See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/1974709

On the Gauge Aspects of Gravity

Article · February 1996

Source: arXiv

CITATIONS READS
123 135

2 authors, including:

Friedrich W. Hehl
University of Cologne
265 PUBLICATIONS   9,468 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Friedrich W. Hehl on 15 May 2019.

The user has requested enhancement of the downloaded file.

 On the Gauge Aspects of Gravity†

Frank Gronwald

and

Friedrich W. Hehl
Institute for Theoretical Physics, University of Cologne
D-50923 Köln, Germany
arXiv:gr-qc/9602013v1 8 Feb 1996

E-mail: [email protected], [email protected]

ABSTRACT

We give a short outline, in Sec. 2, of the historical development of the gauge

idea as applied to internal (U (1), SU (2), . . .) and external (R4 , SO(1, 3), . . .)
symmetries and stress the fundamental importance of the corresponding con-
served currents. In Sec. 3, experimental results with neutron interferometers in
the gravitational field of the earth, as interpreted by means of the equivalence
principle, can be predicted by means of the Dirac equation in an accelerated and
rotating reference frame. Using the Dirac equation in such a non-inertial frame,
we describe how in a gauge-theoretical approach (see Table 1) the Einstein-
Cartan theory, residing in a Riemann-Cartan spacetime encompassing torsion
and curvature, arises as the simplest gravitational theory. This is set in con-
trast to the Einsteinian approach yielding general relativity in a Riemannian
spacetime. In Secs. 4 and 5 we consider the conserved energy-momentum cur-
rent of matter and gauge the associated translation subgroup. The Einsteinian
teleparallelism theory which emerges is shown to be equivalent, for spinless mat-
ter and for electromagnetism, to general relativity. Having successfully gauged
the translations, it is straightforward to gauge the four-dimensional affine group
R4 ⊃ × GL(4, R) or its Poincaré subgroup R4 ⊃ × SO(1, 3). We briefly report
on these results in Sec. 6 (metric-affine geometry) and in Sec. 7 (metric-affine
field equations (111, 112, 113)). Finally, in Sec. 8, we collect some models, cur-
rently under discussion, which bring life into the metric-affine gauge framework
developed.

†
Proc. of the 14th Course of the School of Cosmology and Gravitation on Quantum Gravity, held
at Erice, Italy, May 1995, P.G. Bergmann, V. de Sabbata, and H.-J. Treder, eds. (World Scientific,
Singapore 1996) to be published.
 Contents

1. Introduction
2. Remarks on the history of the gauge idea
2.1. General relativity and Weyl’s U(1)-gauge theory
2.2. Yang-Mills and the structure of a gauge theory
2.3. Gravity and the Utiyama-Sciama-Kibble approach
2.4. E. Cartan’s analysis of general relativity and its consequences
3. Einstein’s and the gauge approach to gravity
3.1. Neutron matter waves in the gravitational field
3.2. Accelerated and rotating reference frame
3.3. Dirac matter waves in a non-inertial frame of reference
3.4. ‘Deriving’ a theory of gravity: Einstein’s method as opposed to the
gauge procedure
4. Conserved momentum current, the heuristics of the translation gauge
4.1. Motivation
4.2. Active and passive translations
4.3. Heuristic scheme of translational gauging
5. Theory of the translation gauge: From Einsteinian teleparallelism to GR
5.1. Translation gauge potential
5.2. Lagrangian
5.3. Transition to GR
6. Gauging of the affine group R4 ⊃
× GL(4, R)
7. Field equations of metric-affine gauge theory (MAG)
8. Model building: Einstein-Cartan theory and beyond
8.1. Einstein-Cartan theory EC
8.2. Poincaré gauge theory PG, the quadratic version
8.3. Coupling to a scalar field
8.4. Metric-affine gauge theory MAG
9. Acknowledgments
10. References
 From a letter of A. Einstein to F. Klein of 1917 March 4 (translation)70 :

“. . . Newton’s theory . . . represents the gravitational field in a seemingly

complete way by means of the potential Φ. This description proves to be
wanting; the functions gµν take its place. But I do not doubt that the day
will come when that description, too, will have to yield to another one,
for reasons which at present we do not yet surmise. I believe that this
process of deepening the theory has no limits . . . ”

1. Introduction

• What can we learn if we look at gravity and, more specifically, at general relativity
theory (GR) from the point of view of classical gauge field theory? This is the question
underlying our present considerations. The answer
• leads to a better understanding of the interrelationship between the metric and affine
properties of spacetime and of the group structure related to gravity. Furthermore, it
• suggests certain classical field-theoretical generalizations of Einstein’s theory, such
as Einstein–Cartan theory, Einsteinian teleparallelism theory, Poincaré gauge theory,
Metric-Affine Gravity, that is, it leads to a deepening of the insight won by GR.
We recently published a fairly technical review article on our results29 . These
lectures can be regarded as a down-to-earth introduction into that subject. We refrain
from citing too many articles since we gave an overviewa of the existing literature in
ref.(29 ).

2. Remarks on the history of the gauge idea

2.1. General relativity and Weyl’s U(1)-gauge theory

Soon after Einstein in 1915/16 had proposed his gravitational theory, namely
general relativity (GR), Weyl extended it in 1918 in order to include – besides grav-
itation – electromagnetism in a unified way. Weyl’s theoretical concept was that of
recalibration or gauge invariance of length. In Weyl’s opinion, the integrability of
length in GR is a remnant of an era dominated by action-at-a-distance theories which
should be abandoned. In other words, if in GR we displace a meter stick from one
point of spacetime to another one, it keeps its length, i.e., it can be used as a standard
a
In the meantime we became aware that in our Physics Reports we should have cited addition-
ally the work of Mistura58 on the physical interpretation of torsion, of Pascual-Sánchez71,72,73 on
teleparallelism, inter alia, of Perlick76 on observers in Weyl spacetimes, and that of Ponomariov and
Obukhov78 on metric-affine spacetimes and gravitational theories with quadratic Lagrangians.
of length throughout spacetime; an analogous argument is valid for a clock. In con-
trast, Weyl’s unified theory of gravitation and electromagnetism of 1918 is set up in
such a way that the unified Lagrangian is invariant under recalibration or re-gauging.
For that purpose, Weyl extended the geometry of spacetime from the (pseudo-)
{}
Riemannian geometry with its Levi-Civita connection Γαβ to a Weyl space with an
additional (Weyl) covector field Q = Qα ϑα , where ϑα denotes the field of coframes of
the underlying four-dimensional differentiable manifold. The Weyl connection one-
form reads
{} 1
ΓWαβ = Γαβ + (gαβ Q − ϑα Qβ + ϑβ Qα ) . (1)
2
The additional freedom of having a new one-form (or covector) field Q at one’s dis-
posal was used by Weyl in order to accommodate Maxwell’s field. He identified the
electromagnetic potential A with Q.
Weyl’s theory turned out to be non-viable, at least in the sense and on the level
of ordinary length and time measurement. However, his concept of gauge invariance
survived in the following way: When quantum mechanics was developed, it became
clear that (in the Schrödinger representation) the wave function Ψ of an electron, for
example, is only determined up to an arbitrary phase φ:

Ψ −→ eiφ Ψ , φ = const . (2)

The set of all phase transformations builds up the one-dimensional Abelian Lie group
U(1) of unitary transformations. If one substitutes, according to (2), the wave func-
tion in the Dirac equation by a phase transformed wave function, no observables will
change; they are invariant under ‘rigid’, i.e., constant phase transformations. This is
an elementary fact of quantum mechanics.
In 1929, Weyl revitalized his gauge idea: Isn’t it against the spirit of field theory
to implement a rigid phase transformation (2) ‘at once’ all over spacetime, he asked.
Shouldn’t we postulate a U(1) invariance under a spacetime dependent change of the
phase instead:
Ψ −→ eiφ(x) Ψ , φ = φ(x) ? (3)
If one does it, the original invariance of the observables is lost under the new ‘soft’
transformations. In order to kill the invariance violating terms, one has to introduce
a compensating potential one-form A with values in the Lie algebra of U(1), which
transforms under the soft transformations in a suitable form. This couples A in a
well determined way to the wave function of the electron, and, if one insists that
the U(1)-potential A has its own physical degrees of freedom, then the field strength
F := dA is non-vanishing and the coupled Dirac Lagrangian has to be amended with
a kinetic term quadratic in F . In this way one can reconstruct the whole classical
Dirac-Maxwell theory from the naked Dirac equation together with the postulate of
soft phase invariance. Because of Weyl’s original terminology, one still talks about
U(1)-gauge invariance – and not about U(1)-phase invariance, what it really is. Thus
 rigid
Conserved Noether’s symmetry
current J theorem of
Lagrangian
dJ=0
Lmat(ψ,dψ)

coupling local
J A gauge
<

symmetry

gauge potential
(connection) A
A
Lmat(ψ,Dψ)
A
DJ=0

Fig. 1. The structure of a gauge theory à la Yang-Mills is depicted in this diagram, which is
adapted from Mills53 . Let us quote some of his statements on gauge theories: ‘The gauge principle,
which might also be described as a principle of local symmetry, is a statement about the invariance
properties of physical laws. It requires that every continuous symmetry be a local symmetry...’ ‘The
idea at the core of gauge theory...is the local symmetry principle: Every continuous symmetry of
nature is a local symmetry.’ The history of gauge theory has been traced back to its beginnings by
O’Raifeartaigh69, who also gave a compact review of its formalism68 .

the electromagnetic potential is an appendage to the Dirac field and not related to
length recalibration as Weyl originally thought.

2.2. Yang-Mills and the structure of a gauge theory

Yang and Mills, in 1954, generalized the Abelian U(1)-gauge invariance to non-
Abelian SU(2)-gauge invariance, taking the (approximately) conserved isotopic spin
current as their starting point, and, in 1956, Utiyama set up a formalism for the
gauging of any semi-simple Lie group, including the Lorentz group SO(1, 3). The
latter group he considered as essential in GR. We will come back to this topic below.
In any case, the gauge principle historically originated from GR as a concept for
removing as many action-at-a-distance concept as possible – as long as the group
under consideration is linked to a conserved current. This existence of a conserved
current of some matter field Ψ is absolutely vital for the setting-up of a gauge theory.
In Fig.1 we sketched the structure underlying a gauge theory: A rigid symmetry of
a Lagrangian induces, via Noether’s theorem, a conserved current J, dJ = 0. It
can happen, however, as it did in the electromagnetic and the SU(2)-case, that a
conserved current is discovered first and then the symmetry deduced by a kind of
a reciprocal Noether theorem (which is not strictly valid). Generalizing from the
gauge approach to the Dirac-Maxwell theory, we continue with the following gauge
procedure:
Extending the rigid symmetry to a soft symmetry amounts to turn the constant
group parameters ε of the symmetry transformation on the fields Ψ to functions of
spacetime, ε → ε(x). This affects the transformation behavior of the matter La-
grangian which usually contains derivatives dΨ of the field Ψ: The soft symmetry
transformations on dΨ generate terms containing derivatives dε(x) of the spacetime-
dependent group parameters which spoil the former rigid invariance. In order to coun-
terbalance these terms, one is forced to introduce a compensating field A = Ai a τa dxi
(a=Lie-algebra index, τa =generators of the symmetry group) – nowadays called gauge
potential – into the theory. The one-form A turns out to have the mathematical mean-
ing of a Lie-algebra valued connection. It acts on the components of the fields Ψ with
respect to some reference frame, indicating that it can be properly represented as the
connection of a frame bundle which is associated to the symmetry group. Thereby it
is possible to replace in the matter Lagrangian the exterior derivative of the matter
field by a gauge-covariant exterior derivative,
A A
d −→ D := d + A , Lmat (Ψ, dΨ) −→ Lmat (Ψ, DΨ) . (4)
This is called minimal coupling of the matter field to the new gauge interaction.
The connection A is made to a true dynamical variable by adding a corresponding
kinematic term V to the minimally coupled matter Lagrangian. This supplementary
term has to be gauge invariant such that the gauge invariance of the action is kept.
A
Gauge invariance of V is obtained by constructing it from the field strength F = DA,
V = V (F ). Hence the gauge Lagrangian V , as in Maxwell’s theory, is assumed to
depend only on F = dA, not, however, on its derivatives dF, d∗d F, . . . Therefore the
field equation will be of second order in the gauge potential A. In order to make
it quasilinear, that is, linear in the second derivatives of A, the gauge Lagrangian
must depend on F no more than quadratically. Accordingly, with the general ansatz
V = F ∧ H, where the field momentum or “excitation” H is implicitly defined by
H = −∂V /∂F , the H has to be linear in F under those circumstances.
By construction, the gauge potential in the Lagrangians couples to the conserved
current one started with – and the original conservation law, in case of a non-Abelian
symmetry, gets modified and is only gauge covariantly conserved,
A
dJ = 0 −→ DJ = 0 , J = ∂Lmat /∂A . (5)
The physical reason for this modification is that the gauge potential itself contributes
a piece to the current, that is, the gauge field (in the non-Abelian case) is charged.
For instance, the Yang-Mills gauge potential B a carries isotopic spin, since the SU(2)-
group is non-Abelian, whereas the electromagnetic potential, being U(1)-valued and
Abelian, is electrically uncharged.

2.3. Gravity and the Utiyama-Sciama-Kibble approach

Let us come back to Utiyama (1956). He gauged the Lorentz group SO(1, 3),
inter alia. Using some ad hoc assumptions, like the postulate of the symmetry of the
connection, he was able to recover GR. This procedure is not completely satisfactory,
as is also obvious from the fact that the conserved current, linked to the Lorentz
group, is the angular momentum current. And this current alone cannot represent
the source of gravity. Accordingly, it was soon pointed out by Sciama and Kibble
(1961) that it is really the Poincaré group R4 ⊃
× SO(1, 3), the semi-direct product of
the translation and the Lorentz group, which underlies gravity. They found a slight
generalization of GR, the so-called Einstein-Cartan theory (EC), which relates – in a
Einsteinian manner – the mass-energy of matter to the curvature and – in a novel way
– the material spin to the torsion of spacetime. In contrast to the Weyl connection
(1), the spacetime in EC is still metric compatible, i.e. governed by a Riemann-Cartanb
(RC) geometry. Torsion is admitted according to
{} 1
ΓRC
αβ = Γαβ − [eα ⌋Tβ − eβ ⌋Tα − (eα ⌋eβ ⌋Tγ )ϑγ ] . (6)
2
Incidentally, approaches to the gauging of the Poincaré group on a fixed Minkowski
background yield effectively similar structures, see Wiesendanger96 .
In order to fix the notation, let us shortly recapitulate the structures emerging
in spacetime geometry. A four-dimensional differential manifold has at each point a
tangent space, spanned by the four basis vectors eα = ei α ∂i , with ∂i as the vectors
tangent to the coordinate lines and α, β = 0, 1, 2, 3, and the cotangent space, spanned
by the four one-forms ϑβ = ej β dxj , with dxj as coordinate one-forms. The two types
of bases are dual to each other eα ⌋ϑβ = δαβ , where ⌋ denotes the interior product, i.e.,
in components, ek α ek β = δαβ . This is the underlying manifold.
On top of it we specify a linear connection. For Einstein14 “... the essential
achievement of general relativity, namely to overcome ‘rigid’ space (ie the inertial
frame), is only indirectly connected with the introduction of a Riemannian metric.
The directly relevant conceptual element is the ‘displacement field’ (Γlik ), which ex-
presses the infinitesimal displacement of vectors. It is this which replaces the par-
allelism of spatially arbitrarily separated vectors fixed by the inertial frame (ie the
equality of corresponding components) by an infinitesimal operation. This makes it
possible to construct tensors by differentiation and hence to dispense with the in-
troduction of ‘rigid’ space (the inertial frame). In the face of this, it seems to be of
b
The terminology is not quite uniform. Borzeskowski and Treder9 , in their critical evaluation of
different gravitational variational principles, call such a geometry a Weyl-Cartan gemetry.
secondary importance in some sense that some particular Γ field can be deduced from
a Riemannian metric...” In this vein, we introduce a linear connection
Γα β = Γiα β dxi , (7)
with values in the Lie-algebra of the linear group GL(4, R). These 64 components
Γiα β (x) of the ‘displacement’ field enable us, as pointed out in the quotation by
Einstein, to get rid of the rigid spacetime structure of special relativity (SR).
In order to be able to recover SR in some limit, the primary structure of a con-
nection of spacetime has to be enriched by the secondary structure of a metric
g = gαβ ϑα ⊗ ϑβ , (8)
with its 10 component fields gαβ (x). At least at the present stage of our knowledge,
this additional postulate of the existence of a metric seems to lead to the only prac-
ticable way to set up a theory of gravity. In some future time one may be able to
‘deduce’ the metric from the connection and some extremal property of the action
function – and some people have tried to develop such type of models, but without
success so far.

2.4. E. Cartan’s analysis of general relativity and its consequences

Besides the gauge theoretical line of development which, with respect to gravity,
culminated in the Sciame-Kibble approach, there was a second line dominated by
E. Cartan’s (1923) geometrical analysis of GR. The concept of a linear connection
as an independent and primary structure of spacetime, see (7), developed gradually
around 1920 from the work of Hessenberg, Levi-Civita, Weyl, Schouten, Eddington,
and others. In its full generality it can be found in Cartan’s work. In particular,
he introduced the notion of a so-called torsion – in holonomic coordinates this is the
antisymmetric and therefore tensorial part of the components of the connection – and
discussed Weyl’s unified field theory from a geometrical point of view.
For this purpose, let us tentatively call
 
gαβ , ϑα , Γα β (9)
the potentials in a gauge approach to gravity and
 
Qαβ , T α , Rα β (10)
the corresponding field strengths. Later, in Sec. 6, inter alia, we will see why this
choice of language is appropriate. Here we defined
Γ
nonmetricity Qαβ := − D gαβ , (11)
Γ
torsion T α := D ϑα = dϑα + Γβ α ∧ ϑβ , (12)
Γ
′′ β ′′
curvature Rα β := β γ β
D Γα = dΓα − Γα ∧ Γγ . (13)
Then symbolically we have
  Γ  
Qαβ , T α , Rα β ∼ D gαβ , ϑα , Γα β . (14)

By means of the field strengths it is straightforward of how to classify the space-

time manifolds of the different theories discussed so far:

GR (1915) : Qαβ = 0 , Tα = 0, Rα β 6= 0 . (15)

γ α β
Weyl (1918) : Qγ =6 0, T = 0, Rα =6 0. (16)
EC (1923/61) : Qαβ = 0 , T α 6= 0 , β
Rα = 6 0. (17)

Note that Weyl’s theory of 1918 requires only a nonvanishing trace of the nonmetric-
ity, the Weyl covector Q := Qγ γ /4. For later use we amend this table with the
Einsteinian teleparallelism (GR||), which was discussed between Einstein and Car-
tan in considerable detail (see Debever12 ) and with metric-affine gravity29 (MAG),
which presupposes the existence of a connection and a (symmetric) metric that are
completely independent from each other (as long as the field equations are not solved):

GR|| (1928) : Qαβ = 0 , Tα =

6 0, Rα β = 0 . (18)
α
MAG (1976) : Qαβ 6= 0 , T = 6 0, Rα β 6= 0 . (19)

Both theories, GR|| and MAG, were originally devised as unified field theories
with no sources on the right hand sides of their field equations. Today, however, we
understand them10,29 as gauge type theories with well-defined sources.
Cartan gave a beautiful geometrical interpretation of the notions of torsion and
curvature. Consider a vector at some point of a manifold, that is equipped with
a connection, and displace it around an infinitesimal (closed) loop by means of the
connection such that the (flat) tangent space, where the vector ‘lives’ in, rolls without
gliding around the loop. At the end of the journey29 the loop, mapped into the tangent
space, has a small closure failure, i.e. a translational misfit. Moreover, in the case
of vanishing nonmetricity Qαβ = 0, the vector underwent a small rotation or – if no
metric exists – a small linear transformation. The torsion of the underlying manifold
is a measure for the emerging translation and the curvature for the rotation (or linear
transformation):

translation −→ torsion T α (20)

β
rotation (lin. transf.) −→ curvature Rα . (21)

Hence, if your friend tells you that he discovered that torsion is closely related to
electromagnetism or to some other nongravitational field – and there are many such
‘friends’ around, as we can tell you as referees – then you say: ‘No, torsion is related
to translations, as had been already found by Cartan in 1923.’ And translations – we
Fig. 2. The neutron interferometer of the COW-experiment11,18 : A neutron beam is split into two
beams which travel in different gravitational potentials. Eventually the two beams are reunited and
their relative phase shift is measured.

hope that we don’t tell you a secret – are, via Noether’s theorem, related to energy-
momentumc, i.e. to the source of gravity, and to nothing else. We will come back to
this discussion in Sec.4.
For the rest of these lectures, unless stated otherwise, we will choose the frame
eα , and hence also the coframe ϑβ , to be orthonormal, that is,
∗
g(eα, eβ ) = oαβ := diag(− + ++) . (22)

Then, in a Riemann-Cartan space, we have the convenient antisymmetries

∗ ∗
ΓRC RC
αβ = −Γβα and RC
Rαβ RC
= −Rβα . (23)

3. Einstein’s and the gauge approach to gravity

3.1. Neutron matter waves in the gravitational field

c
Not long ago, one of the editors of Physical Review D, Lowell S. Brown, decreed that all papers
with (Cartan’s) torsion in the title or the abstract are automatically to be rejected. You may guess
how the authors of such papers reacted. I was always wondering what this clever physicist would
do, if ‘energy-momentum’ appeared in a title. My guess being that he would not even recognize the
need for becoming active again.
 Twenty years ago a new epoch began in gravity: Colella-Overhauser-Werner
measured by interferometric methods a phase shift of the wave function of a neutron
caused by the gravitational field of the earth, see Fig.2. The effect could be predicted
by studying the Schrödinger equation of the neutron wave function in an external
Newtonian potential – and this had been verified by experiment. In this sense noth-
ing really earth-shaking happened. However, for the first time a gravitational effect
had been measured the numerical value of which depends on the Planck constant h̄.
Quantum mechanics was indispensable in deriving this phase shift
m2 g
θgrav = λ A sin α (24)
2πh̄2
(m = mass of the neutron, λ its de Broglie wave length, g = gravitational acceleration,
A = area surrounded by the neutron beams, α = angle between the normal vector of
the area A and the vector g).
It was the availability of nearly perfect single silicon crystals of about 10 cm length
that provided a new tool for X-ray and neutron interferometry. This had first been
demonstrated by Bonse and Hart in 1965 for X-rays. After Bonse (1974) and Rauch,
Treimer, and Bonse (1974), had shown that this device also works for neutrons,
Colella, Overhauser, and Werner (in the following abbreviated by COW) “...used a
neutron interferometer to observe the quantum-mechanical phase shift of neutrons
caused by their interaction with the Earth’s gravitational field”11 . Their experiment
is sketched in Fig.3.
They used neutrons cooled to room temperature such that their resulting mean
velocity vn ≃ 10−5 c is non-relativistic. Their mass is mn = 1.67 × 10−21 kg, and the
de Broglie wave length λn := 2πh̄/p ≈ 0.2 nm. A beam of 1 cm width enters the first
‘ear’ of the interferometer at a Bragg angle in the range of 20◦ to 30◦ . It is coherently
scattered by planes of atoms perpendicular to the surface of the crystal. This Laue
scattering gives rise to a transmitted and a diffracted beam, with opposite Bragg
angles. Due to the Borrman effect, the beam travels through the crystal at first along
the planes and the splitting occurs actually only after it emerges from the ear again.
When the interferometer gets rotated in the gravitational field of the earth, the
upper and lower beams travel at a vertical distance of about 2 cm and encounter a
potential difference of ∆ϕ/Ekin = (mn g△h/(1/2)mn v 2 ) ≈ 103 /(3 × 105 )2 ≈ 10−8 ,
which is only a tiny fraction of the kinetic energy. Nevertheless, this leads to a
measurable effect on the phase of the neutron’s coherent wave which oscillates about
10 cm/λn ≈ 109 times during the horizontal flight. Although the oscillation rate of
the upper beam is ‘redshifted’ merely by a factor of 10−7 , the upper beam manages
to make θgrav ≈ 109 /107 = 100 oscillations more than the lower beam. This phase
shift can be observed by the interference pattern of the recombined beams.
In the actual experiment, side effects have to be taken care of: Gravity produces
distortions in the single crystal. Contributions from this can be eliminated by com-
paring X-ray and neutron interference patterns in the same interferometer. Moreover,
 z
g path 2

~ 2 cm A
path 1
x
~ 6 cm

Fig. 3. COW experiment schematically.

the neutron beam itself is bent into a parabolic path with 4 × 10−7cm loss in altitude.
This yields, however, no significant influence on the phase.
In the COW experiment, the single-crystal interferometer is at rest with respect
to the laboratory, whereas the neutrons are subject to the gravitational potential. In
order to compare this with the effect of acceleration relative to the laboratory frame,
Bonse and Wroblewski8 let the interferometer oscillate horizontally by driving it via
a pair of standard loudspeaker magnets. Thus these experiments of BW and COW
test the effect of local acceleration and local gravity on matter waves and prove its
equivalence up to an accuracy of about 4%.

3.2. Accelerated and rotating reference frame

In order to be able to describe the interferometer in an accelerated frame, we

first have to construct a non-inertial frame of reference. If we consider only mass
points, then a non-inertial frame in the Minkowski space of SR is represented by a
curvilinear coordinate system, as recognized by Einstein13 . Einstein even uses the
names ‘curvilinear co-ordinate system’ and ‘non-inertial system’ interchangeably.
According to the standard gauge model of electro-weak and strong interactions,
a neutron is not a fundamental particle, but consists of one up and two down quarks
which are kept together via the virtual exchange of gluons, the vector bosons of
quantum chromodynamics, in a permanent ‘confinement phase’. For studying the
properties of the neutron in a non-inertial frame and in low-energy gravity, we may
disregard its extension of about 0.7 f m, its form factors, etc. In fact, for our purpose,
it is sufficient to treat it as a Dirac particle which carries spin 1/2 but is structureless
otherwise.
Table 1. Einstein’s approach to GR as compared to the gauge approach: Used are a mass point
m or a Dirac matter field Ψ (referred to a local frame), respectively. IF means inertial frame, NIF
non-inertial frame. The table refers to special relativity up to the second boldface horizontal line.
Below, gravity will be switched on. Note that for the Dirac spinor already the force-free motion in
an inertial frame does depend on the mass parameter m.

Einstein’s approach gauge approach ( → COW)

elementary object in SR mass point m Dirac spinor Ψ(x)

Cartesian coord. system xi holonomic orthon. frame

inertial frame ∗
ds2 = oij dxi dxj eα = δαi ∂i , eα · eβ = oαβ

force-free ∗ ∗
u̇i = 0 (iγ i ∂i − m)Ψ = 0
motion in IF

anholonomic orthon. frame

arbitrary curvilinear
non-inertial frame ′ eα = ei α ∂i
coord. system xi
coframe ϑα = ei α dxi

iγ α ei α (∂i + Γi ) − m Ψ = 0
 
force-free ′ ′ ′ i′
u̇i + uj uk

=0
motion in NIF j ′ k′ Γi := 12 Γi βγ ρβγ Lorentz

i′
ϑα , Γαβ = −Γβα

non-inertial objects j ′ k′
40 16 + 24

R̃(∂{}, {}) = 0 T (∂e, e, Γ) = 0, R(∂Γ, Γ) = 0

constraints in SR
20 24 + 36

∗ i ∗
∗
global IF gij = oij , jk =0 ei α , Γi αβ = (δiα , 0)

R̃ 6= 0 T 6= 0, R 6= 0
switch on gravity
Riemann Riemann − Cartan

local IF ∗ i ∗ ∗
gij |P = oij , jk |P = 0 (ei α , Γαβ α
i )|P = (δi , 0)
(‘Einstein elevator’)

R̃ic− 21 tr(R̃ic) ∼ mass Ric − 21 tr(Ric) ∼ mass

field equations T or + 2 tr(T or) ∼ spin
GR EC
 A Dirac particle has to be described by means of a four-component Dirac spinor.
And this spinor is a half-integer representation of the (covering group SL(2, C) of
the) Lorentz group SO(1, 3). Therefore at any one point of spacetime we need an
orthonormal reference frame in order to be able to describe the spinor. Thus, as soon
as matter fields are to be represented in spacetime, the notion of a reference system
has to be generalized from Einstein’s curvilinear coordinate frame ∂i to an arbitrary,
in general anholonomic, orthonormal frame eα , with eα · eβ = oαβ .
It is possible, of course, to introduce in the Riemannian spacetime of GR arbi-
trary orthonormal frames, too. However, in the heuristic process of setting up the
fundamental structure of GR, Einstein and his followers (for a recent example, see the
excellent text of d’Inverno36 , Secs.9 and 10) restricted themselves to the discussion
of mass points and holonomic (natural) frames. Matter waves and arbitrary frames
are taboo in this discussion. In Table 1, in the middle column, we displayed the Ein-
steinian method. Conventionally, after the Riemannian spacetime has been found and
the dust settled, then electrons and neutron and what not, and their corresponding
wave equations, are allowed to enter the scene. But before, they are ignored. This
goes so far that the well-documented experiments of COW (1975) and BL (1983) –
in contrast to the folkloric Galileo experiments from the leaning tower – seemingly
are not even mentioned in d’Inverno36 (1992).
Prugovečki79 , one of the lecturers here in Erice at our school, in his discussion of
the classical equivalence principle, recognizes the decisive importance of orthonormal
frames (see his page 52). However, in the end, within his ‘quantum general relativity’
framework, the good old Levi-Civita connection is singled out again (see his page
125). This is perhaps not surprising, since he considers only zero spin states in this
context.
We hope that you are convinced by now that we should introduce arbitrary or-
thonormal frames in SR in order to represent non-inertial reference systems for mat-
ter waves – and that this is important for the setting up of a gravitational gauge
theory2,42 . The introduction of accelerated observers and thus of non-inertial frames
is somewhat standard, even if during the Erice school one of the lecturers argued that
those frames are inadmissible. Take the text of Misner, Thorne, and Wheeler57 . In
their Sec.6, you will find an appropriate discussion. Together with Ni30 and in our
Honnef lectures27 we tailored it for our needs.
Suppose in SR a non-inertial observer locally measures, by means of the instru-
ments available to him, a three-acceleration a and a three-angular velocity ω. If the
laboratory coordinates of the observer are denoted by xi , with x as the correspond-
ing three-radius vector, then the non-inertial frame can be written in the succinct
form30,27
" B #
1 ω

e0̂ = 2
∂0 − ×x ∂B ,
1 + a · x/c c
eA = ∂A . (25)
Here ‘naked’ capital Latin letters, A, . . . = 1̂, 2̂, 3̂, denote spatial anholonomic com-
ponents. For completeness we also display the coframe, that is, the one-form basis,
which one finds by inverting the frame (25):
a·x
 
0̂
ϑ = 1+ 2 dx0 = N dx0 ,
c
A
ω

ϑA = dxA + ×x dx0 = dxA + N A dx0 . (26)
c
In the (3 + 1)-decomposition of spacetime, N and N A are known as lapse function
and shift vector, respectively.
Starting with the coframe, we can read off the connection coefficients (since torsion
vanishes in SR) by using Cartan’s first structure equation dϑα = −Γβ α ∧ ϑβ with
Γβ α = Γi β α dxi . We find:
aA
Γ0 0̂A = −Γ0A0̂ = ,
c2
ωC
Γ0AB = −Γ0BA = ǫABC . (27)
c
It is important to note that the first index is holonomic, whereas the second and the
third indices are anholonomic. If we transform the first index, by means of the frame
coefficients ei α , with eα = ei α ∂ i , into an anholonomic one, then we find the totally
anholonomic connection coefficients as follows:
aA  
Γ0̂0̂A = −Γ0̂A0̂ = 2 / 1 + a · x/c2 ,
c
ωC  
Γ0̂AB = −Γ0̂BA = ǫABC / 1 + a · x/c2 . (28)
c
These connection coefficients (28) will enter the Dirac equation referred to a non-
inertial frame.
In order to assure ourselves that we didn’t make mistakes in computing the ‘non-
inertial’ connection (27,28) by hand, we used for checking its correctness the EXCALC
package on exterior differential forms of the computer algebra system REDUCE, see
Puntigam et al.80 and the literature given there.

3.3. Dirac matter waves in a non-inertial frame of reference

The phase shift (24) can be derived from the Schrödinger equation with a Hamilton
operator for a point particle in an external Newton potential. For setting up a grav-
itational theory, however, one better starts more generally in the special relativistic
domain. Thus we have to begin with the Dirac equation in an external gravitational
field or, if we expect the equivalence principle to be valid, with the Dirac equation in
an accelerated and rotating, that is, in a non-inertial frame of reference.
 Take the Minkowski spacetime of SR. Specify Cartesian coordinates. Then the
field equation for a massive fermion of spin 1/2 is represented by the Dirac equation
∗
ih̄γ i ∂i ψ = mcψ , (29)

where the Dirac matrices γ i fulfill the relation

γ i γ j + γ j γ i = 2 oij . (30)

For the conventions and the representation of the γ’s, we essentially follow Bjorken-
Drell7 .
Now we straightforwardly transform this equation from an inertial to an accel-
erated and rotating frame. By analogy with the equation of motion in an arbitrary
frame as well as from gauge theory, we can infer the result of this transformation: In
the non-inertial frame, the partial derivative in the Dirac equation is simply replaced
by the covariant derivative
i
∂i ⇒ Dα := ∂α + σ βγ Γαβγ , ∂α := ei α ∂i ≡ eα , (31)
4
where Γαβγ are the anholonomic components of the connection, see (28), and xi the
Cartesian coordinates of the lab system (which we called xi previously; we drop the
bar for convenience). The anholonomic Dirac matrices are defined by

γ α := ei α γ i ⇒ γ α γ β + γ β γ α = 2 oαβ . (32)

The six matrices σ βγ are the infinitesimal generators of the Lorentz group and fulfill
the commutation relation

[γ α , σ βγ ] = 2i(oαβ γ γ − oαγ γ β ) . (33)

For Dirac spinors, the Lorentz generators can be represented by

σ βγ := (i/2)(γ β γ γ − γ γ γ β ) , (34)

furthermore,
α := γ 0̂ γ with γ = {γ Ξ } . (35)
Then, the Dirac equation, formulated in the orthonormal frame of the accelerated
and rotating observer, reads
ih̄γ α Dα ψ = mcψ . (36)
Although there appears now a ‘minimal coupling’ to the connection, which is caused
by the change of frame, there is no new physical concept involved in this equation.
Only for the measuring devices in the non-inertial frame we have to assume hypotheses
similar to the clock hypothesis. This proviso can always be met by a suitable con-
struction and selection of the devices. Since we are still in SR, torsion and curvature
of spacetime both remain zero. Thus (36) is just a reformulation of the ‘Cartesian’
Dirac equation (29).
The rewriting in terms of the covariant derivative provides us with a rather ele-
gant way of explicitly calculating the Dirac equation in the non-inertial frame of an
accelerated, rotating observer: Using the anholonomic connection components of (28)
as well as α = −i{σ 0̂Ξ }, we find for the covariant derivative:
1 1 i
 
D0̂ = 2
∂0 + 2 a · α − ω·J ,
1 + a · x/c 2c ch̄
DΞ = ∂Ξ . (37)

The total three-angular momentum operator

h̄ ∂ 1 1
J := L + S = x × + h̄ σ = x × p + h̄ σ (38)
i ∂x 2 2
is composed, in the canonical manner, from the orbital piece L and the spin piece S.
The physical effects in our lab frame can be most easily understood by going
over to the Hamiltonian. After multiplying the Dirac equation by β := γ 0̂ and
c(1 + a · x/c2 ), we get
∂ψ
ih̄ = Hψ with H = βmc2 + O + E . (39)
∂t
After substituting the covariant derivatives, the operators O and E, which are odd
and even with respect to β, read, respectively30 :
1
O := cα · p + {(a · x)(p · α) + (p · α)(a · x)} , (40)
2c
E := βm(a · x) − ω · (L + S) . (41)
Up to now, these are mathematically exact results. For later purposes we introduce
O1 = cα · p and O2 = O − O1 .
Similarly as in quantum electrodynamics7 , a non-relativistic approximation can
be obtained by applying successive Foldy-Wouthuysen transformations. After three
such steps we find30 , up to the order of c−2 ,
β 2 β
H ′ = βmc2 + p − 3 2
p4 + βm(a · x) − ω · (L + S)
2m 8m c
β a·x βh̄ 1
+ p· 2 p+ 2
σ · a × p + O( 3 ) . (42)
2m c 4mc c
The different non-inertial effects of a fermion are displayed in Table 2. Besides the
rest mass and the usual kinetic term, we obtain terms which account for the redshift
effect due to acceleration, thereby verifying the BW and, if the acceleration a is
substituted by the gravitational acceleration g, the COW experiments. Moreover the
 Table 2. Inertial effects for a massive fermion of spin 1/2 in non-relativistic approximation.

β m (a · x) Redshift (Bonse-Wroblewski → COW)

−ω · L Sagnac type effect (Heer-Werner et al.)

−ω · S Spin-rotation effect (Mashhoon)

β p · (a · x) p / (2mc2 ) Redshift effect of kinetic energy

β h̄ σ · (a × p) / (4mc2 ) New inertial spin-orbit coupling

‘Sagnac type’ effect occurs in the same manner as in the non-relativistic Schrödinger
equation, and a spin-rotation effect is found which, for the neutron interferometer,
has first been proposed by Mashhoon46 . This term could not have been obtained by
using the Schrödinger equation.
Thus we demonstrated that the Dirac equation in a Minkwoski spacetimed, referred
to a non-inertial frame, yields the BW phase shift and, if the equivalence principle is
assumed, also that of the COW experiment. The claim of our colleague in Erice, that
in SR such non-inertial frames are illegitimate, is thus disproved. Rather we have
shown the usefulness of non-inertial frames ϑα , see (26), with dϑα 6= 0.

3.4. ‘Deriving’ a theory of gravity: Einstein’s method as opposed to the gauge proce-
dure

Now we turn back to Table 1: Einstein’s approach, in the middle column, is

compared with the gauge approach represented in the right column. The basic idea is
that a mechanical system is considered in an inertial and then in a non-inertial frame.
In Einstein’s approach the object under investigation is a point particle and the non-
inertial frame a natural (or coordinate) coframe dxi , in the gauge approach the object
is a Dirac wave function and the non-inertial frame an arbitrary (or anholonomic)
coframe ϑα , with dϑα 6= 0. Einstein’s ‘trick’ was to search for the fields emerging
in the non-inertial frame and to identify them as pseudo-gravitational. ‘Pseudo’,
since the spacetime is a flat and uncontorted Minkowski space without gravitational
fields present. This implies that the pseudo-gravitational fields have to obey some
constraints, otherwise it wouldn’t be guaranteed that the Minkowski background
prevails.
d
These considerations can be generalized to a Riemannian spacetime, see Huang34 and the literature
quoted there.
 The gauge approach had been developed by Utiyama et al. long before BW and
COW set up their experiments. These experiments confirmed the appropriateness of
the earlier theoretical development. Clearly, in some WKB-limit, compare the Foldy-
Wouthuysen technique in Sec. 3.3, the Einsteinian approach can be recovered as a
limiting case of the more general gauge approach.
In the Einstein case the pseudo-gravitational field is represented by the Christoffel
symbol (of the second kind), in the gauge case by the coframe and the Lorentz-
connection. The transition from SR to a spacetime with ‘real’ gravity consists in the
relaxation of the constraints formulated earlier. We are finding a Riemannian and a
Riemann-Cartan spacee, respectively.
But beware! Locally we have to recover the Minkowskian behavior of spacetime:
In the Einsteinian free-fall elevator gravity does not show up. In the Riemann space
this is technically achieved by means of Riemannian normal coordinates, in the case of
the Riemann-Cartan space there exist normal frames which realize an ‘anholonomic’
elevator, as has been first pointed out by von der Heyde31 and developed by one of
us25 , by Modanese & Toller59 , Iliev35 , and Hartley23 . In Hartley’s formulation23 , we
have the following proposition:

Let M be a manifold with metric g and (arbitrary) metric–compatible connection

∇ (Riemann–Cartan space). For any single point P ∈ M, there exist coordinates
{xi } and an orthonormal frame {eα } in a neighborhood of P such that
( )
eα = δαi ∂xi
at P , (43)
Γα β = 0

where Γα β are the connection 1–forms referred to the frame {eα }.

Not too many people know of this theorem, even though it had been proposed
by von der Heyde already twenty years ago. Usually, if we talk to relativists about
the normal frames in a Riemann-Cartan space, they state that those cannot exist,
since torsion, as a tensor, cannot be transformed to zero. In this context it is tacitly
assumed that the starting point are Riemannian normal coordinates and the torsion
is ‘superimposed’. However, since only a natural frame is attached to Riemannian
normal coordinates, one is too restrictive in the discussion right from the beginning.
And, of course, the curvature is also of tensorial nature – and still Riemannian normal
coordinates do exist.
In Table 1 there are two subtleties involved which we want to mention:
In the gauge column we formulated two constraints in SR, namely the vanishing
of torsion and that of curvature. One could well wonder whether we have to relax
e
Recently Hammond22 gave a very pronounced and interesting plea in favor of the existence of torsion
in nature. Only his attempts to derive torsion from some new potential, we don’t find convincing,
since it looks ad hoc to us. Moreover, the coframe is some sort of potential already, see (12) and
Table 5 – and why should we multiply the number of potentials beyond necessity?
both constraints at the same time. We could allow for torsion only, thereby ending
up with a spacetime carrying a teleparallelism, or, alternatively, we could only admit
the emergence of curvature, thereby recovering the Riemannian spacetime of GR. It
is perhaps surprising that also in the teleparallelism case, see Secs.4 and 5 below, by
a suitable choice of a torsion square Lagrangian, one can arrive at a theory which is
equivalent to GR. Nevertheless, there are good reasons for relaxing both constraints:
Firstly, by mimicking the Einsteinian procedure of the middle column of Table 1, we
cannot see any reason why we should lift only one constraint; secondly, by admitting
a Riemann-Cartan space, we can still ‘trivialize’ the gravitational potentials at a
point P and its neighborhood, as found in proposition (43), in spite of the presence
of torsion and curvature. This implies that a Riemann-Cartan space, if described in
terms of suitable coframes, looks locally Minkowskian. Eqs.(43) supply the strongest
reason for taking the four-dimensional Riemann-Cartan space seriously as a model
for spacetime.
A second subtlety is related to the fact that the minimally coupled Dirac equation
(36), in a Riemann-Cartan space, slightly differs from the Euler-Lagrange equation of
the minimally coupled Dirac Lagrangian. However, this shouldn’t cause headaches:
The gauge theoretical set up, see Fig. 1, is so closely linked to the Lagrangian for-
malism – not to speak of the fundamental importance of the Feynman path integral
or of Schwinger’s variational principle – that the minimal coupling procedure should
be implemented on the Lagrangian level.
Eventually concentrating our attention to the last row of Table 1, we read off the
field equations of GR and of the EC-theory: They result from the simplest conceivable
Lagrangian, namely from the curvature scalar of the Riemann or the Riemann-Cartan
spacetime respectively. The EC-theory differs from GR in a very weak spin-contact
interaction which is unmeasurable at the present time. In this sense, the EC-theory
is a viable theory of gravitation.

4. Conserved momentum current, the heuristics of the translation gauge

4.1. Motivation

We have in mind to derive gravity from a symmetry principle. But what is the
right symmetry to derive gravity from? As already indicated in Sec.2, we think that
gravity stems from translation symmetry. Our motivation is the following:
We start from SR. The invariance of the action W = Lmat (Ψ, dΨ) of an isolated
R

material system under rigid spacetime translations yields, by the application of the
Noether theorem, a conserved energy-momentum current three-form Tj via

δLmat 1
Tj := j
= Tklmj dxk ∧ dxl ∧ dxm , d Tj = 0 . (44)
δdx 3!
(One obtains the “usual” energy-momentum tensor Tij from Tj by means of Tij =
ǫi klm Tklmj .) The corresponding charge M := d3 xT0 is conserved in time. In other
R

words: Rigid translational invariance is attributed to the classical field-theoretical

equivalent to mass(-energy density), which is the source of Newton-Einstein gravity.
The analogy to electrodynamics guides us of how to actually generate the grav-
itational interaction. In electrodynamics one finds from rigid U(1)-invariance of an
action W = Lmat (Ψ, dΨ) a conserved electromagnetic current with corresponding
R

electric charge Q. As we discussed in length in Sec.2, it is possible to generate the elec-

tromagnetic interaction by gauging the rigid U(1)-symmetry. Following this example,
we expect the gravitational interaction to emerge from gauging the rigid translational
symmetry. To quote Feynman15 : “...gravity is that field which corresponds to a gauge
invariance with respect to displacement transformations.” But before we gauge the
translation group we should think about...

4.2. Active and passive translations

Whenever we describe quantities with respect to some reference system (a cer-

tain basis, a certain coordinate system) we do encounter the possibility of active or
passive transformations. Actively transforming a quantity or passively transforming
its reference system will usually change the labels of the quantity which are related
to the reference system. How can we decide whether an active, a passive, or even a
mixture of both kinds of transformation took place when those labels of a quantity
changed? Is it necessary to distinguish between active and passive transformations?
For the case of active and passive translations, we will answer this question to the
extent that we show the mathematical equivalence of actively or passively translating
a matter field Ψ in Minkowski spacetime. From this we will later proceed in order to
gauge the translation group.
The field Ψ is thought to be some p-form, possibly Lie-algebra valued. We discard
any internal structure of Ψ since we are only interested in its spacetime properties. For
the Minkowski spacetime we assume for convenience a gauge in which the connection
∗
components are identically zero, Γ = 0. This can be achieved by choosing pseudo-
Cartesian coordinates xi and using the corresponding holonomic (co-)frame. The
total variation δt Ψ of Ψ under translations is the sum of both active and passive
variations:
δt Ψ = δa Ψ + δp Ψ . (45)
An active translation of Ψ is generated by the flow of a vector field ξ = ξ i∂i , that is,
Ψ → ξ ∗ Ψ. In this case the active variation is given by the Lie-derivative lξ ,

δa Ψ = lξ Ψ , lξ = d(ξ⌋) + (ξ⌋)d . (46)

Despite possible exterior indices of Ψ, we don’t have to use a ‘covariant Lie derivative’
∗
since Γ = 0. We will be interested only in infinitesimal variations. Then the Lie
derivative describes the difference between the actively translated ξ ∗ Ψ(xi ) and Ψ(xi +
ξ i ).
Passive translations are generated by coordinate transformations. To find the
passive transformation which corresponds to a given active transformation, we have
to know the explicit form of the operator δp and the actual coordinate transformation
which corresponds to a specific δa = lξ . For the actual coordinate transformation, we
look at the total variation of the coordinate functions xi ,
δt xi = δa xi + δp xi = lξ xi + δp xi = ξ i + δp xi . (47)

The coordinate transformation δxi = δp xi = −ξ i makes δt xi to vanish. Thus the

passive translation δp x = ξ is equivalent to δa x = lξ x (what we intuitively expected
without any calculation). Having this candidate for the appropriate coordinate trans-
formation, we look for an operator δp such that δt Ψ = lξ Ψ + (δp Ψ)(−ξ) = 0. We set

(δp Ψ)(−ξ) = Ψ(xi − ξ i ) − Ψ(xi ) , (48)

i.e., we take the value of Ψ at a point in the coordinate system xi − ξ i and subtract
the value of Ψ at the same point while using the coordinate system xi . Writing
(δp Ψ)(−ξ) = Ψi1 ...ip (xi − ξ i)d(xi1 − ξ ii ) ∧ ... ∧ d(xip − ξ ip ) , (49)

Taylor expanding Ψi1 ...ip (xi − ξ i ), and keeping terms up to order ξ, we can show that
(δp Ψ)(−ξ) = −lξ Ψ . (50)
Thereby we just rederived the definition of the Lie derivative from a passive point
of view, and the coordinate transformation δx = ξ becomes clearly equivalent to an
active translation generated by lξ .

4.3. Heuristic scheme of translational gauging

In order to gauge the translation group we will follow the general gauge scheme
which we set up in Sec.2. Therefore we will begin with a rigidly translation invariant
action W = Lmat (Ψ, dΨ). We expect to have to introduce four gauge potentials
R

Aα in order to extend from rigid to soft translation invariance of W . Each Aα will

compensate one of the four independent soft translations lξα .
We note that soft translations do not commute any longer: Expanding the flow-
generating vectors ξα according to ξα = ei α ∂i , with spacetime dependent functions
ei α = ei α (x), we can use the formula

[lξα , lξβ ] = l[ξα ,ξβ ] (51)

∗ ∗
to show that (the equality = requires the gauge Γ = 0)
∗
[lξα , lξβ ] = −ej [α ei β] ei γ , j ξγ = −Tαβ γ ξγ . (52)
 uR Dvu
[u,v]

R Duv
||
u R T(u,v)

vP vQ|| vQ

Q
P
uP
Fig. 4. On the geometrical interpretation of torsion: It represents a closure failure of infinitesimal
displacements.

The Tαβ γ are the components of the torsion two-form T γ = 21 Tαβ γ ϑα ∧ ϑβ expanded
in the anholonomic coframe ϑα = ei α dxi . This softening of the Lie algebra of the
translation group should be seen in correspondence to the structure

commutator of soft symmetry transformations = field strength , (53)

which is well known from Yang-Mills theory. Therefore it is tempting to view the
torsion tensor as the translational field strength.
The torsion tensor represents a translational misfit, as already indicated in Sec.2.4.
A more detailed discussion is given by de Sabbata and Sivaram82 who also collected
numerous other facts and results on torsion. Torsion measures the noncommutativity
of displacements of points in analogy to the curvature tensor which measures the
noncommutativity of displacements of vectors. This is explained in Fig. 4: Let P be
a point of the (spacetime) manifold M and vP , uP two linearly independent vectors
of TP M. We regard vP , uP as infinitesimally small. Then they define two points,
R and Q, on M. (We can make this mathematically precise by defining R, Q via
the exponential map expP : TP M → M, expP (v) = R, expP (u) = Q.) In other
words: The vector vP displaces P infinitesimally to R, the vector uP displaces P
infinitesimally to Q. The prescription to perform two successive displacements v, u
||
is to first displace P by means of vP to R and second parallel transport uP to uR
 ||
and displace R by means of uR . Fig. 4 shows that the commutator of two successive
displacements won’t vanish in general – the gap between the two resulting points is,
by definition, a measure for the torsion. It is also shown in Fig.4 that this definition
matches the usual textbook definition

T (u, v) := Du v − Dv u − [u, v] . (54)

To make contact with our present notation, we evaluate the components of T with
respect to an arbitrary frame eα :

T (eβ , eγ ) = Deβ eγ − Deγ eβ − [eβ , eγ ]

= Γβγ α eα − Γγβ α eα + Cβγ α eα (55)
= Tβγ α eα .

Hence
Tβγ α = Cβγ α + Γβγ α − Γγβ α , (56)
or, by using the object of anholonomity
1
C α := dϑα = Cβγ α ϑβ ∧ ϑγ , (57)
2
we find:
T α = dϑα + Γβ α ∧ ϑβ . (58)
The three irreducible pieces of the torsion under SO(1, 3)-decomposition are displayed
for later use in Tab.3.
(1)
Table 3. Irreducible decomposition of the torsion T α = T α + (2) T α + (3) T α under the Lorentz
group SO(1, 3)

explicit expression number of components name

(2) T α 1 α
3ϑ ∧ (eβ ⌋T β ) 4 TRATOR

(3) T α − 13 ∗ (ϑα ∧ ∗ (T β ∧ ϑβ )) 4 AXITOR

(1) T α T α − (2) T α − (3) T α 16 TENTOR

From the translational field strength T α , see (52) and (53), we come back to the
translational gauge potential Aα . According to the general gauge scheme, it will
couple to the energy-momentum current according to
δL
Tα = . (59)
δAα
The left hand side of (44) suggests that the gauge process will replace the holonomic
coframe dxi by the potentials Aα . Do the Aα have something to do with an anholo-
nomic coframe ϑα ? A consistent relation between T α , ϑα , and Aα is established if we
assume that
ϑα = δiα dxi + Aα . (60)
Then we obtain
∗
dAα = dϑα = T α . (61)
α ∗
From (61) we get the ‘Bianchi’ identity d T = 0. Comparison to the general Bianchi
identity DT α = Rβ α ∧ ϑβ indicates that the structures found so far are part of a more
general framework. This is, indeed, the case. In order to derive GR from as little
input as possible, we refrain at this point from introducing additional fields, as for
example an independent linear connection Γ. We collected the structures relevant for
the translational gauge scheme in Tab.4.
Table 4. The relevant structures in a gauge approach of the four parameter translation group

number of components

conserved momentum current Tα 4×4

translational gauge potential ϑα ≃ Aα (coframe) 4×4

translational field strength T α (torsion) 4×6

∗ ∗
1st Bianchi identity (Γ = 0) dTα = 0 4×4

5. Theory of the translation gauge: From Einsteinian teleparallelism to

GR

As we argued in the last section, GR should be derivable from gauge ideas in a

fairly straightforward manner. The key to arrive at GR is to start with the conserved
momentum current and to gauge the translation group that is connected with it.

5.1. Translation gauge potential

We commence with the very basics: Consider again a field theory in Minkowski
spacetime (pseudo-Cartesian coordinates) defined by the Lagrangian Lmat = Lmat (Ψ, dΨ).
An explicit dependence of L on the coordinates xi is already forbidden by rigid trans-
lational invariance, which we started from. The coordinates enter more implicitly:
The field Ψ has to be expressed in terms of differential forms in order to build the
Lagrangian 4-form L as the appropriate integrand of the action. Therefore one needs
the differentials dxi as natural (or holonomic) basis for the physical field. They are
invariant under rigid but not under soft translations:
δxi = εi (x) =⇒ δdxi = dεi . (62)
Referring to the equivalence between active and passive transformations, which is
valid at this stage (see Sec.4.2), we view in this approach the translations as passive
transformations. The differentials dxi are no longer sufficient to build up an invariant
Lagrangian. According to the gauge principle, we have to introduce a gauge potential
Aα with transformation behavior δAα = −δiα dεi such that
ϑα := δiα D̂xi := δiα dxi + Aα (63)
transforms like
δϑα = 0 . (64)
The anholonomic one-form basis ϑα serves as an appropriate form basis for La-
grangians Lmat since it automatically incorporates soft translational invariance. One
usually refers to ϑα , instead of Aα , as the translational potential. Normally, a dis-
tinction between ϑα and Aα is not necessary, since their field components differ just
by a constant (1 or 0).

5.2. Lagrangian
∗
The corresponding field strength T α = dϑα can be used to construct a kinematic
supplementary term for ϑα to the Lagrangian. The double role of ϑα as both, a
dynamical gauge potential and an orthonormal frame (defining a new metric via
g = oαβ ϑα ⊗ ϑβ ), explains the transition from Minkowski space to a dynamical
spacetime, which is due to translational invariance. For the kinematic term we make
the quadratic ansatz V = dϑα ∧ Hα . What would be a good choice for Hα ? Eyeing
at Yang-Mills theory, we are tempted to put Hα = 2ℓ12 ∗ dϑα , with ℓ = Planck length.
But we would like to end up with a softly Lorentz invariant theory. The Lagrangian
V = 2ℓ12 ∗ dϑα ∧ dϑα is rigidly but not softly Lorentz invariant, though. Note that
this postulate of soft Lorentz symmetry is not equivalent to a gauging of the Lorentz
group! We won’t cure the lack of Lorentz invariance by the introduction of some new
gauge field Γ for the Lorentz field, but will use it just as a criterion of choosing a
good Lagrangian.
The most general term V quadratic in dϑα is obtained by choosing Hα as
1 ∗  (1) (2) (3)

Hα = a1 dϑα + a2 dϑα + a3 dϑα , (65)
2ℓ2
see also Mielke50,49 . The pieces (I)
dϑα correspond to the irreducible pieces (I)
T α of
the torsion, compare Table 3:

(2) 1 α
dϑα := ϑ ∧ (eβ ⌋dϑβ ) ,
3
(3) 1
dϑα := − ∗ {ϑα ∧ ∗ (dϑβ ∧ ϑβ )} ,
3
(1)
dϑα := dϑα − (2) dϑα − (3) dϑα . (66)

The postulate of soft Lorentz invariance leads to a solution for the constant and real
parameters aI in the following way:
Infinitesimal Lorentz rotations are expressed by δϑα = εα β ϑβ where εαβ = −εβα
are the antisymmetric Lorentz group parameters. It is easy to check that the gauge
Lagrangian V = dϑα ∧ Hα , with Hα given by (65), is invariant under rigid Lorentz
rotations, δV = 0. The general expression for δV reads
! !
∂V ∂V α ∂V
δV = − d ∧ δϑ + d ∧ δϑα . (67)
∂ϑα ∂dϑα ∂dϑα

Hence we have δV = 0 for rigid Lorentz rotations. However, for soft Lorentz rotations
with spacetime-dependent group parameters εαβ = εαβ (x), we get from (67) the
offending term
∂V
δ(soft) V = dεα β ∧ ∧ ϑβ . (68)
∂dϑα
In order to preserve Lorentz invariance, this term has to be canceled, modulo an exact
form. Using the Leibniz rule, we obtain
! !
α ∂V ∂V ∂V
dε β ∧ α
∧ ϑβ = εα β d α
∧ ϑβ − d εα β α ∧ ϑβ . (69)
∂dϑ ∂ϑ ∂ϑ

The second term on the r.h.s. is already exact. From the first term we get the
condition
∂V
[α ∧ ϑβ] = exact form (70)
∂dϑ
for soft Lorentz invariance of V . We plug in the explicit expression for V and obtain,
after some algebra,
∂V 1 1 2 1
   
[α ∧ ϑβ] = a1 − a3 dηαβ − a3 + a1 dϑ[α ∧ ϑβ]
∂dϑ 3 3 3 3
1 1 1
 
+ a1 + a2 − a3 (eγ ⌋dϑγ ) ∧ ηαβ . (71)
6 6 3
The last two terms can be made vanishing by choosing
1
a2 = −2a1 , a3 = − a1 . (72)
2
 PG

U4
Riemann-Cartan
Qα β =0
T α, Rαβ

=0

to
rs
re

io
tu

n
a

=0
rv
cu
V4
W4
GR || .. Riemann
Weitzenbock GR
(teleparallelism) R αβ
Tα

0
to

=
rs

re
io

tu
n=

a
rv
0

cu
M4
Minkowski

SR
Fig. 5. A spacetime with a metric and a metric compatible connection (nonmetricity=0) is called a
Riemann-Cartan space U4 . It can either become a Weitzenböck space W4 , if its curvature vanishes,
or a Riemann space V4 , if the torsion happens to vanish. These different models of spacetime are
the arenas for different gravitational theories.

Then we obtain
∂V a1
[α ∧ ϑβ] = dηαβ . (73)
∂dϑ 2
The constant a1 can be absorbed by a suitable choice of the coupling constant ℓ in V ,
see (65). According to the usual conventions, we put a1 = −1, i.e. V is softly Lorentz
invariant for the choice of parameters
1
a1 = −1 , a2 = 2 , a3 = . (74)
2
Hence
1 1
 
V|| = 2
dϑα ∧ ∗ −(1) dϑα + 2 (2) dϑα + (3) dϑα . (75)
2ℓ 2

The total Lagrangian reads

Ltot = V|| + Lmat (Ψ, dΨ, ϑα ) , (76)

and the field equation δLtot /δϑα = 0 becomes
dHα − Eα = Tα , (77)
where, as before, Tα = δLmat /δϑα denotes the canonical energy-momentum three-form
and
1
Eα := (eα ⌋dϑβ ) ∧ Hβ − eα ⌋(dϑβ ∧ Hβ )
2
1 h i
= (eα ⌋dϑβ ) ∧ Hβ − dϑβ ∧ (eα ⌋Hβ ) , (78)
2
the energy-momentum current of the gauge field.

5.3. Transition to GR

If the Lagrangian (75) is substituted into the field equation (77), then it can be
seen that the antisymmetric piece of the left hand side of (77) vanishes,
ϑ[β ∧ dHα] − ϑ[β ∧ Eα] = 0 . (79)
Therefore the right hand side has to be symmetric, too, i.e. only scalar matter fields or
gauge fields, such as the electromagnetic field, are allowed as material sources, whereas
matter carrying spin cannot be consistently coupled in such a framework. The existing
nontrivial torsion, expressed by dϑα , describes the nontrivial Riemannian geometry
{}
of spacetime. This is because we have tied the Christoffel connection Γ αβ , which is
determined by the metric, to the contortion tensor Kαβ by means of the teleparallel
∗
condition Γαβ = 0:
{} {}
∗ ∗
Γ αβ − Kαβ = Γαβ = 0 =⇒ Γ αβ = Kαβ . (80)
In other words: Meaningful teleparallel theories do not presuppose spinning matter as
a source for nontrivial torsion, in contrast to what is sometimes stated in literature39 .
Doesn’t all this look like general relativity? The Levi-Civita (or Christoffel) con-
nection, corresponding to the metric g = oαβ ϑα ⊗ ϑβ , is given by
{} 1
(eα ⌋dϑβ − eβ ⌋dϑα − (eα ⌋eβ ⌋dϑγ ) ∧ ϑγ ) .
Γ αβ = (81)
2
The corresponding Riemannian curvature reads
{} {} {} {}
Rαβ = d Γ αβ − Γ αγ ∧ Γ γ β . (82)
We use the last two equations to replace on the Lagrangian level the variable dϑα by
{}
Γ αβ . Using (81) and (82), one can prove the remarkable identity
1 {}αβ
R ∧ ηαβ − ℓ2 V = d(ϑα ∧ ∗ dϑα ) , (83)
2
with V given by (75). Therefore one finds that the kinematic term V , with the above
choice of parameters aI , is equivalent to the Hilbert-Einstein action modulo an exact
term. Replacing V in the action S by means of (83) leads, via δLtot /δϑα = 0, to
Einstein’s equation
{} 1 {}
Gα := ηαβγ ∧ Rβγ = ℓ2 Tα . (84)
2
In such a way, we arrive at GR in its original form. Shifting back and forth from the
{}
variable pair (ϑα , Γ α β ) to (ϑα , dϑα ) means shifting back and forth from original GR
to its teleparallel equivalent GR||. This is displayed in Fig.5: A general PG induces
a Riemann-Cartan space U4 with nonvanishing torsion and curvature. Such a U4 can
be reduced to either a Weitzenböck space W4 (curvature = 0) or a Riemann space V4
(torsion = 0), i.e. to the geometries induced by GR|| or GR, respectively.
We wish to point out that both theories can be obtained as special cases within the
framework of the Poincaré gauge theory PG, see our Physics Reports29 , the literature
given there, and the work of Pascual-Sánchez72,73 .

6. Gauging of the affine group R4 ⊃

× GL(4, R)

We recognized that the gauging of the translations yields a theory which, for
spinless matter and for electromagnetism, that is, for symmetric energy-momentum
currents, is equivalent to GR. Thus we have a new understanding of Einstein’s theory.
Why should we generalize GR|| if it is consistent with experiment? Three somewhat
interrelated arguments come to mind:
• The translations represent a subgroup of the Poincaré group. Only to gauge the
translations and to leave the Lorentz subgroup of the Poincaré group untouched,
would seem unnatural. This argument is all the more convincing, since the semi-
direct product structure of the Poincaré group interrelates its two mentioned
subgroups stronger than it were the case for a direct product.

• A Weitzenböck spacetime is a degenerate Riemann-Cartan space. The gauge

arguments against the rigidity of a teleparallelism were already advanced by
Weyl in the twenties against Einstein’s corresponding theory.

• The translational gauge procedure, as it is obvious from the field equation (77)
with (75), works only for spinless matter and for electromagnetism, since the
field equation is symmetric, see (79), and supplies only 10 independent compo-
nents.
We conclude that a gauging of the whole 4 + 6 parameter Poincaré group is manda-
tory. The theory which emerges, the Poincaré gauge theory (PG), is formulated in
a Riemann-Cartan spacetime, the Einstein-Cartan theory (EC) being a degenerate
subcase of it. And the EC is a viable gravitational theory!
 In order to see the built-up of the different structures of spacetime more clearly,
we prefer to gauge48 immediately the 4 + 16 parameter affine group48 A(4, R) =
R4 ⊃
× GL(4, R) — which lacks a metric structure altogether — and to introduce the
metric subsequently. Thus we arrive at the metric-affine gauge theory (MAG) which
encompasses the PG as a subcase. Symbolically, we may write

MAG/nonmetricity = 0 = PG . (85)

The actual gauging of the affine group was done in our Physics Reports29 , and
we follow the presentation given there. A short outline of these results, see Secs.
3.1, 3.2, 3.3, loc. cit., will be given here. We start then in the flat n–dimensional
affine space Rn . The rigid affine group A(n, R) := Rn ⊃ × GL(n, R) is the semidirect
product of the group of n-dimensional translations and n-dimensional general linear
transformations. This transformation group, cf. ref.(40 ) p. 27, acts on an affine n–
vector x = {xα } according to

x → x′ = Λ x + τ , (86)

where Λ = {Λα β } ∈ GL(n, R) and τ = {τ α } ∈ Rn . Thus it is a generalization

of the Poincaré group P := R4 ⊃ × SO(1, 3), with the pseudo–orthogonal group
SO(1, n − 1) being replaced by the general linear group GL(n, R). It is convenient to
work with a Möbius type representation for which we take the same symbol A(n, R):
=
It is that subgroupof GL(n + 1, R) which leaves the n-dimensional hyperplane Rn :=
x
  
=
x= ∈ Rn+1 invariant:
1

Λ τ
  
n
A(n, R) = ∈ GL(n + 1, R) | Λ ∈ GL(n, R), τ ∈ R . (87)
0 1
Thus, by an affine transformation, we obtain
Λ τ x Λx + τ
    
=′
x = = , (88)
0 1 1 1
as is required for the action of the affine group on the flat affine space. The Lie algebra
a(n, R) consists of the generators Pγ , representing n-dimensional translations, and
the Lα β , which span the Lie algebra gl(n, R) of n-dimensional linear transformations.
In a matrix representation we can write the affine gauge group as
Λ(x) τ (x)
  
A(n, R) = | Λ(x) ∈ GL(n, R), τ (x) ∈ T (n, R) . (89)
0 1
Having already had some experience with the Yang-Mills type gauge approach, we
are aware of the need of introducing a gauge potential in order to step from rigid to
soft group invariance. Accordingly, by gauging the affine group, the softening of the
affine group transformations is accompanied by the introduction of the generalized
affine connection (→ potential)
≈ Γ(L) Γ(T ) Γ(L) β α
L β Γ(T )α Pα
   
α
Γ= = . (90)
0 0 0 0
≈ ≈
It is a one–form Γ = Γi dxi and transforms inhomogeneously under an affine gauge
transformation,
≈ A−1 (x) ≈ ≈
Γ −→ Γ′ = A−1 (x) Γ A(x) + A−1 (x)dA(x) , A(x) ∈ A(n, R), (91)

where the transformation is formed with respect to the group element

Λ−1 (x) −Λ−1 (x)τ (x)
 
−1
A (x) = . (92)
0 1
The corresponding affine curvature is given by
≈ ≈ ≈ ≈ R(L) R(T ) dΓ(L) + Γ(L) ∧ Γ(L) dΓ(T ) + Γ(L) ∧ Γ(T )
   
R := dΓ + Γ ∧ Γ = = .
0 0 0 0
(93)
It transforms covariantly under the affine gauge group:
≈ A−1 (x) ≈′ ≈
R −→ R = A−1 (x) R A(x) . (94)
≈ ≈ ≈ Ψ
 
The exterior covariant derivative D := d + Γ∧ acts on an affine p–form Ψ = as
1
follows:
≈ ≈ dΨ + Γ(L) ∧ Ψ + Γ(T ) DΨ + Γ(T )
   
DΨ = = . (95)
0 0
After substitution of (90) and (92), the inhomogeneous transformation law (91)
splits into
A−1 (x)
Γ(L) −→ Γ(L)′ = Λ−1 (x) Γ(L) Λ(x) + Λ−1 (x)dΛ(x) , (96)
and
A−1 (x)
Γ(T ) −→ Γ(T )′ = Λ−1 (x) Γ(T ) + Λ−1 (x)Dτ (x) . (97)
The soft translations τ (x) automatically drop out in (96) due to the one–form struc-
ture of Γ(T ) . Thereby (96) acquires the conventional transformation rule (with the
exterior derivative d) of a Yang–Mills–type connection for GL(n, R). Thus we can
identify Γ(L) = Γ = Γα β Lα β with the linear connection and are basically done with
the gauging of the linear part of the affine group.
Things are more involved for the translational part, though. From what we have
learned by gauging the translation group we expect Γ(T ) to be related to the coframe
ϑ := ϑα Pα , i.e. to a one–form with values in the Lie algebra of Rn . But due to the
 Table 5. Gauge fields in metric-affine gauge theory MAG

Potential Field strength Bianchi identity

metric gαβ Qαβ = −Dgαβ DQαβ = 2R(α µ gβ)µ

coframe ϑα T α = Dϑα DT α = Rµ α ∧ ϑµ

connection Γα β Rα β = dΓα β − Γα µ ∧ Γµ β DRα β = 0

covariant exterior derivative term Dτ (x) := dτ (x)+Γ(L) τ (x) in (97), the translational
part Γ(T ) does not transform as a covector, as is required for the coframe ϑ. To get a
correspondence between Γ(T ) and ϑ, we introduce a vector (vector–valued zero–form)
≈ ξ ξ α Pα ≈ ≈
   
ξ= = which transforms as ξ ′ = A−1 (x) ξ, i.e. as
1 1
A−1 (x)
ξ −→ ξ ′ = Λ−1 (x) (ξ − τ (x)) (98)

under an active affine gauge transformation. Then we define

ϑ := Γ(T ) + Dξ (99)

which transforms as a vector–valued one–form under the A(n, R), as required:

A−1 (x)
ϑ −→ ϑ′ = Λ−1 (x) ϑ . (100)

The geometric and physical meaning of the relation (99), especially the role of the
field ξ, is perhaps not completely satisfactorily clarified, yet. For an expanded dis-
cussion of this point we refer the reader again to ref.(29 ), see also Mistura58 . The
translational piece of the affine connection, namely Γ(T ) , could play a decisive role
in the gravitationally induced phase factor of a matter wave, as was pointed out by
Morales-Técotl et al.60 . Here, we confine ourselves to require the condition

Dξ = 0 (101)
≈
to hold. Then the generalized affine connection Γ on the affine bundle A(M) reduces
to the Cartan connection  (L)
= Γ ϑ

Γ= (102)
0 0
on the bundle L(M) of linear frames. Due to (100), this is not anymore a connection in
the usual sense. We are thus left with the potentials (ϑα , Γα β ) and their corresponding
field strengths (T α , Rα β ).
 For this gauging of the affine group no metric was necessary. If additionally
a metric is given, we recover the metric–affine geometrical arena of Sec. 2.4, as is
summarized in Table 5.
We close this geometric section by introducing a further geometric ingredient, the
“η-bases”, which span the graded algebra of dual exterior forms on each cotangent
space and which turn out to be quite useful in practical calculations.
q First, owing to
the existence of a metric, we can define the scalar density | det gµν | and the familiar
Hodge star operator ∗ . The Hodge star operator maps a p–form Ψ into an (n−p)–form
∗
Ψ by means of the explicit formula

∗ 1 q
Ψ := | det gµν | g α1 γ1 · · · g αp γp ǫα1 ···αp β1 ···βn−p Ψγ1 ···γp ϑβ1 ∧ · · · ∧ ϑβn−p .
(n − p)! p!
(103)
Now we define the g-volume element n–form
q 1 q
η := | det gµν | ϑ1̂ ∧ · · · ∧ ϑn̂ = | det gµν | ǫα1 ···αn ϑα1 ∧ · · · ∧ ϑαn = ∗ 1 , (104)
n!
dual to the unit zero–form. Picking a (pseudo-)orthonormal positively oriented coframe
◦
ϑ α , the g-volume element simplifies to
◦ ◦
η = ϑ 1̂ ∧ · · · ∧ ϑ n̂ . (105)

Having this n–form at our disposal, we can successively contract it by means of

the frame eα , thereby arriving at an (n − 1)–form, an (n − 2)–form, etc., until we
terminate the series with a zero–form:
1
ηα1 := eα1 ⌋η = ηα1 α2 ···αn ϑα2 ∧ · · · ∧ ϑαn = ∗ϑα1 ,
(n − 1)!
1
ηα1 α2 := eα2 ⌋ηα1 = ηα α α ···α ϑα3 ∧ · · · ∧ ϑαn = ∗(ϑα1 ∧ ϑα2 ) ,
(n − 2)! 1 2 3 n
..
. (106)
∗
ηα1 ···αn := eαn ⌋ · · ·⌋eα1 ⌋η = (ϑα1 ∧ · · · ∧ ϑαn ) .

Thus, in four dimensions, which we are concentrating on, we have an η-basis at our
disposal (η, ηα , ηαβ , ηαβγ , ηαβγδ ) that we will meet in numerous applications.

7. Field equations of metric–affine gauge theory (MAG)

The gauge procedure of the affine group led to the identification of the gauge
potentials of the MAG, see Table 5. The material currents are then expected to
couple to these potentials in the Yang-Mills type fashion, as is indicated in Fig.
1, and as we know already from the material energy-momentum current and the
coframe according to the line after (77). If Ψ, as a p-form, represents a matter field
(fundamentally a representation of the SL(4, R) or of some of its subgroups61,62 ), its
first order Lagrangian L will be embedded in metric-affine spacetime by the minimal
coupling procedure, that is, exterior covariant derivatives feature in the kinetic terms
of the Lagrangian instead of only exterior ones. Then the material currents are defined
as follows:
δL δL δL
σ αβ := 2 , Σα := α , ∆α β := . (107)
δgαβ δϑ δΓα β
From GR we expect that σ αβ is the metric (and symmetric) energy-momentum
current of matter (‘Hilbert current’), whereas Secs. 4 and 5 lead us to the believe
that Σα should be the corresponding canonical energy-momentum current (‘Noether
current’). The canonical hypermomentum current ∆α β , which couples to the linear
connection, can be decomposed according to
1 ⌢
∆αβ = ταβ + gαβ ∆ + ∆
րαβ (108)
4
∼ spin current ⊕ dilation current ⊕ shear current ,

with ταβ := ∆[αβ] as (dynamical) spin current, ∆ := ∆γ γ , as dilation current, and

⌢
∆
րαβ as symmetric and tracefree shear current.
If one applies the Noether theorem to the material Lagrangian, then, Σα , ταβ , and
∆ can be identified with the special-relativistic canonical Noether currents of energy-
⌢
momentum, spin, and dilation, respectively. This leaves only the shear current ∆ րαβ ,
see our earlier work21 for a detailed discussion, as a concept that is a bit more remote
from direct observation than the other currents. In any case, the definitions (107)
and their physical interpretations are completely justified by these considerations.
Adding to the matter Lagrangian a metric-affine first order gauge Lagrangian V ,
Z h i
W = V (gαβ , ϑα , Qαβ , T α , Rα β ) + L(gαβ , ϑα , Ψ, DΨ) , (109)

and applying the action principle, we find29 the matter and the gauge field equations
(of Yang-Mills type):
!
∂L ∂L
D − (−1)p = 0, (MATTER) (110)
∂(DΨ) ∂Ψ
!
∂V ∂V
D 2 +2 = − σ αβ , (ZEROTH) (111)
∂Qαβ ∂gαβ
!
∂V ∂V
D + = − Σα , (FIRST) (112)
∂T α ∂ϑα
!
∂V ∂V ∂V
D + ϑα ∧ + 2 gβγ = − ∆α β . (SECOND) (113)
∂Rα β ∂T β ∂Qαγ
In SECOND a Noether identity for V has already been employed. Analogously, in
FIRST, the canonical energy-momentum of the translational gauge potential ϑα can
be expressed explicitly as
∂V ∂V ∂V ∂V
α
= eα ⌋V − (eα ⌋T β ) ∧ β
− (eα ⌋Rβ γ ) ∧ γ
− (eα ⌋Qβγ ) . (114)
∂ϑ ∂T ∂Rβ ∂Qβγ
This structure is known from Minkowski’s energy-momentum tensor of the Maxwellian
field. It is interesting to note that, provided SECOND is fulfilled, FIRST and ZE-
ROTH are equivalent, i.e., one of them is redundant. It is for this reason that we
abstain, for ∂V /∂gαβ , to display a formula similar to (114).
A
The field equation of a Yang-Mills theory reads D (∂VYM /∂F ) = −J. The field
equations ZEROTH, FIRST, and SECOND are of this type. However, because of the
universality of ‘external’ interactions (gravitation), additional tensor-valued gauge
currents (∂V /∂ϑα etc.) surface in the field equations. This is the distinguishing
feature of MAG as compared to gauge theories of internal groups (U(1), SU(2), . . .).
In Yang-Mills theory the non-linearity of the gauge field is hidden in the non-tensorial
A
pieces of the gauge covariant exterior derivative D occurring on the left hand side
A
D (∂VYM /∂F ) of the Yang-Mills equation. In gravitational gauge theory, there are
Γ
additional non-linearities, besides those in D , namely the ones represented by (114)
etc.. This result is non-trivial as a simple argument will show:
Suppose we had a Hilbert-Einstein gauge Lagragian linear in the curvature Rα β .
Then in FIRST the leading term in differential order on its left hand side will vanish
and we are left with (114). The surviving terms are
∂VHE ∂VHE
α
∼ eα ⌋VHE − (eα ⌋Rβ γ ) ∧ ∼ − Σα . (115)
∂ϑ ∂Rβ γ

Since ∂VHE /∂Rβ γ will be a constant, we recover the Einstein three-form (correspond-
ing to the Einstein tensor in Ricci calculus) from this equation, giving substance to
Schrödinger’s dictum83 that the left hand side of Einstein’s equation is, in some sense,
the gravitational energy-momentum tensor. Consequently, Einstein’s field equation
of GR is encapsulated in ∂V /∂ϑα of FIRST and, as such, has a distinctive anti-Yang-
Mills flavor. In contrast, the Einsteinian teleparallelism GR|| , with its torsion-square
LagrangianV||, picks up an essential piece from the proper Yang-Mills term of FIRST,
∂V|| ∂V||
D α
+ eα ⌋V|| − (eα ⌋T β ) ∧ ∼ − Σα , (116)
∂T ∂T β
compare (77). It has much more of the Yang-Mills spirit than GR has – and this is
the reason why GR|| turned up when we gauged the translations in Secs. 4 and 5.
If, within our formalism, one desires to correctly derive the field equation for GR,
see (115), and for GR||, see (116), then one has to put on Lagrange multipliers. In the
case of (115) they have to kill nonmetricity and torsion and for (116) to remove non-
metricity and curvature. The details have been worked out in our earlier review29 , see
also Kopczyński38 . MAG (metric-affine gauge theory, the general framework), PG
(Poincaré gauge theory, vanishing nonmetricity, hence in a Riemann-Cartan space-
time), EC (Einstein-Cartan theory, the PG with the curvature scalar as gravitational
Lagrangian), GR|| (Einsteinian teleparallelism, vanishing nonmetricity, vanishing cur-
vature, hence in a Weitzenböck spacetime, specific torsion square Lagrangian), and
GR (general relativity, vanishing nonmetricity, vanishing torsion, hence in a Rieman-
nian space, curvature scalar as Lagrangian) are different (sub-)cases of this general
scheme.
A further remark is in order: Gauging the affine group yields the gauge poten-
tials (ϑα , Γα β ), see Table 5. If a metric exists on top of that linearly connected
manifold, then a further independent geometrical field variable is at hand. Following
Trautman88 , we are taking (gαβ , ϑα , Γα β ) as independent variables in the action (109).
Because of the redundancy of ZEROTH or FIRST, provided SECOND is fulfilled, one
could argue that one should drop, say, the coframe as independent variable, as done
by Tuckerf and Wang95 , for example. In the earlier metric-affine unified field theories
à la Einstein13 (App.2) and Schrödinger83 , the coframe didn’t even show up since they
worked in a holonomic formalism (unsuitable for representing fermions). Because of
various arguments, however, we feel more comfortable with our procedure: (i) Both
energy-momentum currents, the Hilbertian σ αβ and the Noetherian Σα , have a good
and direct physical interpretation in SR. Dropping a gauge variable means dropping
one of these useful quantities as a fundamental current. (ii) A posteriori, we do find
the identity causing the redundancy mentioned. This seems safer than to assume
something to that extend a priori. (iii) If we dropped the metric as an independent
variable, for example, then we would have to take orthonormal frames as field vari-
ables, giving away a piece of freedom which we had earlier in being able to work with
whatever frame we liked, be it orthonormal or oblique.
Let us finally remind ourselves: As soon as the gauge Lagrangian V is specified
explicitly, we can find the field equations ZEROTH, FIRST, and SECOND by sheer
partial differentiation. Hence the using of the general form of the field equations may
save a lot of work.

8. Model building: Einstein-Cartan theory and beyond

The missing piece within the framework that we finally established in Sec. 7 is the
gauge field Lagrangian V (gαβ , ϑα , Qαβ , T α, Rα β ). The hope is that the model, with
a suitably chosen V – perhaps combined with some symmetry breaking mechanism
which, for example, reduces the linear group GL(4, R) to the Lorentz group SO(1, 3)
– can be consistently quantized. For 1+1 dimensional curvature square models51,64,67
f
We are grateful to Robin W. Tucker (Lancaster) for an interesting discussion on this question.
successful quantization methods are already available, see Kloesch and Strobl37 and
the literature given there.
In testing a new framework, one first wants, in some limit, to recover old ground
where one feels at home. In our case this is GR. There are at least two ways of
how to achieve this: One can choose the curvature scalar à la Hilbert (in exterior
calculus: the corresponding four-form Rαβ ∧ ηαβ à la Trautman88 ) and take Lagrange
multipliers for extinguishing nonmetricity and torsion:
1 αβ 1
VGR = − R ∧ ηαβ + Qαβ ∧ (1) λαβ + T α ∧ (2) λα . (117)
2ℓ2 2
Note that Rα β is the curvature tensor of the independent field variable Γα β , for the
η’s see (106).

8.1. Einstein-Cartan theory EC

A second way is to start with the Einstein-Cartan Lagrangian (here with cosmo-
logical constant)
1 αβ 1 1
VEC = − 2
R ∧ ηαβ + Λ η + Qαβ ∧ λαβ , (118)
2ℓ 2 2
and to derive the corresponding field equations:
1
ηαβγ ∧ Rβγ + Λ ηα = ℓ2 Σα , (119)
2
1
ηαβγ ∧ T γ = ℓ2 ταβ . (120)
2
In components in a holonomic frame they read:
1
Ric ij − gij Ric k k + Λ gij = ℓ2 Σij , (121)
2
Tij k + 2 δ[ik Tj]l l = ℓ2 τij k . (122)

We recover GR for vanishing spin ταβ = 0. In this context only the vanishing of
nonmetricity had to be assumed. The vanishing of torsion, for spinless matter, was
the result of the second field equation (120). We will see below, in Sec. 8.2, that also
the last Lagrange multiplier can be abandoned if one amends the gravitational part
of the Lagrangian with a piece quadratic in the Weyl one-form.
Basically, the two EC field equations (119) and (120) are first order partial differ-
ential equations in Γα β and ϑα , with a spin fluid as source, see ref.(65 ). As we explained
already in the last Sec. 7, this is the anti-Yang-Mills flavor of GR or EC caused by
the absence of an explicit torsion piece in the Lagrangian. A physical consequence is
that in EC we have the usual gravitational interaction of the Newton-Einstein type
plus a very weak (non-propagating) spin contact interaction. Up to the fifties, weak
interaction was also thought to be of contact type (in fact, of a vector-axial vector
type). Later, following gauge ideas, the short-range intermediate W -boson (and the
Z) were postulated that made the weak interaction propagate.

8.2. Poincaré gauge theory PG, the quadratic version

The simplest appearance of explicit torsion pieces is that in the GR|| Lagrangian24,25
studied in Sec. 5:
1 1 1
 
V|| = − 2 T α ∧ ∗ −(1) Tα + 2(2) Tα + (3) Tα + Qαβ ∧ (1) λαβ + Rα β ∧ (2) λα β . (123)
2ℓ 2 2
This can be considered as a starting point for turning to Lagrangians quadratic in
the field strengths. Amongst the simplest model cases is the purely quadratic von der
Heyde et al.32,6 Lagrangian
1  α  1 αβ ∗ 1
VvdH = − 2
T ∧ ϑβ
∧ ∗ (Tβ ∧ ϑα ) − R ∧ Rαβ + Qαβ ∧ λαβ . (124)
2ℓ 2κ 2
This Lagrangian has been ‘derived’ by means of the Gordon decomposition argu-
ment32,25 , see also Rumpf81 . It may have problems with the positivity of the energy,
see refs.(85,41,97 ), but the situation is not completely clear to us. The torsion square
piece in this Lagrangian, in a Weitzenböck spacetime, has a classical Newtonian limit.
It differs from the torsion pieces in the teleparallel Lagrangian (123) by a quadratic
axial torsion piece (then in (123) we had −(3) Tα instead).
A number of exact classical solutions has been found for the model (124), a Kerr-
NUT solution with torsion4 is amongst the most prominent ones. For illustrating the
basic features of such solutions, we display the less complicated subcase, namely the
Baekler-Lee3,43 solution – this is the Reissner-Nordström solution with dynamic tor-
sion – as a fairly transparent example. We choose Schwarzschild coordinates (t, r, θ, φ),
M = mass, q = electric charge, and find the orthonormal coframe
1 1
ϑt̂ = [(Φ + 1)dt + (1 − )dr] ,
2 Φ
1 1
ϑr̂ = [(Φ − 1)dt + (1 + )dr] ,
2 Φ
ϑθ̂ = rdθ ,
ϑϕ̂ = r sin θ dϕ . (125)

The corresponding metric reads:

1 2
ds2 = −Φdt2 + dr + r 2 (dθ2 + sin2 θ dϕ2 ) . (126)
Φ
In the Reissner-Nordström function
2(Mr − q 2 ) κ
Φ := 1 − 2
− 2 r2 (127)
r 4ℓ
the ‘cosmological term’ is induced by the Yang-Mills type curvature square piece
in the Lagrangian (‘strong gravity’, cf. ref.(63 )). If we additionally had a ‘naked’
cosmological constant, we could shift its value by a suitable choice of the ‘strong’
coupling constant κ. Torsion and curvature are given by

t̂ r̂ Mr − 2q 2 t̂
T =T = 3
ϑ ∧ ϑr̂ ,
r
Mr − q 2  t̂ 
T θ̂ = ϑ ∧ ϑθ̂
− ϑr̂
∧ ϑθ̂
,
r3
Mr − q 2  t̂ 
T ϕ̂ = ϑ ∧ ϑϕ̂
− ϑ r̂
∧ ϑ ϕ̂
, (128)
r3
and
κ α Mr − q 2 (4)
Rαβ = ϑ ∧ ϑβ
+ Rαβ , (129)
4ℓ2 r2
respectively, where
(4) (4) κ  t̂ 
Rt̂θ̂ = Rr̂θ̂ := ϑ ∧ ϑθ̂
− ϑr̂
∧ ϑθ̂
,
4ℓ2 
(4) (4) κ 
Rt̂ϕ̂ = Rr̂ϕ̂ := 2 ϑt̂ ∧ ϑϕ̂ − ϑr̂ ∧ ϑϕ̂ , (130)
4ℓ
represent a tracefree symmetric Ricci piece of the curvature two-form. The Coulomb
field shows up in the electromagnetic field strength:
2 q t̂
F = ϑ ∧ ϑr̂ . (131)
ℓ r2
It is a spherically symmetric vacuum solution with Maxwell field, i.e. a Reissner-
Nordström solution with dynamic torsion. We don’t display the solution in its origi-
nal frame but in a suitably rotated one5 such that the torsion two-form (128) has a
‘Coulombic’ look without global extra factors in front of the corresponding expres-
sions. Note the factor of two in the T t̂ component of (128) in the q 2 -piece. One
interesting feature of this solution is that the curvature square Lagrangian supplies
a constant curvature background proportional to κ, see the first piece on the right
hand side of (129). The torsion is – this is not an unexpected feature in the light
of our teleparallelism ‘philosophy’ – induced by ordinary Newton-Einstein gravity, as
can be read off from (128), a fact usually hard to swallow by colleagues who relate
torsion with obscurity. In (128) torsion is visibly the translation field strength, a fact
made possible by the purely quadratic torsion piece in the Lagrangian, without a
Hilbert-Einstein type admixture.
 The von der Heyde Lagrangian (124) is a subcase of the general quadratic PG
Lagrangian25,17,66
3
!
Λ a0 αβ 1 α ∗ X
VQPG = 2η + 2
R ∧ ηαβ + 2 T ∧ a(M ) (M ) Tα
ℓ 2ℓ 2ℓ M =1
6
!
1 αβ ∗ X 1
+ R ∧ b(N ) (N ) Rαβ + Qαβ ∧ λαβ . (132)
2κ N =1 2
Each of the three irreducible torsion and six irreducible curvature pieces contributes to
the Lagrangian with an individual weight. The propagating modes of this Lagrangian
were investigated by Sezgin and van Nieuwenhuizen85 and by Kuhfuss and Nitsch41 .
A subclass of Lagrangians survived their selection criteria motivated by quantum field
theoretical considerations (ghost-freeness, positive energy). Minkevich54,55 , amongst
others, studied Friedmann type cosmological models resulting from such a Lagrangian.

8.3. Coupling to a scalar field

It is near at hand to add a dilaton type massless scalar field to this Lagrangian:
1
VΦgrav = VQPG + dΦ ∧ ∗dΦ . (133)
2
We found for this model a remarkable exact solution, a torsion kink6 . Attributing
distinctive Higgs-like features to the scalar field, we arrive at the more general model
of Floreanini and Percacci16 :
a0
VFP = c1 dΦ ∧ ∗dΦ + c2 dΦ ∧ ∗(eµ ⌋T µ ) + U(Φ) + 2 Φ2 Rαβ ∧ ηαβ (134)
2ℓ
3 6
! !
1 α ∗ X (M ) 1 αβ ∗ X (N ) 1
+ 2
T ∧ aM Tα + R ∧ bN Rαβ + Qαβ ∧ λαβ .
2ℓ M =1 2 N =1 2
The scalar Φ couples to the EC term in a Jordan-Brans-Dicke type way. The explicit
form of the ‘Higgs’ potential U(Φ) is left open. Note the direct coupling dΦ ∧ ∗T
in (134) which is, however, odd in Φ. The authors of (134), on a quantum field
theoretical level, investigated the renormalizability properties of their model.
Being relativists, we would be ill-advised if we didn’t try to give the scalar field Φ
a geometrical meaning, perhaps in the context of the Weyl one-form Q which is of the
type of a gauge potential for dilations anyways. Therefore we lift the last Lagrange
multiplier and now turn to...

8.4. Metric-affine gauge theory MAG

Let us, however, first continue the discussion of above of how to arrive at GR
in spite of relaxing the last constraint and liberating thereby the connection com-
pletely from its dominance by the metric. The naive way, namely just to take a term
proportional to Rαβ ∧ ηαβ , doesn’t work. The projective transformation

Γα β −→ Γα β + δαβ P , (135)

with some one-form field P , leaves the Hilbert-Einstein type Lagrangian invariant.
Consequently, in such a model, the connection would only be determined up to a
one-form (with four components). This is unsatisfactory. Moreover, if one coupled
the gauge fields to matter, then only projectively invariant matter Lagrangians would
be allowed, an a priori constraint without physical justification. Therefore one has to
remove the projective invariance from the gravitational Lagrangian.
Since a projective transformation changes the trace Γγ γ of a connection and this
trace is closely related to the Weyl one-form,
q
Γγ γ = 2 Q + d ln |det gαβ | , dΓγ γ = Rγ γ = 2 dQ , (136)

see Eq.(3.10.13) of ref.(29 ), an obvious way to remove the projective invariance is to

add the square of the Weyl one-form26,47 :
1  αβ ∗

VGR′ = − R ∧ ηαβ + β Q ∧ Q . (137)
2ℓ2
A moment’s reflection will remind ourselves what we have achieved by this innocently
looking Lagrangian (137): Varying metric, frame, and connection independently, in
vacuum – that is, in the absence of matter – yields the Einstein vacuum equation
in a Riemannian spacetime. Generically, if matter is present and supplies energy-
momentum and hypermomentum currents, then the hypermomentum, via the second
field equation (113), turns out to be proportional to the post-Riemannian pieces of
the connection.
The problem with the Lagrangian (137) is that it has the same defect as the EC
Lagrangian. We have Newton-Einstein gravity and no further propagating modes.
The obvious remedy is to add kinetic terms and, since we called the Weyl one-form
already in earlier, we may want to make these modes propagating28,78 :
1  αβ ∗
 α
VGRQ = − 2
R ∧ ηαβ + β Q ∧ Q − dQ ∧ ∗dQ . (138)
2ℓ 2
In our first order formalism the Lagrangian must be expressed in terms of the gauge
potential and their first derivatives. Since Q is already on the level of a field strength,
one may be tempted to forbid the kinetic Q-terms. However, by means of the trace
of the zeroth Bianchi identity, dQ = 21 Rγ γ , we can cast out the pseudo-second order
terms and find the well-behaved first order Lagrangian
1  αβ ∗
 α
VGRQ = − 2
R ∧ ηαβ + β Q ∧ Q − Rα α ∧ ∗Rβ β . (139)
2ℓ 8
This might be the simplest reasonable metric-affine Lagrangian with propagating
dilation modes. Probably it deserves closer investigation. If one amended (139) with
a piece ∼ Rαβ ∧ ∗(3) Zαβ , see (140), then one would expect the simplest shear modes
to arise. Here Zαβ := R(αβ) is the symmetric (post-Riemannian) and, for later use,
Wαβ := R[αβ] the antisymmetric piece of the curvature two-form.
The trace Rβ β of the curvature in (139) represents an irreducible piece of the
curvature, in fact the piece which we numbered as ten and called, in our computer
algebra programs84 , DILCURV. Altogether, in a metric-affine space, the curvature
has eleven irreducible pieces, see ref.(29 ), Table 4. If we recall that the nonmetricity
has four irreducible pieces, then the general quadratic Lagrangian in MAG reads:
3
" !
1 ∗
αβ α
aI (I) Tα
X
VQMA = −a0 R ∧ ηαβ + 2Λ η + T ∧
2ℓ2 I=1
4 4
! !#
(I) α ∗ β ∗ (I) αβ
X X
+ 2 cI Qαβ ∧ ϑ ∧ T + Qαβ ∧ bI Q
I=2 I=1
6 5
!
1 αβ ∗ X
wI (I) Wαβ + zI (I) Zαβ
X
+ R ∧ . (140)
2 I=1 I=1

In spite of this affluence of generality, Tresguerres91,g was able to find two exact
solutions of the corresponding vacuum field equations in a most remarkable piece of
work. He also discussed91 the reasons for introducing the mixed Q ∧ ϑ ∧ ∗ T term.
Generically Tresguerres found a Baekler-Lee type solution with torsion, see (125) to
(131) above, but the dilaton charge takes the place of the electric charge. Thus the
Weyl one-form Q, the quasiMaxwellian potential, has a 1/r-behavior and is closely
interwoven with the torsion vector. In the second solution, the metric is the same,
but, additionally, two types of shear charges emerge. The Tresguerres solutions with
dilation and shear charges are something qualitatively new in gauge models of gravity.
Tucker and Wang95 put β = 0 in (138) or (139),
1 αβ α 1 α
VTW = − 2
R ∧ ηαβ − dQ ∧ ∗dQ = − 2 Rαβ ∧ ηαβ − Rα α ∧ ∗Rβ β , (141)
2ℓ 2 2ℓ 8
i.e. they excluded the massive term of the Weyl field from further consideration.
Rather, they explored the analogy of the Weyl one-form Q to the Maxwell potential
A under these circumstances. Their field equations can be read off from (112), (113),
and (114) straightforwardly as,
∂VTW
− eα ⌋VTW + (eα ⌋Rβ γ ) ∧ = Σα , (142)
∂Rβ γ
!
∂VTW
−D β
= ∆α β , (143)
∂Rα
or
1 αh i
ηαβγ ∧ Rβγ + (eα ⌋Rβ β ) ∧ ∗ Rγ γ − (eα ⌋∗ Rβ β ) ∧ Rγ γ = ℓ2 Σα , (144)
2 8
g
See also Tresguerres89,90,92,93 , Macı́as et al.44 , and, for cosmological models, Minkevich56 .
 1 α
Dη αβ − δβα d ∗Rγ γ = ℓ2 ∆α β . (145)
2 4
Tucker and Wang95 found Baekler-Lee type vacuum solutions with dilation (‘Weyl’)
charge, just as Tresguerres, but, in addition, they presented a highly interesting so-
lution with propagating massless spinor matter as source.
Nevertheless, one should recognize that the Lagrangian (141) is not without prob-
lems. Because the massive piece ∼ Q ∧ ∗ Q is missing, the gauge Lagrangian VTW is
invariant under the special projective transformation
Γα β −→ Γα β + δαβ d p . (146)
Whereas such a type of ‘gauge’ transformation is desirable for an internal U(1)-
connection – like in Maxwell’s theory – it is definitely dangerous in the context of a
dilation transformation. The linear connection is only determined up to the transfor-
mation (146), that is, not all of the 64 components are uniquely determined in the
TW-model, see also the analysis of Teyssandier and Tucker86 .
The case studies of Tresguerres and Tucker-Wang taught us that the Weyl one-
form plays a particular role in representing a dilation type field that ought to be
useful in the context of the breaking of the dilation symmetry. Therefore there were
attempts by Mielke et al., see ref.(29 ) Sec. 6, to superimpose on the Lagrangian
(140) an additional conformal symmetry in order to have a massless theory, free of
dimensionful coupling constants at the beginning. Then one couples to a hypothetical
dilaton field σ,
σ 2 αβ 1 λ
V =− R ∧ ηαβ + (Dσ) ∧ ∗Dσ + σ 4 η , (147)
2 2 4
and breaks the dilation symmetry, thereby arriving at a low energy massive La-
grangian. Analogous approaches were proposed by Gregorash and Papini19,20 , Hochberg
and Plunien33 , and by Poberii77 . Recently Pawlowski and Ra̧czka75,74,94 developed
similar models in a Riemannian spacetime, though. Nevertheless, they required con-
formal invariance for the Lagrangian they started from, coupled to a dilation field
etc.. There are close relationships between the gravitational sectors of these models
which lead us to the belief that the P&R model should be redrafted in the framework
of a Weyl-Cartan, if not of a metric-affine spacetime.
With all these developments from different quarters, we have here – perhaps for
the first time – a consistent framework for gauge models carrying both, non-trivial
torsion and nonmetricity.

9. Acknowledgments

One of the authors would like to thank Venzo de Sabbata, Peter Bergmann, and
H.-J. Treder for the invitation to lecture at this Erice School and his coauthors29 Der-
mott McCrea† (Dublin), Eckehard Mielke (Kiel/Mexico City), and Yuval Ne’eman
(Tel Aviv/Austin) for great help. Moreover, he is very obliged to H.A. Kastrup
(Aachen), G. Süssmann (Munich), Ryszard Ra̧czka (Warsaw), and H. Petry (Bonn)
for seminar invitations where different versions of these lectures were given. We
are most grateful to our colleagues David Hartley (St.Augustin), Yuri Obukhov
(Moscow/Cologne), and Romualdo Tresguerres (Madrid) for help and many clari-
fying discussions.

10. References

1. A. Ali, J. Ellis, and S. Randjbar-Daemi (eds.): Salamfestschrift. A Collection

of Talks from the Conference on Highlights of Particle and Condensed Matter
Physics (World Scientific, Singapore, 1994).
2. J. Audretsch, F.W. Hehl, and C. Lämmerzahl: Matter wave interferometry
and why quantum objects are fundamental for establishing a gravitational the-
ory. In: ‘Relativistic Gravity Research’ Proc., Bad Honnef, Germany 1991, J.
Ehlers and G. Schäfer, eds., Lecture Notes in Physics (Springer) 410 (1992)
368–407.
3. P. Baekler: A spherically symmetric vacuum solution of the quadratic Poincaré
gauge field theory of gravitation with Newtonian and confinement potentials.
Phys. Lett. B99 (1981) 329–332.
4. P. Baekler, M. Gürses, F.W. Hehl, and J.D. McCrea: The exterior gravi-
tational field of a charged spinning source in the Poincaré gauge theory: A
Kerr-Newman metric with dynamic torsion. Phys. Lett. A128 (1988) 245–
250.
5. P. Baekler, R. Hecht, F.W. Hehl, and T. Shirafuji: Mass and spin of exact
solutions of the Poincaré gauge theory. Progr. Theor. Phys. 78 (1987) 16–21.
6. P. Baekler, E.W. Mielke, R. Hecht, and F.W. Hehl: Kinky torsion in a
Poincaré gauge model of gravity coupled to a massless scalar field. Nucl. Phys.
B288 (1987) 800–812.
7. J.D. Bjorken and S. Drell: Relativistic Quantum Mechanics (McGraw-Hill,
New York, 1964).
8. U. Bonse and T. Wroblewski: Measurement of neutron quantum interference.
Phys. Rev. Lett. 51 (1983) 1401–1404.
9. H.-H.v. Borzeszkowski and H.-J. Treder: Mixed field theories and Weyl-Cartan
geometry. Lecture given at this Erice School (1995).
10. Y. M. Cho: Einstein Lagrangian as the translational Yang-Mills Lagrangian.
Phys. Rev. D14 (1976) 2521–2525.
11. R. Colella, A.W. Overhauser, and S.A. Werner: Observation of Gravitation-
ally Induced Quantum Interference. Phys. Rev. Lett. 34 (1975) 1472–1474.
12. R. Debever (ed.): Elie Cartan – Albert Einstein: Lettres sur le Parallélisme
Absolu 1929–1932. Original text with English translation. Académie Royale
 de Belgique and Princeton University Press 1979.
13. A. Einstein: The Meaning of Relativity, 5th ed. (Princeton University Press,
Princeton, 1955).
14. A. Einstein: Preface as of 1955 April 04 in ‘Cinquant’anni di Relatività
1905–1955.’ M. Pantaleo, ed. Edizioni Giuntine and Sansoni Editore, Firenze
(1955). Translation from the German by D. Hartley and the authors. The
German original of the piece we quoted reads as follows: “...die wesentliche
Leistung der allgemeinen Relativitätstheorie, nämlich die Überwindung (des)
‘starren’ Raumes, d.h. des Inertialsytsems, ist nur indirekt mit der Einführung
einer Riemann-Metrik verbunden. Das unmittelbar wesentliche begriffliche
Element ist das die infinitesimale Verschiebung von Vektoren ausdrückende
‘Verschiebungsfeld’ (Γlik ). Dieses nämlich ersetzt den durch das Inertialsys-
tem gesetzten Parallelismus räumlich beliebig getrennter Vektoren (nämlich
Gleichheit entsprechender Komponenten) durch eine infinitesimale Operation.
Dadurch wird die Bildung von Tensoren durch Differentiation ermöglicht und
so die Einführung des ‘starren’ Raumes (Inertialsystem) entbehrlich gemacht.
Dem gegenüber erscheint es in gewissem Sinne von sekundärer Wichtigkeit,
dass ein besonderes Γ-Feld sich aus der Existenz einer Riemann-Metrik de-
duzieren lässt...”
15. R. Feynman: Lectures on Gravitation. Lecture notes by F.B. Morinigo
and W.G. Wagner (California Institute of Technology, Pasadena, California
1962/63).
16. R. Floreanini and R. Percacci: Renormalization-group flow of a dilaton po-
tential. Phys. Rev. D52 (1995) 896–911.
17. H. Goenner: Alternative theories of gravity. In: ‘General Relativity and
Gravitation, GR11’. M.A.H. MacCallum, ed. (Cambridge University Press,
Cambridge, 1987) pp. 262–273.
18. D.M. Greenberger and A.W. Overhauser: The role of gravity in quantum
theory. Scientific American 242 (May 1980) 54–64.
19. D. Gregorash and G. Papini: Weyl–Dirac theory with torsion. Nuovo Cimento
B55 (1980) 37–51.
20. D. Gregorash and G. Papini: Weyl–Dirac theory with torsion. II. Foundations
and conservation equations. Nuovo Cimento B56 (1980) 21–38.
21. F. Gronwald and F.W. Hehl: Stress and hyperstress as fundamental con-
cepts in continuum mechanics and in relativistic field theory. In: ‘Advances in
Modern Continuum Dynamics’, International Conference in Memory of Anto-
nio Signorini, Isola d’Elba, June 1991. G. Ferrarese, ed. (Pitagora Editrice,
Bologna, 1993) pp. 1–32.
22. R.T. Hammond: New fields in general relativity. Contemporary Physics 36
(1995) 103–114.
23. D. Hartley: Normal frames for non-Riemannian connections. Class. Quan-
 tum Grav. 12 (1995) L103–L105.
24. K. Hayashi and T. Shirafuji: New general relativity. Phys. Rev. D19 (1979)
3524–3553.
25. F.W. Hehl: Four lectures on Poincaré gauge theory. In: Proc. of the 6th
Course of the School of Cosmology and Gravitation on ‘Spin, Torsion, Rota-
tion, and Supergravity’, held at Erice, Italy, May 1979, P.G. Bergmann, V. de
Sabbata, eds. (Plenum, New York. 1980) pp. 5–61.
26. F.W. Hehl, G.D. Kerlick, and P. von der Heyde: On a new metric affine
theory of gravitation. Phys. Lett. 63B (1976) 446–448.
27. F.W. Hehl, J. Lemke, and E.W. Mielke: Two lectures on fermions and gravity.
In: ‘Geometry and Theoretical Physics’, J. Debrus and A.C. Hirshfeld (eds.)
(Springer, Berlin, 1991) pp. 56–140.
28. F.W. Hehl, E.A. Lord, L.L. Smalley: Metric–affine variational principles in
general relativity. II. Relaxation of the Riemannian constraint. Gen. Relat.
Grav. 13 (1981) 1037–1056.
29. F.W. Hehl, J.D. McCrea, E.W. Mielke, and Y. Ne’eman: Metric-Affine Gauge
Theory of Gravity: Field Equations, Noether Identities, World Spinors, and
Breaking of Dilation Invariance. Phys. Rep. 258 (1995) 1–171.
30. F.W. Hehl and W.-T. Ni: Inertial effects of a Dirac particle. Phys. Rev.
D42 (1990) 2045–2048.
31. P. von der Heyde: The equivalence principle in the U4 theory of gravitation.
Lettre al Nuovo Cimento 14 (1975) 250–252.
32. P. von der Heyde: Is gravitation mediated by the torsion of spacetime? Z.
Naturf. 31a (1976) 1725–1726.
33. D. Hochberg and G. Plunien: Theory of matter in Weyl spacetime. Phys.
Rev. D43 (1991) 3358–3367.
34. J.C. Huang: Dirac particle, gravity, and inertial effects. Ann. Physik
(Leipzig) 3 (1994) 53–67.
35. B.Z. Iliev: Special bases for derivations of tensor algebras. I. Cases in the
neighborhood and at a point. II. Case along paths. III. Case along smooth
maps with separable points of selfintersection. Preprints E5-92-507, E5-92-508,
E5-92-543 (Joint Institute for Nuclear Research, Dubna, 1992).
36. R. d’Inverno: Introducing Einstein’s Relativity. (Clarendon Press, Oxford,
1992).
37. T. Kloesch and T. Strobl: Classical and quantum gravity in 1+1 dimen-
sions, part I: A unifying approach; part II: The universal coverings. Electronic
archive Los Alamos, gr-qc/9508020 and gr-qc/9511081 (1995).
38. W. Kopczyński: The Palatini principle with constraints. Bull. de l’Academie
Polon. des Sci., Sér. math. astr. et phys. 23 (1975) 467–473.
39. W.Kopczyński: Problems with metric-teleparallel theories of gravitation. J.
Phys. A15 (1982) 493–506.
40. W. Kopczyński and A. Trautman: Spacetime and Gravitation. Translated
from the Polish (Wiley, Chichester and PWN – Polish Scientific Publishers,
Warszawa, 1992).
41. R. Kuhfuss and J. Nitsch: Propagating modes in gauge field theories of grav-
ity. Gen. Relat. Grav. 18 (1986) 1207–1227.
42. C. Lämmerzahl: On the equivalence principle in quantum theory. to appear
in Gen. Rel. Grav. (1996).
43. C.H. Lee: A spherically symmetric electro-vacuum solution of the Poincaré
gauge field theory of gravitation. Phys. Lett. B130 (1983) 257–259.
44. A. Macı́as, E.W. Mielke, and H.A. Morales-Técotl: Torsion and Weyl con-
vector in metric-affine models of gravity, J. Math. Phys. 36 (1995) 5868–5976.
45. A. Macı́as, E.W. Mielke, and H.A. Morales–Técotl: Chiral fermions coupled
to chiral gravity. In: ‘Proc. of CAP 50 aniversary’, E. Lesard et al., eds.
(Springer Verlag, Berlin) in print.
46. B. Mashhoon: Neutron interferometry in a rotating frame of reference.
Phys. Rev. Lett. 61 (1988) 2639–2642.
47. J.D. McCrea, F.W. Hehl, and E.W. Mielke: Mapping Noether identities into
Bianchi identities in general relativistic theories of gravity and in the field
theory of static lattice defects. Int. J. Theor. Phys. 29 (1990) 1185–1206.
48. E.W. Mielke: Geometrodynamics of Gauge Fields — On the geometry of
Yang-Mills and gravitational gauge theories (Akademie–Verlag, Berlin, 1987).
49. E.W. Mielke: Positive–gravitational–energy proof from complex variables?
Phys. Rev. D42 (1990) 3388–3394.
50. E.W. Mielke: Ashtekar’s complex variables in general relativity and its
teleparallelism equivalent. Ann. Phys. (N.Y.) 219 (1992) 78–108.
51. E.W. Mielke, F. Gronwald, Yu.N. Obukhov, R. Tresguerres, and F.W. Hehl:
Towards complete integrability of two dimensional Poincaré gauge gravity.
Phys. Rev. D48 (1993) 3648–3662.
52. E.W. Mielke, Yu.N. Obukhov, and F.W. Hehl: Yang–Mills configurations
from 3D Riemann–Cartan geometry. Phys. Lett. A192 (1994) 153–162.
53. R. Mills: Gauge fields. American J. Physics 57 (1989) 493–507.
54. A.V. Minkevich: Generalized cosmological Friedmann equations without grav-
itational singularity. Phys. Lett. A80 (1980) 232–234.
55. A.V. Minkevich: Generalized cosmological Friedman equation and de Sitter
solution. Phys. Lett. A95 (1983) 422–424.
56. A.V. Minkevich and I.M. Nemenman: On the influence of the gravitating
vacuum on the dynamics of homogeneous isotropic models in gauge theories of
gravity, Class. Quantum Grav. 12 (1995) 1259–1265.
57. C.W. Misner, K.S. Thorne, and J.A. Wheeler: Gravitation. (Freeman, San
Francisco, 1973).
58. L. Mistura: Cartan connection and defects in Bravais lattices. Int. J. Theor.
 Physics 29 (1990) 1207–1218.
59. G. Modanese and M. Toller: Radial gauge in Poincaré gauge field theories.
J. Math. Phys. 31 (1990) 452–458.
60. H.A. Morales–Técotl, A. Macı́as, and E.W. Mielke (1995): Geometrical
phases and translations. In: ‘1st Mexican School on Gravitation and Mathe-
matical Physics’, A. Macı́as et al., eds. (World Scientific, Singapore) in print.
61. Y. Ne’eman: Gravitational interaction of hadrons: Band–spinor representa-
tions of GL(n, R). Proc. Nat. Acad. Sci. (USA) 74 (1977) 4157–4159.
62. Y. Ne’eman: Spinor–type fields with linear, affine and general co–ordinate
transformations. Ann. Inst. Henri Poincaré A28 (1978) 369–378.
63. Y. Ne’eman and Dj. Šijački: Chromogravity explains ‘strong gravity’. In ref.
[1] pp. 540–551.
64. Yu.N. Obukhov: Two-dimensional Poincaré gauge gravity with matter. Phys.
Rev. D50 (1994) 5072–5086.
65. Yu.N. Obukhov and V.A. Korotky: The Weyssenhoff fluid in Einstein-Cartan
theory. Class. Quant. Grav. 4 (1987) 1633–1657.
66. Yu.N. Obukhov, V.N. Ponomariev, and V.V. Zhytnikov: Quadratic Poincaré
gauge theory of gravity: a comparison with the general relativity theory. Gen.
Relat. Grav. 21 (1989) 1107–1142.
67. Yu.N. Obukhov, S.N. Solodukhin, and E.W. Mielke: Coupling of lineal
Poincaré gauge gravity to scalar fields. Class. Quantum Grav. 11 (1994)
3069–3079.
68. L. O’Raifeartaigh: Hidden gauge symmetry. Rep. Progr. Phys. 42 (1979)
159–223.
69. L. O’Raifeartaigh: The dawning of gauge theory. In ref.[1] pp. 577–591.
70. A. Pais: ‘Subtle is the Lord...’ The Science and the Life of Albert Einstein
(Clarendon Press, Oxford, 1982).
71. J.-F. Pascual-Sánchez: lh theory of hadrons and leptons. Lett. Nuovo Cim.
28 (1980) 321–324.
72. J.-F. Pascual-Sánchez: On the limits of Poincaré gauge theory. Phys. Lett.
A108 (1985) 387–390.
73. J.-F. Pascual-Sánchez: Constrained macroscopic limit of Poincaré gauge the-
ory of gravity. Lett. Nuovo Cim. 44 (1985) 326–328.
74. M. Pawlowski and R. Ra̧czka: A unified conformal model for fundamental
interactions without dynamical Higgs field. Found. Phys. 24 (1994) 1305–
1327.
75. M. Pawlowski and R. Ra̧czka: A Higgs–free model for fundamental interac-
tions. Part 1: Formulation of the model. In: ‘Modern Group Theoretical
Methods in Physics’, J. Bertrand et al. (eds.) (Kluwer Academic, Dordrecht,
1995) pp. 221–232.
76. V. Perlick: Observer fields in Weylian spacetime models. Class. Quantum
 Grav. 8 (1991) 1369–1385.
77. E.A. Poberii: Metric-affine scale-covariant gravity. Gen. Relat. Grav. 26
(1994) 1011–1054.
78. V.N. Ponomariov and Ju. Obukhov: The generalized Einstein-Maxwell theory
of gravitation. Gen. Relat. Grav. 14 (1982) 309–330.
79. E. Prugovečki: Principles of Quantum General Relativity (World Scientific,
Singapore, 1995).
80. R. Puntigam, E. Schrüfer, and F.W. Hehl: The use of computer algebra in
Maxwell’s theory. In: ‘Computer Algebra in Science and Engineering’, J.
Fleischer et al. (eds.) (World Scientific, Singapore, 1995) pp. 195–211.
81. H. Rumpf: On the translational part of the Lagrangian in the Poincaré gauge
theory of gravitation. Z. Naturf. 33a (1978) 1224–1225.
82. V. de Sabbata and C. Sivaram: Spin and Torsion in Gravitation (World
Scientific, Singapore, 1994).
83. E. Schrödinger: Space–Time Structure. Reprinted with corrections (Cam-
bridge University Press, Cambridge, 1960).
84. E. Schrüfer, F.W. Hehl, and J.D. McCrea: Exterior calculus on the computer:
The REDUCE–package EXCALC applied to general relativity and the Poincaré
gauge theory. Gen. Rel. Grav. 19 (1987) 197–218.
85. E. Sezgin and P. van Nieuwenhuizen: New ghost-free Lagrangians with prop-
agating torsion. Phys. Rev. D21 (1980) 3269–3280.
86. P. Teyssandier and R.W. Tucker: Gravity, gauges and clocks. Class. Quan-
tum Grav., to appear (1995/96).
87. W. Thirring: Classical Field Theory. A Course in Mathematical Physics 2,
2nd ed. (Springer, New York, 1986).
88. A. Trautman: On the structure of the Einstein–Cartan equations. In: ‘Differ-
ential Geometry’, Symposia Mathematica Vol. 12 (Academic Press, London
1973) p. 139–162.
89. R. Tresguerres: An exact solution of (2 + 1)–dimensional topological gravity
in metric–affine spacetime. Phys. Lett. A168 (1992) 174–178.
90. R. Tresguerres: Topological gravity in a 3–dimensional metric–affine space.
J. Math. Phys. 33 (1992) 4231–4238.
91. R. Tresguerres: Exact vacuum solutions of 4–dimensional metric–affine gauge
theories of gravitation. Z. Phys. C65 (1995) 347–354.
92. R. Tresguerres, Exact static vacuum solution of four-dimensional metric-
affine gravity with nontrivial torsion. Phys. Lett. A 200 (1995) 405–410.
93. R. Tresguerres: Weyl–Cartan model for the non–massive cosmological eras
with a dilatedly polarized vacuum. Phys. Lett. A, submitted (1994).
94. E. Tsantilis, R.A. Puntigam, and F.W. Hehl: A quadratic curvature La-
grangian of Pawlowski and Ra̧czka: A finger exercise with MathTensor. In:
‘Relativity and Scientific Computing – Computer Algebra, Numerics, Visual-
ization –’ F.W. Hehl, R.A. Puntigam, H. Ruder (eds.) (Springer, Berlin, 1996)
to be published.
95. R.W. Tucker and C. Wang: Black holes with Weyl charge and non-
Riemannian waves. Class. Quantum Grav. 12 (1995) 2587–2605.
96. C. Wiesendanger: Poincaré gauge invariance and gravitation in Minkowski
spacetime. Electronic archive Los Alamos, gr-qc/9505049 (1995).
97. V.V. Zhytnikov: Wavelike exact solutions of R + R2 + Q2 gravity. J. Math.
Phys. 35 (1994) 6001–6017.
 This figure "fig1-1.png" is available in "png" format from:

http://arXiv.org/ps/gr-qc/9602013v1

View publication stats