assume that we vary the metric and torsion tensors independently.
Without getting precise at this point, we can think of µµνσ as the spin energy potential of
matter, although that interpretation is not necessary for the present: one may simply view (33)
in analogy to (32). The resulting equations may be put in the form

Gµν − ∇ β (T µνβ + T βµν + T βνµ ) = kT µν (34)
and
Sαβγ = 2k(µγ [αβ] + µ[α gβ]γ ) (35)
where µα = µασσ .
In vacuum (35) shows us that Sαβγ = 0, and we pinpoint the source of a rumour that
propagates more effectively than its source. We see that torsion is proportional to the spin
energy potential, but in vacuum it vanishes. In other words, instead of obeying a differential
equation torsion finds itself trapped in an algebraic one, and in vacuum, must perish. Thus,
exterior to a source (34) reduces to the usual Einstein field equations and torsion is as hard to
detect as material inside the event horizon of a black hole.
Throughout this review many examples of propagating torsion will be examined. However,
even the restrictive case of nonpropagating torsion carries physical interest, and has a few twists.
For example, in standard cosmology the Universe is described by a simple fluid, and although
the discrete stars and galaxies that make up our real Universe paint quite a different picture than
the idealized fluid, that model correctly describes a large tract of empirical information. By
the same reasoning, it is natural to endow this fluid with a spin density to describe the average
effects of spin. While such an effect would not be expected to be large today, in the early
Universe it may have an important role. For example, with a generalized energy–momentum
tensor in the game, Hawking’s energy conditions may be violated and singularities may be
avoided. Thus, there is good motivation for the study of material fluid sources which are
endowed with spin. Although, it proved very difficult to obtain a Lagrangian and variational
principle, the problem was solved Ray and Smalley [170]. More recent work considers each
micro-volume to be viewed as rigid bodies, and thermodynamic properties of this fluid are
investigated [163]. Besides this, it has also been shown that quantum corrections can yield an
effective propagating torsion field from classical non-propagating torsion [36]. A more recent
and definitive paper on fluids, which encompasses spin fluids, has been given in [160].
Nevertheless, the example described above is practically the only formulation in which
torsion vanishes in vacuum, and is a very special case. For example, suppose following the
Brans–Dickian notion, we let G → φG, where G remains constant and the scalar field is
dimensionless. With this, using (30), (31) the vacuum torsion equation becomes
1
Sασ γ = φ[,α gσ ]ν . (36)
φ
Assuming a Lagrangian with the usual kinetic terms for the scalar field, which gives φ full
citizenship in the world of propagating fields, then (36) shows that the torsion retains the same
citizenship, and should be entitled full status of a propagating field.
As another example suppose we consider an alternative Lagrangian,

1 √
Ig = −g(R + BRµν R µν ) d4 x. (37)
2k
Torsion gravity 607

The use of quadratic terms is expected in gauge theories and is required for renormalization [53].
For these reasons let us examine (37), which would be included in more general and complete
formulations. The torsional field equations (still obtained by varying the torsion while holding
the metric tensor fixed) are [70]
T αβγ = BC γ [αβ] (38)
where
∗ ∗ β
C αβγ = − ∇ γ R αβ + g αγ ∇ φ R φβ + 2Rφ S γ φα . (39)
These results have been generalized in [73] to examine the more complete action:

1 √
Ig = −g(R + AR 2 + BRµν R µν + CRµν R νµ ) d4 x (40)
2k
but (39) makes the point. Since the curvature tensor contains the first derivative of the
torsion, (39) and (38) show that torsion, in vacuum, obeys second-order differential equations,
and therefore propagates.
Another physical approach is based in the idea that torsion is derivable from a potential.
As an example, consider
Sασγ = φ,α δσγ − φ,σ δαγ (41)
and define the source according to


δIs = 21 d4 x −gρδφ. (42)

Then variations with respect to the potential φ yield


 φ = −48ρ (43)
which shows, again, that torsion propagates through space.
I should note an important objection to the paragraph above. It could be argued that the
torsion potential is introduced in an ad hoc manner, engineered soley to obtain second-order
differential equations. For a further discussion, and not simply isolated equations with more
suggestion than prediction, one should see section 3 below, where the reader is directed to
original works.
My point is to show that torsion has escaped from Pandora’s box, and only the most
restrictive set of assumptions could have kept it there. For the majority of this review, attention
is focused on four dimensions, although important work and insight has been gained by looking
beyond N = 4, as in [38, 62, 113, 114, 116, 129, 134], and the references contained therein.
Sometimes torsion is like an endangered species: it seems that it should not propagate but
manages to find nourishment from unexpected quarters. For example, in [55], which also
has a discussion on the use of Riemann normal coordinates in spacetime with torsion, it was
shown that a kinetic term in the torsion trace is generated in the low-energy effective action
(one should also consult [83] for a discussion on normal frames).

1.5. Tetrads and anholonomies


I might have left a false impression, it being, that it is easy to generalize or alter general
relativity. It is not! Many theories are cut down by the blade of empiricism, while others never
get off the operating table, beset with internal inconsistencies or inoperable mathematical
pathologies. One reason that such stringent demands are placed on a putative theory is that the
equations of motion, either through the Bianchi identities or conservation laws arising from
608 R T Hammond

symmetry, are attached to the field equations like an electromagnetic field to an electron—you
cannot have one without the other.
At the heart of the equations of motion is the energy–momentum tensor of matter, and at
the heart of matter is the Dirac equation. Thus, even before considering a quantized theory
of gravity, one must couple the geometric left hand to the material right. The problem of
introducing the Dirac spinor into the world of tensors was solved long ago [23] by assuming
that at each point in spacetime a tetrad ei can be introduced, which is an analytic function of
spacetime2 .
The Latin indices label the tetrad, each of which has four components. (This is the notation
of Schouten, but opposite to that of Hehl. Unfortunately there is no consensus or overwhelming
adoption of one way or the other, so the reader of the literature must become bilingual, or a
µ
fast translator.) The tetrad components are indicated by e i where µ is a coordinate label. The
µ j j
co-frame, or dual, tetrads are ej , and by definition are taken to obey the relations e i eµ = δi .
Tensor quantities may be transvected with the tetrads to obtain quantities that are given in
the tetrad frame (warning, this only works, in general, for tensor quantities):
Ai j = eµi ejν Aµν . (44)
It is often necessary to use anholonomic coordinates, which occur when the object of
anholonomity,
4αβ a = e[β,α]
a
, (45)
does not vanish. It is also useful to define the quantity
β
4ab c ≡ eαa e b 4αβ c . (46)
The Minkowski metric is denoted by ηij (= diag(+ − −−)) and is used to raise and lower
tetrad indices. Orthonormal tetrads are introduced according to
ea eb = ηab (47)
and, as a fundamental object, the metric tensor
gµν = eµi eνj ηij (48)
steps aside, letting the tetrad don the mantle it previously wore. In this notation, it would
allowable to use g in place of η on the right-hand side of (48), but it is conventional to use ηij
as the Minkowski metric.
In anholonomic coordinates the covariant derivative is given by
j j
∇i Aj = A ,i +
im Am (49)
j
where a new player enters the field—
im ,
which may be called the anholonomic connection
µ
coefficient, but will be reintroduced below with other names (also, note that ∂i ≡ ei ∂µ ).
µ j
Writing ∇i Aj = e i eν ∇µ Aν we deduce from (1) and (49) the important relation

αβσ =
abm em
σ b a σ
eβ eα − eb,a eβb eαa (50)
which relates the affine connection
αβσ to the anholonomic connection coefficients
abm . One
may also deduce that

[ab] c = Sab c − 4ab c , (51)
which should be compared thoughtfully to (8). Note that even if torsion vanishes does
[ab]c
not, and one should not think that the object of anholonomity is the source of torsion (see
also (190) and the discussion there). It is also useful to invert (50) by transvexing with the
2 The Dirac equation may also be treated in curved space without tetrads by using spinors, see [11] for a review.
Torsion gravity 609

tetrad (and the dual tetrad) to get the connection coefficient in terms of the affine connection,
which gives,
β

abc = eαa e b eσc
αβσ − eαb eαc ,a . (52)

Latin indices are raised and lowered with ηab , so for example
abc = ηmb
amc , and it is
convenient to define the quantity
µab according to


µab ≡ eµn
nab . (53)
µ
This quantity has several aliases. Often written as ωµmn = ηam ηbn
µab , and ωamn = ea ωµmn ,
it is called the spin connection, or the connection 1-form (see below). Assuming that
Qαβγ = 0, as we do throughout majority of this paper (except in metric affine theories where
the nonmetricity is not zero), it follows that


µab = −
µba . (54)

With this, and using (52), one may show that


abc = −4abc + 4acb + 4bca − Kabc . (55)

The first three terms are what Ricci called the rotation coefficients,

γbca = −4abc + 4acb + 4bca . (56)

Later we will see that


µab is also called the rotational gauge potential.
We may obtain the curvature tensor in anholonomic coordinates by casting the tensor
equation (20) into an anholonomic basis

∇[k ∇j ] Ai = − 21 Rj kin An − Skjn ∇n Ai , (57)

and see

Rj ki n = 2(
[j i,k] n +
[j i m
k]m n + 4kj m
mi n ) (58)

where the underlined letters stand immune to the antisymmetrization.


Having planted the seeds of anholonomity, the time is ripe for harvesting the beauty of
differential forms. To begin, we take ea to be a 1-form, or in terms of the basis 1-forms dx ν ,
this is written as eνa dx ν . With this we find

dea = Ωa (59)

where Ωa is the anholonomic 2-form Ωa = 4µνa dx µ ∧ dx ν and the exterior derivative is


dea = eν,µ
a
dx µ ∧ dx ν . Care must be taken to avoid losing an unpolished diamond in the
ore, and we should not allow (59) to go unnoticed. In fact, defining the connection 1-form by
Γbc =
αbc dx α and the torsion 2-form S a = Sµνa dx µ ∧ dx ν , and using (59), (51) becomes

S a = dea + Γba ∧ eb (60)

which is also what Cartan called an equation of structure. We also introduce the curvature
2-form Rab = Rµνab dx µ ∧ dx ν and we see that (58) becomes
1
Rb
2 a
= dΓab + Γnb ∧ Γan (61)

which Cartan called the other equation of structure.


610 R T Hammond

1.6. The Dirac equation


Orthonormal tetrads can be an integral part of general relativity—from simplifying algebra to
providing an alternative physical interpretation of gravitation, yet nothing is more crucial than
their role in describing the Dirac equation in curved space. The main issue that introduces
itself is this: although general covariance can be achieved through the use of tensors, there are
no finite-dimensional representations of the spinor.
Consider the Dirac equation in Cartesian coordinates,
imc
γ µ ∂µ ψ + ψ = 0. (62)

The rule ‘comma goes to semicolon’, or ordinary partial derivative goes to covariant derivative
cannot be directly applied here, but the way around this impediment is well known. We write
µ
the Dirac equation in curved space by taking the gamma matrices to be γ µ = ei γ i , where γ i
is the matrix in Minkowski space. We may then employ a covariant derivative of the form
Dµ = ∂µ +
µ , (63)
and under the transformation ψ → Sψ, where S is a representation of the Lorentz group, and
assume
µ →
µ . In particular, for the infinitesimal transformation,
ψ → (1 + 21 ω(x)ab σ ab )ψ (64)
and
µ
eaµ → (δab + ω(x)ba )eb (65)
we insist that (63) is covariant. Notice that we are considering local Lorentz transformations
since the ωab are spacetime dependent. Also we are considering the spin-1/2 representation
so that σ ab = (1/2)γ [a γ b] , which obeys the Lorentz algebra,
[σ mn , σ ab ] = ηna σ mb − ηma σ nb + ηnb σ am − ηbm σ an (66)
(note that these brackets do not carry the factor of 1/2 that the brackets that surround indices
do, as in (8)). Using these transformations in (63) and using S −1 γ a S = γ a , we find (63) is
covariant if

a →
a = S
a S −1 − S,a S −1 . (67)
Thus, we must find a quantity that transforms according to (67). The result turns out to be

µ = 21 σ ab eνa ∇µ eνb . (68)
Using (66) one may show (68) transforms according to (67). This result may be put in a more
traditional and useful form by noting that
∇µ eνb = −
µa c ηbc eνa (69)
which may be derived by using
∇µ eφb = eφb,µ −
µφσ eσ b (70)
and (52) so that

µ = − 21 σab
µab . (71)
Putting this together we finally write
Dµ ψ = ψ,µ − 41 γa γb
µab ψ. (72)
Now that we have the gauge covariant derivative one might be tempted to think that the
Dirac equation in curved space with torsion is γ µ Dµ ψ + (imc/h̄)ψ = 0, but Nature likes to
Torsion gravity 611

tease us from time to time, and this putative equation is an example of just this. The reason
is this, suppose we want to derive the Dirac equation from a variational principle: the most
natural way to proceed is to start with the Minkowski–Lagrangian and let ∂µ → Dµ , which
gives

δID = δ eLD d4 x = 0 (73)
where  
ih̄c a a 2imc
LD = − (Da ψ)γ ψ − ψγ Da ψ − ψψ (74)
2 h̄
where
Da ψ = ψ ,µ + 41
µab ψγa γb . (75)
A useful identity is
γ a γ b γ c = γ a ηbc + γ c ηab − γ b ηac + i; abcd γ5 γd , (76)
which may be used with (75) in (74) to obtain
 
ih̄c 2imc
L=− ψ ,a γ a ψ − ψγ a ψ,a +
abc =abc − ψψ (77)
2 h̄
where =abc ≡ (1/2)ψγ [a γ b γ c] ψ and ;αβµν is the totally antisymmetric tensor. This result
shows that the interaction part of the Lagrangian contains only the totally antisymmetric part
of the torsion.
The Dirac equation is obtained by taking independent variations with respect to ψ, which
yields
imc
γ a (Da + Sa )ψ + ψ =0 (78)

where the (perhaps) surprising term γ a Sa apparently worms its way into the equation, but is
not really there at all. To see what this really means it is helpful to break the derivative up
according to
Da ψ = ψ;a + 41 Kabc γ b γ c (79)
where the torsion is isolated in the second term and
ψ;a ≡ ψ,a − 41 ωabc γ b γ c ψ. (80)
With this the Dirac equation (78) becomes
imc
γ a (ψ;a + iγ5 bn ) + ψ =0 (81)

where
bµ = ;µαβγ S αβγ . (82)
This is a well known result: a Dirac particle only interacts with the totally antisymmetric
part of the torsion. It is interesting to note that if the torsion happens to be totally antisymmetric,
then the torsion trace vanishes and the Dirac equation (78) becomes
imc
γ a Da ψ + ψ = 0, (83)

which is what we might have guessed in the first place, but is only valid if the torsion is totally
antisymmetric, see [88] for more details.
The Dirac equation will revisit us from time to time, especially when we look for physical
manifestations of torsion in section 7, but first we shall see how the Dirac spinor plays an
essential role in gauge theories. One last point: although it has been shown that only the
totally antisymmetric part of the torsion couples to a spin-1/2 particle, higher spin particles
(spin-3/2) couple to other parts of the torsion, see for example [95].
612 R T Hammond

1.7. Other reviews


The first publication on torsion is about to enjoy its 80th birthday. Obviously, to write a report
that does justice to all publications on torsion risks creating a thick dust collector that resides
permanently on a shelf, up in the corner. The following articles usually focus on one area, and
while this paper attempts to give a more inclusive and broad discussion of recent developments,
many essential details and important references are relegated to these other reviews.
The most widely read (and rightly so) paper, and a good starting point is the paper by
Hehl et al [91]. Although it is now dated, and focuses on the local Poincaré gauge theory,
it is an excellent starting point. More recently, the metric affine gauge theory was described
in great detail in [95]. In a less widely known but very readable paper [93], Hehl discusses
the kinematics of torsion, and the meaning of measurability of torsion, and cosmological
implications. In a review dedicated to torsion of the string theory type, one may consult [81],
and a very elementary exposition is given in [78]. A recent review dedicated to the quantum
aspects of torsion, including unitarity and renormalization issues may be found in [185], and
for a recent review of gauge gravity in general one may consult [17].

2. Gauge theories of gravitation with torsion

2.1. Reasons for a gauge theory of gravity


The first paper that formulates gravitation as a gauge theory was submitted by Utiyama in
1955 [210], only 1 year after the famous paper of Yang and Mills [223]. In fact, Utiyama
developed his theory independently of the work of Yang and Mills, and was shocked to see
so many of his ideas published, and later remarked that he wished he published his work
earlier. For a discussion on this and the early formulations of gauge theory one should see
O’Raifeartaigh’s very nice book [164]. The point is, the underlying beauty of local gauge
invariance is irresistible, and this is perhaps one of the most pervasive reasons for attempting
to formulate gravity as a local gauge theory.
There are more prosaic reasons as well. Local gauge theory seems to be the best way
we can understand Nature: whether we look at the SU (2) × U (1) electroweak theory, the
SU (3) × SU (2) × U (1) standard model, or even electromagnetism as a U (1) theory. Gravity
has always kept its distance from the other fundamental forces, and resisted quite successfully
attempts at quantization. Thus, it was hoped that formulating gravity as a gauge theory might
shed light on the correct path to quantization, and in so doing show us how to develop a fully
unified theory of all of the forces in Nature.

2.2. The beginning: local Poincaré gauge theory


It is sometimes argued that gravity is already a gauge theory of the group of
diffeomorphisms [3], but the first attempt at making gravity a local gauge theory in the
more modern sense was made by Utiyama. Although it is hard to find an argument against
formulating gravity as a gauge theory, it is easy to find one about the appropriate underlying
group. Utiyama assumed it was the homogeneous Lorentz group, and that under the local
(; km = ; km (x)) Lorentz transformation
x k → x k + ; km x m (84)
the tetrad suffered the transformation
µ µ
em → em − ; am eµa . (85)
Torsion gravity 613

This was, and is, an appealing idea. Instead of thinking about metric tensor transformations,
by going to the tetrad, one assumes the effect of the Lorentz transformation is to rotate the
tetrad according to (85). In order to resurrect invariance, in the true spirit of gauge theory, the
partial derivative must be generalized to something of the form
ψ,µ → ψ,µ − 21 Aabµ ;ab ψ (86)
where the quantities Aabµ were interpreted as the gauge field. This was an extremely important
step in the development of a gauge theory of gravitation, especially when one considers its
timeliness (back then). However, nowadays we see some difficulties with the details. In order
to relate Aabµ to the affine connection, Utiyama essentially assumed the affine connection to be
symmetric. Moreover, his conservation law seems only to contain orbital angular momentum,
but the biggest problem is this: the Lorentz group relates to orbital angular momentum while,
in general relativity, the source is energy–momentum. Thus, a major improvement was taken
by Kibble [119] who solved these problems by taking the underlying symmetry group to be the
inhomogeneous Lorentz group, or the Poincaré group. (In fact, after Kibble’s work Utiyama,
with Fukuyama, used the inhomogeneous Lorentz invariance to show that a symmetric second
rank tensor Aµν was required as the gauge field. They further showed that Aµν appears the
same as the metric tensor in the Lagrangian [211].)
The Poincaré transformation includes the translations ; µ ,
x µ → x µ + ; µν x ν + ; ν , (87)
and Kibble considers both the local variation,
δψ = ψ  (x  ) − ψ(x) (88)
and the total variation
δ0 ψ = ψ  (x) − ψ(x) = δψ − δx µ ψ,µ . (89)
Kibble embraces the nonsymmetric nature of the affine connection and shows that spin
gives rise to an antisymmetric part. Kibble’s work, and that of Sciama [181] (who actually
discussed the Poincaré group before Kibble) are discussed in more detail by Hehl et al who give
the most comprehensive formulation of a local Poincaré gauge theory of gravity in their famous
paper of 1976 [91]. In their work, they first consider the active global Poincaré transformation
as
ψ → (1 − ; µ ∂µ + ωαβ fβα )ψ (90)
where the ∂µ are the translation operators and fβα are the rotation generators which obey the
Lorentz algebra (66) and for the Poincaré group are given by fβα = 21 σβα . The theory is made
into a local gauge theory by assuming that the ; µ and ωαβ are functions of space and time.
In order to maintain invariance the ordinary partial derivative must be replaced by the gauge
covariant derivative,
∂µ → Dµ = ∂µ −
µab fab . (91)
As noted above, the Poincaré group has the four translation operators and six rotation
operators, and the tetrads eµi become the translation gauge potential and the
µab are the
rotation gauge potential. The local Poincaré transformation becomes
ψ → (1 − ; µ Dµ + ωαβ fβα )ψ, (92)
and the commutation of the derivative operators gives
β µν
[Da , Db ] = eαa e b (Fαβ fνµ − Fαβn Dn ) (93)
614 R T Hammond

where
Fαβab = 2(∂[α
β]ab +
[αna
β]nb ) = eσa eνb Rαβσ ν (94)
is the rotational field strength and
−Fαβn = 2Sαβn (95)
is the translational field strength. Poincaré gauge theory is characterized by these field strengths
and two gauge potentials
eµn translational gauge potential (96)

µab rotational gauge potential. (97)
The gauge covariant translation operator Dµ which appears here changes the commutation
relations from the Minkowski space results; in particular the algebra does not close, as (93)
shows. The field equations are derived by postulating independent variations of the two gauge
potentials, and we will come back to this point again later.
Most of the work in [91], including an important discussion of the conservation laws,
is performed without choosing the actual Lagrangian. It is noted that if the scalar invariant
R is chosen then one has the result that torsion couples algebraically to the spin density
of matter. This occurs for essentially the same reasons as were discussed in section 1.4.
However, these authors also emphasize that traditional gauge theories contain terms quadratic
in the field, signalling that gravity might follow suit, in which case (as shown in section 1.4)
torsion propagates. Thus, there is nothing about a gauge theory that forces torsion to be a
nonpropagating field, one may say instead that it is the choice of the Lagrangian, but I will
come back to this point later. To enjoy the full beauty and richness of this theory I urge the
reader to study the original work [91] and a review [92].
In the 1980s, Hayashi and Shirafuji [85] further examined Poincaré gauge theory, but
took the notion of quadratic Lagrangians very seriously. They considered the irreducible
decompositions (under the Lorentz group) of the torsion tensor as follows. The trace of the
torsion is defined as
Sµ = Sµσσ . (98)
Using the Young table method they also define the traceless part
tαβγ = Sαβγ + Sβγ α − 13 (Sβ gαγ + Sα gβγ ) + 23 gαβ Sγ (99)
and antisymmetric part
aσ = 13 ;σ αβγ S αβγ . (100)
With these they consider the most general Lagragian quadratic in these irreducible parts of the
torsion,
LT = αtσβγ t σβγ + 4βSσ S σ + γ aσ a σ . (101)
They repeat this procedure for the curvature scalar, and excluding parity violators, get
five terms quadratic in the curvature tensor, or combinations and contractions of the curvature
tensors. Adding the scalar invariant R they arrive at a ten parameter Lagrangian for the
Poincaré gauge theory. The quantum effect of quadratic torsion terms at the one-loop level
has been investigated in [110]. Another issue that arises in gravitation is stability, or whether
the total energy has a lower bound. This issue is discussed in the context of R + T 2 (where
T 2 generically represents terms quadratic in torsion) in [141], where stability is shown to
exist. Properties of more general Lagrangians, and the coupling to matter fields, was discussed
in [175].
Torsion gravity 615

Advantages of this general framework are that torsion propagates, and the use of terms
quadratic in the field strength mimics conventional gauge theory. However, certain difficulties
do arise within this framework that go beyond the general complexity of the field equations.
Although the equations that follow from the variational principle are of second differential
order, they can be cast in a form that are of third differential order, and it becomes difficult to
assess whether the equations are well posed. There is a plethora of new constants with little
guiding physics principles to help elucidate either their magnitude or origin (however, in [112]
Katanaev reduces this number by imposing constraints on the equation of motion).
However, these authors do show how a judicious combination of constants can eliminate
higher derivatives from the theory. In subsequent papers they discuss equations of motion,
limiting cases, and the weak field approximations to their equations [84]. As the years
progressed, various kinds of difficulties emerged, including a sensible Newtonian limit,
tachyonic solutions, and the lack of a well defined initial value problem. These problems
were discussed in more detail and partly solved later [87].
In 1992 Grignani and Nardelli [65] also adopt a Poincaré gauge theory approach, and
claim that their implementation of gauge invariance is more closely aligned with ordinary non-
Abelian gauge theories in that their approach is passive, and does not replace the translational
symmetry in terms of coordinate transformations. In order to accomplish this they must
introduce a set of ‘Poincaré coordinates’ q a (x) which are interpreted as coordinates in the
internal gauge space. In this development, care must be taken to correctly identify the
gauge potential and torsion. For example, unlike the Poincaré theory discussed above, the
translational field strength only reduces to the torsion in the so-called physical gauge, in which
q a (x) = 0. This approach, however, was questioned by Strobl [201] who claims the road to
quantization becomes more difficult in their approach due, in part, to the constraint algebra,
but this claim was rejected by the authors [66]. In another approach, Poincaré gauge theory
with torsion was developed on a Minkowski spacetime, and renormalization properties were
investigated in [219].

2.3. Other gauge theories


Despite the great success of gauge theories in general, a weakness they cannot shroud is the
lack of physical motivation for choosing the particular group. The universal acceptance by
nature to possess rigid Poincaré invariance in particle physics (in Minkowski spacetime) is the
exception, and explains why gauging the rigid Poincaré theory seems so natural. Nevertheless,
many of the hopes and dreams of local Poincaré gauge theory, such as quantization of gravity
and unification with the other interactions, never materialized, and theoreticians, like sailors
looking for a port, searched for other groups that may harbour gravitation more effectively
than Poincaré.
In 1980 Stelle and West [198] formulated a Yang–Mills gauge theory of gravity using the
de Sitter SO(3, 2) group. In this theory they had to introduce a nonpropagating field y A in
the five-dimensional space (here and in what follows A = 1, 2, . . . , 5). In order to reduce to
the four-dimensional spacetime of general relativity, the de Sitter symmetry is spontaneously
broken to SO(3, 1). They obtain, as a special case, the result that torsion does not have to arise
from intrinsic spin, but may be generated by the gradient of the Higgs field. This is especially
interesting when we compare it to the work below, where we envision torsion as being derived
from the gradient of a scalar potential.
A few years later Pagels [165] established a gauge theory of gravity based on the orthogonal
group O(5). The ten gravitation gauge fields
µAB transform as the adjoint representation and
break down such that the six
µab become the connection of four-dimensional spacetime and
616 R T Hammond

the
µ5a becomes the tetrad. Like the I SO(3, 1), vacuum torsion vanishes in this theory due
to the restrictive Lagrangian that is used.
Also gauged supersymmetry, i.e. supergravity, invites the presence of torsion. For
example, van Nieuwenhuizen [212] shows that the torsion field is required to implement
the gauged supersymmetry of the gravitino field, although Chamseddine and West [34] set the
translation field to zero in the theory based on the 14-dimensional supersymmetry group. This
is also discussed in [192]. Burinskii [28] has shown that in N = 2 supergravity, super torsion
exists and gives rise to torsion travelling waves, and interactions between a supersymmetric
particle in spacetime with torsion was investigated in [35]. Supersymmetric sigma models
with a torsion potential were investigated in [105].
In recent years metric affine gravity (MAG) has ignited brush fires across the globe. Fuel
for the fire was provided in 1995 by the compelling arguments in [67], and with more details
in a major review article for MAG [95]. The main generalization of MAG is that it relaxes
the assumptions that the nonmetricity tensor Qµνσ vanishes. From the gauge point of view
there are now three gauge potentials, the translation (tetrad), the field strength of the linear
group and now the metric tensor. The nonmetricity tensor is the corresponding field strength.
In physics we are usually more comfortable with letting a field strength, such as Qµνσ , be
determined from field equations rather than ad hoc assumptions (that is, assume it is zero).
Another reason for considering nonmetric theories is that they provide a natural setting for
conformal invariance. Although this invariance may not hold on the large scale of today, there
is reason to believe that it was important at early times or on the small scale. The fundamental
symmetry in MAG is the affine group, which is the semidirect product of the translation group
R 4 and the linear group GL(4, R). When this group is gauged and a metric is maintained,
MAG results.
Up until 1977 there was a widely known fact which was generally stated as follows:
‘There are no spinorial representations of the diffeomorphisms because sl(4, R) has no double
coverings’3 . Unfortunately this widely known fact was wrong, as shown by Ne’eman [154].
The fact is, there are infinite-dimensional representations, and this opens the door to a much
richer gauge structure. A discussion with references may be found in chapter 4 and appendix C
in [95].
The essence of nonmetricity is that it leads to a change in the length of a vector upon
parallel transport. For example, let L2 = gµν Aµ B ν . Now suppose we examine the change of
L upon parallel transport: this gives
dL2 = −Aµ B ν Qσ µν dx σ , (102)
µ ν
and, if A = B , L is interpreted as the length of the vector and dL is the change in length
upon parallel transport. Weyl embraced this effect and assumed that
Qσ µν = −φσ gµν , (103)
which gives
dL = 21 Lφσ dx σ (104)
where he assumed that φσ was proportional to the electromagnetic potential. It was an appealing
idea, not only in its own right, but that he used this to develop a conformally invariant theory.
Even though Einstein showed Weyl’s theory was flawed on physical grounds, over the years
this theory has smouldered like an underground coal fire, and with MAG, it has flared back to
the surface. Poberii [169] has given a comprehensive treatment of tensor calculus of MAG.
In trying to decide upon the correct Lagrangian of a particular theory it is useful to have
the irreducible decomposition of torsion, as we saw above, and in MAG it is equally helpful
3 From Y Ne’eman, private communication.
Torsion gravity 617

to have similar decompositions. McCrea [145] gives the irreducible decomposition of torsion,
curvature and nonmetricity for the pseudo-orthogonal group. Problems with MAG include the
fact that the standard Einstein–Hilbert action leads to unphysical constraints on the material
sources [159]. These authors consider a 12 parameter Lagrangian and present vacuum and
cosmological solutions to the equations, and the equation of motion of test particles [155],
and other exact solutions have been given by in [208]. It has also been shown that Plebański–
Demiański-like solutions exist for special Lagrangians, and the effect that higher spin massive
particles couple to parts of the torsion beyond simply the axial vector part [60]. Plane wave
solutions have been found in the electrovac sector, with a cosmological constant, which carry
curvature, nonmetricity and electromagnetism in addition to torsion [61], while solitonic
monopole solutions have been demonstrated [140]. The physical implications and constraints
introduced by perfect fluids in MAG has also been investigated [191]. An extensive and useful
review of exact solutions of MAG is given by Hehl and Macias [100].
For an in depth discussion on gauge theories in gravitation one may consult [17], and for a
more selective view one may consult [206]. Gauge theories of gravity have not only generated
much research in their own right, but have rekindled a warm interest in torsion as well. With
the strong physical motivation of local gauge theory, and its spectacular track record, this is not
surprising. The resulting microscope under which torsion has been viewed reveals that there
are no physical or mathematical barriers that will prevent theories of gravitation with torsion.
In fact, it seems much more natural and less restrictive to formulate gravity with torsion than
without. However, gauge theory can be very restrictive, and although this may be viewed as
good, since torsion is automatically forced into playing a given role and therefore testable,
it may also hide the true physical description of torsion. Nature is a murky gem with many
facets, and opposite the shining side that reflects our own prejudices are darker sides we have
not yet penetrated. Over the past decade or two many have tried peering from other sides, as
many other descriptions and theories of torsion have emerged. In the following we shall look
into some other theories of torsion, and in next section torsion that is derived from a potential
will be described.

3. The potential of torsion

3.1. Torsion from a scalar, vector potential


As we have seen, an early perplexing issue was the concept of nonpropagating torsion, and
another head-scratcher concerned the coupling to electromagnetism. For example, suppose we
adopt the principle of minimal coupling, by which, in this context, I mean that when postulating
the electromagnetic field, defined as usual as Fµν = 2A[ν,µ] in Minkowski space, in curved
space we replace ordinary partial derivatives with covariant derivatives (and as always adopt a
metric-independent potential Aµ ). Then we have
Fµν
min
= Fµν − 2Sµν σ Aσ . (105)
This is an unacceptable definition due its lack of gauge invariance under
Aµ → Aµ + λ,µ . (106)
Thus, one had to either abandon minimal coupling, which worked so well elsewhere, or sacrifice
gauge invariance, which is sacrosanct. From another standpoint one may ask: if torsion arises
from spin, and photons have spin, then why does torsion not couple to electromagnetism?
A pleasant compromise was discovered in [103] where the definition of gauge invariance
was generalized to
Aµ → Aµ + eφ δµν φ,ν (107)
618 R T Hammond

Figure 2. Torsion through the decades.

by introducing the scalar field φ and taking the torsion of spacetime to be

Sµν σ = 21 (δνσ φ,µ − δµσ φ,ν ). (108)

Now, under the generalized gauge transformation (107), we see that (105) is invariant. The
authors formulated a Lagrangian and showed that torsion, multiplied by the electromagnetic
field strength, generates electric and magnetic currents. These authors also suggested the apt
tlaplon for the name of the quantum of the torsion field. However, in [156] Ni showed that this
theory was in disagreement with null experiments for difference in acceleration of objects of
differing material. After this work and related discussions, in [15] it was argued that treating
gravity as a SL(2, C) gauge theory, these problems do not really arise, and nowadays the most
accepted result is that torsion does not couple to the electromagnetic potential in the definition
of the field strength. For a modern discussion of torsion–electromagnetic coupling, and one
that includes Einstein’s equivalence principle (see [98]).
As we saw earlier, the reason that torsion propagates is due to the fact that, when it is
assumed to be derived from a potential, the Lagrangian (the curvature scalar) contains products
bilinear in the first derivative of the potential and therefore the field equations are of second
differential order. The advantage of this approach is that we may retain the curvature scalar as
the Lagrangian and are not forced into adopting much more complicated quadratic Lagrangians
and their associated luggage. This problem of minimal coupling with torsion has irrigated the
fields of discussion for decades and will be reviewed in section 7.1.1.
Torsion from a scalar field also arises in work that generalizes the invariant volume element
√ √
−g d4 x. Saa argues for a form given by e−3φ −g d4 x where the torsion is given by its trace,
which is in this case, Sα = φ,α [176]. Equations of motion for this theory were discussed in [57].
De Sabbata also considered a scalar field representation of torsion in [49]. In [195], the
torsion vector is derived from the derivative of a scalar field, where the scalar field is taken to
be the spacetime-dependent G. Possible cosmological or solar system effects are discussed in
section 7.
Torsion gravity 619

Torsion has also been written in terms of a vector field according to


Sµν σ = δνσ θµ − δµσ θν (109)
and of course, we may view (108) as a special case of (109). This was examined in [139].
Nowadays, this trace or vector piece of the torsion is referred to as the trator in computer
algebra work.
Schrödinger used a connection that contained a vector part in his attempt to construct a
unified theory of gravitational and electromagnetism [179], and related research was performed
much later [71], where the antisymmetric part of the Ricci tensor was found to be proportional
to the electromagnetic field tensor. It was shown that the photon could acquire a mass [74],
and that magnetic monopole type solutions were present [76], but nowadays it is generally
accepted that torsion arises from spin, not electromagnetic charge. Nevertheless, the thread
was picked up by Smalley and Krisch in [189], who showed how to maintain gauge invariance
by endowing the torsion with a compensating transformation property. They developed the
Lagrangian and equations of motion for a charged spinning fluid in U4 spacetime. Another
formulation in which torsion is related to electromagnetism was developed in [172], in which
the torsion tensor is proportional to the tensor product of the electromagnetic field tensor and
the particle four velocity.
Vector torsion was also considered in the idea of transposition invariance in [72], and was
of the form
S µνσ = (S µ g νσ − S ν g µσ )/3 (110)
µ
where S is the torsion trace. In this formulation the torsion field, in the static limit, turned
out to be given by
e−µr
S0 = q (111)
r
where we see that the torsion gives rise to a Yukawa potential.
The use of a torsion potential in non-metric theories was also considered in [224] and the
relation between a scalar torsion potential and the Brans–Dicke field was reported in [138].
Also, a parity violating term was introduced into the Lagrangian where the torsion is derived
from the derivative of a scalar field in [106]. It was found that, on the solar system scale, the
theory was indistinguishable from the Brans–Dicke predictions.

3.2. The tensor potential


Another approach that assumes the torsion is derived from a potential, but gives torsion a richer
structure than that provided by a scalar field, assumes that [77]
Sµνσ = ψ[µν,σ ] = 13 (ψµν,σ + ψσ µ,ν + ψνσ,µ ) (112)
where ψµν is the antisymmetric potential. This introduces a new gauge invariance
ψµν → ψµν + ξ[µ,ν] (113)
while providing the natural foundation for propagating torsion.
A short digression concerning first-order versus second-order formalisms must intrude
upon us at this point. Suppose, for the moment, we consider vacuum only so that we have


δ d4 x −gR = 0. (114)

Obviously, we can make no ground until we postulate precisely what the variation is. In a
first-order formalism independent variations of the metric tensor and the connection are taken.
620 R T Hammond

This stems from the Palatini method, and in pure general relativity simplifies the calculations
considerably since one only varies with respect to gµν , and not its derivatives (which are
contained in the connection). Now, assuming Qαµν = 0, (112) with (10) gives
σ

µν σ = {µν } + Sµνσ . (115)
In the following gµν and ψµν are taken as the independent potentials of the gravitational field
and the torsion field. Consequently, independent variations of these quantities are considered.
Since the curvature scalar contains higher-order derivatives of gµν , this must be accounted
for in the variation, and this procedure is therefore usually called a second-order formalism.
The same argument holds when the tetrads are introduced. In that case, instead of taking
independent variations of gµν and ψµν , one varies the tetrad and ψµν , but one still must
consider the derivatives of the tetrad that appear in the Lagrangian. In contrast, let us go back
and look at the Poincaré gauge theory, which uses independent variations of the tetrad (the
translation potential) and the connection (the rotational potential) and at the same time let us
examine (55). This indicates that the rotational gauge potential (the spin connection) already
has derivatives of the tetrad (within this first-order formalism), so it is unnecessary to postulate
a torsion potential. In a second-order formalism, on the other hand, derivatives of the tetrad
are taken into account in the variational principle, and therefore there it is inappropriate to
maintain independent variations of the tetrad derivatives, and therefore a potential is taken as
an independent object4 . This ends the intrusion.
The vacuum equations are given by
µνσ µ
Gµν − S ;σ − 2S αβ S ναβ = 0 (116)
and variations with respect to ψµν yield the torsional field equations
µνσ
S ;σ = 0. (117)
It is helpful to bear in mind that with (112) the torsion is totally antisymmetric and (14)
becomes
µνσ
R [µν] = S ;σ . (118)
A mathematical simplification, or condensation, occurs by defining the quantity φµν =
gµν + ψµν . Since gµν is symmetric and ψµν is antisymmetric, one may consider variations
of φµν in a variational principle. The symmetric and antisymmetric parts of the resulting
equations will be the gravitational field equations (obtained by taking variations of gµν ) and
the torsional field equations (obtained by taking variations of ψµν ).
From a philosophical point of view, in our battle to understand and describe torsion we
have gained some ground and lost some ground. The ground lost refers to the appealing
foundation that local Poincaré gauge theory gives. The ground we gained is that torsion (the
potential) obeys a second-order differential equation and takes its place side by side with other
physical fields. Also, there is more good news to come.
The first hint as to the physical interpretation of torsion comes from the fact that no
spherically symmetric solutions to (116) and (117) exist [75]. This is the first indication
that the source must be a vector quantity, and therefore spin seems the most natural choice.
However, nothing definite can be determined until the equations of motion are derived, and to
obtain these a material action is needed. It has been shown that a suitable action is given by

  µ
dxn dxnν c2
 ν 
µ dxn
IM = mn c 2
dτn gµν + dτn ξn (2ψµν + gµν ) . (119)
n dτn dτn 2 dτn
4 As a warm breeze blew across Jerusalem, Professor Hehl and I discussed this point at the 8th Marcel Grossman

Meeting.
Torsion gravity 621

The sum over n represents the sum over all of the infinitesimal regions of the particle, which,
as is discussed below, must have structure. This is called the minimally coupled action because
it agrees with the Dirac coupling, which is discussed below.
The cornerstone of this formulation was the introduction of the intrinsic vector quantity
ξ µ , which was found necessary in order to provide a definition of the source of torsion. It was
further found that the source had to have structure, and more, if it were assumed to be static,
then

ξnµ = 0. (120)
n

All of the properties are quite natural if one adopts a string as the source, but more on that
later. For the general case the variational principle takes the form
δ(Igeom + IM ) = 0 (121)
with

√ R
Igeom = −g d4 x (122)
2k
variations are taken with respect to gµν and ψµν and


δIM = 21 d4 x −gT µν δφµν . (123)

With this the gravitational field equations are


µ
G(µν) − 2S αβ S ναβ = kT (µν) (124)
and the torsional field equations are
µνσ
S ;σ = −kj µν (125)
µν
where j ≡ (1/2)T . [µν]

It is also useful to present the equation that is obtained by varying φµν , which yields the
16 equations
µνσ µ
Gµν − 3S ;σ − 2S αβ S ναβ = kT µν . (126)
Of course, the symmetric and antisymmetric parts of (126) give back (124) and (125).
The specific form for the source tensor is derived by making sure (119) is equivalent
to (123), which gives
T µν = T (µν) + T [µν] , (127)
dx µ dx ν dx ν)
T (µν) = ρ + σ (µ , (128)
dτ dτ dτ
and
dx ν]
T [µν] = 2σ [µ (129)

where

c 
ρ=√ dτ δ(x − xn )mn (130)
−g n
and

c2 
σµ = √ dτ δ(x − xn )ξnµ . (131)
−g n
622 R T Hammond

The new gauge invariance of the theory leads to the conservation law


1
2
−g d4 x T µν λ[µ,ν] = 0, (132)

and assuming that λµ is arbitrary this gives


µν
j ;ν =0 (133)
(which also follows from the field equations (125)). Also, assuming that IM is invariant under
x µ → x µ + ; µ we have


d4 x −gT µν L; φµν = 0 (134)

where
L; φµν = −; σ,ν φµσ − ; σ,µ φσ ν − ; σ φµν,σ . (135)
This conservation law yields
µν µ
T ;ν = 23 T αβ S αβ (136)
which may also be derived from the Bianchi identity (16).
This equation is the gateway to observation. The equations of motion can be derived using
the Papapetrou method. First define

v0
M µν = τ µν dV (137)
c
where τ µν = T̃ (µν) , parentheses imply symmetrization, dV = d3 x, and the tilde implies

multiplication by −g. Also, consider the definitions

αµν v0
M =− δx α τ µν dV (138)
c

v0
mαµν = δx α j˜µν dV (139)
c

1
J µν = (δx µ τ νo − δx ν τ µ0 ) dV . (140)
c
The method of Papapetrou can be used to derive the equations of motion, both for
translation and for rotation [166]. The translational equations turn out to be

d M µ0 µ µ µ
+ {αβ }M αβ = {αβ },η M ηαβ + 3S αβ ,η mηαβ (141)
dτ v0
where

µM µ0 1
p ≡ 0 = τ µ0 dV (142)
v c
is the particle momentum, and the rotational equation is
d αβ dy α M β0 dy β M α0
J + − = 6S [βµν mα]µν + 2{[β
µν }M
α]µν
. (143)
dτ dτ v 0 dτ v 0
This equation shows that torsion arises from intrinsic spin. To see this, consider the case that
no forces act (fields are zero) so that the right-hand side is set equal to zero, and integrate to
get
J µν + (orbital angular momentum)µν = constant. (144)
Torsion gravity 623

Now, one may set


J µν = Lµν + S µν (145)
where, breaking the test particle into small volumes labelled by i,

µν 1 µ
L = dṼi {δyi ρviν vi0 − (µ ↔ ν)} (146)
c i
and

µν 1  µ
S = dṼi {δzi (σiν vi0 + σi0 viν ) − (µ ↔ ν)}. (147)
2c i
One may infer from these definitions that Lµν is the rotational angular momentum (note the
velocity terms). Thus, from (144) we conclude that S µν must represent intrinsic spin (details
may be found in [81]).
This result not only shows that torsion gives rise to intrinsic spin, we may find an explicit
form for spin from the definition
v α µν
Sγ = J ;αµνγ : (148)
2c
Taking the limit that v → 0 this gives

S= 2 1
r × σ dṼ (149)

where σ is the three-vector associated with (131). This result shows that intrinsic spin arises
from structure, and has nothing to do with velocity.
The translational equation would predict geodesic motion were it not for the two terms on
the right-hand side. The first term is the well known acceleration due to a particle with structure
in a gravitational field, and the second term represents the acceleration of a particle with spin,
embodied by the term mαµν , in a non-uniform torsion field. We may think of this as an electric
dipole in a non-uniform electric field. The equation may be reduced to a low-velocity limit
and one may read out the force on a particle with spin due to a torsion field, it is
c
F = − ∇(b · S ), (150)
2
and the corresponding potential energy
c
U = b·S (151)
2
where b = {bi } and bµ is defined in (82).
In order to assess the physical meaning of torsion it is helpful to look at the low-energy
limit. For this tensor potential, one may recast the equations into three-vector equations by
defining the sources according to N i = 6kj 0i where N = (Ni ) and Ii = 6kjj k , I = (In )
where the subscripts on In are permuted (I2 = 6kj31 etc). Potentials are defined according to
ai = 2ψj k with a = (ai ), Ai = 2ψ0i and A = (Ai ). With these definitions the fields are given
by
b = ∇ × A − ȧ (152)
and
b0 = ∇ · a (153)
while the field equations become
ḃ − ∇b0 = I (154)
624 R T Hammond

Figure 3. Some common features between the vector formulation of the totally antisymmetric
torsion and electromagnetism are displayed in this table.

and
∇ × b = N, (155)
where the overdot means ∂/∂t.
One may compare the vector formulation to that of electromagnetism, and some
similarities are noted in table 3.
A few applications of these equations are helpful to visualize torsion. For example, in
Minkowski space for a static source one has
∇ 2 A = −J (156)
which yields
3kc S × r
A= , (157)
4π r 3
and the field, taking the spin in the z direction, is
3ck S
b= (2 cos(θ )r̂ + sin(θ )θ̂). (158)
4π r 3
This shows explicitly how spin creates the torsion field. One may see that if S is taken to
be the intrinsic spin of an elementary particle, then the torsion goes like b ∼ L2P /r 3 . This is
the minimally coupled value, and below it is shown how torsion may be much larger than this
value (see section 7.1.1). In fact, it is shown that the non-minimally coupled solution contains
the dimensionless constant κ, and the above solution may be generalized by letting k → κk.
To allow for generality, the non-minimal coupling is adopted below.
As another example, using the Green function G(x − x  ), the solution to (125) is

ψ µν (x) = −3kκ d4 x  G(x − x  )j µν (x  ) (159)

which yields, after performing the time integration and expanding the Green function
3kκ ab0
ψ ab = ṁ , (160)
4πcr
where here Latin indices refer to spatial components (not tetrad indices) and the over dot
implies a time derivative.
Now it is useful to consider the vacuum gravitational field equations broken apart as
follows:
o
Gµν = k(T µν + t µν ) (161)
Torsion gravity 625

where
kt µν = 3S µαβ S ναβ − 21 g µν S αβσ Sαβσ (162)
which separates out the stress energy tensor of torsion. With it, one computes the outgoing
energy flux from ct 0i ni , where ni is the unit vector. Integrating over the solid angle and
averaging over the angles yields
P = β m̈ab0 m̈ab0 (163)
where β = 2Gκ /c . This represents the total power radiated from a source.
2 7

One application is the case of spin flipping. Modelling this by S = (h̄/2) cos ωt ẑ for N
coherent spin flips we get [80]
3βc2 h̄2 N 2 ω4
P = . (164)
256
The Dirac equation gives further motivation to consider torsion of the form (112). It was
shown in section 1.6 that a Dirac particle couples only to the totally antisymmetric part of
the torsion, making the rest of the torsion superfluous. Thus it is natural, when one considers
that the fundamental particles are spin- 21 , to consider a formulation in which torsion is totally
antisymmetric from the outset (however, one should note that higher spin particles do couple
to parts of the torsion beyond the axial vector torsion [95]).
In any case, we consider


R
δ e + L d4 x = 0 (165)
2k
and
 
ih̄c a a 2imc
L=− (Da ψ)γ ψ − ψγ Da ψ − ψψ (166)
2 h̄
where
Da ψ = ψ,a − 41
abc γ b γ c ψ (167)
and

abc = −4abc + 4bca − 4cab + Sabc . (168)
In (165) variations are taken with respect to the tetrad eµi , the Dirac adjoint ψ and the torsion
potential ψµν . The resulting field equations are
µνσ ih̄ck mn
Gmn − eµm eνn S ;σ − 2S mab S nab = − M , (169)
2
imc
γ a Da ψ + ψ = 0, (170)

and
αβσ ih̄ck αβσ
S ;σ =− = ;σ (171)
2
where =αβσ = (1/2)ψγ [α γ β γ σ ] ψ. In the above, the energy–momentum tensor of the
gravitational field equations is defined according to
 
ih̄c µ
δ eL d4 x ≡ − e d4 x Mi δeµi , (172)
2
which turns out to be
i
=mn = ψγ(m Dn) ψ − D(m ψγn) + ψ(γ a ba gmn − b(m γn) )ψ. (173)
3
626 R T Hammond

Some of the consequences and implications of these results are discussed in section 7.2. Also,
a more detailed examination of the antisymmmetric field and strings is given in section 8.
Autumn’s planting will predict Spring’s growth, but there are always surprises in the
harvest. Having planted the antisymmetric field as the torsion potential, the surprise we find
is that a scalar field blooms as a required ingredient in the theory. To see how this arises, note
that a solution of (171) is given by
S αβσ = −4π iL2P =αβσ + λαβσ , (174)
αβσ
where λ is a function of integration that must satisfy
αβσ
λ ;σ = 0. (175)
This must be satisfied identically in order that (174) satisfies (171) and therefore
λαβσ = ; αβσ µ ξ,µ (176)
where ξ is a scalar field. To see that this scalar field is required, consider the dual of (174):
bµ = 12πL2P ψγ5 γµ ψ + 6ξ,µ . (177)
µ
The definition of torsion (8) yields the identity b ;µ ≡ 0, which with (177), implies that
,µ Gm
ξ ;µ = −2πL2P (ψγ5 γ µ ψ);µ = 2π i ψγ5 ψ. (178)
c2
This shows that the scalar field cannot be arbitrarily set equal to zero and must therefore be an
integral part of gravity with torsion (of the form (8)). This result has interesting implications.
First, it gives a definite indication of the source of the scalar field: the pseudoscalar invariant.
Also it shows that under certain conditions the scalar field may be approximated by zero.
For example, in scattering where the fermion is taken to be an asymptotically free particle,
the pseudoscalar term vanishes. In another area, in [79] it was shown that by adopting the
low-energy string theory Lagrangian, the scalar field of string theory can sit in for (176).
This result unearthed the clue that provided a generalized definition of torsion to include
the scalar field from the ground floor. To do this, the torsion was defined according to [68]
Sµνσ = Hµνσ + (2/3);µνσ α λ,α (179)

where Hµνσ = ψ[µν,σ ] . The geometrical action, Igeom = (1/2k) d x −gR, becomes
4

√ 1
Igeom = −g (o R − H αβγ Hαβγ + 4λ,α λ,α ) d4 x. (180)
2k
With G → G0 e2φ (G0 is the fixed universal constant) then we have the low-energy string
theory Lagrangian. This result was also derived without using (179), but using a conformal
transformation in [79]. We will pursue the string theory connection further in section 8.

3.3. The early days of the antisymmetric field


Although the antisymmetric field finds a cosy home in general relativity, it has an interesting
history before it settled here. The first serious study of an antisymmetric field seems to be
undertaken in 1938 by Kemmer in [117], who, motivated by Yukawa’s ideas, was trying to
describe the nuclear interaction. About thirty years later it was rediscovered by Ogievetskiĭ and
Polubarinov [161]. They did not consider torsion at all, but were investigating the nature a field
given by (112). The guiding principle appears to have been the consideration of a field that, in a
sense, is complimentary to electromagnetism. They show that the quantum of the field, which
they call the notoph (this predates the tlaplon suggestion, but is not as appropriate) is massless
with zero helicity. These authors also investigate some possible interactions, showing that
Torsion gravity 627

while some lead to processes with no absorption or emission of the notoph, the pseudovector
interaction leads to nontrivial effects, an observation that boomeranged through the following
decades.
Four years later Miyamoto and Nakano [150] considered the tetrad as the translation gauge
field, and linearizing the tetrad, defined a tensor akµ according to
eµk = δµk + akµ . (181)
They identify the symmetric part of aµν as the gravitational field and the antisymmetric part as
Aµν . They consider various physical effects, including hydrogen level shifts and contributions
to the anomalous magnetic moment, using potential coupling to the Dirac particle.
In 1974 Kalb and Ramond [109] used the antisymmetric field to characterize the interaction
of strings, and in the appendix, gave the quantized version of the field theory of the
antisymmetric tensor. Since then, this field has often been called the Kalb–Ramond field.
In the same year that Kalb and Ramond published their seminal paper, Scherk and
Schwarz [177] published the first paper tying the antisymmetric field of string theory to the
torsion of gravity. This remarkable marriage was not, however, made in heaven, and its
sanctity was not universally acknowledged. Their motivation was string theory, which sports
the low-energy effective Lagrangian (this topic is discussed further below)


I= −g d4 x (o R − f (φ)Fµνσ F µνσ − 21 φ ,σ φ,σ ) (182)

which includes a scalar field φ and the antisymmetric field which is denoted by Fµνσ ≡
3A[µν,σ ] , where Aµν is the antisymmetric potential, and I am using their units and notation.
Their remarkable result was that this action could be replaced by


I= −g d4 x R (183)

where R is the full curvature scalar (with torsion) provided that the affine connection is given
by

σ σ 4πG σ σ ,σ 8π Gf (φ) σ
{µν } ≡ { µν } + (δµ φ,ν +δν φ,µ −gµν φ ) + Fµν . (184)
3 3
This was formulated in terms of a non-metric theory where ∇σ gµν ∼ gµν φ,σ , and, in those
days, f (φ) was not determined.
This field, and its coupling to the electromagnetic field, was discussed by Cremmer and
Scherk [41]. Nambu [153], who calls it the Cremmer–Scherk–Kalb–Ramond field, breaks
from gauge invariance and adds a mass term, and besides, discusses a long-range force
generated between strings. Covariant quantization was performed and the Ward identities
were analysed; among other effects it was shown that hidden ghosts were in the theory [204],
and soon after the hidden ghosts were studied in [182] and the theory was further studied using
the BRS formalism [120]. Renormalization of the antisymmetric field coupled to gravity
was investigated, but the antisymmetric field was not taken to be torsion in this work [184].
Also, during this period Hayashi and co-workers [84], in their teleparallel theory (see below
for a discussion of teleparallelism) introduce the antisymmetric field as a torsion potential,
something like (112). In his work de Sabbata [50] also considered an antisymmetric field.
A more detailed examination of the antisymmetric field in string theory was undertaken
in [174], and by then the papers on the antisymmetric field burgeoned quickly [214], although
a few papers that took special interest in this field are [167] who look at the antisymmetric
field in de Sitter space [171], who looks at the Higgs mechanism, while the non-Abelian case
was considered in [69], algebraic renormalization in [135] and an exact solution that violates
cosmic censorship in [123].
628 R T Hammond

Being an integral part of string theory, most of the work and study of the antisymmetric
field comes from that quarter, but the question remains, is the antisymmetric field of string
theory the torsion of gravitation? I will come back to this question again later, but cannot
promise a universally accepted answer. It occurred to me that, following Nambu, we might
call this field the Ogievetskiŏ–Polubarinov–Kalb–Ramond–Cremmer–Scherk–Schwarz field,
although the order of names might need rearranging, everyone is still not represented, and it is
much too long already, so, following custom, I will sometimes refer to it as the antisymmetric
field, or the Kalb–Ramond field, and of course, torsion.

4. Teleparallelism

Torsion is nowhere more required than in teleparallelism. Just as a gravitation field cannot
exist without matter, teleparallelism cannot exist without torsion, and after adopting torsion,
Einstein considered teleparallel unified theories of gravitation and electromagnetism with
much enthusiasm. A space possesses teleparallelism, or absolute parallelism, if the change
in a quantity that is parallel transported along a curve between any two points A and B is
independent of the path taken. One could look at this in two equivalent ways: one is to say
that this implies the curvature tensor is zero. Spacetime with zero curvature is also referred to
as a Weitzenböck spacetime.
A vanishing curvature tensor also implies that the Einstein tensor is zero,
Gµν = 0 (185)
which, with (26) may be written in the suggestive form
o µν
G = 8πtµν (186)
where
8π t µν = Kµν ;σ σ − 2Sν;µ + Kµφ σ Kσ νφ − 2Sφ Kµνφ + gµν (2S σ;σ − 2S σ Sσ − 21 Kαβσ K γ αβ )
(187)
and Kνσν = 2Sσ (which follows from (21)) was used.
At first glance this equation has some appeal. If one interprets t µν as the source, this
formulation naturally geometrizes the material source, and in this approach torsion, in place of
matter, may be viewed as giving rise to Riemannian curvature. However, the separation in (186)
is certainly not unique, and simply because certain terms are dumped on the right-hand side
(and a cosmetic factor of 8π is tossed around) does not turn them into matter. Another issue
relates to the Bianchi identity (16), which, with (185), is identically satisfied. This means that no
additional constraints are placed upon t µν , which means the equations of motion are completely
unspecified. Even more important, of course, is the fact that these are merely identities: the
real requirement is that the full curvature tensor vanishes, and not merely its contractions.
A better way to view teleparallelism is this: consider any orthonormal frame field, and
define the spacetime such that this is parallel, i.e. define
∇µ eφb = 0. (188)
This implies, using (70), that

µν σ = enσ eν,µ
n
. (189)
If the expression is used in the definition of the curvature tensor (12), we find that the curvature
tensor vanishes identically.
In this situation we see that the affine connection, and therefore the geometry, is defined
by the tetrad. In fact, from (189) we have
Sµν σ = 4µν σ (190)
Torsion gravity 629

so in this case the torsion appears to be given by the object of anholonomity, which it is not.
This result only holds in this particular case of orthonormal frames.
With these introductory remarks about teleparallelism aired out, it is time to mention the
more modern approach to this subject, as presented in [84]. These authors observe that if one
considers an action of the form

I = d4 x R (191)

then, since R = 0, it appears that the variational principle is satisfied identically. However
they, consider a more general action by adding

I = d4 x (R + LT ) (192)

where LT is given by (101). They also consider material sources, find vacuum solutions and a
Newtonian limit. To study the motion of a test particle, they assume that the particles follow
along geodesics. In a later publication an analysis of spin precession is given [86].
More recently this problem was reconsidered, from a gauge point of view, where the
Lagrangian is assumed to be proportional to o R, the scalar of torsion free spacetime [46].
Using (28) their geometrical part of the Lagrangian is
L = −S σ Sσ + Sαβγ S αβγ − 2Sαβγ S βγ α (193)
where the total derivative term from (28) is ignored, and the constants have been left out
of (193). They point out that this is the same as that obtained from the translational gauge
theory. To compute the equation of motion, they postulate a separate particle action. A similar
approach is studied in 2 + 1 dimensions [115].
Studying the gauge symmetries, a Lagrangian of the form
L = λµν αβ R µν αβ + LT (194)
αβ
is written where λµν are Lagrange multipliers which are used to enforce the teleparallel
condition. This is studied in the context of a Poincaré gauge theory so variations are performed
with respect to the tetrad, the rotational gauge potential and λµν αβ . This gives the λµν αβ
a new gauge symmetry. The explicit roles of this λ symmetry as well as the Poincaré
symmetry are explained. In other work a BRST-like operator was constructed and studied
in a teleparallel geometry [162]. This was later identified as a translational symmetry [202].
In [147] the Noether identities for Weitzenböck spacetime were considered, and in another
interesting development, in [209] the teleparallel theory was shown to be equivalent to that
with a quadratic spinor Lagrangian. In fact, an equivalence was examined between teleparallel
theories and Ashtekar’s variable in [148]. The relationship between the Yilmaz–Rosen metric
and teleparallel geometry was investigated in [152], and a Poincaré formulation of teleparallel
theory was given in [16].
An old yet forever interesting theme in gravity concerns the localization (or non-
localization) of gravitational energy momentum density. In teleparallel space, the gravitational
field strength is the torsion (in a tetrad basis), and of course, we know that the torsion is a tensor.
In [47] it is shown that there exists a conserved energy–momentum gauge current, which is
shown to be the teleparallel equivalent to Møller’s canonical gravitational energy–momentum
tensor, while the teleparallel version of the Kaluza–Klein theory is given in [48] in which torsion
generates the gravitational field and the electromagnetic field appears as a gauge component
of the torsion tensor. One may also consult [45] for the force equation in a Weitzenböck
spacetime.
The Hamiltonian formulation is given in [142], conservation laws of the Poincaré
symmetry as related to the energy–momentum and angular momentum of teleparallel theories
630 R T Hammond

is discussed in [18]. A problem with some teleparallel theories is that the torsion is so busy
giving rise to the gravitational field, it has no time represent spin. This problem and others
have been investigated by Kopczyński [126].

5. Conformal invariance

The concept of conformal invariance has grabbed the interest of physicists throughout the last
century. From Weyl’s very appealing but non-metric theory to the latest motivations provided
by string theory and MAG, conformal invariance clings to our imagination with the tenacity
of a good idea. Under the conformal transformation
gµν → ωgµν (g µν → g µν /ω) (195)
the curvature tensor is not invariant, showing that general relativity is not conformally invariant.
Weyl was able to provide a compensating field to allow for this invariance, but the cost was
nonmetricity. Within a metric theory, it was noted in [73] that torsion could be used as a
compensating field in the place on the nonmetricity tensor. In fact it was shown that under the
combined transformation (195) and
Sµν σ → Sµν σ + 2λ,[µ δν]
σ
(196)
one has
γ φ
Rαβ → Rαβ + Kγ αγ 4β − Kγ φ 4φ gαβ + 2Kαβ 4φ + Kαφβ 4φ
+ Kφβα 4φ + gαβ (2  λ −  ω/2ω − 8λ,φ λ,φ ) + 4λ,β;α + 8λ,α λ,β
+ ω−1 [−4(ω,α λ,β ) + 4λ,φ ω,φ gαβ − ω,β;α + 3ω,α ω,β /2ω] (197)
where 4σ = 2λ,σ −ω,σ /2ω. This shows that if 4σ = 0 (and therefore also 4σ;σ
= 0) then
Rµν is conformally invariant. For this reason, it is sometimes said that torsion may be viewed
as a gauge field of conformal invariance [73]. However, there are certain objections to this
approach. First of all, the Ricci tensor is defined only in terms of a connection, and not in
terms of a metric, while the conformal transformation obviously is a metric transformation.
So can a conformal metric transformation effect the Ricci tensor? The answer is, this occurs
only after (10) is adopted, at which point the Ricci tensor depends on the metric. Nowadays,
however, the torsion is not viewed as a gauge field of the conformal transformation. Of

course, since −gR is not conformally invariant, one is led to study a Lagrangian density of

the form −gRµν R µν . This was studied in [72] and it was shown that the ensuing torsion
may be interpreted as massive vector field, the field strength of which is the antisymmetric
part of the Ricci tensor. A similar approach was taken a few years later and the effect of a
phase transformation was investigated in an inflationary de Sitter space [121], and the effect
of a deviation from a quartic potential is also discussed [122]. A related approach was also
considered in which the torsion was derived from a scalar potential [224].

6. Equation of motion

The equation of motion of a particle in general relativity (in its original 1915 form) can be
obtained in at least three ways. One way, I, is to postulate that it follows along a geodesic,

δm ds = 0 (198)

which implies
dv σ
+ v α v β {αβ
σ
} = 0. (199)

Torsion gravity 631

Another approach, II, assumes that the particle undergoes motion such that the velocity
undergoes parallel transport along the curve, or equivalently, using the comma goes to
semicolon rule. This also yields (199). All of these, however, are merely postulates. For
example, with torsion, I and II do not agree. With torsion, II leads to
dv σ
+ v α v β
αβ σ = 0 (200)

which is called the autoparallel equation and differs from the geodesic. In general, however,
as emphasized in [91], neither of these is correct since we know that the equations of motion
follow from the field equations due to either the Bianchi identities or the Noether identities,
which would be methods III or IV. Einstein’s original equation
Gµν = kT µν (201)
µν
along with the Bianchi identity G ;ν = 0 imply
µν
T ;ν =0 (202)
which produce the equation of motion. For example, using (202), one has (see equation (44)
of [81])

d M µ0 µ µ
+ {αβ }M αβ = {αβ },η M ηαβ (203)
dτ v0
plus higher-order corrections. The terms in this equation are defined in section (3.2) but as a
reminder, M ηαβ may arise from rotation, or other ‘structure’ of a particle. It is interesting to
note that (203) shows that, in general, different particles fall with different accelerations5 .
Of course, in the case just presented, the deviation from the geodesic is usually small.
However, the point is, equations of motion follow from the field equations and should not be
postulated separately. For example, in section (3.2) it was shown that the equation of motion
was neither geodesic nor autoparallel, and that the additional acceleration may, in some cases,
be the dominate force on the particle. For example, in a torsion field without gravity (141)
yields,

d M µ0 µ
= +3S αβ ,η mηαβ (204)
dτ v0
which gives the exact nature of the torsion force, and its magnitude6 .
In addition to the translational equation, the rotational equation of motion, which also
follows from the field equations, is also important. For example, many theories predict the
following equation in the weak field limit (see, e.g., equation (73) in [77]):
dS
∼S×b (205)
dt
which shows that the effect of the torsion is to cause the spin vector to precess around the
torsion field. These effects may lead to experimental observations of torsion. A more complete
discussion on the equations of motion and torsion, and the Noether identities, is given in [94].
The point is, for a realistic determination of torsion effects, one needs a complete theory,
and not simply suggested interactions. However, due to the fact that spin is so intricately
tied to the Dirac equation, very often the Dirac–Lagrangian is used as the source for the
field equations. In this case one usually looks at the Dirac equation, its Hamiltonian and
5 Note that if one takes general relativity to be based on the principle of equivalence, than this equation violates that

principle for any real particle. In this vein, one may be led to claim, ‘The only thing the principle of equivalence
proves is that it is wrong’.
6 The magnitude is uncertain for the non-minimally coupled case.
632 R T Hammond

the associated interactions to infer the observable interactions with torsion. In the following
sections authors draw on all of these methods in our ongoing quest for observational evidence
of torsion.

7. Physical manifestations and experiments

7.1. General remarks


Torsion is nearly as old as general relativity, but unlike gravitation, torsion did not enjoy the
status of having spawned missions that travelled across the globe to measure its effects, nor did
it have ready-made anomalies such as planets failing to measure up to Newtonian command.
Nevertheless, after decades of research, theoretical evidence mounted and built a platform
on which to conduct experiments. The experimental search for torsion takes us down many
unchartered paths, but we often recognize familiar terrain and we must struggle to separate
possible new torsion effects from those explained by known forces and fields.
One of the problems that faces the experimentalist is the array of theories that are laid out
on the table like a dinner buffet. The issue is exacerbated by the further choice of possible terms
within a given theory that are spread across the dessert tray. For example, if one considers
Poincaré theory with a seven parameter Lagrangian, what constants are important and which
might be zero? Or should one consider the metric affine theory, or string theory torsion?
As stressed by Carroll and Field [30], the torsion is a tensor and therefore we do not have
to restrict the Lagrangian to be simply the curvature scalar. These investigators attempt to
consider the effects of torsion without choosing a specific theory. For example, they discuss
the notion that although there may be no coupling between torsion and gauge fields at the
classical level, a coupling may occur at the quantum level that could place constraints on the
theory. However, they consider that torsion arises as a scalar field, and although their otherwise
general approach is helpful in looking at effects of torsion that may be common to different
theories, we are forced to look deeper.
In the following I will break up torsion effects into three broad areas, which often overlap.
One arena is quantum mechanics, another consists of laboratory tests, and the third concerns
effects on the large scale such as solar system or astrophysical effects. Of course, there is
plenty of overlap.

7.1.1. Minimal versus nonminimal coupling. As far as torsion is concerned, the phrase
‘minimal coupling’ is used in two distinct ways in the literature. One way has already been
explained in section 3, and states, when postulating the definition of the electromagnetic field
the ordinary partial derivative is generalized to be the covariant derivative. This concept may
be extended to the derivative of other fields as well.
The other common usage for minimal coupling usually involves the Dirac equation. For
example, suppose we generalize the Dirac–Lagrangian from (74), which is called the minimally
coupled Lagrangian, to
 
ih̄c a a c̄ 5 µ 2imc
LD = − (Da ψ)γ ψ − ψγ Da ψ − Bψγ γ bµ ψ − ψψ (206)
2 4 h̄
where bµ is defined by (82), and the B term is the non-minimal term. With this, the Dirac
equation is
imc i
γ α ψ;α + ψ = −κ γ5 bσ γσ ψ (207)
h̄ 4
Torsion gravity 633

where κ = B + 1. In fact, when (119) was introduced it was called the minimally coupled
action because it leads to the exact same interaction as the minimally coupled Dirac equation.
However, by replacing ψµν by Bψµν in (119), the classical phenomenological results will
agree with the Dirac results. For example, the torsion field (158) becomes
3ckκ S
b= (2 cos(θ )r̂ + sin(θ )θ̂) (208)
4π r 3
and with (151) the interaction energy is
g
U = 3 [S · S  − 3(S · r )(S  · r )/r 2 ] (209)
r
where g = 3ckκ/8π.
More general non-minimal coupling can be considered. For example, one may add terms
proportional to the trace of the torsion, and look for effects these terms produce. These ideas
are discussed in detail in [185] and [30].
On a final note, one might notice that, using the principle of minimal coupling directly on
the Dirac equation produces (83) but this violates the principle of minimal coupling when it is
applied to the Lagrangian, which yields (78). Thus, the equation we obtain depends on where
the principle of minimal coupling is applied (although the results are identical for the case of
totally antisymmetric torsion).

7.2. Quantum effects


The search for experimentally observed effects of torsion is made extremely difficult, not so
much by its expected weakness (which nevertheless is a significant hurdle) but by the myriad of
theories and formulations of torsion. General investigations can be helpful because their wide
umbrella covers a large swath of ground. On the other hand, specific investigations provide
details of the elusive interaction, and like a gold nugget to the miner, show us where to dig. In
the following both general investigations and some more specific ones will be discussed.
We saw in section 1.6 that torsion couples to the Dirac spinor, and it is therefore natural to
look for quantum effects that might arise from the Dirac equation. An extensive discussion of
the quantum mechanics of torsion, including nonminimal coupling, anomalies, unitarity and
renormalization is given by Shapiro [185]. The reader is advised to consult this recent paper for
a more detailed discussion, but certain general observations may be made here. For example,
prompted by the observation that Dirac particles do not follow geodesics in spacetime with
torsion [8], the Stern–Gerlach type of effect for torsion was compared to the usual magnetic
effect in [44], finding torsion effects 17 orders of magnitude smaller than the magnetic effect.
However, in this case, the torsion was taken to be the gradient of a scalar field, and this result
cannot hold for all theories of torsion. Nevertheless, it is probably true that when a torsion effect
is in direct competition with an electromagnetic effect, the torsion effects will be hopelessly
swamped.
Singh and Ryder [183] performed a detailed analysis starting from the Dirac equation
written in the form
∂ψ
ih̄ = H ψ. (210)
∂t
Generalizing the torsion-free procedure of [99], they perform the Foldy–Wouthuysen
transformation three times. This yields the torsion energy interaction terms explicitly and
they demonstrate several kinds of interactions. In addition to an overall contribution to the
energy, they show a coupling between the orbital angular momentum and torsion, and argue
that this should have an influence on neutron phase shifts, which have been observed in [218].
634 R T Hammond

They also derive a spin coupling and discuss the nature of the coupling that arises with the
zero component of the axial torsion vector. These are very intriguing results, and may give
direction in the search for observable effects of torsion, however, the nature of the torsion field
is left unaddressed.
A very interesting idea has been put forth in [29] that also capitalizes on the sensitivity
of phase measurements, and may even measure torsion of the non-propagating type. Starting
with the Berry phase formula,

d
β = i ψ(t)| |ψ(t) dt (211)
dt
and using the Dirac equation in the form (210), it is shown that torsion gives rise to a phase
shift in the Dirac particle. As expected the effect is small, and the authors do not give any
details about the nature of the torsion source.
Sometimes the effects of torsion may be indirect. For example, in [54] the anomalies in the
standard model in curved space with torsion are studied. Quantum anomalies are also studied
in [9]. While the baryonic anomaly is unaffected, the leptonic current anomaly (assuming left-
handed neutrinos) is changed due to torsion. It is then argued that this effect may lead to new
physical effects. Kleinert argues that torsion should couple to the electroweak vector field, and
discusses how the Meissner–Higgs field arises with scalar torsion [124]. The effect of torsion
in three dimensions in topologically massive gravity with propagating torsion from a Chern–
Simons term has been investigated, but it is difficult to extrapolate to four dimensions [20].
An important element in any theory is the concept of self-consistency and anomalies. The
chiral anomaly demands particular attention with torsion, and is discussed very carefully by
Mielke in [149], and also one should consult [127] and [193]. The trace anomaly in quantum
gravity induced by the conformal factor is studied in [7].
In a generalization of the Brans–Dicke theory, torsion derived from a potential is
considered by Dereli and Tucker in [52]. They focus on the distinction between the geodesic
equation and the autoparallel, which are in general different (they are identical when the torsion
is totally antisymmetric, but do not represent the equation of motion [81]). The original Brans–
Dicke theory was formulated so that the principle of equivalence would be satisfied, but it is
possible to formulate the theory in a more general way. Dereli and Tucker use the orbit of the
planet Mercury to act as a test bed for geodesic motion as compared to autoparallel motion,
and suggest that the autoparallel with torsion describes the planet’s motion.
Another area where putative effects of torsion have been investigated concerns neutrino
oscillations, which has also been studied in a teleparallel theory in [225]. The physical idea
rests on the notion that, if we label the mass eigenstates of the neutrino by |νm , m = 1, 2, 3,
then the flavour eigenstates |νf  where f represents e (electron), µ (muon), and τ , are given
by

|νf  Umf |νm  (212)
m
where Umf is the unitary neutrino mixing matrix. Now consider the Dirac equation (210):
this describes the evolution of the state in time, and the Hamiltonian can be responsible for
changing the flavour of the neutrino in time, or equivalently, through distance. For example,
in [2], Ahluwalia considers the Hamiltonian
h̄2 2 GMm 1
H = mc2 − ∇ − − s·g (213)
2m r 2
where

2G J − 3(J · r̂)r̂
g= 2 (214)
c r3
Torsion gravity 635

is the gravitomagnetic field of the Earth, J is the rotational angular momentum of the Earth
and s is the spin of the neutrino. As time evolves, a phase develops according to, from (210),

ψ(t) = e−i/h̄ H dt
ψ(0). (215)
In the absence of gravity, for a two-neutrino system, the induced phase is φ = (m2 −
m1 )c2 t/2h̄. Gravity introduces a phase proportional the gravitational potential, and the effect
of the rotation is also considered by adding a gravitomagnetic interaction. These results are
very interesting in their own right, but we should move on to torsion. For example, torsion
may be included by replacing the rotational angular momentum of the Earth in (213) by the
torsion, say as in (158). For a neutrino travelling through the sun, for example, the cross section
should be derived and integrated over to obtain the total oscillation length. This calculation is
patiently waiting to be done.
This result was generalized to include torsion in [1], who use (78) as the starting point in
a Schwarzschild geometry, and take the torsion to be a given background field. The explicit
neutrino oscillation dependence on torsion was calculated and discussed, and as expected, was
found to be small. However, in this work the fixed background torsion was assumed to be of
the order of 10−60 cm−1 . However, we see from (158) that torsion is much larger than this
at nuclear and even atomic distances, and from (208) we see that torsion may be much larger
than even this. Thus, although we see that torsion will contribute to neutrino oscillations, the
calculations presented so far are not realistic enough to give us confidence that they give the
whole story, and in fact, torsion effects can be much larger than the scope of current calculations.
Additional quantum observable effects has been investigated in [13, 14], and [186].
An important campaign in the fight to find observable consequences of torsion, or to rule
out possible kinds of torsion, has been waged by Lämmerzahl and co-workers. General tests
of general relativity using quantum mechanics, such as atom interferometry in Hughes–Drever
-type experiments, starting from a generalized Dirac equation, have been described in [130]
and possible couplings to the electromagnetic field has been investigated in [131]. A review
of the principle of equivalence and various kinds of violations are given in [132]. It is pointed
out that there are more possible violations of the principle in the quantum domain than in the
classical domain, and Hughes–Drever, g − 2, and spin polarized material tests are studied.
Constraints on the axial torsion and anomalous couplings have been obtained in [133].
Now let us turn to a specific theory where torsion is derived from a tensor potential, as
described in section 3.2, where the torsion is given by (208). This may be applied to hydrogen,
where the proton and the electron experience a spin–spin interaction over and beyond that
produced by the magnetic spin–spin interaction. It has been shown, assuming the interaction
produces an energy shift smaller than about 10−10 eV (otherwise it would be observed), that κ
is no greater than 3 × 1017 , and using the more sensitive experimental results of Ni [157] one
obtains an upper bound of the order of 1014 [81]. This is discussed in more detail below.
Another application focuses on the possible spin flip of neutrinos. If they are massive, then
a left -handed neutrino may be flipped into a right-handed neutrino. Since these right-handed
neutrinos do not interact within the standard model, they would not be detected in Earth-bound
neutrino detectors. Due to their inability to spawn interactions, they have been dubbed sterile
neutrinos. The amplitude for the spin flip probability can be obtained from the Dirac equation
in the form (207), and is


Sf i = δf i − d4 x ψ f (x)γ5 γ µ bµ (x)ψi (x). (216)
4
If we assume that the initial wavefunction of the neutrino is spin up, and the final spin
down, then the invariant amplitude is given by
636 R T Hammond

i µ
M= j b µ (q ) (217)
4 5
with
µ
j5 = uf γ5 γ µ ui (218)
where

bµ (q ) = exp(iq · x)bµ (x) d3 x. (219)

The cross section for spin flipping was worked out in [79] and is

3κL2P 2
σ = 2πF (220)
4a
where F is the result of a numerical integration and is of the order of 4.
This result was applied to the case of neutrinos travelling outward through the Sun, and it
was found that the probability of spin flip is 50% if κ ≈ 1021 , seven orders of magnitude to big,
from Ni’s result. Thus, significant spin flipping cannot arise from torsion interactions unless
the force is non-universal, that is, unless the coupling constant is much bigger for interactions
between neutrinos and nucleons (or quarks) than between electrons and nucleons, which seems
unlikely.

7.3. Laboratory tests


The gauge invariance provided by (113) insures that the torsion quantum is massless, and
therefore gives rise to a long-range force. Thus, there has been considerable experimental
effort to detect, or bound, such a force in laboratory tests.
In [6], Anselm and Uratsev consider a Lagrangian of arions, massless Goldstone particles,
 √
L= xf mf ( 2GF )1/2 ψ f iγ5 ψf α(x) (221)
f

where xf are unknown parameters of order unity, α(x) is the Goldstone field and mf are√the
fermion masses. In the nonrelativistic limit they obtain (209) where g = xf mf (GF /8π 2)
and GF is the Fermi constant. Their idea was to generate the field by using an electromagnet
with a ferromagnetic core, and alternating the polarity at frequency ν. The detector is a
ferromagnetic material, the electrons of which will experience some degree of spin flipping
due to the arion field, and the resulting emf can be measured. The advantage of using high-
frequency ν is that the electromagnetic effects can be screened out using ordinary copper,
obviating the need of a superconductor. This method is essentially the same idea described
in section 3.2, but they did not have a power radiation formula such as (164) available.
Nevertheless, by using a tuned rf circuit, they estimate that a power as low as P ∼ 10−21 W
could be measured. They also briefly discuss the effect in hydrogen and the nuclear Zeeman
effect.
These ideas were put to the test a few years later in [216], who give the following numbers
for g:
g < 10−2 GF electron–quark interaction
g < 10GF electron–muon interaction (222)
g < 10GF electron–electron interaction.
In their experiment, they placed a Permalloy sample in a superconducting screen, and changed
the magnetization of a ferromagnetic body outside the screen, using a SQUID magnetometer
to measure the sample. They established the result that g < 10−3 GF .
Torsion gravity 637

Moody and Wilczek [151] investigate macroscopic forces that arise from a scalar and
pseudoscalar interaction. They derive a monopole interaction, a monopole–dipole interaction
and the dipole–dipole interaction. These forces are taken to be mediated by a massive particle,
which relates to the range in the classical potential energy formulae. This work is more closely
aligned with particle physics, such as the axion of the spontaneously broken Pecci–Quinn
symmetry, than torsion, but of course all of this work might apply to torsion as well.
In [220], these kinds of interactions were searched for using stored ion spectroscopy. In
particular, they looked for an anomalous energy dependence in the ground state hyperfine
transition of atomic 9 Be+ . They use the dipole–dipole interaction of [151] in the form (leaving
out the Dirac delta function term)

2

−mr m 1  m 3m 3 
V = Te + S·S − + 2 + 2 (S · r̂)(S · r̂) (223)
r2 r3 r r r
where T is the overall coupling constant. For long-range (or low-mass exchange particles,
λ = 1/m) forces, this is equivalent to the magnetic force, and they define α as the ratio of
the interaction given by (223) to the usual magnetic dipole–dipole interaction in this limit.
They tabulate the result of α molecular spectra experiments, NMR, acceleration and induced
magnetism, quoting values from 5×10−5 to approximately 10−14 . Although the term S in (223)
is proportional to intrinsic spin in the torsion interpretation, the authors also comment about
effects in which S might arise from angular momentum. It might also be appropriate to add that
a bound on the monopole–dipole σ · r̂ interaction resulting from the gravitational field of the
Earth and spins in atomic mercury has been reported in [213], who find that the gravitational
energy of this effect is less than 2.2 × 10−21 eV. This interaction is also examined in [173].
Although this is not viewed as a torsion effect, it has interest, as reported in [151] due to the fact
it is the only macroscopic force that violates both parity and time reversal invariance. Ni [157],
who has designed many beautiful experiments to detect, or bound torsion, and others [187],
also discussed the notion of testing for this force in a space-based environment, and ground-
based quantum effects have been investigated in [222]. There have also been experiments
which, while not directly looking for torsion, use spin polarized bodies to search for CPT [19]
violations, where torsion could put a different spin of the interpretations.
One method used to search for spin forces makes use of the torsion pendulum. At the
University of Washington a spin pendulum, pictured in figure 4, is constructed with different
materials so that the net alignment of electrons and therefore a net spin, but zero external
magnetic field.
The spin pendulum is put into the apparatus shown in figure 5 and rotates at constant
angular velocity. With any interaction the external mass distribution will create a time-varying
signal which would be measured by the co-rotating readout system.
Chui and Ni [39] pointed out that for very small applied fields, the permeability of
Permallow, referring to [216], would decrease if the energy required to rearrange a magnetic
domain is smaller than the thermal activation energy. To avoid this problem, Chui and Ni
used terbium fluoride (TbF3 ), a paramagnetic salt, which has a linear response and a large
susceptibility at 4.2 K, the operating temperature of the experiment. They assumed a potential
of the form (209) and included a table of α, as defined above, for various experiments. Their
work has the smallest value and is α = 2.7 × 10−14 , which means the spin interaction is at
least 2.7 × 10−14 of the strength of the magnetic dipole–dipole interaction. This value of α
leads to the value of κ quoted above. The concept of placing limits on torsion interactions
using gyroscopes was discussed in [200].
In [40], another series of experiments involve measuring, or bounding, the terms in a
Yukawa potential (see [40] for references)
638 R T Hammond

Figure 4. This is a drawing of the spin pendulum, which is essentially a torus made from one half
circle of AlNiCo magnets and another half circle of samarium–cobalt magnets. The former gets its
magnetism from spins and the later both from spins and from the electron’s orbital motion. When
these are combined in a torus the external magnetic field vanishes, but the pendulum retains a net
spin. The colours indicate the magnetic material (AlNiCo or SmCo).

Gmm
V =− (1 + αe−(r/λ) ) (224)
r
where λ is the range and α the coupling constant of the new force. Most of these experiments
have been discussed in terms of the ‘fifth force’, but in view of (111), they may also be applied
to a particular brand of torsion. However, as I have said before, it is generally believed that
torsion arises from intrinsic spin (and I agree) so let us return to the search for spin interactions.

7.4. Large-scale tests


So far we have looked at possible physical effects and manifestations of torsion that may
observable either at the quantum level or in Earth-bound laboratory tests. It is possible that
torsion may manifest itself on a larger scale, and a general formulation in terms of the PPN
formalism is given in [64]. They consider the Lagrangian

16π GL = R + R αβγ δ (a1 Rαβγ δ + a2 Rγ δαβ + a3 Rαγβδ )


+ R αβ (a4 Rαβ + a5 Rβα ) + a6 R 2 + S αβγ (a7 Sαβγ + a8 Sγβα ) + a9 Sσ S σ (225)

(note, as occurs in a few papers, they define the torsion tensor as twice that which is defined
here). They consider a Poincaré gauge theory and perform variations with respect to the metric
and torsion. However, they take the source tensor, which is the spin tensor (watch out—they
use Sαβγ for the source and Qαβγ for torsion) to be zero, and emphasize that, unlike PGT with
a linear Lagrangian only, torsion still propagates7 . They expand the metric as
7 Care must be taken with this argument. It still may be true that the source has to be non-zero somewhere in order

to have fields from a localized region.


Torsion gravity 639

Figure 5. The apparatus is shown with some of its thermal panels removed. In the centre of the
picture the lid of the vacuum vessel may be seen lifted from its normal position and supported
on a temporary fixture. Above the lid is an ion pump (coloured red) and behind it is the tube
which provides the upper support for the pendulum’s fibre (which is not yet installed). The spin
pendulum hangs in the gold coloured cage below the lid and is normally surrounded by layers
of electrostatic and magnetic shielding. Near the back is a mirror which reflects the laser beam
from the position readout system into the pendulum. The vacuum vessel sits on a turntable at the
bottom and rotates at a constant angular velocity when making the measurement. Courtesy of the
University of Washington, Seattle.

 n
gµν = ηµν + hµν (226)
n=1
n
where hµν = O; n and
 n
Sαβγ = S αβγ (227)
n=1
n
and S αβγ = O; n . The authors compute the post-Newtonian approximation to order ; 2 , but note
that the totally antisymmetric part of the torsion, which would couple to Dirac particles, enters
640 R T Hammond

at the ; 3 level. Post-Newtonian parameters were also calculated for torsion that is derived from
a scalar field in [58] and it was shown that there are no inconsistencies at the scale of the solar
system.
An astrophysical source might also display torsion effects. For example, the magnetic
field may arise from, or be partially due to, a ferromagnetic state in which the neutrons are
aligned [188]. This alignment would result in the creation of a torsion field. Although the
associated gravitational and electromagnetic field are also strong, this may nevertheless create
a region in which torsion can have measurable effects. For example, wraith-like neutrinos,
which are almost immune to other forces, may fall prey to the elusive touch of torsion. Another
possibility is the radiation of energy due to torsion. Although a formula was presented that
describes the radiation from spin flips, the radiation from a spinning neutron polarized star
has not been calculated. However, it has been has shown how torsion can play another role in
pulsars [37]. Consider β to be defined as the angle between the rotational axis of a pulsar and
the polar magnetic axis. It is known that this angle decreases in time, typically by 10◦ in ten
million years. Part of the reason is due to special relativity [143], the self-magnetic interaction
and torsion. More detailed calculations need to be done, however, to assess the contribution
made by torsion. In another role, it is shown that that torsion forces decrease the effect of
gravity and have a significant influence on the mass of the neutron star [22]. Solar system
effects based on torsion derived from the gradient of a spacetime-dependent G, including the
use of the Viking lander ranging data for temporal variations, have been investigated in [195].

7.5. Cosmological tests


Perhaps it is fitting to end this section with the beginning—the beginning of the Universe. It
was recognized almost 30 years ago that torsion could have profound influence on the very
early evolution of the Universe, and in fact it was shown that spin fluid models could prevent
the singularity and give rise to a bounce Universe, so to speak [125, 199, 205]. However, soon
thereafter Kerlick [118] showed that using the Dirac coupling (instead of using spin fluids)
leads to an enhancement of singularity formation. In addition, he showed that torsion can play a
dominate role in particle creation in the early Universe. From this point of view, one might not
only argue that if it were not for torsion, this paper could not have been written, but furthermore,
there might be no one to read it! A brief review of the early work has been given by Wolf [221]
who studies the effect of torsion on the scale factor in Robertson–Walker cosmology. Although
he studies a special case of radial torsion, calculations show that torsion not only effects the
timescale factor, but can also prevent the singularity upon collapse. Brüggen [24], for torsion
given by (112), investigates the effect of torsion-induced spin flips of neutrinos in the early
Universe. Being made sterile, such neutrinos could alter the helium abundance and have other
effects on the early nucleosynthesis of the Universe. It was calculated, however, even with
large non-minimal coupling, such effects are too small to alter the observed abundances. This,
of course, comes as good news for torsion enthusiasts, for otherwise this torsion would have to
be ruled out. A singularity was also found to be absent when torsion is derived from a scalar
potential and a parity violating term is introduced into the Lagrangian in [107]. The effect of
torsion that arises from the gradient of a spacetime-dependent G was investigated in [195].
The effect of torsion and vacuum quantum fluctuations on cosmological singularities has
been investigated in [25] and phase transitions with torsion induced by external fields was
investigated in [26]. The renormalization group equations in U4 space was examined in [27].
A more general investigation of the role of torsion as regards the strength of the singularity
is given in [111], and the effect of torsion on wormhole dynamics and energy conditions is
discussed in [102].
Torsion gravity 641

7.6. Exact solutions


When torsion is finally measured, one wonders, will it be from a gradual accumulation of
evidence that eventually crushes all doubt, like the discovery of the first black hole, or will it
be relatively quick, like the discovery of the positron or the J /ψ particle? Either way, a firm
underlying theory with hard predictions seems helpful, if not mandatory, and exact solutions,
while not direct evidence, may provide the definitive clues in the observational discovery of
torsion. In this brief section I will give a cross section of a few exact solutions or methods,
which also provide references to a more complete set.
Some of the original methods used to find exact solutions with torsion may be found in [89]
and [90]. A static dust sphere solution of a Schwarzschild type was discovered in [197] in which
the spins were aligned radially, but since the pressure was zero at the centre the solution was
found to be unphysical. Subsequently, a dust model with the spins taken to be zero at the centre
and surface was constructed which yielded a pressure that was maximum at the centre [146].
A Schwarzschild-like solution (i.e. a solution with corrections to the Schwarzschild form) was
given in [217]. In fact, this solution contains confinement potentials of the form r 2 and r 2 ln r
that may be important on the microscopic scale
An exact cosmological solution is given in [207]. In this, symmetries are imposed on
the torsion tensor that lead to a Robertson–Walker metric, and the effects of torsion, in
principle, could be measured by cosmological-type observations, i.e. the Hubble constant
and the deceleration parameter. However, the symmetry imposed upon the torsion disallow its
interpretation as arising from intrinsic spin, since it becomes associated with the four velocity.
Spherically symmetric solutions and a cosmological solution with torsion were reported in [12].
On a very different track, in [144] all possible quadratic actions with torsion are considered
which yield unique solutions for appropriate Cauchy data. In two dimensions, R 2 gravity with
propagating torsion is considered in [128]. Exact torsion solutions corresponding to a de Sitter
type, and other solutions were found using light cone coordinates and conformal coordinates.
Exact solutions for a non-Abelian antisymmetric tensor field are given in [123].
An exact solution in the metric affine theory is given in [215]. The solution is axially
symmetric and carries torsion as well as nonmetricity, and a spherically symmetric solution
was reported in [158]. Other solutions may be found in [96] and [97]. In the context of the
cosmology of the Gödel type, a class of solutions is discussed in [4]. The solutions are very
general, and avoid the adoption of a particular kind of torsion, although it is assumed that the
torsion is polarized along a preferred direction. Torsion in the Gödel Universe, using computer
algebra programs, is discussed in [59]. Under the assumption that the torsion and nonmetricity
can be obtained from a certain class of 1-forms, it is shown in [51] that solutions with torsion
and nonmetricity can be generated from solutions of the Einstein–Maxwell, or Einstein–Proca
type. Using a ten-parameter quadratic Lagrangian, [10] discusses the meaning of torsional
plane waves, and gives solutions.

8. String theory and torsion

The idea of incorporating torsion and cosmic strings has a rich history. If one considers a
long, thin (but not distributional) string with an interior and exterior, it is natural to assume
that the interior may have a torsion source, usually but not always assumed to be a spin density
source, as in [194], and for a more detailed discussion, including the development of a Poincaré
gauge theory, one should consult [196] and the references. It has also been shown that conical
singularities can give rise to distributional sources of torsion [203], and it has been shown that
torsion can exist as a distributional source where the curvature tensor vanishes [136]. This work
642 R T Hammond

was extended to thick strings in [137]. Torsion in domain walls and inflation was investigated
in [43], and, within a principle of equivalence for quantum gravity, torsion appears within the
context of cosmic strings in [5]. Electric and magnetic self-forces with a distributional torsion
are considered in [33] and a general argument concerning affine defects and Einstein–Cartan
theory is presented in [168]. In addition, the notion of spinning strings and cosmology has
been considered by many authors, see [190] and [108] and the references therein.
In this section, however, I would like to zero in on the relationship between the
antisymmetric field of string theory and the torsion of spacetime as given by (112), i.e.
Sµνσ = ψ[µν,σ ] . We have seen that although this theory of torsion adopts the anzatz (112), many
theories of torsion do not, including the Poincaré gauge theory. For this reason, some people
camping in the general relativity park do not see torsion as the antisymmetric field of string
theory, and similarly, some camping in the string theory park do not see the antisymmetric or
Kalb–Ramond field as torsion of spacetime. In this section I will forge ahead assuming that
torsion is given by (112), but urge the reader to bear in mind that these remarks are mostly
based on this one specific theory.
The first big clue came along in (182) which exhibits both the antisymmetric field and the
scalar field explicitly. In the context of a metric theory of gravity (∇σ gµν = 0) this idea was
further refined as in [79], which predicts the precise form of the low-energy effective Lagrangian
of string theory. In addition, when the phenomenological source (119) was introduced, which
was based on the intrinsic vector ξ µ , it was found that the source could not be a point,
but had
µ
to be extended in space, and furthermore that under some general conditions that n ξn = 0
where the sum over n is was taken to be a sum over infinitesimal parts of the body. It may be
shown that this condition arises naturally when the object is taken to be a closed string.
Thus, since both the geometrical and material Lagrangian of gravitation with torsion and
string theory share common features, it is natural to adopt from the outset a theory in which
the material action is given by a string. As usual we take
δ(Igeom + IM ) = 0 (228)
where

√ R
Igeom = −g d4 x (229)
2k
but now assume that
 

IM = µ −γ d2 ζ + η ψµν dσ µν . (230)

In this, ζ0 is the time-like string coordinate and ζ1 is the space-like coordinate, and in this
section a, b sum from 0 to 1. The two-dimensional metric γab is related to the four-dimensional
gµν according to
µ ν
γab = gµν x,a x,b . (231)
The first term in (230) is the Nambu–Goto term, and (231) shows that this gives rise to a
symmetric energy–momentum tensor. The second term is called the Kalb–Ramond term and
introduces torsion into the theory. The string surface element is given by
dσ µν = ; ab x,a
µ ν 2
x,b d ζ (232)
and

√ 0 −1
−γ ; ab = . (233)
1 0
Thus, the second term in (230) provides a very natural coupling between the string and the
torsion potential.
Torsion gravity 643

To obtain the field equations, we perform independent variations of the metric and torsion
potential as before, and obtain the same equations:
µνσ µ
Gµν − 3S ;σ − 2S αβ S ναβ = kT µν (234)
but now the energy–momentum tensor is

µ √
T σν = √ σ ν
d2 ζ −γ δ(x − x(ζ ))x,a x,b (γ ab + λ; ab ) (235)
−g
where λ = 2η/µ.
The first thing I would like to show is that, in this formulation and with no further
assumptions, a closed string automatically gives rise to intrinsic spin. To see this, first let
us consider a string in a given external field, and take (235) to be a test string in the external
field. Then (235) becomes

α β
T αβ → µ dζ δ 3 (x − x(ζ ))x,a x,b (ηab + λ; ab ) (236)

where ηab is the two-dimensional Minkowski metric (this results from the approximation that
the external field is weak). Also, ζ 0 is taken to coincide with the time coordinate of the four-
dimensional spacetime (ζ 0 = x 0 ) and ζ 1 ≡ ζ . Using (236) in the definition of the total angular
momentum tensor (140) and defining the spin vector as before, i.e. as in (148), we have in the
v µ → 0 limit,
Sγ → 21 J αβ ;0αβγ (237)
 β
dx
= 2 dζ ;0αβγ δx α . (238)

Since the equations of rotational momentum are the same as those derived in section 3.2,
this result shows that Sγ is the intrinsic of a particle, and therefore that a string automatically
has spin, and gives rise to torsion. Note that this spin does not come from any motion of
the string, but is due solely to its structure. In fact, if we assume that the string is in a static
configuration (238) gives

S = 2η r × r  dζ ⇒ S = 4η × Area (239)

where r  = dr /dζ , and r is a vector from the centre of mass to a point on the string.
Of course, we have no right to assume that the string is in a static configuration, but (238) holds
nevertheless. The point is, with torsion, strings naturally have spin. The only assumption other
than the natural Kalb–Ramond coupling is that the string is closed (and small). The equations
of motion and the Bianchi identity for strings have been considered in [82], but the effects of
torsion have not yet been fully developed.
Another interesting result is that the intrinsic vector ξ µ that had to be invented in order
to construct the phenomenological matter action (119) may now be interpreted as the tangent
vector to the string. One may argue that this result might have led to the introduction of strings
independently from a purely classical motivation, but I think this is stretching a point.
The equation of motion of the string may also be derived. Conventionally, one often
considers the geodesic postulate, which takes coordinate variations of the string material action.
However, as stressed above, the equations of motion are derived from the field equations, so
let us see how this works without torsion in lowest order, in which case (141) becomes
d Mσ0 σ
+ {αβ }M αβ = 0 (240)
dτ v 0
where
644 R T Hammond
 
√ α β
M αβ ≡ v 0 d̃3 x T αβ = v 0 µ dζ −γ γ ab x,a x,b . (241)

We consider the centre of mass of the small object to be located at the coordinate y µ and that
xµ = yµ + ;µ (242)
µ µ
where ;  y , so that


M αβ = v 0 µ −γ dζ (γ 00 (ẏ α ẏ β + ;˙ α ẏ β + ;˙ β ẏ α + ;˙ α ;˙ β ) + γ 11 ; α ; β ). (243)

The dot represents the derivative with respect to time and the prime the derivative with respect ζ .
As it stands, (243), with (240) gives rise to a very unusual looking equation of motion, but
things can be reduced to a more acceptable form if we look at the time average of (240),
d M σ 0  σ
+ {αβ }M αβ  = 0 (244)
dτ v 0
where the time average of M αβ is defined by

1
M αβ  ≡ dt M αβ . (245)
T
Assuming that the limits of integration range over one period we obtain,
 
µv α v β µv α v 0
M αβ  = 0 d2 ζ γ̃ 00 + d2 ζ γ̃ 00 ;˙ β
v T T
 
µv β v 0 v0 µ √
+ d2 ζ γ̃ 00 ;˙ β − d2 ζ −γ ; α  ; β (246)
T T
where integrations by parts were used, and the tilde represents density with respect to the
two-dimensional metric.
It is well known that, with structure, the equations of motion are underdetermined. For
example, the Bianchi identities give a set of four equations, which is just right for a point particle
in four dimensions, but with structure present, then these four conditions are insufficient to
completely determine the motion. Usually some other condition is imposed, such as the
constraints S µν vν = 0 or S µν pν = 0. Here we shall investigate the consequences of the
constraint  ; β = 0. With this, after some integrations by parts it may be shown that [82]

µv α
M α0  = d2 ζ γ̃ 00 . (247)
T
This leads to the definition of mass of the string as
 
µ µ
m ≡ 0 dζ γ̃ = 0 00
d2 ζ γ̃ 00 (248)
v v T
so that
M αβ  = mv α v β . (249)
With this, (240) becomes
dv σ σ
+ {αβ }v α v β = 0, (250)

a sensible result: when the rapid string fluctuations are averaged out, the string travels along,
in an external field that does not appreciably change over the length and timescales of string,
a simple geodesic.
The next calculation is to redo this but with torsion included. The expectation is that the
equations of motion discussed in section 6, under the averaging process, should be obtained.
Such a result would further confirm the intricate relation between torsion and spin.
Torsion gravity 645

The main point is this: there seems to be either sufficient (from the conservative view)
or compelling theoretical evidence, arising from similarities and common building blocks, to
view torsion of spacetime and the antisymmetric field in string theory to be one and the same8 .
If this association is correct, then it not only gives the antisymmetric field of string theory a
definite interpretation, but any future measurement of torsion would automatically yield an
indirect observation of string theory.

9. Closing thought

Engaging us in extraordinary ways, Nature can be wonderful, exciting, dangerous, perplexing,


nurturing, deadly, comprehensible and incomprehensible. There is a wonderful beauty and
reward to our comprehension of the Universe, but it is balanced by the intrigue and mystery
of yet unsolved problems. Certainly torsion is woven into the fulcrum of this balance.
Some people argue that torsion is not elusive at all, and with every step forward we feel it,
pushing us along. This is the teleparallel view, where matter is described by torsion naturally,
with beauty and grace, in purely geometric terms. Others fret torsion is too small to be seen,
but like an electron, is a fundamental building block of nature. This is the gauge theory point
of view, where torsion enters on an equal footing with gravity, the equivalence broken only by
disparate coupling constants.
We have seen that torsion is called on stage by many directors, from string theory to
supergravity, yet the audience has not yet settled on the correct interpretation of its role.
I believe any direct, or even indirect, observation of torsion would be one of the greatest
breakthroughs in many decades, and would certainly help settle these questions. I hope this
paper will invite and develop interest from both theoreticians and experimentalists to join the
quest in the search for torsion.

Acknowledgments

I am deeply indebted to Friedrich Hehl, not only for his willingness to help and to correct
many errors, but for his patience in explaining both physics and historical developments to
me. I would also like to thank Miluten Blagojević, Sean Carroll, Garcia de Andrade, Michael
Katanaev, Hagen Kleinert, Claus Lämmerzahl, Eckehard Mielke, Yuval Ne’eman, José Pereira,
Terry Pilling, Ilya Shapiro, Larry Smalley, Erik Swanson and Robin Tucker, for their help and
contributions.

References

[1] Adak M, Dereli T and Ryder L H 2001 Preprint gr-qc/0103046


[2] Ahluwalia D V 1997 Preprint gr-qc/970505
[3] Alvarez E 1989 Rev. Mod. Phys. 61 561–603
[4] Aman J E, Neto J B, MacCallum M A H and Reboucas M J 1998 Class. Quantum Grav. 15 1089–101
[5] Anandan J 1996 Phys. Rev. D 53 779–86
[6] Anselm A A and Uraltsev 1982 Phys. Lett. B 116 161–4
[7] Antoniadis I and Odintsov S D 1993 Mod. Phys. Lett. A 8 979–85
[8] Audretsch J 1981 Phys. Rev. D 24 1470–7
[9] Aurilia A and Spallucci E 1990 Phys. Rev. D 42 464–8
[10] Babourave O V, Frolov B N and Klimova E A 1999 Class. Quantum Phys. 16 1149–62
[11] Bade W L and Jehle H 1953 Rev. Mod. Phys. 25 714–28
8 Although it should be mentioned that the equivalence begins to break down at high energy, where the antisymmetric

field of string theory is no longer given simply by (112).


646 R T Hammond

[12] Baekler P, Mielke E W, Hecht R and Hehl F W 1987 Nucl. Phys. B 288 800–12
[13] Bagrov V G, Belov V V, Trifonov A and Yevseyevich A A 1991 Class. Quantum Grav. 8 1349–59
Bagrov V G, Buchbinder I L and Shapiro I L 1992 Sov. J. Phys. 36 5–12
[14] Belyaeb A S and Shapiro I L 1999 Nucl. Phys. B 543 20–46
[15] Benn I M, Derelli T and Tucker R W 1980 Phys. Lett. B 96 100–4
[16] Blagojević M and Vasilic M 2000 Class Quantum Grav. 17 3785–97
[17] Blagojević M 2001 Gravitation and Gauge Symmetries (Bristol, UK: Institute of Physics Publishing) at press
[18] Blagojević M and Vasilić M 2001 Phys. Rev. D 64 044010-1–12
[19] Bluhm R and Kostelecky V 2000 Phys. Rev. Lett. 84 1381–4
[20] Boldo J L, de Moraes L M and Helayel-Neto J A Class Quantum Grav. 17 813–23
[21] Borzeszkowski H and Treder H 1997 Gen. Rel. Grav. 29 455–66
[22] Boyadjiev T, Fiziev P and Yazadiev S 1999 Class. Quantum Grav. 16 2359–80
[23] Brill R and Jeffrey M 1966 J. Math. Phys. 7 238–43
[24] Brüggen M 1999 Gen. Rel. Grav. 31 1935–40
[25] Buchbinder I L, Odintsov S D and Shapiro I L 1985 Phys. Lett. B 162 92–6
[26] Buchinder I L, Odintsov S D and Shapiro I L 1985 Sov. J. Phys. 30 3–9
[27] Buchinder I L and Shapiro I L 1985 Class. Quantum Grav. 7 1197–206
[28] Burinskii A 1999 Class. Quantum Grav. 16 3497–516
[29] Capozziello S, Lambaise G and Stornaiolo C 1999 Europhys. Lett. 48 482–5
[30] Carrol M and Field B 1994 Phys. Rev. D 50 3867–73
[31] Cartan E 1922 C. R. Acad. Sci., Paris 174 593
Cartan E 1922 Ann. Ec. Norm. Sup 40 325
Cartan E 1922 Ann. Ec. Norm. Sup 42 17
[32] Cartan E and Einstein A 1979 Letters of Absolute Parallelism (Princeton, NJ: Princeton University Press)
[33] Carvalho A M, Furtado C and Moraes F 2000 Phys. Rev. D 62 067504-1–4
[34] Chamseddine A H and West P C 1977 Nucl. Phys. B 129 39–44
[35] Chandía O and Zanelli J 1998 Phys. Rev. D 58 045014-1–4
[36] Chang L N, Lebedev O, Loinaz W and Takeuchi T 2000 Phys. Rev. Lett. 85 3765–8
[37] Cheng-min Z, Guo-chen Y, Fang-pei C and Xin-ji W 1992 Gen. Rel. Grav. 24 359–71
[38] Cho Y M and Yoon J H 1993 Phys. Rev. D 47 3465–73
[39] Chui T C P and Ni W 1993 Phys. Rev. Lett. 71 3247–50
[40] Cowsik R, Krishnan N, Tandon S N and Unnikrisknan S 1990 Phys. Rev. 64 336–9
[41] Cremmer E and Scherk J 1974 Nucl. Phys. B 72 117–24
[42] Damour T, Deser S and McCarthy J 1993 Phys. Rev. D 47 1541–56
[43] de Andrade L C 1999 Class. Quantum Grav. 16 2097–103
[44] de Andrade V C 1993 Nuovo Cimento B 108 947–8
[45] de Andrade V C and Pereira J G 1997 Phys. Rev. D 56 4689–95
[46] de Andrade V C and Pereira J G 1998 Gen. Rel. Grav. 30 263–74
[47] de Andrade V C, Guillen L C T and Pereira J G 2000 Phys. Rev. Lett. 84 4533–6
[48] de Andrade V C, Guillen L C T and Pereira J G 2000 Phys. Rev. D 61 084031-1–4
[49] de Sabbata V and Gasperini M 1981 Phys. Rev. D 23 2116–20
[50] de Sabbata V and Gasperini M 1983 Unified Field Theories of More Than 4 Dimensions ed V de Sabbata and
E Schmutzer (Singapore: World Scientific) pp 152–70
[51] Dereli T, Onder M, Schray J, Tucker R and Wang C 1996 Class. Quantum Grav. 13 L103–9
[52] Dereli T and Tucker R W 2001 Preprint gr-qc 0104050
[53] DeWitt B 1965 Dynamical Theory of Groups and Fields (New York: Gordon and Breach)
Stelle K S 1977 Phys. Rev. D 16 953
Szczryba V 1987 Phys. Rev. D 63 351 and the many references therein
[54] Dobado A and Maroto A 1996 Phys. Rev. D 54 5185–94
[55] Dobado A and Maroto A 1999 Class. Quantum Grav. 16 4057–74
[56] Einstein A 1956 The Meaning of Relativity 5th edn (Princeton, NJ: Princeton University Press)
[57] Fiziev P 1998 Gen. Rel. Grav. 30 1341–70
[58] Fiziev P and Yazadjiev S Class. Quantum Grav. 16 3133–6
[59] Fonseca-Neto J B and Reboucas M J 1998 Gen. Rel. Grav. 30 1301–18
[60] Garcia A, Hehl F, Lämmerzahl C, Macias A and Socorro J Class. Quantum Grav. 15 1793–9
[61] Garcia A, Macias A, Puetzfeld D and Socorro J 2000 Phys. Rev. D 62 044021-1–7
[62] German G, Mascias A and Obregon O 1993 Class. Quantum Grav. 10 1045–53
[63] Giannopoulos A 1991 Class. Quantum Phys. 8 571–85
Torsion gravity 647

[64] Gladchencko M S and Zhytnikov V V 1994 Phys. Rev. D 50 5060–71


[65] Grignani G and Nardelli G 1992 Phys. Rev. D 45 2719–28
[66] Grignani G and Nardelli G 1993 Phys. Rev. D 48 5032–5
[67] Gronwald F and Hehl F W 1995 Erice, Quantum Gravity ed P G Bergmann, V de Sabbatra and H-J Treder
(Singapore: World Scientific) pp 148–98 (gr-qc/960213)
[68] Gruver C, Hammond R T and Kelly P 2001 Mod. Phys. Lett. A 16 113–19
[69] Guadagnini E, Maddiore N and Sorella S P 1991 Phys. Lett. B 255 65–73
[70] Hammond R 1988 J. Math. Phys. 30 1115–16
[71] Hammond R 1988 Gen. Rel. Grav. 20 813–27
[72] Hammond R 1990 Class. Quantum Grav. 7 2107–12
[73] Hammond R 1990 J. Math. Phys. 31 2221–4
[74] Hammond R 1991 Gen. Rel. Grav. 23 973–80
[75] Hammond R 1991 Class. Quantum Grav. 8 L75–7
[76] Hammond R 1993 Nuovo Cimento 108 725–38
[77] Hammond R 1994 Gen. Rel. Grav. 26 247–63
[78] Hammond R 1995 Contemp. Phys. 36 103–14
[79] Hammond R 1996 Class. Quantum Grav. 13 L73–9
[80] Hammond R 1997 Gen. Rel. Grav. 29 727–31
[81] Hammond R 1999 Gen. Rel. Grav. 31 233–71
[82] Hammond R 2001 String motion in curved space and the Bianchi identity Gen. Rel. Grav. 33 1897
[83] Hartley D 1995 Class. Quantum Grav. 12 L103–5
[84] Hayashi K and Shirafuji T 1979 Phys. Rev. D 19 3524–53
[85] Hayashi K and Shirafuji T 1980 Prog. Theor. Phys. 64 866–82
Hayashi K and Shirafuji T 1980 Prog. Theor. Phys. 64 883–96
Hayashi K and Shirafuji T 1980 Prog. Theor. Phys. 64 1435–52
[86] Hayashi K, Nomura K and Shirafuji T 1990 Prog. Theor. Phys. 84 1085–99
[87] Hecht R D, Lemke J and Wallner R P 1991 Phys. Rev. D 44 2442–51
[88] Hehl F W and Datta B K 1971 J. Math. Phys. 12 1334–9
[89] Hehl F W 1973 Gen. Rel. Grav. 4 333–49
[90] Hehl F W 1974 Gen. Rel. Grav. 5 491–516
[91] Hehl F W, von der Heyde P, Kerlick G D and Nester J M 1976 Rev. Mod. Phys. 48 393–416
[92] Hehl F, Nitsch J and von der Heyde P 1980 General Relativity and Gravitation ed A Held (New York: Plenum)
pp 329–55
[93] Hehl F 1985 Found. Phys. 15 451–68
[94] Hehl F and McCrea J D 1986 Found. Phys. 16 267
[95] Hehl F W, McCrea J D, Mielke E W and Ne’eman Y 1995 Phys. Rep. 258 1–171
[96] Hehl F W, Obukhov Y, Tresguerres R and Vlachynsky E J 1996 Class. Quantum Grav. 13 3253–9
[97] Hehl F W and Ne’eman Y 1997 Class. Quantum Grav. 14 A251–9
[98] Hehl F W and Obukhov Y 2000 Preprint arXiv:gr-qc/0001010 pp 1–26
[99] Hehl F W and Ni W-T 1990 Phys. Rev. D 42 2045–8
[100] Hehl F W and Macias A 1999 Int. J. Mod. Phys D 8 399–416 (gr-qc/9902076)
[101] Hlavaty 1957 Geometry of Einstein’s Unified Field Theory (Grinlinger, Holland: Noordhoff)
[102] Hochberg D and Visser M 1998 Phys. Rev. D 58 044021-1–14
[103] Hojman S, Rosenbaum M and Ryan M P 1978 Phys. Rev. D 17 3141–6
[104] Hojman S, Marcos R and Ryan M Jr 1979 Phys. Rev. D 19 430–7
[105] Ivanov E A 1996 Phys. Rev. D 53 53–7
[106] Jha R 1993 Gen. Rel. Grav. 25 281–90
[107] Jha R 1994 Gen. Rel. Grav. 26 1067–74
[108] Krisch J P 1996 Gen. Rel. Grav. 28 69
[109] Kalb M and Ramond P 1974 Phys. Rev. 9 2273–84
[110] Kalmykov M and Pronin P I 1995 Gen. Rel. Grav. 27 873–86
[111] Kánnár J 1995 Gen. Rel. Grav. 27 23–34
[112] Katanaev M O 1993 Gen. Rel. Grav. 25 349–60
[113] Kao W F 1993 Phys. Rev. D 47 3639–42
[114] Katanaev M O, Kummer W and Liebl H 1996 Phys. Rev. D 53 5609–10
[115] Kawai T 1993 Phys. Rev. D 48 5668–75
[116] Kawai T 1994 Phys. Rev. D 49 2862–71
[117] Kemmer N 1938 Proc. R. Soc. London, Ser. A 166 127–153
648 R T Hammond

[118] Kerlick G D 1975 Phys. Rev. D 12 3003–6


[119] Kibble T W B 1961 J. Math. Phys. 2 212–21
[120] Kim C B and Lee H Y 1983 Phys. Lett. B 131 99–103
[121] Kim J, Park C J and Yoon Y 1995 Phys. Rev. D 51 562–7
[122] Kim J, Park C J and Yoon Y 1995 Phys. Rev. D 51 4595–7
[123] Kim H 1996 Phys. Lett. B 369 221–5
[124] Kleinert H 1998 Preprint gr-qc/9808022 pp 1–4
[125] Kopczyński W 1972 Phys. Lett. A 39 219
[126] Kopczyński W 1982 J. Phys. A: Math. Gen. 15 493–506
[127] Kreimer D and Mielke E 2001 Phys. Rev. D 63 048501-1–4
[128] Kummer W and Schwarz D 1992 Phys. Rev. D 45 3628–35
[129] Kummer W and Tieber G 1998 Phys. Rev. D 59 044011-1–10
[130] Lämmerzahl C 1998 Class. Quantum Grav. 15 13
[131] Lämmerzahl C 1997 Class. Quantum Grav. 14 1347
[132] Lämmerzahl C 1997 Quantum tests of spacetime structure Proc. Int. School of Cosmology and Gravitation
ed V de Sabbata, G Gillies and P Pronin
[133] Lämmerzahl C 1997 Phys. Lett. A 228 223
[134] Lavrov P M and Moshin P 1999 Class. Quantum Grav. 16 2247–58
[135] Lemes V, Renan R and Sorella S P 1995 Phys. Lett. B 344 162–3
[136] Letelier P 1995 Class. Quantum Grav. 12 2221–3
[137] Letelier P 1995 Class. Quantum Grav. 12 471–8
[138] Lindström U 1985 Gen. Rel. Grav. 18 845–8
[139] Luehr C, Rosenbaum M, Ryan M and Shepley L 1977 J. Math. Phys. 18 965–70
[140] Macias A, Mielke E and Socorro J 1998 Class. Quantum Grav. 15 445–52
[141] Maluf J W and Santos-Silva G 1993 Gen. Rel. Grav. 25 653–62
[142] Maluf J W and da Rocha-Neto J F 2001 Phys. Rev. D 64 084014-1–8
[143] Mashhoon B 1988 Phys. Rev. Lett. 61 2639–42
[144] Mardones A and Zanelli J 1991 Class. Quantum Phys. 8 1545–58
[145] McCrea J 1992 Class. Quantum Grav. 9 553–68
[146] Mehra A L and Gokhroo M K 1992 Gen. Rel. Grav. 24 1011–14
[147] Mielke E W, Hehl F W and McCrea J D 1989 Phys. Lett. A 140 368–72
[148] Mielke E W 1992 Ann. Phys., NY 219 78–108
[149] Mielke E and Kreimer D 1998 Int. J. Mod. Phys. D 7 535–48
[150] Miyamoto S and Nakano T 1971 Prog. Theor. Phys. 45 295–311
[151] Moody J E and Wilczek F 1984 Phys. Rev. D 30 130–7
[152] Muench U, Gronwald F and Hehl F W 1998 Gen. Rel. Grav. 30 933–62
[153] Nambu Y 1976 Phys. Rep. C 23 250–4
[154] Ne’eman Y 1977 Proc. Natl Acad. Sci. USA 74 4157–9
Ne’eman Y 1978 Ann. Inst. Henri Poincaré A 28 369–78
[155] Ne’eman Y and Hehl F 1997 Class. Quantum Grav. 14 A251–9
[156] Ni W-T 1979 Phys. Rev. D 19 2260–3
[157] Ni W-T 1996 Class. Quantum Grav. 13 A135–41
[158] Obukhov Y, Vlachynsky E J, Esser W, Tresguerres R and Hehl F W 1996 Phys. Lett. A 220 1
[159] Obukhov Y, Vlachynsky E J, Esser W and Hehl F W 1997 Phys. Rev. D 56 7769–78
[160] Obukhov Y, Yu N and Tresguerres R 1993 Phys. Lett. A 184 17–22
[161] Ogievetskiĭ V I and Polubarinov I V 1967 Sov. J. Nucl. Phys. 4 156–61
[162] Okubo S 1991 Gen. Rel. Grav. 23 599–607
[163] Oliveira H P 1993 Gen. Rel. Grav. 25 473–81
[164] O’Raifeartaigh L 1997 The Dawning of Gauge Theory (Princeton, NJ: Princeton University Press)
[165] Pagels H 1984 Phys. Rev. D 29 1690–8
[166] Papapetrou A 1948 Proc. R. Soc. A 209 248–58
[167] Pathinayake C, Vilenkin A and Allen B 1988 Phys. Rev. D 37 2872
[168] Petti R J 2001 Gen. Rel. Grav. 33 209–17
[169] Poberii E A 1994 Gen. Rel. Grav. 26 1011–53
[170] Ray J R 1972 J. Math. Phys. 13 1451
Ray J R and Smalley L L 1982 Phys. Rev. Lett. 49 1059
[171] Rey S 1989 Phys. Rev. D 40 3396
[172] Ringermacher H 1994 Class. Quantum Grav. 11 2383–94
Torsion gravity 649

[173] Ritter R C, Winkler L I and Gillies G T 1993 Phys. Rev. Lett. 70 701–4
[174] Rohm R and Written E 1986 Ann. Phys., NY 170 454
[175] Rubilar G 1998 Class. Quantum Grav. 15 239–44
[176] Saa A 1997 Gen. Rel. Grav. 29 205
[177] Scherk J and Schwarz J H 1974 Phys. Lett. B 52 347
[178] Schouten J A 1954 Ricci Calculus (Berlin: Springer)
[179] Schrödinger E 1943 Proc. R. Ir. Acad. A 49 43–58
[180] Schrödinger E 1943 Proc. R. Ir. Acad. A 49 135–48
Schrödinger E 1948 Proc. R. Ir. Acad. A 52 1–8
Schrödinger E 1951 Proc. R. Ir. Acad. A 54 79–85
[181] Sciama D W 1964 Rev. Mod. Phys. 36 463–9
[182] Siegel W 1980 Phys. Lett. B 93 170–2
[183] Singh P and Ryder L H 1997 Class. Quantum Grav. 14 3513–25
[184] Sezgin E and van Nieuwenhuizen P 1980 Phys. Rev. D 22 301–7
[185] Shapiro I L 2001 Preprint hep-th/0103093
[186] Shapiro I L 1999 Photon and Poincaré Group, A Volume in: Contemporary Fundamental Physics
ed V V Dvoeglazov (Nova Science Huntingdon, NY) pp 278–97
[187] Shaul D, Sumner T J, Rocherster G K and Speake C C 1996 Class. Quantum Grav. 13 A107–12
[188] Silverstein S D 1969 Phys. Rev. Lett. 23 139–41
Silverstein S D 1969 Phys. Rev. Lett. 23 453 (erratum)
[189] Smalley L and Krisch J P 1991 J. Math. Phys. 33 1073–81
[190] Smalley L and Krisch J P 1999 Gen. Rel. Grav. 31 799–814
[191] Smalley L 2000 Class. Quantum Grav. 17 1447–53
[192] Sohnius M F and West P C 1981 Phys. Lett. B 105 353–7
[193] Soo C 2000 Phys. Rev. D 59 045006-1–7
[194] Soleng H H 1990 Class. Quantum Grav. 7 999–1007
[195] Soleng H H 1991 Gen. Rel. Grav. 23 1089–112
[196] Soleng H H 1990 Class. Quantum Grav. 14 1129–49
[197] Som M M and Bedran M L 1981 Phys. Rev. D 81 2561–3
[198] Stelle K S and West P C 1980 Phys. Rev. D 21 1466–88
[199] Stewart J and Hájíček P 1973 Nature (Phys. Sci.) 244 96
[200] Stoeger W R 1985 Gen. Rel. Grav. 17 981–8
[201] Strobl T 1993 Phys. Rev. D 48 5029–31
[202] Thienel H P 1993 Gen. Rel. Grav. 25 483–90
[203] Tod K P 1994 Class. Quantum Grav. 11 1331–9
[204] Townsend P K 1979 Phys. Lett. B 88 97–101
[205] Trautman A 1973 Nature (Phys. Sci.) 242 7
[206] Trautman A 1999 Class. Quantum Grav. 16 A157–75
[207] Tsamparlis M 1981 Phys. Rev. D 24 1451–7
[208] Tucker R and Wang C 1998 Class. Quantum Grav. 15 933–54
[209] Tung R S and Nester J M 1999 Phys. Rev. D 60 021501-1–5
[210] Utiyama R 1956 Phys. Rev. 101 1597–607
[211] Utiyama R and Fukuyama T 1971 Prog. Theor. Phys. 45 612–27
[212] van Nieuwenhuizen P 1981 Phys. Rep. C 68 189
[213] Venema B J, Majumder P K, Lamoreaux S K, Heckel B R and Fortson E N 1992 Phys. Rev. Lett. 68 135–8
[214] Vilenkin A 1985 Phys. Rev. 121 263
[215] Vlachynsky E J, Tresguerres R, Obukhov Y N and Hehl F W 1996 Class. Quantum Grav. 13 3253–9
[216] Vorobyov P V and Gitarts Y 1988 Phys. Lett. 208 146–8
[217] Weichuan M, Changgui S, Yuanjie L and Xu T 1995 Gen. Rel. Grav. 27 143–52
[218] Werner S A 1994 Class. Quantum Grav. 11 A207–26
[219] Wiesendanger C 1996 Class. Quantum Grav. 13 681–99
[220] Wineland D J, Bollinger J J, Heinzen D J, Itano W M and Raizen M G 1991 Phys. Lett. Rev. 67 1735–8
[221] Wolf C 1995 Gen. Rel. Grav. 27 1031–42
[222] Xia C Q and Wu Y L 1989 Phys. Lett. A 141 251–6
[223] Yang C N and Mills R L 1954 Phys. Rev. 96 191
[224] Yoon Y 1999 Phys. Rev. D 59 127501-1–3
[225] Zhang C M 2000 Preprint arXiv:gr-qc/0002066 pp 1–2

You might also like