
LIE ALGEBRAIC CARROLL/GALILEI DUALITY 3
This allows us to complete the diagram (1.4) to the following commutative diagram of Lie algebras:
0 0
R R
0c b R0
0s0gR0
0 0
(1.6)
which will recur in other contexts in this work with other Lie algebras playing the rôles of s0,c,gand
b. In fact, this very diagram has already appeared in [14, App. B] in the context of the hamiltonian
description of particle dynamics on the (spatially isotropic) homogeneous galilean spacetimes classified
in [15]. In that context, the Lie algebras which play the rôles of s0,c,gand bare infinite-dimensional.
The two exact rows in the the diagram (1.6) say that b(resp. g) is an extension-by-derivation of c(resp.
s0), whereas the two exact columns say that b(resp. c) is a one-dimensional extension of g(resp. s0).
We could say, borrowing the terminology from the theory of metric Lie algebras, that the Bargmann Lie
algebra bis a double extension1of s0. This is more than a mere analogy and we will see that this is
exactly right for Lie algebras admitting ad-invariant bargmannian structures (to be defined below).
The Bargmann Lie algebra bcan also be defined as the subalgebra of the Poincaré Lie algebra
iso(n+1, 1)in one dimension higher which centralises a null translation. Since the Poincaré transla-
tions commute, they are contained in band hence the (simply-connected) Bargmann Lie group acts
transitively on Minkowski spacetime Mn+2. This fact underlies the geometric Carroll/Galilei duality
in [8].
Choosing a Witt frame (ea,e+,e−)for Minkowski spacetime Mn+2we may express the generat-
ors of the Poincaré algebra as (Lab,L+a,L−a,L+−,Pa,P+,P−). The centraliser of P+is spanned by
(Lab,L+a,Pa,P+,P−)and it is isomorphic to the Bargmann Lie algebra b, with P+playing the rôle of M.
If Xis an element of the Poincaré Lie algebra, we let ξXdenote the corresponding Killing vector field on
Minkowski spacetime. The null vector field ξP+on Minkowski spacetime is not just Killing but actually
parallel. It defines a distribution ξ⊥
P+⊂TMwhich is integrable. The leaves of the corresponding foliation
are null hypersurfaces and are copies of the Carroll spacetime Cn+1. Indeed, they are homogeneous
spaces of the normal subgroup with Lie algebra the ideal of bspanned by (Lab,L+a,Pa,P+), which is
isomorphic to the Carroll Lie algebra c, with P+playing the rôle of H. On the other hand, the null re-
duction [16,17] of Mby the one-parameter subgroup generated by ξP+is isomorphic to Galilei spacetime
Gn+1. Indeed, the quotient is homogeneous under the Lie group whose Lie algebra is the quotient of b
by the line spanned by P+, which is isomorphic to the Galilei Lie algebra g. This results in the following
suggestive diagram
Cn+1Mn+2Gn+1,(1.7)
which is the fundamental example of the geometric Carroll/Galilei duality in [8].
This duality seems to be broken when we take all (spatially isotropic) homogeneous kinematical
spacetimes into consideration. In the classification of [15] there are three other homogeneous carrollian
spacetimes besides the Carroll spacetime: the carrollian limits dSC of de Sitter and AdSC of anti de Sitter
spacetimes, and the lightcone LC, which can be realised as null hypersurfaces in de Sitter, anti de Sitter
and Minkowski spacetimes, respectively [15,18]. On the other hand, there are two one-parameter families
of homogeneous galilean spacetimes. The galilean limit dSG of de Sitter spacetime is one point (γ= −1)
in a continuum dSGγ, for γ∈[−1, 1], of galilean spacetimes. Similarly, the galilean limit AdSG of
anti de Sitter spacetime is one point (χ=0)in a continuum AdSGχ, for χ>0, of galilean spacetimes.
These two continua have a point in common, since limγ→1dSGγ=limχ→∞AdSGχ. It is often convenient
to describe homogeneous spaces infinitesimally via their Klein pairs. The above homogeneous galilean
1To be clear, this proposal expands the definition of a double-extension, transcending its origin in the context of metric
Lie algebras, to a more general notion in which a double extension is the composition of a one-dimensional extension with
an extension-by-derivation; at least in those cases where these two operations commute.