Connections compatible with degenerate metric structures are known to possess peculiar features: on the one hand, the compatibility conditions involve restrictions on the torsion; on the other hand, torsionfree compatible connections are not unique, the arbitrariness being encoded in a tensor field whose type depends on the metric structure. Nonrelativistic structures typically fall under this scheme, the paradigmatic example being a contravariant degenerate metric whose kernel is spanned by a one-form. Torsionfree compatible (i.e., Galilean) connections are characterised by the gift of a two-form (the force field). Whenever the two-form is closed, the connection is said Newtonian. Such a nonrelativistic spacetime is known to admit an ambient description as the orbit space of a gravitational wave with parallel rays. The leaves of the null foliation are endowed with a nonrelativistic structure dual to the Newtonian one, dubbed Carrollian spacetime. We propose a generalisation of this unifying framework by introducing a new non-Lorentzian ambient metric structure of which we study the geometry. We characterise the space of (torsional) connections preserving such a metric structure which is shown to project to (respectively, embed) the most general class of (torsional) Galilean (respectively, Carrollian) connections.
REFERENCES
PLATO “The Republic” Book VII.
By homogeneous Bargmann group, we mean the homogeneous Galilei group in a d + 2 dimensional representation inherited from that of the Bargmann group, cf. Ref. 22.
The term (Galilean, Carrollian, etc.) “manifold” will stand for a smooth manifold endowed with a metric structure and a compatible connection.
We denote the affine Lie algebra in d + 2 dimensions.
Let be a Lie algebra with being a Lie subalgebra (i.e., ) and an ideal (i.e., ). The algebra is the semidirect sum of with if it admits the decomposition as -modules. The subalgebra will be called the homogeneous part of .
The leaf t = 0 is special in that only the homogeneous Carroll algebra acts faithfully.
Actually, when dealing with non-Riemannian geometries, the definition of the vector space will sometimes necessitate a piece of the metric structure (e.g., for Galilean connections and for Carrollian and ambient Galilean connections).
Indeed, an important discrepancy between Riemannian and non-Riemannian geometries lies in the fact that, in the latter case, only a subset of “metric” structures admits torsionfree compatible connections (e.g., Leibnizian structures with closed absolute clock in the Galilean case or invariant Carroll structures in the Carrollian case).
We pass on an important subtlety that arises in nonrelativistic cases (in contradistinction with the relativistic case), namely, that there may not be a canonical choice of a natural origin (and thus of affine map Θ) but rather a class of such origins (dubbed special connections in our terminology). The class of special connections forms an affine space which can be “resolved,” i.e., which is isomorphic to a “simpler” affine space [e.g., the affine space of fields of observers to define special Galilean connections, cf. Proposition (3.9) in Ref. 5, or the one of principal connections to define special Carrollian connections, cf. the Appendix].
In the relativistic case, the presence of a canonical origin (the Levi-Civita connection) ensures that the classification is a vector space. However, in the nonrelativistic cases, the lack of a canonical origin (cf. footnote 57) prevents the classification to possess a natural structure of vector space. In the latter case, a classification is thus an affine space, isomorphic to , obtained by taking the product of by the affine space used in the definition of Θ and then quotienting by the model vector space of the latter (cf. Proposition B.4 in Ref. 5 for precise statements).
Note that the second relation of Proposition 3.10 is invariant under the transformation V ↦V + fξ, for all (and similarly for W) so that for a given field of observers N, the bilinear form defined by relation 2 really acts on and then constrains components.
We pursue with the notation convention used in Ref. 5 and make use of the same symbols for the various spaces and maps encountered in the torsionfree and torsional cases in order to emphasise the similitude in the logic of the arguments.
Note that the tensor in (3.25) is independent of the choice of field of observers N, hence the absence of superscript on .
Recall that in the Lorentzian case, the arbitrariness is encoded in the unconstrained torsion tensor, i.e., in a section of the vector bundle whose fibers have dimension .
Note that relations 1–2 of (*) together with ensure that . Similarly, the fact that guarantees that .
Note that the invariance of γ and N ensure so that the second component of does not depend on A and will then be denoted as .