Canonical Absolute Parallelism Overview

• Canonical Absolute Parallelism is defined by a global frame that induces a unique flat metric connection characterized by nontrivial torsion.
• It integrates concepts from teleparallel gravity, CR geometry, and Lie-group frameworks by encoding geometric data through invariant torsion structures.
• Conformal modifications yield invariant formulations, bridging traditional Riemannian geometries with teleparallel and generalized AP constructions.

Canonical Absolute Parallelism denotes a cluster of closely related constructions centered on a globally defined frame field and the distinguished connection that keeps that frame parallel. In the classical absolute-parallelism setting, an $n$-dimensional manifold $M$ carries $n$ linearly independent global vector fields, written either as $X_i$ or $\lambda_i$, and therefore admits a global coframe $\Omega^i$ or $\lambda_{i\mu}\,dx^\mu$. This data induces a metric, a flat metric connection with nontrivial torsion, and a hierarchy of related tensors and auxiliary connections. In later literatures, the same phrase also appears in teleparallel gravity, Cartan–Tanaka reductions for CR structures, direction-dependent generalized AP geometries, and Lie-group geometry, where “canonical” refers to a preferred frame, connection, coframing, or metric determined by the underlying parallelization (Youssef et al., 2012, Youssef et al., 2016).

1. Foundational definition

An absolute parallelism space, or AP-space, is a parallelizable manifold equipped with a global frame. One formulation takes a pair $(M,X)$, where $M$ is an $n$-dimensional smooth manifold and $M$0 $M$1 are $M$2 independent vector fields defined globally on $M$3. The dual coframe $M$4 is defined by

$M$5

so every vector field $M$6 decomposes globally as

$M$7

An equivalent coordinate formulation uses global vector fields $M$8 with components $M$9 and covariant components $n$0 satisfying

$n$1

The associated metric is then

$n$2

In the global formulation this same metric is written

$n$3

with $n$4 in the default positive-definite presentation; for relativistic applications the frame metric may be replaced by $n$5 (Youssef et al., 2012, Youssef et al., 2016).

The canonical object attached to this data is the unique linear connection for which the global frame is parallel. In coordinate-free form it is characterized by

$n$6

and is given explicitly by

$n$7

In local coordinates this becomes the usual AP formula

$n$8

or, with the $n$9 notation, the standard Weitzenböck-type expression defined directly from the frame. This is the sense in which the connection is canonical: it is determined uniquely by the parallelization itself, with no additional normalization or variational principle required (Youssef et al., 2012, Nwachioma et al., 2017).

2. Canonical connection, torsion, and associated structures

The canonical AP connection is metric and flat, but it is generally not symmetric. Its torsion is the primary nontrivial invariant. In the global formulation,

$X_i$0

and

$X_i$1

Thus, in the parallelization basis, torsion is the obstruction to the frame being holonomic. If

$X_i$2

then the global structure coefficients $X_i$3 directly encode the torsion. The curvature of the canonical connection vanishes identically: $X_i$4 This is the standard teleparallel pattern: flat affine connection, nonzero torsion (Youssef et al., 2012).

The canonical connection sits beside three other natural connections: the dual connection, the symmetric connection, and the Levi-Civita connection of the induced metric. In the global AP treatment they are defined by

$X_i$5

$X_i$6

and the usual metric torsion-free Levi-Civita connection $X_i$7. Their differences from the canonical connection are expressed through torsion $X_i$8 and contortion $X_i$9, where

$\lambda_i$0

The torsion–contortion relations are

$\lambda_i$1

and

$\lambda_i$2

A central structural result is that the curvature tensors of the noncanonical connections can be written entirely in terms of canonical torsion or contortion. For example,

$\lambda_i$3

while the symmetric and Levi-Civita curvatures also admit compact torsion-based formulas. The Wanas tensor is likewise defined globally by

$\lambda_i$4

so it too is a torsion-derived invariant specific to AP geometry (Youssef et al., 2012).

A coordinate treatment of the same geometry emphasizes the same canonical data in index form. The torsion tensor is

$\lambda_i$5

the dual connection is

$\lambda_i$6

the symmetric part is

$\lambda_i$7

and the contortion is

$\lambda_i$8

The relation

$\lambda_i$9

makes explicit that AP torsion measures the deviation from Levi-Civita geometry (Nwachioma et al., 2017).

3. Conformal changes and canonical invariants

A major refinement of canonical AP geometry concerns conformal changes of the parallelization. For an AP-space endowed with its canonical global frame, a conformal rescaling of the AP structure changes the induced metric and the standard AP connections, but suitable modifications of these objects can be made conformally invariant. The work on conformal changes in AP geometry studies the Weitzenböck connection and the Levi-Civita connection of the induced metric and constructs new conformal invariants from them. Its contribution to canonical AP geometry is precisely that it works directly with the canonical frame and the two most natural associated connections and shows how to alter them to obtain tensors and connections unchanged by conformal rescaling of the AP frame (Youssef et al., 2016).

This places conformal geometry inside AP geometry without abandoning the distinguished frame. Rather than treating conformal rescaling as external to the AP structure, the construction identifies combinations of canonical AP data that survive such rescaling. A plausible implication is that, within AP geometry, “canonical” need not mean rigidly fixed up to only diffeomorphism; it can also mean that the frame-determined objects admit invariant reformulations under a prescribed class of frame rescalings. In this sense, conformal AP invariants interpolate between teleparallel frame geometry and Riemannian conformal geometry (Youssef et al., 2016).

4. Teleparallel gravity, preferred frames, and remnant symmetries

In teleparallel gravity, canonical absolute parallelism reappears in vierbein or tetrad language. The basic field is the tetrad $\Omega^i$0, with inverse $\Omega^i$1, and the metric is induced by

$\Omega^i$2

The torsion is

$\Omega^i$3

and the connection is the Weitzenböck connection

$\Omega^i$4

It satisfies

$\Omega^i$5

so the tetrad defines a globally parallel frame. In this setting, the canonical AP connection is the teleparallel connection determined by the preferred tetrad/coframe (Ferraro et al., 2011).

The teleparallel equivalent of general relativity is distinguished by the identity

$\Omega^i$6

so TEGR differs from Einstein–Hilbert only by a total divergence. However, once $\Omega^i$7 is replaced by a nonlinear function $\Omega^i$8, the theory is no longer invariant under local Lorentz transformations of the tetrad. The consequence is that the dynamical variable is not merely the metric but a preferred global tetrad or coframe. In curved FRW cosmologies, the physically correct cosmological frames are therefore the tetrads that actually realize the required parallelization, not arbitrary locally Lorentz-rotated representatives of the same metric (Ferraro et al., 2011).

This loss of local Lorentz invariance leads to the notion of remnant symmetries. In a deliberately non–locally Lorentz invariant 2D pure-tetrad model, the spin connection is not introduced as an independent compensating field, torsion is simply

$\Omega^i$9

and the effective 2D torsion scalar is

$\lambda_{i\mu}\,dx^\mu$0

For a Milne background, the remnant boost functions are fully characterized by three families: an $\lambda_{i\mu}\,dx^\mu$1-dependent family, a $\lambda_{i\mu}\,dx^\mu$2-dependent family, and a mixed family depending on

$\lambda_{i\mu}\,dx^\mu$3

In the vacuum remnant sector, the privileged diads become uniformly accelerated Rindler frames, and the local boost parameter, normally pure gauge in TEGR/GR, becomes dynamically constrained. The same analysis explicitly notes that no Hamiltonian analysis is provided, so “canonical” here refers to preferred frame structure rather than canonical variables in the Dirac sense (Fiorini et al., 2021).

5. Cartan–Tanaka and CR-geometric canonical absolute parallelisms

In several branches of CR geometry, canonical absolute parallelism means a canonical frame or $\lambda_{i\mu}\,dx^\mu$4-structure solving a local equivalence problem. For everywhere $\lambda_{i\mu}\,dx^\mu$5-nondegenerate CR manifolds of hypersurface type in arbitrary odd dimension $\lambda_{i\mu}\,dx^\mu$6, with Levi kernel of arbitrary admissible dimension and regular symbol, the appropriate infinitesimal datum is not an ordinary negatively graded Tanaka symbol but a complex $\lambda_{i\mu}\,dx^\mu$7-bigraded symbol. The corresponding bigraded Tanaka prolongation controls the geometric prolongation procedure. If the bigraded universal algebraic prolongation is finite-dimensional, then to any such CR structure one can assign a canonical absolute parallelism on a bundle of real dimension

$\lambda_{i\mu}\,dx^\mu$8

Two such CR structures are locally equivalent if and only if the associated absolute parallelisms are equivalent, and the symmetry algebra of the flat model is the real part of the bigraded prolongation (Porter et al., 2017).

A concrete five-dimensional instance is Pocchiola’s treatment of $\lambda_{i\mu}\,dx^\mu$9-nondegenerate real hypersurfaces $(M,X)$0 of constant Levi rank $(M,X)$1. In that class, the Levi kernel is spanned by

$(M,X)$2

and $(M,X)$3-nondegeneracy is equivalent to

$(M,X)$4

Cartan’s reduction produces two explicit biholomorphic invariants, $(M,X)$5 and $(M,X)$6. If

$(M,X)$7

the hypersurface is locally biholomorphic to the tube over the light cone. If at least one of $(M,X)$8 or $(M,X)$9 does not vanish on a generic open set, the residual structure group can be completely normalized and the equivalence problem reduces to one between $M$0-structures: in the nonflat case the outcome is a $M$1-dimensional canonical coframing on $M$2, whereas in the flat case two prolongations are required and the result is a $M$3-dimensional $M$4-structure on a prolonged bundle (Pocchiola, 2013).

These CR constructions broaden the meaning of canonical absolute parallelism. The canonical object is no longer the Weitzenböck connection of a given global frame on the base manifold; it is a Cartan-type coframing on a natural bundle whose structure functions encode the local invariants of the CR geometry. What remains constant across both usages is that local equivalence is reduced to equivalence of canonical coframes (Porter et al., 2017, Pocchiola, 2013).

6. Generalizations: direction dependence and Lie-group geometry

Canonical AP geometry also extends to direction-dependent settings. In a Generalized Absolute Parallelism space, or GAP-space, the geometry is placed on the pullback bundle $M$5, and the basic fields $M$6 depend on both position and direction. The induced metric is

$M$7

and the Finsler function is

$M$8

The canonical connection is defined from the Miron connection by

$M$9

or directly by

$n$0

It satisfies the generalized AP-condition

$n$1

and all of its curvature tensors vanish: $n$2 In the FP-Riemannian case, defined by $n$3, the horizontal geometry becomes identical to classical AP geometry (Youssef et al., 2012).

A different generalization reinterprets a Lie group as a flat, globalizable absolute parallelism. If $n$4 is a global trivialization of the frame bundle, with inverse $n$5, the canonical metric is

$n$6

equivalently $n$7. The groupoid arrow determined by the parallelism is

$n$8

and the torsion of the canonical connection is the integrability object

$n$9

For a local Lie group, the primary curvature is

$M$00

and the Riemann curvature tensor $M$01 of the canonical metric satisfies

$M$02

The Ricci tensor of $M$03 is the Killing form: $M$04 $M$05 if and only if the local Lie group is $M$06-step nilpotent, and semisimplicity is equivalent to nondegeneracy of $M$07. This viewpoint makes the canonical metric secondary to the absolute parallelism itself and shows that Lie-algebraic structure can be encoded directly in teleparallel-type torsion data (Ortacgil, 2020).

7. Scope, alternative usages, and common confusions

The phrase “canonical absolute parallelism” is not uniform across the literature. In the global AP literature it denotes the unique flat metric connection parallelizing a chosen global frame (Youssef et al., 2012). In teleparallel gravity it often denotes the preferred tetrad or Weitzenböck connection selected by the field variables, especially when local Lorentz invariance is absent (Ferraro et al., 2011). In CR geometry it means a canonical $M$08-structure or coframing on a natural bundle reducing local equivalence to absolute parallelism (Porter et al., 2017). In path-space treatments it denotes the canonical AP connection together with the associated distinguished path equation

$M$09

derived from the absolutely conserved quantity

$M$10

In the neutral case $M$11, this reduces to

$M$12

which is interpreted there as a spin–torsion interaction (Nwachioma et al., 2017).

An even more idiosyncratic usage appears in a five-dimensional AP theory that identifies a single distinguished second-order frame equation,

$M$13

as the unique parameter-free, singularity-safe AP system. In that paper, “canonical” means uniquely selected by compatibility and singularity-avoidance criteria, together with the claim that the admissible dimension is $M$14 rather than $M$15. This usage is mathematically specific and not identical to the more standard global-connection meaning of canonical AP (Zhogin, 2011).

Taken together, these usages show that Canonical Absolute Parallelism is best understood not as a single universally fixed formalism, but as a family of constructions in which a preferred frame or coframe determines distinguished geometric data. What is stable across the different settings is the primacy of parallelization over metric alone, the presence of a preferred connection or coframing, and the central role of torsion or structure functions in organizing the geometry (Youssef et al., 2012, Ortacgil, 2020).